Construction of a clinical prediction model for overall survival and cancer-specific survival in malignant phyllode tumor of the breast based on the SEER database

Abstract Objective Malignant phyllodes tumor of the breast (MPTB) represents a distinct breast tumor subtype associated with a poor prognosis. The objective of this research was to create and verify a nomogram to predict both overall survival (OS) and breast cancer-specific survival (BCSS) for individuals with a diagnosis of MPTB. Methods From the Surveillance, Epidemiology, and End Results (SEER) database, clinicopathological data of MPTB patients diagnosed between 2000 and 2020 were collected. The study cohort was randomly divided into training (70%) and validation (30%) sets using computer-generated random numbers. We performed Cox regression analyses to determine the independent factors that predict both OS and BCSS in the training cohort and internal validation cohort. Subsequently, a nomogram was developed integrating these significant predictors. Model performance was assessed using metrics such as the calibration curve, area under receiver operating characteristic curve (ROC), and decision curve analysis (DCA). Based on the nomogram scores, patients were categorized into low-risk and high-risk groups using restricted cubic spline (RCS), and survival differences were then assessed through the log-rank test and Kaplan-Meier curves. Results The study encompassed 1692 MPTB patients, randomly allocated into a training cohort (N = 1188, 70%,) and a validation cohort (N = 504, 30%). Eight independent predictors for OS were identified through univariate and multivariate analyses: age, marital status, income, stage, tumor stage, nodal stage, surgery, and chemotherapy. Additionally, six independent predictors for BCSS were identified through the same analytical approach: age, stage, tumor stage, nodal stage, surgery, and chemotherapy. Nomograms were constructed based on these variables to forecast OS and BCSS rates for patients with MPTB. Evaluation of the model’s discriminative ability using ROC demonstrated satisfactory predictive performance for OS and BCSS in both cohorts. Strong concordance between the probabilities observed and those predicted was indicated by the calibration curve. Furthermore, DCA underscored the clinical utility of the nomogram. High-risk patients (scores ≥ 85/118) had significantly reduced OS/BCSS vs. low-risk counterparts. Conclusions In this investigation, a nomogram was effectively constructed and internally confirmed to forecast OS and BCSS among individuals with MPTB. This predictive tool provides clinicians with essential prognostic information to guide their clinical decision-making processes.