Survival data were missing for 3 patients who received tranfusion and
for 27 patients who did not receive transfusion.
Odds ratios were adjusted for baseline characteristics (site, age, race,
weight in kilograms, diabetes mellitus, systolic and diastolic blood pressure,
heart rate at baseline, time from symptom onset to hospitalization, prior
stroke, prior myocardial infarction, sex, history of angina prior to qualifying
episode, hypertension, hyperlipidemia, family history of coronary artery disease,
history of congestive heart failure, peripheral vascular disease, prior percutaneous
coronary intervention [PCI], prior coronary artery bypass graft surgery [CABG],
Killip class, baseline hematocrit, maximum creatine kinase ratio at baseline,
chronic renal insufficiency, ST-segment elevation or depression on initial
electrocardiogram, β-blocker use at baseline, calcium channel blocker
use at baseline, nitrate use at baseline, and current smoking), bleeding events
occurring before the end of each time period, and procedures (PCI and CABG)
occurring before the end of each time period.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Rao SV, Jollis JG, Harrington RA, et al. Relationship of Blood Transfusion and Clinical Outcomes in Patients With Acute Coronary Syndromes. JAMA. 2004;292(13):1555–1562. doi:10.1001/jama.292.13.1555
Context It is unclear if blood transfusion in anemic patients with acute coronary
syndromes is associated with improved survival.
Objective To determine the association between blood transfusion and mortality
among patients with acute coronary syndromes who develop bleeding, anemia,
or both during their hospital course.
Design, Setting, and Patients We analyzed 24 112 enrollees in 3 large international trials of
patients with acute coronary syndromes (the GUSTO IIb, PURSUIT, and PARAGON
B trials). Patients were grouped according to whether they received a blood
transfusion during the hospitalization. The association between transfusion
and outcome was assessed using Cox proportional hazards modeling that incorporated
transfusion as a time-dependent covariate and the propensity to receive blood,
and a landmark analysis.
Main Outcome Measure Thirty-day mortality.
Results Of the patients included, 2401 (10.0%) underwent at least 1 blood transfusion
during their hospitalization. Patients who underwent transfusion were older
and had more comorbid illness at presentation and also had a significantly
higher unadjusted rate of 30-day death (8.00% vs 3.08%; P<.001), myocardial infarction (MI) (25.16% vs 8.16%; P<.001), and death/MI (29.24% vs 10.02%; P<.001)
compared with patients who did not undergo transfusion. Using Cox proportional
hazards modeling that incorporated transfusion as a time-dependent covariate,
transfusion was associated with an increased hazard for 30-day death (adjusted
hazard ratio [HR], 3.94; 95% confidence interval [CI], 3.26-4.75) and 30-day
death/MI (HR, 2.92; 95% CI, 2.55-3.35). In the landmark analysis that included
procedures and bleeding events, transfusion was associated with a trend toward
increased mortality. The predicted probability of 30-day death was higher
with transfusion at nadir hematocrit values above 25%.
Conclusions Blood transfusion in the setting of acute coronary syndromes is associated
with higher mortality, and this relationship persists after adjustment for
other predictive factors and timing of events. Given the limitations of post
hoc analysis of clinical trials data, a randomized trial of transfusion strategies
is warranted to resolve the disparity in results between our study and other
observational studies. We suggest caution regarding the routine use of blood
transfusion to maintain arbitrary hematocrit levels in stable patients with
ischemic heart disease.
