Addressing verification bias in masld screening: A bayesian latent class evaluation of FIB-4 and apri in a low-prevalence occupational cohort

Tóm tắt

Non-invasive tests (NITs), such as the Fibrosis-4 (FIB-4) index and the AST to Platelet Ratio Index (APRI), are recommended for stratifying risk of advanced fibrosis in metabolic dysfunction-associated steatotic liver disease (MASLD), yet their true diagnostic accuracy in primary care is often obscured by verification bias. This study evaluated the operational yield and model-based diagnostic performance of these NITs in a low-prevalence occupational health-check cohort of 238 adults. Operational analysis revealed high overall rule-out yields using pre-specified low-risk thresholds (FIB-4 < 1.30: 93.3%; APRI < 0.50: 92.4%), although the FIB-4 yield declined precipitously in patients aged >60 years old (28.6%). To compensate for the absence of a histological reference standard, a Bayesian Latent Class Analysis (LCA) adjusting for conditional dependence was employed, estimating a true advanced fibrosis prevalence of 2.6% (95% CrI: 0.5%-7.3%). Adjusted for correlated errors, FIB-4 demonstrated superior sensitivity compared to APRI (73.1% vs. 57.8%) while maintaining comparable specificity (92.7% vs. 91.8%). Furthermore, Decision Curve Analysis confirmed that an FIB-4 guided triage strategy maximizes net clinical benefit across relevant threshold probabilities. These findings indicate that FIB-4 is a highly efficient primary care gatekeeping tool, superior to APRI, provided age-adjusted thresholds are applied to older adults.