Abstract
OBJECTIVE: The efficacy of primary tumor surgery on survival in female patients with de novo stage IV breast cancer (BC) remains unclear. Our study endeavored to develop comprehensive competing risk nomograms to predict clinical outcomes and guide precision treatment in these patients. PARTICIPANTS AND METHODS: A total of 12281 patients who had distant metastasis at initial BC diagnosis between 2010 and 2017 in the Surveillance Epidemiology and End Results (SEER) database, were enrolled in this study. First, we assessed the impacts of primary tumor surgery on overall survival (OS) and breast cancer-specific survival (BCSS) using the Kaplan-Meier curves. Then subgroup analyses stratified by different metastatic patterns were performed using Cox and competing risk models (CRM). Based on the filtered independent prognostic parameters by CRM, we established two nomograms to predict the probability of breast cancer-specific death (BCSD) at 1-,2- and 3-year intervals. Furthermore, calibration curves and area under the curves (AUC) were conducted for validation. RESULTS: Kaplan-Meier analysis revealed that surgery was associated with better OS and BCSS (P<0.001). Subgroup analyses demonstrated that in bone-only metastases pattern, relative to breast-conserving surgery (BCS), patients receiving mastectomy had worse prognosis and the poorest survival belonged to non-surgery individuals (BCSS: mastectomy: HR=1.35; 95%CI=1.15-1.60; non-surgery: 2.42; 2.08-2.82; OS: mastectomy: 1.44; 1.23-1.68; non-surgery: 2.40; 2.08-2.78). Additionally, no survival difference was observed between BCS and reconstruction recipients (BCSS: HR=1.10; 95%CI=0.85-1.43; OS: 1.11; 0.86-1.44). Furthermore, patients undergoing BCS possessed similar BCSS with mastectomy recipients as well as reconstruction recipients in viscera metastases pattern, whereas non-surgery individuals had a worse survival (mastectomy: HR=1.04; 95%CI=0.92-1.18; reconstruction: 0.86; 0.69-1.06; non-surgery: 1.83; 1.63-2.05). Two competing risk nomograms of distinct metastatic patterns were established to comprehensively predict the survival of patients. Calibration curves indicated the terrific consistency of the models. Moreover, the AUC values in the training and validation sets were in the range of 0.70-0.80, exhibiting good specificity and sensitivity. CONCLUSION: The surgery implementation was associated with a lower probability of BCSD in de novo stage-IV BC patients. Our nomograms could offer a relatively accurate and individualized prediction of the cumulative incidence rate of BCSD after primary tumor resection.