Transfusion Requirements

in Critical Care
A Pilot Study
Paul C. H\l=e'\bert,MD, MHSc, FRCPC; George Wells, PhD; John Marshall, MD, FRCSC; Claudio Martin, MD, MSc, FRCPC;
Martin Tweeddale, MD, PhD, FRCPC; Giuseppe Pagliarello, MD, FRCSC; Morris Blajchman, MD, FRCPC;
for the Canadian Critical Care Trials Group

Objective.\p=m-\Toevaluate the effects of a restrictive and a liberal red blood cell ANEMIA is common in the critically ill
(RBC) transfusion strategy on mortality and morbidity in critically ill patients. and results in the frequent use of red
blood cell (RBC) transfusions. For de¬
Study Design.\p=m-\Multicenter,prospective, randomized clinical trial.
Patient Population.\p=m-\Sixty-ninenormovolemic critically ill patients admitted to cades, an arbitrary transfusion thresh¬
old of 100 g/L has been widely used in
one of five tertiary level intensive care units with hemoglobin values less than 90
clinical medicine, including treatment of
g/L within 72 hours of admission. critically ill patients.1 Recently, a réévalu¬
Interventions.\p=m-\Patientswere randomly allocated to one of two RBC transfusion ation of this practice was prompted by
strategies. Hemoglobin values were maintained between 100 and 120 g/L in the the fear of transfusion-related infections,
liberal transfusion group and between 70 and 90 g/L in the restrictive group. such as acquired immunodeficiency syn¬
Results.\p=m-\Primarydiagnosis and mean\m=+-\SDage (58.6\m=+-\15vs 59.0\m=+-\21years) drome (AIDS), and the prospect of an
and Acute Physiology and Chronic Health Evaluation II score (20\m=+-\6.2vs 21 \m=+-\7.2) ever-decreasing blood supply.2"4 These
were similar in the restrictive and liberal groups, respectively. Daily hemoglobin concerns resulted in the formulation of
values averaged 90 g/L in the restrictive group vs 109 g/L in the liberal group general recommendations and guidelines
(P<.001). The restrictive group received 2.5 U per patient compared with 4.8 U per by the National Institutes of Health and
the American Association ofBlood Banks
patient in the liberal group. This represents a 48% relative decrease (P<.001) in for the use of RBCs.5,6 The guidelines did
RBC units transfused per patient. The 30-day mortality rate was 24% in the restric- not specifically recommend a transfusion
tive group compared with 25% in the liberal group; the 95% confidence interval threshold in critically ill patients. There
around the absolute difference was \m=-\19% to 21%. Similar observations were noted is reason to believe that the complex
for intensive care unit mortality (P=.76) and 120-day mortality (P>.99). In addition, metabolic and cardiovascular alterations
survival analysis comparing time until death in both groups did not reveal any sig- of critical illness would restrict the ap¬
nificant difference (P=.93) between groups. Organ dysfunction scores were also plicability of any recommendation derived
similar (P=.44). from studies performed in other patient
Conclusion.\p=m-\Inthis small randomized trial, neither mortality nor the develop- populations and in experimental models
ment of organ dysfunction was affected by the transfusion strategy, which suggests that may not have comparable physiologi¬
that a more restrictive approach to the transfusion of RBCs may be safe in critically cal responses.7
Several studies8,9 have suggested that
ill patients. However, the study lacked power to detect small but clinically significant
differences. Therefore, further investigations of RBC transfusion strategies are
augmenting systemic oxygen delivery
may decrease mortality in critically ill
warranted. patients when various treatment mo¬
(JAMA. 1995;273:1439-1444) dalities, including RBC transfusions, are
used. In contrast, studies examining the
role of transfused RBCs in immune
From the Critical Care Programs at the University of nadian Critical Care Trials Group appears at the end of
Ottawa (Ontario) (Drs H\l=e'\bertand Pagliarello); the Uni- this article. modulation10 and in microcirculatory al¬
versity of Toronto (Ontario) (Dr Marshall); the University Presented in abstract form at the 60th Annual Inter- terations1112 have suggested several del¬
of Western Ontario, London, Ontario (Dr Martin); and national Meeting of the Scientific Assembly of the eterious consequences from the admin¬
the University of British Columbia, Vancouver (Dr American College of Chest Physicians, October 30\x=req-\
Tweeddale); the Clinical Epidemiology Unit, University November 3, 1994, New Orleans, La. istration of RBC transfusions in similar
of Ottawa (Drs H\l=e'\bertand Wells); and the Department Reprint requests to the Department of Medicine, Di- patients. Thus, an optimal and safe lower
of Pathology, McMaster University, Hamilton, Ontario vision of Respiratory Medicine, Room N14, Ottawa limit to the transfusion threshold has
(Dr Blajchman). Dr H\l=e'\bertis a Career Scientist with the General Hospital, 501 Smyth Rd, Ottawa, Ontario,
Ontario Ministry of Health. Canada K1H 8L6 (Dr H\l=e'\bert). not been established in critically ill pa¬
A complete list of the study participants from the Ca- tient populations.13

Downloaded From: http://jama.jamanetwork.com/ by a Florida International University Medical Library User on 05/24/2015
 Surveys of perioperative transfusion
practice of anesthesiologists have docu¬
mented both the widespread use of a
transfusion threshold of 100 g/L and the
tremendous variation in individual
transfusion practice.1416 Recently, we
also documented a variation in transfu¬
sion practice among Canadian critical
care practitioners.17 The lack of convinc¬
ing published data and evidence of sig¬
nificant variability in transfusion prac¬
tice prompted the development of this
randomized clinical trial comparing a re¬
strictive transfusion strategy that main¬
tained hemoglobin values between 70
and 90 g/L with a liberal transfusion
strategy that maintained hemoglobin
values between 100 and 120 g/L in nor-
movolemic critically ill patients.
