A retrospective review was carried out to study children, not more than 16 years old, with a confirmed diagnosis of medulloblastoma, who were residents of the Province of Ontario at the time of diagnosis between 1977 and 1987 inclusive. The provincial tumour registry provided the population database. One hundred and eight children with medulloblastoma were identified of whom 72 (67%) were initially treated at University of Toronto Centres and 36 (33%) at other Health Science Centres, hospitals, and Regional Cancer Centres (RCC) in Ontario. The hospital/Cancer Centre records were reviewed. The 5-year relapse-free survival (RFS) for all patients treated in Ontario was 58% (SE = 5%). Those treated in Toronto had a 5-year RFS of 65% (SE = 6%) compared to 44% (SE = 8%) for those treated in other RCCs in the province (P = 0.02). Relapse-free survival for the RCCs ranged from 25 to 60%, with a trend for improved survival with increasing centre size. Univariate analysis of determinants of relapse-free survival for all 108 patients showed the following variables to be significant: T-stage (Tx + T1 + T2 vs. T3A + T3B) P = 0.0004, M-stage (M0 + Mx vs. M1-4) P = 0.0006, extent of resection (total vs. less than total) P = 0.002, radiotherapy (craniospinal irradiation and posterior fossa boost vs. other) P = 0.02, and treatment centre (Toronto centres vs. RCC) P = 0.02. Cases treated at centres outside metropolitan Toronto had a nearly two-fold (relative risk = 1.93; 95% confidence interval = 1.07, 3.47) greater risk of recurrence or death than those seen in Toronto. However, in multivariate analysis this difference was not quite significant (P = 0.07) after controlling for stage (T and M), extent of resection, meningitis, and gender. These data suggest that patients with medulloblastoma should be referred for treatment to large centres with major pediatric neurosurgical and oncology resources.