The outcome of children with acute lymphoblastic leukemia (ALL) and bone marrow relapse has been unsatisfactory largely because of failure to prevent subsequent leukemia relapses. Ninety-six patients were enrolled and received vincristine, prednisone, L-asparaginase, and an anthracycline as reinduction therapy. Ninety-two patients were randomized to receive either daunomycin (DNR) or idarubicin (IDR). After achievement of second complete remission (CR2), maintenance chemotherapy included the same anthracycline, IDR or DNR, high-dose cytarabine, and escalating-dose methotrexate. Compared to DNR (45 mg/m2/week x 3), IDR (12.5 mg/m2/week x 3) was associated with prolonged myelosuppression and more frequent serious infections. Halfway through the study, the dose of IDR was reduced to 10 mg/m2. Overall, second remission was achieved in 71% of patients. Reinduction rate was similar for IDR and DNR. Reasons for induction failure differed; none of 15, 1 of 5, and 5 of 7 reinduction failures were due to infection for DNR, IDR (10 mg/m2), and IDR (12.5 mg/m2), respectively. Two-year event-free survival (EFS) was better among patients who received IDR (12.5 mg/m2) (27 +/- 18%) compared to DNR (10 +/- 8%, P = 0.05) and IDR (10 mg/m2) (6 +/- 12%, P = 0.02). However, after 3 years of follow-up, late events in the high-dose IDR group result in a similar EFS to the lower-dose IDR and DNR groups. In conclusion, IDR is an effective agent in childhood ALL. When used weekly at 12.5 mg/m2 during induction, the EFS outcome during the first 2 years of treatment appears better than lower-dose IDR or DNR (45 mg/m2), although this difference was not sustained at longer periods of follow-up. Increased hematopoietic toxicity seen at this dose might be reduced through the use of supportive measures, such as hematopoietins and intestinal decontamination.