Context Anemia is a common problem in critically ill patients admitted to intensive
care units (ICUs), but the consequences of anemia on morbidity and mortality
in the critically ill is poorly defined.
Objectives To prospectively define the incidence of anemia and use of red blood
cell (RBC) transfusions in critically ill patients and to explore the potential
benefits and risks associated with transfusion in the ICU.
Design Prospective observational study conducted November 1999, with 2 components:
a blood sampling study and an anemia and blood transfusion study.
Setting and Patients The blood sampling study included 1136 patients from 145 western European
ICUs, and the anemia and blood transfusion study included 3534 patients from
146 western European ICUs. Patients were followed up for 28 days or until
hospital discharge, interinstitutional transfer, or death.
Main Outcome Measures Frequency of blood drawing and associated volume of blood drawn, collected
over a 24-hour period; hemoglobin levels, transfusion rate, organ dysfunction
(assessed using the Sequential Organ Failure Assessment score), and mortality,
collected throughout a 2-week period.
Results The mean (SD) volume per blood draw was 10.3 (6.6) mL, with an average
total volume of 41.1 (39.7) mL during the 24-hour period. There was a positive
correlation between organ dysfunction and the number of blood draws (r = 0.34; P<.001) and total
volume drawn (r = 0.28; P<.001).
The mean hemoglobin concentration at ICU admission was 11.3 (2.3) g/dL, with
29% (963/3295) having a concentration of less than 10 g/dL. The transfusion
rate during the ICU period was 37.0% (1307/3534). Older patients and those
with a longer ICU length of stay were more commonly transfused. Both ICU and
overall mortality rates were significantly higher in patients who had vs had
not received a transfusion (ICU rates: 18.5% vs 10.1%, respectively; χ2 = 50.1; P<.001; overall rates: 29.0% vs
14.9%, respectively; χ2 = 88.1; P<.001).
For similar degrees of organ dysfunction, patients who had a transfusion had
a higher mortality rate. For matched patients in the propensity analysis,
the 28-day mortality was 22.7% among patients with transfusions and 17.1%
among those without (P = .02); the Kaplan-Meier log-rank
test confirmed this difference.
Conclusions This multicenter observational study reveals the common occurrence of
anemia and the large use of blood transfusion in critically ill patients.
Additionally, this epidemiologic study provides evidence of an association
between transfusions and diminished organ function as well as between transfusions
Anemia is a common problem in critically ill patients admitted to intensive
care units (ICUs).1 Deleterious effects of
anemia include increased risk of cardiac-related morbidity2
and mortality,2,3 as well as a
generalized decrease in oxygen-carrying capacity. Consequences of anemia may
be compounded in this population since critical illness often increases metabolic
Among the many causes of anemia in the critically ill, some of the most
important are sepsis,4 overt or occult blood
loss (including frequent blood sampling5),
decreased production of endogenous erythropoietin, and immune-associated functional
iron deficiency.6 However, the specific impact
of anemia on ICU patient morbidity and mortality is poorly defined, as is
the optimal hemoglobin level for this population.
Anemia is typically treated with blood transfusions to help maintain
adequate oxygen delivery. Groeger et al7 found
that 16% of patients in medical ICUs and 27% of those in surgical ICUs are
transfused on any given day. In a retrospective chart review in the United
States, 85% of patients with an ICU length of stay greater than 1 week received
at least 1 blood transfusion, and the mean number of units of blood transfused
per patient was 9.5.8 More than two thirds
of ICU transfusions were not associated with acute blood losses.8
While blood transfusions are prevalent among ICU patients, the criteria
for optimal anemia management are not clearly defined.9
A randomized controlled trial by the Canadian Critical Care Trials Group10 comparing restrictive (ie, target hemoglobin level,
7-9 g/dL) and liberal (target level, 10-12 g/dL) transfusion strategies in
ICU patients indicated that the liberal use of transfusions may have resulted
in higher hospital mortality rates.
