Older age is a poor prognosis factor in acute myeloid leukemia (AML). This double-blind trial was designed to test the hypothesis that granulocyte colony-stimulating factor (G-CSF) used as supportive care could improve the treatment of elderly AML patients. Two hundred thirty-four patients 55 or more years of age with a morphologic diagnosis of de novo or secondary AML, French-American-British (FAB) M0-M7, excluding M3, were randomly assigned to a standard induction regimen (daunorubicin at 45 mg/m2 intravenously [IV] on days 1 through 3 and Ara-C at 200 mg/m2 IV continuous infusion on days 1 through 7) plus either placebo or G-CSF (400 microg/m2 IV over 30 minutes once daily). Results are reported here for 211 centrally confirmed cases of non-M3 AML. The two groups were well balanced in demographic, clinical, and hematological parameters, with median ages of 68 years in the G-CSF and 67 years in the placebo groups. The complete response (CR) rate was not significantly better in the G-CSF group: 50% in the placebo and 41% in the G-CSF group (one-tailed P = .89). Median overall survival was also similar, 9 months (95% confidence interval [CI], 7 to 10 months) in the placebo and 6 months (95% CI, 3 to 8 months) in the G-CSF arms (P = .71). We found a significant 15% reduction in the time to neutrophil recovery in the G-CSF group (P = .014). G-CSF had no impact on recovery from thrombocytopenia (P = .80) or duration of first hospitalization (P = .27). When infection complications were evaluated, G-CSF had a beneficial effect on the duration but not on incidence of infection. G-CSF patients had fewer days with fever and shorter duration of antibiotic use. However, there was no difference in the frequency of total documented infections or in the number of fatal infections (19% placebo v 20% G-CSF). In this study of elderly AML patients, G-CSF improved clinical parameters of duration of neutropenia and antibiotic use, but did not change CR rate or survival or shorten hospitalization.