Background: Recent randomized trials evaluating the role of surgery in de novo stage 4 BC yielded mixed results indicating that patient/tumor characteristics might influence the survival. We analyzed the predictors of survival after LRT in stage 4 BC by using National Cancer Data Base (NCDB). Methods: Retrospective analysis of histologically confirmed infiltrative ductal (IDC), lobular (LC) and metaplastic stage 4 BC patients from 2004-2015 using NCDB. The study cohort was divided into 2 subsets based on the therapy received: no LRT and LRT (surgery ± radiation). The LRT subset was randomized into training and validation cohorts. In the training cohort, Cox proportional method was used to calculate HRs based on which a 17 point survival prediction scoring system was developed (Table). Both the cohorts were stratified into 3 groups based on the scores- group (G)1 (0-3), G2 (4-7) and G3 (8-17). Kaplan Meier (KM) method and log-rank test were used to compare survival among the 3 groups. We validated the prognostic score by comparing the OS between the respective groups in each cohort. Results: A total of 67,978 patients met the inclusion criteria (median age 61 y). The patients in LRT subset (21,120) had significantly better survival (median: 45 vs 24 m) (p < .0001). The 3 groups in training cohort showed significant difference in the OS (p < 0.0001) – G1 having better prognosis. The 3-year OS rates of the groups were 84% (G1), 66% (G2), and 38% (G3). On validation, comparable OS was seen between the respective groups in each cohort (p = 0.77). Conclusions: LRT was associated with improved OS in de novo stage 4 BC. Based on the patient/tumor characteristics, we developed a prediction model to characterize the prognostication in patients undergoing LRT for stage 4 BC.