In a randomized trial, 299 evaluable patients with acute myelogenous leukemia, age greater than or equal to 51, were initially randomized to cytosine arabinoside, 100 mg/m2/day, by continuous intravenous infusion for seven days, plus either daunorubicin 45 mg/m2/day x 3 (DA), or m-AMSA 200 mg/m2/day x 3 days (MA). Complete remission (CR) rates were not significantly different, 47% for DA and 42% for MA. Toxicities were similar except that severe hepatic toxicity, serum bilirubin greater than or equal to 7 mg/dl, was more frequent in patients receiving MA (10%) than in patients receiving DA (4%), p less than 0.05. Deaths during induction were significantly more frequent in patients receiving MA (38%) than in patients receiving DA (25%), p = 0.018. Patients achieving a CR received thioguanine, cytosine arabinoside, and daunorubicin (TAD) for three cycles as consolidation. Among evaluable patients, 82/102 (80%) stayed in CR during these three cycles. Patients were then randomized to either no maintenance or to DA every 13 weeks x 4 cycles, at a dose slightly lower than used for induction. Remission duration was similar for the two maintenance programs, 10.7 months for no maintenance and 8.5 months for DA. The percentage of patients evaluable for maintenance achieving three year relapse-free survival was similar for the two maintenance programs, 28% for no maintenance and 21% for DA. However, overall survival was significantly greater (40 vs 12 months, p = 0.007) for patients receiving no maintenance therapy, due to greater survival after recurrence in these patients. At each phase of the study there were substantial numbers of non-evaluable cases, often due to incomplete evaluation of remission status. Of the 379 patients initially entered into the trial, 35% obtained a complete remission. Of all the patients who achieved a complete remission, 61% were both evaluable and in remission upon completion of the maintenance phase. Of these patients who completed the maintenance phase in remission, 15% were relapse free survivors three years following initiation of maintenance therapy. Overall, only 3.2% of patients who entered the trial (35% x 61% x 15%) were continuous relapse-free survivors three years into the maintenance phase.