LETTERS
Reclassification of Acute Respiratory Distress
Syndrome: A Secondary Analysis of the ARDS
Network Trials
To the Editor:
According to the Berlin definition, the acute respiratory distress
syndrome (ARDS) is classified as mild, moderate, and severe by
using an arterial partial pressure of oxygen (PaO2) to fraction of
inspired oxygen (FIO2) threshold of 300, 200, and 100 mm Hg,
respectively (1). Despite the above classification, a PaO2/FIO2
threshold of 150 mm Hg has been used to define severe ARDS
and to determine eligibility in therapeutic randomized,
controlled trials examining interventions, such as prone
positioning (2) and neuromuscular blockade (3). In addition, the
PaO2/FIO2 threshold of 150 mm Hg identified two separate
subgroups of patients, namely, a moderate–severe (100 , PaO2/
FIO2 , 150 mm Hg) and a mild–moderate (150 < PaO2/FIO2 <
200 mm Hg) ARDS subgroup, with different anatomical and
physiological characteristics, according to a recently published
retrospective study (4). The latter study, which included 105
patients with moderate ARDS, did not have the statistical power
to examine whether the two subgroups have different clinical
outcomes (4). Therefore, the clinical significance of the PaO2/FIO2
threshold of 150 mm Hg remains unexplored. We explored
the clinical significance of the above-mentioned threshold by
examining whether patients with moderate–severe ARDS have
worse clinical outcomes than patients with mild–moderate
ARDS.
We undertook a secondary analysis of patient-level data from
the three most recently published ARDS Network (ARDSNet)
randomized controlled clinical trials, namely, the Aerosolized
b2-Agonist for Treatment of Acute Lung Injury (ALTA), Initial
Trophic vs Full Enteral Feeding in Patients With Acute Lung Injury
(EDEN), and Rosuvastatin for Sepsis-Associated Acute Respiratory
Distress Syndrome (SAILS) (5–7). For this analysis, we excluded
individuals from the Ventilation with Lower Tidal Volumes as
Compared with Traditional Tidal Volumes for Acute Lung Injury
and the Acute Respiratory Distress Syndrome (ARMA), Higher
versus Lower Positive End-Expiratory Pressures in Patients with
the Acute Respiratory Distress Syndrome (ALVEOLI), Comparison
of Two Fluid-Management Strategies in Acute Lung Injury
(FACTT), and Efficacy and Safety of Corticosteroids for Persistent
Acute Respiratory Distress Syndrome (LaSRS) ARDSNet trials,
because they were published before 2010 and therefore may
not reflect modern clinical practice (8–11). Subjects from the
Enteral Omega-3 Fatty Acid, g-Linolenic Acid, and Antioxidant
Supplementation in Acute Lung Injury (OMEGA) trial (12) were
included in our analysis as part of the EDEN trial (6). Subjects in
the ARDSNet trials were intubated receiving positive-pressure
mechanical ventilation, had a PaO2/FIO2 of 300 mm Hg or lower, and
presented with bilateral infiltrates on chest radiography consistent
with pulmonary edema, without evidence of left atrial hypertension
(5–7). Deidentified data from ARDSNet trials were provided by the
Biologic Specimen and Data Repository Information Coordinating
Center (BioLINCC) of the National Heart, Lung, and Blood
Institute. The Institutional Review Board at Weill Cornell Medicine
998
determined that our study was exempt from review and the need
for informed consent (no. 1802019008).
Patients having 100 , PaO2/FIO2 < 200 mm Hg at the time of
trial enrollment (i.e., patients classified as having moderate ARDS
according to the Berlin definition) (1) were separated into two
subgroups: a moderate–severe (100 , PaO2/FIO2 , 150 mm Hg)
and a mild–moderate (150 < PaO2/FIO2 < 200 mm Hg) ARDS
subgroup (4). Given that PaO2/FIO2 varies throughout ARDS and
depends on ventilator strategy (13), we chose for subgroup
allocation the PaO2/FIO2 at trial enrollment, when all patients were
ventilated according to a standardized lung-protective strategy,
which included a minimum positive end-expiratory pressure of
5 cm H2O (5–7).
The primary outcome of our analysis was all-cause 60-day
mortality, with patients discharged from hospital with
unassisted breathing before 60 days being considered to be alive
at 60 days. Proportion of patients with rapidly resolving ARDS,
ventilator-free days, and intensive care unit (ICU)–free days
were the secondary outcomes of our analysis. ARDS was
considered rapidly resolving if the patient had a PaO2/FIO2 over
300 mm Hg within 24 hours after enrollment or achieved
unassisted breathing within 24 hours after enrollment and
remained free from assisted breathing for 48 hours or more.
Ventilator-free days and ICU-free days were the number of days
in the first 28 days that a patient was alive and not on a ventilator,
or alive and not in the ICU, respectively.
Categorical variables are presented as count (percentage) and
compared using chi-square or Fisher’s exact test, as appropriate.
Continuous variables are presented as median (interquartile range)
and compared using nonparametric Mann-Whitney U test. Time to
mortality of the compared groups was estimated by a Kaplan-Meier
analysis. All statistical analyses were done with R v3.2.3 (R Core
Team). Two-sided P less than 0.05 was considered to denote
statistical significance.
Of the 1,909 unique patients enrolled in the ARDSNet
randomized controlled trials (5–7), 275 (14.4%) had missing data
on PaO2/FIO2 at trial enrollment, and therefore were excluded from
this analysis. Of the remaining 1,634 patients, 870 (53.2%) had
moderate ARDS, according to the Berlin definition (1). As depicted
in Table 1, there was no difference between patients with moderate–
severe (n = 467) versus mild–moderate (n = 403) ARDS in terms of
baseline clinical characteristics. In addition, there was no difference
between the compared subgroups in terms of time from diagnosis
of ARDS to enrollment. Although driving pressure was not different,
patients with moderate–severe ARDS had greater corrected expired
volume per minute (a surrogate of dead space) and higher plateau
pressure than patients with mild–moderate ARDS.
As depicted in Table 2, all-cause 60-day mortality was not
different between patients with moderate–severe and mild–
moderate ARDS (104/467 [22.3%] vs. 91/403 [22.6%]; P = 0.98).
Consistently, the probability of death over time was not
different between comparators (P = 0.71 by log-rank test;
Figure 1). However, a smaller proportion of patients with
moderate–severe (26/467 [5.6%]) than patients with mild–
moderate ARDS (45/403 [11.2%]; P = 0.004) had the ARDS
resolve rapidly. In addition, patients with moderate–severe
ARDS had fewer ventilator-free days (median [interquartile
AnnalsATS Volume 15 Number 8 | August 2018
 LETTERS

