Background: NSCLC is the most common cause of BM and GPA is used as a prognostic index (PI) for BM. We evaluated its validity in a contemporary large cohort treated at a single tertiary care institution. Preliminary data was presented at ASCO 2013. Methods: IRB approved. NSCLC BM treated between 2000 and 2012 was identified from the Cleveland Clinic Brain Tumor and Neuro-Oncology Center’s database. OS from the diagnosis of NSCLC BM was the primary end point.Cox proportional hazards models were used for data analysis. Results: 958 (54% male) patients were included for final analysis. Median age at BM diagnosis was 62 years (29-92). Karnofsky performance scale (KPS) was 90-100 in 345 (39%), 70-80 in 415 (47%) and <70 in 116 (13%) patients. Single BM was noted in 448 (47%), 2 BM in 169 (18%), 3 BM in 92 (10%), 4 BM in 49 (5%), 5 BM in 56 (6%) and >5 in 136 (14%) patients. Initial therapy included stereotactic radiosurgery (SRS) in 186 (20%) patients, whole brain radiation (WBRT) in 420 (44%), surgery (S) + WBRT in 132 (14%), WBRT + SRS in 138 (15%) and S + SRS in 23 (2%) patients, while 20 patients underwent only S, 11 patients received medical therapy alone, 17 were observed, 1 patient had S, WBRT + SRS. Median OS from diagnosis of BM was 7.8 months (95% C.I. 7.1-8.9).GPA for lung cancer is derived from KPS, number of BM, the presence/absence of extracranial metastases and age. GPA was prognostic for survival with very good separation between each of the groups (p<.0001). Conclusions: Even in the modern era GPA is still a valid PI for OS in NSCLC patients with BM.

*Proportional hazards model stratified by when brain metastases were diagnosed.