METHODS
Study Design
We undertook a prospective, multi-
center, randomized clinical trial to
compare two transfusion strategies in
normovolemic critically ill patients.
Patients were assigned to one of two
treatment groups by consecutive allo¬
cation from a random listing stratified
by center and disease severity using
Acute Physiology and Chronic Health
Evaluation (APACHE) II scores18
(APACHE II scores <15 or >15) and
blocked by balanced groupings of 10.19
Blinding of treatment allocation was
not feasible.
Patient Selection
All patients admitted to one of the
five tertiary-level intensive care units
(ICUs) between March 15, 1993, and
January 30, 1994, were considered eli¬
gible for this study. The accrual period
in each center varied from 4 to 11 months.
We included patients who were expected
to stay more than 24 hours, had a he¬
moglobin value less than or equal to 90
g/L within 72 hours of ICU admission,
and were considered volume resusci¬
tated or normovolemic by the attending
staff. Patients were excluded for any of
the following reasons: (1) less than 16
years of age; (2) pregnant; (3) unable to
receive blood products; (4) active blood
loss at the time of enrollment, defined as
a 30-g/L decrease in hemoglobin in the
preceding 12 hours or requiring at least
3 U of packed RBCs during the same
period; (5) brain dead or not expected to
survive more than 24 hours; or (6) ad¬
mitted following routine cardiac surgi¬
cal interventions. The protocol was ap¬
proved by the research ethics board of
each participating institution. Informed
consent was obtained from either the
patient or the closest family member.
Downloaded From: http://jama.jamanetwork.com/ by a Florida International University Medical Library User on 05/24/2015
Treatment Protocols
Transfusion guidelines for both the
restrictive and liberal treatments were
developed through a consensus-build¬
ing process involving the Transfusion
Requirements in Critical Care (TRICC)
Trial Executive Committee, the Cana¬
dian Critical Care Trials Group, faculty
of the University of Ottawa (Ontario),
Clinical Epidemiology Unit, and infor¬
mation from the national survey of Ca¬
nadian critical care practitioners.17 The
following principles were used in devel¬
oping the two transfusion ranges: (1)
safety of each specific range; (2) current
transfusion practices in Canadian ICUs;
and (3) general acceptability of each spe¬
cific range. Once randomized, a study
patient's hemoglobin level was main¬
tained using allogeneic RBC transfusions
as required. Patients allocated to the
restrictive strategy had their hemoglo¬
bin levels maintained between 70 and 90
g/L, with a transfusion trigger at 70 to
75 g/L. Patients allocated to a liberal
transfusion strategy had their hemoglo¬
bin levels maintained between 100 and
120 g/L, with a transfusion trigger at
100 to 105 g/L. Given that this study
compared transfusion strategies, other
interventions modifying oxygen deliv¬
ery such as the type and quantity of
fluids administered and the use of va-
soactive drugs were documented but not
rigorously controlled in the treatment
protocols.
Outcome Measures
The pilot phase of the TRICC study
was undertaken to determine the fea¬
sibility of a larger multicenter clinical
trial using survival as its primary end-
point. Therefore, the pilot study was not
designed with sufficient power to ex¬
clude small but clinically significant dif¬
ferences in treatment effects. Outcomes
evaluated were mortality rates, includ¬
ing 30-day all-cause mortality, ICU mor¬
tality, hospital mortality, and survival
times; assessments of organ dysfunc¬
tion, including the number of organ fail¬
ures,20·21 the rates of multiple system
organ failure (MSOF) as defined by Hé¬
bert et al,20 and multiple organ dysfunc¬
tion (MOD) score as defined by Marshall
et al22; and related morbidity including
length of ICU and hospital stay.
Statistical Analyses
The final analysis was conducted us¬
ing an intention-to-treat approach. Base¬
line characteristics of patients in the two
treatment arms were compared using
univariate descriptive statistics. The
analysis of hemoglobin values over time
was performed using analysis of vari¬
ance with repeated measures followed
by Tukey's honestly significant differ¬
ence procedure for all pairwise compari¬
sons. The analysis comparing mortality
rates, the number of organs failing per
patient, and the development of MSOF
were performed using Fisher's exact test
comparing the proportion of events in
each treatment group. Kaplan-Meier
survival curves were used to summa¬
rize the time to death experience in both
groups. Survival curves were then com¬
pared using a log rank test statistic.
Continuous outcome measures includ¬
ing MOD scores and total RBC units
transfused were analyzed using an in¬
dependent t test, and lengths of stay
were analyzed using a Wilcoxon rank
sum statistic. Comparisons involving pri¬
mary outcomes were considered statis¬
tically significant using a two-sided of
.05. No adjustments were made for mul¬
tiple comparisons. Absolute values and
confidence intervals (CIs) are reported.
RESULTS
Patient Characteristics
Sixty-nine patients were enrolled in
the study; 33 in the restrictive group,
and 36 in the liberal group. All random¬
ized patients completed the study. Pa¬
tient characteristics, including age, sex,
APACHE II scores, and primary diag¬
nosis, were similar in both groups
(P>.39) (Table 1). Forty percent of pa¬
tients enrolled required a pulmonary ar¬
tery catheter, and 90% required me¬
chanical ventilation; no differences were
observed in comparing the two groups.
