The aim of the study was to characterize the relationship between the pharmacokinetics of methotrexate in serum and concentrations in the cerebrospinal fluid, and to analyse the association to risk of a central nervous system relapse in children with acute lymphoblastic leukaemia. In this retrospective study, 353 children with acute lymphoblastic leukaemia treated with high-dose methotrexate according to the Nordic Society of Pediatric Haematology and Oncology-92 acute lymphoblastic leukaemia protocol were studied. Data from 18 patients with a subsequent central nervous system relapse and 335 event-free patients were available. In 34 patients the methotrexate concentrations were monitored repeatedly during each 24-h methotrexate intravenous infusion and a cerebrospinal fluid sample was taken at the end of the infusion. Using statistics separating interindividual and intraindividual variability, methotrexate concentration in cerebrospinal fluid was found to be significantly dependent upon both serum concentrations at the end of infusion and the area under the concentration curve in serum (P<0.0017 and <0.002, respectively). The logistic regression analysis revealed that high patient median serum methotrexate concentrations at the end of infusion were significantly associated with decreased risk for a central nervous system relapse in the standard risk group (P=0.02) and the number of courses with a calculated cerebrospinal fluid concentration of more than 1 mumol/l (P=0.048) with a decreased risk of a central nervous system relapse in the combined (standard risk and intermediate) risk group. In conclusion, methotrexate concentrations in cerebrospinal fluid are dependent on methotrexate concentrations in serum and serum area under the concentration curve. Multivariate analysis indicates that an increased exposure to methotrexate is related to decreased risk for central nervous system relapse.