Evidence Based Review Article

The Journal of Informed Pharmacotherapy 2003;13:200.

EPO for Critically Ill Patients: Does it Reduce Transfusion Requirements?

Reviewer: Sean Gorman
Reviewer’s e-mail address: skgorman@interchange.ubc.ca 

Reviewer's profession/specialty: Pharmacotherapeutic Specialist -
Critical Care

Original Citation

Corwin HL, Gettinger A, Pearl RG, Fink MP, Levy MM, Shapiro MJ, et al. Efficacy of recombinant human erythropoietin in critically ill patients: a randomized controlled trial. JAMA 2002; 288:2827-35. PubMed Citation

Overall Study Question

This was a prospective, randomized, double-blind, placebo-controlled trial conducted at 65 intensive care units in the United States between December 1998 and June 2001.  The investigators aimed to determine if weekly administration of recombinant human erythropoietin (rHuEPO) to critically ill intensive care unit (ICU) patients would reduce the exposure to allogeneic red blood cells (RBC).

All patients admitted to either a medical, surgical, or a medical/surgical ICU who remained in the ICU for a minimum of 2 days were evaluated for study eligibility prior to ICU day 3.  Inclusion criteria included stay in the ICU for a minimum of 3 days; a minimum age of 18 years; hematocrit (HCT) less than 38%; and informed consent.  Patients were excluded if they had renal failure requiring dialysis; uncontrolled hypertension; new onset or uncontrolled seizures; acute burns; pregnancy or lactation; acute ischemic heart disease; acute gastrointestinal bleeding; prior treatment with rHuEPO; participation in another research protocol; or expected ICU discharge within 48 hours of ICU day 2.

Enrolled patients were randomized to receive rHuEPO 40,000 units or a placebo administered by subcutaneous injection on ICU day 3 and continued once weekly (maximum of 4 doses) for those patients remaining in the unit.  The study drug was withheld if the HCT was 38% prior to the scheduled administration.

All patients were to be followed for 28 days following randomization.  Patients also received a minimum daily dose of 150 mg elemental iron via the enteral route beginning on ICU day 3 unless they were unable to tolerate enteral feeding.  Parenteral iron was given to patients demonstrating an inadequate response to oral iron.  The following RBC transfusion guidelines were established for the purposes of the study: no RBC transfusion if hemoglobin (Hgb) > 90 g/L or Hct > 27% unless there was a specific clinical indication (i.e. active bleeding, ischemia, or other); RBC transfusion for Hgb < 90 g/L or Hct < 27% was left to the physician's discretion.  Thus, the ultimate decision for RBC transfusion was made by the intensive care physician.

The primary outcome was transfusion independence.  This was assessed by comparing the percentage of patients in each treatment group who received any RBC transfusion between study days 1 and 28.  Prospectively identified secondary outcomes were the cumulative RBC units transfused per patient through study day 28; the cumulative mortality through study day 28; the change in Hgb from baseline; and the time to first transfusion or death.

What was the declared relationship between authors and sponsors for this study?

This study was designed by the principal investigators (Drs. H. Corwin and Gettinger) with input from the data coordinating center (CareStat Inc. in collaboration with the Boston University School of Public Health) and was reviewed by the study sponsor, Ortho Biotech Products LP, manufacturer and study supplier of rHuEPO. Drs. H. Corwin, Gettinger, and Pearl have received honoraria from Ortho Biotech. Drs. H. Corwin and Gettinger have received research funding from Ortho Biotech. Drs. H Corwin, Gettinger, Fink, Levy, Shapiro, and Pearl have been paid consultants to Ortho Biotech. Drs. Colton and M. Corwin are partners in CareStat Inc.

Are the Results of the Study Valid?

1. Was assignment of patients randomized?

Yes.  Randomization was stratified by site and entailed the use of computer-generated random numbers.