Seventy percent of patients were ad¬
mitted to the ICU from surgical ser¬
vices. Baseline characteristics that may
have affected oxygen transport, includ¬
ing the number of transfusions (P=.16),
the cumulative fluid balance (P=.10), and
vasoactive drugs administered on ad¬
mission to ICU (P=.13), were similar in
the two groups. Mean serum lactate val¬
ues, an indicator of perfusion, were el¬
evated on admission to ICU in both
groups, but more so in the liberal trans¬
fusion group (2.1 mmol/L) compared with
the restrictive group (3.8 mmol/L)
(P=.03). All other laboratory values were
similar on admission (P>.05). At the time
of randomization, all laboratory values
were similar in both groups (P>.05)
(Table 2).
The Success of Each
Treatment Strategy
Average hemoglobin values were sig¬
nificantly different in the two treatment
arms (P<.001) throughout the study
(Figure 1). The average daily hemoglo¬
bin values averaged 90 g/L in the re¬
strictive transfusion group and 109 g/L
in the liberal transfusion group. An av-
 erage 2.5 U per patient (82 U in 33 pa¬ study patients allocated to the restric¬ Outcome Measures
tients) were transfused using the re¬ tive strategy and 50% ofthe liberal strat¬ The 30-day mortality rate was 24% in
strictive strategy compared with 4.8 U egy attained the target compliance the restrictive group compared with 25%
per patient (174 U in 36 patients) using (P>.05). There were four (6%) inten¬ in the liberal group (P>.99). The ob¬
the liberal strategy; a 48% relative de¬ tional crossovers, two from the restric¬ served 1% difference in mortality has a
crease (P<.001) in the number of RBC tive group to the liberal group and two 95% CI of -19% to 21%. Thus, the 30-
transfusions (Figure 2). from the liberal to the restrictive. In
As a measure of compliance with the day mortality rate may have been as
total, five patients crossed over from much as 21% greater (an 88% relative
treatment strategies, a particular treat¬ one group to the other. Hemoglobin val¬
risk increase) or as much as 19% less (a
ment strategy was considered success¬ ues increased beyond 90 g/L in 16% of
79% relative risk reduction) in the re¬
fully implemented if at least 80% of all patients without additional RBC trans¬ strictive group compared with that in
daily hemoglobin values were main¬ fusions following the implementation of the liberal group. Other mortality rates
tained between 70 and 90 g/L in the the restrictive strategy. In the liberal
patients assigned to the restrictive strat¬ transfusion strategy group, pretransfu-
examined, including ICU mortality (five
[15%] of 33 in the restrictive group vs
egy group or between 100 and 120 g/L sion and posttransfusion hemoglobin val¬
seven [19%] of 36 in the liberal group,
in the patients assigned to the liberal ues were included in the analysis shown
P=.76) and 120-day mortality (13 [54%]
strategy group. Forty-five percent of in Figure 1. of 24 in the restrictive group vs 11 [50%]
of 22 in the liberal group, P>.99), were
Table 1.—Baseline Characteristics of Patients Randomized to Receive Red Blood Cell Transfusions Based not significantly different in the two
on Either a Restrictive or a Liberal Transfusion Strategy* groups (Table 3).
Since all-cause mortality may not be
Restrictive Liberal
Variable (n=33) (n=36) readily influenced by single interven¬
Patient characteristics tions in complex critically ill patients,
Sex, M/F 14/19 19/17 other measures of ICU morbidity and
Age in years, mean±SD 58.6±15 59±21 mortality were examined, including sur¬
APACHE II score, mean±SD 20±6.2 vival times, organ dysfunction scores,
No. of organs failing, No. (%) and lengths of stay. A comparison of
None 12 (36) 20 (56) Kaplan-Meier survival curves using a
One 15 (45) 11(31) log rank test did not demonstrate any
Two to four 6(18) 5(14) differences in survival between treat¬
Para-aortic catheter required, No. (%) 15(45) 14 (39) ment strategies (P=.93) (Figure 3). Or¬
Mechanical ventilation required, No. (%) 29 (88) 33 (92) gan dysfunction measured as the num¬
Admission diagnostic category, No. (%) ber of failing organs, using either a di-
Respiratory failure 13(39) 7(19) chotomous assignment for organ failure20
Trauma 5(15) 11(31) or a MOD score, was similar in the two
Gastrointestinal disease 4(12) 2(6) groups. Other outcomes compared, in¬
Metabolic disease 2(6) 3(8) cluding the percentage of patients di¬
Abdominal and vascular surgery 3(9) 4(11) agnosed with MSOF as well as the ICU
Cardiac and hospital median lengths of stay, were
4(12) 2(6)
Other
also similar in the two groups (Table 3).
