Delay in hematologic recovery after bone marrow transplantation (BMT) can extend and amplify the risks of infection and hemorrhage, compromise patients' survival, and increase the duration and cost of hospitalization. Because current studies suggest that granulocyte-macrophage (GM) colony-stimulating factor (CSF) may potentiate the sensitivity of hematopoietic progenitor cells to G-CSF, we performed a prospective, randomized trial comparing GM-CSF (250 micrograms/m2/d x 14 days) versus sequential GM-CSF x 7 days followed by G-CSF (5 micrograms/kg/d x 7 days) as treatment for primary or secondary graft failure after BMT. Eligibility criteria included failure to achieve a white blood cell (WBC) count > or = 100/microL by day +21 or > or = 300/microL by day +28, no absolute neutrophil count (ANC) > or = 200/microL by day +28, or secondary sustained neutropenia after initial engraftment. Forty-seven patients were enrolled: 23 received GM-CSF (10 unrelated, 8 related allogeneic, and 5 autologous), and 24 received GM-CSF followed by G-CSF (12 unrelated, 7 related allogeneic, and 5 autologous). For patients receiving GM-CSF alone, neutrophil recovery (ANC > or = 500/microL) occurred between 2 and 61 days (median, 8 days) after therapy, while those receiving GM-CSF+G-CSF recovered at a similar rate of 1 to 36 days (median, 6 days; P = .39). Recovery to red blood cell (RBC) transfusion independence was slow, occurring 6 to 250 days (median, 35 days) after enrollment with no significant difference between the two treatment groups (GM-CSF: median, 30 days; GM-CSF+G-CSF; median, 42 days; P = .24). Similarly, platelet transfusion independence was delayed until 4 to 249 days (median, 32 days) after enrollment, with no difference between the two treatment groups (GM-CSF: median, 28 days; GM-CSF+G-CSF: median, 42 days; P = .38). Recovery times were not different between patients with unrelated donors and those with related donors or autologous transplant recipients. Survival at 100 days after enrollment was superior after treatment with GM-CSF alone. Only 1 of 23 patients treated with GM-CSF died versus 7 of 24 treated with GM-CSF+G-CSF who died 16 to 84 days (median, 38 days) after enrollment, yielding Kaplan-Meier 100-day survival estimates of 96% +/- 8% for GM-CSF versus 71% +/- 18% for GM-CSF+G-CSF (P = .026). These data suggest that sequential growth factor therapy with GM-CSF followed by G-CSF offers no advantage over GM-CSF alone in accelerating trilineage hematopoiesis or preventing lethal complications in patients with poor graft function after BMT.(ABSTRACT TRUNCATED AT 400 WORDS)