While recombinant human erythropoietin (rHuEPO) is an effective therapy for anemia in renal failure, most published studies concern benefits in relatively healthy hemodialysis patients. The present study compares intravenous and subcutaneous administration of rHuEPO in an unselected group of 128 hemodialysis patients who were randomized to receive rHuEPO in an initial dose of 150 U/kg/week in three divided doses by subcutaneous or intravenous injection. Following a 4-week placebo run-in period, patients received rHuEPO until their hemoglobin was stable between 105 and 125 g/l for 4 weeks and then followed for a further 24 weeks. Eighty-three patients completed the study, 45 in the subcutaneous and 38 in the intravenous group. There was no difference in mean hemoglobin at any stage between subcutaneous and intravenous patients. Mean rHuEPO dose at the time of stabilization was significantly lower in the subcutaneous group compared to the intravenous (205.9 ± 135.4 vs. 274.1 ± 142.4 U/kg/week; p = 0.019), mean time to hemoglobin target was 9.9 ± 4.5 weeks for the subcutaneous group and 11.9 ± 4.9 weeks for the intravenous group (p = 0.037). Time to stabilization was 14.9 ± 4.7 weeks for the subcutaneous compared to 17.3 ± 3.9 weeks for the intravenous group (p = 0.006). Diabetic patients had higher dose requirements for rHuEPO at all time points and required a longer time to reach stabilization than nondiabetics (18.6 ± 4.6 vs. 15.6 ± 4.3 weeks; p = 0.016). Quality of life estimated by a disease-specific Kidney Disease Questionnaire improved significantly during rHuEPO therapy in both groups. There was no significant change in dialysis prescription throughout the study. No significant changes were seen in the index of removal of urea and protein catabolic rate during the study in either group. Urea generation rate was significantly higher in the intravenous group at 24 weeks of follow-up. Serious adverse events were rare. Seizures occurred in 4.1 % of patients during the study. Episodes of significant hypertension occurred in 40 patients. Thrombotic events related to access and dialysis circuit occurred in 39% of subcutaneous patients and 48% of intravenous patients. Access failure occurred in 23.1 % of subcutaneous and 17.7% of intravenous patients. Access failure was more common in patients with grafts compared to arteriovenous fistulae (p = 0.002). In conclusion, rHuEPO therapy was associated with similar hematologic effects and quality of life benefits in this unselected patient population compared to more selected patients. The adverse event profile was also similar with access failure, particularly for implanted access, occurring more fequently during rHuEPO therapy. Subcutaneous administration appeared at least equivalent to intravenous administration of rHuEPO, and the lower induction dose and shorter induction period might confer some clinical advantages and redu