Two hundred fifty-six children with previously untreated acute nonlymphocytic leukemia (ANLL) were evaluated on a Pediatric Oncology Group (POG) phase III randomized trial of both induction and continuation chemotherapies. Induction therapy compared vincristine, cytarabine, and dexamethasone (VADx) with daunorubicin, cytarabine, and thioguanine (DAT). The complete remission (CR) rate using DAT was superior (82% v 61%, P = .02). Postremission therapy consisted of either "standard" two-cycle therapy or a more intensive four-cycle regimen given for 2 years. Overall, there was no difference in outcome for patients randomized to either continuation regimen. The overall complete continuous remission rate (CCR) for the "best" induction/continuation therapy combination at 2 years was .50 (SE = .06), at 3 years was .35 (.04), and at 4 years was .34 (.05). Analysis of selected clinical and laboratory parameters demonstrated differences in induction responses favoring DAT induction but did not impact eventual disease-free survival. There were two subgroups of patients who responded better to four-cycle continuation therapy. These were patients with French-American-British (FAB) M1/M2 (2-year CCR was .20 v .44, P = .01) and patients older than 10 years at diagnosis (.32 v .62, P = .004).