PURPOSE Locally advanced rectal cancer (LARC) poses significant treatment challenges, particularly in predicting responses to neoadjuvant chemoradiotherapy. Current methods lack accuracy in identifying patients who would benefit most from this approach. METHODS A clinical and magnetic resonance imaging (MRI) –based radiomics-immunological model (Rad-scoreₘulti) was developed to predict treatment outcomes in patients with LARC. The study included 70 patients in cohort A for biomarker validation and 68 patients in cohort B for therapy efficacy evaluation. Radiomics features were extracted from multisequence MRI scans, and machine learning algorithms were applied for feature selection and model development. The predictive performance of the model was evaluated using the AUC, decision curve analysis (DCA), and the Hosmer-Lemeshow test. RESULTS Rad-scoreₘulti, derived from multisequence MRI, demonstrated high predictive accuracy for PD-L1 expression (AUC = 0. 90) and CD8 + T-cell infiltration (P =. 0068). Patients with low Rad-scores exhibited significantly higher CD8 + T-cell infiltration. A nomogram combining clinical and radiomics features showed superior predictive performance for achieving pathological complete response (pCR) compared with clinical models alone. CONCLUSION The Rad-scoreₘulti nomogram offers a noninvasive, effective tool for predicting the efficacy of neoadjuvant therapy in patients with LARC, enhancing personalized treatment planning.
Hu et al. (Wed,) studied this question.