Between January 1983 and November 1987, the Radiation Therapy Oncology Group conducted a prospective, randomized, multi-institutional, dose searching Phase I/II trial to evaluate hyperfractionated radiation therapy in the treatment of supratentorial malignant glioma. Patients with anaplastic astrocytoma, or glioblastoma multiforme, age 18-70 years with a Karnofsky performance status of 40-100 were stratified according to age, Karnofsky performance status, and histology, and were randomized. Initially randomization was to one of three arms: 64.8 Gy, 72.0 Gy, and 76.8 Gy. Fractions of 1.2 Gy were given twice daily, 5 days per week, with intervals of 4 to 8 hr. All patients received bis-chlorethyl nitrosourea (BCNU) 80 mg/m2 on days 3, 4, 5 of radiation therapy and then every 8 weeks for 1 year. After acceptable rates of acute and late effects were found, the randomization was changed to 81.6 Gy and 72.0 Gy with a weighting of 2:1. Out of 466 patients randomized, 435 were analyzed. The distribution of prognostic factors was comparable among the 76.8 Gy arm, 81.6 Gy arm, and the final randomization of the 72 Gy arm. The 64.8 Gy arm and the initial randomization of the 72 Gy arm had somewhat worse prognostic variables. Late radiation toxicity occurred in 1.3-6.8% of the patients, with a modest increase with increasing radiation dose. The best survival occurred in those patients treated with 72 Gy (median survival of 12.8 months overall, and 14 months for the final 72 Gy randomization). The Cox proportional hazards model confirmed the prognostic variables of age, histology and Karnofsky performance status. In addition, the longer interval of 4.5-8 hr was associated with a worse prognosis than the 4-4.4 hr interval (p = 0.0011). The difference in survival between the 81.6 Gy arm and the lower three arms approached significance (p = 0.078) with inferior survival observed in the 81.6 Gy arm. When therapy was evaluated by radiation therapy dose received (60-74.4 Gy compared with 74.5-84.0 Gy), the p value was 0.062 in favor of the lower dose range. Patients with anaplastic astrocytoma treated with 72 Gy by hyperfractionation + BCNU had at least as good a survival as those treated with 60 Gy by conventional fractionation + BCNU on Radiation Therapy Oncology Group protocols 7401 and 7918. This suggests that 72 Gy delivered by 1.2 Gy twice daily is no more toxic than 60 Gy delivered by conventional fractionation.