We read with interest the multicentre study by Guo et al, which developed a late-fusion imaging model (LFM) and a clinical-radiological model (CRM) for hepatocellular carcinoma treated with transarterial chemoembolisation (TACE).1 The multinational cohort, the external validation in Warsaw and the radiogenomic analyses are important strengths. However, several aspects of the manuscript’s clinical interpretation would benefit from clearer definition.

First, the current design seems to support the CRM primarily as a prognostic model within TACE-treated patients rather than as a treatment-selection tool. The boundary between prognostic and predictive use therefore deserves more explicit clarification.2 A prognostic model estimates outcome irrespective of treatment, whereas a predictive model indicates whether treatment effect differs between groups. The study population received first-line TACE alone, and the trial-based biomarker analysis was likewise restricted to the TACE-alone group, yet the manuscript extends its implications to personalised TACE therapy decision-making.

Second, although the CRM showed…