Critical Care NEJM PDF

Critical
Care
Clinical Collections — Critical Care
Source: The New England Journal of Medicine
Table of Contents
CRITICAL CARE — INTRODUCTON
CRITICAL CARE SERIES OPENING EDITORIAL

8 Critical Care — An All-Encompassing Specialty
Simon Finfer and Jean-Louis Vincent
Aug 15, 2013
SEVERE SEPSIS AND SEPTIC SHOCK

11 Case Challenge
13 Review Article: Severe Sepsis and Septic Shock
Derek C. Angus and Tom van der Poll
Aug 29, 2013
25 Correspondence
Nov 21, 2013
27 Case Challenge Answer
Related Content
28 o Protocol-Based Approaches Work? Evidence from ProCESS and
D
Other Trials
29 A Randomized Trial of Protocol-Based Care for Early Septic Shock
The ProCESS Investigators
Mar 18, 2014
40 Editorial: The ProCESS Trial — A New Era of Sepsis Management
Craig M. Lilly
Mar 18, 2014
42 Correspondence
Jul 24, 2014
46 Goal-Directed Resuscitation for Patients with Early Septic Shock
The ARISE Investigators and the ANZICS Clinical Trials Group
Oct 16, 2014
57 Correspondence
Jan 8, 2015
60 Trial of Early, Goal-Directed Resuscitation for Septic Shock
Paul R. Mouncey et al.
Mar 17, 2015
71 Systemic Inflammatory Response Syndrome Criteria in Defining Severe Sepsis
Kirsi-Maija Kaukonen et al.
Mar 17, 2015
continued
2
Downloaded from collections.nejm.org. For personal use only. No other uses without permission.
Copyright © Massachusetts Medical Society. All rights reserved.
RESUSCITATION FLUIDS
82 Case Challenge
84 Review Article: Resuscitation Fluids
John A. Myburgh and Michael G. Mythen
Sep 26, 2013
93 Correspondence
Dec 19, 2013
Related Content
98 Albumin Replacement in Patients with Severe Sepsis or Septic Shock
Pietro Caironi et al.
Apr 10, 2014
108 Correspondence
Jul 3, 2014
CIRCULATORY SHOCK
111 Case Challenge
113 Review Article: Circulatory Shock
Jean-Louis Vincent and Daniel De Backer
Oct 31, 2013
122 Correspondence
Feb 6, 2014
Related Content
125 High versus Low Blood-Pressure Target in Patients with Septic Shock
Pierre Asfar et al.
Apr 24, 2014
136 Editorial: Is there a Good MAP for Septic Shock?
James A. Russell
Apr 24, 2014
139 Correspondence
Jul 17, 2014
VENTILATOR-INDUCED LUNG INJURY

143 Case Challenge
145 Review Article: Ventilator-Induced Lung Injury
Arthur S. Slutsky and V. Marco Ranieri
Nov 28, 2013
156 Correspondence
Mar 6, 2014
continued
3
Back to Table of Contents
VENTILATOR-INDUCED LUNG INJURY (continuted)

Related Content
160 Special Article: Driving Pressure and Survival in the Acute Respiratory Distress Syndrome
Marcelo B.P. Amato et al.
Feb 19, 2015
169 Editorial: Driving Pressure and Respiratory Mechanics in ARDS
Stephen H. Loring and Atul Malhotra
Feb 19, 2015
ACUTE LIVER FAILURE

172 Case Challenge
174 Review Article: Acute Liver Failure
William Bernal and Julia Wendon
Dec 26, 2013
184 Correspondence
Mar 20, 2014
SEDATION AND DELIRIUM IN THE INTENSIVE CARE UNIT

189 Case Challenge
191 Review Article: Sedation and Delirium in the Intensive Care Unit
Michael C. Reade and Simon Finfer
Jan 30, 2014
202 Correspondence
Apr 17, 2014
BLEEDING AND COAGULOPATHIES IN CRITICAL CARE

207 Case Challenge
209 Review Article: Bleeding and Coagulopathies in Critical Care
Beverly J. Hunt
Feb 27, 2014
222 Correspondence
May 29, 2014
Related Content
226 Age of Transfused Blood in Critically Ill Adults
Jacques Lacroix et al.
Mar 17, 2015
continued
4
FEEDING CRITICALLY ILL PATIENTS

236 Case Challenge
238 Review Article: Nutrition in the Acute Phase of Critical Illness
Michael P. Casaer and Greet Van den Berghe
Mar 27, 2014
248 Correspondence
Mar 27, 2014
Related Content
253 Trial of the Route of Early Nutritional Support in Critically Ill Adults
Sheila E. Harvey
Oct 30, 2014
265 Editorial: The Route of Early Nutrition in Critical Illness
Deborah Cook and Yaseen Arabi
Oct 30, 2014
267 Correspondence
Jan 29, 2015
270 Early versus On-Demand Nasoenteric Tube Feeding in Acute Pancreatitis
Olaf J. Bakker et al.
Nov 20, 2014
281 Correspondence
Feb 12, 2015
ICU-ACQUIRED WEAKNESS AND RECOVERY FROM CRITICAL ILLNESS

284 Case Challenge
286 Review Article: ICU-Acquired Weakness and Recovery from Critical Illness
John P. Kress and Jesse B. Hall
Apr 24, 2014
296 Correspondence
Jul 17, 2014
TRAUMATIC INTRACRANIAL HYPERTENSION

301 Case Challenge
303 Review Article: Traumatic Intracranial Hypertension
Nino Stocchetti and Andrew I.R. Maas
May 29, 2014
313 Correspondence
Sep 4, 2014
continued
5
DYING WITH DIGNITY IN THE INTENSIVE CARE UNIT

318 Case Challenge
320 Review Article: Dying with Dignity in the Intensive Care Unit
Deborah Cook and Graeme Rocker
Jun 26, 2014
6
CRITICAL CARE COLLECTION —
INTRODUCTION
the new england
journal of medicine Caring for acute, severe illness in specialized units accounts for a substantial fraction of
today’s hospital-based health care. The NEJM Group’s “Collection on Critical Care” is
nejm journal watch
based on a series of review articles on Critical Care Medicine published in the New England
Cardiology
Dermatology Journal of Medicine between August 2013 and June 2014. This series was not meant to be a
Emergency Medicine comprehensive review of the entire field, but rather covered 11 topics that the series edi-
Gastroenterology tors, Jean-Louis Vincent and Simon Finfer, thought would be of interest to the critical care
General Medicine specialist. (See the editorial that opened the series – the next item in this collection.)
Hospital Medicine
Infectious Diseases When originally published the series was widely read, attracting over 50,000 views of each
Neurology article within the first few months of publication. One of the series’ features was a case
Oncology and Hematology challenge that was published a fortnight before a related review article; the “answers” to
Pediatrics and Adolescent Medicine
that challenge were published with the review article itself. We encourage you to read the
Physician’s First Watch
Psychiatry
case before reading the review, and only then to look at the answer.
Women’s Health
Critical care is changing rapidly. The review articles were current at the time of their
nejm careercenter publication, but to provide an indication of issues that were raised by each article we also
nejm knowledge+ include in this collection the “Letters to the Editor” along with the authors’ replies that
were published in the Journal. Although we have examined the series to be sure that there
are no areas where new consensus has arisen, the onus lies with you the reader to regard
each article as a foundation that was current at the time of publication.
To provide insight into new work that has appeared in the Journal since each review was
published, this collection also contains selected original articles from the Journal (and
related editorials when appropriate) on the topics that were part of the review article
series.
We hope that you find this Critical Care Collection of value for its convenience and utility.
We urge you to follow the Journal closely as we work hard to identify and publish the most
important work in critical care.
— Jeffrey M. Drazen, M.D.

Editor-in-Chief, New England Journal of Medicine
Distinguished Parker B. Francis Professor of
Medicine
Harvard Medical School
July 2015
The n e w e ng l a n d j o u r na l of m e dic i n e
e d i t or i a l
Critical Care — An All-Encompassing Specialty

Simon Finfer, M.D., F.C.I.C.M., and Jean-Louis Vincent, M.D., Ph.D.
The August 29 issue of the Journal will include catering to a patient population that extends to
the first in a series of review articles on critical both extremes of age. In adult ICUs, the average
care. Critical care is a young specialty that is age is increasing and is now commonly well
generally considered to have developed from the over 60 years. Although ICUs admitting patients
successful use of invasive ventilation during the for preplanned brief stays after planned major
1952 polio epidemic in Copenhagen. In his report surgery have very low mortality rates, the rates in
of the response to that epidemic, Ibsen described adult ICUs among patients admitted “for cause”
much more than the use of invasive ventilation; are generally around 15% in developed countries.
he also described collaborative, multidisciplinary In a recent study of Medicare beneficiaries in
care that can serve as a model for critical care the United States, 29.2% of patients had been
services to this day.1 He described managing se- treated in an ICU during the last month of their
vere infections and respiratory failure, providing lives.4 Currently, most deaths in ICUs are ex-
cardiovascular support with resuscitation fluids pected, and ICU clinicians regularly face the de-
and vasopressors, monitoring ventilation by mea- cision of when to change the focus of treatment
suring carbon dioxide, placing nasogastric tubes from attempting to cure to providing palliative
to feed patients, and conducting daily multidis- care. Compassionate care of dying patients re-
ciplinary rounds. He also described the impor- quires that critical care practitioners add yet an-
tance of backup systems when patients’ lives are other essential skill set to their more obvious
so dependent on technology that even brief tech- background knowledge and procedural skills
nical failures will prove fatal.1 designed to sustain life.
From these beginnings, critical care has In 2013, critical care practitioners may rec-
spread to most countries in the world. In many ognize many of the problems faced by Ibsen in
developed societies, the number of critical care 1952. Although we have much more highly de-
beds is increasing while total number of acute veloped technology available, our patients are of-
care hospital beds is decreasing; the proportion ten much older, and many have multiple coexist-
of acute care hospital beds that are intensive ing diseases. Determining how best to use the
care unit (ICU) beds is increasing substantially.2 available technology for our patients’ benefit can
Critical care services consume a high propor- be determined only through high-quality research.
tion of health care budgets. In 2005, critical To the credit of our specialty, large national and
care services in the United States were estimat- international clinical-trial networks are system-
ed to cost $81.7 billion, or 0.66% of the gross atically evaluating both established and new
domestic product.3 treatments in high-quality large-scale trials.5
Although the organization of critical care ser- Most of these trials are funded by competitive,
vices varies from country to country, it is clear peer-reviewed grants, and many of the trial re-
that taken at its broadest definition, critical care ports have been published in the Journal.6-11
is an all-encompassing specialty with almost Although we cannot cover anywhere near the
limitless boundaries. Critical care involves the full range of critical care practice in our series,
use of life-sustaining, high-technology medicine we have invited our authors to address many of
n engl j med 369;7 8

nejm.org august 15, 2013 669
The NewFor
Downloaded from collections.nejm.org. England Journal
personal of Medicine
use only. No other uses without permission.
Downloaded from nejm.orgCopyright
by MJ MEDAS on September
© Massachusetts 29, 2014.
Medical For personal
Society. All rightsuse only. No other uses without permission.
reserved.
Copyright © 2013 Back
Massachusetts Medical Society.
to Table of Contents All rights reserved.
the core issues faced in the ICU. Coming re- 1. Ibsen B. The anaesthetist’s viewpoint on the treatment of
respiratory complications in poliomyelitis during the epidemic
views will address the management of severe in Copenhagen, 1952. Proc R Soc Med 1954;47:72-4.
sepsis, the choice and use of resuscitation fluids, 2. Halpern NA, Pastores SM, Greenstein RJ. Critical care medi-
and the treatment of shock. In addition, they cine in the United States 1985-2000: an analysis of bed numbers,
use, and costs. Crit Care Med 2004;32:1254-9.
will address newer issues that are a product of 3. Halpern NA, Pastores SM. Critical care medicine in the United
our success in supporting older, sicker patients States 2000-2005: an analysis of bed numbers, occupancy rates,
through longer stays in the ICU — problems payer mix, and costs. Crit Care Med 2010;38:65-71.
4. Teno JM, Gozalo PL, Bynum JP, et al. Change in end-of-life
such as the management of delirium, ICU- care for Medicare beneficiaries: site of death, place of care, and
acquired weakness, and recovery from prolonged health care transitions in 2000, 2005, and 2009. JAMA 2013;
critical illness. 309:470-7.
5. Cook D, Brower R, Cooper J, Brochard L, Vincent JL. Multi-
In preparation for the start of the series, we center clinical research in adult critical care. Crit Care Med
have posted a case at NEJM.org that highlights 2002;30:1636-43.
issues raised in the review article on sepsis, the 6. Hébert PC, Wells G, Blajchman MA, et al. A multicenter,
randomized, controlled clinical trial of transfusion require-
first in the series. As the series progresses, each ments in critical care. N Engl J Med 1999;340:409-17. [Erratum,
installment of the case will be accompanied, N Engl J Med 1999;340:1056.]
2 weeks before publication of the review article, 7. The SAFE Study Investigators. A comparison of albumin and
saline for fluid resuscitation in the intensive care unit. N Engl J
by questions about the diagnosis or management Med 2004;350:2247-56.
of the condition to be explored in that month’s 8. NICE-SUGAR Study Investigators. Intensive versus conven-
critical care review article. We encourage you to tional glucose control in critically ill patients. N Engl J Med 2009;
360:1283-97.
follow the case and tell us how you would man- 9. The Acute Respiratory Distress Syndrome Network. Ventila-
age the patient’s treatment. We will post the re- tion with lower tidal volumes as compared with traditional tidal
sults of the online polling to coordinate with volumes for acute lung injury and the acute respiratory distress
syndrome. N Engl J Med 2000;342:1301-8.
publication of the actual review article. 10. Young D, Lamb SE, Shah S, et al. High-frequency oscillation
Disclosure forms provided by the authors are available with for acute respiratory distress syndrome. N Engl J Med 2013;368:
the full text of this article at NEJM.org. 806-13.
11. Ferguson ND, Cook DJ, Guyatt GH, et al. High-frequency
From the George Institute for Global Health and Royal North oscillation in early acute respiratory distress syndrome. N Engl J
Shore Hospital, University of Sydney, Sydney (S.F.); and the De- Med 2013;368:795-805.
partment of Intensive Care Medicine, Université Libre de Brux-
elles, and the Department of Intensive Care, Erasme University N Engl J Med 2013; 369:669
Hospital — both in Brussels (J.L.V.). DOI: 10.1056/NEJMe1304035
Copyright © 2013 Massachusetts Medical Society.
A Role for Finasteride in the Prevention of Prostate Cancer?

Michael LeFevre, M.D., M.S.P.H.
All medical care should seek to achieve one or identify with the challenge that researchers face
more of these three goals: to relieve suffering, to in the ascertainment of cause of death. But at
prevent future suffering, or to prolong life. Pre- any specific age, most single diseases play a rel-
ventive services, by definition, are utilized to atively small role in overall mortality. It is much
prevent future suffering or prolong life. We easier to demonstrate a reduction in disease-
should offer preventive services when science as- specific mortality.
sures us that across the population of patients Prostate cancer is a logical target for a pre-
we serve, we do more good than harm. ventive service, with most of the public dis-
How would we know if a preventive service course about prostate-cancer prevention today
accomplishes one or more of these three goals? focusing on screening. Screening seeks to iden-
All-cause mortality is the most appealing out- tify cancers in asymptomatic persons with the
come in a prevention trial because it clearly re- hope of altering the natural history of those
flects the goal of prolonging life, and it is not cancers that are destined to cause suffering
subject to the difficulties of accurately assigning without doing too much harm in the process. In
a specific cause of death. All clinicians who the multicenter Prostate, Lung, Colorectal, and
struggle with completing a death certificate can Ovarian Cancer Screening Trial1 conducted in
670 n engl j med 369;7 9

nejm.org august 15, 2013
The New England Journal

Downloaded from collections.nejm.org. of Medicine
For personal use only. No other uses without permission.
Downloaded from nejm.org by MJ MEDAS on September
Copyright 29, 2014.
© Massachusetts For personal
Medical use
Society. Allonly. Noreserved.
rights other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
SEVERE SEPSIS AND SEPTIC SHOCK
This area is one of the most controversial in medicine. Over the past 20 years we have become better and better
at recognizing sepsis clinically, understanding its pathobiology and organizing its treatment.
Since the accompanying review article was published, a trio of similarly designed studies, one performed in the
United States, one largely in Australia and New Zealand, and one in the United Kingdom, have been completed
and published in the Journal. These trials show that our current recognition and management of sepsis has ad-
vanced substantially over what it was at the turn of the 21st century. The articles describing these studies have
been included in this collection.
10
Case Challenge
Severe Sepsis and Septic Shock

Simon Finfer, M.D., and Jean-Louis Vincent, M.D., Ph.D., Editors
A 77-year-old man is admitted to the ICU after an emergency laparotomy for a perforated sigmoid colon with
marked peritoneal contamination. He is mechanically ventilated and hypotensive. What should be done first?
Presentation of Case
A 77-year-old man is admitted to the intensive care unit (ICU) of a university hospital from the operating room.
Earlier the same day, he had presented to the emergency department with abdominal pain. His medical history
included treated hypertension and hypercholesterolemia, previous heavy alcohol intake, and mild cognitive im-
pairment. In the emergency department, he was drowsy and confused when roused and was peripherally cold
with cyanosis. The systemic arterial blood pressure was 75/50 mm Hg, and the heart rate was 125 beats per min-
ute. The abdomen was tense and distended. After the administration of 1 liter of intravenous crystalloid to restore
the blood pressure, a computed tomographic scan of the abdomen showed extraluminal gas and suspected extra-
luminal feces consistent with a perforated sigmoid colon. He was treated with intravenous antibiotics and taken to
the operating room for laparotomy. During this procedure, gross fecal peritonitis from a perforated sigmoid colon
was confirmed; resection of the sigmoid colon with closure of the rectal stump and creation of an end colostomy
(Hartmann’s procedure) was performed with extensive peritoneal toilet and washout.
11
On arrival in the ICU, he is still anesthetized, the trachea is intubated, and the lungs are mechanically ventilated
with a fraction of inspired oxygen of 0.4; the arterial blood pressure is supported with a norepinephrine infusion.
When the patient was in the operating room, he received a total of 4 liters of crystalloid. On his arrival in the ICU,
the vital signs are a blood pressure of 88/52 mm Hg, heart rate of 120 beats per minute in sinus rhythm, central ve-
nous pressure of 6 mm Hg, and temperature of 35.6°C. An analysis of arterial blood shows a pH of 7.32, a partial
pressure of carbon dioxide of 28 mm Hg, a partial pressure of oxygen of 85 mm Hg, and a lactate level of 3.0 mmol
per liter.
Case Challenge Question

What therapy should be instituted to reduce this patient’s risk of dying from septic shock?
A. Treatment with intravenous immune globulin (gamma globulin).

B. Treatment with intravenous hydrocortisone (at a dose of 50 mg every 6 hours).
C. Treatment with a hydroxymethylglutaryl coenzyme A reductase inhibitor (statin).
12
review article
Critical Care Medicine

Simon R. Finfer, M.D., and Jean-Louis Vincent, M.D., Ph.D., Editors

Derek C. Angus, M.D., M.P.H., and Tom van der Poll, M.D., Ph.D.
S
From the CRISMA (Clinical Research, Inves- epsis is one of the oldest and most elusive syndromes in medicine.
tigation, and Systems Modeling of Acute Hippocrates claimed that sepsis (σηψις)
´ was the process by which flesh rots,
Illness) Center, Department of Critical
Care Medicine, University of Pittsburgh swamps generate foul airs, and wounds fester.1 Galen later considered sepsis
School of Medicine, Pittsburgh (D.C.A.); a laudable event, necessary for wound healing.2 With the confirmation of germ
and the Center for Experimental and Mo- theory by Semmelweis, Pasteur, and others, sepsis was recast as a systemic infec-
lecular Medicine, Division of Infectious
Diseases, and Center for Infection and tion, often described as “blood poisoning,” and assumed to be the result of the
Immunity Amsterdam, Academic Medical host’s invasion by pathogenic organisms that then spread in the bloodstream.
Center, University of Amsterdam, Am- However, with the advent of modern antibiotics, germ theory did not fully explain
sterdam (T.P.). Address reprint requests
to Dr. Angus at the Department of Criti- the pathogenesis of sepsis: many patients with sepsis died despite successful erad-
cal Care Medicine, University of Pitts- ication of the inciting pathogen. Thus, researchers suggested that it was the host,
burgh, 614 Scaife Hall, 3550 Terrace St., not the germ, that drove the pathogenesis of sepsis.3
Pittsburgh, PA 15261, or at angusdc@
upmc.edu; or to Dr. van der Poll at the In 1992, an international consensus panel defined sepsis as a systemic inflam-
Division of Infectious Diseases, Academ- matory response to infection, noting that sepsis could arise in response to mul-
ic Medical Center, Meibergdreef 9, Rm. tiple infectious causes and that septicemia was neither a necessary condition nor
G2-130, 1105 AZ Amsterdam, the Nether-
lands, or at t.vanderpoll@amc.uva.nl. a helpful term.4 Instead, the panel proposed the term “severe sepsis” to describe
instances in which sepsis is complicated by acute organ dysfunction, and they
This article was updated on November codified “septic shock” as sepsis complicated by either hypotension that is refrac-
21, 2013, at NEJM.org.
tory to fluid resuscitation or by hyperlactatemia. In 2003, a second consensus
N Engl J Med 2013;369:840-51. panel endorsed most of these concepts, with the caveat that signs of a systemic
DOI: 10.1056/NEJMra1208623 inflammatory response, such as tachycardia or an elevated white-cell count, occur
in many infectious and noninfectious conditions and therefore are not helpful in
A correction was made to this article after distinguishing sepsis from other conditions.5 Thus, “severe sepsis” and “sepsis”
publication on November 21, 2013, and is are sometimes used interchangeably to describe the syndrome of infection com-
reflected in this PDF. plicated by acute organ dysfunction.
Incidence a nd C ause s
The incidence of severe sepsis depends on how acute organ dysfunction is defined
and on whether that dysfunction is attributed to an underlying infection. Organ
dysfunction is often defined by the provision of supportive therapy (e.g., mechani-
cal ventilation), and epidemiologic studies thus count the “treated incidence” rath-
er than the actual incidence. In the United States, severe sepsis is recorded in 2% of
patients admitted to the hospital. Of these patients, half are treated in the intensive
care unit (ICU), representing 10% of all ICU admissions.6,7 The number of cases in
the United States exceeds 750,000 per year7 and was recently reported to be rising.8
However, several factors — new International Classification of Diseases, 9th Revision
(ICD-9) coding rules, confusion over the distinction between septicemia and severe
sepsis, the increasing capacity to provide intensive care, and increased awareness
and surveillance — confound the interpretation of temporal trends.
Studies from other high-income countries show similar rates of sepsis in the
ICU.9 The incidence of severe sepsis outside modern ICUs, especially in parts of
840 13
n engl j med 369;9 nejm.org august 29, 2013
The New For

personal useof Medicine
only. No other uses without permission.
Downloaded from nejm.org Copyright
by MJ MEDAS on SeptemberMedical
© Massachusetts 29, 2014. For personal
Society. usereserved.
All rights only. No other uses without permission.
Back to Table of Contents All rights reserved.
critical care medicine
the world in which ICU care is scarce, is largely encoding proteins implicated in the pathogene-
unknown. Extrapolating from treated incidence sis of sepsis, including cytokines and other me-
rates in the United States, Adhikari et al. estimated diators involved in innate immunity, coagula-
up to 19 million cases worldwide per year.10 The tion, and fibrinolysis. However, findings are
true incidence is presumably far higher. often inconsistent, owing at least in part to the
Severe sepsis occurs as a result of both com- heterogeneity of the patient populations stud-
munity-acquired and health care–associated in- ied.19,20 Although a recent genomewide associa-
fections. Pneumonia is the most common cause, tion study21 explored drug responsiveness in
accounting for about half of all cases, followed by sepsis, no such large-scale studies of susceptibil-
intraabdominal and urinary tract infections.7,8,11,12 ity to or outcome of sepsis have been performed.
Blood cultures are typically positive in only one
third of cases, and in up to a third of cases, Cl inic a l Fe at ur e s
cultures from all sites are negative.7,11,13,14 Staphy-
lococcus aureus and Streptococcus pneumoniae are the The clinical manifestations of sepsis are highly
most common gram-positive isolates, whereas variable, depending on the initial site of infec-
Escherichia coli, klebsiella species, and Pseudomonas tion, the causative organism, the pattern of acute
aeruginosa predominate among gram-negative iso- organ dysfunction, the underlying health status
lates.11,14 An epidemiologic study of sepsis of the patient, and the interval before initiation
showed that during the period from 1979 to of treatment. The signs of both infection and or-
2000, gram-positive infections overtook gram- gan dysfunction may be subtle, and thus the
negative infections.15 However, in a more recent most recent international consensus guidelines
study involving 14,000 ICU patients in 75 coun- provide a long list of warning signs of incipient
tries, gram-negative bacteria were isolated in 62% sepsis (Table 1).5 Acute organ dysfunction most
of patients with severe sepsis who had positive commonly affects the respiratory and cardiovas-
cultures, gram-positive bacteria in 47%, and cular systems. Respiratory compromise is classi-
fungi in 19%.12 cally manifested as the acute respiratory distress
Risk factors for severe sepsis are related both syndrome (ARDS), which is defined as hypox-
to a patient’s predisposition for infection and to emia with bilateral infiltrates of noncardiac ori-
the likelihood of acute organ dysfunction if in- gin.22 Cardiovascular compromise is manifested
fection develops. There are many well-known risk primarily as hypotension or an elevated serum
factors for the infections that most commonly lactate level. After adequate volume expansion,
precipitate severe sepsis and septic shock, includ- hypotension frequently persists, requiring the
ing chronic diseases (e.g., the acquired immuno- use of vasopressors, and myocardial dysfunction
deficiency syndrome, chronic obstructive pul- may occur.23
monary disease, and many cancers) and the use The brain and kidneys are also often affected.
of immunosuppressive agents.7 Among patients Central nervous system dysfunction is typically
with such infections, however, the risk factors manifested as obtundation or delirium. Imaging
for organ dysfunction are less well studied but studies generally show no focal lesions, and
probably include the causative organism and the findings on electroencephalography are usually
patient’s genetic composition, underlying health consistent with nonfocal encephalopathy. Criti-
status, and preexisting organ function, along cal illness polyneuropathy and myopathy are
with the timeliness of therapeutic intervention.16 also common, especially in patients with a pro-
Age, sex, and race or ethnic group all influence longed ICU stay.24 Acute kidney injury is mani-
the incidence of severe sepsis, which is higher in fested as decreasing urine output and an in-
infants and elderly persons than in other age creasing serum creatinine level and frequently
groups, higher in males than in females, and requires treatment with renal-replacement ther-
higher in blacks than in whites.7,17 apy. Paralytic ileus, elevated aminotransferase
There is considerable interest in the contribu- levels, altered glycemic control, thrombocytope-
tion of host genetic characteristics to the inci- nia and disseminated intravascular coagulation,
dence and outcome of sepsis, in part because of adrenal dysfunction, and the euthyroid sick syn-
strong evidence of inherited risk factors.18 Many drome are all common in patients with severe
studies have focused on polymorphisms in genes sepsis.5

The NewFor
by MJ MEDAS on September
reserved.
Copyright © 2013 Back
Table 1. Diagnostic Criteria for Sepsis, Severe Sepsis, and Septic Shock.*
Sepsis (documented or suspected infection plus ≥1 of the following)†

General variables
Fever (core temperature, >38.3°C)
Hypothermia (core temperature, <36°C)
Elevated heart rate (>90 beats per min or >2 SD above the upper limit of the normal range for age)
Tachypnea
Altered mental status
Substantial edema or positive fluid balance (>20 ml/kg of body weight over a 24-hr period)
Hyperglycemia (plasma glucose, >120 mg/dl [6.7 mmol/liter]) in the absence of diabetes
Inflammatory variables
Leukocytosis (white-cell count, >12,000/mm3)
Leukopenia (white-cell count, <4000/mm3)
Normal white-cell count with >10% immature forms
Elevated plasma C-reactive protein (>2 SD above the upper limit of the normal range)
Elevated plasma procalcitonin (>2 SD above the upper limit of the normal range)
Hemodynamic variables
Arterial hypotension (systolic pressure, <90 mm Hg; mean arterial pressure, <70 mm Hg; or decrease in systolic
pressure of >40 mm Hg in adults or to >2 SD below the lower limit of the normal range for age)
Elevated mixed venous oxygen saturation (>70%)‡
Elevated cardiac index (>3.5 liters/min/square meter of body-surface area)§
Organ-dysfunction variables
Arterial hypoxemia (ratio of the partial pressure of arterial oxygen to the fraction of inspired oxygen, <300)
Acute oliguria (urine output, <0.5 ml/kg/hr or 45 ml/hr for at least 2 hr)
Increase in creatinine level of >0.5 mg/dl (>44 μmol/liter)
Coagulation abnormalities (international normalized ratio, >1.5; or activated partial-thromboplastin time, >60 sec)
Paralytic ileus (absence of bowel sounds)
Thrombocytopenia (platelet count, <100,000/mm3)
Hyperbilirubinemia (plasma total bilirubin, >4 mg/dl [68 μmol/liter])
Tissue-perfusion variables
Hyperlactatemia (lactate, >1 mmol/liter)
Decreased capillary refill or mottling
Severe sepsis (sepsis plus organ dysfunction)
Septic shock (sepsis plus either hypotension [refractory to intravenous fluids] or hyperlactatemia)¶
* Data are adapted from Levy et al.5

† In children, diagnostic criteria for sepsis are signs and symptoms of inflammation plus infection with hyperthermia or
hypothermia (rectal temperature, >38.5°C or <35°C, respectively), tachycardia (may be absent with hypothermia), and at
least one of the following indications of altered organ function: altered mental status, hypoxemia, increased serum lac-
tate level, or bounding pulses.
‡ A mixed venous oxygen saturation level of more than 70% is normal in newborns and children (pediatric range, 75 to 80%).
§ A cardiac index ranging from 3.5 to 5.5 liters per minute per square meter is normal in children.
¶ Refractory hypotension is defined as either persistent hypotension or a requirement for vasopressors after the adminis-
tration of an intravenous fluid bolus.
Ou t c ome death from septic shock were often in excess of

80% as recently as 30 years ago.25 However, with
Before the introduction of modern intensive care advances in training, better surveillance and
with the ability to provide vital-organ support, monitoring, and prompt initiation of therapy to
severe sepsis and septic shock were typically le- treat the underlying infection and support failing
thal. Even with intensive care, rates of in-hospital organs, mortality is now closer to 20 to 30% in
842 n engl j med 369;9 15

The New EnglandFor

Downloaded from collections.nejm.org. Journal of Medicine
personal use only. No other uses without permission.
Downloaded from nejm.org by MJ MEDAS on©September
Copyright 29, 2014.
Massachusetts For personal
Medical Society. use only. No
All rights other uses without permission.
reserved.
Copyright © 2013 Massachusetts Medical Society. All
Back to Table of Contents rights reserved.
many series.7,26 With decreasing death rates, at- four main classes — toll-like receptors, C-type
tention has focused on the trajectory of recovery lectin receptors, retinoic acid inducible gene 1–like
among survivors. Numerous studies have sug- receptors, and nucleotide-binding oligomerization
gested that patients who survive to hospital dis- domain–like receptors — have been identified,
charge after sepsis remain at increased risk for with the last group partially acting in protein
death in the following months and years. Those complexes called inflammasomes (Fig. 1).31
who survive often have impaired physical or neu- These receptors recognize structures that are
rocognitive functioning, mood disorders, and a conserved among microbial species, so-called
low quality of life.27 In most studies, determining pathogen-associated molecular patterns, result-
the causal role of sepsis in such subsequent disor- ing in the up-regulation of inflammatory gene
ders has been difficult. However, a recent analy- transcription and initiation of innate immunity.
sis of the Health and Retirement Study, involving The same receptors also sense endogenous mol-
a large, longitudinal cohort of aging Americans, ecules released from injured cells, so-called
suggested that severe sepsis significantly acceler- damage-associated molecular patterns, or alarm-
ated physical and neurocognitive decline.28 ins, such as high-mobility group protein B1, S100
proteins, and extracellular RNA, DNA, and his-
Pathoph ysiol o gy tones.32 Alarmins are also released during sterile
injury such as trauma, giving rise to the concept
Host Response that the pathogenesis of multiple organ failure in
As the concept of the host theory emerged, it was sepsis is not fundamentally different from that in
first assumed that the clinical features of sepsis noninfectious critical illness.32
were the result of overly exuberant inflamma-
tion. Later, Bone et al.29 advanced the idea that Coagulation Abnormalities
the initial inflammatory response gave way to a Severe sepsis is almost invariably associated with
subsequent “compensatory antiinflammatory re- altered coagulation, frequently leading to dis-
sponse syndrome.” However, it has become ap- seminated intravascular coagulation.33 Excess
parent that infection triggers a much more com- fibrin deposition is driven by coagulation
plex, variable, and prolonged host response, in through the action of tissue factor, a transmem-
which both proinflammatory and antiinflamma- brane glycoprotein expressed by various cell
tory mechanisms can contribute to clearance of types; by impaired anticoagulant mechanisms,
infection and tissue recovery on the one hand including the protein C system and antithrom-
and organ injury and secondary infections on the bin; and by compromised fibrin removal owing
other.30 The specific response in any patient de- to depression of the fibrinolytic system (Fig. 2).33
pends on the causative pathogen (load and viru- Protease-activated receptors (PARs) form the mo-
lence) and the host (genetic characteristics and lecular link between coagulation and inflamma-
coexisting illnesses), with differential responses tion. Among the four subtypes that have been
at local, regional, and systemic levels (Fig. 1). The identified, PAR1 in particular is implicated in
composition and direction of the host response sepsis.33 PAR1 exerts cytoprotective effects when
probably change over time in parallel with the stimulated by activated protein C or low-dose
clinical course. In general, proinflammatory reac- thrombin but exerts disruptive effects on endo-
tions (directed at eliminating invading pathogens) thelial-cell barrier function when activated by
are thought to be responsible for collateral tissue high-dose thrombin.34 The protective effect of
damage in severe sepsis, whereas antiinflamma- activated protein C in animal models of sepsis is
tory responses (important for limiting local and dependent on its capacity to activate PAR1 and
systemic tissue injury) are implicated in the en- not on its anticoagulant properties.34
hanced susceptibility to secondary infections.
Antiinflammatory Mechanisms
Innate Immunity and Immunosuppression
Knowledge of pathogen recognition has in- The immune system harbors humoral, cellular,
creased tremendously in the past decade. Patho- and neural mechanisms that attenuate the poten-
gens activate immune cells through an interac- tially harmful effects of the proinflammatory
tion with pattern-recognition receptors, of which response (Fig. 1).30 Phagocytes can switch to an

The New
Downloaded from collections.nejm.org. ForEngland
personalJournal of Medicine
Downloaded from nejm.org by MJ MEDAS
Copyright on September
Medical ForAll
Society. personal use only. No other uses without permission.
rights reserved.
Copyright © 2013Back
Massachusetts Medical
to Table of ContentsSociety. All rights reserved.
Proinflammatory response Excessive inflammation causing collateral damage (tissue injury)
Damage-associated
Pathogen factors Perpetuation of inflammation molecular patterns
Load
Virulence
Pathogen-associated
molecular patterns
Cytokines
Proteases
Reactive oxygen species Complement products Coagulation proteases
CLRs
Host–pathogen interaction
TLRs Leukocyte activation Complement activation Coagulation activation Necrotic cell death
Impaired function Inhibition of proinflammatory

Neuroendocrine regulation
of immune cells gene transcription
NLRs
Brain
Vagus Apoptosis of T, B,
RLRs nerve Celiac and dendritic cells
Endosome ganglion
Antiinflammatory cytokines
Liver,
Host cell Spleen Soluble cytokine receptors
intestine
Negative regulators
Expansion of regulatory of TLR signaling
Norepinephrine
T and myeloid Epigenetic regulation
Host factors Hypothalamic– suppressor cells
Environment pituitary– Acetylcholine
adrenal axis
Genetics
Age Inhibition of proinflammatory
Other illnesses cytokine production Impaired
Medications phagocytosis
Adrenal Catecholamines
gland Cortisol
Antiinflammatory response Immunosuppression with enhanced susceptibility to secondary infections
Figure 1. The Host Response in Severe Sepsis. COLOR FIGURE
Draft 6
The host response to sepsis is characterized by both proinflammatory responses (top of panel, in red) and antiinflammatory immunosup- 8/09/13
Author Angus
pressive responses (bottom of panel, in blue). The direction, extent, and duration of these reactions are determined by1 both host factors
Fig #
(e.g., genetic characteristics, age, coexisting illnesses, and medications) and pathogen factors (e.g., microbial Title
load and virulence). In-
flammatory responses are initiated by interaction between pathogen-associated molecular patterns expressed by pathogens and pattern-
ME
recognition receptors expressed by host cells at the cell surface (toll-like receptors [TLRs] and C-type lectin receptors [CLRs]), in the
DE Drazen
endosome (TLRs), or in the cytoplasm (retinoic acid inducible gene 1–like receptors [RLRs] and nucleotide-binding Artist
oligomerization
Knoper
domain–like receptors [NLRs]). The consequence of exaggerated inflammation is collateral tissue damage and necrotic cell PLEASE
AUTHOR death, which
NOTE:
results in the release of damage-associated molecular patterns, so-called danger molecules that perpetuate inflammation at least in part
Figure has been redrawn and type has been reset
Please check carefully
by acting on the same pattern-recognition receptors that are triggered by pathogens. Issue date 8/29/13
antiinflammatory phenotype that promotes tis- T cells. The acetylcholine release targets α7 cho-
sue repair, and regulatory T cells and myeloid- linergic receptors on macrophages, suppressing
derived suppressor cells further reduce inflam- the release of proinflammatory cytokines.36 In
mation. In addition, neural mechanisms can animal models of sepsis,35 disruption of this
inhibit inflammation.35 In the so-called neuroin- neural-based system by vagotomy increases sus-
flammatory reflex, sensory input is relayed ceptibility to endotoxin shock, whereas stimula-
through the afferent vagus nerve to the brain tion of the efferent vagus nerve or α7 cholinergic
stem, from which the efferent vagus nerve acti- receptors attenuates systemic inflammation.
vates the splenic nerve in the celiac plexus, re- Patients who survive early sepsis but remain
sulting in norepinephrine release in the spleen dependent on intensive care have evidence of im-
and acetylcholine secretion by a subset of CD4+ munosuppression, in part reflected by reduced
844 n engl j med 17

369;9 nejm.org august 29, 2013
The New
Downloaded from collections.nejm.org. England
For Journal
Medical For All
Society. personal
rightsuse only. No other uses without permission.
reserved.
Tissue hypoperfusion Loss of barrier function
Increased coagulation Decreased anticoagulation
Endothelial cell
↓Tissue
↑ Tissue factor pathway S1P3 S1P1
factor inhibitor ↓ TM ↓ Endothelial PAR1
Monocyte ↑ S1P3 and
protein C receptor
↓ S1P1
↓ Antithrombin ↓ Protein C
Microcirculation
↓ Activated protein C ↑ Angiopoietin 2

NETs Vasodilatation and ↑ thrombin
↓ Activated
with trapped protein C ↓ Blood pressure
platelets
↑ PAI-1 ↓Fibrinolysis ↓ Red-cell
Cell shrinkage
deformability and cell death
Neutrophil ↓ VE cadherin and
Thrombus ↓Tight junctions
Thrombosis
Capillary leak
and interstitial
edema
Loss of
Tissue hypoperfusion barrier function
Tissue
Release of Mitochondrial
mitochondrial dysfunction
↓Tissue oxygenation
contents
Organ failure
Figure 2. Organ Failure in Severe Sepsis and Dysfunction of the Vascular Endothelium and Mitochondria.
Sepsis is associated with microvascular thrombosis caused by concurrent activation of coagulation (mediated by tissue factor) COLOR FIGURE and im-
pairment of anticoagulant mechanisms as a consequence of reduced activity of endogenous anticoagulant pathways Draft(mediated
6 by acti-
7/24/13
vated protein C, antithrombin, and tissue factor pathway inhibitor), plus impaired fibrinolysis owing to enhanced release
Author Angus of plasminogen
Fig # 2
activator inhibitor type 1 (PAI-1). The capacity to generate activated protein C is impaired at least in part by reduced expression of two
Title
endothelial receptors: thrombomodulin (TM) and the endothelial protein C receptor. Thrombus formation is further facilitated by neu-
trophil extracellular traps (NETs) released from dying neutrophils. Thrombus formation results in tissue hypoperfusion,
ME which is aggra-
vated by vasodilatation, hypotension, and reduced red-cell deformability. Tissue oxygenation is further impaired DEby the loss of barrier
Drazen
Artist Knoper
function of the endothelium owing to a loss of function of vascular endothelial (VE) cadherin, alterations in endothelial cell-to-cell tight
AUTHOR PLEASE NOTE:
junctions, high levels of angiopoietin 2, and a disturbed balance between sphingosine-1 phosphate receptor 1 (S1P1) Figure hasand S1P3
been redrawn within
and type has been reset
the vascular wall, which is at least in part due to preferential induction of S1P3 through protease activated receptor 1 (PAR1)
Issue date 8/29/13
as a result
of a reduced ratio of activated protein C to thrombin. Oxygen use is impaired at the subcellular level because of damage to mitochondria
from oxidative stress.
expression of HLA-DR on myeloid cells.37 These tients who had died of sepsis in the ICU.37 Be-
patients frequently have ongoing infectious foci, sides the spleen, the lungs also showed evidence
despite antimicrobial therapy, or reactivation of of immunosuppression; both organs had en-
latent viral infection.38,39 Multiple studies have hanced expression of ligands for T-cell inhibi-
documented reduced responsiveness of blood tory receptors on parenchymal cells.37 Enhanced
leukocytes to pathogens in patients with sep- apoptosis, especially of B cells, CD4+ T cells,
sis,30 findings that were recently corroborated by and follicular dendritic cells, has been implicat-
postmortem studies revealing strong functional ed in sepsis-associated immunosuppression and
impairments of splenocytes obtained from pa- death.40,41 Epigenetic regulation of gene expres-
n engl j med 369;9 nejm.org 18 29, 2013

august 845
The New EnglandFor
Downloaded from nejm.org by MJ MEDAS on ©
Copyright September 29, 2014.
Medical Society. use
All only.
rightsNo other uses without permission.
reserved.
Table 2. Guidelines for the Treatment of Severe Sepsis and Septic Shock from the Surviving Sepsis Campaign.*
846
Element of Care Grade†
Resuscitation
Begin goal-directed resuscitation during first 6 hr after recognition 1C
Begin initial fluid resuscitation with crystalloid and consider the addition of albumin 1B
Consider the addition of albumin when substantial amounts of crystalloid are required to maintain adequate arterial pressure 2C
Avoid hetastarch formulations 1C
Begin initial fluid challenge in patients with tissue hypoperfusion and suspected hypovolemia, to achieve ≥30 ml of crystalloids per kilogram of body weight‡ 1C
Continue fluid-challenge technique as long as there is hemodynamic improvement UG
Downloaded from nejm.org

Use norepinephrine as the first-choice vasopressor to maintain a mean arterial pressure of ≥65 mm Hg 1B
Use epinephrine when an additional agent is needed to maintain adequate blood pressure 2B
Copyright
Add vasopressin (at a dose of 0.03 units/min) with weaning of norepinephrine, if tolerated UG
Avoid the use of dopamine except in carefully selected patients (e.g., patients with a low risk of arrhythmias and either known marked left ventricular systolic dys- 2C
The
by MJ MEDAS
function or low heart rate)
Infuse dobutamine or add it to vasopressor therapy in the presence of myocardial dysfunction (e.g., elevated cardiac filling pressures or low cardiac output) or on- 1C
Downloaded from collections.nejm.org.
going hypoperfusion despite adequate intravascular volume and mean arterial pressure
The New
© Massachusetts
Avoid the use of intravenous hydrocortisone if adequate fluid resuscitation and vasopressor therapy restore hemodynamic stability; if hydrocortisone is used, ad- 2C
on September
minister at a dose of 200 mg/day
n engl j med19
Massachusetts
ForEngland
Medical
Target a hemoglobin level of 7 to 9 g/dl in patients without hypoperfusion, critical coronary artery disease or myocardial ischemia, or acute hemorrhage 1B
369;9
personal
29, 2014.
Infection control
Journal
Medical
Society.
to Table of Contents
Obtain blood cultures before antibiotic therapy is administered 1C
use only.
ForAll
nejm.org
Perform imaging studies promptly to confirm source of infection UG
Society.
n e w e ng l a n d j o u r na l
Administer broad-spectrum antibiotic therapy within 1 hr after diagnosis of either severe sepsis or septic shock 1B/1C
personal
of Medicine
All
of
rights use
Reassess antibiotic therapy daily for de-escalation when appropriate 1B
rights
Perform source control with attention to risks and benefits of the chosen method within 12 hr after diagnosis 1C
reserved.
august 29, 2013
Respiratory support
reserved.
Use a low tidal volume and limitation of inspiratory-plateau-pressure strategy for ARDS 1A/1B
m e dic i n e
Apply a minimal amount of positive end-expiratory pressure in ARDS 1B

Administer higher rather than lower positive end-expiratory pressure for patients with sepsis-induced ARDS 2C
No other uses without permission.

Use recruitment maneuvers in patients with severe refractory hypoxemia due to ARDS 2C
Use prone positioning in patients with sepsis-induced ARDS and a ratio of the partial pressure of arterial oxygen (mm Hg) to the fraction of inspired oxygen of 2C
<100, in facilities that have experience with such practice
Elevate the head of the bed in patients undergoing mechanical ventilation, unless contraindicated 1B

Use a conservative fluid strategy for established acute lung injury or ARDS with no evidence of tissue hypoperfusion 1C
Use weaning protocols 1A
sion may also contribute to sepsis-associated
ministration of a bolus of 20 ml of crystalloids (or albumin equivalent) per kilogram of body weight during a period of 5 to 10 minutes for hypovolemia (2C); increased use of inotropes
tinuous positive end-expiratory pressure in the presence of respiratory distress and hypoxemia (2C); use of physical examination therapeutic end points, such as capillary refill (2C); ad-
† For all grades, the number indicates the strength of the recommendation (1, recommended; 2, suggested), and the letter indicates the level of evidence, from high (A) to low (D), with
and vasodilators in septic shock with low cardiac output associated with elevated systemic vascular resistance (2C); and use of hydrocortisone only in children with suspected or prov-
UG indicating ungraded. Recommendations that are specific to pediatric severe sepsis include therapy with face-mask oxygen, high-flow nasal cannula oxygen, or nasopharyngeal con-
immunosuppression.42
1C
2C
2C
1A
1B
2B
1B
1B
1B
Organ Dysfunction
Although the mechanisms that underlie organ
Use a protocol-specified approach to blood glucose management, with the initiation of insulin after two consecutive blood glucose levels of >180 mg/dl (10 mmol/ failure in sepsis have been only partially eluci-
Administer oral or enteral feedings, as tolerated, rather than either complete fasting or provision of only intravenous glucose within the first 48 hr after a diagnosis
dated, impaired tissue oxygenation plays a key
role (Fig. 2). Several factors — including hypo-
tension, reduced red-cell deformability, and
microvascular thrombosis — contribute to dimin-
ished oxygen delivery in septic shock. Inflamma-
tion can cause dysfunction of the vascular endo-
thelium, accompanied by cell death and loss of
barrier integrity, giving rise to subcutaneous and
Use the equivalent of continuous venovenous hemofiltration or intermittent hemodialysis as needed for renal failure or fluid overload
body-cavity edema.43 In addition, mitochondrial

damage caused by oxidative stress and other mech-
anisms impairs cellular oxygen use.44 Moreover,
injured mitochondria release alarmins into the
* Data are adapted from Dellinger et al.23 ARDS denotes acute respiratory distress syndrome, and ICU intensive care unit.
extracellular environment, including mitochon-

drial DNA and formyl peptides, which can acti-
vate neutrophils and cause further tissue injury.45
T r e atmen t
Administer a short course of a neuromuscular blocker (<48 hr) for patients with early, severe ARDS
The Surviving Sepsis Campaign, an international

consortium of professional societies involved in
critical care, treatment of infectious diseases,
Address goals of care, including treatment plans and end-of-life planning as appropriate
‡ The guidelines recommend completing the initial fluid resuscitation within 3 hours (UG).
and emergency medicine, recently issued the third

iteration of clinical guidelines for the manage-
Administer stress-ulcer prophylaxis to prevent upper gastrointestinal bleeding
ment of severe sepsis and septic shock (Table 2).23

The most important elements of the guidelines
are organized into two “bundles” of care: an ini-
Use sedation protocols, targeting specific dose-escalation end points
Avoid neuromuscular blockers if possible in patients without ARDS
tial management bundle to be accomplished with-

in 6 hours after the patient’s presentation and a
management bundle to be accomplished in the
ICU.23 Implementation of the bundles is associ-
liter), targeting a blood glucose level of <180 mg/dl
ated with an improved outcome.46,47

Administer prophylaxis for deep-vein thrombosis
The principles of the initial management

bundle are to provide cardiorespiratory resusci-
tation and mitigate the immediate threats of
uncontrolled infection. Resuscitation requires the
of severe sepsis or septic shock
en absolute adrenal insufficiency (2C).
use of intravenous fluids and vasopressors, with

oxygen therapy and mechanical ventilation pro-
Central nervous system support
vided as necessary. The exact components re-

quired to optimize resuscitation, such as the
General supportive care
choice and amount of fluids, appropriate type

and intensity of hemodynamic monitoring, and
role of adjunctive vasoactive agents, all remain the
subject of ongoing debate and clinical trials;
many of these issues will be covered in this se-
ries.23 Nonetheless, some form of resuscitation is
considered essential, and a standardized approach
n engl j med 369;9 nejm.org 20 29, 2013

august 847
The New England
Downloaded from collections.nejm.org. Journal
For of Medicine
Copyright 29, 2014.
Medical userights
Society. All only. No other uses without permission.
reserved.
has been advocated to ensure prompt, effective agents and agents exerting more pleiotropic ef-
management.23 The initial management of in- fects. The specific agents can be divided into
fection requires forming a probable diagnosis, those designed to interrupt the initial cytokine
obtaining cultures, and initiating appropriate cascade (e.g., antilipopolysaccharide or anti–pro-
and timely empirical antimicrobial therapy and inflammatory cytokine strategies) and those de-
source control (i.e., draining pus, if appropriate). signed to interfere with dysregulated coagulation
The choice of empirical therapy depends on (e.g., antithrombin or activated protein C).61 The
the suspected site of infection, the setting in only new agent that gained regulatory approval
which the infection developed (i.e., home, nurs- was activated protein C.62 However, postapproval
ing home, or hospital), medical history, and lo- concern about the safety and efficacy of activated
cal microbial-susceptibility patterns. Inappropri- protein C prompted a repeat study, which did not
ate or delayed antibiotic treatment is associated show a benefit and led the manufacturer, Eli Lilly,
with increased mortality.48,49 Thus, intravenous to withdraw the drug from the market.11 All other
antibiotic therapy should be started as early as strategies thus far have not shown efficacy. With
possible and should cover all likely pathogens. It the recent decision to stop further clinical devel-
has not been determined whether combination opment of CytoFab, a polyclonal anti–tumor ne-
antimicrobial therapy produces better outcomes crosis factor antibody (ClinicalTrials.gov number,
than adequate single-agent antibiotic therapy in NCT01145560), there are no current large-scale
patients with severe sepsis.50-53 Current guide- trials of anticytokine strategies in the treatment
lines recommend combination antimicrobial of sepsis.
therapy only for neutropenic sepsis and sepsis Among the agents with broader immunomod-
caused by pseudomonas species. Empirical anti- ulatory effects, glucocorticoids have received the
fungal therapy should be used only in patients at most attention. Intravenous immune globulin is
high risk for invasive candidiasis.50 also associated with a potential benefit,63 but
The patient should also be moved to an ap- important questions remain, and its use is not
propriate setting, such as an ICU, for ongoing part of routine practice.23 Despite a large num-
care. After the first 6 hours, attention focuses on ber of observational studies suggesting that the
monitoring and support of organ function, use of statins reduces the incidence or improves
avoidance of complications, and de-escalation of the outcome of sepsis and severe infection,64
care when possible. De-escalation of initial broad- such findings have not been confirmed in ran-
spectrum therapy may prevent the emergence of domized, controlled trials, so the use of statins
resistant organisms, minimize the risk of drug is not part of routine sepsis care.23
toxicity, and reduce costs, and evidence from
observational studies indicates that such an ap- PROBLEMS WITH therapeutic development
proach is safe.54 The only immunomodulatory Faced with these disappointing results, many ob-
therapy that is currently advocated is a short servers question the current approach to the de-
course of hydrocortisone (200 to 300 mg per day velopment of sepsis drugs. Preclinical studies
for up to 7 days or until vasopressor support is commonly test drugs in young, healthy mice or
no longer required) for patients with refractory rats exposed to a septic challenge (e.g., bacteria or
septic shock.23 This recommendation is support- bacterial toxins) with limited or no ancillary treat-
ed by a meta-analysis,55 but the two largest stud- ment. In contrast, patients with sepsis are often
ies had conflicting results,56,57 and other clinical elderly or have serious coexisting illnesses, which
trials are ongoing.58,59 may affect the host response and increase the risk
of acute organ dysfunction. Furthermore, death in
se a rch for ne w ther a pie s the clinical setting often occurs despite the use of
antibiotics, resuscitation, and intensive life sup-
Recent Failures port, and the disease mechanisms in such cases
One of the great disappointments during the past are probably very different from those underlying
30 years has been the failure to convert advances the early deterioration that typically occurs in ani-
in our understanding of the underlying biologic mal models in the absence of supportive care.
features of sepsis into effective new therapies.60 There are also large between-species genetic dif-
Researchers have tested both highly specific ferences in the inflammatory host response.65
848 n engl j med 369;9 21


Copyright 29, 2014.Medical
© Massachusetts For personal useAll
Society. only. No reserved.
Back to Table of Contentsrights reserved.
In clinical studies, the enrollment criteria are survivors of sepsis opens up avenues to explore
typically very broad, the agent is administered on agents currently being tested in patients with
the basis of a standard formula for only a short dementia and related conditions.
period, there is little information on how the agent The designs of trials could be modified to
changes the host response and host–pathogen more easily incorporate these ideas. For exam-
interactions, and the primary end point is death ple, the considerable uncertainty at the begin-
from any cause. Such a research strategy is prob- ning of a trial with regard to the appropriate
ably overly simplistic in that it does not select pa- selection of patients and drug-administration
tients who are most likely to benefit, cannot adjust strategy and the possibility of treatment inter-
therapy on the basis of the evolving host response actions may be better handled with the use of
and clinical course, and does not capture poten- a Bayesian design. A trial could commence with
tially important effects on nonfatal outcomes. multiple study groups that reflect the various un-
certainties to be tested but then automatically nar-
NEW STRATEGIES row assignments to the best-performing groups
Consequently, hope is pinned on newer so-called on the basis of predefined-response adaptive
precision-medicine strategies with better preclin- randomization rules. Such designs could be par-
ical models, more targeted drug development, ticularly helpful when testing combination ther-
and clinical trials that incorporate better patient apy or incorporating potential biomarkers of drug
selection, drug delivery, and outcome measure- responsiveness.
ment. For example, options to enrich the pre-
clinical portfolio include the study of animals C onclusions
that are more genetically diverse, are older, or
have preexisting disease. Longer experiments Severe sepsis and septic shock represent one of
with more advanced supportive care would allow the oldest and most pressing problems in medi-
better mimicry of the later stages of sepsis and cine. With advances in intensive care, increased
multiorgan failure, permitting the testing of awareness, and dissemination of evidence-based
drugs in a more realistic setting and perhaps fa- guidelines, clinicians have taken large strides in
cilitating the measurement of outcomes such as reducing the risk of imminent death associated
cognitive and physical functioning. In addition, with sepsis. However, as more patients survive
preclinical studies could be used to screen for sepsis, concern mounts over the lingering se-
potential biomarkers of a therapeutic response quelae of what was previously a lethal event.
for which there are human homologues. Strategies are also needed to reach the many mil-
Activated protein C mutants that lack antico- lions of patients with sepsis who are far from
agulant properties are examples of more target- modern intensive care. At the same time, advanc-
ed drug development and were shown to provide es in molecular biology have provided keen in-
protection from sepsis-induced death in animals, sight into the complexity of pathogen and alarm
without an increased risk of bleeding.66 Bio- recognition by the human host and important
markers such as whole-genome expression pat- clues to a host response that has gone awry.
terns in peripheral-blood leukocytes may aid in However, harnessing that information to provide
stratifying patients into more homogeneous sub- effective new therapies has proved to be difficult.
groups or in developing more targeted therapeu- To further improve the outcome of patients with
tic interventions.67 The insight that severe sepsis sepsis through the development of new therapeu-
can cause immunosuppression raises the possi- tic agents, newer, smarter approaches to clinical-
bility of using immune-stimulatory therapy (e.g., trial design and execution are essential.
interleukin-7, granulocyte–macrophage colony- Dr. Angus reports receiving grant support through his insti-
stimulating factor,68 or interferon-γ69), but ide- tution from Eisai, consulting fees from Idaho Technology, Pfizer,
ally, such therapy would be used only in patients Eisai, MedImmune, BioAegis, and Ferring, and fees from Eli
Lilly for serving as a member of a clinical-trial data and safety
in whom immunosuppression is identified or monitoring board. Dr. van der Poll reports receiving grant sup-
predicted. Thus, such therapies could be deployed port through his institution from Sirtris Pharmaceuticals and
on the basis of laboratory measures, such as consulting fees from Eisai. No other potential conflict of inter-
est relevant to this article was reported.
monocyte HLA-DR expression. In addition, con- Disclosure forms provided by the authors are available with
cern about accelerated neurocognitive decline in the full text of this article at NEJM.org.

The New
Downloaded from collections.nejm.org. ForEngland Journal
Medical ForAll
Society. personal
rights use only. No other uses without permission.
reserved.
References
1. Majno G. The ancient riddle of sigma et al. Infection rate and acute organ dys- 34. Ruf W. New players in the sepsis-pro-
eta psi iota sigma (sepsis). J Infect Dis function risk as explanations for racial tective activated protein C pathway. J Clin
1991;163:937-45. differences in severe sepsis. JAMA 2010; Invest 2010;120:3084-7.
2. Funk DJ, Parrillo JE, Kumar A. Sepsis 303:2495-503. 35. Andersson U, Tracey KJ. Reflex prin-
and septic shock: a history. Crit Care Clin 18. Sørensen TI, Nielsen GG, Andersen ciples of immunological homeostasis.
2009;25:83-101. PK, Teasdale TW. Genetic and environ- Annu Rev Immunol 2012;30:313-35.
3. Cerra FB. The systemic septic response: mental influences on premature death in 36. Rosas-Ballina M, Olofsson PS, Ochani
multiple systems organ failure. Crit Care adult adoptees. N Engl J Med 1988;318:727- M, et al. Acetylcholine-synthesizing T cells
Clin 1985;1:591-607. 32. relay neural signals in a vagus nerve cir-
4. Bone RC, Sibbald WJ, Sprung CL. The 19. Chung LP, Waterer GW. Genetic pre- cuit. Science 2011;334:98-101.
ACCP-SCCM Consensus Conference on disposition to respiratory infection and 37. Boomer JS, To K, Chang KC, et al. Im-
sepsis and organ failure. Chest 1992;101: sepsis. Crit Rev Clin Lab Sci 2011;48:250- munosuppression in patients who die of
1481-3. 68. sepsis and multiple organ failure. JAMA
5. Levy MM, Fink MP, Marshall JC, et al. 20. Namath A, Patterson AJ. Genetic 2011;306:2594-605.
2001 SCCM/ESICM/ACCP/ATS/SIS Inter- polymorphisms in sepsis. Crit Care Nurs 38. Limaye AP, Kirby KA, Rubenfeld GD,
national Sepsis Definitions Conference. Clin North Am 2011;23:181-202. et al. Cytomegalovirus reactivation in
Crit Care Med 2003;31:1250-6. 21. Man M, Close SL, Shaw AD, et al. Be- critically ill immunocompetent patients.
6. Rangel-Frausto MS, Pittet D, Costigan yond single-marker analyses: mining whole JAMA 2008;300:413-22.
M, Hwang T, Davis CS, Wenzel RP. The genome scans for insights into treatment 39. Torgersen C, Moser P, Luckner G, et
natural history of the systemic inflamma- responses in severe sepsis. Pharmacoge- al. Macroscopic postmortem findings in
tory response syndrome (SIRS): a prospec- nomics J 2012 February 7 (Epub ahead of 235 surgical intensive care patients with
tive study. JAMA 1995;273:117-23. print). sepsis. Anesth Analg 2009;108:1841-7.
7. Angus DC, Linde-Zwirble WT, Lidick- 22. Ranieri VM, Rubenfeld GD, Thomp- 40. Hotchkiss RS, Tinsley KW, Swanson
er J, Clermont G, Carcillo J, Pinsky MR. son BT, et al. Acute respiratory distress PE, et al. Depletion of dendritic cells, but
Epidemiology of severe sepsis in the Unit- syndrome: the Berlin Definition. JAMA not macrophages, in patients with sepsis.
ed States: analysis of incidence, outcome, 2012;307:2526-33. J Immunol 2002;168:2493-500.
and associated costs of care. Crit Care 23. Dellinger RP, Levy MM, Rhodes A, et 41. Hotchkiss RS, Tinsley KW, Swanson
Med 2001;29:1303-10. al. Surviving Sepsis Campaign: interna- PE, et al. Sepsis-induced apoptosis causes
8. Lagu T, Rothberg MB, Shieh MS, Pe- tional guidelines for management of se- progressive profound depletion of B and
kow PS, Steingrub JS, Lindenauer PK. vere sepsis and septic shock: 2012. Crit CD4+ T lymphocytes in humans. J Immu-
Hospitalizations, costs, and outcomes of Care Med 2013;41:580-637. nol 2001;166:6952-63.
severe sepsis in the United States 2003 to 24. De Jonghe B, Sharshar T, Lefaucheur 42. Carson WF, Cavassani KA, Dou Y,
2007. Crit Care Med 2012;40:754-6. [Erra- J, et al. Paresis acquired in the intensive Kunkel SL. Epigenetic regulation of im-
tum, Crit Care Med 2012;40:2932.] care unit: a prospective multicenter study. mune cell functions during post-septic
9. Linde-Zwirble WT, Angus DC. Severe JAMA 2002;288:2859-67. immunosuppression. Epigenetics 2011;6:
sepsis epidemiology: sampling, selection, 25. Friedman G, Silva E, Vincent JL. Has 273-83.
and society. Crit Care 2004;8:222-6. the mortality of septic shock changed 43. Goldenberg NM, Steinberg BE, Slutsky
10. Adhikari NK, Fowler RA, Bhagwanjee with time? Crit Care Med 1998;26:2078- AS, Lee WL. Broken barriers: a new take
S, Rubenfeld GD. Critical care and the 86. on sepsis pathogenesis. Sci Transl Med
global burden of critical illness in adults. 26. Kumar G, Kumar N, Taneja A, et al. 2011;3:88ps25.
Lancet 2010;376:1339-46. Nationwide trends of severe sepsis in the 44. Galley HF. Oxidative stress and mito-
11. Ranieri VM, Thompson BT, Barie PS, 21st century (2000-2007). Chest 2011;140: chondrial dysfunction in sepsis. Br J An-
et al. Drotrecogin alfa (activated) in adults 1223-31. aesth 2011;107:57-64.
with septic shock. N Engl J Med 2012; 27. Angus DC, Carlet J. Surviving inten- 45. Zhang Q, Raoof M, Chen Y, et al. Cir-
366:2055-64. sive care: a report from the 2002 Brussels culating mitochondrial DAMPs cause in-
12. Vincent JL, Rello J, Marshall J, et al. Roundtable. Intensive Care Med 2003;29: flammatory responses to injury. Nature
International study of the prevalence and 368-77. 2010;464:104-7.
outcomes of infection in intensive care 28. Iwashyna TJ, Ely EW, Smith DM, Lan- 46. Ferrer R, Artigas A, Levy MM, et al.
units. JAMA 2009;302:2323-9. ga KM. Long-term cognitive impairment Improvement in process of care and out-
13. Abraham E, Reinhart K, Opal S, et al. and functional disability among survivors come after a multicenter severe sepsis edu-
Efficacy and safety of tifacogin (recombi- of severe sepsis. JAMA 2010;304:1787-94. cational program in Spain. JAMA 2008;
nant tissue factor pathway inhibitor) in 29. Bone RC, Grodzin CJ, Balk RA. Sep- 299:2294-303.
severe sepsis: a randomized controlled sis: a new hypothesis for pathogenesis of 47. Levy MM, Dellinger RP, Townsend
trial. JAMA 2003;290:238-47. the disease process. Chest 1997;112:235- SR, et al. The Surviving Sepsis Campaign:
14. Opal SM, Garber GE, LaRosa SP, et al. 43. results of an international guideline-
Systemic host responses in severe sepsis 30. van der Poll T, Opal SM. Host-patho- based performance improvement program
analyzed by causative microorganism and gen interactions in sepsis. Lancet Infect targeting severe sepsis. Crit Care Med
treatment effects of drotrecogin alfa (ac- Dis 2008;8:32-43. 2010;38:367-74.
tivated). Clin Infect Dis 2003;37:50-8. 31. Takeuchi O, Akira S. Pattern recogni- 48. Paul M, Shani V, Muchtar E, Kariv G,
15. Martin GS, Mannino DM, Eaton S, tion receptors and inflammation. Cell Robenshtok E, Leibovici L. Systematic re-
Moss M. The epidemiology of sepsis in 2010;140:805-20. view and meta-analysis of the efficacy of
the United States from 1979 through 32. Chan JK, Roth J, Oppenheim JJ, et al. appropriate empiric antibiotic therapy for
2000. N Engl J Med 2003;348:1546-54. Alarmins: awaiting a clinical response. sepsis. Antimicrob Agents Chemother
16. Angus DC, Wax RS. Epidemiology of J Clin Invest 2012;122:2711-9. 2010;54:4851-63.
sepsis: an update. Crit Care Med 2001;29: 33. Levi M, van der Poll T. Inflammation 49. Kumar A, Roberts D, Wood KE, et al.
Suppl:S109-S116. and coagulation. Crit Care Med 2010;38: Duration of hypotension before initiation
17. Mayr FB, Yende S, Linde-Zwirble WT, Suppl:S26-S34. of effective antimicrobial therapy is the
850 n engl j med 369;9 23

The New England

Downloaded from collections.nejm.org. ForJournal of Medicine
Downloaded from nejm.org by MJ Copyright
MEDAS on©September 29, 2014.
Massachusetts For Society.
Medical personalAll
userights
reserved.
Back to Table of ContentsAll rights reserved.
critical determinant of survival in human severe sepsis and septic shock in adults: septic shock in critically ill adults: a sys-
septic shock. Crit Care Med 2006;34:1589- a systematic review. JAMA 2009;301:2362- tematic review and meta-analysis. Crit
96. 75. Care Med 2007;35:2686-92.
50. Bochud PY, Bonten M, Marchetti O, 56. Annane D, Sebille V, Charpentier C, 64. Yende S, Milbrandt EB, Kellum JA, et
Calandra T. Antimicrobial therapy for pa- et al. Effect of treatment with low doses al. Understanding the potential role of
tients with severe sepsis and septic shock: of hydrocortisone and fludrocortisone on statins in pneumonia and sepsis. Crit Care
an evidence-based review. Crit Care Med mortality in patients with septic shock. Med 2011;39:1871-8.
2004;32:S495-S512. JAMA 2002;288:862-71. 65. Seok J, Warren HS, Cuenca AG, et al.
51. Safdar N, Handelsman J, Maki DG. 57. Sprung CL, Annane D, Keh D, et al. Genomic responses in mouse models
Does combination antimicrobial therapy Hydrocortisone therapy for patients with poorly mimic human inflammatory dis-
reduce mortality in Gram-negative bacte- septic shock. N Engl J Med 2008;358:111- eases. Proc Natl Acad Sci U S A 2013;110:
raemia? A meta-analysis. Lancet Infect Dis 24. 3507-12.
2004;4:519-27. 58. ADjunctive coRticosteroid trEatment 66. Kerschen EJ, Fernandez JA, Cooley
52. Brunkhorst FM, Oppert M, Marx G, et iN criticAlly ilL Patients With Septic Shock BC, et al. Endotoxemia and sepsis mortal-
al. Effect of empirical treatment with (ADRENAL). ClinicalTrials.gov, 2013 (http:// ity reduction by non-anticoagulant acti-
moxifloxacin and meropenem vs merope- clinicaltrials.gov/ct2/show/NCT01448109). vated protein C. J Exp Med 2007;204:2439-
nem on sepsis-related organ dysfunction 59. Hydrocortisone for Prevention of Sep- 48.
in patients with severe sepsis: a random- tic Shock (HYPRESS). ClinicalTrials.gov, 67. Wong HR. Clinical review: sepsis and
ized trial. JAMA 2012;307:2390-9. 2013 (http://www.clinicaltrials.gov/ct2/ septic shock — the potential of gene ar-
53. Paul M, Benuri-Silbiger I, Soares- show/NCT00670254). rays. Crit Care 2012;16:204.
Weiser K, Leibovici L. Beta lactam mono- 60. Angus DC. The search for effective 68. Meisel C, Schefold JC, Pschowski R, et
therapy versus beta lactam-aminoglyco- therapy for sepsis: back to the drawing al. Granulocyte-macrophage colony-stim-
side combination therapy for sepsis in board? JAMA 2011;306:2614-5. ulating factor to reverse sepsis-associated
immunocompetent patients: systematic 61. Webster NR, Galley HF. Immuno- immunosuppression: a double-blind, ran-
review and meta-analysis of randomised modulation in the critically ill. Br J An- domized, placebo-controlled multicenter
trials. BMJ 2004;328:668. [Erratum, BMJ aesth 2009;103:70-81. trial. Am J Respir Crit Care Med 2009;
2004;328:884.] 62. Bernard GR, Vincent JL, Laterre PF, et 180:640-8.
54. Heenen S, Jacobs F, Vincent JL. Anti- al. Efficacy and safety of recombinant hu- 69. Döcke WD, Randow F, Syrbe U, et al.
biotic strategies in severe nosocomial man activated protein C for severe sepsis. Monocyte deactivation in septic patients:
sepsis: why do we not de-escalate more N Engl J Med 2001;344:699-709. restoration by IFN-gamma treatment. Nat
often? Crit Care Med 2012;40:1404-9. 63. Laupland KB, Kirkpatrick AW, Delaney Med 1997;3:678-81.
55. Annane D, Bellissant E, Bollaert PE, A. Polyclonal intravenous immunoglobu- Copyright © 2013 Massachusetts Medical Society.
et al. Corticosteroids in the treatment of lin for the treatment of severe sepsis and
images in clinical medicine

The Journal welcomes consideration of new submissions for Images in Clinical
Medicine. Instructions for authors and procedures for submissions can be found
on the Journal’s website at NEJM.org. At the discretion of the editor, images that
are accepted for publication may appear in the print version of the Journal,
the electronic version, or both.

The For
Downloaded from collections.nejm.org. Newpersonal
Englanduse
Journal
only.ofNoMedicine
other uses without permission.
Downloaded from nejm.org by MJ©MEDAS
Massachusetts 29,Society.
Medical 2014. For
All personal use only. No other uses without permission.
rights reserved.
c or r e sp ondence

2. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed
To the Editor: We would like to address two therapy in the treatment of severe sepsis and septic shock. N Engl
potentially confusing issues concerning venous J Med 2001;345:1368-77.
3. de Oliveira CF, de Oliveira DS, Gottschald AF, et al. ACCM/
oxygen saturation (Svo2) as presented in Table 1 PALS haemodynamic support guidelines for paediatric septic
of the review by Angus and van der Poll (Aug. 29 shock: an outcomes comparison with and without monitoring
issue).1 First, Table 1 suggests that Svo2 is raised central venous oxygen saturation. Intensive Care Med 2008;34:
1065-75.
in “sepsis, severe sepsis, and septic shock.” De- 4. Deep A, Goonasekera CD, Wang Y, Brierley J. Evolution of
pending on the timing of patient presentation haemodynamics and outcome of fluid-refractory septic shock in
and the type of sepsis and septic shock, Svo2 may children. Intensive Care Med 2013;39:1602-9.
5. Dellinger RP, Levy MM, Rhodes A, et al. Surviving Sepsis
indeed be elevated as a result of microcirculatory Campaign: international guidelines for management of severe
shunting or mitochondrial dysfunction. However, sepsis and septic shock, 2012. Intensive Care Med 2013;39:165-
in septic shock, Svo2 can be depressed, reflecting 228.
an increase in the extraction of oxygen due to a DOI: 10.1056/NEJMc1312359
decrease in cardiac output.2-4 Timely treatment
to increase low Svo2 values in adults and children
may improve outcomes2,3 and has therefore been To the Editor: Angus and van der Poll describe
added to current guidelines for the treatment of the clinical and mechanistic characteristics un-
sepsis.5 Second, Table 1 indicates that higher derlying sepsis and septic shock. In the section
Svo2 levels (70 to 80%) are normal in children. of the article on coagulation abnormalities, al-
As pediatric intensivists, we wish to point out though the authors focused on the molecular fac-
that the opposite may be true in children with tors implicated in coagulation, they did not dis-
sepsis, since their cardiac output is more often cuss the critical role of platelets in sepsis.1 Nor
decreased rather than increased.3,4 This effect did they address the diagnostic indications pro-
compromises oxygen delivery and decreases Svo2 vided by measuring the immature platelet frac-
levels.4 tion.2 In a study conducted by myself and others,
Job Calis, M.D., Ph.D. published earlier this year,3 we reported that
Job van Woensel, M.D., Ph.D. changes in this marker of cellular reactivity, pre-
Academic Medical Center viously proposed as an inexpensive daily screen-
Amsterdam, the Netherlands ing tool for bacterial infection in patients with
job.calis@gmail.com
neutrophilia,2 precede the clinical manifestations
Joris Lemson, M.D., Ph.D. of sepsis.
Radboud University Nijmegen Medical Center
Paying attention to this early cellular marker
Nijmegen, the Netherlands
has two advantages: it is easily measured during
No potential conflict of interest relevant to this letter was re-
ported. routine laboratory blood testing, and it can reveal
the presence of sepsis before it becomes clini-
1. Angus DC, van der Poll T. Severe sepsis and septic shock.
N Engl J Med 2013;369:840-51. cally manifest. Given that the immature platelet
2060 n engl j med 369;21 nejm.org

25 november 21, 2013
correspondence
fraction has already provided useful information 3. Silva JM Jr, De Sousa dos Santos S. Sepsis in AIDS patients:
on coagulation in patients admitted to our gen- clinical, etiological and inflammatory characteristics. J Int AIDS
Soc 2013;16:17344.
eral intensive care unit, we suggest that it merits
attention and possibly wide-scale validation as a DOI: 10.1056/NEJMc1312359
new means of screening critically ill patients for
sepsis and providing more immediate treatment. The Authors Reply: We agree with Calis et al.
Roberto A. De Blasi, M.D. that Svo2 can be low in septic shock. The pur-
pose of Table 1, however, was to list diagnostic
Sapienza University of Rome
Rome, Italy criteria for sepsis. A low Svo2, although possible
radbl@libero.it in sepsis, is a feature of all shock states. In con-
No potential conflict of interest relevant to this letter was re- trast, a high Svo2 provides diagnostic information
ported. because it is indicative of impaired peripheral
1. Tyml K. Critical role for oxidative stress, platelets, and coagu- oxygen extraction, which is consistent with the
lation in capillary blood flow impairment in sepsis. Microcir- subset of so-called distributive shock syndromes,
culation 2011;18:152-62.
2. Di Mario A, Garzia M, Leone F, Arcangeli A, Pagano L, Zini G.
such as sepsis. The pediatric qualification was
Immature platelet fraction (IPF) in hospitalized patients with included simply to clarify the point that the nor-
neutrophilia and suspected bacterial infection. J Infect 2009;59: mal range for Svo is higher in children than in
2
201-6.
3. De Blasi RA, Cardelli P, Costante A, Sandri M, Mercieri M,
adults.
Arcioni R. Immature platelet fraction in predicting sepsis in We also agree with De Blasi that platelet
critically ill patients. Intensive Care Med 2013;39:636-43. pathophysiology is important in sepsis and has
DOI: 10.1056/NEJMc1312359 numerous features. There are many potential
novel biomarkers for infection and sepsis, of
which the immature platelet fraction is but one;
To the Editor: Angus and van der Poll did not more research is needed before the most clini-
discuss the importance of coinfection with the cally reliable indicators can be identified.
human immunodeficiency virus (HIV) in patients Moreira points out that HIV infection can
with sepsis. Severe sepsis is an established cause influence the incidence and course of sepsis.
of short-term and long-term mortality in patients We agree that screening for HIV infection may
with HIV.1 In addition, there are unique features be appropriate in some situations, but we are not
of coinfection that may pose challenges in the sure that mandatory HIV screening should be
management of both illnesses: a wider spectrum part of all sepsis workups.
of microbiologic agents that may require initial Please allow us to clarify an error: in Table 2,
broad-spectrum antimicrobial coverage,2 a more the contents of which we adapted from the guide-
severe course of sepsis,3 substantial interaction lines of the Surviving Sepsis Campaign, the state-
among several antiretroviral agents and medica- ment that fluid resuscitation should continue
tions commonly used in the intensive care unit,2 in the presence of hemodynamic improvement is
and insufficient levels of antiretroviral agents due actually ungraded (UG) rather than recommend-
to impaired gastrointestinal absorption, which is ed (1C).
often seen in critically ill patients. Consequently,
Derek C. Angus, M.D., M.P.H.
I think that HIV serologic status should always be
determined as part of the initial testing for sepsis. University of Pittsburgh
Pittsburgh, PA
José Moreira, M.D. angusdc@upmc.edu
Instituto de Pesquisa Clínica Evandro Chagas Tom van der Poll, M.D., Ph.D.
Rio de Janeiro, Brazil
jose.moreira@ipec.fiocruz.br Academic Medical Center
Amsterdam, the Netherlands
ported. Since publication of their article, the authors report no fur-
ther potential conflict of interest.
1. Japiassú AM, Amâncio RT, Mesquita EC, et al. Sepsis is a DOI: 10.1056/NEJMc1312359
major determinant of outcomes in critically ill HIV/AIDS patients.
Crit Care 2012;14(4):R152.
2. Ganesan A, Masur H. Critical care of persons infected with
the human immunodeficiency virus. Clin Chest Med 2013;34:
307-23.

nejm.org november 21, 2013 2061

What therapy should be instituted to reduce this patient’s risk of dying from septic shock?
A. Treatment with intravenous immune globulin (gamma globulin).

B. Treatment with intravenous hydrocortisone (at a dose of 50 mg every 6 hours).
C. Treatment with a hydroxymethylglutaryl coenzyme A reductase inhibitor (statin).
Case Challenge Answer

The patient in the vignette has septic shock, which with current treatment carries a risk of death of 25 to 40%
during the index hospital admission. Available treatments that are thought to reduce the risk of death are limited
to early resuscitation with fluids and vasopressors if required, control of the source of sepsis (with surgery, if indi-
cated), and early administration of appropriate empirical antibiotics.1 With the withdrawal from the market of
recombinant human activated protein C, drotrecogin alfa (activated) (Xigris, Eli Lilly), there are currently no
therapies that are licensed specifically for the treatment of severe sepsis and septic shock. Although the recom-
mendation is not strong, the guidelines of the Surviving Sepsis Campaign recommend treatment with 200 mg of
hydrocortisone per day if adequate fluid resuscitation and treatment with vasopressor agents do not restore hemo-
dynamic stability. However, the guidelines stipulate that glucocorticoids should not be administered for the treat-
ment of sepsis in the absence of shock1. Although meta-analyses have suggested that the use of polyvalent immune
globulin (gamma globulin) might confer a survival advantage in patients with severe sepsis and septic shock, these
conclusions are based on low-quality evidence, and such treatment is not currently recommended.2 Likewise, ob-
servational studies have suggested that patients who are treated with statins may have a reduced risk of severe sep-
sis and a reduced risk of sepsis-associated death. However, other studies have suggested that these observed effects
may be due to a “healthy user” effect and indication bias.3 At present, the effect of statin therapy on the outcome
of patients with severe sepsis and septic shock is not fully understood, but there is no robust evidence supporting
the use of statins as a treatment, and the use of such agents is not currently recommended.
References
1. Dellinger RP, Levy MM, Rhodes A, et al. Surviving Sepsis Campaign: international guidelines for manage-
ment of severe sepsis and septic shock, 2012. Intensive Care Med 2013;39:165–228.
2. Soares MO, Welton NJ, Harrison DA, et al. An evaluation of the feasibility, cost and value of information
of a multicentre randomised controlled trial of intravenous immunoglobulin for sepsis (severe sepsis and septic
shock): incorporating a systematic review, meta-analysis and value of information analysis. Health Technol
Assess 2012;16:1–186.
3. Yende S, Milbrandt EB, Kellum JA, et al. Understanding the potential role of statins in pneumonia and
sepsis. Crit Care Med 2011;39:1871–8.
27
Do Protocol-Based Approaches Work? Evidence from ProCESS and

Other Trials.
The ProCESS trial, published as the series was concluding in the spring of 2014, provides evidence that
protocol-based approaches to treating septic shock are no better than current usual care. An editorial on
the trial concludes that ProCESS “identifies early recognition of sepsis, early administration of antibiotics,
early adequate volume resuscitation, and clinical assessment of the adequacy of circulation as the elements
we should focus on to save lives.” Those results were reaffirmed by the ARISE and the ProMISe trials,
published in the Journal some months later.
28
new england
The
journal of medicine
Original Article
established in 1812 may 1, 2014 vol. 370 no. 18
A Randomized Trial of Protocol-Based Care for Early Septic Shock

The ProCESS Investigators*
A BS T R AC T
Background
In a single-center study published more than a decade ago involving patients pre- The members of the writing committee
senting to the emergency department with severe sepsis and septic shock, mortality (Donald M. Yealy, M.D., John A. Kellum,
M.D., David T. Huang, M.D., Amber E.
was markedly lower among those who were treated according to a 6-hour protocol Barnato, M.D., Lisa A. Weissfeld, Ph.D.,
of early goal-directed therapy (EGDT), in which intravenous fluids, vasopressors, and Francis Pike, Ph.D., University of Pitts-
inotropes, and blood transfusions were adjusted to reach central hemodynamic burgh, Pittsburgh; Thomas Terndrup, M.D.,
Ohio State University, Columbus; Henry
targets, than among those receiving usual care. We conducted a trial to determine E. Wang, M.D., University of Alabama at
whether these findings were generalizable and whether all aspects of the protocol Birmingham, Birmingham; Peter C. Hou,
were necessary. M.D., Brigham and Women’s Hospital,
Boston; Frank LoVecchio, D.O., Maricopa
Methods Medical Center, Phoenix; Michael R. Fil-
bin, M.D., Massachusetts General Hos-
In 31 emergency departments in the United States, we randomly assigned patients pital, and Nathan I. Shapiro, M.D., Beth
with septic shock to one of three groups for 6 hours of resuscitation: protocol-based Israel Deaconess Medical Center — both
EGDT; protocol-based standard therapy that did not require the placement of a in Boston; and Derek C. Angus, M.D.,
M.P.H., University of Pittsburgh, Pitts-
central venous catheter, administration of inotropes, or blood transfusions; or usu- burgh) assume responsibility for the con-
al care. The primary end point was 60-day in-hospital mortality. We tested sequen- tent and integrity of the article. Address
tially whether protocol-based care (EGDT and standard-therapy groups combined) reprint requests to Dr. Angus at the De-
partment of Critical Care Medicine, Uni-
was superior to usual care and whether protocol-based EGDT was superior to pro- versity of Pittsburgh, 3550 Terrace St.,
tocol-based standard therapy. Secondary outcomes included longer-term mortality 614 Scaife Hall, Pittsburgh, PA 15261, or
and the need for organ support. at angusdc@upmc.edu.
Results * A complete list of investigators in the

Protocolized Care for Early Septic Shock
We enrolled 1341 patients, of whom 439 were randomly assigned to protocol-based (ProCESS) study is provided in the Supple-
EGDT, 446 to protocol-based standard therapy, and 456 to usual care. Resuscitation mentary Appendix, available at NEJM.org.
strategies differed significantly with respect to the monitoring of central venous
This article was published on March 18,
pressure and oxygen and the use of intravenous fluids, vasopressors, inotropes, and 2014, at NEJM.org.
blood transfusions. By 60 days, there were 92 deaths in the protocol-based EGDT
group (21.0%), 81 in the protocol-based standard-therapy group (18.2%), and 86 in The Supplementary Appendix for this
article is available at NEJM.org
the usual-care group (18.9%) (relative risk with protocol-based therapy vs. usual (http://www.nejm.org/action/
care, 1.04; 95% confidence interval [CI], 0.82 to 1.31; P = 0.83; relative risk with showSupplements?doi=10.1056%
protocol-based EGDT vs. protocol-based standard therapy, 1.15; 95% CI, 0.88 to 2FNEJMoa1401602&viewType=
Popup&viewClass=Suppl)
1.51; P = 0.31). There were no significant differences in 90-day mortality, 1-year
mortality, or the need for organ support. N Engl J Med 2014;370:1683-93.
DOI: 10.1056/NEJMoa1401602
Conclusions Copyright © 2014 Massachusetts Medical Society.
In a multicenter trial conducted in the tertiary care setting, protocol-based resuscita-

tion of patients in whom septic shock was diagnosed in the emergency department
did not improve outcomes. (Funded by the National Institute of General Medical
Sciences; ProCESS ClinicalTrials.gov number, NCT00510835.)
n engl j med 370;18 nejm.org may 1, 2014 1683

The New England Journal of Medicine
Downloaded from nejm.org by MIKE MULARCZYK on January 15, 29 2015. For personal use only. No other uses without permission.
T
here are more than 750,000 cases Lifesciences, but the company had no other role
of severe sepsis and septic shock in the in the study. Study coordinators at each site en-
United States each year.1 Most patients tered data into a secure Web-based data-collec-
who present with sepsis receive initial care in the tion instrument. The University of Pittsburgh
emergency department, and the short-term mor- Clinical Research, Investigation, and Systems
tality is 20% or more.2,3 In 2001, Rivers et al. re- Modeling of Acute Illness (CRISMA) Center man-
ported that among patients with severe sepsis or aged all the data and generated blinded and un-
septic shock in a single urban emergency depart- blinded reports for the data and safety monitor-
ment, mortality was significantly lower among ing board. We reported the statistical analysis
those who were treated according to a 6-hour plan before the data were unblinded.8 The clini-
protocol of early goal-directed therapy (EGDT) cal coordinating team and investigators at the
than among those who were given standard ther- participating sites remained unaware of the
apy (30.5% vs. 46.5%).4 On the basis of the prem- study-group outcomes until the data were locked
ise that usual care lacked aggressive, timely as- in December 2013. The writing committee vouch-
sessment and treatment, the protocol for EGDT es for the accuracy and completeness of the data
called for central venous catheterization to mon- and for the fidelity of the study to the protocol.
itor central venous pressure and central venous
oxygen saturation (Scvo2), which were used to Sites and Patients
guide the use of intravenous fluids, vasopressors, All the participating sites were academic hospitals
packed red-cell transfusions, and dobutamine in with more than 40,000 emergency department
order to achieve prespecified physiological tar- visits yearly. To be eligible, the study sites had to
gets. In the decade since the publication of that use the measurement of serum lactate levels as
article, there have been many changes in the man- the method for screening for cryptogenic shock
agement of sepsis, raising the question of whether and had to adhere to the Surviving Sepsis Cam-
all elements of the protocol are still necessary.5-7 paign guidelines9,10 for nonresuscitation aspects
To address this question, we designed a multi- of care but could have no routine resuscitation
center trial comparing alternative resuscitation protocols for septic shock and could not routine-
strategies in a broad cohort of patients with septic ly use continuous Scvo2 catheters. We recruited
shock. Specifically, we tested whether protocol- patients in the emergency department in whom
based resuscitation was superior to usual care and sepsis was suspected according to the treating
whether a protocol with central hemodynamic physician, who were at least 18 years of age, who
monitoring to guide the use of fluids, vasopres- met two or more criteria for systemic inflamma-
sors, blood transfusions, and dobutamine was tory response syndrome11 (see the Methods sec-
superior to a simpler protocol that did not in- tion in the Supplementary Appendix), and who
clude these elements. had refractory hypotension or a serum lactate
level of 4 mmol per liter or higher. We defined
Me thods refractory hypotension as a systolic blood pres-
sure that either was less than 90 mm Hg or re-
Study Oversight quired vasopressor therapy to maintain 90 mm Hg
We conducted the multicenter, randomized Pro- even after an intravenous fluid challenge. We ini-
tocolized Care for Early Septic Shock (ProCESS) tially required the fluid challenge to be 20 ml or
trial at 31 hospitals in the United States. The more per kilogram of body weight, administered
institutional review board at the University of over the course of 30 minutes, but in April 2010,
Pittsburgh and at each other participating site we simplified the requirement to a challenge of
approved the registered study protocol, which is 1000 ml or more administered over the course of
available with the full text of this article at 30 minutes. Patients did not have to be in shock
NEJM.org. The National Institute of General Med- on arrival in the emergency department but had
ical Sciences funded the study and convened an to be enrolled in the study in the emergency de-
independent data and safety monitoring board partment within 2 hours after the earliest detec-
(see the Supplementary Appendix, available at tion of shock and within 12 hours after arrival.
NEJM.org). The Scvo2 monitoring equipment for The exclusion criteria are listed in the Methods
the study was loaned to the sites by Edwards section in the Supplementary Appendix. All pa-
1684 n engl j med 370;18 30

nejm.org may 1, 2014

Downloaded from nejm.org by MIKE MULARCZYK
Copyright ©onMassachusetts
January 15, 2015. For personal
Medical use rights
reserved.
Copyright © 2014 Massachusetts Medical Society. All rights
Back to Table of Contents reserved.
Protocol-based Care for Early Septic Shock
tients or their legally authorized representatives gressive than those used for protocol-based
provided written informed consent. Randomiza- EGDT (Fig. S2 in the Supplementary Appendix).
tion was performed with the use of a centralized ProCESS investigators designed the protocol-
Web-based program in variable block sizes of 3, 6, based standard-therapy approach on the basis of
or 9, with stratification according to site and race. a review of the literature, two independent sur-
veys of emergency physician and intensivist
Study Interventions practice worldwide,5,12 and consensus feedback
We randomly assigned patients, in a 1:1:1 ratio, from investigators. Protocol-based standard ther-
to one of three groups: protocol-based EGDT, apy required adequate peripheral venous access
protocol-based standard therapy, or usual care. (with placement of a central venous catheter only
The same trained and dedicated physician-led if peripheral access was insufficient) and ad-
team implemented both the protocol-based EGDT ministration of fluids and vasoactive agents to
and the protocol-based standard-therapy inter- reach goals for systolic blood pressure and
ventions. The team consisted of at least one avail- shock index (the ratio of heart rate to systolic
able physician who was trained in the protocol- blood pressure) and to address fluid status and
guided resuscitation interventions, a study hypoperfusion, which were assessed clinically
coordinator who monitored adherence to proto- at least once an hour. In contrast to the triggers
col instructions and provided timed prompts, in the EGDT protocol, protocol-based standard
and a bedside nurse. All study physicians were therapy recommended packed red-cell transfu-
trained in emergency medicine or critical care sion only if the hemoglobin level was less than
medicine and had completed a Web-based certifi- 7.5 g per deciliter. The protocol for standard
cation examination. The protocol-based care be- therapy mandated administration of fluids un-
gan in the emergency department but could be til the team leader decided that the patient’s
continued elsewhere. Details regarding the train- fluids were replete. The standard-therapy pro-
ing and conduct of the personnel are provided in tocol, like the EGDT protocol, did not specify
the Methods section in the Supplementary Ap- the type of fluid or vasopressor and did not
pendix. In cases in which a team physician was specify nonresuscitation aspects of care, which
the bedside provider before enrollment, care was were provided by the treating physician. We as-
transferred to a nonstudy physician before en- sessed adherence to the EGDT and standard-
rollment. therapy protocols using an algorithm that screened
For patients randomly assigned to protocol- for decision prompts and actions at 2, 4, and
based EGDT, the resuscitation team followed the 6 hours (Fig. S3 and S4 in the Supplementary
protocol outlined in Figure S1 in the Supplemen- Appendix).
tary Appendix, which mimics that used by Rivers For patients in the usual-care group, the bed-
et al.4 The protocol prompted placement of a side providers directed all care, with the study
central venous catheter to monitor pressure and coordinator collecting data but not prompting
Scvo2 and to administer intravenous fluids, vaso- any actions. Lead investigators at a site could not
pressors, dobutamine, or packed red-cell transfu- serve as the bedside treating physician for pa-
sions, as directed. We did not require placement tients in the usual-care group.
of an arterial catheter for blood-pressure monitor-
ing. The protocol in our study, like the protocol Outcome Measures
in the study by Rivers et al., specified the amount The primary outcome of the study was the rate of
and timing, but not the type, of resuscitation fluid. in-hospital death from any cause at 60 days. Sec-
Similarly, the protocol in our study specified ondary mortality outcomes included the rate of
thresholds for vasopressor use but not the specific death from any cause at 90 days and cumulative
choice of vasopressor. The protocol guided only mortality at 90 days and 1 year. Other outcomes
resuscitation, with all other aspects of care, in- included the duration of acute cardiovascular
cluding the choice of antimicrobial agents, given failure (defined as the duration of the need for
at the discretion of the treating physician. vasopressors), acute respiratory failure, and acute
Protocol-based standard therapy also used a renal failure (defined as the duration of mechan-
team approach with a set of 6-hour resuscitation ical ventilation or dialysis during the acute hospi-
instructions, but the components were less ag- talization, truncated at 60 days, in patients who

nejm.org may 1, 2014 1685
The New
Copyright on Medical
© Massachusetts January 15, 2015.All
Society. Forrights
reserved.
had not had a long-term need for ventilation or the use of type 3 tests; and used compound sym-
dialysis before enrollment); the duration of the metry for the covariance structure.
stay in the hospital and intensive care unit; and For other end points, we used Fisher’s exact
hospital discharge disposition (i.e., discharge to test for categorical outcomes and an analysis of
a long-term or other acute care facility, a nursing variance for continuous outcomes. For survival
home, a private home, or other). We collected in- analyses, we generated Kaplan–Meier estimates,
formation on serious adverse events using stan- assessed between-group differences using the
dard federal guidelines.13 log-rank test, and expressed the data as cumula-
tive mortality curves. In prespecified subgroup
Statistical Analysis analyses, we used the Breslow–Day test to assess
We analyzed all data according to the intention- interactions between treatment assignment and
to-treat principle. For the primary outcome, our subgroups defined according to age, sex, race,
design tested sequentially whether protocol-based source of infection, and enrollment criterion (re-
resuscitation (EGDT or standard therapy) was su- fractory hypotension or elevated serum lactate
perior to usual care and, if it was, whether proto- level). We also conducted post hoc subgroup
col-based EGDT was superior to protocol-based analyses according to thirds of values for the
standard therapy. We initially calculated that Acute Physiology and Chronic Health Evaluation
with a sample of 1950 patients, the study would (APACHE) II score, for the baseline serum lactate
have at least 80% power to detect a reduction in level, and for the time from detection of shock
mortality of 6 to 7 percentage points, at an alpha until randomization, using logistic regression
level of 0.05 for both hypotheses, assuming mor- to test for an interaction between treatment as-
tality of 30 to 46% with usual care; interim signment and subgroups. Unless otherwise
analyses were planned after 650 patients and specified, analyses are for tests of differences
1300 patients had been enrolled. The trial did across the three study groups, with P values of
not meet the stopping criteria at the first planned less than 0.05 considered to indicate statistical
interim analysis (after the enrollment of 650 pa- significance. We used SAS software, version 9.3,
tients). Before the second interim analysis, we for all analyses.
observed that the overall mortality was approxi-
mately 20%, which was much lower than antici- R e sult s
pated but consistent with the results of a recent
study involving similar patients.14 After consulta- Patients
tion with the data and safety monitoring board From March 2008 through May 2013, we enrolled
and the National Institute of General Medical 1351 patients (Fig. 1, and Fig. S5 in the Supple-
Sciences, and with the group assignments still mentary Appendix). Ten patients who provided
concealed, we calculated that we would need to informed consent later requested complete with-
enroll a total of 1350 patients to preserve the drawal from the study, leaving a final cohort of
same power for the same absolute risk reduction. 1341 patients for the analysis: 439 in the protocol-
After spending 0.0005 alpha for the first based EGDT group, 446 in the protocol-based
interim analysis, and after recalculation of the standard-therapy group, and 456 in the usual-
sample size (which removed the requirement for care group. The three groups were well matched
a second interim analysis), the alpha level re- at baseline with respect to demographic and clini-
quired for the sequential hypotheses was 0.0494, cal characteristics, as well as the care received be-
with no adjustment for multiple testing. We tested fore randomization (Table 1, and Tables S1, S2,
for between-group differences in the primary out- and S4 in the Supplementary Appendix).
come using Fisher’s exact test. In the event that
protocol-based care (EGDT and standard therapy Adherence to the Protocol
combined) was not superior to usual care, all Adherence to the protocol was high in both pro-
other analyses were to be specified as secondary. tocol-based groups. At 6 hours, incomplete ad-
Because of possible site heterogeneity, we also herence was recorded in 48 of 404 patients in the
conducted a secondary analysis using a general- EGDT group (11.9%) and 19 of 435 patients in
ized linear mixed model in which we allowed for the standard-therapy group (4.4%) who could be
a random effect of study site, with treatment evaluated (Table S3 in the Supplementary Appen-
group as a covariate; assessed significance with dix). In most of the patients who had been ran-
1686 n engl j med 370;18 32

The New EnglandFor

Downloaded from nejm.org by MIKE MULARCZYK on January 15,Medical
Copyright © Massachusetts 2015. For personal
Society. All use only.
rights No other uses without permission.
reserved.
12,707 Patients were screened
8864 Were ineligible

6841 Did not meet inclusion criteria
2659 Did not have hypoperfusion
2294 Did not have refractory hypotension
120 Could not be assessed for refractory
hypotension
993 Did not have suspected infection
621 Did not have ≥2 SIRS criteria
154 Had other reason
2023 Met exclusion criteria
660 Had “Do Not Resuscitate” order
264 Had treating physician who deemed
aggressive care unsuitable
194 Required immediate surgery
165 Had active gastrointestinal hemorrhage
162 Had contraindication to central venous
catheter
95 Had acute coronary syndrome
71 Had major cardiac arrhythmia
71 Had seizure
69 Had acute pulmonary edema
46 Were participating in another interventional
study
42 Had drug overdose
38 Had CD4 count <50/mm3
36 Were transferred from another in-hospital
setting
31 Had acute cerebrovascular event
21 Had absolute neutrophil count <500 mm3
18 Had burn or trauma
12 Had contraindication to blood transfusion
8 Had status asthmaticus
20 Had other reason
2492 Were eligible but excluded
1191 Had study logistic issues
631 Had decreased mental capacity and no LAR
569 Declined to participate
101 Had other reason
1351 Underwent randomization
445 Were assigned to protocol-based 448 Were assigned to protocol-based 458 Were assigned to usual care
EGDT standard therapy 456 Were eligible for analysis
439 Were eligible for analysis 446 Were eligible for analysis 2 Requested removal of all data
6 Requested removal of all data 2 Requested removal of all data
439 Were included in primary 446 Were included in primary 456 Were included in primary
outcome analysis outcome analysis outcome analysis
Figure 1. Screening, Randomization, and Follow-up.

EGDT denotes early goal-directed therapy, LAR legally authorized representative, and SIRS systemic inflammatory
response syndrome.
domly assigned to EGDT, a central venous catheter, which occurred in 30 of the 439 patients in
eter for monitoring of Scvo2 was placed promptly that group (6.8%), included technical difficulties
(Fig. S6A in the Supplementary Appendix). The (10 patients), refusal by the treating clinician (9)
reasons for failure to place a central venous cath- or patient (5), the need for emergency surgery (1),

The New
Copyright on January
© Massachusetts Medical15, 2015.All
Society. Forrights
reserved.
and death (1); no reason was provided in the case group or the usual-care group, central venous
of 4 patients). The mean (±SD) Scvo2 after cath- catheters were placed in 56.5% of the patients
eterization was 71±13%. Although placement of (252 patients) and 57.9% (264 patients) in the
central venous catheters was not required for pa- two groups, respectively; however, placement oc-
tients in the protocol-based standard-therapy curred later than in the EGDT group (P<0.001)
Table 1. Characteristics of the Patients at Baseline.*
Protocol-Based Protocol-Based
EGDT Standard Therapy Usual Care
Characteristic (N = 439) (N = 446) (N = 456)
Age — yr† 60±16.4 61±16.1 62±16.0
Male sex — no. (%) 232 (52.8) 252 (56.5) 264 (57.9)
Residence before admission — no. (%)‡
Nursing home 64 (14.6) 72 (16.1) 73 (16.0)
Other 373 (85.0) 373 (83.6) 382 (83.8)
Charlson comorbidity score§ 2.6±2.6 2.5±2.6 2.9±2.6
Source of sepsis — no. (%)
Pneumonia 140 (31.9) 152 (34.1) 151 (33.1)
Urinary tract infection 100 (22.8) 90 (20.2) 94 (20.6)
Intraabdominal infection 69 (15.7) 57 (12.8) 51 (11.2)
Infection of unknown source 57 (13.0) 47 (10.5) 66 (14.5)
Skin or soft-tissue infection 25 (5.7) 33 (7.4) 38 (8.3)
Catheter-related infection 11 (2.5) 16 (3.6) 11 (2.4)
Central nervous system infection 3 (0.7) 3 (0.7) 4 (0.9)
Endocarditis 1 (0.2) 3 (0.7) 3 (0.7)
Other 28 (6.4) 31 (7.0) 26 (5.7)
Determined after review not to have infection 5 (1.1) 14 (3.1) 12 (2.6)
Positive blood culture — no. (%) 139 (31.7) 126 (28.3) 131 (28.7)
APACHE II score¶ 20.8±8.1 20.6±7.4 20.7±7.5
Entry criterion — no. (%)
Refractory hypotension 244 (55.6) 240 (53.8) 243 (53.3)
Hyperlactatemia‖ 259 (59.0) 264 (59.2) 277 (60.7)
Physiological variables
Systolic blood pressure — mm Hg 100.2±28.1 102.1±28.7 99.9±29.5
Serum lactate — mmol/liter** 4.8±3.1 5±3.6 4.9±3.1
Time to randomization — min
From arrival in the emergency department†† 197±116 185±112 181±97
From meeting entry criteria 72±77 66±38 69±45
* Plus–minus values are means ±SD. There were no significant differences in baseline characteristics across groups (P values
range from 0.10 to 0.96). EGDT denotes early goal-directed therapy.
† Information on age was missing for one patient in the usual-care group.
‡ Information on residence before admission was missing for four patients. The category of nursing home included personal-
care homes, skilled or unskilled assisted-living facilities, and extended-care facilities.
§ The Charlson comorbidity index15 measures the effect of coexisting conditions on mortality, with scores ranging from
0 to 33 and higher scores indicating a greater burden of illness.
¶ Scores on the Acute Physiology and Chronic Health Evaluation (APACHE) II range from 0 to 71, with higher scores in-
dicating greater severity of illness.
‖ Hyperlactatemia was defined as a serum lactate level of 4 mmol per liter or higher. The serum lactate level was higher
than 2 mmol per liter in 346 patients in the protocol-based EGDT group (78.8%), 340 in the protocol-based standard-
therapy group (76.2%), and 359 in the usual-care group (78.7%).
** Data on the baseline serum lactate level were available for 95.5% of the patients overall (1281 of 1341 patients).
†† Not all patients were eligible at the time of arrival in the emergency department.
1688 n engl j med 370;18 34

The New England

For of Medicine
Downloaded from nejm.org by MIKE MULARCZYK on January 15,
Copyright © Massachusetts 2015. For
Medical personal
Society. use only.
All rights No other uses without permission.
reserved.
(Fig. S6B in the Supplementary Appendix) and lation and moderate glycemic control (Table S4
involved serial monitoring of Scvo2 in only a in the Supplementary Appendix). In general, the
small proportion of patients (4.0% [18 patients] condition of the patients in all three groups im-
in the protocol-based standard-therapy group proved over time, with few differences among
and 3.5% [16 patients] in the usual-care groups, the groups. By 6 hours, the target mean arterial
vs. 93.6% [411 patients] in the EGDT group; pressure of 65 mm Hg or higher had been
P<0.001). achieved in more patients in each of the protocol-
based groups than in the usual-care group
Resuscitation (P = 0.02), but the mean heart rate did not differ
During the first 6 hours, the volume of intra- significantly among the groups (P = 0.32) (Table S2
venous fluids administered differed significantly in the Supplementary Appendix). Patients in the
among the groups (2.8 liters in the protocol-based protocol-based EGDT group had a higher mean
EGDT group, 3.3 liters in the protocol-based international normalized ratio at 6 hours (2.2,
standard-therapy group, and 2.3 liters in the usual- vs. 1.7 in the protocol-based standard-therapy
care group (P<0.001) (Table S4 and Fig. S6C in the group and 1.6 in the usual-care group; P = 0.01),
Supplementary Appendix). The volume of fluids whereas patients in the usual-care group had
administered decreased during the 6 hours in all slightly less acidosis at 6 hours and 24 hours (ar-
the groups, but patients in the protocol-based terial pH, 7.31 in each protocol-based group vs.
standard-therapy group received the greatest vol- 7.34 in the usual-care group at 6 hours, and 7.34
ume initially and overall, patients in the usual- in each protocol-based group vs. 7.36 in the
care group received the least volume of fluid, and usual-care group at 24 hours, P = 0.02), but these
patients in the protocol-based EGDT group re- differences did not persist.
ceived fluid at the most consistent rate (P<0.001
for differences in total volume and P = 0.007 for Outcomes
differences over time). Crystalloids were the pre- By day 60, a total of 92 patients in the protocol-
dominant fluid used in all the groups, adminis- based EGDT group (21.0%), 81 in the protocol-
tered in 96% of the patients overall. More patients based standard-therapy group (18.2%), and 86 in
in the two protocol-based groups than in the the usual-care group (18.9%) had died in the hos-
usual-care group received vasopressors (54.9% in pital (Table 2). The 60-day in-hospital mortality
the protocol-based EGDT group and 52.2% in the for the combined protocol-based groups (19.5%
protocol-based standard-therapy group vs. 44.1% [173 of 885 patients]) did not differ significantly
in the usual-care group, P = 0.003) (Table S4 and from that in the usual-care group (relative risk,
Fig. S6D in the Supplementary Appendix). More 1.04; 95% confidence interval [CI], 0.82 to 1.31;
patients in the protocol-based EGDT group than P = 0.83), nor did mortality differ significantly
in the protocol-based standard-therapy group or when the groups were compared separately (with
the usual-care group received dobutamine and P values ranging from 0.31 to 0.89) (Table 2 and
packed red-cell transfusions (dobutamine use, Fig. 2A). There were also no significant differ-
8.0% vs. 1.1% and 0.9%, respectively; P<0.001; ences in 90-day mortality or in the time to death
packed red-cell transfusions, 14.4% vs. 8.3% and up to 90 days and 1 year (P = 0.66 for 90-day mor-
7.5%, respectively; P = 0.001) (Table S4, and Fig. tality and P = 0.70 and P = 0.92 for cumulative
S6D in the Supplementary Appendix). The use of mortality at 90 days and 1 year, respectively)
antibiotics, glucocorticoids, and activated pro- (Table 2 and Fig. 2B). Results were essentially
tein C was similar across the three groups (with unchanged when adjusted for potential site het-
P values ranging from 0.16 to 0.90) (Table S4 in erogeneity (odds of 60-day in-hospital death with
the Supplementary Appendix). protocol-based care vs. usual care, 1.08; 95% CI,
0.85 to 1.38; P = 0.54).
Ancillary Care The incidence of acute renal failure, as indi-
The use of intravenous fluids, vasopressors, do- cated by a new need for renal-replacement ther-
butamine, and blood transfusions between 6 and apy, was higher in the protocol-based standard-
72 hours did not differ significantly among the therapy group than in the other two groups
groups (Table S4 in the Supplementary Appen- (6.0% in the protocol-based standard-therapy
dix). Patients in all three groups had mean values group vs. 3.1% in the protocol-based EGDT
that were consistent with low-tidal-volume venti- group and 2.8% in the usual-care group, P = 0.04),

The NewFor
Downloaded from nejm.org byCopyright
MIKE MULARCZYK on January
© Massachusetts 15,Society.
Medical 2015. For
Allpersonal use only. No other uses without permission.
rights reserved.
Table 2. Outcomes.*
Protocol-based Protocol-based
EGDT Standard Therapy Usual Care
Outcome (N = 439) (N = 446) (N = 456) P Value†
Death — no./total no. (%)
In-hospital death by 60 days: primary outcome 92/439 (21.0) 81/446 (18.2) 86/456 (18.9) 0.83‡
Death by 90 days 129/405 (31.9) 128/415 (30.8) 139/412 (33.7) 0.66
New organ failure in the first week — no./total no. (%)
Cardiovascular 269/439 (61.3) 284/446 (63.7) 256/456 (56.1) 0.06
Respiratory 165/434 (38.0) 161/441 (36.5) 146/451 (32.4) 0.19
Renal 12/382 (3.1) 24/399 (6.0) 11/397 (2.8) 0.04
Duration of organ support — days§
Cardiovascular 2.6±1.6 2.4±1.5 2.5±1.6 0.52
Respiratory 6.4±8.4 7.7±10.4 6.9±8.2 0.41
Renal 7.1±10.8 8.5±12 8.8±13.7 0.92
Use of hospital resources
Admission to intensive care unit — no. (%) 401 (91.3) 381 (85.4) 393 (86.2) 0.01
Stay in intensive care unit among admitted 5.1±6.3 5.1±7.1 4.7±5.8 0.63
patients — days
Stay in hospital — days 11.1±10 12.3±12.1 11.3±10.9 0.25
Discharge status at 60 days — no. (%)
Not discharged 3 (0.7) 8 (1.8) 2 (0.4) 0.82
Discharged to a long-term acute care facility 16 (3.6) 22 (4.9) 22 (4.8)
Discharge to another acute care hospital 8 (1.8) 2 (0.4) 5 (1.1)
Discharged to nursing home 71 (16.2) 93 (20.9) 88 (19.3)
Discharged home 236 (53.8) 227 (50.9) 235 (51.5)
Other or unknown 13 (3.0) 13 (2.9) 18 (3.9)
Serious adverse events — no. (%)¶ 23 (5.2) 22 (4.9) 37 (8.1) 0.32
* Plus–minus values are means ±SD.

† Unless stated otherwise, P values are for a three-group comparison, with the use of Fisher’s exact test for categorical
measures and linear models for continuous and normally distributed measures. Skewed outcomes were analyzed with
the use of nonparametric alternatives.
‡ The P value for the primary analysis was for a comparison between the two protocol-based groups combined and the
usual-care group, with the use of Fisher’s exact test. The three-group comparison, with the use of Fisher’s exact test, was
also nonsignificant (P = 0.55), as was each one of the two-way comparisons (with P values ranging from 0.31 to 0.89).
§ Included in the analysis were patients in whom new organ failure developed in the first week after randomization.
¶ A detailed list of serious adverse events is provided in Table S5 in the Supplementary Appendix.
although the duration of therapy did not differ ences in the length of stay in the hospital or the
significantly across the groups (Table 2). The discharge disposition (Table 2).
rate of admission to the intensive care unit was Reports of potentially serious adverse events
higher in the protocol-based EGDT group than (excluding death) were rare and did not differ
in the other two groups, although among pa- significantly across groups (Table 2, and Table S5
tients who were admitted, there were no signifi- in the Supplementary Appendix). There were no
cant between-group differences in the length of significant interactions between the assigned
stay in the intensive care unit (Table 2). There treatment and any prespecified subgroup with re-
were no significant differences in the incidence spect to the primary outcome of 60-day in-hospital
and duration of cardiovascular failure or respira- mortality or with respect to the secondary mortal-
tory failure, nor were there significant differ- ity outcomes (Table S6 in the Supplementary Ap-
1690 n engl j med 370;18 36

The New England For

Downloaded from nejm.org by MIKE MULARCZYK on January 15, Medical
Copyright © Massachusetts 2015. ForSociety.
personalAlluse only.reserved.
pendix). Similarly, in a post hoc analysis, there

Protocol-based EGDT Protocol-based Usual care
was no evidence of a treatment effect within standard therapy
ranges of values for the APACHE II score, serum
lactate level, or time from meeting the criteria A Cumulative In-Hospital Mortality to 60 Days
for shock to randomization (Table S7 in the Sup- 50
plementary Appendix).
40
Discussion
Mortality (%)
30
In our study, adherence to the two experimental

20
protocols was high, and, as expected, protocol-
based care, as compared with usual care, result- 10 P=0.52 by log-rank test
ed in increased use of central venous catheteriza-
tion, intravenous fluids, vasoactive agents, and 0
blood transfusions. The two protocol-based re- 0 10 20 30 40 50 60
suscitation approaches led to a small but tran- Days
sient improvement in blood pressure by the end No. at Risk
of the resuscitation period but a higher require- Protocol-based EGDT 439 373 356 348 347 347 347
Protocol-based standard therapy 446 389 376 368 366 366 365
ment for intensive care and renal-replacement Usual care 456 396 376 371 371 371 370
therapy. There were no significant differences in
mortality, either overall or in a number of pre-
specified and post hoc subgroups. B Cumulative Mortality to 1 Yr
50
Our results differ from those of Rivers et al.4;
90 days
however, our study was not a direct replication
40
of that study, and there are probably several fac-
tors that contribute to the differences. Although
Mortality (%)
30
the two trials used similar inclusion criteria, the
enrolled populations differed. The study cohorts 20
were similar with respect to many demographic
and clinical characteristics, including the sever- 10 P=0.70 by log-rank test, 90 days
ity of illness (Table S8 in the Supplementary P=0.92 by log-rank test, 1 yr
Appendix), but the cohort in the study by Rivers 0
et al. was slightly older, had higher rates of pre- 0 60 120 180 240 300 365
existing heart and liver disease, and had a higher Days

initial serum lactate level. Although we modified No. at Risk
Protocol-based EGDT 439 289 217 194 175 156 145
the minimum fluid bolus required to establish Protocol-based standard therapy 446 308 212 196 179 158 142
the presence of refractory hypotension, the mean Usual care 456 285 211 199 181 164 139
volume of the bolus that was administered fell
within the range used in the study by Rivers et Figure 2. Cumulative Mortality.
al. (20 to 30 ml per kilogram). The mean initial Panel A shows cumulative in-hospital mortality, truncated at 60 days, and
Scvo2 reported by Rivers et al. was 49%, which Panel B cumulative mortality up to 1 year after randomization.
was lower than that in the ProCESS trial. How-
ever, early central venous catheterization was
considered to be part of usual care in that trial, theless, the cohort in the study by Rivers et al.
allowing Scvo2 readings to be made before ad- may have had, on average, more severe or persis-
ministration of the initial fluid bolus, the re- tent shock than the patients in our cohort. How-
sponse to which was required to establish re- ever, we were unable to show a benefit even
fractory hypotension. In contrast, for patients when we restricted the analyses to the sickest
randomly assigned to the protocol-based EGDT third of our patients — those with the highest
group in our study, we measured Scvo2 only after serum lactate levels and those with the highest
the initial fluid bolus had been administered, APACHE II scores.
making a direct comparison problematic. None- Both trials used the same EGDT protocol
n engl j med 370;18 nejm.org 37

may 1, 2014 1691
The New England
For of Medicine
Medical personal
Society. use only.
reserved.
delivered by a trained, dedicated team at each extent to which any of these strategies offer ad-
site. Rivers et al. reported nearly perfect adher- vantages in settings where septic shock is not
ence but did not provide details regarding the recognized promptly. Third, septic shock occurs
assessment method. Although adherence to the in a heterogeneous population, and care before
protocol was high in our study, we cannot ex- randomization can be variable. Fourth, we had
clude the possibility that the outcome would limited power to address whether particular strat-
have been better if adherence had been perfect. egies were more effective in specific subgroups.
We believe that the rate of adherence in our Two ongoing multicenter trials of EGDT, the
study parallels the likely performance in any Australasian Resuscitation in Sepsis Evaluation
widespread effort targeting the care of patients (ARISE) trial in Australia (ClinicalTrials.gov num-
with septic shock. Furthermore, changes during ber, NCT00975793) and the Protocolised Man-
the past decade in the care of critically ill pa- agement in Sepsis (ProMISe) trial in the United
tients, including the use of lower hemoglobin Kingdom (Current Controlled Trials number,
levels as a threshold for transfusion, the imple- ISRCTN36307479) may offer additional insight.19,20
mentation of lung-protection strategies, and the Finally, in-hospital mortality among patients
use of tighter control of blood sugar, may have requiring life support is strongly influenced by
helped lower the overall mortality and may have varying practices regarding the withdrawal of
reduced the marginal benefit of alternative re- care, which could have influenced our findings.
suscitation strategies.9,10,16,17 In summary, in our multicenter, randomized
In 2010, Jones et al. reported the results of a trial, in which patients were identified early in
randomized trial involving a patient population the emergency department as having septic
similar to ours (Table S8 in the Supplementary shock and received antibiotics and other nonre-
Appendix). That trial showed that an EGDT pro- suscitation aspects of care promptly, we found
tocol that was based on serial measurement of no significant advantage, with respect to mor-
serum lactate levels was not inferior to an EGDT tality or morbidity, of protocol-based resuscita-
protocol that used Scvo2 monitoring.14 In-hospital tion over bedside care that was provided accord-
mortality and the use of intravenous fluids, blood ing to the treating physician’s judgment. We also
transfusions, and dobutamine were similar to found no significant benefit of the mandated
those seen in the ProCESS trial. Other studies use of central venous catheterization and central
showing the benefit of EGDT in adults presenting hemodynamic monitoring in all patients.
to the emergency department with septic shock Supported by a grant (P50 GM076659) from the National Insti-
have been observational and open to potential tute of General Medical Sciences, National Institutes of Health.
confounding.18 Dr. Shapiro reports receiving consulting fees from Thermo
Fisher Scientific, fees for serving on a data and safety monitor-
There are important limitations to our study. ing board from Cumberland Pharmaceuticals, and grant support
First, although we took many steps to ensure from Thermo Fisher Scientific, Rapid Pathogen Screening,
close adherence to the resuscitation protocols, Cheetah Medical, and Astute Medical. Dr. Angus reports receiv-
ing consulting fees from MedImmune, Ferring Pharmaceuticals,
we cannot be sure that elements critical to the and Roche Diagnostics, lecture fees from Pfizer, fees for serving
success of the protocol in the study by Rivers et on a data and safety monitoring board from Eli Lilly, and grant
al. were not lost during dissemination. Second, support through his institution from Eisai. No other potential
conflict of interest relevant to this article was reported.
we enrolled patients who were recognized to be Disclosure forms provided by the authors are available with
in septic shock. Our study does not address the the full text of this article at NEJM.org.
References
1. Angus DC, Linde-Zwirble WT, Lidicker 3. Wang HE, Shapiro NI, Angus DC, resuscitation in the emergency depart-
J, Clermont G, Carcillo J, Pinsky MR. Epi- Yealy DM. National estimates of severe ment — results of a national survey. Crit
demiology of severe sepsis in the United sepsis in United States emergency depart- Care Med 2007;35:2525-32.
States: analysis of incidence, outcome, and ments. Crit Care Med 2007;35:1928-36. 6. Jones AE, Shapiro NI, Roshon M. Im-
associated costs of care. Crit Care Med 4. Rivers E, Nguyen B, Havstad S, et al. plementing early goal-directed therapy in
2001;29:1303-10. Early goal-directed therapy in the treat- the emergency setting: the challenges and
2. Angus DC, van der Poll T. Severe sep- ment of severe sepsis and septic shock. experiences of translating research inno-
sis and septic shock. N Engl J Med 2013; N Engl J Med 2001;345:1368-77. vations into clinical reality in academic
369:840-51. [Erratum, N Engl J Med 2013; 5. Carlbom DJ, Rubenfeld GD. Barriers and community settings. Acad Emerg
369:2069.] to implementing protocol-based sepsis Med 2007;14:1072-8.
1692 n engl j med 370;18 38

The New EnglandFor

reserved.
7. Reade MC, Huang DT, Bell D, et al. RL, Banks SM, Natanson C, Deans KJ. 1999;340:409-17. [Erratum, N Engl J Med
Variability in management of early severe The importance of usual care control 1999;340:1056.]
sepsis. Emerg Med J 2010;27:110-5. groups for safety monitoring and validity 17. The Acute Respiratory Distress Syn-
8. Pike F, Yealy DM, Kellum JA, et al. during critical care research. Intensive drome Network. Ventilation with lower
Protocolized Care for Early Septic Shock Care Med 2008;34:942-7. tidal volumes as compared with traditional
(ProCESS) statistical analysis plan. Crit 13. A handbook for clinical investigators tidal volumes for acute lung injury and
Care Resusc 2013;15:301-10. conducting therapeutic clinical trials the acute respiratory distress syndrome.
9. Dellinger RP, Carlet JM, Masur H, supported by CTEP, DCTD, NCI. Bethes- N Engl J Med 2000;342:1301-8.
et al. Surviving Sepsis Campaign guide- da, MD: National Cancer Institute CTEP, 18. Rivers EP, Katranji M, Jaehne KA,
lines for management of severe sepsis and 2013. et al. Early interventions in severe sepsis
septic shock. Crit Care Med 2004;32:858- 14. Jones AE, Shapiro NI, Trzeciak S, Ar- and septic shock: a review of the evidence
73. [Errata, Crit Care Med 2004;32:1448, nold RC, Claremont HA, Kline JA. Lactate one decade later. Minerva Anestesiol
2169-70.] clearance vs central venous oxygen satu- 2012;78:712-24.
10. Dellinger RP, Levy MM, Carlet JM, et ration as goals of early sepsis therapy: a 19. Huang DT, Angus DC, Barnato A, et al.
al. Surviving Sepsis Campaign: interna- randomized clinical trial. JAMA 2010; Harmonizing international trials of early
tional guidelines for management of se- 303:739-46. goal-directed resuscitation for severe
vere sepsis and septic shock: 2008. Crit 15. Charlson ME, Pompei P, Ales KL, sepsis and septic shock: methodology of
Care Med 2008;36:296-327. [Erratum, Crit MacKenzie CR. A new method of classi- ProCESS, ARISE, and ProMISe. Intensive
Care Med 2008;36:1394-6.] fying prognostic comorbidity in longitu- Care Med 2013;39:1760-75.
11. Bone RC, Balk RA, Cerra FB, et al. dinal studies: development and valida- 20. Reade MC, Delaney A, Bailey MJ, et al.
Definitions for sepsis and organ failure tion. J Chronic Dis 1987;40:373-83. Prospective meta-analysis using individu-
and guidelines for the use of innovative 16. Hébert PC, Wells G, Blajchman MA, al patient data in intensive care medicine.
therapies in sepsis. Chest 1992;101:1644- et al. A multicenter, randomized, con- Intensive Care Med 2010;36:11-21.
55. trolled clinical trial of transfusion re- Copyright © 2014 Massachusetts Medical Society.
12. Minneci PC, Eichacker PQ, Danner quirements in critical care. N Engl J Med
n engl j med 370;18 39nejm.org may 1, 2014 1693

The New
Downloaded from collections.nejm.org. England use
For personal Journal
only.ofNo
Medicine
by MIKE MULARCZYK
© MassachusettsonMedical
JanuarySociety.
15, 2015.
AllFor personal
reserved.
Copyright © 2014 Massachusetts Medical
Back to Table of Contents Society. All rights reserved.
edit or ial
The ProCESS Trial — A New Era of Sepsis Management

Craig M. Lilly, M.D.
The importance of early detection and treatment creased mortality to delays in the administration
for reducing the mortality associated with sepsis of appropriate antibiotics6 suggested that early
has been a tenet of medical training since the administration of antibiotics increased survival
middle ages, when it was noted that “. . . the in all groups of the trial. Indeed, in the ProCESS
physicians say it happens in hectic fever, that in trial, the early or facilitated recognition of sep-
the beginning of the malady it is easy to cure but tic shock, administration of intravenous antibi-
difficult to detect, but in the course of time, not otics, and other best practices were associated
having been either detected or treated in the be- with rates of survival that were higher than pro-
ginning, it becomes easy to detect but difficult jected and higher than predicted on the basis of
to cure.”1,2 The critical role of the clinician in the scores on the Acute Physiology and Chronic
early recognition of sepsis continues to this day Health Evaluation (APACHE) II,7 and a thought-
to be fundamental to our efforts to improve the ful design allowed the sample size of the trial to
rate of survival.3 Identification of the combina- be recalculated to preserve the power of the
tion of signs and symptoms that make up the sys- study to test the primary outcome. One impor-
temic inflammatory response syndrome (SIRS)4 tant contribution of the ProCESS trial is the evi-
in the context of an infection allows the astute dence it provides regarding the ongoing role of
clinician to recognize the malady. early recognition of and antibiotic treatment for
Early recognition of sepsis was incorporated sepsis in improving survival.
into the trial design, prompts, and protocols of The ProCESS trial also provides transforma-
the Protocolized Care for Early Septic Shock tive insights about the treatments for septic
(ProCESS) trial, the results of which are now re- shock that bring generalizable benefits when
ported in the Journal.5 For all the groups in the septic shock is recognized in the first hours af-
trial, the goal was early recognition of sepsis, as ter arrival in the emergency department. The use
specified in the Surviving Sepsis Campaign guide- of central hemodynamic and oxygen-saturation
lines,3 and the design called for early treatment monitoring in the protocol-based early goal-
with antimicrobial agents6 and conservative directed therapy (EGDT) group did not result in
transfusion thresholds; in addition, the patients better outcomes than those that were achieved
received low tidal-volume ventilation and had with clinical assessment of the adequacy of cir-
moderate glycemic control. culation. The finding that adjusting therapies to
Indeed, septic shock was recognized early in surrogate physiological targets measured with
a majority of the patients; 76% of the patients invasive catheters was not required to reduce
received antimicrobial agents by the time they mortality is consistent with the results of a study
underwent randomization, which occurred a that showed that serial measurement of blood
mean of approximately 3 hours after patients’ lactate levels was noninferior to catheter-derived
arrival in the emergency department. The rate of measurements8 and of analyses that have not
intravenous antimicrobial administration 6 hours found benefits of the use of pulmonary-artery
after randomization was approximately 97%, a catheters.9 State legislation and clinical guide-
finding that suggests that notification that sep- lines, including those endorsed by the National
tic shock is present encourages the administra- Quality Forum, should be updated to remove
tion of antibiotics. A study that attributed in- the requirement for central hemodynamic moni-
1750 n engl j med 370;18 40

The New EnglandFor

Society. Alluse only.
reserved.
editorials
toring and to focus on less costly, lower-risk, The ProCESS trial identifies early recognition
and equally effective alternatives. of sepsis, early administration of antibiotics,
The association of the implementation of the early adequate volume resuscitation, and clinical
multifaceted EGDT intervention with significantly assessment of the adequacy of circulation as the
lower mortality in an earlier study10 launched the elements we should focus on to save lives. The
EGDT era of sepsis management. This milestone publication of the ProCESS trial launches the era
study encouraged coordinated efforts3 to improve of early recognition and treatment in the man-
the outcomes in patients with this common11 agement of sepsis.
and life-threatening condition. These efforts Disclosure forms provided by the author are available with the
translated into the earlier identification of septic full text of this article at NEJM.org.
shock and into an increased number of patients From the Division of Pulmonary, Allergy and Critical Care Medi-
receiving earlier administration of a larger vol- cine, University of Massachusetts Medical School, UMass Me-
ume of resuscitation fluid. The ProCESS trial morial Medical Center, Worcester.
allows refinement of the EGDT approach to fluid This article was published on March 18, 2014, at NEJM.org.
administration by defining lower boundaries
that are associated with equivalent outcomes 1. Machiavelli N. Il principe. S.l. [nach Ebert vielleicht Genf];
1550.
and setting limits that are needed to avoid the 2. Idem. The prince. Ann Arbor, MI: Borders Classics, 2006.
twin problems of renal failure from too little 3. Vassalos A, Rooney K. Surviving sepsis guidelines 2012. Crit
fluid and pulmonary dysfunction from fluid over- Care Med 2013;41:e485-6.
4. Levy MM, Fink MP, Marshall JC, et al. 2001 SCCM/ESICM/
load. Another interesting and seemingly paradox- ACCP/ATS/SIS International Sepsis Definitions Conference. Crit
ical finding is that patients in whom sepsis was Care Med 2003;31:1250-6.
managed without a protocol had an outcome as 5. The ProCESS Investigators. A randomized trial of protocol-
based care for early septic shock. N Engl J Med 2014;370:1683-
good as those in patients in whom the sepsis 93.
was managed with the use of a protocol. If one 6. Kumar A, Roberts D, Wood KE, et al. Duration of hypoten-
assumes that the treatments for septic shock, as sion before initiation of effective antimicrobial therapy is the
critical determinant of survival in human septic shock. Crit Care
well as the timing of the treatments, that would Med 2006;34:1589-96.
be administered in all emergency departments, 7. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE
regardless of size or available resources, would II: a severity of disease classification system. Crit Care Med
1985;13:818-29.
be equivalent to those used in the no-protocol 8. Jones AE, Shapiro NI, Trzeciak S, Arnold RC, Claremont HA,
(usual-care) group of the ProCESS trial (which Kline JA. Lactate clearance vs central venous oxygen saturation
included strategies for early recognition of sep- as goals of early sepsis therapy: a randomized clinical trial.
JAMA 2010;303:739-46.
sis), one could come to the dubious conclusion 9. Rajaram SS, Desai NK, Kalra A, et al. Pulmonary artery cath-
that protocols and decision prompts do not have eters for adult patients in intensive care. Cochrane Database Syst
a role in the treatment of septic shock. I prefer Rev 2013;2:CD003408.
10. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed
to think differently. I believe that the prompt- therapy in the treatment of severe sepsis and septic shock. N Engl
ing, serum lactate screening and assessment of J Med 2001;345:1368-77.
SIRS criteria, and reporting of activities that 11. Angus DC, Linde-Zwirble WT, Lidicker J, Clermont G, Car-
cillo J, Pinsky MR. Epidemiology of severe sepsis in the United
were parts of the study by Rivers et al. and the States: analysis of incidence, outcome, and associated costs of
ProCESS trial can be applied in clinical practice care. Crit Care Med 2001;29:1303-10.
to ensure early diagnosis and treatment for all DOI: 10.1056/NEJMe1402564
patients with septic shock. Copyright © 2014 Massachusetts Medical Society.
Left Bundle-Branch Block Myopathy in Heart Failure

Jeffrey J. Goldberger, M.D., M.B.A.
Cardiac-resynchronization therapy (CRT), also factorial, with effects on contractile function,

known as biventricular pacing, has emerged as a β-adrenergic responsiveness, and other cellular
pivotal therapy in selected patients with heart functions.1 The dramatic clinical improvement
failure with reduced ejection fraction. The mech- that has been observed in some patients with
anism of benefit of CRT is complex and not yet New York Heart Association (NYHA) class III or
completely understood, but it is probably multi- IV heart failure2 has prompted the evaluation of

The New
Downloaded from collections.nejm.org. For England
personalJournal of No
use only. Medicine
by MIKE MULARCZYK onMedical
© Massachusetts January Society.
15, 2015.All
For personal
reserved.
c or r e sp ondence
Protocol-Based Care for Early Septic Shock

To the Editor: In the Protocolized Care for Ear- 1119.]
3. Reid F, Lobo DN, Williams RN, Rowlands BJ, Allison SP.
ly Septic Shock (ProCESS) study (May 1 issue),1 (Ab)normal saline and physiological Hartmann’s solution: a ran-
the investigators report that protocol-based early domized double-blind crossover study. Clin Sci (Lond) 2003;104:
goal-directed therapy (EGDT) did not improve 17-24.
4. Yunos NM, Bellomo R, Hegarty C, Story D, Ho L, Bailey M.
the outcome in patients with septic shock. Since Association between a chloride-liberal vs chloride-restrictive
fluid therapy is an essential component of EGDT, intravenous fluid administration strategy and kidney injury in
it would be useful to know what types of fluid critically ill adults. JAMA 2012;308:1566-72.
5. Shaw AD, Bagshaw SM, Goldstein SL, et al. Major complica-
were administered. During the first 72 hours of tions, mortality, and resource utilization after open abdominal
care, patients received about 6.5 liters of intrave- surgery: 0.9% saline compared to Plasma-Lyte. Ann Surg 2012;
nous fluids, and there was a mean increase in the 255:821-9.
serum chloride level from 100 mmol per liter to DOI: 10.1056/NEJMc1406745
between 106 and 108 mmol per liter. In a recent
study,2 a similar increase in the serum chloride To the Editor: In the ProCESS study, all three
level from 103 to 108.5 mmol per liter 60 minutes study groups had an improvement in the rate of
after the infusion of 2 liters of 0.9% sodium chlo- death (ranging from 17 to 19 percentage points),
ride in healthy participants was associated with a as compared with values predicted by their score
40% decrease in renal blood-flow velocity (as on the Acute Physiology and Chronic Health
measured in centimeters per second) and the Evaluation (APACHE) II, which shows that proto-
perfusion of renal cortical tissue. Additional evi- colized care does work. As compared with the
dence supports the adverse renal effects of hy- patients in the Early Goal-Directed Therapy
perchloremia.2-5 Since such adverse effects might study,1 the patients in the ProCESS study had a
have modified the findings, and since the ad- lower severity of illness because the initial lactate
ministration of large amounts of hyperchloremic values were lower and the inclusion criteria in-
0.9% sodium chloride is the probable cause of cluded a 1-liter fluid bolus instead of 20 to 30 ml
the reported hyperchloremia, can the authors per kilogram of body weight.2 The majority of
provide information on the types of fluid that the study sites had preexisting sepsis programs
were administered? that were influenced by the Surviving Sepsis
Hans-Joachim Priebe, M.D. Campaign, as shown by the rate of early central-
University of Freiburg catheter placement of 57% in the usual-care
Freiburg, Germany group, a procedure that has been associated with
hans-joachim.priebe@uniklinik-freiburg.de a 10% reduction in mortality.3 Even the delayed
No potential conflict of interest relevant to this letter was re- introduction of monitoring of central venous
ported.
pressure and central venous oxygen saturation
1. The ProCESS Investigators. A randomized trial of protocol- after the 6-hour avoidance period can still give
based care for early septic shock. N Engl J Med 2014;370:1683-93.
2. Chowdhury AH, Cox EF, Francis ST, Lobo DN. A random-
rise to improved outcomes.4 Moreover, the high
ized, controlled, double-blind crossover study on the effects of likelihood that the usual-care group received pre-
2-L infusions of 0.9% saline and Plasma-Lyte 148 on renal blood existing protocol-driven care, as outlined in the
flow velocity and renal cortical tissue perfusion in healthy volun-
teers. Ann Surg 2012;256:18-24. [Erratum, Ann Surg 2013;258: Surviving Sepsis Campaign, could explain the
low mortality and small between-group differ-
384 n engl j med 371;4 42

nejm.org july 24, 2014
correspondence
Table 1. Differences in Mortality and Key Clinical Values in the EGDT Study and the ProCESS Study.*
Variable EGDT Study ProCESS Study

Protocol-Based
EGDT Control EGDT Standard-Therapy Usual-Care
Group Group Group Group Group
Predicted mortality on the basis of APACHE II 40.3 36.9 38.2 37.5 37.9
score (%)
Actual mortality (%) 30.5 46.5 21.0 18.2 18.9
Lactate (mmol/liter)
At 0 hr 7.7 6.9 4.8 5.0 4.8
At 6 hr 4.3 4.9 NR NR NR
Central venous oxygen saturation (%)
At 0 hr 48.6 49.2 71.0 NA NA
At 6 hr 77.3 66.0 NR NA NA
Central-catheter rate at 6 hr (%) 100 100 93.6 56.5 57.9
* APACHE denotes Acute Physiology and Chronic Health Evaluation, NA not applicable, and NR not reported.
ences (Table 1). Does it make sense to change a mortality may have influenced the ProCESS trial
historically successful protocol that has im- results to some extent, it is important to recog-
proved patient outcomes? The recent continued nize that only half the patients in the ProCESS
endorsement of EGDT by the Surviving Sepsis population were able to be discharged home
Campaign supports the status quo.5 from the hospital, and the 1-year mortality ap-
Angel Coz Yataco, M.D. pears to be nearly double the short-term mortal-
University of Kentucky
ity. Other recent studies have found similar dis-
Lexington, KY cordance. As highlighted in the Journal,3 chronic
angel.coz@uky.edu critical illness (i.e., critically illness in patients
No potential conflict of interest relevant to this letter was re- who neither die in the acute phase nor recover) is
ported. an emerging public health problem that is both
1. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed created and sustained by advances in critical care
therapy in the treatment of severe sepsis and septic shock. N Engl medicine. How many patients in the ProCESS
J Med 2001;345:1368-77.
2. Mikkelsen ME, Miltiades AN, Gaieski DF, et al. Serum lac-
study received the diagnosis of chronic critical
tate is associated with mortality in severe sepsis independent of illness? And should the primary outcome in sep-
organ failure and shock. Crit Care Med 2009;37:1670-7. sis trials incorporate the occurrence of chronic
3. Walkey AJ, Wiener RS, Lindenauer PK. Utilization patterns
and outcomes associated with central venous catheter in septic critical illness as a “poor” outcome?
shock: a population-based study. Crit Care Med 2013;41:1450-7. Stephen Trzeciak, M.D., M.P.H.
4. Castellanos-Ortega A, Suberviola B, García-Astudillo LA,
Cooper University Hospital
Ortiz F, Llorca J, Delgado-Rodríguez M. Late compliance with
Camden, NJ
the sepsis resuscitation bundle: impact on mortality. Shock
trzeciak-stephen@cooperhealth.edu
2011;36:542-7.
5. Surviving Sepsis Campaign. Surviving Sepsis Campaign re- No potential conflict of interest relevant to this letter was re-
sponds to ProCESS trial. Updated May 19, 2014 (http://www ported.
.survivingsepsis.org/SiteCollectionDocuments/SSC-Responds
1. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed ther-
-Process-Trial.pdf).
apy in the treatment of severe sepsis and septic shock. N Engl J
DOI: 10.1056/NEJMc1406745 Med 2001;345:1368-77.
2. Kaukonen KM, Bailey M, Suzuki S, Pilcher D, Bellomo R.
Mortality related to severe sepsis and septic shock among criti-
cally ill patients in Australia and New Zealand, 2000-2012. JAMA
To the Editor: Although it is indeed good news 2014;311:1308-16.
that advances in the field have decreased short- 3. Lamas D. Chronic critical illness. N Engl J Med 2014;370:
175-7.
term sepsis-related mortality over time since the
original report of EGDT,1,2 and that this lower DOI: 10.1056/NEJMc1406745

nejm.org july 24, 2014 385
To the Editor: The ProCESS trial represents a of patients in the control group is evidence that
little-understood paradigm shift. Twenty years sites all followed the EGDT-based resuscitation
ago, when I began practicing emergency medi- guidelines of the Surviving Sepsis Campaign.
cine, the administration of antibiotics without an Central-catheter use is extremely common, espe-
identified source of infection violated accepted cially for patients admitted to the intensive care
practice. We can quibble over monitoring of cen- unit. Furthermore, as we reported, only 3.5% and
tral venous pressure and lactate clearance. The 4.0% of patients in the two control groups under-
critical change, though, is that in this “new era,” went monitoring of central venous oxygen satu-
as described in the editorial accompanying the ration,1 a prerequisite for EGDT. Coz Yataco
article on the ProCESS trial,1 we are urged to suggests that the resuscitation bundle that is
treat sepsis quickly on clinical evidence rather recommended by the Surviving Sepsis Campaign
than on delayed bacteriologic evidence. should remain intact because observational stud-
Antonio J. Dajer, M.D. ies report good outcomes. We contend that clini-
New York Presbyterian–Lower Manhattan Hospital cal guidelines should be modified as robust data
New York, NY emerge from randomized trials, and the Surviv-
tonydajer@aol.com
ing Sepsis Campaign bundles have undergone
ported. numerous changes in the past on the basis of
1. Lilly CM. The ProCESS trial — a new era of sepsis manage- just such a process.4
ment. N Engl J Med 2014;370:1750-1. We agree with Trzeciak that enthusiasm for
DOI: 10.1056/NEJMc1406745 the decline in hospital mortality from sepsis must
be tempered by concern that many patients who
The authors reply: We agree with Priebe that are discharged may die in the following months,
saline can cause hyperchloremia and acidosis. In as we reported, or suffer protracted sequelae.
our study, we did not evaluate different fluid for- That said, in our study, only 45 of 1341 patients
mulations. Saline comprised 93% of intravenous (3.4%) were still undergoing mechanical ventila-
fluids (range, 92 to 96% in the three study tion at the time of discharge. We agree that as-
groups) during the first 6 hours and 83% (range, sessing outcomes beyond short-term mortality
80 to 86%) from 6 hours to 72 hours, which we are key considerations for future sepsis trials.
consider within the scope of usual U.S. practice.1 Derek C. Angus, M.D., M.P.H.
We also agree with Dajer regarding the impor- Donald M. Yealy, M.D.
tance of the early use of antibiotics in patients John A. Kellum, M.D.
who appear to be sick and infected, such as pa- University of Pittsburgh
tients enrolled in our study, although use of these Pittsburgh, PA
drugs in less sick patients may enhance antibi- angusdc@upmc.edu
otic resistance. for the ProCESS Investigators
Modern studies typically report outcomes Since publication of their article, the authors report no further
better than those predicted by the APACHE II potential conflict of interest.
score, presumably because of the many advances 1. Jones D, McEvoy S, Merz TM, et al. International albumin use:
in care since the original APACHE II calibration 1995 to 2006. Anaesth Intensive Care 2010;38:266-73.
2. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE
30 years ago.2 Coz Yataco notes that our cohort II: a severity of disease classification system. Crit Care Med
had some features that suggested a lower severity 1985;13:818-29.
of illness than that in the study by Rivers et al.3 3. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed
therapy in the treatment of severe sepsis and septic shock. N Engl
However, in our reported subgroup analyses, the J Med 2001;345:1368-77.
sickest third of patients on the basis of lactate 4. Dellinger RP, Levy MM, Rhodes A, et al. Surviving Sepsis
levels or APACHE II scores, who were sicker than Campaign: international guidelines for management of severe
sepsis and septic shock: 2012. Crit Care Med 2013;41:580-637.
patients in the cohort study by Rivers et al.,
DOI: 10.1056/NEJMc1406745
showed no benefit from EGDT. Thus, we do not
believe that differences in severity of illness explain
the differences in results between the two trials. The editorialist replies: Dajer astutely points
We disagree that central-catheter use in 57% out that the one key element that the ProCESS
386 n engl j med 371;4 44

correspondence
trial procedures shared with those in the preced- performing investigations before starting anti-
ing EGDT trial1 was that antimicrobials were ad- biotic therapy.
ministered within the first 2 hours after sepsis Craig M. Lilly, M.D.
was identified in circulatory failure. The practice University of Massachusetts Medical School
of deferring antimicrobial therapy in favor of a Worcester, MA
prolonged search or unassailable evidence of the craig.lilly@umassmed.edu
source of the infection has been difficult to jus- Since publication of his article, the author reports no further
potential conflict of interest.
tify given studies in patients with septic shock
that documented mortality benefits from the ad- 1. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed
ministration of antimicrobials to which the patho- therapy in the treatment of severe sepsis and septic shock. N Engl
J Med 2001;345:1368-77.
genic organism was sensitive,2 better outcomes 2. Kumar A, Ellis P, Arabi Y, et al. Initiation of inappropriate
with combination therapy as compared with antimicrobial therapy results in a fivefold reduction of survival
monotherapy therapy,3 and particularly the real- in human septic shock. Chest 2009;136:1237-48.
Kumar A, Zarychanski R, Light B, et al. Early combination
ization that each hour that antimicrobial therapy 3.antibiotic therapy yields improved survival compared with
is deferred has been associated with a 7.6% de- monotherapy in septic shock: a propensity-matched analysis.
crease in survival for patients with septic shock.4 Crit Care Med 2010;38:1773-85.
Kumar A, Roberts D, Wood KE, et al. Duration of hypoten-
The training we received to identify the source of 4.
sion before initiation of effective antimicrobial therapy is the
any serious infection is as valid today as it was critical determinant of survival in human septic shock. Crit Care
when it was first brought to our attention; what Med 2006;34:1589-96.
has changed is the amount of time allotted for DOI: 10.1056/NEJMc1406745

original article
Goal-Directed Resuscitation for Patients

with Early Septic Shock
The ARISE Investigators and the ANZICS Clinical Trials Group*
A bs t r ac t
Background
The members of the writing committee Early goal-directed therapy (EGDT) has been endorsed in the guidelines of the Sur-
(Sandra L. Peake, M.D., Ph.D., Anthony viving Sepsis Campaign as a key strategy to decrease mortality among patients pre-
Delaney, M.D., Ph.D., Michael Bailey,
Ph.D., Rinaldo Bellomo, M.D., Peter A. senting to the emergency department with septic shock. However, its effectiveness
Cameron, M.D., D. James Cooper, M.D., is uncertain.
Alisa M. Higgins, M.P.H., Anna Hold-
gate, M.D., Belinda D. Howe, M.P.H.,
Steven A.R. Webb, M.D., Ph.D., and Patri- Methods
cia Williams, B.N.) assume responsibility In this trial conducted at 51 centers (mostly in Australia or New Zealand), we randomly
for the overall content and integrity of the assigned patients presenting to the emergency department with early septic shock
article. Address reprint requests to Ms.
Belinda Howe at the Australian and New to receive either EGDT or usual care. The primary outcome was all-cause mortality
Zealand Intensive Care Research Centre, within 90 days after randomization.
Alfred Centre, Level 6 (Lobby B), 99 Com-
mercial Rd., Melbourne, VIC 3004, Aus-
tralia, or at anzicrc@monash.edu. Results
Of the 1600 enrolled patients, 796 were assigned to the EGDT group and 804 to the
*The Australasian Resuscitation in Sep- usual-care group. Primary outcome data were available for more than 99% of the
sis Evaluation (ARISE) study is a col-
laboration of the Australian and New patients. Patients in the EGDT group received a larger mean (±SD) volume of intra-
Zealand Intensive Care Society (ANZICS) venous fluids in the first 6 hours after randomization than did those in the usual-
Clinical Trials Group, the Australasian care group (1964±1415 ml vs. 1713±1401 ml) and were more likely to receive vaso-
College for Emergency Medicine, and
the Australian and New Zealand Inten- pressor infusions (66.6% vs. 57.8%), red-cell transfusions (13.6% vs. 7.0%), and
sive Care Research Centre. The affilia- dobutamine (15.4% vs. 2.6%) (P<0.001 for all comparisons). At 90 days after ran-
tions of the writing committee members domization, 147 deaths had occurred in the EGDT group and 150 had occurred in
are listed in the Appendix. A complete
list of investigators in the ARISE study the usual-care group, for rates of death of 18.6% and 18.8%, respectively (absolute
is provided in the Supplementary Ap- risk difference with EGDT vs. usual care, −0.3 percentage points; 95% confidence
pendix, available at NEJM.org. interval, −4.1 to 3.6; P = 0.90). There was no significant difference in survival time,
This article was published on October 1, in-hospital mortality, duration of organ support, or length of hospital stay.
2014, at NEJM.org.
Conclusions
The Supplementary Appendix for this
article is available at NEJM.org (http:// In critically ill patients presenting to the emergency department with early septic
www.nejm.org/stoken/default shock, EGDT did not reduce all-cause mortality at 90 days. (Funded by the National
+domain/ClinicalCollections-PDF/full? Health and Medical Research Council of Australia and the Alfred Foundation; ARISE
redirectUri=/doi/suppl/10.1056/
NEJMoa1404380/suppl_file/ ClinicalTrials.gov number, NCT00975793.)
nejmoa1404380_appendix.pdf)
N Engl J Med 2014;371:1496-506.

