Post-treatment MRI to predict pathological complete response in triple-negative breast cancer following neoadjuvant chemoimmunotherapy.

590 Background: Neoadjuvant chemoimmunotherapy (NACI) has significantly improved pathological complete response (pCR) rates in early-stage triple-negative breast cancer (TNBC). However, the predictive accuracy of post-treatment MRI for pCR remains unexplored. Our objective was to assess the performance of post-treatment MRI in predicting pCR in TNBC patients treated with NACI. Methods: In this prospective, multicenter study (August 2021–June 2024), women with early-stage TNBC were recruited from three centers. Post-treatment dynamic contrast-enhanced (DCE) MRI data were analyzed across multiple vendors. The predictive performance of radiological complete response (rCR) on MRI for pCR was evaluated using the area under the curve (AUC) of receiver operating characteristics. A multivariable logistic regression model incorporating rCR, nodal involvement, and Ki-67 levels was developed and validated. For patients with residual enhancement, a radiomics score was generated using first-order and shape-based features. Results: The study included 175 women in a training set from centers #1 (mean age 49 ± 11 years) and 84 in an external test set from centers #2 and #3 (mean age 52 ± 12 years). MRI rCR achieved an AUC of 0.83 (95% CI: 0.75–0.92) for pCR prediction. A combined model with rCR, nodal status, and Ki-67 levels yielded an AUC of 0.88 (95% CI: 0.81–0.96) in the test set. Among patients with rCR, no nodal involvement, and Ki-67 >30%, the false-positive rate was 3.6% and 3.5% in the training and test sets, respectively, with all cases limited to Residual Cancer Burden-I. For patients with residual enhancement, a model incorporating a radiomics score and lesion count achieved an AUC of 0.80 (95% CI: 0.69–0.90). Conclusions: Post-treatment MRI effectively predicts pCR in early-stage TNBC after NACI, suggesting its potential role in identifying candidates for breast cancer surgery omission trials. Accuracy of rCR for predicting pCR. Set Subgroup Sensitivity (%) Specificity (%) PPV (%) NPV (%) F1 score (%) Training N0 andKi-67>30% 81 (54/67)[71,90] 86 (12/14) [67, 99] 96 (54/56)[92, 99] 48 (12/25)[28, 68] 88 External test N0 andKi-67>30% 82 (28/34)[70, 95] 92 (11/12) [76, 99] 97 (28/29)[90, 99] 65 (11/17)[42, 87] 89 Note-Unless otherwise specified, the data are presented as percentages, with the numbers of participants in parentheses and 95% CIs in brackets; pCR = pathological complete response; PPV = positive predictive value; NPV = negative predictive value.