Table 1. Baseline characteristics of patients with moderate–severe versus mild–moderate acute respiratory distress syndrome

Characteristic Moderate–Severe ARDS Mild–Moderate

ARDS

No. of patients 467 403

Age, yr, median (IQR) 52 (41–62) 53 (42–66)
Male sex, n (%) 231 (49.5) 213 (52.9)
Body mass index, median (IQR) 29 (24–36) 28 (24–34)
Usage of vasopressors, n (%) 227 (49.3) 205 (51.2)
APACHE III score, median (IQR) 92 (73–112) 90 (72–107)
Primary risk factor of ARDS, n (%)
Pneumonia 327 (70.0) 256 (63.5)
Sepsis 69 (14.8) 76 (18.9)
Aspiration 37 (7.9) 42 (10.4)
Trauma 11 (2.4) 13 (3.2)
Multiple transfusions 8 (1.7) 3 (0.7)
Other 18 (3.9) 14 (3.5)
Nonpulmonary organ failure, n (%)
Circulatory 332 (71.1) 281 (69.7)
Coagulation 82 (17.7) 81 (20.3)
Hepatic 73 (16.4) 61 (16.1)
Renal 111 (24.1) 99 (24.7)
Time from diagnosis of ARDS to trial enrollment, 1 (0–1) 1 (0–1)
days, median (IQR)
Ventilator and physiological parameters at
enrollment, median (IQR)
PaO2/FIO2 125 (113–137) 172 (160–185)
Corrected expired volume per minute, L/min 11 (9–14) 10 (8–13)
Plateau pressure, cm H2O 24 (21–28) 22 (19–27)
Driving pressure, cm H2O 14 (11–18) 14 (11–18)
Set positive end-expiratory pressure, cm H2O 10 (8–12) 8 (5–10)
FIO2 0.60 (0.50–0.70) 0.50 (0.40–0.60)
Minute ventilation, L/min 11 (9–13) 11 (9–13)

Definition of abbreviations: APACHE = acute physiology and chronic health evaluation; ARDS = acute respiratory distress syndrome; FIO2 = fraction of
inspired oxygen; IQR = interquartile range; PaO2 = arterial partial pressure of oxygen.
Corrected expired volume per minute was calculated as the measured minute ventilation times arterial partial pressure of carbon dioxide divided by 40.
A PaO2/FIO2 threshold of 150 mm Hg was used to categorize patients into moderate–severe (100 , PaO2/FIO2 , 150 mm Hg) and mild–moderate (150 <
PaO2/FIO2 < 200 mm Hg) ARDS.