The use of invasive procedures for treatment of ischemic heart disease
has more than tripled in the past 2 decades and is likely to increase in high-risk
patients.1 This, coupled with the widespread
use of potent fibrinolytic and antithrombotic drugs,2,3 has
increased the potential for bleeding and blood transfusion among patients
with cardiovascular disease. Approximately 12 million units of blood are transfused
to 3.5 million patients each year in the United States,4 and
although transfusing blood to anemic patients with ischemic heart disease
may theoretically increase oxygen delivery and improve outcomes, there is
no definitive evidence to support such a practice. Some studies actually indicate
no increase in tissue oxygenation with blood transfusion.5-7
Studies of clinical outcomes have shown disparate findings. A randomized
trial found no benefit of liberally transfusing blood in critically ill patients
to maintain a hemoglobin level of 10.0 mg/dL compared with restricting transfusion
to patients in whom the hemoglobin was 7.0 mg/dL or lower.8 A
post hoc analysis of this trial, limited to patients with cardiovascular dis
ease, supported the overall results.9 In contrast,
an observational study of elderly patients with acute myocardial infarction
(MI) found an association between transfusion and improved short-term survival
when hematocrit at admission was 30% to 33% or less.10 This
study did not examine the association between transfusion and outcome in patients
who developed anemia during their hospital course.
Patients hospitalized for an acute coronary syndrome (ACS) are at risk
of developing anemia acutely as a consequence of bleeding. For clinical practice,
a crucial issue is whether blood transfusion is beneficial or harmful for
patients with ischemic heart disease who have developed anemia acutely during
We used detailed clinical data from 3 large international trials of
patients with ACS to determine the association between blood transfusion and
outcomes among patients who developed moderate to severe bleeding, anemia,
or both during their hospitalization.
The institutional review boards of all participating institutions reviewed
and approved the protocols of the Global Use of Strategies to Open Occluded
Coronary Arteries (GUSTO) IIb, Platelet Glycoprotein IIb/IIIa in Unstable
Angina: Receptor Suppression Using Integrilin Therapy (PURSUIT), and Platelet
IIb/IIIa Antagonism for the Reduction of Acute Coronary Syndrome Events in
a Global Organization Network (PARAGON) B trials. All patients enrolled gave
written informed consent.
Clinical data from the multicenter international GUSTO IIb, PURSUIT,
and PARAGON B trials were pooled and included 24 112 patients with ACS.
The details of the trials have been published elsewhere.3,11,12 Briefly,
GUSTO IIb randomly assigned 12 142 patients with ACS to receive either
intravenous heparin or hirudin. For this analysis, we included 8011 patients
from GUSTO IIb without persistent ST-segment elevation on initial electrocardiogram.
PURSUIT randomly assigned 10 948 ACS patients without persistent ST-segment
elevation to receive either eptifibatide or placebo; PARAGON B randomly assigned
5225 ACS patients without ST-segment elevation to either intravenous lamifiban
or placebo. For the current study, the analysis was limited to patients from
the 3 trials who had complete data on transfusion and bleeding occurrence.
Concomitant treatment with aspirin in dose ranges of 80 to 325 mg/d
was recommended by protocol in all 3 trials. Use of antithrombin agents was
also recommended in the PURSUIT andPARAGON B trials and was mandated by protocol
in the GUSTO IIb trial. Use of other medications and procedures was at the
discretion of the treating physicians in all 3 trials.
Bleeding. The GUSTO IIb investigators used
the GUSTO definition of bleeding2 that classifies
bleeding as mild, moderate, severe, or life-threatening. Severe or life-threatening
bleeding was defined as either intracranial hemorrhage or bleeding that caused
hemodynamic compromise and required intervention. Moderate bleeding was defined
as bleeding that required blood transfusion but did not result in hemodynamic
compromise. The PURSUIT investigators used the GUSTO and TIMI (Thrombolysis
in Myocardial Infarction)13 bleeding classifications.
The TIMI classification defines bleeding events as major or minor, where major
bleeding is either intracranial hemorrhage or bleeding associated with a hemoglobin
decrease of 5 g/dL or more (or a hematocrit decrease of ≥ 15%). Minor
bleeding is defined as observed blood loss resulting in a hemoglobin decrease
of 3 g/dL or more (hematocrit decrease of ≥ 10%) or a decrease in
hemoglobin of 4 g/dL (hematocrit decrease of ≥ 12%) if no bleeding
site was identifiable.
