Machine Learning Identifies Biomarkers for Early Detection of MASLD
Overview
This study utilized machine learning on NHANES data to identify key biomarkers predictive of metabolic dysfunction–associated steatotic liver disease (MASLD) and hepatic fibrosis. Random Forest models achieved up to 86% accuracy in predicting hepatic fibrosis, highlighting age, BMI, and insulin as significant biomarkers.
Background
MASLD is the most common chronic liver disorder worldwide, linked closely to metabolic syndrome and cardiovascular disease. Diagnosis traditionally relies on invasive liver biopsy, but noninvasive imaging like FibroScan® offers safer alternatives. However, limitations of imaging underscore the need for reliable biomarkers to enable early detection and monitoring of MASLD and hepatic fibrosis.
Data Highlights
Model
Accuracy (%)
Sensitivity (%)
Specificity (%)
P-value
Steatosis (MASLD)
79.59
Not reported
84.65
< .001
Hepatic Fibrosis (Stiffness)
86.07
98.01
Not reported
< .001
Key Findings
Random Forest models predicted MASLD with 79.59% accuracy and 84.65% specificity.
Hepatic fibrosis prediction models achieved 86.07% accuracy and 98.01% sensitivity.
Age, BMI, and insulin were significant biomarkers for both steatosis and fibrosis.
Other biomarkers assessed included HbA1c, plasma fasting glucose, lipid profiles, liver enzymes, and inflammatory markers.
Machine learning approaches like XGBoost and Recursive Feature Elimination supported the robustness of biomarker selection.
Clinical Implications
Incorporating a panel of demographic, metabolic, lipid, and biochemical biomarkers can enhance early MASLD and hepatic fibrosis detection without invasive biopsy. Machine learning models can support clinicians in risk stratification and monitoring, potentially improving patient outcomes through timely intervention.
Conclusion
This study demonstrates that machine learning applied to routine biomarkers can effectively predict MASLD and hepatic fibrosis, offering a promising noninvasive diagnostic approach to address the growing burden of metabolic liver disease.
References
National Health and Nutritional Examination Survey (NHANES) 2017-2020 Data
American Liver Foundation -- MASLD Prevalence and Public Health Initiatives
by Jolie Boullion, Amanda Husein, Akshat Agrawal, Diensn Xing, Md Ismail Hossain, Md Shenuarin Bhuiyan, Oren Rom, Steven A Conrad, John A Vanchiere, A Wayne Orr, Christopher G Kevil, Mohammad Alfrad Nobel Bhuiyan
