Germ cell tumours (GCTs) of the central nervous system (CNS) encompass various histological subtypes, and their optimal management has been the subject of debate. To indicate a better management strategy for each subtype, we analysed the records of 111 patients (median age 14 years), who underwent treatment since 1970. With a median follow-up duration of 86 months, the probability of surviving 5 years was: 96% for pure germinoma patients, 100% for mature teratoma, 67% for immature teratoma and 69% immature teratoma mixed with germinoma. The probability of cause-specific progression of germinomas producing human chorionic gonadotropin (HCG) was higher than that of non-producing germinomas (P < 0.01). GCTs that included a highly malignant component, such as embryonal carcinoma or yolk sac tumour, exhibited a poor prognosis with 38% chance of 5-year survival. Late adverse effects of therapy included stroke, secondary malignancy and cognitive, endocrinological, auditory and visual dysfunctions. Of 85 survivors with a median follow-up period of 99 months, 58 patients needed hormone replacement therapy, 26 patients showed poor performance status and, to date, only 1 patient has fathered children. Because the outcomes varied widely for each subtype, the traditional categories, that is, germinoma and non-germinomatous GCT as an extrapolation from the gonadal GCTs, are not suitable for appropriately selecting therapeutic regimen for CNS GCTs.