The use of blood includes health risk and has significant administrative,
logistic, and economic implications. Allogeneic blood transfusion is associated
with risk of allergic reaction,11 infection
transmission,12 and immunosuppression.11,13,14 Additionally, blood
transfusion in the critically ill, particularly with "old" stored cells, does
not consistently improve tissue oxygenation.15
The use of blood transfusions for the treatment of anemia in critically
ill patients warrants further evaluation. This prospective observational study
was designed to determine the frequency of blood drawing and the volume of
blood lost from blood drawing, and to estimate the incidence of anemia and
the use of red blood cell (RBC) transfusions in critically ill patients throughout
western Europe. Such data could provide an objective basis for the future
study of anemia management strategies.
This prospective, observational study had 2 components: a blood sampling
study, and an anemia and blood transfusion study. Institutional recruitment
for participation was by open invitation from the study steering committee
to academic and nonacademic ICUs. Institutional participation was limited
to general medical and/or surgical ICUs in western Europe. Additionally, participation
required a minimum of at least 6 beds to limit any variability in staffing,
technology, or practice associated with smaller ICUs. Since this epidemiologic
observational study did not require any deviation from routine medical practice,
institutional review board approval was either waived or expedited in participating
institutions and informed consent was not required.
The purpose of the blood sampling study was to determine the frequency
of blood drawing and the associated volume of blood lost. The number of blood
draws and volume per draw was recorded for all patients in each ICU for a
24-hour period 1 week prior to the anemia and blood transfusion study. Data
were also collected for the Sequential Organ Failure Assessment (SOFA) score.16
Anemia and Blood Transfusion Study
For all patients admitted to participating ICUs during a 2-week period
(November 15, 1999, through November 29, 1999), data were collected daily
for a maximum of 28 days or until hospital discharge, interinstitutional transfer,
or death. Data collected at ICU admission included demographics, type of admission,
type of surgery (if applicable), primary diagnostic category, comorbidities,
history of anemia and recent acute blood loss, Acute Physiology and Chronic
Health Evaluation II (APACHE II) score, SOFA score, and history of recombinant
human erythropoietin therapy. Data collected daily included morning hemoglobin,
hematocrit, RBC count, SOFA score, indication for transfusion, pretransfusion
hemoglobin (if different from the patient's morning value), and number of
units transfused. The primary indication for transfusion (assessed by the
attending physician) and pretransfusion hemoglobin levels were collected for
the first transfusion of the day only.
Data Management and Statistical Evaluation
Data collected using preprinted case report forms (CRFs) were verified
using a series of domain and consistency edits as defined by the study steering
committee. Intensive care units were queried when data values were either
questionable or were missing for required fields. The CRFs were then scanned
electronically using TELEform scanning software (Cardiff Software Inc, Vista,
Calif) and a database was created using SPSS v9.01 (SPSS Inc, Chicago, Ill).
Unanswered queries concerning required fields were treated as missing and
resulted in an overall loss of 0.06% of data.
Descriptive statistics were computed for all study variables. Bivariate
and multivariate distributions were evaluated for normality using standard
statistical methods and by examining stratified distribution plots. Nonparametric
measures of comparison were used for variables evaluated as not normally distributed.
Difference testing between groups was performed using the 2-tailed t test, analysis of variance (with Bonferroni post-hoc analyses), or χ2 test. Significance for main effects was tested at the α = .05
level. Logistic regression was conducted to assess determinants of mortality.
Kaplan-Meier survival analyses were conducted on the entire sample as well
as on a sample of matched pairs of patients (transfused and nontransfused)
who were matched on their propensity to receive a transfusion. Propensity
scores were obtained through logistic regression of patient characteristics
on transfusion status (ie, actual receipt of a transfusion). A greedy matching
technique17 was used to match unique transfused
patients with unique nontransfused patients based on propensity scores. The
best-match propensity score was identical to 6 digits. Once a match was made,
the control patient was removed from the pool. This process was then repeated
using 5-digit matching, then 4-digit matching, etc. If a transfused patient
matched with more than 1 control, a match was randomly selected. The process
proceeded sequentially to a single-digit match on propensity score. If a match
was not obtained at this point, the transfused patient was excluded.
Cohorts reflecting patient groups of specific clinical interest were
created post hoc. In particular, cohort was defined of those patients with
acute bleeding as an indicator for transfusion at any time or having hemorrhagic
shock on admission.
Table 1 describes the country
participation and the characteristics of the participating ICUs.
Twenty-four–hour blood sampling data were collected on 1136 patients
from 145 participating western European ICUs (see list at end of article).