The PARAGON B investigators defined bleeding complications as major
or life-threatening and intermediate. Major or life-threatening bleeding was
defined as any intracranial hemorrhage or bleeding leading to hemodynamic
compromise requiring intervention. Intermediate bleeding was defined as bleeding
requiring transfusion or a decrease in hemoglobin of 5 g/dL or more, or a
decrease in hematocrit ≥ 15% when hemoglobin measurement was unavailable.
Data on the date, time, severity, and location (including unidentifiable)
of each bleeding event were collected prospectively. To be consistent across
trials, the GUSTO definition of bleeding was used for the GUSTO IIb and PURSUIT
trials; for PARAGON B, major or life-threatening bleeding episodes and intermediate
bleeding episodes were considered to be GUSTO severe and moderate bleeding,
respectively. For the purpose of this analysis, only the first moderate or
severe bleeding episode was considered, and the nadir hemoglobin or hematocrit
was defined as the lowest value occurring during the hospitalization if no
transfusion or bleeding occurred. When studying bleeding or transfusion events,
only the nadir level before the event was considered. Nadir hematocrit was
considered a continuous variable.
Transfusion. Data on the number of units of
packed red blood cells and whole blood transfused as well as the date of transfusion
were collected prospectively in each trial.
End Points. The primary end point was 30-day
all-cause mortality. Secondary end points were occurrence of the composite
of 30-day death or MI. Myocardial infarction was defined according to the
protocol of each trial. The GUSTO IIb investigators defined MI as an increase
in creatine kinase–MB (CK-MB) fraction (or total CK, if CK-MB measurement
was unavailable) to greater than the upper limit of normal or at least 2 times
the previous value if it was elevated at enrollment and/or new significant
Q waves in 2 contiguous electrocardiographic (ECG) leads, along with the appropriate
signs and symptoms.
The PURSUIT and PARAGON B investigators defined MI as new chest pain
and ST-segment elevation within 18 hours of enrollment, new or repeat CK-MB
fraction elevation greater than the upper limit of normal after 18 hours,
new Q waves in 2 contiguous ECG leads, or both. Creatine kinase–MB elevations
greater than 3 times the upper limit of normal after percutaneous coronary
intervention and greater than 5 times the upper limit of normal after coronary
artery bypass graft surgery (CABG) were also classified as MI. All end points
were adjudicated by an independent, blinded events committee.
Patient Comparisons. Patients were categorized
according the occurrence of transfusion. Baseline characteristics were compared
using χ2 tests for categorical variables and the nonparametric
Kruskal-Wallis test for continuous variables. Baseline differences with P values <.01 were considered statistically significant.
Kaplan-Meier analysis was used to illustrate 30-day event-free survival for
patients who did and did not undergo transfusion. Analyses were computed using
SAS software, version 8.2 (SAS Institute Inc, Cary, NC).
Modeling of Outcomes. Because blood transfusion
was a postrandomization event that was left to the discretion of the investigator,
the association between transfusion and the primary and secondary end points
could be confounded by patient characteristics and influenced by in-hospital
events such as bleeding and procedures. To control for these biases, we developed
4 statistical models. The first 2 models examined patients’ propensity
to bleed or receive a transfusion and used moderate or severe bleeding and
transfusion, respectively, as the outcomes. Logistic regression using a stepwise
variable selection technique was used in each model and incorporated baseline
demographic characteristics, medical comorbidities, age (as a continuous variable),
sex, body weight (as a continuous variable), presenting characteristics, baseline
hematocrit, site (US vs non-US), and in-hospital medical therapy received
within the 2 weeks prior to randomization as independent variables.
