BACKGROUND: Biomarkers are routinely used to predict responses to systemic therapies, but their utility for predicting responses to local therapy for breast cancer is not known. This study determined whether biomarkers could predict responses to breastconserving therapy (BCT) and mastectomy. METHODS: A review of the Surveillance, Epidemiology, and End Results database identified women diagnosed with early-stage invasive ductal breast cancer and treated with BCT or mastectomy from 1998 to 2008. The estrogen receptor (ER) status and the histologic grade were used to construct 3 biomarker profiles: low risk (ER-positive, low/intermediate grade), intermediate risk (ER-positive, high grade), and high risk (ER-negative, any grade). The primary measured outcome was disease-specific survival (DSS). RESULTS: BCT and mastectomy were performed in 114,486 patients (59.2%) and 79,035 patients (40.8%), respectively. There were 122,420 low-risk patients (63.3%), 34,341 intermediate-risk patients (17.7%), and 36,760 high-risk patients (19.0%). Multivariate analyses were performed separately for patients with low-, intermediate-, and high-risk tumors. The adjusted hazard ratios for DSS for patients who underwent mastectomy versus BCT for low-, intermediate-, and high-risk tumors were 1.66 (95% confidence interval [CI], 1.54-1.79; P <.001), 1.40 (95% CI, 1.29-1.53; P <.001), and 1.27 (95% CI, 1.19-1.35; P <.001), respectively. CONCLUSIONS: Patients with ER-positive, low-grade breast cancers who underwent mastectomy had a 66% increase in disease-specific mortality versus those who underwent BCT. Biomarker profiles defined by the ER status and grade may improve the selection of local therapy for breast cancer.