OBJECTIVE To investigate whether recombinant erythropoietin (rhEPO) reduces the need for transfusion in extremely low birth weight (ELBW) infants (birth weight 500-999 g) and to determine the optimal time for treatment. METHODS In a blinded multicenter trial, 219 ELBW infants were randomized on day 3 to one of 3 groups: early rhEPO group (rhEPO from the first week for 9 weeks, n = 74), late rhEPO group (rhEPO from the fourth week for 6 weeks, n = 74), or control group (no rhEPO, n = 71). All infants received enteral iron (3-9 mg/kg/day) from the first week. The rhEPO beta dose was 750 IU/kg/week. Success was defined as no transfusion and hematocrit levels never below 30%. RESULTS Success rate was 13% in the early rhEPO group, 11% in the late rhEPO group, and 4% in the control group (P =.026 for early rhEPO versus control group). Median transfusion volume was 0.4 versus 0.5 versus 0.7 mL/kg/day (P =.02) and median donor exposure was 1.0 versus 1.0 versus 2.0 (P =.05) in the early rhEPO group, the late rhEPO group, and the control group, respectively. Infection risk was not increased and weight gain was not delayed with rhEPO beta. CONCLUSION Early rhEPO beta treatment effectively reduces the need for transfusion in ELBW infants.