To compare three patient stratification systems predicting survival: recursive partitioning analysis (RPA), score index for radiosurgery in brain metastases (SIR), and a proposed basic score for brain metastases (BS-BM)    Methods and materials We analyzed the outcome of 110 patients treated with Leksell Gamma Knife radiosurgery between December 1999 and January 2003. The BS-BM was calculated by evaluating three main prognostic factors: Karnofsky performance status, primary tumor control, and presence of extracranial metastases.

Results

The median survival was 27.6 months for RPA Class I, 10.7 months for RPA Class II, and 2.8 months for RPA Class III. Using the SIR, the median survival was 27.7, 10.8, 4.6, and 2.4 months for a score of 8–10, 5–7, 4, and 0–3, respectively. The median survival was undefined in patients with a BS-BM of 3 (55% at 32 months) and was 13.1 months for a BS-BM of 2, 3.3 months for a BS-BM of 1, and 1.9 months for a BS-BM of 0 . The backward elimination model in multivariate Cox analysis identified SIR and BS-BM as the only two variables significantly associated with survival (p = 0.031 and p = 0.043, respectively).  Conclusion  SIR and BS-BM were the most accurate for estimating survival. They were specific enough to identify patients with short survival (SIR 0–3 and BS-BM 0). Because of it simplicity, BS-BM is easier to use.