The safety and possible efficacy of recombinant human granulocyte-macrophage colony-stimulating factor (rhGM-CSF) were evaluated in 40 consecutive patients who received transplants from unrelated donors. rhGM-CSF was administered by 2-hour daily intravenous infusion from day 0 to day 20 or day 27 after the marrow infusion. These patients were compared with 78 historical patients who received transplants from unrelated donors who did not receive rhGM-CSF. The rhGM-CSF-treated patients were older (P = .037) and were treated less frequently in laminar air flow rooms (P = .005) than were control patients. However, the rhGM-CSF-treated group had a higher proportion of "good risk" patients with chronic myelogenous leukemia in chronic phase (P = .006) than did the comparison group (P = .017), rendering comparisons of transplant-related complications not meaningful. rhGM-CSF was well tolerated and did not adversely increase the incidence of graft rejection or increase the incidence and severity of acute graft-versus-host disease. The median day the absolute neutrophil count reached 500/mm3 in patients who received rhGM-CSF was day 21, which was not different from that of historical patients. Nevertheless, the numbers of febrile days and septicemic episodes within the first 28 days in patients who received rhGM-CSF were less than in historical patients. The probability of nonrelapse mortality at 1 year in patients who received rhGM-CSF was 22%. In view of the retrospective nature of the control group, we cannot conclusively determine whether rhGM-CSF administration was beneficial. A prospective, randomized controlled study of rhGM-CSF is required to confirm these suggestive data.