In all, 625 patients with acute lymphoblastic leukemia (ALL) entered the Leucémie Aiguë Lymphoblastique de l'Adulte-94 trial from June 1994 to June 1999, and received a 4-week induction therapy followed either by chemotherapy alone or stem cell transplantation (SCT). In a clinical phase II study, 40 patients with standard- or high-risk ALL - except Philadelphia chromosome-positive ALL -, relapsing at least 3 months after the beginning of therapy and who did not receive any SCT, received a rescue protocol combining amsacrine 120 mg/m(2)/day, days 1-3, cytarabine 1 g/m(2)/12 h, days 1-5, and etoposide 100 mg/m(2)/day, days 1-5. All relapses occurred 'on therapy'. In all, 16 patients (40%) achieved a second complete remission. The median time to neutrophil recovery >0.5 x 10(9)/l was 27 days. The median time to platelet recovery >50 x 10(9)/l was 28 days. Extra-hematologic toxicity was mild (only one toxic death from severe infection). The median overall survival was 5.4 months. The median disease-free survival (DFS) was 3.2 months with a 3-year DFS of 12%. Unfavorable prognostic factors for complete remission achievement were: high-risk ALL at diagnosis (P=0.03), and white blood cell count at relapse >or=30 x 10(9)/l (P=0.02). No relationship was found between survival and any characteristics of the disease. Four patients underwent allogeneic SCT (two phenoidentical and two genoidentical) and three patients received autologous SCT. This treatment combining amsacrine, cytarabine, and etoposide was therefore effective and well tolerated in 'on-therapy'-relapsed ALL. However, the median DFS was short requiring the rapid completion of effective intensive postremission therapy.