High-dose methotrexate (500 to 33,600 mg per square meter of body-surface area) with leucovorin rescue is a common component of therapy for acute lymphocytic leukemia. To increase understanding of the relation between the serum concentration and the effect of methotrexate, we conducted a randomized, prospective study of 108 children with "standard-risk" acute lymphocytic leukemia who were treated with 15 doses of methotrexate (1000 mg per square meter) that were infused over 24 hours. The median length of follow-up was 3.5 years from diagnosis for patients still in remission. Variability between patients in methotrexate clearance produced steady-state serum concentrations that ranged from 9.3 to 25.4 microM. Patients with median methotrexate concentrations of less than 16 microM (n = 59) had a lower probability of remaining in remission (P less than 0.05) than patients with concentrations of 16 microM or more (n = 49). Multivariate analyses indicated that patients with methotrexate concentrations of less than 16 microM were 3 times more likely to have any kind of relapse during therapy (P = 0.01) and 7 times more likely to have a hematologic relapse during therapy (P = 0.001). Stepwise Cox's regression identified leukemic-cell DNA content, methotrexate concentration, and hemoglobin as significant prognostic variables for hematologic relapse (P = 0.0005). We conclude that there is a concentration-effect relation for high-dose methotrexate in acute lymphocytic leukemia and that 1000 mg per square meter infused over a period of 24 hours may not be optimal for patients with relatively fast drug clearance.