Between May 1983 and March 1994, 31 patients with AML in second CR underwent BMT. Fifteen underwent allogeneic BMT from an HLA-identical sibling donor and 16 without a donor, unpurged ABMT. Two different preparative regimens were used: CY (120 mg/kg) and 12 Gy of fractioned TBI (19 patients), and Bu (16 mg/kg) and Cy (120 mg/kg) (BuCy2) in 12 patients. Main clinical characteristics including age, sex, length of first remission, FAB type, and number of leukocytes at diagnosis were similar in both groups. A combination of MTX and CsA was used in 13 cases whereas either CsA or MTX alone was employed in the other two patients. With a median follow-up of 5 years the actuarial 5 year probability of disease-free survival (DFS) for the whole group was 39.8% (95% CI: 29.5-50.1%). The 5 year DFS was equivalent for those who received either ABMT (41.6 +/- 14.2%) or allogeneic BMT (40 +/- 15%). Probabilities of relapse and non-relapse mortality for ABMT and allo BMT patients were 48.7 +/- 16.1 and 18.7 +/- 14.3, and 30.1 +/- 19.2 and 40.7 +/- 16.9, respectively. DFS was better in those patients with a longer duration of first CR (> 18 months) 62.5 +/- 14.4 vs 30.4 +/- 17.9%, attributable to a significantly lower relapse rate in this group of patients 16.6 +/- 12.8 vs 57.8 +/- 22.7 (P 0.05). In conclusion, similar results were observed when ABMT and allo BMT were compared for AML in CR2. A higher antileukemic effect associated with the allo BMT is balanced by an increase in transplant-related mortality. Duration of first remission was the most important factor affecting DFS and better outcome was observed for patients with longer CR1.