DOI: 10.1056/NEJMoa1404380
1496 n engl j med 371;1646 nejm.org october 16, 2014
The New England

Downloaded from collections.nejm.org. Journaluse
For personal of only.
Medicine
reserved.
Goal-Directed Resuscitation in Early Septic Shock
S
evere sepsis has a reported annual Me thods
incidence in adults of up to 300 cases per
100,000 population.1-3 Despite decreasing Study Design and Oversight
mortality from sepsis in recent years,4 the risk of From October 5, 2008, to April 23, 2014, we con-
death remains high.5,6 The fundamental princi- ducted this prospective, randomized, parallel-
ples for the management of sepsis include early group trial in 51 tertiary care and nontertiary
recognition, control of the source of infection, care metropolitan and rural hospitals. Most cen-
appropriate and timely administration of antimi- ters were in Australia or New Zealand, with 6 cen-
crobial drugs, and resuscitation with intravenous ters in Finland, Hong Kong, and the Republic of
fluids and vasoactive drugs. Ireland (Table S1 in the Supplementary Appen-
Patients presenting to the emergency depart- dix, available with the full text of this article at
ment account for a large proportion of patients NEJM.org).24 Participating institutions did not
with severe sepsis.7 Reported in-hospital mortality have sepsis-resuscitation protocols at the time
ranges in this subgroup from 20 to 50%.3,8-10 In of site selection, and usual care did not include
2001, a proof-of-concept, randomized trial showed resuscitation guided by measurement of the
that early hemodynamic resuscitation according central venous oxygen saturation (Scvo2).25 The
to a specific protocol termed early goal-directed ARISE study was one of three collaborative, har-
therapy (EGDT) improved outcomes in patients monized studies, along with the ProCESS trial10
presenting to the emergency department with and the Protocolized Management in Sepsis
severe sepsis, as compared with usual therapy.11 (ProMISe) trial (Current Controlled Trials num-
EGDT was subsequently incorporated into the ber, ISRCTN36307479), designed to address the
6-hour resuscitation bundle of the Surviving Sepsis effectiveness of EGDT.24
Campaign guidelines,12-14 and a number of non- The study protocol was approved by the ethics
randomized studies showed a survival benefit with committee at Monash University, which was the
bundle-based care that included EGDT.15-18 De- coordinating center, and at each participating in-
spite such successes, considerable controversy has stitution. The protocol and statistical analysis
surrounded the role of EGDT in the treatment of plan are available at NEJM.org. Prior informed
patients with severe sepsis. Concerns have included written consent or delayed consent was obtained
the potential risks associated with individual ele- from all patients or their legal surrogates. The
ments of the protocol,19,20 uncertainty about the trial was overseen by an independent data and
external validity of the original trial, and the in- safety monitoring committee. Scvo2 monitors
frastructure and resource requirements for imple- were loaned to participating sites by Edwards
menting EGDT.21,22 Lifesciences, which had no other role in the con-
In a randomized trial conducted in 31 academic duct of the study.
centers in the United States (Protocolized Care
for Early Septic Shock [ProCESS]),10 protocol- Study Population
based resuscitation (a combination of EGDT and Patients 18 years of age or older who met the eli-
protocol-based standard therapy) was not associ- gibility criteria within 6 hours after presentation
ated with a survival benefit, as compared with to the emergency department were assessed for
usual care that was not protocol-based. Whether enrollment. Eligibility criteria were a suspected
these results would hold up outside the United or confirmed infection, two or more criteria for
States and across a variety of academic and non- a systemic inflammatory response26 (see the
academic health care settings is unknown; more Methods section in the Supplementary Appen-
evidence is needed to provide clinical direction.23 dix), and evidence of refractory hypotension or
We designed the multicenter Australasian Re- hypoperfusion. Refractory hypotension was de-
suscitation in Sepsis Evaluation (ARISE) study to fined as a systolic blood pressure of less than
test the hypothesis that EGDT, as compared with 90 mm Hg or a mean arterial pressure of less than
usual care, would decrease 90-day all-cause mor- 65 mm Hg after an intravenous fluid challenge of
tality among patients presenting to the emergency 1000 ml or more administered within a 60-min-
department with early septic shock in diverse ute period. Hypoperfusion was defined as a blood
health care settings. lactate level of 4.0 mmol per liter or more. Ran-
n engl j med 371;16 nejm.org

47 october 16, 2014 1497
The New
For personal of No
use only. Medicine
22, 2015.All
For personal
reserved.
domization was required within 2 hours after ful- Study Outcomes

fillment of the final inclusion criterion. The ini- The primary study outcome was death from any
tiation of the first dose of intravenous cause within 90 days after randomization. Sec-
antimicrobial therapy was mandated before ran- ondary and tertiary outcomes included survival
domization. The study exclusion criteria are pro- time from randomization to 90 days; mortality in
vided in the Methods section in the Supplemen- the ICU; mortality at 28 days; in-hospital mortal-
tary Appendix. ity at 60 days; cause-specific mortality at 90 days27;
length of stay in the emergency department, ICU,
Randomization or elsewhere in the hospital; receipt and duration
Eligible patients were randomly assigned in a of mechanical ventilation, vasopressor support,
1:1 ratio to receive either EGDT or usual care for or renal-replacement therapy; destination at the
6 hours. Randomization was stratified accord- time of discharge (for surviving inpatients); limi-
ing to study center with the use of a permuted- tation of therapy (e.g., do-not-resuscitate order)
block method and was performed by means of a at the time of death (for nonsurvivors); and ad-
centralized telephone interactive voice-response verse events.
system that was accessible 24 hours a day. Be-
cause of the nature of the intervention, all pa- Statistical Analysis
tients and clinicians involved in their care were All analyses were conducted according to a sta-
aware of study-group assignments. tistical analysis plan that was reported previous-
ly.28 The sample-size calculation was based on an
Study Treatments assumed in-hospital rate of death in the usual-care
For patients in the usual-care group, decisions group of 28%,25,29 with an increment of 10 per-
about the location of care delivery, investiga- centage points (38%) for the rate of death at 90
tions, monitoring, and all treatments were made days.8,30 Thus, an enrollment of 1600 patients
by the treating clinical team. Scvo2 measurement would have a power of 85 to 90% (at a two-sided
was not permitted during the 6-hour intervention alpha level of 0.05) to detect an absolute risk re-
period. Data were collected regarding insertion duction of 7.6 percentage points (or a relative risk
of invasive monitoring devices, intravenous-fluid reduction of 20%) in the EGDT group, with al-
resuscitation, vasoactive support, red-cell trans- lowance for a plausible range of loss to follow-up.
fusion, mechanical ventilation, and other support- One interim analysis was planned and performed
ive therapy. after the enrollment of 50% of the patients, with
For patients in the EGDT group, the inter- the use of a two-sided, symmetric O’Brien–Flem-
vention was provided by a study team trained in ing design and a two-sided P value of 0.005; this
EGDT delivery. Both the care providers and lo- analysis was reviewed by the independent data
cation of delivery were dependent on local re- and safety monitoring committee.
sources. Thus, investigators used EGDT imple- All analyses were conducted according to the
mentation models based in the emergency intention-to-treat principle. No assumptions were
department, the intensive care unit (ICU), or made for missing or unavailable data. We report
both. A multifaceted intervention was used to continuous variables as means (±SD) or medians
standardize EGDT delivery across sites. De- 24 and interquartile ranges, and categorical variables
tails of EGDT implementation, personnel, and as proportions. We used Student’s t-test or the
location are provided in Table S1 and Figure S1 Wilcoxon rank-sum test to analyze between-group
in the Supplementary Appendix. differences, as appropriate. Fisher’s exact test was
In the EGDT group, an arterial catheter and used for categorical variables, including the pri-
a central venous catheter capable of continuous mary outcome. Absolute and relative risk differ-
Scvo2 measurement (Edwards Lifesciences) were ences with 95% confidence intervals for all-cause
inserted within 1 hour after randomization. The mortality at 90 days are reported. Additional sensi-
resuscitation algorithm was based on the original tivity analyses were performed with the use of
11
EGDT algorithm and was followed until 6 hours multivariable logistic regression adjusted for pre-
after randomization (Fig. S1 in the Supplemen- defined baseline covariates: country, age, score
tary Appendix). on the Acute Physiology and Chronic Health
1498 n engl j med 371;16 48

nejm.org october 16, 2014
The New EnglandFor

reserved.
Evaluation II (APACHE II), systolic blood pressure baseline were similar in the two groups (Table 1,
(<90 mm Hg or ≥90 mm Hg), and presence or and Tables S2 and S3 in the Supplementary Ap-
absence of invasive mechanical ventilation. We pendix).31,32 The criterion for refractory hypoten-
used the Kaplan–Meier method to calculate sur- sion was met by 555 patients (70.0%) in the EGDT
vival time from randomization to 90 days and group and 557 (69.8%) in the usual-case group.
the log-rank test to perform between-group com- The criterion for an elevated lactate level was
parisons. We used Cox proportional-hazards mod- met by 365 patients (46.0%) in the EGDT group
els adjusted for the previously specified baseline and 371 (46.5%) in the usual-care group (Table 1).
covariates to calculate hazard ratios with 95% There was no significant difference in the mean
confidence intervals. Values for length of hospi- intravenous fluid volume that had been infused
tal stay and duration of organ support were log- at baseline, with 2515±1244 ml (34.6±19.4 ml
transformed and analyzed with the use of linear per kilogram) in the EGDT group and 2591±1331
regression and are reported as ratios with 95% ml (34.7±20.1 ml per kilogram) in the usual-care
confidence intervals. group. The median time from presentation to the
We conducted subgroup analyses for the pri- emergency department until randomization was
mary outcome for predefined variables: country, 2.8 hours (interquartile range, 2.1 to 3.9) in the
age (<65 years or ≥65 years), APACHE II score EDGT group and 2.7 hours (interquartile range,
(<25 or ≥25), presence or absence of invasive me- 2.0 to 3.9) in the usual-care group.
chanical ventilation, presence or absence or re-
fractory hypotension, lactate level (<4.0 mmol Microbiologic Data
per liter or ≥4.0 mmol per liter), and intravenous The median time between presentation to the
fluid administration (<20 ml per kilogram of body emergency department and administration of the
weight or ≥20 ml per kilogram). Subgroup analy- first dose of intravenous antimicrobial therapy
ses were performed with the use of logistic re- was similar in the two groups: 70 minutes (inter-
gression, with heterogeneity determined on the quartile range, 38 to 114) in the EGDT group and
basis of interaction between treatment and sub- 67 minutes (interquartile range, 39 to 110) in the
group. Odds ratios with 95% confidence intervals usual-care group. The lungs and urinary tract
for death at 90 days are presented in a forest plot. were the most common locations of infection,
All analyses were performed with the use of and blood cultures were positive in 38% of pa-
SAS software, version 9.3 (SAS Institute). A two- tients in each study group. The numbers of patients
sided P value of 0.05 or less was considered to receiving treatment to control the source of in-
indicate statistical significance, except for the pri- fection up to 72 hours after randomization were
mary outcome, for which a P value of 0.0491 or 78 (9.8%) in the EGDT group and 97 (12.2%) in
less was used. the usual-care group (P = 0.14). Detailed micro-
biologic data are presented in Table S4 in the Sup-
R e sult s plementary Appendix.
Study Patients Interventions and Therapies

We enrolled 1600 patients, with 796 assigned to Patients who were admitted directly from the emer-
the EGDT group and 804 to the usual-care group gency department to the ICU numbered 690 (87.0%)
(Fig. 1). Delayed consent was refused for 9 patients in the EGDT group and 614 (76.9%) in the usual-
(3 in the EGDT group and 6 in the usual-care case group (P<0.001). A central venous catheter for
group), leaving an intention-to-treat population continuous monitoring of the Scvo2 was inserted
of 793 patients and 798 patients, respectively. By during the first 6 hours after randomization in 714
day 90, 1 patient in the usual-care group had re- patients (90.0%) in the EGDT group. The median
voked consent, and 2 patients (1 in each group) time to insertion was 1.1 hours (interquartile range,
were lost to follow-up, leaving a final cohort of 0.7 to 1.6), and the mean Scvo2 was 72.7±10.5%. A
1588 patients for whom the primary outcome was central venous catheter was inserted during the
available: 792 (99.5%) in the EGDT group and first 6 hours in 494 patients (61.9%) in the usual-
796 (99.0%) in the usual-care group. care group. The median time to insertion was 1.2
Demographic and clinical characteristics at hours (interquartile range, 0.4 to 2.6). No patients

nejm.org october 16, 2014 1499
The New
only.ofNo
Medicine
by MIKE MULARCZYK
JanuarySociety.
22, 2015.
AllFor personal
reserved.
3559 Patients met all study inclusion criteria
1959 Were excluded

769 Met ≥1 exclusion criteria
23 Were <18 yr of age
72 Had contraindication to superior vena cava
CVC insertion
6 Had contraindication to blood products
5 Had hemodynamic instability due to active
bleeding
71 Had life expectancy <90 days
43 Were expected to die imminently
329 Had documented limitation of therapy order
or had treating physician who deemed aggres-
sive care unsuitable
116 Had in-patient transfer from another acute
care facility
8 Were confirmed or suspected to be pregnant
96 Were not able to start EGDT within 1 hr after
randomization or to complete 6 hr of EGDT
1190 Were eligible but did not undergo randomization
515 Were outside randomization window
282 Did not have access to a study-team member
18 Were previously recruited into study
39 Were unable to give consent
274 Declined to give consent
62 Had other reasons
796 Were assigned to receive EGDT 804 Were assigned to receive usual care
8 Were excluded
4 Were excluded
1 Was lost to follow-up
1 Was lost to follow-up
6 Declined to provide
3 Declined to provide
delayed consent
delayed consent
1 Withdrew prior consent
792 Were included in intention-to-treat 796 Were included in intention-to-treat

analysis for primary outcome analysis for primary outcome
Figure 1. Enrollment and Outcomes.

CVC denotes central venous catheter, and EGDT early goal-directed therapy.
in the usual-care group received continuous Scvo2 parisons) (Table S5 in the Supplementary Appen-
monitoring during the first 6 hours. dix). Between 6 and 72 hours, the proportion of
The volume of intravenous fluids adminis- patients receiving vasopressor infusions was high-
tered during the first 6 hours was greater in the er in the EGDT group than in the usual-care group
EGDT group than in the usual-care group (58.8% vs. 51.5%, P = 0.004), as was the proportion
(1964±1415 ml vs. 1713±1401 ml, P<0.001) (Ta- of patients receiving dobutamine (9.5% vs. 5.0%,
ble S5 in the Supplementary Appendix). More P<0.001) (Table S5 in the Supplementary Ap-
patients in the EGDT group than in the usual-care pendix).
group received a vasopressor infusion (66.6% vs. EGDT was stopped prematurely in 18 pa-
57.8%), red-cell transfusion (13.6% vs. 7.0%), or tients (2.3%). The median time to cessation was
dobutamine (15.4% vs. 2.6%) (P<0.001 for all com- 3.5 hours (interquartile range, 1.2 to 5.6). The
1500 n engl j med 371;16 50

The New England

For of Medicine
Medical personal
Society. use only.
reserved.
most common reasons were withdrawal of

therapy (5 patients), transfer to the operating
room (2 patients), and interhospital transfer EGDT Usual Care
(3 patients). Characteristic (N = 793) (N = 798)
Age — yr 62.7±16.4 63.1±16.5
Physiological and Laboratory Values Male sex — no. (%) 477 (60.2) 473 (59.3)
At the end of the 6-hour intervention period, the Usual residence — no. (%)
mean arterial pressure was higher in the EGDT Home 749 (94.5) 759 (95.1)
group than in the usual-care group (76.5±10.8 Long-term care facility 44 (5.5) 39 (4.9)
mm Hg vs. 75.3±11.4 mm Hg, P = 0.04). Other Median score on Charlson comorbidity 1 (0–2) 1 (0–2)
physiological and laboratory values were similar index (IQR)†
in the two groups (Fig. S3 and Table S6 in the APACHE II score‡ 15.4±6.5 15.8±6.5
Supplementary Appendix). The proportions of Mechanical ventilation — no. (%)
patients in the EGDT group for whom the indi-
Invasive 71 (9.0) 64 (8.0)
vidual resuscitation goals were achieved at 6 hours
Noninvasive 60 (7.6) 48 (6.0)
or for whom the relevant therapy was delivered
when a goal was not achieved were 99.6% for Vasopressor infusion — no. (%)§ 173 (21.8) 173 (21.7)
saturation of peripheral oxygen, 88.9% for central Total intravenous fluids¶
venous pressure, 94.1% for mean arterial pressure, Volume — ml 2515±1244 2591±1331
and 95.3% for Scvo2 (Fig. S4 in the Supplementary Volume per weight — ml/kg 34.6±19.4 34.7±20.1
Appendix). At 72 hours after randomization, phys- Inclusion criteria‖
iological and laboratory values were similar in Refractory hypotension — no. (%) 555 (70.0) 557 (69.8)
the two groups (Table S6 in the Supplementary
Systolic blood pressure — mm Hg 78.8±9.3 79.6±8.4
Appendix).
Lactate
Primary Outcome ≥4.0 mmol/liter — no. (%) 365 (46.0) 371 (46.5)
By 90 days after randomization, the primary out- Value at time that criterion was 6.7±3.3 6.6±2.8
met — mmol/liter
come (death from any cause) had occurred in 147
Median interval after presentation to
of 792 patients (18.6%) in the EGDT group and emergency department
150 of 796 patients (18.8%) in the usual-care group (IQR) — hr
(P = 0.90) (Table 2, and Table S7 in the Supple- Until final inclusion criterion was 1.4 (0.6–2.5) 1.3 (0.5–2.4)
mentary Appendix). The absolute difference in the met
risk of death for the EGDT group as compared with Until randomization 2.8 (2.1–3.9) 2.7 (2.0–3.9)
the usual care group was −0.3 percentage points
(95% confidence interval [CI], −4.1 to 3.6). The sur- * Plus–minus values are means ±SD. There were no significant differences in
baseline characteristics between the two study groups. EGDT denotes early
vival time did not differ significantly between goal-directed therapy, and IQR interquartile range.
the groups (Fig. 2A). Between-group mortality † Scores on the Charlson comorbidity index range from 0 to 33, with higher
was similar in all the predefined subgroups (Fig. scores indicating a greater burden of disease.
‡ A severity-of-illness score that was based on the Acute Physiology and Chronic
2B). There were no significant between-group dif- Health Evaluation II (APACHE II) variables with the use of data that were re-
ferences in 90-day mortality with the use of mul- corded closest to, but prior to, randomization was calculated to assess base-
tivariable logistic regression and Cox proportion- line equivalence. Scores on the APACHE II range from 0 to 71, with higher
scores indicating more severe disease and a higher risk of death.
al-hazards analysis after adjustment for the § Vasopressor infusions included one or more of the following agents at any
prespecified baseline covariates (Table S8 in the dose for at least 30 minutes within 1 hour before randomization: norepineph-
Supplementary Appendix). rine, epinephrine, dopamine, metaraminol, and phenylephrine.
¶ Total intravenous fluids include fluids administered before arrival at the hos-
pital and during the interval between presentation to the emergency depart-
Secondary and Tertiary Outcomes ment and randomization.
The median length of stay in the emergency de- ‖ Data on systolic blood pressure and lactate are provided only for patients who
met the inclusion criterion for refractory hypotension (a systolic blood pres-
partment after randomization was shorter in sure of <90 mm Hg or a mean arterial pressure of <65 mm Hg after an intra-
the EGDT group than in the usual-care group venous fluid challenge of 1000 ml or more administered within a 60-minute
(1.4 hours [interquartile range, 0.5 to 2.7] vs. period) or the inclusion criterion for hyperlactatemia (a lactate level of ≥4.0
mmol per liter). These values were recorded at the time that the inclusion cri-
2.0 hours [interquartile range, 1.0 to 3.8], P<0.001) terion was met.
(Table 2). Overall, more patients in the EGDT

october 16, 2014 1501
The New England
Medical personal
Society. use only.
reserved.
1502
Table 2. Study Outcomes.
EGDT Usual Care Relative Risk Risk Difference

Variable (N = 793) (N = 798) (95% CI) (95% CI)* P Value
percentage points
Primary outcome: death by day 90 — no./total no. (%) 147/792 (18.6) 150/796 (18.8) 0.98 (0.80 to 1.21) −0.3 (−4.1 to 3.6) 0.90
Secondary outcomes
The
Median duration of stay (IQR)†

Emergency department — hr 1.4 (0.5–2.7) 2.0 (1.0–3.8) <0.001
ICU — days 2.8 (1.4–5.1) 2.8 (1.5–5.7) 0.81
Hospital — days 8.2 (4.9–16.7) 8.5 (4.9–16.5) 0.89
Use and duration of organ support‡
n engl j med 371;16

Invasive mechanical ventilation — no./total no. (%) 238/793 (30.0) 251/798 (31.5) 0.95 (0.82 to 1.11) −1.4 (−6.0 to 3.1) 0.52

Median duration of invasive mechanical ventilation (IQR) — hr 62.2 (23.5–181.8) 65.5 (23.0–157.9) 0.28
nejm.org
Vasopressor support — no./total no. (%) 605/793 (76.3) 525/798 (65.8) 1.16 (1.09 to 1.24) 10.5 (6.1 to 14.9) <0.001
Median duration of vasopressor support (IQR) — hr 29.4 (12.9–61.0) 34.2 (14.0–67.0) 0.24
52of Medicine

of
Renal-replacement therapy — no./total no. (%) 106/793 (13.4) 108/798 (13.5) 0.99 (0.77 to 1.27) −0.2 (−3.5 to 3.2) 0.94
Median duration of renal-replacement therapy (IQR) — hr§ 57.8 (25.3–175.0) 85.9 (29.3–182.9) 0.40
Tertiary outcomes — no./total no. (%)
october 16, 2014

Death by day 28 117/792 (14.8) 127/797 (15.9) 0.93 (0.73 to 1.17) −1.2 (−4.7 to 2.4) 0.53
m e dic i n e

Death by the time of discharge from ICU 79/725 (10.9) 85/661 (12.9) 0.85 (0.64 to 1.13) −2.0 (−5.4 to 1.5) 0.28
Death by the time of discharge from hospital¶ 115/793 (14.5) 125/797 (15.7) 0.92 (0.73 to 1.17) −1.2 (−4.7 to 2.3) 0.53
* Risk differences of less than 1.0 indicate better results in the EGDT group.
† The duration of stay was calculated from the time of randomization, except for the stay in the intensive care unit (ICU), which was calculated from the time of ICU admission.
‡ The duration of organ support was calculated from the time of randomization.
§ Data for renal-replacement therapy were censored at 90 days after randomization.
¶ Data for mortality at the time of hospital discharge (for the index admission) were censored at 60 days after randomization.
Downloaded from nejm.org by MIKE MULARCZYK on January 22, 2015. For personal use only. No other uses without permission.
A Survival
1.00 EGDT
Usual care
Probability of Survival
0.75
0.50
0.25
0.00
0 30 60 90
Days since Randomization
No. at Risk
EGDT 792 677 660 646
Usual care 796 670 657 646
B Subgroup Analyses
P Value for
Subgroup EGDT Usual Care Odds Ratio (95% CI) P Value Interaction
no. of events/total no. (%)
Overall 147/792 (18.6) 150/796 (18.8)
Country 0.38
Australia 126/677 (18.6) 132/679 (19.4) 0.95 (0.72–1.24) 0.70
New Zealand 13/67 (19.4) 8/69 (11.6) 1.84 (0.71–4.76) 0.21
Other 8/48 (16.7) 10/48 (20.8) 0.76 (0.27–2.13) 0.60
Age 0.15
<65 yr 61/393 (15.5) 50/387 (12.9) 1.24 (0.83–1.85) 0.30
≥65 yr 86/399 (21.6) 100/409 (24.4) 0.85 (0.61–1.18) 0.33
APACHE II 0.98
<25 114/720 (15.8) 114/718 (15.9) 1.00 (0.75–1.32) 0.98
≥25 33/72 (45.8) 36/78 (46.2) 0.99 (0.52–1.88) 0.97
Invasive mechanical ventilation 0.25
Yes 19/71 (26.8) 23/64 (35.9) 0.65 (0.31–1.36) 0.25
No 128/721 (17.8) 127/732 (17.3) 1.03 (0.78–1.35) 0.84
Refractory hypotension 0.50
Yes 90/554 (16.2) 97/557 (17.4) 0.92 (0.67–1.26) 0.60
No 57/238 (23.9) 53/239 (22.2) 1.11 (0.72–1.69) 0.65
Hypofusion 0.27
Yes 99/365 (27.1) 93/369 (25.2) 1.10 (0.79–1.54) 0.55
No 48/427 (11.2) 57/427 (13.3) 0.82 (0.55–1.24) 0.35
IV fluid volume before 0.41
randomization
≥20 ml/kg 106/574 (18.5) 104/572 (18.2) 1.02 (0.76–1.37) 0.90
<20 ml/kg 28/181 (15.5) 35/181 (19.3) 0.76 (0.44–1.32) 0.33
0.01 0.1 1.0 10 100
EGDT Better Usual Care Better
Figure 2. Probability of Survival and Subgroup Analyses of the Risk of Death at 90 Days.
Panel A shows Kaplan–Meier estimates of the probability of death at 90 days for patients with septic shock receiving either early goal-
directed therapy (EGDT) or usual care for 6 hours (P = 0.82 by the log-rank test for the between-group difference). Panel B shows the
odds ratio for death at 90 days in the EGDT group, as compared with the usual-care group, among all patients and in predefined sub-
groups. The size of the squares representing odds ratios corresponds to the relative size of the subgroup. The horizontal bars represent
95% confidence intervals. Scores on the Acute Physiology and Chronic Health Evaluation II (APACHE II) range from 0 to 71, with higher
scores indicating more severe disease and a higher risk of death. IV denotes intravenous.
n engl j med 371;16 nejm.org october

53 16, 2014 1503
The New England
Medical personal
Society. use only.
reserved.
group than in the usual-care group received a participating sites were representative of all re-
vasopressor infusion (76.3% vs. 65.8%, P<0.001), gions across Australia and New Zealand, includ-
but the median duration of the infusion did not ing metropolitan and rural centers, with a mix
differ significantly between the two groups (29.4 of EGDT implementation models based in the
hours [interquartile range, 12.9 to 61.0] and 34.2 emergency department, ICU, or both.
hours [interquartile range, 14.0 to 67.0], respec- The rate of death in our study was lower than
tively; P = 0.24). There were no other significant that reported in the original EGDT trial.11 This
between-group differences in secondary or ter- finding is consistent with data showing that in-
tiary outcomes. Subsidiary analyses of secondary hospital mortality for patients who are admitted
and tertiary variables after adjustment for pre- to ICUs with severe sepsis and septic shock has
defined covariates did not alter any of the re- been reduced by 1 percentage point per year dur-
ported findings (Table S8 in the Supplementary ing the past two decades, with the decline be-
Appendix). ginning before the introduction of the Surviving
Sepsis Campaign.4,8,33 Although our study had
Adverse Events entry criteria similar to those in the ProCESS study
There was no significant between-group differ- and the original EGDT trial, it is possible that the
ence in the number of patients with one or more patients in our study had a reduced risk of death
adverse events: 56 patients (7.1%) in the EGDT because of low rates of chronic disease and better
group and 42 patients (5.3%) in the usual-care functional status, as evidenced by the low propor-
group (P = 0.15). A breakdown of specific adverse tion of nursing home residents before random-
events is presented in Table S9 in the Supplemen- ization. Nonetheless, the number of patients
tary Appendix. with septic shock at the time of enrollment was
high, indicating that the target population was
Discussion enrolled. The high number of patients who were
discharged home may also support the small in-
In this randomized trial conducted in a variety of crement in mortality between hospital discharge
health care settings, we found that EGDT, as and 90 days. There was no trend suggesting an
compared with usual care, did not reduce the pri- effect of EGDT in any unadjusted or adjusted
mary outcome of 90-day all-cause mortality, ei- estimates of mortality, and subgroup analyses
ther overall or in any of the prespecified sub- did not indicate that the benefit from EGDT in-
groups, among patients with early septic shock creased with the severity of illness. Although
who presented to the emergency department. contamination of the usual-care group by the
There were also no significant differences in 28- incorporation of some elements of the EGDT
day or in-hospital mortality, duration of organ protocol into usual care may have biased the
support, or length of hospital stay. study results, significant differences in EGDT-
Adherence to the algorithm-directed therapies specific treatments that were administered in the
was very high, and the potentially confounding two groups and the similarity between therapies
effect of the time to the administration of anti- administered in the usual-care group in this
microbial drugs was addressed by the requirement study and those in our pretrial observational
that such drugs be administered before random- study25 indicate that such an effect in the usual-
ization. In addition, the loss to follow-up was care group is unlikely.
minimal. The statistical analysis plan was pub- Our findings agree with those of the ProCESS
lished before recruitment was completed, which trial,10 in which investigators also used a resus-
eliminated the potential for analytical bias.28 citation algorithm that was similar to that used
Although the trial could not be blinded because in the original EGDT trial.11 Although our re-
of the practical requirements of EGDT, the risk sults differ from those in the original trial, they
of bias was minimized through central random- are consistent with previous studies showing
ization, concealment of study-group assignments that bias in small, single-center trials may lead
before randomization to avoid selection bias, and to inflated effect sizes34 that cannot be replicated
the use of a robust primary outcome that would in larger, multicenter studies.35-37 Although the
not be subject to observer bias. The results also ProCESS study did not directly compare proto-
have a high degree of external validity, since col-based EGDT for resuscitation with care that
1504 n engl j med 371;16 54

The New EnglandFor

reserved.
was not protocol-based, the concordance of re- tion practice, did not decrease mortality among
sults between our study and the ProCESS study patients presenting to the emergency department
suggests that EGDT does not offer a survival ad- with early septic shock. Our findings suggest that
vantage in patients presenting to the emergency the value of incorporating EGDT into interna-
department with early septic shock. Whether re- tional guidelines as a standard of care is ques-
suscitation protocols with different goals or dif- tionable.
ferent individual therapies in the EGDT bundle Supported by grants from the National Health and Medical
offer a survival benefit remains to be determined. Research Council of Australia (491075 and 1021165) and the
Alfred Foundation.
In conclusion, the results of our trial show Disclosure forms provided by the authors are available with
that EGDT, as compared with usual resuscita- the full text of this article at NEJM.org.
Appendix
The affiliations of the writing committee members are as follows: the Australian and New Zealand Intensive Care Research Centre,
School of Public Health and Preventive Medicine (S.L.P., A.D., M.B., R.B., D.J.C., A.M.H., B.D.H., S.A.R.W., P.W.) and the School of
Public Health (P.A.C.), Monash University, Melbourne, VIC; University of Adelaide and Queen Elizabeth Hospital, Adelaide, SA (S.L.P.,
P.W.); Royal North Shore Hospital and University of Sydney, Sydney, NSW (A.D.); Austin Hospital, Melbourne, VIC (R.B.); Alfred Hos-
pital Melbourne, VIC (P.A.C., D.J.C); Liverpool Hospital and University of New South Wales, Sydney, NSW (A.H.); and Royal Perth
Hospital and University of Western Australia, Perth, WA (S.A.R.W.) — all in Australia; and Hamad Medical, Doha, Qatar (P.A.C).
References
1. Jawad I, Lukšić I, Rafnsson SB. As- Australia and New Zealand. Crit Care Re- 17. Barochia AV, Cui X, Vitberg D, et al.
sessing available information on the bur- susc 2007;9:8-18. Bundled care for septic shock: an analysis
den of sepsis: global estimates of inci- 9. Quenot J-P, Binquet C, Kara F, et al. of clinical trials. Crit Care Med 2010;38:
dence, prevalence and mortality. J Glob The epidemiology of septic shock in 668-78.
Health 2012;2:010404. French intensive care units: the prospec- 18. Levy MM, Dellinger RP, Townsend
2. Angus DC, Linde-Zwirble WT, Lidick- tive multicenter cohort EPISS study. Crit SR, et al. The Surviving Sepsis Campaign:
er J, Clermont G, Carcillo J, Pinsky MR. Care 2013;17:R65. results of an international guideline-
Epidemiology of severe sepsis in the Unit- 10. Yealy DM, Kellum JA, Huang DT, et al. based performance improvement program
ed States: analysis of incidence, outcome, A randomized trial of protocol-based care targeting severe sepsis. Intensive Care
and associated costs of care. Crit Care for early septic shock. N Engl J Med 2014; Med 2010;36:222-31.
Med 2001;29:1303-10. 370:1683-93. 19. Hayes MA, Timmins AC, Yau EH,
3. Gaieski DF, Edwards JM, Kallan MJ, 11. Rivers E, Nguyen B, Havstad S, et al. Palazzo M, Hinds CJ, Watson D. Elevation
Carr BG. Benchmarking the incidence Early goal-directed therapy in the treat- of systemic oxygen delivery in the treat-
and mortality of severe sepsis in the Unit- ment of severe sepsis and septic shock. ment of critically ill patients. N Engl J
ed States. Crit Care Med 2013;41:1167-74. N Engl J Med 2001;345:1368-77. Med 1994;330:1717-22.
4. Kaukonen K-M, Bailey M, Suzuki S, 12. Dellinger RP, Carlet JM, Masur H, et 20. Hebert PC, Wells G Blajchman MA, et
Pilcher D, Bellomo R. Mortality related to al. Surviving Sepsis Campaign guidelines al. A multicenter, randomized, controlled
severe sepsis and septic shock among for management of severe sepsis and sep- clinical trial of transfusion requirements
critically ill patients in Australia and New tic shock. Crit Care Med 2004;32:858-73. in critical care. N Engl J Med 1999;340:409-
Zealand, 2000-2012. JAMA 2014;311:1308- [Errata, Crit Care Med 2004;32:1449, 17. [Erratum, N Engl J Med 1999;340:1056.]
16. 2169-70.] 21. Peake S, Webb S, Delaney A. Early
5. Annane D, Aegerter P, Jars-Guincestre 13. Dellinger RP, Levy MM, Carlet JM, et goal-directed therapy of septic shock: we
MC, Guidet B. Current epidemiology of al. Surviving Sepsis Campaign: interna- honestly remain skeptical. Crit Care Med
septic shock: the CUB-Réa Network. Am J tional guidelines for management of se- 2007;35:994-5.
Respir Crit Care Med 2003;168:165-72. vere sepsis and septic shock: 2008. Inten- 22. Perel A. Bench-to-bedside review: the
6. Harrison DA, Welch CA, Eddleston sive Care Med 2008;34:17-60. [Erratum, initial hemodynamic resuscitation of the
JM. The epidemiology of severe sepsis in Intensive Care Med 2008;34:783-5.] septic patient according to Surviving Sep-
England, Wales and Northern Ireland, 14. Dellinger RP, Levy MM, Rhodes A, et sis Campaign guidelines — does one size
1996 to 2004: secondary analysis of a al. Surviving Sepsis Campaign: interna- fit all? Crit Care 2008;12:223.
high quality clinical database, the ICNARC tional guidelines for management of se- 23. Surviving Sepsis Campaign responds
Case Mix Programme Database. Crit Care vere sepsis and septic shock: 2012. Crit to ProCESS Trial: updated 19 May 2014
2006;10:R42. Care Med 2013;41:580-637. (http://www.survivingsepsis.org/
7. Wang HE, Shapiro NI, Angus DC, 15. Kortgen A, Niederprüm P, Bauer M. SiteCollectionDocuments/SSC-Responds
Yealy DM. National estimates of severe Implementation of an evidence-based -ProCESS-Trial.pdf).
sepsis in United States emergency depart- “standard operating procedure” and out- 24. Huang DT, Angus DC, Barnato A, et
ments. Crit Care Med 2007;35:1928-36. come in septic shock. Crit Care Med al. Harmonizing international trials of
8. The Australasian Resuscitation in 2006;34:943-9. early goal-directed resuscitation for se-
Sepsis Evaluation (ARISE) Investigators, 16. Jones AE, Focht A, Horton JM, Kline vere sepsis and septic shock: methodolo-
Australian and New Zealand Intensive JA. Prospective external validation of the gy of ProCESS, ARISE, and ProMISe. In-
Care Society (ANZICS) Adult Patient Data- clinical effectiveness of an emergency de- tensive Care Med 2013;39:1760-75.
base (APD) Management Committee. The partment-based early goal-directed thera- 25. Peake SL, Bailey M, Bellomo R, et al.
outcome of sepsis and septic shock pre- py protocol for severe sepsis and septic Australasian Resuscitation of Sepsis Eval-
senting to the emergency departments of shock. Chest 2007;132:425-32. uation (ARISE): a multi-centre, prospec-

55 october 16, 2014 1505
TheFor
Downloaded from collections.nejm.org. New England
personal useJournal of Medicine
Downloaded from nejm.org by MIKE
Copyright © MULARCZYK on January
Massachusetts Medical 22, 2015.
Society. For personal
All rights use only. No other uses without permission.
reserved.
tive, inception cohort study. Resuscitation and New Zealand intensive care units. In- vere sepsis: a comparative meta-analysis.
2009;80:811-8. tensive Care Med 2004;30:589-96. [Erra- Crit Care Med 2014;42:625-31.
26. Bone RC, Balk RA, Cerra FB, et al. tum, Intensive Care Med 2004;30:1252.] 34. Bellomo R, Warrillow SJ, Reade MC.
Definitions for sepsis and organ failure 30. Angus DC, Laterre PF, Helterbrand J, Why we should be wary of single-center
and guidelines for the use of innovative et al. The effect of drotrecogin alfa (acti- trials. Crit Care Med 2009;37:3114-9.
therapies in sepsis. Chest 1992;101:1644- vated) on long-term survival after severe 35. Ioannidis JP. Contradicted and ini-
55. sepsis. Crit Care Med 2004;32:2199-206. tially stronger effects in highly cited clin-
27. The NICE-SUGAR Study Investiga- 31. Charlson ME, Pompei P, Ales KL, ical research. JAMA 2005;294:218-28.
tors. Intensive versus conventional glu- MacKenzie CR. A new method of classify- 36. Zhang Z, Xu X, Ni H. Small studies
cose control in critically ill patients. ing prognostic comorbidity in longitudi- may overestimate the effect sizes in criti-
N Engl J Med 2009;360:1283-97. nal studies: development and validation. cal care meta-analyses: a meta-epidemio-
28. Delaney AP, Peake SL, Bellomo R, et J Chronic Dis 1987;40:373-83. logical study. Crit Care 2013;17:R2.
al. The Australasian Resuscitation in Sep- 32. Knaus WA, Draper EA, Wagner DP, 37. Kjaergard LL, Villumsen J, Gluud C.
sis Evaluation (ARISE) trial statistical Zimmerman JE. APACHE II: a severity of Reported methodologic quality and dis-
analysis plan. Crit Care Resusc 2013;15: disease classification system. Crit Care crepancies between large and small ran-
162-71. Med 1985;13:818-29. domized trials in meta-analyses. Ann In-
29. Finfer S, Bellomo R, Lipman J, French 33. Stevenson EK, Rubenstein AR, Radin tern Med 2001;135:982-9. [Erratum, Ann
C, Dobb G, Myburgh J. Adult-population GT, Wiener RS, Walkey AJ. Two decades of Intern Med 2008;149:219.]
incidence of severe sepsis in Australian mortality trends among patients with se- Copyright © 2014 Massachusetts Medical Society.
an nejm app for iphone

The NEJM Image Challenge app brings a popular online feature to the smartphone.
Optimized for viewing on the iPhone and iPod Touch, the Image Challenge app lets
you test your diagnostic skills anytime, anywhere. The Image Challenge app
randomly selects from 300 challenging clinical photos published in NEJM,
with a new image added each week. View an image, choose your answer,
get immediate feedback, and see how others answered.
The Image Challenge app is available at the iTunes App Store.

56 october 16, 2014
The New England

Downloaded from collections.nejm.org. ForJournal of use
personal Medicine
Downloaded from nejm.org by MIKECopyright
MULARCZYK on January 22,
© Massachusetts 2015.Society.
Medical For personal use only.
reserved.
c or r e sp ondence
Goal-Directed Resuscitation in Septic Shock

To the Editor: The Australasian Resuscitation To the Editor: The ARISE investigators con-
in Sepsis Evaluation (ARISE) Investigators and clude that EGDT, as compared with standard
the Australian and New Zealand Intensive Care care, did not decrease mortality among patients
Society (ANZICS) Clinical Trials Group (Oct. 16 with early septic shock. This conclusion is mis-
issue)1 report that early goal-directed therapy leading, because these data need to be interpret-
(EGDT) did not reduce mortality at 90 days ed in the context of the implementation of the
among patients with early septic shock. During the Surviving Sepsis Campaign guidelines1 and the
72 hours after randomization, both groups in the change in clinical practice worldwide during the
study received a mean of approximately 6500 ml past decade. ARISE started in 2008 — 7 years
(70 to 75 ml per kilogram of body weight) of after the results of the study by Rivers et al. were
combined crystalloids and colloids (see Table S5 reported2 and 4 years after the launch of the Sur-
in the Supplementary Appendix of the article, viving Sepsis Campaign guidelines. The marginal
available at NEJM.org). A footnote in Table S5 and clinically irrelevant differences in fluid and
suggests that 0.9% saline was often used as crystal- vasopressor therapy between the two groups in
loid; rapid isotonic saline infusion predictably the ARISE study show that the “usual-care”
results in hyperchloremic acidosis.2 Hyperchlore- group was treated with awareness of the protocol
mia can be associated with reduced gastric muco- described by Rivers et al. Moreover, if one com-
sal perfusion, renal vasoconstriction, a reduced pares mortality in the ARISE control group with
glomerular filtration rate,3 and even increased mortality among patients in sepsis studies in the
mortality.4 Since the exact nature of fluid therapy early 2000s and acknowledges the recent data
is crucial in EGDT, can the authors provide data from the ANZICS registry,3 one has to conclude
on the specific fluids that patients received both that ARISE reflects the widespread implementa-
before and after randomization? tion of EGDT approaches in routine care. What
Hans-Joachim Priebe, M.D. we see in the ARISE study (and the Protocolized
University Hospital Freiburg Care for Early Septic Shock [ProCESS] trial)4 is
Freiburg, Germany not a failure of protocol-based resuscitation per se,
hans-joachim.priebe@uniklinik-freiburg.de
but an impressive global “learning bias” regard-
ported. ing EGDT in terms of early administration of anti-
biotics and fluids.
1. ARISE Investigators, ANZICS Clinical Trials Group. Goal-
directed resuscitation for patients with early septic shock. N Engl Bernd Saugel, M.D.
J Med 2014;371:1496-506. Daniel A. Reuter, M.D.
2. Prough DS, Bidani A. Hyperchloremic metabolic acidosis is a
University Medical Center Hamburg–Eppendorf
predictable consequence of intraoperative infusion of 0.9% saline.
Hamburg, Germany
Anesthesiology 1999;90:1247-9.
bcs.muc@gmx.de
3. Handy JM, Soni N. Physiological effects of hyperchloraemia
and acidosis. Br J Anaesth 2008;101:141-50. Dr. Saugel reports receiving unrestricted research grants from
4. Silva Junior JM, Neves EF, Santana TC, Ferreira UP, Marti YN, Tensys Medical, travel expenses from CNSystems Medizintech-
Silva JMC. The importance of intraoperative hyperchloremia. nik, and fees for serving on the medical advisory board of Pul-
Rev Bras Anestesiol 2009;59:304-13. sion Medical Systems; and Dr. Reuter, receiving fees for serving
on the medical advisory board of Pulsion Medical Systems and
DOI: 10.1056/NEJMc1413936 consulting fees from Masimo. No other potential conflict of in-
terest relevant to this letter was reported.
188 n engl j med 372;2 57

nejm.org january 8, 2015
correspondence
1. Dellinger RP, Carlet JM, Masur H, et al. Surviving Sepsis Cam- after randomization was similar in the EGDT
paign guidelines for management of severe sepsis and septic
shock. Crit Care Med 2004;32:858-73. [Errata, Crit Care Med and usual-care groups, and there was no evidence
2004;32:1448, 2169-70.] of metabolic acidosis (Table S5 in the Supple-
2. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed ther- mentary Appendix of our article). Furthermore,
apy in the treatment of severe sepsis and septic shock. N Engl J
Med 2001;345:1368-77. the ideal crystalloid to administer in patients
3. Kaukonen KM, Bailey M, Suzuki S, Pilcher D, Bellomo R. with septic shock has not been the subject of
Mortality related to severe sepsis and septic shock among criti- a large multicenter, randomized, controlled
cally ill patients in Australia and New Zealand, 2000-2012.
JAMA 2014;311:1308-16. trial.
4. The ProCESS Investigators. A randomized trial of protocol- There is no evidence that the results of our
based care for early septic shock. N Engl J Med 2014;370:1683- study indicate that EGDT-based resuscitation has
93.
been incorporated into routine care, as Saugel
DOI: 10.1056/NEJMc1413936
and Reuter contend. The ANZICS did not endorse
To the Editor: The ARISE investigators report the updated 2008 Surviving Sepsis Campaign
that the “median time between presentation to guidelines (in particular, the recommendation
the emergency department and administration of for EGDT) because the guidelines did not reflect
the first dose of intravenous antimicrobial ther- usual care in Australia and New Zealand.2 Our
apy was similar in the two groups: 70 minutes pretrial prospective, inception-cohort study con-
(interquartile range, 38 to 114) in the EGDT group ducted in 32 hospitals in Australia and New
and 67 minutes (interquartile range, 39 to 110) in Zealand also established that central venous
the usual-care group.” Would the authors please oxygen saturation (Scvo2) monitoring, a funda-
clarify whether the first administration of anti- mental component of the resuscitation protocol,
microbial therapy occurred by means of an infu- was not performed in patients who presented
sion pump over 30 to 60 minutes or as a bolus? to the emergency department with early septic
Also, does the time to the reported first dose of shock.3 Moreover, before the site selection, we
intravenous antimicrobial therapy refer to the confirmed that all sites that participated in
initiation or the completion of the first dose of ARISE did not include sepsis-resuscitation proto-
the antimicrobial agent? This information is im- cols as part of usual care. There is no indication
portant because the time to the first dose of an- from the trial results that EGDT has subsequent-
tibiotics affects survival among patients with ly been integrated into routine practice in Aus-
severe sepsis or septic shock.1 tralia and New Zealand. After randomization,
Dominique J. Pepper, M.B., Ch.B., M.D. the mean (±SD) volume of intravenous fluid ad-
National Institutes of Health ministered during the first 6 hours in the usual-
Bethesda, MD care group (1713±1401 ml) and the use of dobu-
dominiquepepper@gmail.com
tamine (2.6%) and red-cell transfusion (7.0%)
ported.
were all similar to the same variables in our
pretrial observational study (2218±1224 ml, 2.5%,
1. Kumar A, Roberts D, Wood KE, et al. Duration of hypoten-
sion before initiation of effective antimicrobial therapy is the
and 7.4%, respectively). In addition, no patients
critical determinant of survival in human septic shock. Crit Care in the usual-care group received Scvo2 monitor-
Med 2006;34:1589-96. ing. Published data from the ANZICS Adult Pa-
DOI: 10.1056/NEJMc1413936 tient Database4,5 also show that the progressive
decrease in sepsis-related deaths began before
publication of both the results of the original
The Authors Reply: Priebe1 raises the same EGDT trial in 2001 and the first Surviving Sepsis
concerns in his letter that he raised about hy- Campaign guidelines in 2004.
perchloremic acidosis associated with rapid ad- We agree with Pepper that the time to admin-
ministration of isotonic saline in relation to the istration of antimicrobial agents is an important
ProCESS trial; however, in our study we did not determinant of outcome. Although we only re-
report the type of crystalloid administered or corded the time to initiation of an intravenous
the serum chloride level. In addition, the volume antimicrobial agent, usual practice in Australia
of crystalloid administered during the 72 hours and New Zealand is to administer intravenous

nejm.org january 8, 2015 189
antimicrobial agents as a bolus. The time to 2. Hicks P, Cooper DJ, Webb S, et al. The Surviving Sepsis Cam-
paign: international guidelines for management of severe sepsis
administration was the same in both the EGDT and septic shock: 2008 — an assessment by the Australian and
and usual-care groups. New Zealand Intensive Care Society. Anaesth Intensive Care
2008;36:149-51.
Sandra L. Peake, M.D., Ph.D. 3. Peake SL, Bailey M, Bellomo R, et al. Australasian Resuscita-
Anthony Delaney, M.D., Ph.D. tion of Sepsis Evaluation (ARISE): a multi-centre, prospective,
inception cohort study. Resuscitation 2009;80:811-8.
Rinaldo Bellomo, M.D. 4. ARISE, ANZICS APD Management Committee. The outcome
Monash University of patients with sepsis and septic shock presenting to emergency
Melbourne, VIC, Australia departments in Australia and New Zealand. Crit Care Resusc
2007;9:8-18.
for the ARISE Investigators 5. Kaukonen KM, Bailey M, Suzuki S, Pilcher D, Bellomo R.
Since publication of their article, the authors report no fur- Mortality related to severe sepsis and septic shock among criti-
ther potential conflict of interest. cally ill patients in Australia and New Zealand, 2000-2012.
JAMA 2014;311:1308-16.
1. Priebe H-J. Protocol-based care for early septic shock. N Engl
DOI: 10.1056/NEJMc1413936
J Med 2014;371:384-7.
190 n engl j med 372;2 59

original article
Trial of Early, Goal-Directed Resuscitation

for Septic Shock
Paul R. Mouncey, M.Sc., Tiffany M. Osborn, M.D., G. Sarah Power, M.Sc.,
David A. Harrison, Ph.D., M. Zia Sadique, Ph.D., Richard D. Grieve, Ph.D.,
Rahi Jahan, B.A., Sheila E. Harvey, Ph.D., Derek Bell, M.D., Julian F. Bion, M.D.,
Timothy J. Coats, M.D., Mervyn Singer, M.D., J. Duncan Young, D.M.,
and Kathryn M. Rowan, Ph.D., for the ProMISe Trial Investigators*
A BS T R AC T
Background
Early, goal-directed therapy (EGDT) is recommended in international guidelines for From the Clinical Trials Unit, Intensive
the resuscitation of patients presenting with early septic shock. However, adoption Care National Audit and Research Centre
(P.R.M., G.S.P., D.A.H., R.J., S.E.H., K.M.R.),
has been limited, and uncertainty about its effectiveness remains. Department of Health Services Research
and Policy, London School of Hygiene and
Methods Tropical Medicine (M.Z.S., R.D.G.), and
Faculty of Medicine, Imperial College
We conducted a pragmatic randomized trial with an integrated cost-effectiveness London (D.B.), Department of Acute
analysis in 56 hospitals in England. Patients were randomly assigned to receive Medicine, Chelsea and Westminster Hos-
either EGDT (a 6-hour resuscitation protocol) or usual care. The primary clinical pital NHS Foundation Trust (D.B.), and
Bloomsbury Institute of Intensive Care
outcome was all-cause mortality at 90 days. Medicine, University College London
(M.S.), London, the Department of Inten-
Results sive Care Medicine, University of Birming-
ham, Birmingham (J.F.B.), Department
We enrolled 1260 patients, with 630 assigned to EGDT and 630 to usual care. By of Cardiovascular Sciences, University of
90 days, 184 of 623 patients (29.5%) in the EGDT group and 181 of 620 patients Leicester, Leicester (T.J.C.), and Nuffield
(29.2%) in the usual-care group had died (relative risk in the EGDT group, 1.01; Division of Anaesthetics, University of
Oxford, Oxford (J.D.Y.) — all in the United
95% confidence interval [CI], 0.85 to 1.20; P = 0.90), for an absolute risk reduction Kingdom; and the Departments of Sur-
in the EGDT group of −0.3 percentage points (95% CI, −5.4 to 4.7). Increased treat- gery and Emergency Medicine, Washing-
ment intensity in the EGDT group was indicated by increased use of intravenous ton University at St. Louis, St. Louis
(T.M.O.). Address reprint requests to Dr.
fluids, vasoactive drugs, and red-cell transfusions and reflected by significantly Rowan at the Intensive Care National Audit
worse organ-failure scores, more days receiving advanced cardiovascular support, and Research Centre, Napier House, 24
and longer stays in the intensive care unit. There were no significant differences in High Holborn, London WC1V 6AZ, United
Kingdom, or at kathy.rowan@icnarc.org.
any other secondary outcomes, including health-related quality of life, or in rates of
serious adverse events. On average, EGDT increased costs, and the probability that * A complete list of investigators in the
it was cost-effective was below 20%. Protocolised Management in Sepsis
(ProMISe) trial is provided in the Supple-
mentary Appendix, available at NEJM.org.
Conclusions
In patients with septic shock who were identified early and received intravenous 2015, at NEJM.org.
antibiotics and adequate fluid resuscitation, hemodynamic management according
DOI: 10.1056/NEJMoa1500896
to a strict EGDT protocol did not lead to an improvement in outcome. (Funded by the Copyright © 2015 Massachusetts Medical Society.
United Kingdom National Institute for Health Research Health Technology Assess-
ment Programme; ProMISe Current Controlled Trials number, ISRCTN36307479.)
n engl j med nejm.org 60 1

T
he incidence of severe sepsis and The ProMISe study, which was conducted in a
septic shock in adults is estimated to range setting in which the reported mortality for septic
from 56 to 91 per 100,000 population per shock is high and was designed with an integrat-
year.1 Affected patients have high rates of death, ed economic evaluation, tested the hypothesis
complications, and resource utilization.2-5 that the 6-hour EGDT resuscitation protocol is
Since 2002, the Surviving Sepsis Campaign superior, in terms of clinical and cost-effective-
(SSC) has promoted best practice, including early ness measures, to usual care in patients present-
recognition, source control, appropriate and timely ing with early septic shock to National Health
antibiotic administration, and resuscitation with Service (NHS) emergency departments in En-
intravenous fluids and vasoactive drugs.6-8 Resus- gland.
citation guidance is largely based on a 2001 single-
center, proof-of-concept study by Rivers et al., Me thods
which indicated that protocolized delivery of
6 hours of early, goal-directed therapy (EGDT) to Study Design and Oversight
patients presenting to the emergency department Our study was a pragmatic, open, multicenter,
with early septic shock reduced hospital mortality parallel-group, randomized, controlled trial. The
and hospital stay.9 Such therapy aims to optimize North West London Research Ethics Committee
tissue oxygen transport with the use of continu- approved the study protocol, which is available
ous monitoring of prespecified physiological tar- with the full text of this article at NEJM.org. The
gets — central venous pressure, mean arterial United Kingdom National Institute for Health
pressure, and central venous oxygen saturation Research (NIHR) funded the study and convened
(ScvO2) — to guide delivery of intravenous flu- a trial steering committee and independent data
ids, vasoactive drugs, and red-cell transfusions. monitoring and ethics committee. The Clinical
However, despite the SSC recommendations, Trials Unit at the United Kingdom Intensive Care
the adoption of EGDT has been limited, with National Audit and Research Centre (ICNARC)
concern about the external validity of results managed the study (for details, see the Supple-
from a single center, the complexity of delivery, mentary Appendix, available at NEJM.org). Ed-
the potential risks of the components, and re- wards Lifesciences loaned monitors and provided
sources required for implementation.10,11 training and technical support but had no other
To address these concerns, multicenter trials of role in the study.
EGDT were conducted in the United States (Pro-
tocolized Care for Early Septic Shock [ProCESS] Sites and Patients
trial),12 Australasia (Australasian Resuscitation in The study was conducted in English NHS hospi-
Sepsis Evaluation [ARISE] trial),13 and England tals that did not routinely use EGDT that includ-
(Protocolised Management in Sepsis [ProMISe] ed continuous ScvO2 monitoring. Adults (≥18 years
trial). In all three trials, harmonized methods14 of age) were eligible if within 6 hours after presen-
were used to permit subsequent meta-analysis of tation to the emergency department they had a
data from individual patients.15 The two published known or presumed infection, two or more criteria
studies12,13 reported no benefit for EGDT. How- of the systemic inflammatory response syndrome,16
ever, both reported lower-than-anticipated mortal- and either refractory hypotension (systolic blood
ity, with a 60-day in-hospital mortality of 18.9% pressure, <90 mm Hg; or mean arterial pressure,
(as compared with an anticipated rate of 30 to <65 mm Hg, despite resuscitation with at least
46%) in the ProCESS trial and 90-day mortality of 1 liter of intravenous fluids within 60 minutes) or
18.8% (as compared with an anticipated rate of hyperlactatemia (blood lactate level, ≥4 mmol
38%) in the ARISE trial. Consequently, neither per liter) and did not meet any exclusion criteria
trial could rule out the potential for a 20% relative (see the Methods section in the Supplementary
reduction in 90-day mortality for EGDT, as com- Appendix).
pared with usual care, with a relative risk of 0.94 Randomization had to be completed within 2
(95% confidence interval [CI], 0.77 to 1.15) in the hours after the patient met the inclusion criteria.
ProCESS trial and a relative risk of 0.98 (95% CI, All patients provided written informed consent,
0.80 to 1.21) in the ARISE trial. We based sample- or consent was granted through an agreement
size calculations for the ProMISe study on a rela- with a personal or professional consultee or in-
tive risk reduction of 20%. dependent clinician.17 Patients were assigned in
2 n engl j med
61 nejm.org
TRIAL OF EARLY RESUSCITATION FOR SEPTIC SHOCK
a 1:1 ratio by means of 24-hour telephone ran- Statistical Analysis

domization to receive either EGDT or usual care. Using the ICNARC Case Mix Program Data-
Study-group assignment was performed by base,20 we estimated that 90-day mortality would
means of randomized permuted blocks, with be 40% in the usual-care group. On the basis of
variable block lengths of 4, 6, and 8, and strati- this estimation, we calculated that an enrollment
fied according to site. In all study patients, anti- of 1260 patients would have a power of 80% to
microbial drugs were initiated before random- detect a relative reduction of 20% in risk (abso-
ization. lute risk reduction, 8 percentage points) in the
EGDT group, allowing for a loss to follow-up or
Study Interventions withdrawal of 6%.21
After randomization, the usual-care group con- All analyses were performed according to the
tinued to receive monitoring, investigations, and intention-to-treat principle and were prespecified
treatment as determined by the treating clini- in the statistical analysis plan.22 A P value of less
cians, whereas the EGDT group started the resus- than 0.05 was considered to indicate statistical
citation protocol (Fig. S1 in the Supplementary significance. All tests were two-sided with no
Appendix). For the latter, during the first hour, adjustment for multiple comparisons. Continu-
which was defined as the next whole hour (e.g., ous variables are reported as means and standard
if randomization was performed at 9:24, then by deviations or medians and interquartile ranges.
11 o’clock), a central venous catheter capable of Categorical variables are reported as proportions.
continuous ScvO2 monitoring was placed. The re- We used Fisher’s exact test to compare between-
suscitation protocol was followed for 6 hours (in- group differences in the primary outcome. Rela-
tervention period) with personnel involved and tive and absolute reductions in risk are reported
treatment location decided according to the site. with 95% confidence intervals without adjustment.
At least one trained staff member was available Secondary analyses of the primary outcome in-
throughout the intervention period. Key staff cluded odds ratio with adjustment for Mortality
members were trained before the initiation of rein Emergency Department Sepsis (MEDS) score
cruitment at each site. All other treatment, dur- components, sensitivity analyses for missing data,
ing the intervention period and after, was at the learning-curve analysis, and adherence-adjusted
discretion of the treating clinicians. Blinding to analysis. We conducted prespecified subgroup
study-group assignment was not possible. Dur- analyses by testing interactions between the ef-
ing the intervention period, data were collected fect of EGDT and the degree of protocolized care
prospectively for the EGDT group and retrospec- (in the usual-care group), age, MEDS score,23
tively for the usual-care group to avoid the influ- SOFA score, and the time from presentation at
ence of data collection on treatment delivery. the emergency department to randomization.
In the cost-effectiveness analysis, we reported
Outcome Measures quality-adjusted life-years (QALYs) by combining
The primary clinical outcome was all-cause mor- survival data with quality-of-life scores at 90 days
tality at 90 days. Secondary outcomes were the and estimated incremental net benefits by valuing
score on the Sequential Organ Failure Assess- incremental QALYs at the recommended thresh-
ment (SOFA)18 at 6 hours and 72 hours; receipt of old for a QALY gain (£20,000 [U.S. $28,430]) and
advanced cardiovascular, advanced respiratory, then subtracting the incremental costs from this
or renal support and the number of days in the value.24 Stata/SE software, version 13.0, was used
first 28 days after randomization that were free for all analyses. (Details about methods are pro-
from such support19; length of stay in the emer- vided in the Statistical Analysis section in the
gency department, intensive care unit (ICU), and Supplementary Appendix.)
hospital; duration of survival; all-cause mortality
at 28 days, at hospital discharge, and at 1 year; and R e sult s
health-related quality of life (as measured on the
European Quality of Life–5 Dimensions [EQ-5D] Study Patients
five-level questionnaire), resource use, and costs From February 16, 2011, to July 24, 2014, we
at 90 days and 1 year. Adverse events were moni- screened 6192 patients at 56 sites (including 29%
tored up to 30 days. All definitions are provided that are teaching hospitals), which resulted in
in the Supplementary Appendix. the enrollment of 1260 patients (Tables S1 and
n engl j med
62 nejm.org 3
S2 and Fig. S2 and S3 in the Supplementary Ap- (Fig. S4 in the Supplementary Appendix). Stan-
pendix). Four patients requested complete with- dard central venous catheters were not mandated
drawal and five were ineligible, which left 1251 but were placed in 50.9% of the patients in the
patients in the initial analysis (625 in the EGDT usual-care group, and ScvO2 was measured from
group and 626 in the usual-care group). Eight aspirated blood samples in 6 patients. Arterial
patients withdrew before 90 days, which left catheters were also not mandated but were placed
1243 patients in the analysis of outcomes (623 in in most patients.
the EGDT group and 620 in the usual-care group) EGDT was stopped prematurely in 21 patients
(Fig. 1, and Table S3 in the Supplementary Ap- (median time to cessation, 3 hours) because
pendix). The two study groups were well matched active treatment was withdrawn (9 patients), the
at baseline (Table 1, and Table S4 in the Supple- patient was no longer considered to have sepsis
mentary Appendix). (5 patients), or EGDT was terminated in error
The criterion for refractory hypotension was (3 patients); in addition, 1 patient was transferred
met in 338 patients (54.1%) in the EGDT group to an operating room, 1 patient declined treat-
and 348 patients (55.6%) in the usual-care group, ment, and no reason was provided for 2 patients.
and the criterion for hyperlactatemia was met in Among the 35 patients who died within 6 hours
409 patients (65.4%) and 399 patients (63.7%), (17 in the EGDT group and 18 in the usual-care
respectively. The intravenous-fluid volume before group), 5 in the EGDT group and 6 in the usual-
randomization was similar in the two groups, as care group had withdrawal of active treatment.
were median times from presentation at the Adherence to EGDT ranged from 86 to 95%,
emergency department until inclusion criteria depending on the method of assessment (Fig. S5
were met and until randomization. Only about in the Supplementary Appendix).
two thirds of patients in either group were deemed
likely to be admitted to the ICU from the emer- Intervention Period
gency department if they were not enrolled in During the 6-hour intervention period, patients
the study; those deemed unlikely to be admitted in the EGDT group received more intravenous
were less severely ill. The sites of infection (most fluids than did patients in the usual-care group
commonly lung) were well balanced in the two (Table 2). Hourly fluid volume decreased over the
groups. All patients received antimicrobial drugs 6 hours, but patients in the usual-care group re-
before randomization. ceived a larger initial volume (Fig. S6 in the Sup-
plementary Appendix). Crystalloids were admin-
Adherence to the Protocol istered more frequently than colloids in the two
Most patients in the EGDT group underwent groups. More patients in the EGDT group than in
timely insertion of a central venous catheter capa- the usual-care group received vasopressors and
ble of continuous ScvO2 monitoring. Two cathe- dobutamine. Although more patients in the EGDT
ters that were inserted in error in the usual-care group received red-cell transfusions, larger vol-
group were not used for monitoring ScvO2 (Table 2, umes were transfused in the usual-care group.
and Table S5 in the Supplementary Appendix). In During the 6-hour intervention period, administra-
the EGDT group, reasons for failure of insertion tion of platelets and fresh-frozen plasma was simi-
were as follows: patient no longer met inclusion lar in the two groups, although the volume of
criteria or met exclusion criteria (22 patients), each was higher in the EGDT group (Table 2). At
there was a lapse in the process of care (lack of 6 hours, values for central venous pressure, mean
equipment, staff, beds, communication, or error) arterial pressure, systolic blood pressure, and hemo-
(20 patients), there were technical difficulties or globin were similar in the two groups among pa-
problems with a patient (18 patients), there was a tients in whom they were measured, which hap-
decision by a clinician (9 patients), or the patient pened with greater frequency in the EGDT group
declined to have a catheter inserted but did not (Table S6 in the Supplementary Appendix).
withdraw from the trial (5 patients); in 4 patients,
no reason was provided, and 2 patients died be- After the Intervention Period
fore catheter insertion. The mean (±SD) first Between 6 and 72 hours, the numbers of patients
ScvO2 value recorded (at hour 1) was 70±12% in the two groups receiving intravenous fluids
4 n engl j med
63 nejm.org
6192 Patients met inclusion criteria

841 Had treating physician who deemed aggressive care unsuitable
794 Had do-not-resuscitate order
167 Were not able to start EGDT ≤1 hr after randomization or
complete 6 hr of EGDT
142 Required immediate surgery
74 Had contraindication to central venous catheterization
58 Had major cardiac arrhythmia
57 Had hemodynamic instability from active gastrointestinal
hemorrhage
57 Were transferred from another in-hospital setting
42 Had seizure
25 Had stroke
25 Had acute coronary syndrome
24 Had drug overdose
23 Had acute pulmonary edema
16 Had advance directives restricting implementation of EGDT
16 Were <18 yr of age
15 Were known to have a history of AIDS
15 Were previously enrolled in ProMISe
8 Had injury from burn or trauma
6 Had status asthmaticus
5 Had contraindication to blood transfusion
5 Were known to be pregnant
995 Had study logistic issues
449 Were excluded by clinician
343 Had delayed referral
25 Did not provide reason
630 Were assigned to receive

630 Were assigned to receive EGDT
usual resuscitation
625 Were eligible for analysis 626 Were eligible for analysis
3 Requested removal of all data 1 Requested removal of all data
2 Were ineligible 3 Were ineligible
623 Were included in primary outcome 620 Were included in primary outcome
analysis analysis
2 Withdrew in <90 days 6 Withdrew in <90 days
356 (81% of those eligible for follow-up) 354 (81% of those eligible for follow-up)
Returned EQ-5D questionnaire Returned EQ-5D questionnaire
at 90 days at 90 days
339 Had complete data 332 Had complete data
Figure 1. Enrollment and Outcomes.

AIDS denotes acquired immunodeficiency syndrome, EGDT early, goal-directed therapy, and EQ-5D European Quality
of Life–5 Dimensions.
n engl j med
64 nejm.org 5
EGDT Usual Care

Characteristic (N = 625) (N = 626)
Age — yr 66.4±14.6 64.3±15.5
Male sex — no. (%) 356 (57.0) 367 (58.6)
Refractory hypotension — no. (%) 338 (54.1) 348 (55.6)
Systolic blood pressure — mm Hg 77.7±11.0 78.4±10.2
Mean arterial pressure — mm Hg 58.8±15.8 59.0±10.7
Hyperlactatemia — no. (%) 409 (65.4) 399 (63.7)
Blood lactate level — mmol/liter 7.0±3.5 6.8±3.2
Intravenous fluids administered†
Before hospitalization until randomization — no./total no. (%) 612/625 (97.9) 606/625 (97.0)
Median total before hospitalization until randomization (IQR) — ml 1950 (1000–2500) 2000 (1000–2500)
Supplemental oxygen — no./total no. (%)‡ 397/539 (73.7) 407/542 (75.1)
Median time from presentation in emergency department to randomization (IQR) — hr 2.5 (1.8–3.5) 2.5 (1.8–3.5)
Patient would have been admitted directly from emergency department to ICU if not
enrolled in study
Yes
Patients — no. (%) 419 (67.0) 427 (68.2)
APACHE II score§ 20±6.9 19.0±7.1
No
Patients — no. (%) 206 (33.0) 199 (31.8)
APACHE II score§ 15.0±6.1 15.8±6.5
APACHE II score§ 18.7±7.1 18.0±7.1
MEDS score¶ 8.0±3.4 7.9±3.3
SOFA score‖ 4.2±2.4 4.3±2.4
Severe condition in medical history — no./total no. (%)** 181/622 (29.1) 161/626 (25.7)
Site of infection — no. (%)
Lungs 228 (36.5) 207 (33.1)
Abdomen 40 (6.4) 51 (8.1)
Blood 97 (15.5) 86 (13.7)
Central nervous system 12 (1.9) 9 (1.4)
Soft tissue 39 (6.2) 39 (6.2)
Urinary tract 108 (17.3) 117 (18.7)
Other 21 (3.4) 37 (5.9)
No sepsis†† 4 (0.6) 3 (0.5)
Unknown 76 (12.2) 77 (12.3)
Change from initial antimicrobial drugs by 72 hr — no./total no. (%) 359/615 (58.4) 342/617 (55.4)
* Plus–minus values are means ±SD. There were no significant differences between the two groups except for age (P = 0.01). EGDT denotes
early, goal-directed therapy, and IQR interquartile range.
† Intravenous fluids include crystalloids and colloids measuring more than 20 ml in volume and all blood products.
‡ The use of supplemental oxygen was based on the fraction of inspired oxygen.
§ Scores on the Acute Physiology and Chronic Health Evaluation (APACHE) II range from 0 to 71, with higher scores indicating greater severity
of illness. The APACHE II score was calculated on the basis of the last recorded data before randomization.
¶ Scores on the Mortality in Emergency Department Sepsis (MEDS) scale range from 0 to 27, with higher scores indicating greater severity of
illness. The MEDS score was calculated on the basis of the last recorded data before randomization.
‖ Scores on the Sequential Organ Failure Assessment (SOFA) range from 0 to 24, with higher scores indicating a greater degree of organ
failure. The SOFA score was calculated on the basis of the last recorded data before randomization. The SOFA renal score was based on
the plasma creatinine level only and did not include urine output.
** Severe conditions in the medical history were defined according to the APACHE II score.
†† The lack of sepsis was confirmed after randomization.
6 n engl j med
65 nejm.org
Table 2. Interventions Delivered during and after the 6-Hour Intervention Period.*
Intervention Hour 0 to 6 Hour >6 to 72

EGDT Usual Care EGDT Usual Care
(N = 625) (N = 626) (N = 608) (N = 607)
Supplemental oxygen — no./total no. (%) 558/623 (89.6) 557/625 (89.1) 520/603 (86.2) 515/603 (85.4)
Insertion of central venous catheter with ScvO2
monitoring capability
Patients — no./total no. (%) 545/624 (87.3) 2/625 (0.3) NA NA
Before hour 1 — no./total no. (%) 459/543 (84.5) NA NA NA
Insertion of any central venous catheter
Median time from randomization to insertion 1.1 (0.8–1.5) 1.4 (0.6–2.9) NA NA
(IQR) — hr
Insertion of arterial catheter
Median time from randomization to insertion 1.1 (0.4–1.9) 1.0 (0.2–1.9) NA NA
(IQR) — hr
Median total intravenous fluids (IQR) — ml† 2000 (1150–3000) 1784 (1075–2775) 3623 (1800–6060) 3981 (1895–6291)
Intravenous colloids
Patients — no./total no. (%)† 197/623 (31.6) 180/625 (28.8) 171/603 (28.4) 150/603 (24.9)
Median volume (IQR) — ml 1000 (500–1500) 750 (500–1000) 750 (500–1750) 750 (500–1500)
Intravenous crystalloids
Patients — no./total no. (%)† 584/623 (93.7) 597/625 (95.5) 537/603 (89.1) 543/603 (90.0)
Vasopressor — no./total no. (%) 332/623 (53.3) 291/625 (46.6) 349/603 (57.9) 317/603 (52.6)
Dobutamine — no./total no. (%) 113/623 (18.1) 24/625 (3.8) 107/603 (17.7) 39/603 (6.5)
Red-cell transfusion
Patients — no./total no. (%) 55/623 (8.8) 24/625 (3.8) 76/603 (12.6) 51/603 (8.5)
Platelets
Fresh-frozen plasma
ICU admission — no./total no. (%) 551/625 (88.2) 467/626 (74.6) NA NA
Median time from randomization to ICU 1.2 (0.4–2.8) 1.2 (0.3–2.8) NA NA
admission (IQR) — hr
* ICU denotes intensive care unit, NA not applicable, and ScvO2 central venous oxygen saturation.
† Included in this category is the administration of more than 20 ml of an intravenous fluid.
were similar, but patients in the usual-care group umes were larger in the usual-care group. The
received larger volumes. More patients in the use of vasopressors and dobutamine remained
EGDT group received intravenous colloids, but higher in the EGDT group. Although the num-
overall volumes were similar in the two groups. bers of patients receiving platelets and fresh-fro-
The number of patients receiving intravenous zen plasma were similar in the two groups, the
crystalloids was similar in the two groups, but volume of platelets was larger in the EGDT group,
volumes were larger in the usual-care group. The whereas the volume of fresh-frozen plasma was
number of patients receiving red-cell transfu- higher in the usual-care group (Table 2, and Ta-
sions was higher in the EGDT group, but vol- ble S7 in the Supplementary Appendix). At 72
n engl j med nejm.org

66 7
hours, physiological, biochemical, and SOFA val- (Table 3). Sensitivity analyses for patients with a
ues were similar in the two groups (Table S8 in missing primary outcome (2 in the EGDT group
the Supplementary Appendix). and 6 in the usual-care group) showed relative
risks ranging from 0.99 to 1.03. There was no
Primary Outcome evidence of a learning-curve effect (P = 0.56). In
Mortality at 90 days was not significantly differ- the adherence-adjusted analysis, the relative risk
ent in the two groups, with deaths reported in was 1.02 (95% CI, 0.78 to 1.32; P = 0.90) (Table S9
184 of 623 patients (29.5%) in the EGDT group and Fig. S7 in the Supplementary Appendix).
versus 181 of 620 patients (29.2%) in the usual-
care group, with an unadjusted relative risk in Secondary Outcomes
the EGDT group of 1.01 (95% CI, 0.85 to 1.20; The mean SOFA score at 6 hours, the proportion
P = 0.90), for an absolute risk reduction of −0.3 of patients receiving advanced cardiovascular
percentage points (95% CI, −5.4 to 4.7). After ad- support, and the median length of stay in the
justment for baseline characteristics, the odds ICU were significantly greater in the EGDT group
ratio was 0.95 (95% CI, 0.74 to 1.24; P = 0.73) than in the usual-care group. No other secondary
Table 3. Study Outcomes.*
EGDT Usual Care Incremental Effect

Outcome (N = 625) (N = 626) (95% CI) P Value
Clinical effectiveness
Primary outcome: death from any cause at 90 days — 184/623 (29.5) 181/620 (29.2)
no./total no. (%)
Relative risk 1.01 (0.85 to 1.20) 0.90†
Absolute risk reduction — percentage points −0.3 (−5.4 to 4.7)
Unadjusted odds ratio 1.02 (0.80 to 1.30)
Adjusted odds ratio 0.95 (0.74 to 1.24) 0.73
Secondary outcomes
SOFA score‡
At 6 hr 6.4±3.8 5.6±3.8 0.8 (0.5 to 1.1)§ <0.001
At 72 hr 4.0±3.8 3.7±3.6 0.4 (−0.0 to 0.8)§ 0.056
Receipt of advanced cardiovascular support — 230/622 (37.0) 190/614 (30.9) 1.19 (1.02 to 1.40)¶ 0.026†
no./total no. (%)
Receipt of advanced respiratory support — 179/620 (28.9) 175/615 (28.5) 1.01 (0.85 to 1.21)¶ 0.90†
no./total no. (%)
Receipt of renal support — no./total no. (%) 88/620 (14.2) 81/614 (13.2) 1.08 (0.81 to 1.42)¶ 0.62†
Days free from advanced cardiovascular support up 20.3±11.9 20.6±11.8 −0.3 (−1.5 to 1.0)§ 0.63
to 28 days
Days free from advanced respiratory support up to 19.6±12.1 19.8±12.0 −0.2 (−1.5 to 1.1)§ 0.78
28 days
Days free from renal support up to 28 days 20.6±12.1 20.6±11.9 0.0 (−1.3 to 1.3)§ 0.97
Median length of stay in emergency department 1.5 (0.4 to 3.1) 1.3 (0.4 to 2.9) 0.34‖
(IQR) — hr
Median length of stay in ICU (IQR) — days 2.6 (1.0 to 5.8) 2.2 (0.0 to 5.3) 0.005‖
Median length of stay in hospital (IQR) — days 9 (4 to 21) 9 (4 to 18) 0.46‖
Death from any cause — no./total no. (%)
At 28 days 155/625 (24.8) 152/621 (24.5) 1.01 (0.83 to 1.23)¶ 0.90†
0.95 (0.73 to 1.25)** 0.73
At hospital discharge 160/625 (25.6) 154/625 (24.6) 1.04 (0.86 to 1.26)¶ 0.74†
0.98 (0.75 to 1.29)** 0.90
8 67 j med
n engl nejm.org
Table 3. (Continued.)
EGDT Usual Care Incremental Effect

Cost-effectiveness
Health-related quality of life on EQ-5D at
90 days†† 0.609±0.319 0.613±0.312 −0.004 (−0.051 to 0.044)§ 0.88
Quality-adjusted life-yr up to 90 days 0.054±0.048 0.054±0.048 −0.001 (−0.006 to 0.005)§ 0.85
Costs up to 90 days 0.26
Pounds 12,414±14,970 11,424±15,727 989 (−726 to 2,705)§
Dollars 17,647±21,280 16,239±22,356 1,406 (−1,032 to 3,845)§
Incremental net benefit up to 90 days‡‡ 0.25
Pounds NA NA −1,000 (−2,720 to 720)§
Dollars NA NA −1,422 (−3,866 to 1,023)§
Serious adverse events — no. (%) 30 (4.8) 26 (4.2) 1.16 (0.69 to 1.93)¶ 0.58†
* All values for the incremental effect are for the EGDT group as compared with the usual-care group. Plus–minus values are means ±SD.
† The P value was calculated with the use of Fisher’s exact test.
‡ Renal scores on the Sequential Organ Failure Assessment (SOFA) were based on the plasma creatinine level only. Patients in whom the
variables for SOFA renal and SOFA coagulation scores were not recorded between randomization and 6 hours had these values carried
forward from baseline, if recorded. Scores for 181 patients who died or were discharged before 48 hours (84 in the EGDT group and 97 in
the usual-care group) were not included in SOFA score at 72 hours.
§ This value is the difference between the means.
¶ This value is the relative risk.
‖ The P value was calculated by means of the Wilcoxon rank-sum test.
** This value is the adjusted odds ratio.
†† Scores on the European Quality of Life–5 Dimensions (EQ-5D) questionnaire range from 0 (death) to 1 (perfect health), with higher
scores indicating a better quality of life.
‡‡ The incremental net benefit was calculated according to methods of the National Institute for Health and Care Excellence (NICE) by multi-
plying the mean gain or loss in quality-adjusted life-years by £20,000 ($28,430) and subtracting from this value the incremental cost. The
currency conversion factor that was used was £1 equals $1.4215.
outcomes were significantly different (Table 3,

and Table S10 in the Supplementary Appendix). 1.00
There was no significant difference in the dura-

Usual care
tion of survival between the two groups (P = 0.63
0.75
by the log-rank test; adjusted hazard ratio, 0.94, EGDT
95% CI, 0.79 to 1.11; P = 0.46) (Fig. 2). Mean EQ-5D
0.50
scores and QALYs were similar in the two groups.
The average cost was higher in the EGDT group
(£12,414 [U.S. $17,647]) than in the usual-care 0.25
Adjusted hazard ratio, 0.94 (0.79–1.11); P=0.46
group (£11,424 [U.S. $16,239]), but the difference P=0.63 by log-rank test
was not significant (P = 0.26) (Table 3, and Tables 0.00
S11 through S16 and Fig. S8 in the Supplemen- 0 15 30 45 60 75 90
tary Appendix). The incremental net benefit for Days since Randomization
EGDT as compared with usual care was negative No. at Risk
EGDT 625 492 470 461 449 445 440
and similar across prespecified subgroups and Usual care 626 487 469 464 448 445 439
alternative scenarios that were considered in sen-
sitivity analyses (Tables S17 and S18 and Fig. S9 in Figure 2. Kaplan–Meier Survival Estimates.
the Supplementary Appendix). The probability Shown is the probability of survival for patients with severe sepsis receiving
that EGDT was cost-effective was below 20% early, goal-directed therapy (EGDT) and those receiving usual care at 90 days.
(Fig. S10 in the Supplementary Appendix).

68 9
Subgroup Analyses patients (0.60) than in the general population

There was no significant difference regarding matched for age and sex (0.80).25 On average, the
the effect of EGDT according to prespecified use of EGDT increased costs, and given similar
subgroups as defined by the degree of proto- QALYs in the two groups, the probability that
colized care used in the usual-care group, age, EGDT was cost-effective was below 20%.
MEDS score, SOFA score, or time from presenta- Our study was set in a real-world context and
tion at the emergency department to randomiza- in a large, representative, mixed sample of ap-
tion (P = 0.39 to 0.72 for interaction) (Table S9 proximately one quarter of NHS hospitals in
and Fig. S11 in the Supplementary Appendix). England. Site setup was rapid, and the study re-
cruited the full 1260 patients over a shorter time
Serious Adverse Events period than those of the two similar studies in
At least one serious adverse event was reported in the United States12 and Australasia.13 This factor
30 patients (4.8%) in the EGDT group and 26 minimized the potential for other changes in
patients (4.2%) in the usual-care group (P = 0.58) clinical practice to affect outcomes. Unlike previ-
(Table 3, and Table S19 in the Supplementary Ap- ous studies, our study reports on quality of life and
pendix). Four serious adverse events were report- cost-effectiveness at 90 days. Loss to follow-up
ed as being related to EGDT (two cases of pulmo- was low, and all analyses were conducted ac-
nary edema and one of arrhythmia, which were cording to a prespecified statistical analysis plan
deemed to be probably related, and one case of and included adjusted analyses to address the
myocardial ischemia, which was deemed to be degree of adherence to EGDT and the possibility
definitely related), as compared with four events of the existence of a learning curve for its delivery.
(in three patients) related to usual care (two cases Our study has several limitations. As in all
of pneumothorax and one case of pulmonary studies that enroll patients presenting to emer-
edema, which were deemed to be probably relat- gency departments, recruitment was more chal-
ed, and one case of ventricular fibrillation, which lenging on weekends and during out-of-office
was deemed to be definitely related). hours; overall, only one third of eligible patients
were recruited, although exclusion from the study
Discussion by a clinician was rare. The intervention could
not be blinded, but the risk of bias was mini-
In our study involving adults with early signs of mized through central randomization to ensure
septic shock who presented to emergency depart- the concealment of study-group assignments and
ments in England, there was no significant dif- the use of a primary outcome that was not sub-
ference in mortality at 90 days among those re- ject to observer bias. Since the rate of death was
ceiving 6 hours of EGDT and those receiving lower than anticipated, our study data may not
usual resuscitation. Although the overall rate of apply to settings with higher mortality.
death in the usual-care group was lower than an- Unlike Rivers et al., in their 2001 study, we
ticipated (29% rather than 40%), it is unlikely did not observe a significant reduction in hospi-
that patients in the EGDT group would have a tal mortality with the use of EGDT. Many aspects
relative reduction of more than 15% in risk. The of initial sepsis management have changed dur-
effect of EGDT was not significantly different in ing the past 15 years, as can be seen in compar-
prespecified subgroups. More patients receiving ing the usual-care groups. In our study, as com-
EGDT were admitted to and spent more days in pared with the study by Rivers et al., mortality
the ICU. Treatment intensity was greater in the was substantially reduced, randomization oc-
EGDT group, driven by adherence to the protocol curred later, patients appeared to be less sick at
and indicated by the increased use of central ve- baseline (with lower blood lactate levels and
nous catheters, intravenous fluids, vasoactive Acute Physiology and Chronic Health Evaluation
drugs, and red-cell transfusions. Increased inten- [APACHE] II scores), and all patients received
sity was reflected by significantly higher SOFA antibiotics before randomization. In addition,
scores and more days of receiving advanced car- our patients received much lower volumes of
diovascular support. There were no significant intravenous fluids and more vasoactive drugs
differences in any other secondary outcomes, in- (Table S20 in the Supplementary Appendix).
cluding health-related quality of life, which was The level of adherence to EGDT was good and
substantially poorer in this severely ill group of was equivalent to adherence levels in the ProCESS12
10 n engl j med
69 nejm.org
and ARISE13 trials and higher than reported rates levels of in-hospital survival in patients receiving
of compliance with the SSC guidelines.26 Most usual care that were similar to those achieved
outcomes were similar to those reported in the with EGDT in the earlier study for patients with
ProCESS and ARISE trials, although the rate of septic shock who were identified early and re-
death at 90 days that was reported in our study was ceived intravenous antibiotics and adequate fluid
lower than that in the ProCESS trial but higher than resuscitation. The addition of continuous ScvO2
that in the ARISE trial. Of note, a higher proportion monitoring and strict protocolization did not
of patients in our study than in the ProCESS and improve outcomes in the EGDT group. Our re-
ARISE trials met both of the two inclusion crite- sults complete the planned trio of studies of
ria — refractory hypotension and hyperlactate- EGDT, all of which showed that EGDT was not
mia (Table S21 in the Supplementary Appendix) superior to usual care.
— a factor that is associated in our national ICU The views expressed in this article are those of the authors
database with a doubling of hospital mortality and do not necessarily reflect those of the National Institute for
Health Research, the Health Technology Assessment Programme,
(Table S22 in the Supplementary Appendix). the National Health Service, or the Department of Health.
In conclusion, our results suggest that tech- Supported by a grant from the United Kingdom National In-
niques used in usual resuscitation have evolved stitute for Health Research Health Technology Assessment Pro-
gramme (project number 07/37/47).
over the 15 years since the landmark study by Disclosure forms provided by the authors are available with
Rivers et al.9 In our study, NHS hospitals achieved the full text of this article at NEJM.org.
References
1. Jawad I, Lukšić I, Rafnsson SB. Assess- Early goal-directed therapy in the treat- the European Society of Intensive Care
ing available information on the burden ment of severe sepsis and septic shock. Medicine. Intensive Care Med 1996;22:
of sepsis: global estimates of incidence, N Engl J Med 2001;345:1368-77. 707-10.
prevalence and mortality. J Glob Health 10. Ho BC, Bellomo R, McGain F, et al. 19. Critical Care Minimum Dataset Full
2012;2(1):010404. The incidence and outcome of septic Specification (CCMDS). London: Infor-
2. Levy MM, Artigas A, Phillips GS, et al. shock patients in the absence of early- mation Standards Board for Health and
Outcomes of the Surviving Sepsis Cam- goal directed therapy. Crit Care 2006;10: Social Care, March 2010.
paign in intensive care units in the USA R80. 20. Harrison DA, Brady AR, Rowan K.
and Europe: a prospective cohort study. 11. Chapman M, Gattas D, Suntharalin- Case mix, outcome and length of stay for
Lancet Infect Dis 2012;12:919-24. gam G. Why is early goal-directed therapy admissions to adult, general critical care
3. Kaukonen KM, Bailey M, Suzuki S, successful — is it the technology? Crit units in England, Wales and Northern Ire-
Pilcher D, Bellomo R. Mortality related to Care 2005;9:307-8. land: the Intensive Care National Audit &
severe sepsis and septic shock among 12. The ProCESS Investigators. A ran- Research Centre Case Mix Programme
critically ill patients in Australia and New domized trial of protocol-based care for database. Crit Care 2004;8:R99-R111.
Zealand, 2000-2012. JAMA 2014;311: early septic shock. N Engl J Med 2014; 21. Harvey SE, Elbourne D, Ashcroft J,
1308-16. 370:1683-93. Jones CM, Rowan K. Informed consent in
4. Cuthbertson BH, Elders A, Hall S, et al. 13. The ARISE Investigators and the clinical trials in critical care: experience
Mortality and quality of life in the five ANZICS Clinical Trials Group. Goal- from the PAC-Man Study. Intensive Care
years after severe sepsis. Crit Care 2013; directed resuscitation for patients with Med 2006;32:2020-5.
17:R70. early septic shock. N Engl J Med 2014; 22. Power GS, Harrison DA, Mouncey PR,
5. Torio CM, Andrews RM. Healthcare 371:1496-506. Osborn TM, Harvey SE, Rowan KM. The
Cost and Utilization Project (HCUP) sta- 14. The ProCESS/ARISE/ProMISe Method- Protocolised Management in Sepsis
tistical briefs — national inpatient hospi- ology Writing Committee. Harmonizing (ProMISe) trial statistical analysis plan.
tal costs: the most expensive conditions international trials of early goal-directed Crit Care Resusc 2013;15:311-7.
by payer, 2011. Statistical brief #160. resuscitation for severe sepsis and septic 23. Shapiro NI, Wolfe RE, Moore RB,
Rockville, MD: Agency for Health Care shock: methodology of ProCESS, ARISE, Smith E, Burdick E, Bates DW. Mortality
Policy and Research, 2013. and ProMISe. Intensive Care Med 2013; in Emergency Department Sepsis (MEDS)
6. Dellinger RP, Carlet JM, Masur H, et al. 39:1760-75. score: a prospectively derived and vali-
Surviving Sepsis Campaign guidelines for 15. Reade MC, Delaney A, Bailey MJ, et al. dated clinical prediction rule. Crit Care
management of severe sepsis and septic Prospective meta-analysis using individu- Med 2003;31:670-5.
shock. Crit Care Med 2004;32:858-73. al patient data in intensive care medicine. 24. Guide to the methods of technology
[Errata, Crit Care Med 2004;32:1448, Intensive Care Med 2010;36:11-21. appraisal. London: National Institute for
2169-70.] 16. Bone RC, Balk RA, Cerra FB, et al. Health and Care Excellence, 2013.
7. Dellinger RP, Levy MM, Carlet JM, Definitions for sepsis and organ failure 25. Ara R, Brazier JE. Populating an eco-
et al. Surviving Sepsis Campaign: inter- and guidelines for the use of innovative nomic model with health state utility val-
national guidelines for management of therapies in sepsis. Chest 1992;101:1644- ues: moving toward better practice. Value
severe sepsis and septic shock: 2008. Crit 55. Health 2010;13:509-18.
Care Med 2008;36:296-327. 17. Mental Capacity Act 2005. London: 26. Levy MM, Rhodes A, Phillips GS, et al.
8. Dellinger RP, Levy MM, Rhodes A, The Stationery Office, 2005. Surviving Sepsis Campaign: association
et al. Surviving Sepsis Campaign: interna- 18. Vincent JL, Moreno R, Takala J, et al. between performance metrics and out-
tional guidelines for management of se- The SOFA (Sepsis-related Organ Failure comes in a 7.5-year study. Crit Care Med
vere sepsis and septic shock: 2012. Crit Assessment) score to describe organ dys- 2015;43:3-12.
Care Med 2013;41:580-637. function/failure: on behalf of the Work- Copyright © 2015 Massachusetts Medical Society.
9. Rivers E, Nguyen B, Havstad S, et al. ing Group on Sepsis-Related Problems of
n engl j med
70 nejm.org 11
original article
Systemic Inflammatory Response Syndrome

Criteria in Defining Severe Sepsis
Kirsi-Maija Kaukonen, M.D., Ph.D., Michael Bailey, Ph.D., David Pilcher, F.C.I.C.M.,
D. Jamie Cooper, M.D., Ph.D., and Rinaldo Bellomo, M.D., Ph.D.
A BS T R AC T
BACKGROUND
The consensus definition of severe sepsis requires suspected or proven infection, From the Australian and New Zealand
organ failure, and signs that meet two or more criteria for the systemic inflamma- Intensive Care Research Centre, School
of Public Health and Preventive Medi-
tory response syndrome (SIRS). We aimed to test the sensitivity, face validity, and cine, Monash University (K.-M.K., M.B.,
construct validity of this approach. D.P., D.J.C., R.B.), the Australian and
New Zealand Intensive Care Society Cen-
tre for Outcome and Resource Evaluation
METHODS (D.P.), and the Department of Intensive
We studied data from patients from 172 intensive care units in Australia and New Care, Alfred Hospital (D.P.), Melbourne,
Zealand from 2000 through 2013. We identified patients with infection and organ VIC, and the Intensive Care Unit, Austin
Health, Heidelberg, VIC (R.B.) — all in
failure and categorized them according to whether they had signs meeting two or Australia; and the Neurosurgical Unit,
more SIRS criteria (SIRS-positive severe sepsis) or less than two SIRS criteria (SIRS- Department of Anesthesiology, Intensive
negative severe sepsis). We compared their characteristics and outcomes and as- Care and Pain Medicine, Helsinki Univer-
sity Central Hospital, Helsinki (K.-M.K.).
sessed them for the presence of a step increase in the risk of death at a threshold Address reprint requests to Dr. Bellomo
of two SIRS criteria. at the Australian and New Zealand Inten-
sive Care Research Centre, School of
Public Health and Preventive Medicine,
RESULTS Monash University, Level 6, the Alfred
Of 1,171,797 patients, a total of 109,663 had infection and organ failure. Among Centre, 99 Commercial Rd., Melbourne, VIC
these, 96,385 patients (87.9%) had SIRS-positive severe sepsis and 13,278 (12.1%) 3004, Australia, or at rinaldo.bellomo@
austin.org.au.
had SIRS-negative severe sepsis. Over a period of 14 years, these groups had similar
characteristics and changes in mortality (SIRS-positive group: from 36.1% [829 of This article was published on March 17,
2296 patients] to 18.3% [2037 of 11,119], P<0.001; SIRS-negative group: from 27.7% 2015, at NEJM.org.
[100 of 361] to 9.3% [122 of 1315], P<0.001). Moreover, this pattern remained DOI: 10.1056/NEJMoa1415236
similar after adjustment for baseline characteristics (odds ratio in the SIRS-positive Copyright © 2015 Massachusetts Medical Society.
group, 0.96; 95% confidence interval [CI], 0.96 to 0.97; odds ratio in the SIRS-
negative group, 0.96; 95% CI, 0.94 to 0.98; P = 0.12 for between-group difference).
In the adjusted analysis, mortality increased linearly with each additional SIRS
criterion (odds ratio for each additional criterion, 1.13; 95% CI, 1.11 to 1.15;
P<0.001) without any transitional increase in risk at a threshold of two SIRS criteria.
CONCLUSIONS
The need for two or more SIRS criteria to define severe sepsis excluded one in eight
otherwise similar patients with infection, organ failure, and substantial mortality
and failed to define a transition point in the risk of death. (Funded by the Austra-
lian and New Zealand Intensive Care Research Centre.)
n engl j med nejm.org 71 1

The New England
personal Medicine
MULARCZYK on April 10,Medical
© Massachusetts 2015. For personal
Society. use only.
reserved.
S
evere sepsis is a major cause of admis- ME THODS
sion to the intensive care unit (ICU) and
death.1,2 The criteria according to the sys- STUDY DESIGN
temic inflammatory response syndrome (SIRS) We conducted a retrospective study from January
were described 23 years ago as a clinical expres- 1, 2000, to December 31, 2013, using data from
sion of the host response to inflammation.3 In the Australia and New Zealand Intensive Care
this context and in the presence of symptoms Society (ANZICS) Adult Patient Database (APD),16
meeting two or more SIRS criteria, severe sepsis a high-quality database run by the ANZICS Cen-
was seen as evolving from infection to sepsis, tre for Outcome and Resource Evaluation. The
severe sepsis, and septic shock, in order of in- ANZICS APD includes information on more than
creasing severity. This approach was codified by 90% of all ICU admissions in Australia and New
the consensus statement of the American College Zealand. The Alfred Hospital Human Research
of Chest Physicians and Society of Critical Care Ethics Committee, Melbourne, Australia, ap-
Medicine in 19923 and has been the predominant proved the study with a waiver of informed con-
approach to classifying sepsis.4-11 sent. The data were gathered as a part of routine
However, the need for patients to meet two or quality-assurance benchmarking processes by
more SIRS criteria has been criticized because of means of clinical registry surveillance by collec-
a low specificity for infection12,13 within 24 hours tors in the participating ICUs.
after admission to the ICU.14 Moreover, some pa-
tients (the elderly and those taking medications DEFINITIONS
that affect heart rate, respiratory rate, or body We defined sepsis, severe sepsis, and septic
temperature) may not have symptoms meeting shock according to the American College of
two or more SIRS criteria, despite having infec- Chest Physicians–Society of Critical Care Medi-
tion and organ failure. Thus, the face validity cine consensus definition.3 We used infection-
and sensitivity of two or more SIRS criteria in related diagnoses according to the Acute Physiol-
the diagnosis of severe sepsis remain unclear.15 ogy and Chronic Health Evaluation (APACHE) III
The face validity and sensitivity can, however, be at admission to infer the presence of suspected
indirectly empirically tested by defining the or proven infection (see the Supplementary Ap-
number, characteristics, and outcome of patients pendix, available with the full text of this article
in the ICU who have infection and organ failure at NEJM.org). We defined organ failure in the
and who do not have symptoms meeting two or first 24 hours after ICU admission as a Sequen-
more SIRS criteria but who can be confidently tial Organ Failure Assessment (SOFA) score of
assumed to have severe sepsis on the basis of 3 or higher (on a scale from 0 to 4, with higher
their presentation. Moreover, the construct va- scores indicating more severe organ failure)2,17,18
lidity of the SIRS criteria can be empirically as- (see the Supplementary Appendix).
sessed by testing whether the cutoff value of two We diagnosed severe sepsis if a patient had
criteria represents a significant transitional in- one of the ANZICS APD diagnostic codes for
crease in the risk of death to logically justify its infection and organ failure or one of the follow-
choice (in preference to one or three or four ing prespecified additional diagnostic categories
criteria) to diagnose or define severe sepsis. in the ANZICS APD diagnostic codes for infec-
We hypothesized that in the first 24 hours tion: sepsis due to infection other than from the
after ICU admission, the presence of symptoms urinary tract with organ failure, sepsis due to
meeting two or more SIRS criteria would have urinary tract infection with organ failure, sepsis
low face and construct validity and sensitivity with shock due to infection other than from the
and that the fulfillment of two criteria would urinary tract, or sepsis with shock due to uri-
not identify a transitional increase in an other- nary tract infection. We applied the consensus
wise linear increased risk of death that would be SIRS criteria to all the data analyses (see the
logically expected with each additional criterion. Supplementary Appendix).3
We tested these hypotheses by conducting a All the above criteria were assessed within
study of all ICU admissions in Australia and the first 24 hours after ICU admission. Patient
New Zealand over the past 14 years. follow-up was available only for the duration of
2 n engl j med
72 nejm.org
The New England

For of Medicine
Downloaded from nejm.org by MIKE Copyright
reserved.
SIRS Criteria in Defining Severe Sepsis
hospital stay, and the primary outcome was in- tion was then used to generate each patient’s
hospital mortality. probability of being SIRS-positive.
To investigate the similarities of differences
STUDY POPULATIONS in hospital outcomes over time for all the pa-
Patients with SIRS-positive severe sepsis fulfilled tients with sepsis, logistic-regression models
the following requirements: symptoms meeting were used (see the Supplementary Appendix),
two or more SIRS criteria, an APACHE III diag- with the Australian and New Zealand calibrated
nosis of infection at admission with at least one Risk of Death (ANZROD) model.19 This model
organ failure, or an APACHE III diagnosis of se- has been shown to perform better than the
vere sepsis or septic shock at admission. Patients APACHE III in Australia and New Zealand.20,21
with SIRS-negative severe sepsis fulfilled the fol- To further determine the predictive capacity
lowing requirements: symptoms meeting fewer of using two or more SIRS criteria to identify an
than two SIRS criteria, and an APACHE III diag- increase in the risk of death, SIRS was consid-
nosis of infection at admission with at least one ered first as a dichotomous variable (≥2 SIRS
organ failure or an APACHE III diagnosis of se- criteria vs. 0 to 1 SIRS criterion) and second as
vere sepsis or septic shock at admission. an ordinal variable from 0 to 4, reflecting the
A descriptive analysis was performed on four number of SIRS criteria met. Because the SIRS
prespecified subgroups of patients: patients with criteria are derived from components related to
septic shock, those receiving mechanical ventila- the severity of sepsis in the patient, we deter-
tion, those with acute renal failure, and those mined the predictive capacity of the SIRS criteria
with an APACHE II score of more than 24 (on a by adjusting for the severity markers without
scale from 0 to 71, with higher scores indicating SIRS criteria, in conjunction with year, site of
more severe disease). An additional post hoc admission, and probability of being SIRS-posi-
analysis was performed to compare differences tive. Data were randomly divided into two equal
between community-acquired sepsis and hospi- sets, with the first used for model development
tal-acquired sepsis as determined by the ICU and the second used for model validation. Mod-
admission source. el discrimination was determined with the use
of the area under the curve (AUC), whereas clas-
STATISTICAL ANALYSIS sification error was determined with the use of
Data are presented as numbers and percentages, integrated discrimination improvement and net
means and standard deviations, medians and in- reclassification improvement with results re-
terquartile ranges, or proportions with 95% con- ported specifically for the validation data set.
fidence intervals. Accordingly, chi-square tests Net reclassification improvement was deter-
for equal proportion, Student’s t-test, or the Wil- mined first as category-free improvement and
coxon rank-sum test were used to test differenc- second stratified in quartiles of risk.
es. No assumptions were made for missing data, To determine whether predictors of death dif-
with multivariable analysis performed on data fered significantly between SIRS-positive sepsis
from patients who had complete data only (see and SIRS-negative sepsis, a multivariable logis-
the Supplementary Appendix). tic-regression model for mortality among all the
To identify independent differences at base- patients with sepsis was developed with the use
line that may exist between patients with SIRS- of the LASSO method, with variable inclusion
positive severe sepsis and those with SIRS-nega- determined by means of the Schwarz–Bayesian
tive severe sepsis, we applied multivariable information criterion. Interactions between iden-
logistic regression to data from all the patients tified predictors of mortality and SIRS status
with severe sepsis with SIRS-positive status as were then applied to determine whether the na-
the outcome (see the Supplementary Appendix). ture of the relationship to mortality differed ac-
The multivariable model was developed with the cording to SIRS status.
use of the least absolute shrinkage and selection All the data were analyzed with the use of
operator (LASSO) method, with variable inclu- SAS software, version 9.4 (SAS Institute). A two-
sion determined by the Schwarz–Bayesian infor- sided P value of less than 0.01 was considered to
mation criterion. The resulting prediction equa- indicate statistical significance.
n engl j med
73 nejm.org 3
The For
Downloaded from collections.nejm.org. Newpersonal
Englanduse
Journal
only.ofNoMedicine
Downloaded from nejm.org by MIKE©MULARCZYK
Copyright on April Society.
Massachusetts Medical 10, 2015.All
For personal
reserved.
R E SULT S The annual proportions of patients with SIRS-

positive severe sepsis or SIRS-negative severe sepsis
STUDY PATIENTS among all ICU admissions are presented in Figure
We studied data from 1,171,797 patients in 172 S1 in the Supplementary Appendix, as are the
ICUs. A total of 1,062,134 patients did not have annual proportions of patients with SIRS-posi-
sepsis, and 109,663 had infection and organ dys- tive severe sepsis or SIRS-negative severe sepsis.
function. Of the patients with infection and organ Risk factors for SIRS-positive status are presented
dysfunction, 96,385 patients (87.9%) had SIRS- in Table S1 in the Supplementary Appendix.
positive severe sepsis and 13,278 (12.1%) had
SIRS-negative severe sepsis (Table 1). Patients with SIRS IN SEVERE SEPSIS
SIRS-positive severe sepsis were younger, were The distribution of SIRS criteria is presented in
more severely ill, and had higher mortality than Table 2. The most frequent SIRS criterion that was
those with SIRS-negative severe sepsis. They were met in patients with SIRS-positive severe sepsis
also more likely to have septic shock or acute kid- was an increased heart rate, followed by an in-
ney injury but less likely to have a surgical admis- creased respiratory rate or a low partial pressure
sion or to be discharged home (Table 1). of arterial carbon dioxide (Paco2) and an abnor-
Table 1. Baseline Characteristics and Hospital Outcomes of Patients with Severe Sepsis, According to Status with Respect to Criteria
for the Systemic Inflammatory Response Syndrome (SIRS).*
Patients with Patients with

Characteristic All Patients SIRS-Positive Sepsis SIRS-Negative Sepsis P Value
no. of patients no. of patients no. of patients
with data with data with data
Age — yr 109,663 96,385 13,278 <0.001
Median 66.0 65.8 68.3
Interquartile range 52.2–76.6 51.9–76.4 55.5–77.7
Male sex — no. (%) 109,663 60,484 (55.2) 96,385 52,932 (54.9) 13,278 7552 (56.9) <0.001
Surgical admission — no. (%) 109,663 23,630 (21.5) 96,385 18,441 (19.1) 13,278 5189 (39.1) <0.001
APACHE III score† 105,674 71.7±30.1 93,466 73.7±30.1 12,208 56.7±26.1 <0.001
Risk of death — %‡
APACHE III 105,320 93,142 12,178 <0.001
Median 22 24 11
Interquartile range 9–48 10–50 4–26
ANZROD model
With SIRS components 105,428 93,251 12,177 <0.001
Median 14 16 8
Without SIRS components 105,428 93,251 12,177 <0.001
Median 15 16 9
Duration of stay
In ICU — hr 109,579 96,313 13,266 <0.001
Median 80 85 57
In hospital — days 108,565 95,507 13,058 <0.001
Median 12.9 13.1 11.2
Interquartile range 6.6–25.3 6.7–25.8 5.8–21.9
4 n engl j med
74 nejm.org
The New
by MIKE MULARCZYK
April 10, 2015. For
Society. personal
All rights use only. No other uses without permission.
reserved.
Patients with Patients with

Characteristic All Patients SIRS-Positive Sepsis SIRS-Negative Sepsis P Value
no. of patients no. of patients no. of patients
with data with data with data
Treatment limitations — no. (%)§ 109,663 4,928 (4.5) 96,385 4,535 (4.7) 13,278 393 (3.0) <0.001
Hospital outcome — no. (%) 109,663 96,385 13,278
Death 25,713 (23.4) 23,577 (24.5) 2136 (16.1) <0.001
Discharge
Home 60,292 (55.0) 52,000 (54.0) 8292 (62.4) <0.001
To rehabilitation or long-term 7,781 (7.1) 6,837 (7.1) 944 (7.1) 0.95
care facility
To other hospital 13,168 (12.0) 11,540 (12.0) 1628 (12.3) 0.34
Subgroup — no. (%)
Septic shock 109,663 61,483 (56.1) 96,385 55,876 (58.0) 13,278 5607 (42.2) <0.001
Mechanical ventilation 109,643 58,678 (53.5) 96,380 51,359 (53.3) 13,263 7319 (55.2) <0.001
Acute renal failure 109,556 19,780 (18.1) 96,328 18,229 (18.9) 13,228 1551 (11.7) <0.001
APACHE II score >24¶ 106,621 30,677 (28.8) 94,041 28,778 (30.6) 12,580 1899 (15.1) <0.001
* Plus–minus values are means ±SD. SIRS-positive status was defined if the patient fulfilled at least two SIRS criteria, and SIRS-negative sta-
tus if the patient fulfilled zero or one SIRS criterion. ICU denotes intensive care unit.
† Scores on the Acute Physiology and Chronic Health Evaluation (APACHE) III range from 0 to 299, with higher scores indicating a greater se-
verity of illness.
‡ The risk of death was assessed by means of the APACHE III model, which is based on data from 40 U.S. hospitals, and by means of the
Australian and New Zealand Risk of Death (ANZROD) model,19 which has been shown to perform better than the APACHE III in Australia
and New Zealand.
§ Patients with treatment limitations are admitted to the ICU with the intention of providing limited treatment, which may include, for example,
the withholding of cardiopulmonary resuscitation in case of cardiac arrest, the withholding of dialysis, or the withholding of endotracheal
intubation.
¶ Scores on the APACHE II range from 0 to 71, with higher scores indicating more severe disease.
mal white-cell count. In patients with SIRS-neg- OUTCOMES

ative severe sepsis, the most frequent single crite- Absolute mortality decreased similarly in the two
rion that was met was an abnormal white-cell groups (Fig. 1). The rate of death in SIRS-positive
count, followed by an increased heart rate and patients decreased from 36.1% (in 829 of 2296
increased respiratory rate or a low Paco2 (Table 2). patients) in 2000 to 18.3% (in 2037 of 11,119 pa-
Of the patients with SIRS-negative sepsis, 20% did tients) in 2013 (P<0.001). Over the same period,
not fulfill any SIRS criteria, and 80% fulfilled in SIRS-negative patients, the rate of death de-
one SIRS criterion (Table 2). The distribution of creased from 27.7% (in 100 of 361 patients) to
baseline characteristics among patients with 8.5% (in 112 of 1315 patients) (P<0.001). These
SIRS-negative severe sepsis, stratified according to changes represent an annual rate of absolute de-
the presence of zero or one SIRS criterion, is pre- crease of 1.3 percentage points in each group,
sented in Table S2 in the Supplementary Appen- and a reduction in relative risk of 49.3% (45.7 to
dix. The patients with SIRS-positive severe sepsis 52.6%) and 66.5% (57.5 to 73.6%), respectively.
were significantly more likely than those with There was a very similar annual decline in
SIRS-negative severe sepsis to present with com- adjusted mortality in the two groups (annual
munity-acquired infection (38.1% vs. 28.1%, odds ratio in the SIRS-positive group, 0.96; 95%
P<0.001) or hospital-acquired infection (28.0% confidence interval [CI], 0.96 to 0.97; odds ratio
vs. 18.3%, P<0.001) (Tables S1 and S3 in the Sup- in the SIRS-negative group, 0.96; 95% CI, 0.94 to
plementary Appendix). 0.98; P<0.001 for both comparisons) (P = 0.12 for

75 5
The New England
reserved.
Table 2. Distribution of Signs Meeting SIRS Criteria in Patients with Severe Sepsis, According to SIRS-Positive
and SIRS-Negative Status.*
Patients with SIRS-Positive Patients with SIRS-Negative

All Patients Severe Sepsis Severe Sepsis
Variable (N = 109,663) (N = 96,385) (N = 13,278)
SIRS criterion met — no. (%)†
Abnormal temperature 64,365 (58.7) 62,430 (64.8) 1,935 (14.6)
High 33,059 (30.1) 32,605 (33.8) 454 (3.4)
Low 36,130 (32.9) 34,599 (35.9) 1,531 (11.5)
Increased heart rate 83,493 (76.1) 80,747 (83.8) 2,746 (20.7)
Increased respiratory rate or 76,558 (69.8) 74,043 (76.8) 2,515 (18.9)
decreased Paco2
Abnormal white-cell count 76,823 (70.1) 73,365 (76.1) 3,458 (26.0)
High 64,720 (59.0) 61,602 (63.9) 3,118 (23.5)
Low 12,967 (11.8) 12,616 (13.1) 351 (2.6)
No. of SIRS criteria met
Median 3 3 1
Distribution
>1 96,385 (87.9) 96,385 (100) 0
0 2,624 (2.4) 0 2,624 (19.8)
1 10,654 (9.7) 0 10,654 (80.2)
2 26,820 (24.5) 26,820 (27.8) 0
3 41,315 (37.7) 41,315 (42.9) 0
4 28,250 (25.7) 28,250 (29.3) 0
* P<0.001 for all comparisons between the SIRS-positive group and the SIRS-negative group. Paco2 denotes partial pres-
sure of arterial carbon dioxide.
† SIRS criteria are defined in the Supplementary Appendix. Patients may have more than one criterion.
between-group difference) (Fig. 1 and Table 3). facility increased significantly in each group
The proportion of patients discharged home in- (P<0.001 for both comparisons), but there was
creased in the same way in the two groups (Fig. no significant between-group difference (P = 0.20)
S2 in the Supplementary Appendix). The adjust- (Table 3).
ed annual odds for discharge home increased
significantly in the SIRS-positive group (annual PREDICTION OF MORTALITY
odds ratio, 1.02; 95% CI, 1.01 to 1.02; P<0.001) When the analysis included a modified risk of
but not in the SIRS-negative group (annual odds death (with SIRS components removed), the prob-
ratio, 1.01; 95% CI, 0.99 to 1.02; P = 0.29) (Table ability of being SIRS-positive, year of admission,
3). However, there was no significant difference study site, and SIRS status (positive or negative)
between the two groups in the adjusted rate of as covariates in the logistic-regression analysis,
increase (P = 0.17). The unadjusted rate of dis- being SIRS-positive independently increased the
charge to another hospital appeared to increase risk of death by 26% (odds ratio, 1.26; 95% CI,
similarly in the two groups (Fig. S2 in the Sup- 1.18 to 1.34; P<0.001). In a similar model that
plementary Appendix). However, in the adjusted included the number of SIRS criteria from 0 to 4,
analysis, there was no significant change over a 13% linear increase in mortality was associated
time in either group (Table 3). The adjusted rate with each additional SIRS criterion (odds ratio
of discharge to a rehabilitation or long-term care for each additional criterion, 1.13; 95% CI, 1.11
6 n engl j med
76 nejm.org
The New EnglandFor

Downloaded from nejm.org by MIKE MULARCZYK on April 10, Medical
Copyright © Massachusetts 2015. ForSociety.
personalAlluse only.reserved.
to 1.15; P<0.001) (Table S4 in the Supplementary

Appendix) without any transitional increase in SIRS-positive sepsis SIRS-negative sepsis
risk when two criteria were met (Fig. 2). When a
A Unadjusted Annual Mortality
comparison of discrimination and reclassifica-
40
tion was performed between SIRS criteria as a
binomial variable (≥2 criteria vs. <2 criteria) and 35
SIRS criteria as a continuous variable (0 to 4 cri- 30
Patients Who Died (%)

teria), there were modest but significant im-
25
provements noted for the continuous model in
the AUC, integrated discrimination improvement, 20
and net reclassification improvement when pa- 15
tients were stratified into quartiles of risk (Table
10
S4 in the Supplementary Appendix).
5
PREDICTION OF MORTALITY ACCORDING TO 0
SIRS-POSITIVE OR SIRS-NEGATIVE STATUS
99
00
01
02
03
04
05
06
07
08
09
10
11
12
13
19
20
20
20
20
20
20
20
20
20
20
20
20
20
20
There were several predictors of mortality among
patients with SIRS-positive sepsis or SIRS-severe B Adjusted Annual Odds of Death
sepsis (Table S5 in the Supplementary Appendix). 1.6
Of 16 identified predictors of mortality, the only 1.4
variable that differed significantly between the two
groups was illness severity (Table S5 in the Supple- 1.2
mentary Appendix).
Odds of Death
1.0
0.8
DISCUSSION
0.6
We studied the sensitivity, face validity, and con- 0.4

struct validity of the rule of using two or more
0.2
SIRS criteria for the diagnosis of severe sepsis in
the first 24 hours after ICU admission. We found 0.0
99
00
01
02
03
04
05
06
07
08
09
10
11
12
13
that the SIRS-criteria rule missed one patient in
19
20
20
20
20
20
20
20
20
20
20
20
20
20
20
eight with severe sepsis. Such patients with SIRS-
negative severe sepsis had lower but still substan- Figure 1. Mortality among Patients with Severe Sepsis, According to Status
with Respect to Criteria for the Systemic Inflammatory Response Syndrome
tial mortality, as compared with patients with (SIRS).
SIRS-positive sepsis, and the incidence, propor- Patients were categorized according to whether they had symptoms meet-
tion, and mortality decreased over time almost ing two or more SIRS criteria (SIRS-positive sepsis) or symptoms meeting
identically to the rates among patients with SIRS- less than two SIRS criteria (SIRS-negative sepsis). Panel A shows the un-
positive sepsis. In addition, their unadjusted and adjusted annual mortality among patients in the two groups from 2000
adjusted discharge rates to a rehabilitation or long- through 2013, and Panel B shows the adjusted annual odds of death. The
bars represent 95% confidence intervals.
term care facility were also similar, as were predic-
tors of mortality. Finally, in the adjusted analysis,
mortality increased linearly with each additional
SIRS criterion from 0 to 4 without any transitional the observation that an increase in the number
increase in risk at a threshold of two criteria. of SIRS criteria met was associated with a worse
SIRS criteria were described more than two outcome in critically ill patients, regardless of
decades ago, and the signs meeting these crite- infection status.15,22
ria have been assumed to indicate a clinical re- The presence of signs meeting two or more
sponse to inflammation.3 Infection was seen to SIRS criteria is common in all patients in the
require the presence of such signs to help define ICU but is not specific for infection.15 Among
the transition to sepsis, severe sepsis, and septic patients in the emergency department, 38% of
shock.15,22,23 This approach was supported by those with SIRS-positive severe sepsis have an

77 7
The New England
personal Medicine
MULARCZYK on April 10,
© Massachusetts 2015. For
Medical personal
Society. use only.
reserved.
Table 3. Annual Change in Hospital Outcomes from 2000 through 2013 and Between-Group Differences.
P Value for
Patients with SIRS-Positive Patients with SIRS-Negative Between-Group
Hospital Outcome Severe Sepsis Severe Sepsis Comparison
P Value for P Value for
Adjusted Odds Change over Adjusted Odds Change over
Ratio (95% CI) Study Period Ratio (95% CI) Study Period
Death 0.96 (0.96–0.97) <0.001 0.96 (0.94–0.98) <0.001 0.12
Discharge
Home 1.02 (1.01–1.02) <0.001 1.01 (0.99–1.02) 0.29 0.17
To other hospital 1.00 (1.00–1.01) 0.50 1.00 (0.98–1.02) 0.86 0.88
To rehabilitation or long-term care facility 1.06 (1.05–1.07) <0.001 1.07 (1.05–1.09) <0.001 0.20
To other hospital 1.00 (1.00–1.01) 0.50 1.00 (0.98–1.02) 0.86 0.88
infection,24 as compared with 21% of those with types of the same process. Moreover, on adjust-
SIRS-negative severe sepsis.24 In our patients ed analysis, mortality increased linearly with
with SIRS-positive severe sepsis, the most fre- each additional SIRS criterion from 0 to 4, with
quent signs meeting SIRS criteria were an inno transitional increase in risk at two criteria to
creased heart rate and increased respiratory rate, justify the use of two criteria as definitional
which is in agreement with the results of a cutoff point.
prospective study of SIRS prevalence among pa- Our study has several strengths. It investigates
tients with infection in the ICU.15 These observa- the effect of SIRS criteria within the first 24
tions support the external validity of our study. hours after ICU admission on the diagnosis of
Our study, however, assessed the prevalence of severe sepsis over a period of 14 years. Second,
signs meeting individual SIRS criteria. Among our the study is large. Third, the SIRS data consist
patients who had infection and organ failure, signs of physiological or laboratory measurements
meeting two or more SIRS criteria occurred in that were prospectively collected for routine
87.9% of patients within 24 hours after ICU ad- quality-surveillance purposes and are therefore
mission, supporting their presence in the major- unlikely to be biased. Fourth, our findings are
ity of patients with severe sepsis. Regardless of broadly consistent with the limited existing lit-
how we assess their sensitivity, these criteria fail erature and include data from 172 ICUs, which
to identify one in eight patients with severe sep- increases external validity, and data obtained in
sis. Moreover, we found that the cutoff point of 2013, which increases contemporary relevance.
two SIRS criteria does not define any specific Our study also has some limitations. The
transition point for risk. data were collected primarily for quality-control
The presence of symptoms meeting two or purposes and not for study purposes. We could
more SIRS criteria is believed to have excellent study symptoms meeting SIRS criteria only dur-
sensitivity but low specificity for severe sepsis.15 ing the first 24 hours in the ICU as recorded
Our study showed that it may also have limited either every 30 minutes or every 60 minutes on
sensitivity. The application of the criteria in the the observation charts. Thus, we cannot com-
first 24 hours after ICU admission (the period ment on their absence or presence before or af-
when recruitment into sepsis trials is most com- ter this period or for short intervals between
mon) would exclude approximately one in eight observations. We cannot define a population at
ICU patients with infection and organ failure. risk for sepsis to assess the SIRS criteria for
However, it may also decrease the specificity of their diagnostic value. However, this was not the
these criteria. These patients have substantial aim of our study. The accuracy of specific diag-
mortality and their epidemiologic data are iden- nostic coding of infections was not indepen-
tical to these of patients with classic SIRS-posi- dently monitored. However, on-site audits of
tive severe sepsis,25 which suggests that these data quality, validation rules built into data-col-
two groups represent different clinical pheno- lection software, and regular education sessions
8 n78
engl j med nejm.org
The New
by MIKE MULARCZYK
April 10,Society.
2015. For
All personal use only. No other uses without permission.
rights reserved.
to Table of Contents Society. All rights reserved.
in any clinical context. Our definition of organ

A Unadjusted Mortality
failure related only to the first 24 hours. Thus,
35
patients who might have had organ failure on
Patients Who Died (%) 30 the second day of the ICU stay were missed. Our
25 definition of organ failure was based on the
SOFA score. Other definitions exist,26 which we
20
did not have sufficient data to assess. Our defi-
15 nition of cardiovascular failure relied on the
10 presence of hypotension and a diagnosis of shock
and did not include the use of vasopressor drugs.
5
However, such criteria were applied equally to
0 both groups, which makes bias unlikely.
0 1 2 3 4
In this epidemiologic study, the requirement
No. of SIRS Criteria Met
of two or more SIRS criteria for the diagnosis of
B Adjusted Odds of Death severe sepsis excluded a sizable group of patients
2.0 in the ICU with infection and organ failure.
1.8 These patients with SIRS-negative severe sepsis
1.6 had substantial mortality and, over a period of
1.4 more than a decade, had epidemiologic charac-
Odds of Death
1.2 teristics and changes that were essentially iden-

1.0 tical to those of patients with SIRS-positive se-
0.8 vere sepsis, providing indirect empirical evidence
0.6
that these two groups of patients represent sepa-
0.4
rate phenotypes of the same condition. Moreover,
0.2
the risk of death in the two groups increased
0.0
0 1 2 3 4 linearly with each additional SIRS criterion from
No. of SIRS Criteria Met 0 to 4, without a transitional increase in risk at
two criteria that would justify this consensus
Figure 2. Mortality among Patients with Severe Sepsis,
According to Number of SIRS Criteria Met. cutoff point. Our findings challenge the sensitiv-
The bars represent 95% confidence intervals. ity, face validity, and construct validity of the rule
regarding two or more SIRS criteria in diagnosing
or defining severe sepsis in patients in the ICU.
make systematic bias in diagnostic coding un- Supported by the Australian and New Zealand Intensive Care
likely. Some of the patients who had severe sep- Research Centre, which is supported by an enabling grant
sis after surgery may have had SIRS because of (0606978) from the Australian National Health and Medical
Research Council.
the surgery rather than because of sepsis. How- Disclosure forms provided by the authors are available with
ever, this problem would apply to such patients the full text of this article at NEJM.org.
REFERENCES
1. Gaieski DF, Edwards JM, Kallan MJ, Effect of eritoran, an antagonist of MD2- 303:341-8. [Erratum, JAMA 2010;303:
Carr BG. Benchmarking the incidence TLR4, on mortality in patients with severe 1698.]
and mortality of severe sepsis in the Unit- sepsis: the ACCESS randomized trial. 8. Sprung CL, Annane D, Keh D, et al.
ed States. Crit Care Med 2013;41:1167- JAMA 2013;309:1154-62. Hydrocortisone therapy for patients with
74. 5. Ranieri VM, Thompson BT, Barie PS, septic shock. N Engl J Med 2008;358:111-
2. Brun-Buisson C, Meshaka P, Pinton P, et al. Drotrecogin alfa (activated) in adults 24.
Vallet B. EPISEPSIS: a reappraisal of the with septic shock. N Engl J Med 2012; 9. Mouncey PR, Osborn TM, Power GS,
epidemiology and outcome of severe sep- 366:2055-64. et al. Trial of early, goal-directed resusci-
sis in French intensive care units. Inten- 6. Perner A, Haase N, Guttormsen AB, et tation for septic shock. N Engl J Med.
sive Care Med 2004;30:580-8. al. Hydroxyethyl starch 130/0.42 versus DOI: 10.1056/NEJMoa1500896.
3. American College of Chest Physi- Ringer’s acetate in severe sepsis. N Engl J 10. The ARISE Investigators and the
cians/Society of Critical Care Medicine Med 2012;367:124-34. [Erratum, N Engl J ANZICS Clinical Trials Group. Goal-
Consensus Conference: definitions for Med 2012;367:481.] directed resuscitation for patients with
sepsis and organ failure and guidelines 7. Annane D, Cariou A, Maxime V, et al. early septic shock. N Engl J Med 2014;
for the use of innovative therapies in sep- Corticosteroid treatment and intensive in- 371:1496-506.
sis. Crit Care Med 1992;20:864-74. sulin therapy for septic shock in adults: a 11. Holst LB, Haase N, Wetterslev J, et al.
4. Opal SM, Laterre P-F, Francois B, et al. randomized controlled trial. JAMA 2010; Lower versus higher hemoglobin thresh-
n engl j med
79 nejm.org 9
The New
use only. Medicine
by MIKE ©MULARCZYK
MassachusettsonMedical
April 10, 2015. All
Society. Forrights
reserved.
old for transfusion in septic shock. function/failure: on behalf of the Work- 22. Rangel-Frausto MS, Pittet D, Costigan
N Engl J Med 2014;371:1381-91. ing Group on Sepsis-Related Problems of M, Hwang T, Davis CS, Wenzel RP. The
12. Vincent JL. Dear SIRS, I’m sorry to say the European Society of Intensive Care natural history of the systemic inflamma-
that I don’t like you. . . . Crit Care Med Medicine. Intensive Care Med 1996;22: tory response syndrome (SIRS): a pro-
1997;25:372-4. 707-10. spective study. JAMA 1995;273:117-23.
13. Vincent JL, Opal SM, Marshall JC, 18. Karlsson S, Varpula M, Ruokonen E, 23. Alberti C, Brun-Buisson C, Chevret S,
Tracey KJ. Sepsis definitions: time for et al. Incidence, treatment, and outcome et al. Systemic inflammatory response
change. Lancet 2013;381:774-5. of severe sepsis in ICU-treated adults in and progression to severe sepsis in criti-
14. Levy MM, Fink MP, Marshall JC, et al. Finland: the Finnsepsis study. Intensive cally ill infected patients. Am J Respir Crit
2001 SCCM/ESICM/ACCP/ATS/SIS Inter- Care Med 2007;33:435-43. Care Med 2005;171:461-8.
national Sepsis Definitions Conference. 19. Paul E, Bailey M, Pilcher D. Risk pre- 24. Liao MM, Lezotte D, Lowenstein SR,
Intensive Care Med 2003;29:530-8. diction of hospital mortality for adult pa- et al. Sensitivity of systemic inflammato-
15. Sprung CL, Sakr Y, Vincent JL, et al. tients admitted to Australian and New ry response syndrome for critical illness
An evaluation of systemic inflammatory Zealand intensive care units: development among ED patients. Am J Emerg Med
response syndrome signs in the Sepsis and validation of the Australian and New 2014;32:1319-25.
Occurrence in Acutely Ill Patients (SOAP) Zealand Risk of Death model. J Crit Care 25. Kaukonen KM, Bailey M, Suzuki S,
study. Intensive Care Med 2006;32: 2013;28:935-41. Pilcher D, Bellomo R. Mortality related to
421-7. 20. Pilcher D, Paul E, Bailey M, Huckson severe sepsis and septic shock among
16. Stow PJ, Hart GK, Higlett T, et al. De- S. The Australian and New Zealand Risk critically ill patients in Australia and New
velopment and implementation of a high- of Death (ANZROD) model: getting mor- Zealand, 2000-2012. JAMA 2014;311:
quality clinical database: the Australian tality prediction right for intensive care 1308-16.
and New Zealand Intensive Care Society units. Crit Care Resusc 2014;16:3-4. 26. Improving global outcomes (KDIGO):
Adult Patient Database. J Crit Care 21. Paul E, Bailey M, van Lint A, Pilcher V. clinical practice guideline for acute kid-
2006;21:133-41. Performance of APACHE III over time in ney injury. Kidney Int Suppl 2012;2:1-138.
17. Vincent JL, Moreno R, Takala J, et al. Australia and New Zealand: a retrospec- Copyright © 2015 Massachusetts Medical Society.
The SOFA (Sepsis-related Organ Failure tive cohort study. Anaesth Intensive Care
Assessment) score to describe organ dys- 2012;40:980-94.
10 n engl j80
med nejm.org
The New England

Downloaded from collections.nejm.org. Journal ofuse
For personal Medicine
Copyright on April 10,
© Massachusetts 2015. Society.
reserved.
RESUSCITATION FLUIDS
When a patient is in shock, one of the first tasks of the critical care physician is to replete the circulating
volume. This section covers this important area.
81
Case Challenge
Resuscitation Fluids
A 77-year old man is in the ICU with septic shock and is on mechanical ventilation. Despite administration of
4 liters of crystalloid, he remains hypotensive with signs of inadequate intravascular volume. What should be
administered next?
A 77-year-old man whose medical history includes treated hypertension and hypercholesterolemia, previous
heavy alcohol intake, and mild cognitive impairment is admitted to the intensive care unit (ICU) of a university
hospital from the operating room after a Hartmann’s procedure performed for fecal peritonitis caused by a perfo-
rated sigmoid colon. He has received a total of 4 liters of crystalloid fluids, which were administered in the oper
ating room. (In the first voting installment, there were 7823 votes regarding treatments that might decrease this
patient’s risk of dying from septic shock, with the use of hydrocortisone favored by 78% of respondents, gamma
globulin by 11%, and a statin by 9%. In addition, many of the respondents suggested that the patient should receive
further fluid resuscitation.)
On arrival in the ICU, the arterial blood pressure is 88/52 mm Hg with marked respiratory variation, the heart
rate is 120 beats per minute in sinus rhythm, the central venous pressure is 6 mm Hg, and the temperature
is 35.6°C. He is peripherally cool, with prolonged capillary refill. Arterial blood gas results while the patient is
82
receiving mechanical ventilation with a fraction of inspired oxygen of 0.4 are as follows: pH, 7.32; partial pressure
of carbon dioxide, 28 mm Hg; partial pressure of oxygen, 85 mm Hg; and lactate, 3.0 mmol per liter. Results of
serum biochemical analysis are as follows: sodium, 142 mmol per liter; potassium, 4.4 mmol per liter; chloride,
109 mmol per liter; urea, 22.0 mg per deciliter (7.9 mmol per liter); creatinine, 2.3 mg per deciliter (203 µmol per
liter); and albumin, 23 g per liter. The urine output during the last 2 hours in the operating room was 28 ml.

