Development and validation of prediction models for 5-year and 10-year ipsilateral breast tumor recurrence after breast-conserving surgery.

575 Background: Ipsilateral breast tumor recurrence (IBTR) remains a critical concern for patients undergoing breast-conserving surgery (BCS). Reliable prediction tools for IBTR risk can support personalized surgical strategies and adjuvant treatment decisions, especially in the era of evolving systemic therapies. This study aimed to develop and validate prediction models for 5-year and 10-year IBTR. Methods: This multi-center retrospective cohort study included 10,089 women who underwent partial mastectomy for invasive breast cancer between 2008 and 2017. Cases involving conversion to mastectomy, use of neoadjuvant chemotherapy, bilateral/multiple cancers, or missing key data were excluded. Prediction models were developed using Cox proportional hazards regression and validated via bootstrap resampling. Model performance was assessed using Harrell’s C-index, Brier scores, calibration plots, and goodness-of-fit tests. The cumulative incidence of IBTR, which served as the baseline for the prediction model, was calculated using the Fine and Gray model, treating death as a competing risk. Results: The median age of patients was 55 years [interquartile range (IQR): 46–65]. During a median follow-up of 8.9 years (IQR: 6.4–10.8), IBTR occurred in 292 patients (3.1%). The initial model, based on variables from Sanghani et al. (JCO 2010), achieved a Harrell’s C-index of 0.70. Incorporating hormonal receptor status, HER2 status, radiotherapy, and targeted therapy as predictors reduced the C-index to 0.60, despite their clinical relevance. Importantly, the inclusion of these factors improved calibration, demonstrating better alignment between predicted and observed IBTR probabilities. The final Cox model exhibited strong clinical and statistical robustness (p < 0.001), providing individualized IBTR risk estimates. Cox-Snell residual analysis confirmed goodness-of-fit, with the cumulative hazard closely following the 45-degree line up to 0.3, indicating reliable model performance for observed events. While hazard ratios (HRs) for chemotherapy and radiotherapy were consistent with results of EBCTCG meta-analyses (MA), HR for endocrine therapy was lower than reported in MA. Consequently, HRs from MA were adopted to account for treatment effects in our prediction model. Conclusions: We have developed and validated a new prediction model for 5-year and 10-year IBTR using Cox regression and bootstrap methods. A web-based tool is under development to enable individualized risk assessment and treatment planning. Future research will focus on external validation and the integration of genetic and novel therapeutic data to enhance model robustness and clinical utility.