Between 1985 and 1989, 38 children with newly diagnosed medulloblastoma entered our therapeutic protocol. After surgery and postoperative staging assessments, patients were assigned to risk groups. Eleven with "standard-risk" (SR) tumors were treated with radiation therapy alone, while 27 with "high-risk" (HR) tumors received radiation therapy plus adjuvant chemotherapy with vincristine, methotrexate, VM-26, and 1-(2-chloroethyl)-3-cyclohexyl-1-nitrosourea (CCNU). After a minimum follow-up of 5 years (range 5-9 years) 21/38 children had developed a recurrence or progression of their disease and 19/38 patients had died. Five-year event-free survival rates and 5-year total survival rates for all 38 patients were 47.4% and 50% respectively. The event-free survival rates at 5 years for SR and HR patients separately were 27.3% and 55.6%, respectively. The corresponding 5-year total survival rates were 27.3% and 59.3%. The differences were not statistically significant. Univariate analysis showed age at diagnosis to be the most important prognostic factor. Infants aged 5 years or less had a significantly shorter event-free survival time than older patients (P = 0.00897). Similar effects were found when total survival time was considered. There were significant differences in outcome in patients receiving different doses of radiation, suggesting a dose-response relationship. A Cox stepwise multivariate analysis showed age at diagnosis as the only independent prognostic factor. Variables relating to treatment entered the model, suggesting that chemotherapy could play an important role in determining outcome.