Bone marrow suppression and anemia are frequent side effects of treatment of the acquired immunodeficiency syndrome (AIDS) with zidovudine (formerly azidothymidine [AZT]). We conducted a randomized, double-blind, placebo-controlled clinical trial of recombinant human erythropoietin (100 U per kilogram of body weight thrice weekly by intravenous bolus) in 63 patients with AIDS treated with zidovudine (29 in the erythropoietin group and 34 in the placebo group). Reductions in the number of units of red cells transfused and the number of patients given transfusions per month became apparent in the second and third months of the trial. The reductions were observed in patients with endogenous erythropoietin levels less than or equal to 500 IU per liter at base line, but not in patients whose levels were greater than 500 IU per liter at the beginning of the study. Although the hematocrit and hemoglobin level were not used as the primary criteria of efficacy because the patients received transfusions when their physicians decided that they needed them, a significantly higher rate of increase in the hematocrit was observed in the patients treated with recombinant human erythropoietin whose levels of endogenous erythropoietin were less than or equal to 500 IU per liter (0.00353 points per week) than in the patients given placebo (0.00116 points per week). This effect was not seen in patients with higher levels of endogenous erythropoietin. Serious side effects did not occur more often in the group treated with erythropoietin than in the placebo group. We conclude that recombinant human erythropoietin may be useful in patients with AIDS treated with zidovudine, although the indicators for its use remain to be clarified.