Randomized trial investigates predictive modeling for treatment response and survival in hepatocellular carcinoma, suggesting distinct biological drivers.
Key Points
This study aims to develop a multi-faceted predictive system to distinguish between treatment response and overall survival in hepatocellular carcinoma.
Multicenter study with 246 patients receiving combined interventional therapies and targeted immunotherapy.
Integration of clinical data, dual-phase CT radiomics, and multimodal fusion algorithms to develop two models: PRIME-R for treatment response and PRIME-S for overall survival.
Model performance validated externally, using SHapley Additive exPlanations and matched imaging-transcriptomic data.
PRIME-R achieved an AUC of 0.85 for predicting treatment response, while PRIME-S had a C-index of 0.72 for predicting overall survival.
Risk stratification confirmed with a hazard ratio of 3.31 (95% CI: 2.06–5.31, P < 0.001).
Distinct biological drivers were identified: PRIME-R linked to tumor morphology and immune activation; PRIME-S associated with systemic inflammation and metabolic adaptation.