To test whether variables at diagnosis can identify patients with acute nonlymphoblastic leukemia (ANL) for whom bone marrow transplantation (BMT) is more likely to be of benefit and those for whom continued chemotherapy is a better approach, the association of 15 clinical and laboratory factors with outcome was investigated among 220 patients (ages 1 to 53 years) treated with cyclophosphamide and total body irradiation (TBI) followed by allogeneic BMT, and among 392 patients (ages 13 to 50) administered intensive chemotherapy. In the BMT group, female sex, younger age, the absence of hepatitis during induction, a larger percentage of circulating blasts, and a shorter duration of symptoms were associated with longer survival, whereas only female sex and younger age favorably influenced disease-free survival (DFS). In the chemotherapy group, younger age, lower WBC at diagnosis, a single successful induction course, and the absence of circulating promyelocytes were associated with longer survival, whereas only a lower WBC and a lower percentage of peripheral neutrophils were associated with longer DFS. Estimated regression coefficients for treatment-by-prognostic-factor interactions were used to characterize subgroups of patients in which one treatment or the other produced better outcomes. BMT and chemotherapy produced similar durations of survival in a subset of patients characterized by many or all of the following: older age, male sex, achievement of complete remission (CR) after one induction, and absence of circulating blast cells at presentation. These data suggest that, using pretreatment variables, subgroups of patients can be identified for whom either BMT or continued chemotherapy is most likely to be beneficial.