Assessing the patient to have inadequate intravascular volume, what option for fluid resuscitation would you
choose to be administered over the next 30 minutes to one hour?
A. A total of 1 liter of normal saline (0.9% sodium chloride).

B. A total of 1 liter of Ringer’s lactate (Hartmann’s solution).
C. A total of 500 ml of 6% hydroxyethyl starch (130/0.42).
D. A total of 500 ml of 4% human albumin solution.
83
review article

John A. Myburgh, M.B., B.Ch., Ph.D., and Michael G. Mythen, M.D., M.B., B.S.
F
luid resuscitation with colloid and crystalloid solutions is a From the University of New South Wales,
ubiquitous intervention in acute medicine. The selection and use of resuscita- the Division of Critical Care and Trauma,
George Institute for Global Health, and
tion fluids is based on physiological principles, but clinical practice is deter- the Department of Intensive Care Medi-
mined largely by clinician preference, with marked regional variation. No ideal cine, St. George Hospital — all in Sydney
resuscitation fluid exists. There is emerging evidence that the type and dose of (J.A.M.); and the National Institute for
Health Research, University College Lon-
resuscitation fluid may affect patient-centered outcomes. don Hospitals Biomedical Research Cen-
Despite what may be inferred from physiological principles, colloid solutions do tre, London (M.G.M.). Address reprint
not offer substantive advantages over crystalloid solutions with respect to hemody- requests to Dr. Myburgh at the Depart-
ment of Intensive Care Medicine, St.
namic effects. Albumin is regarded as the reference colloid solution, but its cost George Hospital, Gray St., Kogarah 2217,
is a limitation to its use. Although albumin has been determined to be safe for Sydney, NSW, Australia, or at jmyburgh@
use as a resuscitation fluid in most critically ill patients and may have a role in georgeinstitute.org.au.
early sepsis, its use is associated with increased mortality among patients with N Engl J Med 2013;369:1243-51.
traumatic brain injury. The use of hydroxyethyl starch (HES) solutions is associ- DOI: 10.1056/NEJMra1208627
ated with increased rates of renal-replacement therapy and adverse events among
patients in the intensive care unit (ICU). There is no evidence to recommend the
use of other semisynthetic colloid solutions.
Balanced salt solutions are pragmatic initial resuscitation fluids, although there
is little direct evidence regarding their comparative safety and efficacy. The use of
normal saline has been associated with the development of metabolic acidosis and
acute kidney injury. The safety of hypertonic solutions has not been established.
All resuscitation fluids can contribute to the formation of interstitial edema,
particularly under inflammatory conditions in which resuscitation fluids are used
excessively. Critical care physicians should consider the use of resuscitation fluids
as they would the use of any other intravenous drug. The selection of the specific
fluid should be based on indications, contraindications, and potential toxic effects
in order to maximize efficacy and minimize toxicity.
His t or y of Fluid R e susci tat ion
In 1832, Robert Lewins described the effects of the intravenous administration of

an alkalinized salt solution in treating patients during the cholera pandemic. He
observed that “the quantity necessary to be injected will probably be found to de-
pend upon on the quantity of serum lost; the object being to place the patient in
nearly his ordinary state as to the quantity of blood circulating in the vessels.”1 The
observations of Lewins are as relevant today as they were nearly 200 years ago.
Asanguinous fluid resuscitation in the modern era was advanced by Alexis
Hartmann, who modified a physiologic salt solution developed in 1885 by Sidney
Ringer for rehydration of children with gastroenteritis.2 With the development of
blood fractionation in 1941, human albumin was used for the first time in large
quantities for resuscitation of patients who were burned during the attack on Pearl
Harbor in the same year.
Today, asanguinous fluids are used in almost all patients undergoing general

september 26, 1243
The New England
Medical personalAll
userights
reserved.
volume. Since venous return is in equilibrium

A Capillary with cardiac output, sympathetically mediated
lumen Red responses regulate both efferent capacitance
cells
(venous) and afferent conductance (arterial) cir-
culations in addition to myocardial contractility.3
Adjunctive therapies to fluid resuscitation, such
as the use of catecholamines to augment cardiac
contraction and venous return, need to be con-
Endothelial cell
sidered early to support the failing circulation.4
In addition, changes to the microcirculation in
vital organs vary widely over time and under dif-
Plasma ferent pathologic states, and the effects of fluid
proteins administration on end-organ function should be
considered along with effects on intravascular
Equilibrium
volume.
Healthy The Ph ysiol o gy of Fluid

glycocalyx R e susci tat ion
For decades, clinicians have based their selection

of resuscitation fluids on the classic compart-
Endothelial cell ment model — specifically, the intracellular fluid
compartment and the interstitial and intravascular
B components of the extracellular fluid compart-
ment and the factors that dictate fluid distribu-
tion across these compartments. In 1896, English
Plasma physiologist Ernest Starling found that capillar-
proteins
ies and postcapillary venules acted as a semiper-
meable membrane absorbing fluid from the in-
terstitial space.5 This principle was adapted to
identify the hydrostatic and oncotic pressure gra-
Leaky
Damaged dients across the semipermeable membrane as the

glycocalyx principal determinants of transvascular exchange.6
Recent descriptions have questioned these
Endothelial cell classic models.7 A web of membrane-bound gly-
coproteins and proteoglycans on the luminal
Figure 1. Role of the Endothelial Glycocalyx Layer in the Use of Resuscitation
side of endothelial cells has been identified as
Fluids. COLOR FIGURE
the endothelial glycocalyx layer8 (Fig. 1). The
The structure and function of the endothelial glycocalyx layer, a web of
Rev4
membrane-bound glycoproteins and proteoglycans on endothelial cells, are
08/16/13 subglycocalyx space produces a colloid oncotic
key determinants of membrane permeability Author
in various Dr. Myburgh
vascular organ sys- pressure that is an important determinant of
tems. Panel A shows a healthy endothelialFig #
glycocalyx 1
layer, and Panel B transcapillary flow. Nonfenestrated capillaries
Title and resultant effect on per-
shows a damaged endothelial glycocalyx layer throughout the interstitial space have been iden-
meability, including the development of interstitial
ME edema in some patients,
tified, indicating that absorption of fluid does
particularly those with inflammatory conditions
DE (e.g., sepsis).
not occur through venous capillaries but that
Artist Daniel Muller
AUTHOR PLEASE NOTE:
fluid from the interstitial space, which enters
anesthesia for majorFigure
surgery, in patients
has been redrawn with
and type has been reset through a small number of large pores, is re-
severe trauma and burns, and in patients in the turned to the circulation primarily as lymph that
Issue date
ICU. It is one of the 09/26/2013interven-
most ubiquitous is regulated through sympathetically mediated
tions in acute medicine. responses.9
Fluid therapy is only one component of a The structure and function of the endothelial
complex hemodynamic resuscitation strategy. It glycocalyx layer are key determinants of mem-
is targeted primarily at restoring intravascular brane permeability in various vascular organ sys-
1244 n engl j med 369;13

85 nejm.org september 26, 2013
The New
use only. Medicine
Medical ForAll
rights reserved.
tems. The integrity, or “leakiness,” of this layer, T y pe s of R e susci tat ion Fluid
and thereby the potential for the development of
interstitial edema, varies substantially among Globally, there is wide variation in clinical prac-
organ systems, particularly under inflammatory tice with respect to the selection of resuscitation
conditions, such as sepsis,10 and after surgery or fluid. The choice is determined largely by regional
trauma, when resuscitation fluids are common- and clinician preferences that are based on insti-
ly used. tutional protocols, availability, cost, and com-
mercial marketing.11 Consensus documents about
The Ide a l R e susci tat ion Fluid the use of resuscitation fluids have been devel-
oped and directed primarily at specific patient
The ideal resuscitation fluid should be one that populations,12-14 but such recommendations have
produces a predictable and sustained increase in been based largely on expert opinion or low-
intravascular volume, has a chemical composi- quality clinical evidence. Systematic reviews of
tion as close as possible to that of extracellular randomized, controlled trials have consistently
fluid, is metabolized and completely excreted shown that there is little evidence that resuscita-
without accumulation in tissues, does not pro- tion with one type of fluid as compared with an-
duce adverse metabolic or systemic effects, and other reduces the risk of death15 or that any solu-
is cost-effective in terms of improving patient tion is more effective or safer than any other.16
outcomes. Currently, there is no such fluid avail-
able for clinical use. Albumin
Resuscitation fluids are broadly categorized Human albumin (4 to 5%) in saline is considered
into colloid and crystalloid solutions (Table 1). to be the reference colloidal solution. It is pro-
Colloid solutions are suspensions of molecules duced by the fractionation of blood and is heat-
within a carrier solution that are relatively inca- treated to prevent transmission of pathogenic
pable of crossing the healthy semipermeable viruses. It is an expensive solution to produce and
capillary membrane owing to the molecular weight distribute, and its availability is limited in low-
of the molecules. Crystalloids are solutions of and middle-income countries.
ions that are freely permeable but contain con- In 1998, the Cochrane Injuries Group Albu-
centrations of sodium and chloride that determin Reviewers published a meta-analysis com-
mine the tonicity of the fluid. paring the effects of albumin with those of a
Proponents of colloid solutions have argued range of crystalloid solutions in patients with
that colloids are more effective in expanding hypovolemia, burns, or hypoalbuminemia and
intravascular volume because they are retained concluded that the administration of albumin
within the intravascular space and maintain col- was associated with a significant increase in the
loid oncotic pressure. The volume-sparing effect rate of death (relative risk, 1.68; 95% confidence
of colloids, as compared with crystalloids, is interval [CI], 1.26 to 2.23; P<0.01).17 Despite its
considered to be an advantage, which is con- limitations, including the small size of the in-
ventionally described in a 1:3 ratio of colloids cluded studies, this meta-analysis caused sub-
to crystalloids to maintain intravascular volume. stantial alarm, particularly in countries that
Semisynthetic colloids have a shorter duration used large amounts of albumin for resuscitation.
of effect than human albumin solutions but are As a result, investigators in Australia and New
actively metabolized and excreted. Zealand conducted the Saline versus Albumin
Proponents of crystalloid solutions have ar- Fluid Evaluation (SAFE) study, a blinded, ran-
gued that colloids, in particular human albumin, domized, controlled trial, to examine the safety
are expensive and impractical to use as resusci- of albumin in 6997 adults in the ICU.18 The
tation fluids, particularly under field-type condi- study assessed the effect of resuscitation with
tions. Crystalloids are inexpensive and widely 4% albumin, as compared with saline, on the
available and have an established, although un- rate of death at 28 days. The study showed no
proven, role as first-line resuscitation fluids. significant difference between albumin and sa-
However, the use of crystalloids has classically line with respect to the rate of death (relative
been associated with the development of clini- risk, 0.99; 95% CI, 0.91 to 1.09; P = 0.87) or the
cally significant interstitial edema. development of new organ failure.

nejm.org september 26, 2013 1245
The New
only.ofNo
Medicine
Downloaded from nejm.org by MJ ©
Copyright MEDAS on September
Medical 2014. For
rights reserved.
Copyright © 2013 Massachusetts
Back Medical Society. All rights reserved.
Table 1. Types and Compositions of Resuscitation Fluids.*
1246
Human
Variable Plasma Colloids Crystalloids
4%
Succinylated 3.5% Compounded Balanced
4% Modified Urea-Linked 0.9% Sodium Salt
Albumin Hydroxyethyl Starch Fluid Gelatin Gelatin Saline Lactate Solution
10% 6% 6% 6%
(200/0.5) (450/0.7) (130/0.4) (130/0.42)

Trade name Albumex Hemohes Hextend Voluven Volulyte Venofundin Tetraspan Gelofusine Haemaccel Normal Hartmann’s or PlasmaLyte
saline Ringer’s lactate

Colloid source Human Potato Maize Maize Maize Potato Potato Bovine Bovine
donor starch starch starch starch starch starch gelatin gelatin
Copyright
by MJ ©
Osmolarity 291 250 308 304 308 286 308 296 274 301 308 280.6 294
(mOsm/liter)
The
MEDAS
Sodium 135–145 148 154 143 154 137 154 140 154 145 154 131 140
Copyright © 2013
(mmol/liter)
Back
Potassium 4.5–5.0 3.0 4.0 4.0 5.1 5.4 5.0
Massachusetts
The New
(mmol/liter)
on September
n engl j med87
Calcium 2.2–2.6 5.0 2.5 6.25 2.0
Medical
Massachusetts
(mmol/liter)
For personal
369;13
England use
Magnesium 0.8–1.0 0.9 1.5 1.0 3.0
(mmol/liter)
Journal
29,Society.
2014. For
Chloride 94–111 128 154 124 154 110 154 118 120 145 154 111 98
nejm.org
only.ofNo
(mmol/liter)
Acetate 34 24 27
Allpersonal
Medicine
(mmol/liter)
of
Lactate 1–2 28 29
rights reserved.
(mmol/liter)
Malate 5
september 26, 2013

(mmol/liter)
Medical Society. All rights reserved.

m e dic i n e
Gluconate 23
(mmol/liter)

Bicarbonate 23–27
(mmol/liter)
Octanoate 6.4
(mmol/liter)
* To convert the values for potassium to milligrams per deciliter, divide by 0.2558. To convert the values for calcium to milligrams per deciliter, divide by 0.250. To convert the values for

magnesium to milligrams per deciliter, divide by 0.4114.
Additional analyses from the SAFE study pro- lenge physiologically based concepts about the
vided new insights into fluid resuscitation among efficacy of albumin and its role as a resuscitation
patients in the ICU. Resuscitation with albumin solution. In acute illness, it appears that the hemo-
was associated with a significant increase in the dynamic effects and effects on patient-centered
rate of death at 2 years among patients with outcomes of albumin are largely equivalent to
traumatic brain injury (relative risk, 1.63; 95% those of saline. Whether specific populations of
CI, 1.17 to 2.26; P = 0.003).19 This outcome has patients, particularly those with severe sepsis,
been attributed to increased intracranial pres- may benefit from albumin resuscitation remains
sure, particularly during the first week after in- to be determined.
jury.20 Resuscitation with albumin was associated
with a decrease in the adjusted risk of death at Semisynthetic Colloids
28 days in patients with severe sepsis (odds ra- The limited availability and relative expense of
tio, 0.71; 95% CI, 0.52 to 0.97; P = 0.03), suggest- human albumin have prompted the development
ing a potential, but unsubstantiated, benefit in and increasing use of semisynthetic colloid solu-
patients with severe sepsis.21 No significant be- tions during the past 40 years. Globally, HES solu-
tween-group difference in the rate of death at tions are the most commonly used semisynthetic
28 days was observed among patients with hypo- colloids, particularly in Europe.11 Other semisyn-
albuminemia (albumin level, ≤25 g per liter) thetic colloids include succinylated gelatin, urea-
(odds ratio, 0.87; 95% CI, 0.73 to 1.05).22 linked gelatin–polygeline preparations, and dex-
In the SAFE study, no significant difference tran solutions. The use of dextran solutions has
in hemodynamic resuscitation end points, such as largely been superseded by the use of other semi-
mean arterial pressure or heart rate, was observed synthetic solutions.
between the albumin and saline groups, although HES solutions are produced by hydroxyethyl
the use of albumin was associated with a signifi- substitution of amylopectin obtained from sor-
cant but clinically small increase in central venous ghum, maize, or potatoes. A high degree of sub-
pressure. The ratio of the volumes of albumin to stitution on glucose molecules protects against
the volumes of saline administered to achieve hydrolysis by nonspecific amylases in the blood,
these end points was observed to be 1:1.4. thereby prolonging intravascular expansion, but
In 2011, investigators in sub-Saharan Africa this action increases the potential for HES to
reported the results of a randomized, controlled accumulate in reticuloendothelial tissues, such
trial — the Fluid Expansion as Supportive Ther- as skin (resulting in pruritus), liver, and kidney.
apy (FEAST) study23 — comparing the use of The use of HES, particularly high-molecular-
boluses of albumin or saline with no boluses of weight preparations, is associated with alterations
resuscitation fluid in 3141 febrile children with in coagulation — specifically, changes in visco-
impaired perfusion. In this study, bolus resusci- elastic measurements and fibrinolysis — al-
tation with albumin or saline resulted in similar though the clinical consequences of these effects
rates of death at 48 hours, but there was a sig- in specific patient populations, such as those
nificant increase in the rate of death at 48 hours undergoing surgery or patients with trauma, are
associated with both therapies, as compared undetermined.25 Study reports have questioned
with no bolus therapy (relative risk, 1.45; 95% CI, the safety of concentrated (10%) HES solutions
1.13 to 1.86; P = 0.003). The principal cause of with a molecular weight of more than 200 kD
death in these patients was cardiovascular col- and a molar substitution ratio of more than 0.5 in
lapse rather than fluid overload or neurologic patients with severe sepsis, citing increased rates
causes, suggesting a potentially adverse interac- of death, acute kidney injury, and use of renal-
tion between bolus fluid resuscitation and com- replacement therapy.26,27
pensatory neurohormonal responses.24 Although Currently used HES solutions have reduced
this trial was conducted in a specific pediatric concentrations (6%) with a molecular weight of
population in an environment in which critical 130 kD and molar substitution ratios of 0.38 to
care facilities were limited or absent, the results 0.45. They are available in various types of crys-
call into question the role of bolus fluid resusci- talloid carrier solutions. HES solutions are wide-
tation with either albumin or saline in other ly used in patients undergoing anesthesia for
populations of critically ill patients. major surgery, particularly as a component of
The observations in these key studies chal- goal-directed perioperative fluid strategies,28 as

The New personalJournal
of Medicine
Copyright
Downloaded from nejm.org © Massachusetts
by MJ MEDAS Medical
on September Society.
29, 2014. ForAll rights reserved.
Back to Table of Medical
ContentsSociety. All rights reserved.
a first-line resuscitation fluid in military the- in high-quality randomized, controlled trials to

aters,29 and in patients in the ICU.11 Because of date. In light of current evidence of the lack of
the potential that such solutions may accumu- clinical benefit, potential nephrotoxicity, and
late in tissues, the recommended maximal daily increased cost, the use of semisynthetic colloids
dose of HES is 33 to 50 ml per kilogram of body for fluid resuscitation in critically ill patients is
weight per day. difficult to justify.
In a blinded, randomized, controlled trial
involving 800 patients with severe sepsis in the Crystalloids
ICU,30 Scandinavian investigators reported that Sodium chloride (saline) is the most commonly
the use of 6% HES (130/0.42), as compared with used crystalloid solution on a global basis, par-
Ringer’s acetate, was associated with a signifi- ticularly in the United States. Normal (0.9%) sa-
cant increase in the rate of death at 90 days line contains sodium and chloride in equal con-
(relative risk, 1.17; 95% CI, 1.01 to 1.30; P = 0.03) centrations, which makes it isotonic as compared
and a significant 35% relative increase in the rate with extracellular fluid. The term “normal saline”
of renal-replacement therapy. These results are comes from the studies of red-cell lysis by Dutch
consistent with previous trials of 10% HES physiologist Hartog Hamburger in 1882 and
(200/0.5) in similar patient populations.27 1883, which suggested that 0.9% was the concen-
In a blinded, randomized, controlled study, tration of salt in human blood, rather than the
called the Crystalloid versus Hydroxyethyl Starch actual concentration of 0.6%.35
Trial (CHEST), involving 7000 adults in the ICU, The strong ion difference of 0.9% saline is
the use of 6% HES (130/0.4), as compared with zero, with the result that the administration of
saline, was not associated with a significant dif- large volumes of saline results in a hyperchlore-
ference in the rate of death at 90 days (relative mic metabolic acidosis.36 Adverse effects such as
risk, 1.06; 95% CI, 0.96 to 1.18; P = 0.26). How- immune37 and renal38 dysfunction have been
ever, the use of HES was associated with a sig- attributed to this phenomenon, although the
nificant 21% relative increase in the rate of renal- clinical consequences of these effects is unclear.39
replacement therapy.31 Concern about sodium and water overload
Both the Scandinavian trial and CHEST associated with saline resuscitation has resulted
showed no significant difference in short-term in the concept of “small volume” crystalloid
hemodynamic resuscitation end points, apart resuscitation with the use of hypertonic saline
from transient increases in central venous pres- (3%, 5%, and 7.5%) solutions. However, the
sure and lower vasopressor requirements with early use of hypertonic saline for resuscitation,
HES in CHEST. The observed ratio of HES to particularly in patients with traumatic brain in-
crystalloid in these trials was approximately jury, has not improved either short-term or long-
1:1.3, which is consistent with the ratio of albu- term outcomes.40
min to saline reported in the SAFE study18 and Crystalloids with a chemical composition that
in other recent blinded, randomized, controlled approximates extracellular fluid have been termed
trials of HES.32,33 “balanced” or “physiologic” solutions and are de-
In CHEST, HES was associated with increases rivatives of the original Hartmann’s and Ringer’s
in urine output in patients at low risk for acute solutions. However, none of the proprietary so-
kidney injury but with parallel increases in se- lutions are either truly balanced or physiologic41
rum creatinine levels in patients at increased (Table 1).
risk for acute kidney injury. In addition, the use Balanced salt solutions are relatively hypo-
of HES was associated with an increased use of tonic because they have a lower sodium concen-
blood products and an increased rate of adverse tration than extracellular fluid. Because of the
events, particularly pruritus.31 instability of bicarbonate-containing solutions
Whether these results are generalizable to the in plastic containers, alternative anions, such as
use of other semisynthetic colloid solutions, such lactate, acetate, gluconate, and malate, have been
as gelatin or polygeline preparations, is unused. Excessive administration of balanced salt
known. A recent observational study has raised solutions may result in hyperlactatemia, meta-
concern about the risk of acute kidney injury bolic alkalosis, and hypotonicity (with com-
associated with the use of gelatin solutions.34 pounded sodium lactate) and cardiotoxicity (with
However, these solutions have not been studied acetate). The addition of calcium in some solu-
1248 n engl j med 369;13 89

nejm.org september 26, 2013
The New EnglandFor

Medical Society.use
All only.
reserved.
tions may generate microthrombi with citrate- fection, renal-replacement therapy, blood trans-
containing red-cell transfusions. fusion, and acidosis-associated investigations.
Given the concern regarding an excess of In a single-center, sequential, observational
sodium and chloride associated with normal ICU study,45 the use of a chloride-restrictive fluid
saline, balanced salt solutions are increasingly strategy (using lactated and calcium-free balanced
recommended as first-line resuscitation fluids solutions) to replace chloride-rich intravenous
in patients undergoing surgery,13 patients with fluids (0.9% saline, succinylated gelatin, or 4%
trauma,14 and patients with diabetic ketoacido- albumin) was associated with a significant de-
sis.42 Resuscitation with balanced salt solutions crease in the incidence of acute kidney injury
is a key element in the initial treatment of pa- and the rate of renal-replacement therapy. Given
tients with burns, although there is increasing the widespread use of saline (>200 million liters
concern about the adverse effects of fluid over- per year in the United States alone), these data
load, and a strategy of “permissive hypovolemia” suggest that a randomized, controlled trial ex-
in such patients has been advocated.43 amining the safety and efficacy of saline, as
A matched-cohort observational study com- compared with a balanced salt solution, is war-
pared the rate of major complications in 213 pa- ranted.
tients who received only 0.9% saline and 714 pa-
tients who received only a calcium-free balanced D ose a nd Volume s
salt solution (PlasmaLyte) for replacement of fluid
losses on the day of surgery.44 The use of bal- The requirements for and response to fluid re-
anced salt solution was associated with a signifi- suscitation vary greatly during the course of any
cant decrease in the rate of major complications critical illness. No single physiological or bio-
(odds ratio, 0.79; 95% CI, 0.66 to 0.97; P<0.05), chemical measurement adequately reflects the
including a lower incidence of postoperative in- complexity of fluid depletion or the response to
Table 2. Recommendations for Fluid Resuscitation in Acutely Ill Patients.
Fluids should be administered with the same caution that is used with any intravenous drug.
Consider the type, dose, indications, contraindications, potential for toxicity, and cost.
Fluid resuscitation is a component of a complex physiological process.
Identify the fluid that is most likely to be lost and replace the fluid lost in equivalent volumes.
Consider serum sodium, osmolarity, and acid–base status when selecting a resuscitation fluid.
Consider cumulative fluid balance and actual body weight when selecting the dose of resuscitation fluid.
Consider the early use of catecholamines as concomitant treatment of shock.
Fluid requirements change over time in critically ill patients.
The cumulative dose of resuscitation and maintenance fluids is associated with interstitial edema.
Pathological edema is associated with an adverse outcome.
Oliguria is a normal response to hypovolemia and should not be used solely as a trigger or end point for fluid resuscita-
tion, particularly in the post-resuscitation period.
The use of a fluid challenge in the post-resuscitation period (≥24 hours) is questionable.
The use of hypotonic maintenance fluids is questionable once dehydration has been corrected.
Specific considerations apply to different categories of patients.
Bleeding patients require control of hemorrhage and transfusion with red cells and blood components as indicated.
Isotonic, balanced salt solutions are a pragmatic initial resuscitation fluid for the majority of acutely ill patients.
Consider saline in patients with hypovolemia and alkalosis.
Consider albumin during the early resuscitation of patients with severe sepsis.
Saline or isotonic crystalloids are indicated in patients with traumatic brain injury.
Albumin is not indicated in patients with traumatic brain injury.
Hydroxyethyl starch is not indicated in patients with sepsis or those at risk for acute kidney injury.
The safety of other semisynthetic colloids has not been established, so the use of these solutions is not recommended.
The safety of hypertonic saline has not been established.
The appropriate type and dose of resuscitation fluid in patients with burns has not been determined.

The New
For personal ofNo
use only. Medicine
Medical 2014. For
rights reserved.
fluid resuscitation in acute illness. However, sys- Although the use of resuscitation fluids is one
tolic hypotension and particularly oliguria are of the most common interventions in medicine,
widely used as triggers to administer a “fluid no currently available resuscitation fluid can be
challenge,” ranging from 200 to 1000 ml of crys- considered to be ideal. In light of recent high-
talloid or colloid for an adult patient. quality evidence, a reappraisal of how resuscita-
The use of crystalloid and colloid resuscitation fluids are used in acutely ill patients is now
tion fluids, often prescribed by the most junior required (Table 2). The selection, timing, and
members of the clinical team, in addition to doses of intravenous fluids should be evaluated
hypotonic “maintenance” fluids, results in in- as carefully as they are in the case of any other
creased cumulative doses of sodium and water intravenous drug, with the aim of maximizing
over time.46 These increases are associated with efficacy and minimizing iatrogenic toxicity.
the development of interstitial edema with resul-
tant organ dysfunction.47 Dr. Myburgh reports receiving travel support and lecture fees,
as well as grant support to his institution, from Fresenius Kabi;
Associations between increased cumulative and Dr. Mythen, receiving consulting fees paid to himself and
positive fluid balance and long-term adverse his institution by Aqix, grant support to his institution from
outcomes have been reported in patients with Fresenius Kabi, lecture fees and travel support from Baxter,
B. Braun, and Fresenius Kabi, and fees for the development and
sepsis.48 In trials of liberal versus goal-directed presentation of educational materials from B. Braun. Dr. Mythen
or restrictive fluid strategies in patients with the also reports organizing nonprofit educational conferences for
acute respiratory distress syndrome (particularly the Evidence-Based Perioperative Medicine (EBPOM) project,
which has received financial support from Baxter, B. Braun,
in perioperative patients),49,50 restrictive fluid Beacon Pharmaceuticals, Cosmed, Cranlea, Deltex, Edwards,
strategies were associated with reduced morbid- Fresenius Kabi, Imacor, LiDCO, Masimo, Medgraphics, Orion
ity. However, since there is no consensus on the Pharma, Pulsion, and Vitalograph. No other potential conflict of
interest relevant to this article was reported.
definition of these strategies, high-quality trials Disclosure forms provided by the authors are available with
in specific patient populations are required.46 the full text of this article at NEJM.org.
References
1. Lewins R. Saline injections into the improved paradigm for prescribing intra- for fluid resuscitation. Cochrane Database
veins. London Medical Gazette. April 7, venous fluid therapy. Br J Anaesth 2012; Syst Rev 2012;7:CD001319.
1832:257-68. 108:384-94. 17. Cochrane Injuries Group Albumin
2. Lee JA. Sydney Ringer (1834-1910) and 10. Lee WL, Slutsky AS. Sepsis and endo- Reviewers. Human albumin administra-
Alexis Hartmann (1898-1964). Anaesthe- thelial permeability. N Engl J Med 2010; tion in critically ill patients: systematic
sia 1981;36:1115-21. 363:689-91. review of randomised controlled trials.
3. Funk DJ, Jacobsohn E, Kumar A. The 11. Finfer S, Liu B, Taylor C, et al. Resus- BMJ 1998;317:235-40.
role of venous return in critical illness citation fluid use in critically ill adults: 18. The SAFE Study Investigators. A com-
and shock. I. Physiology. Crit Care Med an international cross-sectional study in parison of albumin and saline for fluid
2013;41:255-62. 391 intensive care units. Crit Care 2010;14: resuscitation in the intensive care unit.
4. Persichini R, Silva S, Teboul JL, et al. R185. N Engl J Med 2004;350:2247-56.
Effects of norepinephrine on mean sys- 12. Dellinger RP, Levy MM, Carlet JM, 19. Idem. Saline or albumin for fluid resus-
temic pressure and venous return in hu- et al. Surviving Sepsis Campaign: interna- citation in patients with traumatic brain
man septic shock. Crit Care Med 2012;40: tional guidelines for management of se- injury. N Engl J Med 2007;357:874-84.
3146-53. vere sepsis and septic shock: 2008. Crit 20. Cooper DJ, Myburgh J, Finfer S, et al.
5. Starling EH. On the absorption of flu- Care Med 2008;36:296-327. [Erratum, Crit Albumin resuscitation for traumatic brain
ids from connective tissue spaces. J Physi- Care Med 2008;36:1394-6.] injury: is intracranial hypertension the
ol 1896;19:312-26. 13. Powell-Tuck J, Gosling P, Lobo DN, cause of increased mortality? J Neurotrau-
6. Krogh A, Landis EM, Turner AH. The et al. British Consensus Guidelines on In- ma 2013 March 21 (Epub ahead of print).
movement of fluid through the human travenous Fluid Therapy for Adult Surgi- 21. Finfer S, McEvoy S, Bellomo R, McAr-
capillary wall in relation to venous pres- cal Patients (GIFTASUP). March 2011 thur C, Myburgh J, Norton R. Impact of
sure and to the colloid osmotic pressure (http://www.bapen.org.uk/pdfs/bapen_ albumin compared to saline on organ
of the blood. J Clin Invest 1932;11:63-95. pubs/giftasup.pdf). function and mortality of patients with
7. Levick JR, Michel CC. Microvascular 14. Advanced Trauma Life Support (ATLS) severe sepsis. Intensive Care Med 2011;
fluid exchange and the revised Starling for doctors. Chicago: American College 37:86-96.
principle. Cardiovasc Res 2010;87:198-210. of Surgeons Committee on Trauma, 2012 22. Finfer S, Bellomo R, McEvoy S, et al.
8. Weinbaum S, Tarbell JM, Damiano (http://www.facs.org/trauma/atls/index Effect of baseline serum albumin concen-
ER. The structure and function of the .html). tration on outcome of resuscitation with
endothelial glycocalyx layer. Annu Rev 15. Perel P, Roberts I. Colloids versus albumin or saline in patients in intensive
Biomed Eng 2007;9:121-67. crystalloids for fluid resuscitation in criti- care units: analysis of data from the Sa-
9. Woodcock TE, Woodcock TM. Revised cally ill patients. Cochrane Database Syst line versus Albumin Fluid Evaluation (SAFE)
Starling equation and the glycocalyx Rev 2012;6:CD000567. study. BMJ 2006;333:1044.
model of transvascular fluid exchange: an 16. Bunn F, Trivedi D. Colloid solutions 23. Maitland K, Kiguli S, Opoka R, et al.
1250 n engl j med 369;13 91

The New EnglandFor

Copyright 29, 2014.
reserved.
Mortality after fluid bolus in African chil- penetrating trauma in a randomized con- fluid resuscitation in diabetic ketoacido-
dren with shock. N Engl J Med 2011;364: trolled study: the FIRST trial (Fluids in sis. J Crit Care 2012;27:138-45.
2483-95. Resuscitation of Severe Trauma). Br J An- 43. Arlati S, Storti E, Pradella V, Bucci L,
24. Maitland K, George E, Evans J, et al. aesth 2011;107:693-702. Vitolo A, Pulici M. Decreased fluid vol-
Exploring mechanisms of excess mortal- 33. Guidet B, Martinet O, Boulain T, et al. ume to reduce organ damage: a new
ity with early fluid resuscitation: insights Assessment of hemodynamic efficacy and approach to burn shock resuscitation?
from the FEAST trial. BMC Med 2013;11: safety of 6% hydroxyethylstarch 130/0.4 A preliminary study. Resuscitation 2007;72:
68. vs. 0.9% NaCl fluid replacement in patients 371-8.
25. Hartog CS, Reuter D, Loesche W, Hof- with severe sepsis: the CRYSTMAS study. 44. Shaw AD, Bagshaw SM, Goldstein SL,
mann M, Reinhart K. Influence of hy- Crit Care 2012;16(3):R94. et al. Major complications, mortality, and
droxyethyl starch (HES) 130/0.4 on hemo- 34. Bayer O, Reinhart K, Sakr Y, et al. Re- resource utilization after open abdominal
stasis as measured by viscoelastic device nal effects of synthetic colloids and crys- surgery: 0.9% saline compared to Plasma-
analysis: a systematic review. Intensive talloids in patients with severe sepsis: Lyte. Ann Surg 2012;255:821-9.
Care Med 2011;37:1725-37. a prospective sequential comparison. Crit 45. Yunos NM, Bellomo R, Hegarty C,
26. Schortgen F, Lacherade JC, Bruneel F, Care Med 2011;39:1335-42. Story D, Ho L, Bailey M. Association
et al. Effects of hydroxyethylstarch and 35. Awad S, Allison SP, Lobo DN. The between a chloride-liberal vs chloride-
gelatin on renal function in severe sepsis: history of 0.9% saline. Clin Nutr 2008;27: restrictive intravenous fluid administration
a multicentre randomised study. Lancet 179-88. strategy and kidney injury in critically ill
2001;357:911-6. 36. Morgan TJ, Venkatesh B, Hall J. Crys- adults. JAMA 2012;308:1566-72.
27. Brunkhorst FM, Engel C, Bloos F, et al. talloid strong ion difference determines 46. Corcoran T, Rhodes JE, Clarke S, Myles
Intensive insulin therapy and pentastarch metabolic acid-base change during acute PS, Ho KM. Perioperative fluid manage-
resuscitation in severe sepsis. N Engl J normovolaemic haemodilution. Intensive ment strategies in major surgery: a strati-
Med 2008;358:125-39. Care Med 2004;30:1432-7. fied meta-analysis. Anesth Analg 2012;
28. Brandstrup B, Svendsen PE, Rasmus- 37. Kellum JA, Song M, Li J. Science re- 114:640-51.
sen M, et al. Which goal for fluid therapy view: extracellular acidosis and the im- 47. Cordemans C, De Laet I, Van Regen-
during colorectal surgery is followed by mune response: clinical and physiologic mortel N, et al. Fluid management in
the best outcome: near-maximal stroke implications. Crit Care 2004;8:331-6. critically ill patients: the role of extravas-
volume or zero fluid balance? Br J Anaesth 38. Hadimioglu N, Saadawy I, Saglam T, cular lung water, abdominal hyperten-
2012;109:191-9. Ertug Z, Dinckan A. The effect of differ- sion, capillary leak, and fluid balance.
29. McSwain NE, Champion HR, Fabian ent crystalloid solutions on acid-base bal- Ann Intensive Care 2012;2:Suppl 1:S1.
TC, et al. State of the art of fluid resusci- ance and early kidney function after kid- 48. Boyd JH, Forbes J, Nakada TA, Walley
tation 2010: prehospital and immediate ney transplantation. Anesth Analg 2008; KR, Russell JA. Fluid resuscitation in sep-
transition to the hospital. J Trauma 2011; 107:264-9. tic shock: a positive fluid balance and ele-
70:Suppl:S2-S10. 39. Handy JM, Soni N. Physiological ef- vated central venous pressure are associ-
30. Perner A, Haase N, Guttormsen AB, fects of hyperchloraemia and acidosis. Br ated with increased mortality. Crit Care
et al. Hydroxyethyl starch 130/0.42 versus J Anaesth 2008;101:141-50. Med 2011;39:259-65.
Ringer’s acetate in severe sepsis. N Engl J 40. Cooper DJ, Myles PS, McDermott FT, 49. The National Heart, Lung, and Blood
Med 2012;367:124-34. [Erratum, N Engl J et al. Prehospital hypertonic saline re- Institute Acute Respiratory Distress Syn-
Med 2012;367:481.] suscitation of patients with hypotension drome (ARDS) Clinical Trials Network.
31. Myburgh JA, Finfer S, Bellomo R, et and severe traumatic brain injury: a ran- Comparison of two fluid-management
al. Hydroxyethyl starch or saline for fluid domized controlled trial. JAMA 2004;291: strategies in acute lung injury. N Engl J
resuscitation in intensive care. N Engl J 1350-7. Med 2006;354:2564-75.
Med 2012;367:1901-11. 41. Guidet B, Soni N, Della Rocca G, et al. 50. Murphy CV, Schramm GE, Doherty
32. James MF, Michell WL, Joubert IA, Ni- A balanced view of balanced solutions. JA, et al. The importance of fluid manage-
col AJ, Navsaria PH, Gillespie RS. Resusci- Crit Care 2010;14:325. ment in acute lung injury secondary to
tation with hydroxyethyl starch improves 42. Chua HR, Venkatesh B, Stachowski E, septic shock. Chest 2009;136:102-9.
renal function and lactate clearance in et al. Plasma-Lyte 148 vs 0.9% saline for Copyright © 2013 Massachusetts Medical Society.
2013 icmje recommendations

The International Committee of Medical Journal Editors (ICMJE) has revised
and renamed its Uniform Requirements. The new ICMJE Recommendations
for the Conduct, Reporting, Editing and Publication of Scholarly Work
in Medical Journals are available at www.icmje.org.

The New
For personal use only.ofNo
Medicine
© Massachusetts 29, Society.
Medical 2014. ForAllpersonal use only. No other uses without permission.
rights reserved.
Back Medical Society. All rights reserved.
c or r e sp ondence
N Engl J Med 2012;366:1423-32.
To the Editor: Myburgh and Mythen (Sept. 26 3. Dengue: guidelines for diagnosis, treatment, prevention and
issue)1 suggest that it is difficult to justify using control. Geneva: World Health Organization, 2009.
4. Wills BA, Dung NM, Loan HT, et al. Comparison of three
semisynthetic colloid solutions for fluid resusci- f luid solutions for resuscitation in dengue shock syndrome.
tation in critically ill patients. However, although N Engl J Med 2005;353:877-89.
the adverse effects of colloids have been clearly 5. Lam PK, Tam DT, Diet TV, et al. Clinical characteristics of
dengue shock syndrome in Vietnamese children: a 10-year pro-
shown, information on potential benefits remains spective study in a single hospital. Clin Infect Dis 2013;57:1577-
sparse. Colloids are frequently used for resuscita- 86.
tion of patients with dengue shock syndrome, a 6. Bhatt S, Gething PW, Brady OJ, et al. The global distribution
and burden of dengue. Nature 2013;496:504-7.
potentially fatal complication of dengue charac-
DOI: 10.1056/NEJMc1313345
terized by substantial plasma leakage and hypo-
volemic shock.2,3 At our hospital in Ho Chi Minh
City, more than 200 children or young adults are To the Editor: The article on resuscitation fluids
admitted annually with dengue shock syndrome, omitted mention of critically ill obstetrical pa-
of whom approximately 30 to 40% require resus- tients, a special population at risk because of the
citation with colloid solutions to achieve cardio- increased cardiac output and reduced total pe-
vascular stability but less than 1% die.4,5 Given ripheral resistance that restrict application of clas-
the huge global burden of disease6 and the limited sical principles of fluid resuscitation.1 Special
access to ventilatory support in many dengue- conditions include postpartum hemorrhage and
endemic areas, resuscitation with crystalloids the aortocaval compression syndrome, preeclamp-
alone might well result in increased global mor- sia, and the HELLP syndrome (hemolysis, elevated
bidity and mortality secondary to fluid overload liver-enzyme levels, and a low platelet count).2
and respiratory failure. This population of rela- Although no randomized trials have been
tively young, previously healthy patients present- conducted, a recent review3 proposes that large-
ing with critical dysfunction of a single organ is volume crystalloid resuscitation may worsen ob-
markedly different from that seen in most criti- stetrical bleeding by increasing intravascular hy-
cal care units; could you comment on what type drostatic pressure and dislodging the fresh clots
of fluid to use for cases of dengue shock syn- at sites of endothelial injury. Other authors sug-
drome that have failed to respond to resuscita- gest early use of blood products and the limiting
tion with crystalloid solutions? of traditional fluid resuscitation with large vol-
Trieu T. Huynh, M.D. umes of crystalloid products to restrict further
V. Chau Nguyen, M.D., Ph.D. coagulopathy and bleeding.4 Do the authors have
Hospital for Tropical Diseases any advice on special approaches to resuscitation
Ho Chi Minh City, Vietnam in this population?
Bridget A. Wills, M.D., D.M. Nikolaos Vrachnis, M.D.
Oxford University Clinical Research Unit
Ho Chi Minh City, Vietnam Dimitrios Zygouris, M.D.
bwills@oucru.org Zoe Iliodromiti, M.D.
No potential conflict of interest relevant to this letter was re- University of Athens Medical School
ported. Athens, Greece
1. Myburgh JA, Mythen MG. Resuscitation fluids. N Engl J Med nvrachnis@hotmail.com
2013;369:1243-51. No potential conflict of interest relevant to this letter was re-
2. Simmons CP, Farrar JJ, van Vinh Chau N, Wills B. Dengue. ported.

nejm.org december 19, 2013 2455
1. Chesnutt AN. Physiology of normal pregnancy. Crit Care the circulation and enter the interstitium. Col-
Clin 2004;20:609-15.
2. Nelson-Piercy C, Mackillop L, Williams DJ, Williamson C, de loids and hyperosmolar solutions are known to
Swiet M, Redman C. Maternal mortality in the UK and the need trigger both anaphylactic and anaphylactoid re-
for obstetric physicians. BMJ 2011;343:d4993. actions through different mechanisms.1 Comple-
3. Ickx BE. Fluid and blood transfusion management in obstet-
rics. Eur J Anaesthesiol 2010;27:1031-5. ment activation and activation of the bradykinin
4. Pacheco LD, Saade GR, Costantine MM, Clark SL, Hankins cascade as well as IgE-dependent mechanisms
GD. The role of massive transfusion protocols in obstetrics. Am have been implicated in reactions induced by
J Perinatol 2013;30:1-4.
gelatin-based or starch-based colloids, and IgE-
DOI: 10.1056/NEJMc1313345 independent degranulation of mast cells and
basophils has been shown on exposure to hyper-
To the Editor: In the article by Myburgh and osmolar solutions.2,3 Moreover, it was recently
Mythen on resuscitation fluids, trauma, which is reported that sensitization to gelatin contained
the leading cause of death among young, produc- in red meat represents a risk factor for gelatin
tive persons, was not addressed.1 The authors do colloid allergy, because most patients who are
not mention prehospital settings, where fluid re- allergic to red meat are sensitized to gelatin,
suscitation is routine. and a subgroup will be clinically allergic to
We have recently introduced plasma as the both.2
resuscitation fluid of choice for hemorrhaging Therefore, World Allergy Organization guide-
trauma patients in the prehospital setting, a use lines recommend infusion of isotonic or normal
not discussed in the article. This is unlike its use saline during anaphylaxis, because colloids (main-
as a supplement to blood-product transfusion in ly gelatins) and hyperosmolar solutions may
hospital settings.2 In fact, plasma (i.e., in lyophi- themselves cause histamine release and worsen
lized form) as a volume expander meets the re- the ongoing reaction.4
quirements suggested by the authors for the Emanuel Della-Torre, M.D.
“ideal” resuscitation fluid, although of course Mona-Rita Yacoub, M.D., Ph.D.
not without disadvantages. Our initial experience Giselda Colombo, M.D.
indeed supports it as an improved resuscitation
San Raffaele Hospital
fluid.3 Could the authors comment on this spe- Milan, Italy
cial case? dellatorre.emanuel@hsr.it
Elon Glassberg, M.D., M.H.A. No potential conflict of interest relevant to this letter was re-
ported.
Roy Nadler, M.D.
Yitshak Kreiss, M.D., M.H.A.. M.P.A. 1. Laxenaire MC, Charpentier C, Feldman L. Anaphylactoid re-
actions to colloid plasma substitutes: incidence, risk factors,
Israel Defense Forces Medical Corps mechanisms: a French multicenter prospective study. Ann Fr
Ramat Gan, Israel Anesth Reanim 1994;13:301-10. (In French.)
No potential conflict of interest relevant to this letter was re- 2. Mullins RJ, James H, Platts-Mills TA, Commins S. Relation-
ported. ship between red meat allergy and sensitization to gelatin and
galactose-α-1,3-galactose. J Allergy Clin Immunol 2012;129(5):
1. Injuries and Violence Prevention Department. The injury chart 1334.e1-1342.e1.
book: a graphical overview of the global burden of injuries. 3. Gulliksson M, Palmberg L, Nilsson G, Ahlstedt S, Kumlin M.
Geneva: World Health Organization, 2002. Release of prostaglandin D2 and leukotriene C4 in response to
2. Glassberg E, Nadler R, Rasmussen TE, et al. Point-of-injury hyperosmolar stimulation of mast cells. Allergy 2006;61:1473-9.
use of reconstituted freeze dried plasma as a resuscitative fluid: 4. Simons FE, Ardusso LR, Bilò MB, et al. World Allergy Orga-
a special report for prehospital trauma care. J Trauma Acute nization anaphylaxis guidelines: summary. J Allergy Clin Im-
Care Surg 2013;75:Suppl 2:S111-S114. munol 2011;127(3):587.e1-22–593.e1-e22.
3. Glassberg E, Nadler R, Gendler S, et al. Freeze-dried plasma
at the point of injury: from concept to doctrine. Shock 2013; DOI: 10.1056/NEJMc1313345
40:444-50.
DOI: 10.1056/NEJMc1313345
The authors reply: These letters highlight as-
pects of resuscitation fluids in specific patient
To the Editor: In addition to the issues reviewed populations for which evidence is limited and
by Myburgh and Mythen, clinicians should con- that are unlikely to be studied in randomized,
sider the allergenic properties of resuscitation controlled trials.
fluids, particularly in the management of ana- Huynh et al. raise the role of colloids in chil-
phylaxis, a common clinical setting in acute dren with severe dengue shock syndrome, which
medicine in which large volumes of fluids leave is characterized by a transient capillary perme-
2456 n engl j med 369;25 94

nejm.org december 19, 2013
correspondence
ability leak syndrome. There is limited evidence ated with the development of clinically signifi-
that semisynthetic colloids are more effective cant interstitial edema in patients with capillary
than crystalloids in restoring hemodynamic sta- permeability leak syndromes such as sepsis,
bility.1 Given the evidence of harm associated dengue, pregnancy, and anaphylaxis. The effi-
with hydroxyethyl starch (HES) solutions in pa- cacy and safety of restrictive fluid-resuscitation
tients with severe sepsis, their role in dengue strategies in these patients as well as in those
must be questioned. Although albumin has an with trauma or burns require evaluation in ran-
unsubstantiated beneficial effect in sepsis,2 there domized, controlled trials to determine both the
is no evidence to recommend its use in the den- appropriate type and dose of resuscitation fluids
gue shock syndrome. Furthermore, the availabil- in critically ill patients.
ity of albumin is usually restricted in regions
John A. Myburgh, M.B., B.Ch., Ph.D.
where dengue is prevalent.
Vrachnis et al. highlight the selection of re- University of New South Wales
Sydney, NSW, Australia
suscitation fluids in critically ill obstetrical pa- jmyburgh@georgeinstitute.org.au
tients. When hypovolemia is primarily due to
severe hemorrhage, resuscitation with fresh blood Michael G. Mythen, M.D., M.B., B.S.
and blood-component therapy is the mainstay of University College London
fluid resuscitation. HES solutions are not recom- London, United Kingdom
mended for use in pregnancy; for this reason, Since publication of their article, Dr. Myburgh reports that
these patients were excluded from the Crystal- his institution (the George Institute for Global Health) has re-
ceived travel expenses from Baxter Healthcare, and Dr. Mythen
loid versus Hydroxyethyl Starch Trial.3 reports having received an honorarium from Edwards Life-
Glassberg et al. discuss the role of lyophilized sciences and an honorarium and accommodation from Deltex
plasma in the early resuscitation of persons with Medical and sponsorship for the Acute Dialysis Quality Initia-
tive. No further potential conflict of interest relevant to this
traumatic hemorrhage. The selection of resusci- letter was reported.
tation fluids under field situations requires prag-
matic considerations that primarily relate to the 1. Wills BA, Dung NM, Loan HT, et al. Comparison of three
f luid solutions for resuscitation in dengue shock syndrome.
availability and stability of fluids in addition to N Engl J Med 2005;353:877-89.
the efficacy and safety of hemodynamic resusci- 2. Finfer S, McEvoy S, Bellomo R, McArthur C, Myburgh J, Nor-
tation. On the basis of current evidence, crystal- ton R. Impact of albumin compared to saline on organ function
and mortality of patients with severe sepsis. Intensive Care Med
loids are the fluids of choice in these patients,4 2011;37:86-96.
and the use of new and emerging fluids such as 3. Myburgh JA, Finfer S, Bellomo R, et al. Hydroxyethyl starch
lyophilized plasma in these patients requires or saline for fluid resuscitation in intensive care. N Engl J Med
2012;367:1901-11.
validation in a randomized, controlled trial. 4. American College of Surgeons Committee on Trauma. Ad-
Della-Torre et al. highlight the association be- vanced trauma life support for doctors. 2012 (http://www.facs
tween semisynthetic and hyperosmolar solutions .org/trauma/atls/index.html).
5. Ford SA, Kam PC, Baldo BA, Fisher MM. Anaphylactic or
and the development of hypersensitivity reactions. anaphylactoid reactions in patients undergoing cardiac surgery.
These fluids, particularly gelatin solutions,5 J Cardiothorac Vasc Anesth 2001;15:684-8.
should not be used for the treatment of anaphy-
laxis, which often involves administration of DOI: 10.1056/NEJMc1313345
large volumes of fluid.
The administration of these fluids with exces-
sive volumes of asanguinous fluids may be associ-


Assessing the patient to have inadequate intravascular volume, what option for fluid resuscitation would you
choose to be administered over the next 30 minutes to one hour?
A. A total of 1 liter of normal saline (0.9% sodium chloride).

B. A total of 1 liter of Ringer’s lactate (Hartmann’s solution).
C. A total of 500 ml of 6% hydroxyethyl starch (130/0.42).
D. A total of 500 ml of 4% human albumin solution.

The patient in the vignette has septic shock that has not been corrected by fluid loading with 4 liters of crystalloid.
His oliguria, tachycardia, low central venous pressure, and respiratory variation in arterial blood pressure are
signs of intravascular volume depletion, and further fluid resuscitation is indicated. Current guidelines recom-
mend the use of crystalloid fluids for the initial resuscitation of patients with severe sepsis or septic shock.1 Suit-
able fluids include normal saline or a balanced salt solution such as Ringer’s lactate (Hartmann’s solution), Ring-
er’s acetate, and PlasmaLyte. The relative risks and benefits of different crystalloid solutions have not been
assessed in large clinical trials; administration of normal saline may cause or worsen hyperchloremic acidosis, and
its use has been associated with an increased risk of renal injury.2, 3 The patient has a metabolic acidosis with a se-
rum chloride level at the upper limit of the normal range, and for that reason, a balanced salt solution may be pre-
ferred. Resuscitation with hydroxyethyl starch (HES) is associated with an increased risk of acute kidney injury
and death in patients with severe sepsis4 and an increased risk of the need for renal-replacement therapy for criti-
cally ill patients in general.5 For those reasons, HES solutions should not be used. Administration of albumin is
recommended for patients with severe sepsis who do not have a response to resuscitation with crystalloid fluids.1
Although a definitive beneficial effect of albumin has not been confirmed, its use is supported by data from the
Saline versus Albumin Fluid Evaluation (SAFE) trial and a subsequent meta-analysis.6, 7
References
1. Dellinger RP, Levy MM, Rhodes A, et al. Surviving Sepsis Campaign: international guidelines for manage-
ment of severe sepsis and septic shock, 2012. Intensive Care Med 2013;39:165–228.
2. Shaw AD, Bagshaw SM, Goldstein SL, et al. Major complications, mortality, and resource utilization after
open abdominal surgery: 0.9% saline compared to Plasma-Lyte. Ann Surg 2012;255:821–9.
3. Yunos NM, Bellomo R, Hegarty C, Story D, Ho L, Bailey M. Association between a chloride-liberal vs

chloride-restrictive intravenous fluid administration strategy and kidney injury in critically ill adults.
JAMA 2012;308:1566–72.
4. Perner A, Haase N, Guttormsen AB, et al. Hydroxyethyl starch 130/0.42 versus Ringer’s acetate in severe
sepsis. N Engl J Med 2012;367:124–34. [Erratum, N Engl J Med 2012;367:481.]
5 Myburgh JA, Finfer S, Bellomo R, et al. Hydroxyethyl starch or saline for fluid resuscitation in intensive care.
N Engl J Med 2012;367:1901–11.
96
6. The SAFE Study Investigators. A comparison of albumin and saline for fluid resuscitation in the intensive care
unit. N Engl J Med 2004;350:2247–56.
7. Delaney AP, Dan A, McCaffrey J, Finfer S. The role of albumin as a resuscitation fluid for patients with sepsis:
A systematic review and meta-analysis. Crit Care Med 2011;39:386–91.
97
original article
Albumin Replacement in Patients

with Severe Sepsis or Septic Shock
Pietro Caironi, M.D., Gianni Tognoni, M.D., Serge Masson, Ph.D.,
Roberto Fumagalli, M.D., Antonio Pesenti, M.D., Marilena Romero, Ph.D.,
Caterina Fanizza, M.Stat., Luisa Caspani, M.D., Stefano Faenza, M.D.,
Giacomo Grasselli, M.D., Gaetano Iapichino, M.D., Massimo Antonelli, M.D.,
Vieri Parrini, M.D., Gilberto Fiore, M.D., Roberto Latini, M.D.,
and Luciano Gattinoni, M.D., for the ALBIOS Study Investigators*
A BS T R AC T
From Dipartimento di Fisiopatologia Medi- BACKGROUND

co-Chirurgica e dei Trapianti, Fondazione
Istituto di Ricovero e Cura a Carattere Sci-
Although previous studies have suggested the potential advantages of albumin ad-
entifico (IRCCS) Ca’ Granda–Ospedale ministration in patients with severe sepsis, its efficacy has not been fully established.
Maggiore Policlinico, Università degli Studi
di Milano (P.C., G.I., L.G.), Dipartimento di METHODS
Anestesia, Rianimazione e Terapia del Do-
lore, Fondazione IRCCS Ca’ Granda– In this multicenter, open-label trial, we randomly assigned 1818 patients with se-
Ospedale Maggiore Policlinico (P.C., L.C., vere sepsis, in 100 intensive care units (ICUs), to receive either 20% albumin and
L.G.), IRCCS–Istituto di Ricerche Farmaco-
logiche Mario Negri (S.M., R.L.), Diparti-
crystalloid solution or crystalloid solution alone. In the albumin group, the target
mento di Scienze della Salute, Università serum albumin concentration was 30 g per liter or more until discharge from the
degli Studi di Milano Bicocca (R.F., A.P.), ICU or 28 days after randomization. The primary outcome was death from any cause
and Dipartimento di Emergenza–Urgenza,
Azienda Ospedaliera S. Paolo–Polo Uni-
at 28 days. Secondary outcomes were death from any cause at 90 days, the number
versitario (G.I.), Milan, Consorzio Mario of patients with organ dysfunction and the degree of dysfunction, and length of
Negri Sud, Santa Maria Imbaro (G.T., M.R., stay in the ICU and the hospital.
C.F.), Anestesiologia e Rianimazione, Di-
partimento Emergenza–Urgenza, Chirurg-
ia Generale e dei Trapianti, Policlinico Uni- RESULTS
versitario S. Orsola Malpighi, Bologna During the first 7 days, patients in the albumin group, as compared with those in
(S.F.), Dipartimento di Emergenza–Urgen-
za, Azienda Ospedaliera S. Gerardo, Mon- the crystalloid group, had a higher mean arterial pressure (P = 0.03) and lower net
za (G.G.), Policlinico Universitario A. Ge- fluid balance (P<0.001). The total daily amount of administered fluid did not differ
melli, Università Cattolica, Rome (M.A.), significantly between the two groups (P = 0.10). At 28 days, 285 of 895 patients
Ospedale del Mugello–Azienda Sanitaria
di Firenze, Florence (V.P.), and Ospedale S. (31.8%) in the albumin group and 288 of 900 (32.0%) in the crystalloid group had
Croce, Moncalieri (G.F.) — all in Italy. Ad- died (relative risk in the albumin group, 1.00; 95% confidence interval [CI], 0.87 to
dress reprint requests to Dr. Gattinoni at 1.14; P = 0.94). At 90 days, 365 of 888 patients (41.1%) in the albumin group and 389
the Dipartimento di Fisiopatologia Medi-
co-Chirurgica e dei Trapianti, Fondazione of 893 (43.6%) in the crystalloid group had died (relative risk, 0.94; 95% CI, 0.85 to
IRCCS Ca’ Granda–Ospedale Maggiore 1.05; P = 0.29). No significant differences in other secondary outcomes were ob-
Policlinico, Università degli Studi di Mila-
no, Via F. Sforza 35, 20122 Milan, Italy, or at
served between the two groups.
gattinon@policlinico.mi.it.
*Investigators of the Albumin Italian CONCLUSIONS
Outcome Sepsis (ALBIOS) study are In patients with severe sepsis, albumin replacement in addition to crystalloids, as
listed in the Supplementary Appendix,
available at NEJM.org.
compared with crystalloids alone, did not improve the rate of survival at 28 and 90
days. (Funded by the Italian Medicines Agency; ALBIOS ClinicalTrials.gov number,
2014, at NEJM.org. NCT00707122.)
N Engl J Med 2014;370:1412-21.
DOI: 10.1056/NEJMoa1305727
1412 98370;15
n engl j med nejm.org april 10, 2014
Albumin Replacement in Severe Sepsis or Septic Shock
F
or decades, human albumin has been study, were responsible for its execution and for
administered to patients to provide adequate the data analysis, made the decision to submit the
oncotic pressure and intravascular volume.1 manuscript for publication, and assume respon-
In 1998, however, a report from the Cochrane In- sibility for the fidelity of the study to the protocol
juries Group Albumin Reviewers indicated that (available at NEJM.org).
the administration of albumin may be potential- The trial was funded by the Italian Medicines
ly harmful in critically ill patients, as compared Agency, which had no role in the conduct of the
with the administration of crystalloid solutions.2 study, the reporting of the data, or the supply of
Subsequent meta-analyses reported contradictory study fluids. Albumin administered during the
findings.3,4 study was provided by each participating institu-
To clarify this issue, a large, double-blind, tion as part of the clinical treatment of critically
randomized trial (the Saline versus Albumin Fluid ill patients. The study protocol and the informed-
Evaluation [SAFE] study)5 was conducted, in which consent process were approved by the ethics
4% albumin solution was compared with normal committee at each participating institution. Written
saline as fluid replacement in critically ill pa- informed consent or deferred consent was ob-
tients, with results indicating that albumin ad- tained from each patient.
ministration was safe. A predefined subgroup Randomization was performed centrally, with
analysis showed that patients with severe sepsis the use of a computer-generated and blinded as-
receiving albumin were at a lower, although not signment sequence. Randomization was stratified
significantly lower, risk for death than those according to the participating ICU and the interval
receiving normal saline. In addition, a subsequent between the time that the patient met the clinical
study pointed out a potential benefit of maintain- criteria for severe sepsis and randomization. The
ing serum albumin at a level of more than 30 g conduct of the trial was overseen by the data and
per liter in critically ill patients.6 safety monitoring board, which performed an in-
There is a convincing rationale for the poten- terim analysis after the enrollment of 700 patients.
tial advantages of albumin administration during
severe sepsis.7 Albumin is the main protein re- PATIENTS
sponsible for plasma colloid osmotic pressure8; Patients 18 years of age or older who met the
it acts as a carrier for several endogenous and ex- clinical criteria for severe sepsis14 within the pre-
ogenous compounds,9 with antioxidant and anti- vious 24 hours at any time during their stay in
inflammatory properties, and as a scavenger of the ICU were enrolled in the study after being
reactive oxygen10,11 and nitrogen12 species and screened for eligibility criteria. Details of the in-
operates as a buffer molecule for acid–base equi- clusion and exclusion criteria are provided in the
librium.13 We therefore conducted a randomized, Supplementary Appendix.
controlled trial to investigate the effects of the
administration of albumin and crystalloids, as STUDY TREATMENTS
compared with crystalloids alone, targeting a se- Patients were randomly assigned to receive either
rum albumin level of 30 g per liter or more in a 20% albumin and crystalloid solution (albumin
population of patients with severe sepsis. group) or crystalloid solution alone (crystalloid
group) from randomization until day 28 or dis-
ME THODS charge from the ICU, whichever came first. Dur-
ing the early phase of volume resuscitation, flu-
STUDY OVERSIGHT AND DESIGN ids were administered in both groups according
We conducted the Albumin Italian Outcome Sep- to early goal-directed therapy.15
sis (ALBIOS) study — an investigator-initiated, After randomization, patients in the albumin
multicenter, open-label, randomized, controlled group received 300 ml of 20% albumin solution.
trial — in 100 intensive care units (ICUs) in Italy. From day 1 until day 28 or ICU discharge (which-
The members of the steering committee (see the ever came first), 20% albumin was administered
Supplementary Appendix, available with the full on a daily basis, to maintain a serum albumin
text of this article at NEJM.org) designed the level of 30 g per liter or more. In both groups,

nejm.org april 10, 2014 1413
crystalloids were administered whenever it was use of renal-replacement therapy, the incidence of
clinically indicated by the attending physician. acute kidney injury, the duration of mechanical
The administration of synthetic colloids was not ventilation, and the time to suspension of the ad-
allowed. All other treatments were at the discre- ministration of vasopressor or inotropic agents.
tion of the attending physician.
STATISTICAL ANALYSIS
OUTCOMES We originally determined that a sample of 1350 pa-
The primary outcome measure was death from tients would provide the study with 80% power
any cause at 28 days after randomization. The to detect an absolute between-group difference
principal secondary outcome measure was death of 7.5 percentage points in mortality at 28 days,
from any cause at 90 days after randomization. on the basis of an estimated baseline mortality of
Additional secondary outcomes were the number 45%, with a two-sided P value of less than 0.05
of patients with organ dysfunction and the de- indicating statistical significance. The study pro-
gree of dysfunction and the length of stay in the tocol specified the possibility of increasing the
ICU and the hospital. The severity of systemic sample to 1800 patients on the basis of a recom-
illness was assessed with the use of the Simpli- mendation by the data and safety monitoring
fied Acute Physiology Score, with scores ranging board during an interim analysis.
from 0 to 163 and higher scores indicating more All the analyses were conducted on an inten-
severe illness.16 Organ function was assessed tion-to-treat basis. Binary outcomes were com-
daily with the use of the Sequential Organ Failure pared with the use of the chi-square test, and
Assessment (SOFA) score,17 which ranges from continuous outcomes with the use of the Wilcoxon
0 to 4 for each of five components (respiratory, rank-sum test. Comparisons of fluid volumes
coagulation, liver, cardiovascular, and renal com- and physiological data over time were performed
ponents), with higher scores indicating more with the use of a two-factor analysis of variance
severe organ dysfunction (Table S1 in the Sup- for repeated measurements. We calculated sur-
plementary Appendix). New organ failures were vival estimates according to the Kaplan–Meier
defined as a change in a component score dur- method and compared them using a log-rank
ing the study from a baseline score of 0, 1, or 2 test. We performed an adjusted analysis using
to a score of 3 or 4.5,18,19 Tertiary outcomes, which robust Poisson regression for binary outcomes.
were assessed in post hoc analyses, included the In a post hoc analysis, the primary and principal
Albumin Group Crystalloid Group

Age — yr
Median 70 69
Interquartile range 57–77 59–77
Female sex — no. (%) 360 (39.9) 357 (39.4)
Body-mass index† 27±6 27±6
Reason for ICU admission — no. (%)
Medical 511 (56.6) 518 (57.1)
Elective surgery 69 (7.6) 58 (6.4)
Emergency surgery 323 (35.8) 331 (36.5)
Preexisting condition — no. (%)‡
Liver disease 13 (1.4) 14 (1.5)
COPD 113 (12.5) 108 (11.9)
Chronic renal failure 44 (4.9) 32 (3.5)
Immunodeficiency 115 (12.7) 128 (14.1)
Congestive or ischemic heart disease 149 (16.5) 165 (18.2)
SAPS II score§
Median 48 48
1414 n engl j med 370;15 100

nejm.org april 10, 2014
Albumin Group Crystalloid Group

Physiological variable¶
Heart rate — beats/min 105±22 106±20
Mean arterial pressure — mm Hg 74±16 73±15
Central venous pressure — mm Hg 10.0±4.9 9.8±4.7
Urine output — ml/hr
Median 50 50
Lactate — mmol/liter
Median 2.3 2.5
Interquartile range 1.4–4.2 1.6–4.3
Serum albumin — g/liter 24.1±6.3 24.2±6.2
Hemoglobin — g/dl 10.9±2.1 11.0±2.0
Central venous oxygen saturation — %
Median 73 73
SOFA score‖
Median 8 8
Organ dysfunction — no. (%)**
1 organ 188 (20.8) 208 (22.9)
2 organs 361 (40.0) 303 (33.4)
3 organs 236 (26.1) 248 (27.3)
4 organs 89 (9.9) 115 (12.7)
5 organs 29 (3.2) 33 (3.6)
Shock — no. (%)†† 565 (62.6) 570 (62.8)
Mechanical ventilation — no. (%) 709 (78.5) 737 (81.3)
Fluid administration in previous 24 hr — no. (%)
Albumin 153 (16.9) 176 (19.4)
Synthetic colloids 452 (50.1) 479 (52.8)
* Plus–minus values are means ±SD. There were no significant differences between the two groups except with respect
to central venous oxygen saturation (P = 0.02) and number of patients with organ dysfunction (P = 0.04). COPD denotes
chronic obstructive pulmonary disease, and ICU intensive care unit.
† The body-mass index is the weight in kilograms divided by the square of the height in meters.
‡ Among preexisting conditions, liver disease was defined as the presence of cirrhosis, portal hypertension, or previous
episodes of liver insufficiency; immunodeficiency as the concurrent presence of immunosuppressive diseases or receipt
of immunosuppressive therapies; and congestive or ischemic heart disease as New York Heart Association class II.
§ The Simplified Acute Physiology Score (SAPS II)16 was used to assess the severity of systemic illness at baseline.
Scores range from 0 to 163, with higher scores indicating more severe illness.
¶ Data on central venous pressure were available for 841 patients in the albumin group and 858 in the crystalloid
group; data on lactate level, for 874 and 867, respectively; data on serum albumin level, for 821 and 813, respectively;
data on hemoglobin level, for 893 and 894, respectively; and data on central venous oxygen saturation, for 798 and
802, respectively.
‖ The Sequential Organ Failure Assessment (SOFA) score17 includes subscores ranging from 0 to 4 for each of five
components (respiratory, coagulation, liver, cardiovascular, and renal components), with higher scores indicating
more severe organ dysfunction. The scoring was modified by excluding the assessment of cerebral failure (the
Glasgow Coma Scale), which was not performed in these patients, and by decreasing to 65 mm Hg the mean arterial
pressure threshold for a cardiovascular subscore of 1, for consistency with the hemodynamic targets as defined
according to the early goal-directed therapy.15
** Organ dysfunctions were defined as a SOFA score of 2 or more on the respiratory component; 2 or more on the co-
agulation component; 2 or more on the liver component; 1, 3, or 4 on the cardiovascular component; and 2 or more
on the renal component.5 A score of 2 on the cardiovascular component was not included because that score is as-
signed for the use of vasopressor drugs at low doses (a condition not considered to be strictly related to cardiovascu-
lar dysfunction).
†† Shock at the time of randomization was defined as a score of 3 or 4 on the cardiovascular component of the SOFA.5
n engl j med 370;15 101

A Figure 1. Serum Albumin Levels through Day 28

34 and Net Fluid Balance through Day 7.
32 Panel A shows the serum albumin concentration
30 through day 28 in patients receiving albumin and crys-
talloids or crystalloids alone. Day 0 was defined as the
28
Serum Albumin (g/liter)
time of randomization. Data are medians, with I bars

26
indicating interquartile ranges. The P value is for the
24 between-group comparison performed with the use
22 of a two-factor analysis of variance for repeated mea-
20 surements to test time (29 days for serum albumin,
18 including day 0) and group effects. Panel B shows the
16 net fluid balance through day 7 for patients receiving
14
albumin and crystalloids or crystalloids alone. The
Albumin
Crystalloids daily net fluid balance was calculated as the difference
P<0.001
between the total amount of administered fluid (in-
0
0 4 8 12 16 20 24 28 cluding 20% albumin; crystalloids; other blood prod-
ucts, such as packed red cells, fresh-frozen plasma, or
Study Day
platelets; and other fluids) and the total amount of
No. at Risk excreted fluid each day (including urinary output and
Albumin 821 677 483 335 239 198 148 107 other fluid losses, such as fluids potentially removed
Crystalloids 813 663 464 303 217 159 130 104
with continuous renal-replacement therapy, fluids lost
as feces, aspirated gastric content, drainage fluids,
B and insensible perspiration). For day 1, the net fluid
5000 Albumin balance was computed from the time of randomiza-
Crystalloids tion to day 1, which averaged 16 hours in the two
4000
study groups. The horizontal line in the boxes indi-
Net Fluid Balance (ml)
3000 cates the median, the top and bottom of the box the
interquartile range, and I bars the 5th and 95th per-
2000 centile range. The P value is for the between-group
comparison performed with the use of the two-factor
1000
analysis of variance for repeated measurements to test
0 time (7 days) and group effects.
−1000
ly assigned to receive 20% albumin and crystal-
−2000
P<0.001 loid solution (910 patients) or crystalloid solution
−3000 alone (908) for fluid replacement. Per protocol,
1 2 3 4 5 6 7
patient enrollment was stratified according to
Study Day
the interval between the time the patient met
No. at Risk the clinical criteria for severe sepsis and random-
Albumin 840 789 742 701 639 586 542
Crystalloids 844 795 735 685 635 587 529 ization: 6 hours or less (579 patients [31.8%])
versus more than 6 hours (1239 [68.2%]). A total
of 8 patients were excluded from the analysis
secondary outcomes were assessed in patients (2 patients in the albumin group owing to with-
who had septic shock and those who did not drawal of consent, and 5 in the albumin group
have septic shock at the time of enrollment. and 1 in the crystalloid group owing to a ran-
Heterogeneity of treatment effects among sub- domization error) (Fig. S1 in the Supplementary
groups was assessed with the use of the test for Appendix).
a common relative risk. SAS software, version After follow-up, data regarding death at 90 days
9.2 (SAS Institute), was used for all the analyses. were available for 888 of 903 patients (98.3%) in
the albumin group and for 893 of 907 (98.5%) in
R E SULT S the crystalloid group. Baseline characteristics were
similar between the two study groups, except for a
STUDY POPULATION slight imbalance in the number of patients with
From August 2008 through February 2012, a total organ dysfunction and values of central venous
of 1818 patients with severe sepsis were random- oxygen saturation (Table 1). The primary site of
1416 102
n engl j med 370;15 nejm.org april 10, 2014
infection, the type of identified microorganism, group and the crystalloid group during the 90
and the proportion of patients receiving anti- days after randomization (P = 0.39) (Fig. 2).
biotics were similar in the two groups (Table S2 No significant difference was observed between
in the Supplementary Appendix). the two study groups with respect to either the
number of newly developed organ failures or the
FLUID THERAPY AND TREATMENT EFFECTS median SOFA score (Table 2). Analysis of the
During the first 7 days, the albumin group, as SOFA score for each organ system revealed that,
compared with the crystalloid group, received a as compared with the crystalloid group, the al-
significantly larger volume of 20% albumin solu- bumin group had a lower cardiovascular score
tion (P<0.001) and less crystalloid solution (P = 0.03), a higher coagulation score (P = 0.04),
(P<0.001). In the albumin group, the administra- and a higher liver score (P = 0.02). No significant
tion of 20% albumin solution accounted for a differences were observed in other secondary
median daily average of 4.3% (interquartile and tertiary outcomes, with the exception of the
range, 2.9 to 5.8) of the total administered fluids. time to suspension of the administration of va-
The total daily amount of administered fluids in sopressor or inotropic agents, which was shorter
the first 7 days did not differ significantly bein the albumin group than in the crystalloid group
tween the albumin group and the crystalloid (P = 0.007) (Table 2).
group (3738 ml [interquartile range, 3174 to In subgroup analyses, no significant differ-
4437] and 3825 ml [interquartile range, 3205 to ence was observed in the prespecified subgroups
4533], respectively; P = 0.10) (Table S3 in the Sup- that were stratified according to the interval be-
plementary Appendix). tween the time the patient met the clinical crite-
The serum albumin level was significantly ria for severe sepsis and randomization (Fig. S3
higher in the albumin group than in the crystal- in the Supplementary Appendix). Conversely, a
loid group from day 1 to day 28 (P<0.001) (Fig. significant difference was observed in a post hoc
1A). During the first 7 days, patients in the al- subgroup analysis that included 1121 patients
bumin group had a significantly lower heart rate with septic shock, as compared with 660 with-
than those in the crystalloid group (P = 0.002), as out septic shock, at the time of enrollment (rela-
well as a significantly higher mean arterial pres- tive risk with septic shock, 0.87; 95% CI, 0.77 to
sure (P = 0.03) (Table S4 and Fig. S2 in the 0.99; relative risk without septic shock, 1.13;
Supplementary Appendix). Daily net fluid bal- 95% CI, 0.92 to 1.39; P = 0.03 for heterogeneity)
ances were lower in the albumin group than in (Fig. S3 in the Supplementary Appendix). Adjust-
the crystalloid group (P<0.001) (Fig. 1B). The ment for baseline covariates did not signifi-
median cumulative net fluid balance was also cantly modify these results (Table S6 in the
significantly lower in the albumin group than in Supplementary Appendix).
the crystalloid group (347 ml [interquartile
range, −3266 to 4042] vs. 1220 ml [interquartile DISCUSSION
range, −2767 to 5034], P = 0.004) (Table S5 in the
Supplementary Appendix). The main results of this large-scale trial provide
evidence regarding both the efficacy and the
OUTCOMES safety of the use of human albumin during se-
At 28 days after randomization, 285 of 895 pa- vere sepsis — an interventional strategy that has
tients (31.8%) in the albumin group and 288 of long been debated.21,22 The addition of albumin
900 (32.0%) in the crystalloid group had died to crystalloids during the first 28 days of treat-
(relative risk in the albumin group, 1.00; 95% ment to maintain a serum albumin level of 30 g
confidence interval [CI], 0.87 to 1.14; P = 0.94) per liter or more is safe but does not provide a
(Table 2). At 90 days of follow-up, 365 of 888 survival advantage over crystalloids alone, over a
patients (41.1%) in the albumin group and 389 of follow-up period of 90 days. Similar findings
893 (43.6%) in the crystalloid group had died were observed in the subgroup stratified accord-
(relative risk, 0.94; 95% CI, 0.85 to 1.05; P = 0.29). ing to the interval between the time the patient
No significant difference in the probability of met the clinical criteria for severe sepsis and
survival was observed between the albumin treatment application.
n engl j med 370;15 103

The findings in our trial may appear to con- dysfunction.4,6 Similar beneficial effects were
tradict those of the predefined subgroup analysis also suggested by a large meta-analysis, which
from the SAFE study,5 which suggested a sur- concluded that the use of albumin-containing
vival advantage with an albumin-based strategy solutions could be associated with lower mortal-
during severe sepsis. The plausibility of this hy- ity than that seen with other fluid regimens.24
pothesis was supported by the significant hemo- Our results confirm that administration of
dynamic advantages observed23 and by further albumin produces small but significant hemody-
investigations showing that the correction of namic advantages. A significantly greater pro-
hypoalbuminemia reduced the severity of organ portion of patients in the albumin group than in
Table 2. Outcomes.
Relative Risk
Outcome Albumin Group Crystalloid Group (95% CI) P Value
Primary outcome: death at 28 days — no./total no. (%) 285/895 (31.8) 288/900 (32.0) 1.00 (0.87–1.14) 0.94
Secondary outcomes
Death at 90 days — no./total no. (%) 365/888 (41.1) 389/893 (43.6) 0.94 (0.85–1.05) 0.29
New organ failures — no./total no. (%)* 0.99
None 372/836 (44.5) 383/841 (45.5)
1 organ 283/836 (33.9) 287/841 (34.1)
2 organs 130/836 (15.6) 123/841 (14.6)
3 organs 40/836 (4.8) 36/841 (4.3)
4 organs 10/836 (1.2) 11/841 (1.3)
5 organs 1/836 (0.1) 1/841 (0.1)
SOFA score† — 0.23
Median 6.00 5.62
SOFA subscore†
Cardiovascular — 0.03
Median 1.20 1.42
Respiratory — 0.63
Median 2.00 2.00
Renal — 0.15
Median 0.83 0.75
Coagulation — 0.04
Median 0.64 0.50
Liver — 0.02
Median 0.28 0.20
Length of stay — days
In ICU — 0.42
Median 9 9
In hospital‡ — 0.65
Median 20 20
1418 104
Relative Risk
Outcome Albumin Group Crystalloid Group (95% CI) P Value
Tertiary outcomes§
Renal-replacement therapy — no./total no. (%)¶ 222/903 (24.6) 194/907 (21.4) 0.11
Acute kidney injury — no./total no. (%)‖ 183/834 (21.9) 190/837 (22.7) 0.71
Duration of mechanical ventilation — days** — 0.50
Median 6 6
Time to suspension of vasopressor or inotropic — 0.007
agents — days††
Median 3 4
* New organ failures were defined by a change in a specific component of the SOFA17 from a score of 0, 1, or 2 at baseline to a score of 3
or 4 during the study period.5,17,18
† The values are the median and interquartile range of the SOFA score, representing the average of the daily SOFA scores for each individu-
al patient during his or her study period (including the SOFA score at baseline). No imputation was performed for missing data.
‡ The length of stay in the hospital included the length of stay in the ICU.
§ Tertiary outcomes were analyzed in post hoc analyses.
¶ Included are patients with any form of renal-replacement therapy prescribed by the attending physician during the study period, including
patients with chronic renal failure at baseline.
‖ Acute kidney injury was defined according to the risk, injury, failure, loss, and end-stage kidney injury (RIFLE) criteria20 for acute kidney in-
jury on the basis of daily incremental increases in serum creatinine levels from baseline during the study period.
** The duration of ventilatory support includes only the time during the study period, which was not necessarily the total duration of ventila-
tory support.
†† The time to the suspension of vasopressor or inotropic agents was assessed as the number of days of administration of vasopressor or
inotropic agents in patients for whom such treatment was ongoing at baseline. Data were available for 582 patients in the albumin group
and 576 in the crystalloid group.
the crystalloid group reached the targeted mean tration during severe sepsis. The incidence of
arterial pressure within 6 hours after random- new organ failures during the study was simi-
ization (Table S7 in the Supplementary Appendix). lar in the two groups. We observed slightly
During the first 7 days, the mean arterial pres- higher average SOFA subscores for liver and
sure was higher, whereas the heart rate and net coagulation in the albumin group, indicating a
fluid balance were lower, in the albumin group higher serum bilirubin and a lower platelet
than in the crystalloid group. Moreover, the av- count, respectively, than were observed in the
erage cardiovascular SOFA subscore over the crystalloid group. Nonetheless, the absolute
course of the study period was lower in the al- excess in the serum bilirubin concentration in
bumin group, and the time to the suspension of the albumin group was marginal (median, 1.0
inotropic or vasopressor agents was shorter, in- mg per deciliter [interquartile range, 0.6 to 1.7]
dicating a decreased use of vasopressors. These vs. 0.9 mg per deciliter [interquartile range, 0.5
effects were obtained with similar amounts of to 1.5], P<0.001) and was probably related to
administered fluids in the two study groups. the methods used to prepare albumin solu-
These findings confirm a physiological advan- tions, which may be inefficient in clearing bili-
tage of albumin administration during severe rubin content from plasma.21,27 The slight re-
sepsis, including a larger fluid distribution within duction in platelet counts in the albumin group
the intravascular compartment and, in addition, may be a further marker of a larger expansion
possible effects of albumin as a scavenger of ni- of the intravascular compartment in this group
tric oxide,12 mediating peripheral vasodilatation than in the crystalloid group, with a consequent
during sepsis.25,26 dilution of the hemoglobin content (Table S4 in
The secondary outcomes also provide a de- the Supplementary Appendix).
tailed profile of the safety of albumin adminis- Post hoc univariate and multivariate analyses
n engl j med 370;15 nejm.org105

april 10, 2014 1419
edly lower than that administered in the SAFE

1.0 study, since our goal was to correct hypoalbu-
minemia and not to directly replace intravascu-
0.9
lar volume. Second, the observed mortality at 28
0.8 days was lower than originally expected, thereby
increasing the likelihood that the study was un-
0.7
Albumin
derpowered. Finally, only approximately one third
of the patients were enrolled during the early phase
0.6
of severe sepsis.
Crystalloids In conclusion, the use of albumin in addition
0.5
to crystalloids to correct hypoalbuminemia, as
0.4 compared with the use of crystalloids alone, in
patients with severe sepsis during their stay in
0.3
the ICU did not provide a survival benefit at 28
P=0.39 or 90 days, despite improvements in hemodynamic
0.0
0 10 20 30 40 50 60 70 80 90
variables. The clinical benefit of albumin that
Days since Randomization
was seen in the post hoc analysis of the subgroup
of patients with septic shock warrants further
No. at Risk
Albumin 903 733 647 597 567 556 545 535 529 523 confirmation.
Crystalloids 907 729 652 598 676 551 538 521 511 504
Supported by the Italian Medicines Agency.
Figure 2. Probability of Survival from Randomization through Day 90. Disclosure forms provided by the authors are available with
the full text of this article at NEJM.org.
The graph shows the Kaplan–Meier estimates for the probability of sur-
We thank Nicola Bottino, M.D., Giuseppe Breda, M.D.,
vival among patients receiving albumin and crystalloids and among those Davide Chiumello, M.D., Stefania Crotti, M.D., Sergio De
receiving crystalloids alone. The P value was calculated with the use of the Chiara, M.D., Alfredo Lissoni, M.D., Francesca Pagan, M.D.,
log-rank test. Mauro Panigada, M.D., Riccarda Russo, M.D., Monica Savioli,
M.D., Alberto Sicignano, M.D., Daniela Tubiolo, M.D., and all
the residency and nursing staff of the Rianimazione Generale
Emma Vecla, Fondazione Istituto di Ricovero e Cura a Carattere
of data from the 1121 patients with septic shock Scientifico (IRCCS) Ca’ Granda–Ospedale Maggiore Policlinico,
showed significantly lower mortality at 90 days Milan (coordinating center), for support in monitoring the ran-
in the albumin group than in the crystalloid group. domization process and study compliance; Paola Bruzzone,
M.D., Federico Polli, M.D., and Federica Tallarini, M.D., of the
Conversely, in the subgroup of patients with se- Dipartimento di Anestesia, Rianimazione e Terapia del Dolore,
vere sepsis without shock, mortality appeared to Fondazione IRCCS Ca’ Granda–Ospedale Maggiore Policlinico,
be higher among those who were treated with Milan, and Tommaso Mauri, M.D., of the Dipartimento di
Medicina Perioperatoria e Terapia Intensiva, Azienda
albumin than among those treated with crystal- Ospedaliera S. Gerardo di Monza, for help in data-quality as-
loids alone, although the difference was far from sessment; Luisella Pirozzi of the Consorzio Mario Negri Sud,
significant. This analysis was not prespecified, Santa Maria Imbaro, and Marina Leonardelli of the
Dipartimento di Anestesia, Rianimazione e Terapia del Dolore,
and therefore it may be characterized by well- Fondazione IRCCS Ca’ Granda–Ospedale Maggiore Policlinico,
known biases. Nonetheless, a state of shock asso- Milan, for support in monitoring study compliance and retriev-
ciated with severe sepsis represents a well-defined ing outcome data; Paolo Cadringher and Eleonora Carlesso,
M.Sc., of the Dipartimento di Fisiopatologia Medico-Chirurgica
clinical entity. Moreover, if the oncotic, anti- e dei Trapianti, Fondazione IRCCS Ca’ Granda–Ospedale
inflammatory, and nitric oxide–scavenging prop- Maggiore Policlinico, Università degli Studi, Milan, for pri-
erties of albumin are of clinical importance, mary support in creating and managing electronic clinical re-
search forms; Jean-Louis Vincent, M.D., of the Department of
these may be maximally exploited in the condi- Intensive Care, Erasme University Hospital, Université Libre de
tions that are the most severe, such as cardiovas- Bruxelles, Brussels, Peter M. Suter, M.D., of the University of
cular dysfunction. Geneva, Geneva, Maria Grazia Valsecchi of the Dipartimento
di Medicina Clinica e Prevenzione, Università degli Studi di
Our trial has certain limitations. First, we in- Milano-Bicocca, Milan, and Amedeo Santosuosso of the
cluded the use of albumin solutions with a greater Dipartimento di Giurisprudenza, Università degli Studi di
concentration than those used in the SAFE study Pavia, Pavia, for serving on the data and safety monitoring
board; the study patients and their relatives for their participa-
(20% vs. 4%). Consequently, the volume of albu- tion; and the physicians and nursing staff of the participating
min solution that was administered was mark- centers for their cooperation.
1420 106370;15
REFERENCES
1. Erstad BL, Gales BJ, Rappaport WD. 10. King TP. On the sulfhydryl group of center, prospective study. Crit Care Med
The use of albumin in clinical practice. human plasma albumin. J Biol Chem 1998;26:1793-800.
Arch Intern Med 1991;151:901-11. 1961;236:PC5. 19. Moreno R, Vincent JL, Matos R, et al.
2. Cochrane Injuries Group Albumin 11. Quinlan GJ, Margarson MP, Mumby The use of maximum SOFA score to quan-
Reviewers. Human albumin administra- S, Evans TW, Gutteridge JM. Administra- tify organ dysfunction/failure in intensive
tion in critically ill patients: systematic tion of albumin to patients with sepsis care: results of a prospective, multicentre
review of randomised controlled trials. syndrome: a possible beneficial role in study. Working Group on Sepsis related
BMJ 1998;317:235-40. plasma thiol repletion. Clin Sci (Lond) Problems of the ESICM. Intensive Care
3. Wilkes MM, Navickis RJ. Patient sur- 1998;95:459-65. Med 1999;25:686-96.
vival after human albumin administration: 12. Stamler JS, Jaraki O, Osborne J, et al. 20. Kellum JA, Bellomo R, Ronco C. The
a meta-analysis of randomized, controlled Nitric oxide circulates in mammalian concept of acute kidney injury and the RIFLE
trials. Ann Intern Med 2001;135:149-64. plasma primarily as an S-nitroso adduct criteria. Contrib Nephrol 2007;156:10-6.
4. Vincent JL, Dubois MJ, Navickis RJ, of serum albumin. Proc Natl Acad Sci 21. Vincent JL. Relevance of albumin in
Wilkes MM. Hypoalbuminemia in acute U S A 1992;89:7674-7. modern critical care medicine. Best Pract
illness: is there a rationale for interven- 13. Reeves RB. Temperature-induced Res Clin Anaesthesiol 2009;23:183-91.
tion? A meta-analysis of cohort studies changes in blood acid-base status: Don- 22. Vincent JL, Gottin L. Type of fluid in
and controlled trials. Ann Surg 2003; nan rCl and red cell volume. J Appl Physi- severe sepsis and septic shock. Minerva
237:319-34. ol 1976;40:762-7. Anestesiol 2011;77:1190-6.
5. Finfer S, Bellomo R, Boyce N, French 14. Bone RC, Balk RA, Cerra FB, et al. Def- 23. The SAFE Study Investigators. Impact of
J, Myburgh J, Norton R. A comparison of initions for sepsis and organ failure and albumin compared to saline on organ func-
albumin and saline for fluid resuscitation guidelines for the use of innovative thera- tion and mortality of patients with severe
in the intensive care unit. N Engl J Med pies in sepsis. Chest 1992;101:1644-55. sepsis. Intensive Care Med 2011;37:86-96.
2004;350:2247-56. 15. Rivers E, Nguyen B, Havstad S, et al. 24. Delaney AP, Dan A, McCaffrey J, Fin-
6. Dubois MJ, Orellana-Jimenez C, Melot Early goal-directed therapy in the treat- fer S. The role of albumin as a resuscita-
C, et al. Albumin administration im- ment of severe sepsis and septic shock. tion fluid for patients with sepsis: a sys-
proves organ function in critically ill hy- N Engl J Med 2001;345:1368-77. tematic review and meta-analysis. Crit
poalbuminemic patients: a prospective, 16. Le Gall JR, Lemeshow S, Saulnier F. Care Med 2011;39:386-91.
randomized, controlled, pilot study. Crit A new Simplified Acute Physiology Score 25. Hollenberg SM, Piotrowski MJ, Par-
Care Med 2006;34:2536-40. (SAPS II) based on a European/North Amer- rillo JE. Nitric oxide synthase inhibition
7. Quinlan GJ, Martin GS, Evans TW. ican multicenter study. JAMA 1993;270: reverses arteriolar hyporesponsiveness to
Albumin: biochemical properties and 2957-63. [Erratum, JAMA 1994;271:1321.] endothelin-1 in septic rats. Am J Physiol
therapeutic potential. Hepatology 2005;41: 17. Vincent JL, Moreno R, Takala J, et al. 1997;272:R969-R974.
1211-9. The SOFA (Sepsis-related Organ Failure 26. Titheradge MA. Nitric oxide in septic
8. Weil MH, Henning RJ, Puri VK. Col- Assessment) score to describe organ dys- shock. Biochim Biophys Acta 1999;1411:
loid oncotic pressure: clinical significance. function/failure. Intensive Care Med 437-55.
Crit Care Med 1979;7:113-6. 1996;22:707-10. 27. McCann KB, Vucica Y, Famulari S,
9. Sudlow G, Birkett DJ, Wade DN. The 18. Vincent JL, de Mendonça A, Cantraine Bertolini J. Effect of processing methods
characterization of two specific drug F, et al. Use of the SOFA score to assess on colouration of human serum albumin
binding sites on human serum albumin. the incidence of organ dysfunction/failure preparations. Biologicals 2009;37:32-6.
Mol Pharmacol 1975;11:824-32. in intensive care units: results of a multi- Copyright © 2014 Massachusetts Medical Society.
specialties and topics at nejm.org

Specialty pages at the Journal’s website (NEJM.org) feature articles in cardiology,
endocrinology, genetics, infectious disease, nephrology, pediatrics, and many other
medical specialties. These pages, along with collections of articles on clinical and
nonclinical topics, offer links to interactive and multimedia content and feature
recently published articles as well as material from the NEJM archive (1812–1989).

107 april 10, 2014 1421
c or r e sp ondence
To the Editor: The Albumin Italian Outcome with different control fluids and patient popula-
Sepsis study conducted by Caironi et al. (April 10 tions.4 The result is also consistent with the
issue)1 is the third large-scale, randomized trial observed reduction in mortality associated with
to compare albumin with crystalloids in adult albumin use among patients with spontaneous
patients with severe sepsis. The first such trial bacterial peritonitis, a disease that shares im-
was the Saline versus Albumin Fluid Evaluation portant pathophysiological features with severe
study.2 In addition, the Early Albumin Resuscita- sepsis.5
tion during Septic Shock study has been complet- Christian J. Wiedermann, M.D.
ed and its mortality results published.3 Central Hospital of Bolzano
In all three trials, mortality was lower among Bolzano, Italy
patients receiving albumin, and the respective christian.wiedermann@asbz.it
relative risks coincided closely, ranging from Michael Joannidis, M.D.
0.87 to 0.94 (Fig. 1). Although the effect did not Medical University of Innsbruck
Innsbruck, Austria
attain statistical significance in any of the indi-
Dr. Wiedermann reports receiving lecture fees from CSL
vidual trials, the pooled relative risk in all three Behring, Baxter, and the Plasma Protein Therapeutics Association;
trials is 0.92 (95% confidence interval [CI], 0.84 and Dr. Joannidis, lecture fees from Baxter, Fresenius, Gambro,
to 1.00; P = 0.046), indicating a significant re- Orion Pharma, CLS Behring, and B. Braun Melsungen. No other
potential conflict of interest relevant to this letter was reported.
duction in mortality associated with albumin
use among adults with severe sepsis. This result 1. Caironi P, Tognoni G, Masson S, et al. Albumin replacement
in patients with severe sepsis or septic shock. N Engl J Med
supports the conclusion of a previous meta- 2014;370:1412-21.
analysis involving predominantly small trials
Relative
Trial Albumin Crystalloids Weight Relative Risk (95% CI) P Value
no. of patients who died/total no. %
SAFE2 185/603 217/615 30.41 0.87 (0.74–1.02)
ALBIOS1 365/888 389/893 54.90 0.94 (0.85–1.05)
EARSS3 96/399 103/393 14.69 0.92 (0.72–1.17)
All trials 646/1890 709/1901 100.00 0.92 (0.84–1.00) 0.046
0.5 1.0 2.0
Albumin Better Crystalloids Better
Figure 1. Meta-Analysis of Mortality in Large-Scale Randomized Trials Comparing Albumin with Crystalloids in Adult
Patients with Severe Sepsis.
A fixed-effect model was used in the analysis. The size of the squares indicates the data points from the individual
trials scaled according to the percentage of total weight (with individual trial weight equaling the proportion of total
patients receiving albumin multiplied by the number of deaths in the crystalloids group), and the diamond indicates
the pooled findings. The dashed line indicates the pooled relative risk. The proportion of variation attributable to
heterogeneity (I2) was 0% (P = 0.71). ALBIOS denotes Albumin Italian Outcome Sepsis, CI confidence interval,
EARSS Early Albumin Resuscitation during Septic Shock, and SAFE Saline versus Albumin Fluid Evaluation.

2. The SAFE Study Investigators. Impact of albumin compared

to saline on organ function and mortality of patients with severe
leading to insufficient pharmacologic activity, as
sepsis. Intensive Care Med 2011;37:86-96. nitric oxide modulation,2 antioxidant action,3
3. Charpentier J, Mira J-P. Efficacy and tolerance of hyperoncotic and anti-immunosuppressive action.4 Although
albumin administration in septic shock patients: the EARSS
study. Intensive Care Med 2011;37:Suppl 1:S115.
this mechanism is speculative, the available data
4. Delaney AP, Dan A, McCaffrey J, Finfer S. The role of albu- suggest that in patients with the most severe ex-
min as a resuscitation fluid for patients with sepsis: a system- pression of sepsis albumin supplementation may
atic review and meta-analysis. Crit Care Med 2011;39:386-91.
5. Salerno F, Navickis RJ, Wilkes MM. Albumin infusion im-
represent an option.
proves outcomes of patients with spontaneous bacterial perito- Pietro Caironi, M.D.
nitis: a meta-analysis of randomized trials. Clin Gastroenterol Fondazione Istituto di Ricovero e Cura a Carattere Scientifico
Hepatol 2013;11:123-30. (IRCCS) Ca’ Granda–Ospedale Maggiore Policlinico
DOI: 10.1056/NEJMc1405675 Milan, Italy
The Authors Reply: Wiedermann and Joannidis Gianni Tognoni, M.D.

Fondazione Mario Negri Sud
suggest that there is a survival advantage associ- Santa Maria Imbaro, Italy
ated with albumin use in patients with severe Luciano Gattinoni, M.D.
sepsis. Formal meta-analyses will confirm or re- Fondazione IRCCS Ca’ Granda–Ospedale Maggiore Policlinico
fute their findings. Our study was not limited to Milan, Italy
the resuscitation phase but included albumin gattinon@policlinico.mi.it
Since publication of their article, the authors report no fur-
supplementation for 28 days after enrollment.
The improvement in survival associated with al-
bumin use was concentrated among patients 1. Quinlan GJ, Martin GS, Evans TW. Albumin: biochemical
properties and therapeutic potential. Hepatology 2005;41:1211-9.
with septic shock (1121 patients; 90-day mortal- 2. Stamler JS, Jaraki O, Osborne J, et al. Nitric oxide circulates
ity, 43.6% in the albumin group vs. 49.9% in the in mammalian plasma primarily as an S-nitroso adduct of se-
crystalloid group; relative risk, 0.87; 95% CI, 0.77 rum albumin. Proc Natl Acad Sci U S A 1992;89:7674-7.
3. Quinlan GJ, Margarson MP, Mumby S, Evans TW, Gutteridge
to 0.99; P = 0.03). These data, together with the JM. Administration of albumin to patients with sepsis syn-
lack of effects in patients enrolled with early sep- drome: a possible beneficial role in plasma thiol repletion. Clin
sis, suggest that there are beneficial effects as- Sci (Lond) 1998;95:459-65.
4. O’Brien AJ, Fullerton JN, Massey KA, et al. Immunosuppres-
sociated with albumin use in relation to its ancil- sion in acutely decompensated cirrhosis is mediated by prosta-
lary functions, rather than only to its primary glandin E2. Nat Med 2014;20:518-23.
oncotic properties.1 We may speculate that the DOI: 10.1056/NEJMc1405675
benefits of albumin administration manifest
mainly when endogenous albumin is function-
ally exhausted owing to the severity of the injury,
78 n engl j med 371;1109

CIRCULATORY SHOCK
There are four basic forms of circulatory shock: hypovolemic, cardiogenic, obstructive, and distributive. This
section covers our understanding of the physiological changes that characterize each and provides guidance on
how to recognize and manage them.
110
Case Challenge
Circulatory Shock
A 77-year-old man was admitted to the intensive care unit after colon resection, complicated by fecal peritonitis.
He is undergoing mechanical ventilation and has received 1 liter of Hartmann’s solution. Blood pressure is
now 85/50 mm Hg; the heart rate is 105 beats per minute. What should be done next?
A 77-year-old man is admitted to the intensive care unit (ICU) of a university hospital from the operating room
after resection of the rectosigmoid colon with closure of the rectal stump and formation of an end colostomy for
the treatment of fecal peritonitis caused by a perforated sigmoid colon. The patient’s medical history includes
treated hypertension and hypercholesterolemia, previous heavy alcohol intake, and mild cognitive impairment.
He has received a total of 4 liters of crystalloid in the operating room. (In the previous installment of this case
there were 9253 votes on the options for fluid resuscitation. Normal saline was chosen by 36% of respondents,
albumin by 30%, Ringer’s lactate by 26%, and hydroxyetheyl starch by 6%.)
When the patient arrives in the ICU, the arterial blood pressure is 88/52 mm Hg (mean arterial pressure,
64 mm Hg), the heart rate is 120 beats per minute in sinus rhythm, and the central venous pressure is 6 mm Hg.
The results of serum biochemical analyses are as follows: sodium, 142 mmol per liter; potassium, 4.4 mmol per
liter; chloride, 109 mmol per liter; urea, 22.0 mg per deciliter (7.9 mmol per liter); albumin, 23 g per liter; and
111
c reatinine, 2.3 mg per deciliter (203 µmol per liter). He has now been in the ICU for an hour. He is still undergoing
mechanical ventilation and has received 1 liter of Hartmann’s solution (consisting of sodium, 131 mmol per liter;
potassium, 5.4 mmol per liter; chloride, 111 mmol per liter; lactate, 29 mmol per liter; and calcium, 2 mmol per
liter, for a total osmolarity of 280.6 mOsm per liter) and 500 ml of 5% human albumin solution. The arterial
blood pressure is now 85/50 mm Hg (mean arterial pressure, 62 mm Hg), the heart rate is 105 beats per minute
in sinus rhythm, the central venous pressure is 9 mm Hg, the capillary refill time is estimated at 1 second, and
the measured urine output for the past hour is 35 ml.

What monitoring strategies would you use to guide hemodynamic support for this patient?
A. An echocardiographic assessment of cardiac function.

B. A pulmonary-artery catheterization for complete, continuous monitoring.
C. Measurement of variation in pulse pressure during mechanical ventilation.
D. Measurement of central venous oxygen saturation with a goal of more than 70%.
112
review article

Circulatory Shock
Jean-Louis Vincent, M.D., Ph.D., and Daniel De Backer, M.D., Ph.D.
S
From the Department of Intensive Care, hock is the clinical expression of circulatory failure that
Erasme Hospital, Université Libre de Brux- results in inadequate cellular oxygen utilization. Shock is a common condi-
elles, Brussels. Address reprint requests to
Dr. Vincent at the Department of Intensive tion in critical care, affecting about one third of patients in the intensive care
Care, Erasme University Hospital, Rte. de unit (ICU).1 A diagnosis of shock is based on clinical, hemodynamic, and bio-
Lennik 808, B-1070 Brussels, Belgium, or chemical signs, which can broadly be summarized into three components. First,
at jlvincen@ulb.ac.be.
systemic arterial hypotension is usually present, but the magnitude of the hypoten-
N Engl J Med 2013;369:1726-34. sion may be only moderate, especially in patients with chronic hypertension. Typi-
DOI: 10.1056/NEJMra1208943
cally, in adults, the systolic arterial pressure is less than 90 mm Hg or the mean
arterial pressure is less than 70 mm Hg, with associated tachycardia. Second, there
are clinical signs of tissue hypoperfusion, which are apparent through the three
“windows” of the body2: cutaneous (skin that is cold and clammy, with vasocon-
striction and cyanosis, findings that are most evident in low-flow states), renal
(urine output of <0.5 ml per kilogram of body weight per hour), and neurologic
(altered mental state, which typically includes obtundation, disorientation, and
confusion). Third, hyperlactatemia is typically present, indicating abnormal cellular
oxygen metabolism. The normal blood lactate level is approximately 1 mmol per liter,
but the level is increased (>1.5 mmol per liter) in acute circulatory failure.
Pathoph ysiol o gic a l Mech a nisms
Shock results from four potential, and not necessarily exclusive, pathophysiological
mechanisms3: hypovolemia (from internal or external fluid loss), cardiogenic fac-
tors (e.g., acute myocardial infarction, end-stage cardiomyopathy, advanced valvular
heart disease, myocarditis, or cardiac arrhythmias), obstruction (e.g., pulmonary
embolism, cardiac tamponade, or tension pneumothorax), or distributive factors
(e.g., severe sepsis or anaphylaxis from the release of inflammatory mediators)
An interactive (Fig. 1A and the interactive graphic, available at NEJM.org). The first three mech-
graphic showing anisms are characterized by low cardiac output and, hence, inadequate oxygen trans-
initial assessment of
shock is available
port. In distributive shock, the main deficit lies in the periphery, with decreased
at NEJM.org systemic vascular resistance and altered oxygen extraction. Typically, in such cases
cardiac output is high, although it may be low as a result of associated myocardial
depression. Patients with acute circulatory failure often have a combination of these
mechanisms. For example, a patient with distributive shock from severe pancreatitis,
anaphylaxis, or sepsis may also have hypovolemia and cardiogenic shock from
myocardial depression.
Differ en t i a l Di agnosis
Septic shock, a form of distributive shock, is the most common form of shock
among patients in the ICU, followed by cardiogenic and hypovolemic shock;
obstructive shock is relatively rare (Fig. 1B and 1C). In a trial involving more than
1726 n engl j med 113

369;18 nejm.org october 31, 2013
The New
Medical For All
Society. personal
reserved.
1600 patients with shock who were randomly as- fuse (fluid resuscitation), and pump (administra-
signed to receive either dopamine or norepineph- tion of vasoactive agents).
rine, septic shock occurred in 62% of the patients,
cardiogenic shock in 16%, hypovolemic shock in Ventilatory Support
16%, other types of distributive shock in 4%, and The administration of oxygen should be started im-
obstructive shock in 2%.4 mediately to increase oxygen delivery and prevent
The type and cause of shock may be obvious pulmonary hypertension. Pulse oximetry is often
from the medical history, physical examination, unreliable as a result of peripheral vasoconstric-
or clinical investigations. For example, shock tion, and precise determination of oxygen require-
after traumatic injury is likely to be hypovolemic ments will often require blood gas monitoring.
(due to blood loss), but cardiogenic shock or Mechanical ventilation by means of a mask
distributive shock may also occur, alone or in rather than endotracheal intubation has a lim-
combination, caused by such conditions as car- ited place in the treatment of shock because
diac tamponade or spinal cord injury. A full clini- technical failure can rapidly result in respiratory
cal examination should include assessment of and cardiac arrest. Hence, endotracheal intuba-
skin color and temperature, jugular venous dis- tion should be performed to provide invasive
tention, and peripheral edema. The diagnosis mechanical ventilation in nearly all patients with
can be refined with point-of-care echocardio- severe dyspnea, hypoxemia, or persistent or wors-
graphic evaluation, which includes assessment ening acidemia (pH, <7.30). Invasive mechanical
for pericardial effusion, measurement of left and ventilation has the additional benefits of reduc-
right ventricular size and function, assessment for ing the oxygen demand of respiratory muscles
respiratory variations in vena cava dimensions, and decreasing left ventricular afterload by in-
and calculation of the aortic velocity–time inte- creasing intrathoracic pressure. An abrupt de-
gral, a measure of stroke volume. Whenever pos- crease in arterial pressure after the initiation of
sible, focused echocardiography should be per- invasive mechanical ventilation strongly suggests
formed as soon as possible in any patient hypovolemia and a decrease in venous return.
presenting with shock (Fig. 1A).5,6 The use of sedative agents should be kept to a
minimum to avoid further decreases in arterial
pressure and cardiac output.
Ini t i a l A pproach
t o the Pat ien t in Sho ck
Fluid Resuscitation
Early, adequate hemodynamic support of patients Fluid therapy to improve microvascular blood
in shock is crucial to prevent worsening organ flow and increase cardiac output is an essential
dysfunction and failure. Resuscitation should be part of the treatment of any form of shock. Even
started even while investigation of the cause is patients with cardiogenic shock may benefit
ongoing. Once identified, the cause must be cor- from fluids, since acute edema can result in a
rected rapidly (e.g., control of bleeding, percuta- decrease in the effective intravascular volume.
neous coronary intervention for coronary syn- However, fluid administration should be closely
dromes, thrombolysis or embolectomy for massive monitored, since too much fluid carries the risk
pulmonary embolism, and administration of anti- of edema with its unwanted consequences.
biotics and source control for septic shock). Pragmatic end points for fluid resuscitation
Unless the condition is rapidly reversed, an are difficult to define. In general, the objective is
arterial catheter should be inserted for monitor- for cardiac output to become preload-indepen-
ing of arterial blood pressure and blood sam- dent (i.e., on the plateau portion of the Frank–
pling, plus a central venous catheter for the infu- Starling curve), but this is difficult to assess
sion of fluids and vasoactive agents and to guide clinically. In patients receiving mechanical ventila-
fluid therapy. The initial management of shock tion, signs of fluid responsiveness may be identi-
is problem-oriented, and the goals are therefore fied either directly from beat-by-beat stroke-volume
the same, regardless of the cause, although the measurements with the use of cardiac-output
exact treatments that are used to reach those monitors or indirectly from observed variations
goals may differ. A useful mnemonic to describe in pulse pressure on the arterial-pressure tracing
the important components of resuscitation is the during the ventilator cycle. However, such bedside
VIP rule7: ventilate (oxygen administration), in- inferences have some limitations8 — notably,

114 october 31, 2013 1727
The New
For personal ofNo
use only. Medicine
Medical 2014. For
rights reserved.
Source: The New England Journal of Medicine The n e w e ng l a n d j o u r na l of m e dic i n e
A B Types of shock
Arterial hypotension
62%
Distributive (septic)
Absent Signs of tissue hypoperfusion Present
Brain
Altered mental
Chronic
state
hypotension?
Syncope Circulatory
(if transient) shock 4% 2%
Distributive Obstructive
Tachycardia (nonseptic)
16% 16%
Skin Elevated Estimate cardiac Cardiogenic Hypovolemic
blood output or SvO2
Mottled,
clammy lactate
Normal Low
or high
Kidney
Oliguria CVP
Low High
Echocardiography In tamponade: pericardial

effusion, small right and
Normal cardiac Small cardiac left ventricles, dilated
Large ventricles and
chambers and (usually) chambers and normal inferior vena cava; in
poor contractility
preserved contractility or high contractility pulmonary embolism or
pneumothorax: dilated right
ventricle, small left ventricle
Distributive shock Hypovolemic shock Cardiogenic shock Obstructive shock
Distributive shock Hypovolemic shock Cardiogenic shock Obstructive shock
Loss of Obstruction
Vasodilatation plasma or
blood
volume
Ventricular Pericardial
failure tamponade
Figure 1. Initial Assessment of Shock States.

COLOR FIGURE
Shown is an algorithm for the initial assessment of a patient in shock (Panel A), relative frequencies of the main types of shock (Panel B),
and schematic representations of the four main types of shock (Panel C). The algorithm starts with the most common Draft 9presentation 10/10/13
Author Vincent
(i.e., arterial hypotension), but hypotension is sometimes minimal or absent. CVP denotes central venous pressure, Fig #
and
1
SvO2 mixed
venous oxygen saturation. Title
ME
DE Drazen
Artist Knoper
AUTHOR PLEASE NOTE:
1728 n engl j med 369;18

115 nejm.org october 31, 2013 Issue date
10/31/13
The New
use only. Medicine
rights reserved.
Source: The New England Journal of Medicine Critical Care Medicine
that the patient must receive ventilation with a vasopressor temporarily while fluid resuscita-
relatively large tidal volumes, have no spontane- tion is ongoing, with the aim of discontinuing it,
ous breathing effort (which usually requires the if possible, after hypovolemia has been corrected.
administration of sedatives or even muscle relax- Adrenergic agonists are the first-line vaso-
ants), and be free of major arrhythmia and right pressors because of their rapid onset of action,
ventricular dysfunction. A passive leg-raising testhigh potency, and short half-life, which allows
is an alternative method9 but requires a rapid- easy dose adjustment. Stimulation of each type
response device, since the effect is transient. of adrenergic receptor has potentially beneficial
Regardless of the test used, there remains a gray and harmful effects. For example, β-adrenergic
zone in which it is difficult to predict a patient’s
stimulation can increase blood flow but also in-
response to intravenous fluids. creases the risk of myocardial ischemia as a result
A fluid-challenge technique should be used to of increased heart rate and contractility. Hence, the
determine a patient’s actual response to fluids, use of isoproterenol, a pure β-adrenergic agent, is
while limiting the risks of adverse effects. A fluid
limited to the treatment of patients with severe
challenge incorporates four elements that should bradycardia. At the other extreme, α-adrenergic
be defined in advance.10 First, the type of fluid stimulation will increase vascular tone and blood
must be selected. Crystalloid solutions are the pressure but can also decrease cardiac output
first choice, because they are well tolerated and and impair tissue blood flow, especially in the
cheap. The use of albumin to correct severe hy- hepatosplanchnic region. For this reason, phenyl-
poalbuminemia may be reasonable in some pa- ephrine, an almost pure α-adrenergic agent, is
tients.11 (A detailed examination of the choice of rarely indicated.
resuscitation fluids was provided in a previous We consider norepinephrine to be the vaso-
article in this series12 and thus is not included in
pressor of first choice; it has predominantly
this review.) Second, the rate of fluid adminis- α-adrenergic properties, but its modest β-adrener-
tration must be defined. Fluids should be in- gic effects help to maintain cardiac output.
fused rapidly to induce a quick response but not Administration generally results in a clinically
so fast that an artificial stress response develops;
significant increase in mean arterial pressure,
typically, an infusion of 300 to 500 ml of fluid with little change in heart rate or cardiac output.
is administered during a period of 20 to 30 min- The usual dose is 0.1 to 2.0 µg per kilogram of
utes.13 Third, the objective of the fluid challengebody weight per minute.
must be defined. In shock, the objective is usu- Dopamine has predominantly β-adrenergic
ally an increase in systemic arterial pressure, effects at lower doses and α-adrenergic effects at
although it could also be a decrease in heart rate higher doses, but its effects are relatively weak.
or an increase in urine output. Finally, the safetyDopaminergic effects at very low doses (<3 µg per
limits must be defined. Pulmonary edema is the kilogram per minute, given intravenously) may
most serious complication of fluid infusion. Al- selectively dilate the hepatosplanchnic and renal
though it is not a perfect guideline, a limit in circulations, but controlled trials have not shown
central venous pressure of a few millimeters of a protective effect on renal function,14 and its
mercury above the baseline value is usually set to routine use for this purpose is no longer recom-
prevent fluid overload.13 mended. Dopaminergic stimulation may also
Stimulation of the patient and any other have undesired endocrine effects on the hypo-
change in therapy should be avoided during the thalamic–pituitary system, resulting in immuno-
test. Fluid challenges can be repeated as required suppression, primarily through a reduction in
but must be stopped rapidly in case of non- the release of prolactin.
response in order to avoid fluid overload. In a recent randomized, controlled, double-
blind trial, dopamine had no advantage over nor-
Vasoactive Agents epinephrine as the first-line vasopressor agent;
Vasopressors moreover, it induced more arrhythmias and was
If hypotension is severe or if it persists despite associated with an increased 28-day rate of
fluid administration, the use of vasopressors is death among patients with cardiogenic shock.4
indicated. It is acceptable practice to administer Administration of dopamine, as compared with

116 october 31, 2013 1729
The New
use only. Medicine
rights reserved.
norepinephrine, may also be associated with ties, dobutamine is less likely to induce tachycar-
higher rates of death among patients with septic dia than isoproterenol. An initial dose of just a
shock.15 Hence, we no longer recommend dopa- few micrograms per kilogram per minute may
mine for the treatment of patients with shock. substantially increase cardiac output. Intravenous
Epinephrine, which is a stronger agent, has doses in excess of 20 µg per kilogram per minute
predominantly β-adrenergic effects at low doses, usually provide little additional benefit. Dobuta-
with α-adrenergic effects becoming more clini- mine has limited effects on arterial pressure, al-
cally significant at higher doses. However, epi- though pressure may increase slightly in patients
nephrine administration can be associated with with myocardial dysfunction as the primary ab-
an increased rate of arrhythmia16,17 and a decrease normality or may decrease slightly in patients
in splanchnic blood flow16 and can increase blood with underlying hypovolemia. Instead of routine
lactate levels, probably by increasing cellular me- administration of a fixed dose of dobutamine to
tabolism.16,18 Prospective, randomized studies have increase oxygen delivery to supranormal, prede-
not shown any beneficial effects of epinephrine termined levels, the dose should be adjusted on
over norepinephrine in septic shock.17,18 We re- an individual basis to achieve adequate tissue
serve epinephrine as a second-line agent for se- perfusion. Dobutamine may improve capillary
vere cases.13 perfusion in patients with septic shock, indepen-
The use of other strong vasopressor agents as dent of its systemic effects.23
continuous infusions (e.g., angiotensin or meta- Phosphodiesterase type III inhibitors, such as
raminol) has largely been abandoned. Nonselec- milrinone and enoximone, combine inotropic and
tive inhibition of nitric oxide has not been shown vasodilating properties. By decreasing the me-
to be beneficial in patients with cardiogenic tabolism of cyclic AMP, these agents may rein-
shock19 and is detrimental in patients with sep- force the effects of dobutamine. They may also
tic shock.20 be useful when β-adrenergic receptors are down-
Vasopressin deficiency can develop in pa- regulated or in patients recently treated with
tients with very hyperkinetic forms of distribu- beta-blockers. However, phosphodiesterase type
tive shock, and the administration of low-dose III inhibitors may have unacceptable adverse ef-
vasopressin may result in substantial increases fects in patients with hypotension, and the long
in arterial pressure. In the Vasopressin and Sep- half-lives of these agents (4 to 6 hours) prevent
tic Shock Trial (VASST), investigators found that minute-to-minute adjustment. Hence, intermit-
the addition of low-dose vasopressin to norepi- tent, short-term infusions of small doses of
nephrine in the treatment of patients with septic phosphodiesterase III inhibitors may be prefer-
shock was safe21 and may have been associated able to a continuous infusion in shock states.
with a survival benefit for patients with forms of Levosimendan, a more expensive agent, acts
shock that were not severe and for those who primarily by binding to cardiac troponin C and
also received glucocorticoids.22 Vasopressin should increasing the calcium sensitivity of myocytes,
not be used at doses higher than 0.04 U per min- but it also acts as a vasodilator by opening ATP-
ute and should be administered only in patients sensitive potassium channels in vascular smooth
with a high level of cardiac output. muscle. However, this agent has a half-life of
Terlipressin, an analogue of vasopressin, has several days, which limits the practicality of its
a duration of action of several hours, as com- use in acute shock states.
pared with minutes for vasopressin. For this
reason, we do not believe it offers an advantage Vasodilators
over vasopressin in the ICU. Vasopressin deriva- By reducing ventricular afterload, vasodilating
tives with more selective V1-receptor activity are agents may increase cardiac output without in-
currently being studied. creasing myocardial demand for oxygen. The
major limitation of these drugs is the risk of de-
Inotropic Agents creasing arterial pressure to a level that compro-
We consider dobutamine to be the inotropic mises tissue perfusion. Nevertheless, in some
agent of choice for increasing cardiac output, re- patients, prudent use of nitrates and possibly
gardless of whether norepinephrine is also being other vasodilators may improve microvascular
given. With predominantly β-adrenergic proper- perfusion and cellular function.24
1730 n engl j med 369;18 117

The New England

For of Medicine
Copyright 29, 2014.
Medical use rights
reserved.
Mech a nic a l Supp or t A

Mechanical support with intraaortic balloon
counterpulsation (IABC) can reduce left ventricu-
lar afterload and increase coronary blood flow.
However, a recent randomized, controlled trial
showed no beneficial effect of IABC in patients
with cardiogenic shock,25 and its routine use in
cardiogenic shock is not currently recommended.
Venoarterial extracorporeal membrane oxygenation
(ECMO) may be used as a temporary lifesaving
measure in patients with reversible cardiogenic
shock or as a bridge to heart transplantation.26
B
G oa l s of Hemody na mic Supp or t
Arterial Pressure
The primary goal of resuscitation should be not
only to restore blood pressure but also to provide
adequate cellular metabolism, for which the cor-
rection of arterial hypotension is a prerequisite.
Restoring a mean systemic arterial pressure of
65 to 70 mm Hg is a good initial goal, but the
level should be adjusted to restore tissue perfu-
sion, assessed on the basis of mental status, skin
appearance, and urine output, as described above. Figure 2. Sidestream Dark-Field Images of Sublingual
In patients with oliguria, in particular, the effects Microcirculation in a Healthy Volunteer and a Patient
of a further increase in arterial pressure on urine with Septic Shock.
output should be assessed regularly, unless acute The microcirculation in the healthy volunteer is charac-
renal failure is already established. Conversely, a terized by dense capillaries that are consistently perfused
mean arterial pressure lower than 65 to 70 mm Hg (Panel A, arrows), whereas in the patient with septic
shock, the density of the capillaries is diminished, and
may be acceptable in a patient with acute bleeding many of the capillaries have stopped or intermittent
who has no major neurologic problems, with the flow (Panel B, arrows).
aim of limiting blood loss and associated coagu-
lopathy, until the bleeding is controlled.
is needed will vary among patients and in the
Cardiac Output and Oxygen Delivery same patient over time.
Since circulatory shock represents an imbalance Measurements of mixed venous oxygen satu-
between oxygen supply and oxygen requirements, ration (SvO2) may be helpful in assessing the
maintaining adequate oxygen delivery to the tis- adequacy of the balance between oxygen de-
sues is essential, but all the strategies to achieve mand and supply; SvO2 measurements are also
this goal have limitations. After correction of hy- very useful in the interpretation of cardiac out-
poxemia and severe anemia, cardiac output is the put.27 SvO2 is typically decreased in patients with
principal determinant of oxygen delivery, but the low-flow states or anemia but is normal or high
optimal cardiac output is difficult to define. Car- in those with distributive shock. Its surrogate,
diac output can be measured by means of various central venous oxygen saturation (ScvO2), which
techniques, each of which has its own benefits is measured in the superior vena cava by means
and drawbacks.6 Absolute measures of cardiac of a central venous catheter, reflects the oxygen
output are less important than monitoring trends saturation of the venous blood from the upper
in response to interventions such as a fluid chal- half of the body only. Under normal circumstances,
lenge. The targeting of a predefined cardiac out- ScvO2 is slightly less than SvO2, but in critically
put is not advisable, since the cardiac output that ill patients it is often greater. Rivers et al.28 found

118 october 31, 2013 1731
The New
only.ofNo
Medicine
Medical 2014. For
rights reserved.
Microcirculatory Variables
Salvage Optimization Stabilization De-escalation The development of handheld devices for orthog-
onal polarization spectral (OPS) imaging and its
Obtain a Provide Provide organ Wean from successor, sidestream dark-field (SDF) imaging,
is providing new means of directly visualizing
Phase Focus
minimal adequate support vasoactive

acceptable oxygen agents
blood pressure availability
the microcirculation and evaluating the effects of
interventions on microcirculatory flow in easily
Perform Optimize Minimize Achieve a accessible surfaces, such as the sublingual area.30
lifesaving cardiac output, complications negative
measures SvO2, lactate fluid balance
Microcirculatory changes, including decreased
capillary density, a reduced proportion of perfused
Figure 3. Four Phases in the Treatment of Shock. capillaries, and increased heterogeneity of blood
COLOR FIGURE
The salvage phase focuses on achieving a blood pressure and cardiac outflow, have been identified in various types of circu-
Draft 6 10/15/13
put compatible with immediate survival and performing
Author Vincent
lifesaving proce- latory shock (Fig. 2), and the persistence of these
dures to treat the underlying cause of shock. The Fig
optimization
# 3 phase focus- alterations is associated with worse outcomes.31
es on promoting cellular oxygen availability and monitoring
Title cardiac output, Near-infrared spectroscopy is a technique that
mixed venous oxygen saturation (SvO2), and lactate levels. The stabilization
ME uses near-infrared light to determine tissue oxy-
phase focuses on preventing organ dysfunction, even
DE after
Drazen hemodynamic
stability has been achieved. The de-escalation phase
Artist focuses
Knoper on weaning
gen saturation from the fractions of oxyhemo-
the patient from vasoactive agents and providing treatmentsAUTHOR PLEASEto help NOTE: globin and deoxyhemoglobin. Analysis of the
achieve a negative fluid balance. Please check carefully changes in tissue oxygen saturation during a
Issue date 10/31/13
brief episode of forearm ischemia can be used to
quantify microvascular dysfunction32; such altera-
that in patients presenting to the emergency tions are associated with worse outcomes.33 Vari-
department with septic shock, a treatment algo- ous therapeutic interventions have been shown
rithm targeting an ScvO2 of at least 70% during to have an effect on these microcirculatory vari-
the first 6 hours was associated with decreased ables, but whether therapy that is guided by
rates of death. The robustness of this finding is monitoring or targeting the microcirculation
currently being evaluated in three multicenter can improve outcomes requires further study
trials. (ClinicalTrials.gov numbers, NCT00975793 and cannot be recommended at this time.
and NCT00510835; and Current Controlled Trials
number, ISRCTN36307479).
Ther a peu t ic Pr ior i t ie s
a nd G oa l s
Blood Lactate Level
An increase in the blood lactate level reflects ab- There are essentially four phases in the treatment
normal cellular function. In low-flow states, the of shock, and therapeutic goals and monitoring
primary mechanism of hyperlactatemia is tissue need to be adapted to each phase (Fig. 3). In the
hypoxia with development of anaerobic metabo- first (salvage) phase, the goal of therapy is to
lism, but in distributive shock, the pathophysiol- achieve a minimum blood pressure and cardiac
ogy is more complex and may also involve increased output compatible with immediate survival. Mini-
glycolysis and inhibition of pyruvate dehydroge- mal monitoring is needed; in most cases, invasive
nase. In all cases, alterations in clearance can be monitoring can be restricted to arterial and cen-
due to impaired liver function. tral venous catheters. Lifesaving procedures (e.g.,
The value of serial lactate measurements in surgery for trauma, pericardial drainage, revascu-
the management of shock has been recognized larization for acute myocardial infarction, and anti-
for 30 years.29 Although changes in lactate take biotics for sepsis) are needed to treat the under-
place more slowly than changes in systemic arte- lying cause. In the second (optimization) phase, the
rial pressure or cardiac output, the blood lactate goal is to increase cellular oxygen availability, and
level should decrease over a period of hours with there is a narrow window of opportunity for inter-
effective therapy. In patients with shock and a ventions targeting hemodynamic status.28 Ade-
blood lactate level of more than 3 mmol per liter, quate hemodynamic resuscitation reduces inflam-
Jansen et al.24 found that targeting a decrease mation, mitochondrial dysfunction, and caspase
of at least 20% in the blood lactate level over a activation.34,35 Measurements of SvO2 and lactate
2-hour period seemed to be associated with re- levels may help guide therapy, and monitoring of
duced in-hospital mortality. cardiac output should be considered. In the third
1732 n engl j med119

369;18 nejm.org october 31, 2013
The New
Medical ForAll
Society. personal
reserved.
(stabilization) phase, the goal is to prevent organ sential so that aggressive management can be
dysfunction, even after hemodynamic stability has started. Appropriate treatment is based on a good
been achieved. Oxygen supply to the tissues is no understanding of the underlying pathophysiolog-
longer the key problem, and organ support be- ical mechanisms. Treatment should include cor-
comes more relevant. Finally, in the fourth (de- rection of the cause of shock and hemodynamic
escalation) phase, the goal is to wean the patient stabilization, primarily through fluid infusion
from vasoactive agents and promote spontaneous and administration of vasoactive agents. The pa-
polyuria or provoke fluid elimination through the tient’s response can be monitored by means of
use of diuretics or ultrafiltration to achieve a neg- careful clinical evaluation and blood lactate mea-
ative fluid balance. surements; microvascular evaluation may be fea-
sible in the future.
C onclusions
No potential conflict of interest relevant to this article was
reported.
Circulatory shock is associated with high mor- Disclosure forms provided by the authors are available with
bidity and mortality. Prompt identification is es- the full text of this article at NEJM.org.
References
1. Sakr Y, Reinhart K, Vincent JL, et al. tion fluid for patients with sepsis: a sys- tric oxide synthase inhibitor 546C88: ef-
Does dopamine administration in shock tematic review and meta-analysis. Crit fect on survival in patients with septic
influence outcome? Results of the Sepsis Care Med 2011;39:386-91. shock. Crit Care Med 2004;32:21-30.
Occurrence in Acutely Ill Patients (SOAP) 12. Myburgh JA, Mythen MG. Resuscita- 21. Russell JA, Walley KR, Singer J, et al.
Study. Crit Care Med 2006;34:589-97. tion fluids. N Engl J Med 2013;369:1243- Vasopressin versus norepinephrine infu-
2. Vincent JL, Ince C, Bakker J. Circula- 51. sion in patients with septic shock. N Engl
tory shock — an update: a tribute to Pro- 13. Dellinger RP, Levy MM, Rhodes A, et J Med 2008;358:877-87.
fessor Max Harry Weil. Crit Care 2012; al. Surviving Sepsis Campaign: interna- 22. Russell JA, Walley KR, Gordon AC, et
16:239. tional guidelines for management of severe al. Interaction of vasopressin infusion,
3. Weil MH, Shubin H. Proposed reclas- sepsis and septic shock: 2012. Crit Care corticosteroid treatment, and mortality of
sification of shock states with special ref- Med 2013;41:580-637. septic shock. Crit Care Med 2009;37:811-8.
erence to distributive defects. Adv Exp 14. Bellomo R, Chapman M, Finfer S, 23. De Backer D, Creteur J, Dubois MJ, et
Med Biol 1971;23:13-23. Hickling K, Myburgh J. Low-dose dopa- al. The effects of dobutamine on micro-
4. De Backer D, Biston P, Devriendt J, et mine in patients with early renal dysfunc- circulatory alterations in patients with
al. Comparison of dopamine and norepi- tion: a placebo-controlled randomised trial. septic shock are independent of its sys-
nephrine in the treatment of shock. N Engl Lancet 2000;356:2139-43. temic effects. Crit Care Med 2006;34:
J Med 2010;362:779-89. 15. De Backer D, Aldecoa C, Njimi H, Vin- 403-8.
5. Labovitz AJ, Noble VE, Bierig M, et al. cent JL. Dopamine versus norepinephrine 24. Jansen TC, van Bommel J, Schoon-
Focused cardiac ultrasound in the emer- in the treatment of septic shock: a meta- derbeek FJ, et al. Early lactate-guided
gent setting: a consensus statement of the analysis. Crit Care Med 2012;40:725-30. therapy in intensive care unit patients:
American Society of Echocardiography 16. Levy B, Perez P, Perny J, Thivilier C, a multicenter, open-label, randomized con-
and American College of Emergency Phy- Gerard A. Comparison of norepinephrine- trolled trial. Am J Respir Crit Care Med
sicians. J Am Soc Echocardiogr 2010;23: dobutamine to epinephrine for hemo- 2010;182:752-61.
1225-30. dynamics, lactate metabolism, and organ 25. Thiele H, Zeymer U, Neumann FJ, et
6. Vincent JL, Rhodes A, Perel A, et al. function variables in cardiogenic shock: al. Intraaortic balloon support for myo-
Clinical review: update on hemodynamic a prospective, randomized pilot study. cardial infarction with cardiogenic shock.
monitoring — a consensus of 16. Crit Care Crit Care Med 2011;39:450-5. N Engl J Med 2012;367:1287-96.
2011;15:229. 17. Annane D, Vignon P, Renault A, et al. 26. Combes A, Leprince P, Luyt CE, et al.
7. Weil MH, Shubin H. The “VIP” ap- Norepinephrine plus dobutamine versus Outcomes and long-term quality-of-life
proach to the bedside management of epinephrine alone for management of of patients supported by extracorporeal
shock. JAMA 1969;207:337-40. septic shock: a randomised trial. Lancet membrane oxygenation for refractory car-
8. Marik PE, Cavallazzi R, Vasu T, Hirani 2007;370:676-84. [Erratum, Lancet 2007; diogenic shock. Crit Care Med 2008;36:
A. Dynamic changes in arterial waveform 370:1034.] 1404-11.
derived variables and fluid responsiveness 18. Myburgh JA, Higgins A, Jovanovska A, 27. Vincent JL. Understanding cardiac
in mechanically ventilated patients: a sys- Lipman J, Ramakrishnan N, Santamaria J. output. Crit Care 2008;12:174.
tematic review of the literature. Crit Care A comparison of epinephrine and norepi- 28. Rivers E, Nguyen B, Havstad S, et al.
Med 2009;37:2642-7. nephrine in critically ill patients. Inten- Early goal-directed therapy in the treat-
9. Cavallaro F, Sandroni C, Marano C, et sive Care Med 2008;34:2226-34. ment of severe sepsis and septic shock.
al. Diagnostic accuracy of passive leg rais- 19. Alexander JH, Reynolds HR, Stebbins N Engl J Med 2001;345:1368-77.
ing for prediction of fluid responsiveness AL, et al. Effect of tilarginine acetate in 29. Vincent JL, Dufaye P, Berré J, Leeman
in adults: systematic review and meta- patients with acute myocardial infarction M, Degaute JP, Kahn RJ. Serial lactate de-
analysis of clinical studies. Intensive Care and cardiogenic shock: the TRIUMPH terminations during circulatory shock.
Med 2010;36:1475-83. randomized controlled trial. JAMA 2007; Crit Care Med 1983;11:449-51.
10. Vincent JL, Weil MH. Fluid challenge 297:1657-66. 30. De Backer D, Hollenberg S, Boerma C,
revisited. Crit Care Med 2006;34:1333-7. 20. López A, Lorente JA, Steingrub J, et al. et al. How to evaluate the microcircula-
11. Delaney AP, Dan A, McCaffrey J, Fin- Multiple-center, randomized, placebo- tion: report of a round table conference.
fer S. The role of albumin as a resuscita- controlled, double-blind study of the ni- Crit Care 2007;11:R101.
n engl j med 369;18 120

The New
Medical ForAll
Society. personal
reserved.
31. Sakr Y, Dubois MJ, De Backer D, O(2) saturation response. Intensive Care sepsis and septic shock. Crit Care Med
Creteur J, Vincent JL. Persistent micro- Med 2008;34:1600-7. 2007;35:2016-24.
circulatory alterations are associated 33. Creteur J, Carollo T, Soldati G, Buchele 35. Corrêa TD, Vuda M, Blaser AR, et al.
with organ failure and death in patients G, De Backer D, Vincent JL. The prognostic Effect of treatment delay on disease sever-
with septic shock. Crit Care Med 2004;32: value of muscle StO2 in septic patients. In- ity and need for resuscitation in porcine
1825-31. tensive Care Med 2007;33:1549-56. fecal peritonitis. Crit Care Med 2012;40:
32. Gómez H, Torres A, Polanco P, et al. 34. Rivers EP, Kruse JA, Jacobsen G, et al. 2841-9.
Use of non-invasive NIRS during a vascu- The influence of early hemodynamic opti- Copyright © 2013 Massachusetts Medical Society.
lar occlusion test to assess dynamic tissue mization on biomarker patterns of severe
1734 n engl j med 369;18 121

The New EnglandFor

Downloaded from nejm.org by MJ MEDAS on ©September
Copyright 29, 2014.
reserved.
c or r e sp ondence
Circulatory Shock
4. Samsky MD, Patel CB, DeWald TA, et al. Cardiohepatic inter-
To the Editor: In their review of circulatory actions in heart failure: an overview and clinical implications.
J Am Coll Cardiol 2013;61:2397-405.
shock, Vincent and De Backer (Oct. 31 issue)1 re-
port that hypoperfusion can be apparent through DOI: 10.1056/NEJMc1314999
the skin (cold extremities), the kidney (low uri-

nary output), and the brain (altered mental state). To the Editor: In their initial assessment of
We believe that the liver should also be included shock states, the authors do not reference the
in this list of key organs that are markers of obtaining of an electrocardiogram (ECG) in the
shock. In cardiogenic shock, in which hypoper- algorithm presented. The ECG is critical in evalu-
fusion is usually associated with increased cen- ating a patient with circulatory shock and may
tral venous pressure, hepatomegaly can be clin- immediately alter workup and management. It
ically apparent by palpation, and abdominal should be obtained before the echocardiogram is
discomfort due to stretching of the liver capsule obtained. The ECG may show evidence of an
may be a symptom. There is often an increase in acute ST-segment elevation myocardial infarction,
levels of conjugated bilirubin, alkaline phos- resulting in urgent transfer of the patient to the
phatase, and aminotransferases that mimics cardiac catheterization laboratory with possible
the increase observed in cholelithiasis.2 Altera- revascularization. Sinus tachycardia with the
tion of coagulation can be present with an in- S1Q3T3 pattern (prominence of the S wave in lead
crease in the international normalized ratio.3 I and a Q-wave and T-wave inversion in lead III)
This cardio–hepatic interaction in shock gener- on the ECG may imply acute pulmonary embo-
ally results in the so-called hypoxic hepatitis or lism.1 Low voltage with electrical alternans is in-
acute cardiogenic liver injury and has important dicative of pericardial tamponade.1 Loss of voltage
prognostic implications.3,4 Can the authors com- in the lateral lead could suggest left-sided pneu-
ment on hepatic dysfunction as an indicator of mothorax.1 Bradyarrhythmias or tachyarrhythmias
shock? (e.g., marked sinus bradycardia, complete heart
Enrico Ammirati, M.D. block with slow ventricular response, rapid atrial
Fabrizio Oliva, M.D. fibrillation, and ventricular tachycardia) can be
Maria Frigerio, M.D. important contributors to circulatory shock.
Ospedale Niguarda Ca’ Granda Ian Joffe, M.D.
Milan, Italy Our Lady of Lourdes Medical Center
enrico.ammirati@ospedaleniguarda.it Camden, NJ
ian_joffe@hotmail.com
ported. No potential conflict of interest relevant to this letter was re-
ported.
1. Vincent J-L, De Backer D. Circulatory shock. N Engl J Med
1. Chou T-C, Ramaiah LS. Electrocardiography in clinical prac-
2013;369:1726-34.
tice: adult and pediatric. 4th ed. Philadelphia: W.B. Saunders, 1996.
2. Nikolaou M, Parissis J, Yilmaz MB, et al. Liver function ab-
normalities, clinical profile, and outcome in acute decompen- DOI: 10.1056/NEJMc1314999
sated heart failure. Eur Heart J 2013;34:742-9.
3. Raurich JM, Llompart-Pou JA, Ferreruela M, et al. Hypoxic
hepatitis in critically ill patients: incidence, etiology and risk
factors for mortality. J Anesth 2011;25:50-6.
574 n engl j med 370;6 122

nejm.org february 6, 2014
correspondence
To the Editor: In their review article on circula- results. The monitoring of liver perfusion would
tory shock, the authors provide little focus on be of interest, but it is difficult to find a simple
evidence-based treatment of patients who have test. The monitoring of oxygen saturation in the
shock with a cardiogenic cause. Early revascular- hepatic vein is too invasive and has therefore been
ization of patients in cardiogenic shock is men- virtually abandoned. Techniques that measure the
tioned only briefly in a short sentence among clearance of indocyanine green are not performed
other causes of shock that must be corrected. routinely. Bilirubin levels increase relatively late
There are no references to the landmark SHOCK in the development of liver dysfunction, and in-
(Should We Emergently Revascularize Occluded creases in aminotransferase and lactate dehydro-
Coronaries for Cardiogenic Shock) trial,1,2 which genase levels are not specific to liver ischemia.
showed that a strategy of early revascularization Other enzymes are not good markers of altered
in patients with cardiogenic shock complicating liver perfusion. Clinical and biologic signs of liver
acute myocardial infarction by means of acute congestion that is associated with raised venous
percutaneous coronary intervention or coronary- pressure are not specific to cardiogenic and ob-
artery bypass surgery saves 13 lives for every 100 structive types of shock, because these signs can
patients treated. In a comprehensive review of also be observed in heart failure without shock.
circulatory shock, urgent revascularization Joffe is correct in underlining the value of the
should be highlighted as an evidence-based life- ECG, which should be obtained routinely. The
saving procedure in shock caused by an acute ECG can help to identify some causes of shock
myocardial infarction. (myocardial infarction being a straightforward
Geir Ø. Andersen, M.D., Ph.D. example), but its usefulness in diagnosing pul-
Oslo University Hospital, Ullevål monary embolism or pneumothorax is more
Oslo, Norway limited. As we are sure that readers can appreci-
g.o.andersen@medisin.uio.no
ate, there are many tests that could be done to
ported. establish the cause of shock; to review them all
would require a whole textbook. As noted by
1. Hochman JS, Sleeper LA, Webb JG, et al. Early revasculariza-
tion in acute myocardial infarction complicated by cardiogenic Andersen, space limitations prevented us from
shock. N Engl J Med 1999;341:625-34. discussing all aspects of treatment according to
2. Hochman JS, Sleeper LA, White HD, et al. One-year survival the underlying causes of shock.
following early revascularization for cardiogenic shock. JAMA
2001;285:190-2. Jean-Louis Vincent, M.D., Ph.D.
Daniel De Backer, M.D., Ph.D.
DOI: 10.1056/NEJMc1314999
Erasme University Hospital
Brussels, Belgium
The authors reply: We agree with Ammirati jlvincen@ulb.ac.be
and colleagues that it would be nice to have some Since publication of their article, the authors report no fur-
assessment of gastrointestinal perfusion. Gastric
DOI: 10.1056/NEJMc1314999
tonometry has been tried but was abandoned be-
cause of the many difficulties in interpretation of

nejm.org february 6, 2014 575

What monitoring strategies would you use to guide hemodynamic support for this patient?
A. An echocardiographic assessment of cardiac function.

