Within 3 weeks of definitive surgery, 571 adult patients with histologically confirmed, supratentorial malignant gliomas were randomly assigned to receive one of three chemotherapy regimens: BCNU (1,3-bis(2-chloroethyl)-1-nitrosourea) alone, alternating courses (every 8 weeks) of BCNU and procarbazine, or BCNU plus hydroxyurea alternating with procarbazine plus VM-26 (epipodophyllotoxin). Patients accrued in 1980 and 1981 were to receive 6020 rads of whole-brain radiotherapy concurrent with the first course of chemotherapy. Patients accrued in 1982 and 1983 were randomly assigned to receive either whole-brain irradiation as above, or 4300 rads of whole-brain radiotherapy plus 1720 rads coned down to to the tumor volume. The data were analyzed for the total randomized population and separately for the 510 patients, termed the "Valid Study Group (VSG)," who met protocol eligibility specifications (including central pathology review), 80% of whom had glioblastoma multiforme. The median survival times from time of randomization for the three chemotherapy groups of the VSG ranged from 11.3 to 13.8 months, and 29% to 37% of the patients survived for 18 months (life-table estimate); the differences between these groups were not statistically significant. Survival differences between the radiotherapy groups were small and not statistically significant. It is concluded that, for malignant glioma, giving part of the radiotherapy by coned-down boost is as effective as full whole-brain irradiation, and that multiple-drug chemotherapy as outlined in this protocol conferred no significant survival advantage over BCNU alone.