We randomly assigned eight concurrently symptom-free premature infants (birth weight less than or equal to 1250 gm) at high risk of requiring erythrocyte transfusions for anemia of prematurity to 6 weeks of intensive treatment with either subcutaneous recombinant human erythropoietin (r-HuEPO group) or a placebo (control group). Treatment with r-HuEPO was initiated at a dose of 100 units/kg per day 5 days a week, and was increased to 200 units/kg per day after 2 or 3 weeks if target reticulocyte counts were not achieved. All patients were given supplemental oral iron therapy at a dose of 6 mg/kg per day, as tolerated. Mean reticulocyte counts in r-HuEPO-treated and control infants were 64,600 versus 67,500 cells/mm3 at entry; were 245,600 versus 78,000 cells/mm3 after 1 week; and averaged 262,600 versus 136,400 cells/mm3 during the study. Mean reticulocyte counts in r-HuEPO-treated infants were 251,200 cells/mm3 during the week when r-HuEPO, 100 units/kg per day, was given, and were 269,500 cells/mm3 after the dose was increased to 200 units/kg per day. Mean hematocrit values at entry were 33.4% in babies who received r-HuEPO versus 33.6% in the control subjects, and were 31.4% in r-HuEPO-treated and 25.2% in the control subjects at the end of treatment. One r-HuEPO-treated and three control babies received transfusions during the study; the total volume of blood given was 17 ml in the r-HuEPO group and 101 ml in the control subjects. The percentage of hemoglobin F increased in infants not given transfusions. We conclude that r-HuEPO stimulates endogenous erythropoiesis in small premature babies who are receiving supplemental oral iron therapy. A controlled multicenter trial has been undertaken to confirm these promising preliminary observations.