Should Patients with Acute Coronary Syndromes Receive Blood Transfusions?

Abstract & Commentary

Synopsis: Post hoc analysis of 3 large multicenter trials of patients with acute coronary syndromes demonstrates a strong association between transfusion and mortality.

Source: Rao SV, et al. JAMA. 2004;292(13):1555-1562.

Rao and associates aimed to determine the association between blood transfusion and mortality in patients with acute coronary syndromes (ACS), with attention to presence of bleeding and/or anemia. The analysis was based on data from 3 previously published large multicenter randomized controlled trials of ACS treatment. Patients with complete data on bleeding occurrence and transfusion were included. Only episodes of moderate or severe bleeding were considered in this analysis. Usual demographics, history and laboratory data were available. The primary outcome was 30-day all-cause mortality. The statistical approach was sophisticated and thorough. Five models were created: 2 stepwise logistic regression models that identified predictors of bleeding or transfusion and derived propensity scores for these outcomes; Cox proportional hazards model with transfusion as a time-dependent covariate; landmark analysis, using transfusion as a time-fixed covariate to analyze 24-hour units for the first 7 days; multivariate logistic regression model examining interaction between nadir hematocrit (Hct) and the association of death and transfusion.

Data from 24,112 hospitalized patients were analyzed: 2401 (10%) of these patients received at least 1 transfusion. There were several differences between patients who received blood and those who did not: for instance, the former were more likely to be older (median age, 69 vs 64 years), female, black and have comorbidities. There was a significantly greater risk of death at 30 days in patients receiving transfusion. This association persisted after adjusting for bleeding and transfusion propensity, baseline characteristics and nadir Hct: hazard ratio for death in patients receiving transfusion was 3.94 (95% CI, 3.26-4.75). Similarly, landmark analysis adjusting for baseline characteristics and nadir Hct supported a trend towards increased mortality in patients receiving transfusion but this was not statistically significant. Finally, there was no association between death and transfusion in patients with nadir Hct 25% or less but a significantly increased risk at nadir Hct > 25%. Analyses were repeated after excluding patients who underwent coronary artery bypass grafting (CABG) and results were unchanged.

Rao et al conclude that blood transfusion in patients with ACS is associated with increased 30-day mortality. They suggest that "a hematocrit as low as 25% may be tolerated without blood transfusion in otherwise stable patients with ischemic heart disease." They make note of conflicting observational data in the literature and the limitations of their post hoc analysis, summarizing that "our study . . . should not be considered as evidence to change practice; rather, it should be considered as evidence that caution is warranted when making transfusion decisions."

Comment by Saadia R. Akhtar, MD, MSc

Rao and colleagues have produced a thoughtful analysis that further challenges our usual transfusion practices. Traditional teaching has been that blood transfusion to maintain Hct at least 30% or hemoglobin (Hgb) 10g/dL or more may improve oxygen delivery and thus be beneficial for critically ill patients. More recently, risks of transfusion and safety of lower transfusion thresholds in general ICU populations have been well-documented.1,2 It is less clear whether these lower thresholds should apply to patients with ACS. Two large studies have previously tried to address this issue.3,4

The first was a post-hoc subset analysis of the Transfusion Requirements in Critical Care Trial (TRICC). TRICC was the randomized controlled study demonstrating safety of restrictive (threshold Hgb, 7 g/dL) transfusion strategies.1 The subset consisted of 357 patients with cardiovascular disease as a primary or secondary diagnosis or as a functionally limiting comorbidity. Though 70% of these patients had a history of coronary artery disease, only about 25% had a primary admitting diagnosis of ischemic heart disease or cardiac arrest. Others’ diagnoses ranged from pulmonary embolism to hypertensive urgency. The study found no difference in 30-day mortality of patients with cardiovascular disease and restrictive vs liberal transfusion strategies.3

In contrast, the work of Wu and colleagues4 suggested lower short-term mortality with blood transfusion in specific populations. Their retrospective Medicare database review examined the association between transfusion and mortality in patients at least 65 years age with acute myocardial infarction (AMI). 78,974 patients were identified and categorized by admission Hct. Patients with bleeding complications and those undergoing CABG were excluded. 3680 patients (4.7%) received transfusions. Lower admission Hct was associated with higher 30-day mortality. Furthermore, transfusion in patients with admission Hct 33% or less was associated with reduced 30-day mortality: adjusted odds ratio for death 0.69 (95% CI, 0.53-0.89) in patients with admission Hct 30.1 to 33%.4

Differences in patient populations (age and primary diagnoses), methods for patient identification (randomized trials with a different primary outcome variable vs database screening), exclusion criteria (bleeding complications) and statistical analyses (factors included in multivariate models and approach to addressing survivor bias) are some of the reasons for the dissimilar findings between the 2 prior studies and that of Rao et al. Most of all, it is essential to remember that all 3 reports are retrospective, secondary analyses. They are quite valuable for hypothesis-generation but do not in themselves provide maximally robust evidence to guide practice. Recent medical history provides several examples (such as hormone replacement therapy for post-menopausal women) of post hoc analyses and observational studies having "led us astray" until well-designed, adequately-powered, randomized prospective trials provided more clear evidence. It is apparent that such a trial is justified to answer the question of whether and when to transfuse patients with ACS.

For now, we must implement the practice that is supported by good evidence: that of transfusion threshold as defined by TRICC (Hgb at least 7 g/dL) for general medical-surgical critically ill patients. I strongly suggest considering Rao et al’s recommendation, to lower transfusion threshold (perhaps to Hct 25%) in otherwise stable patients with ACS. In reality, each of us must use the information at hand in combination with our best clinical experience and judgment to define this threshold until further data are available.


1. Hebert PC, et al. N Engl J Med. 1999:340:409-417.

2. Vincent JL, et al. JAMA. 2002:288:1499-1507.

3. Hebert PC, et al. Crit Care Med. 2001:29(2):227-234.

4. Wu WC, et al. N Engl J Med. 2001:345(17):1230-1236.

Saadia R. Akhtar, MD, MSc, Pulmonary and Critical Care Medicine, Yale University, School of Medicine, is Associate Editor for Critical Care Alert.