Recursive partitioning analysis of prognostic factors for patients with four or more intracranial metastases treated with radiosurgery.

The purpose of this study was to devise a new recursive partitioning analysis (RPA) of patients with four or more intracranial metastases treated with a single radiosurgery procedure to identify a class of patients with extended survival. 205 patients underwent Gamma Knife radiosurgery for four or more intracranial metastases (median = 5, range 4-18) during one session. The median total treatment volume was 6.8 cc (range 0.6-51.0 cc). Radiosurgery was used as sole management (17% of patients), or in combination with WB-RT (46%), or after failure of WB-RT (38%). The median marginal radiosurgery dose was 16 Gy (range 12-20 Gy). RPA assessed the effects of age, Karnofsky >70, extracranial disease, visceral metastases, number of metastases, total treatment volume, history of breast and melanoma primaries on survival. The median overall survival after radiosurgery for all patients was 8 months. RPA identified a favorable subgroup of 78 patients (43% of the series) with a total treatment volume <7 cc and < 7 brain metastases (Class 1), with a median survival of 13 months. This subgroup's survival was significantly better (p <0.00005) than the remaining patients (Class 2) (n=111) with a median survival of 6 months. In conclusion, RPA of multiple brain metastasis patients identified 2 distinct cohorts of patients. Class 1 patients have a total treatment volume <7 cc and < 7 metastases (4-6) with favorable survival after Radiosurgery and Class 2 patients have a total treatment volume > or = 7 cc and/or > or = 7 metastases and have a significantly poorer survival.