[en] Purpose/Objective: It is thought that focally increased dose results in improved survival in selected patients with malignant glioma, and many patients receive radiosurgery as part of their therapy. However, the association between selection and outcome is poorly understood. We performed a retrospective multi-institutional analysis to determine factors associated with better or worse survival and complications. The 9 participating Gamma Knife facilities provided a broad range of selection criteria and clinical and technical data that could be related to outcome. Materials and Methods: A total of 202 tumors were treated in 189 patients (176 patients with unifocal tumor; 13 patients with bifocal tumors treated on the same day), with 30% of patients treated for primary tumor and 70% for recurrent tumor. Pathological diagnoses included glioblastoma (51%), anaplastic astrocytoma (23%), astrocytoma (23%), and pilocytic astrocytoma (3%). Median (range) characteristics were: age 47 years (2-84 years), KPS 90 (40-100), minimum tumor dose 16 Gy (5-37.5 Gy), prescription isodose percent 50% (20-90%), number of isocenters per tumor 5 (1-15), and median treatment volume 64 cc (0.3-96.0 cc). Brachytherapy selection criteria were satisfied in 65% of patients. The primary endpoints of the study were survival and complications. Median follow-up of surviving patients was 65 weeks after radiosurgery (maximum 341 weeks). Results: Acute complications were seen in 9% of evaluated patients, chronic complications in 17% of evaluated patients followed at least one year, and chronic steroid dependency in 36% of evaluated patients followed at least one year. Long-term KPS was found to improve or remain stable in approximately half of surviving patients. No factors were identified that were significantly associated with risk of chronic complications. Risk of acute complications and chronic corticosteroid dependency were associated with older patient age. Multivariate analysis showed that significantly improved survival was associated with 5 variables: lower pathologic grade, younger age, increased Karnofsky performance status (KPS), smaller tumor volume, and unifocal tumor. A hazard ratio model that is independent of the technical details of radiosurgery suggests that variations in reported survival following radiosurgery may be explained in part by differences in distributions of these variables. Survival from the date of radiosurgery was not found to be significantly related to technical parameters associated with radiosurgery (dose, isocenters, prescription percent, inhomogeneity), extent of pre-radiosurgery surgery, or whether typical brachytherapy selection criteria were satisfied. Conclusion: Patients with a poor constellation of 5 clinical variables are unlikely to obtain significant benefit from radiosurgery; patients with a favorable constellation may obtain a benefit similar to that of brachytherapy. These 5 variables should be incorporated in the design of future studies