BACKGROUND: We sought to identify characteristics of metastatic breast cancer (MBC) patients who may benefit most from primary tumor resection. METHODS: Recursive partitioning analysis (RPA) was used to categorize non-surgical patients with de novo MBC in the NCDB (2010-2015) into 3 groups (I/II/III) based on 3-year overall survival (OS). After bootstrapping (BS), group-level profiles were applied, and the association of surgery with OS was estimated using Cox proportional hazards models. RESULTS: All patients benefitted from surgery (median OS, surgery vs no surgery): 72.7 vs 42.9 months, 47.3 vs 30.4 months, 23.8 vs 14.4 months (all p < 0.001) in BS-groups I, II, and III, respectively. After adjustment, surgery remained associated with improved OS (HR 0.52, 95% CI 0.50-0.55). The effect of surgery on OS differed quantitatively across groups. CONCLUSION: Prognostic groups may inform the degree of benefit from surgery, with the greatest benefit seen in those with the most favorable survival.