Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Erythropoietin for anaemia with cancer therapy


Individuals with cancer commonly have anaemia because of changes in the production of erythropoietin caused by the cancer or its treatment. When haemoglobin levels fall, the anaemia may be treated by transfusion. There can be problems with this, because blood supplies are often limited, and because increasingly we are aware of infections from transfusion. Treatment of blood for transfusion is increasingly complicated and expensive.

So can treatment with exogenous erythropoietin reduce the need for transfusion associated with cancer or cancer treatment? A new systematic review suggests that it can [1], and also demonstrates good standards of systematic review in an awfully difficult area.


Controlled trials, including randomised studies and nonrandomised studies with concurrent or historical controls were used, but the latter had to show comparability between groups. In the event only a few of these were found, and largely not used for analysis. Included studies had to enrol patients with existing anaemia because of chemotherapy or radiotherapy, or have nonanaemic patients beginning a course of therapy.

The main outcome sought was transfusion, though others, including quality of life, were examined. Higher quality randomised trials were defined as those that were both randomised and double blind, with intention to treat analysis and with fewer than 10% patient exclusions.


There were 18 randomised trials with 1,700 patients, about half in adults with haemoglobin levels less than 100 g/L. There were three trials enrolling 108 children with haemoglobin less than 100 g/L.

Fourteen adult trials reported on transfusions, and all found that fewer patients were transfused with erythropoietin. For seven of twelve studies with sufficient data for computation of odds ratios there was a statistical difference between erythropoietin and control. Without erythropoietin, and with initial haemoglobin below 100 g/L, between 36% and 69% of patients were transfused. With erythropoietin between 20 and 53% were transfused. The range of differences in individual trials was 9% to 45%. Similar results were found in the three trials enrolling children and with adults with higher initial haemoglobin levels.

Odds ratios and numbers needed to treat were given (Table 1). In twelve randomised trials the number needed to treat to prevent one transfusion was 4.4. It was 5.2 in higher quality trials and 2.6 in lower quality trials. Confidence intervals did not overlap, and this probably just achieved statistical significance.

Table 1: Erythropoietin for anaemia - effect of quality of study on efficacy measure

Type of study Number of studies Odds ratio (95%CI) NNT (95%CI)
All randomised 12 0.4 (0.3 to 0.5) 4.4 (3.6 to 6.1)
Higher quality 5 0.5 (0.3 to 0.6) 5.2 (3.8 to 8.4)
Lower quality 7 0.1 (0.06 to 0.3) 2.6 (2.1 to 3.8)
Outcome was avoidance of transfusion Subcutaneous EPO was 300-450 units/kg/week


This study is interesting because of its detail as well as its result, and because it highlights an important area in cancer and palliative therapy. There is information on number of units transfused (the range of savings is 0 to 1.3 units a month for all adults, and 1.1 to 2.2 units a month for children), and on some quality of life outcomes, though there was little information on the latter.

Though methodologically sound, the systematic review falls down on one thing: it fails to give us details plus results of the individual studies. This means that if we have other ideas about how data may be analysed, we have to get all the papers and start again. That's not always a bad thing, but irritates because we cannot do a quick check of the results.

It is another example of the importance of minimising bias. For instance, in studies that were not blinded, a knowledge of treatment actually received could influence clinical decisions about the necessity for a transfusion. That may be one reason why higher quality studies had a lower estimate of efficacy than lower quality studies. Pooling information from all studies without some sensitivity analysis can never be safe, and we have to wonder why we still have to bang on about it. In any event, here the lesson is learned again.

Erythropoietin does not come cheap, though. The estimated US cost per chemotherapy cycle is given as US$3,700 to $6,600. The evidence on efficacy needs to be balanced against this. There is some interesting work for someone on cost-effectiveness, and on trials that describe more fully beneficial effects on quality of life.


  1. J Seidenfeld et al. Epoetin treatment of anemia associated with cancer therapy: a systematic review and meta-analysis of controlled clinical trials. Journal of the National Cancer Institute 2001 93: 1204-1214.

Additional comment from the authors of the review

You are correct that our published review did not include the detailed results abstracted from the individual studies. This paper summarized one section of a more extensive systematic review and evidence report on epoetin in oncology patients. The full report was published by the U.S. Agency for Healthcare Research and Quality (AHRQ) before the JNCI paper was published. As cited in the JNCI paper, detailed evidence tables are included in the full AHRQ evidence report. Therefore, we accepted the journal reviewers' advice to condense our paper by deleting similar tables from the manuscript we first submitted.

Your readers might want to know that zipped PDF files of this and all other AHRQ evidence reports (with full evidence tables) can be downloaded free of charge from the Agency's web site ( Printed copies also are available from the Agency's Publications Clearinghouse (1-800-358-9295).

Thanks again for your interest in our work.

Jerome Seidenfeld, PhD
Associate Director, Technology Evaluation Center
Blue Cross and Blue Shield Association
225 N. Michigan Avenue
Chicago, IL 60601-7680

312.297.6044 (voice)
312.297.6320 (FAX)


previous or next story in this issue