The recent study by Lin and colleagues investigating a liver steatosis–related polygenic risk score (PRS) for hepatocellular carcinoma (HCC) after sustained virologic response (SVR) to direct-acting antiviral therapy addresses an important and clinically relevant question 1. The integration of genetic susceptibility with long-term outcomes is timely, and the authors are to be commended for their careful consideration of competing events in the analysis. We would like to offer several methodological considerations that may help refine the interpretation of the findings, particularly with respect to risk estimation and potential clinical application. First, the authors appropriately used Fine–Grey subdistribution hazard models to account for death and liver transplantation as competing events, thereby quantifying effects on the cumulative incidence of HCC. In contrast, the non-linear association between PRS and HCC risk was examined using restricted cubic splines within a Cox proportional hazards framework. In competing-risk settings, standard Cox models typically estimate cause-specific hazards rather than cumulative incidence 2. As a result, spline-based hazard ratios derived from Cox models are not directly comparable to the subdistribution hazard ratios reported elsewhere in the manuscript. Without explicit clarification of this distinction, the observed non-linear patterns may be interpreted as reflecting absolute risk gradients 3. Greater alignment of the estimands, or parallel presentation within a consistent competing-risk framework, would enhance interpretability. Second, the proposed high-risk threshold for PRS-5 (≥ 0.66, corresponding to the 90th percentile) merits careful contextualization. Percentile-based cut-offs are reasonable for within-cohort stratification and are commonly used in genetic studies. However, the manuscript does not describe any internal validation procedures, such as resampling-based stability assessment or optimism correction, to evaluate the robustness of this threshold 4, 5. Although the decile-based sensitivity analysis suggests that excess risk is concentrated in the highest decile, this approach does not fully address the reproducibility or generalizability of the proposed cut-off, particularly given the relatively small number of HCC events observed in the cohort (38 events overall, with 8 events in the highest PRS decile). Explicitly framing this threshold as exploratory would help avoid overinterpretation in clinical settings 6. Third, post-SVR HCC is a surveillance-dependent endpoint 7. In routine practice, imaging frequency and follow-up intensity often vary according to fibrosis stage, cirrhosis status, and perceived clinical risk 8. While the models adjusted for several markers of liver disease severity, the manuscript does not report whether surveillance intensity differed between individuals classified as high versus low PRS risk. Additional clarification on surveillance patterns would help distinguish biological risk associated with genetic susceptibility from potential differences in HCC detection. Recent studies have highlighted the role of biomarkers and genetic factors in HCC risk stratification post-SVR. Caviglia et al. showed that PIVKA-II is a valuable serum biomarker for predicting HCC in HCV-related cirrhosis, while El-Serag et al. found that DAA treatment reduces HCC risk compared to untreated patients 9, 10. These findings support the need for personalised surveillance based on genetic and biomarker factors, especially in cirrhotic patients. These considerations do not detract from the novelty of incorporating steatosis-related genetic risk into post-SVR HCC assessment. Rather, they help define the current boundaries of inference and may guide future studies aimed at translating PRS-based risk stratification into clinical practice. Zhejin Shi, Haoyi Li: conceptualization and manuscript draft. Zichen Yu: critical revision for important intellectual content, final approval. The authors have nothing to report. The authors declare no conflicts of interest. This article is linked to Lin et al. papers. To view these articles, visit https://doi.org/10.1111/apt.70566 and https://doi.org/10.1111/apt.70610. Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
Shi et al. (Sat,) studied this question.