7(19)
Past medical history, No. (%)
Chronic anemia or renal failure 5(15) 1(3) COMMENT
Ischemie heart disease or CHF 5(15) 4(11) In this study, hemoglobin values were
Previous ICU admission 6(18) 2(6)
Previous transfusions 15 (45)
successfully maintained in one of two
13(36) predefined target ranges following the
*APACHE indicates Acute Physiology and Chronic Health Evaluation; CHF, congestive heart failure; and ICU, implementation of guidelines for the ad¬
intensive unit.
care ministration of allogeneic RBCs. The re-

Table 2.—Variables Related to Oxygen Delivery and Laboratory Values on Admission to the Intensive Care Unit and at the Time of Randomization*

Restrictive (n=33) Liberal (n=36)

Variable Admission Randomization Admission Randomization
Oxygen delivery measures
Hemoglobin values, g/L 97±14 84 ±7.8 93±15 83±5.7
Cumulative total fluid intake, L 3.1 + 1.7 6.7±3.8t 4.1 ±3.3 9.0±4.7f
No. (%) of patients taking vasoactlve drugs 0(0) 3(9) 0(0) 3(8)
Lactate, mmol/L 2.1±1.3t 1.6+1.5 3.8±2.54t 1.9 + 1.5
Laboratory values
White blood cell count, X107L 15±6.7 11 ±5.2 14±8.7 12±5.5
Platelet count, x109/L 257±131 208±112 207±113 153±84
Creatlnine, µmol/L (mg/dL) 165±193(1.87±2.18) 162±189(1.83±2.14) 135±107(1.53±1.21) 137±108(1.55±1.22)
Sodium, mmol/L 138±7.3 139±8.4 140±7.4 141 ±7.9
Total bilirubin, µ . (mg/dL) 21±26(1.23±1.52) 21 ±23 (1.23±1.35) 24±37 (1.40+2.16 25±44 (1.46+2.57)
Alanine aminotransferase, U/L 70±134 50±103 42±61 63±118
'Variables reported as means±SD unless otherwise specified. All comparisons were between the restrictive and liberal strategies at comparable times.
tP<.05.

Downloaded From: http://jama.jamanetwork.com/ by a Florida International University Medical Library User on 05/24/2015

125
100 4
•g 75- \ Randomization
50
• Liberal
25- ° Restrictive
8 12 16 20
Time, d
Figure 1.—Average dally hemoglobin values. Daily
mean hemoglobin value in each study group mea¬
sured at least once a day in each patient. These val¬
ues are significantly different for each day and for the
20-day period shown (all P<.001 using analysis of
variance followed by Tukey's honestly significant
differences procedure for pairwise comparisons).
The daily means underestimate the differences be¬
tween study groups in that both pretransfusion and
posttransfusion values are included. The graph was
truncated at 20 days because of the limited number
of patients available beyond this time.
20
Liberal
Restrictive
16
8 (1 ß) of 80% and a level of significance
>10
0 1-3 4-6 7-10
No. of Red Blood Cell Units Transfused
Figure 2.—Total number of transfusions per patient
following a restrictive or liberal transfusion strategy.
This graph represents the total number of red blood
cell units transfused per patient in each treatment
group during the study. Significant differences
(P<.001) are noted between the restrictive and lib¬
eral transfusion strategy groups, particularly in the
number of patients receiving less than 4 U.
strictive transfusion strategy resulted
in a 48% relative decrease in RBC trans¬
fusions compared with the more liberal
strategy. The results of this study also
failed to demonstrate that the restrained
use of RBCs increase the risk of signifi¬
cant harm. Specifically, there were no
significant differences between groups
in mortality rates, development of or¬
gan dysfunction or failure, length of ICU
or hospital stay, or intervention rates,
which suggests that both treatment
strategies were comparable.
The lack of a demonstrable treatment
effect may indicate true equivalence, but
may also result from inadequate power
ofthe study. Table 3 lists 95% CIs around
differences based on the study results
for all major study outcomes. As ex¬
pected, all 95% CIs are wide and, as a
result, include zero as well as clinically
meaningful differences. By manipulât-
Downloaded From: http://jama.jamanetwork.com/ by a Florida International University Medical Library User on 05/24/2015
Table 3.—Outcomes of All Patients Enrolled in the TRICC Trial Pilot
Treatment
Restrictive
Variable (n=33)
Mortality rates
30-day 8 (24%)
120-dayt 13(54%)
ICU 5(15%)
Organ failure and dysfunction
Organ dysfunction score, mean ± SD 9.3±3.6
MSOF (=:3 organ failures) 9 (27%)
Median (interquartile range) length of stay, d
ICU 6(3-12)
Hospital 38(25-62)
*TRICC indicates Transfusion Requirements in Critical Care; ICU, intensive care unit; MSOF, multiple system
organ failure; 95% confidence intervals relate to observed difference of the indicated measure.
•( Because of unavailability for follow-up, n=24 in the restrictive group, and n=22 in the liberal group.
ing the sample size equation, it is also
possible to estimate differences in out¬
comes that might be detected in 69 pa¬
tients. Using 25% as the 30-day mor¬
tality rate in the liberal strategy, we
may have missed a true increase in mor¬
tality as great as 56% (a 133% relative
risk increase) or a decrease as small as
2% (a 92% relative risk reduction). More
than 1000 patients would be required to
demonstrate a 25% relative risk reduc¬
tion (24% to 19%) assuming a power
(a) of .05. An analysis using either the
-
CIs or the sample size equation sug¬
gests that this study has insufficient
power to exclude clinically meaningful
differences following the implementa¬
tion of transfusion guidelines.
The treatment protocols were designed
to separate patients into two distinct
groups, to be easily administered, and to
represent current practice patterns. Dif¬
ferences in average daily hemoglobin val¬
ues and the number of RBC units trans¬
fused following the implementation ofboth
treatment strategies were statistically
and clinically significant. On any given
day, 70% to 100% of hemoglobin values
were within the prespecified ranges with
no between-group differences. In addi¬
tion, even though every attempt was made
to minimize crossovers, there were four
(6%) intentional crossovers from one
treatment strategy to another; one pa¬
tient in the restrictive group received
additional blood in a deliberate attempt
to augment oxygen delivery, another pa¬
tient was given blood following a change¬
over in resident staff, and two patients
allocated to the liberal strategy were not
given blood in a timely fashion. Exclud¬
ing the four crossovers from the analysis
did not substantially affect the outcomes.
