To assess whether erythropoietin (EPO) treatment is safe and reduces the need for transfusion, we randomized 44 preterm infants to an EPO group and a comparable control (CON) group. EPO 150 U/kg was given s.c. twice weekly for 6 wk from the 1st wk of life. Hematologic parameters, transfusion requirements, and growth were followed during therapy and for 6 mo thereafter. To better assess in which neonates EPO treatment was effective, we classified retrospectively the EPO and CON groups into uncomplicated neonates (EPO A: n = 9, birth weight = 1247 ± 126 g, gestational age = 29.8 ± 1.5 wk; CON A: n = 7, birth weight = 1217 ± 145 g, gestational age = 29.9 ± 1.5 wk) and neonates requiring artificial ventilation (EPO B: n = 16, birth weight = 1169 ± 249 g, gestational age = 28.1 ± 2 wk; CON B: n = 12, birth weight = 1173 ± 215 g, gestational age = 28.3 ± 2 wk). There were significant differences in reticulocytes between both uncomplicated and ventilated neonates in the EPO group compared with respective control groups. However, the need for transfusion was significantly less in the uncomplicated EPO group (EPO A: 0.44 ± 0.73 versus CON A: 1.28 ± 0.75, p< 0.05) but not in the neonates on ventilation (EPO B: 8.25 ± 5 versus CON B: 7.75 ± 3.7). In conclusion, early EPO administration reduces the need for transfusion in uncomplicated premature neonates. Although stimulation of erythropoiesis was apparent in both uncomplicated and complicated neonates, the end-result of increased need for transfusion in complicated neonates was related to altered indication of transfusion. These infants probably require further or longer EPO administration after weaning from ventilation and improvement of clinical condition.