Abstract

Background: Metabolic dysfunction-associated steatohepatitis (MASH) diagnosis and scoring depend on invasive liver biopsies by measuring metabolic dysfunction-associated steatotic liver disease (MASLD) activity score (NAS) and fibrosis score. However, a quantitative estimation of the combined effect of clinically available laboratory measures and transient elastography parameters (controlled attenuation parameter [CAP] and liver stiffness score [LSM]) as a screening / diagnostic test for advanced MASH has been suggested. In this study, we aimed to develop such a non-invasive classifier for MASH (NAS>4 and fibrosis stage >3 [F3]) in lieu of histological assessment by liver biopsies in adult patients with MASLD.

Methods: We used a retrospective cohort study of 258 adult patients with MASLD who had undergone biopsy and transient elastography (FibroScan^®) assessment. A predictive logistic regression-based algorithm utilizing clinical and controlled attenuation parameter (CAP) and liver stiffness measurement (LSM) measured by FibroScan^® devices was developed and internally validated to predict MASH(NAS4/F3) (confirmed with liver biopsy). Fractional polynomial forms of the predictors were evaluated, and the final model was built with L1 penalization (LASSO) using 10-fold cross-validation to obtain the optimum penalty factor. The developed model was examined for calibration (calibration plot), discrimination (AUROC), and performance (Brier score, Sensitivity, Specificity, positive predictive value [PPV], negative predictive value [NPV]). Optimal Cutpoints for ‘rule-in’ and ‘rule-out’ criteria were developed. Optimism and unbiased estimates from internal validation were obtained using 2000 bootstrapped samples. All tests were two-sided with a significant p-value set at 0.05.

Results: Our predictive algorithm (FAST-3) utilized five clinical variables (sex, body mass index (BMI), aspartate aminotransferase (AST), bilirubin, and platelet count), controlled attenuation parameter (CAP), and liver stiffness measurement (LSM). The classifier demonstrated good calibration (calibration plot), discrimination (AUROC: 0.77 (0.71 – 0.83), and performance (Brier score 0.13 (0.11 – 0.16)). Optimal cutpoints for ‘rule-in’ and ‘rule-out’ criteria were evaluated and showed excellent PPV and NPV (PPV: 45% and NPV: 95%).

Conclusion: The potential utility for a non-invasive classifier (FAST-3) as a replacement for invasive liver biopsies to identify patients with advanced fibrosis in MASH (NAS4/F3) is enormous. We demonstrate excellent performance and internal validity for this index. The results from this study should be validated in a multi-center prospective cohort study, especially with high representation from Hispanic and other at-risk minority groups.