B. A pulmonary-artery catheterization for complete, continuous monitoring.
C. Measurement of variation in pulse pressure during mechanical ventilation.
D. Measurement of central venous oxygen saturation with a goal of more than 70%.

The patient who is described in the vignette has circulatory shock, and the clinical scenario suggests that sepsis is
the predominant cause. However, the patient is elderly, with multiple risk factors for vascular disease and previous
heavy alcohol intake, so the shock state may have several contributory causes. Sepsis causes vasodilatation, hypo-
volemia owing to capillary leak, and myocardial depression. In addition, he may have a cardiogenic component
from undiagnosed myocardial ischemia or alcoholic cardiomyopathy. Focused echocardiography should be per-
formed as soon as possible to assess the left and right ventricular size and function, respiratory variations in vena
cava dimensions, and the aortic velocity–time integral as a measure of stroke volume. Although the optimal cardiac
output in patients with septic shock is not known and may differ both among patients and in the same patient
over time, the restoration of cellular oxygen delivery and reversal of shock are predicated on achieving an adequate
cardiac output. Cardiac output can be measured or monitored with the use of a variety of invasive and noninva-
sive techniques, and although use of a pulmonary-artery catheter is the standard technique, it has not been shown
to improve the outcome. In a patient who is still under anesthesia, the variation in pulse pressure throughout the
respiratory cycle can indicate that the patient is likely to have a response to intravenous fluids. However, it is im-
portant that there is no breathing effort, no major arrhythmia, and a relatively large tidal volume. In one study
involving patients with septic shock in an emergency department,1 a resuscitation strategy targeting a central
venous oxygen saturation of at least 70% (a surrogate for cardiac output monitoring) during the initial resuscita-
tion resulted in a significant reduction in mortality. The robustness of this finding is now being tested in three
multicenter trials. Monitoring of the blood lactate level may be an equally effective way to ensure an adequate
cardiac output.2
References
1. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed therapy in the treatment of severe sepsis and septic
shock. N Engl J Med 2001;345(19):1368–1377.
2. Jones AE, Shapiro NI, Trzeciak S, et al. Lactate clearance vs central venous oxygen saturation as goals of early
sepsis therapy: a randomized clinical trial. JAMA 2010;303(8):739–746.
124
new england
The
journal of medicine
ORIGINAL ARTICLE
established in 1812 april 24, 2014 vol. 370 no. 17
High versus Low Blood-Pressure Target in Patients with Septic Shock

Pierre Asfar, M.D., Ph.D., Ferhat Meziani, M.D., Ph.D., Jean-François Hamel, M.D., Fabien Grelon, M.D.,
Bruno Megarbane, M.D., Ph.D., Nadia Anguel, M.D., Jean-Paul Mira, M.D., Ph.D., Pierre-François Dequin, M.D., Ph.D.,
Soizic Gergaud, M.D., Nicolas Weiss, M.D., Ph.D., François Legay, M.D., Yves Le Tulzo, M.D., Ph.D.,
Marie Conrad, M.D., René Robert, M.D., Ph.D., Frédéric Gonzalez, M.D., Christophe Guitton, M.D., Ph.D.,
Fabienne Tamion, M.D., Ph.D., Jean-Marie Tonnelier, M.D., Pierre Guezennec, M.D., Thierry Van Der Linden, M.D.,
Antoine Vieillard-Baron, M.D., Ph.D., Eric Mariotte, M.D., Gaël Pradel, M.D., Olivier Lesieur, M.D.,
Jean-Damien Ricard, M.D., Ph.D., Fabien Hervé, M.D., Damien du Cheyron, M.D., Ph.D., Claude Guerin, M.D., Ph.D.,
Alain Mercat, M.D., Ph.D., Jean-Louis Teboul, M.D., Ph.D., and Peter Radermacher, M.D., Ph.D.,
for the SEPSISPAM Investigators*
A BS T R AC T
Background
The Surviving Sepsis Campaign recommends targeting a mean arterial pressure of The authors’ affiliations are listed in the
at least 65 mm Hg during initial resuscitation of patients with septic shock. Appendix. Address reprint requests to
Dr. Asfar at the Department of Medical
However, whether this blood-pressure target is more or less effective than a higher Intensive Care and Hyperbaric Medicine,
target is unknown. University Hospital of Angers, 4 rue Larrey,
F-49933 Angers CEDEX 9, France, or at
Methods piasfar@chu-angers.fr.
In a multicenter, open-label trial, we randomly assigned 776 patients with septic * Additional investigators in the Sepsis and
shock to undergo resuscitation with a mean arterial pressure target of either 80 to Mean Arterial Pressure (SEPSISPAM) trial
are listed in the Supplementary Appen-
85 mm Hg (high-target group) or 65 to 70 mm Hg (low-target group). The primary dix, available at NEJM.org.
end point was mortality at day 28.
Results 2014, at NEJM.org.
At 28 days, there was no significant between-group difference in mortality, with N Engl J Med 2014;370:1583-93.
deaths reported in 142 of 388 patients in the high-target group (36.6%) and 132 of DOI: 10.1056/NEJMoa1312173
388 patients in the low-target group (34.0%) (hazard ratio in the high-target group, Copyright © 2014 Massachusetts Medical Society.
1.07; 95% confidence interval [CI], 0.84 to 1.38; P = 0.57). There was also no sig-
nificant difference in mortality at 90 days, with 170 deaths (43.8%) and 164 deaths
(42.3%), respectively (hazard ratio, 1.04; 95% CI, 0.83 to 1.30; P = 0.74). The occur-
rence of serious adverse events did not differ significantly between the two groups
(74 events [19.1%] and 69 events [17.8%], respectively; P = 0.64). However, the inci-
dence of newly diagnosed atrial fibrillation was higher in the high-target group
than in the low-target group. Among patients with chronic hypertension, those in
the high-target group required less renal-replacement therapy than did those in the
low-target group, but such therapy was not associated with a difference in mortality.
Conclusions
Targeting a mean arterial pressure of 80 to 85 mm Hg, as compared with 65 to
70 mm Hg, in patients with septic shock undergoing resuscitation did not result in
significant differences in mortality at either 28 or 90 days. (Funded by the French
Ministry of Health; SEPSISPAM ClinicalTrials.gov number, NCT01149278.)
n engl j med 370;17 nejm.org april 24, 2014 1583
125
S
eptic shock is characterized by ar- Me thods
terial hypotension despite adequate fluid
resuscitation. The guidelines of the Surviv- Study Design
ing Sepsis Campaign1 recommended initial re- From March 2010 through December 2011, we
suscitation with vasopressors to reverse hypoten- enrolled patients at 29 centers in France. The
sion, with a mean arterial pressure target of at study was approved for all centers by the ethics
least 65 mm Hg (grade 1C, indicating a strong committee at the Angers University Hospital.
recommendation with a low level of evidence). Written informed consent was obtained from all
This recommendation is based on the findings of patients, their next of kin, or another surrogate
small studies, which showed no significant dif- decision maker, as appropriate. If patients were
ferences in lactate levels or regional blood flow unable to provide informed consent and neither
when the mean arterial pressure was elevated to their next of kin nor other designated person was
more than 65 mm Hg in patients with septic available, a procedure for inclusion in the study
shock.2,3 in emergency situations was applied. A definitive
However, as emphasized by the Surviving post hoc consent form was ultimately obtained
Sepsis Campaign guidelines, for patients with from patients who survived but had been initially
atherosclerosis or previous hypertension, a higher treated on the basis of the emergency consent.
blood-pressure target may be better. According- Randomization was performed with the use
ly, values for mean arterial pressure exceeding of a computer-generated assignment sequence in
65 mm Hg are frequently observed, as confirmed a centralized, blinded fashion and was stratified
by data from large, prospective, randomized, according to whether patients had chronic hyper-
controlled trials that focused on resuscitation of tension (i.e., had been receiving antihypertensive
patients with septic shock, which showed that treatment or had a history of arterial hyperten-
patients had mean arterial pressures in the range sion). Given the pragmatic character of the trial,
of 75 to 95 mm Hg 24 hours after inclusion.4-8 it was impossible to obtain details on the pa-
Moreover, a large, retrospective study showed that tients’ adherence to the antihypertensive drug
a mean arterial pressure of more than 75 mm Hg regimen or the adequacy of the antihypertensive
may be required to maintain kidney function.9 treatment during the inclusion time window.
The notion that a higher blood pressure can be Patients, research staff members, and members
useful was confirmed in a small, prospective, of the safety and writing committees were un-
observational study.10 Finally, a study of physio- aware of the study-group assignments.
logical mechanisms of chronic arterial hyperten-
sion showed that such hypertension causes a Study Oversight
rightward shift in cerebral pressure-flow auto- The data and safety monitoring committee over-
regulation, which might justify targeting a high- saw the trial conduct and the safety of the pa-
er mean arterial pressure.11 tients, with interim analyses performed after the
Since the selection of effective blood-pressure inclusion of 200, 400, and 600 patients. Data
targets is still controversial, we conducted a multi- were collected by the investigators and analyzed
center, randomized, stratified, open-label trial by the data-management committee. The steering
involving patients with septic shock to deter- committee vouches for the accuracy of the data,
mine whether targeting a mean arterial pressure the completeness of the analysis, and the fidelity
of 80 to 85 mm Hg would decrease 28-day mor- of the study to the protocol, which is available
tality, as compared with targeting a mean arte- with the full text of this article at NEJM.org.
rial pressure of 65 to 70 mm Hg. We also postu- Members of the steering committee made the de-
lated that the beneficial effects of a higher target cision to submit the manuscript for publication.
would be more pronounced among patients with The writing committee (the first author and the
chronic hypertension. Therefore, at randomiza- last three authors) had full access to all the data
tion, patients were stratified according to wheth- and collaborated with all the investigators in the
er they had a history of chronic hypertension. writing of the manuscript. All the drugs used in
1584 n engl j med 370;17 126

High vs. Low Blood-Pressure Target in Septic Shock
the study were purchased from the manufacturers, antibiotics unless judged necessary by the attend-
which had no role in the study. ing physician. Any use of the above-mentioned
drugs after study entrance was recorded.
Study Patients Renal-replacement therapy was initiated if at
Patients older than 18 years of age were enrolled least one of the following criteria was present:
if they had septic shock that was refractory to anuria, hyperkalemia with electrocardiographic
fluid resuscitation, if they required vasopressors changes, pure metabolic acidosis with a pH of
(norepinephrine or epinephrine) at a minimum less than 7.2, or a blood urea nitrogen level of
infusion rate of 0.1 μg per kilogram per minute, more than 84 mg per deciliter (30 mmol per li-
and if they were evaluated within 6 hours after ter) or a creatinine level of more than 5.65 mg
the initiation of vasopressors. Refractoriness to per deciliter (499 μmol per liter). Administration
fluid resuscitation was defined as a lack of re- of sedative and analgesic drugs or muscle relaxants
sponse to the administration of 30 ml of normal was left to the discretion of the clinician; doses
saline per kilogram of body weight or of colloids were reassessed at least daily to achieve values
or was determined according to a clinician’s as- ranging from −3 to 0 on the Richmond Agitation–
sessment of inadequate hemodynamic results on Sedation Scale (which ranges from −5 to 4, with
the basis of values obtained during right-heart lower scores indicating deeper sedation, 0 indi-
catheterization, pulse-pressure measurement, cating a calm and responsive patient, and higher
stroke-volume measurement, or echocardiogra- scores indicating increasing agitation); all doses of
phy (although study investigators did not record sedative and analgesic drugs were recorded daily.
the values for these variables). Septic shock was After enrollment, patients were assigned to
defined by the presence of two or more diagnos- vasopressor treatment that was adjusted to
tic criteria of the systemic inflammatory re- maintain a mean arterial pressure of 80 to 85
sponse syndrome, proven or suspected infection, mm Hg (high-target group) or 65 to 70 mm Hg
and sudden dysfunction of at least one organ.12 (low-target group). The target mean arterial
Exclusion criteria were legal protection (i.e., in- pressure was to be maintained for a maximum
competence to provide consent and no guardian of 5 days or until the patient was weaned from
or incarceration), no affiliation with the French vasopressor support; after that, the target pres-
health care system, pregnancy, recent participa- sure was determined by the attending physician.
tion in another biomedical study or another in- For patients in whom the assigned target pres-
terventional study with mortality as the primary sure was not reached despite the administration
end point, or an investigator’s decision not to of increasing doses of vasopressors, group as-
resuscitate. signments were not modified, and data analysis
was conducted on an intention-to-treat basis.
Study Treatments (The vasopressor-weaning strategy is described
Fluid resuscitation was performed as recom- in the Supplementary Appendix, available at
mended by the French intensive care societies,13 NEJM.org.)
with norepinephrine administered as a first-line In the high-target group, a reduction in vaso-
vasopressor, except at one center, in which epi- pressor doses to maintain a mean arterial pres-
nephrine was used. The use of activated protein C sure of 65 to 70 mm Hg was recommended if
and hydrocortisone was left to the discretion of any of the prespecified serious adverse events
the attending physician, and the following treat- that were potentially related to an increased rate
ments were prohibited: the use of diuretics, ex- of vasopressor infusion occurred. These events
cept for compelling indications, such as hypox- were as follows: clinically relevant bleeding (i.e.,
emia attributed to symptomatic sodium and transfusion requirements of at least 2 units of
water overload or life-threatening hyperkalemia; packed red cells), myocardial infarction (defined
the use of nonsteroidal antiinflammatory drugs; as typical electrocardiographic changes, with a
the use of iodinated contrast agents unless nec- concomitant increase in troponin, and segmen-
essary for imaging; and the use of nephrotoxic tal echocardiographic hypokinesia or akinesia,
n engl j med 370;17 127

with the infarction confirmed, when possible, size in one group was less than 30. We used
by means of coronary angiography), major ven- chi-square tests or Fisher’s exact test to com-
tricular arrhythmia, poorly tolerated supraven- pare qualitative variables, as appropriate. All
tricular arrhythmia, mesenteric ischemia, and comparisons were also performed with the
distal-limb ischemia. Data analysis for serious use of the entire sample with stratification (as
adverse events was performed for all patients on prespecified) according to the presence or ab-
an intention-to-treat basis. sence of chronic hypertension. Multiple logistic-
regression analyses were conducted in the
Study Outcomes intention-to-treat population to adjust for known
The primary outcome was death from any cause risk factors for acute kidney injury, such as
by 28 days after inclusion. Secondary outcomes chronic renal failure or the use of diuretics,
were 90-day mortality, days alive and free from vancomycin, aminoglycosides, iodine-containing
organ dysfunction by day 28, and the length of contrast material, or long-term use of nonsteroidal
stay in the intensive care unit (ICU) and hospital. antiinflammatory drugs.
Survival by day 28 without organ support was Interim analyses were performed for the pri-
defined as the number of days without catechol- mary outcome of 28-day mortality, according to
amine infusion, mechanical ventilation, or renal- the Haybittle–Peto method. Statistical signifi-
replacement therapy.14 Serious adverse events cance was indicated by a P value of 0.001 in the
were recorded and classified as cardiac, ische- three interim analyses and a two-sided P value of
mic, or other. 0.0492 in the final analysis. To detect a possible
interaction between group and stratum covari-
Statistical Analysis ates, logistic-regression analyses were performed
We determined that the enrollment of 800 pa- for dichotomous dependent variables, whereas
tients would provide a power of 80% to show an analysis-of-variance models were used for con-
absolute between-group difference of 10 percent- tinuous dependent variables. All analyses were
age points in the primary outcome, at a two-sid- performed with the use of Stata software, ver-
ed alpha level of 0.05, assuming a rate of death of sion 12.1.
45%. We decided not to compensate for dropouts
caused by the withdrawal of consent. All analy- R e sult s
ses were performed by the study statistician be-
fore the randomization code was broken, in line Study Population
with both the International Conference on We enrolled 776 patients and followed them for
Harmonization–Good Clinical Practice guide- 90 days; we conducted the analyses according to
lines and our statistical analysis plan (which is the group to which the patients were randomly
available in the protocol). assigned (Fig. 1). Baseline characteristics were
The analyses were performed in the inten- similar in the two groups (Table 1, and Table S1
tion-to-treat population, which was defined as and Fig. S1 in the Supplementary Appendix).
all patients who had undergone randomization Overall, 167 of 388 patients (43.0%) in the high-
except for those who did not provide consent for target group and 173 of 388 (44.6%) in the low-
the use of their data. We used Cox regression target group had a history of chronic hyperten-
models to calculate between-group differences sion. All the enrolled patients were critically ill,
in mortality at 28 days and 90 days. We analyzed as defined by the Simplified Acute Physiology
Schoenfeld residuals to test the assumption of Score (SAPS) II and Sequential Organ Failure
proportional hazards and used the Kaplan– Assessment (SOFA) score, serum lactate levels,
Meier method to calculate survival curves. We and norepinephrine infusion rates at study entry.
expressed quantitative variables as means (±SD) During the 5 protocol-specified days, the mean
and used t-tests to compare them when the arterial pressures in the low-target group were
sample size in each group was 30 or more (in significantly lower than those in the high-target
accordance with the central limit theorem) and group, yet they exceeded the target values of 65 to
the Wilcoxon rank-sum test when the sample 70 mm Hg (Fig. 2).
1586 n engl j med 370;17 128

4098 Patients were assessed for eligibility
3298 Were not eligible

1682 Did not meet inclusion criteria
858 Had shock lasting >6 hr
353 Had insufficient fluid challenge
269 Had catecholamine dose <0.1 µg/kg/min
202 Declined to participate
369 Had limitation of therapy
258 Were enrolled in another study
175 Were close to death
74 Were under guardianship
28 Were not affiliated with health care system
18 Were <18 yr
1 Was pregnant
298 Had unknown reasons
97 Did not have attending physician register
for the protocol
90 Were overlooked by investigator
90 Were in nonseptic or mixed shock
42 Had cardiac arrest before inclusion
28 Were weaned from vasopressor support
before inclusion
28 Did not have access to computer server for
randomization
20 Met unreported exclusion criteria
800 Patients were eligible
2 Signed consent but did not

undergo randomization
396 Were assigned to low-target group 402 Were assigned to high-target group
14 Were excluded
8 Were excluded 8 Withdrew consent
6 Withdrew consent 5 Were under guardianship
2 Were under guardianship 1 Had electronic tag for
tracking prisoners
388 Were included in 90-day follow-up 388 Were included in 90-day follow-up
and analysis and analysis
Figure 1. Screening, Randomization, and Follow-up of Study Patients.
Vasopressor Use and Fluid Balance group than in the low-target group (Table 2, and
The infusion rates of vasopressors were signifi- Table S2A in the Supplementary Appendix). A to-
cantly higher, and the duration of vasopressor tal of 64 patients (16.5%) in the high-target
treatment significantly longer, in the high-target group and 40 patients (10.3%) in the low-target
n engl j med 370;17 nejm.org april 24, 2014 1587

129
group (P = 0.01) did not reach targets for mean Primary Outcome
arterial pressure because of the attending physi- At 28 days, there was no significant between-
cian’s decision to limit the vasopressor infusion group difference in the rate of death, with deaths
rates. In 14 patients (3.6%) in the high-target reported in 142 of 388 patients (36.6%) in the
group, vasopressor infusion rates were adjusted high-target group and 132 of 388 patients
downward to maintain a mean arterial pressure (34.0%) in the low-target group (hazard ratio in
of 65 to 70 mm Hg because of adverse effects. the high-target group, 1.07; 95% confidence in-
Values for total fluid administration and total terval [CI], 0.84 to 1.38; P = 0.57). There was also
urine output during the 5 days specified in the no significant between-group difference in mor-
protocol were similar in the two study groups tality at 90 days, with 170 deaths (43.7%) and 164
(Table S2B in the Supplementary Appendix). deaths (42.3%), in the two groups, respectively
Low-Target Group High-Target Group

Age — yr 65±15 65±13
Male sex — no. (%) 250 (64.4) 267 (68.8)
Simplified Acute Physiology Score II† 57.2±16.2 56.1±15.5
Sequential Organ Failure Assessment score‡ 10.8±3.1 10.7±3.1
Recent surgical history — no. (%)
Elective 5 (1.3) 2 (0.5)
Emergency 55 (14.2) 47 (12.1)
Preexisting conditions — no. (%)
Ischemic heart disease 39 (10.1) 39 (10.1)
Chronic heart failure 53 (13.7) 59 (15.2)
Chronic obstructive pulmonary disease 47 (12.1) 58 (14.9)
Chronic kidney disease 30 (7.7) 20 (5.2)
Chronic kidney disease requiring long-term dialysis 12 (3.1) 5 (1.3)
Liver cirrhosis 28 (7.2) 29 (7.5)
Diabetes 90 (23.2) 75 (19.3)
Cancer or autoimmune disease 135 (34.8) 142 (36.6)
Chronic arterial hypertension 173 (44.6) 167 (43.0)
Source of infection — no. (%)
Lung 200 (51.5) 202 (52.1)
Abdomen 67 (17.3) 65 (16.8)
Urinary tract 44 (11.3) 44 (11.3)
Other§ 73 (18.8) 72 (18.6)
Community-acquired infection — no. (%) 253 (65.2) 262 (67.5)
Hemodynamic and biochemical variables
Mean arterial pressure — mm Hg 73±14 74±15
Heart rate — beats/min 103±24 104±27
Arterial pH 7.30±0.13 7.30±0.12
Serum lactate level — mmol/liter 3.7±3.7 3.3±3.2
Fluid therapy before inclusion — ml 2946±1360 2973±1331
1588 n engl j med 370;17 130


Vasoactive drug infusions at randomization — no. (%)
Norepinephrine 368 (94.8) 373 (96.1)
Epinephrine 20 (5.2) 15 (3.9)
Dobutamine 21 (5.4) 16 (4.1)
Median vasopressor dose at randomization — µg/kg/min (IQR)
Norepinephrine 0.35 (0.20–0.61) 0.40 (0.20–0.62)
Epinephrine 0.23 (0.17–0.32) 0.22 (0.13–0.64)
Mechanical ventilation — no. (%) 286 (73.7) 308 (79.4)
Pao2:Fio2 ratio — mm Hg 198±120 199±126
Acute kidney injury — no./total no. (%)¶ 188/386 (48.7) 173/384 (45.1)
Serum creatinine at inclusion — mg/dl 1.96±1.39 1.93±1.47
* Plus–minus values are means ±SD. The target mean arterial pressure was 80 to 85 mm Hg in the high-target group
and 65 to 70 mm Hg in the low-target group. None of the differences between the two groups were significant at base-
line. To convert the values for creatinine to micromoles per liter, multiply by 88.4. Fio2 denotes fraction of inspired oxy-
gen, IQR interquartile range, and Pao2 partial pressure of oxygen in arterial blood.
† The Simplified Acute Physiology Score II is based on 17 variables; scores range from 0 to 163, with higher scores indi-
cating more severe disease.
‡ The score on the Sequential Organ Failure Assessment (SOFA) includes subscores ranging from 0 to 4 for each of five
components (circulation, lungs, liver, kidneys, and coagulation). Aggregated scores range from 0 to 20, with higher
scores indicating more severe organ failure.
§ Other sources of infection included blood, soft tissue, skin, central nervous system, bones and joints, cardiac system,
reproductive organs, and unknown sources.
¶ Acute kidney injury was defined as a renal SOFA score of 2 or more (plasma creatinine level, >1.9 mg per deciliter [168 µmol
per liter]; or urinary output, <500 ml per day).
(hazard ratio, 1.04; 95% CI, 0.83 to 1.30; P = 0.74) the total number of cardiac adverse events did
(Table 2 and Fig. 3). not differ between the groups, the incidence of
In addition, there were no significant differ- newly diagnosed atrial fibrillation was signifi-
ences in the secondary outcomes: need for me- cantly higher in the high-target group, with
chanical ventilation, length of stay in the ICU events reported in 26 patients (6.7%) in the high-
and hospital, and the SOFA score by day 7 (Table target group and 11 patients (2.8%) in the low-
2, and Tables S2C and S2D in the Supplementary target group (P = 0.02). The frequencies of is-
Appendix). However, in patients with chronic chemic events and bleeding complications were
arterial hypertension, there was a significant similar in the two study groups.
interaction between study group and hyperten-
sion stratum with respect to the doubling of the Discussion
blood creatinine level (P = 0.009) and with re-
spect to the need for renal-replacement therapy In this multicenter, randomized, open-label
(P = 0.04). Multivariate logistic-regression analy- trial, we compared the strategy of targeting a
sis indicated that none of the potentially neph- high mean arterial pressure (80 to 85 mm Hg)
rotoxic therapies influenced this result. with the strategy of targeting a low pressure (65
to 70 mm Hg) in patients with septic shock. The
Adverse Events high-target group received significantly higher
There was no significant difference between the doses of vasopressor catecholamines over a signi-
two study groups in the overall incidence of seri- ficantly longer time period, but we found no sig-
ous adverse events (P = 0.64) (Table 2, and Table nificant difference in 28-day mortality. There was
S2E in the Supplementary Appendix). Although no significant between-group difference in early

131 april 24, 2014 1589
trial was designed. The absolute reduction of 10

90 percentage points in mortality was chosen in
High-target group our study because the trials that were available
Mean Arterial Pressure (mm Hg)
in the literature when the protocol was designed

85
in 2008 had tested the hypothesis of absolute
reductions of 20 percentage points,5 15 percent-
80 age points,4 and 10 percentage points8 in rates
of death. Two other trials that were published
75 after we started recruiting patients tested the
hypothesis of an absolute mortality reduction of
Low-target group 7 percentage points7,15 and 10 percentage points.16
70
Hence, the anticipated risk reduction in our
0 1 2 3 4 5
study was close to the risk reductions tested in
Days previous studies. However, our observed rate of
death at 28 days was lower than the rate in some
Figure 2. Mean Arterial Pressure during the 5-Day Study Period.
other studies, although it was in line with the
Mean arterial pressures were significantly lower in the low-target group than
in the high-target group during the 5 protocol-specified days (P = 0.02 by rate in more recent trials, in which death rates
repeated-measures regression analysis), although the values exceeded the ranging from 25 to 57% were reported.7,15
target values of 80 to 85 mm Hg in the high-target group and 65 to 70 mm Hg Nevertheless, the lower-than-expected rate of
in the low-target group. The I bars represent 95% confidence intervals. death led to an underpowered study. Therefore,
we may not have detected differences in the in-
cidence of some adverse events, especially rare
fluid balance, and the fluid balance was lower events such as myocardial infarction.
than those reported previously,7,8 possibly be- Septic shock is a major risk factor for atrial
cause our population of patients differed from fibrillation,17 and in our study, atrial fibrillation
those in previous studies or because of more re- was significantly more common in the high-
strictive protocols for fluid administration in target group than in the low-target group. This
France. In addition, there were no significant adverse effect might be related to the signifi-
between-group differences in the overall rates of cantly higher doses of catecholamine and the
organ dysfunction or death at 90 days. However, longer duration of catecholamine infusions in
in patients with a history of chronic arterial hy- the high-target group. However, given the small
pertension, targeting a mean arterial pressure of number of episodes of atrial fibrillation, other
80 to 85 mm Hg reduced both the incidence of a confounding factors cannot be ruled out. The
doubling of the blood creatinine level and the association between atrial fibrillation and septic
rate of renal-replacement therapy. There was no shock should be considered only as a hypothe-
significant between-group difference in the over- sis-generating concept for future trials.
all rate of serious adverse events, but patients in At randomization, patients were stratified ac-
the high-target group had significantly more cording of the presence or absence of chronic
episodes of atrial fibrillation. hypertension. More than 40% of the patients
No differences in the primary and secondary reported having a history of chronic hyperten-
outcomes were observed between the two sion, which is in line with rates in previous stud-
groups. Our study was prospectively powered to ies.18 Among patients with chronic hyperten-
detect an absolute difference of 10 percentage sion, a rightward shift of the curve for organ
points in the rate of death on the basis of an pressure-flow autoregulation is expected, which
expected rate of 45% in the low-target group, at means that an increased mean arterial pressure
an alpha level of 0.05 and a beta level of 0.20, could hypothetically result in improved organ
with the use of a two-tailed test. The expected perfusion11 and, eventually, in improved survival
overall death rate in our study was consistent rates. No significant differences in adverse effects
with the rates among patients with septic shock between patients with chronic hypertension and
that were reported in previous multicenter trials those without chronic hypertension were evi-
(37%,5 39%,8 47%,4 and 49%6) at the time the dent. The results in the subgroup with chronic
1590 n engl j med132

370;17 nejm.org april 24, 2014
Table 2. Clinical Results, Primary and Secondary Outcomes, and Serious Adverse Events.

Variable (N = 388) (N = 388) P Value
Cumulative fluid intake from day 1 to day 5 — liters 10.0 (5.8–14.0) 10.5 (5.5–14.0) 0.89
Cumulative urine output from day 1 to day 5 — liters 6.7 (2.9–10.7) 6.9 (2.4–10.7) 0.87
Cumulative fluid balance from day 1 to day 5 — liters 2.8 (0.0–6.2) 2.4 (0.0–6.0) 0.74
Median dose of norepinephrine (IQR) — µg/kg/min
Day 1 0.45 (0.17–1.21) 0.58 (0.26–1.80) <0.001
Day 2 0.16 (0.03–0.48) 0.38 (0.14–0.90) <0.001
Day 3 0.02 (0.00–0.16) 0.14 (0.01–0.50) <0.001
Day 4 0.00 (0.00–0.05) 0.03 (0.00–0.22) <0.001
Day 5 0.00 (0.00–0.03) 0.01 (0.00–0.15) <0.001
Duration of catecholamine infusion — days 3.7±3.2 4.7±3.7 <0.001
Primary outcome: death at day 28 — no. (%)* 132 (34.0) 142 (36.6) 0.57
Secondary outcomes — no./total no. (%)
Death at day 90† 164 (42.3) 170 (43.8) 0.74
Survival at day 28 without organ support‡ 241 (62.1) 235 (60.6) 0.66
Doubling of plasma creatinine 161 (41.5) 150 (38.7) 0.42
No chronic hypertension 71/215 (33.0) 85/221 (38.5) 0.32
Chronic hypertension 90/173 (52.0) 65/167 (38.9) 0.02
Renal-replacement therapy from day 1 to day 7 139 (35.8) 130 (33.5) 0.50
No chronic hypertension 66/215 (30.7) 77/221 (34.8) 0.36
Chronic hypertension 73/173 (42.2) 53/167 (31.7) 0.046
Serious adverse events — no. (%)
Any 69 (17.8) 74 (19.1) 0.64
Acute myocardial infarction§ 2 (0.5) 7 (1.8) 0.18
Atrial fibrillation 11 (2.8) 26 (6.7) 0.02
Ventricular fibrillation or tachycardia 15 (3.9) 22 (5.7) 0.24
Digital ischemia 9 (2.3) 10 (2.6) 0.82
Mesenteric ischemia 9 (2.3) 9 (2.3) 1.00
Bleeding 42 (10.8) 31 (8.0) 0.22
* The hazard ratio for death at 28 days was 1.07 (95% confidence interval [CI], 0.84 to 1.38) in the high-target group, as
compared with the low-target group.
† The hazard ratio for death at 90 days was 1.04 (95% CI, 0.83 to 1.30) in the high-target group, as compared with the
low-target group.
‡ Organ support refers to the use of vasopressors, mechanical ventilation, or renal-replacement therapy.
§ Acute myocardial infarction was defined as typical electrocardiographic changes, with a concomitant increase in tropo-
nin, and segmental echocardiographic hypokinesia or akinesia, with the infarction confirmed, when possible, by means
of coronary angiography.
hypertension may indicate that targeting a high- design, investigators were invited to follow these
er mean arterial pressure is acceptable because guidelines in the low-target group. However, the
it was not associated with greater harms. observed mean arterial pressures in the low-target
The guidelines of the Surviving Sepsis Cam- group (target range, 65 to 70 mm Hg) were for the
paign recommend targeting a mean arterial pres- most part between 70 and 75 mm Hg. Similarly,
sure of at least 65 mm Hg. According to our study the observed values in the high-target group

133 april 24, 2014 1591
pressure is about 70 mm Hg, as recently re-

100
ported by Poukkanen et al.19 In that study, pa-
tients spent more than 75% of the time at a
Cumulative Survival (%)

75 mean arterial pressure of more than 70 mm Hg.
Low-target group
Finally, the generalizability of our trial results
High-target group
may be limited because of the frequent use of
50
glucocorticoids and activated protein C and
because of the large number of patients who
25 were excluded because of the narrow inclusion
window.
0 In conclusion, among patients with septic
0 28 60 90 shock, 28-day and 90-day mortality did not dif-
Day fer significantly between those who were treated
No. at Risk to reach a target mean arterial pressure of 80 to
Low target 379 256 233 225
High target 375 249 227 219
85 mm Hg and those who were treated to reach
a target of 65 to 70 mm Hg.
Figure 3. Kaplan–Meier Curves for Cumulative Survival. Presented in part at the annual meeting of Société de Réanima-
tion de Langue Française, Paris, January 16–18, 2013; the Interna-
Data for the survival analysis, which was performed in the intention-to treat tional Symposium on Intensive Care and Emergency Medicine,
population, were censored at 90 days. There was no significant difference Brussels, March 19–22, 2013; the annual meeting of the European
in survival between the high-target group and the low-target group (P = 0.57 Society of Intensive Care Medicine, Paris, October 5–9, 2013; and
at 28 days; P = 0.74 at 90 days). the annual meeting of the German Interdisciplinary Society for
Intensive and Emergency Medicine, Leipzig, December 4–6, 2013.
Supported by the French Ministry of Health.
were also higher (between 85 and 90 mm Hg) than Dr. Asfar reports receiving lecture fees from LFB; Dr. Meziani,
receiving grant support from LFB; Dr. Mira, receiving honoraria
the predefined target range of 80 to 85 mm Hg. and travel support from LFB and Eisai and honoraria from
Thus, the target between-group difference was AstraZeneca; Dr. Weiss, receiving lecture fees from Gore; Dr.
well maintained. Whether higher achieved mean Gonzalez, receiving travel support from Merck Sharp and
Dohme and Novartis; and Dr. Mercat, receiving grant support
arterial pressures in the two groups influenced from Covidien, Maquet, and General Electric, consulting fees
the results is impossible to ascertain. However, from Air Liquide Medical Systems, and fees for serving on a
given the pragmatic nature of the trial, these steering committee from Faron Pharmaceuticals and being
named as an inventor on a patent related to methods of evaluat-
data were not recorded as protocol violations. ing a patient for positive end-expiratory pressure therapy (US
In addition, the higher mean arterial pressures 12/834,354), which is to be licensed to General Electric. No
in the two groups may reflect the reluctance of other potential conflict of interest relevant to this article was
reported.
some attending physicians to decrease the vaso- Disclosure forms provided by the authors are available with
pressor infusion rate when the mean arterial the full text of this article at NEJM.org.
Appendix
The authors’ affiliations are as follows: the Departments of Medical Intensive Care (P.A., A.M.) and Surgical Intensive Care (S.G.) and
the Clinical Research Center (J.-F.H.), University Hospital of Angers, Angers (P.A.), the Department of Medical Intensive Care, Nouvel
Hôpital Civil, Strasbourg University, Strasbourg (F.M.), the Medical and Surgical Intensive Care Unit, Le Mans Hospital, Le Mans
(F. Grelon), the Department of Medical and Toxicological Intensive Care, Lariboisière University Hospital (B.M.), the Department of
Medical Intensive Care, Cochin University Hospital (J.-P.M.), the Department of Medical Intensive Care, Georges Pompidou European
Hospital (N.W.), and the Department of Intensive Care, Saint Louis Hospital (E.M.), Paris, the Department of Medical Intensive Care,
Le Kremlin Bicêtre University Hospital, Le Kremlin Bicêtre (N.A., J.-L.T.), the Department of Medical Intensive Care, Tours University
Hospital, Tours (P.-F.D.), the Department of Medical Intensive Care, Saint Brieuc Hospital, Saint Brieuc (F.L.), the Department of Infec-
tious Diseases and Medical Intensive Care, Rennes University Hospital, Rennes (Y.L.T.), the Department of Medical Intensive Care,
Nancy University Hospital, Nancy (M.C.), the Department of Medical Intensive Care, Poitiers University Hospital, Poitiers (R.R.), the
Department of Medical and Surgical Intensive Care, Avicenne Teaching Hospital, Bobigny (F. Gonzalez), the Department of Medical
Intensive Care, Nantes University Hospital, Nantes (C. Guitton), the Department of Medical Intensive Care, Rouen University Hospital,
Rouen (F.T.), the Department of Medical Intensive Care, Brest University Hospital, Brest (J.-M.T.), the Department of Medical Intensive
Care, Versailles University Hospital, Versailles (P.G.), the Department of Intensive Care, Lille University Hospital, Lille (T.V.D.L.), the
Department of Medical and Surgical Intensive Care, Boulogne Billancourt University Hospital, Boulogne Billancourt (A.V.-B.), the De-
partment of Intensive Care, Avignon Hospital, Avignon (G.P.), the Department of Medical and Surgical Intensive Care, La Rochelle Saint
Louis Hospital, La Rochelle (O.L.), the Department of Medico-Surgical Intensive Care, University Paris Diderot, Colombes (J.-D.R.), the
Department of Medical and Surgical Intensive Care, Quimper Hospital, Quimper (F.H.), the Department of Medical Intensive Care, Caen
University Hospital, Caen (D.C.), and the Department of Medical Intensive Care, Lyon University Hospital, Lyon (C. Guérin) — all in
France; and the Department of Anesthesiology, Ulm University Hospital, Ulm, Germany (P.R.).
1592 134
References
1. Dellinger RP, Levy MM, Rhodes A, et al. Comparison of dopamine and norepi- 14. Heyland DK, Muscedere J, Drover J,
al. Surviving Sepsis Campaign: interna- nephrine in the treatment of shock. Jiang X, Day AG. Persistent organ dys-
tional guidelines for management of se- N Engl J Med 2010;362:779-89. function plus death: a novel, composite
vere sepsis and septic shock: 2012. Crit 8. Russell JA, Walley KR, Singer J, et al. outcome measure for critical care trials.
Care Med 2013;41:580-637. Vasopressin versus norepinephrine infu- Crit Care 2011;15:R98.
2. LeDoux D, Astiz ME, Carpati CM, sion in patients with septic shock. N Engl 15. Ranieri VM, Thompson BT, Barie PS,
Rackow EC. Effects of perfusion pressure J Med 2008;358:877-87. et al. Drotrecogin alfa (activated) in adults
on tissue perfusion in septic shock. Crit 9. Dünser MW, Takala J, Ulmer H, et al. with septic shock. N Engl J Med 2012;366:
Care Med 2000;28:2729-32. Arterial blood pressure during early sepsis 2055-64.
3. Bourgoin A, Leone M, Delmas A, Gar- and outcome. Intensive Care Med 2009;35: 16. Perner A, Haase N, Guttormsen AB,
nier F, Albanèse J, Martin C. Increasing 1225-33. et al. Hydroxyethyl starch 130/0.42 versus
mean arterial pressure in patients with 10. Badin J, Boulain T, Ehrmann S, et al. Ringer’s acetate in severe sepsis. N Engl J
septic shock: effects on oxygen variables Relation between mean arterial pressure Med 2012;367:124-34. [Erratum, N Engl
and renal function. Crit Care Med 2005; and renal function in the early phase of J Med 2012;367:481.]
33:780-6. shock: a prospective, explorative cohort 17. Walkey AJ, Wiener RS, Ghobrial JM,
4. Rivers E, Nguyen B, Havstad S, et al. study. Crit Care 2011;15:R135. Curtis LH, Benjamin EJ. Incident stroke
Early goal-directed therapy in the treat- 11. Strandgaard S, Olesen J, Skinhoj E, and mortality associated with new-onset
ment of severe sepsis and septic shock. Lassen NA. Autoregulation of brain circu- atrial fibrillation in patients hospitalized
N Engl J Med 2001;345:1368-77. lation in severe arterial hypertension. Br with severe sepsis. JAMA 2011;306:2248-
5. Annane D, Vignon P, Renault A, et al. Med J 1973;1:507-10. 54.
Norepinephrine plus dobutamine versus 12. Bone RC, Balk RA, Cerra FB, et al. 18. Sprung CL, Annane D, Keh D, et al.
epinephrine alone for management of sep- Definitions for sepsis and organ failure Hydrocortisone therapy for patients with
tic shock: a randomised trial. Lancet 2007; and guidelines for the use of innovative septic shock. N Engl J Med 2008;358:111-
370:676-84. [Erratum, Lancet 2007;370: therapies in sepsis. Chest 1992;101:1644- 24.
1034.] 55. 19. Poukkanen M, Wilkman E, Vaara ST,
6. López A, Lorente JA, Steingrub J, et al. 13. Pottecher T, Calvat S, Dupont H, et al. Hemodynamic variables and progres-
Multiple-center, randomized, placebo-con- Durand-Gasselin J, Gerbeaux P. Haemo- sion of acute kidney injury in critically ill
trolled, double-blind study of the nitric dynamic management of severe sepsis: patients with severe sepsis: data from the
oxide synthase inhibitor 546C88: effect recommendations of the French Intensive prospective observational FINNAKI study.
on survival in patients with septic shock. Care Societies (SFAR/SRLF) Consensus Crit Care 2013;17:R295.
Crit Care Med 2004;32:21-30. Conference, 13 October 2005, Paris, Copyright © 2014 Massachusetts Medical Society.
7. De Backer D, Biston P, Devriendt J, et France. Crit Care 2006;10:311.

n engl j med 370;17135

edi t or i a l
Is There a Good MAP for Septic Shock?

James A. Russell, M.D.
As fundamental as the issue is, there is no clear, Reviews1,2 and sepsis guidelines3 recommend
high-level evidence to determine the most effec- a target MAP of more than 65 mm Hg in patients
tive mean arterial pressure (MAP) for resuscita- with septic shock. However, these recommenda-
tion of patients with septic shock. During hypo- tions are based on low-quality evidence. Accord-
tension, three salient aspects of cardiovascular ingly, Asfar et al.4 now report in the Journal the
physiology are important. First, the autoregula- results of a large, randomized, controlled trial
tion of cerebral blood flow maintains an ade- of targeting a low MAP (65 to 70 mm Hg) versus
quate level over a wide range of pressures until a high MAP (80 to 85 mm Hg) among patients
a critical pressure (about 50 mm Hg) is reached; with septic shock at 29 centers in France. By
below this level, as the pressure falls, so does 28 days, there was no significant difference in
cerebral blood flow. Second, chronic hyperten- mortality (the primary end point) between the
sion shifts the autoregulatory curve of the rela- low-target group and the high-target group. In
tionship between pressure and perfusion so that the prospectively defined stratum of patients
perfusion is decreased at a higher critical pres- with chronic hypertension (more than 40% of
sure; long-term antihypertensive treatment re- the patients), those in the high-target group had
stores the autoregulatory curves toward normal. less renal dysfunction and need for renal-replace-
Third, there are differences among organs in the ment therapy than did those in the low-target
critical point for oxygen delivery: gut oxygen de- group. However, there was a safety concern,
livery decreases earlier in shock than oxygen since patients in the high-target group had an
delivery to other organs. The consequence of increased risk of new atrial fibrillation, which is
inadequate perfusion is ischemic injury to the independently associated with an increased risk
kidney, gut, brain, and myocardium, followed by of stroke.5 Unfortunately, stroke was not evalu-
multiple organ dysfunction and death. ated by Asfar and colleagues.
The goal of cardiovascular resuscitation of The strengths of this trial are that the inter-
septic shock is to improve organ perfusion, often vention was assigned by blinded randomization
by increasing the MAP. Adequate fluid resuscita- in a multicenter context in intensive care units;
tion is limited when the administration of fluids the two study groups were well balanced; the
causes edema (e.g., acute lung injury). Vasopres- intervention was a pragmatic, real-world study;
sors are added if fluid resuscitation does not and the difference in ranges of MAP values be-
restore adequate perfusion, but such therapy is tween the two groups was significant. One limi-
limited by excessive vasoconstriction and organ tation was that the interventions targeting a low
ischemia. A common clinical problem is distin- or high MAP were not blinded.
guishing between ischemia that is caused by in- How do these results align with those of oth-
adequate resuscitation (e.g., as manifested by er randomized, controlled trials of vasoactive
oliguria) and ischemia that is caused by exces- interventions in shock? I compared intervention
sive vasoconstriction. application and outcomes of this study with
1648 n engl j med 370;17 136

Source: The New England Journal of Medicine editorials
both published and unpublished data regarding shock,4,6,10 severe sepsis,9 and shock7,8) and trial
the norepinephrine control groups of large, location,4,6-10 since mortality associated with
randomized, controlled trials of shock6-10 septic shock varies according to country,11 mak-
(Table 1). As compared with the findings in the ing comparisons between countries and conti-
study by Asfar et al., the target and actual MAP nents difficult.
on day 1 in trials in which vasoactive agents The findings of Asfar et al. have at least three
were used were similar to those in the low- major clinical implications. First, they show that
target group and lower than those in the high- there is no indication for routinely targeting a
target group. Asfar and colleagues used less high MAP in patients with septic shock, since
fluid and higher doses of norepinephrine than there was no significant between-group differ-
were used in some trials8-10 but administered ence in mortality, although patients in the high-
less norepinephrine and lower fluid levels than target group had an increased rate of atrial fibril-
were used in another trial.7 Thus, there is vari- lation. Second, a high MAP target may decrease
ability in how fluids and vasopressors are used the risk of renal injury and the need for renal-
to achieve target MAPs in randomized, con- replacement therapy (number needed to treat of
trolled trials. Finally, rates of death in the trial 9.5 to prevent one patient from needing renal-
by Asfar el al. were similar to those in two tri- replacement therapy) in patients with hyperten-
als,6,10 higher than those in one trial,8 and lower sion. I make this point because of the well-known
than those in two other trials.7,9 These mortality risks and costs of renal-replacement therapy.
differences may arise from differences in the in- Thus, there are several target MAPs for septic
clusion criteria used in these trials (septic shock, depending on the circumstances of the
Table 1. Comparison of Norepinephrine Control Groups and Intervention Groups in Randomized, Controlled Trials of Vasoactive Agents
in Shock.*
Renal- Death
Mean Arterial Pressure Norepinephrine Replacement at 28
Trial (MAP) Infusion Rate Fluid Balance Therapy Days
Actual, Actual, Days
Target Day 0 Day 1 Day 0 Day 1 Day 2 Day 0 Day 1 0 to 4
mm Hg μg/kg/min ml %
Asfar et al.4†
Low-target group
All patients 65 to 70 74 74 0.35 0.45 0.16 1603 1016 2,800 35.8 34
Patients with chronic 65 to 70 42.2
hypertension
High-target group
All patients 80 to 85 74 84 0.40 0.58 0.38 1595 1106 2,400 33.5 36
Patients with chronic 80 to 85 31.7
hypertension
Annane et al.6‡ 70 70 80 0.94 1.09 0.65 1586 172 −2,767 24.8 34
De Backer et al.7 MD 58 76 0.54 0.82 0.68 2100 1700 8,300 17.0§ 48
Myburgh et al.8¶ MD 70 73 0.26 0.17 0.07 2232 1782 5,712 22.1 26
Rivers et al.9‖ 65 76 81 NA NA NA 3500 NA 10,602** NA 49
10
Russell et al. 65 to 75 72 73 0.28 0.20 0.08 1500 2500 11,000 43.6 39
* MD indicates clinician judgment, and NA not available.

† Listed are results for both the control (low-target) group and the intervention (high-target) group.
‡ In the study by Annane et al., fluid balance was reported for survivors only.
§ The percentage was 19.9% in the subgroup with septic shock.
¶ In the study by Myburgh et al., if no target MAP was ordered, the default target MAP was 70 mm Hg.
‖ In the study by Rivers et al., the control group received a combination treatment.
** Fluid balance was averaged over 3 days rather than 4.

april 24, 2014 1649
patient. In some randomized, controlled trials 5. Walkey AJ, Wiener RS, Ghobrial JM, Curtis LH, Benjamin EJ.
Incident stroke and mortality associated with new-onset atrial
(and in clinical practice), practitioners use more fibrillation in patients hospitalized with severe sepsis. JAMA
fluid (increasing the risk of acute lung injury), 2011;306:2248-54.
whereas others use more vasopressors (increas- 6. Annane D, Vignon P, Renault A, et al. Norepinephrine plus
dobutamine versus epinephrine alone for management of septic
ing the risk of renal injury). Indeed, methods for shock: a randomised trial. Lancet 2007;370:676-84. [Erratum,
targeting a MAP among patients in septic shock Lancet 2007;370:1034.]
are probably critical to the success of the strategy 7. De Backer D, Biston P, Devriendt J, et al. Comparison of do-
pamine and norepinephrine in the treatment of shock. N Engl J
and deserving of greater investigation. Med 2010;362:779-89.
Disclosure forms provided by the author are available with the 8. Myburgh JA, Higgins A, Jovanovska A, Lipman J, Rama-
full text of this article at NEJM.org. krishnan N, Santamaria J. A comparison of epinephrine and
norepinephrine in critically ill patients. Intensive Care Med
From the Center for Heart Lung Innovation and the Division of 2008;34:2226-34.
Critical Care Medicine, St. Paul’s Hospital, University of British 9. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed
Columbia, Vancouver, Canada. therapy in the treatment of severe sepsis and septic shock.
N Engl J Med 2001;345:1368-77.
This article was published on March 18, 2014, at NEJM.org. 10. Russell JA, Walley KR, Singer J, et al. Vasopressin versus
norepinephrine infusion in patients with septic shock. N Engl J
1. Vincent J-L, De Backer D. Circulatory shock. N Engl J Med Med 2008;358:877-87.
2013;369:1726-34. 11. Levy MM, Artigas A, Phillips GS, et al. Outcomes of the
2. Angus DC, van der Poll T. Severe sepsis and septic shock. Surviving Sepsis Campaign in intensive care units in the USA
N Engl J Med 2013;369:840-51. [Erratum, N Engl J Med 2013; and Europe: a prospective cohort study. Lancet Infect Dis 2012;
369:2069.] 12:919-24.
Campaign: international guidelines for management of severe DOI: 10.1056/NEJMe1402066
sepsis and septic shock, 2012. Intensive Care Med 2013;39:165-
228.
4. Asfar P, Meziani F, Hamel J-F, et al. High versus low blood-
pressure target in patients with septic shock. N Engl J Med 2014;
370:1583-93.
1650 n engl j med 370;17 138

c or r e sp ondence
High versus Low Blood-Pressure Target in Septic Shock

To the Editor: As noted by Asfar et al. (April 24 1. Asfar P, Meziani F, Hamel J-F, et al. High versus low blood-
pressure target in patients with septic shock. N Engl J Med 2014;
issue),1 a major concern related to targeting a 370:1583-93.
relatively low mean arterial blood pressure in pa- 2. Strandgaard S, Olesen J, Skinhøj E, Lassen NA. Autoregula-
tients with septic shock is that the brain is at risk tion of brain circulation in severe arterial hypertension. Br Med
J 1973;1:507-10.
for ischemia in this condition. The risk of poten- 3. Gofton TE, Young GB. Sepsis-associated encephalopathy.
tial brain ischemia is particularly pertinent Nat Rev Neurol 2012;8:557-66.
among patients with a history of arterial hyper- 4. Drummond JC. The lower limit of autoregulation: time to
revise our thinking? Anesthesiology 1997;86:1431-3.
tension, because their cerebral autoregulatory 5. Berg RMG, Plovsing RR, Ronit A, Bailey DM, Holstein-
curve may be shifted to the right.2 Indeed, cerebral Rathlou NH, Møller K. Disassociation of static and dynamic
ischemia may contribute to sepsis-associated en- cerebral autoregulatory performance in healthy volunteers after
lipopolysaccharide infusion and in patients with sepsis. Am J
cephalopathy and persistent neurocognitive dys- Physiol Regul Integr Comp Physiol 2012;303:R1127-R1135.
function.3 However, neither the primary nor the DOI: 10.1056/NEJMc1406276
secondary outcomes in the study by Asfar et al.
involved any measures of neurocognitive function.
A mean arterial blood pressure of approxi- To the Editor: Asfar and colleagues found that
mately 65 to 90 mm Hg describes the lower among patients with septic shock, a high blood-
limit of autoregulation in healthy persons,4 and pressure target, as compared with a low one, re-
the same level appears to be present in critically sulted in fewer patients requiring renal-replace-
ill patients with sepsis who do not have a his- ment therapy. Although the analyses stratified
tory of hypertension.5 Before the findings by patients according to the presence or absence of
Asfar et al. are implemented in major guidelines, chronic hypertension and were adjusted for sev-
we think that further studies should address the eral known risk factors for acute kidney injury,
way in which cerebral hemodynamics are spe- the influence of intraabdominal pressure was not
cifically affected in patients with sepsis and a considered in the analyses. Intraabdominal hyper-
history of hypertension, as well as whether target- tension (pressure >12 mm Hg)1 occurs common-
ing a relatively high mean arterial blood pressure ly in critically ill adults and in those with septic
improves long-term neurocognitive function. shock.2 In a study in animals, when intraabdom-
Ronan M.G. Berg, M.D., Ph.D. inal pressure was raised above 20 mm Hg, renal
University Hospital Rigshospitalet blood flow and the glomerular filtration rate de-
Copenhagen, Denmark creased to less than 25% of the normal rates.3 A
ronan@dadlnet.dk prospective cohort study involving 123 patients
Ronni R. Plovsing, M.D., Ph.D. in the intensive care unit identified intraabdomi-
Hvidovre Hospital nal hypertension as the second-best predictor of
Copenhagen, Denmark
the development of acute renal failure, after the
Kirsten Møller, M.D., Ph.D. presence of shock.2 Because the benefits of co-
University Hospital Rigshospitalet variate adjustment in randomized trials are likely
Copenhagen, Denmark
to outweigh the risks,4 we wonder whether intra-
ported. abdominal hypertension should have been con-
sidered as a potential confounder in the study
282 n engl j med 371;3139

correspondence
and whether future trials examining the influ- widely used in nonindustrialized countries. It
ence of resuscitative goals on the risk of acute may be tempting to increase the mean arterial
kidney failure and related outcomes should con- pressure target to improve renal function,4 but
sider this possibility. only in norepinephrine-treated patients.
Andrew W. Kirkpatrick, M.D. Antoine Kimmoun, M.D.
Derek J. Roberts, M.D. Nicolas Ducrocq, M.D.
University of Calgary Bruno Levy, M.D., Ph.D.
Calgary, AB, Canada Service de Réanimation Médicale Brabois
andrew.kirkpatrick@albertahealthservices.ca Vandoeuvre lès Nancy, France
Jan De Waele, M.D. b.levy@chu-nancy.fr
Ghent University Hospital No potential conflict of interest relevant to this letter was re-
Ghent, Belgium ported.
Dr. Kirkpatrick reports serving on an advisory board for Lan- 1. De Backer D, Biston P, Devriendt J, et al. Comparison of do-
theus Medical Imaging, receiving a research grant from Kinetic pamine and norepinephrine in the treatment of shock. N Engl J
Concepts, and receiving a NanoMaxx ultrasonography machine Med 2010;362:779-89.
from Sonosite for research use; and Dr. De Waele, receiving 2. Myburgh JA, Higgins A, Jovanovska A, Lipman J, Rama-
consulting fees from Smith and Nephew, Kinetic Concepts, krishnan N, Santamaria J. A comparison of epinephrine and
Pfizer, and AstraZeneca. No other potential conflict of interest norepinephrine in critically ill patients. Intensive Care Med 2008;
relevant to this letter was reported. 34:2226-34.
1. Kirkpatrick AW, Roberts DJ, De Waele J, et al. Intra-abdom-
Campaign: international guidelines for management of severe
inal hypertension and the abdominal compartment syndrome:
sepsis and septic shock: 2012. Crit Care Med 2013;41:580-637.
updated consensus definitions and clinical practice guidelines
4. Russell JA. Is there a good MAP for septic shock? N Engl J
from the World Society of the Abdominal Compartment Syn-
Med 2014;370:1649-51.
drome. Intensive Care Med 2013;39:1190-206.
2. Dalfino L, Tullo L, Donadio I, Malcangi V, Brienza N. Intra- DOI: 10.1056/NEJMc1406276
abdominal hypertension and acute renal failure in critically ill
patients. Intensive Care Med 2008;34:707-13.
3. Harman PK, Kron IL, McLachlan HD, Freedlender AE, Nolan The Authors Reply: Berg et al. correctly state
SP. Elevated intra-abdominal pressure and renal function. Ann that neither imaging for cerebral ischemia nor
Surg 1982;196:594-7. assessment of neurocognitive function was part
4. Kahan BC, Jairath V, Doré CJ, Morris TP. The risks and re-
wards of covariate adjustment in randomized trials: an assess- of the secondary-outcome assessments. In some
ment of 12 outcomes from 8 studies. Trials 2014;15:139. recruiting centers, scores on the Glasgow Coma
DOI: 10.1056/NEJMc1406276 Scale and Confusion Assessment Method for the
ICU were recorded. Because such scores are part
To the Editor: Asfar et al. suggest that a mean of a substudy, these data are not reported in our
arterial pressure target of 80 to 85 mm Hg might article, which presents the primary study results.
decrease the risk of renal injury during septic Kirkpatrick et al. note that we did not assess
shock among patients with chronic hyperten- the possible confounding role of intraabdominal
sion. Nevertheless, the choice of vasopressor is hypertension in the development of renal failure.
not discussed in their study, in which 95% of the When the trial was designed, the assessment of
patients were treated with norepinephrine. We intraabdominal hypertension was not part of the
think that the results would probably have been guidelines of the Surviving Sepsis Campaign.
different if the patients had been treated with Therefore, it was not recorded, in keeping with
dopamine or epinephrine. Indeed, when used in the pragmatic nature of the trial. Moreover, only
patients with shock, dopamine has been shown 14% of the patients had a history of recent sur-
to markedly increase the rate of arrhythmias,1 gery, and therefore intraabdominal hypertension
whereas the use of epinephrine,2,3 as compared was most likely to have been infrequent. Finally,
with norepinephrine, has been associated with a for the study to be methodologically rigorous,
higher rate of withdrawal from study treatment stratification of the patients according to the pres-
due to lactic acidosis or arrhythmias. In these ence or absence of intraabdominal hypertension
latter two studies, vasopressor agents were used would have been mandatory at inclusion, with a
to obtain a mean arterial pressure of 70 mm Hg. risk of an underpowered subsequent analysis.
It is likely that targeting a high pressure with We agree with Kimmoun et al. that our results
dopamine or epinephrine would be associated are limited with regard to the vasopressors used;
with an increased rate of adverse events. Epi- we used norepinephrine. Targeting a high mean
nephrine is cheaper than norepinephrine and is blood pressure with the use of epinephrine or

dopamine might have caused an increase in the Since publication of their article, the authors report no fur-
incidence of adverse effects such as new-onset
arrhythmia1 or lactic acidosis.2 In contrast, one 1. De Backer D, Biston P, Devriendt J, et al. Comparison of do-
pamine and norepinephrine in the treatment of shock. N Engl J
might speculate that infusing vasopressin to in- Med 2010;362:779-89.
crease the mean blood pressure might have been 2. Levy B, Bollaert PE, Charpentier C, et al. Comparison of nor-
associated with opposite findings: in patients epinephrine and dobutamine to epinephrine for hemodynamics,
lactate metabolism, and gastric tonometric variables in septic
with catecholamine-resistant vasodilatory shock, shock: a prospective, randomized study. Intensive Care Med 1997;
vasopressin significantly decreased the incidence 23:282-7.
of a new onset of atrial fibrillation,3 and a post 3. Dünser MW, Mayr AJ, Ulmer H, et al. Arginine vasopressin
in advanced vasodilatory shock: a prospective, randomized, con-
hoc analysis of the Vasopressin and Septic Shock trolled study. Circulation 2003;107:2313-9.
Trial showed a reduced progression to renal 4. Gordon AC, Russell JA, Walley KR, et al. The effects of vaso-
failure in vasopressin-treated patients at risk for pressin on acute kidney injury in septic shock. Intensive Care
Med 2010;36:83-91.
kidney injury.4
Pierre Asfar, M.D., Ph.D. DOI: 10.1056/NEJMc1406276
University Hospital of Angers
Angers, France
piasfar@chu-angers.fr
Jean-Louis Teboul, M.D., Ph.D.
Le Kremlin-Bicêtre University Hospital
Le Kremlin-Bicêtre, France
Peter Radermacher, M.D., Ph.D.
Ulm University Hospital
Ulm, Germany
284 n engl j med 371;3 141

VENTILATOR-INDUCED LUNG INJURY
Since the widespread use of mechanical ventilator assist began in the 1970s, continuing until early in the 21st
century, critical care physicians tried to duplicate normal breathing patterns and gas exchange. This section
reviews ideas suggesting that much of the morbidity from mechanical ventilator support may derive from trying
to replicate this normal physiology.
142
Case Challenge
Ventilator-Induced Lung Injury

A 77-year-old man is undergoing mechanical ventilation after severe sepsis and circulatory shock. He has a
positive fluid balance, and echocardiography has been performed, showing normal left ventricular function.
The pulmonary-artery occlusion pressure is 11 mm Hg. What strategy is indicated for mechanical ventilation?
Our story continues. A 77-year-old man whose medical history includes treated hypertension and hypercholester-
olemia, previous heavy alcohol intake, and mild cognitive impairment was admitted to the intensive care unit
(ICU) of a university hospital after a resection of the rectosigmoid colon with closure of the rectal stump and for-
mation of an end colostomy performed for fecal peritonitis caused by a perforated sigmoid colon. (In the previous
installment of this case, there were 4748 votes on monitoring strategies for circulatory support, and more than one
strategy could be selected. Echocardiographic assessment of cardiac function was chosen by 41% of respondents,
pulmonary-artery catheterization for continuous monitoring by 27%, measurement of variation in pulse pressure
during mechanical ventilation by 27%, and measurement of central venous oxygen saturation with a goal of more
than 70% saturation by 51%.)
Now we must deal with setting the patient’s ventilator. The patient is 178 cm tall and weighs 60 kg. On arrival in
the ICU, he was hypotensive with a poor urine output. His arterial blood pressure has been supported with intra-
venous fluids and a norepinephrine infusion. After 24 hours in the ICU, he has a positive fluid balance of 2 liters.
143
Echocardiography shows normal left ventricular size and function, and a pulmonary-artery catheter shows that
the pulmonary-artery occlusion pressure is 11 mm Hg. The patient is sedated and undergoing volume-cycled
mechanical ventilation. Current ventilator settings are a tidal volume of 450 ml, a respiratory rate of 20 breaths
per minute, a fraction of inspired oxygen (FiO2) of 0.6, and a positive end-expiratory pressure (PEEP) of 5 cm of
water. Chest radiography shows bilateral diffuse alveolar infiltrates, current arterial blood gases are pH 7.25, the
partial pressure of arterial oxygen (PaO2) is 74 mm Hg (9.8 kPa), the partial pressure of arterial carbon dioxide
(PaCO2) is 55 mm Hg (7.3 kPa), and the plateau airway pressure (at a time of zero tracheal airflow) is 32 cm of
water.

What mechanical ventilation strategy would provide the best support for this patient?
A. Continue with the current ventilator settings, but increase the respiratory rate and tidal volume to normalize the
PaCO2.
B. Continue with the current ventilator mode, but decrease the tidal volume to 360 ml and increase the respiratory
rate to 22 breaths per minute.
C. Change the ventilator mode to pressure-controlled synchronized intermittent mandatory ventilation with pres-
sure support (SIMV/PS) with a PEEP of 10 cm of water, with a pressure control of 30 cm of water and pressure
support of 20 cm of water above the PEEP.
D. Continue with the current ventilator mode, increase the PEEP to 10 cm of water, increase the respiratory rate to
22 breaths per minute, and decrease the current tidal volume slightly to keep the plateau pressure at 30 cm of
water or less.
144
review article


Arthur S. Slutsky, M.D., and V. Marco Ranieri, M.D.
T
From the Keenan Research Center, Li Ka he purpose of mechanical ventilation is to rest the respiratory
Shing Knowledge Institute, St. Michael’s muscles while providing adequate gas exchange. Ventilatory support proved
Hospital, and the Department of Medi-
cine and Interdepartmental Division of to be indispensable during the 1952 polio epidemic in Copenhagen, decreasing
Critical Care Medicine, University of To- mortality among patients with paralytic polio from more than 80% to approximately
ronto — both in Toronto (A.S.S.); and 40%.1 Despite the clear benefits of this therapy, many patients eventually die after
Dipartimento di Anestesia e Medicina
degli Stati Critici, Ospedale S. Giovanni the initiation of mechanical ventilation, even though their arterial blood gases may
Battista Molinette, Università di Torino, have normalized.
Turin, Italy (V.M.R.). Address reprint re- This mortality has been ascribed to multiple factors, including complications
quests to Dr. Slutsky at St. Michael’s Hos-
pital, 30 Bond St., Toronto, ON M5B 1W8, of ventilation such as barotrauma (i.e., gross air leaks), oxygen toxicity, and hemo-
Canada, or at slutskya@smh.ca. dynamic compromise.2,3 During the polio epidemic, investigators noted that me-
chanical ventilation could cause structural damage to the lung.4 In 1967, the term
This article was updated on April 24,
2014, at NEJM.org. “respirator lung” was coined to describe the diffuse alveolar infiltrates and hyaline
membranes that were found on postmortem examination of patients who had
N Engl J Med 2013;369:2126-36.
DOI: 10.1056/NEJMra1208707 undergone mechanical ventilation.5 More recently, there has been a renewed focus
Copyright © 2013 Massachusetts Medical Society. on the worsening injury that mechanical ventilation can cause in previously dam-
A correction was made to this article aged lungs and the damage it can initiate in normal lungs. This damage is character-
after publication on April 24, 2014, and is ized pathologically by inflammatory-cell infiltrates, hyaline membranes, increased
reflected in this PDF. vascular permeability, and pulmonary edema. The constellation of pulmonary con-
sequences of mechanical ventilation has been termed ventilator-induced lung injury.
The concept of ventilator-induced lung injury is not new. In 1744, John Fothergill
discussed a case of a patient who was “dead in appearance” after exposure to coal
fumes and who was successfully treated by mouth-to-mouth resuscitation.6 Fother-
gill noted that mouth-to-mouth resuscitation was preferable to using bellows because
“the lungs of one man may bear, without injury, as great a force as those of an-
other man can exert; which by the bellows cannot always be determin’d.” Fother-
gill clearly understood the concept that mechanical forces generated by bellows (i.e.,
a ventilator) could lead to injury.
However, it was not until early in this century that the clinical importance of
ventilator-induced lung injury in adults was confirmed by a study showing that a
ventilator strategy designed to minimize such injury decreased mortality among
patients with the acute respiratory distress syndrome (ARDS).7 Given the clinical
importance of ventilator-induced lung injury, this article will review mechanisms
underlying the condition, its biologic and physiological consequences, and clinical
strategies to prevent it and mitigate its effects.
PATHOPH YSIOL O GIC A L FE AT UR E S
Pressures in the Lung

During a lifetime, a person will take approximately 500 million breaths. For each
breath, the pressure necessary to inflate the lungs comprises the pressure to over-
come airway resistance and inertance (a measure of the pressure gradient required
2126 145
n engl j med 369;22 nejm.org november 28, 2013
The New
Medical ForAll
Society. personal
reserved.
to accelerate the gas) and the pressure to over- pressure) (Fig. 1). Thus, lung volume and trans-
come the elastic properties of the lung. When pulmonary pressure are inextricably linked.
airflow is zero (e.g., at end inspiration), the prin- Regional lung overdistention is a key factor
cipal force maintaining inflation is the transpul- in generating ventilator-induced lung injury. Since
monary pressure (alveolar pressure minus pleural there is no well-accepted clinical method of
A Normal spontaneously breathing B Normal anesthetized, paralyzed C Patient with stiff chest wall, on
person, at end inspiration patient on mechanical ventilation, mechanical ventilation, at end
at end inspiration inspiration
Palv = 0 cm H2O Palv = 9 cm H2O Palv = 30 cm H2O
Ppl = −8 cm H2O Ppl = 1 cm H2O

Ppl = 25 cm H2O
Ptp = 0 − (−8) = +8 cm H2O Ptp = 9 − 1 = +8 cm H2O Ptp = 30 − 25 = +5 cm H2O
D Trumpet player while playing a note E Patient with marked respiratory distress, on noninvasive
ventilation, at end inspiration
Palv = 150 cm H2O Palv = 10 cm H2O
Ppl = 140 cm H2O

Ppl = −15 cm H2O
Ptp = 150 − 140 = +10 cm H2O Ptp = 10 − (−15) = +25 cm H2O
Figure 1. Intrathoracic Pressures and Lung Stretching.

Panel A shows end inspiration in a patient with normal lung function who is breathing spontaneously (with an open
glottis); the alveolar pressure (Palv) is 0, and the pleural pressure (Ppl) is negative (−8 cm of water), creating a transpul-
COLOR FIGURE
monary pressure (Ptp) of +8 cm of water (Palv minus Ppl). Panel B shows the same lung while the patient
Draft 6
undergoes11/8/13
general anesthesia and positive-pressure ventilation with the use of the same tidal volume Author as in PanelSlutskyA. The lung
would be similarly stretched, with an alveolar pressure of 9 cm of water and a pleural pressure Fig # of 1 1 cm of water for a
transpulmonary pressure of +8 cm of water. Panel C shows end inspiration in a patient with severe Title obesity, massive
Ventilator Induced as-
Lung
Injury
cites, or pleural effusions, who may have a very stiff chest wall. In such patients, much of theME pressure that is applied
by the ventilator will be used to distend the chest wall rather than the lung. As such, the plateauDE pressureDrazen may be high,
but so will the pleural pressure, and hence there may not be an increase in transpulmonary pressureArtist Nwith
Koscalaccompany-
ing lung overdistention. Panel D shows a musician playing a trumpet, which can result in airwayFigure pressures
AUTHOR PLEASE of NOTE:
as much
has been redrawn and type has been reset
as 150 cm of water. However, because of the positive pleural pressure developed by the respiratory muscles, the pres-
sure across the lung will not exceed normal values. Panel E shows a patient with marked dyspnea Issue date who11/28/13
is undergoing
a type of mechanical ventilation that requires the active contraction of the respiratory muscles to initiate the assist-
ed breath (e.g., noninvasive ventilation or pressure-support ventilation). In such cases, there may be large negative
swings in pleural pressure, leading to a very high transpulmonary pressure, even though the airway pressure is only
10 cm of water. (Additional details about the importance of transpulmonary pressure are discussed by Rahn et al.8)
n engl j med 369;22 146

The New
Medical ForAll
Society. personal
reserved.
measuring regional overdistention, limiting in- ing ventilator-induced lung injury. Other factors
flation pressure during mechanical ventilation (e.g., respiratory acidosis, respiratory frequency,
is used as a surrogate strategy to limit overdis- pulmonary vascular pressures, and body temper-
tention. This is currently a reasonable thera- ature) have been shown experimentally to be rel-
peutic approach, but it is important to understand evant to ventilator-induced lung injury, but these
from a physiological standpoint the usefulness factors probably represent second-order effects
and limitations of the various pressures that are and will not be addressed further in this review.
measured.
Alveolar pressure is relatively easy to estimate Ventilation at High Lung Volumes
clinically as the airway pressure during a period Ventilator-induced lung injury can occur because
of zero flow; in a patient undergoing mechanical of ventilation at high (absolute) lung volumes,
ventilation who is not making spontaneous leading to alveolar rupture, air leaks, and gross
breathing efforts, the airway pressure that is barotrauma (e.g., pneumothorax, pneumomedias-
measured during a period when airflow is stopped tinum, and subcutaneous emphysema3) (Fig. 2).
at end inspiration is called the plateau pressure. The term barotrauma can be misleading, because
Unfortunately, pleural pressure — the other vari- the critical variable leading to the air leaks is re-
able needed to calculate transpulmonary pressure gional lung overdistention, not high airway pres-
— is more complicated. There is a gravitational sure per se (Fig. 1D).
gradient in pleural pressure, and it can be esti- More subtle injury that is manifested as pulmo-
mated in the broader clinical setting only by nary edema can occur as a result of lung over-
measurement of esophageal pressure.9 This mea- distention. In a classic experiment, Webb and
surement is somewhat cumbersome to perform Tierney14 ventilated rats with very high peak air-
and yields only approximate results. Therefore, way pressures (and therefore overdistention) and
the plateau pressure is the most common vari- zero positive end-expiratory pressure (PEEP). Hy-
able used in a clinical setting to indicate lung
overdistention. However, there are nuances re- Figure 2 (facing page). Lung Injury Caused by Forces
quired in interpreting the plateau pressure. If Generated by Ventilation at Low and High Lung Volumes.
the patient is not making respiratory efforts, the When ventilation occurs at low lung volumes, lung in-
plateau pressure represents the pressure that is jury can be caused by the opening and closing of lung
units (atelectrauma) as well as by other mechanisms.
distending the lungs plus the chest wall. In a This injury is magnified when there is increased lung
patient with a stiff chest wall (e.g., a patient with inhomogeneity, as shown on computed tomography
a pleural effusion or massive ascites), a large frac- (Panel A), especially in patients with the acute respira-
tion of ventilator-delivered pressure is dissipated tory distress syndrome (ARDS) who have surfactant
in inflating the chest wall rather than the lung. dysfunction, pulmonary edema, and atelectasis.11 In
addition, ventilation may be very inhomogeneous, a
Thus, a high airway pressure — in this case, the status that may be partially or fully reversed by the use
plateau pressure — may not be indicative of ex- of positive end-expiratory pressure (PEEP), as shown in
cessive pulmonary stretching forces (i.e., elevated a ventilated ex vivo rat lung (see video in Slutsky and
transpulmonary pressure) (Fig. 1C). Hudson12). At high lung volumes, overdistention can
By analogy, when a musician plays the trumpet, lead to gross barotrauma (air leaks)13 (Panel B). Over-
distention can also lead to increased alveolar–capillary
airway pressure can reach 150 cm of water,10 but permeability and gross pulmonary edema. Ventilation
pneumothorax is uncommon, because pleural at both high and low lung volumes has structural,
pressure is also elevated and there is no overdis- physiological, biologic, and systemic effects (Panel C).
tention (Fig. 1D). In contrast, during noninvasive Mediators that are released into the lung can cause
ventilation, if the patient is markedly distressed further lung injury, recruit neutrophils to the lung, or
set the stage for the development of pulmonary fibro-
and generating very large negative pleural pres- sis. In addition, the increased alveolar–capillary permea-
sures, transpulmonary pressure (and hence lung bility associated with ventilator-induced lung injury can
stretching) may be extremely high, despite low lead to translocation of mediators, lipopolysaccharides,
airway pressures (Fig. 1E). and bacteria into the systemic circulation, potentially
leading to multiple-organ dysfunction and death. PaCO2
denotes partial pressure of arterial carbon dioxide, PaO2
Physical Forces
partial pressure of arterial oxygen, and PMN polymor-
The following sections deal with major physical phonuclear leukocytes.
factors that are thought to be important in produc-

147 november 28, 2013
The New EnglandFor

Copyright 29, 2014.
reserved.
A Ventilation at low lung volume B Ventilation at high lung volume

End expiration End inspiration Normal Hyperinflation
CT Image CT Image
Atelectrauma Lung inhomogeneity Air leaks Overdistention
C Structural consequences Barotrauma

Sloughing of
bronchial epithelium
Hyaline
membranes Alveolus
Epithelial–mesenchymal
transformations
Alveolus
Pulmonary
edema
Surfactant dysfunction
Atelectasis
Fibroproliferation
Increased
alveolar–capillary
permeability
Alveolus
Biologic alterations Physiological abnormalities

Increased concentrations of:
Hydroxyproline Increased physiological
Transforming growth factor-† PMN
TNF-’
Interleukin-8 dead space
†-catenin
Release of mediators: IL-1β
IL-1† IL-6
Decreased compliance
Tumor necrosis factor ’ (TNF-’)
†-catenin
PAM
Interleukin-6 (IL-6)
Interleukin-1† (IL-1†) Decreased PaO2
Increased PaCO2
Recruitment of:
Pulmonary alveolar macrophages (PAMs)
Neutrophils
Activation of epithelium
and endothelium
LPS
Capillary
Systemic effects Bacteria Mediators
Death
Translocation of:
Lipopolysaccharides (LPS) Multiorgan
Bacteria Multiple mechanisms dysfunction
Various mediators (e.g., increased apoptosis)

november 28, 2013 COLOR FIGURE 2129
Draft 7 11/8/13
The New England
For of Medicine
Author Slutsky
Medical userights
Society. All only. No other uses withoutFig
reserved. permission.
# 2
Copyright © 2013 Massachusetts Medical Society. All rights reserved. Title Ventilator Induced Lung
Back to Table of Contents Injury
ME
DE Drazen
poxemia developed in the animals, and post- stretching forces in lung parenchyma at margins
mortem examination revealed perivascular and between aerated and atelectatic regions could be
alveolar edema. Edema did not develop in ani- up to four to five times as high as those in other
mals that underwent ventilation with the same lung regions.
peak airway pressure but with the addition of a
PEEP of 10 cm of water, showing an interaction Biologic Forces
between overdistention and low end-expiratory The physical forces described above may cause the
lung volume with respect to lung injury. The pre- release of various intracellular mediators21 either
cise mechanisms underlying this interaction have directly (by injuring various cells) or indirectly
not been completely elucidated. (by transducing these forces into activation of
Dreyfuss et al.15 found that pulmonary edema cell-signaling pathways in epithelial, endothelial,
developed in animals undergoing ventilation with or inflammatory cells). Some mediators may di-
high tidal volumes, whereas such edema did not rectly injure the lung; others may set the stage for
develop in animals undergoing ventilation with subsequent development of pulmonary fibrosis.22
similar airway pressures but with straps around Additional mediators may act as homing mole-
their abdomens and chests that reduced the tidal cules recruiting cells (e.g., neutrophils) to the
volumes. Thus, their experiments showed that lung, and such cells can then release more injuri-
volume (i.e., lung stretching), not airway pres- ous molecules (Fig. 2).
sure, was the most important factor in determin- This process has been termed biotrauma.23
ing injury, a finding that led them to coin the The translocation of mediators,24 bacteria,25 or
term “volutrauma.” lipopolysaccharide26 from the airspaces into the
Although ventilator-induced lung injury is a systemic circulation may occur in lungs that
well-accepted term, it may be a misnomer. The have increased alveolar–capillary permeability,
key factor causing injury is lung overdistention, which is inherent in the case of ARDS or which
which may be caused by factors other than a is induced by volutrauma or epithelial microtears.
ventilator. For example, Mascheroni et al.16 in- This translocation may lead to subsequent multi-
jected sodium salicylate into the cisterna magna organ dysfunction and death27 (Fig. 2).
of spontaneously breathing sheep, causing a
marked increase in minute ventilation and alve- Cl inic a l M a nagemen t
olar overdistention with each breath. Hypoxemia
developed in the animals, along with stiff lungs The recognition of the importance of ventilator-
and severe morphologic pulmonary derangements induced lung injury has led to a marked change
consistent with lung injury observed during me- in the philosophy underlying the provision of
chanical ventilation. Such effects did not develop mechanical ventilation. Whereas previously the
in animals that were treated with sodium salicy- goals of mechanical ventilation were to maintain
late but underwent controlled ventilation without gas exchange while minimizing the work of
excessive lung stretching. breathing, an additional goal has been estab-
lished: to provide gas exchange that sustains life
Ventilation at Low Lung Volumes while minimizing ventilator-induced lung injury.
Ventilation that occurs at low (absolute) lung vol- In practice, this means that setting the venti-
umes can also cause injury through multiple lator often entails difficult tradeoffs. For exam-
mechanisms, including repetitive opening and ple, is it better to use a smaller tidal volume and
closing of airways and lung units,17,18 effects on let the partial pressure of arterial carbon dioxide
surfactant function,19 and regional hypoxia. This (Paco2) increase despite the associated risks (e.g.,
type of injury, which is characterized by epithe- increased intracranial hypertension from respi-
lial sloughing, hyaline membranes, and pulmo- ratory acidosis) or use larger tidal volumes to nor-
nary edema, has been termed “atelectrauma.”17 malize the Paco2 but increase the risk of lung
Atelectrauma is amplified in lungs in which injury? Whereas previously the answer might have
there are marked heterogeneities in ventilation. been to increase the tidal volume, current phi-
In a classic study, Mead et al.20 noted that the losophy has shifted to a stronger focus on pro-

149 november 28, 2013
The New England

Medical personalAll
userights
reserved.
Source: The New England Journal of Medicine critical care medicine
tection of the lung with the use of smaller tidal dressed these tradeoffs in patients with ARDS
volumes. and concluded that a higher PEEP was associated
with an absolute reduction of 5 percentage points
Ventilation Strategies in the rate of death among patients who had worse
Various ventilation strategies have been used to oxygenation, defined as a ratio of the partial pres-
minimize lung injury: low tidal volumes to limit sure of arterial oxygen to the fraction of inspired
overdistention, higher PEEPs to prevent injury oxygen (Pao2:Fio2) of 200 mm Hg or less.31
from low lung volume (atelectrauma), and recruit- Given the importance of transpulmonary pres-
ment maneuvers (i.e., procedures that are used to sure in lung injury, an obvious approach would be
reinflate collapsed lung units) that involve sus- to use transpulmonary pressure to set the PEEP,
tained application of an airway pressure of more with the use of esophageal pressure as a surrogate
than approximately 35 cm of water. The increase for pleural pressure. However, the interpretation
in pressure can inflate atelectatic lung regions of absolute esophageal pressure is difficult be-
and minimize ventilation heterogeneity. Studies cause of cardiac artifacts, the uneven distribu-
addressing these interventions are summarized tion of pleural pressure (i.e., no single value of
briefly below. pleural pressure describes the entire lung), and
esophageal distortion and contraction (especially
Low Tidal Volumes in supine patients).9 Nevertheless, this approach
Patients with ARDS often have relatively non- has been studied in patients with ARDS. In a pilot
aerated dependent lung regions (i.e., regions that study, Talmor et al.32 set the PEEP to achieve an
are lower from a gravitational perspective than end-expiratory transpulmonary pressure of 0 to
other regions and hence are more likely to be col- 10 cm of water, while limiting end-inspiratory
lapsed) and relatively normally aerated nondepen- transpulmonary pressure to 25 cm of water. They
dent lung regions. Because there is a smaller vol- found improved oxygenation and a trend toward
ume available for ventilation, this condition has lower 28-day mortality. These data are promis-
led to the term “baby lung.”28 The implication is ing, but a larger trial that shows improved clini-
that a decreased tidal volume (i.e., one that might cally important outcomes would be needed be-
be normal for a baby) should be used to prevent fore this approach could be recommended.
overinflation of the relatively small, normally aer- Recruitment maneuvers should theoretically
ated regions. In a seminal study that built on pre- reduce ventilator-induced lung injury.33 Although
vious studies,29,30 the ARDS Network investiga- such maneuvers were used in some trials that
tors compared a control strategy that used a tidal were included in the meta-analysis described
volume of 12 ml per kilogram of predicted body above31 and were implemented in a protective
weight with a low-tidal-volume strategy that used strategy that increased the number of lungs re-
6 ml per kilogram of predicted body weight.7 The trieved from heart-beating donors,34 the role of
low-tidal-volume strategy was associated with an recruitment maneuvers in clinical practice re-
absolute reduction of 9 percentage points in the mains uncertain because of questions about its
rate of death (39.8% vs. 31.0%). effect on outcomes and concerns regarding com-
plications (e.g., hemodynamic compromise or
High PEEP and Recruitment Maneuvers pneumothorax).35
Pulmonary edema and end-expiratory alveolar col-
lapse characterize several forms of respiratory High-Frequency Oscillatory Ventilation
failure. In these situations, a low PEEP may be High-frequency oscillatory ventilation (HFOV) is
insufficient to stabilize alveoli and keep them a technique in which very small tidal volumes
open, thereby increasing the likelihood of venti- (sometimes less than the anatomic dead space)
lator-induced lung injury from atelectrauma. Con- are applied at high frequencies (up to 15 per sec-
versely, a higher PEEP has potentially adverse ef- ond). Theoretically, this technique should be ideal
fects, including impairment of venous return and for minimizing ventilator-induced lung injury.36
pulmonary overdistention. A recent meta-analysis In a meta-analysis of eight randomized, con-
of patient-level data in randomized trials31 ad- trolled trials involving a total of 419 adults with
n engl j med 369;22 nejm.org november 28, 2013 2131

150
The New
Medicine
Medical 2014. For
rights reserved.
ARDS,37 HFOV-treated patients had significantly instead use extracorporeal membrane oxygenation
lower mortality than did patients treated with (ECMO).46 It is also possible to combine mechan-
conventional ventilation (risk ratio, 0.77; P = 0.03),
ical ventilation with partial extracorporeal sup-
which suggested that HFOV might improve sur- port; with this approach, the intensity of ventila-
vival and is unlikely to cause harm.37 However, tion that is needed to sustain life is decreased,
since two recent large multicenter trials involv- and carbon dioxide is removed through an extra-
ing patients with ARDS did not show improved corporeal circuit.47 The advantages of this hybrid
outcomes with HFOV,38,39 this type of ventilation strategy are a decreased rate of complications, as
cannot be recommended as first-line therapy in compared with full ECMO, and a decreased rate
such patients.40 of lung injury because tidal volumes can be re-
duced. Preliminary data have supported this ap-
Adjunctive Strategies proach,48,49 but further studies are required to
One goal of mechanical ventilation is to help show which mode of extracorporeal support to
meet the gas-exchange demands of the patient. use, when to apply it, and which, if any, patients
Thus, one nonspecific approach that might limit might benefit.
ventilator-induced lung injury is to decrease a pa-
tient’s metabolic demands, thereby decreasing Pharmacologic Interventions
the required minute ventilation and decreasing Neuromuscular Blocking Agents
breathing efforts. Other specific approaches are Because of extreme dyspnea, patients with ARDS
discussed below. often “fight the ventilator,” which may aggravate
ventilator-induced lung injury.50 One therapeu-
Prone Position tic approach is to administer a neuromuscular
About 70% of patients with ARDS and hypoxemia
have improved oxygenation when they are placed
Figure 3 (facing page). Ventilatory Strategies.
in a prone position.41 Possible mechanisms for
Shown are strategies for the use of a ventilator in a
this effect include increased end-expiratory lung patient with ARDS (Panel A), a heart-beating organ donor
volume, better ventilation–perfusion matching, (Panel B), and a patient with normal lungs (Panel C).
less effect of the mass of the heart on the lower A protective ventilation strategy is defined as one in
lobes, and improved regional ventilation.41 Most which the goal is to minimize the injury that can be
important, as has been shown in studies in ani- caused by mechanical ventilation; components of this
strategy include minimization of end-inspiratory stretch-
mals,42,43 the prone position should minimize lung ing and minimization of injury caused by ventilation at
injury by increasing homogeneity of ventilation. low lung volumes. A protective lung strategy includes a
A recent meta-analysis44 of seven trials involv- protective ventilation strategy plus approaches to mini-
ing a total of 1724 patients showed that prone mize derecruitment of the lung (e.g., the use of contin-
positioning lowered absolute mortality by ap- uous positive airway pressure during apnea tests and
the use of closed circuits during suctioning). There is
proximately 10 percentage points in the subgroup currently no evidence showing that any mode of venti-
of patients with ARDS and severe hypoxemia lation is better than any other in delivering the tidal
(Pao2:Fio2 ratio, <100 mm Hg). Patients who were volume of 6 ml per kilogram of predicted body weight
treated with prone positioning had an increased (PBW) or limiting the plateau pressure. There is less
number of potentially preventable complications, evidence for the strategies for heart-beating organ do-
nors and patients with normal lungs in the intensive
including pressure ulcers, endotracheal-tube ob- care unit (ICU) than for the strategies for patients with
struction, and chest-tube dislodgement. In a ARDS7 and for anesthetized patients undergoing major
recent ly completed trial involving 466 patients abdominal surgery.56 Rescue therapy refers to treat-
with ARDS who had a Pao2:Fio2 ratio of less than ments that may improve oxygenation in life-threaten-
150 mm Hg while receiving an Fio2 of 0.60 or ing situations but for which there are insufficient data
clearly showing improved clinical outcomes. Some of
more, the rate of death at 28 days was 32.8% these treatments have been shown to be ineffective in
among those who were treated in the supine posi- terms of clinical outcomes (e.g., the use of nitric oxide
tion and 16.0% among those treated in the prone and high-frequency ventilation), whereas others have
position.45 not been adequately evaluated (e.g., extracorporeal sup-
port). Their use should be carefully considered before
they are implemented. PEEP denotes positive end-
Partial or Total Extracorporeal Support
expiratory pressure, and P/F ratio of the partial pressure
One approach to preventing ventilator-induced of arterial oxygen to the fraction of inspired oxygen.
lung injury is to avoid mechanical ventilation and
2132 n engl j med 369;22 151

nejm.org november 28, 2013
The New England

Medical personalAll
userights
only. reserved.
A B C
• Patient with normal lungs in ICU
Patient with ARDS Heart-beating organ donor • Anesthetized patient undergoing
major abdominal surgery, at high risk
for complications
Protective lung strategy Protective lung strategy Protective ventilation strategy
• Tidal volume = 6 ml/kg PBW • Tidal volume = 6–8 ml/kg PBW • Tidal volume = 6–8 ml/kg PBW
• Plateau pressure ≤≤30
30 cm
cm of
of water
water • Plateau pressure <20 cm of water
• PEEP and inspired oxygen fraction table • PEEP at 88–10
10 cm of water • PEEP at 44–8 cm of water
• Recruitment maneuvers every
30 minutes for anesthetized patients
• Apnea tests using continuous

positive airway pressure
• Closed airway suctioning

• Recruitment maneuvers after
disconnecting Caveat:
If patient has acute brain
injury, monitor for
respiratory acidosis
If patient has stiff chest wall If P/F≤ 200 mm Hg

(e.g., with massive ascites or
pleural effusion)consider
• Allowing plateau pressure >30

• Use of higher PEEP levels
• Measuring esophageal pressure Consider use of higher
to set pressures If P/F <150 mm Hg
PEEP strategy
Consider using prone position Consider using neuromuscular

if staff are well trained blocking agent if early ARDS
in procedure (approximately <48 hours)
If patient still is in distress or

has extreme hypoxemia,
consider rescue therapies
COLOR FIGURE

november 28, 2013 Draft 6 2133
11/8/13
Author Slutsky
The New England
personal use only. No other uses without 3 permission.
Fig #
Medical personalAll
userights
only. reserved.
Title Ventilator Induced Lung
Injury
Back to Table of Contents ME
DE Drazen
Artist N Koscal
blocking agent to ensure patient–ventilator syn- patient has a markedly stiff chest wall (e.g., be-
chrony and facilitate limitations on tidal volume cause of massive ascites), the lungs may not be
and pressure. In a recent multicenter, placebo- overdistended at 30 cm of water or even at higher
controlled, randomized trial involving 340 pa- plateau pressures (Fig. 1C). In this situation, in
tients with ARDS and a Pao2:Fio2 ratio of less than a patient with hypoxemia, higher plateau pres-
150 mm Hg, Papazian et al.51 found that the ad- sures may be appropriate, even though the ARDS
justed 90-day mortality was lower among those Network study did not recommend such adjust-
who received a neuromuscular blocking agent for ments. Another possibility is to consider mea-
48 hours than among those who received placebo, suring esophageal pressure to help set the venti-
without any increase in residual muscle weak- latory strategy.
ness. The precise mechanism for the decreased When clinicians use assisted modes of venti-
mortality is unclear,50 but a previous study showed latory support (i.e., mechanical ventilation that
reduced serum cytokine levels among patients requires active contraction of respiratory muscles)
receiving a neuromuscular blocking agent.52 In or noninvasive ventilation, it is important for them
the study by Papazian et al., the divergence in to be aware that large tidal volumes may be de-
mortality between groups occurred relatively late livered despite relatively low airway pressures.
(approximately 16 days after the initiation of treat- From a theoretical perspective, all patients re-
ment), which could be explained by a decreased ceiving ventilator support should benefit from
rate of multiorgan failure due to biotrauma.27 strategies that minimize ventilator-induced lung
injury. However, it is important to bear in mind
Antiinflammatory Agents and Stem Cells the effects of any strategy to reduce lung injury
Pharmacologic interventions that are aimed at on other important physiological or clinical phe-
minimizing biotrauma have not been reported in nomena (e.g., the effect of a lung-protective strat-
humans, but antiinflammatory strategies53 and egy on hemodynamics). Figure 3 summarizes
the use of mesenchymal stem cells54 have been clinical scenarios in which lung-protective venti-
investigated in studies in animals. The key ad- lation has been shown to have substantial or pos-
vantage of such therapies in the prevention of the sible benefit. Of particular interest is whether
consequences of ventilator-induced lung injury, as patients with relatively normal lungs should
compared with their use in other inflammatory undergo ventilation with low tidal volumes. Un-
conditions (e.g., sepsis), is that the therapy could injured (normal) lungs tolerate relatively large
be delivered before the inciting agent is initiated tidal volumes delivered at relatively low pressures
(i.e., immediately before mechanical ventilation). as long as the stress and strains applied are be-
These treatments remain experimental and of un- low an injurious threshold.56 The specific thresh-
proven benefit. olds are uncertain, but a recent meta-analysis
showed that ventilation with smaller tidal vol-
umes in patients without ARDS may be associ-
A r e a s of Uncer ta in t y
a nd R ec om mendat ions ated with improved outcomes.57 Although these
data suggest that lung-protective ventilation strat-
Although the trials described above help clini- egies should be adopted widely, the ideal ventila-
cians make difficult tradeoffs, they often do not tion strategy remains to be determined, and more
address the complexity of many clinical situa- definitive studies are necessary before the use of
tions. In these cases, it is advisable for clinicians such strategies becomes standard practice.
to integrate underlying physiological principles Dr. Slutsky reports receiving payment for serving on an advi-
with trial data. For example, in the ARDS Net- sory board at Ikaria, receiving consulting fees from Gambro,
GlaxoSmithKline, Maquet Medical, Novalung, and Hemodec
work study, investigators used ventilation with a and lecture fees from Dräger, having an equity interest in
tidal volume of 6 ml per kilogram of predicted Apeiron, and receiving royalties through his institution from
Maquet Medical; and Dr. Ranieri, receiving payment for serving
body weight and limited the plateau pressure to on advisory boards at Hemodec and Maquet, receiving consult-
30 cm of water.7 However, since regional overdis- ing fees from Hemodec, Maquet, Faron Pharmaceuticals, and
tention occurs in some patients despite these set- Covidien, and receiving payment for work on a patent owned by
Dräger and the University of Turin. No other potential conflict
tings, the use of a reduced tidal volume (or re- of interest relevant to this article was reported.
duced PEEP) may be warranted.55 Disclosure forms provided by the authors are available with
Conversely, the plateau-pressure limit of 30 cm We thank Drs. Roy Brower and Rolf Hubmayr for their review
of water may be too low in some patients. If a of an earlier version of the manuscript.

153 november 28, 2013
The New EnglandFor

Copyright 29, 2014.
Medical use rights
reserved.
References
1. Lassen HC. A preliminary report on 16. Mascheroni D, Kolobow T, Fumagalli strategy on mortality in the acute respira-
the 1952 epidemic of poliomyelitis in Co- R, Moretti MP, Chen V, Buckhold D. Acute tory distress syndrome. N Engl J Med 1998;
penhagen with special reference to the respiratory failure following pharmaco- 338:347-54.
treatment of acute respiratory insufficien- logically induced hyperventilation: an ex- 31. Briel M, Meade M, Mercat A, et al.
cy. Lancet 1953;1:37-41. perimental animal study. Intensive Care Higher vs lower positive end-expiratory
2. Nash G, Blennerhassett JB, Pontoppi- Med 1988;15:8-14. pressure in patients with acute lung injury
dan H. Pulmonary lesions associated with 17. Slutsky AS. Lung injury caused by and acute respiratory distress syndrome:
oxygen therapy and artificial ventilation. mechanical ventilation. Chest 1999;116: systematic review and meta-analysis. JAMA
N Engl J Med 1967;276:368-74. 1 Suppl:9S-15S. 2010;303:865-73.
3. Macklin MT, Macklin CC. Malignant 18. Bilek AM, Dee KC, Gaver DP III. 32. Talmor D, Sarge T, Malhotra A, et al.
interstitial emphysema of the lungs and Mechanisms of surface-tension-induced Mechanical ventilation guided by esopha-
mediastinum as an important occult com- epithelial cell damage in a model of pul- geal pressure in acute lung injury. N Engl
plication in many respiratory diseases and monary airway reopening. J Appl Physiol J Med 2008;359:2095-104.
other conditions: an interpretation of the 2003;94:770-83. 33. Lachmann B. Open up the lung and
clinical literature in the light of laboratory 19. Albert RK. The role of ventilation- keep the lung open. Intensive Care Med
experiment. Medicine 1944;23:281-358. induced surfactant dysfunction and atel- 1992;18:319-21.
4. Avignon PD, Hedenstrom G, Hedman ectasis in causing acute respiratory dis- 34. Mascia L, Pasero D, Slutsky AS, et al.
C. Pulmonary complications in respirator tress syndrome. Am J Respir Crit Care Effect of a lung protective strategy for or-
patients. Acta Med Scand Suppl 1956;316: Med 2012;185:702-8. gan donors on eligibility and availability of
86-90. 20. Mead J, Takishima T, Leith D. Stress lungs for transplantation: a randomized
5. Respirator lung syndrome. Minn Med distribution in lungs: a model of pulmo- controlled trial. JAMA 2010;304:2620-7.
1967;50:1693-705. nary elasticity. J Appl Physiol 1970;28:596- 35. Brower RG, Morris A, MacIntyre N,
6. Fothergill J. Observations of a case 608. et al. Effects of recruitment maneuvers in
published in the last volume of the medi- 21. Tremblay L, Valenza F, Ribeiro SP, Li J, patients with acute lung injury and acute
cal essays of recovering a man dead in Slutsky AS. Injurious ventilatory strategies respiratory distress syndrome ventilated
appearance, by distending the lungs with increase cytokines and c-fos m-RNA ex- with high positive end-expiratory pressure.
air. Philosophical Transactions 1744;43: pression in an isolated rat lung model. Crit Care Med 2003;31:2592-7. [Erratum,
275-81. J Clin Invest 1997;99:944-52. Crit Care Med 2004;32:907.]
7. The Acute Respiratory Distress Syn- 22. Cabrera-Benítez NE, Parotto M, Post 36. Ferguson ND, Slutsky AS. High-fre-
drome Network. Ventilation with lower M, et al. Mechanical stress induces lung quency ventilation is the optimal physio-
tidal volumes as compared with tradition- fibrosis by epithelial-mesenchymal tran- logical approach to ventilate ARDS pa-
al tidal volumes for acute lung injury and sition. Crit Care Med 2012;40:510-7. tients. J Appl Physiol 2008;104:1230-1.
the acute respiratory distress syndrome. 23. Tremblay LN, Slutsky AS. Ventilator- 37. Sud S, Sud M, Friedrich JO, et al. High
N Engl J Med 2000;342:1301-8. induced injury: from barotrauma to bio- frequency oscillation in patients with
8. Rahn H, Otis AB, Chadwick LE, Fenn trauma. Proc Assoc Am Physicians 1998; acute lung injury and acute respiratory dis-
WO. The pressure-volume diagram of the 110:482-8. tress syndrome (ARDS): systematic review
thorax and lung. Am J Physiol 1946;146: 24. Ranieri VM, Suter PM, Tortorella C, and meta-analysis. BMJ 2010;340:c2327.
161-78. et al. Effect of mechanical ventilation on 38. Ferguson ND, Cook DJ, Guyatt GH,
9. Zin WA, Milic-Emili J. Esophageal pres- inflammatory mediators in patients with et al. High-frequency oscillation in early
sure measurement. In: Hamid Q, Shan- acute respiratory distress syndrome: a ran- acute respiratory distress syndrome. N Engl
non J, Martin J, eds. Physiologic basis of domized controlled trial. JAMA 1999;282: J Med 2013;368:795-805.
respiratory disease. Hamilton, ON, Canada: 54-61. 39. Young D, Lamb SE, Shah S, et al.
BC Decker, 2005:639-47. 25. Nahum A, Hoyt J, Schmitz L, Moody J, High-frequency oscillation for acute respi-
10. Bouhuys A. Physiology and musical Shapiro R, Marini JJ. Effect of mechanical ratory distress syndrome. N Engl J Med
instruments. Nature 1969;221:1199-204. ventilation strategy on dissemination of 2013;368:806-13.
11. Ware LB, Matthay, MA. The acute re- intratracheally instilled Escherichia coli in 40. Malhotra A, Drazen JM. High-frequen-
spiratory distress syndrome. N Engl J Med dogs. Crit Care Med 1997;25:1733-43. cy oscillatory ventilation on shaky ground.
2000;342:1334-49. 26. Murphy DB, Cregg N, Tremblay L, et al. N Engl J Med 2013;368:863-5.
12. Slutsky AS, Hudson LD. PEEP or no Adverse ventilatory strategy causes pulmo- 41. Piedalue F, Albert RK. Prone position-
PEEP — lung recruitment may be the so- nary-to-systemic translocation of endo- ing in acute respiratory distress syndrome.
lution. N Engl J Med 2006;354:1839-41. toxin. Am J Respir Crit Care Med 2000;162: Respir Care Clin N Am 2003;9:495-509.
13. Maunder RJ, Pierson DJ, Hudson LD. 27-33. [Erratum, Am J Respir Crit Care Med 42. Broccard A, Shapiro RS, Schmitz LL,
Subcutaneous and mediastinal emphy- 2002;166:1517.] Adams AB, Nahum A, Marini JJ. Prone
sema: pathophysiology, diagnosis, and 27. Slutsky AS, Tremblay LN. Multiple positioning attenuates and redistributes
management. Arch Intern Med 1984;144: system organ failure: is mechanical venti- ventilator-induced lung injury in dogs.
1447-53. lation a contributing factor? Am J Respir Crit Care Med 2000;28:295-303.
14. Webb HH, Tierney DF. Experimental Crit Care Med 1998;157:1721-5. 43. Valenza F, Guglielmi M, Maffioletti M,
pulmonary edema due to intermittent pos- 28. Gattinoni L, Pesenti A. The concept of et al. Prone position delays the progres-
itive pressure ventilation with high infla- “baby lung.” Intensive Care Med 2005;31: sion of ventilator-induced lung injury in
tion pressures: protection by positive end- 776-84. rats: does lung strain distribution play a
expiratory pressure. Am Rev Respir Dis 29. Hickling KG, Henderson SJ, Jackson R. role? Crit Care Med 2005;33:361-7.
1974;110:556-65. Low mortality associated with low volume 44. Sud S, Friedrich JO, Taccone P, et al.
15. Dreyfuss D, Soler P, Basset G, Saumon pressure limited ventilation with permis- Prone ventilation reduces mortality in pa-
G. High inflation pressure pulmonary sive hypercapnia in severe adult respira- tients with acute respiratory failure and
edema: respective effects of high airway tory distress syndrome. Intensive Care Med severe hypoxemia: systematic review and
pressure, high tidal volume, and positive 1990;16:372-7. meta-analysis. Intensive Care Med 2010;
end-expiratory pressure. Am Rev Respir 30. Amato MB, Barbas CS, Medeiros DM, 36:585-99.
Dis 1988;137:1159-64. et al. Effect of a protective-ventilation 45. Guérin C, Reignier J, Richard J-C, et al.
n engl j med 369;22 154

The New
use only. Medicine
rights reserved.
Prone positioning in severe acute respi- tracorporeal CO2 removal versus “conven- Mesenchymal stem cells enhance recovery
ratory distress syndrome. N Engl J Med tional” protective ventilation (6 ml/kg) in and repair following ventilator-induced
2013;368:2159-68. severe ARDS: the prospective randomized lung injury in the rat. Thorax 2012;67:496-
46. Brodie D, Bacchetta M. Extracorporeal Xtravent-study. Intensive Care Med 2013; 501.
membrane oxygenation for ARDS in adults. 39:847-56. 55. Grasso S, Stripoli T, De Michele M, et al.
N Engl J Med 2011;365:1905-14. 50. Slutsky AS. Neuromuscular blocking ARDSnet ventilatory protocol and alveolar
47. Gattinoni L, Kolobow T, Agostoni A, agents in ARDS. N Engl J Med 2010;363: hyperinflation: role of positive end-expira-
et al. Clinical application of low frequency 1176-80. tory pressure. Am J Respir Crit Care Med
positive pressure ventilation with extra- 51. Papazian L, Forel JM, Gacouin A, et al. 2007;176:761-7.
corporeal CO2 removal (LFPPV-ECCO2R) Neuromuscular blockers in early acute re- 56. Futier E, Constantin J-M, Paugam-
in treatment of adult respiratory distress spiratory distress syndrome. N Engl J Med Burtz C, et al. A trial of intraoperative
syndrome (ARDS). Int J Artif Organs 1979; 2010;363:1107-16. low-tidal-volume ventilation in abdominal
2:282-3. 52. Forel JM, Roch A, Marin V, et al. Neuro- surgery. N Engl J Med 2013;369:428-37.
48. Terragni PP, Del Sorbo L, Mascia L, et muscular blocking agents decrease inflam- 57. Serpa Neto A, Cardoso SO, Manetta
al. Tidal volume lower than 6 ml/kg en- matory response in patients presenting JA, et al. Association between use of lung-
hances lung protection: role of extracor- with acute respiratory distress syndrome. protective ventilation with lower tidal vol-
poreal carbon dioxide removal. Anesthe- Crit Care Med 2006;34:2749-57. umes and clinical outcomes among pa-
siology 2009;111:826-35. 53. Uhlig S, Uhlig U. Pharmacological in- tients without acute respiratory distress
49. Bein T, Weber-Carstens S, Goldmann terventions in ventilator-induced lung inju- syndrome: a meta-analysis. JAMA 2012;
A, et al. Lower tidal volume strategy (ap- ry. Trends Pharmacol Sci 2004;25:592-600. 308:1651-9.
proximately 3 ml/kg) combined with ex- 54. Curley GF, Hayes M, Ansari B, et al. Copyright © 2013 Massachusetts Medical Society.
Lioness, Blind in One Eye, with Loss of Depth Perception (and Prey), Okavango Delta, Botswana Jayesh Mehta, M.D.

155 november 28, 2013
The New England

Downloaded from nejm.org by MJ MEDAS ©
Copyright onMassachusetts
September 29,Medical
2014. For personal
reserved.
c or r e sp ondence

To the Editor: Slutsky and Ranieri (Nov. 28 is- but not ventilator-associated lung injury. As
sue)1 note that ventilator-induced lung injury may compared with injury induced by mechanical
be minimized by means of lung-protective me- ventilation, ventilator-associated lung injury is
chanical ventilation or by avoiding it altogether defined as lung injury in patients with preexist-
with the use of extracorporeal life-support tech- ing lung disease1 (such as the acute respiratory
niques2 such as by extracorporeal removal of car- distress syndrome) who are undergoing mechan-
bon dioxide or extracorporeal membrane oxygen- ical ventilation.
ation (ECMO). Are there differences in the lung-protective
However, there is no consensus regarding ven- strategies2,3 that should be used to prevent and
tilator settings during extracorporeal respiratory ameliorate ventilator-induced and ventilator-
support. For example, with removal of extracor- associated lung injury?
poreal carbon dioxide, there is usually some S. Mohammadreza Hashemian, M.D.
degree of mechanical ventilation, whereas pa- Shahid Beheshti University of Medical Sciences
tients undergoing ECMO can receive support Tehran, Iran
without mechanical ventilation. While we await smrhashemian@sbmu.ac.ir
the results of controlled trials regarding this Seyed A. Mohajerani, M.D.

question,3,4 are the authors aware of data to help Chronic Respiratory Diseases Research Center
clinicians decide how to set a ventilator when a Tehran, Iran
patient is receiving extracorporeal respiratory
Hamidreza R. Jamaati, M.D.
support?
National Research Institute of Tuberculosis and Lung Disease
Tehran, Iran
Kiran Shekar, M.B., B.S.
John F. Fraser, Ph.D. ported.
Prince Charles Hospital 1. Pinhu L, Whitehead T, Evans T, Griffiths M. Ventilator-asso-

Brisbane, QLD, Australia ciated lung injury. Lancet 2003;361:332-40.
kiran_shekar@health.qld.gov.au 2. Fan E, Villar J, Slutsky AS. Novel approaches to minimize
No potential conflict of interest relevant to this letter was re- ventilator-induced lung injury. BMC Med 2013;11:85.
ported. 3. Bein T, Weber-Carstens S, Goldmann A, et al. Lower tidal
volume strategy (≈3 ml/kg) combined with extracorporeal CO2
1. Slutsky AS, Ranieri VM. Ventilator-induced lung injury. removal versus ‘conventional’ protective ventilation (6 ml/kg) in
N Engl J Med 2013;369:2126-36. severe ARDS: the prospective randomized Xtravent study. In-
2. Sorbo LDCM, Fan E. Extracorporeal life support for adults tensive Care Med 2013;39:847-56.
with severe acute respiratory failure. Lancet Respir Med 2013.
DOI: 10.1056/NEJMc1400293
DOI: 10.1016/S2213-2600(13)70197-8.
3. Strategies for Optimal Lung Ventilation in ECMO for
ARDS: the SOLVE ARDS Study (http://clinicaltrials.gov/ct2/
show/NCT01990456). The Authors Reply: Shekar and Fraser are cor-
rect in noting that there are few outcome data
4. Extracorporeal Membrane Oxygenation for Severe Acute Re-
spiratory Distress Syndrome (EOLIA) (http://clinicaltrials.gov/
on how best to set a ventilator when patients are
show/NCT01470703).
receiving ECMO. The specific strategy will de-
DOI: 10.1056/NEJMc1400293
pend on how ECMO is being used. For example,
low-flow ECMO (e.g., removal of extracorporeal
To the Editor: In their review, Slutsky and carbon dioxide) is increasingly being studied as
Ranieri discuss ventilator-induced lung injury a means of removing a portion of the patient’s
972 n engl j med 370;10 156

nejm.org march 6, 2014
Clinical Collections — Critical Care correspondence
metabolic carbon dioxide production to allow ventilatory strategy. For this reason, many au-
less injurious ventilation (i.e., to minimize ven- thors use the term “ventilator-associated lung
tilator-induced lung injury); in this situation, injury” when referring to humans. Nonethe-
the goal is to use a lower tidal volume (e.g., 3 ml less, the basic underlying mechanisms causing
per kilogram of predicted body weight), plateau lung injury are the same for both ventilator-
pressure (e.g., <25 cm of water), or both while induced lung injury and ventilator-associated
removing adequate amounts of arterial carbon lung injury, and hence the lung-protective strat-
dioxide.1,2 During full extracorporeal support in egies should be based on similar principles.
adults, the ventilator settings recommended by Arthur S. Slutsky, M.D.
the Extracorporeal Life Support Organization
St. Michael’s Hospital
are to minimize ventilator support with the use Toronto, ON, Canada
of a respiratory rate of 4 or 5 breaths per min- slutskya@smh.ca
ute, modest levels of positive end-expiratory V. Marco Ranieri, M.D.
pressure (PEEP) (e.g., 10 cm of water), and low Ospedale S. Giovanni Battista Molinette
inflation pressures (e.g., 10 cm of water above Turin, Italy
the PEEP).3 Other groups have reported venti- Since publication of their article, the authors report no fur-
lator strategies with somewhat higher PEEP lev- ther potential conflict of interest.
els (10 to 20 cm of water) to more fully recruit 1. Terragni PP, Del Sorbo L, Mascia L, et al. Tidal volume lower
the lung.4,5 than 6 ml/kg enhances lung protection: role of extracorporeal
Hashemian et al. ask whether there are carbon dioxide removal. Anesthesiology 2009;111:826-35.
2. Bein T, Weber-Carstens S, Goldmann A, et al. Lower tidal
differences in lung-protective strategies for volume strategy (≈3 ml/kg) combined with extracorporeal CO2
ventilator-induced lung injury versus ventilator- removal versus ‘conventional’ protective ventilation (6 ml/kg) in
associated lung injury. Our view of the defini- severe ARDS: the prospective randomized Xtravent study. Inten-
sive Care Med 2013;39:847-56.
tion of ventilator-associated lung injury is 3. Extracorporeal Life Support Organization (ELSO) Patient
somewhat different than that presented by Specific Supplements to the ELSO General Guidelines. April 2009
Hashemian and colleagues. We suggest that the (http://www.elso.med.umich.edu/WordForms/ELSO%20Pt%20
Specific%20Guidelines.pdf).
injury caused by the ventilator is always venti- 4. Patroniti N, Zangrillo A, Pappalardo F, et al. The Italian
lator-induced lung injury. In studies in animals, ECMO network experience during the 2009 influenza A(H1N1)
this form of injury is (relatively) easy to iden- pandemic: preparation for severe respiratory emergency out-
breaks. Intensive Care Med 2011;37:1447-57.
tify, since one has a control group as a com- 5. Pham T, Combes A, Rozé H, et al. Extracorporeal membrane
parison. In patients, however, it is often more oxygenation for pandemic influenza A(H1N1)-induced acute re-
problematic to determine how much of the spiratory distress syndrome: a cohort study and propensity-
matched analysis. Am J Respir Crit Care Med 2013;187:276-85.
observed lung injury is due to the underlying
disease process and how much is due to the DOI: 10.1056/NEJMc1400293
n engl j med 370;10 nejm.org march 6, 2014 973

157

What mechanical ventilation strategy would provide the best support for this patient?
A. Continue with the current ventilator settings, but increase the respiratory rate and tidal volume to normalize the
PaCO2.
B. Continue with the current ventilator mode, but decrease the tidal volume to 360 ml and increase the respiratory
rate to 22 breaths per minute.
C. Change the ventilator mode to pressure-controlled synchronized intermittent mandatory ventilation with pres-
sure support (SIMV/PS) with a PEEP of 10 cm of water, with a pressure control of 30 cm of water and pressure
support of 20 cm of water above the PEEP.
D. Continue with the current ventilator mode, increase the PEEP to 10 cm of water, increase the respiratory rate to
22 breaths per minute, and decrease the current tidal volume slightly to keep the plateau pressure at 30 cm of
water or less.

The patient who is described in the vignette satisfies the diagnostic criteria for the acute respiratory distress syn-
drome (ARDS), since he has severe sepsis (a recognized risk factor for ARDS), bilateral alveolar infiltrates, and a
PaO2:FiO2 (P:F) ratio of less than 200 mm Hg; the hemodynamic data suggest that he does not have cardiogenic
pulmonary edema. In recent years, there has been a broad consensus that mechanical ventilation of patients with
ARDS should focus on preventing ventilator-induced lung injury through the avoidance of excessive lung stretch-
ing and high airway pressures rather than on normalizing gas exchange. Such strategies have been shown to re-
duce mortality.1
In a seminal study, the ARDS Network investigators found that a low-tidal-volume strategy targeting a tidal
volume of 6 ml per kilogram of predicted body weight and a plateau airway pressure of 30 cm of water or less
reduced mortality, as compared with a control strategy using a tidal volume of 12 ml per kilogram.1 This study
confirmed previously published observational data and the findings of a smaller randomized trial. Early ARDS
is characterized by atelectasis affecting the dependent portions of the lung, with the delivered tidal volume dis-
tributed to a smaller-than-usual volume of still aerated lung.2 At least some of the atelectatic-dependent lung is
“recruitable,” meaning that it can potentially be reaerated, resulting in better gas exchange and reducing the stress
placed on ventilated lung units. As a result, other ventilation strategies, such as the used of higher levels of PEEP,3,4
recruitment maneuvers,4 or the placement of patients in the supine position alternating with the prone position,5
have been investigated both to recruit atelectatic lung and to reduce mortality. Currently, there are no consistent
outcome data to support the use of levels of PEEP that are higher than those used in the ARDS Network study of
low tidal volumes or to support the routine use of lung-recruitment maneuvers.
The patient should undergo mechanical ventilation with the use of a strategy designed to minimize ventilator-
induced lung injury. In this patient, it would be reasonable to target a tidal volume of 440 ml, since although his
actual weight is 60 kg, his predicted body weight is 73 kg (the size of the lungs is proportional to the predicted
body weight); a slight decrease in tidal volume would be reasonable to limit the plateau airway pressure to 30 cm
of water and to increase the PEEP to 10 cm of water in view of the severity of the hypoxemic respiratory failure.
158
A recent trial showed a marked reduction in mortality among patients with severe ARDS (defined as a P:F ratio of
<150 mm Hg after at least 12 hours of optimized ventilation) who were placed in the prone position for sessions of
at least 16 hours, as compared with patients who were placed in the supine position.6 Treating patients in the prone
position poses many medical and nursing challenges. Thus, it should be considered in patients with severe ARDS
who have not had a response to treatment with standard ventilator approaches but only in units in which the rele-
vant staff members have training and experience in its use.
References
1. The Acute Respiratory Distress Syndrome Network. Ventilation with lower tidal volumes as compared with
traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. N Engl J Med
2000;342(18):1301–1308.
2. Gattinoni L, Caironi P, Cressoni M, et al. Lung recruitment in patients with the acute respiratory distress
syndrome. N Engl J Med 2006;354(17):1775–1786.
3. The National Heart, Lung, and Blood Institute ARDS Clinical Trials Network. Higher versus lower positive
end-expiratory pressures in patients with the acute respiratory distress syndrome. N Engl J Med
2004;351(4):327–336.
4. Meade MO, Cook DJ, Guyatt GH, et al. Ventilation strategy using low tidal volumes, recruitment maneuvers,
and high positive end-expiratory pressure for acute lung injury and acute respiratory distress syndrome: a
randomized controlled trial. JAMA 2008;299(6):637–645.
5. Taccone P, Pesenti A, Latini R, et al. Prone positioning in patients with moderate and severe acute respiratory
distress syndrome: a randomized controlled trial. JAMA 2009;302(18):1977–1984.
6. Guérin C, Reignier J, Richard J-C, et al. Prone positioning in severe acute respiratory distress syndrome.
N Engl J Med 2013;368(23):2159–2168.
159
Special Article
Driving Pressure and Survival in the Acute

Respiratory Distress Syndrome
Marcelo B.P. Amato, M.D., Maureen O. Meade, M.D., Arthur S. Slutsky, M.D.,
Laurent Brochard, M.D., Eduardo L.V. Costa, M.D., David A. Schoenfeld, Ph.D.,
Thomas E. Stewart, M.D., Matthias Briel, M.D., Daniel Talmor, M.D., M.P.H.,
Alain Mercat, M.D., Jean-Christophe M. Richard, M.D.,
Carlos R.R. Carvalho, M.D., and Roy G. Brower, M.D.
A BS T R AC T
BACKGROUND
Mechanical-ventilation strategies that use lower end-inspiratory (plateau) airway From the Cardio-Pulmonary Depart-
pressures, lower tidal volumes (VT), and higher positive end-expiratory pressures ment, Pulmonary Division, Heart Insti-
tute (Incor), University of São Paulo
(PEEPs) can improve survival in patients with the acute respiratory distress syndrome (M.B.P.A., E.L.V.C., C.R.R.C.), and the
(ARDS), but the relative importance of each of these components is uncertain. Research and Education Institute, Hospi-
Because respiratory-system compliance (CRS) is strongly related to the volume of tal Sirio-Libanês (E.L.V.C.) — both in São
Paulo; the Departments of Clinical Epide-
aerated remaining functional lung during disease (termed functional lung size), miology and Biostatistics and Medicine,
we hypothesized that driving pressure (ΔP = VT/CRS), in which VT is intrinsically nor- McMaster University, Hamilton, ON
malized to functional lung size (instead of predicted lung size in healthy persons), (M.O.M., T.E.S., M.B.), and the Keenan
Research Centre for Biomedical Science,
would be an index more strongly associated with survival than VT or PEEP in patients St. Michael’s Hospital (A.S.S., L.B.), and
who are not actively breathing. the Interdepartmental Division of Critical
Care Medicine and Department of Medi-
cine, University of Toronto (A.S.S., L.B.),
METHODS Toronto — all in Canada; the Massachu-
Using a statistical tool known as multilevel mediation analysis to analyze individual setts General Hospital Biostatistics Cen-
data from 3562 patients with ARDS enrolled in nine previously reported randomized ter, Harvard Medical School (D.A.S.),
and Department of Anesthesia, Critical
trials, we examined ΔP as an independent variable associated with survival. In the Care, and Pain Medicine, Beth Israel
mediation analysis, we estimated the isolated effects of changes in ΔP resulting Deaconess Medical Center and Harvard
from randomized ventilator settings while minimizing confounding due to the Medical School (D.T.) — both in Boston;
the Basel Institute for Clinical Epidemiol-
baseline severity of lung disease. ogy and Biostatistics, University Hospi-
tal Basel, Basel, Switzerland (M.B.); the
RESULTS Department of Intensive Care and Hyper-
baric Medicine, Angers University Hospi-
Among ventilation variables, ΔP was most strongly associated with survival. A 1-SD tal, Angers (A.M.), the Emergency De-
increment in ΔP (approximately 7 cm of water) was associated with increased partment, General Hospital of Annecy,
mortality (relative risk, 1.41; 95% confidence interval [CI], 1.31 to 1.51; P<0.001), Annecy (J.-C.M.R.), and INSERM UMR
955, Creteil (J.-C.M.R.) — all in France;
even in patients receiving “protective” plateau pressures and VT (relative risk, 1.36; and the Division of Pulmonary and Criti-
95% CI, 1.17 to 1.58; P<0.001). Individual changes in VT or PEEP after randomiza- cal Care Medicine, Johns Hopkins Uni-
tion were not independently associated with survival; they were associated only if versity School of Medicine, Baltimore
(R.G.B.). Address reprint requests to Dr.
they were among the changes that led to reductions in ΔP (mediation effects of ΔP,
Amato at Faculdade de Medicina, Univer-
P = 0.004 and P = 0.001, respectively). sidade de São Paulo., Av. Dr. Arnaldo
455, sala 2144 (2nd Fl.), 01246-903, São
CONCLUSIONS Paulo, Brazil, or at amato.marcelo.bp@
gmail.com.
We found that ΔP was the ventilation variable that best stratified risk. Decreases
in ΔP owing to changes in ventilator settings were strongly associated with increased N Engl J Med 2015;372:747-55.
DOI: 10.1056/NEJMsa1410639
survival. (Funded by Fundação de Amparo e Pesquisa do Estado de São Paulo and Copyright © 2015 Massachusetts Medical Society.
others.)

nejm.org February 19, 2015 747
M
echanical-ventilation strategies ARDS from four early randomized clinical trials
that use lower end-inspiratory (plateau) testing various strategies of volume-limited ven-
airway pressures, lower tidal volumes tilation.1,19-21 We next tested and refined this
(VT), and higher positive end-expiratory pressures model with data from a validation cohort of 861
(PEEPs) — collectively termed lung-protective patients from a large, randomized trial2 comparing
strategies — have been associated with survival lower versus higher VT values. Finally, we retested
benefits in randomized clinical trials involving the model with data from a more recent valida-
patients with the acute respiratory distress syn- tion cohort of 2365 patients with ARDS enrolled
drome (ARDS).1-4 The different components of lung in four randomized trials comparing higher-PEEP
protection in those strategies, such as lower VT, versus lower-PEEP strategies4,10-12,22 (Table 1, and
lower plateau pressure, and higher PEEP, can all Tables S1 and S2 and Fig. S1 in the Supplemen-
reduce mechanical stresses on the lung, which are tary Appendix, available with the full text of this
thought to induce ventilator-induced lung injury.5-9 article at NEJM.org).
Clinical trials, however, have reported conflicting
responses to the manipulation of separate com- Independent Variables and Outcomes
ponents of lung protection,10-14 and clinicians of- The primary outcome (the dependent variable)
ten face a dilemma when the optimization of one was survival in the hospital at 60 days (Cox sur-
component negatively affects another (for instance, vival model). Data from patients who were dis-
increasing PEEP may increase plateau pressure), charged home before day 60 were censored at
with unknown net consequences.15 day 60, with the patients considered to be alive
To minimize ventilator-induced lung injury, at day 60.
most studies have scaled VT to predicted body The independent variables tested as predic-
weight to normalize VT to lung size. However, in tors included treatment group (lung-protective
patients with ARDS, the proportion of lung avail- [i.e., varying variables such as VT, PEEP, and
able for ventilation is markedly decreased, which is plateau pressures with an intention to protect] vs.
reflected by lower respiratory-system compliance control assignment), characteristics of patients,
(CRS).13,16-18 Therefore, we hypothesized that nor- baseline severity of illness (e.g., risk according to
malizing VT to CRS and using the ratio as an index the Acute Physiology and Chronic Health Evalu-
indicating the “functional” size of the lung would ation [APACHE] or Simplified Acute Physiology
provide a better predictor of outcomes in patients Score [SAPS] and the ratio of the partial pressure
with ARDS than VT alone. This ratio, termed the of arterial oxygen to the fraction of inspired oxy-
driving pressure (ΔP = VT/CRS), can be routinely gen [Pao2:Fio2]), and ventilation variables (e.g., VT
calculated for patients who are not making inspira- and plateau pressure) averaged over the first 24
tory efforts as the plateau pressure minus PEEP. hours after randomization (Table S3 in the
To determine whether data from previous stud- Supplementary Appendix). In a separate analy-
ies are consistent with this hypothesis, we com- sis, we averaged individual ventilation data over
bined individual data from patients involved in the first 3 days and observed no predictive ad-
nine randomized trials comparing ventilation vantage of this approach (Tables S4, S5, and S6
strategies in patients with ARDS.1,2,10-12,19-22 We used in the Supplementary Appendix). Patients who
both a standard risk analysis with multivariate received pressure-support ventilation or had respi-
adjustments and a multilevel mediation analysis23,24 ratory rates that were higher than the ventilator
and examined the extent to which a change in settings (suggesting the presence of ventilatory
ΔP (or other variables) resulting from a change efforts) were excluded. Both conditions account-
in ventilator settings could be statistically linked ed for less than 3% of our sample. Barotrauma
to effects on survival, independent of the under- was defined as pneumothorax requiring chest-
lying severity of the lung injury and of the spe- tube drainage during the first 28 days after
cific lung-protection protocol. randomization.
Derivation and Validation of a Survival

Me thods Prediction Model
Derivation and Validation Cohorts Variables that had a significant univariate asso-
We derived a survival-prediction model with the ciation with survival were entered into a forward
use of data from a cohort of 336 patients with stepwise multivariate analysis and then into a
748 n engl j med 372;8 161

nejm.org February 19, 2015
Driving Pressure and Survival in ARDS
Table 1. Multivariate Cox Regression Model for 60-Day Hospital Survival.*
High-VT vs. Low-VT Trials High-PEEP vs. Low-PEEP Trials Combined Analysis
Variable (N = 1020) (N = 2060) (N = 3080)
Relative Risk Relative Risk Relative Risk

(95% CI) P Value (95% CI) P Value (95% CI) P Value
Model 1
Trial — <0.001 — 0.83 — <0.001
Age 1.51 (1.36–1.69) <0.001 1.64 (1.50–1.79) <0.001 1.59 (1.48–1.70) <0.001
Risk of death† 1.34 (1.20–1.49) <0.001 1.41 (1.29–1.54) <0.001 1.38 (1.29–1.48) <0.001
Arterial pH at entry 0.69 (0.63–0.77) <0.001 0.68 (0.63–0.74) <0.001 0.68 (0.64–0.72) <0.001
Pao2:Fio2 at entry 0.85 (0.77–0.95) 0.004 0.88 (0.80–0.96) 0.005 0.87 (0.81–0.93) <0.001
Day 1 ΔP 1.35 (1.24–1.48) <0.001 1.50 (1.35–1.68) <0.001 1.41 (1.31–1.51) <0.001
Model 2 (including all the variables
in model 1)
Day 1 ΔP 1.32 (1.19–1.47) <0.001‡ 1.51 (1.35–1.68) <0.001‡ 1.40 (1.30–1.51) <0.001‡
Day 1 VT 1.04 (0.95–1.14) 0.42§ 1.05 (0.90–1.23) 0.52§ 1.02 (0.95–1.10) 0.58§
Model 3 (including all the variables
in model 1)
Day 1 ΔP 1.36 (1.24–1.49) <0.001‡ 1.50 (1.34–1.68) <0.001‡ 1.41 (1.32–1.52) <0.001‡
Day 1 PEEP 0.97 (0.80–1.18) 0.78§ 0.99 (0.91–1.09) 0.90§ 1.03 (0.95–1.11) 0.51§
* Relative risks are the adjusted relative risks of death associated with a 1-SD increment in the given variable. Values higher than 1 indicate in-
creased mortality. Day 1 values are for the first 24 hours after randomization. The values used for standard deviations were as follows: age,
17 years; risk of death, 26%; arterial pH, 0.09; Pao2:Fio2, 60; driving pressure (ΔP), 7; positive end-expiratory pressure (PEEP), 5 cm of wa-
ter; and tidal volume (VT), 2 ml per kilogram of predicted body weight. By normalizing relative risk in this way, we were able to compare the
strength of the association of different variables with survival as the relative risk per se (using 1/relative risk when the relative risk was <1).
For instance, in the combined analysis, ΔP had a stronger association with survival (relative risk, 1.4) than did Pao2:Fio2 (1/relative
risk = 1/0.87 = 1.15). Although it is not shown in the table, the variables day 1 plateau pressure, day 1 respiratory-system compliance, and day
1 mean airway pressure were tested before and after inclusion of ΔP in model 1 and showed no significant association with survival (see
Section II.6, Table S8, in the Supplementary Appendix). CI denotes confidence interval.
† The risk of death was calculated according to the equations of the Acute Physiology and Chronic Health Evaluation (APACHE) II, APACHE
III, or Simplified Acute Physiology Score (SAPS) II, depending on the trial.
‡ The P value is for the test of inclusion of the variable in the model in which the variables in model 1 plus the extra covariate in the line be-
low were previously included.
§ The P value is for the test of inclusion of the variable in the model (the net contribution of the variable to predictive power in a likelihood ra-
tio test) in which the variables in model 1 plus ΔP were previously included.
backward stepwise multivariate analysis. Vari- for one variable (e.g., PEEP) but distinct mean
ables that were consistently found to be associ- levels for another ranking variable (e.g., driving
ated with survival with the use of both modeling pressure).
procedures were included in the final derivation
model. We adjusted all analyses for the trial vari- Mediation Analysis
able (Fig. S2 in the Supplementary Appendix). To investigate whether ΔP was more than a base-
The derivation model (model 1) was subsequently line risk predictor, we conducted a mediation
tested in each of the validation cohorts, as well analysis,24,25 searching for key variables that could
as in the combined data set. To show that the be linked to positive outcomes after randomiza-
prognostic information provided by ΔP was in- tion. When mediation analysis is applied to ran-
dependent of PEEP and plateau-pressure values, domized controlled trials, the goal is to determine
we resampled the combined data set (see Section whether a specific variable, strongly affected by
III.3 in the Supplementary Appendix), producing treatment-group assignment, has an effect on out-
subgroups of patients with matched mean levels comes that explains in whole or in part the effects
n engl j med 372;8 nejm.org February 19, 2015

162 749
resulting from treatment-group assignment.24,25 atic changes in ventilation parameters due to

For the relevant fraction of the effect in which treatment-group assignment might affect abso-
such a variable (the “mediator” in the model) is lute values of elastance but would not affect the
implicated, the correlation with outcomes must ranking of individual elastance values within the
exceed that of treatment group, typically exhibit- respective study groups. In the Supplementary
ing an independent, dose–response relationship Appendix (Section II.4, Fig. S3), we present a
(i.e., larger mediator changes are associated with sensitivity analysis addressing this assumption.
stronger survival effects). For example, in the In addition to the covariates of model 1, we
lower-VT studies, we tested whether survival was entered baseline respiratory system tidal elas-
better explained by specific ventilatory variables tance in all regression models used for the me-
than by treatment group (the treatment group in diation analysis, a procedure that intrinsically
these studies incorporated an intention-to-treat filtered out the potential confounding caused by
bundle including various recommendations, such differences in the severity of underlying lung
as VT reduction, plateau-pressure limitation, and disease. Accordingly, the mediation analysis ex-
acidosis management). We tested four mediator clusively addressed the effect of variations in ΔP
candidates: VT, plateau pressure, PEEP, and ΔP. The related to strategy — that is, variations in ΔP
first three variables were explicit targets in the superimposed by changes in ventilator settings
protocols, whereas ΔP, which was a dependent after randomization.
variable in these studies, was the variable we
hypothesized a priori to be the key mediator. Fol- R e sult s
lowing standard procedures for mediation analy-
sis, we examined each mediator candidate through Building and Testing the Prediction Model
a sequence of four logical tests, ultimately assess- In univariate analyses in the derivation cohort,
ing whether variations in the mediator ex- several significant associations were detected
plained the mean benefit of the randomly as- between independent predictor variables and
signed treatment group, as well as assessing the survival (Table S3 in the Supplementary Appen-
dose–response effect on outcomes. dix). Two baseline variables (risk according to
We used R software, version 2.10.1, with the APACHE or SAPS and arterial pH) and two ven-
R Package for Causal Mediation Analysis (R Proj- tilator variables (Fio2 and ΔP) were significantly
ect for Statistical Computing),23,24 in which a me- associated with survival after multivariate ad-
diation proportion is estimated, indicating how justment.
much of the whole risk reduction in the treatment The test of this preliminary model in our first
group can be explained by the indirect path in validation cohort showed that baseline Pao2:Fio2
which treatment-group assignment drives a change could replace the information associated with the
in the mediator and the change in the mediator Fio2 variable (Table S7 in the Supplementary Ap-
then affects the outcome (see the Supplementary pendix), with the advantage of being externally
Appendix). We calculated an average causal me- validated.26 We also observed that age was a strong,
diation effect,24 which expressed the indepen- independent predictor of survival even though it
dent hazard (relative risk) associated with this is a component of the APACHE score. After con-
indirect path. Other analyses were conducted servatively including the trial covariate, our final
with the use of SPSS software, version 20 (SPSS). model included six variables (Table 1, model 1);
To avoid possible biases due to differences in in this model, ΔP predicted survival as accurately
the severity of the underlying respiratory disor- as did risk according to APACHE or SAPS.
der, we preadjusted all mediation models ac- Treatment-group assignment was not inde-
cording to the baseline respiratory system tidal pendently associated with survival in model 1
elastance (the reciprocal of tidal compliance). and was omitted from Table 1. This variable was
For the lower-VT trials, this calculation was not considered separately in our mediation analysis.
possible, because baseline data were frequently Testing of model 1 in the second validation co-
missing. Thus, we used the elastance ranks within hort showed a strong association with survival
each treatment group (calculated after random- (P<0.001), with all covariates conferring similar
ization) for each trial, assuming that the system- relative risks in the two cohorts.