range] = 18 [0–23] vs. 21 [0–25]; P = 0.001) and fewer ICU-free
days (16 [0–21] vs. 18 [1–22]; P = 0.007) than patients with
mild–moderate ARDS.
We found that patients with moderate–severe as opposed to
mild–moderate ARDS have worse clinical outcomes other than
mortality. Our analysis tested the recent proposal by Gattinoni and
colleagues (14) for splitting of moderate ARDS into categories of
moderate–severe and mild–moderate ARDS using a PaO2/FIO2
threshold of 150 mm Hg, with the biological rationale being that
pulmonary edema (as assessed by lung weight) increased
significantly only at PaO2/FIO2 less than 150 mm Hg (15).
Similar to the study by Maiolo and colleagues (4), we found
that patients with moderate–severe ARDS had higher plateau
pressure and more dead space than patients with mild–moderate
ARDS. This difference in physiological parameters (Table 1) was
accompanied by differences in clinical outcomes, such as rapid
resolution of ARDS, ventilator-free days, and ICU-free days
(Table 2). We found no difference between the compared

Table 2. Outcomes of patients with moderate–severe versus mild–moderate acute respiratory distress syndrome

Outcome Moderate–Severe ARDS (n = 467) Mild–Moderate ARDS (n = 403) P Value

60-d mortality, n (%) 104 (22.3) 91 (22.6) 0.98

Rapidly resolving ARDS, n (%) 26 (5.6) 45 (11.2) 0.004
Ventilator-free days, median (IQR) 18 (0–23) 21 (0–25) 0.001
ICU-free days, median (IQR) 16 (0–21) 18 (1–22) 0.007

Definition of abbreviations: ARDS = acute respiratory distress syndrome; ICU = intensive care unit; IQR = interquartile range.
An arterial partial pressure of oxygen (PaO2) to fraction of inspired oxygen (FIO2) threshold of 150 mm Hg was used to categorize patients into moderate–severe
(100 , PaO2/FIO2 , 150 mm Hg) and mild–moderate (150 < PaO2/FIO2 < 200 mm Hg) ARDS.

