Prognostic factors and grading systems for overall survival in patients treated with radiosurgery for brain metastases: variation by primary site.

OBJECT: The authors conducted a study to determine whether prognostic factors and the applicability of prognostic systems vary by primary tumor site in patients treated with radiosurgery for brain metastases.
METHODS: The authors evaluated data obtained in patients who underwent radiosurgery with or without whole-brain radiotherapy (WBRT) from 1991 to 2005 for newly diagnosed brain metastases. Four groups were analyzed: 1) all primary sites combined, 2) breast, 3) lung, and 4) melanoma primary sites. Kaplan-Meier, log-rank, Cox proportional hazard uni- and multivariate analysis, and recursive partitioning analysis (RPA) were used to assess prognostic factors and 4 prognostic systems: Radiation Therapy Oncology Group (RTOG) RPA, Graded Prognostic Assessment (GPA), basic score for brain metastases (BSBM), and the newly proposed Golden grading system (GGS). The GGS divides patients into 4 prognostic groups by age >or= 65 years, Karnofsky Performance Scale (KPS) score < 70, and known presence of extracranial metastases.
RESULTS: Data acquired in 479 newly diagnosed patients with 1664 lesions were analyzed. The median survival time from diagnosis of brain metastases was 12.1 months; the median follow-up was 25.4 months in 73 patients who were censored. Survival and prognostic factors were equivalent for 369 patients treated with radiosurgery compared with 110 patients treated with radiosurgery and WBRT, so these subsets were combined. Multivariate analysis of all primary sites combined demonstrated age < 65 years, KPS score >or= 70, no known extracranial metastases, and <or= 3 brain metastases were associated with longer survival, and primary tumor control was not. In subgroup multivariate analysis of patients with breast, lung, or melanoma primaries, favorable factors included only primary tumor control in 87 patients with breast primary; age < 65 years, no known extracranial metastases, and <or= 3 brain metastases in 169 patients with lung primary; and KPS >or= 70 years, primary tumor control, and <or= 3 brain metastases in 137 patients with melanoma primary. The median survival for <or= 3 versus > 3 metastases was 15.6 and 16.9 months, respectively, for breast, 16.5 and 11.3 months for lung, and 9.0 and 5.7 months for melanoma. Analysis of the 4 prognostic systems (RTOG RPA, BSBM, GPA, and GGS) showed that each prognostic system's clinical applicability varied depending on primary tumor site. The RPA confirmed that GGS and primary tumor site are significant variables for prognosis.
CONCLUSIONS: Favorable prognostic factors for patients with newly diagnosed brain metastases treated with radiosurgery vary by primary site. The 4 prognostic grading systems analyzed were applicable to different primary sites depending on which prognostic factors each individual system incorporated. Therefore, the authors recommend further development and use of primary-specific prognostic systems.