PURPOSE: To evaluate the usefulness of whole brain radiotherapy (WBRT) and of the Radiation Therapy Oncology Group recursive partitioning analysis (RPA) for brain metastases among patients receiving stereotactic radiosurgery (SRS). METHODS AND MATERIALS: A retrospective analysis was performed on 135 patients who underwent linear accelerator (Linac) (n = 73) or Gamma Knife (n = 62) SRS for newly diagnosed brain metastases at the Cleveland Clinic Foundation between 8/89 and 12/98. Univariate and multivariate analyses were performed to evaluate the effects of age, primary site, control of the primary, interval to development of brain metastases (disease-free interval [DFI]), number of brain metastases, presence of extracranial metastases, Karnofsky performance status (KPS), treatment of brain metastases, and RPA class on overall survival. RESULTS: Application of the RPA classification revealed 29 patients fit the criteria for class I, 96 for class II, and 10 for class III. All of the patients underwent SRS. Fifty-seven patients also received WBRT at the time of initial presentation (SRS and immediate WBRT), and 78 patients received WBRT only if CNS relapse occurred (SRS alone). The median survival for all patients was 7.9 months (range: 1.1-90.1), and was 11.2 months for RPA class I compared to 6. 9 months for RPA classes II-III (p = 0.016). Median survival was 10. 5 months following SRS alone compared to 6.4 months following SRS and WBRT (p = 0.07). On univariate analysis, KPS >/= 80% (p = 0.002) and absence of systemic disease (p = 0.013) were also associated with longer survival, whereas control of the primary, DFI, and number of brain metastases did not have an impact. Multivariate analysis revealed only RPA class (p = 0.023) to be an independent predictor for overall survival, whereas treatment group (p = 0.079) was only marginally significant. At 2 years, immediate WBRT improved control at the original site of metastases (80% vs. 52%, p = 0.03) and prevention of new metastatic sites within the brain, 74% vs. 48% (p = 0.06). The 2-year intracranial disease-free survival was 60% following SRS and WBRT compared to only 34% following SRS alone (p = 0.03). CONCLUSIONS: Despite the inherent biases to select more favorable patients for SRS, the RPA class retains its prognostic value. Omission of WBRT from the initial management was not detrimental in terms of overall survival; however, progressive disease occurred in over 50% of patients treated in this manner. Further studies are required to determine which, if any, patients should be considered for SRS with WBRT held in reserve.