Because the likelihood of receiving a transfusion may vary over time,
2 further models were developed. One used Cox proportional hazards regression
to determine the association between transfusion and 30-day death and incorporated
transfusion as a time-dependent covariate. The use of transfusion as a time-dependent
covariate enables accounting for survivor bias (ie, not living long enough
to undergo blood transfusion) and for the possibility that the timing of the
transfusion relative to the outcome may be influential (eg, if the transfusion
occurred after MI). The model was then adjusted for baseline variables found
to be predictive of 30-day death among patients with non–ST-segment
elevation ACS,14 propensity for bleeding and
transfusion from the models described herein, and nadir hematocrit. Because
of the influence of CABG on transfusion practice and mortality, we repeated
the analysis by censoring patients at the time of CABG.
The final model of 30-day death incorporated transfusion as a “time-fixed”
covariate in a landmark analysis.15,16 With
this approach, the follow-up time is divided into periods of interest. Patient
survival is then described with standard techniques conditional on the patient
being alive and not yet having received a transfusion at the start of the
period. All procedures and bleeding events that occurred prior to the end
of each interval are included in the analysis. This approach provides a general
trend of the adjusted association between the independent variable (transfusion)
and dependent variable (30-day mortality) over time.
For the purposes of this study, the analysis was performed on the first
seven 24-hour periods after trial enrollment because the majority of events
(transfusions, bleeding events, and procedures) occurred during this time
interval. The analysis for each time period compares outcomes between patients
who did and did not undergo transfusion within the discrete 24-hour period
and then adjusts for differences between these 2 populations. The analysis
adjusted for baseline characteristics, nadir hematocrit occurring prior to
the end of each interval, and bleeding and invasive procedures that occurred
prior to the end of each interval.
This approach has several advantages. First, it deals with nonproportionality
because the analysis is fitted to a restricted time period. Second, it minimizes
survivor bias because the analysis is performed on data captured within relatively
short time intervals (24 hours in the case of our study) among patients who
are alive at the start of each period. Third, it incorporates other covariates
that may be time-dependent, such as invasive procedures and bleeding. In light
of prior work showing an association between transfusion and lower mortality
at certain hematocrit levels in elderly persons,10 we
also explored the interactions between age and transfusion and baseline and
nadir hematocrit and transfusion in the landmark analyses.
We also examined the predicted probability of 30-day death among patients
undergoing and not undergoing transfusion using a multivariable logistic regression
model that incorporated nadir hematocrit as a continuous variable and adjusted
for baseline characteristics. The association of nadir hematocrit with 30-day
mortality was evaluated using restricted cubic splines. It appeared that this
association followed 2 lines, 1 below a nadir hematocrit value of 25% and
1 above a nadir hematocrit value of 25%. Therefore, a linear spline transformation
with a nadir hematocrit value of 25% as the knot point was used. The 2 continuous
components of this transformation were added to the model along with transfusion
use and the interaction of transfusion with nadir hematocrit. Owing to the
influence of coronary artery bypass surgery on transfusion practice and mortality,
we repeated the analysis censoring at the time of CABG. To account for survival
bias, we also repeated the analysis excluding patients who died within the
first 5 days.
From the 3 trials, 24 112 patients had complete data on bleeding,
transfusion, and outcomes. Of these, 2401 (10.0%) underwent transfusion of
at least 1 unit of whole blood or packed red blood cells. Table 1 displays the baseline characteristics of patients who did
and did not receive a transfusion. Compared with those who did not receive
blood, patients who received a blood transfusion were older, more often female,
more often black, and had lower median body weight. They also had more medical
comorbidities at presentation, with a higher proportion having ST-segment
depression on the initial ECG and Killip class greater than II. The median
baseline and nadir hematocrit measurements for patients who received a transfusion
were 39.9% and 29.0%, respectively. For patients who did not receive blood,
the values were 41.7% and 37.6%, respectively.
The first 2 regression models examined significant baseline predictors
of moderate or severe bleeding and blood transfusion. Table 2 shows the baseline characteristics that were most associated
with bleeding and blood transfusion during hospitalization. Similar baseline
characteristics were associated with both bleeding and blood transfusion.
