To investigate the efficacy and safety of granulocyte colony-stimulating factor (G-CSF) in patients with acute myelogenous leukaemia, a multicentre randomized study was performed. From October 1993 to September 1996, 270 patients with newly diagnosed acute myelogenous leukaemia were randomized to G-CSF or control groups after remission induction therapy. The G-CSF group received G-CSF (Filgrastim) from 48 h after the completing chemotherapy until the absolute neutrophil count exceeded 1.5 x 10(9)/l. The control group did not receive G-CSF unless severe infection occurred. There were 245 evaluable patients (120 and 125 in the G-CSF and control groups respectively). The complete remission rate was similar in the G-CSF and control groups (80.8% versus 76.8%), as was the 5-year probability of disease-free survival (34.5% versus 33.6%) and overall survival (42.7% versus 35.6%). Neutrophil recovery was significantly faster in the G-CSF group than in the control group (12 d versus 18 d, P = 0.0001). The median duration of febrile neutropenia was significantly shorter in the G-CSF group than in the control group (3 d versus 4 d, P = 0.0001). In conclusion, prophylactic administration of G-CSF after remission induction therapy for acute myelogenous leukaemia is safe and useful even in patients without infection on completing chemotherapy.