Assessing Treatment Heterogeneity in Predictive Models for HCC Retreatment Response

We have thoroughly reviewed the innovative study by Wang et al. on the VITAL/VITAL-P score for predicting retreatment response in hepatocellular carcinoma after transarterial chemoembolisation [1]. This work successfully established a predictive model based on MRI imaging features, providing novel insights for individualised treatment decision-making in liver cancer. However, we would like to raise a critical methodological concern that may affect the interpretability of the model: the confounding effect introduced by treatment heterogeneity was not adequately controlled. As shown in Table 1 of the article, patients' “retreatment” strategies varied significantly, encompassing both locoregional therapies (TACE) and systemic treatments. This variation in treatment choice is itself an independent prognostic factor [2]. Under such circumstances, the association between the VITAL-P score and poor treatment response may be confounded by indication bias: clinicians may tend to select more aggressive or alternative treatment regimens for patients with poorer imaging features (i.e., higher VITAL scores). This makes it difficult to discern whether poor outcomes stem from inherent tumour refractoriness or from a mismatch between the chosen treatment strategy and the tumour's biological profile [3]. Key validation analysis: For example, conduct validation analysis within subgroups receiving only specific treatment regimens to control for treatment heterogeneity. This will enable a more accurate assessment of the intrinsic relationship between imaging features and tumour biological behaviour. Prospective validation design: Future studies should directly validate the value of this score as a “treatment stratification biomarker.” For example, patients with high VITAL-P scores could be randomly assigned to different treatment intensity groups to determine whether treatment strategies based on this score can improve patient prognosis [4]. We believe that through in-depth exploration and validation of treatment heterogeneity issues, the VITAL/VITAL-P scoring system will be able to transition more robustly from retrospective findings to clinical application, ultimately achieving the goal of guiding individualised treatment decisions. Hao Tan and Jiacheng Gao: conceptualisation, writing – original draft. Xiaofeng Luan and Hangyu Liu: supervision, writing – review and editing. The authors have nothing to report. The authors have nothing to report. The authors declare no conflicts of interest. This article is linked to Wang et al. paper. To view this article, visit https://doi.org/10.1111/liv.70481. Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.