Therefore, we have established that the
two transfusion strategies can be suc¬
cessfully implemented in critically ill pa¬
tients and result in the separation of he¬
moglobin values into two discrete groups.
A negative study result may have been
Study*
Strategy
95% Confidence Interval
Liberal
(n=36) Lower Upper
9 (25%) -19% 21% >.99
-9% 25% >.99
11(50%)
7 (19%) 22% .76
10.0±3.8 -1 (mean) 2.4 (mean) .44
6 (17%) -9% 29% .38
9(5-12) -2 (median) 3 (median) .53
31(13-64) -25 (median) 6 (median) .66
a consequence of the inappropriate se¬
lection of patients. The prespecified eli¬
gibility criteria may have resulted in the
selection of a low-risk group in which sig¬
nificant benefit or harm from treatment
would not have been expected, a group in
which the treatment impact has limited
biological plausibility, or a heterogeneous
group of patients in which any potential
harm or benefit would be obscured by the
range of results for each outcome exam¬
ined. In examining the characteristics of
the 69 patients enrolled in this study, sev¬
eral baseline variables, including diagnos¬
tic categories, APACHE II scores, and
ages, indicated that patients were at high
risk of adverse events. From the mean
APACHE II scores alone, the overall pre¬
dicted mortality of the cohort was in ex¬
cess of 20%, which was confirmed by our
results.
In this small study, clinically impor¬
tant imbalances in baseline risk factors
may have occurred by chance alone.
Even though we attempted to minimize
such imbalances by stratifying using
APACHE II score, nonsignificant in¬
equalities in the number of organs fail¬
ing and in lactate levels at baseline were
still observed. Interestingly, a greater
number of single organ failures were
observed in the restrictive group while
higher lactate levels were noted in the
liberal group. Thus, the distribution of
these two indicators of disease severity
favored different treatment strategies.
This lack of consistency suggests that
imbalances in these risk factors were
indeed a chance occurrence.
We also examined patient character¬
istics to determine the possibility of an
abnormal linear relationship between
oxygen delivery and oxygen consump¬
tion, termed "pathologic supply depen¬
dency" in both groups.21,23"28 Two prospec¬
tive studies21,23 have documented in¬
creased mortality rates in patients who
exhibited pathologic supply dependency.
It is hypothesized that this phenomenon
is indicative of tissue hypoxia eventually

100
80
60
40
•
Liberal
,20 » Restrictive
10 15
Time, d
Figure 3.—Intensive care unit survival over 20 days
in the restrictive and liberal transfusion strategy
groups. Kaplan-Meier survival curves for both study
groups were compared using a log rank test. The
time to death in days was not significantly different
for both treatment strategies (P=.93). The survival
status of the 25th percentile was 30 days in the re¬
strictive group and 23 days in the liberal group.
contributing to the evolution ofirrevers¬
ible MSOF followed by death. To pre¬
vent this cascade of events, it has been
suggested that oxygen delivery be in¬
creased or maintained at high levels in
critically ill patients. Because hemoglo¬
bin is the principal oxygen carrier in
blood, a transfusion threshold of at least
100 g/L in critically ill patients such as
that used in the liberal transfusion strat¬
egy implemented in this study has been
recommended to maintain or augment
oxygen deliveryP^ It follows that de¬
creased hemoglobin levels maintained in
the restrictive transfusion strategy group
would result in increased organ dysfunc¬
tion scores and mortality rates.2126 The
acuity of illness, the diagnostic catego¬
ries, and the elevated serum lactate lev¬
els on admission to the ICU lead us to
believe that patients enrolled in this
study were at high risk of ongoing tissue
damage from ischemia as a result of
pathologic supply dependency. In the 33
patients randomly allocated to lower he¬
moglobin values, we were unable to docu¬
ment any increase in mortality rates,
MOD scores, or MSOF rates as a result
oftissue ischemia compared with patients
in the liberal transfusion group.
Selecting an overly heterogeneous co¬
hort might also explain our inability to
demonstrate a treatment difference be¬
tween groups. If the selection process
identified a diverse group of patients with
various disease processes, then any po¬
tential effect of transfusions would be
lost because of unpredictable responses
to the intervention. The conflicting lit¬
erature made it difficult to identify one
distinct group of patients at higher risk
in which to expect a treatment benefit.
Further, selective eligibility criteria
would significantly hamper accrual. We
concluded that a study limited to a highly
selected patient group would jeopardize
Downloaded From: http://jama.jamanetwork.com/ by a Florida International University Medical Library User on 05/24/2015
the feasibility of the full study and com¬
promise the generalizibility ofthe results.
Therefore, we elected to use more per¬
missive eligibility criteria. In adopting
such criteria, there is a risk of assem¬
bling an overly heterogeneous cohort. In
practice, however, the eligibility criteria
used in the pilot study selected a rela¬
tively uniform cohort with more than 75%
of patients having one of five clinical di¬
20 agnoses.
