The aim of this study was to evaluate the feasibility and efficacy of a protocol that includes "sandwich" chemotherapy, that is, chemotherapy alternated with radiotherapy, and reduced doses of supratentorial irradiation in children with medulloblastoma. Between March 1985 and September 1988, 70 successive children with newly diagnosed medulloblastoma from eight centers were treated in this prospective nonrandomized study. Patients were assigned to two risk groups. Group A included patients with macroscopically complete or subtotal excision, no brainstem involvement, no atypical cells in the cerebrospinal fluid, normal myelography, and who were more than 2 years of age. Group B patients encompassed those who did not fit the criteria for Group A. Two children were excluded from analysis after histological review confirmed ependymoma. Thus, a population of 68 children was selected, with 31 in Group A and 37 in Group B. Treatment consisted of two courses of the "eight drugs in 1 day" ("8/1") regimen followed by two courses of high-dose methotrexate (12 g/m2). Radiotherapy was begun during the 7th week after surgery in Group A and during the 5th week in Group B. In patients older than 2 years, the median radiation dose to the posterior fossa, the spinal axis, and the brain was 54 Gy, 36 Gy, and 27 Gy, respectively. Group B patients received postirradiation chemotherapy with four 8/1 courses monthly. The median time from surgery to radiation therapy was 50 days (range 21 to 141 days). One fatality due to chicken pox on Day 102 and one World Health Organization Grade IV infection occurred. The estimated 5- and 7-year disease-free survival (DFS) rates were 62% and 59%, respectively. These were 74% and 62% in Group A and 57% and 57% in Group B. Patient age, extent of resection, and radiation dose to the whole brain had no prognostic value. Patients with metastasis had a nonsignificant trend for a worse prognosis than patients with nonmetastatic disease (7-year DFS 45% vs. 68%, p = 0.11). In Group B, the 7-year DFS rates for children who received more or less than 30 Gy to the brain were 69% and 52% respectively (p = 0.15). There were recurrences in the posterior fossa (37%), spine (20%), and brain (20%). After a review of radiotherapeutic treatments, only one supratentorial failure could be blamed on reduction of the supratentorial radiation dose. This "sandwich" chemotherapy appeared to be feasible and did not show adverse survival data when compared to other series.