Beyond Genetic Enhancement: The Enhanced Steatosis Index Demands Risk‐Stratified Thresholds and Subgroup‐Specific Solutions

We read the Liver International article by Alyousifi and his colleagues with great interest, and we congratulate the authors on their elegant development and validation of the Enhanced Steatosis Index (ESI) [1]. The integration of a polygenic risk score (PRS) with readily available clinical parameters represents a thoughtful advancement toward personalised screening for metabolic dysfunction-associated steatotic liver disease (MASLD). The rigorous two-cohort validation approach and transparent reporting are commendable. However, we propose two critical considerations that, if addressed, would significantly enhance the clinical translatability and robustness of this promising tool. First, the single-threshold strategy may not accommodate the heterogeneity of clinical decision-making contexts. The authors appropriately determined an optimal threshold of 0.255 using Youden's index in ESI-2, yielding a positive predictive value (PPV) of only 0.55 and a negative predictive value (NPV) of 0.90 in the UK Biobank test set. While the authors also present performance at 90% sensitivity and 90% specificity thresholds (0.162 and 0.462 respectively), the manuscript does not provide guidance on when to apply each threshold in practice. For instance, primary care screening requires maximal sensitivity to minimise false negatives, whereas hepatology referral decisions demand higher specificity to avoid unnecessary invasive testing [2, 3]. We recommend developing risk-stratified, scenario-specific thresholds anchored to clinical actions. The Decision Curve Analysis already suggests net benefit varies across probability thresholds; extending this to define distinct “screen-triage-confirm” zones (e.g., < 0.162: routine care; 0.162–0.462: lifestyle intervention; > 0.462: specialist referral) would transform the ESI from a static score into a dynamic decision-support system. Such stratification would also mitigate the high false-positive rate that could overwhelm healthcare systems. Second, the significant performance attenuation in high-risk subgroups warrants urgent attention and solution development. As reported in Results, the ESI-2 discriminative ability declines markedly in individuals with obesity (AUC 0.74 vs. 0.81 in non-obese) and type 2 diabetes (AUC 0.76 vs. 0.84 in non-diabetic patients). Since these are precisely the populations in whom accurate MASLD detection is most clinically impactful, this paradoxical finding undermines the model's utility. The authors speculate but do not elucidate whether this reflects predictor saturation or unique pathophysiology in these groups. We encourage subgroup-specific model recalibration or development of interaction terms (e.g., BMI × diabetes status) that might restore discriminative capacity, consistent with TRIPOD guidance on subgroup reporting [4]. In addition, reporting clinical outcomes (such as the false negative rate of diabetic patients) will quantify the potential missed diagnoses. Before clinical deployment, external validation in cohorts enriched with these high-risk characteristics (such as weight loss surgery clinics or diabetes registration systems) is crucial, as these environments represent the most likely real-world application points. In summary, the ESI model represents a significant methodological advance. Addressing these two dimensions-background threshold adaptation and subgroup performance optimization—will bridge the gap between statistical performance and true clinical impact, ensuring that this innovative tool serves those who need it most. J.Z. writing-review and editing, writing-original Draft; L.L. writing-review and editing. The authors have nothing to report. During the preparation of this work the author(s) used ChatGPT in order to check for grammar/spelling. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. The authors have nothing to report. The authors declare no conflicts of interest. The authors have nothing to report.