To study the safety and efficacy of administering human recombinant erythropoietin (rHuEPO) to infants with anaemia of prematurity, a combined phase I/II trial of weekly intravenous injections for 4 weeks was undertaken. We treated 16 infants with 10, 25, 50, 100 or 200 units/kg body weight in groups of two to four patients per dose level. They were all born prematurely (mean gestational age: 29 weeks; range 27-32), had a mean post-natal age of 42 days (range: 25-59) and haemoglobin concentration of 87 g/l (range: 72-94) when treatment was started. Four patients (25%) needed a transfusion during the trial, one at day 7 treated with 10 units/kg and 3 at days 15, 25, 29 with 100 units/kg. In the others, a progressive rise in mean haemoglobin values was seen in each group after 21 days of treatment, without a dose-dependent effect. A positive change in absolute reticulocyte counts with a peak after 7-14 days of therapy was observed with low (25-50 units/kg) but not with higher doses, with a significant difference at day 14 between 25 and 100 units/kg (P less than 0.01). A dose-limiting severe neutropenia (absolute neutrophil count less than 0.5 x 10(9)/l) occurred transiently in five patients, with doses greater than 25 units/kg. No infectious complication and no sign of iron deficiency were observed. Weekly low doses of rHuEPO appear safe, convenient to administer and able to induce a reticulocytic response in infants with anaemia of prematurity. A phase III placebo-controlled trial is needed to confirm these results. Neutropenia associated with rHuEPO administration in infants might be related to their stage of human ontogeny.