The number and volume of blood samples drawn in the 24-hour period varied
widely across patients. A total of 45.9% of patients had 5 or more blood samples
drawn. The mean (SD) number of draws per patient was 4.6 (3.2), the mean volume
per draw was 10.3 (6.6) mL, and the average total volume was 41.1 (39.7) mL
for the 24-hour period. There was a significant positive correlation between
organ dysfunction (as measured by the SOFA score) and number of blood draws
(r = 0.34; P<.001) and
total volume drawn (r = 0.28; P<.001).
Anemia and Blood Transfusion Study
Anemia and blood transfusion study data were collected on 3534 patients
from 146 ICUs. Mean (SD) patient age was 61 (17) years, with 33.4% older than
70 years. The majority of patients (62%) were men. Table 2 presents the primary clinical data on admission to the ICU.
The mean admitting APACHE II score was 14.8 (7.9) and the mean admitting SOFA
score was 5.2 (3.8). Overall mean ICU length of stay was 4.5 (6.7) days (≤1
day, 46.6% of patients; 2-3 days, 21.4%; 4-7 days, 14.2%; and >7 days, 17.8%).
The mean (SD) admitting hemoglobin level was 11.3 (2.3) g/dL, with 63%
(2078/3295) of patients having an admitting hemoglobin level less than 12
g/dL and 29% (963/3295) less than 10 g/dL. Admitting hemoglobin was correlated
with age (r = −0.11; P
< .001). The mean admitting hemoglobin level was 11.7 g/dL for patients
younger than 50 years (n = 726), but 11.0 g/dL in patients aged 80 through
90 years (n = 298), and 9.9 g/dL in patients older than 90 years (n = 23)
(F = 6.6; P<.001).
On admission, approximately 13% of patients had a recent history of
anemia; of those patients, 36.6% had an admitting hemoglobin less than 10
g/dL. Of all patients with an admitting hemoglobin level less than 10 g/dL,
48.9% had neither a history of anemia nor of acute bleeding.
Hemoglobin patterns over the 28-day period, stratified by the day 1
hemoglobin level, are illustrated in Figure
1, revealing a convergence of hemoglobin levels over time, irrespective
of the admitting hemoglobin level. Hemoglobin evolution was similar both for
the cohort of patients without acute bleeding (75% of patients, n = 2655)
and the cohort with acute bleeding (data not shown). There were weak negative
correlations between admitting hemoglobin level and admitting SOFA score (r = −0.21, P<.001) and
ICU length of stay (r = −0.25, P<.001).
The transfusion rate (proportion of 3534 patients transfused) during
the ICU period was 37.0%, and the post-ICU transfusion rate was 12.7% (proportion
of all patients initially admitted to the ICU). The overall transfusion rate
(including both ICU and post-ICU timeframes) was 41.6% over the 28-day period.
The mean (SD) age of blood transfused in the ICUs was 16.2 (6.7) days. Leukocyte-depleted
blood was used most of the time by 46% of the ICUs surveyed, some of the time
in 35%, and never in 19%. Transfusion rates differed significantly by hospital
type, with community hospitals transfusing at a rate of 36.4%, regional hospitals
at a rate of 39.8%, and academic hospitals at the highest rate of 44.2% (χ2 = 12.1; P<.002), which reflected the differences
in severity of the patient populations as measured by mean admitting SOFA
score across the hospital types (4.5 for patients in community hospitals,
5.1 for patients in regional hospitals, and 5.4 for patients in academic hospitals; F = 9.2; P<.001).
Two thirds of patients transfused within 24 hours prior to ICU admission
had subsequent transfusions during the survey. Patients admitted for emergency
surgery were transfused more frequently (57.5%), followed by those admitted
for trauma (48.0%), elective surgery (42.1%), and medical reasons (32.0%).
These differences were statistically significant (χ2 = 115.0; P<.001).
Transfusion rates increased consistently by age group, from 29.9% (18-30
years) to 54.2% (>80 years) (χ2 = 63.2; P<.001). However, differences in the mean number of units transfused
by age, during both the ICU stay and throughout the 28-day survey, were not
Transfused patients were older, had higher admitting SOFA and APACHE
II scores, lower admitting hemoglobin levels, and longer hospital lengths
of stay (Table 3).