Letters 999

Moderate-severe ARDS
1.00 Mild-moderate ARDS
Survival probability
0.75
0.50
0.25
Log-rank
p = 0.71
0.00
0 20 40
Days since enrollment
Number at risk
467 388 368
403 338 322
0 20 40
Days since enrollment
Figure 1. Kaplan-Meier curves of mortality of patients with moderate–
severe versus mild–moderate acute respiratory distress syndrome (ARDS).
An arterial partial pressure of oxygen (PaO2) to fraction of inspired oxygen
(FIO2) threshold of 150 mm Hg was used to categorize patients into
moderate–severe (100 , PaO2/FIO2 , 150 mm Hg) and mild–moderate
(150 < Pa O2/F IO2 < 200 mm Hg) ARDS. Patients discharged home
considered alive at 60 days.
subgroups in terms of mortality (Table 2 and Figure 1). Two
explanations could be given for the latter finding. First, at PaO2/
FIO2 greater than 100 mm Hg, mortality of patients with ARDS
may be mainly determined by factors other than severity of
hypoxemia (e.g., by underlying condition). Second, our choice to
use a cohort of patients enrolled in clinical trials rather than
observational studies might bias our finding on mortality toward
the null. Subjects enrolled in clinical trials may have less
comorbidity, be more homogenous, and have lower mortality
than those in observational studies. We acknowledge it as a
limitation of our analysis.
Another limitation of our analysis is the lack of data on PaO2/
FIO2 at trial enrollment in one-seventh of patients. Missing data
are a common issue in studies of critical illness, and excluding
those subjects is a usual and acceptable approach (16). We too
chose the latter approach (i.e., performing an available-case
analysis) over more sophisticated approaches to handling missing
data, such as multiple imputation, and this might have biased our
results toward the null (17). Finally, although the study by Maiolo
and colleagues (4) showed that “recruitability” was very different
between the moderate–severe and mild–moderate subgroups,
suggesting differential treatment responsiveness in mechanical
ventilation interventions, we did not explore the predictive
validity of the PaO2/FIO2 threshold of 150 mm Hg (i.e., whether it
predicts different responses to therapy). Rather, we only explored
the prognostic validity of the PaO2/FIO2 threshold of 150 mm Hg
(i.e., whether it predicts different outcomes). Therefore, whether
there should be a distinction between moderate–severe and mild–
moderate ARDS based on treatment responsiveness remains an
open question.
1000
LETTERS
In conclusion, our analysis of data from patients enrolled in
high-quality ARDSnet trials showed no difference between patients
with moderate–severe and mild–moderate ARDS in terms of
mortality, even though there were differences in rapid resolution of
ARDS, ventilator-free days, and ICU-free days.
Author disclosures are available with the text of this article at
www.atsjournals.org.
Acknowledgment: The authors are grateful to the investigators of ARDS
Network (ARDSNet) trials for collecting the data on which this analysis was
based, and the Biologic Specimen and Data Repository Information
Coordinating Center (BioLINCC) of the National Heart, Lung, and Blood
Institute (NHLBI) for providing these data. This letter was prepared using
Aerosolized b2-Agonist for Treatment of Acute Lung Injury (ALTA), Initial
60
Trophic vs Full Enteral Feeding in Patients With Acute Lung Injury (EDEN),
and Rosuvastatin for Sepsis-Associated Acute Respiratory Distress
Syndrome (SAILS) research materials obtained from the BioLINCC and
does not necessarily reflect the opinions or views of the ARDSNet or the
363 NHLBI.
312
Clara Oromendia, M.S.
60 Weill Cornell Medicine
New York, New York
Ilias I. Siempos, M.D.
Weill Cornell Medicine
New York, New York
and
University of Athens Medical School
Athens, Greece
References
1 Ranieri VM, Rubenfeld GD, Thompson BT, Ferguson ND, Caldwell E,
Fan E, et al.; ARDS Definition Task Force. Acute respiratory distress
syndrome: the Berlin definition. JAMA 2012;307:2526–2533.
2 Guérin C, Reignier J, Richard JC, Beuret P, Gacouin A, Boulain T,
et al.; PROSEVA Study Group. Prone positioning in severe acute
respiratory distress syndrome. N Engl J Med 2013;368:2159–2168.
3 Papazian L, Forel JM, Gacouin A, Penot-Ragon C, Perrin G, Loundou A,
et al.; ACURASYS Study Investigators. Neuromuscular blockers in
early acute respiratory distress syndrome. N Engl J Med 2010;363:
1107–1116.
4 Maiolo G, Collino F, Vasques F, Rapetti F, Tonetti T, Romitti F, et al.
Reclassifying acute respiratory distress syndrome. Am J Respir
Crit Care Med 2018;197:1586–1595.
5 Matthay MA, Brower RG, Carson S, Douglas IS, Eisner M, Hite D, et al.;
National Heart, Lung, and Blood Institute Acute Respiratory Distress
Syndrome (ARDS) Clinical Trials Network. Randomized, placebo-
controlled clinical trial of an aerosolized b₂-agonist for treatment of
acute lung injury. Am J Respir Crit Care Med 2011;184:561–568.
6 Rice TW, Wheeler AP, Thompson BT, Steingrub J, Hite RD, Moss M,
et al.; 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.
7 Truwit JD, Bernard GR, Steingrub J, Matthay MA, Liu KD, Albertson TE,
et al.; National Heart, Lung, and Blood Institute ARDS Clinical Trials
Network. Rosuvastatin for sepsis-associated acute respiratory
distress syndrome. N Engl J Med 2014;370:2191–2200.
8 Brower RG, Matthay MA, Morris A, Schoenfeld D, Thompson BT,
Wheeler A; 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;342:1301–1308.
9 Brower RG, Lanken PN, MacIntyre N, Matthay MA, Morris A,
Ancukiewicz M, et al.; National Heart, Lung, and Blood Institute
ARDS Clinical Trials Network. Higher versus lower positive
AnnalsATS Volume 15 Number 8 | August 2018
 LETTERS
end-expiratory pressures in patients with the acute respiratory
