One hundred twenty-one anemic, transfusion-dependent patients with multiple myeloma (MM) or low-grade non-Hodgkin's lymphoma (NHL) were randomly allocated to receive (a) recombinant human erythropoietin (rhEPO) 10,000 U/d subcutaneously 7 days a week (fixed dose group) (n = 38), or (b) rhEPO 2,000 U/d subcutaneously for 8 weeks followed by step-wise escalation of the rhEPO dose (titration group) (n = 44), or (c) no rhEPO therapy (control group) (n = 39). The total treatment period was 24 weeks. There were no differences between the three groups with regard to baseline clinical, demographic, or health status measures. The cumulative response frequency, defined as elimination of the transfusion need in combination with an increase in the hemoglobin concentration by >20 g/L, was 60% in both rhEPO treatment groups and 24% in the control group (P = .01 and .02, respectively, log rank test). For patients in the titration group the response rate on the first dose level (2,000 U/d) was only 14%. Cox's univariate regression analysis revealed that an inadequately low endogenous erythropoietin concentration in relation to the degree of anemia and a baseline platelet concentration > or = 100 x 10(9)/L were significant predictors for response to rhEPO therapy (P < .01). Multivariate regression analysis showed that relative erythropoietin concentration was the most important factor and the platelet count had no additional influence on response. Treatment with rhEPO was well tolerated. We conclude that treatment with rhEPO may be indicated in anemic MM and NHL patients with a relative erythropoietin deficiency. An initial dose of 5,000 U/d subcutaneously may be recommended.