Risk factor analysis and predictive model construction of lean MAFLD: a cross-sectional study of a health check-up population in China

AimCardiovascular disease morbidity and mortality rates are high in patients with metabolic dysfunction-associated fatty liver disease (MAFLD). The objective of this study was to analyze the risk factors and differences between lean MAFLD and overweight MAFLD, and establish and validate a nomogram model for predicting lean MAFLD. MethodsThis retrospective cross-sectional study included 4363 participants who underwent annual health checkup at Yuyao from 2019 to 2022. The study population was stratified into three groups: non-MAFLD, lean MAFLD (defined as the presence of fatty liver changes as determined by ultrasound in individuals with a BMI < 25 kg/m(2)), and overweight MAFLD (BMI >= 25.0 kg/m(2)). Subsequent modeling analysis was conducted in a population that included healthy subjects with < 25 kg/m(2) (n = 2104) and subjects with lean MAFLD (n = 849). The study population was randomly split (7:3 ratio) to a training vs. a validation cohort. Risk factors for lean MAFLD was identify by multivariate regression of the training cohort, and used to construct a nomogram to estimate the probability of lean MAFLD. Model performance was examined using the receiver operating characteristic (ROC) curve analysis and k-fold cross-validation (k = 5). Decision curve analysis (DCA) was applied to evaluate the clinical usefulness of the prediction model. ResultsThe multivariate regression analysis indicated that the triglycerides and glucose index (TyG) was the most significant risk factor for lean MAFLD (OR: 4.03, 95% CI 2.806-5.786). The restricted cubic spline curves (RCS) regression model demonstrated that the relationships between systolic pressure (SBP), alanine aminotransferase (ALT), serum urate (UA), total cholesterol (TCHO), triglyceride (TG), triglyceride glucose (TyG) index, high density lipoprotein cholesterol (HDLC), and MAFLD were nonlinear and the cutoff values for lean MAFLD and overweight MAFLD were different. The nomogram was constructed based on seven predictors: glycosylated hemoglobin A1c (HbA1c), serum ferritin (SF), ALT, UA, BMI, TyG index, and age. In the validation cohort, the area under the ROC curve was 0.866 (95% CI 0.842-0.891), with 83.8% sensitivity and 76.6% specificity at the optimal cutoff. The PPV and NPV was 63.3% and 90.8%, respectively. Furthermore, we used fivefold cross-validation and the average area under the ROC curve was 0.866 (Figure S3). The calibration curves for the model's predictions and the actual outcomes were in good agreement. The DCA findings demonstrated that the nomogram model was clinically useful throughout a broad threshold probability range. ConclusionsLean and overweight MAFLD exhibit distinct metabolic profiles. The nomogram model developed in this study is designed to assist clinicians in the early identification of high-risk individuals with lean MAFLD, including those with a normal BMI but at metabolic risk, as well as those with abnormal blood lipid, glucose, uric acid or transaminase levels. In addition, this model enhances screening efforts in communities and medical screening centers, ultimately ensuring more timely and effective medical services for patients.