2. Were all patients who entered the trial properly accounted for and attributed at its conclusion?

Yes. During a 30-month period, 33,685 patients were screened, and 1,302 (3.9%) were enrolled.  A study flow chart was provided that accounted for all patients screened and enrolled in the trial.  Of the 1,302 patients randomized, 31 (2.4%) discontinued treatment (2.9% rHuEPO vs. 1.8% placebo) and 36 (2.8%) were lost to follow-up after hospital discharge (2.6% rHuEPO vs. 2.9% placebo).  Analysis was conducted as per the intention-to-treat principle. 

3. Were patients, their clinicians, and study personnel 'blind' to treatment?

Yes.  This study was double-blinded and randomization was performed at the data coordinating center.  However, it is not known whether clinician blinding was maintained in all cases.  The authors state that the syringes were prepared in either the hospital’s pharmacy or in the patient care area.   No information was provided as to who actually prepared the syringes in these areas.  The randomization schedule appears to have been concealed since randomization was performed by the data coordinating center.
The mean pre-transfusion Hgb for the first and each subsequent RBC transfusion was also similar between study groups, which suggests that physician blinding was maintained.

4. Were the groups similar at the start of the trial?

Overall, there were no major differences in baseline demographic or medical history characteristics.  Baseline Hgb was not different between the groups, however, there were differences between groups with regards to baseline iron concentration and % transferrin saturation.  Patients in the rHuEPO group had higher baseline iron concentrations (6.48 umol/L vs. 5.03 umol/L) and % transferrin saturation (20.3% vs. 17.2%).  Patients randomized to rHuEPO also appeared to have higher baseline ferritin concentrations than placebo recipients (1536 pmol/L vs.  1262 pmol/L).  These differences have the potential to introduce noise into the interpretation of the results and could skew the results in favor of rHuEPO.  The authors state that these differences are not clinically meaningful and multivariate analysis took these variables into account. 
There was also a higher proportion of postoperative and primary hematological disease as an admitting diagnosis patients in the rHuEPO group, although this would not be expected to have a significant effect on the study outcome.  The investigators performed a logistic regression analysis to adjust the calculated odds ratio (OR) for RBC transfusion to account for differences in baseline characteristics between groups.

5. Aside from the experimental intervention, were the groups treated equally?

Yes.  Although there was no description of phlebotomy practices, it appears as though transfusion practices were similar between groups.  The mean Hgb immediately prior to a RBC transfusion was 85 g/L for both groups.  There were identical proportions of patients in each group (21%) who received a RBC transfusion at Hgb > 90 g/L or Hct > 27%.  It is not known whether there were differences in proportions of patients who received iron.

What were the Results?

1. How large was the treatment effect? 

Three hundred and twenty-eight of the 650 (50.5%) patients in the rHuEPO group received at least one RBC transfusion during the 28-day follow up period, as compared to 394/652 (60.4%) patients in the placebo group [p < 0.001;ARR = 9.9% (95%CI 4.5% - 15.3% (my calculations)); OR, 0.67 (95%CI 0.54 - 0.83), NNT 11 (95% CI 7 - 22) (my calculations)).  For this analysis, patients who had not received a transfusion by the time they were withdrawn from the study or lost to follow-up were assumed to have received no RBC transfusions.

There was only a slight difference in the proportion of withdrawals/patients lost to follow-up between groups (5.5% for rHuEPO vs. 4.7% for placebo).  An additional analysis was performed whereby patients who were withdrawn from the study or lost to follow-up were assumed to have received RBC transfusions.  Under these conditions, the investigators reported that 53.4% of the rHuEPO group were transfused vs. 63.3% of the placebo group [no p value reported; ARR = 9.9%; OR, 0.66 (0.53-0.83), with an NNT of 10 (7 - 23) (my calculations)].  To account for differences in baseline characteristics between groups, the investigators performed a logistic regression analysis to adjust the OR for RBC transfusion. This adjustment revealed similar results [Adjusted OR 0.65; (0.51-0.83)].

The mean number of units of RBC transfused per patient was 2.4 (SD 4.79) for the rHuEPO group vs. 3.0 (SD 5.42) for the placebo group (no p value reported).  The median number of units of RBC transfused per patient was 1 (IQR 0-3) for the rHuEPO group versus 2 (IQR 0-4) for the placebo group (p < 0.001).  There did not appear to be a meaningful difference in total number of days alive between the two groups (16,247 for rHuEPO vs. 16,235 for placebo). 

