Should Critically Ill Patients Receive Erythropoietin?

Abstract & Commentary

Synopsis: In this large, randomized trial, critically ill patients who received weekly doses of recombinant human erythropoietin had higher hemoglobin levels and required fewer transfusions than placebo-treated patients.

Source: Corwin HL, et al. Efficacy of recombinant human erythropoietin in critically ill patients. A randomized controlled trial. JAMA. 2002;288(22):

Recombinant human erythropoietin (rHuEPO) is being heavily promoted for use in critically ill patients. This multicenter, randomized, controlled clinical trial, sponsored by the drug’s manufacturer (Ortho Biotech), sought to determine whether weekly injections of rHuEPO (40,000 U) would raise hemoglobin levels in this setting and thereby reduce erythrocyte transfusions. At 65 US medical centers, Corwin and colleagues screened 33,685 ICU patients for inclusion in the study. Of these, 9674 met eligibility criteria (age 18 or older; hematocrit < 38%; anticipated ICU stay at least 3 days; and informed consent), and 1302 were randomized to receive either rHuEPO or placebo weekly for 3 doses starting on ICU day 3. All patients also received either oral or parenteral iron. Criteria for transfusion were not rigidly enforced but Corwin et al recommended a hematocrit threshold of 27% (hemoglobin 9 g/dL) in the absence of a specific indication such as bleeding or ischemia. The primary end point was the percentage of patients in each group who received any red blood cell transfusion within 28 days. Secondary end points were cumulative units transfused per patient, 28-day mortality, change in hemoglobin from baseline, and time to first transfusion or death.

The patients in the rHuEPO and placebo groups were well matched. Of the 650 patients who received rHuEPO, 328 (50.5%) received at least 1 red blood cell transfusion during the 28-day study period, as compared to 394 of the 652 placebo-treated patients (60.4%; P < 0.001; OR, 0.67; 95% CI, 0.54-0.83). Statistical analysis demonstrated a 19% reduction in units transfused per day alive (P = 0.04). The mean increase in hemoglobin concentration from baseline to final determination was significantly greater in patients who received rHuEPO (1.32 vs 0.94 g/dL; P < 0.001). There were no differences in 28-day mortality (14% and 15% in rHuEPO and placebo groups, respectively) or other adverse events examined. The mean pretransfusion hemoglobin levels were 8.53 and 8.57 g/dL for the 2 groups, and 21% of patients in each group underwent transfusion at a hemoglobin level greater than 9 g/dL (hct > 27%).

Comment by David J. Pierson, MD

This study shows that, under the conditions of this study, when administered weekly at a dose of 40,000 U (at a cost of about $400 per dose), rHuEPO reduces by approximately 10% the number of adult general ICU patients receiving red blood cell transfusion. A number-needed-to-treat analysis cited in an accompanying editorial1 showed that, among patients staying in the ICU for at least 3 days, 10 patients would need to be treated with rHuEPO in order to prevent 1 patient from receiving a transfusion during the next 28 days. The study was not sufficiently powered to detect differences in survival, although the death rates (111 patients in the rHuEPO group compared with 120 patients in the placebo group) make it unlikely that even a much larger study would show such a difference.

As clinicians, what should we do with these findings? Should we prescribe rHuEPO for all patients who remain in the ICU for 72 hours and are likely to be there for several more days? Such a practice has been adopted by some of my colleagues and is certainly defensible on the basis of available evidence if avoidance of transfusion is considered an important enough goal to justify the cost.

Three points raised by Carson in the editorial cited above1 are worth noting, however. One is that the risks associated with red blood cell transfusion in the United States in 2003 are very small. Current estimates of infection risk are 1 case of hepatitis C per 1.935 million transfused U, and 1 case of HIV transmission per 2.135 million U.2 Although more common, such adverse effects as allergic reactions, febrile nonhemolytic transfusion reactions, and erythrocyte or leukocyte/platelet alloimmunization are not usually serious; more serious complications are both rare and more likely to be due to human error than to a problem with the blood supply.

The other points have to do with the present study and how well its results apply to our own practices. First, patients on whom the data were collected represented only 13% of those eligible: nearly two-thirds of patients meeting the entry criteria were not approached by Corwin et al for inclusion, and half of all patients or surrogates asked to participate declined to do so. This raises concerns about generalizability. I would have to take it on a certain amount of faith that the study’s results would be the same if I applied the same regimen across the board to my patients that met the listed entry criteria. And, second, Corwin et al did not control transfusion practice, and the clinicians managing the patients applied a fairly liberal (by current standards) hemoglobin threshold of about 8.5 g/dL. According to several recent studies, it appears to be safe to use a transfusion threshold of around 7 g/dL in the absence of severe cardiac disease or other specific contraindication. The transfusion reductions achieved in this study were not large, and might thus have been achievable simply by applying this lower transfusion cut-off.

The availability of rHuEPO is an important advance, which has greatly benefited some patient populations. In the ICU its role remains unclear, however, because of cost considerations, differences of opinion about the indications and hazards of red blood cell transfusion, and uncertainty about which patients might benefit most from its use.


1. Carson JL. Should patients in intensive care units receive erythropoietin? JAMA. 2002;288(22): 2884-2886.

2. Dodd RY, et al. Current prevalence and incidence of infectious disease markers and estimated window-period risk in the American Red Cross blood donor population. Transfusion. 2002;42:975-979.

Dr. Pierson is Professor of Medicine University of Washington Medical Director Respiratory Care Harborview Medical Center Seattle.