Development and internal validation of a multivariable prognostic prediction model in advanced breast cancer patients...

A significant degree of contention persists regarding the optimal therapeutic approaches for managing individuals with advanced breast cancer. The objective of this study was to develop predictive models aimed at estimating the prognosis of patients and internally validating their predictive performance, as well as to precisely ascertain which subsets of patients would derive the most substantial benefits from specific treatment modalities, thereby enabling more personalized and effective therapeutic strategies.

We retrospectively analyzed advanced breast cancer cases from the Surveillance, Epidemiology, and End Results (SEER) database (2010-2015) for prognostic model development and internal validation. Eligibility: female ≥15 years at diagnosis, histopathologically confirmed primary breast cancer, American Joint Committee on Cancer (AJCC) 6th Edition stage IIB, IIIA, IIIB, IIIC, or TxNxM1. Cases with incomplete data were excluded. The distribution of clinical characteristic variables and treatment modalities was evaluated using chi-squared tests. Kaplan-Meier univariate analysis and Cox proportional hazards regression analysis were performed to screen for prognostic variables that significantly influence overall survival and breast cancer-specific survival. We constructed the nomograms to evaluate the prognosis of patients using R version 4.2.0. and plotted receiver operating characteristic (ROC) curves, calibration curves and decision curve analysis (DCA) curves to comprehensively assess the predictive performance of these nomograms.

A total of 27,593 patients were ultimately included in our study cohort. We constructed nomograms to predict the prognosis of each subgroup of patients, which included clinical variables such as age, race, grade, T-stage, N-stage, surgery, radiation, chemotherapy, estrogen receptor (ER) status, progesterone receptor (PR) status, human epidermal growth factor receptor 2 (HER2) status, etc. The model exhibited robust performance: ROC curve analysis showed area under the curve (AUC) values exceeding 75% for 3-, 5-, and 10-year survival across all patient subsets. Calibration curves demonstrated excellent agreement between predicted and actual survival probabilities, confirming the model's reliability. DCA validated good clinical utility, as the model yielded higher net benefit within the target risk threshold range.

This study provided accurate prognostic information of advanced breast cancer patients and constructed models which could provide a reference for clinicians in enabling personalized treatment among these patients.