In a double-blind placebo-controlled study we showed a 3-fold decrease in blood transfusions (BTFs) given to preterm infants with anaemia of prematurity who received recombinant erythropoietin. However, only 50% of placebo recipients required a BTF. Data from the placebo group indicated that either mean daily weight gain < or = 7.5 g/day before study entry or haematocrit < or = 50% at birth was associated with BTFs (P < 0.001). We calculated that giving erythropoietin to patients in the treatment group with either of these variables prevented 24 of 28 BTFs and that it would cost R184 to prevent 1 BTF. The cost of each BTF was R187 (blood filtered to remove white cells and reduce cytomegalovirus transmission). Therefore, the costs of the two treatments were similar, but as the risk of transmitting infection is lower with erythropoietin, we recommend its use in selected preterm infants.