To evaluate the effect of recombinant human erythropoietin (rHuEpo) on the haemoglobin level and transfusion requirement in low-risk myelodysplastic syndromes (MDS), 87 patients were enrolled in a randomized double-blind placebo-controlled study, 44 patients were assigned to epoetin alpha (150 U/kg/d s.c. for 8 weeks) and 43 to placebo arms. MDS types were homogenous in both groups: refractory anaemia (RA) 47.7-48.8%. refractory anaemia with ringed sideroblasts (RAS) 20.5-25.6%, refractory anaemia with excess of blasts (RAEB) (blasts < 10%) 31.8-25.6%, 14/38 evaluable patients responded to epoetin alpha versus 4/37 to placebo (P=0.007). 50% of RA responded to epoetin alpha versus 5.9% to placebo (P=0.0072), RAS 37.5% v 18.2% (P=0.6) and RAEB 16.7% v 11.1% (P=1.00). 60% of non-pretransfused patients responded to epoetin alpha (Hb 8.35< or = 0.73 to 10.07+/-1.87 g/dl), whereas a slight decrease was observed in the placebo group (8.4+/-0.66 to 8.19+/-0.92 g/dl) (P=0.0004). Percentage of transfused patients was similar in both arms. Basal erythropoietin (Epo) serum levels > 200 mU/l predicted for a non-response. At week 4 sTfR levels were increased > 50% in responders (P=0.013), whereas an increase < 18% predicted for non-response (P=0.006). Leucocyte and platelet counts were not influenced by epoetin alpha treatment. Adverse events occurred in 31.8% of the rHuEpo-treated versus 42.99%) of the placebo-treated patients (P=0.2), and seven patients did not complete the course. In conclusion, rHuEpo was effective in the treatment of low-risk MDS. RA subtype, no transfusions prior to rHuEpo therapy, and low basal Epo levels were associated with higher probability of response. Soluble transferrin receptor level at the fourth week was an early predictor of response.