163 February 19, 2015
Source: The New England Journal of Medicine Driving Pressure and Survival in ARDS
Independence of Information mediation analysis23 with the use of trial as a

Even though ΔP is mathematically linked to CRS random effect, initially pooling the five VT stud-
and VT, no other ventilation variable conferred ies and then pooling the four PEEP studies (Fig.
independent predictive information to any survival S8 through S11 in the Supplementary Appendix).
model when ΔP was already a covariate. In con- A consistency analysis (Table S9 in the Supple-
trast, ΔP always conferred strong, nonredundant mentary Appendix) testing moderated mediation
predictive information when it was included in also suggested that there was consistency across
models preadjusted for other ventilator variables trials.
(Table 1, models 2 and 3; and Table S8 in the Reductions in ΔP after randomization were
Supplementary Appendix, models 2 through 5). significantly associated with better survival in
This observation was consistent in the derivation, both cohorts (step 2 of mediation analysis) (Fig.
validation, and combined cohorts. Higher ΔP pre- S8 and S9 in the Supplementary Appendix), in-
dicted lower survival consistently across trials dependently of baseline elastance of the respira-
(P = 0.13 for heterogeneity) (Fig. S4 in the Sup- tory system, and had similar effect sizes in both
plementary Appendix). cohorts (relative risk for VT trials, 0.62; 95% CI,
0.52 to 0.74; relative risk for PEEP trials, 0.57;
Risk Priority of ΔP 95% CI, 0.42 to 0.72).
Figure 1 shows that in the pooled sample (includ- For the VT and PEEP trials, treatment-group
ing 3562 patients), higher plateau pressures were assignment was an independent predictor of
observed in patients with higher ΔP or higher survival. Except for ΔP, however, no mediation
PEEP, but with different consequences (resampling candidate consistently passed through the step-
A vs. B): higher mortality was noted only when wise mediation tests (Fig. S10 and S11 in the
higher plateau pressures were observed in patients Supplementary Appendix). VT per se was not a
with higher ΔPs. Similarly, the protective effects significant mediator in the VT trials (P = 0.68 for
of higher PEEP were noted only when there were the average causal mediation effect), and PEEP
associated decreases in ΔP (resampling B vs. C). was not a significant mediator in the PEEP trials
In addition, at constant levels of plateau pressure (P = 0.50). In contrast, ΔP mediated 75% of the
(Fig. S5 in the Supplementary Appendix), we benefits due to treatment-group assignment in
observed that VT was a strong predictor of sur- the VT trials (P = 0.004 for the average causal
vival when normalized to CRS (i.e., ΔP) but not mediation effect) and 45% of these benefits in
when normalized to predicted body weight. the PEEP trials (P = 0.001). This was enough to
We also found a strong association between suppress the significance of the direct effect of
ΔP and survival even though all the ventilator the randomized treatment group, classically char-
settings that were used were lung-protective acterizing complete mediation.
(relative risk of death, 1.36; 95% confidence in- Thus, although ΔP was not an explicit target,
terval [CI], 1.17 to 1.58; P<0.001).2,11 In contrast, survival benefits in the VT trials were proportional
further reductions in plateau pressures or VT be- to reductions in ΔP driven by treatment-group
low these thresholds (plateau pressures ≤30 cm assignment rather than to reductions in VT (tested
of water and VT ≤7 ml per kilogram of predicted as a continuous variable). Similarly, the survival
body weight) had no effect on survival (Fig. S6 benefits observed in the PEEP trials occurred in
in the Supplementary Appendix). relation to reductions in ΔP rather than in rela-
Figure 2 shows the increase in the risk of tion to numerical increments in PEEP.
death as a function of progressive percentiles of
ΔP in the combined population. There was also Discussion
an increase in the odds of pneumothorax requir-
ing drainage as a function of progressive percen- In trials of mechanical ventilation involving pa-
tiles of ΔP but not of VT (Fig. S7 in the Supple- tients with ARDS, in which VT and PEEP were
mentary Appendix). included as independent variables, the dependent
quantity ΔP was the variable that was most
Test of Mediation strongly associated with survival. Although cau-
After observing that ΔP was associated with out- sality can be inferred only from direct controlled
comes in each study, we performed a multilevel trials, we found, using a statistical approach that

164 February 19, 2015 751
Resampling A: Resampling B: Resampling C:

Matched PEEP Matched ∆P Matched Plateau Pressure
40 40 40
Airway Pressure (cm of water)
30 30 30
∆P
∆P
∆P
20 20 20
10 10 PEEP 10 PEEP
PEEP
0 0 0
576 607 635 574 597 566 577 696 568 582 604 574 592 622 597
No. of Patients in Subsample
2.0 2.0 2.0
1.4 1.4 1.4

Multivariate Relative Risk
of Death in the Hospital
1.0 1.0 1.0
0.7 0.7 0.7
0.5 P<0.001 0.5 P=0.61 0.5 P<0.001
0.0 0.0 0.0

S1 S2 S3 S4 S5 S1 S2 S3 S4 S5 S1 S2 S3 S4 S5
A B C
Contrast Contrast
Higher plateau pressure: Not always risky Higher PEEP: Not always protective
Figure 1. Relative Risk of Death in the Hospital across Relevant Subsamples after Multivariate Adjustment — Survival Effect of
Ventilation Pressures.
Using double stratification procedures (obtaining subgroups of patients with matched mean levels for one variable but very different
mean levels for another ranking variable; see Section III.3 in the Supplementary Appendix for details), we partitioned our data set into
five distinct subsamples (each including approximately 600 patients with the acute respiratory distress syndrome [ARDS]) and calculat-
ed the relative risk (adjusted mortality) for each subsample in comparison with the mean risk in the combined population. The upper
stacked-bar diagrams illustrate the mean values for positive end-expiratory pressure (PEEP), plateau pressure, and driving pressure (ΔP)
observed in each subsample. The error bars represent 1 standard deviation. Each resampling (A, B, and C) produced subsamples with
similar mean values for one ventilator variable but very distinct values for the two other variables. At the bottom, the respective relative
risks for death in the hospital are shown, calculated for each subsample after multivariate adjustment (at the patient level) for the five
covariates (trial, age, risk of death according to the Acute Physiology and Chronic Health Evaluation [APACHE] or Simplified Acute Physi-
ology Score [SAPS], arterial pH at entry, and Pao2:Fio2 at entry) specified in model 1. Error bars represent 95% confidence intervals. A
relative risk of 1 represents the mean risk of the pooled population, which had an adjusted survival rate of 68% at 60 days. Note that a
lower survival rate was observed among patients with higher ΔP and higher survival was observed among patients with lower ΔP, inde-
pendent of concomitant variations in PEEP and plateau pressure.
adjusted for the effect of underlying lung disease by random treatment-group assignment were ben-
on the mechanical characteristics of the lung, eficial only if associated with decreases in ΔP.
that ΔP was a critical mediator of the benefits of No other ventilation variable had such a mediat-
various interventions. Our analyses indicated ing effect.
that reductions in VT or increases in PEEP driven We identified the striking correlations between
752 n engl j med 372;8

165 nejm.org February 19, 2015
Driving Pressure and Survival in ARDS
VT and survival or between VT and barotrauma

only when we scaled VT to individual CRS values
(ΔP = VT/CRS) (Fig. S5 in the Supplementary Ap- 2.5
pendix). This scaling has a strong physiological P<0.001
Multivariate Relative Risk

of Death in the Hospital
basis. In patients with ARDS, CRS is directly re- 2.0
lated to functional lung size (the volume of aer-

ated lung available for tidal ventilation).17,18 These 1.5
observations suggest that the aerated lung in a
patient with ARDS is not “stiff” but is small, 1.0
with nearly normal specific compliance (compli-
ance per unit of lung volume) in preserved areas. 0.5
The rationale underlying our mediation anal-
0.0
ysis was that ΔP was the surrogate for cyclic 5 10 15 20 25 30 35
lung strain that was most accessible and easiest Median VT
(10th–90th percentile) — ∆P (cm of water)
to calculate27; ΔP is defined as the amount of mg/kg of predicted
cyclic parenchymal deformation imposed on body weight 6.0 (5.9–7.5) 6.1 (5.8–9.2) 8.0 (5.7–12.1)
ventilated, preserved lung units. We also postu-
lated that cyclic strain predicts lung injury better Figure 2. Relative Risk of Death in the Hospital versus ΔP in the Combined
than VT. Implicitly, we hypothesized that the Cohort after Multivariate Adjustment.
functional lung size during disease is better The combined cohort (with 1249 death events) was partitioned into 15
quantified by CRS than by predicted body weight. quantiles of ΔP, and the relative risk for each quantile was calculated in
relation to the mean risk of the combined population (assumed to be 1).
Under such conditions, especially when CRS var- The mean risk and 95% confidence intervals (error bars) for each percen-
ies considerably among patients, cyclic strain, tile were calculated after multivariate adjustment at the patient level (Cox
ventilator-induced lung injury, and survival proportional-hazards model) for the five covariates (trial, age, risk of death
should all be correlated with ΔP rather than according to APACHE or SAPS, arterial pH at entry, and Pao2:Fio2 at entry)
with V T. specified in model 1. The gray zone represents the 95% confidence interval
for the Cox regression (dashed line) across the whole population when ΔP
Although this mediation analysis cannot esis considered as a continuous variable.
tablish causality, experimental studies provide a
plausible link between ΔP and ventilator-induced
lung injury. Many studies suggest that cell and
tissue damage are more closely related to the be relevant to patients with extremely low chest-
amplitude of cyclic stretch than to the maximal wall compliances.22,31
level of stretch — that is, lung tissue can undergo The Acute Respiratory Distress Syndrome
sustained stretching without damage.5,7,8,27-30 Network (ARDSNet) trial2 is often viewed as
Our study has a number of limitations. First, showing that low VT values per se decrease mor-
our conclusions are valid only for ventilation in tality from ARDS. However, our analyses sug-
which the patient is not making respiratory ef- gest that the efficacy of this strategy is also
forts. It is difficult to interpret ΔP in actively critically dependent on other components of the
breathing patients. Second, we studied a rela- lung-protective bundle (e.g., plateau-pressure lim-
tively narrow range of variables. Thus, extrapola- itation, respiratory-rate modification, and hyper-
tions to patients with plateau pressures greater capnia). For example, when low VT values were
than 40 cm of water, PEEPs less than 5 cm of introduced into the lung, improved survival was
water, or respiratory rates greater than 35 breaths observed only when large changes in ΔP (the
per minute are not warranted. Finally, we did not dependent variable during volume control) were
directly estimate the cyclic gradient of pressures avoided.
across the lung (transpulmonary ΔP), which is Our findings might also explain why studies
the probable effector of parenchymal injury. Be- of higher PEEPs did not show consistent survival
cause a large fraction of ΔP is typically applied benefits4,10-12; PEEP increments might be protec-
to inflate the lung in patients with severe ARDS, tive only when the increased PEEP values result
ΔP was probably a reasonable surrogate for trans- in a change in lung mechanics so that the same
pulmonary ΔP. However, this approach may not VT can be delivered with a lower ΔP. This hypoth-

February 19, 2015 753
esis is consistent with recent physiological stud- analysis. Clinical trials need to be designed in
ies suggesting that the benefits of PEEP are found which ventilator changes are linked to achieve
mainly in patients with greater lung recruitabil- changes in ΔP, in order to determine whether
ity,15 with some harm reported when PEEP caused our observations can be translated into changes
overdistention.15,32,33 Well-known devastating ef- that may be implemented at the bedside.
fects of zero-PEEP ventilation7,8 have been related Supported by Fundação de Amparo e Pesquisa do Estado de
to progressive atelectasis, decreased lung com- São Paulo, Conselho Nacional de Pesquisa e Desenvolvimento,
and Financiadora de Estudos e Projetos (FINEP).
pliance, and ultimately higher ΔP.34 Disclosure forms provided by the authors are available with
Finally, our work is a post hoc observational the full text of this article at NEJM.org.
References
1. Amato MBP, Barbas CSV, Medeiros acute respiratory distress syndrome. Evaluation of a ventilation strategy to pre-
DM, et al. Effect of a protective-ventila- N Engl J Med 2004;351:327-36. vent barotrauma in patients at high risk
tion strategy on mortality in the acute re- 11. Mercat A, Richard J-CM, Vielle B, et for acute respiratory distress syndrome.
spiratory distress syndrome. N Engl J Med al. Positive end-expiratory pressure set- N Engl J Med 1998;338:355-61.
1998;338:347-54. ting in adults with acute lung injury and 21. Brower RG, Shanholtz CB, Fessler HE,
2. The Acute Respiratory Distress Syn- acute respiratory distress syndrome: a et al. Prospective, randomized, controlled
drome Network. Ventilation with lower randomized controlled trial. JAMA 2008; clinical trial comparing traditional versus
tidal volumes as compared with tradition- 299:646-55. reduced tidal volume ventilation in acute
al tidal volumes for acute lung injury and 12. Meade MO, Cook DJ, Guyatt GH, et al. respiratory distress syndrome patients.
the acute respiratory distress syndrome. Ventilation strategy using low tidal vol- Crit Care Med 1999;27:1492-8.
N Engl J Med 2000;342:1301-8. umes, recruitment maneuvers, and high 22. Talmor D, Sarge T, Malhotra A, et al.
3. Villar J, Kacmarek RM, Pérez-Méndez positive end-expiratory pressure for acute Mechanical ventilation guided by esopha-
L, Aguirre-Jaime A. A high positive end- lung injury and acute respiratory distress geal pressure in acute lung injury. N Engl
expiratory pressure, low tidal volume ven- syndrome: a randomized controlled trial. J Med 2008;359:2095-104.
tilatory strategy improves outcome in JAMA 2008;299:637-45. 23. Imai K, Keele L, Tingley D, Yamamoto
persistent acute respiratory distress syn- 13. Terragni PP, Rosboch G, Tealdi A, et T. Causal mediation analysis using R. In:
drome: a randomized, controlled trial. al. Tidal hyperinflation during low tidal Vinod HD, ed. Advances in social science
Crit Care Med 2006;34:1311-8. volume ventilation in acute respiratory research using R. New York: Springer,
4. Briel M, Meade M, Mercat A, et al. distress syndrome. Am J Respir Crit Care 2010:129-54.
Higher vs lower positive end-expiratory Med 2007;175:160-6. 24. Imai K, Keele L, Tingley D, Yamamoto
pressure in patients with acute lung injury 14. Grasso S, Stripoli T, De Michele M, et T. Unpacking the black box of causality:
and acute respiratory distress syndrome: al. ARDSnet ventilatory protocol and al- learning about causal mechanisms from
systematic review and meta-analysis. veolar hyperinflation: role of positive experimental and observational studies.
JAMA 2010;303:865-73. end-expiratory pressure. Am J Respir Crit Am Polit Sci Rev 2011;105:765-89.
5. Webb HH, Tierney DF. Experimental Care Med 2007;176:761-7. 25. Shrout PE, Bolger N. Mediation in ex-
pulmonary edema due to intermittent 15. Grasso S, Fanelli V, Cafarelli A, et al. perimental and nonexperimental studies:
positive pressure ventilation with high in- Effects of high versus low positive end- new procedures and recommendations.
flation pressures: protection by positive expiratory pressures in acute respiratory Psychol Methods 2002;7:422-45.
end-expiratory pressure. Am Rev Respir distress syndrome. Am J Respir Crit Care 26. Ranieri VM, Rubenfeld GD, Thomp-
Dis 1974;110:556-65. Med 2005;171:1002-8. son BT, et al. Acute respiratory distress
6. Dreyfuss D, Soler P, Basset G, Saumon 16. Gattinoni L, Pesenti A, Avalli L, Ros- syndrome: the Berlin Definition. JAMA
G. High inflation pressure pulmonary si F, Bombino M. Pressure-volume curve 2012;307:2526-33.
edema: respective effects of high airway of total respiratory system in acute respi- 27. Protti A, Andreis DT, Monti M, et al.
pressure, high tidal volume, and positive ratory failure: computed tomographic Lung stress and strain during mechanical
end-expiratory pressure. Am Rev Respir scan study. Am Rev Respir Dis 1987;136: ventilation: any difference between stat-
Dis 1988;137:1159-64. 730-6. ics and dynamics? Crit Care Med 2013;41:
7. Muscedere JG, Mullen JB, Gan K, 17. Gattinoni L, Pesenti A, Baglioni S, Vi- 1046-55.
Slutsky AS. Tidal ventilation at low air- tale G, Rivolta M, Pelosi P. Inflammatory 28. Verbrugge SJ, Sorm V, van ’t Veen A,
way pressures can augment lung injury. pulmonary edema and positive end-expi- Mouton JW, Gommers D, Lachmann B.
Am J Respir Crit Care Med 1994;149: ratory pressure: correlations between im- Lung overinflation without positive end-
1327-34. aging and physiologic studies. J Thorac expiratory pressure promotes bacteremia
8. Tremblay L, Valenza F, Ribeiro SP, Li J, Imaging 1988;3:59-64. after experimental Klebsiella pneumoniae
Slutsky AS. Injurious ventilatory strate- 18. Gattinoni L, Pesenti A. The concept of inoculation. Intensive Care Med 1998;24:
gies increase cytokines and c-fos m-RNA “baby lung.” Intensive Care Med 2005;31: 172-7.
expression in an isolated rat lung model. 776-84. 29. Tschumperlin DJ, Oswari J, Margulies
J Clin Invest 1997;99:944-52. 19. Brochard L, Roudot-Thoraval F, Roupie AS. Deformation-induced injury of alveo-
9. Slutsky AS, Ranieri VM. Ventilator- E, et al. Tidal volume reduction for pre- lar epithelial cells: effect of frequency,
induced lung injury. N Engl J Med 2013; vention of ventilator-induced lung injury duration, and amplitude. Am J Respir Crit
369:2126-36. in acute respiratory distress syndrome. Care Med 2000;162:357-62.
10. Brower RG, Lanken PN, MacIntyre N, Am J Respir Crit Care Med 1998;158: 30. Garcia CS, Rocco PR, Facchinetti LD,
et al.Higher versus lower positive end- 1831-8. et al. What increases type III procollagen
expiratory pressures in patients with the 20. Stewart TE, Meade MO, Cook DJ, et al. mRNA levels in lung tissue: stress in-

167 February 19, 2015
Source: The New England Journal of Medicine Driving Pressure and Survival in ARDS
duced by changes in force or amplitude? 32. Vieira SR, Puybasset L, Lu Q, et al. A acute respiratory distress syndrome.
Respir Physiol Neurobiol 2004;144:59- scanographic assessment of pulmonary N Engl J Med 2006;354:1775-86.
70. morphology in acute lung injury: signifi- 34. Seah AS, Grant KA, Aliyeva M, Allen
31. Ranieri VM, Brienza N, Santostasi S, cance of the lower inflection point de- GB, Bates JH. Quantifying the roles of
et al. Impairment of lung and chest wall tected on the lung pressure-volume curve. tidal volume and PEEP in the pathogene-
mechanics in patients with acute respira- Am J Respir Crit Care Med 1999;159:1612- sis of ventilator-induced lung injury. Ann
tory distress syndrome: role of abdominal 23. Biomed Eng 2011;39:1505-16.
distension. Am J Respir Crit Care Med 33. Gattinoni L, Caironi P, Cressoni M, et Copyright © 2015 Massachusetts Medical Society.
1997;156:1082-91. al. Lung recruitment in patients with the

168 February 19, 2015 755
edi t or i a l
Driving Pressure and Respiratory Mechanics in ARDS

Stephen H. Loring, M.D., and Atul Malhotra, M.D.
In this issue of the Journal, Amato et al.1 use data overlooked when practitioners focus on the pla-
from previously published trials to determine teau pressure without considering the effect of
whether it is possible to predict outcomes in the chest wall in determining lung expansion
patients with the acute respiratory distress syn- and stress.6,7 High transpulmonary pressures can
drome (ARDS) on the basis of the settings of cause lung injury resembling ARDS or gross baro-
their mechanical ventilators or parameters de- trauma in the form of pneumothorax. Indeed,
rived from monitoring the mechanics of the ven- abundant data have shown that low Vt, and con-
tilation achieved. Previous articles published in sequently lower plateau and transpulmonary
the Journal had shown that a lung-protective pressures, improve survival.3 Importantly, ΔP
strategy — that is, limiting the tidal volume (Vt) limitation may not be helpful for patients who
and plateau pressure while providing relatively are actively breathing and who have pleural-pres-
high positive end-expiratory pressure (PEEP), can sure decreases during inspiration as a result of
improve survival in ARDS,2,3 thus demonstrating their own efforts to breathe in that result in
the importance of respiratory mechanics in de- high transpulmonary pressures. Second, atelec-
termining outcomes in patients.4 Lung-protective trauma,8 caused by the repetitive collapse and
ventilation strategies maintain alveolar aeration, reexpansion of lung units, has been shown to be
prevent overexpansion of the lung, and limit damaging. Lung collapse can result from surfac-
driving pressure (ΔP, which can be calculated as tant dysfunction, in which case surfactant fails
ventilator-measured plateau pressure minus ap- to have its physiologic effect and the surface
plied PEEP) and thereby are thought to reduce tension of alveolar-lining fluid becomes high,
ventilator-induced lung injury. promoting alveolar collapse. Collapse can also
Amato et al. focus on ΔP as a predictor of out- occur when elevated pleural pressures — for ex-
come in ARDS. Because ΔP is the tidal increase in ample, caused by pleural effusions, obesity, or
static transrespiratory pressure, it is proportional ascites — effectively compress the lung exter-
to Vt, with respiratory-system elastance (the in- nally.6 Applying adequate PEEP can help to pre-
verse of compliance) being the constant of pro- vent collapse of the lung at end exhalation and
portionality; elastance reflects the severity and thus prevent atelectrauma.4,8,9
extent of lung injury. Thus, ΔP is determined by The ability of ΔP to predict outcome is attrib-
variables known to predict or affect mortality in utable to the fact that the variables that define it
ARDS. The authors conducted a statistical media- are themselves highly predictive of survival. As
tion analysis of the aforementioned data, in which the authors emphasize, previous studies were not
variations of Vt, PEEP, ΔP, and respiratory-system designed to assess ΔP as an independent varia-
compliance were assessed to determine which of ble, and thus the findings reported by Amato
the operator-set or measured variables was most et al. should be considered hypothesis-generat-
closely linked to outcomes. They concluded that ΔP ing rather than definitive. The authors argue for
was the variable most closely related to survival. the “baby lung” concept, in which some portion
Several concepts are important in the consid- of the lung in patients with ARDS is collapsed
eration of these findings. First, transpulmonary or flooded and thus does not participate in gas
pressure (the pressure difference from airway exchange, leaving the rest of the lung (i.e., the
opening to pleural space) is the relevant distend- “baby lung”) to effect gas exchange.10 If this is
ing pressure for the lung.5 This concept is often the case, limiting ΔP may be a way to scale the

delivered breath to the size of the lung that is From the Department of Anesthesia, Critical Care, and Pain
Medicine, Beth Israel Deaconess Medical Center, and Harvard
available to participate in gas exchange, rather Medical School — both in Boston (S.H.L.); and the Division of
than scaling to body size, which may be less bio- Pulmonary Critical Care and Sleep Medicine, University of
logically relevant. Although the concept of limit- California, San Diego, La Jolla (A.M.).
ing ΔP is appealing, the question of whether the 1. Amato MBP, Meade MO, Slutsky AS, et al. Driving pressure
manipulation of ΔP rather than Vt is beneficial and survival in the acute respiratory distress syndrome. N Engl J
Med 2015;372:747-55.
remains. Designing prospective, randomized tri- 2. Amato MB, Barbas CS, Medeiros DM, et al. Effect of a pro-
als to assess the independent role of high versus tective-ventilation strategy on mortality in the acute respiratory
low ΔP in clinical outcomes will be complicated distress syndrome. N Engl J Med 1998;338:347-54.
3. The Acute Respiratory Distress Syndrome Network. Ventila-
and will require consideration of the effect that tion with lower tidal volumes as compared with traditional tidal
limiting ΔP has on Vt and subsequent minute volumes for acute lung injury and the acute respiratory distress
ventilation, as indicated by levels of carbon di- syndrome. N Engl J Med 2000;342:1301-8.
4. Malhotra A. Low-tidal-volume ventilation in the acute respi-
oxide in arterial blood, as well as the fact that a ratory distress syndrome. N Engl J Med 2007;357:1113-20.
given ΔP would have very different effects de- 5. Mead J, Takishima T, Leith D. Stress distribution in lungs:
pending on the PEEP level chosen (e.g., a PEEP a model of pulmonary elasticity. J Appl Physiol 1970;28:596-608.
6. Loring SH, O’Donnell CR, Behazin N, et al. Esophageal pres-
of 5 cm of water vs. 15 cm of water). sures in acute lung injury: do they represent artifact or useful
Is a strategy in which ventilators are set to information about transpulmonary pressure, chest wall mechan-
limit ΔP superior to our current approach? We ics, and lung stress? J Appl Physiol (1985) 2010;108:515-22.
7. Talmor D, Sarge T, Malhotra A, et al. Mechanical ventilation
strongly urge caution in accepting the idea that guided by esophageal pressure in acute lung injury. N Engl J Med
limiting ΔP is what we should do at the bedside 2008;359:2095-104.
now. Instead, the meta-analytic findings report- 8. Slutsky AS, Ranieri VM. Ventilator-induced lung injury. N Engl
J Med 2013;369:2126-36. [Erratum, N Engl J Med 2014;370:1668-9.]
ed by Amato et al. form the basis for a robust 9. Mercat A, Richard JC, Vielle B, et al. Positive end-expiratory
debate regarding how to design a controlled trial pressure setting in adults with acute lung injury and acute respira-
to be sure the idea of limiting ΔP is correct. tory distress syndrome: a randomized controlled trial. JAMA
2008;299:646-55.
Although the design of such a trial will not be 10. Gattinoni L, Pesenti A. The concept of “baby lung.” Intensive
easy, the problem is important. In the words of Care Med 2005;31:776-84.
Piet Hein, “Problems worthy of attack prove DOI: 10.1056/NEJMe1414218
their worth by hitting back.” Copyright © 2015 Massachusetts Medical Society.
Disclosure forms provided by the authors are available with
776 n engl j med 372;8 170

ACUTE LIVER FAILURE
Whether caused by the primary condition leading to a patient’s hospitalization or as a result of shock or drugs,
it is not uncommon for hepatic failure to develop in the ICU. This section focuses on the common causes of
hepatic failure, how to recognize them, and how to treat them.
171
Case Challenge
Acute Liver Failure

After septic shock and multiple complications, a 77-year-old man is intubated with his condition stabilized
when suddenly his ALT increases to 3402 U per liter, his AST to 5322 U per liter, and his bilirubin to 3.8 mg
per deciliter. How should his liver failure be treated?
heavy alcohol intake, and mild cognitive impairment was admitted to the intensive care unit (ICU) of a university
hospital from the operating room after resection of the rectosigmoid colon with closure of the rectal stump and
formation of an end colostomy performed for fecal peritonitis due to a perforated sigmoid colon. On arrival in the
ICU, he was in septic shock. He received fluid resuscitation with Hartmann’s solution and 5% human albumin
solution. His blood pressure was supported with a norepinephrine infusion, and he was treated with ampicillin,
amikacin, and metronidazole. He underwent mechanical ventilation with the use of a lung-protective strategy
and was sedated with propofol. (In the previous installment of this case, there were 2783 votes on the avoidance
of ventilator-induced lung injury. Most respondents [65%] chose to continue with the current ventilator mode,
increase the positive end-expiratory pressure to 10 cm of water, increase the respiratory rate to 22 breaths per
172
minute, and decrease the current tidal volume slightly to keep the plateau pressure at 30 cm of water or less, an
approach consistent with the ARDS Network strategy for low tidal volume. A total of 15% of respondents chose
to decrease the tidal volume to 360 ml and increase the respiratory rate to 22 breaths per minute; 15% chose to
change the ventilator mode to pressure-controlled synchronized intermittent mandatory ventilation with pressure
support; and 3% chose to increase the respiratory rate and tidal volume to normalize the partial pressure of arterial
carbon dioxide.) The postoperative analgesia was a fentanyl infusion, with the dose adjusted by the nurses, and 1 g
of intravenous acetaminophen every 6 hours.
Over the next 2 days, his condition stabilized, but on the 4th day after surgery, the alanine aminotransferase
(ALT) level had increased from within the normal range (7 to 55 U per liter) at admission to 3402 U per liter, and
the aspartate aminotransferase (AST) level had increased to 5322 U per liter (normal range, 8 to 48 U per liter).
The international normalized ratio was 4.3, and the bilirubin level was 3.8 mg per deciliter (65 µmol per liter)
(normal range, 0.1 to 1.0 mg per deciliter [1.7 to 17.0 µmol per liter]).

What strategy would you use to treat this patient’s acute liver failure?
A. Reduce the acetaminophen dose to 1 g every 12 hours and continue supportive therapy.
B. Cease acetaminophen and continue supportive therapy.
C. Cease acetaminophen, start treatment with acetylcysteine, and continue supportive therapy.
D. Continue acetaminophen, since it is probably not associated with the patient’s liver failure.
173
review article
CRITICAL CARE MEDICINE
Acute Liver Failure

William Bernal, M.D., and Julia Wendon, M.B., Ch.B.
A
cute liver failure is a rare but life-threatening critical ill- From the Liver Intensive Therapy Unit,
ness that occurs most often in patients who do not have preexisting liver Institute of Liver Studies, King’s College
London, London. Address reprint re-
disease. With an incidence of fewer than 10 cases per million persons per quests to Dr. Bernal at the Liver Intensive
year in the developed world, acute liver failure is seen most commonly in previ- Therapy Unit, Institute of Liver Studies,
ously healthy adults in their 30s and presents unique challenges in clinical manage- King’s College London, Denmark Hill
Campus, King’s College Hospital, Den-
ment. The clinical presentation usually includes hepatic dysfunction, abnormal mark Hill, London SE5 9RS, United King-
liver biochemical values, and coagulopathy; encephalopathy may develop, with mul- dom, or at william.bernal@kcl.ac.uk.
tiorgan failure and death occurring in up to half the cases (Fig. 1).1-3
The Supplementary Appendix for this
The rarity of acute liver failure, along with its severity and heterogeneity, has article is available at NEJM.org (http://
resulted in a very limited evidence base to guide supportive care.4 However, rates www.nejm.org/action/showSupplements?
of survival have improved substantially in recent years through advances in critical doi=10.1056%2FNEJMra1208937)
care management and the use of emergency liver transplantation.5 In this review, N Engl J Med 2013;369:2525-34.
we outline the causes and clinical manifestations of acute liver failure and discuss DOI: 10.1056/NEJMra1208937
current approaches to patient care. Copyright © 2013 Massachusetts Medical Society.
The Cl inic a l Probl em

Definition and Presentation
The original term “fulminant hepatic failure,” defined as “a severe liver injury, po-
tentially reversible in nature and with onset of hepatic encephalopathy within
8 weeks of the first symptoms in the absence of pre-existing liver disease,”6 remains
relevant today. More modern definitions recognize distinct disease phenotypes and
quantify the interval between the onset of symptoms and the development of en-
cephalopathy7 (Fig. 2). This interval provides clues to the cause of disease, likely
complications, and prognosis with supportive medical care alone.8-10 In hyperacute
cases, this interval is a week or less, and the cause is usually acetaminophen toxic-
ity or a viral infection. More slowly evolving, or subacute, cases may be confused
with chronic liver disease and often result from idiosyncratic drug-induced liver
injury or are indeterminate in cause. Patients with subacute causes, despite having
less marked coagulopathy and encephalopathy, have a consistently worse outcome
with medical care alone than those in whom the illness has a more rapid onset.
Causes
Acute liver failure is much less common in the developed world than in the develop-
ing world, where viral infections (hepatitis A, B, and E) are the predominant causes.
Public health measures (e.g., vaccination and improved sanitation) are among the
factors resulting in the reduced incidence of these infections in the United States
and much of Western Europe, where drug-induced liver injury is the most common
cause of acute liver failure (Fig. 3).
Viruses
Globally, hepatitis A and E infections are probably responsible for the majority of
cases of acute liver failure, with rates of death of more than 50% reported from the
n engl j med 369;26 nejm.org december 174

26, 2013 2525
The New England
personal Medicine
Downloaded from nejm.org by MJCopyright
MEDAS on September 29, 2014.
© Massachusetts ForSociety.
Medical personalAll
use only.reserved.
Source: The New England Journal of MedicineT h e n e w e ng l a n d j o u r na l of m e dic i n e
Brain
Hepatic encephalopathy
Cerebral edema
Intracranial hypertension
Lungs
Acute lung injury Heart
Acute respiratory distress syndrome High output state
Frequent subclinical myocardial injury
Liver
Pancreas
Loss of metabolic function
Pancreatitis, particularly in
Decreased:
acetaminophen-related disease
Gluconeogenesis hypoglycemia
Lactate clearance lactic acidosis
Ammonia clearance hyperammonemia Adrenal gland
Synthetic capacity coagulopathy Inadequate glucocorticoid production
contributing to hypotension
Kidney
Bone marrow
Frequent dysfunction or failure
Frequent suppression, particularly
in viral and seronegative disease
Portal hypertension
May be prominent in subacute disease
and confused with chronic liver disease
Circulating leukocytes
Impaired function, with immunoparesis
contributing to high risk of sepsis Systemic inflammatory response
High energy expenditure or
rate of catabolism
Figure 1. Clinical Features of Acute Liver Failure.
developing world.11,12 Acute liver failure may also tive in prevention.14 Other rare viral causes of acute
occur after hepatitis B infection,13 which is a com- liver failure include herpes simplex virus, cytomeg-
mon cause in some Asian and Mediterranean alovirus, Epstein–Barr virus, and parvoviruses.15
countries. Particularly poor survival has been
seen in patients with reactivation of previously Drug-Induced Liver Injury
stable subclinical infection with the hepatitis B Drug-induced liver injury is responsible for ap-
virus without established chronic liver disease. proximately 50% of cases of acute liver failure in
This scenario is most common in patients with the United States.16,17 Such injury may be dose-
treatment-induced immunosuppression during or dependent and predictable, as exemplified by
after therapy for cancer. The identification of acetaminophen-induced hepatotoxicity, which is
at-risk patients and the use of antiviral prophy- the most common cause of acute liver failure in
laxis before the initiation of chemotherapy, im- the United States. It may also be idiosyncratic,
munotherapy, or glucocorticoid therapy are effec- unpredictable, and probably independent of dose.
2526 n engl j med 369;26 nejm.org december 26, 2013

175
The New England For
Copyright 29, 2014.
© Massachusetts For personal
Medical Society.use
Allonly. Noreserved.
Although acute liver failure after acetamino-

A O’Grady System
phen ingestion can occur after consumption of
Hyper-
a single large dose, the risk of death is greatest acute Acute Subacute
with substantial drug ingestion staggered over
hours or days rather than at a single time point. 0 1 2 4 8 12
Acute liver failure is more common with late Weeks from Jaundice to Encephalopathy
presentation to medical attention because of un-
intentional rather than deliberate self-poisoning.18 B Bernuau System
Malnourished patients and patients with alco- Fulminant Subfulminant
holism are at increased risk.19 Acetaminophen is
also a potential cofactor for hepatic injury in 0 1 2 4 8 12
patients taking the drug for the relief of symp- Weeks from Jaundice to Encephalopathy
toms from hepatic illness of other causes.20,21
Idiosyncratic drug-induced liver injury is rare, C Japanese System
even among patients who are exposed to poten- Fulminant Late-Onset
tially hepatotoxic medication, and few patients
with drug-induced liver injury have progression Subclass:
Acute Subacute
to encephalopathy and acute liver failure.22 Fac-
tors such as an older age, increased elevations in
0 1 2 4 8 12
blood aminotransferase and bilirubin levels, and
coagulopathy are associated with an increased Weeks from Jaundice to Encephalopathy
risk of death.17,23 Figure 2. Classification Systems for Acute Liver Failure.

Data are from O’Grady et al.,8 Bernuau et al.,9 and Mochida et al.10 In the
Other Causes Japanese system, the late-onset period is 8 to 24 weeks.
Acute ischemic hepatocellular injury, or hypoxic
hepatitis, may occur in critically ill patients with
primary cardiac, circulatory, or respiratory failure. virus or exposure to a toxin. These cases often
It may be caused by severe sepsis accompanied by follow a subacute presentation, and rates of sur-
signs of cardiac failure and major, transient eleva- vival are poor without transplantation.
tions in blood aminotransferase levels.24,25 This
condition primarily requires supportive cardio- Fo cus of Cr i t ic a l C a r e
respiratory management rather than specific inter-
ventions targeted at the liver injury. The prognosis Initial Care
depends on both the cause of hepatic hypoxia and Recognition of hepatic injury may be delayed if
the severity of liver injury. A similar liver-injury confusion or agitation is the dominant present-
pattern may also be seen in drug-induced liver in- ing sign, particularly in hyperacute cases in which
jury caused by recreational drugs such as MDMA jaundice is minimal or in subacute cases, which
(3,4-methylenedioxy-N-methylamphetamine, also may be mistaken for chronic liver disease. Early
known as ecstasy) or cocaine. discussion with specialists at a liver center may
Other causes of acute liver failure are neoplastic be crucial to guide management (Table 1) and ex-
infiltration, acute Budd–Chiari syndrome, heat- pedite the safe transfer of suitable patients.
stroke, mushroom ingestion, and metabolic dis- Early restoration of intravascular volume and
eases such as Wilson’s disease.15,16 Acute liver systemic perfusion may prevent or mitigate the
failure that occurs during pregnancy may require severity of organ failure. In patients with severe
early delivery of the fetus; management should be acetaminophen poisoning, the interval between
discussed with specialists at a referral center that drug ingestion and treatment with acetylcysteine
has capabilities for both neonatal care and inten- is closely related to the outcome.18,26 Acetylcys-
sive management of the mother’s liver disease. teine has complex antioxidant and immunologic
In many cases, the cause of acute liver failure effects that may also benefit patients with non–
remains unknown, despite intensive investigation; acetaminophen-related acute liver failure. In a
potential causes include infection with a novel randomized, controlled study involving such
n engl j med 369;26 nejm.org december

176 26, 2013 2527
The New England
Medical personalAll
userights
only. reserved.
United Kingdom
HAV Germany
HAV
Japan
2%
Other HEV 4% Other HAV
7% 1% HEV, NT 7% 7%
HBV HEV
Unknown 5% HBV 1%
17% Other
18%
28%
Aceta- Unknown HBV
Other drugs Aceta- minophen 34% 42%
minophen Unknown
11% 21% 15%
57%
Other drugs
14% Other Aceta-
drugs minophen
9% 0%
Bangladesh
Other
United States 0%
Unknown HAV
HAV HEV, NT
6% 3%
4%
Other drugs
HBV
Other 3%
7% HBV
19% India Aceta- 13%
Sudan Other minophen
Unknown Aceta- HAV
0%
18% minophen Other HAV 7% 2% HEV
39% 27% 0% 75%
HEV
5%
Other Unknown
drugs HBV HEV
31%
13% 22% 44%
Unknown
38% HBV
15%
Other Other Aceta-
drugs drugs minophen
8% 1% 0%
Acetaminophen
0%
Figure 3. Worldwide Causes of Acute Liver Failure.

HAV denotes hepatitis A virus, HBV hepatitis B virus, HEV hepatitis E virus, and NT not tested.
patients, treatment with intravenous acetylcyste- hepatic regeneration.30 In the absence of an evi-
ine improved survival rates, but only among pa- dence base to guide practice, we administer antibi-
tients with low-grade encephalopathy.27 otics preemptively in patients who have coagulopa-
Encephalopathy may progress rapidly, partic- thy and organ failure or encephalopathy and those
ularly in patients with hyperacute disease. For in whom illness progression is considered likely.
patients with progression to agitation or coma, High standards of infection control should be prac-
we recommend early endotracheal intubation and ticed to minimize the risks of nosocomial sepsis.
sedation for airway control in order to facilitate Overt bleeding is uncommon in patients with
general care, control oxygen and carbon dioxide acute liver failure and reflects a balanced hemo-
levels, and prevent aspiration pneumonitis, al- static defect. In most cases, the loss of hepatic
though practice varies according to center. synthesis of procoagulant factors is paralleled by
A low arterial blood pressure with systemic the loss of hepatically derived anticoagulants.
vasodilatation with or without confirmed sepsis Functional testing indicates no major bleeding
is common in patients with acute liver failure tendency and may even indicate the presence of
and is associated with more severe encephalopa- a procoagulant state.31,32 Since serial evaluation
thy and increased mortality.28,29 A later pattern of laboratory coagulation variables (e.g., the inter-
of functional immunosuppression may be seen national normalized ratio and prothrombin time)
with secondary nosocomial sepsis and impaired is central to prognostic evaluation, the adminis-

177 december 26, 2013
The New EnglandFor

Medical Society. use
All only.
reserved.
Table 1. Common Management Issues and Condition-Specific Elements of Care in Acute Liver Failure.*
Organ System and Common

Conditions Assessment Specific Elements of Care
Cardiovascular system
Hypotension Invasive monitoring for all conditions;
echocardiography for low cardiac
output and right ventricular failure
Intravascular volume depletion Correction of volume depletion
Vasodilatation Vasopressors
Low cardiac output and right Inotropic support
ventricular failure
Hepatic system
Evolving hepatic dysfunction Serial biochemical and coagulation testing Intravenous acetylcysteine
Respiratory system
Risk of aspiration pneumonitis Neurologic observation to monitor level Early tracheal intubation for depressed level
of consciousness of consciousness
Metabolic and renal systems
Hypoglycemia Serial biochemical testing Maintain normoglycemia
Hyponatremia Active fluid management
Renal dysfunction, lactic aci- Renal-replacement therapy
dosis, hyperammonemia
Impaired drug metabolism Review drug administration
Central nervous system
Progressive encephalopathy Neurologic observation; monitoring of Treatment of fever and hyponatremia; screening
serum ammonia level; transcranial for sepsis
ultrasonography; consideration of High-grade encephalopathy: endotracheal intu-
intracranial-pressure monitoring bation; avoidance of Paco2 of <30 mm Hg or
>45 mm Hg; target for serum sodium, 145–
150 mmol/liter; risk assessment for intracra-
nial hypertension
Intracranial hypertension Interventions for pressure surges: osmotherapy
(mannitol, hypertonic saline); temperature
control; rescue therapies (indomethacin,
thiopentone)
Hematologic system
Coagulopathy Laboratory coagulation testing No routine correction of coagulation abnormali-
ties, only for invasive procedures (including
platelets and fibrinogen)
Immunologic system
High risk of sepsis Clinical evaluation Antibiotic prophylaxis
* Paco2 denotes partial pressure of arterial carbon dioxide.
tration of coagulation factors should be avoided, since specific therapies may be available for some
except when needed to treat bleeding or before causes of acute liver failure (Table S1 in the Supple-
invasive procedures. mentary Appendix, available with the full text of
this article at NEJM.org). However, inappropriately
Subsequent Care prolonged investigation and medical therapy may
The severity of illness, rapidity of change, and extent make transplantation impossible if surgery be-
of extrahepatic organ involvement require early crit- comes contraindicated because of the progression
ical care. The cause of liver injury should be sought, of multiorgan failure and development of sepsis.
n engl j med 369;26 nejm.org december

178 26, 2013 2529
The New England
For of Medicine
Copyright 29, 2014.
Medical userights
reserved.
C ompl ic at ions In patients with subacute presentations, even low-

grade encephalopathy indicates an extremely poor
Cardiorespiratory Dysfunction prognosis, whereas in hyperacute disease, high
Circulatory dysfunction and hypotension are grades of encephalopathy may clearly indicate a
common in patients with acute liver failure and poor prognosis. The goal of clinical strategies is to
are often multifactorial in origin. The effective prevent the onset of encephalopathy, limit its se-
blood volume may initially be low owing to poor verity when it develops, and reduce the risk of ce-
oral intake and fluid losses through vomiting rebral edema. Intracranial hypertension from se-
and the development of vasodilatation, leading to vere cerebral edema remains a feared complication
a condition consistent primarily with hypovole- and is a leading cause of death worldwide among
mic shock. patients with acute liver failure. In many centers,
Approaches to cardiovascular support in pa- intracranial hypertension is seen in only a minor-
tients with acute liver failure do not differ mark- ity of patients. However, among patients in whom
edly from those used in patients with other intracranial hypertension develops, the rate of sur-
critical illnesses and focus on early restoration vival without transplantation remains poor.5
of circulating volume, systemic perfusion, and The pathogenesis of encephalopathy and ce-
oxygen delivery. In patients who continue to rebral edema in acute liver failure is only partly
have hypotension despite volume repletion, nor- understood; there is evidence that both systemic
epinephrine is the preferred vasopressor, with or and local inflammation and circulating neuro-
without adjunctive use of vasopressin or vaso- toxins, particularly ammonia, play a role.35,36
pressin analogues.33 Myocardial function should Encephalopathy can be precipitated by infection
be assessed by means of echocardiography, since and may occur in patients with low systemic
hypoxic hepatitis may result from impaired car- blood pressure and vasodilatation.37,38 Inflam-
diac function. Relative adrenal insufficiency may matory mediators may trigger or worsen en-
be present in patients with cardiovascular insta- cephalopathy through the alteration of cerebral
bility and is associated with increased mortality, endothelial permeability to neurotoxins or the
but whether supplemental glucocorticoids im- initiation of inflammatory responses and altered
prove survival is unclear.34 cerebral blood flow.39
Although endotracheal intubation is often re- In liver failure, the normal detoxification of
quired to manage a reduced level of conscious- ammonia to urea is impaired, and levels of cir-
ness, respiratory dysfunction is uncommon early culating ammonia increase. There is a close re-
in the clinical course of acute liver failure. It is lationship between an elevated arterial ammonia
more common later, during the phase of hepatic level and the development of encephalopathy,
regeneration or in association with nosocomial with the risk of intracranial hypertension great-
sepsis. The goals of respiratory care are similar to est when there is a sustained level of ammonia
those in other critical illnesses; hyperventilation to of 150 to 200 μmol per liter (255 to 340 μg per
induce hypocapnia may be used for emergent con- deciliter).37,40 Ammonia increases intracellular os-
trol of intracranial hypertension if the condition is molarity through its cerebral metabolism to glu-
associated with cerebral hyperemia, but sustained tamine and induces changes in neurotransmitter
hyperventilation should be avoided. Spontaneous synthesis and release and in mitochondrial func-
hyperventilation is averted by means of appropri- tion; altered cerebral function and swelling re-
ate sedation and mandatory modes of ventilation. sult.35,36 The speed of development of hyper-
ammonemia is such that the usual osmotic
Neurologic Conditions compensatory mechanisms are ineffective in
The central place of encephalopathy in the defini- cases of acute liver failure — in contrast to
tion of acute liver failure reflects its key prognostic cases of subacute or chronic disease, in which
importance, and its development reflects critically these compensatory mechanisms are function-
impaired liver function (Table S2 in the Supple- ing and intracranial hypertension is uncom-
mentary Appendix). However, depending on the mon.35,36 Treatments that are used in chronic
speed with which encephalopathy develops, its liver disease may be inappropriate in acute liver
presence has differential prognostic importance. failure. In particular, the role of neomycin, rif-

179 december 26, 2013
The New England

Medical personalAll
userights
reserved.
axamin, and other nonabsorbable antibiotics is level of at least 150 μmol per liter despite treat-
unclear, and treatment with lactulose is poten- ment, an age of 35 years or less, and concurrent
tially deleterious. renal or cardiovascular organ failure.37,38,40
Neurologic care focuses on the prevention of We treat sustained increases in intracranial
infection, the maintenance of stable cerebral per- pressure with a bolus of intravenous hypertonic
fusion, and the control of circulating ammonia saline (at a dose of 20 ml of 30% sodium chlo-
and its cerebral metabolism. The drug l-ornithine– ride or 200 ml of 3% sodium chloride, keeping
l-aspartate enhances ammonia detoxification to serum sodium at <150 mmol per liter) or man-
glutamine in muscle. However, in a large, ran- nitol (at a dose of 2 ml of 20% solution per kilo-
domized, controlled trial, the drug did not lower gram of body weight, maintaining serum osmo-
circulating ammonia levels, reduce the severity of lality at <320 mOsm per liter). Hypothermia at
encephalopathy, or improve survival rates among 32 to 34°C may be used in patients with resistant
patients with acute liver failure.41 cases, and a bolus of intravenous indomethacin
In patients with established encephalopathy, (at a dose of 0.5 mg per kilogram) may be used
treatment is focused on minimizing the risk of when cerebral hyperemia is also present.46
intracranial hypertension by lowering cerebral
ammonia uptake and metabolism through the Renal Dysfunction
use of sedation and prophylactic osmotherapy. Substantial renal dysfunction may occur in more
In a randomized, controlled trial involving pa- than 50% of patients with acute liver failure. This
tients with high-grade encephalopathy, treatment complication is more common in the elderly and
with intravenous hypertonic saline solution de- in patients with acetaminophen-induced acute
layed the onset of intracranial hypertension.42 liver failure.47 Although renal dysfunction is as-
Hypothermia affects multiple processes involved sociated with increased mortality, the resolution
in the development of cerebral edema; by slow- of liver failure is accompanied by a return to pre-
ing body metabolism, it lowers systemic produc- existing levels in most cases.48 In patients requir-
tion of ammonia and cerebral uptake and me- ing renal-replacement therapy, continuous rather
tabolism, in addition to having hemodynamic than intermittent forms are generally used to
stabilizing effects and reducing cerebral blood achieve greater metabolic and hemodynamic sta-
flow.35 Clinical observations have suggested that bility.49 In addition to indications for the use of
moderate hypothermia (32 to 33°C) improves renal-replacement therapy in other forms of criti-
hemodynamics and controls refractory intracra- cal illness, such therapy may be used to control
nial hypertension, but a multicenter trial of pro- hyperammonemia and other biochemical and
phylactic moderate hypothermia (34°C) in pa- acid–base disturbances.
tients with high-grade encephalopathy did not
show a delay in or reduced severity of intracra- T r e atmen t
nial hypertension.43,44 A pragmatic approach to
temperature management is to avoid fever and Metabolic and Nutritional Support
maintain a core body temperature of 35 to 36°C. The goal of treatment is to achieve overall meta-
The most effective mode of neurologic moni- bolic and hemodynamic stability, with the rea-
toring to guide therapy in patients with high- sonable, though yet unproven, idea that such
grade encephalopathy is not clear. Direct mea- therapy will greatly improve conditions for he-
surement of intracranial pressure is associated patic regeneration and minimize the risk of com-
with uncommon but definite risks, particularly plications. In patients with acute liver failure,
intracranial hemorrhage.45 In view of the poten- this type of support is provided as it is for other
tial complications and the decreasing incidence critically ill patients, with specific caveats. Patients
of intracranial hypertension, we monitor intra- with acute liver failure are at increased risk for
cranial pressure only in patients with clinical hypoglycemia, which can be prevented by an intra-
signs or evidence of evolving intracranial hyper- venous glucose infusion. Large-volume infusions
tension. Other indicators of increased risk in- of hypotonic fluids, which may result in hypona-
clude an arterial ammonia concentration of tremia and cerebral swelling, should be avoided.
more than 200 μmol per liter or a sustained Patients with acute liver failure have high energy

180 december 26, 2013 2531
The New
For personal ofNo
use only. Medicine
Medical 2014. For
rights reserved.
expenditure and protein catabolism, requiring nu- though the need for improved identification of
tritional support to preserve muscle bulk and im- candidates for transplantation is clear.
mune function.50,51 Pragmatically, in patients with
encephalopathy, we administer 1.0 to 1.5 g of en- Transplantation
teral protein per kilogram per day while frequently Although transplantation is a treatment option
measuring blood ammonia levels, with a lowered for some specific causes of acute liver failure,
protein load for short periods in patients with such treatment is not universally available, and
worsening hyperammonemia or otherwise at high less than 10% of liver transplantations are per-
risk for intracranial hypertension. formed in patients with acute liver failure.52,55 In
such patients, especially those who are at risk for
Prognostic Evaluation intracranial hypertension, intraoperative and post-
Early identification of patients who will not sur- operative management is challenging, and rates of
vive with medical therapy alone is of great practi- survival are consistently lower than those associ-
cal importance in identifying potential candi- ated with elective liver transplantation. However,
dates for transplantation. Since the progression outcomes have improved over time, with registry
of multiorgan failure results in deterioration in data reporting current rates of survival after
many patients who are awaiting transplantation, transplantation of 79% at 1 year and 72% at
candidates for transplantation should be identi- 5 years.55 Most deaths after transplantation for
fied as quickly as possible.52 acute liver failure occur from infection during
Various prognostic evaluation systems, most the first 3 postoperative months. The risk of
of which have features derived from analyses of death is higher among older recipients and
historical patient cohorts that were treated without among those receiving older or partial grafts or
transplantation, are in use worldwide. Although grafts from donors without an identical ABO
the details of these systems differ, they share com- blood group.55,56 Early impaired liver-graft
mon features (Table 2). The presence of encepha- function is poorly tolerated in critically ill pa-
lopathy is a key indicator, with further consider- tients and predisposes them to intracranial hyper-
ation given to the patient’s age and the severity tension and sepsis.56
of liver injury, as assessed by the presence of
coagulopathy or jaundice. The most well charac- Other Therapies
terized evaluation system is the King’s College The limited availability of liver transplantation
Criteria, with meta-analyses confirming that has led to the evaluation of other therapies in
these criteria have clinically acceptable specificity patients with advanced disease. Hepatocyte trans-
but more limited sensitivity.53,54 To address these plantation involves intraportal or intraperitoneal
limitations, a wide variety of alternate prognostic infusion of isolated human hepatocytes to aug-
systems and markers have been proposed. To ment liver function. The procedure has been used
date, none have achieved universal acceptance, successfully in neonates and children with inborn
errors of metabolism, but to date the experience
Table 2. Criteria for the Selection of Patients with Acute Liver Failure in pediatric acute liver failure has been limited.57
for Transplantation.* The cell mass that is infused represents only 5%
King’s College Clichy Japanese
of the theoretical liver mass, which is insufficient
Factor Criteria Criteria Criteria in patients with massive hepatic necrosis, and the
Age† Yes Yes Yes technique remains experimental.
Cause Yes No No
Other therapies seek to support the failing
liver through the removal of circulating toxic
Encephalopathy† Yes Yes Yes
mediators, to stabilize the clinical conditions of
Bilirubin level Varies No Yes the patients while they await definitive trans-
Coagulopathy† Yes Yes Yes plantation, or to facilitate native liver regenera-
tion. Among such extracorporeal liver-assist de-
* The King’s College criteria are from O’Grady et al.,8 the Clichy criteria from
Bernuau et al.,9 and the Japanese criteria from Mochida et al.10 Yes indicates vices are nonbiologic dialysis-based systems for
that the factor is included as a criterion, and No that the factor is not included; systemic detoxification and bioartificial devices
Varies indicates that the criterion is used only in cases not associated with that incorporate hepatic cells of porcine or human
acetaminophen.
† This factor is common to all prognostic models. origin to replace both detoxification and synthetic
functions.58,59 The most extensively studied device
181
2532 n engl j med 369;26 nejm.org december 26, 2013
The New England
Copyright © Massachusetts Journal
Medical of Medicine
Society. All rights reserved.
Downloaded from nejm.org by MJ MEDAS on September 29, 2014. For
Back to Table of Contentspersonal use only. No other uses without permission.
is the molecular adsorbent recirculating sys- except on secondary analysis.61 For now, the use
tem, with case series suggesting biochemical of extracorporeal devices should be restricted
improvements during its use.59 A multicenter, to clinical trials. Preliminary reports suggest
randomized, controlled trial involving patients that high-volume plasma exchange may be a
with acute liver failure showed no survival ben- promising therapy.62
efit, but the study was confounded by a trans- Dr. Wendon reports receiving fees for board membership
plantation rate of 75% soon after enrollment.60 from Pulsion and Excalenz and lecture fees from Fresenius and
The porcine hepatocyte–based HepatAssist de- Asahi Kasei. No other potential conflict of interest relevant to
this article was reported.
vice appeared to be safe in a randomized, con- Disclosure forms provided by the authors are available with
trolled trial but did not show a survival benefit the full text of this article at NEJM.org.
References
1. Escorsell A, Mas A, de la Mata M. prognostic factors in acute liver failure in 25. Henrion J. Hypoxic hepatitis. Liver Int
Acute liver failure in Spain: analysis of India. J Viral Hepat 2003;10:224-31. 2012;32:1039-52.
267 cases. Liver Transpl 2007;13:1389-95. 14. Torres HA, Davila M. Reactivation of 26. Heard KJ. Acetylcysteine for acetamin-
2. Bower WA, Johns M, Margolis HS, hepatitis B virus and hepatitis C virus in ophen poisoning. N Engl J Med 2008;359:
Williams IT, Bell BP. Population-based patients with cancer. Nat Rev Clin Oncol 285-92.
surveillance for acute liver failure. Am J 2012;9:156-66. 27. Lee WM, Hynan LS, Rossaro L, et al.
Gastroenterol 2007;102:2459-63. [Erratum, 15. Ichai P, Samuel D. Etiology and prog- Intravenous N-acetylcysteine improves
Am J Gastroenterol 2008;103:255.] nosis of fulminant hepatitis in adults. transplant-free survival in early stage
3. Kumar R, Shalimar, Bhatia V, et al. Liver Transpl 2008;14:Suppl 2:S67-S79. non-acetaminophen acute liver failure.
Antituberculosis therapy-induced acute liver 16. Ostapowicz G, Fontana RJ, Schiodt Gastroenterology 2009;137:856-64.
failure: magnitude, profile, prognosis, and FV, et al. Results of a prospective study of 28. Vaquero J, Polson J, Chung C, et al.
predictors of outcome. Hepatology 2010; acute liver failure at 17 tertiary care cen- Infection and the progression of hepatic
51:1665-74. ters in the United States. Ann Intern Med encephalopathy in acute liver failure. Gas-
4. Lee WM, Stravitz RT, Larson AM. In- 2002;137:947-54. troenterology 2003;125:755-64.
troduction to the revised American Asso- 17. Reuben A, Koch DG, Lee WM. Drug- 29. Karvellas CJ, Pink F, McPhail M, et al.
ciation for the Study of Liver Diseases induced acute liver failure: results of a Predictors of bacteraemia and mortality
Position Paper on acute liver failure 2011. U.S. multicenter, prospective study. Hepa- in patients with acute liver failure. Inten-
Hepatology 2012;55:965-7. tology 2010;52:2065-76. sive Care Med 2009;35:1390-6.
5. Bernal W, Hyyrylainen A, Gera A, et al. 18. Craig DGN, Bates CM, Davidson JS, 30. Berry PA, Antoniades CG, Hussain
Lessons from look-back in acute liver fail- Martin KG, Hayes PC, Simpson KJ. Stag- MJ, et al. Admission levels and early
ure? A single centre experience of 3300 pa- gered overdose pattern and delay to hos- changes in serum interleukin-10 are pre-
tients. J Hepatol 2013;59:74-80. pital presentation are associated with dictive of poor outcome in acute liver fail-
6. Trey C, Davidson CS. The management adverse outcomes following paracetamol- ure and decompensated cirrhosis. Liver
of fulminant hepatic failure. Prog Liver induced hepatotoxicity. Br J Clin Pharma- Int 2010;30:733-40.
Dis 1970;3:282-98. col 2012;73:285-94. 31. Agarwal B, Wright G, Gatt A, et al.
7. Wlodzimirow KA, Eslami S, Abu- 19. Myers RP, Shaheen AAM, Li B, Dean Evaluation of coagulation abnormalities
Hanna A, Nieuwoudt M, Chamuleau RA. S, Quan H. Impact of liver disease, alco- in acute liver failure. J Hepatol 2012;57:
Systematic review: acute liver failure — hol abuse, and unintentional ingestions 780-6.
one disease, more than 40 def initions. on the outcomes of acetaminophen over- 32. Stravitz RT, Lisman T, Luketic VA, et
Aliment Pharmacol Ther 2012;35:1245-56. dose. Clin Gastroenterol Hepatol 2008;6: al. Minimal effects of acute liver injury/
8. O’Grady JG, Schalm SW, Williams R. 918-25. acute liver failure on hemostasis as as-
Acute liver failure: redefining the syn- 20. Rezende G, Roque-Afonso AM, Samuel sessed by thromboelastography. J Hepatol
dromes. Lancet 1993;342:273-5. [Erratum, D, et al. Viral and clinical factors associ- 2012;56:129-36.
Lancet 1993;342:1000.] ated with the fulminant course of hepati- 33. Eefsen M, Dethloff T, Frederiksen H-J,
9. Bernuau J, Rueff B, Benhamou JP. Ful- tis A infection. Hepatology 2003;38:613-8. Hauerberg J, Hansen BA, Larsen FS. Com-
minant and subfulminant liver failure: 21. Khandelwal N, James LP, Sanders C, parison of terlipressin and noradrenalin
definitions and causes. Semin Liver Dis Larson AM, Lee WM. Unrecognized acet- on cerebral perfusion, intracranial pres-
1986;6:97-106. aminophen toxicity as a cause of indeter- sure and cerebral extracellular concentra-
10. Mochida S, Nakayama N, Matsui A, minate acute liver failure. Hepatology tions of lactate and pyruvate in patients
Nagoshi S, Fujiwara K. Re-evaluation of 2011;53:567-76. with acute liver failure in need of inotropic
the Guideline published by the Acute Liver 22. Chalasani N, Fontana RJ, Bonkovsky support. J Hepatol 2007;47:381-6.
Failure Study Group of Japan in 1996 to HL, et al. Causes, clinical features, and 34. Etogo-Asse F-E, Vincent RP, Hughes
determine the indications of liver trans- outcomes from a prospective study of SA, et al. High density lipoprotein in pa-
plantation in patients with fulminant drug-induced liver injury in the United tients with liver failure: relation to sepsis,
hepatitis. Hepatol Res 2008;38:970-9. States. Gastroenterology 2008;135:1924-34. adrenal function and outcome of illness.
11. Hoofnagle JH, Nelson KE, Purcell RH. 23. Björnsson E, Olsson R. Outcome and Liver Int 2012;32:128-36.
Hepatitis E. N Engl J Med 2012;367:1237- prognostic markers in severe drug-induced 35. Vaquero J. Therapeutic hypothermia
44. liver disease. Hepatology 2005;42:481-9. in the management of acute liver failure.
12. Wasley A, Fiore A, Bell BP. Hepatitis A 24. Lescot T, Karvellas C, Beaussier M, Neurochem Int 2012;60:723-35.
in the era of vaccination. Epidemiol Rev Magder S. Acquired liver injury in the in- 36. Desjardins P, Du T, Jiang W, Peng L,
2006;28:101-11. tensive care unit. Anesthesiology 2012;117: Butterworth R. Pathogenesis of hepatic
13. Khuroo MS, Kamili S. Aetiology and 898-904. encephalopathy and brain edema in acute
n engl j med 369;26 182

The New
use only. Medicine
rights reserved.
liver failure: role of glutamine redefined. randomized, controlled trial. J Hepatol 54. McPhail MJW, Wendon JA, Bernal W.
Neurochem Int 2012;60:690-6. 2011;54:Suppl:S26. Meta-analysis of performance of Kings’
37. Bernal W, Hall C, Karvellas CJ, 45. Vaquero J, Fontana RJ, Larson AM, et College Hospital Criteria in prediction of
Auzinger G, Sizer E, Wendon J. Arterial al. Complications and use of intracranial outcome in non-paracetamol-induced acute
ammonia and clinical risk factors for pressure monitoring in patients with acute liver failure. J Hepatol 2010;53:492-9.
encephalopathy and intracranial hyper- liver failure and severe encephalopathy. 55. Germani G, Theocharidou E, Adam R,
tension in acute liver failure. Hepatology Liver Transpl 2005;11:1581-9. et al. Liver transplantation for acute liver
2007;46:1844-52. 46. Wendon J, Lee W. Encephalopathy and failure in Europe: outcomes over 20 years
38. Kitzberger R, Funk GC, Holzinger U, cerebral edema in the setting of acute liver from the ELTR database. J Hepatol 2012;
et al. Severity of organ failure is an inde- failure: pathogenesis and management. 57:288-96.
pendent predictor of intracranial hyper- Neurocrit Care 2008;9:97-102. 56. Bernal W, Cross TJS, Auzinger G, et al.
tension in acute liver failure. Clin Gastro- 47. Leithead JA, Ferguson JW, Bates CM, Outcome after wait-listing for emergency
enterol Hepatol 2009;7:1000-6. et al. The systemic inflammatory response liver transplantation in acute liver failure:
39. Butterworth RF. Hepatic encephalop- syndrome is predictive of renal dysfunction a single centre experience. J Hepatol 2009;
athy: a central neuroinflammatory disor- in patients with non-paracetamol-induced 50:306-13.
der? Hepatology 2011;53:1372-6. acute liver failure. Gut 2009;58:443-9. 57. Hughes RD, Mitry RR, Dhawan A.
40. Kumar R, Shalimar, Sharma H, et al. 48. O’Riordan A, Brummell Z, Sizer E, et Current status of hepatocyte transplanta-
Persistent hyperammonemia is associated al. Acute kidney injury in patients admitted tion. Transplantation 2012;93:342-7.
with complications and poor outcomes in to a liver intensive therapy unit with para- 58. Stutchfield BM, Simpson K, Wigmore
patients with acute liver failure. Clin Gas- cetamol-induced hepatotoxicity. Nephrol SJ. Systematic review and meta-analysis of
troenterol Hepatol 2012;10:925-31. Dial Transplant 2011;26:3501-8. survival following extracorporeal liver
41. Acharya SK, Bhatia V, Sreenivas V, 49. Davenport A. Continuous renal replace- support. Br J Surg 2011;98:623-31.
Khanal S, Panda SK. Efficacy of L-orni- ment therapies in patients with liver dis- 59. Tritto G, Davies NA, Jalan R. Liver re-
thine L-aspartate in acute liver failure: a ease. Semin Dial 2009;22:169-72. placement therapy. Semin Respir Crit
double-blind, randomized, placebo-con- 50. Walsh TS, Wigmore SJ, Hopton P, Rich- Care Med 2012;33:70-9.
trolled study. Gastroenterology 2009;136: ardson R, Lee A. Energy expenditure in 60. Saliba F, Camus C, Durand F, et al.
2159-68. acetaminophen-induced fulminant hepatic Albumin dialysis with a noncell artificial
42. Murphy N, Auzinger G, Bernal W, failure. Crit Care Med 2000;28:649-54. liver support device in patients with acute
Wendon J. The effect of hypertonic sodi- 51. Plauth M, Cabré E, Riggio O, et al. liver failure: a randomized, controlled
um chloride on intracranial pressure in ESPEN guidelines on enteral nutrition: trial. Ann Intern Med 2013;159:522-31.
patients with acute liver failure. Hepatol- liver disease. Clin Nutr 2006;25:285-94. 61. Demetriou AA, Brown RS Jr, Busuttil
ogy 2004;39:464-70. 52. Simpson KJ, Bates CM, Henderson RW, et al. Prospective, randomized, mul-
43. Jalan R, Olde Damink SW, Deutz NE, NC, et al. The utilization of liver trans- ticenter, controlled trial of a bioartificial
Hayes PC, Lee A. Moderate hypothermia plantation in the management of acute liver in treating acute liver failure. Ann
in patients with acute liver failure and liver failure: comparison between aceta- Surg 2004;239:660-7.
uncontrolled intracranial hypertension. minophen and non-acetaminophen etiol- 62. Larsen FS, Schmidt LE, Wendon J,
Gastroenterology 2004;127:1338-46. ogies. Liver Transpl 2009;15:600-9. et al. Liver assisting with high-volume
44. Larsen FS, Murphy N, Bernal W, et al. 53. O’Grady JG, Alexander GJ, Hayllar KM, plasma exchange in patients with acute
Prophylactic effect of mild hypothermia Williams R. Early indicators of prognosis liver failure. Hepatology 2010;52:376A.
to prevent brain edema in patients with in fulminant hepatic failure. Gastroenter- abstract.
acute liver failure: results of a multicenter, ology 1989;97:439-45. Copyright © 2013 Massachusetts Medical Society.
images in clinical medicine

The Journal welcomes consideration of new submissions for Images in Clinical
Medicine. Instructions for authors and procedures for submissions can be found
on the Journal’s website at NEJM.org. At the discretion of the editor, images that
are accepted for publication may appear in the print version of the Journal,
the electronic version, or both.

183 december 26, 2013
The New England

Downloaded from collections.nejm.org. For Journal
personalofuse
Medicine
Downloaded from nejm.org by MJ MEDAS on
Copyright September 29, Medical
© Massachusetts 2014. ForSociety.
personal
Alluse only.reserved.
c or r e sp ondence
Acute Liver Failure

To the Editor: In their review article, Bernal To the Editor: Regarding the causes of acute
and Wendon (Dec. 26 issue)1 state that lactulose liver failure discussed by Bernal and Wendon, we
is potentially dangerous in the treatment of acute would like to point out that a number of causes
liver failure. Although its side-effect profile may of this condition should always be considered
make it unacceptable for patients in whom liver because specific treatments are available. Auto-
transplantation is being contemplated, in resource- immune hepatitis can present abruptly with hep-
poor settings where liver transplantation is not atocellular failure, constituting 2 to 16% of all
feasible, one has to fall back on other treatments. cases of acute liver failure.1 This presentation can
Lactulose is a first-line treatment to minimize make the diagnosis of this disorder difficult be-
the production and absorption of ammonia from cause of low serum globulin and IgG levels, the
the gut and is a well-established treatment for frequent absence of autoantibodies, and atypi-
hepatic encephalopathy.2 The American Associa- cal histologic hallmarks (centrilobular necrosis
tion for the Study of Liver Diseases recommends or perivenulitis2 rather than interface hepatitis).
its use in the early stages of encephalopathy.3 Al- A characteristic pattern of heterogeneous hepatic
though the data are limited and not uniformly hypoattenuation on unenhanced computed tomog-
positive, lactulose can be used to treat early stag- raphy has been described recently3 in autoim-
es of encephalopathy, and it may help prolong mune acute liver failure, as compared with the
survival.4 diffuse hypoattenuation that is seen in viral acute
Bimal K. Agrawal, M.D. liver failure. High-dose prednisolone is the first-
Maharishi Markandeshwar Institute of Medical Sciences line therapy,1 although the absence of improve-
and Research ment may be an indication for transplantation.
Mullana, India
In addition, complicated forms of malaria,
onlybimal@gmail.com
typhoid, leptospirosis, rickettsial infection, den-
Usha Agrawal, M.D. gue, and hepatic amoebiasis can cause jaundice
National Institute of Pathology
New Delhi, India
with encephalopathy and mimic acute liver failure.
No potential conflict of interest relevant to this letter was re- However, specific antimicrobial agents improve
ported. survival to 75% without liver transplantation.4
1. Bernal W, Wendon J. Acute liver failure. N Engl J Med 2013; Spiking fever, splenomegaly, mild elevations of
369:2525-34. aspartate aminotransferase (AST) and alanine
2. Zafirova Z, O’Connor M. Hepatic encephalopathy: current aminotransferase (ALT) (<5 times the upper
management strategies and treatment, including management
and monitoring of cerebral edema and intracranial hypertension limit of the normal range), a ratio of AST to ALT
in fulminant hepatic failure. Curr Opin Anaesthesiol 2010;23: of more than 1, a ratio of lactate dehydrogenase
121-7. to ALT of more than 1, and a normal value for
3. Lee WM, Stravitz T, Larson AM. Introduction to the re-
vised American Association for the Study of Liver Diseases the international normalized ratio are clues that
position paper on acute liver failure 2011 (http://www.aasld.org/ strongly suggest the consideration of such compli-
practiceguidelines/Documents/AcuteLiverFailureUpdate2011.pdf). cated infections, rather than acute liver failure.4
4. Larson AM. Diagnosis and management of acute liver failure.
Curr Opin Gastroenterol 2010;26:214-21. Harpal S. Dhaliwal, M.D., D.M.
DOI: 10.1056/NEJMc1400974 Preetmohinder Singh, M.D., D.N.B.
Postgraduate Institute of Medical Education and Research
Chandigarh, India
hsdhaliwalpgi@yahoo.com
n engl j med 370;12 184

nejm.org march 20, 2014 1167
No potential conflict of interest relevant to this letter was re- be seen in some systemic infections, and in such
ported.
cases the early administration of targeted anti-
1. Czaja AJ. Acute and acute severe (fulminant) autoimmune microbial medication is central to effective man-
hepatitis. Dig Dis Sci 2013;58:897-914.
2. Stravitz RT, Lefkowitch JH, Fontana RJ, et al. Autoimmune agement. Autoimmune processes may be impor-
acute liver failure: proposed clinical and histological criteria. tant in the pathogenesis of liver injury in acute
Hepatology 2011;53:517-26. liver failure due to a number of causes, including
3. Yasui S, Fujiwara K, Okitsu K. Importance of computed to-
mography imaging features for the diagnosis of autoimmune new presentations of autoimmune hepatitis2;
acute liver failure. Hepatol Res 2012;42:42-50. however, this cause of acute liver failure is very
4. Deepak NA, Patel ND. Differential diagnosis of acute liver uncommon, and clinical management may be
failure in India. Ann Hepatol 2006;5:150-6.
challenging.3 Although some patients may have
DOI: 10.1056/NEJMc1400974
a response to immunosuppressive therapy, a key
issue is that inappropriately prolonged therapy
The Authors Reply: In response to Agrawal and in an attempt to achieve medical control of dis-
Agrawal: therapies for acute liver failure must be ease may preclude successful and definitive trans-
tailored to the resources available locally, partic- plantation, owing to the development of treat-
ularly in settings where liver transplantation is ment-related sepsis and other complications.4
not available. As stated in our review, since the William Bernal, M.D.
evidence base for care is very limited, the use of Julia Wendon, M.B., Ch.B.
most therapies is based on opinion. Although the King’s College London
use of lactulose may be beneficial in some pa- London, United Kingdom
tients with cirrhosis and low-grade encephalopa- william.bernal@kcl.ac.uk.
thy, its role in critically ill patients with acute Since publication of their article, the authors report no fur-
liver failure is not established. Its use may be
deleterious because patients with acute liver fail- 1. Stravitz RT, Kramer AH, Davern T, et al. Intensive care of
patients with acute liver failure: recommendations of the U.S.
ure frequently have ileus that may be worsened, Acute Liver Failure Study Group. Crit Care Med 2007;35:2498-
particularly if oral fluid intake is inadequate. 508.
There are no clinical data to suggest a prolonga- 2. Stravitz RT, Lefkowitch JH, Fontana RJ, et al. Autoimmune
acute liver failure: proposed clinical and histological criteria.
tion of survival, and we and others do not recom- Hepatology 2011;53:517-26.
mend the use of lactulose for the great majority 3. Czaja AJ. Review article: the management of autoimmune
of patients.1 hepatitis beyond consensus guidelines. Aliment Pharmacol Ther
2013;38:343-64.
Dhaliwal and Singh bring up autoimmune 4. Ichai P, Duclos-Vallée J-C, Guettier C, et al. Usefulness of
hepatitis and specific infections as causes of corticosteroids for the treatment of severe and fulminant forms
acute liver failure; space considerations prevent- of autoimmune hepatitis. Liver Transpl 2007;13:996-1003.
ed us from an exhaustive discussion of all of DOI: 10.1056/NEJMc1400974

these in our review. Severe liver involvement may
1168 n engl j med 370;12 185


What strategy would you use to treat this patient’s acute liver failure?
A. Reduce the acetaminophen dose to 1 g every 12 hours and continue supportive therapy.
B. Cease acetaminophen and continue supportive therapy.
C. Cease acetaminophen, start treatment with acetylcysteine, and continue supportive therapy.
D. Continue acetaminophen, since it is probably not associated with the patient’s liver failure.

The biochemical disturbance reflects severe acute hepatic necrosis, with the concurrent elevations in ALT and
AST indicating a hepatocellular origin. The magnitude and clinical significance of hepatic injury are reflected by
the concurrent coagulation disturbance through impairment of synthetic function and a reduction in the produc-
tion and release of liver-derived coagulation factors.
The clinical vignette illustrates the complexity of acute liver failure in clinical practice, in which preexisting medi-
cal conditions may enhance susceptibility and multiple pathologic processes may simultaneously contribute to he-
patic injury.
In this case, the presence of hepatic steatosis or even cirrhosis from alcohol use may result in greater sensitivity to
hepatotoxic injury and compromise subsequent hepatic regeneration. Cognitive impairment and secondary mal-
nutrition may also result in glutathione deficiency and enhance acetaminophen hepatotoxicity. Although the dose
of acetaminophen is within the range recommended in standard guidelines, treatment needs to be individualized
on the basis of the patient’s weight and potential for hepatic sensitivity and with consideration given to the in-
creased bioavailability of intravenous preparations.
Although the administration of regular doses of acetaminophen alone in an unwell and predisposed patient, such
as the man in the vignette, may result in substantial hepatotoxic liver injury,1 other clinical factors need to be con-
sidered. These include an ischemic hepatitis from right-heart dysfunction and liver congestion and decreased in-
flow resulting from hypotension or a low cardiac output state. Complications from intraabdominal sepsis (includ-
ing acute portal-vein thrombosis, portal pyemia, and hepatic abscesses), drug-induced hepatotoxic injury from
other agents, and global liver hypoperfusion from elevated intraabdominal pressure may also be contributory. The
time course of illness is probably too rapid for this condition to represent reactivation of a carrier state of hepatitis
B virus or other infection.
Regardless of whether acetaminophen is the primary cause of liver injury, it should be discontinued. Treatment
with acetylcysteine should be initiated, even if other causes are being considered, since it is unlikely to do harm.2,3
Consideration should also be given to the administration of vitamin K, since coagulation abnormalities may re-
flect a deficiency in this vitamin rather than impaired hepatocellular synthetic function.
Other potential causes of liver injury should be investigated; echocardiography should be used to assess cardiac
structure and function; cardiac output should be optimized. Ultrasonography and Doppler studies should be per-
formed to assess possible hepatic vascular occlusion. To minimize the possibility of other drug injuries, all non
essential medications should be discontinued. Intraabdominal pressure should be determined and addressed.
186
Accurate prognostication on the basis of a single set of hepatic laboratory values is difficult, although the patient’s
age and severity of illness clearly place him at increased risk for death. Evaluating the contribution of liver dys-
function to the risk will depend on determining its dynamics though changes in laboratory values over time.
Hyperacute liver injury may resolve rapidly as hepatic regeneration occurs, but a more worrisome picture is that
of a failure of regeneration, as evidenced by persistent coagulopathy and progressive and worsening jaundice.
References
1. Claridge LC, Eksteen B, Smith A, Shah T, Holt AP. Acute liver failure after administration of paracetamol
at the maximum recommended daily dose in adults. BMJ 2010;341:c6764.
2. Harrison PM, Wendon JA, Gimson AES, Alexander GJM, Williams R. Improvement by acetylcysteine of
hemodynamics and oxygen transport in fulminant hepatic failure. N Engl J Med 1991;324(26):1852–1857.
3. Keays R, Harrison PM, Wendon JA, et al. Intravenous acetylcysteine in paracetamol induced fulminant
hepatic failure: a prospective controlled trial. BMJ 1991;303:1026–1029.
187
SEDATION AND DELIRIUM

IN THE INTENSIVE CARE UNIT
Delirium is common in the ICU. Adding to this, the procedures to which patients are subjected there may be
difficult to tolerate, prompting them to try removing lines and tubes. How these conditions are recognized and
treated is the basis of this review.
188
Case Challenge
Sedation and Delirium in the Intensive Care Unit

A 77-year old man is undergoing mechanical ventilation following an emergency laparotomy, complicated
by septic shock and then liver failure. He is receiving norepinephrine to support his blood pressure plus
continuous morphine for pain. Does he need to be sedated? If so, by what method?
hospital from the operating room after a Hartmann’s procedure that was performed for fecal peritonitis due to a
perforated sigmoid colon. On arrival in the ICU, he was in septic shock. He is undergoing mechanical ventilation
with the use of a low-tidal-volume protocol with positive end-expiratory pressure (PEEP). His arterial blood pres-
sure is supported with a norepinephrine infusion. His acute liver failure is being treated by removal of acetamino-
phen from his treatment regimen. (In the previous installment of this case, there were 3120 votes on strategies for
treating acute liver failure. A majority of the respondents [66%] recommended the cessation of acetaminophen
with the initiation of acetylcysteine and the continuation of supportive therapy, whereas 28% favored the former
strategy without the initiation of acetylcysteine, 2% favored a reduction in the acetaminophen dose, and 3% fa-
vored continuation of acetaminophen.) Analgesia is being provided by a continuous morphine infusion.
189

What sedation should be provided to this patient?
A. Intermittent intravenous lorazepam.

B. Continuous intravenous infusions of propofol to facilitate daily cessation of sedation to assess the ongoing need
for sedation and analgesia.
C. An intravenous infusion of alpha-2-adrenoceptor agonist such as dexmedetomidine with daily cessation to
assess the ongoing need for sedation.
D. I would not sedate this patient.
190
review article
Sedation and Delirium in the Intensive

Care Unit
Michael C. Reade, M.B., B.S., D.Phil., and Simon Finfer, M.D.
P
From the Burns, Trauma and Critical Care atients in intensive care units (ICUs) are treated with many in-
Research Centre, University of Queens terventions (most notably endotracheal intubation and invasive mechanical
land, and Joint Health Command, Aus
tralian Defence Force, Brisbane (M.C.R.); ventilation) that are observed or perceived to be distressing. Pain is the most
and the George Institute for Global common memory patients have of their ICU stay.1 Agitation can precipitate acciden-
Health, and Royal North Shore Hospital, tal removal of endotracheal tubes or of intravascular catheters used for monitoring
University of Sydney, Sydney (S.F.) — all
in Australia. Address reprint requests to or administration of life-sustaining medications. Consequently, sedatives and analge-
Dr. Reade at Level 9, University of Queens sics are among the most commonly administered drugs in ICUs.
land Health Sciences Building, Royal Bris Early intensive care practice evolved from intraoperative anesthetic care at a
bane and Women’s Hospital, Brisbane,
QLD 4029, Australia, or at m.reade@ time when mechanical ventilation was delivered by rudimentary machines that
uq.edu.au. were not capable of synchronizing with patients’ respiratory efforts. As a result,
N Engl J Med 2014;370:444-54. deep sedation was commonly used until a patient was able to breathe without as-
DOI: 10.1056/NEJMra1208705 sistance. Developments over the past 30 years, including microprocessor-controlled
ventilators that synchronize with patients’ own respiratory efforts and new,
shorter-acting sedative and analgesic medications, have dramatically changed this
approach. Equally important has been the recognition that pain, oversedation, and
delirium are issues that if undetected and untreated are distressing to patients and
associated with increased morbidity and mortality.
Just as the concept of the “triad of anesthesia” underscores the pharmacody-
namic interactions among hypnotics, analgesics, and muscle relaxants and the
recognition that the simultaneous administration of agents of each class permits
the use of lower doses of drugs of all classes, the concept of the “ICU triad” rec-
ognizes that pain, agitation, and delirium — and therefore approaches to their
management — are inextricably linked (Fig. 1). According to the principle that it
is better to treat disease than to mask it, sedatives should be used only when pain
and delirium have been addressed with the use of specific pharmacologic and
nonpharmacologic strategies.
Pa in, A na l ge si a , a nd Sedat ion in the ICU
Prospective studies confirm that the majority of patients who are treated in ICUs
have pain,1 which makes the assessment of pain and provision of adequate analge-
sia essential components of ICU care. The short-term consequences of untreated
pain include higher energy expenditure and immunomodulation.2,3 Longer-term,
untreated pain increases the risk of post-traumatic stress disorder.4 Assessing
whether a patient in the ICU is in pain may be difficult. The reference standard for
the assessment of pain is self-reporting by the patient, but patients in the ICU may
not be sufficiently interactive to give valid responses. Physiological indicators such
as hypertension and tachycardia correlate poorly with more intuitively valid mea-
sures of pain,5 but pain scales such as the Behavioral Pain Scale6 and the Critical
444 191
n engl j med 370;5 nejm.org january 30, 2014
The New
Medical ForAll
rights reserved.
Neurologic diagnosis (e.g., head injury) Observable and occult metabolic abnormalities
Preexisting mental impairment Withdrawal

from chronic
Medical comorbidity psychoactive
medications
Sleep (e.g., benzo-
Severity of illness
deprivation diazepines,
Substance opioids)
Advanced age
Delirium abuse or
withdrawal
Noise
Sedatives
Elements of
routine ICU care
(e.g., turning, physical
therapy)
Anxiety
Endotracheal
Agitation; (appropriate
Pain unpleasant
tube or pathologic)
awareness
Tissue injury
Frustration
(e.g., surgery, trauma,
pressure areas)
Lack of
homeostasis
Vascular Physical
(e.g., thirst,
access Affective component restraint
hunger, dyspnea)
(e.g., “this pain means
I’m more likely to die”) Inability to
Ventilator
communicate
dyssynchrony
Figure 1. Causes and Interactions of Pain, Agitation, and Delirium.

Drugs and other treatments for pain, agitation, and delirium form an “ICU triad” cognitive management analogous
to the “triad of anesthesia,” which highlights interactions among hypnotics, analgesics, and muscle relaxants to en
courage balanced anesthesia. The “ICU triad” concept highlights that changing one element is unlikely to be as ef
fective as a coordinated approach.
Care Pain Observation Tool7 provide structured Evidence from randomized, controlled trials
and repeatable assessments and are currently the consistently supports the use of the minimum
best available methods for assessing pain. possible level of sedation. In a landmark trial
A minority of ICU patients have an indication that compared routine daily interruption of seda-
for continuous deep sedation, for reasons such tive infusions with discretionary interruption by
as the treatment of intracranial hypertension, treating clinicians, patients whose sedation was
severe respiratory failure, refractory status epi- routinely interrupted received less sedation over-
lepticus, and prevention of awareness in patients all and spent fewer days undergoing mechanical
treated with neuromuscular blocking agents. This ventilation and fewer days in the ICU.8 Although
review will focus on the remaining overwhelm- the trial was too small to assess differences in
ing majority of patients undergoing mechanical mortality or discharge destination, the observed
ventilation for whom the use of sedatives and reductions in the duration of mechanical ventila-
analgesics should be minimized, with the goal tion and length of stay in the ICU were associated
that they be calm, lucid, pain-free, interactive, with a nonsignificant reduction in mortality and a
and cooperative with their care. nonsignificant increase in the proportion of pa-

192 january 30, 2014 445
The New
only.ofNo
Medicine
Medical 2014. For
rights reserved.
tients who were discharged to their own homes.8 monly used in the ICU are the benzodiazepines
A subsequent larger multicenter trial combined midazolam and lorazepam (and to a lesser ex-
the daily interruption of sedation with daily spon-tent, diazepam), the short-acting intravenous an-
taneous breathing trials.9 Daily interruption of esthetic agent propofol, and dexmedetomidine.12
sedation was associated with reduced adminis- Remifentanil, an opioid, is also used as a sole
tration of a benzodiazepine sedative, reduced agent because of its sedative effects. Benzodiaz-
duration of mechanical ventilation, reduced length epines act through γ-aminobutyric acid type A
of stay in the ICU, and significantly increased (GABA a) receptors, as in part does propofol,
survival. In contrast, when daily interruption of whereas dexmedetomidine is an α2-adrenoceptor
sedation was added to a protocol for sedation agonist, and remifentanil is a μ-opioid receptor
practice that already sought to minimize the agonist (Table 1). Marked differences in prescrib-
level of sedation, the total sedative dose was in- ing patterns among countries suggest that the
creased and there was no clinical benefit.10 choice of agent is determined more by tradition
These conflicting results are open to a number and familiarity than by evidence-based practice.
of interpretations, including that daily interrup- If minimizing the depth and duration of se-
tion is beneficial only when it results in a reduc-dation is accepted as a desirable goal, then the
tion in the total dose of sedative administered. use of a short-acting agent with an effect that
The conflicting findings also highlight that the can be rapidly adjusted such as propofol or remi-
results of daily interruption of sedation may be fentanil should offer advantages over longer-
context-specific and will depend on the popula- acting agents or agents with active metabolites.
tion being studied, protocol adherence, and As compared with benzodiazepines, propofol has
management of the control group. A random- not been shown to reduce mortality but may
ized, controlled trial in which all patients under-result in a reduction in the length of stay in the
going mechanical ventilation received morphine ICU.18 Dexmedetomidine may also have advan-
for the treatment of pain in an “analgesia first” tages over benzodiazepines, since it produces
approach compared a protocol of no sedation analgesia, causes less respiratory depression, and
with the routine use of sedation with daily inter- seemingly provides a qualitatively different type
ruption.11 Patients who were assigned to the of sedation in which patients are more interac-
protocol of no sedation had shorter stays in the tive and so potentially better able to communi-
ICU and the hospital and more days without cate their needs.19 As compared with lorazepam
mechanical ventilation. and midazolam, dexmedetomidine resulted in less
The consistent message from all these sedation-delirium and a shorter duration of mechanical
interruption trials is that minimizing sedation ventilation but not reduced stays in the ICU or
among patients in the ICU provides clinical hospital.19-21 When two short-acting and titrat-
benefit. Further support comes from a prospec- able drugs such as propofol and dexmedetomi-
tive, multicenter, longitudinal cohort study show- dine were compared, there was no significant
ing that the depth of sedation was independently difference in the time spent at the target seda-
associated with the duration of mechanical tion level and no difference in either the duration
ventilation, in-hospital mortality, and rates of of mechanical ventilation or ICU stay.19
death within 180 days.12 In a randomized, con- Remifentanil has a half-life of 3 to 4 minutes
trolled trial, the use of lighter sedation resultedthat is independent of the infusion duration or
in more ventilator-free and ICU-free days.13 In organ function. It has been investigated as a
comparison with deep sedation, the use of lighter sedative agent in ICUs predominantly among sur-
sedation did not increase the rate of short-term gical patients. It has been compared with mid-
adverse events, and long-term psychiatric out- azolam alone, midazolam with fentanyl, fentanyl
comes were either unaffected or improved.13-16 alone, and morphine.22-25 Although remifentanil
has been associated with a reduced duration of
Choice of Sedat i v e Agen t mechanical ventilation and ICU stay in these
small trials, it has not yet been evaluated in a
Despite at least 90 trials comparing sedative reg- large, heterogeneous population of critically ill
imens,17 in general, no sedative drug is clearly patients and is currently not a common choice in
superior to all others. Sedatives that are com- most ICUs.

193 january 30, 2014
The New EnglandFor

Copyright 29, 2014.
Medical use rights
reserved.
Table 1. Sedatives and Analgesics in Common Use in the ICU.*
Drug (Brand Name) Mechanism of Action Typical Adult Dose Pharmacokinetic Properties Adverse Effects
Midazolam (Versed) GABAA agonist Bolus, 1 to 5 mg; infusion, Halflife, 3 to 11 hr; active metabolite accumulates Possibly a higher risk of delirium and
1 to 5 mg/hr with prolonged infusion; metabolized by hepatic tolerance than with certain other sedatives
oxidation, with renal excretion of active metabolite
Lorazepam (Ativan) GABAA agonist Bolus, 1 to 4 mg; infusion, Slower onset (5 to 20 min) than that of midazolam or Possibly a higher risk of delirium and
1 to 5 mg/hr diazepam (2 to 5 min); halflife, 8 to 15 hr; metab tolerance than with certain other sedatives
olized by hepatic glucuronidation, with no active
metabolites, so offset may be more predictable
than that of midazolam in critical illness

Diazepam (Valium; GABAA agonist Bolus, 1 to 5 mg Halflife, 20 to 120 hr; metabolized by hepatic Poorly soluble in water, so prolonged peripheral
Diazemuls) desmethylation and hydroxylation; active intravenous infusion may cause phlebitis;
metabolite accumulates in renal failure possibly a higher risk of delirium and
tolerance than certain other sedatives
Downloaded from nejm.org by MJCopyright

Propofol (Diprivan) Halflife, 30 to 60 min after infusion; longer after Vasodilatation or negative inotropy causing
MEDAS on
GABAA agonist, with other 50 to 200 mg/hr or
effects, including on 1 to 3 mg/kg/hr prolonged infusion because of redistribution hypotension or bradycardia; propofol
n engl j med 370;5

glutamate and canna from fat stores; metabolized by hepatic glucu infusion syndrome (lactic acidosis,
binoid receptors ronidation and hydroxylation arrhythmia, and cardiac arrest), mostly
associated with prolonged infusion rates
© September
of >4 to 5 mg/kg/hr; hypertriglyceridemia;
The New England

Massachusetts
nejm.org
pancreatitis
Dexmedetomidine α2Agonist 0.2 to 1.5 μg/kg/hr Halflife, 2 hr; does not accumulate with prolonged Transient hypertension, then hypotension;
29, 2014.
ForJournal
194
(Precedex) infusion; metabolized by hepatic glucuronida bradycardia, dry mouth, nausea
Medical
personal
tion and oxidation, with no active metabolites
of use
Remifentanil (Ultiva) μOpioid agonist (also 0.5 to 2 μg/kg/min; loading Halflife, 3 to 4 min; does not accumulate with Nausea, constipation, respiratory depression,
ForSociety.
with κopioid agonist dose of 0.4 to 0.8 μg/kg prolonged infusion; metabolized by plasma bradycardia

january 30, 2014
Medicine
effects) may be considered esterases and so is unaffected by organ function
personalAll
Fentanyl (Sublimaze) μOpioid agonist (also 20 to 100 μg/hr; loading Halflife, 1.5 to 6 hr; highly fat soluble, so rapid onset Nausea, constipation, respiratory depression,
userights
with κopioid agonist dose of 50 to 100 μg but accumulates with prolonged infusion; metab skeletalmuscle rigidity with high bolus
effects) may be considered olized by hepatic oxidation; no active metabolites doses
Morphine (Roxanol; μOpioid agonist (also 1 to 5 mg/hr; loading dose Halflife, 3 to 7 hr; more water soluble, so slower Nausea, constipation, respiratory depression,

only.reserved.
Duramorph) with κopioid and of 2 to 5 mg may be onset than fentanyl with less accumulation; histamine release and consequent vaso
δopioid agonist considered metabolized by hepatic glucuronidation to dilatation and hypotension, itch
effects) morphine6glucuronide (10%) (20 times as
active as parent drug) and morphine3glucuronide
(90%) (inactive as an analgesic but causes neuro
excitation, at least in animal models), both with
renal excretion

Hydromorphone μOpioid agonist (also 0.5 to 2 mg/hr; loading Halflife, 1.5 to 3.5 hr; 7 to 11 times as potent as Nausea, constipation, respiratory depression
(Dilaudid) with κopioid and dose of 0.4 to 1.5 mg morphine; metabolized by hepatic glucuroni

δopioid agonist may be considered dation to hydromorphone3glucuronide, with
effects) effects similar to those of morphine3
glucuronide
* GABAA denotes γaminobutyric acid type A.
447
clude an advanced age and the presence of more

Pr e v en t ion a nd T r e atmen t than one condition associated with coma, fol-
of Del ir ium
lowed by treatment with sedative medications, a
The Diagnostic and Statistical Manual of Mental Disor- neurologic diagnosis, and increased severity of
ders, 4th edition (DSM-IV),26 lists four domains of illness.31 A diagnosis of delirium is associated
delirium: disturbance of consciousness, change with increased mortality (estimated as a 10%
in cognition, development over a short period, increase in the relative risk of death for each day
and fluctuation. Delirium is defined by the Na- of delirium32) and decreased long-term cognitive
tional Institutes of Health as “sudden severe con- function.33
fusion and rapid changes in brain function that There are two distinct forms of delirium,
occur with physical or mental illness.” The most hypoactive and agitated (or hyperactive). When
common feature of delirium, thought by many to individual patients intermittently have both forms,
be its cardinal sign, is inattention. Delirium is a it is termed mixed delirium. The hypoactive
nonspecific but generally reversible manifesta- form is characterized by inattention, disordered
tion of acute illness that appears to have many thinking, and a decreased level of consciousness
causes, including recovery from a sedated or without agitation. Pure agitated delirium affects
oversedated state. less than 2% of patients with delirium in the ICU.34
The pathophysiology of delirium that is associ- Patients with hypoactive delirium are the least
ated with critical illness remains largely unchar- likely to survive, but those who do survive may
acterized and may vary depending on the cause. have better long-term function than those with
The increased risk associated with the use of agitated or mixed delirium.33 Separating the ef-
GABA a agonists and anticholinergic drugs led to fects of delirium status from those of illness
the suggestion that the GABAergic and choliner- severity with respect to the risk of death is dif-
gic neurotransmitter systems play a contributory ficult, since patients with more serious illnesses
role. In particular, central cholinergic deficiency are at increased risk for both delirium and death.
may be a final common pathway. Alternative Association studies typically adjust for illness
hypotheses include excess dopaminergic activity severity on admission to the ICU rather than at
and direct neurotoxic effects of inflammatory the time that delirium is diagnosed. Although the
cytokines. Currently, these hypotheses are un- association between delirium and a worse out-
proven, making pharmacologic management come is clear, a causal relationship has not been
strategies largely empirical. established. Currently, the evidence that specific
Studies using magnetic resonance imaging treatment of delirium may improve outcomes is
have shown a positive association between the tenuous.
duration of delirium in the ICU and both cere-
bral atrophy and cerebral white-matter disrup- A sse ssmen t a nd Moni t or ing
tion.27,28 These preliminary investigations indi- of Sedat ion a nd Del ir ium
cate either that delirium in the ICU gives rise to
alterations in brain structure or that the presence Although ICU practice is characterized by close
of such cerebral atrophy and white-matter disrup- monitoring of carefully administered care, sur-
tion renders patients more susceptible to delirium. veys that have been conducted in various coun-
Regardless of the cause and the underlying tries have shown that the depth of sedation fre-
pathophysiology, delirium is now recognized as a quently goes unmonitored.35 This finding is
frequent and serious event in critically ill patients. surprising and unacceptable, since evidence sug-
There is no diagnostic blood, electrophysiologi- gests that the routine monitoring of sedation
cal, or imaging test for delirium, which there- may improve patients’ outcomes.36
fore remains a clinical diagnosis. Estimates for
the incidence of delirium in the ICU range from Sedation Scales
16%29 to 89%,30 with the reported incidence af- Of the sedation scales described, the Riker
fected both by the characteristics of the popula- Sedation–Agitation Scale37 and the Richmond
tion being studied and by the diagnostic criteria Agitation–Sedation Scale38 are the most com-
used. Risk factors that have been identified in- monly reported, but in head-to-head comparison,
448 n engl j med 370;5 195

The New EnglandFor

Copyright 29, 2014.
reserved.
neither is demonstrably superior39 (Table 2). Forters of their patients who have the condition,
the majority of patients undergoing mechanical whereas active screening by research nurses
ventilation in an ICU, an appropriate target is aidentified delirium in up to 64% of patients who
score of 3 to 4 on the Riker Sedation–Agitation were considered to be delirious by a psychiatrist,
Scale (which ranges from 1 to 7, with scores of a geriatrician, or a neurologist.40 Scales with re-
<4 indicating deeper sedation, a score of 4 indi-spect to delirium in the ICU apply the four DSM-IV
cating an appearance of calm and cooperative- domains defining delirium in general medical
ness, and scores of ≥5 indicating increasing agi-and psychiatric patients to those in the ICU whose
tation) or a score of −2 to 0 on the Richmond severity of illness can rapidly fluctuate, who re-
Agitation–Sedation Scale (which ranges from −5 ceive multiple analgesics and sedatives, and who
to +4, with more negative scores indicating deep-are unable to speak owing to endotracheal intuba-
er sedation and more positive scores indicating tion. Two scales are in common use, the Confu-
increasing agitation, and with 0 representing thesion Assessment Method for the ICU (CAM-ICU)41
appearance of calm and normal alertness). and the Intensive Care Delirium Screening Check-
list (ICDSC)29 (Table 3). The CAM-ICU reports a
Identifying Delirium dichotomous assessment at a single time point,
In routine practice, ICU staff members typically whereas the ICDSC lists signs that can be ob-
do not diagnose delirium in almost three quar- served over a period of time. Although such
Table 2. Sedation Scales for Patients in the ICU.
Scale and Scoring Method Description

Riker Sedation–Agitation Scale (SAS)*
Dangerous agitation (score of 7) Pulling at endotracheal tube, trying to remove catheters, climbing over bed rail,
striking at staff, thrashing from side to side
Very agitated (score of 6) Requiring restraint and frequent verbal reminding of limits, biting endotracheal tube
Agitated (score of 5) Anxious or physically agitated, calming at verbal instruction
Calm and cooperative (score of 4) Calm, easily rousable, follows commands
Sedated (score of 3) Difficult to arouse but awakens to verbal stimuli or gentle shaking; follows
simple commands but drifts off again
Very sedated (score of 2) Arouses to physical stimuli but does not communicate or follow commands,
may move spontaneously
Cannot be aroused (score of 1) Minimal or no response to noxious stimuli, does not communicate or follow
commands
Richmond Agitation–Sedation Scale (RASS)†
Combative (score of 4) Overtly combative, violent, immediate danger to staff
Very agitated (score of 3) Pulls or removes tubes or catheters; aggressive
Agitated (score of 2) Frequent nonpurposeful movement, fights ventilator
Restless (score of 1) Anxious but movements not aggressive or vigorous
Alert and calm (score of 0) Alert and calm
Drowsy (score of −1) Not fully alert but has sustained awakening (eye opening or eye contact) to
voice (≥10 sec)
Light sedation (score of −2) Briefly awakens with eye contact to voice (<10 sec)
Moderate sedation (score of −3) Movement or eye opening to voice but no eye contact
Deep sedation (score of −4) No response to voice but movement or eye opening to physical stimulation
Cannot be aroused (score of −5) No response to voice or physical stimulation
* Data are from Riker et al.37

† Data are from Sessler et al.38

The New
Medical ForAll
rights reserved.
scales dichotomize delirium as being either pres-

Table 3. Scoring Systems for the Diagnosis of Delirium
in Critically Ill Patients.* ent or absent, although it would seem to be in-
tuitive that delirium has different degrees of se-
System, Scoring Method, and Criteria verity. The CAM-ICU and ICDSC are currently the
Confusion Assessment Method for the ICU (CAM-ICU)† two accepted methods for identifying a condition
Scoring is positive or negative according to the presence that otherwise frequently goes undiagnosed.44
or absence of criteria listed
Patient must be sufficiently awake (RASS score, −3 or more)
for assessment according to the following criteria: Pr e v en t ion a nd T r e atmen t
An acute change from mental status at baseline or of Del ir ium
fluctuating mental status during the past 24 hr
(must be true to be positive) Prevention
More than 2 errors on a 10point test of attention to
voice or pictures (must be true to be positive) There is some evidence that delirium can be pre-
If the RASS is not 0 and the above two criteria are vented. Outside the ICU, repeated reorientation,
positive, the patient is delirious noise reduction, cognitive stimulation, vision and
If the RASS is 0 and the above two criteria are posi
tive, test for disorganized thinking using 4 yes/no
hearing aids, adequate hydration, and early mo-
questions and a 2step command; >1 error means bilization can reduce the incidence of delirium in
the patient is delirious; ≤1 error excludes delirium hospitalized patients.45 Haloperidol prophylaxis
Intensive Care Delirium Screening Checklist (ICDSC)‡ in patients undergoing hip surgery reduced the
A score of ≥4 is positive for delirium (with scores of 1 to severity and duration of delirium.46 Among pa-
3 termed “subsyndromal delirium”) tients in the ICU, the duration of delirium was
Patient must show at least a response to mild or moderate cut in half with early mobilization during inter-
stimulation. Then score 1 point for each of the
following features, as assessed in the manner ruptions in sedation.47
thought appropriate by the clinician: Pharmacologic studies of delirium prevention
Anything other than “normal wakefulness” include trials comparing one sedative–analgesic
Inattention
regimen with another and studies of antipsychotic
Disorientation
Hallucination drugs administered with the specific intent of
Psychomotor agitation preventing delirium. Four placebo-controlled
Inappropriate speech or mood trials have evaluated pharmacologic prophylaxis
Disturbance in sleep or wake cycle of delirium; low-dose haloperidol48 and low-dose
Fluctuation in symptoms
risperidone49 both reduced the incidence of de-
* RASS denotes Richmond Agitation–Sedation Scale. lirium, as did a single low dose of ketamine
† Data are from Ely et al.41 during the induction of anesthesia.50 However,
‡ Data are from Bergeron et al.29 these trials were conducted among patients un-
dergoing elective surgical procedures, and it is
not clear whether their results can be extrapo-
scales are essential in objectively diagnosing de- lated to the general ICU population. In contrast,
lirium for research purposes, it is not clear that the cholinesterase inhibitor rivastigmine was
the use of these scales is more sensitive than un- ineffective in preventing delirium.51
structured assessments made by trained bedside Sedation with dexmedetomidine rather than
nurses who are prompted to look for delirium. benzodiazepines appears to reduce the incidence
Some studies have shown a high sensitivity when of delirium in the ICU. In a multicenter, ran-
such assessments are performed by bedside domized trial predominantly involving medical
nurses,42 whereas other studies have shown con- patients in the ICU, the administration of dex-
flicting results.43 Used alone (without an accom- medetomidine or midazolam resulted in similar
panying sedation scale), none of the published proportions of time within the target range of
scales distinguish hyperactive from hypoactive −2 to +1 on the Richmond Agitation–Sedation
delirium, and none of the published scales quan- Scale among patients, but those assigned to re-
tify the relative importance of individual elements ceive dexmedetomidine had a reduced risk of
of the scales despite recognition that specific delirium and spent less time undergoing me-
treatments may shorten the duration of some ele- chanical ventilation.21 As compared with a loraz-
ments and prolong the duration of others. All the epam infusion, sedation with dexmedetomidine
450 n engl j med 370;5 197

The New England

For of Medicine
Copyright 29, 2014.
Medical use rights
reserved.
Yes Specific indication for sedation No

Status epilepticus
Intracranial hypertension
Severe respiratory failure with or Assess pain and treat with opioid
without neuromuscular blockade or other drug or technique
No
Pain controlled
Assess pain and treat with opioid
or other drug or technique
Yes
No
Assess for delirium
Pain controlled
Yes
Mainly hyperactive Mainly hypoactive No
delirium delirium delirium
Target sedation to indication:
• Seizure control Yes
• Acceptable intracranial
pressure Treat with antidelirium Treat with nonpharmacologic
• Tolerance of hypercarbia medication (or measures (e.g., physical therapy,
or necessary ventilator nonpharmacologic earplugs or quiet room,
settings measures) cognitive stimulation,
• No awareness when being No repeated reorientation)
treated with neuromuscular
blocking agent
Delirium controlled
Regularly assess the need
for this level of sedation Yes
The target sedation level is
likely to be best communicated Assess need for sedative medication to achieve target RASS score of –2 to 0
using the RASS scale (lightly sedated but responsive at least to voice)
Target sedation to Do not use sedative

RASS score of –2 to 0 medication
Reassess analgesic, antidelirium, and sedative requirement regularly

(e.g., every 4 hr or with observed change)
Figure 2. Algorithm for the Coordinated Management of Pain, Agitation, and Delirium.
The application of this algorithm — which combines the use of analgesics, antidelirium agents, sedatives, and non
pharmacologic techniques — will depend on the individual situation for each patient. For example, patients under
going surgery who are suitable candidates for rapid extubation and who have little risk of delirium may be treated
only with analgesia and rapidly diminishing sedation. RASS denotes Richmond Agitation–Sedation Scale, which
ranges from −5 to +4, with more negative scores indicating deeper sedation and more positive scores indicating in
creasing agitation, and with 0 representing the appearance of calm and normal alertness.
resulted in more time at the target level of seda- level of sedation. The rates of the composite end
tion and longer survival without delirium or point of agitation, anxiety, or delirium were
coma.20 In a multicenter European trial, patients lower with dexmedetomidine than with propo-
were randomly assigned to continue treatment fol, but the rates with dexmedetomidine were
with their current sedative (midazolam or pro- equivalent to those with midazolam. When de-
pofol) or to switch to sedation with up to 1.4 μg lirium was assessed with the use of the CAM-
of dexmedetomidine per kilogram of body weight ICU 48 hours after sedation was discontinued,
per hour.19 There were no between-group differ- there were no significant differences among the
ences in the proportion of time at the target groups.

198 january 30, 2014 451
The New
only.ofNo
Medicine
Medical 2014. For
rights reserved.
Treatment the “pain, agitation, and delirium” (PAD) guide-

There is very little evidence to guide the manage- lines44 and the “spontaneous awakening and
ment of established delirium, and most existing breathing coordination, attention to the choice
trials were categorized by the investigators as pi- of sedation, delirium monitoring, and early mo-
lot studies. Only one small placebo-controlled bility and exercise” (ABCDE) bundle.56 These
trial supports the efficacy of a drug treatment for guidelines emphasize improving team commu-
established delirium in patients in the ICU. In a nication in the ICU, standardizing care process-
study of 36 patients who were randomly assigned es, and prioritizing methods to lighten sedation
to treatment with quetiapine or placebo, delirium and facilitate early mobilization and extubation.
resolved faster in patients who received quetia- Each guideline recognizes the conceptual evolu-
pine. The use of quetiapine also increased the tion from spontaneous-breathing trials and in-
number of patients who were discharged to their terruption of sedation to a comprehensive ap-
own home or to rehabilitation.52 A study of 103 proach to monitoring and managing pain,
patients who were randomly assigned to receive agitation, and delirium.
regular haloperidol, ziprasidone, or placebo
showed no significant differences in the number C onclusions
of days that patients survived without delirium or
coma.53 The single study comparing haloperidol Accumulating evidence suggests that the man-
with an atypical antipsychotic (olanzapine) agement of sedation and delirium can have an
showed equivalent efficacy.54 None of these trials important effect on the outcomes of patients
distinguished between hyperactive and hypoac- who are treated in ICUs. Currently available data
tive delirium. suggest that the best outcomes are achieved with
In a pilot study comparing dexmedetomidine the use of a protocol in which the depth of seda-
with haloperidol in patients with hyperactive de- tion and the presence of pain and delirium are
lirium, dexmedetomidine was associated with a routinely monitored, pain is treated promptly
shorter time to extubation and shorter length of and effectively, the administration of sedatives is
stay in the ICU.55 This finding is supported by a kept to the minimum necessary for the comfort
randomized trial of dexmedetomidine versus and safety of the patient, and early mobilization
midazolam in which patients with delirium at is achieved whenever possible (Fig. 2).
the time of enrollment had a more rapid resolu-
tion of delirium if they were assigned to receive Dr. Reade reports receiving grant support through his institu-
dexmedetomidine than if they were assigned to tion from Hospira. Dr. Finfer reports receiving grant support
receive midazolam.21 However, definitive evidence through his institution from Fresenius Kabi; being a member of
the International Sepsis Forum (ISF) council, which has received
supporting the use of dexmedetomidine for the funding from Eisai, Siemens, Agennix, AstraZeneca, BD Diag-
treatment of delirium is not currently available. nostics, bioMérieux, BRAHMS/Thermo Fisher, Eli Lilly, Roche,
Spectral, Toray, Philips, Apex, Ferring, BioCritica, and Plasma
Protein Therapeutics Association during his membership; and
Qua l i t y Improv emen t receiving consulting fees from Edwards (paid to ISF), lecture
Technique s fees from Eli Lilly and PPTA (donated to ISF), and travel support
from the ISF. No other potential conflict of interest relevant to
this article was reported.
Frameworks that facilitate the aforementioned Disclosure forms provided by the authors are available with
approaches have been developed. These include the full text of this article at NEJM.org.
References
1. Stein-Parbury J, McKinley S. Patients’ diator of the tumor-promoting effects of Care Pain Observation Tool and vital
experiences of being in an intensive care surgery in rats. Pain 2001;90:191-9. signs for the detection of pain in sedated
unit: a select literature review. Am J Crit 4. Myhren H, Ekeberg O, Toien K, Karls- and mechanically ventilated critically ill
Care 2000;9:20-7. son S, Stokland O. Posttraumatic stress, adults: a pilot study. Intensive Crit Care
2. Swinamer DL, Phang PT, Jones RL, anxiety and depression symptoms in pa- Nurs 2011;27:46-52.
Grace M, King EG. Effect of routine ad- tients during the first year post intensive 6. Payen JF, Bru O, Bosson JL, et al. As-
ministration of analgesia on energy ex- care unit discharge. Crit Care 2010;14(1): sessing pain in critically ill sedated pa-
penditure in critically ill patients. Chest R14. tients by using a behavioral pain scale.
1988;93:4-10. 5. Gélinas C, Tousignant-Laflamme Y, Crit Care Med 2001;29:2258-63.
3. Page GG, Blakely WP, Ben-Eliyahu S. Tanguay A, Bourgault P. Exploring the va- 7. Gélinas C, Fillion L, Puntillo KA, Viens
Evidence that postoperative pain is a me- lidity of the bispectral index, the Critical- C, Fortier M. Validation of the critical-care
452 n engl j med 370;5199

The New England

Downloaded from collections.nejm.org. Journal of
For personal useMedicine
Downloaded from nejm.org by MJ MEDAS©on
Copyright September 29,Medical
Massachusetts 2014. For personal
Society. use only.
reserved.
pain observation tool in adult patients. et al. Dexmedetomidine vs midazolam for tion. Am J Respir Crit Care Med 2009;
Am J Crit Care 2006;15:420-7. sedation of critically ill patients: a ran- 180:1092-7.
8. Kress JP, Pohlman AS, O’Connor MF, domized trial. JAMA 2009;301:489-99. 33. van den Boogaard M, Schoonhoven L,
Hall JB. Daily interruption of sedative in- 22. Muellejans B, Matthey T, Scholpp J, Evers AW, van der Hoeven JG, van Achter-
fusions in critically ill patients undergo- Schill M. Sedation in the intensive care berg T, Pickkers P. Delirium in critically
ing mechanical ventilation. N Engl J Med unit with remifentanil/propofol versus ill patients: impact on long-term health-
2000;342:1471-7. midazolam/fentanyl: a randomised, open- related quality of life and cognitive func-
9. Girard TD, Kress JP, Fuchs BD, et al. label, pharmacoeconomic trial. Crit Care tioning. Crit Care Med 2012;40:112-8.
Efficacy and safety of a paired sedation 2006;10:R91. 34. Peterson JF, Pun BT, Dittus RS, et al.
and ventilator weaning protocol for me- 23. Breen D, Karabinis A, Malbrain M, et Delirium and its motoric subtypes: a study
chanically ventilated patients in intensive al. Decreased duration of mechanical ven- of 614 critically ill patients. J Am Geriatr
care (Awakening and Breathing Con- tilation when comparing analgesia-based Soc 2006;54:479-84.
trolled trial): a randomised controlled sedation using remifentanil with standard 35. Soliman HM, Melot C, Vincent JL.
trial. Lancet 2008;371:126-34. hypnotic-based sedation for up to 10 days Sedative and analgesic practice in the in-
10. Mehta S, Burry L, Cook D, et al. Daily in intensive care unit patients: a random- tensive care unit: the results of a European
sedation interruption in mechanically ised trial. Crit Care 2005;9:R200-R210. survey. Br J Anaesth 2001;87:186-92.
ventilated critically ill patients cared for 24. Dahaba AA, Grabner T, Rehak PH, 36. De Jonghe B, Bastuji-Garin S, Fangio
with a sedation protocol: a randomized List WF, Metzler H. Remifentanil versus P, et al. Sedation algorithm in critically ill
controlled trial. JAMA 2012;308:1985-92. morphine analgesia and sedation for me- patients without acute brain injury. Crit
11. Strøm T, Martinussen T, Toft P. A pro- chanically ventilated critically ill patients: Care Med 2005;33:120-7.
tocol of no sedation for critically ill pa- a randomized double blind study. Anes- 37. Riker RR, Picard JT, Fraser GL. Pro-
tients receiving mechanical ventilation: thesiology 2004;101:640-6. spective evaluation of the Sedation-Agita-
a randomised trial. Lancet 2010;375:475- 25. Spies C, MacGuill M, Heymann A, tion Scale for adult critically ill patients.
80. et al. A prospective, randomized, double- Crit Care Med 1999;27:1325-9.
12. Shehabi Y, Bellomo R, Reade MC, et al. blind, multicenter study comparing remi- 38. Sessler CN, Gosnell MS, Grap MJ, et al.
Early intensive care sedation predicts long- fentanil with fentanyl in mechanically The Richmond Agitation-Sedation Scale:
term mortality in ventilated critically ill ventilated patients. Intensive Care Med validity and reliability in adult intensive
patients. Am J Respir Crit Care Med 2012; 2011;37:469-76. care unit patients. Am J Respir Crit Care
186:724-31. 26. Diagnostic and statistical manual of Med 2002;166:1338-44.
13. Treggiari MM, Romand JA, Yanez ND, mental disorders, 4th ed. text rev.: DSM- 39. Pun BT, Dunn J. The sedation of criti-
et al. Randomized trial of light versus IV-TR. Arlington, VA: American Psychiat- cally ill adults — part 1: assessment: the
deep sedation on mental health after crit- ric Association, 2011. first in a two-part series focuses on as-
ical illness. Crit Care Med 2009;37:2527- 27. Gunther ML, Morandi A, Krauskopf E, sessing sedated patients in the ICU. Am J
34. et al. The association between brain vol- Nurs 2007;107:40-8.
14. Kress JP, Gehlbach B, Lacy M, Pliskin umes, delirium duration, and cognitive 40. van Eijk MM, van Marum RJ, Klijn IA,
N, Pohlman AS, Hall JB. The long-term outcomes in intensive care unit survivors: de Wit N, Kesecioglu J, Slooter AJ. Com-
psychological effects of daily sedative in- the VISIONS cohort magnetic resonance parison of delirium assessment tools in a
terruption on critically ill patients. Am J imaging study. Crit Care Med 2012;40: mixed intensive care unit. Crit Care Med
Respir Crit Care Med 2003;168:1457-61. 2022-32. 2009;37:1881-5.
15. Strøm T, Stylsvig M, Toft P. Long-term 28. Morandi A, Rogers BP, Gunther ML, 41. Ely EW, Inouye SK, Bernard GR, et al.
psychological effects of a no-sedation pro- et al. The relationship between delirium Delirium in mechanically ventilated pa-
tocol in critically ill patients. Crit Care duration, white matter integrity, and cog- tients: validity and reliability of the Con-
2011;15:R293. nitive impairment in intensive care unit fusion Assessment Method for the Inten-
16. Jackson DL, Proudfoot CW, Cann KF, survivors as determined by diffusion tensive Care Unit (CAM-ICU). JAMA 2001;
Walsh TS. The incidence of sub-optimal sor imaging: the VISIONS prospective co- 286:2703-10.
sedation in the ICU: a systematic review. hort magnetic resonance imaging study. 42. Vasilevskis EE, Morandi A, Boehm L,
Crit Care 2009;13:R204. Crit Care Med 2012;40:2182-9. et al. Delirium and sedation recognition
17. Roberts DJ, Haroon B, Hall RI. Seda- 29. Bergeron N, Dubois MJ, Dumont M, using validated instruments: reliability of
tion for critically ill or injured adults in Dial S, Skrobik Y. Intensive Care Delirium bedside intensive care unit nursing as-
the intensive care unit: a shifting para- Screening Checklist: evaluation of a new sessments from 2007 to 2010. J Am Geri-
digm. Drugs 2012;72:1881-916. screening tool. Intensive Care Med 2001; atr Soc 2011;59:Suppl 2:S249-S255.
18. Ho KM, Ng JY. The use of propofol for 27:859-64. 43. Reade MC, Eastwood GM, Peck L,
medium and long-term sedation in criti- 30. Ely EW, Girard TD, Shintani AK, et al. Bellomo R, Baldwin I. Routine use of the
cally ill adult patients: a meta-analysis. Apolipoprotein E4 polymorphism as a Confusion Assessment Method for the In-
Intensive Care Med 2008;34:1969-79. genetic predisposition to delirium in crit- tensive Care Unit (CAM-ICU) by bedside
19. Jakob SM, Ruokonen E, Grounds RM, ically ill patients. Crit Care Med 2007;35: nurses may underdiagnose delirium. Crit
et al. Dexmedetomidine vs midazolam or 112-7. Care Resusc 2011;13:217-24.
propofol for sedation during prolonged 31. van den Boogaard M, Pickkers P, 44. Barr J, Fraser GL, Puntillo K, et al.
mechanical ventilation: two randomized Slooter AJ, et al. Development and vali- Clinical practice guidelines for the man-
controlled trials. JAMA 2012;307:1151- dation of PRE-DELIRIC (PREdiction of agement of pain, agitation, and delirium
60. DELIRium in ICu patients) delirium pre- in adult patients in the intensive care unit.
20. Pandharipande PP, Pun BT, Herr DL, diction model for intensive care patients: Crit Care Med 2013;41:263-306.
et al. Effect of sedation with dexmedeto- observational multicentre study. BMJ 2012; 45. Vidán MT, Sanchez E, Alonso M,
midine vs lorazepam on acute brain dys- 344:e420. Montero B, Ortiz J, Serra JA. An interven-
function in mechanically ventilated pa- 32. Pisani MA, Kong SY, Kasl SV, Murphy tion integrated into daily clinical practice
tients: the MENDS randomized controlled TE, Araujo KL, Van Ness PH. Days of de- reduces the incidence of delirium during
trial. JAMA 2007;298:2644-53. lirium are associated with 1-year mortal- hospitalization in elderly patients. J Am
21. Riker RR, Shehabi Y, Bokesch PM, ity in an older intensive care unit popula- Geriatr Soc 2009;57:2029-36.

The New
use only. Medicine
rights reserved.
46. Kalisvaart KJ, de Jonghe JF, Bogaards 50. Hudetz JA, Patterson KM, Iqbal Z, lirium: the MIND randomized, placebo-
MJ, et al. Haloperidol prophylaxis for el- et al. Ketamine attenuates delirium after controlled trial. Crit Care Med 2010;38:
derly hip-surgery patients at risk for de- cardiac surgery with cardiopulmonary by- 428-37.
lirium: a randomized placebo-controlled pass. J Cardiothorac Vasc Anesth 2009;23: 54. Skrobik YK, Bergeron N, Dumont M,
study. J Am Geriatr Soc 2005;53:1658- 651-7. Gottfried SB. Olanzapine vs haloperidol:
66. 51. Gamberini M, Bolliger D, Lurati Buse treating delirium in a critical care setting.
47. Schweickert WD, Pohlman MC, Pohl- GA, et al. Rivastigmine for the prevention Intensive Care Med 2004;30:444-9.
man AS, et al. Early physical and occupa- of postoperative delirium in elderly pa- 55. Reade MC, O’Sullivan K, Bates S, Gold-
tional therapy in mechanically ventilated, tients undergoing elective cardiac surgery smith D, Ainslie WR, Bellomo R. Dexme-
critically ill patients: a randomised con- — a randomized controlled trial. Crit detomidine vs. haloperidol in delirious,
trolled trial. Lancet 2009;373:1874-82. Care Med 2009;37:1762-8. agitated, intubated patients: a randomised
48. Wang W, Li HL, Wang DX, et al. Halo- 52. Devlin JW, Roberts RJ, Fong JJ, et al. open-label trial. Crit Care 2009;13:R75.
peridol prophylaxis decreases delirium Efficacy and safety of quetiapine in criti- 56. Vasilevskis EE, Pandharipande PP,
incidence in elderly patients after noncar- cally ill patients with delirium: a prospec- Girard TD, Ely EW. A screening, preven-
diac surgery: a randomized controlled tive, multicenter, randomized, double- tion, and restoration model for saving the
trial. Crit Care Med 2012;40:731-9. blind, placebo-controlled pilot study. Crit injured brain in intensive care unit survi-
49. Prakanrattana U, Prapaitrakool S. Ef- Care Med 2010;38:419-27. vors. Crit Care Med 2010;38:Suppl:S683-
ficacy of risperidone for prevention of 53. Girard TD, Pandharipande PP, Carson S691.
postoperative delirium in cardiac surgery. SS, et al. Feasibility, efficacy, and safety of Copyright © 2014 Massachusetts Medical Society.
Anaesth Intensive Care 2007;35:714-9. antipsychotics for intensive care unit de-

201 january 30, 2014
The New England

Medical personalAll
userights
reserved.
c or r e sp ondence
Sedation and Delirium in Intensive Care

To the Editor: We agree with Reade and Finfer 4. Bourne RS, Mills GH. Melatonin: possible implications for
the postoperative and critically ill patient. Intensive Care Med
(Jan. 30 issue)1 that prevention of delirium in pa- 2006;32:371-9.
tients in the intensive care unit (ICU) is clearly 5. Bourne RS, Mills GH, Minelli C. Melatonin therapy to im-
preferable to treatment after the fact. prove nocturnal sleep in critically ill patients: encouraging re-
sults from a small randomised controlled trial. Crit Care 2008;
Reduction or prevention of sleep deprivation 12:R52.
may be an additional measure to reduce the risk of DOI: 10.1056/NEJMc1402402
delirium in the ICU.2 Noise, light, and other fac-
tors may be associated with poor sleep quality. A
reduction in these factors and promotion of posi- To the Editor: Delirium in patients in acute
tive sleep-hygiene behaviors are associated with a care settings is common, underrecognized, costly,
reduction in the incidence of delirium or coma.3 and potentially deadly. Pharmacologic approach-
Because most sedatives degrade sleep architec- es to prevent or treat it are modestly successful at
ture, their use may actually increase the risk of best and have clear potential to harm. In con-
delirium, rather than reduce it.2 In addition to trast, family-based interventions to comfort con-
minimizing the use of sedative agents, other fused patients predate modern medicine. Their
therapies warrant mention. Critically ill patients use and feasibility are discussed in the medical
have a documented loss of a normal circadian literature,1-3 but data from large, high-quality
rhythm associated with melatonin secretion.4 Ex- trials are lacking. As volunteer “specialists” who
ogenous melatonin may be an effective counter- know and understand the patient, families are
measure to reduce abnormalities in sleep archi- often motivated to help prevent and resolve de-
tecture, and preliminary data to this effect are lirium. Future research should focus on this
encouraging.5 Data are lacking from studies to common-sense, pragmatic, inexpensive, and hu-
evaluate the effects of melatonin on sleep, delir- mane approach to prevent and treat delirium.
ium, and acute brain dysfunction in critically ill Bert Govig, M.D., M.P.H.
patients. We hope that future studies of delirium McGill University
in the ICU include this therapy. Amos, QC, Canada
bert@govig.ca
Alexander H. Flannery, Pharm.D. No potential conflict of interest relevant to this letter was re-
Kevin W. Hatton, M.D. ported.
Barbara Phillips, M.D., M.S.P.H. 1. Wright LM, Leahey M. Families and life-threatening illness:
University of Kentucky HealthCare assumptions, assessment, and interventions. In: Leahey M,
Lexington, KY Wright LM, eds. Families and life-threatening illness. Spring-
alex.flannery@uky.edu house, PA: Springhouse, 1987:45-58.
2. Leahey M, Harper-Jaques S. Family-nurse relationships: core
assumptions and clinical implications. J Fam Nurs 1996;2:133-51.
ported.
3. Rosenbloom-Brunton DA, Henneman EA, Inouye SK. Feasi-
1. Reade MC, Finfer S. Sedation and delirium in the intensive bility of family participation in a delirium prevention program
care unit. N Engl J Med 2014;370:444-54. for hospitalized older adults. J Gerontol Nurs 2010;36:22-33.
2. Weinhouse GL, Schwab RJ, Watson PL, et al. Bench-to-bed- DOI: 10.1056/NEJMc1402402
side review: delirium in ICU patients — importance of sleep
deprivation. Crit Care 2009;13:234.
3. Kamdar BB, King LM, Collop NA, et al. The effect of a qual-
The Authors Reply: We agree with Flannery et al.
ity improvement intervention on perceived sleep quality and cog-
nition in a medical ICU. Crit Care Med 2013;41:800-9. that process-of-care interventions to promote

202 april 17, 2014
correspondence
sleep in critically ill patients are feasible, although obvious that the benefit of family involvement is
considerable effort may be required to attain at likely and the risk low; thus, without waiting for
best 60 to 80% compliance.1 Preliminary data such evidence, we would encourage critical care
suggest that such efforts may reduce the inci- practitioners to afford the families and friends
dence of delirium,2 but data from definitive stud- of critically ill patients as much access as pos-
ies are lacking. More effort is warranted both to sible, when feasible, to involve them in efforts to
test the effectiveness of “bundles” of therapies maintain patients’ orientation and normal sleep–
for delirium prevention and management (espe- wake cycles, and to assist in early mobilization
cially nonpharmacologic approaches) and to op- as recovery occurs.
timize methods of translating them into routine Michael C. Reade, M.B., B.S., D.Phil.
clinical practice. A melatonin agonist approved University of Queensland
by the Food and Drug Administration has shown Brisbane, QLD, Australia
m.reade@uq.edu.au
promise; it was associated with a reduction in the
incidence and frequency of delirium in a small Simon Finfer, M.D.
George Institute for Global Health
trial in which 24 of the 67 patients enrolled were Sydney, NSW, Australia
in the ICU.3 Once again, more definitive evidence Since publication of their article, the authors report no fur-
is awaited. ther potential conflict of interest.
Govig advocates family-based interventions in 1. Kamdar BB, Yang J, King LM, et al. Developing, implement-
critical illness, and it is highly likely that these ing, and evaluating a multifaceted quality improvement inter-
vention to promote sleep in an ICU. Am J Med Qual 2013 Novem-
measures can benefit critically ill patients — as ber 22 (Epub ahead of print).
has already been shown with family involvement 2. Kamdar BB, King LM, Collop NA, et al. The effect of a qual-
in early mobilization.4 However, testing this ap- ity improvement intervention on perceived sleep quality and cog-
nition in a medical ICU. Crit Care Med 2013;41:800-9.
proach in clinical trials poses considerable dif- 3. Hatta K, Kishi Y, Wada K, et al. Preventive effects of ramelteon
ficulty, including the assignment of patients to on delirium: a randomized placebo-controlled trial. JAMA Psy-
a control group with less family involvement. chiatry 2014 February 14 (Epub ahead of print).
4. Rukstele CD, Gagnon MM. Making strides in preventing
Cluster-randomization trials or trials using his- ICU-acquired weakness: involving family in early progressive
torical controls could facilitate the encourage- mobility. Crit Care Nurs Q 2013;36(1):141-7.
ment of family involvement without affecting a DOI: 10.1056/NEJMc1402402
standard-care control group, but they still might Correspondence Copyright © 2014 Massachusetts Medical Society.
not allay concerns about external validity. It seems
n engl j med 370;16 203


What sedation should be provided to this patient?
A. Intermittent intravenous lorazepam.

B. Continuous intravenous infusions of propofol to facilitate daily cessation of sedation to assess the ongoing need
for sedation and analgesia.
C. An intravenous infusion of alpha-2-adrenoceptor agonist such as dexmedetomidine with daily cessation to as-
sess the ongoing need for sedation.
D. I would not sedate this patient.

