To investigate radiation necrosis in patients treated for glioma in terms of incidence, outcomes, predictive and prognostic factors. Records were reviewed for 426 patients followed up until death or for at least 3 years. Logistic regression analysis was performed to identify predictive and prognostic factors. Multivariate survival analysis was conducted using Cox proportional hazards regression. Separate analyses were performed for the subset of 352 patients who received a biologically effective dose (BED) > or =85.5 Gy2 (> or =45 Gy/25 fractions) who were at highest risk for radionecrosis. Twenty-one patients developed radionecrosis (4.9%). Actuarial incidence plateaued at 13.3% after 3 years. In the high-risk subset, radiation parameters confirmed as risk factors included total dose (p < 0.001), BED (p < 0.005), neuret (p < 0.001), fraction size (p = 0.028), and the product of total dose and fraction size (p = 0.001). No patient receiving a BED <96 Gy2 developed radionecrosis. Subsequent chemotherapy significantly increased the risk of cerebral necrosis (p = 0.001) even when adjusted for BED (odds ratio OR, 5.8; 95% confidence interval CI, 1.6-20.3) or length of follow-up (OR, 5.4; 95% CI, 1.5-19.3). Concurrent use of valproate appeared to delay the onset of necrosis (p = 0.013). The development of radionecrosis did not affect survival (p = 0.09). Cerebral necrosis is unlikely at doses below 50 Gy in 25 fractions. The risk increases significantly with increasing radiation dose, fraction size, and the subsequent administration of chemotherapy.