OBJECTIVE To assess the effect of recombinant human erythropoietin (r-HuEPO) on anemia in patients with the acquired immunodeficiency syndrome (AIDS) who are receiving zidovudine therapy. DESIGN Combined analysis of four 12-week, randomized, double-blind, controlled clinical trials. SETTING Multiple centers in the United States. PATIENTS Two hundred and ninety-seven anemic (hematocrit < 30%) patients with AIDS who were receiving zidovudine therapy. Of the 297 patients, 255 were evaluable for efficacy, but all patients were included in analysis of safety. INTERVENTION Patients were randomly assigned to receive either r-HuEPO (100 to 200 U/kg body weight) or placebo, intravenously or subcutaneously, three times per week for up to 12 weeks. MEASUREMENTS Changes in mean hematocrit, transfusion requirement, and quality of life. RESULTS Sixty-nine percent of patients had endogenous serum erythropoietin levels less than or equal to 500 IU/L, and 31% had erythropoietin levels greater than 500 IU/L. In patients with low erythropoietin levels (< or equal to 500 IU/l), r-HuEPO therapy decreased the mean number of units of blood transfused per patient when compared with placebo (3.2 units and 5.3 units, respectively; P = 0.003) and increased the mean hematocrit from the baseline level (4.6 percentage points and 0.5 percentage points, respectively; P <0.001). Overall quality of life improved in patients on r-HuEPO therapy (P = 0.13). Patients with erythropoietin levels greater than 500 IU/L showed no benefit from r-HuEPO in any outcome variable. Placebo and r-HuEPO recipients did not differ in the incidence of adverse effects or opportunistic infections. CONCLUSION Therapy with r-HuEPO can increase the mean hematocrit and decrease the mean transfusion requirement in anemic patients with AIDS who are receiving zidovudine and have endogenous low erythropoietin levels (< or equal to 500 IU/L). Such therapy is of no apparent benefit in patients whose endogenous erythropoietin levels are higher than 500 IU/L.