We report the first randomized study assessing the efficacy and safety of daunorubicin (DNR) continuous infusion (CI) compared to the more conventional 30-min infusion (i.v.) in newly diagnosed adult acute lymphoblastic leukemia (ALL). Seventy-seven patients were initially randomized to receive either a 24-h CI DNR (60 mg/m2 days 2-4) (40 patients) or bolus DNR at the same dosage (37 patients) with vincristine (2 mg i.v. days 1, 8, 15) and oral prednisone (60 mg/m2 days 1-15), without hematopoietic growth factor support, as an induction regimen. The distribution of adverse prognostic factors was comparable in the two-induction arm. Acute toxicity was more important in the CI arm. Gram negative infection (9 vs 1 gram negative septicemia, P = 0.01) and infection-related deaths (6 vs 1 deaths, P = NS) occurred more frequently in the CI arm during the induction treatment than in the i.v. arm, leading to the study interruption. Neutropenia but not thrombopenia duration was significantly longer in the CI arm than in the i.v. arm (18 days vs 14 days, P > 0.05 and 16 days vs 12 days, P > 0.05, respectively). Despite a similar CR rate according to the method of DNR administration (68% in the CI DNR arm vs 76% in the i.v. arm after the first course), there was a trend toward higher freedom from relapse (FFR) after DNR CI (48% vs 28% in the i.v. arm at 5 years, P = NS), suggesting that despite this high toxicity, DNR CI may improve the CR quality and decrease further the residual disease.