Patterns of failure were studied in two consecutive randomized trials of intensified induction therapy carried out by the Australian Leukaemia Study Group (ALSG) between 1984 and 1991 to determine the impact of dose intensification. Patients received standard dose cytarabine and daunorubicin (7-3), 7-3 plus etoposide (7-3-7) or 7-3 plus high-dose cytarabine (HIDAC-3-7) chemotherapy. Patients with FAB M3 morphology were excluded. Time to failure (TTF) was defined as the time from randomization to induction death or removal from study for non-responders, or to relapse or death in complete response (CR) for complete responders. An estimated 86% of 470 de novo patients with acute myeloid leukaemia failed within 10 years of randomization, as a result of death in induction in 17% of the randomized patients, failure to achieve CR in a further 17%, relapse in 44% and death in CR in 8% of patients. An estimated 66% of patients failed as a result of refractory disease or relapse within that period (disease-related failures). Multifactor analysis identified age and peripheral blast count as the most significant pretreatment factors associated with overall TTF. These factors, together with cytogenetics, were significantly associated with disease-related failures. High-dose cytarabine in induction significantly decreased the disease-related failure rate as did allogeneic transplantation in first CR. The impact of high-dose cytarabine did not depend on the cytogenetic risk group.