Agile 3+ Score in CHB–MASLD: A Paradigm Shift or a Context‐Specific Victory?

Dear Editor,

1.

We commend Yu et al. [1]. for demonstrating that the Agile 3+ score outperforms FIB‐4, NFS, APRI, and even standalone LSM in staging fibrosis and predicting liver‐related events (LREs) among virally suppressed chronic hepatitis B (CHB) patients with concurrent metabolic dysfunction‐associated steatotic liver disease (MASLD). Their study astutely reveals limitations of conventional noninvasive tests (NITs) when metabolic and viral injuries coexist, providing timely evidence that a composite elastography‐based algorithm enhances diagnostic and prognostic precision.

However, the study design underpinning Agile 3+’s superiority also defines its boundaries. The score was trained and validated in cohorts meeting stringent criteria: (1) HBV DNA persistently suppressed below 20 IU/mL via potent antivirals, (2) alcohol use and other comorbidities rigorously excluded, and (3) hepatic steatosis defined by cardiometabolic criteria rather than ultrasonographic estimation. While these inclusion criteria maximize internal validity, they constrain external generalizability. In resource‐limited settings—where HBV viral load monitoring is intermittent, alcohol history unreliable, and controlled attenuation parameter (CAP) unavailable—can the original cutoffs retain accuracy? [2] We hypothesize not, given that both liver stiffness measurement (LSM) and platelet counts fluctuate with ongoing viral replication, necroinflammation, and hypersplenism—variables intentionally stabilized in this study.

A second, subtler caveat involves the score's biological weighting. Diabetes and AST/ALT ratio reflect metabolic activity, whereas LSM and platelets primarily indicate fibrosis. Their fusion in Agile 3+ boosts sensitivity but may increase false positives in patients with acute steatohepatitis or uncontrolled diabetes exhibiting transient LSM elevations [3]. Though the authors report a reassuringly low 15.5% gray‐zone rate, this was observed in a biopsy‐enriched tertiary cohort with 28% bridging fibrosis prevalence. In primary care, where advanced fibrosis prevalence falls below 5%, the positive predictive value of the “rule‐in” threshold (≥ 0.679) would decline without Bayesian prior adjustment or sequential testing.

Lastly, while 8‐year AUROCs exceeding 0.83 are notable, they partly reflect extended follow‐up in a high‐risk, event‐rich cohort. Unreported competing‐risk analyses pose concerns: Cardiovascular deaths—common in MASLD—could censor LREs and attenuate real‐world hazard ratios. We urge the authors to provide fine‐gray subdistribution models to verify prognostic persistence after accounting for nonliver mortality.

In summary, Agile 3+ represents not a universal fibrometer but a context‐specific tool optimized for CHB patients with pharmacologically controlled viremia and metabolic‐driven disease progression. To prevent Agile 3+ from becoming another elegant score confined by its own success, we emphasize the necessity of prospective validation across diverse healthcare settings, the development of transparent model‐retraining protocols, and the creation of an open‐access calculator that incorporates local prevalence for dynamic posterior probability updates. Extending its application to viremic, alcohol‐exposed, or HIV‐coinfected populations without appropriate recalibration risks perpetuating the very “one‐size‐fits‐all” NIT fallacies that the authors rightly critique. Therefore, prior to adoption by global guidelines, these strategic steps are essential to ensure robust and generalizable clinical implementation.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

We will be able to provide data as needed.