Kaplan-Meier and Cox Regression Analyses. Figure 1 shows the Kaplan-Meier curves for 30-day
mortality among patients who did and did not receive blood transfusion. Table 3 shows the unadjusted rates of 30-day
death, MI, and composite death/MI among patients who did and did not receive
a transfusion. For all 3 outcomes, the rates were significantly higher among
patients who received a transfusion (30-day death, 8.00% for patients who
received a transfusion vs 3.08% for patients who did not; P<.001; 30-day MI, 25.16% vs 8.16%; P<.001;
30-day composite death/MI, 29.24% vs 10.02%; P<.001).
Table 3 also shows the results
of the Cox model that examined the association between blood transfusion as
a time-dependent covariate and 30-day death and 30-day composite death/MI.
After adjustment for baseline characteristics, blood transfusion was associated
with a hazard ratio for death of 3.54 (95% confidence interval, 2.96-4.23)
for 30-day death. After adjustment for baseline characteristics, bleeding
and transfusion propensity, and nadir hematocrit, blood transfusion was associated
with a hazard ratio for death of 3.94 (95% confidence interval, 3.26-4.75).
Landmark Analysis. Figure 2 shows the results of the landmark analysis that adjusted
for baseline characteristics, nadir hematocrit occurring prior to the end
of each time period, bleeding events that occurred prior to the end of each
time period, and invasive procedures (cardiac catheterization, percutaneous
coronary intervention, and/or CABG) that occurred before the end of each time
period. During the first 7 days after randomization, there was a trend association
between blood transfusion and increased 30-day mortality. In the landmark
analysis, there were no significant interactions between transfusion and age
or transfusion and baseline or nadir hematocrit.
Predicted Probabilities of 30-Day Death. Table 4 shows the adjusted predicted probability
of 30-day death with and without transfusion by nadir hematocrit. The interaction
between nadir hematocrit and transfusion was significant (P = .003) such that there was no significant association
between transfusion and 30-day mortality at a nadir hematocrit of 25% or less.
However, at a nadir hematocrit higher than 25%, transfusion was associated
with a significantly higher odds of 30-day death. The results were unchanged
after excluding patients who underwent CABG or those who died within the first
5 days of follow-up.
The results of our study show that blood transfusion in the setting
of anemia during hospitalization for ACS is associated with increased 30-day
mortality. This association persisted across the 3 different analytical methods
we used. The increased risk of death associated with transfusion was present
after adjustment for demographic characteristics and in-hospital events such
as bleeding and invasive procedures. When included as a time-dependent covariate
in the Cox model, blood transfusion was associated with a higher risk of death.
In the landmark analysis, the odds ratios showed a trend toward increased
mortality with transfusion after adjustment for both baseline and nadir hematocrit.
When hematocrit level was included as a continuous variable in the logistic
regression model, we found an association between transfusion and increased
30-day mortality at a nadir hematocrit above 25%. This suggests that a hematocrit
as low as 25% may be tolerated without blood transfusion in otherwise stable
patients with ischemic heart disease.
Our findings differ from those of Wu et al,10 who
analyzed data from an administrative database and found that blood transfusion
was associated with lower 30-day mortality among elderly patients with MI
if the admission hematocrit was 30% or lower. There are many likely reasons
for the disparity between our study and that of Wu et al. First, Wu et al
used hematocrit measurement at admission in their analysis, whereas we examined
the association among anemia developing during the hospitalization (ie, nadir
hematocrit), transfusion, and mortality. The latter is a critical issue for
clinical practice. Given that many current therapies for ACS rely on mechanisms
that increase the risk of bleeding (antithrombotic medications and invasive
procedures), a fundamental problem facing clinicians is whether to use transfusion
in patients who are otherwise stable but have developed anemia as a consequence
of medications, procedures, or both. We included in-hospital procedures and
bleeding events, which are important drivers of transfusion, in our landmark
analysis, while Wu et al did not.
Second, Wu et al used an observational data set based on Medicare claims
data. Although the clinical information was abstracted from hospital records,
data on transfusion were likely derived from claims that may have been incomplete.