To minimize the potential for selec¬
tion bias, blinding of the randomization
process was used. However, there was
agreement that double blinding of the
treatment strategies would not be fea¬
sible. Despite the inability to blind the
treatment strategies, we were unable
to discern any significant differences be¬
tween patient groups in total cumula¬
tive fluid balances, use of vasoactive
drugs or other medications, and patient
interventions. Therefore, there is no evi¬
dence that physicians attempted to in¬
crease oxygen delivery using other
means or that they instituted other
therapies because of the patient assign¬
ment to different transfusion strategies.
Several groups of investigators have
advocated the need for maintaining el¬
evated hemoglobin levels in critically ill
patients.1,24"28 Shoemaker and colleagues24
recommend hematocrit of 0.33 (hemo¬
globin values of 110 g/L) as part of re¬
suscitation protocols in high-risk perio¬
perative patients. They appear to have
adopted this "ideal" hemoglobin level
based solely on one descriptive uncon¬
trolled clinical study. The only other spe¬
cific recommendation1 suggests that all
volume-resuscitated patients with he¬
moglobin values less than 70 g/L be
transfused. That author1 also recom¬
mends doing a careful cardiopulmonary
assessment prior to transfusing criti¬
cally ill patients with hemoglobin levels
greater than 70 g/L.
Recent concerns regarding transfu¬
sion-transmissible diseases such as
AIDS and hepatitis have prompted a
critical review of transfusion practices
and the formulation.6,6 Thus, allogeneic
RBC transfusion guidelines have been
proposed by various agencies.5,6 Such
guidelines do not specifically address the
needs of critically ill patients other than
recommending that clinical judgment
should prevail. Organizations also ac¬
knowledge the absence of controlled
clinical trials evaluating transfusion
practices that would document the risks
and benefits of different approaches.
In our review of the literature, we
were unable to identify any randomized
clinical trials that evaluated transfusion
strategies in the critically ill, although
we identified two small randomized clini¬
cal trials comparing transfusion strate-
gies in patients undergoing coronary ar¬
tery bypass grafting.29·30 One study30
compared a "liberal" group (n=18) that
had hematocrits maintained at 0.32 with
a "conventional" group (n=20) in which
hematocrits were maintained at less than
0.25. The authors concluded that there
were no adverse consequences associ¬
ated with profound normovolemic ane¬
mia. In the second study, Weisel and
colleagues29 randomized 27 patients to
receive either colloids and blood (n=13)
or crystalloids alone (n=14). The signifi¬
cant decrease in transfusion require¬
ments noted was without an associated
increase in the rate of postoperative com¬
plications. In six patients, a statistically
significant increase in the recovery time
of myocardial lactate and oxygen ex¬
traction was observed in the crystalloid
group following coronary revasculariza-
tion. The clinical relevance and accu¬
racy of the myocardial metabolism data
are suspect given the lack of association
with clinical complications, the small
number of patients, and the large num¬
ber of variables potentially confounding
the results. As with this study, the in¬
ability of the two coronary artery by¬
pass grafting trials to demonstrate dif¬
ferences in complication rates also may
have been a function of small sample
size.
Several observational studies also
have been reported.31"37 Among the stud¬
ies relevant to the critically ill, there are
numerous reports of severe anemia be¬
ing well tolerated in high-risk surgical
patients. Descriptive studies in patients
refusing allogeneic RBC transfusion32*4·37
and in regions experiencing hmited blood
supplies35·36 report that most patients
survive surgical interventions even with
hemoglobin levels as low as 45 g/L. The
two small randomized clinical trials and
the large number of observational stud¬
ies suggest that hemoglobin values may
be safely maintained between 70 to 90
g/L in low-risk cardiac and noncardiac
surgical patients. Our observations also
support these conclusions.
In summary, this is the first random¬
ized clinical trial in critically ill patients
attempting to document the safety of
maintaining hemoglobin values between
70 and 90 g/L. These preliminary re¬
sults suggest that a restrictive approach
to the administration of RBCs in nor¬
movolemic critically ill patients is fea¬
sible. However, the small sample size
limits our ability to rule out small but
clinically relevant benefits or harm of
RBC transfusion strategies. Larger clini¬
cal trials are needed prior to the wide¬
spread implementation of treatment
guidelines that advocate the lowering of
transfusion thresholds in critically ill pa¬
tients.
 Study participants from the Canadian Critical
Care Trials Group: G. Wood, MD, Critical Care
Medicine, Richardson House, Kingston, Ontario; J.
Calvin, MD, Rush-Presbyterian-St Luke's Medical
Center, Chicago, Ill; D. Cook, MD, MSc, St Joseph's
Hospital, Hamilton, Ontario; W. Demajo, MD, The
Toronto (Ontario) Hospital; P. Dodek, MD, St Paul's
Hospital, Vancouver, British Columbia; C. Bradley,
MD, Hamilton (Ontario) General Hospital; L. Cham-
pion, MD, University Hospital, London, Ontario; H.