The overall mean pretransfusion hemoglobin levels were 8.4 (1.3) g/dL
for the entire population and 8.5 (1.1) g/dL for those without acute bleeding. Table 4 shows the distribution of indications
for transfusion as well as the pretransfusion hemoglobin levels for both the
ICU and the post-ICU periods. There were no significant differences in the
mean pretransfusion hemoglobin levels by age, primary admitting diagnosis,
or indication for transfusion (data not shown). Pretransfusion hemoglobin
concentrations exceeded 9 g/dl in less than 30% of cases.
Higher transfusion rates were associated with increased ICU length of
stay (Table 5). Mean ICU length
of stay for transfused patients was 7.2 (median = 3) days compared with 2.6
(median = 1) days for nontransfused patients. Nearly 70% of all transfused
patients received their first transfusion within the first 2 days in the ICU.
Patients whose ICU length of stay was longer than 7 days had an overall transfusion
rate of 73.3%, with a transfusion rate of 52.6% during the first 2 days in
The association between transfusion rates and degree of organ failure,
as assessed by the SOFA score, was statistically significant (χ2 = 358.5; P<.001). When transfused and nontransfused
patients were paired within strata of their admitting SOFA scores, transfused
patients had lower mean admitting hemoglobin levels than nontransfused patients
at every admitting SOFA category (t tests significant
at P<.05 within each SOFA category). Within each
admitting SOFA category, the nontransfused groups consistently showed greater
improvement in organ function over time than the transfused groups (data not
shown). Mean daily SOFA scores were significantly higher in transfused patients
than in nontransfused patients across the entire 28-day survey (P<.01 on all 28 days). Transfused patients both with and without
bleeding had similar SOFA evolution (data not shown). The last recorded SOFA
score, irrespective of length of stay, was significantly lower in the nontransfused
group than in the transfused group (2.8 vs 4.8; t
= −12.8; P<.001).
ICU mortality was 13.5% and 28-day overall mortality was 20.2%. Both
ICU and overall mortality rates were significantly higher for transfused vs
nontransfused patients (ICU: 18.5% vs 10.1%, respectively; χ2
= 50.1; P<.001; overall: 29.0% vs 14.9%, respectively; χ2 = 88.1; P<.001). Higher 28-day mortality
rates in transfused patients than in nontransfused patients persisted across
all SOFA categories in all but the sickest patients (data not shown).
The 28-day survival curves differed significantly by admitting hemoglobin
category in transfused patients (Kaplan-Meier log-rank = 30.3; P<.001) but not in nontransfused patients. Mortality rates were
higher in transfused patients than in nontransfused patients at every hemoglobin
category. There was a dose-response relationship between the number of RBC
units transfused and mortality, with patients receiving the most units having
the highest mortality rates (Table 6).
Hierarchical logistic regression was conducted to determine the associations
between various patient characteristics and death in the ICU. In the first
step, patients' admission characteristics (age, admitting hemoglobin level,
APACHE II score, and SOFA score) were entered into the model; in the second
step, transfusion status (yes/no) was entered. In this model (Table 7), the associations of transfusion, APACHE II, SOFA, and
age with mortality were statistically significant but admitting hemoglobin
level was not. Receipt of a transfusion in the ICU increased a patient's odds
of dying by a factor of 1.37 (95% confidence interval, 1.02-1.84).
To examine the specific association between transfusion and mortality,
propensity scores18 were used to adjust for
differences in all observed background characteristics in the estimation of
the effects of transfusion.19 A logistic regression
model was used to generate a propensity score that reflected the probability
of being transfused for each patient. Variables chosen were age, sex, admission
type, diagnosis on admission, admitting SOFA score, admitting APACHE II score,
day 1 hemoglobin concentration, recent history of anemia, recent acute blood
loss, whether the patient was in shock on admission, and hospital length of
stay. The logistic regression model was able to correctly classify 76% of
patients and produced a Nagelkerke R2
of 0.415. The probability of being transfused ranged from 0.01 to 0.97.
The matching process yielded 516 unique transfused and 516 unique nontransfused
patients who were matched by their propensity scores. Postmatched comparisons
of patient characteristics (from Table 3) are presented in Table 8
and illustrate that the differences between transfused and nontransfused patients
along these parameters were no longer evident.