Our analysis was performed on information from clinical trials databases in
which data collection, especially bleeding and transfusion data, was meticulous.
Third, Wu et al excluded patients younger than 65 years, those with
bleeding within 48 hours of admission, and those who underwent open-heart
surgery. In our analysis, we included all patients, regardless of age, bleeding
events, or procedures, for whom all clinical information was complete.
Finally, Wu et al attempted to control for survival bias (ie, living
long enough to receive a transfusion) in a secondary analysis by excluding
patients who died within 48 hours of admission. This eliminated the association
between transfusion and improved mortality in patients with a hematocrit of
30% to 33%. We believe that our statistical methods were robust because we
performed our analysis first by including transfusion as a time-dependent
covariate and second by using a landmark analysis. Both methods not only minimized
survivor bias but the landmark analysis also included other time-dependent
events such as bleeding and procedures.
Our results also run counter to conventional clinical thinking about
cardiac function and anemia. Mild to moderate anemia (hemoglobin level of
7.0-10.0 g/dL) increases cardiac output, primarily through reduced blood viscosity
leading to reduced afterload. Under these conditions, myocardial oxygen demand
does not change.17 The myocardium has a high
oxygen-extraction ratio, however, and can augment oxygen delivery only by
increasing coronary blood flow. Such an increase might not be possible in
patients with fixed coronary stenoses. Considerable experimental model data
suggest that a hemoglobin level of 7 g/dL is tolerated without myocardial
ischemia if there is no obstructive coronary artery disease. With coronary
artery obstruction, however, ischemia can occur with even mild anemia in experimental
circumstances.18 Furthermore, prior observational
studies have shown an association between anemia and increased mortality in
patients with cardiovascular disease.10,19 In
this circumstance, there are no definitive data that show that treating anemia
with blood transfusion either mitigates myocardial ischemia or improves survival.
While clinical studies suggest that increasing hemoglobin level via
transfusion increases oxygen delivery,5,7,20 studies
also show that measures of tissue oxygenation either decrease or do not change.5-7 Increasing oxygen delivery
through transfusion leads to increases in oxygen utilization by tissues only
at severe levels of anemia. At higher but subnormal hematocrit levels, this
relation does not appear to exist—as delivery increases, tissue uptake
decreases and tissue utilization of oxygen remains constant.
The reason for this paradox (greater oxygen delivery but no improvement
in tissue use) is unclear. Alterations in erythrocyte nitric oxide biology
in stored blood may be a partial explanation. Nitric oxide (NO) is essential
for oxygen exchange,21 but the half-life of
NO in erythrocytes is believed to be short. Red blood cells in stored blood,
once depleted of NO, may function as NO “sinks,” promoting vasoconstriction,
platelet aggregation, and ineffective oxygen delivery. Moreover, red blood
cells in stored blood are low in 2,3-diphosphoglyceric acid and have high
oxygen affinity,17 which may further impair
the delivery of oxygen to hypoxic tissues. Also, administration of blood to
patients with coronary artery disease may lead to increases in inflammatory
mediators that are associated with exacerbation of myocardial ischemia.22 All of this, in aggregate, may act to promote myocardial
ischemia rather than mitigate it.
Previous randomized studies support the conclusion that blood transfusion
may, at best, be neutral with respect to survival or, at worst, be associated
with either decreased survival or worsening cardiac function. Fortune et al5 examined the effect of maintaining a hematocrit of
30% vs 40% on hemodynamic variables in 25 patients with trauma, acute hemorrhage,
or both. They found no differences in cardiac index, heart rate, or left ventricular
stroke index between the groups. Johnson et al23 compared
a liberal transfusion strategy (hematocrit of 32%) vs a conservative strategy
(hematocrit of 25%) in 38 patients undergoing elective CABG. They found no
adverse effects with the conservative strategy and reported better exercise
tolerance in this group. Bush et al24 preoperatively
randomized 99 patients undergoing elective aortic and infrainguinal arterial
reconstruction procedures to receive either a liberal transfusion strategy
(maintain a hemoglobin level ≥10 g/dL) or a restrictive strategy (transfusion
only for hemoglobin level <9.0 g/dL). In an intention-to-treat analysis,
there was no difference in myocardial ischemia, MI, or death between the strategies.