Devitt, MD, Sunnybrook Health Science Centre,
Toronto, Ontario; G. Ford, MD, Calgary (Alberta)
General Hospital; G. Fox, MD, Memorial University
of Newfoundland, St John's; M. Girotti, MD, Victo-
ria Hospital, London, Ontario; R. Hall, Victoria
General Hospital, Halifax, Nova Scotia; P. H\l=e'\bert;
J. Hewson, MD, Hamilton General Hospital; H.
Fuller, MD, St Joseph's Hospital, Hamilton, Ontario;
B. Guslits, MD, Henry Ford Hospital, Detroit, Mich;
S. Hamilton, MD, University Hospital, Edmonton,
Alberta; M. Heule, MD, Misericordia Hospital,
Edmonton, Alberta; D. Heyland, MD, Hamilton
General Hospital; T. Hillers, MD, MSc, Hamilton
General Hospital; P. Houston, MD, The Toronto
Hospital; J. Kronick, MD, Children's Hospital, Lon-
don, Ontario; A. Laupacis, MD, MSc, Ottawa (On-
tario) Civic Hospital; D. Leasa, MD, University
Hospital, London, Ontario; R. Hodder, MD, MSc,
References
1. Cane RD. Hemoglobin: how much is enough?
Crit Care Med. 1990;18:1046-1047.
2. Surgenor DM, Wallace EL, Hao SHS, Chapman
RH. Collection and transfusion ofblood in the United
States, 1982-1988. N Engl J Med. 1990;322:1646-1651.
3. Welch HG, Meehan KR, Goodnough LT. Pru-
dent strategies for elective red blood cell transfu-
sion. Ann Intern Med. 1992;116:393-402.
4. Surgenor DM, Wallace EL, Hale SG, Gilpatrick
MW. Changing patterns of blood transfusions in
four sets of United States hospitals, 1980 to 1985.
Transfusion. 1988;28:513-518.
5. Silberstein LE, Kruskall MS, Stehling LC, et al.
Strategies for the review of transfusion practices.
JAMA. 1989;262:1993-1997.
6. Consensus Conference. Perioperative red blood
cell transfusion. JAMA. 1988;260:2700-2703.
7. Abel FL. Myocardial function in sepsis and en-
dotoxin shock. Am J Physiol. 1989;257:R1265-R1281.
8. Boyd O, Ground M, Bennett D. A randomized
clinical trial of the effect of deliberate perioperative
increase of oxygen delivery on mortality in high\x=req-\
risk surgical patients. JAMA. 1993;270:2699-2707.
9. Tuchschmidt J, Fried J, Astiz ME, Rackow E.
Elevation of cardiac output and oxygen delivery
improves outcome in septic shock. Chest. 1992;102:
216-220.
10. Nichols RL, Smith JW, Klein DB, et al. Risk of
infection after penetrating abdominal trauma.
N Engl J Med. 1984;311:1065-1070.
11. Marik PE, Sibbald WJ. Effect of stored-blood
transfusion on oxygen delivery in patients with
sepsis. JAMA. 1993;269:3024-3029.
12. Messmer K, Sunder-Plassmann L, Klovekorn
WP, Holper K. Circulatory significance of hemo-
dilution: rheological changes and limitations. Adv
Microcirc. 1972;4:1-77.
13. Crosby ET. Perioperative haemotherapy, I: in-
dications for blood component transfusion. Can J
Anesth. 1992;39:695-707.
14. Stehling L, Esposito B. An analysis of the ap-
propriateness of intraoperative transfusion. Anesth
Downloaded From: http://jama.jamanetwork.com/ by a Florida International University Medical Library User on 05/24/2015
Ottawa Civic Hospital; R. Hill, MD, MSc, Health
Science Centre, Calgary, Alberta; J. LaCroix, MD,
H\l=o^\pitalSainte-Justine, Montreal, Quebec; N. Lazar,
MD, General Division, Toronto, Ontario; S. Magder,
MD, Royal Victoria Hospital, Montreal, Quebec; J.
Marshall; S. Moffatt, MD, Kingston (Ontario) Gen-
eral Hospital; L. Oppenheimer, MD, Health Science
Center, Winnipeg, Manitoba; L. Passerini, MD, H\l=o^\-
tel Dieu de Montreal (Quebec); P. Powles, MD, St
Joseph's Hospital, Hamilton, Ontario; R. McIntyre,
MD, Ottawa Civic Hospital; T. Noseworthy, MD,
MHA, Royal Alexandra Hospital, Edmonton, Al-
berta; J. Pagliarello, MD, Ottawa Civic Hospital; S.
Peters, MD, Health Science Centre, St John's,
Newfoundland; D. Roberts, MD, Health Science
Centre, Winnipeg, Manitoba; T. Rosenal, MD, Cal-
gary General Hospital; J. Russell, MD, St Paul's
Hospital, Vancouver, British Columbia; J. D. Sand-
ham, MD, Jewish General Hospital, Montreal, Que-
bec; P. Walker, MD, The Toronto Hospital; R. Wigle,
MD, Kingston General Hospital; P. Roy, MD, Victo-
ria General Hospital, Halifax, Nova Scotia; R. Sae-
schke, MD, St Joseph's Hospital, Hamilton, Ontario;
D. W. Shragge, MD, Hamilton, Ontario; T. R. Todd,
MD (chair), The Toronto Hospital; J. Watters, MD,
Ottawa Civic Hospital; T. Winton, MD, The Toronto
Hospital; M. Tweeddale; M. Heule, MD, Misericor-
dia Hospital, Edmonton, Ontario; A. Slutsky, MD,
Analg. 1989;68:S1-S278.
15. Stehling LC. Preoperative blood ordering. Int
Anesthesiol Clin. 1982;20:45-57.
16. Stehling LC, Ellison N, Faust RJ, Grotta AW,
Moyers JR. A survey oftransfusion practices among
anesthesiologists. Vox Sang. 1987;52:60-62.
17. H\l=e'\bertPC, Schweitzer I, Wells GA, et al. A
survey of red cell transfusion practices in Canadian
critical care practitioners. Clin Invest Med. 1994;
17:B22. Abstract.
