We studied the effectiveness of risk-directed therapy for infants younger than 13 months of age with acute lymphoblastic leukaemia (ALL). Fifty-five infants were assigned to different treatment programs (from December 1995 to December 1998) on the basis of their MLL gene status at diagnosis. Forty-two cases (76.3%) had a rearranged MLL gene (MLL+) and were treated with remission induction therapy followed by sequential intensive chemotherapy, including multiple genotoxic agents (MLL9601 protocol). Haematopoietic stem cell transplantation (HSCT) was attempted if suitable donors were available. Thirteen infants (23.7%) were classified as MLL- and treated for 2.5 years with intensive chemotherapy for high-risk B-ALL (MLL9602 protocol). Complete remission was induced in 38 of the 42 infants (90.5%) with MLL+ ALL and in all 13 patients (100%) with MLL- disease. In the MLL+ subgroup, the estimated event-free survival (EFS) rate at 3 years post diagnosis was 34.0% +/- 7.5%, compared with 92.3% +/- 7.4% in the MLL- subgroup (overall comparison, P = 0.001 by log-rank analysis). Both age less than 6 months (hazard ratio = 6.87, 95% CI = 0.91-52.3; P = 0.013) and central nervous system (CNS) involvement at diagnosis (hazard ratio = 2.92 95% CI = 1.29-6.63; P = 0.015) were significant independent predictors of an inferior outcome. These findings indicate a strategic advantage in classifying infant ALL as either MLL+ or MLL- early in the clinical course and selecting therapy accordingly. Standard chemotherapy for high-risk B-lineage ALL appeared adequate for MLL- cases. Novel therapeutic initiatives are warranted for infants with MLL+ disease, particularly those with initial CNS leukaemic involvement or age less than 6 months, or both.