Between October 1988 and December 1992, 167 patients with leukemia receiving marrow transplants from HLA-identical donors and conditioned with cyclophosphamide (120 mg/kg) were randomized to additional treatment with either busulfan (16 mg/kg, n = 88) or total body irradiation (TBI; n = 79). The busulfan-treated patients had an increased cumulative incidence of veno-occlusive disease of the liver, ie, 12% compared with 1% in the TBI group (P = .009). Furthermore, hemorrhagic cystitis occurred in 24% of the busulfan patients versus 8% in the TBI patients (P = .003). In patients with advanced disease beyond first remission or first chronic phase, transplantation-related mortality was 62% among the busulfan-treated patients compared with 12% among the TBI recipients (P = .002). These differences between the two groups were statistically significant in multivariate analysis. Seizures were seen in 6% of the busulfan-treated patients and were absent in the TBI group (P = .03). Grade II-IV of acute graft-versus- host disease (GVHD) was similar in the two groups, but grade III-IV and chronic disease was more common in the busulfan-treated group (P = .04). Death associated with GVHD occurred in 17% of the busulfan- treated group and 2% of the TBI group (P = .003). Patients treated with busulfan had a 3-year actuarial survival of 62%, which was worse than the 76% among those treated with TBI (P < .03). In multivariate analysis, poor survival was associated with advanced disease (P < .0001), no posttransplant septicemia (P = .0006), grade II-IV GVHD (P = .006), and busulfan treatment (P < .02). The incidence of relapse did not differ between the two groups. Relapse-free survival was also similar in the two treatment groups on analysis of data from all patients, children, patients with early disease, and those with acute myeloid leukemia, acute lymphoblastic leukemia, and chronic myeloid leukemia. However, in adults (P = .05) and patients with advanced disease (P = .005), leukemia-free survival was significantly better in those treated with TBI. We conclude that patients treated with busulfan have more early toxicity and an increased transplant-related mortality in patients with advanced disease. TBI is therefore the treatment of choice, especially in adults and patients with advanced disease. However, busulfan is an acceptable alternative for patients with early disease and for those in whom TBI is not feasible.