distress syndrome. N Engl J Med 2004;351:327–336.
10 Wiedemann HP, Wheeler AP, Bernard GR, Thompson BT, Hayden D,
deBoisblanc B, et al.; National Heart, Lung, and Blood
Institute Acute Respiratory Distress Syndrome (ARDS)
Clinical Trials Network. Comparison of two fluid-management
strategies in acute lung injury. N Engl J Med 2006;354:2564–
2575.
11 Steinberg KP, Hudson LD, Goodman RB, Hough CL, Lanken PN,
Hyzy R, et al.; National Heart, Lung, and Blood Institute
Acute Respiratory Distress Syndrome (ARDS) Clinical Trials
Network. Efficacy and safety of corticosteroids for persistent
acute respiratory distress syndrome. N Engl J Med 2006;354:1671–
1684.
12 Rice TW, Wheeler AP, Thompson BT, deBoisblanc BP, Steingrub J,
Rock P; NIH NHLBI Acute Respiratory Distress Syndrome Network of
Investigators. Enteral omega-3 fatty acid, gamma-linolenic acid, and
antioxidant supplementation in acute lung injury. JAMA 2011;306:
1574–1581.
13 Villar J, Pérez-Méndez L, López J, Belda J, Blanco J, Saralegui I,
et al.; HELP Network. An early PEEP/FiO2 trial identifies different
Response to Cookstove Trials and Tribulations: What
Is Needed to Decrease the Burden of Household
Air Pollution?
To the Editor:
We thank Mortimer and Balmes for their timely critique of the
current evidence on household air pollution (1) and their call for a
more holistic approach to reduce its burden on public health
globally. We agree that oversimplification of the potential causes of
household air pollution is unlikely to result in effective solutions,
and that we need to step back from the focus on simple
determinants of household air pollution to consider the complexity
of the problem within local contexts.
The single focus on cleaner-burning cookstoves as an
intervention to improve respiratory disease fails to account for the
many other contributors to household air pollution, such as motor
vehicles, dust, and rubbish burning. In addition, many cookstove
studies fail to take adequate account of local experiences of air
pollution, which are specific to time and place. Miscommunications
between researchers and local communities, which occur when
local perspectives are underrepresented and one-size-fits-all
interventions are implemented across different geographical and
sociocultural contexts, can often result in the failure of well-
intentioned interventions and unforeseen adverse effects (2). An
example is when traditional cultural cooking preferences, when
faced with unreliable cookstove equipment, lead to many
households reverting back to cooking on open fires, and
where inadequate understandings of the local context lead to
problems with disposal of spent batteries from cleaner-burning
cookstoves.
A situated understanding of local schemes of perception is
imperative to inform household air pollution interventions (3).
This includes interdisciplinary working across boundaries between
scientific evidence and practical, lived experience, to position
researchers in dialogue with local communities around household
air pollution and respiratory disease. Such an approach would
Letters
degrees of lung injury in patients with acute respiratory
distress syndrome. Am J Respir Crit Care Med 2007;176:795–
804.
14 Gattinoni L, Marini JJ, Quintel M. Time to rethink the approach to
treating acute respiratory distress syndrome. JAMA 2018;319:
664–666.
15 Gattinoni L, Caironi P, Cressoni M, Chiumello D, Ranieri VM,
Quintel M, et al. Lung recruitment in patients with the acute
respiratory distress syndrome. N Engl J Med 2006;354:1775–
1786.
16 Freund Y, Lemachatti N, Krastinova E, Van Laer M, Claessens YE,
Avondo A, et al.; French Society of Emergency Medicine
Collaborators Group. Prognostic accuracy of sepsis-3 criteria
for in-hospital mortality among patients with suspected infection
presenting to the emergency department. JAMA 2017;317:301–
308.
17 White IR, Carlin JB. Bias and efficiency of multiple imputation
compared with complete-case analysis for missing covariate values.
Stat Med 2010;29:2920–2931.
Copyright © 2018 by the American Thoracic Society
allow existing knowledges to be interwoven, so that all
perspectives are altered and represented (4) in the co-creation
of new culturally relevant interventions (5). Close working
with communities would support the development of new
household air pollution interventions that are informed by
local perceptions, values, and beliefs, as well as biomedical
science, and are therefore feasible, acceptable, and importantly,
inspiring for their target populations. Such interventions are
far more likely to be adhered to, and thus to be effective in
reducing air pollution and in achieving meaningful and sustained
health benefits.
Author disclosures are available with the text of this letter at
www.atsjournals.org.
Cindy M. Gray, B.Sc., B.A., Ph.D.
Emily Colarte, B.Sc.
Clara Young, B.A.
University of Glasgow
Glasgow, United Kingdom
ORCID ID: 0000-0002-4295-6110 (C.M.G.).
References
1 Mortimer K, Balmes JR. Cookstove trials and tribulations: what is needed
to decrease the burden of household air pollution? Ann Am Thorac
Soc 2018;15:539–541.
2 Hahn RA, Inhorn MC. Anthropology and public health: bridging
differences in culture and society. New York: Oxford University Press;
2009.
3 Zuiderent-Jerak T. Situated intervention: sociological experiments in
health care. Cambridge: MIT Press; 2015.
4 Smith L. Decolonizing methodologies: research and indigenous peoples.
London: University of Otago Press; 2012.
5 Panter-Brick C, Clarke SE, Lomas H, Pinder M, Lindsay SW. Culturally
compelling strategies for behaviour change: a social ecology model
and case study in malaria prevention. Soc Sci Med 2006;62:2810–
2825.
Copyright © 2018 by the American Thoracic Society
1001