A prospective, randomized, double-blind placebo-controlled trial was designed to evaluate the impact of granulocyte/macrophage-colony-stimulating factor (GM-CSF) on the efficacy of sequential high-dose cytosine arabinoside/mitoxantrone chemotherapy (S-HAM) in adult patients with high-risk myelodysplastic syndromes (MDS). GM-CSF or placebo was given subcutaneously once daily at a dose of 250 microg/m2, starting 48 h prior to chemotherapy, and continued until neutrophil recovery. Owing to high toxicity and slow patient recruitement the study was closed and unblinded after 31 patients had been enrolled; 15 were randomized to receive placebo and 16 to receive GM-CSF. A total of 29 patients were evaluable for response; their median age was 57 years. Ten patients achieved a complete remission (34.5%), 9 patients had persistent MDS (31%), 10 patients died within 6 weeks after the onset of treatment (early death) (34.5%). The median remission duration was 190 days (range: 2.5-45 months). Among the 29 evaluable patients no significant differences could be found between the two study arms regarding complete remission rate [GM-CSF: 31% (5/16) versus placebo: 38% (5/13) P = 0.45], rate of persistent MDS [GM-CSF: 25% (4/16) versus 38% (5/13) P = 0.35), early death rate [44% (7/16) versus 23% (3/13) P = 0.22] and remission duration (GM-CSF: 87 days versus placebo 221 days). Duration of granulocytopenia (median: 33 days with GM-CSF) versus 35 days with placebo) and frequency of infectious episodes were not significantly influenced by GM-CSF. The small number of patients finally analyzed means that no definite conclusions about the effect of GM-CSF can be reached.