The investigators report the results of an analysis accounting for time at risk for transfusion.  However, this analysis was performed post-hoc and therefore should only be considered hypothesis-generating.  The transfusion rate/days alive for the rHuEPO group was 0.098 (SE 0.0074) vs. 0.121 (SE 0.0085). The relative reduction in transfusion rate/days alive was 19%. 

The absolute reduction in transfusion rate/days alive was 0.02 (95% CI 0.04 - 0.001); the clinical significance of this reduction is unknown.  The effect estimate for the ratio of transfusion rates (rHuEPO/placebo) was 0.81 (95% CI 0.79-0.83); p = 0.04).  There was no statistically significant difference between groups for median ICU-free days (18 days rHuEPO group vs. 17 days placebo group; p = 0.25).  Also, there was no difference between groups in terms of median ventilator-free days (22 days rHuEPO vs. 20 days placebo; p = 0.27).  However, this study was not powered to show a small difference in length of stay outside of the ICU or mechanical ventilation. 

The mean increase in Hgb from baseline to final determination was significantly greater for patients who received rHuEPO (13.2 g/L (SD 20) versus 9.4 g/L (SD 19) for placebo; p<0.001).  The clinical significance of this small difference is likely negligible.

A Kaplan-Meier plot was constructed to determine whether a difference existed between groups with regards to the composite endpoint of time to first transfusion or death.  A difference did exist between the 2 treatment groups in favor of rHuEPO commencing near the end of the first week following randomization (p = 0.001).

There was no difference in 28-day mortality between the 2 groups (14% in rHuEPO vs. 15% in placebo; p = 0.61).  Although not stated in the methods as an endpoint, there did not appear to be a significant difference in reported serious adverse events between the groups.

Subgroup analyses for proportions of patients undergoing transfusion and the incidence of mortality were performed based on admitting diagnosis, age, and APACHE II score (ICU severity of illness scale).  These subgroups were selected for analysis post-hoc and therefore should be considered hypothesis generating only.  The reduction in the percentage of rHuEPO patients undergoing transfusion was consistent across all subgroups analyzed.  However, overall mortality varied widely by subgroup, with an increase in mortality with advanced age and higher APACHE II scores (indicating more severe illness).  Multivariate analysis demonstrated no treatment or treatment-by-baseline variable interaction with mortality.

2. How precise was the estimate of the treatment effect?

For the primary endpoint of proportion of patients who received any RBC transfusion during the 28-day follow-up, the ARR was 9.9% and the 95% CI was 4.5% to 15.3% (my calculations).  The OR was 0.67 (95%CI 0.54 - 0.83) and the NNT was 11 (95% CI 7 - 22) (my calculations)).

Will the Results Help Me in Caring for My Patients?

1. Can the results be applied to my patient care?

The sample of patients included in this study appears to be representative of the general ICU population. The APACHE II scores were approximately 20, which is similar to the scores in a previous landmark study examining RBC transfusion strategies. (1)  However, 71% of screened patients were ineligible for enrollment into this study. The major reason for ineligibility was an expected ICU stay of < 48 hours.  Other excluded populations of note were patients with acute ischemic heart disease, acute gastrointestinal bleeds, and burns.  The baseline Hgb in each group in this study was approximately 100 g/L, which is higher than patients included in a previous study examining anemia in the ICU (82 g/L). (1)  The higher baseline Hgb may reflect a relatively healthier population in this study.