18. Knaus WA, Draper EA, Wagner DP, Zimmer-
man JE. APACHE II: a severity of disease clas-
sification system. Crit Care Med. 1985;13:818-829.
19. Friedman LM, Furgerg CD, Demets DL. Fun-
damentals of Clinical Trials. 2nd ed. Littleton,
Mass: PSG Publishing Co Inc; 1985.
20. H\l=e'\bertPC, Drummond AJ, Singer J, Bernard
Gordon R, Russell JE. A simple multiple system
organ failure scoring system predicts mortality of
patients who have sepsis syndrome. Chest. 1993;
104:230-235.
21. Bihari D, Smithies M, Gimson A, Tinker J. The
effects of vasodilation with prostacyclin on oxygen
delivery and uptake in critically ill patients. N Engl
J Med. 1987;317:397-403.
22. Marshall JC. A scoring system for multiple or-
gan dysfunction syndrome. In: Vincent J-L, Rein-
hart K, Eyrich K, Sprung C, eds. Sepsis: Current
Perspectives in Pathophysiology and Therapy. Ber-
lin, Germany: Springer-Verlag; 1994:38-49.
23. Gutierrez G, Pohil RJ. Oxygen consumption is
linearly related to O2 supply in critically ill patients.
J Crit Care. 1986;1:45-53.
24. Czer LSC, Shoemaker WC. Optimal hemato-
crit value in critically ill postoperative patients.
Surg Gynecol Obstet. 1978;147:363-368.
25. Shoemaker WC, Appel PL, Kram HB. Role of
oxygen debt in the development of organ failure,
sepsis, and death in high-risk surgical patients.
Chest. 1992;102:208-215.
26. Shoemaker WC, Appel PL, Kram HB, Wax-
man K, Lee TS. Prospective trial of supranormal
Mt Sinai Hospital, Toronto, Ontario; J. Granton, MD,
Vancouver, British Columbia, B. Wilson, MD, Den-
ver, Colo; A. Kirby, MD, St Joseph Hospital, Lon-
don, Ontario; G. Rocker, MD, Victoria General Hos-
pital, Halifax, Nova Scotia; T. Hillers, MD, MSc,
Hamilton General Hospital; K. Inman, MSc, Victo-
ria Hospital Research Institute, London, Ontario; P.
Boiteau, MD, Mt Sinai Hospital, Toronto, Ontario; P.
Kernerman, MD, Toronto, Ontario; G. Doig, DVM,
London, Ontario; and T. Stewart, MD, Wellesley
Hospital, Toronto, Ontario.
This work was supported by the Canadian Red
Cross Society, Blood Services, Ottawa, Ontario,
and the Physicians' Services Incorporated, North
York, Ontario.
The authors wish to thank Irwin Schweitzer,
MSc, for coordinating this research initiative, and
Diane Ferland, RN, Merrilee Lowen, RN, Deborah
Foster, RN, Denise Foster, RN, and XiangRu Lu,
MD, for their help in completing this project. In
addition, we wish to thank the Canadian Critical
Care Trials Group, and in particular Thomas R.
Todd, MD (chair), for their ongoing support of this
initiative. We also wish to acknowledge Fiona
Daigle and the data management team at the Uni¬
versity of Ottawa, Clinical Epidemiology Unit.
values of survivors as therapeutic goals in high-risk
surgical patients. Chest. 1988;94:1176-1186.
27. Shoemaker WC, Kram HB, Appel PL. Therapy
of shock based on pathophysiology, monitoring and
outcome prediction. Crit Care Med. 1990;18:S19\x=req-\
S25.
28. Czer LSC, Shoemaker WC. Myocardial perfor-
mance in critically ill patients: response to whole
blood transfusions as a prognostic measure. Crit
Care Med. 1980;8:710-715.
29. Weisel RD, Charlesworth DC, Mickleborough
LL, et al. Limitations of blood conservation. J Tho-
rac Cardiovasc Surg. 1984;88:26-38.
30. Johnson RG, Thurer RL, Kruskall MS, et al.
Comparison of two transfusion strategies after elec-
tive operations for myocardial revascularization.
J Thorac Cardiovasc Surg. 1992;104:308-314.
31. Bayer WL, Coenen WM, Jenkins DC, Zucker
ML. The use of blood and blood components in 1769
patients undergoing open-heart surgery. Ann Tho-
rac Surg. 1980;29:117-122.
32. Gollub S, Bailey CP. Management of major sur-
gical blood loss without transfusion. JAMA. 1966;
198:149-152.
33. Carson JL, Spence RK, Poses RM, Bonavita G.
Severity of anaemia and operative mortality and
morbidity. Lancet. 1988;1:727-729.
34. Spence RK, Carson JA, Poses R, et al. Elective
surgery without transfusion: influence of preop-
erative hemoglobin level and blood loss on mortal-
ity. Am J Surg. 1990;159:320-324.
35. Fullerton WT, Turner AG. Exchange transfu-
sion in treatment of severe anemia in pregnancy.
Lancet. 1962;1:75-78.
36. Alexiu O, Mircea N, Balaban M, Furtunescu B.
Gastro-intestinal haemorrhage from peptic ulcer:
an evaluation of bloodless transfusion and early
surgery. Anaesthesia. 1975;30:609-615.
37. Kawaguchi A, Bergsland J, Subramanian S. To-
tal bloodless open heart surgery in the pediatric
age group. Circulation. 1984;70:1-30.