Multimodal Imaging-Based Prediction of Pathological Complete Response After Neoadjuvant Chemotherapy for Breast...

The objective of this study was to develop a nomogram and assess the predictive value of imaging features-derived from ultrasound, mammography, and contrast-enhanced magnetic resonance imaging (MRI) of primary breast lesions-in combination with clinicopathological factors and serological tumor markers, for predicting pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) in patients with breast cancer (BC). Retrospective analysis was used in 294 breast cancer patients who received NAC and subsequently underwent surgery at the Harbin Medical University Cancer Hospital from 2017 to 2023. Patients were randomly assigned to a training cohort (n = 206) or a validation cohort (n = 88) in a 7:3 ratio. Data collected included preoperative imaging features of the primary breast lesion from conventional ultrasound, mammography, and contrast-enhanced MRI, as well as clinicopathological factors and serological tumor markers. After comparing the baseline characteristics between the two cohorts, univariate analysis was performed on the training cohort. Variables with significant results in the univariate analysis were incorporated into a multivariate logistic regression model. Backward stepwise selection was employed to identify independent risk factors of nonpathological complete response (non-pCR). A nomogram was constructed based on the final multivariate model. The model's discriminatory power was evaluated using the receiver operating characteristic (ROC) curve, and its calibration was assessed with a calibration plot and the Hosmer Lemeshow goodness-of-fit test. Of 294 enrolled patients, 87 (29.6%) achieved pCR. Univariate analysis in the training cohort identified multiple factors potentially associated with non-pCR. These factors included clinical pathological markers such as ER, PR, HER2, and Ki-67 status; ultrasound features including tumor location, distance from the nipple, hyperechoic halo, posterior echo, and calcification; mammographic characteristics encompassing mass margin, microcalcification, distribution and morphology of microcalcification, asymmetry, density of asymmetry, and other signs; and contrast-enhanced MRI parameters like background parenchymal enhancement (BPE) and mass margin. Multivariate logistic regression analysis subsequently demonstrated that ER, HER2, Ki-67, tumor location, distance from the nipple, morphology of microcalcification, and mass margin on contrast-enhanced MRI independently predicted non-pCR (p < 0.05). A nomogram incorporating these independent predictors showed excellent discrimination; the training cohort's AUC was 0.833 (95% CI 0.772-0.893), the validation cohort's AUC was 0.749 (95% CI 0.640-0.857). This robust predictive model represents a significant step toward individualized treatment strategies by accurately forecasting the likelihood of pCR following neoadjuvant chemotherapy in breast cancer patients.