Early Identification of Breast Cancer Patients Achieving Radiological Complete Response to NAC: A Clinicopathological and MRI-Based Approach

Background Radiologic complete response (rCR) after neoadjuvant chemotherapy (NAC) is increasingly recognized as a prognostic indicator in breast cancer, yet predictive models integrating baseline clinicopathological and multiparametric MRI features remain underdeveloped. We aimed to develop and validate a combined model for pretreatment prediction of rCR. Methods This retrospective study analyzed 286 consecutive breast cancer patients who received NAC between 2016 and 2023. rCR was strictly defined as complete absence of enhancement on post-treatment DCE-MRI per RECIST 1.1 criteria. Pretreatment clinicopathological variables (including ER, PR, HER2 status, Ki-67 index, and serum CA15-3 level) and multiparametric MRI characteristics (tumor size, morphology, enhancement kinetics, and ADC values) were evaluated. Variable selection was performed using multivariable logistic regression with variance inflation factor restriction (VIF < 5) to identify independent predictors. Model performance was assessed via receiver operating characteristic (ROC) analysis. Results Significant differences in ER/PR/HER2 status, Ki-67, CA15-3, tumor diameter, and morphology were observed between rCR (13.3%, 38/286) and non-rCR groups (all P < 0.05). Multivariate analysis identified high Ki-67 (OR = 9.009), low CA15-3 (OR = 0.098), tumor diameter < 3.15 cm (OR = 0.778), and non-irregular morphology (OR = 0.148) as independent predictors (all P < 0.05). The combined model achieved an AUC of 0.772 (sensitivity = 63.2%, specificity = 83.1%). Conclusions We developed a clinically applicable model combining readily available pretreatment clinicopathological and MRI features that effectively stratifies patients by likelihood of achieving rCR after NAC. This tool may facilitate early identification of NAC responders, potentially optimizing treatment strategies and reducing unnecessary chemotherapy exposure. Further validation in prospective, multicenter cohorts is warranted to confirm its generalizability.