Abstract:Background and Aims The prognostic value of surgical intervention in patients with de novo metastatic breast cancer (dnMBC) has long been controversial. Some patients may benefit from local surgical treatment, but there is currently no effective method to identify those who would benefit from surgery. Therefore, this study was conducted to analyze the relationship between local surgery and prognosis in dnMBC patients, construct a prognostic prediction model, and determine the potential beneficiary group.Methods Data of pathologically diagnosed dnMBC cases from 2010 to 2019 were obtained from the SEER database. Patients were divided into surgery and non-surgery groups based on whether they received surgery on the primary breast lesion. Propensity score matching (1∶1) was used to balance baseline characteristics between the two groups. The matched cases were randomly split into training and validation sets in a 7∶3 ratio. A multivariate Cox proportional hazards model was employed to analyze independent prognostic factors for breast cancer-specific survival (BCSS) and to construct a prediction model. The model's discrimination, calibration, and clinical utility were evaluated using the C-index, time-dependent area under the curve (AUC), calibration curves, and decision curve analysis (DCA) in both the training and validation sets. Prognostic risk scores were calculated based on the prediction model, and patients were categorized into low, intermediate, and high-risk groups. The relationship between surgical treatment and prognosis in each risk group was analyzed using the Kaplan-Meier method.Results After matching, 2 034 patients were included in the analysis, with a median age of 56 (48-63) years. The training set comprised 1 441 cases, and the validation set comprised 593 cases. The median follow-up was 27 (11-48) months, during which 963 breast cancer-related deaths (47.35%) occurred. The estimated 3-year BCSS rates for surgery and non-surgery patients were 53.7% (95% CI=50.3%-57.3%) and 63.1% (95% CI=59.9%-66.5%), respectively. Survival analysis showed that surgery significantly improved BCSS in dnMBC patients (HR=0.72, 95% CI=0.63-0.82, P<0.001). Multivariate Cox model analysis indicated that race, histological grade, tumor T stage, brain metastasis, lung metastasis, bone metastasis, liver metastasis, hormone receptor status, HER-2 status, local surgery, and chemotherapy were independent prognostic factors (all P<0.05). A prognostic prediction model was constructed based on these independent prognostic factors, and the model was validated in both the training and validation sets. The validation results showed that the C-index was 0.707 (95% CI=0.685-0.728) and 0.705 (95% CI=0.672-0.738), respectively; the time-dependent AUC values were all between 0.7 and 0.8; the 1-, 2-, and 3-year calibration curves in both sets indicated a high concordance between predicted and actual survival rates; DCA demonstrated a good clinical net benefit of the prediction model. According to the prediction model, patients were divided into low, medium, and high-risk groups. Survival analysis revealed that surgery improved BCSS in the low-risk group (training set: P<0.000 1; validation set: P=0.003 8), while no improvement in prognosis was observed in the medium and high-risk groups.Conclusion The prognostic prediction model developed based on the clinicopathologic characteristics of dnMBC patients can stratify patients and assess the potential benefit of local surgery. This model requires further validation and optimization in prospective studies.