To determine the factors that govern their response to erythropoietin (EPO), we conducted a cross-sectional study of all patients in four outpatient hemodialysis facilities in Brooklyn, NY, who had end-stage renal disease (ESRD) and human immunodeficiency virus (HIV) infection and were receiving recombinant EPO. We also compared the hematocrit and EPO requirements of these patients to those of a control group of hemodialysis patients without HIV infection. We documented known duration of HIV infection, and total CD4 count was measured once. In both groups, hematocrit was measured weekly for 5 weeks and a mean value calculated for each subject. Transferrin saturation was measured twice and a mean value calculated for each subject. Intensity of hemodialysis was assessed by measuring both percent reduction of urea and serum albumin concentration twice; mean values were calculated for each subject. Twenty-nine (88%) of 33 study subjects had acquired immunodeficiency syndrome. Mean known duration of HIV infection was 49 +/- 32.5 months (median, 48 months), and mean total CD4 count was 143 +/- 152.4 cells/mm3 (median, 72 cells/mm3). Mean hematocrit in the study subjects was 27.4% +/- 4.7% compared with 27.6% +/- 3.7% in the controls (P = 0.69). Mean thrice-weekly EPO dose was higher in the study subjects (90 +/- 52 U/kg body weight) than in the controls (62 +/- 36 U/Kg body weight) (P = 0.001). Among the study subjects, hematocrit had direct univariate correlations with serum albumin concentration (r = 0.43; P = 0.02), transferrin saturation (r = 0.4; P = 0.03), and percent reduction of urea (r = 0.4; P = 0.02), but not with total CD4 count (r = -0.05; P = 0.8) or known duration of HIV infection (r = -0.11; P = 0.55). There was an inverse correlation between hematocrit and dose of EPO (r = -0.5; P = 0.003). Multiple regression analysis showed that transferrin saturation (P = 0.01) and percent reduction of urea (P = 0.003) had direct correlations with hematocrit after adjustment for other factors. There was an inverse relationship between hematocrit and dose of EPO (P = 0.0006). We conclude that in patients with ESRD and HIV infection receiving hemodialysis, the response to EPO (hematocrit) is modulated by the dose of EPO, quantity of hemodialysis, and transferrin saturation, but not by the severity of HIV disease. Hemodialysis patients infected with HIV receive a higher dose of EPO than those without HIV infection.