We conducted a randomized multicenter trial comparing low-dose cytarabine (LD ARA-C) (20 mg/m2 for 21 days) with an intensive chemotherapy (rubidazone [a daunorubicin-derived agent], 100 mg/m2 for 4 days, ARA-C 200 mg/m2 for 7 days) in 87 patients over 65 years of age with de novo acute nonlymphocytic leukemia (ANLL). Forty-one patients received LD ARA-C and 46 received intensive chemotherapy. The number of complete remissions (CRs) but also of early deaths was higher in the intensive chemotherapy group, while partial remissions (PRs) and failures were more frequent in the LD ARA-C group (P less than .001). Infectious complications during induction treatment were more numerous and more severe in the intensive chemotherapy group (P less than .01). Patients treated with LD ARA-C required fewer RBC transfusions (P less than .02), fewer platelet transfusions (P less than .01), and had a shorter hospital stay for induction treatment (P less than .01). Overall survival and CR duration were not significantly different in either group. In the LD ARA-C group, the survival of patients with PR and those of patients in CRs was identical. We conclude that in a selected group of elderly patients with de novo ANLL a higher number of CRs may be obtained with intensive chemotherapy, but that with LD ARA-C, the number of early deaths is lower, and long-lasting PRs are obtained, resulting in a similar overall survival.