We analyzed retrospectively data from 1696 patients with AML transplanted in Europe from January 1987 to December 1992 and reported to the acute leukemia EBMT registry. Groups of patients were analyzed according to age (adults and children) and status at transplant (first remission = CR1; second remission = CR2). (1) 1114 adult patients were transplanted in CR1; 516 received an allograft; 598 received an autograft. Following alloBMT, the transplant-related mortality (TRM) was significantly higher (27 vs 13%, P < 10(-4)), the relapse incidence (RI) lower (25 vs 52%, P < 10(-4)) and the leukemia-free survival (LFS) better (55 vs 42%, P = 0.006). Favorable prognostic factors for alloBMT were a FAB type other than M4-M5, a donor-recipient combination excluding a female donor to a male recipient, and a younger age. Favorable prognostic factors for ABMT were a younger age of the patients at time of transplant, the AML3 FAB type, and a longer interval from CR1 to ABMT. (2) 288 adult patients were transplanted in CR2: 98 received an allograft; 190 received an autograft. The TRM was higher following allogeneic BMT (32 vs 20%, P = 0.02) and the RI lower (42 vs 63%, P = 0.001). The LFS was not significantly different (alloBMT: 39%; ABMT: 30%, P = 0.22). (3) 242 children were transplanted in CR1; 129 received an allograft; 113 received an autograft. Following alloBMT, the RI was lower (25 + 5 vs 48 + 6%, P < 10(-4)), and the LFS better (68 vs 47%, P = 0.002). The use of TBI was a favorable prognostic factor in allografted patients with a lower RI and a better LFS. (4) The number of children transplanted in CR2 was too small for a comparative analysis. These results confirm that both allogeneic and autologous BMT are suitable curative approaches for AML. They favor the use of an HLA identical related allogeneic transplant when available, especially in younger patients, over ABMT with unpurged marrow. The role of purging in ABMT could not be addressed in this study.