Dear Editor,
We read with great interest the study by Su et al. examining the prognostic value of the Steatosis-Associated Fibrosis Estimator (SAFE) score in predicting hepatocellular carcinoma (HCC) across diverse chronic liver disease (CLD) etiologies [
1]. The inclusion of both hospital and community cohorts strengthens the generalizability of their findings, particularly in addressing the surveillance gap among nonviral liver disease populations. We commend the authors for validating the SAFE score in large datasets and for demonstrating its potential utility in identifying high-risk individuals beyond the traditional cirrhosis-based models.
However, we wish to raise a statistical consideration with direct clinical implications for future studies. The authors used subdistribution hazard models to derive adjusted subdistribution hazard ratios (aSHRs) for HCC risk, which is an appropriate choice for accounting for competing mortality. However, their reported annual incidence rates and cumulative risks in patients with SAFE ≥100 do not reflect the etiology-specific risk heterogeneity that would influence real-world screening decisions.
For instance, in the MASLD subgroup, only 1.1% of patients with high SAFE scores developed HCC, equating to an annual incidence below the cost-effectiveness threshold of 0.4%. However, the authors suggested surveillance for all patients with SAFE ≥100, regardless of etiology. This conflation of relative hazard with absolute risk may inadvertently promote over-surveillance in low-incidence subgroups, exposing patients to unnecessary testing without a demonstrable benefit.
Moreover, the diagnostic performance of SAFE ≥100, while sensitive, exhibited a low positive predictive value (PPV) across all cohorts (e.g., 4% [0.04] at 5 years in the overall cohort). Although a high negative predictive value (NPV) supports SAFE <100 as a useful exclusion tool, the low PPV raises concerns about its use as a standalone criterion for initiating long-term HCC surveillance [
2]. Clinically actionable thresholds require not only discrimination but also a favorable balance between benefits and potential harm, especially in primary care or community settings, where surveillance resources may be constrained.
Finally, SAFE was developed for fibrosis detection, not HCC prediction [
3]. The statistical reuse of this fibrosis model without recalibration for HCC-specific risk may amplify misclassification at the clinical decision point. Although the calibration plots provided by the authors are informative, they do not offer disease-specific accuracy and may overstate SAFE’s predictive reliability when pooled across etiologies.
Nonetheless, this study makes an important contribution by highlighting the need for scalable HCC risk stratification tools in non-viral liver disease. The SAFE score, especially its ability to rule out high-risk patients, represents a meaningful step toward accessible surveillance frameworks. We encourage further refinement of risk thresholds based on absolute HCC incidence by etiology and integration with other prognostic variables to guide its implementation.
FOOTNOTES
-
Authors’ contributions
Prajnasini Satapathy: Conceptualization, Methodology, Writing-Original Draft, Writing-Review & Editing. Rachana Mehta: Writing-Original Draft, Writing-Review & Editing. Ranjana Sah: Validation, Supervision, Project Administration, Writing-Original Draft, Writing-Review & Editing.
-
Conflicts of Interest
The authors have no conflicts to disclose.
Abbreviations
subdistribution hazard ratios
negative predictive value
positive predictive value
Steatosis-Associated Fibrosis Estimator
REFERENCES
- 1. Su TH, Yang SS, Lee MH, Kao WY, Huang SC, Chen FF, et al. High Steatosis-Associated Fibrosis Estimator scores predict hepatocellular carcinoma in viral and non-viral hepatitis and metabolic dysfunction-associated steatotic liver disease. Clin Mol Hepatol 2025;31:796-809.
- 2. Lee HC. Noninvasive diagnostic criteria for hepatocellular carcinoma. Clin Mol Hepatol 2012;18:174-177.
- 3. Li G, Lin H, Pimsiri Sripongpun, Liang Y, Zhang X, Wong VWS, et al. SAT-413 - Diagnostic and prognostic performance of the SAFE score in non-alcoholic fatty liver disease. J Hepatol 2023;78(Suppl 1):S663-S664.
Citations
Citations to this article as recorded by
