Four hundred ninety evaluable patients were treated on an induction regimen consisting of two to four courses of cytosine arabinoside plus an anthracycline. Overall, 78% of patients went into remission, 10% died during induction, and 12% were induction failures. For the first 152 patients, courses consisted of 7 days continuous infusion with cytosine arabinoside (Ara-C, 100 mg/m2) and 3 days of doxorubicin (30 mg/m2). Because of unacceptable toxicity, particularly for children less than 3 years of age, the anthracycline was changed to daunorubicin, and the doses of both induction drugs for children under 3 was reduced. For children aged 3 years and older the change in anthracycline was associated with a significant increase in induction failures (7% to 16%, P = .04) and a decrease in deaths (15% to 8%, P = .09). For younger children, for whom doses were also changed, the effect was greater: Mortality decreased from 29% to 1% (P < .0001), and the remission induction rate increased from 66% to 88% (P = .005). The therapy modifications also influenced survival following remission induction: Daunorubicintreated patients, aged 3 years and over, did significantly better than those given doxorubicin (P = .03), but the opposite was seen in younger children (P = .06). Gastrointestinal and skin toxicities and septicemia were significantly more common when doxorubicin was being used, but the extent of myelosuppression was similar for the two anthracyclines.