This paper reports the long-term, disease-free survival of children with high-risk (HR) acute lymphoblastic leukemia (ALL) treated with two different chemotherapy regimens. HR was defined in children with ALL, younger than 2 or older than 10 years, with or without initial extrahematopoietic disease, with a leukocyte count higher than 25 x 10(9)L, or with L3 FAB type. The first group, LAL1, included 67 patients; they received induction with vincristine (VCR) and prednisone (PDN), early consolidation with 3 cycles of 6-mercaptopurine (6MP), citarabine (ARA) and VCR: central nervous system (CNS) prophylaxis (PFX) consisted of both chemotherapy in induction, consolidation and maintenance, radiotherapy (RT) in consolidation. Maintenance was given with 6MP, methotrexate (MTX), VCR adriamycin (ADR), ARA, cyclophosphamide (CFA) and PDN. The second group, LAL2, with 45 patients, received induction with VCR, PDN, CFA, epirrubicin (EPI), L-Asparaginase (L-ASP); early and late consolidation with 6MP, ARA, VCR, carmustine (BCNU), CFA, EPI, MTX and teniposide 8VM26): CNS PFX consisted of both chemotherapy in induction, consolidation and maintenance, RT in consolidation, maintenance with 6MP, MTX, EPI, CFA, ARA VM26 and BCNU. At the time of diagnosis, both groups were comparable. Disease-free survival probability, for LAL1 group was 0.41 at 14 years and for LAL2 group 0.34 at 8 years (p = 0.45). In the LAL1 group there were three failures and 20 relapses, and in the LAL2 group, there were two failures and 22 relapses. CNS relapses were one and seven in LAL1 and LAL2 groups respectively (p = 0.04). In the LAL2, group relapses were more frequent in patients with dose reduction or difered dose due to marrow toxicity (p = 0.02). We believe that the increase in CNS relapse in the LAL2 group was caused by the late administration of CNS PFX. We also believe that although intensive chemotherapy can increase long-term survival, dose adjustments due to marrow toxicity have a negative effect on long-term, relapse-free survival.