Among the matched patients, 205 died during the 28-day follow-up period.
Mortality of the transfused patients was 22.7% vs 17.1% for nontransfused
patients (χ21 = 5.1; P
= .02). In other words, of those who died, 57.1% were transfused, and 42.9%
were not. Figure 2 illustrates the
different survival curves for the matched transfused and nontransfused patients.
The Kaplan-Meier log-rank statistic testing the equality of the survival distributions
for the matched groups was 3.99 (df = 1; P = .05), indicating significantly different survival patterns.
This large, epidemiologic study in European ICUs validates the common
occurrence of anemia in critically ill patients and also reports that lower
mean hemoglobin levels were associated with higher SOFA scores, longer lengths
of stay, and higher mortality rates.
Our findings also underline that blood loss through blood sampling is
considerable, averaging 41 mL per day. Sicker patients had more blood draws
per 24 hours resulting in a higher total loss of blood than patients of lower
acuity. Mechanisms to reduce blood sampling losses can include use of point-of-care
testing20 and microchemistry techniques, use
of tubing that eliminates the need for discard,21
a thorough knowledge and communication of the minimum volume needed for each
laboratory test, and storage of blood specimens for potential subsequent use.22 Education and guideline development on such sparing
mechanisms may lead to greater implementation.
The overall transfusion rate during the 28-day study period was 42%,
with an ICU transfusion rate of 37%. While most transfusions were given within
the first week of ICU admission, many transfusions continued to be given throughout
the 28-day follow-up period. In patients with ICU length of stay longer than
1 week, 73% received a blood transfusion. This is slightly lower than the
rate found in a 1995 study by Corwin et al,8
in which 85% of US patients with length of stay longer than 1 week received
a blood transfusion.
The overall mean pretransfusion hemoglobin level was 8.4 g/dL, a value
perhaps lower than expected. Actually, the distribution of pretransfusion
hemoglobin levels indicated that transfusions were given at a hemoglobin level
greater than 9 g/dL in less than 30% of cases. The important results of the
Canadian Critical Care Trials Group10 have
probably influenced practice. Surprisingly, there was no significant difference
in the mean pretransfusion hemoglobin by age, primary admitting diagnosis,
or indication for transfusion. Even patients whose transfusion indication
was acute bleeding had pretransfusion hemoglobin levels similar to those of
patients who had other indications for transfusion. Hence, it seems that the
hemoglobin value is still the major determinant of the decision to transfuse.
While elderly patients had pretransfusion hemoglobin levels similar
to those of younger patients, they had lower daily mean hemoglobin levels
than younger patients (ie, had a higher incidence of reaching the transfusion
trigger) and higher transfusion rates than younger patients. In addition to
acute blood loss, common causes of anemia in elderly patients can include
iron deficiency, chronic disorders, cancer, and nutritional or metabolic disorders.23 Aging itself may influence erythropoiesis, with possible
contributors being a reduction in pluripotent hematopoetic stem cell reserve,
reduced production of growth factors, reduced sensitivity of stem cells and
progenitors to growth factors, and microenvironmental abnormalities.24
Transfused patients had longer ICU length of stay, more severe organ
failure, and higher mortality rates than nontransfused patients, but these
associations may be explained by different underlying clinical conditions.
An altered or blunted erythropoietic response characteristic of anemia of
chronic disease may also contribute to anemia in these sickest patients.1,4
We found that mortality rates differed significantly by organ failure
and transfusion status. As expected, mortality increased in all patients as
organ failure increased, but mortality was higher for transfused patients
at all levels of organ dysfunction with the exception of those with the most
severe organ failure.
Transfused patients had higher mortality rates at every admitting hemoglobin
level when compared with nontransfused patients. Mortality rates also differed
significantly by admitting hemoglobin group for transfused, but not for nontransfused,
patients. There was a dose-response relationship with increasing mortality
rates as the number of RBC units transfused increased. Receipt of a blood
transfusion increased the risk of dying by a factor of 1.4 (all other variables
being equal). Matching patients by their propensity scores for receipt of
a transfusion (and hence controlling for age, admission type, diagnosis, admitting
SOFA score, admitting APACHE II score, and length of stay) likewise revealed
higher mortality rates in transfused patients.