The mean pre-transfusion Hgb (~85 g/L) was similar between the groups in this study.  The clinicians in this study used a higher Hgb value for initiating RBC transfusions than that suggested in the previous TRICC trial (70 g/L). (1)  This landmark trial was published while the current study was underway and this may partially explain the differences in transfusion practices.   Assuming that clinicians have adopted a more restrictive (i.e. lower Hgb value) transfusion threshold since the publication of the TRICC study, the more liberal (i.e. higher Hgb value) transfusion threshold seen in this study may weaken its external validity.  However, this is unlikely since this threshold is within the highly variable range observed in practice in the US and Western Europe.(2,3)  This study also revealed that 21% of all patients underwent transfusion at Hgb > 90 g/L or Hct > 27%.  Transfusion thresholds can widely fluctuate among ICU clinicians and there remains considerable debate surrounding the optimal transfusion threshold. (4)

There was a statistically significant difference between groups in terms of both the primary (i.e. receipt of any transfusion) and secondary (i.e. cumulative RBC transfusions) clinical endpoints.  In my opinion, the secondary endpoint is more clinically meaningful.  That said, the differences between groups do not appear to be clinically meaningful and I would not apply the results of this study to my patients.

2. Were all clinically important outcomes considered?

Yes.  Exposure to transfused RBC units was documented.  Mortality rate was also captured, although it is unclear how exposure to RBC transfusion and rHuEPO would affect this outcome.  Any change in Hgb from baseline was also captured and shown to be significantly greater for patients receiving rHuEPO, however the clinical significance of this surrogate endpoint is debatable.  All patients were monitored for adverse drug reactions.

3. Are the likely treatment benefits worth the potential harms and costs?

This study reveals that for every 11 critically ill patients with an average baseline Hgb of 100 g/L treated with 40,000 units of rHuEPO sc every week for 3-4 weeks, one patient will avoid exposure to RBC transfusions.  It also shows that patients who are treated with the aforementioned regimen require 0.6 fewer units of RBC compared to those not treated with rHuEPO.  These reported benefits are of marginal clinical significance.  This study revealed no major differences in adverse events between rHuEPO and placebo.  The cost of three doses of 40,000 units rHuEPO is approximately $1600 (CAN).  A rough estimate of the cost of 1 unit of RBC is approximately $270 (CAN).  Thus, there is a significant difference in direct cost associated with these therapies. 
Another important factor determining the relative utility of rHuEPO in the ICU is the clinical significance of avoiding RBC transfusions.  Approximately 0.5% to 3% of all RBC transfusions are associated with adverse events, but the majority of them are minor reactions with no significant consequences. (5)  In light of the small risk associated with a blood transfusion and the questionable clinical significance of an absolute reduction of 10% in having any RBC transfusion after receiving the costly rHuEPO, this therapy should not be adopted for routine use in the ICU.  This drug should be reserved for patients who are unable to receive RBC transfusions for any reason.


Anemia in the critically ill is a common problem that results from a multitude of factors including numerous blood draws, occult gastrointestinal bleeding, and blood loss caused by surgical techniques. (6)  The remedy of choice for correcting anemia is RBC transfusion.  However, the actual benefits of increasing Hgb through RBC transfusion have not been clearly elucidated and this therapy is not devoid of adverse events. A potential option for treatment of anemia that may spare RBC transfusion in the ICU population is rHuEPO.  Endogenous erythropoietin triggers RBC production by bone marrow stimulation, however secretion has been shown to be reduced in patients with inflammatory disorders and sepsis. (6)  A previous well-designed study examined the role of prophylactic rHuEPO on the number of RBC transfusions required in a medical-surgical ICU population. (7) The dosing regimen of rHuEPO was 300 U/kg/day for 5 days followed by alternate day dosing for 2-6 weeks.  This previous study showed a reduction in number of units of RBC transfusions required, however mortality, length of stay, and adverse effects did not differ between groups.

The transfusion threshold is an arbitrarily determined value below which the deleterious effects of reduced oxygen-carrying capacity are expected.  The investigators in this study utilized a transfusion threshold of 85 g/L, which may be a higher Hgb level than would be expected based on the results of the TRICC trial. (1)  Also, 21% of patients in each group underwent transfusion at a Hgb > 90 g/L or Hct > 27%, which appears high. This study may have been able to show a similar reduction in RBC transfusions simply by mandating a lower transfusion threshold of 70 g/L.