The largest trial to date comparing aggressive and conservative transfusion
strategies randomized 838 critically ill patients to a restrictive transfusion
strategy (transfusion for hemoglobin <7.0 g/dL) or a liberal strategy (transfusion
for hemoglobin <10.0 g/dL).8 In an intention-to-treat
analysis, there was no difference in 30-day all-cause mortality between the
2 groups. There also were significantly more MIs and cases of pulmonary edema
with the liberal transfusion strategy. When the subgroup of patients with
coronary artery disease was analyzed separately, there was no difference in
30-day mortality between the study groups.9 Further
post hoc analysis of patients with MI and unstable angina revealed a trend
toward better survival with maintenance of a higher hematocrit level, but
this finding was not statistically significant. Our study, which was much
larger, had the statistical power to determine the association between transfusion
and outcome in patients with ischemic heart disease and supports the results
observed in the randomized trial.
There are several limitations to our study. First, our study was a post
hoc analysis of prospectively collected data within the context of multiple
clinical trials. As such, transfusion was a postrandomization event and any
attempt to draw associations between postrandomization variables and outcome
has the potential for bias. Indeed, one reason transfusion was associated
with a worse outcome was that all of the bias could not be adjusted for in
the analysis. Although we repeated the analysis using several rigorous statistical
methods and found similar results, there may still be unmeasured confounders
that might account for the finding of increased mortality with transfusion.
Second, we could not explore the indications for or the appropriateness of
blood transfusion in our analyses because this information was not captured
in our database. Third, the patients in our study were all participants in
a clinical trial and therefore may not reflect the real-world population of
patients with ACS, which may include patients with different comorbidities
in whom transfusion decisions may be more complicated. Finally, because our
study was not randomized, it should not be considered as evidence to change
practice; rather, it should be considered as evidence that caution is warranted
when making transfusion decisions.
In our study, blood transfusion in the setting of ACS was associated
with an increased risk of short-term mortality. This risk persisted despite
adjustment for patient characteristics, including baseline and nadir hematocrit,
bleeding, and in-hospital procedures. Given the disparity in results between
our study and other observational studies of transfusion and outcome, a randomized
trial of transfusion strategies in anemic patients with ACS is warranted to
guide clinical practice. Until then, we caution against the routine use of
blood transfusion to maintain arbitrary hematocrit levels in stable patients
with ischemic heart disease.
Corresponding Author: Sunil V. Rao, MD,
Duke Clinical Research Institute, PO Box 17969, 2400 Pratt St, Durham, NC
Author Contributions: Dr Rao had full access
to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Rao, Harrington,Moliterno,
Pieper, Stamler, Califf.
Acquisition of data: Harrington, Granger, Newby,Moliterno,
Analysis and interpretation of data: Rao, Jollis,Harrington,
Newby, Armstrong, Moliterno,Lindblad, Pieper, Topol, Califf.
Drafting of the manuscript: Rao, Topol.
Critical revision of the manuscript for important
intellectual content: Jollis, Harrington, Granger, Newby, Armstrong,
Moliterno, Lindblad, Pieper, Topol,Stamler, Califf.
Statistical analysis: Rao, Jollis, Granger,
Obtained funding: Harrington, Armstrong, Topol,Califf.
Administrative, technical, or material support:Harrington,
Study supervision: Rao, Harrington, Granger,Moliterno,
Funding/Support: This work was supported by
the Duke Clinical Research Institute, Durham, NC.
Role of the Sponsor: The study’s sponsor
had no role in the design and conduct of the study, in the collection, analysis,
and interpretation of the data, or in the preparation, review, or approval
of the manuscript.