Of the deleterious effects of RBC transfusion, the likely contributing
factor to mortality in the critically ill population is related to immunosuppression
as opposed to allergic reaction or infectious transmission. Red blood cell
transfusion has been associated with a higher incidence of postoperative infection
in various populations.25-27
The use of leukodepleted blood may result in shorter lengths of stay and fewer
complications.28,29 At the time
of this study, the use of deleukocyted blood in participating institutions
was variable. Variability in storage techniques may also contribute to RBC
deformability that may influence cell oxygen availability.
Due to the risks associated with blood transfusions, alternative treatments
of anemia in the critically ill are being explored. The administration of
exogenous human recombinant erythropoietin (epoetin alfa) has been shown to
raise reticulocyte counts26,27
and hematocrit levels, and to reduce the total number of units of transfused
blood required in critically ill patients.28
Interpretation of findings from this observational study should be limited
to identification and description of associations between variables of interest.
No intervention was implemented, no attempt to control or influence provision
of care was made, and the physicians were not blinded to the study purpose.
Additionally, the illnesses and treatments of the critically ill are complex
and confounded by a multitude of factors. Moreover, data collection in studies
of this design is necessarily limited, and not all factors that may have influenced
physicians' transfusing practice may not have been collected. Therefore, no
causal inferences for associations can be drawn.
The findings from this study reveal that anemia in critically ill patients
is common. Transfusions are also common, especially in elderly patients and
those with longer ICU lengths of stay. Even though transfusions currently
seem to be given primarily in response to decreased hemoglobin levels, the
trigger for blood transfusions averaged only 8.4 g/dL. This study provides
evidence of the associations between transfusion and diminished organ function
as well as between low hemoglobin level, transfusion, and mortality. Clearly,
further study is needed to clarify the place of transfusion in critically
ill patients and to evaluate other treatments for anemia that reduce the need
for blood transfusions, such as the use of recombinant human erythropoietin
von Ahsen N, Muller C, Serke S, Frei U, Eckardt KU. Important role of nondiagnostic blood loss and blunted erythropoietic
response in the anemia of medical intensive care patients. Crit Care Med.1999;27:2630-2639.Google Scholar
Nelson AH, Fleisher LA, Rosenbaum SH. Relationship between postoperative anemia and cardiac morbidity in
high-risk vascular patients in the intensive care unit. Crit Care Med.1993;21:860-866.Google Scholar
Hébert PC, Wells G, Tweeddale M.
et al. for the Transfusion Requirements in Critical Care (TRICC) Investigators
and the Canadian Critical Care Trials Group. Does transfusion practice affect mortality in critically ill patients? Am J Respir Crit Care Med.1997;155:1618-1623.Google Scholar
Rogiers P, Zhang H, Leeman M.
et al. Erythropoietin response is blunted in critically ill patients. Intensive Care Med.1997;23:159-162.Google Scholar
Viljoen M, Coetzee IH, Roux JJ, Pretorius JP. Anemia in surgical intensive care patients. Haematologica.1994;79:19-24.Google Scholar
Groeger JS, Guntupalli KK, Strosberg M.
et al. Descriptive analysis of critical care units in the United States: patient
characteristics and intensive care unit utilization. Crit Care Med.1993;21:279-291.Google Scholar
Corwin HL, Parsonnet KC, Gettinger A. RBC transfusion in the ICU: is there a reason? Chest.1995;108:767-771.Google Scholar
Goodnough LT, Despotis GJ, Hogue Jr CW, Ferguson Jr TB. On the need for improved transfusion indicators in cardiac surgery. Ann Thorac Surg.1995;60:473-480.Google Scholar
Hébert PC, Wells G, Blajchman MA.
et al. for the Transfusion Requirements in Critical Care Investigators, Canadian
Critical Care Trials Group. A multicenter, randomized, controlled clinical trial of transfusion
requirements in critical care. N Engl J Med.1999;340:409-417.Google Scholar
Brunson ME, Alexander JW. Mechanisms of transfusion-induced immunosuppression. Transfusion.1990;30:651-658.Google Scholar
Stevens CE, Aach RD, Hollinger FB.