The patient in the vignette has undergone major surgery in the form of a laparotomy, which makes assessment of
his pain and provision of adequate analgesia essential components of his care.
Assessing this patient’s pain may be difficult initially, since he is unlikely to be sufficiently interactive to give valid
self-reports of pain. Physiological indicators, such as hypertension and tachycardia, correlate poorly with more
valid measures of pain.
Traditionally, patients who are undergoing mechanical ventilation in the ICU have been sedated to ensure ventila-
tor synchrony and to prevent self-harm through the accidental removal of vascular access lines or the endotrache-
al tube. Benzodiazepines have been the most commonly used drugs, but more recently the short-acting anesthetic
agent propofol and the α2-adrenoceptor agonist dexmedetomidine have become popular agents to sedate patients
in the ICU. Our patient’s age and his history of previous heavy alcohol intake and mild cognitive impairment
place him at high risk for delirium, which is associated with reduced survival and worse long-term cognitive func-
tion. Current opinion favors maintaining the minimum possible level of sedation and routinely monitoring the
depth of sedation and targeting the lightest possible level of sedation. Some practitioners monitor sedation using a
scoring system such as the Riker Sedation–Agitation Scale (SAS) or the Richmond Agitation–Sedation Scale
(RASS). Daily interruption of sedation with spontaneous-breathing trials has previously been shown to reduce the
duration of mechanical ventilation.1 However, a recent trial indicated that for adults who are undergoing mechan-
ical ventilation managed with protocolized light sedation, the addition of daily sedation interruption did not re-
duce the duration of mechanical ventilation or ICU stay.2 Many practitioners prefer to avoid sedation altogether. A
randomized, controlled trial involving patients undergoing mechanical ventilation who received morphine for the
treatment of pain in an “analgesia first” approach compared a protocol of no sedation with the routine use of se-
dation with daily sedation interruption. In this trial, patients who were assigned to receive no sedation had shorter
stays in the ICU and in the hospital and more days without mechanical ventilation.3
If a sedative agent is needed for agitation, the use of short-acting agents in the minimum possible dose appears to
be beneficial, although current evidence has not identified one sedative drug that is superior to all others. The
avoidance of benzodiazepines may reduce the risk of delirium.
204
References
1. Girard TD, Kress JP, Fuchs BD, et al. Efficacy and safety of a paired sedation and ventilator weaning protocol
for mechanically ventilated patients in intensive care (Awakening and Breathing Controlled trial): a ran-
domised controlled trial. Lancet 2008;371:126-34.
2. Mehta S, Burry L, Cook DJ, et al. Daily sedation interruption in mechanically ventilated critically ill patients
cared for with a sedation protocol: a randomized controlled trial. JAMA 2012;308:1985-92.
3. Strøm T, Martinussen T, Toft P. A protocol of no sedation for critically ill patients receiving mechanical venti-
lation: a randomised trial. Lancet 2010;375:475-80.
205
BLEEDING AND COAGULOPATHIES

IN CRITICAL CARE
Everyone in the ICU seems to be clotting too much or too little. This section provides guidance on assessing
coagulation in your ICU patient, along with ways of treating coagulopathies once recognized.
206
Case Challenge
Bleeding and Coagulopathies in Critical Care

A 77-year old man is undergoing mechanical ventilation after an emergency colon resection, complicated by
septic shock and then acute liver failure. Nurses note slight bleeding from the surgical site and from around
arterial and central venous catheters. What should be done about the bleeding?
hospital from the operating room after a Hartmann’s procedure performed for fecal peritonitis due to a perforated
sigmoid colon. On arrival in the ICU, he was undergoing mechanical ventilation and was in septic shock. He re-
ceived fluid resuscitation with Hartmann’s solution and 5% human albumin solution. His blood pressure was sup-
ported with a norepinephrine infusion. In addition to intravenous antibiotics, his treatment included subcutane-
ous unfractionated heparin (5000 U twice daily). On the day after admission to the ICU, his blood pressure is
stable and supported by a reducing dose of norepinephrine. Analgesia is being provided by a continuous morphine
infusion. However, despite adequate analgesia, he intermittently requires sedation in the form of low-dose propo-
fol by infusion to tolerate the endotracheal tube. (In the previous installment of this case, there were 3894 votes on
strategies for treating delirium and the use of sedation. A total of 15% of respondents favored the use intermittent
207
intravenous lorazepam, 28% favored continuous intravenous infusion of propofol to facilitate daily cessation of se-
dation, 36% favored an intravenous infusion of an α2-adrenoceptor agonist such as dexmedetomidine with daily
cessation, and 18% favored no sedation.)
The patient’s nurse notes slight bleeding from his abdominal wound and from around his arterial and central ve-
nous catheters. Laboratory test results from that morning include normal liver-function tests, a hemoglobin level
of 8.9 g per deciliter, a platelet count of 54,000 per cubic millimeter, an international normalized ratio of 1.6, an
activated partial-thromboplastin time (APTT) of 52 seconds, and a fibrinogen level of 1.7 g per liter.

How would you assess and manage his bleeding?
A. Cease heparin and transfuse packed red cells, platelets, cryoprecipitate, and fresh-frozen plasma or prothrombin
complex concentrate.
B. Cease heparin and transfuse cryoprecipitate and fresh-frozen plasma or prothrombin complex concentrate.
C. Cease heparin, transfuse platelets, administer an antifibrinolytic agent (e.g., epsilon-aminocaproic acid or
tranexamic acid).
D. Cease heparin and repeat laboratory tests in 8 to 12 hours.
208
review article
Bleeding and Coagulopathies

in Critical Care
Beverley J. Hunt, M.D.
T
he definition of coagulopathy is “a condition in which the From King’s College London and Guy’s
blood’s ability to clot is impaired.” However, for some clinicians, the term and St. Thomas’ Trust — both in London.
Address reprint requests to Dr. Hunt at
also covers thrombotic states, and because of the complexity of the hemo- the Thrombosis and Haemophilia Centre,
static pathways, the two conditions can exist simultaneously. Some practitioners St Thomas’ Hospital, Westminster Bridge
would consider that mildly abnormal results on coagulation screening without Rd., London SE1 7EH, United Kingdom,
or at beverley.hunt@gstt.nhs.uk.
bleeding can also indicate a coagulopathy. This review is confined to the original
definition of coagulopathy as given above. Such states are common in patients in N Engl J Med 2014;370:847-59.
DOI: 10.1056/NEJMra1208626
the intensive care unit (ICU) and require a clinicopathological approach to ensure Copyright © 2014 Massachusetts Medical Society.
that the correct diagnosis is made and the appropriate treatment administered. The
lack of evidence for managing coagulopathies in critical care is striking. This re-
view will highlight selected areas in which there is high-quality evidence and at the
same time point out areas for which there is poor evidence. In the latter case, there
is little consensus on management.
Differ en t i a l Di agnosis
A medical history taking and physical examination are vital, since many different
conditions can produce similar laboratory abnormalities. For example, end-stage
liver failure and disseminated intravascular coagulation produce thrombocytopenia
and similar changes in standard tests of coagulation, and yet the management of
and prognosis for these conditions are very different. A peripheral-blood smear is
a vital investigation tool in most cases to confirm a low platelet count and the pres-
ence or absence of other diagnostic features, such as red-cell fragmentation, plate-
let morphologic abnormalities, or evidence of dysplasia or hematinic deficiency.
Table 1 and Figure 1 highlight the relationship between laboratory findings and
various coagulopathies.
Once it has been determined that the underlying cause is not a response to
therapeutic agents meant to modify the coagulation response (e.g., treatment with
vitamin K antagonists, heparinoids, or direct factor Xa or IIa inhibitors), practitioners
need to evaluate the pattern of bleeding, which may include widespread petechiae
and mucosal bleeding in platelet disorders, generalized oozing from de-epithelialized
surfaces, and fast bleeding from damaged major vessels.
M a nagemen t of C oagul opathie s
The first principle of the management of coagulopathies in critical care is to avoid

the correction of laboratory abnormalities with blood products unless there is a
clinical bleeding problem, a surgical procedure is required, or both.
Major Bleeding
The lack of good-quality evidence is most marked in the use of blood components to
manage major bleeding. When blood components were introduced into critical care
practice decades ago, their benefit was never assessed in randomized clinical trials.

february 27, 2014 847
The New England
Medical personal
Society. use only.
reserved.
Table 1. Laboratory Findings in Various Platelet and Coagulation Disorders in the ICU.
Prothrombin Activated Partial- Fibrinogen d-Dimer Bleeding Platelet Findings on

Condition Time Thromboplastin Time Level Level Time Count Blood Smear
Vitamin K deficiency or use Normal or
of vitamin K antagonist Prolonged mildly prolonged Normal Unaffected Unaffected Unaffected
Aspirin or thienopyridines Unaffected Unaffected Unaffected Unaffected Prolonged Unaffected
Liver failure
Early stage Prolonged Unaffected Unaffected Unaffected Unaffected Unaffected
End stage Prolonged Prolonged Low Increased Prolonged Decreased
Uremia Unaffected Unaffected Unaffected Unaffected Prolonged Unaffected
Disseminated intravascular Prolonged Prolonged Low Increased Prolonged Decreased Fragmented
coagulation red cells
Thrombotic thrombocytopenic Unaffected Unaffected Unaffected Unaffected Prolonged Very low Fragmented
purpura red cells
Hyperfibrinolysis Prolonged Prolonged Low Very high Possibly Unaffected
prolonged
Later, concern about transfusion-transmitted infec- an increase by a factor of 6 in the multiple organ
tions (human immunodeficiency virus infection, dysfunction syndrome.5
hepatitis, and a new variant of Creutzfeldt–Jakob The critical transfusion ratio of fresh-frozen
disease) and limitations in the blood supply led to plasma to red cells in the management of major
a more restrictive use of blood components. bleeding is not known. This question is being
In the absence of randomized, controlled trials, evaluated in the North American Pragmatic, Ran-
retrospective studies of military casualties1 and, domized Optimal Platelets and Plasma Ratios
later, similar studies of civilian casualties2 show- study (ClinicalTrials.gov number, NCT01545232).
ing improved survival with transfusion of 1 U of This multicenter, randomized trial is comparing
fresh-frozen plasma for each unit of red cells the effect of various ratios of blood products
have resulted in earlier administration of an in- administered to trauma patients who are pre-
creased number of units of fresh-frozen plasma. dicted to require massive transfusion (>10 U of
However, these studies have been criticized, packed red cells within the next 24 hours) on
particularly for methodologic flaws that include rates of death at 24 hours and 30 days. In the
survival bias (i.e., patients who did not survive interim, a North American–European divide in
were not transfused with fresh-frozen plasma) the practice of using blood components to sup-
and heterogeneity between studies.3 port hemostasis has emerged. Although in North
Despite the lack of evidence that bleeding America there has been increased use of fresh-
after surgery and gastrointestinal or obstetric frozen plasma in patients with major hemor-
hemorrhage are associated with hemostatic rhage, some European practitioners have aban-
changes similar to those in acute traumatic co- doned the use of fresh-frozen plasma, relying on
agulopathy, the early use of a transfusion ratio the exclusive use of factor concentrates on the
of fresh-frozen plasma to red cells of 1:1 or 1:2 basis of rotational-elastometry–guided interven-
has become widespread. This increased use of tion with prothrombin complex concentrate,
plasma is not risk-free, since the incidence of factor XIII, and fibrinogen.6 In contrast, other
transfusion-related acute lung injury is increased,4 practitioners believe that the treatment of major
as may be the risk of the acute respiratory dis- hemorrhage should begin with fibrinogen sup-
tress syndrome (ARDS). In one study involving plementation with tranexamic acid, a synthetic
trauma patients requiring a nonmassive transfu- derivative of the amino acid lysine that acts as an
sion (<10 U of packed red cells within 12 hours antifibrinolytic agent by competitively inhibiting
after admission), the administration of more than plasminogen, with red cells and intravenous fluid
6 units of fresh-frozen plasma, as compared used on an as-needed basis.7
with no transfusion, was associated with an in- Fibrinogen is a critical molecule in coagula-
crease by a factor of 12 in the rate of ARDS and tion. It is the protein that ultimately forms fibrin,
848 n engl j med210

370;9 nejm.org february 27, 2014
The New
Copyright © Massachusettson Medical
January 15, 2015.All
Society. Forrights
reserved.
the ligand for platelet aggregation, and in pa- outcomes in patients with major bleeding. Future
tients with major bleeding, it is required to a randomized, controlled trials should assess the
larger extent than any other hemostatic protein.8 overall benefit and safety, including the rate of
In such patients, this requirement reflects in- hospital-acquired venous thromboembolism.10,11
creased consumption, loss, dilution, and fibrin- Similarly, the use of recombinant factor VIIa,
ogenolysis. On the basis of these multiple roles, which has been shown to reduce the use of red
even in the absence of evidence from random- cells in bleeding but not to reduce mortality,
ized, controlled trials, guidelines for the man- needs further evaluation. Data from placebo-
agement of traumatic bleeding now indicate that controlled trials have shown that the off-label
the trigger level for supplementing fibrinogen use of recombinant factor VIIa significantly in-
should be 1.5 to 2.0 g per liter rather than 1.0 g creased the risk of arterial thrombosis.12,13
per liter.9 It is unknown whether early fibrino- Tranexamic acid should be administered to
gen supplementation and the use of prothrom- all patients with major bleeding after trauma. This
bin complex concentrate, as compared with the recommendation is supported by a large, random-
use of fresh-frozen plasma, improves clinical ized, controlled trial, the Clinical Randomization
History: Rule out inherited defect or use of antithrombotic drugs.
Examination: Is bleeding general or local? General bleeding Local bleeding
Coagulation screening and full blood count
Low platelet count with normal Low platelet count and Normal platelet count with Low platelet count with
results on coagulation screening fragmented red coagulation deficiencies coagulation deficiencies
cells (microangiopathic
hemolytic anemia)
Disseminated
Reduced
intravascular
survival
Failed production of coagulation
coagulation factors
Failed production Reduced Underlying

of platelets survival disorder
Increased splenic Schistocyte
pooling
Stem cell
Factor VIII
Antibody-mediated Consumption of Widespread
Acquired hemophilia platelets and fibrin deposition
platelet destruction
clotting factors
Splenomegaly
Megakaryocytosis
Thrombocytopenia
and coagulation factor Microvascular
deficiency thrombotic obstruction
Release of
platelets
Megakaryocyte Bleeding
Organ failure
Figure 1. Causes of Bleeding among Patients in the ICU.

COLOR FIGURE
After the presence of inherited disorders and the use of antithrombotic drugs have been ruled out, the first major question (“Is the bleeding
Draft 7 2/11/14
general or local?”), combined with a platelet count and coagulation screening, will assist in the identification of the pathogenesis of bleeding.
Author Hunt
Fig # 1
Title Bleeding and coagulopathies
in critical care
n engl j med 370;9 nejm.org february
211 27, 2014 ME 849
DE Drazen
The New England
Artist N Koscal
Medical personal
Society. use only.
reserved. AUTHOR PLEASE NOTE:
Copyright © 2014 Massachusetts Medical Society. All rights reserved. Figure has been redrawn and type has been reset
Issue date 02/27/2014
of an Antifibrinolytic in Significant Hemorrhage of vitamin K1 (at a dose of at least 1 mg orally

(CRASH-2) study, in which 20,000 trauma patients daily or 10 mg intravenously weekly) for critical
with bleeding or at risk for major bleeding were care patients at risk.
randomly assigned to receive either tranexamic
acid or placebo. Patients who received tranexamic Disseminated In t r ava scul a r
acid within 3 hours after injury had a one-third C oagul at ion
reduction in deaths from bleeding.14 After a sec-
ondary analysis of their data, the CRASH-2 in- Disseminated intravascular coagulation is a clini-
vestigators recommended that tranexamic acid copathological diagnosis21 of a disorder that is de-
be administered as soon as possible after injury, fined by the International Society on Thrombosis
since the drug ceased to confer benefit and ap- and Hemostasis (ISTH) as “an acquired syndrome
peared to be associated with increased mortality characterized by the intravascular activation of co-
if it was administered more than 3 hours after agulation with loss of localization arising from
injury.15 Reassuringly for a hemostatic drug, the different causes.” This condition typically origi-
incidence of thrombosis after trauma was not nates in the microvasculature and can cause dam-
increased in the study patients. Strong evidence age of such severity that it leads to organ dysfunc-
that tranexamic acid reduced the need for blood tion (Fig. 2). It can be identified on the basis of a
transfusion in surgery has been available for scoring system developed by the ISTH (Table 2).
years, although the effect of tranexamic acid on Disseminated intravascular coagulation usually
thromboembolic events and mortality in such presents as hemorrhage, with only 5 to 10% of
patients remains uncertain.16 cases presenting with microthrombi (e.g., digital
ischemia) alone. Whether the condition presents
with a thrombotic or bleeding episode depends
Hemos tat ic Supp or t for
In va si v e Pro cedur e s on its cause and host defenses. Sepsis is the
most common cause of disseminated intravascu-
There is no supportive evidence for the prophy- lar coagulation in critical care; systemic infec-
lactic use of fresh-frozen plasma to correct ab- tion with a range of bacteria from Staphylococcus
normal results on coagulation screening (for pro- aureus to Escherichia coli is known to be associated
thrombin time, activated partial-thromboplastin with this condition. The complex pathophysiology
time, and fibrinogen) before an invasive proce- is mediated by pathogen-associated molecular pat-
dure. Coagulation screening has no predictive terns, which generate an inflammatory response
value for later bleeding, and the use of fresh-fro- in the host through signaling at specific recep-
zen plasma may not correct abnormal results on tors. For example, signaling by means of toll-like
coagulation tests. There is currently no consensus and complement receptors initiates intracellular
on what coagulation-screening results should trig- signaling, which results in the synthesis of sev-
ger the use of fresh-frozen plasma, which has led eral proteins (including proinflammatory cyto-
to variation in the use of fresh-frozen plasma by kines). These proteins trigger hemostatic chang-
critical care physicians.17,18 Thrombin generation, es, leading to the up-regulation of tissue factor22
although delayed, is normal or enhanced19 when and impairment of physiologic anticoagulants
the prothrombin ratio is 1.5 or less, so I suggest and fibrinolysis. Tissue factor plays a critical role
that a prothrombin ratio of 1.5 or less is satisfac- in this process, as shown in meningococcal septi-
tory for the insertion of a central venous or an cemia, in which the level of tissue factor on mono-
arterial catheter in patients in whom direct com- cytes at presentation may be predictive of sur-
pression can assist with hemostasis without the vival.23 Another study of meningococcal sepsis
need for prophylactic supplementation with fresh- showed that a large amount of tissue factor was
frozen plasma. found on monocyte-derived circulating micropar-
As a general rule, dietary intake of vitamin K, ticles.24 The up-regulation of tissue factor activates
which is necessary for the formation of coagu- coagulation, leading to the widespread deposition
lation factors II, VII, IX, and X, may be inade- of fibrin and to microvascular thromboses and
quate in critical care settings.20 Despite the may contribute to multiple organ dysfunction.
lack of high-quality evidence and the inability Complex abnormalities of the physiologic anti-
of vitamin K to correct a coagulopathy caused coagulants occur, and pharmacologic doses of
by liver disease, I recommend supplementation activated protein C, antithrombin, and tissue
850 n engl j med 370;9 212

The New EnglandFor

Society. use only.
reserved.
Sepsis
Microparticles
Monocyte
Tissue factor Microvascular

expression Fibrin deposition
Proinflammatory thrombus
cytokines
Insufficient
Impairment of removal
anticoagulant
mechanisms
Plasminogen
activator inhibitor 1
Endothelial cell Activation of
endothelial cells
Figure 2. Pathogenesis of Disseminated Intravascular Coagulation in Sepsis.

Through the generation of proinflammatory cytokines and the activation of monocytes, bacteria cause the up-regulation of tissue factor
COLOR FIGURE
as well as the release of microparticles expressing tissue factor, thus leading to the activation of coagulation. Proinflammatory cytokines
Draft 4
also cause the activation of endothelial cells, a process that impairs anticoagulant mechanisms and down-regulates fibrinolysis by 1/30/14
gener-
Author Hunt
ating increased amounts of plasminogen activator inhibitor. Fig # 2
Title Bleeding and coagulopathies
in critical care
ME
DE Drazen
Artist N Koscal
Table 2. Diagnostic Scoring System for Disseminated Intravascular Coagulation (DIC).*
AUTHOR PLEASE NOTE:
Risk assessment: Does the patient have an underlying disorder known to be associated with overt DIC? Please check carefully

If yes, proceed with this algorithm
If no, do not use this algorithm
Order global coagulation tests (prothrombin time, platelet count, fibrinogen, fibrin-related marker)
Score the test results as follows:
Platelet count: 50,000 to 100,000 per mm3, 1 point; <50,000 per mm3, 2 points
Elevated fibrin-related marker (e.g., d-dimer, fibrin degradation products): no increase, 0 points; moderate increase, 2 points; strong
increase, 3 points
Prolonged prothrombin time: <3 sec, 0 points; ≥3 sec but <6 sec, 1 point; ≥6 sec, 2 points
Fibrinogen level: ≥1 g per liter, 0 points; <1 g per liter, 1 point
Calculate the score as follows:
≥5 points: compatible with overt DIC; repeat scoring daily
<5 points: suggestive of nonovert DIC; repeat scoring within next 1 to 2 days
* Data are adapted from Toh and Hoots21 on the basis of the scoring system developed by the International Society on Thrombosis and Hemostasis.
factor pathway inhibitor appeared to be benefi- activated protein C,25 antithrombin,26 and tissue
cial in a study of endotoxemia in animals. These factor pathway inhibitor27 in patients with sepsis.
promising studies led to major randomized, con- However, the studies showed no reductions in the
trolled trials of the supplementation of physio- rates of death and increased bleeding episodes.
logic anticoagulants with pharmacologic doses of The consumption of the coagulation proteins

213 27, 2014 851
The New England
reserved.
The cornerstone for managing this condition

Table 3. Differential Diagnosis of Thrombocytopenia in the ICU.
remains the management of the underlying cause
First, rule out pseudothrombocytopenia by asking the following question: (e.g., sepsis). Further management may not be
Is the blood sample clotted? necessary in patients with mild abnormalities in
Check for EDTA-dependent platelet antibodies by collecting the sample in coagulation and no evidence of bleeding. Guide-
an anticoagulant (e.g., citrate) lines for management are based mainly on ex-
If pseudothrombocytopenia has been ruled out, ask the following: pert opinion, which suggests replacement of
Is the patient taking drugs that could lower the platelet count? coagulation proteins and platelets in patients
Check for receipt of: who are bleeding. Platelet transfusion is indi-
Heparin, which may be associated with heparin-induced thrombocytopenia cated to maintain a platelet level of more than
IIb/IIIa inhibitors (e.g., abciximab, eptifibatide, tirofiban) 50,000 per cubic millimeter, along with the admin-
Adenosine diphosphate (ADP)–receptor antagonists (e.g., clopidogrel) istration of fresh-frozen plasma to maintain a pro-
Acute alcohol toxicity thrombin time and activated partial-thromboplas-
Does the patient have a hematinic deficiency (particularly, acute folate deficiency)? tin time of less than 1.5 times the normal control
Does the patient have any of the following: time and a source of fibrinogen to maintain a fi-
Sepsis (especially consider) brinogen level of more than 1.5 g per liter.28
Human immunodeficiency virus (HIV) infection The use of antifibrinolytic agents is contra-
Disseminated intravascular coagulation indicated in the management of disseminated in-
Major blood loss and hemodilution travascular coagulation, since the fibrinolytic sys-
Mechanical fragmentation tem is required in recovery to ensure the dissolution
Post-cardiopulmonary bypass of the widespread fibrin. Some guidelines recom-
Intraaortic balloon pump mend the administration of therapeutic doses of
Renal dialysis unfractionated heparin in patients with a throm-
Extracorporeal membrane oxygenation botic phenotype (e.g., gangrene),28 but such rec-
Immune-mediated disorder ommendations remain controversial because of
Immune thrombocytopenic purpura the difficulties in monitoring treatment in a pa-
Antiphospholipid syndrome tient who already has a prolonged activated partial-
Post-transfusion purpura
thromboplastin time; in addition, heparin admin-
Microangiopathic hemolytic anemia
istration may provoke hemorrhage. Currently, there
Disseminated intravascular coagulation
is insufficient clinical evidence to make a firm
recommendation on the use of heparin in patients
Thrombotic thrombocytopenic purpura
with disseminated intravascular coagulation.
Hemolytic–uremic syndrome
Hypersplenism
Other disorder
Thrombo c y t openi a
Myelodysplastic syndrome
Pathophysiological Mechanisms
Cancer
Thrombocytopenia may arise because of de-
Hereditary thrombocytopenia
creased production or increased destruction (im-
mune or nonimmune) of platelets, as well as from
and platelets produces a bleeding tendency, with sequestration in the spleen. Among patients who
thrombocytopenia, a prolonged prothrombin time are admitted to an ICU, the condition occurs in
and activated partial-thromboplastin time, hypo- about 20% of medical patients and a third of sur-
fibrinogenemia, and elevated levels of fibrin deg- gical patients. The cause of this condition is of-
radation products, such as d-dimers. The physi- ten multifactorial. Patients with thrombocytope-
ologic anticoagulants are also consumed in the nia tend to be sicker, with higher illness-severity
process of inhibiting the many activated coagu- scores, than those who are admitted with normal
lation factors.19 In fulminant disseminated intra- platelet counts.29 Table 3 lists the differential di-
vascular coagulation, the consumption and dimin- agnoses of thrombocytopenia in the critical care
ished supply of platelets and coagulation proteins setting. Given the long list, it is important to iden-
usually results in oozing at vascular access sites tify patients in whom thrombocytopenia requires
and wounds but occasionally causes profuse specific and urgent action (e.g., heparin-induced
hemorrhage. thrombocytopenia and thrombotic thrombocyto-
852 n engl j med214

The New
Society. Forrights
reserved.
penic purpura). Drug-induced thrombocytopenia Post-Transfusion Purpura

is a diagnostic challenge, because critically ill Post-transfusion purpura is a rare bleeding dis-
patients often receive multiple medications that order caused by a platelet-specific alloantibody
can cause thrombocytopenia. (usually, anti–human platelet antigen 1a [HPA-1a])
A platelet threshold of 10,000 per cubic milli- in the recipient. HPA-1a reacts with donor plate-
meter for platelet transfusion in patients who lets, destroying them and also the recipient’s
are in stable condition is both hemostatically own platelets. The majority of affected patients
efficacious and cost-effective in reducing platelet- are multiparous women who have been sen-
transfusion requirements.30 Patients with sustained sitized during pregnancy. Treatments for post-
failure of platelet production, such as those with transfusion purpura include intravenous immune
myelodysplasia or aplastic anemia, may remain (gamma) globulin, glucocorticoids, and plasma-
free of serious hemorrhage, with counts below pheresis. High-dose intravenous immune globu-
5000 to 10,000 per cubic millimeter. However, a lin (2 g per kilogram of body weight administered
higher platelet transfusion trigger should be set over either 2 or 5 days) produces an increased
in patients with other hemostatic abnormalities platelet count in about 85% of patients. Large
or increased pressure on platelet turnover or numbers of platelet transfusions may be required
platelet function. If the patient is actively bleed- to control severe bleeding before there is a re-
ing, then a platelet count of 50,000 per cubic mil- sponse to intravenous immune globulin. There is
limeter should be maintained. Among patients limited evidence that the use of HPA-1a–negative
who have or are at risk for bleeding in the central platelets is more effective than the use of plate-
nervous system or who are undergoing neurosur- lets from random donors.33
gery, a platelet count of more than 100,000 per
cubic millimeter is often recommended, although Thrombotic Microangiopathies
data are lacking to support this recommenda- Profound thrombocytopenia and microangiopathic
tion.29,30 hemolytic anemia (red-cell fragmentation) charac-
Standard platelet counts are produced by terize the thrombotic microangiopathies, which
cell counters that categorize the cells according includes three major disorders: thrombotic throm-
to size, but large platelets may be the same size bocytopenic purpura, the hemolytic–uremic syn-
as red cells and thus be categorized as such. drome, and the HELLP syndrome (characterized
Therefore, an immunologic method of platelet by pregnancy-related hemolysis, elevated liver-
counting, in which platelet antigens are labeled enzyme levels, and low platelet count). The major-
with markers that can be detected with the use ity of cases of thrombotic thrombocytopenic pur-
of flow cytometry, may be helpful in providing pura are due to a deficiency of a disintegrin and
a true count.31 Since platelet transfusions may metalloproteinase with thrombospondin type 1
lead to immune platelet refractoriness owing motif 13 (ADAMTS13), a disorder that may be
to the formation of anti-HLA antibodies, the hereditary or caused by autoimmune destruction.
use of HLA-matched platelets, if available, The absence of ADAMTS13 results in the persis-
should produce better platelet counts after trans- tence of high-molecular-weight von Willebrand
fusion. factor, which can cause spontaneous platelet ag-
gregation when the protein is subjected to high
Immunologic Causes shear stress. The rate of death in untreated cases
As a general rule, an abrupt reduction in platelet is nearly 95%, but with early plasmapheresis, the
counts with a history of recent surgery suggests survival rate is 80 to 90%. The use of rituximab,
an immunologic cause or adverse transfusion re- a chimeric monoclonal antibody against the sur-
action (post-transfusion purpura or drug-induced face B-cell protein CD20, which leads to destruc-
thrombocytopenia). Heparin-induced thrombo- tion of those cells, has been shown to reduce the
cytopenic thrombosis is an uncommon, transient, rate of recurrence of the autoimmune form of
drug-induced, autoimmune prothrombotic disor- this disorder from 57% to 10%.34 Thrombotic
der caused by the formation of IgG antibodies thrombocytopenic purpura is a medical emer-
that cause platelet activation by the formation of gency and in untreated cases is associated with a
antibodies to complexes of platelet factor 4 and rate of death of 90%, usually from myocardial
heparin.32 infarction due to platelet thrombi in the coronary

The New
Medicine
by MIKE MULARCZYK
JanuarySociety.
15, 2015.
AllFor personal
reserved.
arteries. Thus, an active diagnosis of this disor- However, in parallel with the reduction in
der or failure to rule it out should lead to urgent coagulation factors, there is a similar reduction
plasmapheresis. in the production of physiologic anticoagulants.
Thus, patients with chronic liver disease and a
L i v er Dise a se prolonged prothrombin time are no longer con-
sidered to have a deficiency of coagulation fac-
Thrombopoietin and most hemostatic proteins tors, since their coagulation is rebalanced and
are synthesized in the liver. Thus, reduced he- thrombin generation is usually normal.35 In
patic synthetic function results in prolongation such cases, there is no need to treat prolonged
of the screening tests of coagulation (particularly coagulation times in the absence of bleeding. If
the prothrombin time) and reduced platelet counts, bleeding does occur in liver disease, then con-
although levels of factor VIII and von Willebrand sensus guidelines, which are largely based on
factor are increased.35 Acute alcohol intake in- consensus expert opinion, recommend blood-
hibits platelet aggregation. In chronic liver dis- component management as determined by the
ease, there is also increased fibrinolytic potential results of testing of the platelet count, prothrom-
due to the failure of the liver to metabolize tissue bin time, activated partial-thromboplastin time,
plasminogen activator. In cholestatic liver dis- thrombin time, and fibrinogen. In a recent ran-
ease, there is reduced absorption of lipid-soluble domized, controlled trial, investigators com-
vitamins, so reduced amounts of the vitamin K– pared a liberal red-cell transfusion strategy
dependent coagulation factors II, VII, IX, and X (hemoglobin level, <9 g per deciliter) and a re-
are produced. Furthermore, in liver disease, the strictive strategy (hemoglobin level, <7 g per
failure of the normal enzymatic removal of si- deciliter) in patients with acute upper gastro-
alic acid from fibrinogen results in dysfibrino- intestinal bleeding. Patients who were treated
genemia36 (Fig. 3A). with the restrictive strategy had longer survival
A Hepatic failure
Acute alcohol
Liver failure and drug intake
Splenic sequestration Platelet dysfunction or

Splenomegaly with or without thrombocytopenia
hypersplenism
↓ Production of ↓ Production of Sialic acid not removed ↑ t-PA due to ↓ Thrombopoietin

coagulation factors to physiologic from fibrinogen impaired metabolism
about 30% (except anticoagulants to
fibrinogen and factor VIII) about 30% OH
OH
100% 100%
O OH
HO COOH
30% 30%
O
H3C N
H OH
↑Fibrinolysis
↓ α2-Antiplasmin
Hemostatic balance preserved
Impaired polymerization of fibrin
Plasminogen Plasmin
Poor coagulation reserve ↓ TAFI

if bleeding occurs
Dysfibrinogenemia Fibrin Fibrin-degradation products
Fibrinogen Fibrinogen-degradation products

and ↓fibrinogen levels
Routine testing shows ↓Functional fibrinogen
prolonged PT or APTT ↑Total fibrinogen Factor V ↓Factor V levels
854 n engl j med216

COLOR FIGURE
The New
use only. Medicine
other uses without permission.Draft 6 2/11/14
Society. Forrights
personal use only. No otherAuthor
reserved. uses without
Hunt permission.
Copyright © 2014 Massachusetts Medical Society. All rights reserved. Fig # 3a
Back to Table of Contents Title Bleeding and coagulopathies
in critical care
ME
B Renal disease
Uremic toxins
Diseased produced
kidney Erythropoiesis
Normal axial red-cell flow, Anemia, resulting in loss of axial Platelet dysfunction due to
hematocrit ≥30% red-cell flow, hematocrit <30% platelet adhesion and granule release
Subendothelial matrix von Willebrand
factor
Glycoprotein Ib Adhesion
Platelet
Fibrinogen
Aggregation
Glycoprotein
IIb/IIIa
Granule release
Figure 3. Hemostasis in Hepatic Failure and Renal Disease.

Liver failure (Panel A) leads to complex hemostatic changes, since the liver is the producer of coagulation factors, physiologic anticoagu-
lants, and thrombopoietin, as well as the site of the metabolism of sialic acid residues from fibrinogen, activated coagulation factors, and
tissue plasminogen activator. These defects result in poor coagulation reserve, dysfibrinogenemia, and increased fibrinolytic potential.
In renal failure (Panel B), decreased production of erythropoietin produces anemia, which results in a loss of axial flow so that the bleeding
time is prolonged. The accumulation of uremic toxins results in platelet dysfunction. APTT denotes activated partial-thromboplastin time,
PT prothrombin time, TAFI thrombin-activatable fibrinolysis inhibitor, and t-PA tissue plasminogen activator.
COLOR FIGURE
Draft 5 2/11/14
Author Hunt
Fig # 3b
Title Bleeding and Coagulotherapies
(6 weeks) and a lower rate of rebleeding than R ena l Dise a se ME
in Critical Care
did those who were treated with the liberal DE Drazen

strategy.37 In this study, portal circulation Uremic bleeding typically presents with ecchy-
Artist N Koscal
AUTHOR PLEASE NOTE:
pressures increased significantly among pa- moses, purpura, epistaxis, and bleeding from beenPlease
Figure has redrawn and type has been reset
check carefully
tients in the liberal-strategy group. Although puncture sites due to impaired platelet function.
there are no similar studies addressing chang- The platelet dysfunction is a result of complex
es in coag ulopathy or thrombocytopenia, it changes that include dysfunctional von Wille-
seems sensible to adopt a restrictive approach brand factor, decreased production of throm-
to the use of fresh-frozen plasma and platelets boxane, increased levels of cyclic AMP and cyclic
in patients with acute upper gastrointestinal GMP, uremic toxins, anemia, and altered plate-
bleeding. The role of tranexamic acid in pa- let granules, all of which are necessary for ade-
tients with gastrointestinal bleeding is under quate formation of a platelet plug (Fig. 3B). The
investigation in the ongoing randomized, con- anemia that commonly accompanies renal dis-
trolled Hemorrhage Alleviation with Tran- ease leads to the loss of laminar flow in arteri-
examic Acid–Intestinal System (HALT-IT) trial oles so that red cells no longer push platelets
(NCT01658124). In patients with liver disease and plasma to the endothelium, leading to pro-
and laboratory tests indicating abnormal syn- longation of the bleeding time; treatment of the
thesis of coagulation factors, vitamin K should anemia partially corrects this problem. There is
be routinely administered to aid in the synthe- also some evidence of impaired fibrinolysis in
sis of coagulation factors. patients with renal disease.

217 27, 2014 855
The New England
reserved.
In the past, the bleeding time was considered proliferative disorders,44 or the breakdown of high-
to be the most useful clinical test of coagulation molecular-weight von Willebrand factor multimers
in patients with renal disease, but much of the owing to high intravascular or extracorporeal cir-
evidence supporting testing and treatment was cuit shear stresses, may also occur in patients in
derived from poor-quality studies performed more the ICU. This disorder can also be caused by
than 30 years ago. We now know that dialysis, shear stresses on blood flow in extracorporeal
especially peritoneal dialysis, improves platelet circuits, such as those caused by extracorporeal
function. Erythropoietin, cryoprecipitate, conju- membrane oxygenation44 and left ventricular assist
gated estrogens, desmopressin, and tranexamic devices. Intravascular shear stress from aortic-
acid have all independently been shown to reduce valve stenosis can cause acquired von Willebrand’s
bleeding time.38,39 In the past decade, citrate has disease, leading to gastrointestinal bleeding
risen in popularity as a replacement anticoagulant (Heyde’s syndrome).45
in continuous renal-replacement therapy, with a Acquired von Willebrand’s disease is treated
reduction in bleeding, although data on its safety with the use of either desmopressin, which stimu-
in patients with liver failure are lacking.40 lates the release of residual stores of von Wille-
brand factor by endothelial cells, or von Willebrand
Fibr inoly t ic Bl eeding factor concentrates, with the latter considered to be
the more effective therapy.46 The use of antifibri-
Excessive fibrinolysis that threatens clot integrity nolytic agents may be considered to alleviate muco-
is known as hyperfibrinolysis.41 Abnormal fibri- cutaneous bleeding. Acquired von Willebrand’s
nolytic activity may be overlooked as a cause of disease due to high shear stresses requires the
bleeding, particularly in liver disease, and the removal of the cause of the condition whenever
condition is difficult to diagnose because of the possible.
absence of a specific routine assay. Clinical sus-
picion should be high in cases in which bleeding Bl eeding A sso ci ated w i th
continues despite hemostatic replacement therapy, A n t i thrombo t ic Ther a py
platelet levels are relatively conserved but fibrino-
gen levels are disproportionately low, and d-dimer Table 4 summarizes the current antithrombotic
levels are disproportionately high for disseminat- drugs, their mechanisms of action, and reversibil-
ed intravascular coagulation. Thromboelastogra- ity.47,48 It is difficult to treat a bleeding patient
phy, which may help differentiate fibrinolytic ac- who is receiving an oral anticoagulant such as
tivation from coagulation factor deficiency, is a dabigatran and rivaroxaban, since there is no
crude tool, since it detects only the most marked specific antidote. Studies that have evaluated the
changes.42 Fibrinolytic bleeding should be con- reversal of the new oral anticoagulants have been
sidered particularly in patients with liver disease limited to reversal of drug effect with the use of
and disseminated cancers. The use of tranexamic recombinant activated factor VII and prothrom-
acid, either by infusion or orally (depending on bin complex concentrate. Current evidence sug-
the severity of the problem and the state of the gests that prothrombin complex concentrate may
patient), is beneficial in controlling bleeding. be the best option and that it reverses the effects
of rivaroxaban better than the effects of dabiga-
Von W il l ebr a nd’s Dise a se tran.49,50 General measures such as stopping the
antithrombotic medication, documenting the time
If unexplained bleeding occurs, consideration and amount of the last drug dose, and noting the
should be given to the late presentation of an presence of renal and hepatic impairment are
inherited bleeding disorder. A personal and fam- suggested. Management may be aided by obtain-
ily history of easy bruising and bleeding should ing a full blood count and hemostatic screening,
be sought. Occasionally, a condition such as mild along with a specific laboratory test to measure
von Willebrand’s disease may present with per- the antithrombotic effect of the drug, if avail-
sistent oozing after an injury or surgery.43 able. If the medication has been recently ingested
Acquired von Willebrand’s disease, which can and there is no specific antidote, oral activated
be caused by several potential mechanisms due charcoal may be given to absorb any residual
to autoantibodies, myeloproliferative and lympho- drug in the stomach.
856 n engl j med 370;9 218

The New EnglandFor

reserved.
Table 4. Common Antithrombotic Agents, Mechanisms of Action, and Reversibility.
Procedure for
Agent Mechanism of Action Site of Clearance Half-Life Immediate Reversal
Aspirin Irreversible cyclooxygenase 20 min but effect will persist Platelet transfusion; consider use of desmopressin
inhibitor for 5 days
Clopidogrel, prasugrel, Hepatic 6 to 15 hr Platelet transfusion
P2Y12 antagonists
ticagrelor
Unfractionated heparin Indirect anti-Xa and anti-IIa Cellular and (at higher doses) 45–90 min Protamine (at a dose of 1 mg) neutralizes 80–100 U
Downloaded from nejm.org by

effect; increases the action renal unfractionated heparin

of antithrombin by factor
of 10,000
Copyright
Low-molecular-weight Same as for unfractionated Renal Approximately 4 hr, with vari- Protamine reverses 60% of effect; consider the use of
heparin heparin but mainly anti- ability among products recombinant activated factor VII if there is continued
Xa effect life-threatening bleeding and the time frame suggests
there is residual effect

MIKE MULARCZYK
n engl j med 370;9
Danaparoid A heparinoid with ratio of Renal 24 hr No specific antidote; plasmapheresis may be considered
The New
© Massachusetts
anti-Xa to anti-IIa of >20 for critical bleeding
Fondaparinux Synthetic pentasaccharide Renal 17–20 hr No specific antidote; use of recombinant activated factor
219
ForEngland
on Medical
nejm.org
with indirect anti-Xa VII should be considered for critical bleeding
effect
January 15,
personalJournal
Bivalirudin Direct antithrombin effect Proteolysis by thrombin (80%) 25 min; 1 hr in renal failure No specific antidote; hemodialysis, hemofiltration, or
Society.
use only.
with 20% renal excretion plasmapheresis may be considered for critical bleeding
2015.All
Argatroban Direct thrombin inhibitor Hepatic 45 min No specific antidote
Forrights
Vitamin K antagonists Reduction in functional levels Hepatic Varies according to drug, Intravenous vitamin K (1 to 5 mg) and prothrombin com-
of Medicine
february 27, 2014
(e.g., warfarin, phen- of vitamin K–dependent with phenprocoumon plex concentrate (25 to 50 U/kg); use of fresh frozen
personal
procoumon, acenocou- clotting factors (II, VII, the longest and aceno- plasma only if prothrombin complex concentrate is
marol, phenindione) IX, and X) coumarol the shortest not available
reserved.
Dabigatran A direct thrombin inhibitor 80% renal 13 hr (range, 11–22 hr); No specific antidote; use of oral activated charcoal if
with creatinine clearance administered within 2 hr after receipt of drug; con-

<30 ml/min, 22–35 hr sider hemofiltration, hemodialysis; if life-threatening
bleeding, consider prothrombin complex concentrate,
activated prothrombin complex concentrate, and

recombinant activated factor VII
Rivaroxaban, apixaban, Direct anti-Xa inhibition Hepatic and renal Rivaroxaban, 7–9 hr; No specific antidote; if life-threatening bleeding, same
edoxaban apixaban, 9–14 hr as for dabigatran

857
C onclusions quired bleeding disorders is very striking and

points to the need for studies to address the
The management of bleeding in critically ill pa- many evidence gaps that currently exist.
tients remains a major clinical challenge. The Dr. Hunt reports receiving consulting fees from Haemonetics
and serving as medical director of Lifeblood: the Thrombosis
cause of a bleeding problem may be complex and Charity, which receives funds from pharmaceutical companies
only partially understood, with limited diagnos- to help educate health care professionals and the public. No
tic tools and management strategies currently other potential conf lict of interest relevant to this article was
reported.
available. The absence of robust evidence from Disclosure forms provided by the author are available with
clinical trials to guide the management of ac- the full text of this article at NEJM.org.
References
1. Borgman MA, Spinella PC, Perkins evaluate the optimal source and dose of Committee on Disseminated Intravascu-
JG, et al. The ratio of blood products fibrinogen in managing bleeding. Crit lar Coagulation of the International Soci-
transfused affects mortality in patients Care 2011;15:1006. ety on Thrombosis and Haemostasis: a
receiving massive transfusions at a com- 12. Simpson E, Lin Y, Stanworth S, 5-year overview. J Thromb Haemost 2007;5:
bat support hospital. J Trauma 2007;63: Birchall J, Doree C, Hyde C. Recombinant 604-6.
805-13. factor VIIa for the prevention and treat- 22. Øvstebø R, Aass HC, Haug KB, et al.
2. Holcomb JB, del Junco DJ, Fox EE, et al. ment of bleeding in patients without hae- LPS from Neisseria meningitidis is crucial
The Prospective, Observational, Multicenter, mophilia. Cochrane Database Syst Rev for inducing monocyte- and microparticle-
Major Trauma Transfusion (PROMMTT) 2012;3:CD005011. associated tissue factor activity but not for
study: comparative effectiveness of a time- 13. Levi M, Levy JH, Andersen HF, Truloff tissue factor expression. Innate Immun
varying treatment with competing risks. D. Safety of recombinant activated factor 2012;18:580-91.
JAMA Surg 2013;148:127-36. VII in randomized clinical trials. N Engl J 23. Osterud B, Flaegstad T. Increased tis-
3. Rajasekhar A, Gowing R, Zarychanski Med 2010;363:1791-800. [Erratum, N Engl sue thromboplastin activity in monocytes
R, et al. Survival of trauma patients after J Med 2011;365:1944.] of patients with meningococcal infection:
massive red blood cell transfusion using a 14. Shakur H, Roberts I, Bautista R, et al. related to an unfavourable prognosis.
high or low red blood cell to plasma Effects of tranexamic acid on death, vas- Thromb Haemost 1983;49:5-7.
transfusion ratio. Crit Care Med 2011;39: cular occlusive events, and blood transfu- 24. Nieuwland R, Berckmans RJ, McGregor
1507-13. sion in trauma patients with significant S, et al. Cellular origin and procoagulant
4. MacLennan S, Williamson LM. Risks of haemorrhage (CRASH-2): a randomised, properties of microparticles in meningo-
fresh frozen plasma and platelets. J Trauma placebo-controlled trial. Lancet 2010;376: coccal sepsis. Blood 2000;95:930-5.
2006;60:Suppl:S46-S50. 23-32. 25. Ranieri VM, Thompson BT, Barie PS,
5. Inaba K, Branco BC, Rhee P, et al. Im- 15. CRASH-2 Collaborators. The impor- et al. Drotrecogin alfa (activated) in adults
pact of plasma transfusion in trauma pa- tance of early treatment with tranexamic with septic shock. N Engl J Med 2012;366:
tients who do not require massive trans- acid in bleeding trauma patients: an ex- 2055-64.
fusion. J Am Coll Surg 2010;210:957-65. ploratory analysis of the CRASH-2 ran- 26. Afshari A, Wetterslev J, Brok J, Møller
6. Innerhofer P, Westermann I, Tauber domised controlled trial. Lancet 2011;377: AM. Antithrombin III for critically ill pa-
H, et al. The exclusive use of coagulation 1096-101. tients. Cochrane Database Syst Rev 2008;
factor concentrates enables reversal of 16. Ker K, Edwards P, Perel P, Shakur H, 3:CD005370.
coagulopathy and decreases transfusion Roberts I. Effect of tranexamic acid on 27. Abraham E, Reinhart K, Opal S, et al.
rates in patients with major blunt trauma. surgical bleeding: systematic review and Efficacy and safety of tifacogin (recombi-
Injury 2013;44:209-16. cumulative meta-analysis. BMJ 2012;344: nant tissue factor pathway inhibitor) in
7. Ziegler B, Schimke C, Marchet P, e3054. severe sepsis: a randomized controlled
Stögermüller B, Schöchl H, Solomon C. 17. Desborough M, Stanworth S. Plasma trial. JAMA 2003;290:238-47.
Severe pediatric blunt trauma — success- transfusion for bedside, radiologically guid- 28. Levi M, Toh CH, Thachil J, Watson
ful ROTEM-guided hemostatic therapy ed, and operating room invasive procedures. HG. Guidelines for the diagnosis and man-
with fibrinogen concentrate and no ad- Transfusion 2012;52:Suppl 1:20S-29S. agement of disseminated intravascular co-
ministration of fresh frozen plasma or 18. Hall DP, Lone NI, Watson DM, Stan- agulation. Br J Haematol 2009;145:24-33.
platelets. Clin Appl Thromb Hemost 2013; worth SJ, Walsh TS. Factors associated 29. Vanderschueren S, De Weerdt A, Mal-
19:453-9. with prophylactic plasma transfusion be- brain M, et al. Thrombocytopenia and
8. Hiippala S. Replacement of massive fore vascular catheterization in non- prognosis in intensive care. Crit Care Med
blood loss. Vox Sang 1998;74:Suppl 2:399- bleeding critically ill adults with pro- 2000;28:1871-6.
407. longed prothrombin time: a case-control 30. Slichter SJ. Evidence-based platelet
9. Spahn DR, Cerny V, Coats TJ, et al. study. Br J Anaesth 2012;109:919-27. transfusion guidelines. Hematology Am
Management of bleeding following major 19. Collins PW, Macchiavello LI, Lewis SJ, Soc Hematol Educ Program 2007:172-8.
trauma: a European guideline. Crit Care et al. Global tests of haemostasis in criti- 31. Segal HC, Harrison P. Methods for
2007;11:R17. [Erratum, Crit Care 2007;11: cally ill patients with severe sepsis syn- counting platelets in severe thrombocyto-
414.] drome compared to controls. Br J Haematol penia. Curr Hematol Rep 2006;5:70-5.
10. Rourke C, Curry N, Khan S, et al. Fi- 2006;135:220-7. 32. Kelton JG, Arnold DM, Bates SM.
brinogen levels during trauma hemor- 20. Holmes MV, Hunt BJ, Shearer MJ. The Nonheparin anticoagulants for heparin-
rhage, response to replacement therapy, role of dietary vitamin K in the manage- induced thrombocytopenia. N Engl J Med
and association with patient outcomes. ment of oral vitamin K antagonists. Blood 2013;368:737-44.
J Thromb Haemost 2012;10:1342-51. Rev 2012;26:1-14. 33. Allen DL, Samol J, Benjamin S, Verjee
11. Stanworth SJ, Hunt BJ. The desperate 21. Toh CH, Hoots WK. The scoring sys- S, Tusold A, Murphy MF. Survey of the use
need for good-quality clinical trials to tem of the Scientific and Standardisation and clinical effectiveness of HPA-1a/5b-
858 n engl j med 370;9 220

The New England

personal Medicine
reserved.
negative platelet concentrates in proven Tranexamic acid inhibits fibrinolysis, short- a few answers, still many questions. Br J
or suspected platelet alloimmunization. ens the bleeding time and improves plate- Haematol 2013;161:177-82.
Transfus Med 2004;14:409-17. let function in patients with chronic renal 46. Tiede A, Rand JH, Budde U, Ganser A,
34. Scully M, Hunt BJ, Benjamin S, et al. failure. Thromb Haemost 1999;82:1250-4. Federici AB. How I treat the acquired
Guidelines on the diagnosis and manage- 40. Zhang Z, Hongying N. Efficacy and von Willebrand syndrome. Blood 2011;
ment of thrombotic thrombocytopenic safety of regional citrate anticoagulation 117:6777-85.
purpura and other thrombotic microangi- in critically ill patients undergoing con- 47. Kaatz S, Kouides PA, Garcia DA, et al.
opathies. Br J Haematol 2012;158:323-35. tinuous renal replacement therapy. Inten- Guidance on the emergent reversal of oral
35. Tripodi A, Mannucci PM. The coagu- sive Care Med 2012;38:20-8. thrombin and factor Xa inhibitors. Am J
lopathy of chronic liver disease. N Engl J 41. Hunt BJ, Segal H. Hyperfibrinolysis. Hematol 2012;87:Suppl 1:S141-S145. [Erra-
Med 2011;365:147-56. J Clin Pathol 1996;49:958. [Erratum, J Clin tum, Am J Hematol 2012;87:748.]
36. Roberts LN, Patel RK, Arya R. Hae- Pathol 1997;50:357.] 48. Makris M, Van Veen JJ, Tait CR, Mum-
mostasis and thrombosis in liver disease. 42. Raza I, Davenport R, Rourke C, et al. ford AD, Laffan M. Guideline on the man-
Br J Haematol 2010;148:507-21. The incidence and magnitude of fibrino- agement of bleeding in patients on anti-
37. Villanueva C, Colomo A, Bosch A, et al. lytic activation in trauma patients. J Thromb thrombotic agents. Br J Haematol 2013;
Transfusion strategies for acute upper gas- Haemost 2013;11:307-14. 160:35-46.
trointestinal bleeding. N Engl J Med 2013; 43. Cuker A, Connors JM, Katz JT, Levy 49. Lazo-Langner A, Lang ES, Douketis J.
368:11-21. [Erratum, N Engl J Med 2013; BD, Loscalzo J. A bloody mystery. N Engl Clinical management of new oral anti-
368:2341.] J Med 2009;361:1887-94. coagulants: a structured review with em-
38. Hedges SJ, Dehoney SB, Hooper JS, 44. Heilmann C, Geisen U, Beyersdorf F, et phasis on the reversal of bleeding compli-
Amanzadeh J, Busti AJ. Evidence-based al. Acquired von Willebrand syndrome in cations. Crit Care 2013;17:230.
treatment recommendations for uremic patients with extracorporeal life support 50. Harper P, Young L, Merriman E. Bleed-
bleeding. Nat Clin Pract Nephrol 2007;3: (ECLS). Intensive Care Med 2012;38:62-8. ing risk with dabigatran in the frail elderly.
138-53. 45. Franchini M, Mannucci PM. Von Wil- N Engl J Med 2012;366:864-6.
39. Mezzano D, Panes O, Muñoz B, et al. lebrand disease-associated angiodysplasia: Copyright © 2014 Massachusetts Medical Society.


The New
© Massachusetts Medical15, 2015. All
Society. Forrights
reserved.
c or r e sp ondence
Bleeding and Coagulopathies in Critical Care

To the Editor: Hunt (Feb. 27 issue)1 discusses factor for adverse outcomes. Several studies have
the management of several coagulopathies in shown that the development of a progressive
critical care patients; however, there is no mention thrombocytopenia during a medical or surgical
of the effect of human immunodeficiency virus ICU stay is associated with increased morbidity
(HIV) infection on thrombotic microangiopa- and mortality; this enhanced risk is greater
thies, especially thrombotic thrombocytopenic among patients with worsening thrombocytope-
purpura (TTP). It is estimated that HIV-infected nia beyond the first week of an ICU stay.1,2 Con-
persons have an incidence of TTP 40 times that versely, a decreased risk of death was observed
of the general population.2 In some cohorts, among patients whose platelet count recovered
more than 80% of TTP cases are related to HIV during the ICU admission.3 Because platelets
infection.3 Moreover, HIV infection is postulated have multiple functions in innate and adaptive
as a direct precipitating factor in TTP, through immunity,4 falling platelet counts in ICU patients
several mechanisms.4-6 Could Hunt comment on may be associated with multiorgan failure.
the role of routine HIV testing in patients with Michele Bibas, M.D.
TTP? Is special treatment indicated? National Institute for Infectious Diseases Lazzaro Spallanzani
José Moreira, M.D. Rome, Italy
michele.bibas@inmi.it
Instituto de Pesquisa Clínica Evandro Chagas
Rio de Janeiro, Brazil
ported.
jose.moreira@ipec.fiocruz.br
No potential conflict of interest relevant to this letter was re- 1. Parker RI. Etiology and significance of thrombocytopenia in
ported. critically ill patients. Crit Care Clin 2012;28:399-411.
2. Hui P, Cook DJ, Lim W, Fraser GA, Arnold DM. The fre-
1. Hunt BJ. Bleeding and coagulopathies in critical care. N Engl quency and clinical significance of thrombocytopenia compli-
J Med 2014;370:847-59. cating critical illness: a systematic review. Chest 2011;139:271-8.
2. Hymes KB, Karpatkin S. Human immunodeficiency virus 3. Greinacher A, Selleng K. Thrombocytopenia in the intensive
infection and thrombotic microangiopathy. Semin Hematol 1997; care unit patient. Hematology Am Soc Hematol Educ Program
34:117-25. 2010;2010:135-43.
3. Visagie GJ, Louw VJ. Myocardial injury in HIV-associated 4. Semple JW, Italiano JE Jr, Freedman J. Platelets and the im-
thrombotic thrombocytopenic purpura (TTP). Transfus Med mune continuum. Nat Rev Immunol 2011;11:264-74.
2010;20:258-64.
4. Meiring M, Webb M, Goedhals D, Louw V. HIV-associated DOI: 10.1056/NEJMc1403768
thrombotic thrombocytopenic purpura — what we know so far.
Eur Oncol Haematol 2012;8:89-91.
To the Editor: We think the article by Hunt
5. Lederman MM, Funderburg NT, Sekaly RP, Klatt NR, Hunt
PW. Residual immune dysregulation syndrome in treated HIV
infection. Adv Immunol 2013;119:51-83.
should have addressed two issues. First, a remark-
able number of patients in intensive care medi-
6. Miller RF, Scully M, Cohen H, et al. Thrombotic thrombocy-
cine are treated with oral antiplatelet medication
topaenic purpura in HIV-infected patients. Int J STD AIDS 2005;
16:538-42.
or with extracorporeal devices (i.e., renal dialysis
DOI: 10.1056/NEJMc1403768 and extracorporeal lung or cardiac support) as
standard therapies for their underlying condi-
To the Editor: Hunt did not address the signifi- tions.1,2 These treatments subsequently result in
cance of changing platelet counts during an ad- thrombocytopathy with impaired platelet func-
mission to an intensive care unit (ICU) as a risk tion, and this warrants vigilance to adapt the
2148 n engl j med 370;22222

correspondence
therapeutic approach. Second, the importance of come. Meier et al. point out that the combination
modern monitoring devices (point-of-care diag- of platelet dysfunction and thrombocytopenia in
nostics) and their specific significance in all critically ill patients requires careful hemostatic
bleeding conditions should be discussed, because monitoring. But how to monitor coagulopathies
these diagnostic tools substantially improve the in critical care is an unresolved issue. Conven-
management of bleeding conditions, save resourc- tional testing with the prothrombin time and
es for the treatment of bleeding disorders, and activated partial-thromboplastin time — which
potentially improve the outcome of patients in were developed to detect inherited coagulation
critical care medicine.2,3 disorders, not to manage acquired coagulopa-
Jens Meier, M.D., Ph.D. thies — have never been validated in the latter.
Janek Henes, M.D. Point-of-care testing has been available for
Peter Rosenberger, M.D., Ph.D. many years and is increasingly sophisticated.
University Hospital Tübingen Assays of clot viscoelasticity (thromboelastogra-
Tübingen, Germany phy and rotational thromboelastometry) give a
peter.rosenberger@med.uni-tuebingen.de more global picture of hemostasis than conven-
No potential conflict of interest relevant to this letter was re- tional coagulation screening. Different types of
ported.
point-of-care testing with different protocols are
1. Brodie D, Bacchetta M. Extracorporeal membrane oxygena- being used by an increasing number of units. As
tion for ARDS in adults. N Engl J Med 2011;365:1905-14.
2. Stone GW, Witzenbichler B, Weisz G, et al. Platelet reactivity Meier et al. note, point-of-care testing has prog-
and clinical outcomes after coronary artery implantation of nostic value in sepsis.3 Despite these prognosti-
drug-eluting stents (ADAPT-DES): a prospective multicentre reg- cating abilities, the fundamental issue remains:
istry study. Lancet 2013;382:614-23.
3. Adamzik M, Görlinger K, Peters J, Hartmann M. Whole there are no internationally validated algorithms
blood impedance aggregometry as a biomarker for the diagnosis to guide the use of pharmacologic agents and
and prognosis of severe sepsis. Crit Care 2012;16:R204. blood-component therapy in treatment. Accord-
DOI: 10.1056/NEJMc1403768 ing to systematic reviews, there is inadequate
evidence that the use of point-of-care testing
The Author Replies: My review had word-count reduces morbidity or mortality among patients
and reference limitations, so only selected areas with severe bleeding 4 or has an added value in
could be highlighted. I therefore welcome discus- improving the management of septic coagulop-
sion of other issues. athy.5 A robust point-of-care algorithm that im-
Moreira asks about best-practice management proves outcome in coagulopathies in critical care
of HIV-related TTP. TTP may be the initial pre- remains much needed.
senting feature of HIV or may occur in those with Beverley J. Hunt, M.D.
low CD4 cell counts after nonadherence to anti- Guy’s and St. Thomas’ NHS Foundation Trust
viral treatment; thus, current guidelines recom- London, United Kingdom
mend testing for HIV in all new cases of TTP.1 beverley.hunt@gstt.nhs.uk
Management requires treatment for both TTP Since publication of her article, the author reports no further
potential conflict of interest.
and HIV infection, because patients with both
conditions may have an underlying ADAMTS13 1. Scully M, Hunt BJ, Benjamin S, et al. Guidelines on the diag-
nosis and management of thrombotic thrombocytopenic purpura
deficiency due to an acquired antibody, therefore and other thrombotic microangiopathies. Br J Haematol 2012;
needing urgent plasma exchange, whereas TTP 158:323-35.
remission is dependent on improving the immune 2. Ross CL, Hunt BJ, Wyncoll D, Hodgkiss A, Peters B. HIV,
thrombotic thrombocytopaenic purpura and rituximab in a vio-
status of the HIV-infected patient and requires lent noncompliant patient. Blood Coagul Fibrinolysis 2009;20:
highly active antiretroviral therapy. The care of 157-60.
such patients can be particularly challenging but 3. Adamzik M, Görlinger K, Peters J, Hartmann M. Whole blood
impedance aggregometry as a biomarker for the diagnosis and
rewarding.2 Other patients who have HIV infec- prognosis of severe sepsis. Crit Care 2012;16:R204.
tion and TTP without ADAMTS13 deficiency have 4. Thrombelastography (TEG) or thromboelastometry (ROTEM)
an increased mortality.1 to monitor haemotherapy versus usual care in patients with mas-
sive transfusion. Cochrane Database Syst Rev 2011;3:CD007871.
Bibas emphasizes that progressive thrombo- 5. Müller MC, Meijers JC, Vroom MB, Juffermans NP. Utility of
cytopenia during a critical care stay is a poor thromboelastography and/or thromboelastometry in adults with
prognostic factor, but there is no evidence that sepsis: a systematic review. Crit Care 2014;18:R30.
treating the thrombocytopenia will improve out- DOI: 10.1056/NEJMc1403768
n engl j med 370;22 223

nejm.org may 29, 2014 2149

How would you assess and manage his bleeding?
A. Cease heparin and transfuse packed red cells, platelets, cryoprecipitate, and fresh-frozen plasma or prothrombin
complex concentrate.
B. Cease heparin and transfuse cryoprecipitate and fresh-frozen plasma or prothrombin complex concentrate.
C. Cease heparin, transfuse platelets, administer an antifibrinolytic agent (e.g., epsilon-aminocaproic acid or
tranexamic acid).
D. Cease heparin and repeat laboratory tests in 8 to 12 hours.

The laboratory test results indicate the patient has thrombocytopenia with mildly abnormal results on coagulation
testing. In addition, there is slight bleeding from the surgical wound and skin-puncture sites. However, the patient
does not have worsening shock, which would be suggestive of substantial intraabdominal bleeding.
Coagulation abnormalities are very common in critically ill patients, and more than one cause may be present.
Our patient has septic shock, which is the most common cause of disseminated intravascular coagulation (DIC).
The laboratory results are consistent with early or low-grade DIC with a reduced platelet count and prolonged
prothrombin time. Additional information should be sought from a blood smear to confirm thrombocytopenia,
as well as checking levels of d-dimer and fibrin degradation products.
DIC is a clinicopathological diagnosis with the prerequisite that the patient must have an underlying disorder
known to be associated with DIC. The scoring system of the International Society on Thrombosis and Hemostasis
can be used to confirm that the laboratory diagnostic criteria have been met.1 The management of DIC should be
focused primarily on treatment of the underlying cause. General guidelines for the management of coagulopathy
and bleeding are based mainly on expert opinion and suggest replacement of coagulation proteins and platelets in
patients who are bleeding. Guidelines suggest platelet transfusion to maintain a platelet level of more than 50,000
per cubic millimeter and fresh-frozen plasma to maintain a prothrombin time and APTT of less than 1.5 times
the normal control time, as well as a source of fibrinogen to maintain a fibrinogen level of more than 1.5 g per liter.
Antifibrinolytic agents are contraindicated in the management of DIC, since the fibrinolytic system is needed
to ensure dissolution of widespread fibrin during the recovery phase. Other possible contributing causes should
be considered. The prolonged prothrombin time could be due to liver dysfunction associated with his previous
alcohol intake, and the low platelet count could be due to hypersplenism associated with portal hypertension, but
the normal results on other liver testing and the lack of history of cirrhosis are against this possibility.
The patient has received unfractionated heparin, and if he has previously been exposed to heparin, a diagnosis
of heparin-induced thrombocytopenia (HIT) is a possibility. If he has not previously been exposed to heparin,
then HIT is very unlikely after only 1 day of exposure. Screening tests for HIT, such as detection of platelet factor
4 antibodies on enzyme-linked immunosorbent assay, are now widely available, but false positives may occur in
up to 80% of patients, and a confirmatory test (e.g., serotonin release assay) should be performed if the screening
test is positive.2
Since it is likely that our patient has DIC associated with only minor bleeding, it is reasonable to simply withhold
heparin and repeat the laboratory tests in 8 to 12 hours, or sooner if there is evidence of increased bleeding.
224
REFERENCES
1. Toh CH, Hoots WK. The scoring system of the Scientific and Standardisation Committee on Disseminated Intra
vascular Coagulation of the International Society on Thrombosis and Haemostasis: a 5-year overview. J Thromb
Haemost 2007;5:604-6.
2. Berry C, Tcherniantchouk O, Ley EJ, et al. Overdiagnosis of heparin-induced thrombocytopenia in surgical ICU
patients. J Am Coll Surg 2011;213:10-7.
225
original article
Age of Transfused Blood in Critically

Ill Adults
Jacques Lacroix, M.D., Paul C. Hébert, M.D., Dean A. Fergusson, Ph.D.,
Alan Tinmouth, M.D., Deborah J. Cook, M.D., John C. Marshall, M.D.,
Lucy Clayton, M.Sc., Lauralyn McIntyre, M.D., Jeannie Callum, M.D.,
Alexis F. Turgeon, M.D., Morris A. Blajchman, M.D., Timothy S. Walsh, M.D.,
Simon J. Stanworth, F.R.C.P., Helen Campbell, D.Phil., Gilles Capellier, M.D.,
Pierre Tiberghien, M.D., Laurent Bardiaux, M.D., Leo van de Watering, M.D.,
Nardo J. van der Meer, M.D., Elham Sabri, M.Sc., and Dong Vo, B.Eng.,
for the ABLE Investigators and the Canadian Critical Care Trials Group*
From Centre Hospitalier Universitaire

(CHU) Sainte-Justine, Université de A BS T R AC T
Montréal (J.L., L.C.) and Centre de Re-
cherche du Centre Hospitalier de
l’Université de Montréal, Montreal Background
(P.C.H.), Ottawa Hospital Research In-
stitute, University of Ottawa, Ottawa Fresh red cells may improve outcomes in critically ill patients by enhancing oxy-
(D.A.F., A.T., L.M., E.S., D.V.), McMas- gen delivery while minimizing the risks of toxic effects from cellular changes and
ter University, Hamilton, ON (D.J.C.,
M.A.B.), University of Toronto, Toron-
the accumulation of bioactive materials in blood components during prolonged
to (J.C.M., J.C.), and Centre de Recher- storage.
che du CHU de Québec, Université
Laval, Quebec, QC (A.F.T.) — all in Methods
Canada; University of Edinburgh
(T.S.W.) and NHS Blood and Trans- In this multicenter, randomized, blinded trial, we assigned critically ill adults to
plant–Oxford University Hospitals receive either red cells that had been stored for less than 8 days or standard-issue
NHS Trust, University of Oxford, Ox-
ford (S.J.S., H.C.) — both in the United
red cells (the oldest compatible units available in the blood bank). The primary
Kingdom; Université de Franche- outcome measure was 90-day mortality.
Comté, Besançon (G.C., P.T.) and
Établissement Français du Sang, La Results
Plaine St. Denis (P.T., L.B.) — both in
France; and Sanquin Blood Supply, Between March 2009 and May 2014, at 64 centers in Canada and Europe, 1211 pa-
Amsterdam (L.W.), Amphia Hospital, tients were assigned to receive fresh red cells (fresh-blood group) and 1219 patients
Breda and Oosterhout (N.J.M.), and
TIAS School for Business and Society–
were assigned to receive standard-issue red cells (standard-blood group). Red cells
Tilburg University, Tilburg (N.J.M.) — were stored a mean (±SD) of 6.1±4.9 days in the fresh-blood group as compared
all in the Netherlands. Address reprint with 22.0±8.4 days in the standard-blood group (P<0.001). At 90 days, 448 patients
requests to Dr. Hébert at the Centre de
Recherche du Centre Hospitalier de
(37.0%) in the fresh-blood group and 430 patients (35.3%) in the standard-blood
l’Université de Montréal, 900 Saint Denis group had died (absolute risk difference, 1.7 percentage points; 95% confidence
St., Montreal, QC H2X 0A9, Canada, or at interval [CI], –2.1 to 5.5). In the survival analysis, the hazard ratio for death in the
paul.hebert.chum@ssss.gouv.qc.ca.
fresh-blood group, as compared with the standard-blood group, was 1.1 (95% CI,
Drs. Lacroix, Hébert, Fergusson, and Tin- 0.9 to 1.2; P = 0.38). There were no significant between-group differences in any of
mouth contributed equally to this article.
the secondary outcomes (major illnesses; duration of respiratory, hemodynamic, or
*A complete list of investigators in the renal support; length of stay in the hospital; and transfusion reactions) or in the
Age of Blood Evaluation (ABLE) study is
subgroup analyses.
provided in the Supplementary Appen-
dix, available at NEJM.org.
Conclusions
2015, at NEJM.org. Transfusion of fresh red cells, as compared with standard-issue red cells, did not
decrease the 90-day mortality among critically ill adults. (Funded by the Cana-
N Engl J Med 2015;372:1410-8.
DOI: 10.1056/NEJMoa1500704 dian Institutes of Health Research and others; Current Controlled Trials number,
Copyright © 2015 Massachusetts Medical Society. ISRCTN44878718.)
1410 226372;15
Age of Tr ansfused Blood in Critically Ill Adults
B
lood transfusions are adminis- with stratification according to study site. Critically
tered frequently and may have unintended ill patients were assigned in a 1:1 ratio to one of the
consequences in critically ill patients.1-4
two study groups, with the use of permuted blocks
Current regulations permit the storage of red of varying sizes of 6, 8, or 10. Randomization was
cells for up to 42 days, but prolonged storage hasinitiated by blood-transfusion personnel when the
been associated with changes that may render first red-cell transfusion was requested. Only the
red cells ineffective as oxygen carriers and that study statistician at the coordinating center had
lead to the accumulation of substances that have knowledge of the randomization codes. An opaque
untoward biologic effects.5-8 sticker was affixed over the expiration and collec-
A systematic review of 18 observational stud- tion dates on the blood units, or the labels were
ies involving a total of 409,840 patients and changed, so that the medical team would be un-
three randomized, controlled trials involving a aware of the treatment-group assignments. Blood-
total of 126 patients suggested that the transfu- transfusion technologists refrained from releasing
sion of older red cells, as compared with newer information on storage duration to all clinical and
red cells, was associated with a 16% increase in research personnel.
the risk of death.9 However, a recent randomized The conduct of the trial and the safety of par-
trial did not document adverse consequences on ticipants were overseen by the data and safety
oxygenation, immunologic, or coagulation vari- monitoring committee, whose members reviewed
ables in 50 patients undergoing mechanical interim analyses after each consecutive group of
ventilation who received red-cell units that had 500 patients had been followed for 90 days. We
been stored for a median of 4.0 days, as compared adopted O’Brien–Fleming group-sequential stop-
with 50 patients who received blood that had ping rules for the four interim analyses. All data
been stored for 26.5 days.10 In the Age of Blood management and statistical analyses were per-
Evaluation (ABLE) pilot trial involving 66 pa- formed by the Methods Centre at the Ottawa Hos-
tients, 27% of the patients who received fresh pital Research Institute in Ottawa. Clinical coordi-
blood, as compared with 13% assigned to stan- nation was conducted by the Research Center of
dard-issue blood, died or had a life-threatening Sainte-Justine Hospital in Montreal. The trial was
complication (P = 0.31).11 conducted with the support of the Canadian Criti-
Blood-transfusion services typically provide cal Care Trials Group.
the oldest compatible red cells as an inventory- The protocol, including details of trial conduct
management approach (“first in, first out”) to and the statistical analysis plan, has been pub-
minimize waste of blood components. This prac- lished previously12 and is available with the full
tice is amplified at large centers, because suppliers
text of this article at NEJM.org. The study protocol
send fresh red cells to small centers or remote was approved by the local or regional research eth-
locations where usage is low. Frequently, unused ics board for each participating institution. The
older red cells are returned from these centers study was designed by the authors, who vouch for
for redistribution to larger centers, thus expos- the accuracy and completeness of the data and for
ing some of the sickest patients in any system to the fidelity of this report to the study protocol. No
the oldest available blood. We hypothesized that one who is not an author contributed to the writing
among critically ill adults, fresh red cells stored
of the paper. There was no support from a com-
for less than 8 days would be superior to stan- mercial entity for this study. At sites where deferred
dard-issue red cells, resulting in lower mortalityconsent was permitted, written informed consent
at 90 days. was obtained from the patient or surrogate deci-
sion maker as soon as possible after enrollment. If
Me thods consent was declined, the study intervention was
stopped, but we explicitly requested to keep all
Study Design and Oversight information on patients who had been enrolled in
The ABLE trial was a multicenter, randomized, the trial and ascertain 90-day follow-up data.
blinded trial comparing red cells stored for less
than 8 days with standard-issue red cells. Ran- Study Population
domization was performed with the use of a cen- We enrolled critically ill adults from tertiary care
tralized computer-generated assignment sequence, intensive care units (ICUs) at 64 centers (26 in
n engl j med 372;15 227

Canada, 20 in the United Kingdom, 10 in France, and occasionally by contact with primary care phy-
7 in the Netherlands, and 1 in Belgium) (see the sicians or by review of vital-statistics registries.
Supplementary Appendix, available at NEJM.org). We also collected information on several sec-
We selected study sites at which all red-cell units ondary outcomes, including organ dysfunction.13
were leukoreduced before storage and suspended We recorded infections, including nosocomial
in saline–adenine–glucose–mannitol (SAGM) ad- pneumonia, deep-tissue infections (peritonitis and
ditive solutions. mediastinitis), and bacteremia, which we catego-
We screened patients 18 years of age or older rized according to Centers for Disease Control
who were admitted to participating ICUs. Pa- criteria.14 We examined the length of stay in the
tients were eligible if a first red-cell transfusion ICU and in the hospital as well as the duration
was prescribed within 7 days after admission to of respiratory, hemodynamic, or renal support.
the ICU and if they were expected to require Adverse events and transfusion reactions were
invasive or noninvasive mechanical ventilation recorded daily. Patient data were abstracted from
for at least 48 hours. Reasons for exclusion of hospital charts. Information on the duration of
patients are listed in Figure S1 in the Supple- red-cell storage was provided by blood-bank tech-
mentary Appendix. nologists.
Intervention Statistical Analysis

We compared the use of fresh red cells that had We estimated that a total sample of 2266 patients
been stored for less than 8 days with the usual would be needed for the study to have 90% power
transfusion practice, whereby blood-transfusion to detect an absolute difference in risk of 5 per-
services issue the oldest available compatible centage points from a baseline mortality of at
blood. Participating blood-transfusion services least 25%, at a type I error rate of 5% (two-sided
in collaboration with blood centers agreed to alpha). With an anticipated rate of loss to follow-
maintain an inventory of fresh red cells for pa- up at 90 days of 5%, we calculated that 2510 pa-
tients assigned to the fresh-blood group. We an- tients would have to undergo randomization.15
ticipated that compatible red cells stored for less All the statistical analyses were based on the
than 8 days might not always be available to pa- intention-to-treat principle. We also performed
tients assigned to receive fresh red cells. In such two prespecified analyses of 90-day mortality:
instances, the protocol specified that patients as- first, we performed a per-protocol analysis that
signed to the fresh-blood group receive the fresh- was limited to patients who received at least one
est compatible red cells (i.e., not the oldest com- red-cell transfusion after randomization; second,
patible, as issued in the standard-blood group). as a sensitivity analysis, we restricted our analysis
For this clinical trial, we defined adherence to to patients in the fresh-blood group who received
the transfusion protocol for the fresh-blood only units stored for less than 8 days versus pa-
group as the transfusion of red-cell units that tients in the standard-blood group who received
had been stored for less than 8 days and for the only units stored for more than 7 days. We calcu-
standard-blood group as the transfusion of the lated the absolute risk difference (risk in the
oldest compatible red cells. We administered the fresh-blood group minus risk in the standard-
study interventions until hospital discharge, blood group) with 95% confidence intervals for
death, or up to 90 days after randomization, all mortality analyses, including 90-day mortal-
whichever occurred first. There were no other ity. A positive value suggested increased mortal-
trial interventions related to red cells or other ity or number of events in the fresh-blood group,
blood products. We did not mandate or monitor whereas a negative value suggested a greater num-
any bedside transfusion guidelines. All decisions ber of events in the standard-blood group. We
regarding patient care were at the discretion of compared continuous measures, including length
the attending physicians and the clinical team. of stay in the hospital or ICU and duration of
respiratory, hemodynamic, or renal support, using
Outcomes the Wilcoxon rank-sum test.
The primary outcome measure was 90-day all-cause We used multivariable logistic-regression mod-
mortality. We ascertained survival status by direct els to calculate absolute risk differences while
contact with clinical teams, patients, or families adjusting for possible independent confounding
1412 n engl j med 372;15 228

variables, including center, patient age, sex, co- Prespecified subgroup analyses of 90-day all-
existing illnesses, illness-severity score on the cause mortality were performed according to age
Acute Physiology and Chronic Health Evaluation (<40 years, 40 to <50 years, 50 to <60 years, or
II (APACHE II; range, 0 to 71, with higher scores ≥60 years), APACHE II score (<20 vs. ≥20), num-
indicating a greater risk of death),16 and the Mul- ber of red-cell units transfused (1 to 3 vs. >3), and
tiple Organ Dysfunction Score (range, 0 to 24, admission category (medical, surgical, or trau-
with higher scores indicating more severe organ ma). The same multivariable logistic-regression
dysfunction).17 We also compared Kaplan–Meier models were used in each subgroup stratum.
survival curves, with time to death censored at Dichotomous data are presented as numbers
90 days as the outcome. We compared treatment and percentages, whereas continuous data are
effects using a log-rank test, followed by Cox expressed as means and standard deviations or
proportional-hazards modeling with the same medians and interquartile ranges, as appropriate.
explanatory variables used in the logistic-regres- We report 95% confidence intervals. P values have
sion models. This approach was used for the not been adjusted for multiple comparisons. Data
secondary outcomes of all-cause mortality at 28 were analyzed with the use of SAS software, ver-
days, as well as mortality in the hospital or ICU. sion 9.1 (SAS Institute).
Table 1. Baseline Characteristics.*
Fresh Blood Standard Blood Total

Characteristic (N = 1206) (N = 1206) (N = 2412)
Age — yr 61.3±16.7 61±16.7 61.2±16.7
Male sex — no. (%) 682 (56.6) 643 (53.3) 1325 (54.9)
Coexisting illness — no. (%) 512 (42.5) 514 (42.6) 1026 (42.5)
APACHE II score† 21.9±7.7 21.6±7.6 21.8±7.6
Length of stay in ICU — days 2.4±2.0 2.4±2.1 2.4±2.1
Time from hospitalization to ICU admission — days 2.6±6.6 2.7±6.5 2.6±6.5
Organ injury and support
MODS‡ 5.0±3.1 4.7±3.1 4.9±3.1
Invasive mechanical ventilation — no. (%) 1176 (97.5) 1174 (97.3) 2350 (97.4)
Renal-replacement therapy — no. (%)§ 324 (26.9) 354 (29.4) 678 (28.1)
Vasoactive support — no. (%)¶ 750 (62.2) 765 (63.4) 1515 (62.8)
Type of admission — no. (%)
Emergency 1164 (96.5) 1169 (96.9) 2333 (96.7)
Elective 42 (3.5) 37 (3.1) 79 (3.3)
Major admission category — no. (%)‖
Medical 845 (70.1) 867 (71.9) 1712 (71.0)
Surgical 175 (14.5) 152 (12.6) 327 (13.6)
Trauma
Brain injury 110 (9.1) 106 (8.8) 216 (9.0)
No brain injury 76 (6.3) 80 (6.6) 156 (6.5)
* Plus–minus values are means ±SD. ICU denotes intensive care unit.
† Scores on the Acute Physiology and Chronic Health Evaluation II (APACHE II) range from 0 to 71, with higher scores
indicating a higher risk of death.16
‡ The Multiple Organ Dysfunction Score (MODS) ranges from 0 to 24, with higher scores indicating more severe organ
dysfunction.17
§ Renal-replacement therapy included continuous renal-replacement therapy, peritoneal dialysis, or hemodialysis.
¶ Vasoactive support was defined as continuous vasoactive drug infusion for hemodynamic support (excluding dopa-
mine infusion at a dose of ≤5 µg per kilogram of body weight per minute).
‖ Major admission category was missing for one patient in the standard-blood group.

229 april 9, 2015 1413
R e sult s duration of storage was 6.1±4.9 days in the fresh-

blood group versus 22.0±8.4 days in the stan-
Patients dard-blood group (P<0.001). The rate of adher-
From March 2009 through May 2014, a total of ence to the transfusion protocol was 95.4% for
19,196 patients were eligible for inclusion. Of these, all red cells transfused, with 100% of patients in
16,605 patients (86.5%) met at least one exclusion the standard-blood group receiving only stan-
criterion; the patient or a surrogate decision maker dard-issue red cells and 84.0% of patients in the
declined consent in 81 instances. Therefore, 2510 fresh-blood group receiving only red cells stored
patients underwent randomization; 80 (3.2%) were for less than 8 days (Table 2, and Fig. S2 in the
withdrawn after randomization because we were Supplementary Appendix). In the fresh-blood
unable to obtain primary outcome data, leaving group, all the patients received the freshest red
2430 patients (1211 in the fresh-blood group and cells available (i.e., there were no protocol viola-
1219 in the standard-blood group) in the intention- tions). Only 6.6% of the patients in the fresh-
to-treat analysis (Fig. S1 in the Supplementary Ap- blood group received more than 1 red-cell unit
pendix). Baseline data were available for 1412 of the that had been stored for more than 7 days, and
1430 patients with primary outcome data. Of these only 4.6% received more than 2 units that had
2430 patients, 94 (3.9%) did not receive any red-cell been stored for more than 7 days. Most patients
transfusions. The two study groups had similar in the fresh-blood group (238 of 249 patients
characteristics at baseline (Table 1). The overall rate [95.6%]) who received only 1 red-cell transfusion
of loss to follow-up was 3.2% at 90 days. received exclusively red-cell units that had been
stored for less than 8 days (Table S2 in the Sup-
Intervention plementary Appendix). Cointerventions were
A total of 5198 red-cell units were given to patients similar in the two groups before and after ran-
in the fresh-blood group and 5210 to patients in domization (Table 1, and Table S3 in the Supple-
the standard-blood group (Table 2). The average mentary Appendix).
Table 2. Anemia and Red-Cell Transfusions.*
Variable Fresh Blood Standard Blood P Value

Hemoglobin level in ICU
No. of patients evaluated 1207 1206
Level before first transfusion — g/dl 7.69±1.28 7.64±1.09 0.27
Lowest level after randomization — g/dl 7.34±1.46 7.31±1.41 0.61
Red-cell transfusions after randomization
Patients who received at least one transfusion — no./total no. (%)† 1163/1207 (96.4) 1173/1208 (97.1) 0.30
Time from randomization to first transfusion — hr 10.3±16.2 9.7±16.2 0.43
No. of red-cell units per patient who received at least one transfusion 4.3±5.2 4.3±5.5 0.98
Duration of storage of all red-cell units — days 6.1±4.9 22.0±8.4 <0.001
Adherence to transfusion protocol — no./total no. (%)
Patients‡ 977/1163 (84.0) 1206/1206 (100) <0.001
Red-cell units§ 4723/5198 (90.9) 5210/5210 (100) <0.001

† Baseline data on 2 patients in the standard-blood group who received a transfusion were missing. Data on transfusion were missing for 36
patients (15 patients in the fresh-blood group and 21 patients in the standard-blood group).
‡ Patients assigned to the fresh-blood group were considered to be adherent to the transfusion protocol if they received red-cell transfusions
only with units stored for less than 8 days; patients assigned to the standard-blood group were considered to be adherent if they received
transfusions of the oldest available compatible red cells.
§ For red-cell units, adherence was defined as (number of red-cell units transfused that were stored ≤7 days) ÷ (total number of red-cell units
transfused) among patients assigned to the fresh-blood group and as (number of standard-issue red-cell units transfused) ÷ (total number
of red-cell units transfused) among patients assigned to the standard-blood group.
1414 n engl j med 372;15

230 nejm.org april 9, 2015
Source: The New England JournalAge
of Medicine
of Tr ansfused Blood in Critically Ill Adults
A Primary Outcome and Secondary Outcomes Related to Death and Major Illnesses
Fresh Standard
Outcome Blood Blood Absolute Risk Difference (95% CI)
no. of patients/total no. (%) percentage points
Primary outcome: death by day 90 448/1211 (37.0) 430/1219 (35.3) 1.7 (−2.1 to 5.5)
Secondary outcomes
Death
In ICU 324/1214 (26.7) 295/1217 (24.2) 2.5 (−1.0 to 5.9)
In hospital 403/1212 (33.3) 386/1211 (31.9) 1.4 (−2.3 to 5.1)
By day 28 371/1214 (30.6) 353/1225 (28.8) 1.7 (−1.9 to 5.4)
Major Illnesses
Multiple organ dysfunction syndrome 162/1206 (13.4) 157/1207 (13.0) 0.4 (−2.3 to 3.1)
Acute respiratory distress syndrome 69/1206 (5.7) 80/1207 (6.6) −0.9 (−2.8 to 1.0)
Cardiovascular failure 61/1206 (5.1) 51/1207 (4.2) 0.8 (−0.8 to 2.5)
Cardiac ischemia or infarction 54/1206 (4.5) 44/1207 (3.6) 0.8 (−0.7 to 2.4)
Deep-vein thrombosis or pulmonary 43/1206 (3.6) 43/1207 (3.6) 0.0 (−1.5 to 1.5)
embolism
Nosocomial infection 411/1206 (34.1) 378/1207 (31.3) 2.8 (−0.9 to 6.5)
Acute transfusion reaction 4/1206 (0.3) 6/1207 (0.5) −0.2 (−0.7 to 0.3)
−10.0 −5.0 0.0 5.0 10.0
Fresh Blood Standard Blood

Better Better
B Other Secondary Outcomes

Fresh Standard
Outcome Blood Blood Mean Difference (95% CI)
Mean (±SD) value
MODS
Highest score 6.4±3.2 6.2±3.2 0.2 (−0.1 to 0.4)
Delta score 1.4±1.8 1.4±1.9 −0.1 (−0.2 to 0.1)
Duration of supportive care (days)
Mechanical ventilation 15.0±18.0 14.7±14.9 0.3 (−1.1 to 1.6)
Cardiac or vasoactive drugs 7.1±10.2 7.5±11.2 −0.4 (−1.2 to 0.5)
Extrarenal epuration 2.5±10.1 2.3±8.3 0.2 (−0.6 to 0.9)
Length of stay (days)
In ICU 15.3±15.4 15.3±14.8 0.1 (−1.2 to 1.3)
In hospital 34.4±39.5 33.9±38.8 0.5 (−2.6 to 3.7)
−10.0 −5.0 0.0 5.0 10.0
Fresh Blood Standard Blood

Better Better
Figure 1. Forest Plot of Absolute Risk Differences in Primary and Secondary Outcomes.
All estimates have 95% confidence intervals that include zero but fall on the “standard blood better” side of the line. The Multiple Organ
Dysfunction Score (MODS) ranges from 0 to 24, with higher scores indicating more severe organ dysfunction.17 The delta MODS is the
difference between the MODS at randomization and the highest daily MODS observed thereafter. ICU denotes intensive care unit.
Primary Outcome Secondary Analyses

At 90 days after randomization, 448 of 1211 pa- The survival analysis of the time to death showed
tients (37.0%) in the fresh-blood group and 430 a hazard ratio in the fresh-blood group, as com-
of 1219 patients (35.3%) in the standard-blood pared with the standard-blood group, of 1.1 (95%
group had died. The unadjusted absolute risk dif- CI, 0.9 to 1.2) (P = 0.38) (Fig. 2). No significant
ference was 1.7 percentage points (95% confi- difference in mortality was observed between the
dence interval [CI], −2.1 to 5.5), and the adjusted groups on the basis of age, number of units
risk difference was also 1.7 percentage points transfused, APACHE II score, or admission cate-
(95% CI, −1.6 to 4.9) (Fig. 1). gory (Fig. S3 in the Supplementary Appendix).

april 9, 2015 1415
critically ill patients. Not only were the primary

100
outcomes similar in the two study groups, but
the results were consistent in all per-protocol and
subgroup analyses. These findings have impor-
75
tant implications for the critical care and blood-
Survival (% of patients)
Standard blood
Fresh blood
transfusion communities. We surmise that the
50
use of fresh red cells is not justified at this time.
We might also infer that changes to red cells or
the storage medium that have been documented
25
in many laboratory studies18 may have limited
clinical consequences.
P=0.38 by log-rank test The results of our trial are consistent with
0 those of seven randomized, controlled trials that
0 20 40 60 80 100 compared various durations of red-cell storage.
Days Five pilot trials did not detect clinically impor-
tant clinical consequences of prolonged red-cell
Figure 2. Kaplan–Meier Survival Analysis of Time to Death in the Intention-
storage.10,11,19-21 Moreover, in two larger trials,
to-Treat Population.
transfusion of fresh red cells, as compared with
The intention-to-treat population included 2430 patients. The hazard ratio
in the fresh-blood group, as compared with the standard-blood group, was standard-issue red cells, did not reduce the com-
1.1 (95% CI, 0.9 to 1.2). plications of prematurity in very-low-birth-weight
infants22 or reduce the rates of organ failure or
adverse events among 1098 patients undergoing
No significant differences were observed with elective cardiac surgical procedures.23,24
respect to major illnesses, duration of respirato- Our findings are not consistent with those of
ry, hemodynamic, or renal support, or length of some previous observational studies that sug-
stay in the ICU or hospital. Acute transfusion re- gested that prolonged red-cell storage may be
actions occurred in four patients in the fresh- deleterious.5,9,25-27 More than 40 observational
blood group and six patients in the standard-blood studies have examined the effect of red-cell stor-
group (Table S3 in the Supplementary Appendix). age on various clinical outcomes, including mor-
We performed a per-protocol analysis of the tality, rates of infection, and length of stay in the
primary outcome that included only patients hospital. Although the initial studies showed an
who received a transfusion. At 90 days, 423 of association between longer red-cell storage and
1153 patients (36.7%) in the fresh-blood group adverse outcomes, these associations may have
and 398 of 1163 patients (34.2%) in the stan- been spurious owing to sicker patients receiving
dard-blood group had died (absolute risk differ- more units with longer storage, the overlap be-
ence, 2.5 percentage points; 95% CI, −1.4 to 6.4). tween comparison groups in the age of the red
We also performed a sensitivity analysis of the cells transfused, and the inclusion of transfu-
primary outcome in which we compared the sions that occurred after the clinical events.28
outcomes of the 967 patients in the fresh-blood Results from more recent studies have been
group who received only red cells that had been more balanced, with many studies showing no
stored for less than 8 days versus the outcomes significant association between increased dura-
in the 1084 patients in the standard-blood group tion of red-cell storage and a worse outcome or
who received red cells that had been stored for showing worse outcomes with fresh red cells.29-31
more than 7 days. In this sensitivity analysis, the Our trial has a number of strengths. It was suf-
number of deaths at 90 days was 357 (36.9%) in ficiently large to detect clinically important differ-
the fresh-blood group and 370 (34.1%) in the ences in 90-day mortality. In addition, we enrolled
standard-blood group (absolute risk difference, a wide spectrum of critically ill patients, ensuring
2.8 percentage points; 95% CI, −1.4 to 6.9). broad applicability of our findings. Bias was mini-
mized by concealed randomization, blinded study-
Discussion group assignments, and a rate of loss to follow-
up of less than 5%. The between-group difference
The ABLE study did not show any benefit attrib- in the duration of red-cell storage was statistically
utable to the transfusion of fresh red cells in and clinically significant. The primary outcome
1416 232372;15
was clinically relevant and important to both is supplied in Canada and Europe. Red cells
patients and medical decision makers. suspended in additive solution 3 (AS-3), which
Limitations of this trial include the possibil- are supplied in the United States, are similar.
ity that some groups of critically ill patients who However, there are differences in the method of
are particularly vulnerable to the adverse conse- production and in storage solutions. The storage
quences of prolonged red-cell storage were under- solutions result in similar, though not identical,
represented in the trial. Overall, most patients in vitro red-cell defects,32 suggesting that our
received transfusions according to a restrictive results may be generalizable. Finally, we asked
transfusion strategy, with a mean pretransfusion “Is fresh blood better than old blood?” rather
hemoglobin level of 7.7 g per deciliter, even in the than “Is old blood bad?”33 Our trial does not ad-
absence of guidelines or protocols. Thus, expo- dress the issue of whether the use of red cells
sure to any red cells in this trial was much less stored for very prolonged periods (35 to 42 days)
than what would be the case at centers that are results in harm.
still opting for more liberal use of blood transfu- In conclusion, we did not detect any clinically
sions. Although the testing of shorter storage important improvements in primary or second-
times might have led to different results, we ary outcomes among critically ill adults who re-
selected a 7-day storage threshold for the fresh- ceived transfusions of fresh red cells.
blood group because blood-transfusion consul- Supported by peer-reviewed grants from the Canadian Insti-
tants suggested that post-trial implementation tutes of Health Research (177453), Fonds de Recherche du Qué-
bec–Santé (24460), the National Institute for Health Research
of a shorter storage period would not be feasible, Evaluation, Trials, and Studies Coordinating Centre Health
especially given the requirement for infectious- Technology Assessment Program, and the French Ministry of
disease testing (which takes up to 72 hours) and Health Programme Hospitalier de Recherche Clinique (12.07,
2011) and by funding from Établissement Français du Sang and
the finite number of blood donors. Sanquin Blood Supply. Dr. Tinmouth is supported by an Ottawa
We included only centers that used leukore- Hospital Department of Medicine research award, Dr. Cook is a
duced red cells; therefore, whether leukocytes Research Chair of the Canadian Institutes of Health Research,
and Dr. Turgeon is supported by a Career Award from Fonds de
worsen the degradation of red cells during stor- Recherche du Québec–Santé.
age or have other toxic effects that are enhanced Disclosure forms provided by the authors are available with
by prolonged storage is uncertain. Only SAGM- the full text of this article at NEJM.org.
We thank the patients, their family members, the site investi-
suspended red cells were used in the trial, be- gators, the research coordinators, and the blood-transfusion
cause this is the standard red-cell product that personnel who participated in this trial.
References
1. Hébert PC, Wells G, Blajchman MA, cells: understanding immune and vascu- pathophysiology and therapy. Berlin:
et al. A multicenter, randomized, con- lar reactivity. Transfusion 2011;51:894-900. Springer-Verlag, 1994:38-49.
trolled clinical trial of transfusion re- 8. Aubron C, Nichol A, Cooper DJ, Bel- 14. Garner JS, Jarvis WR, Emori TG,
quirements in critical care. N Engl J Med lomo R. Age of red blood cells and trans- Horan TC, Hughes JM. CDC definitions
1999;340:409-17. [Erratum, N Engl J Med fusion in critically ill patients. Ann Inten- for nosocomial infections, 1988. Am J In-
1999;340:1056.] sive Care 2013;3:2. fect Control 1988;16:128-40. [Erratum,
2. Lacroix J, Hébert PC, Hutchison JS, et 9. Wang D, Sun J, Solomon SB, Klein Am J Infect Control 1988;16:177.]
al. Transfusion strategies for patients in HG, Natanson C. Transfusion of older 15. Lachin JM. Introduction to sample
pediatric intensive care units. N Engl J stored blood and risk of death: a meta- size determination and power analysis for
Med 2007;356:1609-19. analysis. Transfusion 2012;52:1184-95. clinical trials. Control Clin Trials 1981;
3. Holst LB, Haase N, Wetterslev J, et al. 10. Kor DJ, Kashyap R, Weiskopf RB, et 2:93-113.
Lower versus higher hemoglobin thresh- al. Fresh red blood cell transfusion and 16. Knaus WA, Draper EA, Wagner DP,
old for transfusion in septic shock. short-term pulmonary, immunologic, and Zimmerman JE. APACHE II: a severity of
N Engl J Med 2014;371:1381-91. coagulation status: a randomized clinical disease classification system. Crit Care
4. The ProCESS Investigators. A ran- trial. Am J Respir Crit Care Med 2012; Med 1985;13:818-29.
domized trial of protocol-based care for 185:842-50. 17. Marshall JC, Cook DJ, Christou NV,
early septic shock. N Engl J Med 2014; 11. Hébert PC, Chin-Yee I, Fergusson D, Bernard GR, Sprung CL, Sibbald WJ. Mul-
370:1683-93. et al. A pilot trial evaluating the clinical tiple organ dysfunction score: a reliable
5. Tinmouth A, Fergusson D, Yee IC, effects of prolonged storage of red cells. descriptor of a complex clinical outcome.
Hébert PC. Clinical consequences of red Anesth Analg 2005;100:1433-8. Crit Care Med 1995;23:1638-52.
cell storage in the critically ill. Transfu- 12. Lacroix J, Hébert PC, Fergusson D, et 18. Hess JR. Red cell changes during stor-
sion 2006;46:2014-27. al. The Age of Blood Evaluation (ABLE) age. Transfus Apher Sci 2010;43:51-9.
6. Bennett-Guerrero E, Veldman TH, randomized controlled trial: study de- 19. Walsh TS, McArdle F, McLellan SA, et
Doctor A, et al. Evolution of adverse sign. Transfus Med Rev 2011;25:197-205. al. Does the storage time of transfused
changes in stored RBCs. Proc Natl Acad 13. Marshall JC. A scoring system for red blood cells influence regional or
Sci U S A 2007;104:17063-8. multiple organ dysfunction syndrome. In: global indexes of tissue oxygenation in
7. Spinella PC, Sparrow RL, Hess JR, Vincent JL, Reinhart K, Eyrich K, Sprung anemic critically ill patients? Crit Care
Norris PJ. Properties of stored red blood C, eds. Sepsis: current perspectives in Med 2004;32:364-71.
n engl j med 372;15 233

20. Heddle NM, Cook RJ, Arnold DM, et age age is not associated with a signifi- sion recipients. Transfus Med Rev 2013;
al. The effect of blood storage duration on cant difference in multiple organ dys- 27:36-43.
in-hospital mortality: a randomized con- function score or mortality in transfused 30. Kaukonen KM, Vaara ST, Pettilä V,
trolled pilot feasibility trial. Transfusion cardiac surgery patients. Transfusion et al. Age of red blood cells and outcome
2012;52:1203-12. 2014;54:Suppl:15A. abstract. in acute kidney injury. Crit Care 2013;17:
21. Aubron C, Syres G, Nichol A, et al. A 25. van de Watering L. Red cell storage R222.
pilot feasibility trial of allocation of fresh- and prognosis. Vox Sang 2011;100:36-45. 31. Heddle NM, Eikelboom J, Liu Y, Barty
est available red blood cells versus stan- 26. Vamvakas EC. Meta-analysis of clini- R, Cook RJ. Exploratory studies on the
dard care in critically ill patients. Trans- cal studies of the purported deleterious age of transfused blood and in-hospital
fusion 2012;52:1196-202. effects of “old” (versus “fresh”) red mortality in patients with cardiovascular
22. Fergusson DA, Hébert P, Hogan DL, blood cells: are we at equipoise? Transfu- diagnoses. Transfusion 2015;55:364-72.
et al. Effect of fresh red blood cell trans- sion 2010;50:600-10. 32. Antonelou MH, Kriebardis AG, Sta-
fusions on clinical outcomes in prema- 27. Lelubre C, Vincent JL. Relationship moulis KE, Economou-Petersen E, Mar-
ture, very low-birth-weight infants: the between red cell storage duration and garitis LH, Papassideri IS. Red blood cell
ARIPI randomized trial. JAMA 2012;308: outcomes in adults receiving red cell aging markers during storage in citrate-
1443-51. transfusions: a systematic review. Crit phosphate-dextrose-saline-adenine-glucose
23. Steiner ME, Assmann SF, Levy JH, et Care 2013;17:R66. -mannitol. Transfusion 2010;50:376-89.
al. Addressing the question of the effect 28. van de Watering L. Pitfalls in the cur- 33. Bennett-Guerrero E, Stafford-Smith
of RBC storage on clinical outcomes: the rent published observational literature M, Waweru PM, et al. A prospective, double-
Red Cell Storage Duration Study (RECESS) on the effects of red blood cell storage. blind, randomized clinical feasibility trial
(Section 7). Transfus Apher Sci 2010;43: Transfusion 2011;51:1847-54. of controlling the storage age of red blood
107-16. 29. Middelburg RA, van de Watering LM, cells for transfusion in cardiac surgical pa-
24. Steiner ME, Triuzi DJ, Assinavy SY, et Briët E, van der Bom JG. Storage time of tients. Transfusion 2009;49:1375-83.
al. Randomized trial results: red cell stor- red blood cells and mortality of transfu- Copyright © 2015 Massachusetts Medical Society.


234 april 9, 2015
FEEDING CRITICALLY ILL PATIENTS
Meeting the basic nutritional needs of patients would seem straightforward, but each approach has strengths
and weaknesses that can help or compromise a patient’s well-being. Unfortunately this is an area where the
amount of high-quality evidence is not as great as one would hope, and consequently a lot of decisions are
based on expert opinion. This review examines the evidence base on feeding patients in the acute phase of
illness and, on that basis, provides recommendations.
235
Case Challenge
Feeding Critically Ill Patients

A previously well-nourished 77-year-old man is undergoing mechanical ventilation after an emergency colon
resection, complicated by septic shock and then acute liver failure. What strategy would you use to provide
nutrition for this patient?
A well-nourished 77-year-old man whose medical history includes treated hypertension and hypercholesterol-
emia, previous heavy alcohol intake, and mild cognitive impairment was admitted to the intensive care unit (ICU)
of a university hospital from the operating room after a Hartmann’s procedure (resection of the rectosigmoid
colon with closure of the rectal stump and formation of an end colostomy) performed for fecal peritonitis due to
a perforated sigmoid colon. On arrival in the ICU, he was in septic shock. He is undergoing mechanical ventila-
tion with the use of a low-tidal-volume protocol with positive end-expiratory pressure (PEEP). His arterial blood
pressure is supported with a norepinephrine infusion. Analgesia is provided by a continuous morphine infusion.
Slight bleeding from the surgical site and from the areas around arterial and central venous catheters is most likely
due to low-grade disseminated intravascular coagulation and does not currently merit any specific treatment other
than withholding previously prescribed heparin and repeating the laboratory tests in 8 to 12 hours. (In the previ-
ous installment of this case, there were 4272 votes on strategies for treating the patient’s bleeding and possible
236
coagulopathy. All the respondents voted for stopping heparin, with 45% favoring repeated laboratory testing
in 8 to 12 hours, 33% favoring transfusion of cryoprecipitate and fresh-frozen plasma or prothrombin complex
concentrate, 11% favoring transfusion of packed red cells, platelets, cryoprecipitate, and fresh-frozen plasma
or prothrombin complex concentrate, and 9% favoring transfusion of platelets plus the administration of an
antifibrinolytic agent, such as epsilon-aminocaproic acid or tranexamic acid.)

What strategy would you use to provide nutrition for this patient?
A. Initiate total parenteral nutrition as soon as possible after the patient’s arrival in the ICU.
B. Await the return of bowel sounds and then initiate enteral nutrition.
C. Initiate enteral nutrition within 24 to 48 hours after ICU admission and then initiate parenteral nutrition on
day 7 if the caloric target is not being met.
D. Initiate parenteral nutrition as soon as possible after the patient’s arrival in the ICU and then start enteral
nutrition once bowel sounds return.
237
review article
Nutrition in the Acute Phase

of Critical Illness
Michael P. Casaer, M.D., Ph.D., and Greet Van den Berghe, M.D., Ph.D.
C
ritically ill patients requiring vital organ support in the in- From the Clinical Department and Labo-
tensive care unit (ICU) commonly have anorexia and may be unable to feed ratory of Intensive Care Medicine, Divi-
sion of Cellular and Molecular Medicine,
volitionally by mouth for periods ranging from days to months. Unless such Katholieke Universiteit Leuven, Leuven,
patients are provided with macronutrients in the form of enteral or parenteral nutri- Belgium. Address reprint requests to Dr.
tion, they accumulate an energy deficit that rapidly reaches proportions that con- Van den Berghe at the Clinical Department
and Laboratory of Intensive Care Medi-
tribute to lean-tissue wasting and that are associated with adverse outcomes.1 The cine, Katholieke Universiteit Leuven, Her-
catabolic response to acute critical illness is much more pronounced than that estraat 49, B-3000 Leuven, Belgium, or at
evoked by fasting in healthy persons, since the energy deficit in acutely ill patients greet.vandenberghe@med.kuleuven.be.
is often superimposed on immobilization and pronounced inflammatory and endo- N Engl J Med 2014;370:1227-36.
crine stress responses. Severe skeletal-muscle wasting and weakness occurring dur- DOI: 10.1056/NEJMra1304623
ing critical illness are associated with a prolonged need for mechanical ventilation
and rehabilitation.2
In many studies, the degree of energy deficit accumulating in critically ill pa-
tients is strongly associated with the duration of stay in the ICU, which, in turn,
is associated with an increased incidence of infectious complications and risk of
death.1 Until recently, however, the causality of these associations remained un-
clear, since the majority of studies that formed the basis of published recommen-
dations for feeding ICU patients were either observational or small interventional
studies.3,4 Recently, the field of critical care nutrition has been revived by the
findings of several randomized, controlled trials, which have opened a new debate
on nutritional practice in the ICU.
For this review, we focus on evidence from randomized, controlled studies that
met the following criteria: study-group assignments were made in a blinded fash-
ion, the sample size was sufficiently large to detect a prespecified treatment effect,
there was a clear delineation of the way in which patients were selected and fol-
lowed, there was a statistical analysis plan with end points defined a priori, and
there was an intention-to-treat analysis with adequate handling of competing risks
(Table 1). In this review, we integrate this newer evidence with older high-level
evidence to provide suggestions for feeding during the acute phase of critical ill-
ness. In cases in which such evidence does not exist, we identify areas of uncer-
tainty that require further investigation.
En ter a l Nu t r i t ion
Timing of Initiation
Anorexia is part of the acute physiologic response to severe illness that can be either
adaptive or maladaptive. Studies in animals and humans have shown a trophic ef-
fect of enteral nutrients on the integrity of the gut mucosa, a finding that has pro-
vided the rationale for instituting enteral nutrition early during critical illness. In
observational studies, patients in the ICU who were fed early through the enteral
n engl j med 370;13 nejm.org march 27,238

2014 1227
The New England
For of Medicine
Medical personal
Society. use only.
reserved.
1228
Table 1. Nutritional Interventions for Critically Ill Patients, According to Data from Randomized, Controlled Trials.
Nutritional Intervention Rationale High-Quality Evidence* References

Enteral feeding versus no artificial feeding Prevention of lean-tissue wasting, weakness, NA NA
in acute critical illness and infections to enhance recovery
Increased enteral feeding during first week Early prevention of caloric deficit to enhance No clear benefit in well-nourished patients Martin et al.,5 Doig et al.,6 National Heart,
in ICU recovery Lung, and Blood Institute Acute
Respiratory Distress Syndrome Clinical

Trials Network,7 Needham et al.,8 Arabi
The
et al.,9 Montejo et al.,10† Reignier et al.11

Use of prokinetic agents or postpyloric Early prevention of caloric deficit to enhance Inconclusive Davies et al.,12 Acosta-Escribano et al.13‡
MULARCZYK
feeding recovery
Supplemental parenteral feeding during Early prevention of caloric deficit to enhance No clear benefit and potential harm with Casaer et al.,14 Heidegger et al.,15 Doig
first week in ICU recovery higher doses in unselected patients re- et al.16

ceiving some enteral nutrition; inconclu-

The New England
© Massachusetts
sive in patients with contraindication to
n engl j med 370;13
on January 15,
enteral nutrition
ForJournal
Medical
239
Use of increased protein (>0.8 g/kg/day) Sparing of endogenous protein to enhance NA NA
Medical
personal
recovery
2015. For
nejm.org
Use of glutamine Resupply of a conditional deficiency to Inconclusive and potentially harmful in Wernerman et al.,17 Andrews et al.,18
Society.
Society.
reduce mortality higher doses Heyland et al.19

of Medicine
All
of
Use of antioxidants Prevention of organ failure No clear benefit; for selenium, possibly de- Heyland et al.,19 Alhazzani et al.20
personal
pendent on dose and preexisting defi-
rights
ciency status
All rights
march 27, 2014
use only.
Use of antiinflammatory lipids Prevention of organ failure Inconclusive Pontes-Arruda et al.,21,22 Umpierrez
et al.,23§ Rice et al.24¶
reserved.
m e dic i n e
reserved.
* The quality of randomized, controlled trials was considered to be high when the study had blinded study-group assignments, a sample size sufficiently large to detect the hypothesized
treatment effect, a diagram showing how patients were selected, prespecified end points, and an intention-to-treat analysis. NA denotes not available because evidence is lacking from
high-quality randomized, controlled trials.
† No diagram showing how patients were selected was included in the study by Montejo et al.10
‡ No information on the way in which investigators concealed study-group assignments was included in the study by Acosta-Escribano et al.13
§ No sample-size calculation and only partial outcome data in the intention-to-treat population were included in the study by Pontes-Arruda et al.,22 and no diagram showing how

patients were selected was included in the study by Umpierrez et al.23
¶ The study by Rice et al.24 was stopped prematurely for futility at the first interim analysis.

route have had a better outcome than those who penditure received more nutrition than did con-
were not.25 However, the inability to provide en- trol patients, who were fed with a calculated
teral nutrition early may be a marker of the sever- energy target. The intervention led to a trend
ity of illness (i.e., patients who can be fed enter- toward reduced hospital mortality but a signifi-
ally are less ill than those who cannot) rather cant increase in infections and in the length of
than a mediator of complications and poor out- stay in the ICU30 (Fig. 1).
comes. Because of interruptions in feeding for a vari-
Hence, the first question to address is whether ety of reasons and delayed gastric emptying,
data from methodologically sound, randomized, patients often receive less than the prescribed
controlled trials support the initiation of enteral amount of enteral nutrition. Failure to deliver
feeding early in the acute phase of critical ill- the prescribed nutrition has been considered to
ness. A meta-analysis of six small trials involv- be one of the reasons that the use of enteral
ing a total of 234 patients in the ICU showed a nutrition has not improved the outcome in criti-
survival benefit with the immediate initiation of cally ill patients. This hypothesis was supported
enteral nutrition, as compared with delayed ini- by a small, randomized, controlled trial involv-
tiation.26 Unfortunately, the quality of the indi- ing patients with traumatic brain injury, which
vidual studies in this meta-analysis was poor. showed that delivering enteral nutrition to reach
For example, in three of the studies, the com- an estimated energy target immediately after
parator group received parenteral nutrients with- ICU admission, rather than to reach gradually
in 24 hours after ICU admission, a factor that increasing targets over the first week in the ICU,
made the interpretation of the results difficult.27 resulted in a reduced rate of infection.31
The rationale for the early initiation of enteral To address these issues, approaches were suc-
nutrition rather than parenteral nutrition is the cessfully developed to deliver larger amounts of
lower risk of infection with enteral nutrition that enteral nutrition earlier during critical illness.5
was observed in older randomized, controlled Two large, cluster-randomized trials5,6 enrolling
trials that were conducted in an era in which 462 and 1118 patients, respectively, investigated
investigators may have underestimated the harm the effect of such protocols on clinical out-
of pronounced hyperglycemia and overfeeding in comes. In the two studies, the protocols in-
critically ill patients.28 However, large, high-quality, creased the amount of nutrition delivered. In the
randomized, controlled trials supporting an out- smaller study, implementation of the protocol
come benefit with early enteral nutrition versus resulted in a decreased length of hospital stay
delayed nutrition during the acute phase of and a nonsignificant reduction in hospital mor-
critical illness have not been performed. tality. In the larger study, in which participating
centers were not allowed to cross over, imple-
Estimation of Energy Requirements mentation of the protocol resulted in an earlier
The energy requirements of critically ill patients initiation of feeding and an increased attain-
continue to be debated.29 Such requirements are ment of caloric goals, but this did not provide
often estimated with the use of characteristics of any benefit in terms of either mortality or length
the patients before the onset of illness.29 Other of stay in the ICU or hospital.6
observers argue that energy requirements differ In the more recent Trophic vs. Full-Energy
per patient and per day in the ICU and thus Enteral Nutrition in Mechanically Ventilated Pa-
should be individually estimated on a daily basis tients with Acute Lung Injury (EDEN) trial, inves-
from measurements of oxygen consumption and tigators addressed the question of how much
carbon dioxide production, as determined with enteral nutrition should be administered early
the use of indirect calorimetry.29 However, this during critical illness7 (Fig. 1). In this study, 1000
method is often technically difficult or impossi- relatively young and well-nourished patients with
ble to perform in critically ill patients.29 acute lung injury who were in the ICU and who
In the Tight Calorie Control Study (TICACOS) were considered to be eligible for enteral nutri-
involving 130 patients, those who were undergo- tion were randomly assigned to receive either just
ing mechanical ventilation in an ICU and who a small amount of (trophic) enteral feeding for
received nutrition that was guided by indirect 1 week in the ICU or full enteral feeding from
calorimetry to estimate the resting energy ex- the time of admission onward. The first 272 pa-
n engl j med 370;13 240

nejm.org march 27, 2014 1229
The New
For personal ofNo
use only. Medicine
© Massachusetts JanuarySociety.
15, 2015.
AllFor personal
reserved.
1230
Full (N=492) Early PN(N=686) Early PN(N=2312) Full (N=153) REE (N=65)
Trophic (N=508) Standard (N=686) Late PN (N=2328) EN only(N=152) Calculated(N=65)
2500 2500 2500 2500 2500
2000 2000 2000 2000 2000
1500 1500 1500 1500 1500
(kcal)
1000 1000 1000 1000 1000
Total Energy
500 500 500 500 500
0 0 0 0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
ICU Day

EDEN Trial Early PN Trial EPaNIC Trial SPN Trial TICACOS
Copyright
(N=1000) (N=1372) (N=4640) (N=305) (N=130)
The
Type of Patients Medical (acute lung injury) Mixed medical and surgical Mixed medical and surgical Mixed medical and surgical Mixed medical and surgical
Eligible for EN EN relatively contraindicated (unselected) (on day 4)
(short term) With nutritional risk (NRS, ≥3) Eligible for EN but <60% target
MULARCZYK
New Infections in ICU Unaffected Unaffected More with early PN Between day 9 and day 28: More with REE
© Massachusetts
less with SPN
From randomization to day 28:
The New England

on January
unaffected
n engl j med 370;13
Medical
For personal
Duration of Mechanical Unaffected Shorter with early PN Longer with early PN Unaffected Longer with REE
Journal
Ventilation
15, 2015.
Society.
241 nejm.org
Length of Stay in ICU Unaffected Unaffected Longer with early PN Unaffected Longer with REE
useofonly.
ForAll
Unaffected Unaffected Unaffected Unaffected Unaffected
Medicine
Mortality in ICU
of
(60-day mortality: unaffected) (trend toward reduced
personal
hospital mortality)
rights use
march 27, 2014
Figure 1. Comparison of Macronutrient Intake and Outcomes of Five Randomized, Controlled Trials Evaluating Nutrition during Critical Illness.
reserved.
Shown are the outcomes with respect to new infections, duration of mechanical ventilation, duration of stay in the intensive care unit (ICU), and ICU mortality in the five trials.

m e dic i n e
The primary end points were the duration of mechanical ventilation in the Trophic vs. Full-Energy Enteral Nutrition in Mechanically Ventilated Patients with Acute Lung Injury
(EDEN) trial,7 60-day mortality in the Early Parenteral Nutrition (Early PN) trial,16 length of stay in the ICU in the Impact of Early Parenteral Nutrition Completing Enteral Nutrition
in Adult Critically Ill Patients (EPaNIC) trial,14 rate of new infections from randomization to day 28 in the Impact of Supplemental Parenteral Nutrition on Infection Rate, Duration
of Mechanical Ventilation, and Rehabilitation in ICU Patients (SPN) trial,15,41 and hospital mortality in the Tight Calorie Control Study (TICACOS).29 (In TICACOS, feeding was

guided by either indirect calorimetry to estimate the resting energy expenditure [REE] or a calculated energy target.) The comparison reveals no benefit with respect to survival or
intensive care dependency with enhanced enteral nutrition (EN) or parenteral nutrition (PN) early in a critical illness. The apparent differences among the studies with respect to
other outcomes may be related to differences in the dose of macronutrients or the route of macronutrient delivery. Solid curves represent days on which patients were not allowed
to receive parenteral nutrition, according to the study protocols. Dashed curves represent days on which patients were allowed to receive parenteral nutrition. All curves represent
mean values for total energy intake. In the SPN trial, values were reported as the percentage of the target value and have been converted to kilocalories to facilitate comparisons.
NRS denotes Nutritional Risk Score, which ranges from 0 to 7, with higher values indicating a higher nutritional risk.

tients who were enrolled underwent concomitant The administration of prokinetic agents such
randomization in the OMEGA trial, which inves- as metoclopramide or erythromycin has been ad-
tigated the effect of n–3 fatty acids in patients vocated to improve gastric emptying. Erythromy-
with acute lung injury. In accordance with the cin may be more effective than metoclopramide,
protocol of that trial, patients received an addi- but benefits with respect to clinical outcome have
tional amount of lipids or an isocaloric protein not been shown.32,33 Prokinetic-resistant impaired
dose.24 The EDEN trial was a large study with a gastric emptying is sometimes considered to be an
high methodologic standard. Although the pa- indication to bypass the stomach and to deliver
tients in the group receiving trophic feeding nutrition directly beyond the pylorus. Although
accumulated a substantially greater nutritional postpyloric feeding may allow increased amounts
deficit than did the group receiving full enteral of enteral nutrition to be given early, findings
feeding for 1 week, there was no between-group from randomized, controlled trials are inconclu-
difference in acute or long-term functional out- sive regarding the effect on clinical outcome.12,13
comes.7,8 These results are consistent with those
of a smaller randomized, controlled trial involv- Pa r en ter a l Feeding
ing 240 patients, in which patients with a variety
of illnesses who were in the ICU were assigned When clinicians rely solely on the enteral route to
either to an approach that allowed underfeeding feed patients in the ICU, the numbers of calories
or to one that targeted full feeding. Patients who that are administered often do not meet the cal-
were assigned to the approach that allowed un- culated targets. This discrepancy can be due to
derfeeding received fewer calories but had outside effects associated with enteral feeding or to
comes that were at least as good as those in the lack of a functional gastrointestinal tract.
patients assigned to early full feeding.9 The clinical question remains whether parenteral
nutrition should be initiated in such patients
Gastric Residual Volume and, if so, when.
Among patients in the ICU, gastric emptying is The findings of observational and small in-
often slow or impaired, which can result in large tervention studies have been inconclusive.28,34
gastric residual volumes with enteral feeding. Although European guidelines have recommended
Since regurgitation of gastric content can lead to the early initiation (within 48 hours after admis-
aspiration pneumonia, enteral feeding is often sion to the ICU) of parenteral nutrition so that
discontinued in patients who are found to have the accumulating nutritional deficit is prevented
large gastric residual volumes. For this reason, as soon as possible, American and Canadian
there is a longstanding controversy about wheth- guidelines have advised allowing hypocaloric
er the presence of large gastric residues is accept- enteral nutrition for 1 week in well-nourished
able. Two recent randomized, controlled trials patients before considering parenteral nutrition.3,4
addressed this question. The Gastric Residual The latter advice was based on the observation of
Volume during Enteral Nutrition in ICU Patients complications (e.g., liver-function abnormalities,
(REGANE) trial (involving 329 patients) showed hyperglycemia, hypertriglyceridemia, and infec-
that gastric residual volumes up to 500 ml could tions) associated with parenteral nutrition and
be safely tolerated.10 The Effect of Not Monitor- overfeeding reported in older studies.28,35-37
ing Residual Gastric Volume on the Risk of In a large, randomized, controlled trial — the
Ventilator-Associated Pneumonia in Adults Re- Impact of Early Parenteral Nutrition Completing
ceiving Mechanical Ventilation and Early Enteral Enteral Nutrition in Adult Critically Ill Patients
Feeding (NUTRIREA 1) trial (involving 449 pa- (EPaNIC) trial14 — patients in the two groups re-
tients) showed that the omission of the measure- ceived tight glucose control (target blood glucose
ment of gastric residual volumes did not increase level, 80 to 110 mg per deciliter [4.4 to 6.1 mmol
the incidence of aspiration or related complica- per liter]) and parenteral trace elements and vi-
tions.11 Interestingly, the two studies showed tamins, which were administered until the ini-
that allowing large gastric residual volumes in- tiation of adequate enteral feeding (Fig. 1). Patients
creased the amount of enteral feeding that was who received early parenteral nutrition to supple-
administered early during critical illness but did ment insufficient enteral nutrition were given
not affect clinical outcomes. parenteral glucose for 2 days (approximately 100 g

242 march 27, 2014 1231
The New
only.ofNo
Medicine
by MIKE MULARCZYK
JanuarySociety.
15, 2015.
AllFor personal
reserved.
on day 1 and 180 g on day 2), which was fol- ized intervention had no significant effect on
lowed from day 3 onward by the administration other clinical end points. In contrast to earlier
of commercially available all-in-one parenteral trials, the administration of parenteral nutrition
nutrition preparations delivering on average 40 g was not associated with an increased risk of in-
of protein per liter per day. The study enrolled fection. However, it was also not associated with
4640 critically ill patients with an average score a net clinical benefit.
of 23 on the Acute Physiology and Chronic Health In a third trial, the Early Parenteral Nutrition
Evaluation II (APACHE II) (on which scores range study,16 investigators addressed another prag-
from 0 to 71, with higher scores indicating more matic question: should parenteral nutrition be
severe disease) who were admitted after cardiac administered early to the subgroup of critically
surgery, medical complications after surgery, mul- ill patients who have a relative contraindication
tiple or cerebral trauma, sepsis, or other life- to early enteral nutrition? In this study, 1372 pa-
threatening emergencies. After 1 week in the ICU, tients from 31 ICUs were assigned to receive ei-
the delivered enteral nutrition in the two study ther early parenteral nutrition (as an all-in-one
groups was approximately 20% of the estimated preparation containing 33 g of protein per liter)
energy requirements. Unexpectedly, patients who within 24 hours after ICU admission or standard
received insufficient enteral nutrition had an therapy at the discretion of the treating physi-
earlier live discharge from the ICU and hospital, cian (Fig. 1). Patients were enrolled an average of
a lower incidence of new ICU infections and of 13.8 hours after ICU admission, and early paren-
ICU-acquired weakness,38 and a lower duration of teral nutrition was started within an hour after
vital-organ support than did patients receiving enrollment. In the standard-therapy group, 253 of
insufficient enteral nutrition supplemented with 686 patients (36.9%) received parenteral nutrition
parenteral nutrition.14 There were substantial cost during the first few days in the ICU, with 27.1%
savings in the group not receiving the parenteral of them receiving parenteral nutrition after a
nutrition, which were explained largely by a re- mean of 1.99 days and another 7.0% receiving
duced need for antibacterial and antifungal supplemental parenteral nutrition with enteral
drugs.39 Results were consistent regardless of nutrition after a mean of 5.58 days. The protocol
the type or severity of illness.14,40 advised study physicians to prescribe parenteral
In a second trial, the Impact of Supplemental trace elements and vitamins only for the group
Parenteral Nutrition on Infection Rate, Duration receiving early parenteral nutrition. There was no
of Mechanical Ventilation, and Rehabilitation in significant between-group difference in the pri-
ICU Patients (SPN) study,15 investigators addressed mary end point, 60-day mortality, but the duration
the pragmatic question of how to treat patients of mechanical ventilation (a tertiary outcome) was
who are eligible to be fed enterally but who shorter in the group receiving early parenteral
cannot tolerate full enteral feeding after 3 days nutrition.16
(12% of admissions in the SPN study) (Fig. 1). It remains unknown whether early parenteral
The study enrolled 305 patients who were not yet nutrition is beneficial for patients who have an
receiving 60% of the energy goal on the fourth absolute and more prolonged contraindication
day in the ICU, a target that was determined on to enteral nutrition. Since the avoidance of par-
the basis of indirect calorimetry performed in enteral feeding results in prolonged fasting,
65% of the patients. Patients were randomly as- such patients are often excluded from studies.
signed to receive supplemental parenteral nutri- A meta-analysis of seven randomized, controlled
tion or enteral nutrition alone from day 4 in the trials published between 1981 and 1994 (involv-
ICU to day 8. The mean (±SD) between-group ing a total of 798 patients) showed that paren-
difference in energy intake over the 4 interven- teral nutrition, as compared with no feeding,
tion days was about 25% of the target (103±18% was associated with a higher rate of infection.42
vs. 77±27%). The incidence of infections between In a post hoc subgroup analysis of the EPaNIC
day 9 and day 28 was lower among patients as- trial, 517 patients who were admitted to the ICU
signed to the parenteral-nutrition group. How- with a surgical contraindication for enteral feed-
ever, the rate of infection during the first 28 days ing had fewer infections and an increased likeli-
of the ICU stay, which was the primary end point hood of earlier live discharge from the ICU if
of the study, was unaffected.15,41 The random- they were not assigned to receive early parenteral
1232 n engl j med 370;13 243

The New EnglandFor

reserved.
nutrition.14 For these patients, starvation was slow recruitment showed no difference in the
tolerated for 1 week in the ICU and resulted in rate of death or organ dysfunction with gluta-
improved clinical outcomes.14 Thus, the most mine supplementation.17
effective time at which the initiation of paren- In two recent high-quality, randomized, con-
teral nutrition can produce a clear clinical bene- trolled trials, investigators studied the effects
fit during critical illness remains unclear. of two doses of glutamine in critically ill pa-
tients. In the Scottish Intensive Care Glutamine
Sel ec t ion of M acronu t r ien t s or Selenium Evaluative Trial (SIGNET),18 involv-
ing 500 patients, investigators evaluated the ef-
Amino Acids fects of a glutamine dose of 0.1 to 0.2 g per kilo-
Another controversial topic is the preferred amino gram per day, whereas in the Reducing Deaths
acid content of enteral and parenteral prepara- Due to Oxidative Stress (REDOXS) trial,19 involv-
tions. Gluconeogenesis, which uses amino acids ing 1223 patients, investigators evaluated a glu-
generated from the breakdown of skeletal mus- tamine dose of 0.6 to 0.8 g per kilogram per day.
cle, cannot be fully suppressed by exogenous glu- The SIGNET trial showed no benefit of low-dose
cose during critical illness. This phenomenon can glutamine administered parenterally to patients
be explained, in part, by the complex stress re- receiving parenteral feeding.18 The REDOXS trial
sponse, which causes hepatic insulin resistance. showed an absolute increase of 6.5 percentage
It was therefore inferred that the administration points in the rate of death at 6 months among
of exogenous protein could induce a protein- patients with organ failure who received early
sparing effect in critically ill patients through the high-dose parenteral nutrition plus enteral glu-
stimulation of protein synthesis. However, analy- tamine treatment.19 These results challenge the
ses of the association between protein intake and concept of conditional glutamine deficiency, and
outcome have shown conflicting results.40,43 robust and consistent trial data do not support
Only one randomized, controlled trial, involving routine glutamine supplementation.
50 patients who were being treated with renal- Arginine supplementation in the postoperative
replacement therapy, has studied the effect of in- period may decrease the rate of infectious compli-
creasing the protein dose. In this highly selected cations and the length of stay in the hospital.
population, increased amounts of protein altered However, current evidence does not support its
the calculated nitrogen balance but not the clini- use during critical illness.46
cal outcome.44 Hence, the most effective protein-
to-energy ratio for critically ill patients remains Lipids
unknown. The type of lipid used in nutritional formulations
Glutamine is the most abundant nonessential may affect inflammation. The n–3 fatty acids that
free amino acid. It is synthesized predominant- are present in fish oil have been shown to have
ly in skeletal muscle; low glutamine levels have antiinflammatory effects, n–9 fatty acids that are
been associated with a poor outcome in critical present in olive oil have a more neutral immune
illness. Low glutamine levels were considered effect, and n–6 fatty acids in soybean oil are pro-
to be the consequence of muscle wasting, since inflammatory.47
with the loss of muscle mass, the production of On the basis of low circulating levels of n–3
glutamine may not match increased glutamine fatty acids in patients with acute lung injury and
requirements of immune cells, enterocytes, the proinflammatory properties of n–6 fatty ac-
and hepatocytes. Thus, glutamine was labeled a ids, the lipid profile of nutrients for such pa-
“conditionally essential” amino acid during crit- tients was hypothesized to contribute to the
ical illness, which led to the hypothesis that development or worsening of acute lung injury.
glutamine supplementation would improve out- In three pioneering studies involving a total of
comes. A meta-analysis of the early randomized, 411 patients, a modified enteral-feeding program
controlled trials involving a total of 485 patients with increased ratios of n–3 fatty acids to n–6
suggested that glutamine supplementation might fatty acids resulted in reduced rates of death and
decrease the risk of infection, the length of stay new organ failure, along with reduced durations
in the hospital, and the risk of death.45 A Scan- of stay in the ICU, as compared with a feeding
dinavian trial that was stopped early because of program that was based on lipids present in

244 march 27, 2014 1233
The New
only.ofNo
Medicine
by MIKE MULARCZYK
JanuarySociety.
15, 2015.
AllFor personal
reserved.
corn or canola oil.21 A later trial showed similar tially fatal complications, such as cardiac failure,
clinical effects in patients with sepsis.22 The lactic acidosis, arrhythmia, and respiratory fail-
most recent and largest trial, the OMEGA study, ure.49 As a result, the routine administration of
was stopped prematurely for futility when it intravenous micronutrients is justified during
showed no benefit with the enteral administra- the acute phase of critical illness until full en-
tion of n–3 fatty acids plus antioxidant supple- teral intake is reached.
ments in 272 patients.31 In addition, this study The administration of pharmacologic doses
showed fewer ventilator-free days and longer of trace elements (selenium, copper, manganese,
stays in the ICU among patients in the n–3 zinc, and iron) and vitamins (E, C, and beta caro-
group. Parenteral preparations based on fish oil tene) has been proposed to reduce oxidative cel-
have not been shown to benefit patients in the lular damage and organ failure in critically ill
ICU,48 and the use of olive oil (an n–9 fatty acid), patients. Despite encouraging results from the
as compared with soybean oil, did not affect ei- meta-analysis of early studies,50 an adequately
ther inflammation or outcomes in a trial involv- powered, high-quality, randomized, controlled
ing 100 patients in the ICU.23 Currently, the lack trial showed no such benefit.19 Selenium supple-
of high-quality evidence precludes any recom- mentation may be beneficial in populations in
mendation on the use of specific lipids in criti- which selenium deficiency is prevalent, and the
cally ill patients. potential for benefit is supported by a recent
meta-analysis.20 Thus, it is possible that factors
Sel ec t ion of Micronu t r ien t s such as the dose, the presence or absence of
deficiency before the onset of illness, and the
Micronutrients (consisting of trace elements, vita- type of critical illness may determine the benefit
mins, and electrolytes) are administered to critically of the intervention.20
ill patients to prevent deficiencies and associated
complications. After depletion of micronutrient C onclusions
stores during starvation, the reinitiation of nu-
trition can result in a refeeding syndrome, most Recent methodologically sound and adequately
typically unmasking deficiencies in thiamine, powered randomized, controlled trials have not
potassium, and phosphate that may cause poten- generated unequivocal evidence that feeding pro-
tocols targeting full-replacement nutrition early
in the course of critical illness result in clinical
Table 2. Recommendations for Clinical Nutritional Practice in the ICU and for benefits. The findings of the EPaNIC and EDEN
Future Research. trials raise concern that targeting full-replace-
Recommendations for clinical practice
ment feeding early in critical illness does not
provide benefit and may cause harm in some
Allow hypocaloric enteral feeding in the acute phase of critical illness for up
to 7 days in previously well-nourished patients. populations or settings. The Early Parenteral Nu-
trition and Supplemental Parenteral Nutrition
Note that current evidence does not support glutamine supplementation
early in critical illness. trials suggest that the use of parenteral nutrition
Supply micronutrients to prevent refeeding syndrome.* in itself may not increase the risk of infectious
complications. These new insights limit the
Recommendations for future research
number of nutritional interventions that can be
Investigate mechanisms underlying benefit or harm from the administration
of macronutrients early during critical illness.
confidently recommended for daily critical care
practice (Table 2). In light of the new evidence
Identify biomarkers of the anabolic recovery phase to guide initiation of more
aggressive feeding. from trials that included few if any severely mal-
nourished patients, it seems reasonable to initi-
Validate scoring systems to identify patients who could benefit most from
early nutrition. ate some gastric feeding, while also providing
Identify the potential role of glutamine as part of parenteral nutrition for
micronutrients, once the patient’s condition is
critically ill patients after recovery from acute organ failure. stabilized and to allow hypocaloric macronutri-
ent intake during the first week of critical illness.
* This recommendation is based only on reports of potentially lethal adverse Whether patients with preexisting malnutrition
events from the Centers for Disease Control and Prevention, not on the re-
sults of high-quality, randomized, controlled trials. should be treated differently is uncertain.14,42
It does not appear to be desirable to interfere
1234 245
n engl j med 370;13 nejm.org march 27, 2014
The New
Society. Forrights
reserved.
with the early catabolic response to critical ill- anisms. In addition, research focusing on bio-
ness, either with macronutrients or, as shown markers and on scoring systems should aim to
previously,51 with anabolic hormones. Although identify patients who are able to effectively use
the unfavorable effect of large amounts of macro- macronutrients for recovery and thus are likely
nutrients and growth factors during acute illness to benefit from more aggressive earlier nutrition.
might be explained by their suppressive effects
on pathways of cell-damage removal that recycle No potential conflict of interest relevant to this article was
reported.
substrates from clearing debris,38 more research Disclosure forms provided by the authors are available with
is needed to unravel the exact underlying mech- the full text of this article at NEJM.org.
References
1. Alberda C, Gramlich L, Jones N, et al. 10. Montejo JC, Miñambres E, Bordejé L, ity in patients with sepsis syndrome: a
The relationship between nutritional in- et al. Gastric residual volume during systematic review and meta-analysis of
take and clinical outcomes in critically ill enteral nutrition in ICU patients: the randomized controlled trials. Crit Care
patients: results of an international multi- REGANE study. Intensive Care Med 2010; Med 2013;41:1555-64.
center observational study. Intensive Care 36:1386-93. 21. Pontes-Arruda A, Demichele S, Seth
Med 2009;35:1728-37. [Erratum, Intensive 11. Reignier J, Mercier E, Le Gouge A, et al. A, Singer P. The use of an inflammation-
Care Med 2009;35:1821.] Effect of not monitoring residual gastric modulating diet in patients with acute
2. De Jonghe B, Bastuji-Garin S, Durand volume on risk of ventilator-associated lung injury or acute respiratory distress
MC, et al. Respiratory weakness is associ- pneumonia in adults receiving mechani- syndrome: a meta-analysis of outcome
ated with limb weakness and delayed cal ventilation and early enteral feeding: a data. JPEN J Parenter Enteral Nutr 2008;
weaning in critical illness. Crit Care Med randomized controlled trial. JAMA 2013; 32:596-605.
2007;35:2007-15. 309:249-56. 22. Pontes-Arruda A, Martins LF, de Lima
3. Martindale RG, McClave SA, Vanek 12. Davies AR, Morrison SS, Bailey MJ, et al. SM, et al. Enteral nutrition with eico-
VW, et al. Guidelines for the provision A multicenter, randomized controlled trial sapentaenoic acid, γ-linolenic acid and anti-
and assessment of nutrition support ther- comparing early nasojejunal with nasogas- oxidants in the early treatment of sepsis:
apy in the adult critically ill patient: Soci- tric nutrition in critical illness. Crit Care results from a multicenter, prospective,
ety of Critical Care Medicine and Ameri- Med 2012;40:2342-8. randomized, double-blinded, controlled
can Society for Parenteral and Enteral 13. Acosta-Escribano J, Fernández-Vivas study: the INTERSEPT study. Crit Care
Nutrition: executive summary. Crit Care M, Grau Carmona T, et al. Gastric versus 2011;15:R144.
Med 2009;37:1757-61. transpyloric feeding in severe traumatic 23. Umpierrez GE, Spiegelman R, Zhao V,
4. Singer P, Berger MM, Van den Berghe brain injury: a prospective, randomized et al. A double-blind, randomized clinical
G, et al. ESPEN guidelines on parenteral trial. Intensive Care Med 2010;36:1532-9. trial comparing soybean oil-based versus
nutrition: intensive care. Clin Nutr 2009; 14. Casaer MP, Mesotten D, Hermans G, olive oil-based lipid emulsions in adult
28:387-400. et al. Early versus late parenteral nutrition medical-surgical intensive care unit pa-
5. Martin CM, Doig GS, Heyland DK, in critically ill adults. N Engl J Med 2011; tients requiring parenteral nutrition. Crit
Morrison T, Sibbald WJ. Multicentre, 365:506-17. Care Med 2012;40:1792-8.
cluster-randomized clinical trial of algo- 15. Heidegger CP, Berger MM, Graf S, et 24. Rice TW, Wheeler AP, Thompson BT,
rithms for critical-care enteral and paren- al. Optimisation of energy provision with deBoisblanc BP, Steingrub J, Rock P. En-
teral therapy (ACCEPT). CMAJ 2004;170: supplemental parenteral nutrition in criti- teral omega-3 fatty acid, gamma-linolenic
197-204. cally ill patients: a randomised controlled acid, and antioxidant supplementation in
6. Doig GS, Simpson F, Finfer S, et al. clinical trial. Lancet 2013;381:385-93. acute lung injury. JAMA 2011;306:1574-
Effect of evidence-based feeding guide- 16. Doig GS, Simpson F, Sweetman EA, et 81. [Erratum, JAMA 2012;307:563.]
lines on mortality of critically ill adults: al. Early parenteral nutrition in critically 25. Artinian V, Krayem H, DiGiovine B.
a cluster randomized controlled trial. JAMA ill patients with short-term relative con- Effects of early enteral feeding on the out-
2008;300:2731-41. traindications to early enteral nutrition: come of critically ill mechanically venti-
7. The National Heart, Lung, and Blood a randomized controlled trial. JAMA 2013; lated medical patients. Chest 2006;129:
Institute Acute Respiratory Distress Syn- 309:2130-8. 960-7.
drome (ARDS) Clinical Trials Network. 17. Wernerman J, Kirketeig T, Andersson 26. Doig GS, Heighes PT, Simpson F,
Initial trophic vs full enteral feeding in B, et al. Scandinavian glutamine trial: a Sweetman EA, Davies AR. Early enteral
patients with acute lung injury: the EDEN pragmatic multi-centre randomised clini- nutrition, provided within 24 h of injury
randomized trial. JAMA 2012;307:795- cal trial of intensive care unit patients. or intensive care unit admission, signifi-
803. Acta Anaesthesiol Scand 2011;55:812-8. cantly reduces mortality in critically ill
8. Needham DM, Dinglas VD, Morris PE, 18. Andrews PJ, Avenell A, Noble DW, et patients: a meta-analysis of randomised
et al. Physical and cognitive performance al. Randomised trial of glutamine, sele- controlled trials. Intensive Care Med
of acute lung injury patients one year after nium, or both, to supplement parenteral 2009;35:2018-27.
initial trophic vs full enteral feeding: nutrition for critically ill patients. BMJ 27. Koretz RL. Enteral nutrition: a hard
EDEN trial follow-up. Am J Respir Crit 2011;342:d1542. look at some soft evidence. Nutr Clin
Care Med 2013 June 27 (Epub ahead of 19. Heyland D, Muscedere J, Wischmeyer Pract 2009;24:316-24.
print). PE, et al. A randomized trial of glutamine 28. Peter JV, Moran JL, Phillips-Hughes J.
9. Arabi YM, Tamim HM, Dhar GS, et al. and antioxidants in critically ill patients. A metaanalysis of treatment outcomes of
Permissive underfeeding and intensive in- N Engl J Med 2013;368:1489-97. [Erratum, early enteral versus early parenteral nutri-
sulin therapy in critically ill patients: a N Engl J Med 2013;368:1853.] tion in hospitalized patients. Crit Care
randomized controlled trial. Am J Clin 20. Alhazzani W, Jacobi J, Sindi A, et al. Med 2005;33:213-20.
Nutr 2011;93:569-77. The effect of selenium therapy on mortal- 29. Frankenfield DC, Ashcraft CM. Esti-

246 march 27, 2014 1235
The New
For personal ofNo
use only. Medicine
15, 2015.
AllFor personal
reserved.
mating energy needs in nutrition support O’Keefe GE. The risk for bloodstream in- 44. Scheinkestel CD, Kar L, Marshall K, et
patients. JPEN J Parenter Enteral Nutr fections is associated with increased par- al. Prospective randomized trial to assess
2011;35:563-70. enteral caloric intake in patients receiving caloric and protein needs of critically Ill,
30. Singer P, Anbar R, Cohen J, et al. The parenteral nutrition. Crit Care 2007;11: anuric, ventilated patients requiring con-
Tight Calorie Control Study (TICACOS): a R114. tinuous renal replacement therapy. Nutri-
prospective, randomized, controlled pilot 37. Grau T, Bonet A, Rubio M, et al. Liver tion 2003;19:909-16.
study of nutritional support in critically ill dysfunction associated with artificial nu- 45. Novak F, Heyland DK, Avenell A,
patients. Intensive Care Med 2011;37: trition in critically ill patients. Crit Care Drover JW, Su X. Glutamine supplementa-
601-9. 2007;11:R10. tion in serious illness: a systematic review
31. Taylor SJ, Fettes SB, Jewkes C, Nelson 38. Hermans G, Casaer MP, Clerckx B, et of the evidence. Crit Care Med 2002;30:
RJ. Prospective, randomized, controlled al. Effect of tolerating macronutrient def- 2022-9.
trial to determine the effect of early en- icit on the development of intensive-care 46. Marik PE, Zaloga GP. Immunonutri-
hanced enteral nutrition on clinical out- unit acquired weakness: a subanalysis of tion in critically ill patients: a systematic
come in mechanically ventilated patients the EPaNIC trial. Lancet Respir Med review and analysis of the literature. In-
suffering head injury. Crit Care Med 2013;1:621-9. tensive Care Med 2008;34:1980-90.
1999;27:2525-31. 39. Vanderheyden S, Casaer MP, Kesteloot 47. Wanten GJ, Calder PC. Immune mod-
32. Nguyen NQ, Chapman MJ, Fraser RJ, K, et al. Early versus late parenteral nutri- ulation by parenteral lipid emulsions. Am
Bryant LK, Holloway RH. Erythromycin is tion in ICU patients: cost analysis of the J Clin Nutr 2007;85:1171-84.
more effective than metoclopramide in EPaNIC trial. Crit Care 2012;16:R96. 48. Manzanares W, Dhaliwal R, Jure-
the treatment of feed intolerance in criti- 40. Casaer MP, Wilmer A, Hermans G, witsch B, Stapleton RD, Jeejeebhoy KN,
cal illness. Crit Care Med 2007;35:483-9. Wouters PJ, Mesotten D, Van den Berghe G. Heyland DK. Parenteral fish oil lipid emul-
33. MacLaren R, Kiser TH, Fish DN, Role of disease and macronutrient dose in sions in the critically ill: a systematic review
Wischmeyer PE. Erythromycin vs meto- the randomized controlled EPaNIC trial: and meta-analysis. JPEN J Parenter Enteral
clopramide for facilitating gastric empty- a post hoc analysis. Am J Respir Crit Care Nutr 2013 April 22 (Epub ahead of print).
ing and tolerance to intragastric nutrition Med 2013;187:247-55. 49. Marik PE, Bedigian MK. Refeeding
in critically ill patients. JPEN J Parenter 41. Heidegger CP, Berger MM, Thibault hypophosphatemia in critically ill pa-
Enteral Nutr 2008;32:412-9. R, Zingg W, Pichard C. Supplemental par- tients in an intensive care unit: a prospec-
34. Kutsogiannis J, Alberda C, Gramlich enteral nutrition in critically ill patients tive study. Arch Surg 1996;131:1043-7.
L, et al. Early use of supplemental paren- — authors’ reply. Lancet 2013;381:1716-7. 50. Manzanares W, Dhaliwal R, Jiang X,
teral nutrition in critically ill patients: re- 42. Braunschweig CL, Levy P, Sheean PM, Murch L, Heyland DK. Antioxidant micro-
sults of an international multicenter ob- Wang X. Enteral compared with paren- nutrients in the critically ill: a systematic
servational study. Crit Care Med 2011;39: teral nutrition: a meta-analysis. Am J Clin review and meta-analysis. Crit Care 2012;
2691-9. Nutr 2001;74:534-42. 16(2):R66.
35. Van den Berghe G, Wouters P, 43. Allingstrup MJ, Esmailzadeh N, 51. Takala J, Ruokonen E, Webster NR, et
Weekers F, et al. Intensive insulin thera- Wilkens Knudsen A, et al. Provision of al. Increased mortality associated with
py in critically ill patients. N Engl J Med protein and energy in relation to mea- growth hormone treatment in critically ill
2001;345:1359-67. sured requirements in intensive care pa- adults. N Engl J Med 1999;341:785-92.
36. Dissanaike S, Shelton M, Warner K, tients. Clin Nutr 2012;31:462-8. Copyright © 2014 Massachusetts Medical Society.
clinical trial registration

The Journal requires investigators to register their clinical trials
in a public trials registry. The members of the International Committee
of Medical Journal Editors (ICMJE) will consider most reports of clinical
trials for publication only if the trials have been registered.
Current information on requirements and appropriate registries
is available at www.icmje.org/faq_clinical.html.
1236 n engl j med 370;13 nejm.org march 27, 2014

247
The New England Journal of Medicine
Downloaded from nejm.org by MIKE MULARCZYK on January 15, 2015. For personal use only. No other uses without permission.
Source: The New England Journal of MedicineT h e n e w e ng l a n d j o u r na l of m e dic i n e
c or r e sp ondence
Nutrition in the Acute Phase of Critical Illness

3. Barr J, Fraser GL, Puntillo K, et al. Clinical practice guide-
To the Editor: With respect to the article by
lines for the management of pain, agitation, and delirium in adult
Casaer and Van den Berghe (March 27 issue)1 on patients in the intensive care unit. Crit Care Med 2013;41:263-306.
nutrition in the acute phase of critical illness: 4. Sawh SB, Selvaraj IP, Danga A, Cotton AL, Moss J, Patel PB.
we suggest the use of the peripherally acting Use of methylnaltrexone for the treatment of opioid-induced con-
stipation in critical care patients. Mayo Clin Proc 2012;87:255-9.
μ opiate antagonist methylnaltrexone bromide 5. Hewitt K, Lin H, Faraklas I, Morris S, Cochran A, Saffle J.
(Relistor, Salix Pharmaceuticals) to facilitate en- Use of methylnaltrexone to induce laxation in acutely injured pa-
teral nutrition in patients in the intensive care tients with burns and necrotizing soft-tissue infections. J Burn
Care Res 2014;35(2):e106-e111.
unit (ICU) who are receiving mu opiates for an-
DOI: 10.1056/NEJMc1404896
algesia and sedation.2,3 This agent has been ap-
proved for the treatment of opioid-induced con-
stipation in patients with advanced illness who To the Editor: Body-mass index (BMI, the weight
are receiving palliative care when standard agents in kilograms divided by the square of the height
prove inadequate. However, we and others have in meters) can be used to select patients for nu-
used it in critically ill patients in the burn unit, trition support. In a prospective, observational
the cardiovascular unit, and the medical ICU to study1 of more than 2700 patients who were re-
facilitate upper and lower gastrointestinal motil- ceiving mechanical ventilation, an increase of
ity.4,5 In our own retrospective study,4 six of seven 1000 calories per day reduced mortality and in-
patients without bowel motility for 3 to 5 days creased the number of ventilator-free days; these
had almost instant restoration of bowel function effects were seen only in patients with a BMI of
after the subcutaneous administration of methyl- 35 or more or less than 25. No benefit was seen
naltrexone — 3.5 days before eight patients who in patients with a BMI in the range between 25 and
received conventional therapy. There was also a less than 35. A prospective, randomized trial2
clinically significant effect on gastric residuals. showed that the administration of parenteral nu-
We believe that there is justification for a pro- trition in severely malnourished patients reduced
spective study designed to determine whether noninfectious surgical complications. The article
methylnaltrexone facilitates feeding in these crit- by Casaer and Van den Berghe cites five studies
ically ill patients. with conflicting results. None of these studies
Jonathan Moss, M.D., Ph.D. were stratified according to BMI. I interpret these
University of Chicago data to mean that patients with high or low
Chicago, IL BMIs, as noted above, should receive nutritional
jm47@midway.uchicago.edu
supplementation, whereas those in the middle
Parind Patel, M.B., B.S. range should receive no or minimal nutrition
Imperial College support for the first 7 days of their illness.
London, United Kingdom
Dr. Moss reports receiving consulting fees from Progenics
James M. Cohen, M.D., C.N.S.P.
Pharmaceuticals and Salix Pharmaceuticals and receiving royal- Memorial Regional Hospital
ties through the University of Chicago. No other potential con- Hollywood, FL
flict of interest relevant to this letter was reported. nstconsultants@aol.com
1. Casaer MP, Van den Berghe G. Nutrition in the acute phase ported.
of critical illness. N Engl J Med 2014;370:1227-36.
2. Moss J, Rosow CE. Development of peripheral opioid antag- 1. Alberda C, Gramlich L, Jones N, et al. The relationship be-
onists’ new insights into opioid effects. Mayo Clin Proc 2008; tween nutritional intake and clinical outcomes in critically ill pa-
83:1116-30. tients: results of an international multicenter observational study.
2446 n engl j med 370;25 nejm.org june 19, 2014

248
correspondence
Intensive Care Med 2009;35:1728-37. [Erratum, Intensive Care fining and improving current practice. JPEN J Parenter Enteral
Med 2009;35:1821.] Nutr 2013 August 23 (Epub ahead of print).
2. The Veterans Affairs Total Parenteral Nutrition Cooperative 4. Heyland DK, Cahill N, Day AG. Optimal amount of calories
Study Group. Perioperative total parenteral nutrition in surgical for critically ill patients: depends on how you slice the cake! Crit
patients. N Engl J Med 1991;325:525-32. Care Med 2011;39:2619-26.
DOI: 10.1056/NEJMc1404896 DOI: 10.1056/NEJMc1404896
To the Editor: The suggestion by Caesar and

Van den Berghe to allow hypocaloric enteral The Authors Reply: We agree with Moss and
feeding in critically ill patients for up to 7 days is Patel that the use of opioids for analgesia could
flawed. The Trophic vs. Full-Energy Enteral Nu- contribute to delayed gastric emptying and there-
trition in Mechanically Ventilated Patients with fore to difficulty with enteral feeding in ICU pa-
Acute Lung Injury (EDEN)1 trial and the Impact tients. Assessment of the effect of peripherally
of Early Parenteral Nutrition Completing Enteral acting μ opiate receptor antagonists to improve
Nutrition in Adult Critically Ill Patients (EPaNIC)2 enteral nutrition would require a randomized,
trial showed the iatrogenic effects of nutritional controlled trial involving patients presenting with
supplementation with glucose loading in young, gastrointestinal motility failure while receiving
well-nourished, less sick patients with a critical treatment with opioids.
illness of short duration and nominal caloric The statement by Cohen that early nutritional
debt. It is misguided to infer that patients with support would be indicated in critically ill pa-
high nutritional risk and multiple organ failure tients with a BMI of less than 25 or a BMI of 35
who may need prolonged critical care should be or more was based on associative data but not
allowed to accumulate caloric debt for a week supported by available evidence from high-quality
when more than 50% of patients in the EPaNIC randomized, controlled trials. Indeed, a similar
study were in the ICU for less than 3 days.3 The subgroup of patients (BMI <25 or ≥40) was well
biologic plausibility of benefit from the accumu- represented in the EPaNIC trial (which included
lation of caloric debt was low, given the enroll- 1989 participants with a BMI in these ranges),
ment of patients with low nutritional risk, a factor and the benefit of early parenteral nutrition was
for which neither sample size nor randomization also absent in this large subgroup.1 In addition,
can compensate. Large observational data sets4 the suggestion that patients with severe malnu-
and data from trials with patients at high nutri- trition should receive early parenteral nutrition
tional risk strongly point to benefit from the early during acute critical illness is not based on evi-
provision of protein and energy.3 Nutritional risk dence, since these patients are not well repre-
scores designed for critically ill populations have sented in randomized, controlled trials. The
been developed.4 Further trials evaluating the EPaNIC trial excluded patients with a BMI of
benefits of protein and energy provision are less than 17, the Veterans Affairs study cited by
needed to stratify critically ill patients according Cohen included only 33 severely malnourished
to nutritional risk. patients, and the Early (Parenteral Nutrition) PN
Adam Rahman, M.D. trial enrolled only 46 patients with a BMI of less
than 18.5.2
Western University
London, ON, Canada Rahman and Corbett state that our recommen-
dation to allow hypocaloric enteral feeding for
Chelsea Corbett, B.Sc., R.D.
up to 7 days in the ICU would be flawed, since
Alberta Health Services
Calgary, AB, Canada
the relevant randomized, controlled trials con-
chelsea.corbett@albertahealthservices.ca ducted to date — more specifically, the EPaNIC
No potential conflict of interest relevant to this letter was re- and EDEN trials — would not have included
ported. patients with high nutritional risk or a high se-
1. Casaer MP, Mesotten D, Hermans G, et al. Early versus late verity of illness. They also state that the nutri-
parenteral nutrition in critically ill adults. N Engl J Med 2011;365: tion administered in these trials contained exces-
506-17. sive amounts of glucose and inadequate amounts
2. Rice TW, Wheeler AP, Thompson BT, et al. Initial trophic vs
full enteral feeding in patients with acute lung injury: the EDEN of amino acids. The available evidence disproves
randomized trial. JAMA 2012;307:795-803. these three points. First, in the EPaNIC trial,
3. Rahman A, Martin C, Heyland DK. Nutrition therapy for the early parenteral nutrition failed to improve the
critically ill surgical patient with aortic aneurysmal rupture: de-
n engl j med 370;25

249
nejm.org june 19, 2014 2447
outcome in the preplanned subgroup of 863 pa- Since publication of their article, the authors report no fur-
tients with a very high nutritional risk.3 Second,
the assumption that more severely ill patients 1. Casaer MP, Mesotten D, Hermans G, et al. Early versus late
parenteral nutrition in critically ill adults. N Engl J Med 2011;
would benefit from early enhanced feeding was 365:506-17.
proven wrong; when subgroups were defined 2. Doig GS, Simpson F, Sweetman EA, et al. Early parenteral
according to severity of illness on admission, it nutrition in critically ill patients with short-term relative contra-
indications to early enteral nutrition: a randomized controlled
was clear that early parenteral nutrition caused trial. JAMA 2013;309:2130-8.
the most harm in the most severely ill subgroup, 3. Casaer MP, Wilmer A, Hermans G, Wouters PJ, Mesotten D,
whereas the intervention did not alter the out- Van den Berghe G. Role of disease and macronutrient dose in the
randomized controlled EPaNIC trial: a post hoc analysis. Am J
come in the least severely ill patients.3 In addi- Respir Crit Care Med 2013;187:247-55.
tion, the administration of early parenteral nutri- 4. Hermans G, Casaer MP, Clerckx B, et al. Effect of tolerating
tion aggravated rather than reduced muscle macronutrient deficit on the development of intensive-care unit
acquired weakness: a subanalysis of the EPaNIC trial. Lancet
weakness in the sickest patients requiring pro- Respir Med 2013;1:621-9.
longed intensive care.4 Third, a retrospective 5. Derde S, Vanhorebeek I, Güiza F, et al. Early parenteral nu-
analysis showed that it was the dose of amino trition evokes a phenotype of autophagy deficiency in liver and
skeletal muscle of critically ill rabbits. Endocrinology 2012;153:
acids, not the amount of glucose, that ex- 2267-76.
plained the harm evoked by early parenteral
DOI: 10.1056/NEJMc1404896
nutrition, an observation that is completely in Correspondence Copyright © 2014 Massachusetts Medical Society.
line with the results from a study of experi-
mentally induced critical illness in rabbits.3,5
Michael P. Casaer, M.D., Ph.D.
Greet Van den Berghe, M.D., Ph.D.
KU Leuven University
Leuven, Belgium
greet.vandenberghe@med.kuleuven.be
2448 n engl j med 370;25 250

nejm.org june 19, 2014

What strategy would you use to provide nutrition for this patient?
A. Initiate total parenteral nutrition as soon as possible after the patient’s arrival in the ICU.
B. Await the return of bowel sounds and then initiate enteral nutrition.
C. Initiate enteral nutrition within 24 to 48 hours after ICU admission and then initiate parenteral nutrition on
day 7 if the caloric target is not being met.
D. Initiate parenteral nutrition as soon as possible after the patient’s arrival in the ICU and then start enteral
nutrition once bowel sounds return.

