The purpose of this study was to estimate the net cost effect to Medicare of the increasing use of recombinant human erythropoietin (EPO) instead of red blood cell transfusions or androgens in the management of anemia for the approximately 100,000 hemodialysis patients in the U.S. End-Stage Renal Disease (ESRD) program. A computerized decision model that takes into account the effectiveness and possible side effects of transfusions, androgens, and EPO and predicts 1- and 5-yr direct medical costs to Medicare associated with each therapy was constructed. Probability estimates for clinical events were derived from the literature. Costs were assigned by use of the amounts Medicare pays providers of ESRD care for: (1) use of EPO, transfusions, and androgens; and (2) health care services related to the treatment of anemia (including complications of treatment and possible reductions in morbidity). For every 10,000 hemodialysis patients treated with EPO, net Medicare expenditures will be much greater than if only transfusions are used by $42,530,000 at 1 yr (6% of ESRD program costs) and by $118,050,000 at 5 yr and also much greater than if androgens are used (by $42,700,000 at 1 yr and $118,370,000 at 5 yr). The increase in cost was highly sensitive to the dose of EPO; moderately sensitive to changes in estimated anemia response rates for EPO, frequency of EPO-induced vascular access clotting, and reduction in cardiovascular or overall morbidity; and slightly sensitive to transfusion rates, estimated anemia response rates for androgens, frequency of EPO-induced seizure or hypertensive complications (stroke, myocardial infarction), frequency of transfusion-related viral infection, and frequency of androgen-induced virilization. Considering both effectiveness and side effects of alternative treatments for the anemia of ESRD, it was projected that the increasing use of EPO will markedly increase the cost to Medicare of ESRD medical care.