Seizures, hypertensive encephalopathy, transient ischemic attacks, and thrombosis of hemodialysis accesses occurred in early clinical trials with recombinant human erythropoietin. To determine if these events may be caused by the increased hematocrit value or some direct effect of the recombinant human hormone, 10 transfusion-dependent hemodialysis patients were divided into two groups of five according to their serum ferritin concentration: group A, < 800 μg/liter, and group B, > 800 μg/liter. After a month of placebo administration, recombinant human erythropoietin was given (150 U/kg intravenously thrice weekly) for four months and then stopped for one month. Hematocrit values were maintained at 0.33 ± 0.02 (mean ± SD) by dose adjustment in group A and at 0.26 ± 0.02 by thrice weekly phlebotomies in group B, who received a constant dose of erythropoietin. Viscosity increased from subnormal to normal in group A (P < 0.02) and cerebral blood flow decreased from above normal to normal (P < 0.02). In group B minor, statistically insignificant, changes in viscosity and reciprocal changes in cerebral blood flow also occurred. There was no change in either group in transcutanous oxygen tension. Bleeding time decreased toward normal in both groups during recombinant human erythropoietin administration but the changes did not reach statistical significance. Fibrinogen levels were increased in all patients but remained unchanged. No other significant coagulation-related changes were observed. Recombinant erythropoietin in the dosage and schedule of administration described in this study did not lead directly or indirectly to changes likely to precipitate seizures or intravascular thrombosis.