During the period 1976-1981, 3241 children were enrolled on three major studies of acute lymphoblastic leukemia by participating institutions of the Children's Cancer Study Group. Each study included a different method of central nervous system (CNS) prophylaxis: (1) standard therapy with cranial irradiation, 2400 rads, and intrathecal methotrexate at 12 mg/m2 six times during consolidation (CCG-141); (2) a modification of CCG-141 in which the intrathecal methotrexate was initiated during induction (CCG-141A); and (3) a reduced cranial irradiation dose of 1800 rads with intrathecal methotrexate given at the same frequency as a CCG-141A, with or without maintenance intrathecal methotrexate, but with a dosage regimen derived from CNS volume considerations rather than based on body surface area (CCG-160 series). Strategy 3, a change in the intrathecal methotrexate dosage, has resulted in the lowest incidence of CNS leukemia to date (p less than 0.007). The cumulative 3-yr CNS relapse rate has decreased from 8%-10% to 2%-5% in average-risk patients (p less than 0.02; life table estimate) and from 23%-27% to 6% in high-risk patients (p less than 0.0002; life table estimate), despite a reduction in the cranial irradiation dose from 2400 to 1800 rads. Maintenance intrathecal chemotherapy has had a marginal effect among patients randomized to receive this additional therapy (p = 0.06). The overall outcome has been an increase in the continuous complete remission rate (p = 0.04) but not in the estimated 3-yr continuous hematologic remission or survival rates.