Development and Validation of a Machine Learning–Based Prediction Model for Cardiovascular Disease in Patients with Metabolic Dysfunction–Associated Fatty Liver Disease

BACKGROUND/AIM: Metabolic dysfunction-associated fatty liver disease (MAFLD) is strongly associated with increased cardiovascular disease (CVD) risk. However, traditional cardiovascular risk assessment tools may inadequately capture the complex pathophysiology linking hepatic and CVD in MAFLD patients. This study aimed to develop and validate machine learning models to predict CVD risk in MAFLD patients. MATERIALS AND METHODS: This cross-sectional study analyzed NHANES 2017-2023 data, with participants randomly split into training (70%) and validation (30%) sets. RESULTS: A total of 6828 adults with MAFLD were included (13.1% with prevalent CVD). Least Absolute Shrinkage and Selection Operator regression identified 14 key predictors, with age, nonalcoholic fatty liver disease Fibrosis Score (NFS), and albumin emerging as the most influential. Critically, liver-specific markers (NFS: mean |SHAP| = 0.0299; albumin: 0.0289) demonstrated superior predictive capacity compared to traditional cardiovascular risk factors (hypertension: 0.0145; diabetes: 0.0107; smoking: 0.0035). SHapley Additive exPlanations analysis revealed that older age, higher NFS (particularly > -1.0), and lower albumin (