Pre-treatment Contrast-enhanced Cone Beam Breast CT Imaging Features Combining with Clinicopathological Characteristics to Predict the Response of Neoadjuvant Chemotherapy: A Preliminary Feasibility Study

Abstract

Background To explore the association between pre-treatment contrast-enhanced cone beam breast CT (CE-CBBCT) imaging features and pathological complete response (pCR) after neoadjuvant chemotherapy (NAC), and to develop a predictive nomogram combining with clinicopathological characteristics. Methods A total of 183 female patients with stage II or III breast cancer underwent CE-CBBCT before NAC followed by surgery between August 2020 and September 2023 were enrolled, whose CE-CBBCT images and clinicopathological records were reviewed. All patients were randomly divided into the development cohort (n = 128) and the validation cohort (n = 55) at a ratio of 7:3. Univariate and multivariate binary logistic regression analysis were performed to identify the independent factors associated with pCR in the development cohort. A nomogram was developed based on the combined model, the receiver operating characteristic (ROC) curves, calibration curves and decision curve analysis (DCA) curves were used to evaluate and validate the predictive ability of the nomogram in the two cohorts. Results Univariate analysis showed that margin of mass (p = 0.018), distribution (p = 0.046) and morphology (p = 0.014) of calcifications, adjacent vessel sign (AVS, p = 0.001), molecular subtypes (p = 0.000), proportion of tumor-infiltrating lymphocytes (TILs, p = 0.000), and CA125 (p = 0.018) were all associated with pCR. In multivariate analyses, linear or segmental distribution of calcifications (odds ratio, OR = 6.06), AVS-positivity (OR = 0.11), HER2 enriched (OR = 10.34), TILs (OR = 1.06), and CA125 (OR = 0.93) were independent factors in the combined model. The predictive ability of the combined model (area under curve, AUC = 0.886) was superior to the clinicopathological model (AUC = 0.804; p = 0.014) and CE-CBBCT imaging model (AUC = 0.812; p = 0.047). The nomogram based on the combined model showed good discrimination (AUC: 0.886 vs. 0.820; p = 0.333) and calibration abilities (p value: 0.997 vs. 0.147) in the development and the validation cohort. Conclusion A nomogram based on pre-treatment CE-CBBCT features combining with clinicopathological characteristics is feasible and reliable for the prediction of pCR, which could contribute to the realization of clinical individualized therapy.