Decoding tumor heterogeneity with imaging biomarkers predicts response to TACE plus immunotherapy and targeted...

This study aims to quantify intratumoral heterogeneity (ITH) and identify prognostic imaging biomarkers in patients with HCC treated with transarterial chemoembolization combined with immune checkpoint inhibitor plus molecular targeted therapy (TACE-ICI-MTT).

This multicenter cohort study included 742 patients with unresectable HCC who received first-line TACE-ICI-MTT from January 2018 to December 2022. Radiomic features representing global tumor regions (GTRs) and ITH were extracted from pretreatment CT scans. A composite GTR-ITH score was generated using principal component analysis to integrate both GTR- and ITH-related features. Ensemble learning with multiple machine learning algorithms was employed to predict treatment response. Model performance was evaluated by the area under the receiver operating characteristic curve (AUC). Following feature selection, 17 GTR- and 27 ITH-related radiomic features were retained to construct the GTR-ITH score. The model demonstrated high discriminative performance, with AUCs of 0.94 in the training set, 0.82 in the internal validation set, and 0.83 in the independent test set. The GTR-ITH score was strongly associated with treatment response (OR: 34.39; p<0.001) and independently predicted overall survival (HR: 0.63; p=0.004). Patients classified as GTR-ITH low-risk consistently showed significantly prolonged progression-free survival and overall survival. To reveal the biological relevance of the radiomic score, immune infiltration patterns were characterized using bulk RNA sequencing data from the Cancer Imaging Archive. The GTR-ITH low-risk group also exhibited an immune-inflamed microenvironment characterized by enriched plasma cells and M1 macrophages, and reduced M2 macrophage infiltration.

An imaging radiomic biomarker that captures both global and intratumoral heterogeneity robustly predicts response and survival in patients with HCC treated with TACE-ICI-MTT, and reflects underlying immune microenvironment phenotypes.