Recombinant human erythropoietin (rHuEPO) has been advocated for the treatment of anaemia in patients submitted to cancer chemotherapy. We used decision analysis to compare the cost-effectiveness of rHuEPO supplemented with red blood cell (RBC) transfusions with conventional treatment with RBC transfusions alone. At baseline, we analysed the use of rHuEPO as secondary prophylaxis according to effectiveness estimates from a community-based oncology study. In order to reduce the probability of transfusions from 21.9% to 10.4%, and the number of RBC units per patient per month from 0.55 to 0.29, 150 units kg(-1) s.c. rHuEPO three times per week for 4 months resulted in an incremental cost of $189,652 per quality-adjusted life year (QALY). In patients treated with cisplatin-containing chemotherapy, rHuEPO added $190,142 per QALY. In a hypothetical scenario of a transfusion pattern that maintained the same haemoglobin level of rHuEPO-responsive patients, the marginal cost of rHuEPO was always greater than $100,000 per QALY. Results were stable with regard to variations in the probability of blood-borne infections, quality of life of responding patients and cancer-related mortality. The additional cost could be lowered to less than $100,000 per QALY by saving 4.5 RBC units over 4 months for any patient treated. In conclusion, according to current use, rHuEPO is not cost-effective in the treatment of chemotherapy-induced anaemia. More tailored utilization of the drug and better consideration of predictive response indicators may lead to an effective, blood-sparing alternative.