Anorexia, which is part of the acute physiologic response to severe illness, may be adaptive or maladaptive. Studies
in animals and humans have shown a trophic effect of enteral nutrients on the integrity of the gut mucosa, and
this provides the rationale for instituting enteral nutrition early during critical illness. A further rationale comes
from the observation that patients who are fed early through the enteral route have a better outcome than those
who are not. However, an inability to establish early enteral feeding may be a marker of an increased severity of
illness, and as yet there are no large, high-quality trials supporting an outcome benefit from early enteral nutrition
during the acute phase of critical illness versus delayed nutrition.
It is very difficult to achieve full replacement feeding through the enteral route during the first week of critical ill-
ness, particularly in severely ill patients, but the role of parenteral nutrition in patients in whom enteral feeding is
either contraindicated or unsuccessful remains controversial. In the Early Parenteral Nutrition Completing Enter-
al Nutrition in Adult Critically Ill Patients (EPaNIC) trial, investigators found that adding parenteral nutrition
may cause harm,1 whereas the Early Parenteral Nutrition (EPN) trial and the Impact of Supplemental Parenteral
Nutrition on Infection Rate, Duration of Mechanical Ventilation, and Rehabilitation in ICU Patients (SPN) trial
did not suggest harm but also did not provide robust evidence that early initiation of parenteral nutrition was
beneficial.2,3
The patient described in the vignette does not appear to be malnourished. He has septic shock caused by fecal
peritonitis and has had an end colostomy formed. In the absence of gross abdominal distention, it may be reason-
able to commence trophic4 enteral nutrition through a nasogastric tube and to withhold parenteral nutrition,
thereby allowing hypocaloric macronutrient intake during the first week. In the event that full enteral nutrition
cannot be established within the first week, consideration should be given to adding supplementary parenteral nu-
trition. However, the most effective time at which the initiation of parenteral nutrition can produce a clear clinical
benefit during critical illness remains unclear, and if a prolonged delay in establishing enteral nutrition seems
likely or if a patient is not well-nourished at presentation, earlier institution of parenteral nutrition can be
justified.
REFERENCES
1. Casaer MP, Mesotten D, Hermans G, et al. Early versus late parenteral nutrition in critically ill adults. N Engl
J Med 2011;365:506-17.
251
2. Doig GS, Simpson F, Sweetman EA, et al. Early parenteral nutrition in critically ill patients with short-term relative
contraindications to early enteral nutrition: a randomized controlled trial. JAMA 2013;309:2130-8.
3. Heidegger CP, Berger MM, Graf S, et al. Optimisation of energy provision with supplemental parenteral nutrition
in critically ill patients: a randomised controlled clinical trial. Lancet 2013;381:385-93.
4. National Heart, Lung, and Blood Institute Acute Respiratory Distress Syndrome (ARDS) Clinical Trials Network.
Initial trophic vs full enteral feeding in patients with acute lung injury: the EDEN randomized trial. JAMA
2012;307:795-803.
252
new england
The
journal of medicine
ORIGINAL ARTICLE
established in 1812 october 30, 2014 vol. 371 no. 18
Trial of the Route of Early Nutritional Support

in Critically Ill Adults
Sheila E. Harvey, Ph.D., Francesca Parrott, M.Sci., David A. Harrison, Ph.D., Danielle E. Bear, M.Res.,
Ella Segaran, M.Sc., Richard Beale, M.B., B.S., Geoff Bellingan, M.D., Richard Leonard, M.B., B.Chir.,
Michael G. Mythen, M.D., and Kathryn M. Rowan, Ph.D., for the CALORIES Trial Investigators*
A bs t r ac t
Background
Uncertainty exists about the most effective route for delivery of early nutritional From the Clinical Trials Unit, Intensive
Care National Audit and Research Centre
support in critically ill adults. We hypothesized that delivery through the paren- (S.E.H., F.P., D.A.H., K.M.R.), the Depart-
teral route is superior to that through the enteral route. ments of Nutrition and Dietetics (D.E.B.)
and Adult Critical Care (D.E.B., R.B.),
Methods Guy’s and St. Thomas’ NHS Foundation
Trust, the Department of Intensive Care,
We conducted a pragmatic, randomized trial involving adults with an unplanned Imperial College Healthcare NHS Trust
admission to one of 33 English intensive care units. We randomly assigned patients (E.S., R.L.), the Division of Asthma, Allergy
and Lung Biology, King’s College London
who could be fed through either the parenteral or the enteral route to a delivery (R.B.), National Institute for Health Re-
route, with nutritional support initiated within 36 hours after admission and con- search Biomedical Research Centre, Uni-
tinued for up to 5 days. The primary outcome was all-cause mortality at 30 days. versity College London Hospitals NHS
Foundation Trust and University College
London (G.B., M.G.M.), and the Depart-
Results ment of Surgery and Cancer, Imperial
We enrolled 2400 patients; 2388 (99.5%) were included in the analysis (1191 in the College London (R.L.) — all in London.
Address reprint requests to Dr. Rowan at
parenteral group and 1197 in the enteral group). By 30 days, 393 of 1188 patients the Intensive Care National Audit and
(33.1%) in the parenteral group and 409 of 1195 patients (34.2%) in the enteral Research Centre, Napier House, 24 High
group had died (relative risk in parenteral group, 0.97; 95% confidence interval, Holborn, London WC1V 6AZ, United King-
dom, or at kathy.rowan@icnarc.org.
0.86 to 1.08; P = 0.57). There were significant reductions in the parenteral group, as
compared with the enteral group, in rates of hypoglycemia (44 patients [3.7%] vs. * A complete list of the investigators and
74 patients [6.2%]; P = 0.006) and vomiting (100 patients [8.4%] vs. 194 patients committee members in the CALORIES
trial is provided in the Supplementary
[16.2%]; P<0.001). There were no significant differences between the parenteral Appendix, available at NEJM.org.
group and the enteral group in the mean number of treated infectious complica-
This article was published on October 1,
tions (0.22 vs. 0.21; P = 0.72), 90-day mortality (442 of 1184 patients [37.3%] vs. 464 2014, at NEJM.org.
of 1188 patients [39.1%], P = 0.40), in rates of 14 other secondary outcomes, or in
N Engl J Med 2014;371:1673-84.
rates of adverse events. Caloric intake was similar in the two groups, with the target DOI: 10.1056/NEJMoa1409860
intake not achieved in most patients. Copyright © 2014 Massachusetts Medical Society.
Conclusions
We found no significant difference in 30-day mortality associated with the route of
delivery of early nutritional support in critically ill adults. (Funded by the United
Kingdom National Institute for Health Research; CALORIES Current Controlled
Trials number, ISRCTN17386141.)
n engl j med 371;18 nejm.org october 30, 2014 1673
253
N
utritional support is standard Sites and Patients
for critically ill patients and requires a The study was conducted in 33 adult general
complex calculus of timing, route of de- ICUs in England participating in the national
livery, and the amount and type of nutrients that clinical audit for adult critical care coordinated
are administered — all of which may affect pa- by ICNARC.12 Patients who were at least 18 years
tient outcomes. The interpretation of published of age were eligible if they were expected to re-
meta-analyses of trials comparing nutritional quire nutritional support for at least 2 days, as
support through the parenteral route versus the determined by a clinician within 36 hours after
enteral route in critically ill patients1-3 is compli- an unplanned ICU admission that was expected
cated by small sample sizes, variable quality, se- to last at least 3 days. Patients were excluded if
lection bias, lack of standardized definitions, they could not be fed through either the paren-
and interventions that combine multiple ele- teral or the enteral route, had received nutritional
ments of nutritional support (e.g., timing and support in the past 7 days, had a gastrostomy or
route). Currently, the enteral route is the main- jejunostomy in situ, were pregnant, or were not
stay, largely on the grounds of physiological ra- expected to be in the United Kingdom for the
tionale and modest evidence suggesting an as- next 6 months. (A detailed list of inclusion and
sociation with fewer infections,2,4,5 yet it can also exclusion criteria is provided in the Supplemen-
be associated with gastrointestinal intolerance tary Appendix.) All patients or their consultees
and underfeeding.6,7 The parenteral route, provided written informed consent or agreement
though more invasive, more often secures deliv- according to the provisions of the United King-
ery of the intended nutrition6 but has been asso- dom Mental Capacity Act of 2005.
ciated with greater risks and rates of complica- Using a 24-hour telephone randomization sys-
tions.1-3 However, these studies have not tem, we assigned patients in a 1:1 ratio to receive
considered improvements in delivery, formula- early nutritional support through the parenteral
tion, and monitoring of parenteral nutrition.8,9 route or the enteral route. We used a computer-
Although recent studies have evaluated sup- ized minimization algorithm with a random
plemental parenteral nutrition,10,11 the most effec- component to balance patients according to ICU,
tive route for early nutritional support in criti- age (<65 years or ≥65 years), surgical status (sur-
cally ill patients is unknown. In the CALORIES gery <24 hours before ICU admission or no sur-
trial, we tested the hypothesis that the paren- gery <24 hours before ICU admission), and the
teral route is superior to the enteral route for the presence or absence of severe malnutrition.
delivery of early nutritional support in adults
who had an unplanned admission to an intensive Study Interventions
care unit (ICU) and who could be fed through Nutritional support was initiated as soon as pos-
either route. sible after randomization (within 36 hours after
admission) and used exclusively for 5 days (120
Me thods hours) or until transition to exclusive oral feed-
ing, discharge from the ICU, or death (termed
Study Design and Oversight the intervention period). Oral feeding could be
Our study was a pragmatic, open, multicenter, initiated if clinically indicated during the inter-
parallel-group, randomized, controlled trial. The vention period. Patients in the parenteral group
North West London Research Ethics Committee received nutrition through a central venous cath-
approved the study protocol, which is available eter with a dedicated lumen positioned in accor-
with the full text of this article at NEJM.org. The dance with National Health Service guidelines.13
United Kingdom National Institute for Health Patients in the enteral group received nutrition
Research funded the study and convened a trial through a nasogastric or nasojejunal tube posi-
steering committee and independent data and tioned in accordance with U.K. guidelines.14,15
safety monitoring committee. The Clinical Trials Energy targets were set at 25 kcal per kilo-
Unit at the U.K. Intensive Care National Audit gram of actual body weight per day, with a goal
and Research Centre (ICNARC) managed the of reaching the target within 48 to 72 hours.
study (for details, see the Supplementary Appen- Protein or amino acid targets were set according
dix, available at NEJM.org). to local practice. Glycemic control was main-
1674 n engl j med 371;18 254

Early Nutritional Support in Critically Ill Adults
tained in accordance with international guide- We performed all statistical analyses using
lines (with a target level for serum glucose of the intention-to-treat principle on the basis of a
<180 mg per deciliter [10 mmol per liter]).16 prespecified statistical analysis plan.19 A P value
Calories from non-nutritional sources (e.g., proof 0.05 was considered to indicate statistical
pofol) were included in the calculations of total significance. All tests were two-sided, and there
calories. All other treatments and nutritional was no adjustment for multiple variables. Con-
support were provided according to local prac- tinuous variables are reported as means and
tice guidelines and at the clinicians’ discretion. standard deviations or as medians and inter-
(Details about the study interventions are pro- quartile ranges. Categorical variables are reported
vided in the Supplementary Appendix.) as proportions. The time to the initiation of ex-
clusive oral feeding was analyzed as the time to
Outcome Measures event, with censoring of data for all patients
We report the evaluation of clinical effectiveness, who died while receiving nutritional support.
including the primary outcome at 30 days and all We used Fisher’s exact test to compare between-
secondary outcomes within 90 days after ran- group differences in the primary outcome. Ab-
domization. The primary outcome was all-cause solute and relative risks are reported with 95%
mortality at 30 days. Secondary outcomes were confidence intervals without adjustment. Re-
the duration of organ support, treated infectious ported as a secondary analysis is the adjusted
and noninfectious complications, length of stay odds ratio from multilevel logistic regression
in the ICU and hospital, the duration of survival, after adjustment for age, ICNARC Physiology
and mortality at the time of discharge from the Score,20 surgical status, degree of malnutrition,
ICU and from the hospital, at 90 days, and at and a site-level random effect. A sensitivity ap-
1 year. Adverse events were monitored for 30 days. proach was taken for missing data with respect
(Definitions for all outcomes are provided in the to the primary outcome by assuming that all
Supplementary Appendix.) patients with missing outcomes survived in the
enteral group and died in the parenteral group,
Statistical Analysis with the analyses then repeated with the oppo-
We assumed a baseline 30-day mortality of 32% site assumptions. Missing baseline data were
for patients receiving nutrition through the en- imputed in adjusted analyses with the use of
teral route. On the basis of our updated meta- multivariate imputation by means of chained
analysis, we estimated that the patients in the equations.21
parenteral group, as compared with those in the For secondary outcomes, we used Fisher’s
enteral group, would have a potential relative risk exact test to analyze binary outcomes, t-tests to
reduction of approximately 20% in the primary analyze the number of infectious complications
outcome (Fig. S1 in the Supplementary Appen- and duration of organ support (with bootstrap-
dix). On the basis of these estimates, we deter- ping for anticipated non-normality in the latter22),
mined that an enrollment of 2400 patients would and Wilcoxon rank-sum tests to analyze the
have a power of 90% to detect a 20% relative risk length of stay, stratified according to survival.
reduction (absolute risk reduction, 6.4 percent- Unadjusted relative risks and adjusted odds ra-
age points) in the parenteral group with a two- tios are reported for all mortality outcomes. We
sided alpha level of 0.05, assuming that 2% of used the log-rank test to compare Kaplan–Meier
patients would cross over to the other group or curves for 90 days with no adjustment and a Cox
have a protocol violation and that 2% of patients proportional-hazards model to compare survival
would be lost to follow-up or withdraw from the with adjustment.
study.17 The likelihood-ratio test was used to assess
A single, planned interim analysis was per- interactions between groups and prespecified
formed and was reviewed by the data and safety subgroups in adjusted multilevel logistic-regres-
monitoring committee at the point when 30-day sion models. Subgroups were defined according
outcomes for the first 1200 patients were avail- to age quartiles, the presence or absence of mal-
able. A Haybittle–Peto stopping rule (P<0.001) nutrition, quartiles of predicted risk of death,20,23
was used to guide recommendations for early the presence or absence of mechanical ventila-
termination.18 tion, the presence or absence of cancer, and the

255 october 30, 2014 1675
11,108 Patients met inclusion criteria

2025 Had contraindication
1648 Received nutritional support within past
7 days
1368 Were in ICU >36 hr
504 Were receiving palliative care
428 Were admitted with PEG, PEJ, or needle
or surgical jejunostomy in situ
99 Were expected to stay in UK <6 mo
84 Were previously enrolled in CALORIES
34 Were known to be pregnant
5 Had burns
716 Were excluded by clinician
252 Did not have research staff available
209 Were participating in another study
75 Did not have parenteral or enteral
product available
25 Had site error
64 Had no reason given
1200 Were assigned to the parenteral group 1200 Were assigned to the enteral group
9 Requested removal 3 Requested removal

of all data of all data
1155 Received assigned nutritional support 1167 Received assigned nutritional support
36 Did not receive assigned nutritional 30 Did not receive assigned nutritional
support support
24 Received no nutritional support 26 Received no nutritional support
12 Received nutritional support through 4 Received nutritional support through
the enteral route the parenteral route
3 Were lost to follow-up 2 Were lost to follow-up

before 30 days before 30 days
1188 Were included in the primary 1195 Were included in the primary
outcome analysis outcome analysis
Figure 1. Screening, Randomization, and Follow-up.

Patients who requested removal of all data or were lost to follow-up (12 in the parenteral group and 5 in the enteral
group) were not included in the primary outcome analysis. ICU denotes intensive care unit, PEG percutaneous endo-
scopic gastrostomy, and PEJ percutaneous endoscopic jejunostomy.
time from ICU admission to the initiation of model24 with an instrumental variable for the
nutritional support (<24 hours or ≥24 hours). study group, assuming a linear relationship be-
We repeated the primary analysis after adjust- tween the degree of adherence (proportion of
ment for adherence using a structural-mean the intervention period that the assigned route

256 october 30, 2014
was used) and treatment effect to estimate the Adverse Events

efficacy of early nutritional support delivered One or more serious adverse events were report-
through the parenteral route, as compared with ed in 58 patients (4.9%) in the parenteral group
the enteral route.25 All analyses were performed and 58 patients (4.8%) in the enteral group
with the use of Stata/SE software, version 13.0. (P = 1.00) (Table 2, and Table S7 in the Supple-
mentary Appendix). There were five serious, un-
R e sult s expected adverse events that were deemed by the
site investigator to be potentially related to the
Patients study treatment in 4 patients (1 with ischemic
From June 22, 2011, to March 2, 2014, we bowel and hypoglycemia and 1 each with upper
screened 11,108 patients at 33 sites in England gastrointestinal hemorrhage and anterolateral
(Table S1 in the Supplementary Appendix). Of myocardial infarction in the parenteral group
these patients, 2400 were enrolled, including 12 and 1 with lower gastrointestinal hemorrhage in
patients who subsequently withdrew from the the enteral group).
study (Table S2 in the Supplementary Appendix),
which resulted in an intention-to-treat popula- Primary Outcome
tion of 2388 patients (1191 in the parenteral By 30 days, 393 of 1188 patients (33.1%) in the
group and 1197 in the enteral group) (Fig. 1). parenteral group and 409 of 1195 patients
Baseline characteristics of the patients were sim- (34.2%) in the enteral group had died, with no
ilar in the two study groups (Table 1, and Table significant between-group difference, even after
S3 in the Supplementary Appendix). adjustment for baseline variables (relative risk in
the parenteral group, 0.97; 95% confidence inter-
Adherence and Nutritional Support val [CI], 0.86 to 1.08; absolute risk reduction,
Overall, 97% of the patients received early nutri- 1.15; 95% CI, −2.65 to 4.94; P = 0.57) (Table 3).
tional support through the assigned route, and The results were similar after the inclusion of
rates of nonadherence to the protocol were simi- 5 patients with missing data for the 30-day out-
lar in the two groups (Table 2). Initiation was come (relative risks, 0.96 and 0.97 after the ap-
delayed for 37 patients (3.1%) in the parenteral plication of extreme assumptions).
group and 41 (3.4%) in the enteral group. Cross-
over occurred in 81 patients (6.8%) in the paren- Secondary Outcomes
teral group and 18 patients (1.5%) in the enteral There were significant reductions in the paren-
group during the intervention period; most of teral group, as compared with the enteral group,
the crossover occurred toward the end of the 120 in rates of hypoglycemia (44 patients [3.7%] vs.
hours. Patients in the enteral group were more 74 patients [6.2%]; absolute risk reduction, 2.5 per-
likely than those in the parenteral group to have centage points; 95% CI, 0.8 to 4.2; P = 0.006) and
complete days without nutrition (Table S4 and vomiting (100 patients [8.4%] vs. 194 patients
Fig. S2 in the Supplementary Appendix). [16.2%]; absolute risk reduction, 7.8 percentage
Nutritional support was initiated early and is points; 95% CI, 5.2 to 10.4; P<0.001). However,
summarized in Table 2 and in Tables S5 and S6 there were no significant differences between the
in the Supplementary Appendix. In the enteral parenteral group and the enteral group for the 16
group, the mean gastric residual volume (253 other secondary outcomes, including the mean
ml per 24 hours, with cutoff volumes of 200 to number of infectious complications (0.22 vs.
500 ml in local protocols) was higher and more 0.21; difference, 0.01; 95% CI, −0.04 to 0.06;
patients received prokinetic agents than in the P = 0.72) and 90-day mortality (442 of 1184 pa-
parenteral group. Scores on the Sequential Or- tients [37.3%] vs. 464 of 1188 patients [39.1%];
gan Failure Assessment (SOFA)26 and amounts relative risk, 0.96; 95% CI, 0.86 to 1.06; P = 0.40)
of calories and proteins or amino acids are pro- (Table 3, and Table S8 in the Supplementary Ap-
vided in Figure 2. The target nutritional value of pendix). There was no significant difference in
25 kcal per kilogram per day was not achieved the duration of survival up to 90 days (P = 0.98 by
for the majority of patients in the two study the log-rank test; adjusted hazard ratio, 0.94;
groups, although caloric intake was similar in the 95% CI, 0.82 to 1.07, P = 0.33) (Fig. S3 in the Sup-
two groups. plementary Appendix).
n engl j med 371;18 257

Parenteral Group Enteral Group

Age — yr 63.3±15.1 62.9±15.4
Male sex — no. (%) 689 (57.9) 725 (60.6)
Severe coexisting illness — no./total no. (%)
Liver 29/1181 (2.5) 34/1193 (2.8)
Renal 20/1181 (1.7) 15/1193 (1.3)
Respiratory 34/1181 (2.9) 23/1193 (1.9)
Cardiovascular 11/1181 (0.9) 14/1193 (1.2)
Immunodeficiency 78/1181 (6.6) 95/1193 (8.0)
Surgery <24 hr before ICU admission — no. (%)† 162 (13.6) 167 (14.0)
APACHE II‡
Acute Physiology Score 15.1±6.2 15.2±6.2
Total score 19.6±6.9 19.6±7.0
Median predicted risk of death (IQR)§ 0.34 (0.18–0.52) 0.34 (0.18–0.52)
ICNARC
Physiology Score¶ 25.6±8.0 25.8±7.8
Median predicted risk of death (IQR)‖ 0.42 (0.22–0.65) 0.43 (0.23–0.65)
Mechanical ventilation — no./total no. (%) 979/1178 (83.1) 993/1185 (83.8)
SOFA score** 9.5±3.4 9.6±3.3
Subjective assessment of severe malnutrition — no. (%)† 151 (12.7) 152 (12.7)
Actual or estimated body-mass index†† 27.7±7.4 28.2±7.5
Degree of malnutrition — no./total no. (%)‡‡
High 74/1152 (6.4) 81/1161 (7.0)
Moderate 8/1152 (0.7) 10/1161 (0.9)
None 1070/1152 (92.9) 1070/1161 (92.2)
* Plus–minus values are means ±SD. There were no significant differences between the two groups. ICU denotes inten-
sive care unit, and IQR interquartile range.
† This characteristic was included in the minimization algorithm.
‡ On the Acute Physiology and Chronic Health Evaluation II (APACHE II), the Acute Physiology Score, which is based
on data regarding physiological function that were obtained during the first 24 hours after admission to the ICU,
ranges from 0 to 60, with higher scores indicating greater severity of illness. The total score, which is based on acute
physiology, age, and severe coexisting illnesses, ranges from 0 to 71, with higher scores indicating greater severity of
illness.
§ This value is the predicted risk of death before discharge from an acute care hospital in the United Kingdom on the
basis of a 2013 recalibration. There were insufficient data to calculate the predicted risk of death for 29 patients in the
parenteral group and 24 in the enteral group.
¶ Scores for physiological function on the Intensive Care National Audit and Research Centre (ICNARC) model range
from 0 to 100, with higher scores indicating a greater severity of illness. This score was based on data regarding phys-
iological function that were obtained during the first 24 hours after admission to the ICU.
‖ There were insufficient data to calculate the predicted risk of death for one patient in the parenteral group.
** Scores on the Sequential Organ Failure Assessment (SOFA) range from 0 to 24, with higher scores indicating a great-
er degree of organ failure. The SOFA score was calculated with the use of data obtained within 24 hours before ran-
domization.
†† The body-mass index (BMI) is the weight in kilograms divided by the square of the height in meters. This value was
based on estimated weight or height for 1552 patients (780 in the parenteral group and 772 in the enteral group) and
was not available for 24 patients (14 in the parenteral group and 10 in the enteral group).
‡‡ A high degree of malnutrition was defined as a BMI of less than 18.5 or a weight loss of more than 10% over the pre-
vious 6 months, and a moderate degree of malnutrition was defined as a BMI of less than 20 and a weight loss of
more than 5%.

258 october 30, 2014
Table 2. Nonadherence, Clinical Management, and Serious Adverse Events.*
Parenteral Group Enteral Group

Variable (N = 1191) (N = 1197)
Any nonadherence to delivery of nutritional support during intervention
period — no. (%) 150 (12.6) 127 (10.6)
Median time from ICU admission to initiation of early nutritional support 24 (17–30) 22 (16–28)
(IQR) — hr
Total calories received during intervention period — kcal/kg 89±44 74±44
Total protein received during intervention period — g/kg 3±2 3±2
Gastric residual volume — ml†
Total aspirated during intervention period 35±265 958±1312
Total replaced during intervention period 24±170 618±863
Patients receiving prokinetic drug during intervention period — no. (%)† 26 (2.2) 426 (35.6)
Blood glucose during intervention period — mg/dl
Daily lowest 127±25 118±26
Daily highest 183±43 181±45
Patients receiving insulin during intervention period — no./total no. (%) 694/1184 (58.6) 668/1191 (56.1)
Patients receiving vasoactive agents during intervention period — 958/1184 (80.9) 1007/1191 (84.6)
no./total no. (%)
Median no. of days from randomization to initiation of exclusive oral 14 (5–36) 13 (5–32)
feeding (IQR)
Serious adverse events — no. (%)‡
Any 58 (4.9) 58 (4.8)
Specified§
Abdominal distention 1 (0.1) 2 (0.2)
Electrolyte disturbance 8 (0.7) 5 (0.4)
Hypoglycemic episode 5 (0.4) 3 (0.3)
Ischemic bowel 8 (0.7) 11 (0.9)
Jaundice 1 (0.1) 1 (0.1)
Pneumothorax 1 (0.1) 1 (0.1)
Elevated liver enzymes 3 (0.3) 7 (0.6)
Regurgitation or aspiration 2 (0.2) 4 (0.3)
Vomiting 1 (0.1) 1 (0.1)
Unspecified¶ 39 (3.3) 30 (2.5)
* Plus–minus values are means ±SD. Data completeness for nutritional support ranged from 97 to 100%. The interven-
tion period extended from the initiation of nutritional support to 120 hours or until transition to exclusive oral feeding,
discharge from the ICU, or death.
† Data in this category were recorded only for patients receiving nutritional support through the enteral route, including
patients who crossed over from the parenteral group.
‡ Adverse events were assessed as serious if they involved prolonging of hospitalization, resulted in persistent or sub-
stantial disability or incapacity, or were life-threatening or fatal. P = 1.00 for the between-group difference.
§ The following specified serious adverse events did not occur in either study group: abdominal pain, hemopneumotho-
rax, hepatomegaly, hyperosmolar syndrome, hypersensitivity reaction (anaphylactic reaction), nausea requiring treat-
ment, or vascular catheter-related infection.
¶ A list of individual unspecified serious adverse events is provided in Table S7 in the Supplementary Appendix.
Subgroup and Secondary Analyses (Fig. S4 in the Supplementary Appendix). The re-
There were no significant interactions between sults were similar after adjustment for nonadher-
study group and any prespecified subgroup with ence (relative risk of death at 30 days, 0.96; 95%
respect to 30-day mortality (P = 0.15 to P = 0.83) CI, 0.85 to 1.09; P = 0.55).

259 october 30, 2014 1679
A SOFA Score B Protein Intake

Parenteral route Enteral route Parenteral route Enteral route
20 2.0
15 1.5
Protein (g per kg)

Score
10 1.0
5 0.5
0 0.0
Baseline 1 2 3 4 5 6 1 2 3 4 5 6
Days from Initiation of Early Nutritional Support Days from Initiation of Early Nutritional Support
C Caloric Intake D Caloric Target Met

Parenteral route Enteral route Parenteral route Enteral route
Parenteral route adjusted Enteral route adjusted
40 40
30 30
Calories (kcal per kg)
Patients (%)
20 20
10 10
0 0
1 2 3 4 5 6 1 2 3 4 5 6
Days from Initiation of Early Nutritional Support Days from Initiation of Early Nutritional Support
Figure 2. Daily SOFA Score and Protein and Caloric Intake from Days 1 to 6.
Panel A shows the Sequential Organ Failure Assessment (SOFA) scores for the 24 hours before randomization (baseline) and from day 1
to day 6. SOFA scores range from 0 to 24, with higher scores indicating a greater degree of organ failure. Also shown are the total
amounts of protein (Panel B) and calories (Panel C) received per kilogram of actual or estimated body weight. Panel D shows the per-
centage of patients in whom the target of 25 kcal per kilogram of body weight per day was met. The dark-colored bars show the percent-
age of patients in whom the target was met after adjustment for partial days of feeding (e.g., after the death of a patient). Day 1 data are
the values from the time of initiation to the end of the day of initiation of nutritional support. In Panels A to C, the horizontal lines within
the boxes indicate medians, the lower and upper ends of the boxes the 25th and 75th percentiles, respectively, and the I bars the 1st and
99th percentiles.
Discussion age, the degree of existing malnutrition, the se-

verity of illness, or the timing of the initiation of
Our study showed that among adults with an un- nutritional support. The enteral route was associ-
planned ICU admission for whom early nutri- ated with significantly more episodes of hypogly-
tional support could be provided through either cemia and vomiting, but there were no signifi-
the parenteral or the enteral route, there was no cant between-group differences in the duration
significant difference in mortality at 30 days ac- of organ support, the number of infectious com-
cording to the route of delivery. In addition, there plications, the length of stay in the ICU or total
was no significant interaction on the basis of hospital stay, or the duration of survival up to
1680 n engl j med 371;18

260 nejm.org october 30, 2014
Table 3. Primary and Secondary Outcomes.*
Absolute Difference
Parenteral Group Enteral Group between Groups Relative Risk
Outcome (N = 1191) (N = 1197) (95% CI) (95% CI) P Value
Primary outcome: death within 30 days 393/1188 (33.1) 409/1195 (34.2) 1.15 (−2.65 to 4.94)† 0.97 (0.86 to 1.08)‡ 0.57§
— no./total no. (%)
Secondary outcomes
No. of days alive and free of
specified organ support
up to 30 days¶
Free of advanced respiratory 14.3±12.1 14.3±12.2 0.04 (−0.94 to 1.01) 0.94
support
Free of advanced cardiovascular 18.9±13.5 18.5±13.6 0.41 (−0.63 to 1.53) 0.44
support
Free of renal support 19.1±13.9 18.8±14.0 0.26 (−0.85 to 1.47) 0.66
Free of neurologic support 19.2±13.8 18.9±14.0 0.34 (−0.81 to 1.36) 0.57
Free of gastrointestinal support 13.0±11.7 13.2±11.8 −0.12 (−1.05 to 0.80) 0.81
No. of treated infectious complica- 0.22±0.60 0.21±0.56 0.01 (−0.04 to 0.06) 0.72
tions per patient‖
Noninfectious complications —
no./total no. (%)
Episodes of hypoglycemia 44/1191 (3.7)** 74/1197 (6.2)†† 2.49 (0.75 to 4.22)† 0.006§
Elevated liver enzymes 212/1191 (17.8) 179/1197 (15.0) −2.85 (−5.81 to 0.12)† 0.07§
Nausea requiring treatment 44/1191 (3.7) 53/1197 (4.4) 0.73 (−0.85 to 2.32)† 0.41§
Abdominal distention 78/1191 (6.5) 99/1197 (8.3) 1.72 (−0.38 to 3.82)† 0.12§
Vomiting 100/1191 (8.4) 194/1197 (16.2) 7.81 (5.20 to 10.43)† <0.001§
New or substantially worsened 181/1190 (15.2) 179/1195 (15.0) −0.23 (−3.10 to 2.64)† 0.91§
pressure ulcers
Median no. of days in the ICU 8.1 (4.0–15.8) 7.3 (3.9–14.3) 0.15
(IQR)‡‡
Median no. of days in acute care 17 (8–34) 16 (8–33) 0.32
hospital (IQR)§§
Death — no./total no. (%)¶¶
In the ICU 317/1190 (26.6) 352/1197 (29.4) 0.91 (0.80 to 1.03) 0.13§
In acute care hospital 431/1185 (36.4) 450/1186 (37.9) 0.96 (0.86 to 1.06) 0.44§
By 90 days 442/1184 (37.3) 464/1188 (39.1) 0.96 (0.86 to 1.06) 0.40§

† This value is the absolute risk reduction between event rates.
‡ The adjusted odds ratio from multiple logistic regression was 0.95 (95% CI, 0.79 to 1.13; P = 0.55).
§ This P value was calculated with the use of Fisher’s exact test.
¶ Data on the number of days alive and free of organ support were not available for 5 patients in the parenteral group and 2 in the enteral group.
‖ Infectious complications in 224 of 262 patients (85.5%) in the parenteral group and in 215 of 253 patients (85.0%) in the enteral group
were confirmed on laboratory testing.
** A total of 25 these episodes occurred during the first 6 days, and the mean (±SD) of the lowest blood glucose levels was 57±15 mg per
deciliter (3.2±0.8 mmol per liter).
†† A total of 48 of these episodes occurred during the first 6 days, and the mean of the lowest blood glucose levels was 54±15 mg per deciliter
(3.0±0.8 mmol per liter).
‡‡ The number of days in the ICU was not available for 1 patient in the parenteral group.
§§ The numbers of days in an acute care hospital were not available for 6 patients in the parenteral group and 11 in the enteral group.
¶¶ The adjusted odds ratios from multiple logistic regression were 0.86 (95% CI, 0.71 to 1.04; P = 0.12) for death in the ICU, 0.93 (95% CI,
0.78 to 1.11; P = 0.43) for death in an acute care hospital, and 0.93 (95% CI, 0.77 to 1.11; P = 0.39) for death by 90 days.
n engl j med 371;18 nejm.org october

261 30, 2014 1681
90 days. The target delivery of 25 kcal per kilo- ment, the results are generalizable only to the
gram per day was not reached in a majority of the specific population that we studied. It was rigor-
patients in each study group. ously conducted, with the study groups well
There is debate about both the route of deliv- balanced at baseline and early initiation of nu-
ery and the dose of clinically supplied nutrition, tritional support, as intended. Blinding was
and the aim of our study was to address solely deemed to be impractical and, although the
the question of route. Set in a real-world context, primary outcome was objective, some of the
our study had two major findings. First, there secondary outcomes, though defined and objec-
was no significant difference in outcome be- tively assessed, may have been more vulnerable
tween the two study groups. Although there was to observer bias. We selected an objective, docu-
a trend toward more gastrointestinal side effects mented clinical definition of a new infection
with the enteral route, the reported increase in — one that was laboratory-confirmed or for
infectious complications that have been associ- which there was sufficient conviction to treat the
ated with the parenteral route was not observed. patient. Though the measurement of the residu-
Possible contributory reasons are improvements al gastric volume has recently been questioned,29
in current management of vascular access27 and the overall feeding performance through the
prevention of ventilator-associated pneumonia,28 enteral route in our study was similar to their
as well as developments in feeding technology. findings, and the rates of vomiting and infec-
Second, there was no significant difference in tious complications in our study were lower.
effective nutritional delivery, since patients in the How do our findings compare with those in
two groups did not receive the caloric targets. other recent trials on nutritional support in the
Although these findings are consistent with critically ill? In the Early Parenteral Nutrition
those of previous trials evaluating the enteral Completing Enteral Nutrition in Adult Critically
route in critically ill patients,6,7 the presump- Ill Patients (EPaNIC) trial,10 which was con-
tion is that the parenteral route is more reliable ducted in two hospitals (seven units) recruiting
in guaranteeing delivery.10,11 Possible contribu- patients who required parenteral nutrition and
tory reasons as to why the parenteral route did using tight glycemic control, investigators found
not meet its caloric target include lack of avail- an association between supplemental parenteral
ability of nutritional product, content (the use of nutrition delivered within 48 hours after presen-
commercially available rather than individually tation and an increased number of infectious
titrated product), delivery (delays or interruptions episodes and days of mechanical ventilation,
in delivery for procedures, transfers, patient fac- less hypoglycemia, and no difference in 90-day
tors, etc.), and clinical preference. However, the mortality. These differences were found for sub-
amount of nutrition that was delivered was con- groups of cardiac surgical patients and other
sistent with amounts in previous studies in critically ill patients. Post hoc analyses suggest-
which delivery also fell short of the target in this ed a dose–response relationship between an in-
population, suggesting that there are substantial creased amount of parenteral nutrition and an
practical and organizational impediments for increased rate of infectious episodes.30 Despite
both routes of delivery, at least during an initial important differences between our study and the
5-day period. However, the similar between- EPaNIC trial and between the two study groups
group caloric intake reinforces the design of our in EPaNIC, our results potentially support their
study to focus on the evaluation of the delivery hypothesis that among patients receiving early
route uncomplicated by dose. supplemental parenteral nutrition, the dose is
Our study was conducted in ICUs in England more associated with harm than is the route of
that had preexisting, established protocols for delivery. In a trial conducted at two sites, Hei-
the delivery of parenteral and enteral nutrition, degger et al.11 found no difference in the rate of
prevention of bloodstream infections and venti- infection between day 8 and day 28 among pa-
lator-associated pneumonia, and glycemic con- tients who started receiving individually opti-
trol. It was designed as a pragmatic effectiveness mized parenteral nutrition to supplement inad-
study and represents the reality of current criti- equate enteral intake on day 4 and patients
cal care practice in the English National Health receiving only enteral nutrition.11 In a trial con-
Service. Although the study had a large enroll- ducted at 31 sites involving patients with relative
1682 n engl j med 371;18 262

contraindications to enteral nutrition, Doig et al.31 cally administered, is neither more harmful nor
found no differences in 60-day mortality and the more beneficial than such support through the
number of infectious episodes but fewer days of enteral route.
mechanical ventilation in patients receiving early
parenteral nutrition, as compared with standard The views expressed in this article are those of the authors
and do not necessarily reflect those of the Health Technology
care. However, in the standard-care group, 27% of Assessment Program of the National Institute for Health Re-
patients received early parenteral nutrition and search, National Health Service, or the Department of Health.
41% received no nutritional support.31 In draw- Supported by a grant from the Health Technology Assessment
Program of the United Kingdom National Institute for Health
ing any comparison, it must be noted that in our Research (project no. 07/52/03).
study, we asked a different research question in Ms. Bear reports receiving consulting fees from Nutricia,
a different population of critically ill patients. lecture fees from Nestle Nutrition, travel support from Nutricia
and Baxter, and grant support through her institution from
Our study leaves unanswered the question of Corpak MedSystems UK; Ms. Segaran, receiving travel support
nutritional dose and the determination of energy from Abbott Nutrition; Dr. Beale, receiving consulting and lec-
and protein or amino acid requirements for criti- ture fees from Nestle Nutrition and grant support from Frese-
nius Kabi (all through his institution), and negotiating an un-
cally ill patients. We specifically did not com- restricted research grant from Nestle Nutrition (through the
pare an individualized enteral feeding regimen European Society of Intensive Care Medicine); and Dr. Mythen,
(that allows for increased amounts in patients receiving lecture fees from Fresenius Kabi and Baxter and grant
support from Fresenius Kabi and holding a patent related to
who can tolerate enteral feeding) with its paren- methods and apparatus for guiding medical care based on de-
teral equivalent (with individualized monitoring tected gastric function through Medical Defence Technologies.
of metabolic and protein balance). However, No other potential conflict of interest relevant to this article was
reported.
our findings suggest that early nutritional sup- Disclosure forms provided by the authors are available with
port through the parenteral route, as it is typi- the full text of this article at NEJM.org.
References
1. Gramlich L, Kichian K, Pinilla J, Ro- 7. Engel JM, Muhling J, Junger A, Meng- webarchive.nationalarchives.gov.uk/
dych NJ, Dhaliwal R, Heyland DK. Does es T, Karcher B, Hempelmann G. Enteral 20130107105354/http://www.dh.gov.uk/
enteral nutrition compared to parenteral nutrition practice in a surgical intensive prod_consum_dh/groups/dh_digitalassets/
nutrition result in better outcomes in care unit: what proportion of energy ex- @dh/@en/documents/digitalasset/dh
critically ill adult patients? A systematic penditure is delivered enterally? Clin Nutr _078120.pdf).
review of the literature. Nutrition 2004; 2003;22:187-92. 14. Reducing the harm caused by mis-
20:843-8. 8. Boitano M, Bojak S, McCloskey S, Mc- placed nasogastric feeding tubes — pa-
2. Heyland DK, Dhaliwal R, Drover JW, Caul DS, McDonough M. Improving the tient safety alert. London: National Patient
Gramlich L, Dodek P. Canadian clinical safety and effectiveness of parenteral nu- Safety Agency, February 2005 (http://
practice guidelines for nutrition support trition: results of a quality improvement www.nrls.npsa.nhs.uk/resources/?EntryId
in mechanically ventilated, critically ill collaboration. Nutr Clin Pract 2010;25: 45=59794).
adult patients. JPEN J Parenter Enteral 663-71. 15. How to confirm the correct position
Nutr 2003;27:355-73. 9. Naylor CJ, Griffiths RD, Fernandez of nasogastric feeding tubes in infants,
3. Simpson F, Doig GS. Parenteral vs. RS. Does a multidisciplinary total paren- children and adults — interim advice for
enteral nutrition in the critically ill pa- teral nutrition team improve patient out- healthcare staff. London: National Patient
tient: a meta-analysis of trials using the comes? A systematic review. JPEN J Par- Safety Agency, 2005 (http://www.nrls.npsa
intention to treat principle. Intensive Care enter Enteral Nutr 2004;28:251-8. .nhs.uk/resources/?EntryId45=59794).
Med 2005;31:12-23. 10. Casaer MP, Mesotten D, Hermans G, 16. Dellinger RP, Levy MM, Rhodes A, et
4. Martindale RG, McClave SA, Vanek et al. Early versus late parenteral nutrition al. Surviving Sepsis Campaign: interna-
VW, et al. Guidelines for the provision in critically ill adults. N Engl J Med 2011; tional guidelines for management of se-
and assessment of nutrition support ther- 365:506-17. vere sepsis and septic shock: 2012. Crit
apy in the adult critically ill patient: Soci- 11. Heidegger CP, Berger MM, Graf S, et Care Med 2013;41:580-637.
ety of Critical Care Medicine and Ameri- al. Optimisation of energy provision with 17. Harvey S, Harrison DA, Singer M, et
can Society for Parenteral and Enteral supplemental parenteral nutrition in criti- al. Assessment of the clinical effective-
Nutrition: executive summary. Crit Care cally ill patients: a randomised controlled ness of pulmonary artery catheters in
Med 2009;37:1757-61. clinical trial. Lancet 2013;381:385-93. management of patients in intensive care
5. Singer P, Berger MM, Van den Berghe 12. Harrison DA, Brady AR, Rowan K. (PAC-Man): a randomised controlled trial.
G, et al. ESPEN guidelines on parenteral Case mix, outcome and length of stay for Lancet 2005;366:472-7.
nutrition: intensive care. Clin Nutr 2009; admissions to adult, general critical care 18. Grant AM, Altman DG, Babiker AB, et
28:387-400. units in England, Wales and Northern Ire- al. Issues in data monitoring and interim
6. De Jonghe B, Appere-De-Vechi C, land: the Intensive Care National Audit analysis of trials. Health Technol Assess
Fournier M, et al. A prospective survey of and Research Centre Case Mix Programme 2005;9:1-238.
nutritional support practices in intensive Database. Crit Care 2004;8:R99-R111. 19. Harvey SE, Parrott F, Harrison DA,
care unit patients: what is prescribed? 13. High impact intervention no 1: cen- Mythen M, Rowan KM. The CALORIES
What is delivered? Crit Care Med 2001; tral venous catheter care bundle. London: trial statistical analysis plan. Crit Care
29:8-12. National Health Service, 2007 (http:// Resusc (in press).
n engl j med 371;18 263

Source: The New England Journal of Medicine Early Nutritional Support in Critically Ill Adults
20. Harrison DA, Parry GJ, Carpenter JR, 25. Greenland S. An introduction to in- 29. Reignier J, Mercier E, Le Gouge A, et al.
Short A, Rowan K. A new risk prediction strumental variables for epidemiologists. Effect of not monitoring residual gastric
model for critical care: the Intensive Int J Epidemiol 2000;29:722-9. volume on risk of ventilator-associated
Care National Audit and Research Centre 26. Vincent JL, Moreno R, Takala J, et al. pneumonia in adults receiving mechani-
(ICNARC) model. Crit Care Med 2007;35: The SOFA (Sepsis-related Organ Failure cal ventilation and early enteral feeding: a
1091-8. Assessment) score to describe organ dys- randomized controlled trial. JAMA 2013;
21. van Buuren S, Boshuizen HC, Knook function/failure. Intensive Care Med 1996; 309:249-56.
DL. Multiple imputation of missing blood 22:707-10. 30. Casaer MP, Wilmer A, Hermans G,
pressure covariates in survival analysis. 27. Bion J, Richardson A, Hibbert P, et al. Wouters PJ, Mesotten D, Van den Berghe
Stat Med 1999;18:681-94. ‘Matching Michigan’: a 2-year stepped G. Role of disease and macronutrient dose
22. Thompson SG, Barber JA. How should interventional programme to minimise in the randomized controlled EPaNIC
cost data in pragmatic randomised trials central venous catheter-blood stream in- trial: a post hoc analysis. Am J Respir Crit
be analysed? BMJ 2000;320:1197-200. fections in intensive care units in England. Care Med 2013;187:247-55.
23. Knaus WA, Draper EA, Wagner DP, BMJ Qual Saf 2013;22:110-23. 31. Doig GS, Simpson F, Sweetman EA, et
Zimmerman JE. APACHE II: a severity of 28. High impact intervention no 5: care al. Early parenteral nutrition in critically
disease classification system. Crit Care bundle for ventilated patients (or tracheos- ill patients with short-term relative con-
Med 1985;13:818-29. tomy where appropriate). London: Nation- traindications to early enteral nutrition:
24. Fischer K, Goetghebeur E, Vrijens B, al Health Service, 2007 (http://webarchive a randomized controlled trial. JAMA
White IR. A structural mean model to al- .nationalarchives.gov.uk/20130107105354/ 2013;309:2130-8.
low for noncompliance in a randomized http://www.dh.gov.uk/prod_consum_dh/ Copyright © 2014 Massachusetts Medical Society.
trial comparing 2 active treatments. Bio- groups/dh_digitalassets/@dh/@en/
statistics 2011;12:247-57. documents/digitalasset/dh_078124.pdf).
receive immediate notification when an article

is published online first
To be notified by e-mail when Journal articles

are published Online First, sign up at NEJM.org.
1684 n engl j med 371;18

264 nejm.org october 30, 2014
edi t or i a l
The Route of Early Nutrition in Critical Illness

Deborah Cook, M.D., and Yaseen Arabi, M.D.
Randomized trials may be broadly categorized enteral formulas in intensive care units (ICUs),
as explanatory or pragmatic. Explanatory trials blocked feeding tubes, delayed reinsertions, sus-
address the question, ‘‘Can this intervention work pension of nutrient delivery on the basis of gastric
under ideal conditions?” Pragmatic trials ask, residual volumes, vomiting or aspiration, challeng-
‘‘Does this intervention work under usual condi- es in small-bowel access among patients with
tions?’’ Usual conditions, of course, are context- feeding intolerance, and insufficient consultation
dependent. In the United Kingdom, the CALORIES with dietitians.3
trial by Harvey et al.,1 reported in the Journal, Might intravenous nutrition be more reliable,
addresses the clinical effectiveness of early en- better meet caloric needs, and confer better out-
teral nutrition, as compared with early parenteral comes than enteral nutrition? The association be-
nutrition, in critically ill patients without contra- tween parenteral nutrition and increased risk of
indications to either route of delivery. This trial, infection may be attenuated today by contempo-
commissioned and sponsored by the Health rary management of central venous catheters and
Technology Assessment Program of the National avoidance of hyperglycemia. In a meta-analysis
Institute for Health Research, was pragmatic in of 13 trials of parenteral nutrition, as compared
design and conduct. with enteral nutrition provided by Harvey et al.
To maximize the detection of differences bein a Supplementary Appendix to their study, there
tween groups, investigators who are designing was no significant reduction in mortality among
an explanatory trial might select malnourished patients receiving parenteral nutrition (relative
patients, maintain blinding with the use of a risk, 0.82; 95% confidence interval [CI], 0.60 to
double-dummy technique, establish near-perfect 1.11),1 whereas in a meta-analysis of 9 trials in
compliance, and minimize random and systemic which analyses were performed on the intention-
errors by adjudicating nonmortality outcomes. to-treat principle, there was a significant bene-
Although explanatory trials are enlightening, fit favoring parenteral nutrition (odds ratio, 0.51;
they often overestimate treatment effects ob- 95% CI, 0.27 to 0.97).4 Thus, Harvey et al. hypoth-
served in practice. Accordingly, the National esized that parenteral nutrition would be superior
Health Service sought to understand the real- to enteral nutrition in the CALORIES trial.
world consequences of early nutrition versus There is conspicuous variation among prac-
early parenteral nutrition. tices regarding the use and timing of parenteral
The adage “use the gut when possible” is based nutrition,5 and guidelines provide differing recom-
on a physiological rationale, including the fact that mendations. European guidelines recommend the
immune-enhancing benefits of enteral nutrition initiation of parenteral nutrition within 48 hours
have been observed in mechanistic studies. In ad- after admission if enteral nutrition is contraindi-
dition, feeding by the enteral route is often more cated or in patients with feeding intolerance.6
convenient, and indirect evidence suggests that North American guidelines generally recommend
early initiation of enteral nutrition has been asso- the initiation of parenteral nutrition after the
ciated with fewer infections than late initiation.2 first week in the ICU.7,8 International guidelines
However, obstacles to starting and sustaining en- for patients with sepsis suggest the use of enteral
teral nutrition abound. Barriers include the ten- nutrition alone rather than parenteral nutrition
dency for prioritization of other aspects of care alone or in conjunction with enteral nutrition
over nutrition, shortages of feeding pumps and during the first week.9

265 october 30, 2014
editorials
In the CALORIES trial, investigators in 33 The CALORIES trial weaves an important

English ICUs used a 24-hour telephone random- thread into the complex tapestry of evidence
ization system, performed analyses based on the that informs the practice of nutritional support
intention-to-treat principle, and had excellent in the ICU. Among patients without contraindi-
follow-up. Staff members monitored nutritional cations to either route of delivery, when enteral
adequacy and nutritional complications using nutrition and parenteral nutrition are initiated
accepted guidelines. Selected sites had estab- early and with similar caloric and protein doses,
lished protocols for the use of enteral nutrition clinical outcomes appear to be similar. The lack
and parenteral nutrition that included serum glu- of benefit associated with exclusive early paren-
cose targets (<180 mg per deciliter [10 mmol per teral nutrition raises questions about inflated es-
liter]) and the use of infection-prevention meth- timates of benefit in previous smaller trials.
ods. Sites were committed to protocol adherence, Meanwhile, bedside decisions about the route of
screening logs, and timely data submission, the nutritional support during critical illness will
last of which is imperative for legitimate safety probably continue to be influenced by availability,
monitoring. access, assessment of individualized nutritional
On average, nutritional support was initiated needs, perceived costs, and prevailing practices.
within 24 hours after admission to the ICU. The Disclosure forms provided by the authors are available with
targeted intake was 25 kcal per kilogram of body the full text of this article at NEJM.org.
weight per day, a goal that was adapted for pa- From the Departments of Medicine, Critical Care, and Clinical
tients at the extremes of body-mass index (BMI). Epidemiology and Biostatistics, McMaster University, and St.
Joseph’s Healthcare, Hamilton, ON, Canada (D.C.); and the
The amount of protein calories, product selection,
Department of Intensive Care, King Saud bin Abdulaziz Univer-
and the use of supplements were determined by sity for Health Sciences, King Abdullah International Medical
local practices. Remarkably, sites achieved simi- Research Center, Riyadh, Saudi Arabia (Y.A.).
lar protein and caloric intake, and rates of ad- This article was published on October 1, 2014, at NEJM.org.
herence were similar, although patients in the
parenteral group received more micronutrients, 1. Harvey SE, Parrott F, Harrison DA, et al. Trial of the route of
early nutritional support in critically ill adults. N Engl J Med
especially selenium. However, the premise that 2014;371:1673-84.
parenteral nutrition would secure better nutri- 2. Canadian Clinical Practice Guidelines: 2.0: early vs. delayed
tional delivery was not realized. nutritional intake. Kingston, ON, Canada: Critical Care Nutri-
tion, March 2013 (http://www.criticalcarenutrition.com/docs/
Today, many nutrition trials evaluate mortal- cpgs2012/2.0.pdf).
ity and acquired infections, which are the most 3. Cahill NE, Murch L, Cook D, Heyland DK. Barriers to feed-
commonly used primary outcomes in random- ing critically ill patients: a multicenter survey of critical care
nurses. J Crit Care 2012;27:727-34.
ized trials conducted in ICUs since 2007.10 In the 4. Simpson F, Doig GS. Parenteral vs. enteral nutrition in the
CALORIES trial, no significant between-group critically ill patient: a meta-analysis of trials using the intention
difference was observed in the rate of infection to treat principle. Intensive Care Med 2005;31:12-23.
5. Heyland DK, Dhaliwal R, Wang M, Day AG. The prevalence
or death at each prespecified time point up to of iatrogenic underfeeding in the nutritionally ‘at-risk’ critically
90 days. Predictably, patients in the enteral group ill patient: results of an international multicenter prospective
had more frequent vomiting, had a higher mean study. Clin Nutr 2014 July 19 (Epub ahead of print).
6. Singer P, Berger MM, Van den Berghe G, et al. ESPEN Guide-
gastric residual volume, and received more pro- lines on Parenteral Nutrition: intensive care. Clin Nutr 2009;
kinetic drugs. 28:387-400.
Could the profile of the study population ex- 7. Martindale RG, McClave SA, Vanek VW, et al. Guidelines for
the provision and assessment of nutrition support therapy in the
plain the similar findings in the two groups? adult critically ill patient: Society of Critical Care Medicine and
Although patients were at high risk for death American Society for Parenteral and Enteral Nutrition: Executive
(80% received mechanical ventilation or vasopres- Summary. Crit Care Med 2009;37:1757-61.
8. Dhaliwal R, Cahill N, Lemieux M, Heyland DK. The Cana-
sors), 90% had no malnutrition, as defined on dian Critical Care Nutrition guidelines in 2013: an update on
the basis of the BMI and recent weight loss; current recommendations and implementation strategies. Nutr
subjectively, 13% were malnourished. The use of Clin Pract 2014;29:29-43.
a scoring system to identify patients who may Campaign: international guidelines for management of severe
be most nutrition-responsive may serve in future sepsis and septic shock: 2012. Crit Care Med 2013;41:580-637.
explanatory trials. However, the administration 10. Harhay MO, Wagner J, Ratcliffe SJ, et al. Outcomes and sta-
tistical power in adult critical care randomized trials. Am J Res-
of nutrition — an archetypal basic life support pir Crit Care Med 2014;189:1469-78.
— clearly needs to be evaluated in more prag- DOI: 10.1056/NEJMe1411474
matic trials. Copyright © 2014 Massachusetts Medical Society.
n engl j med 371;18 266

T h e n e w e ng l a n d j o u r na l o f m e dic i n e
c or r e sp ondence
Trial of the Route of Early Nutritional Support

in Critically Ill Adults
1. Harvey SE, Parrott F, Harrison DA, et al. Trial of the route of
To the Editor: In reporting the results of the early nutritional support in critically ill adults. N Engl J Med 2014;
CALORIES trial, Harvey and colleagues (Oct. 30 371:1673-84.
2. Heidegger CP, Berger MM, Graf S, et al. Optimisation of
issue)1 indicate that the route of delivery of early energy provision with supplemental parenteral nutrition in criti-
nutritional support in the intensive care unit cally ill patients: a randomised controlled clinical trial. Lancet
(ICU) does not alter patient outcomes. This mes- 2013;381:385-93.
3. Heyland D, Muscedere J, Wischmeyer PE, et al. A random-
sage contradicts the widely held belief that the ized trial of glutamine and antioxidants in critically ill patients.
enteral route, which is more physiological, is to N Engl J Med 2013;368:1489-97. [Erratum, N Engl J Med 2013;
be preferred. However, we think that there is an- 368:1853.]
other implicit message: this study suggests that DOI: 10.1056/NEJMc1414479
the role of nutritional support in the ICU should

be reconsidered. Past evidence has led researchers To the Editor: The external validity of the
to implement an overzealous approach to nutri- study by Harvey et al. is questionable in our
tional support in patients in the ICU. Given the view. The inclusion criteria do not adhere to
results from the present study, we should proba- standards of care1,2 based on updated literature
bly take a step backward. There is still an unan- (see www.criticalcarenutrition.com). These stan-
swered question regarding which critically ill dards call for the preferential use of enteral nu-
patients should receive early nutritional support. trition in patients who are able to receive it. The
Some recent trials1-3 suggest that such patients area in which experts and guidelines disagree is
may be those with depleted body stores due to the timing of supplemental parenteral nutrition
malnutrition rather than all those who are at nu- in patients who are unable to receive sufficient en-
tritional risk as a consequence of critical illness. teral nutrition without unacceptable side effects.3,4
We believe that targeting early nutritional support The lack of advantage of early parenteral nutri-
to the right patients constitutes a key point that tion in the CALORIES study is not surprising,
should be addressed. given that some patients may not have needed any
Once that question is addressed, we could nutritional support and the most appropriate route
focus again on timing, the route of delivery, was not assessed in other patients in this trial.
protein and caloric targets, and nutrients that Both the low protein intake and the low caloric
have putative pharmacologic activity. intake, as well as the low number of patients per
Emanuele Cereda, M.D., Ph.D. center, are other major concerns. In our view, the
Riccardo Caccialanza, M.D. primary outcome, all-cause mortality at 30 days,
Fondazione IRCCS Policlinico San Matteo was also unlikely to be related to the efficacy of
Pavia, Italy the therapeutic interventions. The contribution
e.cereda@smatteo.pv.it
of this study would have been much more valu-
ported. able after a selection of patients in whom the
267
adequacy of nutrition mattered and could be cost-effectiveness of early parenteral nutrition

properly evaluated. as compared with early enteral nutrition at 1 year.
Jean-Charles Preiser, M.D., Ph.D. I do not find mention of this in the article.
Erasme University Hospital Madhusudan Ramamurthy, M.B., B.S.
Brussels, Belgium
Sakra World Hospital
Vincent Fraipont, M.D. Bengaluru, India
madhusudanr@ymail.com
Centre Hospitalier Régional de la Citadelle
Liege, Belgium No potential conflict of interest relevant to this letter was re-
ported.
Didier Quilliot, M.D., Ph.D. DOI: 10.1056/NEJMc1414479
Centre Hospitalier Universitaire de Nancy
Nancy, France
ported.
The Authors Reply: CALORIES was a pragmatic
trial evaluating the effectiveness of the paren-
1. Singer P, Berger MM, Van den Berghe G, et al. ESPEN guide-
lines on parenteral nutrition: intensive care. Clin Nutr 2009;28:
teral route as compared with the enteral route for
387-400. early nutritional support in critically ill adults.
2. Martindale RG, McClave SA, Vanek VW, et al. Guidelines for Nutritional support was initiated within 36 hours
the provision and assessment of nutrition support therapy in the
adult critically ill patient: Society of Critical Care Medicine and
after unplanned admission to one of a represen-
American Society for Parenteral and Enteral Nutrition: executive tative sample of ICUs in England and was used
summary. Crit Care Med 2009;37:1757-61. exclusively for 120 hours.
3. Casaer MP, Mesotten D, Hermans G, et al. Early versus late
parenteral nutrition in critically ill adults. N Engl J Med 2011;
The suggestion by Cereda and Caccialanza that
365:506-17. critically ill patients should receive early nutri-
4. Heidegger CP, Berger MM, Graf S, et al. Optimisation of tional support only if they are malnourished is
energy provision with supplemental parenteral nutrition in criti-
cally ill patients: a randomised controlled clinical trial. Lancet
interesting but not directly relevant to our trial.
2012;381:385-93. Therefore, we do not believe that our data can be
DOI: 10.1056/NEJMc1414479 interpreted either to support or to refute their
hypothesis, which we agree merits further con-
sideration.
To the Editor: In my view, the CALORIES trial Preiser and colleagues appear to advocate
fails to support the hypothesis that the paren- opinion-based guidelines and a meta-analysis of
teral route is superior to the enteral route for the small, older, and methodologically compromised
delivery of early nutritional support in adults. Ac- studies, rather than evidence from a large, rigor-
cording to the trial design, even patients who ous, randomized, controlled trial. We disagree
were assigned to the parenteral group received with such an approach. Contrary to their asser-
the benefits of enteral nutrition. A significant tion, all patients recruited into the CALORIES
number of patients in the parenteral group re- trial met standard criteria to receive nutritional
ceived enteral nutrition both during the 120-hour support, as our article made clear. Moreover, ef-
period and after it. Among 700 patients in the fectiveness, not efficacy, was tested in our prag-
parenteral group, only 13 patients (1.9%) exclu- matic trial. Despite their belief that mortality
sively received parenteral nutrition after the in- was unlikely to be affected by the route of nutri-
tervention. Extrapolation of such a short dura- tional support, the meta-analysis by Simpson and
tion of parenteral nutrition with or without Doig1 suggested otherwise. The energy intake
enteral nutritional support to estimate the effect and protein intake in our pragmatic trial were
on mortality at 30 days is not ideal. Data about similar to or greater than those in other studies
coexisting diseases such as diabetes mellitus, of nutritional support in clinical practice.2 In
hypertension, congestive heart failure, and chronic addition, the number of eligible patients recruit-
obstructive pulmonary disease — all of which ed per center was greater than that in another
may have an important effect on secondary out- recent, large, multicenter trial of nutritional sup-
come measures — are not mentioned. The sum- port in intensive care.3
mary of the original protocol (available with the Ramamurthy’s statement that benefits are as-
full text of the article at NEJM.org) mentions a sociated with the use of the enteral route rather
primary objective to estimate the incremental than the parenteral route is not supported by our
268
correspondence
trial. Our aim was to evaluate the effect of the Since publication of their article, the authors report no fur-
route for the delivery of early nutritional sup-
1. Simpson F, Doig GS. Parenteral vs. enteral nutrition in the
port, and the number of patients in the paren- critically ill patient: a meta-analysis of trials using the intention
teral group who were fed enterally during the to treat principle. Intensive Care Med 2005;31:12-23.
120-hour intervention period was small. Base- 2. Alberda C, Gramlich L, Jones N, et al. The relationship be-
tween nutritional intake and clinical outcomes in critically ill
line characteristics, including coexisting condi- patients: results of an international multicenter observational
tions, were well balanced between the groups study. Intensive Care Med 2009;35:1728-37. [Erratum, Intensive
and were summarized in Table 1 of our article. Care Med 2009;35:1821.]
3. Doig GS, Simpson F, Sweetman EA, et al. Early parenteral
A 1-year follow-up study is under way to assess nutrition in critically ill patients with short-term relative contra-
longer-term outcomes and to provide an inte- indications to early enteral nutrition: a randomized controlled
grated economic evaluation. trial. JAMA 2013;309:2130-8.
DOI: 10.1056/NEJMc1414479
Sheila E. Harvey, Ph.D. Correspondence Copyright © 2015 Massachusetts Medical Society.
Intensive Care National Audit and Research Centre
Ella Segaran, M.Sc.
Richard Leonard, M.B., B.Chir.
Imperial College Healthcare NHS Trust

original article
Early versus On-Demand Nasoenteric Tube

Feeding in Acute Pancreatitis
O.J. Bakker, S. van Brunschot, H.C. van Santvoort, M.G. Besselink, T.L. Bollen,
M.A. Boermeester, C.H. Dejong, H. van Goor, K. Bosscha, U. Ahmed Ali, S. Bouwense,
W.M. van Grevenstein, J. Heisterkamp, A.P. Houdijk, J.M. Jansen, T.M. Karsten,
E.R. Manusama, V.B. Nieuwenhuijs, A.F. Schaapherder, G.P. van der Schelling,
M.P. Schwartz, B.W.M. Spanier, A. Tan, J. Vecht, B.L. Weusten, B.J. Witteman,
L.M. Akkermans, M.J. Bruno, M.G. Dijkgraaf, B. van Ramshorst,
and H.G. Gooszen, for the Dutch Pancreatitis Study Group
A BS T R AC T
BACKGROUND
Early enteral feeding through a nasoenteric feeding tube is often used in patients The authors’ full names, academic de-
with severe acute pancreatitis to prevent gut-derived infections, but evidence to sup- grees, and affiliations are listed in the
Appendix. Address reprint requests to
port this strategy is limited. We conducted a multicenter, randomized trial compar- Dr. Bakker at University Medical Center
ing early nasoenteric tube feeding with an oral diet at 72 hours after presentation Utrecht, P.O. Box 85500, HP G04.228,
to the emergency department in patients with acute pancreatitis. 3508 GA Utrecht, the Netherlands.
N Engl J Med 2014;371:1983-93.
METHODS DOI: 10.1056/NEJMoa1404393
We enrolled patients with acute pancreatitis who were at high risk for complications
on the basis of an Acute Physiology and Chronic Health Evaluation II score of 8 or
higher (on a scale of 0 to 71, with higher scores indicating more severe disease), an
Imrie or modified Glasgow score of 3 or higher (on a scale of 0 to 8, with higher
scores indicating more severe disease), or a serum C-reactive protein level of more
than 150 mg per liter. Patients were randomly assigned to nasoenteric tube feeding
within 24 hours after randomization (early group) or to an oral diet initiated 72 hours
after presentation (on-demand group), with tube feeding provided if the oral diet was
not tolerated. The primary end point was a composite of major infection (infected
pancreatic necrosis, bacteremia, or pneumonia) or death during 6 months of follow-up.
RESULTS
A total of 208 patients were enrolled at 19 Dutch hospitals. The primary end point
occurred in 30 of 101 patients (30%) in the early group and in 28 of 104 (27%) in
the on-demand group (risk ratio, 1.07; 95% confidence interval, 0.79 to 1.44;
P = 0.76). There were no significant differences between the early group and the on-
demand group in the rate of major infection (25% and 26%, respectively; P = 0.87)
or death (11% and 7%, respectively; P = 0.33). In the on-demand group, 72 patients
(69%) tolerated an oral diet and did not require tube feeding.
CONCLUSIONS
This trial did not show the superiority of early nasoenteric tube feeding, as com-
pared with an oral diet after 72 hours, in reducing the rate of infection or death
in patients with acute pancreatitis at high risk for complications. (Funded by the
Netherlands Organization for Health Research and Development and others;
PYTHON Current Controlled Trials number, ISRCTN18170985.)

november 20, 2014 1983
A
cute pancreatitis is the most com- tive prevention of infection with early tube feed-
mon gastrointestinal disease leading to ing has passed.7 To address this problem in the
hospital admission, and its incidence con- management of acute pancreatitis, we compared
tinues to rise.1-4 Most patients with acute pancre- the effects of early nasoenteric tube feeding with
atitis recover uneventfully and are discharged those of an oral diet started at 72 hours, with a
after a few days.5,6 In 20% of patients, the dis- switch to nasoenteric tube feeding only in the
ease course is complicated by major infection, case of insufficient oral intake.
such as infected pancreatic necrosis, which is
associated with a mortality of 15%.7-11 ME THODS
A meta-analysis of eight randomized trials
involving 348 patients showed that nasoenteric STUDY PARTICIPANTS
tube feeding, as compared with total parenteral The protocol of the Pancreatitis, Very Early Com-
nutrition, reduced the rate of infections and pared with Selective Delayed Start of Enteral
mortality among patients with severe pancreati- Feeding (PYTHON) trial has been published pre-
tis.12 These infections are thought to be medi- viously33 and is available with the full text of this
ated by bacterial translocation from the gut, article at NEJM.org. The study was conducted ac-
provoked by disturbed intestinal motility, bacte- cording to the protocol. Adults with a first epi-
rial overgrowth, and increased mucosal perme- sode of acute pancreatitis who were at high risk
ability.13-18 Nasoenteric tube feeding is believed for complications (i.e., patients predicted to have
to stimulate intestinal motility — thus reducing severe pancreatitis) were eligible to undergo ran-
bacterial overgrowth — and may increase domization. Patients were considered to be at
splanchnic blood flow, which helps to preserve high risk for complications if, within 24 hours
the integrity of the gut mucosa.19,20 Total paren- after presentation to the emergency department,
teral nutrition lacks the trophic effect of enteric the Acute Physiology and Chronic Health Evalua-
feeding and is associated with central venous tion (APACHE) II34 score was 8 or higher (on a
catheter–related infections as well as metabolic scale of 0 to 71, with higher scores indicating
complications.21 more severe disease), if the Imrie or modified
A meta-analysis of randomized trials involv- Glasgow score35 was 3 or higher (on a scale of
ing acutely ill patients admitted to the hospital 0 to 8, with higher scores indicating more severe
for indications other than pancreatitis showed a disease), or if the serum C-reactive protein level
22% reduction in the rate of major infection was more than 150 mg per liter.36 These assess-
when nasoenteric tube feeding was started very ments predict the development of complications
early (≤36 hours after admission or surgery) as during the course of the disease. Pancreatitis was
compared with a later start.22 Similarly, nonran- diagnosed if at least two of the three following
domized studies of acute pancreatitis have shown features were present: typical abdominal pain, a
that nasoenteric tube feeding started within 48 serum lipase or amylase level that was more than
hours after admission, as compared with a start 3 times the upper limit of the normal range, or
after 48 hours, significantly reduced the rate of characteristic findings on cross-sectional imag-
major infection and in some studies even re- ing of the abdomen. The exclusion criteria are
duced mortality.23-26 given in the Supplementary Appendix, available
On the basis of these potential benefits, at NEJM.org.
American and European nutritional societies
recommend routine early nasoenteric tube feed- STUDY DESIGN AND OVERSIGHT
ing in all patients with severe pancreatitis.27-29 The PYTHON trial was a multicenter, random-
Guidelines from gastroenterologic and pancreatic ized, controlled superiority trial performed in six
societies, however, state that, regardless of dis- university medical centers and 13 large teaching
ease severity, tube feeding is indicated when hospitals of the Dutch Pancreatitis Study Group.
patients are not able to tolerate an oral diet for up Patients were randomly assigned in a 1:1 ratio
to 7 days.30,31 Unfortunately, it takes 3 to 4 days either to nasoenteric tube feeding initiated with-
after admission to make this assessment,32 and in 24 hours after randomization (the early group)
by that time the window of opportunity for effec- or to an oral diet starting at 72 hours (the on-
1984 n engl j med 371;21 271

Early vs. On-Demand Tube Feeding in Acute Pancreatitis
demand group). Randomization was performed after the placement of a nasojejunal tube, and
centrally by the study coordinator with the use of the same procedure was followed as in the early
a Web-based system that used permuted-block group.
randomization with a concealed, varying block
size. Randomization was stratified according to END POINTS
treatment center and a dichotomized APACHE II The primary end point was a composite of major
score (<13 vs. ≥13); the latter stratification factor infection or death within 6 months after random-
was used because patients with an APACHE II ization. Major infection was defined as infected
score of 13 or higher are at increased risk for pancreatic necrosis, bacteremia, or pneumonia
major infection. (for definitions, see Box S1 in the Supplementary
All the patients or their legal representatives Appendix). Predefined secondary end points in-
provided written informed consent. The study cluded the development of necrotizing pancreati-
protocol was approved by the institutional review tis as diagnosed on the basis of computed tomog-
board of the University Medical Center Utrecht raphy (CT) performed 5 to 7 days after admission
and by all the participating centers. All the au- (because pancreatic parenchymal necrosis may
thors vouch for the veracity and completeness of take up to 72 hours to develop) and the develop-
the data and data analyses. The sponsors were ment of organ failure after randomization.
not involved in the design or conduct of the
study or in the preparation of the manuscript or DATA COLLECTION AND END-POINT ASSESSMENT
the decision to submit it for publication. Dieticians registered the caloric intake and calcu-
lated energy-intake targets during the first week
STUDY PROCEDURES after admission on the basis of actual body
Patients underwent randomization within 24 hours weight. All CT studies were interpreted by an
after presentation to the emergency department. author who is an experienced radiologist and
Those assigned to early nasoenteric feeding re- who was unaware of the treatment assignments.
ceived a nasojejunal feeding tube as soon as pos- An adjudication committee, consisting of four
sible but not later than 24 hours after randomiza- pancreatic surgeons and a gastroenterologist who
tion. Feeding tubes were placed endoscopically were unaware of the treatment assignments, in-
or radiologically, according to local practice. Naso- dividually evaluated each patient for the occur-
enteric feeding was administered as Nutrison rence of the primary end point before interim
Protein Plus (Nutricia). After tube placement, and final analyses. Disagreements with respect
feeding was started at 20 ml per hour during the to major infection were resolved during a plenary
first 24 hours and was gradually increased (see consensus meeting.
the Supplementary Appendix). In the two study
groups, full nutrition was defined as an energy PATIENT SAFETY
target of 25 kcal per kilogram of body weight per An independent data and safety monitoring com-
day for patients in the intensive care unit (ICU) mittee evaluated the progress of the trial and ex-
and 30 kcal per kilogram per day for patients in amined safety end points after the completion of
the ward.28,37,38 follow-up in each consecutive group of 25 patients.
Patients assigned to an oral diet did not re- Adverse events were listed and presented to the
ceive nutrition by any means other than that data and safety monitoring committee in an un-
provided by standard intravenous fluids during blinded fashion.
the first 72 hours after presentation to the
emergency department. Exceptions were made STATISTICAL ANALYSIS
for patients who requested oral food during this The expected incidence of the primary end point
72-hour period. At 72 hours, all the patients in in the on-demand group was based on data from
the on-demand group were given an oral diet. individual patients in the placebo group of a pre-
If an oral diet was not tolerated, it was offered vious randomized trial.39 For the early group,
again after 24 hours. If an oral diet still was data from randomized trials comparing nasoen-
not tolerated after 96 hours from the time of teric with parenteral nutrition were used to esti-
presentation, nasoenteric feeding was started mate the incidence.33,40-42 The sample-size calcu-
n engl j med 371;21 272

lation was based on an expected reduction in the incorrectly diagnosed with acute pancreatitis
primary composite end point associated with (2 patients had gastric carcinoma and 1 had in-
early tube feeding from 40 to 22%.33 We esti- testinal volvulus). A total of 101 patients in the
mated that a sample of 208 patients would pro- early group and 104 in the on-demand group
vide the study with at least 80% power, at a two- were included in the intention-to-treat analysis.
sided alpha level of 5% and assuming a 1% loss Baseline characteristics, presented in Table 1,
to follow-up. Analysis was based on the inten- were equally distributed between the groups ex-
tion-to-treat method, with the exclusion only of cept for the mean (±SD) BMI (29±5 in the early
patients for whom the adjudication committee, group vs. 27±5 in the on-demand group, P = 0.01).
whose members were unaware of the treatment Details regarding the number of calories de-
assignments, decided before any analysis that the livered during the first week after admission and
diagnosis of acute pancreatitis was incorrect. the timing of feeding are shown in Figure 1, and
Predefined subgroups included patients with in Table S1 in the Supplementary Appendix. As
an APACHE II score below 13 and those with a specified by the protocol, patients in the early
score of 13 or higher at randomization. Two post group received feeding earlier than those in the
hoc subgroup analyses were performed: one for on-demand group. Nasoenteric tube feeding in
patients with the systemic inflammatory re- the early group was started a median of 8 hours
sponse syndrome (SIRS, as defined by the Con- after randomization and a median of 23 hours
sensus Conference criteria of the American Col- after presentation to the emergency department,
lege of Chest Physicians–Society of Critical Care as compared with initiation of an oral diet 64
Medicine) at randomization, because such pa- hours after randomization and 72 hours after
tients are at high risk for complications,43 and presentation in the on-demand group (P<0.001)
one for a low or high body-mass index (BMI; the (Table S1 in the Supplementary Appendix). A
weight in kilograms divided by the square of the total of 5 of 104 patients (5%) assigned to on-
height in meters), since the BMI differed sig- demand feeding requested and received food
nificantly between the two treatment groups at within the first 72 hours after presentation.
baseline.
An interim analysis of the primary end point OUTCOMES
was performed after 50% of the patients had Primary End Point
completed 6 months of follow-up. The interim The primary composite end point of major infec-
analysis was performed by an independent statis- tion or death occurred in 30 patients (30%) in the
tician, who was unaware of the treatment assign- early group, as compared with 28 (27%) in the
ments, applying the Peto approach with symmet- on-demand group (risk ratio, 1.07; 95% confi-
ric stopping boundaries at a P value of less than dence interval [CI], 0.79 to 1.44; absolute risk
0.001.33,44 difference, 3 percentage points; 95% CI, −9 to 15;
For the final analyses, a two-sided P value of P = 0.76). Major infections occurred in 25% of the
less than 0.05 was considered to indicate statis- patients in the early group and in 26% of those in
tical significance. P values were not adjusted for the on-demand group (P = 0.87) (Table 2). Mor-
multiple testing. tality was 11% in the early group, as compared
with 7% in the on-demand group (P = 0.33), and
R E SULT S most of the deaths were related to persistent
multiple organ failure (defined as failure of two
ENROLLMENT AND RANDOMIZATION or more organs on ≥3 consecutive days).
From August 2008 through June 2012, a total of
867 patients were assessed for eligibility (Fig. S1 in Secondary End Points
the Supplementary Appendix). A total of 208 pa- Necrotizing pancreatitis developed in 63% of
tients (24%) were enrolled and randomly assigned the patients in the early group and in 62% of
to early nasoenteric tube feeding (102 patients) or those in the on-demand group. A total of 18%
an oral diet with tube feeding if required (106). of the patients in the early group and 19% of
The adjudication committee excluded 3 patients those in the on-demand group required ICU ad-
who had undergone randomization and had been mission (Table 2).

273 november 20, 2014
Early On-Demand
Tube Feeding Tube Feeding
Female sex — no. (%) 46 (46) 45 (43)
Age — yr 65±16 65±15
Cause of pancreatitis — no. (%)
Gallstones 59 (58) 56 (54)
Alcohol abuse 14 (14) 23 (22)
Other 28 (28) 25 (24)
Body-mass index — no./total no. (%)†
<25 20/99 (20) 33/103 (32)
25 to <35 69/99 (70) 67/103 (65)
≥35 10/99 (10) 3/103 (3)
Disease severity
APACHE II score‡
Mean 11±4 11±5
≥13§ 32 (32) 29 (28)
Imrie or modified Glasgow score¶
Median 2 2
Range 0–6 0–5
C-reactive protein — mg/liter
Median 70 75
SIRS — no. (%)‖ 63 (62) 70 (67)
Respiratory failure — no. (%) 30 (30) 27 (26)
Multiple organ failure — no. (%)** 6 (6) 5 (5)
Duration — hr
Onset of symptoms to presentation at the emergency
department
Median 12 13
Presentation at the emergency department
to randomization
Median 13 11
* Plus–minus values are means ±SD. There were no significant between-group differences at baseline, except for body-
mass index (P = 0.01).
† The body-mass index is the weight in kilograms divided by the square of the height in meters.
‡ Scores on the Acute Physiology and Chronic Health Evaluation (APACHE) II range from 0 to 71, with higher scores
indicating more severe disease.34
§ Patients with an APACHE II score of 13 or higher constituted a predefined subgroup.
¶ Imrie or modified Glasgow35 scores range from 0 to 8, with higher scores indicating more severe disease.
‖ The systemic inflammatory response syndrome (SIRS) was diagnosed with the use of the Consensus Conference criteria
of the American College of Chest Physicians–Society of Critical Care Medicine.
** Organ failure was defined as a modified Marshall score of 2 or more (on a scale of 0 to 12, with higher scores indicating
more severe disease), as proposed in the revised Atlanta classification of acute pancreatitis.45 Multiple organ failure
was defined as failure of two or more organs on the same day.
n engl j med 371;21 274

primary end point did not differ significantly

30 between the two treatment groups (Table S3 in
Early tube
feeding the Supplementary Appendix). Post hoc subgroup
On-demand analyses also did not show a significant between-
tube feeding
25
group difference in the primary end point for
patients with SIRS at randomization or those
with a BMI of less than 25 or 35 or more (Table
Calories Delivered (kcal/kg of weight/day)
S3 in the Supplementary Appendix). No signifi-

20 cant differences were observed in health care
utilization except for the number of tube place-
ments (145 tube placements in the early group
15 vs. 57 in the on-demand group, P<0.001) (Table
S4 in the Supplementary Appendix).
DISCUSSION
10
This multicenter, randomized trial involving pa-

tients with acute pancreatitis who were at high
5 risk for complications did not show that an early
start of nasoenteric tube feeding was superior to
the introduction of an oral diet after 72 hours,
with tube feeding only if required, in reducing
0
0 1 2 3 4 5 6 7 the composite end point of major infection or
Days after Admission death. With the oral diet and on-demand tube-
feeding strategy, only approximately one third of
Figure 1. Calories Delivered with the Use of Early versus On-Demand Naso- patients required a nasojejunal feeding tube.
enteric Tube Feeding. The absolute between-group difference in the
Each rectangle shows the mean value (horizontal line) and 95% confidence primary end point was 3 percentage points, with
interval (top and bottom of the rectangle). the 95% confidence interval ranging from 9 per-
centage points lower to 15 percentage points
higher. These findings do not support clinical
In the on-demand group, 32 patients (31%) guidelines recommending the early start of naso-
required nasoenteric tube feeding; 72 patients enteric tube feeding in all patients with severe
(69%) tolerated an oral diet and did not require acute pancreatitis in order to reduce the risks of
tube feeding (Table 3). In 9 of these 32 patients infection and death. However, this trial was not
(28%), tube feeding was prompted by the use of powered to exclude a substantial benefit of early
mechanical ventilation. The on-demand tube- feeding.
feeding strategy reduced the number of days to The results of our trial differ from those of
full tolerance of an oral diet (9 days with the previous trials and observational studies.12,23-26
early strategy vs. 6 days with the on-demand Previous trials showed an improved outcome
strategy, P = 0.001). Gastrointestinal events oc- after early nasoenteric tube feeding as compared
curred frequently, but the frequency did not dif- with total parenteral nutrition. This may be ex-
fer significantly between the groups. plained in part by complications associated
Attenuation of the acute inflammatory re- with providing total parenteral nutrition, such as
sponse was hypothesized to be part of the ben- catheter-related infections.21 The negative out-
eficial effect of early feeding. However, such an come of our study, as compared with the out-
effect did not occur (Fig. S3 in the Supplemen- comes in these previous trials, is not explained
tary Appendix). by differences in the timing of early tube feeding
In a predefined subgroup analysis restricted or the severity of pancreatitis in the study par-
to patients with an APACHE II score of 13 or ticipants. The timing of early nasoenteric tube
higher at randomization, the occurrence of the feeding in our study was similar to the timing

275 november 20, 2014
Table 2. Primary and Secondary End Points, According to the Intention-to-Treat Analysis.*
Early On-Demand
Tube Feeding Tube Feeding Risk Ratio
Primary composite end point: infection 30 (30) 28 (27) 1.07 (0.79–1.44) 0.76
or death — no. (%)
Secondary end points
Infection — no. (%)† 25 (25) 27 (26) 0.97 (0.70–1.34) 0.87
Infected pancreatic necrosis 9 (9) 15 (14) 0.74 (0.43–1.26) 0.28
Bacteremia 17 (17) 18 (17) 0.98 (0.68–1.43) 1.00
Pneumonia 12 (12) 13 (12) 0.97 (0.63–1.50) 1.00
Death — no. (%) 11 (11) 7 (7) 1.27 (0.85–1.89) 0.33
Necrotizing pancreatitis — no. (%)‡ 64 (63) 65 (62) 1.06 (0.77–1.47) 0.76
CT severity index§ 4±2 4±3 — 0.29
ICU admission after randomization 18 (18) 20 (19) 0.95 (0.66–1.38) 0.86
— no. (%)
Mechanical ventilation — no. (%) 12 (12) 14 (13) 0.93 (0.60–1.44) 0.84
New-onset organ failure — no./total
no. at risk (%)¶
Single organ failure 26/67 (39) 31/73 (42) 0.92 (0.65–1.32) 0.73
Persistent single organ failure 10/67 (15) 10/73 (14) 1.05 (0.65–1.70) 1.00
Multiple organ failure 7/67 (10) 6/73 (8) 1.14 (0.67–1.95) 0.77
Persistent multiple organ failure 4/67 (6) 4/73 (5) 1.05 (0.51–2.14) 1.00
* Plus-minus values are means ±SD. Risk ratios are for early tube feeding as compared with on-demand tube feeding.
ICU denotes intensive care unit.
† Patients may have had more than one type of infection.
‡ Necrotizing pancreatitis was defined as pancreatic parenchymal necrosis or extrapancreatic necrosis.45,46 In nine patients
(9%) in the early group and seven (7%) in the on-demand group, no CT was performed.
§ Scores on the CT severity index range from 0 to 10, with higher scores indicating more extensive pancreatic or extra-
pancreatic necrosis.
¶ New-onset organ failure was defined as organ failure that was not present at randomization. Persistent organ failure
was defined as organ failure present on 3 or more consecutive days (>48 hours). Multiple organ failure was defined
as failure of two or more organs on the same day.
in the previous studies. In addition, we used quality of the trials that form the basis for these
similar criteria for enrolling patients at high risk general ICU recommendations has been criti-
for complications, and we observed similar rates cized.48 Thus, for critically ill patients in general
of major infection and death. and for those with acute pancreatitis specifical-
Previous observational studies investigating ly, large, high-quality, randomized, controlled
the initiation of nasoenteric tube feeding within trials that show an improved outcome with early
48 hours after admission, as compared with enteral feeding are lacking.49
initiation more than 48 hours after admission, There are several possible explanations for
cannot differentiate between cause and effect the negative result of our study. First, early en-
(i.e., less severely ill patients may have been fed teral feeding may not be as effective as we antici-
earlier). This is in line with a recently revived pated. Our hypothesis was that the trophic effect
debate on the presumed benefit of early enteral of early enteral feeding would stabilize the integ-
feeding in critically ill patients in general. Early rity of the gut mucosa, reducing inflammation
enteral feeding is recommended in most current and improving the outcome. Early enteral feed-
ICU guidelines.38,47 However, the methodologic ing was not associated with a reduction in any
n engl j med 371;21 276

Table 3. Nutrition Tolerance and Gastrointestinal Events.*
Early On-Demand
Tube Feeding Tube Feeding Risk Ratio
Nutrition variable
Need for nasoenteric feeding tube — no. (%) NA 32 (31) — —
Dislodging of nasoenteric feeding tube — no./total no. (%)† 38/99 (38) 14/32 (44) 0.95 (0.77–1.16) 0.68
Obstruction of nasoenteric feeding tube — no./total no. (%)† 11/99 (11) 4/32 (12) 0.97 (0.70–1.33) 0.76
Need for insertion of nasogastric tube for decompression — no. (%)‡ 19 (19) 23 (22) 0.90 (0.62–1.30) 0.61
Need for parenteral nutrition — no. (%)‡ 5 (5) 10 (10) 0.66 (0.32–1.37) 0.28
Days from admission to full tolerance of oral diet§ — 0.001
Median 9 6
Gastrointestinal event — no. (%)¶
Nausea 32 (32) 37 (36) 0.91 (0.68–1.24) 0.66
Vomiting 19 (19) 26 (25) 0.82 (0.57–1.20) 0.31
Aspiration‖ 0 4 (4) — 0.12
Ileus** 10 (10) 11 (11) 0.96 (0.60–1.54) 1.00
Diarrhea 21 (21) 29 (28) 0.81 (0.57–1.17) 0.26
* NA denotes not applicable.

† Dislodging or obstruction of the nasogastric tube was noted in case-record forms by the attending physician or nurse. The denominator is
the number of patients who had a feeding tube inserted. Two patients in the early group declined tube feeding.
‡ The need for a nasogastric tube to be inserted for decompression or the need for parenteral nutrition was indicated by the attending physician.
§ Full tolerance of an oral diet was defined as tolerance of the oral diet without receipt of any other type of nasoenteric or parenteral nutrition.
¶ Gastrointestinal events were assessed during each day of the hospital stay.
‖ Data are for suspected aspiration as noted by a physician or nurse in the case-record forms.
** Ileus was diagnosed by the attending physician and noted in the case-record forms.
of the variables indicating inflammation (Fig. S3 even earlier. In a trial involving a small number
in the Supplementary Appendix). We did not of patients at one center, it would be possible to
evaluate gut permeability and bacterial translo- start nasogastric tube feeding some hours ear-
cation on the basis of the serum intestinal fatty lier by using a feeding tube that could be placed
acid–binding protein level or endotoxin expo- at the bedside. In daily practice, however, we
sure.17,50 Therefore, we cannot determine whether believe that an earlier start of tube feeding
gut permeability was influenced by early feeding would not be feasible. Starting an oral diet later
in a subset of our patients. Increased gut perme- in the on-demand group in order to increase the
ability and bacterial translocation may be restrict- difference in timing between the two study
ed to patients with acute pancreatitis who have groups would not be ethical because it would put
multiple organ failure,14 a subgroup that ac- patients at risk for malnutrition.
counted for only a small fraction of the patients A third explanation for the negative result
in this trial. However, a study of acute pancreati- may be that the study was too small to detect a
tis in which the rates of multiple organ failure difference between the two groups. To our
and death were similar to the rates in our study knowledge, this is the largest trial of nutrition
did show an increase in gut permeability and in patients with acute pancreatitis that has been
endotoxin exposure in most patients with severe performed so far, but the wide confidence inter-
acute pancreatitis.51 val for the primary end point may indicate that
Another possibility is that tube feeding in the an even larger trial is needed.
early group in our trial should have been started Fourth, the widely accepted scoring systems
1990 n engl j med 371;21 nejm.org november 20, 2014

277
for prediction of severity in acute pancreatitis In conclusion, our trial did not show the hy-
are only moderately accurate.52 In early-interven- pothesized benefit of early nasoenteric tube
tion studies in acute pancreatitis, it is therefore feeding in patients with acute pancreatitis who
unavoidable that mild or moderate disease will were at high risk for complications. The observa-
develop in a proportion of patients who were tion that the clinical outcomes of early tube
classified at presentation as having severe pan- feeding were similar to those of an oral diet initi-
creatitis. Nevertheless, at randomization, approx- ated at 72 hours, with tube feeding only if re-
imately one third of our patients had organ quired, challenges the concept of the gut mucosa–
failure and two thirds had SIRS. Organ failure is preserving effect of early enteral feeding during
one of the determinants of severe pancreatitis, acute pancreatitis.
and SIRS is increasingly recognized as an early
Supported by grants from the Netherlands Organization for
indicator of severe pancreatitis.30,45 Health Research and Development, the ZonMw Health Care Ef-
A feeding tube frequently causes discomfort, ficiency Research Program (170992902), and Nutricia (08/KR/
excessive gagging, or esophagitis and is often AB/002).
Disclosure forms provided by the authors are available with
dislodged or becomes obstructed, which neces- the full text of this article at NEJM.org.
sitates the replacement of the feeding tube.53,54 We thank Vera Zeguers for assistance as a study research
If tube feeding were restricted to patients who nurse, all the medical and nursing staff at the participating cen-
ters for assistance in the enrollment and care of the patients in
could not tolerate an oral diet, this would result this study, and the patients and their families for contributions
in substantial avoidance of discomfort and costs. to the study.
appendix
The authors’ full names and academic degrees are as follows: Olaf J. Bakker, M.D., Sandra van Brunschot, M.D., Hjalmar C. van Santvoort,
M.D., Ph.D., Marc G. Besselink, M.D., Ph.D., Thomas L. Bollen, M.D., Marja A. Boermeester, M.D., Ph.D., Cornelis H. Dejong, M.D., Ph.D.,
Harry van Goor, M.D., Ph.D., Koop Bosscha, M.D., Ph.D., Usama Ahmed Ali, M.D., Stefan Bouwense, M.D., Wilhelmina M. van
Grevenstein, M.D., Ph.D., Joos Heisterkamp, M.D., Ph.D., Alexander P. Houdijk, M.D., Ph.D., Jeroen M. Jansen, M.D., Ph.D., Thom M.
Karsten, M.D., Ph.D., Eric R. Manusama, M.D., Ph.D., Vincent B. Nieuwenhuijs, M.D., Ph.D., Alexander F. Schaapherder, M.D., Ph.D.,
George P. van der Schelling, M.D., Ph.D., Matthijs P. Schwartz, M.D., Ph.D., B.W. Marcel Spanier, M.D., Ph.D., Adriaan Tan, M.D.,
Ph.D., Juda Vecht, M.D., Ph.D., Bas L. Weusten, M.D., Ph.D., Ben J. Witteman, M.D., Ph.D., Louis M. Akkermans, M.D., Ph.D., Marco
J. Bruno, M.D., Ph.D., Marcel G. Dijkgraaf, Ph.D., Bert van Ramshorst, M.D., Ph.D., and Hein G. Gooszen, M.D., Ph.D.
The authors’ affiliations are as follows: the Department of Surgery, University Medical Center Utrecht, Utrecht (O.J.B., H.C.S., U.A.A.,
W.M.G., L.M.A.), the Departments of Gastroenterology and Hepatology (S.B.) and Surgery (M.G.B., M.A.B.) and the Clinical Research
Unit (M.G.D.), Amsterdam Medical Center, and the Department of Gastroenterology, Onze Lieve Vrouwe Gasthuis (J.M.J.), Amsterdam,
the Departments of Radiology (T.L.B.), Gastroenterology (B.L.W.), and Surgery (B.R.), St. Antonius Hospital Nieuwegein, Nieuwegein,
the Department of Surgery, Maastricht University Medical Center and NUTRIM School for Nutrition, Toxicology, and Metabolism,
Maastricht (C.H.D.), the Departments of Surgery (H.G., S.B.) and Operation Room/Evidence-Based Surgery (H.G.G.), Radboud Univer-
sity Medical Center, and the Department of Gastroenterology, Canisius Wilhelmina Hospital (A.T.), Nijmegen, the Department of
Surgery, Jeroen Bosch Hospital, Hertogenbosch (K.B.), the Department of Surgery, St. Elisabeth Hospital, Tilburg (J.H.), the Depart-
ment of Surgery, Medical Center Alkmaar, Alkmaar (A.P.H.), the Department of Surgery, Reinier de Graaf Gasthuis, Delft (T.M.K.), the
Department of Surgery, Medical Center Leeuwarden, Leeuwarden (E.R.M.), the Department of Surgery, University Medical Center
Groningen, Groningen (V.B.N.), the Department of Surgery, Leiden University Medical Center, Leiden (A.F.S.), the Department of Surgery,
Amphia Hospital, Breda (G.P.S.), the Department of Gastroenterology, Meander Medical Center, Amersfoort (M.P.S.), the Department of
Gastroenterology, Rijnstate Hospital, Arnhem (B.W.M.S.), the Department of Gastroenterology, Isala Clinics, Zwolle (J.V.), the Depart-
ment of Gastroenterology, Hospital Gelderse Vallei Ede, Ede (B.J.W.), and the Department of Gastroenterology and Hepatology, Erasmus
University Medical Center, Rotterdam (M.J.B.) — all in the Netherlands.
REFERENCES
1. Peery AF, Dellon ES, Lund J, et al. for the years 1995-2005. World J Gastro- 7. Besselink MG, van Santvoort HC,
Burden of gastrointestinal disease in the enterol 2013;19:3018-26. Boermeester MA, et al. Timing and im-
United States: 2012 update. Gastroenter- 4. Yadav D, Lowenfels AB. Trends in the pact of infections in acute pancreatitis. Br
ology 2012;143:1179-87. epidemiology of the first attack of acute J Surg 2009;96:267-73.
2. Fagenholz PJ, Castillo CF, Harris NS, pancreatitis: a systematic review. Pancreas 8. Wu BU, Johannes RS, Kurtz S, Banks
Pelletier AJ, Camargo CA Jr. Increasing 2006;33:323-30. [Erratum, Pancreas 2007; PA. The impact of hospital-acquired infec-
United States hospital admissions for 34:174.] tion on outcome in acute pancreatitis.
acute pancreatitis, 1988-2003. Ann Epide- 5. Singh VK, Bollen TL, Wu BU, et al. An Gastroenterology 2008;135:816-20.
miol 2007;17:491-7. assessment of the severity of interstitial 9. Rodriguez JR, Razo AO, Targarona J,
3. Spanier B, Bruno MJ, Dijkgraaf MG. pancreatitis. Clin Gastroenterol Hepatol et al. Debridement and closed packing for
Incidence and mortality of acute and 2011;9:1098-103. sterile or infected necrotizing pancreati-
chronic pancreatitis in the Netherlands: 6. Whitcomb DC. Acute pancreatitis. tis: insights into indications and outcomes
a nationwide record-linked cohort study N Engl J Med 2006;354:2142-50. in 167 patients. Ann Surg 2008;247:294-9.
n engl j med 371;21 278

10. van Santvoort HC, Bakker OJ, Bollen 24. Petrov MS, Pylypchuk RD, Uchugina 38. Kreymann KG, Berger MM, Deutz NE,
TL, et al. A conservative and minimally AF. A systematic review on the timing of et al. ESPEN guidelines on enteral nutri-
invasive approach to necrotizing pancreati- artificial nutrition in acute pancreatitis. tion: intensive care. Clin Nutr 2006;25:
tis improves outcome. Gastroenterology Br J Nutr 2009;101:787-93. 210-23.
2011;141:1254-63. 25. Sun JK, Mu XW, Li WQ, Tong ZH, Li J, 39. Besselink MG, van Santvoort HC,
11. van Santvoort HC, Besselink MG, Zheng SY. Effects of early enteral nutri- Buskens E, et al. Probiotic prophylaxis in
Bakker OJ, et al. A step-up approach or tion on immune function of severe acute predicted severe acute pancreatitis: a ran-
open necrosectomy for necrotizing pan- pancreatitis patients. World J Gastroen- domised, double-blind, placebo-controlled
creatitis. N Engl J Med 2010;362:1491-502. terol 2013;19:917-22. trial. Lancet 2008;371:651-9. [Erratum,
12. Al-Omran M, Albalawi ZH, Tashkandi 26. Wereszczynska-Siemiatkowska U, Lancet 2008;371:1246.]
MF, Al-Ansary LA. Enteral versus par- Swidnicka-Siergiejko A, Siemiatkowski A, 40. Gupta R, Patel K, Calder PC, Yaqoob
enteral nutrition for acute pancreatitis. Dabrowski A. Early enteral nutrition is P, Primrose JN, Johnson CD. A randomised
Cochrane Database Syst Rev 2010;1: superior to delayed enteral nutrition for clinical trial to assess the effect of total
CD002837. the prevention of infected necrosis and enteral and total parenteral nutritional
13. Ammori BJ, Leeder PC, King RF, et al. mortality in acute pancreatitis. Pancreas support on metabolic, inflammatory and
Early increase in intestinal permeability 2013;42:640-6. oxidative markers in patients with pre-
in patients with severe acute pancreatitis: 27. McClave SA, Martindale RG, Vanek dicted severe acute pancreatitis (APACHE
correlation with endotoxemia, organ fail- VW, et al. Guidelines for the provision and II > or =6). Pancreatology 2003;3:406-13.
ure, and mortality. J Gastrointest Surg assessment of nutrition support therapy 41. Oláh A, Pardavi G, Belágyi T, Nagy A,
1999;3:252-62. in the adult critically ill patient: Society of Issekutz A, Mohamed GE. Early nasojeju-
14. Besselink MG, van Santvoort HC, Re- Critical Care Medicine (SCCM) and Amer- nal feeding in acute pancreatitis is associ-
nooij W, et al. Intestinal barrier dysfunc- ican Society for Parenteral and Enteral ated with a lower complication rate. Nu-
tion in a randomized trial of a specific Nutrition (A.S.P.E.N.). JPEN J Parenter En- trition 2002;18:259-62.
probiotic composition in acute pancreati- teral Nutr 2009;33:277-316. 42. Petrov MS, Kukosh MV, Emelyanov
tis. Ann Surg 2009;250:712-9. 28. Meier R, Ockenga J, Pertkiewicz M, NV. A randomized controlled trial of en-
15. Fritz S, Hackert T, Hartwig W, et al. et al. ESPEN guidelines on enteral nutri- teral versus parenteral feeding in patients
Bacterial translocation and infected pan- tion: pancreas. Clin Nutr 2006;25:275-84. with predicted severe acute pancreatitis
creatic necrosis in acute necrotizing pan- 29. Mirtallo JM, Forbes A, McClave SA, shows a significant reduction in mortality
creatitis derives from small bowel rather Jensen GL, Waitzberg DL, Davies AR. Inter- and in infected pancreatic complications
than from colon. Am J Surg 2010;200: national consensus guidelines for nutri- with total enteral nutrition. Dig Surg 2006;
111-7. tion therapy in pancreatitis. JPEN J Par- 23:336-44.
16. Nieuwenhuijs VB, Verheem A, van enter Enteral Nutr 2012;36:284-91. 43. Singh VK, Wu BU, Bollen TL, et al.
Duijvenbode-Beumer H, et al. The role of 30. IAP/APA evidence-based guidelines for Early systemic inflammatory response
interdigestive small bowel motility in the the management of acute pancreatitis. syndrome is associated with severe acute
regulation of gut microflora, bacterial Pancreatology 2013;13:Suppl 2:e1-e15. pancreatitis. Clin Gastroenterol Hepatol
overgrowth, and bacterial translocation 31. Tenner S, Baillie J, DeWitt J, Vege SS. 2009;7:1247-51.
in rats. Ann Surg 1998;228:188-93. American College of Gastroenterology 44. Chan AW, Tetzlaff JM, Gøtzsche PC,
17. Rahman SH, Ammori BJ, Holmfield J, guideline: management of acute pancre- et al. SPIRIT 2013 explanation and elabo-
Larvin M, McMahon MJ. Intestinal hypo- atitis. Am J Gastroenterol 2013;108:1400- ration: guidance for protocols of clinical
perfusion contributes to gut barrier fail- 15. trials. BMJ 2013;346:e7586.
ure in severe acute pancreatitis. J Gastro- 32. Banks PA, Freeman MLy. Practice 45. Banks PA, Bollen TL, Dervenis C, et al.
intest Surg 2003;7:26-35. guidelines in acute pancreatitis. Am J Classification of acute pancreatitis —
18. Van Felius ID, Akkermans LM, Boss- Gastroenterol 2006;101:2379-400. 2012: revision of the Atlanta classifica-
cha K, et al. Interdigestive small bowel 33. Bakker OJ, van Santvoort HC, van tion and definitions by international con-
motility and duodenal bacterial over- Brunschot S, et al. Pancreatitis, very early sensus. Gut 2013;62:102-11.
growth in experimental acute pancreati- compared with normal start of enteral 46. Bakker OJ, van Santvoort H, Besselink
tis. Neurogastroenterol Motil 2003;15:267- feeding (PYTHON trial): design and ratio- MG, et al. Extrapancreatic necrosis without
76. nale of a randomised controlled multi- pancreatic parenchymal necrosis: a sepa-
19. Marik PE. What is the best way to feed center trial. Trials 2011;12:73. rate entity in necrotising pancreatitis? Gut
patients with pancreatitis? Curr Opin Crit 34. Knaus WA, Draper EA, Wagner DP, 2013;62:1475-80.
Care 2009;15:131-8. Zimmerman JE. APACHE II: a severity of 47. Heyland DK, Dhaliwal R, Drover JW,
20. McClave SA, Heyland DK. The physi- disease classification system. Crit Care Gramlich L, Dodek P. Canadian clinical
ologic response and associated clinical Med 1985;13:818-29. practice guidelines for nutrition support
benefits from provision of early enteral 35. Corfield AP, Cooper MJ, Williamson in mechanically ventilated, critically ill
nutrition. Nutr Clin Pract 2009;24:305-15. RC, et al. Prediction of severity in acute adult patients. JPEN J Parenter Enteral
21. Ziegler TR. Parenteral nutrition in the pancreatitis: prospective comparison of Nutr 2003;27:355-73.
critically ill patient. N Engl J Med 2009; three prognostic indices. Lancet 1985;2: 48. Koretz RL, Lipman TO. The presence
361:1088-97. 403-7. and effect of bias in trials of early enteral
22. Marik PE, Zaloga GP. Early enteral nu- 36. Neoptolemos JP, Kemppainen EA, nutrition in critical care. Clin Nutr 2014;
trition in acutely ill patients: a systematic Mayer JM, et al. Early prediction of sever- 33:240-5.
review. Crit Care Med 2001;29:2264-70. ity in acute pancreatitis by urinary tryp- 49. Casaer MP, Van den Berghe G. Nutri-
[Erratum, Crit Care Med 2002;30:725.] sinogen activation peptide: a multicentre tion in the acute phase of critical illness.
23. Li JY, Yu T, Chen GC, et al. Enteral study. Lancet 2000;355:1955-60. N Engl J Med 2014;370:1227-36.
nutrition within 48 hours of admission 37. Dickerson RN, Vehe KL, Mullen JL, 50. Pan L, Wang X, Li W, Li N, Li J. The
improves clinical outcomes of acute pan- Feurer ID. Resting energy expenditure in intestinal fatty acid binding protein diag-
creatitis by reducing complications: a meta- patients with pancreatitis. Crit Care Med nosing gut dysfunction in acute pancreati-
analysis. PLoS One 2013;8(6):e64926. 1991;19:484-90. tis: a pilot study. Pancreas 2010;39:633-8.
1992 n engl j med 371;21 279

51. Rahman SH, Ammori BJ, Larvin M, 52. Mounzer R, Langmead CJ, Wu BU, et of administration. Gut 1986;27:Suppl 1:
McMahon MJ. Increased nitric oxide ex- al. Comparison of existing clinical scor- 47-50.
cretion in patients with severe acute pan- ing systems to predict persistent organ 54. Marik PE. Aspiration pneumonitis
creatitis: evidence of an endotoxin medi- failure in patients with acute pancreati- and aspiration pneumonia. N Engl J Med
ated inflammatory response? Gut 2003; tis. Gastroenterology 2012;142:1476-82. 2001;344:665-71.
52:270-4. 53. Jones BJ. Enteral feeding: techniques Copyright © 2014 Massachusetts Medical Society.
n engl j med 371;21 280

Source: The New England Journal of Medicine T h e n e w e ng l a n d j o u r na l of m e dic i n e
c or r e sp ondence
Early versus On-Demand Tube Feeding in Pancreatitis

To the Editor: Bakker et al. (Nov. 20 issue)1 Definition, reporting, and interpretation of composite outcomes
in clinical trials: systematic review. BMJ 2010;341:c3920.
found no benefit of early enteral nutrition in pa- 4. Guo Q, Li A, Xia Q, et al. The role of organ failure and infec-
tients with predicted severe acute pancreatitis. tion in necrotizing pancreatitis: a prospective study. Ann Surg
Clinical trials involving patients with this condi- 2014;259:1201-7.
5. Besselink MG, van Santvoort HC, Boermeester MA, et al.
tion are hampered by the low positive predictive Timing and impact of infections in acute pancreatitis. Br J Surg
value of current prognostic scoring systems, re- 2009;96:267-73.
sulting in the inclusion of many patients who DOI: 10.1056/NEJMc1415356
ultimately have mild acute pancreatitis and do

not require early enteral nutrition.2 A composite To the Editor: The study by Bakker et al. requires
end point allowed for sample-size reduction but careful interpretation. First, the actual delivery
ultimately resulted in an underpowered study, site of enteral nutrition is unclear. The authors
owing to the inequality between the individual did not disclose in the article that the original
end points.3 Death and infection have vastly dif- ethics committee approval was for tube feeding
ferent clinical significance, given that persistent into the stomach, not the jejunum. Furthermore,
organ failure, not infection, is the primary cause feeding tubes were dislodged in 40% of the pa-
of death in patients with severe acute pancreati- tients, which many would consider to be a rather
tis.4,5 The timing, type, and volume of fluid ad- high rate for leading research centers. Second,
ministered were not detailed in this study. Indi- only one third of the study patients had actual, as
vidual centers across this consortium may have opposed to predicted, severe or critical acute
different practices with regard to fluid resuscita- pancreatitis (i.e., persistent organ failure, infect-
tion that potentially biased the results toward the ed pancreatic necrosis, or both).1 Hence, two
null hypothesis. Enteral nutrition has consistent- thirds of the patients in the study were not posed
ly been shown to have a benefit in patients with to benefit from tube feeding.2 Third, the lack of
severe acute pancreatitis, but we are still no closer superiority of feeding within 24 hours after pre-
to optimizing patient selection and the timing of sentation versus feeding at 72 hours (or later) af-
enteral nutrition. ter presentation does not mean that the latter
Robert A. Moran, M.D. should be deemed the preferred nutritional strat-
Ruben Hernaez, M.D., Ph.D. egy. The weight of evidence indicates that the
Vikesh K. Singh, M.D. most effective time to start feeding in patients
Johns Hopkins Medical Institutions with acute pancreatitis is between 24 and 72
Baltimore, MD hours after presentation, and the exact timing of
vsingh1@jhmi.edu
feeding in an individual patient is influenced by
ported. the adequacy of intravenous fluid resuscitation
1. Bakker OJ, van Brunschot S, van Santvoort HC, et al. Early and opiate use.3-5
versus on-demand nasoenteric tube feeding in acute pancreati- Maxim S. Petrov, M.D., Ph.D.
tis. N Engl J Med 2014;371:1983-93.
2. Mounzer R, Langmead CJ, Wu BU, et al. Comparison of ex-
John A. Windsor, M.B., Ch.B., M.D.
isting clinical scoring systems to predict persistent organ failure University of Auckland
in patients with acute pancreatitis. Gastroenterology 2012;142: Auckland, New Zealand
1476-82. max.petrov@gmail.com
3. Cordoba G, Schwartz L, Woloshin S, Bae H, Gøtzsche PC. No potential conflict of interest relevant to this letter was re-
ported.

281 february 12, 2015 683
1. Dellinger EP, Forsmark CE, Layer P, et al. Determinant-based In response to Petrov and Windsor: the first
classification of acute pancreatitis severity: an international
multidisciplinary consultation. Ann Surg 2012;256:875-80. version of the protocol indeed included nasogas-
2. Petrov MS. Gastric feeding and “gut rousing” in acute pan- tric instead of nasojejunal feeding. After critical
creatitis. Nutr Clin Pract 2014;29:287-90. appraisal of the available evidence at the time,
3. Petrov MS, Pylypchuk RD, Uchugina AF. A systematic review
on the timing of artificial nutrition in acute pancreatitis. Br J Nutr we decided to switch to nasojejunal feeding to
2009;101:787-93. minimize the risk of aspiration. This decision
4. Sun JK, Li WQ, Ke L, et al. Early enteral nutrition prevents was made before the start of patient recruitment
intra-abdominal hypertension and reduces the severity of severe
acute pancreatitis compared with delayed enteral nutrition: and hence did not influence outcome (for details,
a prospective pilot study. World J Surg 2013;37:2053-60. see www.isrctn.com/ISRCTN18170985). The rate
5. Petrov MS, Windsor JA. Nutritional management of acute of tube dislocation is similar to rates found in
pancreatitis: the concept of ‘gut rousing.’ Curr Opin Clin Nutr
Metab Care 2013;16:557-63. the literature.3,4
DOI: 10.1056/NEJMc1415356 Our study did not show that starting an oral
diet 72 hours after presentation is the most ef-
fective strategy for all patients with acute pan-
The Authors Reply: The limited accuracy of creatitis. However, our results show that routine
prognostic scoring systems in acute pancreatitis early tube feeding in all patients at high risk for
is addressed in the letters by Moran et al. and by severe pancreatitis does not improve outcome
Petrov and Windsor. However, currently there is and that the implementation of on-demand tube
no better tool available during triage, when clini- feeding will reduce patient discomfort and costs.
cians need to make decisions regarding any type Olaf J. Bakker, M.D.
of early intervention. This limitation does not
University Medical Center Utrecht
undermine the validity of the study results. The Utrecht, the Netherlands
sample size was estimated on the basis of studies
Marc G. Besselink, M.D., Ph.D.
that used similar predictive scoring systems in
similar patient populations with similar clinical Amsterdam Medical Center
Amsterdam, the Netherlands
outcomes.1 Furthermore, this approach reflects
how clinical decision making for patients with Hein G. Gooszen, M.D., Ph.D.
acute pancreatitis is done in everyday practice Radboud University Medical Center
Nijmegen, the Netherlands
and therefore, in our opinion, is relevant to prac-
Since publication of their article, the authors report no fur-
ticing clinicians. ther potential conflict of interest.
Moran et al. question the composition of the
Bakker OJ, van Santvoort HC, van Brunschot S, et al. Pancre-
primary end point. Our study was designed to 1. atitis, very early compared with normal start of enteral feeding
show a reduction in the rate of infection because (PYTHON trial): design and rationale of a randomised con-
infection has a major effect on the outcome of trolled multicenter trial. Trials 2011;12:73.
Wu BU, Johannes RS, Kurtz S, Banks PA. The impact of
patients with acute pancreatitis.2 No significant 2.hospital-acquired infection on outcome in acute pancreatitis.
differences were found in the individual compo- Gastroenterology 2008;135:816-20.
nents of the primary end point, which strongly 3. Seder CW, Stockdale W, Hale L, Janczyk RJ. Nasal bridling
suggests that the proposed inequality does not decreases feeding tube dislodgment and may increase caloric in-
take in the surgical intensive care unit: a randomized, controlled
hamper the overall power of the study. trial. Crit Care Med 2010;38:797-801.
In the complex setting of a large, multicenter 4. Gerritsen A, Besselink MG, Cieslak KP, et al. Efficacy and
trial, subtle differences in treatment among cen- complications of nasojejunal, jejunostomy and parenteral feeding
after pancreaticoduodenectomy. J Gastrointest Surg 2012;16:
ters may occur, including with regard to fluid 1144-51.
resuscitation. However, a randomized study de- DOI: 10.1056/NEJMc1415356
sign and stratification according to study center
balance out such differences between treatmen
groups.
684 n engl j med 372;7 282

ICU-ACQUIRED WEAKNESS AND RECOVERY

FROM CRITICAL ILLNESS
Critical illness can continue to reverberate in patients’ lives long after discharge. One such sequel is ICU-
acquired weakness, which can affect overall physical function, and is particularly troublesome in the elderly.
The following review offers guidance for understanding and evaluating this problem.
283
Case Challenge
ICU-Acquired Weakness and Recovery from Critical Illness

A 77-year-old man is on mechanical ventilation in the ICU after an emergency colon resection, complicated
by septic shock and acute liver failure. Since it appears his stay in the ICU will be prolonged, what measures
would you take to optimize his long-term recovery?
A well-nourished 77-year-old man whose medical history includes treated hypertension and hypercholesterol-
emia, previous heavy alcohol intake, and mild cognitive impairment is admitted to the intensive care unit (ICU)
of a university hospital from the operating room after a Hartmann’s procedure (resection of the rectosigmoid
colon with closure of the rectal stump and formation of an end colostomy) performed for fecal peritonitis due to
a perforated sigmoid colon. On arrival in the ICU, he was in septic shock. He is undergoing mechanical ventila-
tion with the use of a low-tidal-volume protocol with positive end-expiratory pressure (PEEP). His arterial blood
pressure is supported with a norepinephrine infusion. Analgesia is provided by a continuous morphine infusion.
Enteral nutrition was started on the day after ICU admission, and target intake was achieved on day 6. Parenteral
nutrition was not used. (In the previous installment of this case, there were 2906 votes on strategies for feeding
this critically ill patient. A majority of respondents [53%] favored initiating enteral nutrition within 24 to 48 hours
after ICU admission and then starting parenteral nutrition on day 7 if the caloric target was not being met.
284
nother 30% favored initiating parenteral nutrition as soon as possible after the patient’s arrival in the ICU and
A
then starting enteral nutrition once bowel sounds return, whereas 11% favored awaiting the return of bowel
sounds and then initiating enteral nutrition. Only 4% favored initiating total parenteral nutrition as soon as possi-
ble after the patient’s arrival in the ICU.)

