The purpose of this study was to determine prognostic factors for patients with melanoma brain metastases that can be recommended for patient selection for clinical trials. A retrospective review was conducted of 65 patients irradiated for brain metastases from 1990 to 1997. Pretreatment factors analyzed for influence on survival included age, stage, Karnofsky Performance Status (KPS), extracranial metastases, the number and location of brain lesions, disease-free interval from initial diagnosis, total dose of radiation, and number of fractions administered. Prognosis was also analyzed by Radiation Therapy Oncology Group recursive partitioning analyses (RPA) classes. The data were analyzed using the Kaplan-Meier method. Median survival was 4 months. RPA class distribution was I-25%, II-48%, and III-28% with a median survival of 6.5, 3.5, and 2.5 months, respectively (p = 0.0098 by log-rank test). KPS less than 70% (p = 0.0039), and the presence of extracranial metastases (p = 0.03), predicted a worse prognosis on univariate analysis. Both factors remained significant on multivariate analysis. The prognosis of patients receiving radiotherapy for brain metastases is related to RPA class, the presence of extracranial metastases, and KPS. These criteria should be employed in selecting patients for aggressive protocol treatment, or for more protracted brain irradiation off protocol.