The increased incidence of vascular access thrombosis (VAT) associated with recombinant human erythropoietin (r-HuEPO) therapy is multifactorial and controversial. Sixty-three hemodialysis patients who received ≥ 12 weeks of r-HuEPO therapy were prospectively followed for the incidence of VAT. Those who experienced VAT (Group 1) were compared with those who did not (Group 2). The patients initially received r-HuEPO 50 U/kg intravenously three times a week. The dose was adjusted for a target hematocrit (t-Hct) of 30–33%. Twenty-five of the 63 patients (40%) experienced VAT (Group 1). These patients were older (mean, 54.0 ± 14.0 years versus 47.3 ± 15.0 years, p = 0.08). There was no difference between Groups 1 and 2 with respect to the baseline Hct level (21.5 ± 2.9% versus 21.4 ± 4.3%), the number of patients who achieved t-Hct (20 versus 28), and the mean time to reach t-Hct (18.6 ±15.1 weeks versus 16.9 ± 16.2 weeks). However, 20 of 25 Group 1 patients (80%) were diabetic compared with only 18 of 38 Group 2 patients (47%, p = 0.0169, by Fisher's exact test). In addition, the types of vascular access differed markedly between the two groups: arteriovenous (AV) grafts/AV fistulae/Permcaths, Group 1: 21/3/1 versus Group 2: 15/21/2, p = 0.0018. It was concluded that the occurrence of VAT in r-HuEPO treated patients was not related to the patient's hematologic response to the drug, but rather, it depended upon the integrity of the patient's vasculature and the type of vascular access used.