We tested the validity of two prognostic indices for stereotactic radiosurgically (SRS)-treated patients with brain metastases (BMs) from five major original cancer categories. The two indices are Diagnosis-Specific Graded Prognostic Assessment (DS-GPA) and our Modified Recursive Partitioning Analysis (RPA). Forty-six hundred and eight BM patients underwent gamma knife SRS during the 1998-2011 period. Primary cancer categories were non-small cell lung cancer (NSCLC, 2827 patients), small cell lung cancer (SCLC, 460), gastro-intestinal cancer (GIC, 582), breast cancer (BC, 547) and renal cell cancer (RCC, 192). There were statistically significant survival differences among patients stratified into four groups based on the DS-GPA systems (p < 0.001) in all five original cancer categories. In the NSCLC category, there were statistically significant mean survival time (MST) differences (p < 0.001) among the four groups without overlapping of 95 % confidence intervals (CIs) between any two pairs of groups with the DS-GPA system. However, among the SCLC, GIC, BC and RCC categories, MST differences between some pairs of groups failed to reach statistical significance with this system. There were, however, statistically significant MST differences (p < 0.001) among the three groups without overlapping of 95 % CIs between any two pairs of groups with the Modified RPA system in all five categories. The DS-GPA system is applicable to our set of patients with NSCLC only. However, the Modified RPA system was shown to be applicable to patients with five primary cancer categories. This index should be considered when designing future clinical trials involving BM patients.