Personalized Prediction of Pathological Complete Response in Breast Cancer Neoadjuvant Therapy: A Nomogram Combining Quantitative MRI Biomarkers and Molecular Subtypes

Zhendong Shi¹Xiaoxing Bian¹Hanyan Zhu¹Chunyan Li¹Jie Meng¹Xiaomin Qian²Peng Zhou¹Jin Zhang^1*

• ¹Tianjin Medical University Cancer Institute and Hospital, Tianjin, China
• ²Tianjin Medical University, Tianjin, China

Purpose: In this study, we aimed to determine the diagnostic performance of MRI in assessing neoadjuvant therapy (NAT) response, investigate determinants of its accuracy, and develop a nomogram for predicting pathological complete response (pCR) following NAT. Methods: A retrospective analysis was conducted on 554 female patients who received NAT between January 2019 and December 2022 and underwent MRI scans pre-and post-treatment. Clinicopathological and MRI characteristics were collected. Univariable logistic regression identified predictors of diagnostic accuracy. Patients were then randomly allocated to training (n=388, 70%) and validation (n=166, 30%) cohorts. Using multivariable logistic regression in the training cohort, we identified independent predictors of pCR and constructed a predictive nomogram. Model performance was assessed in both cohorts using receiver operating characteristic (ROC) curves, area under the curve (AUC), and goodness-of-fit tests. Results: The overall accuracy of breast MRI in evaluating NAT response was 77.44%. Multivariable analysis identified three factors independently associated with reduced MRI accuracy: ER-negative status, absence of ductal carcinoma in situ (DCIS), and coexistence of mass lesions with non-mass enhancement (NME). Independent predictors of pCR included: ER-negative, HER2-positive, without the presence of DCIS, the coexistence of mass lesions and NME on pre-NAT MRI, radiologic complete remission (rCR), smaller tumor size, and increasing/plateau TIC on post-NAT MRI. The predictive nomogram demonstrated robust discrimination, with AUC values of 0.894 (95% CI: 0.857–0.932) in the training cohort and 0.888 (95% CI: 0.841–0.935) in the validation cohort. Conclusion: Breast MRI accuracy was reduced in ER-negative tumors, those lacking DCIS, and lesions exhibiting coexistent mass and NME. A clinicopathological-MRI integrated nomogram demonstrated robust predictive performance for pCR after NAT completion, potentially aiding in surgical strategy planning.