et al. Hepatitis B virus antibody in blood donors and the occurrence of non-A,
non-B hepatitis in transfusion recipients: an analysis of the Transfusion-Transmitted
Viruses Study. Ann Intern Med.1984;101:733-738.Google Scholar
Blumberg N, Heal JM, Murphy P, Agarwal MM, Chuang C. Association between transfusion of whole blood and recurrence of cancer. Br Med J (Clin Res Ed).1986;293:530-533.Google Scholar
Blumberg N, Heal JM. Immunomodulation by blood transfusion: an evolving scientific and clinical
challenge. Am J Med.1996;101:299-308.Google Scholar
Marik PE, Sibbald WJ. Effect of stored-blood transfusion on oxygen delivery in patients with
sepsis. JAMA.1993;269:3024-3029.Google Scholar
Vincent JL, Moreno R, Takala J.
et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe
organ dysfunction/failure. Intensive Care Med.1996;22:707-710.Google Scholar
Gum PA, Thamilarasan M, Watanabe J, Blackstone EH, Lauer MS. Aspirin use and all-cause mortality among patients being evaluated
for known or suspected coronary artery disease. JAMA.2001;286:1187-1194.Google Scholar
Rosenbaum PR, Rubin DB. Reducing bias in observational studies using subclassification on the
propensity score. J Am Stat Assoc.1984;79:516-524.Google Scholar
Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med.1997;127:757-763.Google Scholar
Schallom L. Point of care testing in critical care. Crit Care Nurs Clin North Am.1999;11:99-106.Google Scholar
Silver MJ, Li YH, Gragg LA, Jubran F, Stoller JK. Reduction of blood loss from diagnostic sampling in critically ill
patients using a blood-conserving arterial line system. Chest.1993;104:1711-1715.Google Scholar
Dale JC, Pruett SK. Phlebotomy—a minimalist approach. Mayo Clin Proc.1993;68:249-255.Google Scholar
Ania BJ, Suman VJ, Fairbanks VF, Rademacher DM, Melton III LJ. Incidence of anemia in older people: an epidemiologic study in a well
defined population. J Am Geriatr Soc.1997;45:825-831.Google Scholar
Balducci L, Hardy CL. Anemia of aging: a model of erythropoiesis in cancer patients. Cancer Control.1998;5:17-21.Google Scholar
Jensen LS, Andersen AJ, Christiansen PM.
et al. Postoperative infection and natural killer cell function following
blood transfusion in patients undergoing elective colorectal surgery. Br J Surg.1992;79:513-516.Google Scholar
Jensen LS, Hokland M, Nielsen HJ. A randomized controlled study of the effect of bedside leucocyte depletion
on the immunosuppressive effect of whole blood transfusion in patients undergoing
elective colorectal surgery. Br J Surg.1996;83:973-977.Google Scholar
Tartter PI, Mohandas K, Azar P, Endres J, Kaplan J, Spivack M. Randomized trial comparing packed red cell blood transfusion with and
without leukocyte depletion for gastrointestinal surgery. Am J Surg.1998;176:462-466.Google Scholar
Gott JP, Cooper WA, Schmidt FE.
et al. Modifying risk for extracorporeal circulation: trial of four antiinflammatory
strategies. Ann Thorac Surg.1998;66:747-753.Google Scholar
van de Watering LM, Hermans J, Houbiers JG.
et al. Beneficial effects of leukocyte depletion of transfused blood on postoperative
complications in patients undergoing cardiac surgery: a randomized clinical
trial. Circulation.1998;97:562-568.Google Scholar
Gabriel A, Kozek S, Chiari A.
et al. High-dose recombinant human erythropoietin stimulates reticulocyte
production in patients with multiple organ dysfunction syndrome. J Trauma.1998;44:361-367.Google Scholar
van Iperen CE, Gaillard CA, Kraaijenhagen RJ, Braam BG, Marx JJ, van de Wiel A. Response of erythropoiesis and iron metabolism to recombinant human
erythropoietin in intensive care unit patients. Crit Care Med.2000;28:2773-2778.Google Scholar
Corwin HL, Gettinger A, Rodriguez RM.
et al. Efficacy of recombinant human erythropoietin in the critically ill
patient: a randomized, double-blind, placebo-controlled trial. Crit Care Med.1999;27:2346-2350.Google Scholar