Mathematical Prediction Models for Sentinel Node Status in Early-Stage Breast Cancer: Protocol for a Systematic Review

Background The status of the axilla remains a significant prognostic factor and influences adjuvant systemic and locoregional treatment choices in early-stage breast cancer (EBC). Sentinel node (SN) biopsy continues to be the preferred technique for establishing axillary nodal status in clinically node-negative EBC. A multivariable prediction model with adequate accuracy and generalizability has been explored as a potential alternative to SN. Objective This systematic review aims to evaluate the predictive performance, methodological quality, and risk of bias associated with the available mathematical models (MMs), excluding artificial intelligence (AI)–based models, for predicting SN status in patients with EBC. Methods A systematic search will be conducted across PubMed, Cochrane CENTRAL, and Embase to identify studies reporting the development of an SN status prediction model. Only studies that report SN status using mathematical modeling techniques will be included. Two independent reviewers will screen the search results and extract data from the included articles. The primary outcome of this systematic review is to evaluate the methodological adequacy and generalizability of individual MMs and compare the reported predictive performances of methodologically robust MMs. The secondary objective is to identify key predictive factors contributing to SN status prediction in MMs. A narrative synthesis of all the included studies will be undertaken. The details of this protocol are accessible on PROSPERO, where it was registered on January 23, 2025. Ethics approval is not required for this study because only published data will be analyzed. Results Funding for this review was obtained in 2022. The literature search was completed on December 15, 2023, and screening began in December 2023. Data extraction and assessment using the Prediction Model Risk of Bias Assessment Tool was completed by December 2025, with synthesis planned for March 2026. Of the 3458 screened records, 122 (3.5%) were selected for data extraction. Results will be prepared for submission for a peer review and publication in mid-2026. Conclusions This review will provide a consolidated evaluation of non–machine learning MMs for predicting SN status in EBC. By clarifying the predictive performance and methodological quality of traditional statistical approaches, the findings will serve as a benchmark against which emerging AI-based tools can be compared. This review is also expected to identify predictors that consistently contribute to accurate modeling, informing the development of future statistical and AI-enhanced prediction tools. Trial Registration PROSPERO CRD42025637632; https://www.crd.york.ac.uk/PROSPERO/view/CRD42025637632 International Registered Report Identifier (IRRID) DERR1-10.2196/82523