Advertisement

Explainable AI Model Bests FIB-4 for Detecting MASLD Fibrosis, Predicting Outcomes

Published on: 

FibroX showed superior performance, interpretability, and potential cost savings relative to FIB-4, improving risk stratification and management in MASLD.

New research is shedding light on the potential utility of FibroX, an explainable AI model, for improving advanced fibrosis detection and predicting all-cause and cardiovascular mortality in patients with metabolic dysfunction-associated steatotic liver disease (MASLD).1

Data on its use relative to fibrosis-4 index (FIB-4) were presented at Digestive Disease Week (DDW) 2025 by Basile Njei, MD, MPH, PhD, an assistant professor at Yale School of Medicine, demonstrating the superior performance, interpretability, and potential cost savings of the novel AI model.1

“Noninvasive liver disease assessment tools like FIB-4 are used to identify patients with metabolic dysfunction-associated steatotic liver disease at risk for advanced fibrosis,” Njei and colleagues wrote.1 “However, their accuracy is limited.”

The most common cause of chronic liver disease, MASLD is projected to become the leading indication for liver transplant in the US. Findings from a recent decision analytical modeling study predict a steady increase in the prevalence of MASLD from 33.7%, or 86.3 million people, in 2020 to 41.4%, or 121.9 million people, by 2050.2

FibroX, an explainable AI model, was developed to address some of the shortcomings of current noninvasive tests by improving advanced fibrosis detection and prognostic value for all-cause and cardiovascular mortality.1

To test its utility relative to FIB-4, investigators used a derivation cohort of adults with MASLD (n = 1487) from NHANES 2017–2020. MASLD was identified by the presence of hepatic steatosis, defined as CAP ≥ 274 dB/m, and ≥ 1 cardiometabolic criterion, excluding individuals with significant alcohol use or viral hepatitis. Fibrosis stage was determined using a 2-step method: blood-based noninvasive liver disease assessment tools, including FIB-4 and NAFLD fibrosis score, followed by vibration-controlled transient elastography (VCTE), in line with the 2024 AASLD guidelines.1

The FibroX model was built using eXtreme Gradient Boosting (XGBoost) and validated internally with 5-fold cross-validation. A 95% specificity cutoff was employed to reduce false positives. To evaluate clinical performance, 2 hepatologists conducted a chart review of 100 MASLD patients with liver biopsy results. Additionally, FibroX was further validated using an external cohort (n = 753) from NHANES 1988–1994 to predict all-cause and cardiovascular mortality over a 30-year follow-up.1

Explainability was ensured using Shapley Additive Explanations (SHAP), and a cost minimization analysis was performed to assess the economic impact of FibroX as a first-line screening tool in the US.1

Results showed FibroX outperformed FIB-4 in the derivation cohort for detecting advanced fibrosis (AUROC, 0.97 vs 0.62; P <.001). In biopsy-confirmed cases, FibroX showed superior accuracy (AUROC, 0.84 vs 0.75; P <.001). Calculation times were 13 seconds for FIB-4 compared with 23 seconds for FibroX when variables were preloaded.1

In NHANES 1988–1994, FibroX (C-statistic: 0.80) and FIB-4 (C-statistic: 0.79) predicted all-cause mortality, but investigators noted only FibroX predicted cardiovascular mortality (adjusted hazard ratio, 2.76; 95% CI, 1.23–2.12).1

SHAP analysis highlighted platelet count, age, HbA1c, AST, and GFR as key cardiovascular mortality predictors. Cost analysis showed FibroX could prevent 16.5 million unnecessary VCTEs and save $3.3 billion in US healthcare costs.1

“FibroX is a highly accurate and explainable AI model that improves the detection of advanced liver fibrosis in MASLD and enhances long-term cardiovascular mortality predictions compared to FIB-4,” investigators concluded.1 “Its superior performance, interpretability, and potential cost savings underscore its value for integration into clinical workflows to improve risk stratification and management in MASLD patients.”

References
  1. Njei B, Ilagan-Ying YC, Boateng S, et al. FIBROX: AN EXPLAINABLE AI MODEL FOR ACCURATE PREDICTION OF ADVANCED LIVER FIBROSIS AND CARDIOVASCULAR MORTALITY IN MASLD. Abstract presented at Digestive Disease Week 2025 in San Diego, CA from May 3 - May 6, 2025.
  2. Le P, Tatar M, Dasarathy S, et al. Estimated Burden of Metabolic Dysfunction–Associated Steatotic Liver Disease in US Adults, 2020 to 2050. JAMA Netw Open. doi:10.1001/jamanetworkopen.2024.54707

Advertisement
Advertisement