Since it is likely that this patient will have a prolonged stay in the ICU, what measures would you take to
optimize his long-term recovery?
A. Maintain a blood glucose level of 80 to 110 mg per deciliter (4.4 to 6.1 mmol per liter) throughout his ICU stay.
B. Administer human growth hormone to improve nitrogen balance and preserve muscle mass and strength.
C. Minimize sedation and institute early active and passive mobilization even while he is still undergoing
mechanical ventilation.
285
review article
ICU-Acquired Weakness and Recovery

from Critical Illness
John P. Kress, M.D., and Jesse B. Hall, M.D.
W
From the Department of Medicine, Section eakness acquired in the intensive care unit (ICU) is caused
of Pulmonary and Critical Care, University by many different pathophysiological mechanisms that are not mutually
of Chicago, Chicago. Address reprint re-
quests to Dr. Kress at the University of Chi- exclusive. This is not surprising, given the diverse diseases that precipi-
cago, Department of Medicine, Section of tate critical illness, the drugs used during its management, and the consequences
Pulmonary and Critical Care, 5841 S. Mary- of protracted immobility. Nonetheless, conceptualization of this entity is valuable,
land Ave., MC 6026, Chicago, IL 60637, or at
jkress@medicine.bsd.uchicago.edu. since weakness in survivors of critical illness is common and is associated with long-
standing consequences that dramatically affect recovery. Moreover, as survival rates
N Engl J Med 2014;370:1626-35.
DOI: 10.1056/NEJMra1209390 among patients in the ICU increase, ICU-acquired weakness will have increasing
Copyright © 2014 Massachusetts Medical Society. relevance for care providers outside the ICU. This article provides an overview of the
condition and its effect on recovery after critical illness.
Early descriptions of weakness in critical illness include reports by Osler1 of neuro-
muscular dysfunction in patients with sepsis and reports by Olsen2 of peripheral
neuropathy complicating protracted coma. In 1977, myopathy was described in a
patient with status asthmaticus3 who received high doses of hydrocortisone and
simultaneous neuromuscular blockade.
These descriptions were followed by a landmark study by Bolton and colleagues4
of polyneuropathy in patients in the ICU. This condition, termed critical illness poly-
neuropathy, is characterized by a primary axonal degeneration, without demyelination,
that typically affects motor nerves more than sensory nerves. The predominant spinal
cord finding is loss of anterior horn cells due to axonal degeneration.5 Electro-
physiological studies show that nerve conduction velocity is preserved, but there
are reductions in the amplitudes of compound muscle action potentials (CMAPs)
and sensory-nerve action potentials. Over the past two decades, improvements in
survival after discharge from the ICU probably have led to increased awareness of
ICU-acquired weakness. The magnitude of neuromuscular impairment in the in-
creasing population of patients undergoing post-ICU rehabilitation has come to
the attention of health care providers, patients, and families.
Fe at ur e s of ICU-Ac quir ed W e a k ne ss
Critical illness polyneuropathy affects the limbs (particularly the lower extremities)
in a symmetric pattern. Weakness is most notable in proximal neuromuscular areas
(e.g., the shoulders and hip girdle). In addition, involvement of the respiratory
muscles can occur and can impede weaning from mechanical ventilation.6 Facial
and ocular muscles are rarely involved. Creatine kinase levels are not elevated in
patients with critical illness polyneuropathy. The condition is distinct from the
Guillain–Barré syndrome because demyelination is not seen.
As distinct from a secondary myopathy resulting from muscle denervation,
critical illness myopathy is a primary myopathy.7 It is often difficult to distinguish
this myopathy from critical illness polyneuropathy by means of a simple bedside
1626 286
The New
Society. Forrights
reserved.
examination. Both conditions are manifested by

Table 1. Electrophysiological Features of ICU-Acquired
limb and respiratory-muscle weakness and retained Weakness.*
sensory function; however, because altered mental
status is common in patients in the ICU, sensory Critical illness polyneuropathy
examination may be difficult to perform. In pa- Normal to minimally reduced nerve conduction
velocity
tients with critical illness myopathy, electrophysi-
ological studies show reduced amplitude and Reduced CMAP amplitude
increased duration of CMAPs. Histopathological Reduced compound SNAP amplitude
analysis shows selective loss of thick filaments Critical illness myopathy
in muscle, which reflects myosin loss as well as Normal to minimally reduced nerve conduction
muscle necrosis.8 Creatine kinase levels may be velocity
mildly elevated, except that in occasional cases of Reduced CMAP amplitude
necrotizing myopathy, such levels are markedly Reduced muscle excitability on direct stimulation
elevated.
Increased CMAP duration
Though it is tempting to categorize weakness
Normal SNAP
after recovery from critical illness as either my-
opathy or neuropathy, there is evidence of overlap * CMAP denotes compound muscle action potential, and
between these pathophysiological processes.9,10 SNAP sensory-nerve action potential.
Critical illness myopathy occurs more frequently
than critical illness neuropathy, and it is associ-
ated with a higher rate of recovery.9 Further- the term “ICU-acquired weakness” be applied in
more, although specific polyneuropathies, myop- cases in which a patient is noted to have clini-
athies, or both contribute to physical dysfunction cally detected weakness with no plausible cause
in critically ill patients, other variables, such as other than critical illness.
drug effects (e.g., from the use of glucocorti- ICU-acquired weakness, with documented poly-
coids11 or neuromuscular blocking agents12), meta- neuropathy, myopathy, or both, can be subclas-
bolic effects (e.g., hyperglycemia13), joint contrac- sified. Critical illness polyneuropathy refers to
tures,14 and muscle wasting from catabolism and ICU-acquired weakness with electrophysiological
physical inactivity also contribute to ICU-acquired evidence of an axonal polyneuropathy. Critical
weakness.15 Table 1 lists the electrophysiological illness myopathy refers to ICU-acquired weak-
features of ICU-acquired weakness. ness with myopathy that is documented electro-
physiologically or histologically. Critical illness
neuromyopathy refers to electrophysiological or
Di agnosis of ICU-Ac quir ed
W e a k ne ss histologic findings of both critical illness poly-
neuropathy and critical illness myopathy.18
Given the complexities of ICU-acquired weak- The diagnosis of ICU-acquired weakness is
ness, a standard definition and list of nosologic made with the use of the Medical Research
features are both elusive and difficult to apply Council (MRC) scale for grading the strength of
readily in clinical practice. For instance, electro- various muscle groups in the upper and lower
myography and nerve-conduction studies are dif- extremities. The scale ranges from 0 to 5, with
ficult to perform in many critically ill patients. higher scores indicating greater muscle strength19;
The former requires that patients are awake and a combined score of less than 48 is diagnostic
able to contract their muscles voluntarily; patients of ICU-acquired weakness.11 Patients with ICU-
in the ICU are often not able to do so. Nerve- acquired weakness according to the MRC exami-
conduction studies can be confounded by prob- nation should undergo serial evaluations, and if
lems such as tissue edema. There is controversy persistent deficits are noted, electrophysiological
regarding the need — outside of research set- studies, muscle biopsy, or both are warranted. Pa-
tings — to establish the diagnosis of ICU- tients with persistent coma after discontinuation
acquired weakness on the basis of formal elec- of sedation should undergo studies of the cen-
trophysiological criteria.16,17 The nosologic scheme tral nervous system such as cranial computed
proposed by Stevens and colleagues18 is a reason- tomography or magnetic resonance imaging. If
able classification. These investigators suggest that such studies are normal, electrophysiological

287 april 24, 2014 1627
The New
Medicine
by MIKE MULARCZYK
JanuarySociety.
15, 2015.
AllFor personal
reserved.
Table 2. Diseases and Syndromes Causing Weakness in Patients in the ICU. Pathoph ysiol o gic a l Mech a nisms
Condition and Pathogenesis Clinical Manifestations The pathophysiological mechanisms of ICU-
Guillain–Barré syndrome acquired weakness are believed to be multi-
Autoimmune mechanism Diarrheal prodrome (e.g., Campylobacter factorial. In critical illness polyneuropathy,
jejuni infection) there is speculation that a dysfunctional micro-
Myelin loss Autonomic instability, ascending paralysis, circulation leads to neuronal injury and axonal
ventilatory failure degeneration.20 Supporting this notion is the
Myasthenia gravis observation that patients with critical illness
Autoimmune mechanism Bulbar palsy polyneuropathy have E-selectin expression in the
Antibodies against neuro- Ventilatory failure
peripheral-nerve vascular endothelium, which sug-
muscular junction gests endothelial-cell activation as described in
Porphyria models of sepsis, with microvascular leak and
alterations in the microvascular environment.21
Porphyrin accumulation Motor axonal neuropathy, autonomic in-
with resulting nerve stability, abdominal pain, psychiatric Hyperglycemia may exacerbate this problem,
toxicity manifestations perhaps by inducing neural mitochondrial dys-
Eaton–Lambert syndrome function.22 Indeed, there is evidence that aggres-
Paraneoplastic syndrome Proximal weakness sive glucose control may reduce the risk of criti-
cal illness polyneuropathy (and myopathy) and
Presynaptic calcium- Autonomic instability result in a decreased need for prolonged mechan-
channel antibodies ical ventilation.13,23 However, because of the in-
Amyotrophic lateral sclerosis creased risk of adverse events associated with
Upper and lower motor- Bulbar involvement, fasciculations, aggressive glucose control,24,25 it is not recom-
neuron degeneration muscle atrophy mended solely for the prevention of critical ill-
Vasculitic neuropathy ness polyneuropathy.
Occlusion of vasa Paresthesia, numbness, pain, weakness Rich and Pinter26 suggest that inactivation of
nervorum vessels sodium channels may contribute to critical illness–
Cell and immune-complex Distal polyneuropathy related neuropathy, myopathy, or both. They re-
mediated ported changes in fast sodium channels leading
Cervical myelopathy to inexcitability of muscle fibers in an animal
Mechanical injury to Paresthesia, numbness, pain, weakness model of critical illness myopathy. The same in-
cervical spinal cord vestigators reported electrocardiographic abnor-
Botulism malities, including decreased QRS amplitudes and
Botulinum toxin Bowel dysfunction, bladder dysfunction, or increased QRS duration,27 in patients with septic
both, inhibition of acetylcholine release shock; in patients who recovered from septic shock,
from presynaptic membrane, bulbar the electrocardiographic findings returned to nor-
palsy, descending paralysis, autonomic
dysfunction mal. These observations, which are consistent with
the notion of a channelopathy, are intriguing
because of the strong association between severe
studies, a muscle biopsy, or both should be per- sepsis and ICU-acquired weakness.
formed. The MRC scale has important limita- Other conditions that are generic to the ICU
tions, such as poor discrimination and a poten- may adversely affect the peripheral nervous system.
tial ceiling effect. Better bedside tools are needed A catabolic state with skeletal-muscle wasting is
to more precisely identify the presence of ICU- common and occurs rapidly in critically ill pa-
acquired weakness. Of course, weakness can tients, particularly those with sepsis.28 Greater
have a broad differential diagnosis. A detailed mitochondrial dysfunction has been observed in
review of specific diseases leading to weakness the skeletal muscle of patients who died from
in patients in the ICU is beyond the scope of this critical illness than in that of survivors of criti-
article. Table 2 summarizes common diseases that cal illness.29 Systemic inflammation and oxida-
cause weakness in these patients, and Figure 1 tive stress are associated with reduced muscle
shows a diagnostic algorithm for the evaluation strength.30 Diaphragm wasting, which has been
of weakness in critically ill patients. extensively described in animal models of full
1628 n engl j med

288370;17 nejm.org april 24, 2014
The New
Society. Forrights
reserved.
Interrupt use of sedatives and analgesics
Patient awake and able to follow instructions?
Yes No
Continue to withhold
Assess MRC score
sedatives or analgesics
Score <48, indicating Patient remains

Normal findings Focal deficit
symmetric weakness unresponsive
Serial examinations, Deficit in peripheral Deficit in central

physical and occupational neuromuscular system nervous system
therapy
CT of brain, MRI of brain, EEG,

CSF studies
Physical and occupational No change in condition
Improvement
therapy, observation or deterioration
NCS, EMG, muscle biopsy
Preserved nerve conduction velocity, Reduced amplitude and increased

no demyelination duration of compound muscle action
Reduced amplitude of motor and potentials
compound sensory-nerve action Selective loss of thick filaments in muscle
potentials and muscle necrosis
Conditions may coexist

(critical illness neuromyopathy)
Critical illness polyneuropathy Critical illness myopathy
Figure 1. Diagnostic Algorithm for the Evaluation of Weakness in Critically Ill Patients.
The Medical Research Council (MRC) scale grades the strength of various muscle groups in the upper and lower extremities from 0 to 5,
with higher scores indicating greater muscle strength; a combined score of less than 48 is diagnostic of ICU-acquired weakness. CT denotes
computed tomography, CSF cerebrospinal fluid, EEG electroencephalography, EMG electromyography, MRI magnetic resonance imaging,
and NCS nerve conduction study.
mechanical ventilation,31 has also been described creased levels of myosin heavy chain, and increased
in studies in humans.32,33 In the diaphragm, atrophic protein kinase B–forkhead box subgroup
mechanical ventilation is associated with increased O (AKT-FOXO) signaling.34 Levine and colleagues32
activity of the ubiquitin–proteasome system, de- noted atrophy in diaphragm myofibers within as

april 24, 2014 1629
The New England
Medical personal
Society. use only.
reserved.
system failure are important risk factors.11,37 Hyper-

Case Control
glycemia in the ICU was reported to be an inde-
A B
pendent risk factor in one trial.13 Glucocorticoid
therapy has been reported to be a risk factor in
some trials.8 For instance, De Jonghe et al.11 and
Fiber Size
Herridge et al.38 reported that glucocorticoid use
during an ICU stay predicted ICU-acquired weak-
50 µm 50 µm ness. Other trials have not identified glucocorti-
coid administration as a risk factor.22,28 In fact,
C D Hermans et al.23 noted that glucocorticoids were
associated with a protective effect with regard to
Slow Myosin
Heavy Chain ICU-acquired weakness. Despite the inconsistent
findings in this area, it seems reasonable to min-
50 µm 50 µm
imize the use of glucocorticoids in patients in the
E F ICU, given the numerous known complications
associated with these drugs. Neuromuscular
Fast Myosin blocking agents used in conjunction with gluco-
Heavy Chain corticoids are associated with muscle weakness
50 µm 50 µm in patients with status asthmaticus.3,12 However,
Papazian et al.39 reported improved sur vival with-
out an increase in ICU-acquired weakness with
Figure 2. Changes in Fiber Size and Expression of Slow-Twitch and Fast-
Twitch MyosinICM HeavyAUTHOR:
Chains inKress Mechanical1stVentilation.
Patients Undergoing RETAKE short-term use of neuromuscular blocking agents
Shown are representative
REG F FIGURE: 1 of 3
diaphragm-biopsy specimens (hematoxylin
2nd
and (i.e., <48 hours) in patients with a severe form of
3rd
eosin) from patients
CASE undergoing mechanical ventilation Revised
(duration of mechan- the acute respiratory distress syndrome (ARDS).
in case patients, 18 to Line
ical ventilationEMail 69 hours;4-C
duration SIZE
in controls, 2 to The reason for these findings is not clear, though
ARTIST: ts H/T H/T
3 hours). As compared
Enon with diaphragm-biopsy specimens22p3obtained from some researchers have proposed that early neuro-
Combo
controls, specimens obtained from case patients showed smaller slow-twitch
AUTHOR, PLEASE NOTE: 32 muscular blockade may protect the diaphragm
and fast-twitch fibers. Reproduced from Levine et al.
Figure has been redrawn and type has been reset.
Please check carefully.
from injury due to patient–ventilator dyssynchrony.
In a study reported by De Jonghe and col-
little as 18 hours after complete diaphragmatic
ISSUE: xx-xx-xx leagues,11 ICU-acquired weakness was more than
inactivity in humans (Fig. 2). Jaber and colleagues33 four times as likely to develop in women as in
reported a loss of diaphragmatic strength in pa- men. Again, the reasons for this finding are not
tients in the ICU within hours after the initiation clear, though the investigators speculated that
of mechanical ventilation, which progressed over a the smaller muscle mass in women might be a
period of 6 days of ongoing measurement. Muscle predisposing factor.
inactivity is a potent stimulus for protease activa- There is some evidence that immobility in the
tion leading to muscle protein breakdown and ICU is associated with muscle wasting. In the con-
activation of the ubiquitin–proteasome pathway of ventional model of ICU care for patients with the
proteolysis. Immobility of the extremities, which is most severe illness, patients are inactive and bed-
common in patients in the ICU, is related to un- ridden for a prolonged period, particularly during
derlying illness in conjunction with prescribed bed treatment with mechanical ventilation, and pro-
rest and sedative-related immobility. The problems longed sedation and immobilization are tradi-
related to bed rest, including severe neuromus- tionally the rule. As noted above, inactivity of
cular deconditioning, have been recognized for
many decades.35 Figure 3 shows the pathophysi- Figure 3 (facing page). Pathophysiological Mechanisms
ological mechanisms of ICU-acquired weakness. of ICU-Acquired Weakness.
Panel A shows skeletal-muscle wasting. Possible mecha-
R isk Fac t or s nisms include microvascular ischemia, catabolism, and
immobility. Panel B shows polyneuropathy with axonal
degeneration. Possible mechanisms include microvas-
The reported incidence of ICU-acquired weakness cular injury with resulting nerve ischemia, dysfunction
ranges from 25 to 100%.11,36 It is clear that sepsis, of sodium channels, and injury to nerve mitochondria.
persistent systemic inflammation, and multiorgan
1630 290
The New
Society. Forrights
reserved.
Massachusetts
to Table ofMedical
ContentsSociety. All rights reserved.
COLOR FIGURE
Draft 3 04/03/14 1631

april 24, 2014
Author Kress
Fig # 2 The New England
personal Medicine
Downloaded from
Title nejm.org by MIKE Copyright
reserved.
ME Copyright © 2014 Massachusetts Medical Society. All rights reserved.
DE Back to Table of Contents
Artist SBL
cal risk factors and pathophysiological features

Table 3. Clinical Risk Factors and Pathophysiological Features of Critical
Illness Polyneuropathy and Critical Illness Myopathy. of critical illness polyneuropathy and myopathy
are summarized in Table 3.
Variable Reference
Clinical risk factors of both critical R ec ov er y from Cr i t ic a l Il l ne ss
illness polyneuropathy and
critical illness myopathy
Survival among patients in the ICU has improved
Female sex De Jonghe et al.11
dramatically over the past 20 years. Indeed, the
Sepsis Garnacho-Montero et al.28 observation that extremely ill patients often sur-
Catabolic state Trojaborg et al.,15 Garnacho-Montero vive their illnesses has led to a relatively new fo-
et al.28
cus of clinical investigation on patients who sur-
Multiorgan system failure De Jonghe et al.11 vive critical illness. Rehabilitation after critical
Systemic inflammatory response Jaber et al.,33 Levine et al.34 illness is arduous and often frustratingly slow,
syndrome
particularly in elderly patients. The greatest burdens
Long duration of mechanical De Jonghe et al.11 that survivors of critical illness face are related
ventilation
to neuromuscular dysfunction and neuropsycho-
Immobility Levine et al.,32 Papazian et al.,39 logical maladjustment.42,43 Patients who survive
Iwashyna et al.41
respiratory failure, circulatory failure (e.g., in as-
Hyperglycemia Van den Berghe et al.13
sociation with ARDS42 or sepsis10), or both seem
Glucocorticoids De Jonghe et al.11 to have these problems with the greatest frequen-
Neuromuscular blocking agents MacFarlane and Rosenthal,3 cy and intensity. Indeed, in patients who require
Leatherman et al.12 prolonged mechanical ventilation, neuromuscu-
Pathophysiological processes lar recovery is typically prolonged and incom-
Critical illness polyneuropathy plete. Studies show that up to 65% of such pa-
Motor nerves affected more than Bolton et al.4 tients have functional limitations after discharge
sensory nerves from the hospital.38,44 Neuromuscular abnormal-
Secondary denervation muscle Bolton et al.4 ities may last for many years in some patients.45
injury (myopathy) As mentioned earlier, in the past, routine
Proposed mechanisms features of general care provided in the ICU in-
Nerve ischemia Bolton20 cluded liberal use of sedation and immobiliza-
Nerve microvascular injury Bolton,20 Fenzi et al.21 tion of the patient, which were thought to be
Nerve mitochondrial injury Van den Berghe et al.22
necessary for facilitating interventions to nor-
malize physiological function by artificial means.
Sodium channelopathy Rich and Pinter26
Recently, there has been a paradigm shift away
Critical illness myopathy from this approach toward a “less is more” phi-
Primary myopathy — selective Derde et al.8 losophy for patients in the ICU.46
myosin loss, muscle necrosis
(e.g., ubiquitin–proteasome
This paradigm shift is consistent with the
proteolysis) observation that long-term physical problems in
Mitochondrial dysfunction Carré et al.29 survivors of critical illness, particularly those
Oxidative stress Reid and Moylan30
with respiratory failure, may result from the
protracted ICU stay and period of immobiliza-
Sodium channelopathy Rich and Pinter26
tion during which the patient is receiving organ
support that is essential for survival. Functional
disability, a natural consequence of weakness, is
the diaphragm is associated with rapid atrophy a frequent and long-lasting complication in sur-
and dysfunction.32,33 Griffiths et al.40 evaluated vivors of critical illness.41 Herridge et al.38 fol-
the effects of continuous passive range of motion lowed a cohort of survivors of ARDS for 5 years
in one leg in critically ill patients with respira- after hospital discharge. Muscle weakness and
tory failure who were receiving neuromuscular functional impairment were frequently observed
blockade, with the contralateral leg serving as at 1 year, and recovery from physical dysfunction
the control. In the mobilized leg, muscle-fiber was incomplete even 5 years after discharge.42 At
atrophy and wasting were attenuated. The clini- the 5-year assessment, the results of 6-minute
1632 292
The New
Society. Forrights
reserved.
walk tests were still only 70% of the predicted point that ambulation during mechanical venti-
results. Elderly survivors of critical illness ap- lation44,52,53 and even extracorporeal membrane
pear to fare particularly poorly. Sacanella et al.47 oxygenation therapy54 is now feasible. Protocols
reported a significant reduction in functional designed to mobilize patients in the ICU consist
status in elderly survivors of critical illness who of sequential strategies from lesser to greater
had had normal functional status at ICU admis- complexity of activities, depending on a patient’s
sion; functional status had not returned to base- ability to perform these activities. Such strate-
line 1 year after discharge. gies are similar to the approach used by physical
The focus on rehabilitation of critically ill therapists who care for patients outside the ICU.
patients should begin in the ICU and continue Bailey and colleagues52 reported on mobiliza-
all the way to recovery at home. This is particu- tion in a large cohort of patients who required
larly important because the burden of illness mechanical ventilation. These investigators de-
affects not only the patient but his or her family scribed a “culture change,” whereby activity was
or other caregivers as well. Accordingly, preparing encouraged as soon as the condition of patients
both the patient and family or other caregivers was hemodynamically stable with modest ventila-
for the difficult task of recovery may help lessen tor settings (i.e., fraction of inspired oxygen ≤0.6
this burden. and positive end-expiratory pressure ≤10 cm of
Given the substantial physical dysfunction water). Ambulation, even during mechanical ven-
noted in survivors of critical illness, several in- tilation, was feasible with few adverse events
vestigators have evaluated the effect of ICU care (e.g., device removal or deterioration of vital
that is focused on optimizing early physical ac- signs). Burtin and colleagues55 reported the ben-
tivity in spite of the severity of illness. This efits of a cycling exercise session with the use of
strategy involves minimizing sedation and en- a bedside cycle ergometer attached to the foot of
listing the early involvement of a diverse, multi- the bed. The ergometer could provide passive
disciplinary group of clinicians, including physical range-of-motion exercise for patients who were
and occupational therapists, nurses, respiratory not sufficiently alert to participate and active
therapists, and patient care technicians, with the resistance for those who were capable of pedal-
goal of getting patients up and out of bed. The ing. These investigators noted improved strength
discontinuation of deep sedation is a critical first and physical function at hospital discharge in the
step in optimizing patient activity and aware- patients who were randomly assigned to cycle
ness. For example, studies have shown that the ergometry.
interruption of daily sedative use during me- Morris et al.53 reported the results of a pro-
chanical ventilation increases the percentage of spective, nonrandomized trial of mobilization in
days during which patients are awake and able patients who were receiving mechanical ventila-
to follow commands.48 De Jonghe et al.49 noted tion. The mobilization group was out of bed
that use of a sedation algorithm designed to al- faster and had shorter hospital stays. There were
low patients to be more alert was associated no differences in the duration of mechanical
with a 50% reduction in pressure sores, presum- ventilation or in the condition of the patient at
ably because of reductions in sedative-related hospital discharge. At 1 year after discharge, the
immobilization. In a before-and-after quality- mobilization group was almost half as likely to
improvement project, Needham et al.50 reported have died or been rehospitalized; this suggests
that changes in ICU care to reduce sedative use a better long-term recovery profile.56 In 2009,
improved in-hospital activity levels in a group of Schweickert et al.44 performed a prospective ran-
patients in the medical ICU. domized, blinded trial of physical and occupa-
Currently, therapeutic interventions to pre- tional therapy from the inception of respiratory
vent or attenuate ICU-acquired weakness and failure. This immediate mobilization strategy
functional impairment after critical illness are led to a near doubling of functional indepen-
limited to moderate glucose control, as noted dence at hospital discharge, despite the fact that
above, and early mobilization. In 1975, Burns hospital length of stay did not differ between the
and Jones51 described a walker device designed intervention and control groups. The patients
for patients who require mechanical ventila- who were randomly assigned to immediate mo-
tion. Technological methods have evolved to the bilization spent fewer days receiving mechanical
n engl j med 370;17 293

The New
use only. Medicine
15, 2015.All
For personal
reserved.
ventilation and had less delirium and greater cannot necessarily be prevented, aggressive treat-
maximal walking distances. In addition, more ment of such conditions is nevertheless impor-
patients in the early-mobilization group were tant to minimize subsequent morbidity. Other
discharged to home after hospitalization. Inde- risk factors, such as severe hyperglycemia, can be
pendent predictors of functional independence attenuated with the use of insulin therapy with
at ICU discharge were younger age, absence of careful monitoring to avoid hypoglycemia. Early
sepsis, and randomized assignment to early mobilization of patients in the ICU, although not
physical and occupational therapy. Despite an a traditional approach, has become established
early, aggressive approach to mobilization in a as an evidence-based strategy to reduce the de-
study population that included patients with conditioning and dysfunction so commonly seen
acute lung injury, morbidly obese patients, pain survivors of critical illness. For this strategy to
tients with shock requiring vasoactive infusions, be successful, ongoing attention to minimizing
and patients receiving ongoing renal-replacement the use of sedation is important. In addition,
therapy, there were very few adverse events, none care providers in the ICU must acknowledge the
of which were consequential with respect to importance of a multidisciplinary care model to
outcome. Indeed, the most common barriers to optimize the efficacy of early mobilization. Al-
early mobility were failure to awaken from seda- though preliminary data are encouraging, more
tion, agitation when sedation was discontinued, research clearly is needed to determine whether
and scheduled diagnostic tests.57 the benefits of mobilizing patients in the ICU are
sustained after hospital discharge.
C onclusions
Dr. Kress reports receiving lecture fees from Hospira and
Many survivors of critical illness have considerable the France Foundation and fees for expert testimony in medi-
cal malpractice cases. Dr. Hall reports receiving fees for expert
functional impairment; a substantial contributor testimony in a patent case (Hospira v. Sandoz) and in medical
to such impairment is ICU-acquired weakness. malpractice cases on behalf of a patient, physicians, and hos-
Recovery is often slow and incomplete in such pitals. No other potential conflict of interest relevant to this
article was reported.
patients, particularly those who are elderly. Al- Disclosure forms provided by the authors are available with
though some of the risk factors, such as sepsis, the full text of this article at NEJM.org.
References
1. Osler W. The principles and practice Muscle atrophy and preferential loss of source utilization and ambulatory status at
of medicine, designed for the use of prac- myosin in prolonged critically ill patients. discharge. Disabil Rehabil 2011;33:105-12.
titioners and students of medicine. New Crit Care Med 2012;40:79-89. 15. Trojaborg W, Weimer LH, Hays AP.
York: D. Appleton, 1915:114-8. 9. Koch S, Spuler S, Deja M, et al. Critical Electrophysiologic studies in critical ill-
2. Olsen CW. Lesions of peripheral nerves illness myopathy is frequent: accompanying ness associated weakness: myopathy or
developing during coma. JAMA 1956;160: neuropathy protracts ICU discharge. J Neu- neuropathy — a reappraisal. Clin Neuro-
39-41. rol Neurosurg Psychiatry 2011;82:287-93. physiol 2001;112:1586-93.
3. MacFarlane IA, Rosenthal FD. Severe 10. Latronico N. Neuromuscular alterations 16. Bednarik J, Lukas Z, Vondracek P.
myopathy after status asthmaticus. Lancet in the critically ill patient: critical illness Critical illness polyneuromyopathy: the
1977;2:615. myopathy, critical illness neuropathy, or electrophysiological components of a com-
4. Bolton CF, Gilbert JJ, Hahn AF, Sib- both? Intensive Care Med 2003;29:1411-3. plex entity. Intensive Care Med 2003;29:
bald WJ. Polyneuropathy in critically ill 11. De Jonghe B, Sharshar T, Lefaucheur 1505-14.
patients. J Neurol Neurosurg Psychiatry JP, et al. Paresis acquired in the intensive 17. Morris C, Trinder JT. Electrophysiology
1984;47:1223-31. care unit: a prospective multicenter study. adds little to clinical signs in critical ill-
5. Zochodne DW, Bolton CF, Wells GA, JAMA 2002;288:2859-67. ness polyneuropathy and myopathy. Crit
et al. Critical illness polyneuropathy: a com- 12. Leatherman JW, Fluegel WL, David Care Med 2002;30:2612.
plication of sepsis and multiple organ fail- WS, Davies SF, Iber C. Muscle weakness 18. Stevens RD, Marshall SA, Cornblath
ure. Brain 1987;110:819-41. in mechanically ventilated patients with DR, et al. A framework for diagnosing and
6. De Jonghe B, Bastuji-Garin S, Shar- severe asthma. Am J Respir Crit Care Med classifying intensive care unit-acquired
shar T, Outin H, Brochard L. Does ICU- 1996;153:1686-90. weakness. Crit Care Med 2009;37:Suppl:
acquired paresis lengthen weaning from 13. van den Berghe G, Wouters P, Weekers S299-S308.
mechanical ventilation? Intensive Care Med F, et al. Intensive insulin therapy in criti- 19. Kleyweg RP, van der Meché FG, Schmitz
2004;30:1117-21. cally ill patients. N Engl J Med 2001;345: PI. Interobserver agreement in the assess-
7. Latronico N, Fenzi F, Recupero D, et al. 1359-67. ment of muscle strength and functional
Critical illness myopathy and neuropathy. 14. Clavet H, Hébert PC, Fergusson DA, abilities in Guillain-Barré syndrome.
Lancet 1996;347:1579-82. Doucette S, Trudel G. Joint contractures in Muscle Nerve 1991;14:1103-9.
8. Derde S, Hermans G, Derese I, et al. the intensive care unit: association with re- 20. Bolton CF. Neuromuscular manifes-
1634 n engl j med 370;17294

The New England

personal Medicine
reserved.
tations of critical illness. Muscle Nerve idly progressive diaphragmatic weakness 46. Slutsky AS. Neuromuscular blocking
2005;32:140-63. and injury during mechanical ventilation agents in ARDS. N Engl J Med 2010;363:
21. Fenzi F, Latronico N, Refatti N, Riz- in humans. Am J Respir Crit Care Med 1176-80.
zuto N. Enhanced expression of E-selectin 2011;183:364-71. 47. Sacanella E, Perez-Castejon JM, Nico-
on the vascular endothelium of peripheral 34. Levine S, Biswas C, Dierov J, et al. In- las JM, et al. Functional status and quality
nerve in critically ill patients with neuro- creased proteolysis, myosin depletion, and of life 12 months after discharge from a
muscular disorders. Acta Neuropathol atrophic AKT-FOXO signaling in human medical ICU in healthy elderly patients:
2003;106:75-82. diaphragm disuse. Am J Respir Crit Care a prospective observational study. Crit Care
22. Van den Berghe G, Schoonheydt K, Med 2011;183:483-90. 2011;15(2):R105.
Becx P, Bruyninckx F, Wouters PJ. Insulin 35. Dock W. The evil sequelae of complete 48. Kress JP, Pohlman AS, O’Connor MF,
therapy protects the central and periph- bed rest. JAMA 1944;125:1083-5. Hall JB. Daily interruption of sedative infu-
eral nervous system of intensive care pa- 36. Bednarík J, Vondracek P, Dusek L, sions in critically ill patients undergoing
tients. Neurology 2005;64:1348-53. Moravcova E, Cundrle I. Risk factors for mechanical ventilation. N Engl J Med
23. Hermans G, Wilmer A, Meersseman critical illness polyneuromyopathy. J Neu- 2000;342:1471-7.
W, et al. Impact of intensive insulin ther- rol 2005;252:343-51. 49. De Jonghe B, Cook D, Griffith L, et al.
apy on neuromuscular complications and 37. Druschky A, Herkert M, Radespiel- Adaptation to the Intensive Care Environ-
ventilator dependency in the medical in- Troger M, et al. Critical illness polyneu- ment (ATICE): development and valida-
tensive care unit. Am J Respir Crit Care ropathy: clinical findings and cell culture tion of a new sedation assessment instru-
Med 2007;175:480-9. assay of neurotoxicity assessed by a pro- ment. Crit Care Med 2003;31:2344-54.
24. The NICE-SUGAR Study Investigators. spective study. Intensive Care Med 2001; 50. Needham DM, Korupolu R, Zanni JM,
Hypoglycemia and risk of death in criti- 27:686-93. et al. Early physical medicine and rehabili-
cally ill patients. N Engl J Med 2012;367: 38. Herridge MS, Cheung AM, Tansey tation for patients with acute respiratory
1108-18. CM, et al. One-year outcomes in survivors failure: a Quality Improvement Project.
25. Idem. Intensive versus conventional glu- of the acute respiratory distress syn- Arch Phys Med Rehabil 2010;91:536-42.
cose control in critically ill patients. N Engl drome. N Engl J Med 2003;348:683-93. 51. Burns JR, Jones FL. Letter: Early am-
J Med 2009;360:1283-97. 39. Papazian L, Forel JM, Gacouin A, et bulation of patients requiring ventilatory
26. Rich MM, Pinter MJ. Crucial role of al. Neuromuscular blockers in early acute assistance. Chest 1975;68:608.
sodium channel fast inactivation in muscle respiratory distress syndrome. N Engl J 52. Bailey P, Thomsen GE, Spuhler VJ, et
fibre inexcitability in a rat model of criti- Med 2010;363:1107-16. al. Early activity is feasible and safe in re-
cal illness myopathy. J Physiol 2003;547: 40. Griffiths RD, Palmer TE, Helliwell T, spiratory failure patients. Crit Care Med
555-66. MacLennan P, MacMillan RR. Effect of 2007;35:139-45.
27. Rich MM, McGarvey ML, Teener JW, passive stretching on the wasting of 53. Morris PE, Goad A, Thompson C, et al.
Frame LH. ECG changes during septic muscle in the critically ill. Nutrition 1995; Early intensive care unit mobility therapy
shock. Cardiology 2002;97:187-96. 11:428-32. in the treatment of acute respiratory fail-
28. Garnacho-Montero J, Madrazo-Osuna J, 41. Iwashyna TJ, Ely EW, Smith DM, ure. Crit Care Med 2008;36:2238-43.
Garcia-Garmendia JL, et al. Critical illness Langa KM. Long-term cognitive impair- 54. Turner DA, Cheifetz IM, Rehder KJ,
polyneuropathy: risk factors and clinical ment and functional disability among et al. Active rehabilitation and physical
consequences: a cohort study in septic pa- survivors of severe sepsis. JAMA 2010;304: therapy during extracorporeal membrane
tients. Intensive Care Med 2001;27:1288-96. 1787-94. oxygenation while awaiting lung trans-
29. Carré JE, Orban JC, Re L, et al. Sur- 42. Herridge MS, Tansey CM, Matté A, plantation: a practical approach. Crit Care
vival in critical illness is associated with et al. Functional disability 5 years after Med 2011;39:2593-8.
early activation of mitochondrial biogen- acute respiratory distress syndrome. N Engl 55. Burtin C, Clerckx B, Robbeets C, et al.
esis. Am J Respir Crit Care Med 2010;182: J Med 2011;364:1293-304. Early exercise in critically ill patients en-
745-51. 43. Hopkins RO, Jackson JC. Short- and hances short-term functional recovery.
30. Reid MB, Moylan JS. Beyond atrophy: long-term cognitive outcomes in intensive Crit Care Med 2009;37:2499-505.
redox mechanisms of muscle dysfunction care unit survivors. Clin Chest Med 2009; 56. Morris PE, Griffin L, Berry M, et al.
in chronic inflammatory disease. J Physiol 30:143-53. Receiving early mobility during an inten-
2011;589:2171-9. 44. Schweickert WD, Pohlman MC, Pohl- sive care unit admission is a predictor of
31. Powers SK, Shanely RA, Coombes JS, man AS, et al. Early physical and occupa- improved outcomes in acute respiratory
et al. Mechanical ventilation results in pro- tional therapy in mechanically ventilated, failure. Am J Med Sci 2011;341:373-7.
gressive contractile dysfunction in the dia- critically ill patients: a randomised con- 57. Pohlman MC, Schweickert WD, Pohl-
phragm. J Appl Physiol 2002;92:1851-8. trolled trial. Lancet 2009;373:1874-82. man AS, et al. Feasibility of physical and
32. Levine S, Nguyen T, Taylor N, et al. 45. Fletcher SN, Kennedy DD, Ghosh IR, occupational therapy beginning from ini-
Rapid disuse atrophy of diaphragm fibers et al. Persistent neuromuscular and neuro- tiation of mechanical ventilation. Crit Care
in mechanically ventilated humans. N Engl physiologic abnormalities in long-term Med 2010;38:2089-94.
J Med 2008;358:1327-35. survivors of prolonged critical illness. Crit Copyright © 2014 Massachusetts Medical Society.
33. Jaber S, Petrof BJ, Jung B, et al. Rap- Care Med 2003;31:1012-6.
posting presentations from medical meetings online

Online posting of an audio or video recording of an oral presentation at
a medical meeting, with selected slides from the presentation, is not considered
prior publication. Authors should feel free to call or send e-mail to the
Journal’s Editorial Offices if there are any questions about this policy.
295nejm.org
n engl j med 370;17 april 24, 2014 1635
Downloaded from collections.nejm.org.TheForNew England
personal use Journal
only. Noofother
Medicine
uses without permission.
Copyright MULARCZYK
© Massachusetts on January
Medical 15, 2015.
Society. For personal
All rights reserved.use only. No other uses without permission.
Copyright ©Back
2014toMassachusetts Medical Society. All rights reserved.
Table of Contents
c or r e sp ondence
ICU-Acquired Weakness and Recovery from Critical Illness

To the Editor: Kress and Hall (April 24 issue)1 age-associated appendicular muscle wasting. Even
describe neuromuscular problems commonly ob- without the necessity of an ICU stay, sarcopenia
served in the intensive care unit (ICU) during is an independent risk factor for reduced surviv-
critical illness. Critical illness polyneuropathy is al.2 The presence of chronic illnesses such as
a generalized axonal neuropathy with predomi- heart failure3 or chronic obstructive pulmonary
nance in the lower extremities.2 Entrapment neu- disease4 further increases the likelihood of mus-
ropathy is also quite common in the ICU. Loss of cle wasting. Increased metabolic needs and ad-
subcutaneous fat during a long stay in the ICU justed rehabilitation measures should be con-
makes peripheral nerves susceptible to compres- sidered when muscle mass and strength are
sion injury, particularly at the fibular head (lead- evaluated in patients in the ICU who are elderly
ing to diminished power of the anterior tibialias) or who have preexisting illness.
and at the ulnar groove (leading to diminished Stephan von Haehling, M.D., Ph.D.
power in the intrinsic hand muscles). The prob- Charité Medical School
lem can be avoided simply by awareness and Berlin, Germany
proper limb positioning. stephan.von.haehling@charite.de
Robert Kalb, M.D. ported.
Perelman School of Medicine at the University of Pennsylvania
Philadelphia, PA 1. Morley JE, Kim MJ, Haren MT, Kevorkian R, Banks WA.
kalb@email.chop.edu Frailty and the aging male. Aging Male 2005;8:135-40.
2. Landi F, Cruz-Jentoft AJ, Liperoti R, et al. Sarcopenia and
No potential conflict of interest relevant to this letter was re- mortality risk in frail older persons aged 80 years and older: re-
ported. sults from ilSIRENTE study. Age Ageing 2013;42:203-9.
1. Kress JP, Hall JB. ICU-acquired weakness and recovery from 3. Fülster S, Tacke M, Sandek A, et al. Muscle wasting in pa-
critical illness. N Engl J Med 2014;370:1626-35. tients with chronic heart failure: results from the Studies Inves-
2. Bolton CF, Gilbert JJ, Hahn AF, Sibbald WJ. Polyneuropathy tigating Co-morbidities Aggravating Heart Failure (SICA-HF).
in critically ill patients. J Neurol Neurosurg Psychiatry 1984;47: Eur Heart J 2013;34:512-9.
1223-31. 4. Sergi G, Coin A, Marin S, et al. Body composition and rest-
ing energy expenditure in elderly male patients with chronic
DOI: 10.1056/NEJMc1406274 obstructive pulmonary disease. Respir Med 2006;100:1918-24.
DOI: 10.1056/NEJMc1406274
To the Editor: Kress and Hall propose that re-

habilitation of critically ill patients should begin The Authors Reply: We agree with Kalb that
in the ICU. The authors name sepsis, systemic peripheral-nerve injury is an avoidable problem,
inflammation, multiorgan failure, hyperglycemia, particularly in superficial nerves that are in close
glucocorticoid use, and female sex as risk factors contact with bones. Proper limb positioning, as
for ICU-acquired weakness; however, they do not well as early mobility and minimization of seda-
discuss the patient’s skeletal-muscle status be- tion, are practical solutions for this problem.
fore ICU admission. This factor is largely ne- Von Haehling raises an important issue, par-
glected in clinical practice, in which 5 to 13% of ticularly given the increasing percentage of elderly
patients who are 60 to 70 years of age and 11 to patients, chronically ill patients, or both in whom
50% of patients who are 80 years of age or older critical illness develops. Unfortunately, since ICU
are affected by sarcopenia,1 which is defined as admission is usually an unexpected event, clini-
296
correspondence
cians often have incomplete knowledge about John P. Kress, M.D.

preexisting functional status. Nevertheless, seek- Jesse B. Hall, M.D.
ing information about baseline functional status University of Chicago
Chicago, IL
and skeletal-muscle mass is important when jkress@medicine.bsd.uchicago.ed
addressing ICU-acquired weakness. Previous Since publication of their article, the authors report no fur-
medical records, a physical examination, and ther potential conflict of interest.
discussions with family members can help to DOI: 10.1056/NEJMc1406274
Correspondence Copyright © 2014 Massachusetts Medical Society.
identify patients who are likely to have preexist-
ing sarcopenia.


Since it is likely that this patient will have a prolonged stay in the ICU, what measures would you take to
optimize his long-term recovery?
A. Maintain a blood glucose level of 80 to 110 mg per deciliter (4.4 to 6.1 mmol per liter) throughout his ICU stay.
B. Administer human growth hormone to improve nitrogen balance and preserve muscle mass and strength.
C. Minimize sedation and institute early active and passive mobilization even while he is still undergoing
mechanical ventilation.

Prolonged treatment in an ICU, and particularly prolonged mechanical ventilation, is associated with a loss of
muscle mass and a marked reduction in muscle strength, which is referred to as ICU-acquired weakness. ICU-
acquired weakness is a result of a number of pathophysiological mechanisms, and more than one mechanism may
affect an individual patient during treatment in an ICU. Early studies identified critical-illness polyneuropathy,
which is characterized by symmetric weakness affecting the proximal limb muscles through denervation and de-
nervation atrophy. Critical-illness polyneuropathy may also affect the respiratory muscles and delay weaning from
mechanical ventilation. Critical-illness myopathy is a primary myopathy that may be difficult to distinguish from
critical-illness polyneuropathy. Critical-illness myopathy is more common than critical-illness polyneuropathy
but presents with similar weakness of limb and respiratory muscles. The two conditions can coexist, and although
they can be characterized and differentiated by electrophysiological studies and muscle-biopsy analysis, such eval-
uations are often not technically possible in a critically ill patient. (Coexisting polyneuropathy and myopathy are
termed critical-illness neuromyopathy.) Risk factors for ICU-acquired weakness include prolonged inflammation
or sepsis, immobility, prolonged mechanical ventilation, uncontrolled hyperglycemia, and treatment with gluco-
corticoids or neuromuscular-blocking agents.
There are no specific treatments that have been shown to prevent ICU-acquired weakness. In one study, targeting
normoglycemia (glucose concentration, 80 to 110 mg per deciliter) in critically ill medical patients was associated
with a reduced incidence of critical-illness neuromyopathy and a reduction in the need for prolonged mechanical
ventilation.1 These results were supported by a retrospective study from the same authors2 but have not been repli-
cated by others. Because of the perceived adverse effects of targeting normoglycemia, current guidelines recom-
mend starting treatment with insulin in patients with a blood glucose level of more than 180 mg per deciliter
(10 mmol per liter) with a target of 140 mg per deciliter (7.8 mmol per liter) to 180 mg per deciliter.3,4 In a placebo-
controlled trial evaluating the hypothesis that growth hormone would improve nitrogen balance and lead to better
clinical outcomes, the use of growth hormone in critically ill adults increased mortality significantly.5
Current strategies to limit ICU-acquired weakness and optimize recovery from critical illness are focused around
minimizing sedation and targeting early mobilization of patients, even while they are still undergoing mechanical
ventilation.6 In patients who cannot be mobilized, passive exercises may also improve recovery.7
REFERENCES
1. Hermans G, Wilmer A, Meersseman W, et al. Impact of intensive insulin therapy on neuromuscular complications
and ventilator dependency in the medical intensive care unit. Am J Respir Crit Care Med 2007;175:480-9.
298
2. Hermans G, Schrooten M, Van Damme P, et al. Benefits of intensive insulin therapy on neuromuscular compli
cations in routine daily critical care practice: a retrospective study. Crit Care 2009;13:R5.
3. Qaseem A, Humphrey LL, Chou R, Snow V, Shekelle P. Use of intensive insulin therapy for the management of
glycemic control in hospitalized patients: a clinical practice guideline from the American College of Physicians.
Ann Intern Med 2011;154:260-7.
4. American Diabetes Association. Standards of medical care in diabetes — 2012. Diabetes Care 2012;
35:Suppl 1:S11-S63.
5. Takala J, Ruokonen E, Webster NR, et al. Increased mortality associated with growth hormone treatment in
critically ill adults. N Engl J Med 1999;341:785-92.
6. Morris PE, Goad A, Thompson C, et al. Early intensive care unit mobility therapy in the treatment of acute
respiratory failure. Crit Care Med 2008;36:2238-43.
7. Burtin C, Clerckx B, Robbeets C, et al. Early exercise in critically ill patients enhances short-term functional
recovery. Crit Care Med 2009;37:2499-505.
299
TRAUMATIC INTRACRANIAL HYPERTENSION
Our ability to identify and manage intracranial hypertension in the ICU is rapidly evolving. This review covers
our understanding of the problem as well as provides suggestions for its treatment.
300
Case Challenge
Traumatic Intracranial Hypertension

After transfer to a surgical floor following a complicated, long ICU stay for septic shock, a frail 77-year-old
man is found on the floor, unable to speak, and with decreased movement on the right side. CT reveals an
acute subdural hematoma, hemorrhagic contusion, and displacement of midline intracranial structures. What
is the most appropriate management of this patient’s traumatic brain injury?
hospital from the operating room after a Hartmann’s procedure (resection of the rectosigmoid colon with closure
of the rectal stump and formation of an end colostomy) performed for fecal peritonitis due to a perforated sigmoid
colon. After surgery, the patient undergoes mechanical ventilation with the use of a low-tidal-volume protocol
with positive end-expiratory pressure (PEEP) for acute respiratory distress syndrome and is treated for septic
shock. Complications during his ICU stay include mild disseminated intravascular coagulation and acute hepatic
necrosis caused by acetaminophen treatment. Despite active mobilization, a degree of ICU-acquired weakness has
developed. (In the previous installment of this case, there were 1937 votes on how to treat the patient’s weakness.
A large majority of the respondents [94%] favored minimizing sedation and instituting early active and passive
301
mobilization, even while the patient is still undergoing mechanical ventilation. Another 4% of respondents
favored maintaining a blood glucose level of 80 to 110 mg per deciliter [4.4 to 6.1 mmol per liter] throughout
his ICU stay, whereas 1% favored administering human growth hormone to improve his nitrogen balance and
preserve muscle mass and strength.)
After 15 days in the ICU, the patient is weaned from mechanical ventilation and transferred to the surgical floor.
On the third morning after discharge from the ICU, he is found on the floor of his room, having apparently fallen
from his bed. On examination, he opens his eyes but is not able to speak coherently. He has decreased movement
on the right side. His pupils are both midsized and reactive to light. Cranial computed tomography (CT) reveals
an acute subdural hematoma in the left hemisphere with underlying hemorrhagic contusion and 5-mm displace-
ment of the midline intracranial structures.

What is the most appropriate management of this patient’s traumatic brain injury?
A. Conservative management, since there is no prospect for functional recovery.

B. Evacuation of the acute subdural hematoma, placement of an external ventricular drain, and admission to the
ICU for active management of increased intracranial pressure.
C. Evacuation of the acute subdural hematoma and admission to the ICU for management without monitoring
of intracranial pressure.
302
review article


Nino Stocchetti, M.D., and Andrew I.R. Maas, M.D., Ph.D.
A
n elevation in intracranial pressure can be a medical or From the Department of Pathophysiology
surgical emergency. There are many possible conditions that can lead to and Transplantation, Milan University, and
Fondazione Istituto di Ricovero e Cura a
elevated intracranial pressure on either an acute or a chronic basis (Table 1). Carattere Scientifico (IRCCS) Cà Granda–
In this article, we focus on the increased intracranial pressure that occurs in pa- Ospedale Maggiore Policlinico — both in
tients after traumatic brain injury, since this is an area in which there are both Milan (N.S.); and the Department of Neu-
rosurgery, Antwerp University Hospital–
physiological and clinical data. University of Antwerp, Edegem, Belgium
Traumatic brain injury is a medical and social problem worldwide, with an (A.I.R.M.). Address reprint requests to
estimated 10 million cases leading to hospitalization or death each year.1 In low- Dr. Stocchetti at Fondazione IRCCS Cà
Granda–Ospedale Maggiore Policlinico,
and medium-income countries, in which the use of motor-powered transportation Via F. Sforza 35, 20122 Milan, Italy, or at
is increasing, the incidence of this condition is rising2 and involves predominantly stocchet@policlinico.mi.it; or to Dr. Maas
young men. In contrast, in richer countries, the epidemiology of traumatic brain in- at the Department of Neurosurgery,
Antwerp University Hospital–University of
jury is changing because of two main factors: the rate of traffic incidents is decreas- Antwerp, Wilrijkstraat 10, 2650 Edegem,
ing owing to successful enforcement of safety laws and preventive measures, whereas Belgium, or at andrew.maas@uza.be.
the aging of the population makes such injury in the elderly more frequent.2-4 N Engl J Med 2014;370:2121-30.
Falls are a frequent cause of injury in older patients, often resulting in contusive DOI: 10.1056/NEJMra1208708
brain injury. Elderly patients often have multiple coexisting illnesses and are fre- Copyright © 2014 Massachusetts Medical Society.
quently taking a number of medications, including anticoagulants and platelet

aggregation inhibitors.4 These medications may contribute to the development of
hematomas and hemorrhagic expansion of contusions. In the past, it was gener-
ally thought that traumatic brain injury in elderly persons was associated with a
uniformly bad outcome, and as a consequence more aggressive treatment was
often withheld.5 Although increasing age is unequivocally an independent predic-
tor of a poor outcome,6 recent data suggest that favorable outcomes, at least in the
short term, are no longer uncommon among elderly patients7; however, these pa-
tients require sustained, high-level medical care and rehabilitation.
Pathoph ysiol o gy
Normal intracranial pressure in adults is below 15 mm Hg, with transient increases due
to coughing or sneezing. Intracranial-pressure values that are sustained above 20 mm Hg
are considered to be pathologic in adults and an indication for intensifying treatment
in patients with traumatic brain injury. Under normal conditions, the total volume
within the skull remains constant and is determined by the sum of the cerebrospinal
fluid, blood, and brain-tissue compartments. The volume of these compartments is
tightly regulated, and cerebral blood flow is kept constant by autoregulation. When
additional volume is added to the system, compensatory mechanisms (e.g., displace-
ment of cerebrospinal fluid to the spinal subarachnoid space and compression of the
cerebral venous bed) operate to keep intracranial pressure constant.
The relationship between intracranial volume and intracranial pressure is expo-
nential. Initially, pressure increases only slightly with increasing volume, but when
the buffering capabilities of the system are exceeded, intracranial pressure rises

may 29, 2014 2121
The New England
Medical personal
Society. use only.
reserved.
Table 1. Main Mechanisms Causing Increased Intracranial Pressure Associated with Common Medical Conditions.*
Disturbed
Condition Mass Effect Edema Vasodilatation Circulation of CSF
Traumatic brain injury + + +
Subarachnoid hemorrhage + + ++
Cerebral venous thrombosis + ++
Anoxic–ischemic encephalopathy +
Brain tumor + +
Brain infarction after acute occlusion +
of middle cerebral artery
Spontaneous intracerebral hematoma + +
Abscess + +
Meningitis +
Idiopathic intracranial hypertension +?
Acute liver encephalopathy + +
Acute hypoosmolar syndromes +
Hypertensive encephalopathy +
Reye’s syndrome +
Craniosynostosis†
* A single plus sign indicates that the mechanism is relevant, and two plus signs indicate that the mechanism is particu-
larly relevant. The question mark indicates that the mechanism has not been clearly identified. CSF denotes cerebro-
spinal fluid.
† Craniosynostosis is associated with inadequate skull growth.
steeply. This explains the rapid deterioration that blood pressure minus the intracranial pressure.
is frequently seen in patients with a traumatic The cerebral perfusion pressure is the driving
intracranial hematoma. force behind cerebral blood flow, but levels that
Both intracranial and systemic events contribute are required for adequate flow vary among pa-
to increased intracranial pressure after traumatic tients. As the cerebral perfusion pressure de-
brain injury (Table 2). In the first hours after trau- creases, cerebral blood flow may become insuf-
ma, expansion of hematomas is the main threat,8 ficient for adequate brain-tissue perfusion and
whereas in the following days, other mechanisms, oxygenation.9,10 Ischemia will induce further cyto-
including water accumulation, disrupted autoreg- toxic edema and result in even higher intracra-
ulation, ischemia, and contusion expansion, lead nial pressure. Adverse effects of increased intra-
to further increases in intracranial pressure. cranial pressure and low cerebral perfusion
The direct consequences of increased intra- pressure on mortality and long-term outcome
cranial pressure on the brain can be broadly dif- have been documented in many studies.11-13
ferentiated as mechanical or vascular. When a mass These insights, while providing a clear motivation
lesion develops, a pressure gradient originating for the monitoring and treatment of raised intra-
from this area will cause distortion of brain tissue, cranial pressure, illustrate the complex interac-
midline shift, and displacement of the brain tissue tions among pressure, flow, and metabolism.
in a medial or caudal direction (herniation) (Fig. 1).
Herniation is a medical emergency, requiring Moni t or ing of
prompt treatment to prevent irreversible and of- In t r acr a ni a l Pr e ssur e
ten fatal damage to the brain stem.
Vascular effects of increased intracranial pres- The practice of continuous monitoring of intra-
sure are caused by impaired cerebral perfusion cranial pressure started with the pioneering
pressure, which is defined as the mean arterial work of Guillaume and Janny in France14 and

304 may 29, 2014
The New EnglandFor

reserved.
Lundberg in Sweden.15 A catheter, usually referred T r e atmen t

to as an external ventricular drain, was placed in
the lateral ventricle and connected, through a Guidelines advocate the early treatment of in-
fluid-filled system, to a transducer. This method creased intracranial pressure, since increased
is still considered the standard of care for the severity and longer duration of raised intracra-
measurement of intracranial pressure. Noninva- nial pressure are associated with a poor out-
sive monitors of intracranial pressure currently come.11,12 The accepted threshold for treatment
have insufficient validation for clinical practice, is an intracranial pressure of 20 mm Hg. In all
but other invasive methods (e.g., intraparenchymal patients with increased intracranial pressure, a
probes, microstrain gauge transducers, and fiber- repeat CT scan should be considered to exclude
optic catheters) are gaining in popularity because surgically treatable lesions. Before initiating
of their ease of use.16 These techniques, however, therapy to reduce increased intracranial pres-
do not allow for drainage of cerebrospinal fluid, sure, practitioners need to exclude erroneous
which removes an effective means of reducing in- measurements and systemic causes that can be
tracranial pressure. Catheters may also be placed
in the subdural space (e.g., after evacuation of an
Table 2. Causes of and Possible Therapies for Increased Intracranial Pressure
epidural or subdural hematoma); such techniques in Traumatic Brain Injury.
permit monitoring of intracranial pressure, but
they do not permit drainage of cerebrospinal fluid, Variable Possible Therapeutic Approach
and measurements are less reliable than those Intracranial cause
obtained with an external ventricular drain. Hematoma Surgical evacuation, decompres-
International guidelines recommend the moni- sive craniectomy
toring of intracranial pressure in all patients with Epidural
survivable severe traumatic brain injury and ab- Acute subdural
normalities on computed tomography (CT) ob- Intracerebral
tained at the time of admission,17,18 as well as in
Contusion Surgical evacuation, decompres-
selected patients (e.g., those who are over the age sive craniectomy
of 40 years with hypotension or abnormal flex- Disturbance in cerebrospinal fluid Drainage of cerebrospinal fluid
ion or extension in response to pain) with a
With ventricular enlargement
normal CT scan. The insertion of intracranial
catheters carries risks of hemorrhage and in- Without ventricular enlargement
fection. Ventricular catheters, with deeper brain Edema Hyperosmolar fluids, decom-
pressive craniectomy
penetration, are more risky. Reported rates of
hemorrhage are variable (1 to 7% for ventricular Cytotoxic (intracellular)
catheters; less than that for intraparenchymal Vasogenic (extracellular)
probes),19,20 and only rarely do such hemorrhages Vasodilatation (increased intravascular Mild hyperventilation, barbiturates
require surgical evacuation.18 Placement of an blood volume, causing swelling)
intracerebral catheter is relatively contraindicated Seizures Antiepileptic medication
in patients with coagulopathy (i.e., increased pro- Extracranial cause
thrombin time, partial thromboplastin time, or Airway obstruction Airway clearance, possible tracheal
international normalized ratio or a platelet count intubation
of <100,000 per microliter).21 Routine cultures of Hypoxemia Oxygenation and ventilation
catheter tips of external ventricular drains may Hypercarbia Ventilation
reveal bacterial colonization, but fortunately the
Hypertension associated with pain Analgesia and sedation
rate of invasive infection is lower than that of
Coughing or straining Sedation, paralysis
colonization.22 The risk of infection is higher
with ventricular catheters than with parenchymal Jugular venous obstruction (pneumo- Draining of pneumothorax, correc-
thorax, neck compression) tion of neck position
probes, with reported rates of infection ranging
Abdominal distention Nasogastric tube
from 1 to 27%. Risk factors include longer dura-
tion of monitoring, the presence of an open skull Fever Antipyretic drugs
fracture with leakage of cerebrospinal fluid, and Hypoosmolarity Hyperosmolar fluids
leakage around the ventriculostomy site.22,23
n engl j med 370;22 305may 29, 2014

nejm.org 2123
The New England
Downloaded from collections.nejm.org. For Journal
personalofuse
Medicine
MULARCZYK on January Medical
© Massachusetts 15, 2015.Society.
For personal use only.
reserved.
A Intracranial pressure under normal conditions, B Acute subdural hematoma, axial section
sagittal section
Midline
Hematoma
Intracranial causes of increased

intracranial pressure
At equilibrium: • Mass lesions (e.g., traumatic
The sum of the 3 intracranial hematomas, tumors)
components (cerebrospinal fluid + • Edema
cerebral tissue + blood content) is constant, • Vasodilatation
corresponding to a normal (10–15 mm Hg) • Disturbed central spinal fluid
intracranial pressure. circulation
30 Normal intracranial pressure tracing 30 Intracranial pressure tracing showing a progressive increase
ICP (mm Hg)
ICP (mm Hg)
20 20
10 10
0 0
0 5 10 15 20 25 30 35 40 45 50 55 13 14 15 16 17 18 19 20 21 22 23 24
Time (sec) Time (hr)
C Intracranial pressure monitoring D Herniation, coronal section

by ventricular catheter
Monitor
Hematoma
Collecting system
Falcine
herniation
Catheter
Midline
shift
Lateral ventricle Uncal

Central herniation
herniation
2124 n engl j med

306370;22 nejm.org may 29, 2014
The New
use only. Medicine
15, 2015.All
For personal
reserved.
Figure 1 (facing page). Intracranial Pressure under

Normal and Abnormal Conditions. Verify ICP measurement
Under normal conditions, the intracranial pressure (ICP) Check calibration of monitor
remains constant at 10 to 15 mm Hg, fluctuating with
cardiac and respiratory cycles, as shown in the normal
trace recording (Panel A). Since the cranium is a rigid Exclude herniation
container, the sum of the various intracranial volumes
(brain tissue, cerebrospinal fluid, and blood) must re- Check pupils
main constant. Cerebrospinal fluid is continuously formed If herniation:
and reabsorbed, with the circulation indicated by blue Perform emergency procedures
arrows. Several intracranial and systemic causes may to reduce ICP (tracheal intu-
bation, ventilation, adminis-
alter the intracranial components and cause pathologic tration of hyperosmolar fluids)
increases in intracranial pressure — for example, a trau- Perform immediate diagnostics
matic left subdural hematoma that compresses the brain (computed tomography)
and shifts the lateral ventricles to the right, as shown Perform surgical intervention
when appropriate
on computed tomography (Panel B). The hematoma
volume cannot be compensated by buffering systems,
and there is a corresponding increase in intracranial
pressure, which can be recorded through a catheter Vasodilatation of cerebral vessels
inserted in a lateral ventricle (also allowing the with- Fever
drawal of cerebrospinal fluid) (Panel C). The catheter Seizures
is connected to a collecting system, to which cerebro- Hypercarbia
Hypoxemia
spinal fluid can be drained, and to a monitor, where Hypotension
the trace recording of intracranial pressure is displayed.
Intracranial hypertension may cause compression and
displacement of the cerebral tissue from areas of higher Increased venous pressure
pressure toward areas of lower resistance (Panel D). Neck torsion or compression
Brain herniation occurs in three main ways. First, a Pneumothorax
Check and treat systemic causes Ventilator asynchrony
hemisphere is displaced medially against the falx, re- of increased ICP Increased abdominal pressure
sulting in falcine herniation. Second, a monolateral
pressure gradient pushes the medial edge of the tem-
poral lobe (uncus) through the tentorial foramen, re- Increased arterial pressure
sulting in uncal herniation. In this syndrome, the third Pain
cranial nerve and the posterior cerebral artery are com- Bladder distention
pressed, causing unilateral pupillary dilation, a lack of
reactivity to light, and infarction. The brain stem is Cellular edema
distorted and compressed, with early impairment of Hyponatremia
consciousness. Third, a bilateral, homogeneous in-
crease in intracranial pressure in the supratentorial
space displaces the brain downward through the ten-
Increase sedation
torial foramen, resulting in central transtentorial her- Drain CSF
niation. The brain stem is compressed and displaced Administer hyperosmolar fluids
downward without signs of lateralization and with bi- Initiate or intensify ICP-directed
lateral pupillary abnormalities. treatment
Induce mild hypocarbia
(hyperventilation)
Induce metabolic suppression
(barbiturates or increased sedation)
rapidly corrected. A practical algorithm is pre- Induce hypothermia
sented in Figure 2. Perform decompressive craniectomy
Medical Therapy Figure 2. Algorithm for the Treatment of Increased Intracranial Pressure (ICP).
During the past 10 years, management of in- CSF denotes cerebrospinal fluid.
creased intracranial pressure has evolved toward
standardized strategies that use a “staircase” ap-
proach with an escalating treatment intensity resulting from vasodilatation, making maintenance
(Fig. 3).24,25 Sedation and analgesia are used to of normovolemia a prerequisite. An additional
treat pain and agitation26 and to prevent arterial advantage of sedation is to minimize the risk of
hypertension and patient–ventilator dyssynchrony. seizures.27,28
Sedation increases the risk of arterial hypotension Hyperosmolar agents reduce brain volume

may 29, 2014 2125
The New England
personal Medicine
reserved.
Therapy Levels of
Steps Evidence Treatment Risk
Infection or delayed hematoma
Subdural effusion
8 Not reported Decompressive craniectomy
Hydrocephalus and syndrome
of the trephined
Hypotension and increased number

7 Level II Metabolic suppression (barbiturates)
of infections
6 Level III Hypothermia Fluid and electrolyte disturbances and infection
5 Level III Induced hypocapnia Excessive vasoconstriction and ischemia
Negative fluid balance

Hyperosmolar therapy
4 Level II Hypernatremia
Mannitol or hypertonic saline
Kidney failure
3 Not reported Ventricular CSF drainage Infection
2 Level III Increased sedation Hypotension
Intubation
Not Coughing, ventilator asynchrony,
1 Normocarbic
reported ventilator-associated pneumonia
ventilation
Figure 3. Staircase Approach to the Treatment of Increased Intracranial Pressure.

The level of therapy in patients with raised intracranial pressure is increased step by step, with more aggressive interventions when there
is no response. The sequence of interventions may vary among different institutions; every intervention is associated with adverse effects.
Shown are the levels of evidence that underpin various approaches to treatment. Levels of evidence are based on the criteria for classifi-
cation of evidence, as used in international guidelines.24 The revised guidelines for the management of severe traumatic brain injury and
the surgical guidelines for the management of such injury do not contain any evidence on ventricular drainage of cerebrospinal fluid (CSF)
or the use of decompressive craniectomy. Level I evidence25 shows that decompressive craniectomy is effective in reducing intracranial
pressure but may worsen the long-term outcome and is associated with several complications. Among them is the syndrome of the tre-
phined, in which a sunken skull flap develops with a (poorly understood) neurologic deterioration.
and intracranial pressure through multiple mech- plasma concentration. Comparisons between
anisms. In the first minutes of infusion, mannitol mannitol and hypertonic saline for the treatment
and hypertonic saline expand the plasma volume, of increased intracranial pressure have not shown
decrease blood viscosity, and reduce the cerebral a clear superiority of one option over the other.29
blood volume.29 Once plasma osmolarity in- Induced arterial blood hypocarbia (hyperventi-
creases, a gradient across the blood–brain bar- lation) reduces intracranial pressure at the expense
rier is established, and water is extracted from of decreasing cerebral blood flow as a result of
the brain. This effect may last for up to several vasoconstriction.30 Hyperventilation carries a seri-
hours, until the osmotic equilibrium is reestab- ous risk of cerebral ischemia. For this reason, cur-
lished. The integrity of the blood–brain barrier rent guidelines recommend additional monitor-
is a prerequisite for the efficacy of hyperosmolar ing for cerebral ischemia (e.g., by the monitoring
agents. Mannitol is an osmotic diuretic and may of oxygen saturation in the jugular bulb and of
cause dehydration and hypovolemia. Hypertonic brain-tissue oxygenation) when hyperventilation
saline may cause abrupt increases in the sodium is used.31
2126 308370;22
n engl j med nejm.org may 29, 2014
The New
Society. Forrights
reserved.
Barbiturates depress the cerebral metabolism the outflow pressure than by actual brain pres-
and reduce cerebral blood flow, causing a pro- sure, so that accurate monitoring of intracranial
portional decrease in cerebral blood volume and pressure is not possible with continuous drainage
a decrease in intracranial pressure. Initial enthu- of cerebrospinal fluid.39 New ventricular catheters,
siasm for barbiturate therapy has been tempered including those with a miniature pressure trans-
by the recognition of serious side effects, including ducer at the tip, may offer more accurate reading
cardiac depression, arterial hypotension, and an during drainage but at a higher monetary cost.
increased risk of infection.32 The administration of Intermittent drainage when the intracranial pres-
barbiturates is generally reserved for refractory in- sure exceeds 20 mm Hg should be performed
tracranial hypertension, after other therapies have against a pressure gradient of approximately 10 cm
been tried and have failed.12 of water. Drainage of cerebrospinal fluid through
Mild hypothermia (32 to 34°C) is effective in lumbar catheters is not recommended in patients
decreasing intracranial pressure,33 but studies with increased intracranial pressure because of the
on the clinical benefit are contradictory, and risk of herniation.40
current evidence does not support its general use Much interest has focused on decompressive
in patients with traumatic brain injury.34,35 The craniectomy. The concept of decompressive crani-
effects of hypothermia are complex, and adverse ectomy is to provide a larger reserve to compen-
effects are a particular problem in patients with sate for increased intracranial volume. The actual
traumatic brain injury who may require cooling volume that is gained depends on the diameter
for many days to control refractory intracranial of the craniectomy. The presumed benefits of
hypertension. decompressive craniectomy have been challenged
by the results of the Decompressive Craniectomy
Surgical Therapy (DECRA) trial.25 In this study, which compared
The surgical therapy of raised intracranial pressure decompressive craniectomy performed within
includes the evacuation of mass lesions, drainage 72 hours after traumatic brain injury with maxi-
of cerebrospinal fluid, and decompressive craniec- mal medical therapy in patients with diffuse
tomy. Rapid detection and timely evacuation of brain injury whose intracranial pressure ex-
an intracranial hematoma is a cornerstone in the ceeded 20 mm Hg for 15 minutes or longer in a
management of traumatic brain injury. Guide- 1-hour period, mortality was similar in the two
lines on the surgical management of epidural or groups. However, the rate of unfavorable neuro-
acute subdural hemorrhage and of brain contu- logic outcomes was significantly higher among
sions have been published, but all of them are patients undergoing decompressive craniectomy.
based on class III evidence and are heavily fo- After adjustment for baseline data, such as the re-
cused on volumetric criteria.36 However, the sur- activity of pupils, the between-group difference in
gical evacuation of an intracerebral or subdural outcome was not significant. The generalizability
hematoma may be motivated not only by volume of these results is limited because of the highly
or mass effect but also by mitigation of a toxic selected patient population and because the study
effect. In a study of experimental rodent models examined the effect only in patients with diffuse
of cerebral contusions, Tanaka et al.37 found meta- injuries.
bolic disturbances, with a massive increase in the Decompressive craniectomy is not without risk,
production of excitatory amino acids and subse- and adverse effects are common.41 In the ongoing
quent ischemic damage, in the cortex underlying Randomized Evaluation of Surgery with Craniec-
blood clots. Possible benefits of surgical excision tomy for Uncontrollable Elevation of Intracranial
in the clinical situation are suggested by an analy- Pressure (RESCUEicp) study (Current Controlled
sis of 182 patients with cerebral contusions reg- Trials number, ISRCTN66202560),42 which is com-
istered in the Japan Neurotrauma Databank.38 paring medical therapy with decompressive crani-
Currently, there are no data from randomized, ectomy, patients with a sustained elevation in in-
controlled trials to support this approach. tracranial pressure (>25 mm Hg for more than
Drainage of cerebrospinal fluid is a simple and 1 hour and up to 12 hours) that is resistant to initial
effective approach to reducing increased intracra- medical therapy are randomly assigned to undergo
nial pressure. During withdrawal of cerebrospinal surgery or intensive medical therapy, including the
fluid, the pressure reading is determined more by use of barbiturates. The results of this trial will
n engl j med 370;22 309

nejm.org may 29, 2014 2127
The New
For personal ofNo
use only. Medicine
15, 2015.
AllFor personal
reserved.
provide further evidence to define the role of de- monitoring of intracranial pressure have persis-
compressive craniectomy in traumatic brain injury. tently increased intracranial pressure.12 In this
trial, the median and mean percentages of record-
R e think ing of Moni t or ing ings of intracranial pressure that were 20 mm Hg
or higher were only 7% and 20%, respectively. This
Despite the strong motivational reasons for mon- low incidence confounds sample-size calculations
itoring intracranial pressure, the concept of such and probably renders the trial underpowered.
monitoring in patients with traumatic brain in- A major asset of the study by Chesnut et al. is
jury has been challenged. In a comparison of two that it forms an incentive to rethink our con-
approaches to the monitoring of patients with cepts of the monitoring of intracranial pressure
traumatic brain injury at two Dutch centers — and the treatment of intracranial hypertension.
one that used frequent monitoring of intracranial The main intent of such monitoring is to better
pressure and the other that did not use such target early and specific therapy to the patients
monitoring — patients in the two institutions who may benefit most while not exposing others
who survived beyond 24 hours after injury had to unnecessary risks. This intent is in line with
similar outcomes at 1 year.43 On the basis of a the concept that improved disease characteriza-
large trauma database, Shafi et al.44 found an tion will facilitate targeted management and
association between monitoring of intracranial individualized approaches.48 Furthermore, we
pressure and a worse outcome. The two studies should recognize that current treatments that
have substantial methodologic limitations be- are initiated in response to intracranial hyper-
cause of study design, selection bias, and inade- tension are simplistic and often ill-focused. Cur-
quate adjustment for confounders. In contrast to rent protocol-driven approaches use a “one size
these trials, two other studies — one a historical fits all,” staircase approach with an escalating
comparison45 and the other an analysis of a large intensity of therapy, regardless of the underlying
database46 — show a reduced rate of death asso- pathophysiological features. More individualized
ciated with monitoring of intracranial pressure. approaches may be preferable, but we are not yet
Chesnut et al.47 recently reported the results able to identify specific causes of increased intra-
of a clinical trial aimed at determining the po- cranial pressure with sufficient reliability. The
tential benefit of monitoring intracranial pres- only study in the literature reporting a targeted
sure. Patients with severe traumatic brain injury approach is based on findings in 17 patients.49
were randomly assigned to a therapeutic proto- Since the outcomes for patients with severe trau-
col driven by monitoring of intracranial pressure matic brain injury have not improved substan-
or to a therapy that was determined on the basis tially during the past 20 to 30 years, developing
of clinical and radiologic findings. No signifi- and advocating a new approach may be consid-
cant between-group difference in outcome was ered to be long overdue.
shown. Although the study provided the first
randomized comparison of the treatment of pa- C onclusions
tients with traumatic brain injury with or with-
out monitoring of intracranial pressure, we do After traumatic injury, the brain is vulnerable to
not consider the findings to be conclusive evi- a range of threats that may be successfully treat-
dence against the use of such monitoring. Con- ed if they are recognized promptly and therapy is
ceptually, any improvement in outcome must started early. Among such threats is the rapid
result from differences in treatment as a result development of increased intracranial pressure,
of monitoring of intracranial pressure and not which is especially relevant, since it is associated
from the monitoring alone. The question then with increased morbidity and mortality and is
becomes how the monitoring influenced treat- amenable to treatment. In the absence of evi-
ment, a question that is difficult to answer since dence to the contrary, increased intracranial
the two groups received aggressive therapy to pressure should be detected and treated prompt-
lower the intracranial pressure. It is also likely ly (e.g., with surgical removal of intracranial he-
that the trial had an inadequate sample size. In matomas) and, whenever possible, should be pre-
clinical practice, at most 50 to 60% of patients vented with appropriate intensive care.
with traumatic brain injury who are undergoing Currently, invasive monitoring is the only reli-

310 may 29, 2014
The New EnglandFor

reserved.
able method for detecting and monitoring raised search will be needed to reduce the massive and
intracranial pressure in daily practice. The tech- growing burden of traumatic brain injury.
nique is controversial, since it clearly carries risks
Dr. Stocchetti reports receiving consulting fees from Orsan
and side effects and its usefulness has not been Medical Technologies and lecture fees from BHR Pharma; and
conclusively shown. Further uncertainty remains Dr. Maas, receiving consulting fees through his institution from
concerning specific medical and surgical therapies NeuroVive, Sanofi Aventis, and Shire. No other potential con-
flict of interest relevant to this article was reported.
for intracranial hypertension. Ongoing trials42,50 Disclosure forms provided by the authors are available with
will provide important new data, but more re- the full text of this article at NEJM.org.
References
1. Langlois JA, Rutland-Brown W, Wald matic brain injury. J Neurosurg 2008;109: prophylaxis. J Neurol Neurosurg Psychi-
MM. The epidemiology and impact of 678-84. atry 2000;69:381-4.
traumatic brain injury: a brief overview. 14. Guillaume J, Janny P. Manométrie intra- 24. Brain Trauma Foundation, American
J Head Trauma Rehabil 2006;21:375-8. crânienne continue: intérêt de la méthode Association of Neurological Surgeons
2. Maas AI, Stocchetti N, Bullock R. et premiers résultats. Rev Neurol (Paris) (AANS), Congress of Neurological Sur-
Moderate and severe traumatic brain inju- 1951;84:131-42. geons (CNS), AANS/CNS Joint Section on
ry in adults. Lancet Neurol 2008;7:728-41. 15. Lundberg N. Continuous recording Neurotrauma and Critical Care. Guidelines
3. Thompson HJ, McCormick WC, Kagan and control of ventricular fluid pressure for the management of severe traumatic
SH. Traumatic brain injury in older adults: in neurosurgical practice. Acta Psychiatr brain injury. Methods. J Neurotrauma 2007;
epidemiology, outcomes, and future impli- Scand Suppl 1960;36(149):1-193. 24:Suppl 1:S3-S6.
cations. J Am Geriatr Soc 2006;54:1590-5. 16. Brain Trauma Foundation, American 25. Cooper DJ, Rosenfeld JV, Murray L,
4. Roozenbeek B, Maas AIR, Menon DK. Association of Neurological Surgeons et al. Decompressive craniectomy in dif-
Changing patterns in the epidemiology of (AANS), Congress of Neurological Surgeons fuse traumatic brain injury. N Engl J Med
traumatic brain injury. Nat Rev Neurol (CNS), AANS/CNS Joint Section on Neuro- 2011;364:1493-502.
3013;9:231-6. trauma and Critical Care. Guidelines for 26. Brain Trauma Foundation, American
5. Teasdale G, Jennett B. Assessment and the management of severe traumatic brain Association of Neurological Surgeons
prognosis of coma after head injury. Acta injury. VII. Intracranial pressure monitor- (AANS), Congress of Neurological Sur-
Neurochir (Wien) 1976;34:45-55. ing technology. J Neurotrauma 2007;24: geons (CNS), AANS/CNS Joint Section on
6. Mushkudiani NA, Engel DC, Steyer- Suppl 1:S45-S54. Neurotrauma and Critical Care. Guidelines
berg EW, et al. Prognostic value of demo- 17. Idem. Guidelines for the management for the management of severe traumatic
graphic characteristics in traumatic brain of severe traumatic brain injury. VI. Indica- brain injury. XI. Anesthetics, analgesics,
injury: results from the IMPACT study. tions for intracranial pressure monitoring. and sedatives. J Neurotrauma 2007;24:
J Neurotrauma 2007;24:259-69. J Neurotrauma 2007;24:Suppl 1:S37-S44. Suppl 1:S71-S76.
7. Stocchetti N, Paternò R, Citerio G, 18. Maas AI, Dearden M, Teasdale GM, et 27. Vespa PM, Nuwer MR, Nenov V, et al.
Beretta L, Colombo A. Traumatic brain al. EBIC-guidelines for management of Increased incidence and impact of non-
injury in an aging population. J Neuro- severe head injury in adults: European convulsive and convulsive seizures after
trauma 2012;29:1119-25. Brain Injury Consortium. Acta Neurochir traumatic brain injury as detected by con-
8. Zumkeller M, Behrmann R, Heissler (Wien) 1997;139:286-94. tinuous electroencephalographic monitor-
HE, Dietz H. Computed tomographic cri- 19. Bauer DF, Razdan SN, Bartolucci AA, ing. J Neurosurg 1999;91:750-60.
teria and survival rate for patients with Markert JM. Meta-analysis of hemorrhagic 28. Olivecrona M, Zetterlund B, Rodling-
acute subdural hematoma. Neurosurgery complications from ventriculostomy place- Wahlström M, Naredi S, Koskinen LO.
1996;39:708-12. ment by neurosurgeons. Neurosurgery 2011; Absence of electroencephalographic seizure
9. Gopinath SP, Robertson CS, Contant 69:255-60. activity in patients treated for head injury
CF, et al. Jugular venous desaturation and 20. Poca MA, Sahuquillo J, Arribas M, with an intracranial pressure-targeted ther-
outcome after head injury. J Neurol Neuro- Báguena M, Amorós S, Rubio E. Fiberoptic apy. J Neurosurg 2009;110:300-5.
surg Psychiatry 1994;57:717-23. intraparenchymal brain pressure monitor- 29. Ropper AH. Hyperosmolar therapy for
10. Nordström CH, Reinstrup P, Xu W, ing with the Camino V420 monitor: reflec- raised intracranial pressure. N Engl J Med
Gardenfors A, Ungerstedt U. Assessment tions on our experience in 163 severely 2012;367:746-52.
of the lower limit for cerebral perfusion head-injured patients. J Neurotrauma 2002; 30. Stocchetti N, Maas AI, Chieregato A,
pressure in severe head injuries by bed- 19:439-48. van der Plas AA. Hyperventilation in
side monitoring of regional energy metab- 21. Slichter SJ. Evidence-based platelet head injury: a review. Chest 2005;127:
olism. Anesthesiology 2003;98:809-14. transfusion guidelines. Hematology (Am 1812-27.
11. Marmarou A, Anderson RL, Ward JD, Soc Hematol Educ Program) 2007:172-8. 31. Brain Trauma Foundation, American
et al. Impact of ICP instability and hypo- 22. Brain Trauma Foundation, American Association of Neurological Surgeons
tension on outcome in patients with se- Association of Neurological Surgeons (AANS), Congress of Neurological Sur-
vere head trauma. J Neurosurg 1991;75: (AANS), Congress of Neurological Sur- geons (CNS), AANS/CNS Joint Section on
Suppl:S59-S66. geons (CNS), AANS/CNS Joint Section on Neurotrauma and Critical Care. Guidelines
12. Stocchetti N, Zanaboni C, Colombo A, Neurotrauma and Critical Care. Guidelines for the management of severe traumatic
et al. Refractory intracranial hypertension for the management of severe traumatic brain injury. XIV. Hyperventilation. J Neu-
and “second-tier” therapies in traumatic brain injury. IV. Infection prophylaxis. rotrauma 2007;24:Suppl 1:S87-S90.
brain injury. Intensive Care Med 2008;34: J Neurotrauma 2007;24:Suppl 1:S26-S31. 32. Roberts I, Sydenham E. Barbiturates
461-7. 23. Rebuck JA, Murry KR, Rhoney DH, for acute traumatic brain injury. Cochrane
13. Vik A, Nag T, Fredriksli OA, et al. Michael DB, Coplin WM. Infection related Database Syst Rev 2012;12:CD000033.
Relationship of “dose” of intracranial to intracranial pressure monitors in adults: 33. Polderman KH. Induced hypothermia
hypertension to outcome in severe trau- analysis of risk factors and antibiotic and fever control for prevention and treat-

311 may 29, 2014 2129
The New
use only. Medicine
Society. Forrights
reserved.
ment of neurological injuries. Lancet 2008; EC, Anderson FA. Erroneous measurement sive management on mortality in severe
371:1955-69. of intracranial pressure caused by simulta- head injury. J Neurosurg 1982;56:498-503.
34. Clifton GL, Miller ER, Choi SC, et al. neous ventricular drainage: a hydrodynamic 46. Farahvar A, Gerber LM, Chiu YL, Car-
Lack of effect of induction of hypothermia model study. Neurosurgery 1989;24:348-54. ney N, Härtl R, Ghajar J. Increased mortal-
after acute brain injury. N Engl J Med 40. Grady MS. Lumbar drainage for in- ity in patients with severe traumatic brain
2001;344:556-63. creased intracranial pressure. J Neurosurg injury treated without intracranial pressure
35. Clifton GL, Valadka A, Zygun D, et al. 2009;110:1198-9. monitoring. J Neurosurg 2012;117:729-34.
Very early hypothermia induction in pa- 41. Stiver SI. Complications of decom- 47. Chesnut RM, Temkin N, Carney N,
tients with severe brain injury (the National pressive craniectomy for traumatic brain et al. A trial of intracranial-pressure mon-
Acute Brain Injury Study: Hypothermia II): injury. Neurosurg Focus 2009;26(6):E7. itoring in traumatic brain injury. N Engl J
a randomised trial. Lancet Neurol 2011;10: 42. RESCUEicp: Randomised Evaluation Med 2012;367:2471-81.
131-9. of Surgery with Craniectomy for Uncon- 48. National Research Council Committee
36. Bullock MR, Chesnut R, Ghajar J, et al. trollable Elevation of Intra-Cranial Pressure on a Framework for Developing a New
Guidelines for the surgical management home page (http://www.rescueicp.com). Taxonomy of Disease. Toward precision
of traumatic brain injury. Neurosurgery 43. Cremer OL, van Dijk GW, van Wensen medicine. Washington, DC: National Acad-
2006;58:Suppl:S7-S60. E, et al. Effect of intracranial pressure emies Press, 2011.
37. Tanaka H, Katayama Y, Kawamata T, monitoring and targeted intensive care on 49. Miller JD, Piper IR, Dearden NM.
Tsubokawa T. Excitatory amino acid re- functional outcome after severe head in- Management of intracranial hypertension
lease from contused brain tissue into sur- jury. Crit Care Med 2005;33:2207-13. in head injury: matching treatment with
rounding brain areas. Acta Neurochir Suppl 44. Shafi S, Diaz-Arrastia R, Madden C, cause. Acta Neurochir (Wien) 1993;S57:
(Wien) 1994;60:524-7. Gentilello L. Intracranial pressure moni- 152-9.
38. Kawamata T, Katayama Y. Surgical toring in brain-injured patients is associ- 50. Eurotherm3235Trial: International
management of early massive edema caused ated with worsening of survival. J Trauma Standard Randomised Controlled Trial
by cerebral contusion in head trauma pa- 2008;64:335-40. Number Register 34555414 (http://www
tients. Acta Neurochir Suppl 2006;96:3-6. 45. Saul TG, Ducker TB. Effect of intra- .controlled-trials.com/ISRCTN34555414).
39. Wilkinson HA, Yarzebsji J, Wilkinson cranial pressure monitoring and aggres- Copyright © 2014 Massachusetts Medical Society.

2130 312nejm.org
n engl j med 370;22 may 29, 2014
The New England

MULARCZYK on January Medical
© Massachusetts 15, 2015.Society.
For personal use only.
reserved.
c or r e sp ondence
To the Editor: We think that Stocchetti and To the Editor: Stocchetti and Maas highlight
Maas, in their review article (May 29 issue),1 the factors that contribute to increased intracra-
should have provided more information on the nial hypertension. However, their article over-
use of transcranial Doppler ultrasonography for looks head position as a major factor that affects
noninvasive monitoring of intracranial pressure. intracranial hypertension in patients with trau-
With the use of this approach, the pulsatility in- matic brain injury.
dex, which is dependent on distal vascular resis- Patients with traumatic brain injury benefit
tance, is strongly correlated with intracranial- from increased cerebral venous outflow, which
pressure values in various clinical conditions.2,3 results from elevation of the head and maintain-
This approach also allows evaluation of the auto- ing the head in a neutral position. Elevation by
regulation of cerebral blood flow, which cannot 15 to 30 degrees is also associated with a shift
be evaluated by means of invasive devices.4 How- of the cerebrospinal fluid from the intracranial
ever, transcranial Doppler ultrasonography can- compartment to the spinal compartment. The
not replace invasive monitors of intracranial consequent reduction in intracranial pressure,
pressure, since it is not a continuous-monitoring cerebral blood flow, or both after this elevation
device. Its best role might be as a screening tool of the head is favorable in most patients.1,2 If
to determine which patients have elevated intra- this maneuver is performed during monitoring
cranial pressure.5 of systemic arterial and intracranial pressure, to
maintain appropriate cerebral perfusion pressure,
Jean Cotte, M.D.
the transducers must be appropriately set to zero
Pierre Esnault, M.D.
at the level of the foramen of Monro or (more
Bertrand Prunet, M.D.
practically) at the level of the external auditory
Saint Anne Military Teaching Hospital
Toulon, France
meatus.
jean.cotte@gmail.com Ata Mahmoodpoor, M.D.
No potential conflict of interest relevant to this letter was re- Samad E.J. Golzari, M.D.
ported.
Tabriz University of Medical Sciences
Tabriz, Iran
1. Stocchetti N, Maas AIR. Traumatic intracranial hyperten-
dr.golzari@hotmail.com
sion. N Engl J Med 2014;370:2121-30.
2. Wakerley BR, Kusuma Y, Yeo LL, et al. Usefulness of trans- No potential conflict of interest relevant to this letter was re-
cranial Doppler-derived cerebral hemodynamic parameters in ported.
the noninvasive assessment of intracranial pressure. J Neuroim-
aging 2014 March 5 (Epub ahead of print). 1. Feldman Z, Kanter MJ, Robertson CS, et al. Effect of head
3. Gura M, Elmaci I, Sari R, Coskun N. Correlation of pulsatil- elevation on intracranial pressure, cerebral perfusion pressure,
ity index with intracranial pressure in traumatic brain injury. and cerebral blood flow in head-injured patients. J Neurosurg
Turk Neurosurg 2011;21:210-5. 1992;76:207-11.
4. Panerai RB. Transcranial Doppler for evaluation of cerebral 2. Ng I, Lim J, Wong HB. Effects of head posture on cerebral
autoregulation. Clin Auton Res 2009;19:197-211. hemodynamics: its influences on intracranial pressure, cerebral
5. Tazarourte K, Atchabahian A, Tourtier JP, et al. Pre-hospital perfusion pressure, and cerebral oxygenation. Neurosurgery
transcranial Doppler in severe traumatic brain injury: a pilot 2004;54:593-7.
study. Acta Anaesthesiol Scand 2011;55:422-8.
DOI: 10.1056/NEJMc1407775
DOI: 10.1056/NEJMc1407775

313 september 4, 2014 965
The Authors Reply: Cotte and coworkers un- perfusion pressure induced by elevation of the
derscore the potential of transcranial Doppler head. In the study by Feldman et al. that was
ultrasonography for noninvasive assessment of cited by Mahmoodpoor and Golzari, 5 patients
intracranial pressure. The evidence supporting (23%) had reduced cerebral blood flow owing to
transcranial Doppler ultrasonography as an in- elevation of the head, although no adverse ef-
dicator of elevated intracranial pressure is ex- fects on cerebral perfusion pressure and cerebral
tremely controversial. This method has been blood flow were reported when data were aver-
used to identify children who are at risk for ele- aged for the whole group.
vated intracranial pressure, with unreliable re- It is our practice to check whether moderate
sults.1 After initial enthusiasm about the use of elevation of the head (by 15 to 30 degrees) may
this method in adults, worrisome results have contribute to control of intracranial pressure
been reported, especially in patients with pre- without dangerous reductions in cerebral perfu-
existing vascular diseases.2 A comprehensive sion pressure in individual patients. In many
analysis by the Cambridge Group3 involving 290 cases, as indicated by our colleagues, slight ele-
patients with head injury concluded that “overall vation of the head helps, but in some cases it
the value of transcranial Doppler ultrasonogra- does not. Perhaps this is a good example of in-
phy to assess intracranial pressure noninvasively dividualized therapy, which we advocated in our
is very limited.” As Cotte and coauthors admit, criticism of “one size fits all” approaches.
transcranial Doppler ultrasonography cannot re- Nino Stocchetti, M.D.
place invasive monitoring of intracranial pres- Milan University
sure. In our opinion, its potential for screening Milan, Italy
patients who are at risk for elevated intracranial
Andrew I.R. Maas, M.D., Ph.D.
pressure remains insufficiently proved.
Antwerp University Hospital
Mahmoodpoor and Golzari recommend ele- Edegem, Belgium
vation of the head. We fully agree that venous Since publication of their article, Dr. Stocchetti reports re-
outflow from the brain is extremely important ceiving an honorarium from Johnson & Johnson. No further
and that any cause of elevated venous pressure, potential conflict of interest relevant to this letter was reported.
as in the case of neck torsion or compression, 1. Figaji AA, Zwane E, Fieggen AG, Siesjo P, Peter JC. Transcra-
should be avoided, as we proposed in Table 2 of nial Doppler pulsatility index is not a reliable indicator of intra-
cranial pressure in children with severe traumatic brain injury.
our article and in our algorithm for the treat- Surg Neurol 2009;72:389-94.
ment of intracranial pressure (Fig. 2 of the arti- 2. Ahmad M, Legrand M, Lukaszewicz AC, Charlier P, Mateo J,
cle). However, we strongly suggest that cerebral Payen D. Transcranial Doppler monitoring may be misleading in
prediction of elevated ICP in brain-injured patients. Intensive
perfusion pressure should be preserved when Care Med 2013;39:1150-1.
intracranial pressure is treated. Rosner and 3. Zweifel C, Czosnyka M, Carrera E, de Riva N, Pickard JD,
Coley4 warned that elevation of the head was as- Smielewski P. Reliability of the blood flow velocity pulsatility
index for assessment of intracranial and cerebral perfusion pres-
sociated with a risk of reducing intracranial sures in head-injured patients. Neurosurgery 2012;71:853-61.
pressure at the expense of an even greater reduc- 4. Rosner MJ, Coley IB. Cerebral perfusion pressure, intracra-
tion in cerebral perfusion pressure. In their se- nial pressure, and head elevation. J Neurosurg 1986;65:636-41.
ries involving 18 patients, 4 had further increas- DOI: 10.1056/NEJMc1407775

es in intracranial pressure after the decrease in
966 n engl j med 371;10 314


What is the most appropriate management of this patient’s traumatic brain injury?
A. Conservative management, since there is no prospect for functional recovery.

B. Evacuation of the acute subdural hematoma, placement of an external ventricular drain, and admission to the
ICU for active management of increased intracranial pressure.
C. Evacuation of the acute subdural hematoma and admission to the ICU for management without monitoring
of intracranial pressure.

After 15 days of treatment in an ICU, our patient is likely to have important sequelae, including decreased muscle
strength and impaired locomotor function. Falls among hospitalized patients are an important safety issue, with
an estimated occurrence rate of 2.1 to 8.4 falls per 1000 bed-days, according to one study,1 which showed that 1 in
200 falls may result in severe harm to the patient.
Our patient has had an acute traumatic brain injury as a result of his fall. When he is first examined, his score
on the Glasgow Coma Scale is 9. He has lateralizing signs suggestive of either an acute stroke or a traumatic intra-
cranial hemorrhage affecting the left cerebral hemisphere. His CT scan confirms an acute traumatic subdural
hematoma. Traumatic brain injury has a bimodal age distribution, with the incidence highest in young adults
and the elderly; among the latter, falls are an increasingly common cause. Although an older age is strongly and
independently associated with a poor outcome after traumatic brain injury, acceptable recovery is now recognized
as possible after sustained, high-level medical care and rehabilitation.
In the absence of an advance directive to the contrary, it would be reasonable to evacuate the patient’s acute sub-
dural hematoma and place an intracranial-pressure monitor. Although monitoring of intracranial pressure has
not been shown to improve outcomes,2 international guidelines recommend such monitoring and the use of the
measurements as a guide to treatment for all patients with severe traumatic brain injury with potential for
recovery.3
Strategies for the management of increased intracranial pressure generally involve a stepwise approach, starting
with mechanical ventilation to ensure normocapnia, sedation, and drainage of cerebrospinal fluid (in patients
in whom a ventricular drain has been placed). Second-line measures include osmolar therapy and induced hypo-
capnia and hypothermia with metabolic suppression (barbiturate coma), with decompressive craniectomy being
reserved for young and previously well patients with refractory intracranial hypertension. Since our patient is
elderly and has an uncertain prognosis, a reasonable approach would be to sedate and mechanically ventilate
the patient and place a drain for the removal of cerebrospinal fluid, with assessment at 48 to 72 hours for signs
suggesting a favorable recovery. In the absence of signs of recovery or if refractory intracranial hypertension de-
velops, changing the focus of treatment to palliation would be indicated.
REFERENCES
1. Healey F, Scobie S, Oliver D, Pryce A, Thomson R, Glampson B. Falls in English and Welsh hospitals: a national
observational study based on retrospective analysis of 12 months of patient safety incident reports. Qual Saf Health
Care 2008;17:424-30.
315
2. Chesnut RM, Temkin N, Carney N, et al. A trial of intracranial-pressure monitoring in traumatic brain injury.
N Engl J Med 2012;367:2471-81.
3. Brain Trauma Foundation. Guidelines for the management of severe traumatic brain injury. J Neurotrauma
2007;24:Suppl 1:S1-106.
316
DYING WITH DIGNITY

IN THE INTENSIVE CARE UNIT
Our efforts to stem illness are not always successful. This article covers the steps that one can follow to make
sure that a patient and his or her family are prepared for the end of life when it is clear that a patient is unlikely
to survive.
317
Case Challenge
Dying with Dignity in the Intensive Care Unit

After a long ICU stay because of septic shock and multiple complications, a frail 77-year-old man had a fall
and suffered an acute subdural hematoma and hemorrhagic contusion. Twelve days after evacuation of the
subdural hematoma, he remains in a coma and is still receiving mechanical ventilation. How should decisions
be made about further treatment?
heavy alcohol intake, and mild cognitive impairment required 15 days of treatment in the intensive care unit
(ICU) of a university hospital for septic shock due to fecal peritonitis from a perforated sigmoid colon. After
surgery, he was placed on a mechanical ventilator. Complications during his ICU stay included mild disseminated
intravascular coagulation and acute hepatic necrosis associated with acetaminophen treatment. After being trans-
ferred to the surgical floor, he had an unwitnessed fall from his bed, and cranial computed tomography showed an
acute subdural hematoma with underlying hemorrhagic contusion on the left side and 5-mm displacement of the
midline intracranial structures. (In the previous installment of this case, there were 2898 votes on how to treat the
patient’s traumatic brain injury. More than two thirds of the respondents [67%] favored evacuation of the acute
subdural hematoma, placement of an external ventricular drain, and admission to the ICU for active management
318
of increased intracranial pressure. Another 26% favored evacuation of the acute subdural hematoma and admis-
sion to the ICU without monitoring of intracranial pressure, whereas 6% opted for conservative management with
the rationale that there was no prospect for functional recovery.)
He was taken to the operating room and underwent a craniotomy for evacuation of the subdural hematoma.
Twelve days after this procedure, during which he received no sedative medications, he remains in a coma with
a best motor response of abnormal flexion on the left side. He is still receiving mechanical ventilation. The con-
sensus opinion of treating clinicians is that he will most likely not make a functional recovery.

How should decisions be made about this patient’s further treatment?
A. Meet with the family and tell them there is no hope of recovery and you will be discontinuing mechanical
ventilation.
B. Meet with the family with the goal of ascertaining the patient’s wishes regarding the continuation of active
medical treatment and whether the likely outcome of continued treatment — at best, placement in a nursing
home with a high level of care — would be consistent with the patient’s wishes.
C. Meet with the family and ask them whether they would like the patient to have a tracheostomy.
D. Meet with the family and tell them you will perform a tracheostomy.
319
review article

Dying with Dignity in the Intensive Care Unit

Deborah Cook, M.D., and Graeme Rocker, D.M.
T
From the Departments of Medicine, Clin- he traditional goals of intensive care are to reduce the
ical Epidemiology, and Biostatistics, Mc- morbidity and mortality associated with critical illness, maintain organ func-
Master University, Hamilton, ON (D.C.),
and the Department of Medicine, Dalhousie tion, and restore health. Despite technological advances, death in the intensive
University, Halifax, NS (G.R.) — both in care unit (ICU) remains commonplace. Death rates vary widely within and among
Canada. Address reprint requests to Dr. countries and are influenced by many factors.1 Comparative international data are lack-
Cook at the McMaster University Health
Sciences Center, Rm. 2C11, 1200 Main St. W, ing, but an estimated one in five deaths in the United States occurs in a critical care bed.2
Hamilton, ON, Canada, L8N 3Z5, or at In this review, we address the concept of dignity for patients dying in the ICU.
debcook@mcmaster.ca. When the organ dysfunction of critical illness defies treatment, when the goals of care
N Engl J Med 2014;370:2506-14. can no longer be met, or when life support is likely to result in outcomes that are
DOI: 10.1056/NEJMra1208795 incongruent with patients’ values, ICU clinicians must ensure that patients die with
dignity. The definition of “dying with dignity” recognizes the intrinsic, unconditional
quality of human worth but also external qualities of physical comfort, autonomy,
meaningfulness, preparedness, and interpersonal connection.3 Respect should be
fostered by being mindful of the “ABCDs” of dignity-conserving care (attitudes,
behaviors, compassion, and dialogue)4 (Table 1). Preserving the dignity of patients,
avoiding harm, and preventing or resolving conflict are conditions of the privilege
and responsibility of caring for patients at the end of life. In our discussion of prin-
ciples, evidence, and practices, we assume that there are no extant conflicts between
the ICU team and the patient’s family. Given the scope of this review, readers are
referred elsewhere for guidance on conflict prevention and resolution in the ICU.5,6
The concept of dying with dignity in the ICU implies that although clinicians
may forgo some treatments, care can be enhanced as death approaches. Funda-
mental to maintaining dignity is the need to understand a patient’s unique per-
spectives on what gives life meaning in a setting replete with depersonalizing
devices. The goal is caring for patients in a manner that is consistent with their
values at a time of incomparable vulnerability, when they rarely can speak for
themselves.7 For example, patients who value meaningful relationships may de-
cline life-prolonging measures when such relationships are no longer possible.
Conversely, patients for whom physical autonomy is not crucial may accept tech-
nological dependence if it confers a reasonable chance of an acceptable, albeit
impaired, outcome.8 At issue is what each patient would be willing to undergo for
a given probability of survival and anticipated quality of life.
On the Need for Pa l l i at i v e C a r e
The coexistence of palliative care and critical care may seem paradoxical in the
technological ICU. However, contemporary critical care should be as concerned
with palliation as with the prevention, diagnosis, monitoring, and treatment of life-
threatening conditions.
2506 320370;26
n engl j med nejm.org june 26, 2014
The New
use only. Medicine
Downloaded from nejm.org© on
Copyright January 28, 2015.
Massachusetts ForSociety.
Medical personalAll
userights
only.reserved.
The World Health Organization defines pal-

Table 1. Examples of the ABCDs of Dignity-Conserving Care.*
liative care as “an approach that improves the
quality of life of patients and their families Attitudes and assumptions can affect practice.
facing the problems associated with life-threat- Reflect on how your own life experiences affect the way in which you
ening illness, through the prevention and relief provide care.
of suffering by means of early identification Be aware that other clinicians’ attitudes and assumptions can affect their
and impeccable assessment and treatment of approach to patients.
pain and other problems, physical, psychosocial Teach learners to be mindful of how their perspectives and presumptions
can shape behaviors.
and spiritual.”9 Palliative care, which is essen-
tial regardless of whether a medical condition Behaviors should always enhance patient dignity.
is acute or chronic and whether it is in an early Demonstrate with nonverbal methods how patients and their families are
or a late stage, can also extend beyond the pa- important to you.
tient’s death to bereaved family members10 Do not rush; sit down and make eye contact when talking with patients
and their families.
(Fig. 1).
Turn off digital devices and avoid jargon when talking with patients and
their families.
El ici t ing the Va lue s of Pat ien t s
Compassion is sensitivity to the suffering of another and the desire to
relieve it.
Sometimes it is too late. A precipitating event
Elicit the personal stories that accompany your patient’s illness.
prompting an ICU admission that occurs within
a protracted downward trajectory of an illness Acknowledge the effect of sickness on your patient’s broader life
experience.
may be irreversible. When clinicians who are car-
Recognize and relieve suffering.
ing for a patient in such a scenario have not pre-
viously explored whether the patient would want Dialogue should acknowledge personhood beyond the illness.
to receive basic or advanced life support, the Explore the values that are most important to your patients.
wishes of the patient are unknown, and invalid Ask who else should be involved to help your patients through difficult
assumptions can be anticipated. Effective ad- times.
vance care planning, which is often lacking in Encourage patients and their families to reflect and reminisce.
such circumstances, elicits values directly from
* This approach is adapted from Chochinov.4
the patient, possibly preventing unnecessary suf-
fering associated with the use of unwelcome in-
terventions and thereby preserving the patient’s rative from patients (or more commonly, from
dignity at the end of life. family members) about relationships, activities,
Regardless of the rate and pattern of decline and experiences treasured by the patient. The
in health, by the time that patients are in the use of engaging, deferential questions, such as
ICU, most cannot hold a meaningful conversa- “Tell me about your . . .” or “Tell us what is im-
tion as a result of their critical condition or portant to . . . ,” is essential. Clinician guidance
sedating medications. In such cases, family for constructing an authentic picture of the in-
members or other surrogates typically speak capacitated patient’s values is offered in the Fa-
for them. In decisions regarding the withdraw- cilitated Values History,8 a framework that pro-
al of life support, the predominant determi- vides clinicians with strategies for expressing
nants are a very low probability of survival, a empathy, sensitively depicting common scenarios
very high probability of severely impaired cog- of death, clarifying the decision-making role of
nitive function, and recognition that patients surrogates, eliciting and summarizing values most
would not want to continue life support in such relevant to medical decision making, and link-
circumstances if they could speak for them- ing these values explicitly to care plans.
selves.11 Probabilistic information is thus often
more important than the patient’s age, coexist- C om munic at ion
ing medical conditions, or illness severity in
influencing decisions about life-support with- Before a critical illness develops, patients’ percep-
drawal. tions about what matters most for high-quality
Discussions can be initiated by eliciting a nar- end-of-life care vary, but human connections are
n engl j med 370;26 nejm.org june 26, 2014

321 2507
The New England
For of Medicine
Downloaded from nejm.org on January
Copyright 28, 2015. For personal
© Massachusetts Medicaluse only. All
Society. No rights
reserved.
mation but assumes the primary responsibility

for decision making. At the other end of the con-
tinuum, the patient makes the decisions, and the
Intensity of Care
Curative care physician has an advisory role. In North America

and in some parts of Europe,17 the archetype is
Palliative care
the shared decision-making model, in which phy-
sicians and patients or their surrogates share in-
formation with one another and participate jointly
in decision making.18
Bereavement
Admission Time
Although preferences for decision-making roles
Death
to ICU vary among family members,19 physicians do not
always clarify family preferences.20 Family mem-
Figure 1. Curative and Palliative Approaches to Care throughout a Critical Illness.
bers may lack confidence about their surrogate
This diagram, which is adapted from a policy statement of the American
decision-maker role, regardless of the decision-
Thoracic Society,10 illustrates the relative intensity of curative and palliative
approaches to the care of patients at different stages of a critical illness. making model, if they have had no experience
In the palliative care model, the intensity of care increases at the end of life, as a surrogate or no prior dialogue with the
and support of the patient’s family continues beyond the patient’s death. patient about treatment preferences.21 Decision-
making burden is postulated as a salient source
of strain among family members of patients
key. Many seriously ill elderly patients cite effec- who are dying in the ICU; anxiety and depres-
tive communication, continuity of care, trust in sion are also prevalent.22,23
the treating physician, life completion, and avoid-
ance of unwanted life support.12 After critical ill- Prov iding Pro gnos t ic
ness develops, most patients or their surrogates Infor m at ion
find themselves communicating with unfamiliar
clinicians in a sterile environment at a time of Valid prognostic information is a fundamental
unparalleled distress. Challenges in communica- component of end-of-life discussions. Under-
tion are magnified when patients die at an early standing the predicted outcome of the critical
stage of critical illness, before rapport has been illness and recognizing the uncertainty of that
well established. prediction are helpful in making decisions that
Clear, candid communication is a determinant reflect the patient’s values. However, when it
of family satisfaction with end-of-life care.13 comes to prognosticating for seriously ill pa-
Notably, measures of family satisfaction with tients, families and physicians sometimes dis-
respect to communication are higher among agree.24 In one study, surrogate decision mak-
family members of patients who die in the ICU ers for 169 patients in the ICU were randomly
than among those of ICU patients who survive, assigned to view one of two videos of a simu-
perhaps reflecting the intensity of communica- lated family conference about a hypothetical
tion and the accompanying respect and compas- patient.25 The videos varied only according to
sion shown by clinicians for the families of dying whether the prognosis was conveyed in numeri-
patients.14 The power of effective communica- cal terms (“10% chance of survival”) or qualita-
tion also includes the power of silence.15 Family tive terms (“very unlikely to survive”). Numeri-
satisfaction with meetings about end-of-life care cal prognostic statements were no better than
in the ICU may be greater when physicians talk qualitative statements in conveying the progno-
less and listen more.16 sis. However, on average, surrogates estimated
twice as often as physicians that the patient
Decision M a k ing would survive.
In another study, when 80 surrogates of pa-
Decision-making models for the ICU vary inter- tients in the ICU interpreted 16 prognostic state-
nationally but should be individualized. At one ments, interviews suggested an “optimism bias,”
end of the continuum is a traditional parental in which the surrogates were likely to interpret
approach, in which the physician shares infor- the physicians’ grim prognostication as positive
2508 n engl j med

322370;26 nejm.org june 26, 2014
The New
Downloaded from nejm.org©on
Medical personalAll
userights
reserved.
with respect to the patient’s condition.26 Clini- Prov iding Hol is t ic C a r e

cians should recognize that family members
who are acting as spokespersons for patients in Cultivating culturally and spiritually sensitive
the ICU are often “living with dying” as they care is central to the palliative approach. The pil-
face uncertainty while maintaining hope.27 Hope lars of both verbal and nonverbal communica-
should be respected during prognostic disclo- tion are crucial. Conscious nonverbal communi-
sure28 while a realistic view is maintained, an cation is rarely practiced yet can be as powerful
attitude that is aptly expressed by the simple but as verbal communication during end-of-life de-
profound notion of “hoping for the best but precision making. Physicians should be aware of
paring for the worst.”29 the cultural landscape reflecting an institution’s
catchment area, how cultural norms can influ-
M a k ing R ec om mendat ions ence admissible dialogue, and what is desirable
versus dishonoring in the dying process.33
Physicians in the ICU sometimes make recommen- The meaning assigned to critical illness, par-
dations to forgo the use of life-support technology. ticularly when death looms, is frequently inter-
In one study involving surrogates of 169 critically preted through a spiritual lens. For many people,
ill patients, 56% preferred to receive a physician’s critical illness triggers existential questions about
recommendation on the use of life support, 42% purpose (of life, death, and suffering), relation-
preferred not to receive such a recommendation, ships (past, present, and future), and destiny.
and 2% stated that either approach was accept- Clinicians should be able to pose questions about
able.30 A recent survey of ICU physicians showed spiritual beliefs that may bear on experiences
that although more than 90% were comfortable with respect to illness. Introductory queries can
making such recommendations and viewed them open doors, such as “Many people have beliefs
as appropriate, only 20% reported always providing that shape their lives and are important at times
recommendations to surrogates, and 10% reported like this. Is there anything that you would like
rarely or never doing so.31 In this study, delivering me to know?”34 A useful mnemonic for obtain-
such recommendations was associated with per- ing ancillary details is SPIRIT, which encom-
ceptions about the surrogate’s desire for, and agree- passes acknowledgment of a spiritual belief
ment with, the physician’s recommendations. system, the patient’s personal involvement with
Other potential influences are uncertainty, per- this system, integration with a spiritual commu-
sonal values, and litigation concerns. nity, ritualized practices and restrictions, impli-
Asking families about their desire for recom- cations for medical care, and terminal-events
mendations from physicians can be a starting planning 34 (Table 2).
point for shared deliberations about care plans.32 Although it is unrealistic to expect that clini-
Eliciting preferences for how patients or their cians will be familiar with the views of all the
families wish to receive information, particularly world religions regarding death, they should be
recommendations concerning life support, is not cognizant of how belief systems influence end-
an abnegation of responsibility but rather an ap- of-life care.35 Physicians may recommend differ-
proach that is likely to engender trust. Physi- ent approaches to similar situations, depending
cians should judiciously analyze each situation on their religious and cultural backgrounds, as
and align their language and approach with the has been self-reported36 and documented in ob-
preferred decision-making model, understand in- servational studies.37 Insensitivity to faith-based
terpersonal relationships, and avoid overempha- preferences for discussion and decision making
sizing a particular point of view. For example, in may amplify the pain and suffering of both pa-
the shared decision-making model of care for tients and their families. Clinicians should un-
dying patients, family discussions typically include derstand how spirituality can influence coping,
a review of the patient’s previous and present either positively or negatively.38 Chaplains are
status and prognosis, elicitation of the patient’s indispensable for addressing and processing
values, presentation of the physician’s recommen- existential distress, conducting life review, and
dations, deliberations, and joint decision making facilitating comforting prayers, rituals, or other
about ongoing levels of care. observances.

323 june 26, 2014 2509
The New
Medicine
Copyright on January 28, 2015.
© Massachusetts ForSociety.
Medical personalAll
use only.reserved.
delay death so that the patient can accomplish

Table 2. Taking a Spiritual History.*
short-term life goals.41 Whatever approach is used,
S for spiritual belief system individualized pharmacologic therapy, which de-
P for personal spirituality pends on prevailing levels of analgesia and seda-
I for integration with a spiritual community tion at the time of decisions to forgo life sup-
R for ritualized practices and restrictions
port, should ensure preemptive, timely alleviation
of dyspnea, anxiety, pain, and other distressing
I for implications for medical care
symptoms.42 Clinicians can mitigate the stress of
T for terminal-events planning family members by discussing what is likely to
* This approach is adapted from Maugans.34 The mnemonic
happen during the dying process (e.g., unusual
SPIRIT can be used to elicit a spiritual history from a patient sounds, changes in skin color, and agonal breath-
as part of the goal of providing sensitive, compassionate ing). Physician attendance is paramount to re-
end-of-life care.
evaluate the patient’s comfort and talk with the
family as needed (Table 4).
The Fina l S teps
C onsequence s for Cl inici a ns
If a shift is made in the goals of care from cure
to comfort, it should be orchestrated with grace Dying patients and their families in the ICU are
and should be individualized to the needs of the not alone in their suffering. For some clinicians,
patient.39 Before proceeding with end-of-life views about the suitability of advanced life sup-
measures, it is necessary to prepare staff mem- port that diverge from those of the patient or
bers and the patient’s room, as well as the pa- family can be a source of moral distress. Clini-
tient (Table 3). The panoply of basic and ad- cians who detect physical or psychic pain and
vanced life-support equipment and the other negative symptoms may suffer indirectly,
mechanics of their deployment or discontinua- yet deeply. Vicarious traumatization results from
tion are chronicled in multiple studies, as well repeated empathic engagement with sadness and
as in discussion documents, consensus state- loss,43 particularly when predisposing character-
ments from professional organizations, and istics amplify clinicians’ response to this work-
task-force reports.10,17,32,40 Strategies should be place stress. Clinicians should be aware of how
openly discussed and informed by the same bal- their emotional withdrawal or lability and “com-
ance of benefits, burdens, and respect for the passion fatigue” can jeopardize the care of dying
preferences of patients and their surrogates that patients and their families.
apply to other aspects of end-of-life care.10 Informal debriefing or case-based rounds,44
There is no single, universally accepted tech- local meetings with other professionals, modi-
nical approach. Admissible strategies in most set- fied work assignments, and other strategies may
tings include variations and combinations of non- help clinicians to cope with the distress.45 Formal
escalation of current interventions, withholding of bereavement counseling that is designed espe-
future interventions, and withdrawal of some or cially for involved clinicians can enhance aware-
all interventions, except those needed for com- ness about vicarious traumatization and encour-
fort. When life-support measures are withdrawn, age adaptive personal and professional coping
the process of withdrawal — immediate or grad- strategies.
ual discontinuation — must be considered care-
fully. Mechanical ventilation is the most com-
End - of-L ife C a r e a s a
mon life-support measure that is withdrawn.11 Qua l i t y-Improv emen t Ta rge t
However, even in the case of mechanical ventila-
tion, legal or faith-based requirements, societal Palliative care is now a mainstream matter for
norms, and physician preferences influence de- quality-improvement agendas in many ICUs. A de-
cisions about withdrawal.32 The initiation of non- cade ago, the Robert Wood Johnson Foundation
invasive ventilation with clear objectives for pa- Critical Care End-of-Life Peer Workgroup and
tients who are not already undergoing mechanical 15 associated nurse–physician teams in North
ventilation can sometimes reduce dyspnea and America conducted a review of reported practices
2510 n engl j med 370;26 324

nejm.org june 26, 2014
The New EnglandFor

on January©28, 2015. For personal
Massachusetts Medicaluse only. No
Society. Allother uses
rights without permission.
reserved.
Table 3. Practical Preparatory Procedures to Ensure Patient Dignity before Withdrawal of Life Support.
Prepare staff members

Review the planned procedures in detail with all relevant staff members.
Ensure that the referring physician is aware of the plans, if not already engaged.
Ensure that spiritual care services are offered, if not already engaged.
Remind staff members that all their actions should ensure the dignity of the patient.
Remind staff members that the patient and family are the unit of care.
Prepare a staffing schedule to maximize the continuity of care during the dying process, if possible.
Ensure that the bedside nurse has not been assigned to care for another acutely ill patient, if possible.
Ensure that the bedside nurse is experienced in palliative care; if not, change the assignment or arrange for
supervision to be provided by a nurse experienced in palliative care.
Ensure that physicians are readily available and do not abandon the patient or family.
Introduce the relevant housestaff members to the patient and family.
Introduce the respiratory therapist to the patient and family, when applicable.
Ensure that staff members minimize unnecessary noise immediately outside the room.
Prepare the patient’s room
Consider the comfort of the patient and family (e.g., lighting, temperature, personal items).
Liberalize visiting restrictions (e.g., timing, duration, number of visitors).
Remove unnecessary equipment.
Bring additional chairs into the room, if necessary.
Secure a quiet room for the family away from the bedside.
Prepare the patient
Position the patient as comfortably as possible.
Honor requests for cultural, spiritual, and religious rituals.
Dim the lighting on screens required for monitoring (e.g., electrocardiography).
Discontinue unnecessary monitoring (e.g., oximetry), unnecessary devices (e.g., feeding tubes), unnecessary tests
(e.g., blood work), and unnecessary treatments (e.g., enteral nutrition).
Discontinue medications that do not provide comfort and provide those that do.
Ensure that the patient is as calm and distress-free as possible before proceeding to withdraw life support.
for end-of-life care and named seven key domains conflicts, ethics consultations helped with con-
for quality improvement: patient- and family- flict resolution and reduced the duration of non-
centered decision making, communication, con- beneficial treatments that the patients received.48
tinuity of care, emotional and practical support, In a subsequent cluster-randomized trial involv-
symptom management, spiritual support, and ing 2318 patients in which investigators evaluat-
emotional and organizational support for ICU ed a five-component, clinician-focused end-of-life
clinicians.46 More than 100 potential interven- strategy,49 there were no significant differences
tions were identified as part of this project, di- between groups with respect to family satisfac-
rected at patients and their families, clinicians, tion with care, family or nurse ratings of the
ICUs, and health care systems. Candidate quality quality of dying, time to withdrawal of mechani-
indicators and “bundled indicators” can facili- cal ventilation, length of stay in the ICU, or other
tate measurement and performance feedback in palliative care indicators.
evaluating the quality of palliative care in ICU Favorable assessments of palliative care inter-
settings.47 ventions in the ICU are beginning to emerge. In
In a multicenter, randomized trial involving one study, family members of 126 dying patients
critically ill patients who were facing value-related in 22 ICUs were randomly assigned to participate
n engl j med 370;26 325

The New
use only. Medicine
Copyright © on January 28, 2015.
Massachusetts ForSociety.
Medical personalAll
use only.reserved.
Table 4. Considerations and Cautions in the Withdrawal of Life Support.*
Variable Considerations Cautions

Discontinuation of renal-replace- Confers a low risk of physical distress Death may take several days if this is
ment therapy the only advanced life support
withdrawn
Discontinuation of inotropes Confers no risk of physical distress Death may not occur quickly if the pa-
or vasopressors Death may occur quickly if the patient tient requires low doses, particular-
requires high doses, with or with- ly if mechanical ventilation is
out withdrawal of mechanical ven- ongoing
tilation
Weaning from inotropes Confers no risk of physical distress May prolong the dying process, partic-
or vasopressors ularly if the patient requires low
doses and this is the only life sup-
port withdrawn
Discontinuation of mechanical Confers risk of dyspnea Preemptive sedation is typically need-
ventilation Death may occur quickly if the patient ed to blunt air hunger due to rapid
requires high pressure settings or changes in mechanical ventilation
high oxygen levels Death may not occur quickly if the pa-
tient requires low pressure settings
or low oxygen levels
Weaning from mechanical Confers low risk of dyspnea May prolong the dying process, partic-
ventilation ularly if the patient requires low
pressure settings or low oxygen
levels and this is the only life sup-
port withdrawn
Extubation Confers risk of dyspnea Informing families about possible
Avoids discomfort and suctioning physical signs after extubation can
of endotracheal tube prepare and reassure them
Can facilitate oral communication Secretions may cause noisy breathing,
Allows for the most natural which may be reduced with the use
appearance of glycopyrrolate; the use of gluco-
corticoids may reduce stridor
Airway obstruction may occur; jaw
thrust or repositioning of the pa-
tient may help
Not advised if the patient has hemop-
tysis
* The choice regarding the type and dose of medications depends on prevailing levels of analgesia and sedation at the
time of the decision, the mode and sequence of the planned withholding or withdrawal of life support, and myriad other
factors.42 These factors preclude any specific dose recommendations. Physician availability for the family during the dy-
ing process is as important as individualized adjustment of medication.
in a standard end-of-life family conference or to were assigned to the standard-conference group,

participate in a proactive family conference and re- with no significant between-group difference in
ceive a brochure on bereavement.50 The mnemonic the length of stay in the ICU or the hospital.
“VALUE” framed the five objectives of the pro- Caregivers in the proactive-conference group, as
active family conference: value and appreciate compared with the standard-conference group,
what family members say, acknowledge the fam- were less negatively affected by the experience
ily members’ emotions, listen to their concerns, and were less likely to have anxiety, depression,
understand who the patient was in active life by and symptoms of post-traumatic stress 90 days
asking questions, and elicit questions from the after the patients’ deaths.
family members. Patients whose family mem-
bers were assigned to the proactive-conference C onclusions
group were treated with significantly fewer non-
beneficial interventions after the family confer- Palliative care in the ICU has come of age. Its
ence than were those whose family members guiding principles are more important than ever

326 june 26, 2014
The New EnglandFor

Copyright © 28, 2015. For personal
Massachusetts Medicaluse only. No
Society. All other
rightsuses without permission.
reserved.
in increasingly pluralistic societies. Ensuring port the bereaved family members. Ensuring death
that patients are helped to die with dignity begs with dignity in the ICU epitomizes the art of med-
for reflection, time, and space to create connec- icine and reflects the heart of medicine. It de-
tions that are remembered by survivors long after mands the best of us.
a patient’s death. It calls for humanism from all No potential conflict of interest relevant to this article was
reported.
clinicians in the ICU to promote peace during the Disclosure forms provided by the authors are available with
final hours or days of a patient’s life and to sup- the full text of this article at NEJM.org.
References
1. Adhikari NK, Fowler RA, Bhagwanjee 14. Wall RJ, Curtis JR, Cooke CR, Engel- methods to disclose prognosis to surro-
S, Rubenfeld GD. Critical care and the berg RA. Family satisfaction in the ICU: gate decision makers in intensive care
global burden of critical illness in adults. differences between families of survivors units. Am J Respir Crit Care Med 2010;
Lancet 2010;376:1339-46. and nonsurvivors. Chest 2007;132:1425- 182:905-9.
2. Angus DC, Barnato AE, Linde-Zwirble 33. 26. Zier LS, Sottile PD, Hong SY, Weiss-
WT, et al. Use of intensive care at the end 15. Lilly CM, Daly BJ. The healing power field LA, White DB. Surrogate decision
of life in the United States: an epidemio- of listening in the ICU. N Engl J Med 2007; makers’ interpretation of prognostic in-
logic study. Crit Care Med 2004;32:638- 356:513-5. formation: a mixed-methods study. Ann
43. 16. McDonagh JR, Elliott TB, Engelberg Intern Med 2012;156:360-6.
3. Proulx K, Jacelon C. Dying with dignity: RA, et al. Family satisfaction with family 27. Sinuff T, Giacomini M, Shaw R, Swin-
the good patient versus the good death. Am conferences about end-of-life care in the ton M, Cook DJ. “Living with dying”: the
J Hosp Palliat Care 2004;21:116-20. intensive care unit: increased proportion evolution of family members’ experience
4. Chochinov HM. Dignity and the es- of family speech is associated with in- of mechanical ventilation. Crit Care Med
sence of medicine: the A, B, C, and D of creased satisfaction. Crit Care Med 2004; 2009;37:154-8.
dignity conserving care. BMJ 2007;335: 32:1484-8. 28. Simpson C. When hope makes us vul-
184-7. 17. Carlet J, Thijs LG, Antonelli M, et al. nerable: a discussion of patient-healthcare
5. Prendergast TJ. Conflicts, negotiations Challenges in end-of-life care in the ICU: provider interactions in the context of
and resolution: a primer. In: Rocker G, statement of the 5th International Con- hope. Bioethics 2004;18:428-47.
Puntillo K, Azoulay E, Nelson J, eds. End sensus Conference in Critical Care: Brus- 29. Back AL, Arnold RM, Quill TE. Hope
of life care in the ICU: from advanced dis- sels, Belgium, April 2003. Intensive Care for the best, and prepare for the worst.
ease to bereavement. Oxford, England: Med 2004;30:770-84. Ann Intern Med 2003;138:439-43.
Oxford University Press, 2010:202-8. 18. Charles C, Whelan T, Gafni A. What 30. White DB, Evans LR, Bautista CA,
6. Choong K, Cupido C, Nelson E, et al. do we mean by partnership in making de- Luce JM, Lo B. Are physicians’ recommen-
ACCADEMY: a framework for resolving cisions about treatment? BMJ 1999;319: dations to limit life support beneficial or
disagreement during end of life care in the 780-2. burdensome? Bringing empirical data to
critical care unit. Clin Invest Med 2010; 19. Heyland DK, Cook DJ, Rocker GM, et the debate. Am J Respir Crit Care Med
33:E240-E253. al. Decision-making in the ICU: perspec- 2009;180:320-5.
7. Drazen JM. Decisions at the end of tives of the substitute decision-maker. 31. Brush DR, Rasinski KA, Hall JB, Alex-
life. N Engl J Med 2003;349:1109-10. Intensive Care Med 2003;29:75-82. ander GC. Recommendations to limit life
8. Scheunemann LP, Arnold RM, White 20. White DB, Braddock CH III, Bereknyei support: a national survey of critical care
DB. The facilitated values history: helping S, Curtis JR. Toward shared decision physicians. Am J Respir Crit Care Med
surrogates make authentic decisions for making at the end of life in intensive care 2012;186:633-9.
incapacitated patients with advanced ill- units: opportunities for improvement. 32. Truog RD, Campbell ML, Curtis JR, et
ness. Am J Respir Crit Care Med 2012;186: Arch Intern Med 2007;167:461-7. al. Recommendations for end-of-life care in
480-6. 21. Majesko A, Hong SY, Weissfeld L, the intensive care unit: a consensus state-
9. World Health Organization. WHO defi- White DB. Identifying family members ment by the American College [corrected]
nition of palliative care (http://www.who.int/ who may struggle in the role of surrogate of Critical Care Medicine. Crit Care Med
cancer/palliative/definition/en/). decision maker. Crit Care Med 2012;40: 2008;36:953-63.
10. Lanken PN, Terry PB, Delisser HM, et 2281-6. 33. Bowman K. Cultural issues and palli-
al. An official American Thoracic Society 22. Azoulay E, Pochard F, Kentish-Barnes ative care in the ICU. In: Rocker G, Pun-
clinical policy statement: palliative care N, et al. Risk of post-traumatic stress tillo K, Azoulay E, Nelson J, eds. End of
for patients with respiratory diseases and symptoms in family members of intensive life care in the intensive care unit: from
critical illnesses. Am J Respir Crit Care care unit patients. Am J Respir Crit Care advanced disease to bereavement. Oxford,
Med 2008;177:912-27. Med 2005;171:987-94. England: Oxford University Press, 2010.
11. Cook D, Rocker G, Marshall J, et al. 23. Pochard F, Darmon MI, Fassier T, et 34. Maugans TA. The SPIRITual history.
Withdrawal of mechanical ventilation in al. Symptoms of anxiety and depression Arch Fam Med 1996;5:11-6.
anticipation of death in the intensive care in family members of intensive care unit 35. Bülow HH, Sprung CL, Reinhart K,
unit. N Engl J Med 2003;349:1123-32. patients before discharge or death: a pro- et al. The world’s major religions’ points
12. Heyland DK, Dodek P, Rocker G, et al. spective multicenter study. J Crit Care 2005; of view on end-of-life decisions in the in-
What matters most in end-of-life care: per- 20:90-6. tensive care unit. Intensive Care Med
ceptions of seriously ill patients and their 24. Cox CE, Martinu T, Sathy SJ, et al. Ex- 2008;34:423-30.
family members. CMAJ 2006;174:627-33. pectations and outcomes of prolonged 36. Vincent JL. Forgoing life support in
13. Heyland DK, Rocker GM, Dodek PM, mechanical ventilation. Crit Care Med western European intensive care units:
et al. Family satisfaction with care in the 2009;37:2888-94, quiz 2904. the results of an ethical questionnaire.
intensive care unit: results of a multiple 25. Lee Char SJ, Evans LR, Malvar GL, Crit Care Med 1999;27:1626-33.
center study. Crit Care Med 2002;30:1413-8. White DB. A randomized trial of two 37. Sprung CL, Cohen SL, Sjokvist P, et al.
n engl j med 370;26 327

The New
Downloaded from nejm.org©on
Medical personalAll
userights
only. reserved.
Back to Table of ContentsSociety. All rights reserved.
End-of-life practices in European inten- 42. Kompanje EJO, van der Hoven B, Pronovost PJ. Improving comfort and
sive care units: the Ethicus Study. JAMA Bakker J. Anticipation of distress after communication in the ICU: a practical
2003;290:790-7. discontinuation of mechanical ventilation new tool for palliative care performance
38. Sulmasy DP. Spirituality, religion, and in the ICU at the end of life. Intensive measurement and feedback. Qual Saf
clinical care. Chest 2009;135:1634-42. Care Med 2008;34:1593-9. Health Care 2006;15:264-71.
39. Cook DJ, Giacomini M, Johnson N, 43. Kuhl D. What dying people want: 48. Schneiderman LJ, Gilmer T, Teetzel HD,
Willms D. Life support in the intensive practical wisdom for the end of life. New et al. Effect of ethics consultations on non-
care unit: a qualitative investigation of York: PublicAffairs, 2002. beneficial life-sustaining treatments in the
technological purposes. CMAJ 1999;161: 44. Hough CL, Hudson LD, Salud A, Lahey intensive care setting: a randomized con-
1109-13. T, Curtis JR. Death rounds: end-of-life dis- trolled trial. JAMA 2003;290:1166-72.
40. Davidson JE, Powers K, Hedayat KM, cussions among medical residents in the 49. Curtis JR, Nielsen EL, Treece PD, et al.
et al. Clinical practice guidelines for sup- intensive care unit. J Crit Care 2005;20:20-5. Effect of a quality-improvement interven-
port of the family in the patient-centered 45. Nelson JE, Angus DC, Weissfeld LA, tion on end-of-life care in the intensive
intensive care unit: American College of et al. End-of-life care for the critically ill: care unit: a randomized trial. Am J Respir
Critical Care Medicine Task Force 2004- a national intensive care unit survey. Crit Crit Care Med 2011;183:348-55.
2005. Crit Care Med 2007;35:605-22. Care Med 2006;34:2547-53. 50. Lautrette A, Darmon M, Megarbane
41. Curtis JR, Cook DJ, Sinuff T, et al. 46. Clarke EB, Curtis JR, Luce JM, et al. B, et al. A communication strategy and
Noninvasive positive pressure ventilation Quality indicators for end-of-life care in brochure for relatives of patients dying in
in critical and palliative care settings: un- the intensive care unit. Crit Care Med the ICU. N Engl J Med 2007;356:469-78.
derstanding the goals of therapy. Crit 2003;31:2255-62. [Erratum, N Engl J Med 2007;357:203.]
Care Med 2007;35:932-9. 47. Nelson JE, Mulkerin CM, Adams LL, Copyright © 2014 Massachusetts Medical Society.
2514 328 nejm.org

n engl j med 370;26 june 26, 2014
The New England

Downloaded from collections.nejm.org. Journaluse
For personal of Medicine
Copyright 28, 2015. For
© Massachusetts personal
Medical use only.
Society. No other
All rights uses without permission.
reserved.

How should decisions be made about this patient’s further treatment?
A. Meet with the family and tell them there is no hope of recovery and you will be discontinuing mechanical
ventilation.
B. Meet with the family with the goal of ascertaining the patient’s wishes regarding the continuation of active
medical treatment and whether the likely outcome of continued treatment — at best, placement in a nursing
home with a high level of care — would be consistent with the patient’s wishes.
C. Meet with the family and ask them whether they would like the patient to have a tracheostomy.
D. Meet with the family and tell them you will perform a tracheostomy.

Most patients who die in an ICU do not die suddenly of unexpected complications. Most deaths are predicted,
and the exact timing of death is dependent on a decision to withhold or withdraw specific medical interventions.
Decision-making models with respect to the withholding or withdrawing of life-prolonging interventions vary
internationally. At one extreme, the physician shares information but takes a dominant role in decision making.
At the other extreme, patient autonomy is seen as paramount, and the physician plays a solely advisory role. A
shared decision-making model is the archetype in North America, in parts of Europe, and in Australasia. Some
patients may have prepared an advance directive regarding life-prolonging treatment; such a directive should be
respected and in some jurisdictions is legally binding. Whatever the decision-making model, the goal should be
to ascertain and respect the patient’s wishes. When treatment is to be withheld or withdrawn in the expectation
of death, the goal should be to individualize care so that it is congruent with the patient’s physical, spiritual, and
cultural needs.
Our patient did not have an advance directive, but his family members were unanimous in their agreement that
he would not want life-prolonging measures continued unless there was a realistic expectation of recovery to his
previous level of function. Having developed a trusting relationship with the ICU staff during his prolonged stay,
they agreed that further active treatment was not consistent with the patient’s wishes. After a further 24 hours
during which all family members who wished to visit were given the opportunity, the patient’s endotracheal tube
was removed, morphine was administered to avert dyspnea, and he died peacefully in an isolation room in the
ICU with his close family present.
329
NEJM Group combines gold-standard content with new technology to deliver innovative
products and services that advance knowledge, learning, practice, and professional
development. NEJM Group products include the New England Journal of Medicine, NEJM
Journal Watch, and NEJM Knowledge+, and are leaders in providing the information
health care professionals need to improve health care quality and patient outcomes.
Visit https://collections.nejm.org for more Clinical Collections from NEJM Group.
© 2015 Copyright Massachusetts Medical Society. NEJM Group is a division of the Massachusetts Medical Society.
All rights reserved.

Critical Care NEJM PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Critical Care NEJM PDF

Uploaded by

Copyright:

Available Formats

Critical

CRITICAL CARE — INTRODUCTON

CRITICAL CARE SERIES OPENING EDITORIAL

SEVERE SEPSIS AND SEPTIC SHOCK

VENTILATOR-INDUCED LUNG INJURY

VENTILATOR-INDUCED LUNG INJURY (continuted)

ACUTE LIVER FAILURE

SEDATION AND DELIRIUM IN THE INTENSIVE CARE UNIT

BLEEDING AND COAGULOPATHIES IN CRITICAL CARE

FEEDING CRITICALLY ILL PATIENTS

ICU-ACQUIRED WEAKNESS AND RECOVERY FROM CRITICAL ILLNESS

TRAUMATIC INTRACRANIAL HYPERTENSION

DYING WITH DIGNITY IN THE INTENSIVE CARE UNIT

— Jeffrey M. Drazen, M.D.

Critical Care — An All-Encompassing Specialty

n engl j med 369;7 8

A Role for Finasteride in the Prevention of Prostate Cancer?

670 n engl j med 369;7 9

The New England Journal

SEVERE SEPSIS AND SEPTIC SHOCK

Severe Sepsis and Septic Shock

Case Challenge Question

A. Treatment with intravenous immune globulin (gamma globulin).

Critical Care Medicine

Severe Sepsis and Septic Shock

The New For

n engl j med 369;9 14

Sepsis (documented or suspected infection plus ≥1 of the following)†

* Data are adapted from Levy et al.5

Ou t c ome death from septic shock were often in excess of

842 n engl j med 369;9 15

The New EnglandFor

n engl j med 369;9 16

Proinflammatory response Excessive inflammation causing collateral damage (tissue injury)

Impaired function Inhibition of proinflammatory

Antiinflammatory response Immunosuppression with enhanced susceptibility to secondary infections

Figure 1. The Host Response in Severe Sepsis. COLOR FIGURE

844 n engl j med 17

Tissue hypoperfusion Loss of barrier function

Increased coagulation Decreased anticoagulation

↓ Activated protein C ↑ Angiopoietin 2

n engl j med 369;9 nejm.org 18 29, 2013

Downloaded from nejm.org

Downloaded from collections.nejm.org.

Apply a minimal amount of positive end-expiratory pressure in ARDS 1B

No other uses without permission.

only. No other uses without permission.

sion may also contribute to sepsis-associated

body-cavity edema.43 In addition, mitochondrial

extracellular environment, including mitochon-

The Surviving Sepsis Campaign, an international

and emergency medicine, recently issued the third

ment of severe sepsis and septic shock (Table 2).23

tial management bundle to be accomplished with-

ated with an improved outcome.46,47

The principles of the initial management

en absolute adrenal insufficiency (2C).

use of intravenous fluids and vasopressors, with

vided as necessary. The exact components re-

choice and amount of fluids, appropriate type

n engl j med 369;9 nejm.org 20 29, 2013

848 n engl j med 369;9 21

The New England Journal

— Jeffrey M. Drazen, M.D.