Approximately 0.5% to 3% of all transfusions result in some adverse events, but the majority of them are minor reactions with no significant consequences. (5) The minor adverse effects associated with RBC transfusions include allergic reactions, febrile nonhemolytic reactions, red blood cell alloimmunization, and leukocyte/platelet alloimmunization. These events are of little clinical consequence in the majority of patients. (8) Rarely, more severe reactions can occur, such as acute hemolytic reactions, delayed hemolytic reactions, transfusion-related acute lung injury, and transmission of infectious agents. (8) Canadian Blood Services performs donor screening and highly sensitive laboratory screening tests on all blood for syphilis, hepatitis B and C, HIV 1 and 2, and Human T-Cell lymphotropic virus I and II. (9)  For example, the risk of transmission of HIV and HCV in Canada through blood transfusion between 1987-1996 was estimated to be 1 in 752,000 and 1 in 225,000 donations, respectively. (10)  The risk of becoming infected with West Nile Virus, Creutzfeldt-Jakob Disease, and Variant Creutzfeldt-Jakob Disease also appears to be very low. (9)  The incidence of transfusion-transmitted bacterial reaction is estimated to be 1 in 500,000 units of red cells. (5)  Therefore the risk of becoming infected with any infectious agent is exceedingly low.  This study revealed no difference between groups in the incidence of sepsis, pneumonia, or abscess development.

Since there is evidence suggesting an extremely low risk for experiencing clinically meaningful transfusion-related adverse events, the usefulness of transfusion-independence as a primary endpoint is questionable.  Is it practical to aim for complete RBC transfusion avoidance in critically ill patients who reside in an ICU for greater than 2 days?  What may be of greater importance is the total exposure to RBC transfusions, as the risk of experiencing an adverse event will increase with repeated exposure.  While patients receiving rHuEPO did have a statistically significant lower median number of transfusions, it did not appear to differ from a clinical perspective. 

The significant cost of rHuEPO, its marginal effect on transfusion avoidance, and its lack of effect on other clinically meaningful outcomes should preclude routine use in the ICU.  There are no data to support using rHuEPO to spare RBC transfusions in times of blood supply shortages.

Reviewer Competing Interests

None declared.


  1. Hebert PC, Wells G, Blajchman MA, et al. A multicenter, randomized, controlled clinical trial of Transfusion Requirements in Critical Care (TRICC). N Engl J Med 1999;340:409-17.
  2. Corwin HL, Abraham E, Fink MP, et al. Anemia and blood transfusion in the critically ill: current clinical practice in the US - The CRIT Study [abstract]. Crit Care Med 2001;29(suppl):A2.
  3. Vincent JL, Baron JF, Gattinoni L, et al. Anemia and blood transfusion in critically ill patients. JAMA 2002;288:1499-1507.
  4. Hebert PC, Wells G, Martin C, et al. Variation in red cell transfusion practice in the intensive care unit: a multicenter cohort study. Crit Care 1999;3:57-63.
  5. Health Canada. Health Care Acquired Infections Division, Transfusion-Transmitted Injuries Section. Available from URL: http://www.hc-sc.gc.ca/pphb-dgspsp/hcai-iamss/tti-it/risks_e.html. Accessed January 10, 2003.
  6. Darveau M, Notebaert E, Denault AY, et al. Recombinant human erythropoietin use in intensive care. Ann Pharmacother 2002;36:1068-74.
  7. Corwin H, 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;27:2346-50.
  8. Carson J. Should patients in intensive care units receive erythropoietin? JAMA 2002;288:2884-2886.
  9. Canadian Blood Services, Safety and Testing. Available from URL: http://www.bloodservices.ca/centreapps/internet/uw_v502_mainengine.nsf/web/DCA8A4C099A9527785256AB100561392?OpenDocument.  Accessed January 10, 2003.
  10. Chiavetta JA, Maki E, Gula CA, and Newman A. Estimated risk of transfusion-transmitted infection in the Canadian blood supply (1987-1996). Vox Sang 2000; 78:P360 (abstract).

Copyright © 2003 by the Journal of Informed Pharmacotherapy. All rights reserved.