Therapy with human recombinant erythropoietin (EPO) has been accepted as effective for renal anemia in dialysis patients. However, studies in rats have shown that correcting anemia with EPO may affect the progression of renal dysfunction. In humans, however, the effect of EPO on residual renal function is a matter of controversy. We, therefore, investigated whether the long-term administration of EPO to predialysis patients influences residual renal function. Anemic patients at the predialysis stage with a serum creatinine (Cr) concentration ranging from 2 to 4 (average 2.9) mg/dl and a hematocrit (Ht) of less than 30% were randomly assigned to two groups which consisted of anemic patients not treated with EPO (group I, untreated anemic controls, n = 31) and anemic patients treated with EPO (group II, treated anemics, n = 42). Patients with nonsevere or moderate anemia (Ht > 30%) with a Cr ranging from 2 to 4 (average 2.6) mg/dl were also recruited as nonanemic controls (group III, untreated nonanemic controls, n = 35). Blood pressure was controlled to the same degree among the three groups by combined treatment with calcium antagonists and angiotensin-converting enzyme inhibitors. All patients were kept strictly on a low-protein (0.6 g/kg/day) and a low-salt (7 g/day) diet. The degree of control of dietary protein and blood pressure and the frequency of angiotensin-converting enzyme inhibitor administration were comparable among the three groups. The primary end point for each patient was a doubling of the baseline Cr which yielded cumulative renal survival rates which were plotted against time. Ht rose significantly from 27.0 ± 2.3 to 32.1 ± 3.2% in group II (n = 42, p < 0.001) with a rate of increase of 0.4 ± 0.06%/week. However, it declined from 27.9 ± 1.8 to 25.3 ± 1.9% in group I (n = 31, p < 0.001) and from 35.9 ± 3.5 to 32.2 ± 3.9% in group III (n = 35, p < 0.001). Cr doubled in 26 patients (84%) in group I as compared with 22 (52%) in group II and 21 (60%) in group III. The cumulative renal survival rates in groups II and III were significantly better than that in group I: p = 0.0003 (group I vs. group II) and p = 0.0024 (group I vs. group III). However, there was no difference in the renal survival rate between groups II and III (p = 0.3111). The better survival rate obtained in group II was attributable to the better survival rate for the nondiabetic patients in this group. The present study suggests that anemia, per se, is a factor in the progression of end-stage renal failure and that reversal of anemia by EPO can retard the progression of renal failure, especially in nondiabetic patiens, provided that blood pressure control, rate of increase in Ht, and dietary protein restriction are appropriate.