Construction and evaluation of a diagnostic model for metabolic dysfunction-associated steatotic liver disease based on advanced glycation end products and their receptors

Abstract

Background: Effective biomarkers for the diagnosis of metabolic dysfunction-associated steatotic liver disease (MASLD) remain limited. This study aims to evaluate the potential of advanced glycation end products (AGEs) and their endogenous secretory receptor (esRAGE) as non-invasive biomarkers for diagnosing MASLD, to explore differences between obese and non-obese MASLD patients, and to develop a novel diagnostic model based on these biomarkers.

Methods: This study enrolled 341 participants, including 246 MASLD patients (118 non-obese, 128 obese) and 95 healthy controls. Serum AGEs and esRAGE levels were measured by ELISA. Key predictors were identified using the Lasso algorithm, and a diagnostic model was developed with logistic regression and visualized as nomograms. Diagnostic accuracy and utility were evaluated through the area under the curve (AUC), bootstrap validation, calibration curves, and decision curve analysis (DCA).

Results: Serum AGEs and esRAGE levels were significantly higher in MASLD patients compared to controls. Moreover, obese MASLD patients had higher esRAGE levels than non-obese ones, but no significant difference in AGEs levels was found. A diagnostic model incorporating age, WC, BMI, ALT, TG, HDL, AGEs, and esRAGE achieved an AUC of 0.963, with 94.3% sensitivity and 85.3% specificity. The AUC for bootstrap internal validation was 0.963 (95% CI: 0.944-0.982). Calibration curves showed strong predictive accuracy, and DCA demonstrated high net clinical benefit.

Conclusion: Serum AGEs and esRAGE serve as non-invasive biomarkers for distinguishing MASLD patients. We developed and validated diagnostic models for MASLD, offering valuable tools to identify at-risk populations and improve prevention and treatment strategies.

Keywords: MASLD; advanced glycation end products; diagnostic model; esRAGE; non-obese.

FIGURE 1

Participant selection flowchart.

FIGURE 2

Comparative analysis of AGEs and esRAGE levels in normal controls versus patients with MASLD. (A) Patients with MASLD had considerably higher levels of AGEs (P < 0.001). (B) MASLD patients had considerably higher levels of esRAGE (P < 0.001). (C) A heatmap depicting the correlation analysis among all candidate predictors showed statistical significance, indicated by asterisks (*) and double asterisks (**), in the correlations between variables. (D) Scatterplot of AGEs versus esRAGE, with a positive correlation (P = 0.0164).

FIGURE 3

AGEs and esRAGE levels in obese and non-obese MASLD patients are compared. (A) No substantial variations in AGE levels were seen between non-obese and obese MASLD patients. (B) Patients with MASLD who were obese had elevated levels of esRAGE (P < 0.05). (C) A heatmap depicting the correlation analysis among all candidate predictors showed statistical significance, indicated by asterisks (*) and double asterisks (**), in the correlations between variables. (D) AGEs and esRAGE did not show a statistically significant link, according to Spearman’s correlation analysis.

FIGURE 4

Factor selection was conducted using Lasso binary logistic regression models. (A) The logarithm of the lambda (λ) values for the 28 candidate variables in the Lasso model was determined. (B) log (Lambda) value that was most appropriate in the LASSO model.

FIGURE 5

ROC curves for predictors and MASLD diagnostic models. (A–H) ROC curves for predictors Age, WC, BMI, ALT, TG, HDL, AGEs, esRAGE models. (I) ROC curves for diagnostic models.

FIGURE 6

Validation of the diagnostic models. (A) Bootstrap internally validated ROC curves. (B) Nomogram for predicting MASLD risk classification. (C) Model calibration curves. (D) When compared to a “all treatment” or “no treatment strategy, DCA demonstrates the net advantage of using the model to diagnose MASLD at different decision thresholds.

FIGURE 7

Comparison of the MASLD Score with MASLD-related scoring systems. (A) ROC curve analysis of the MASLD score, HSI, K-NAFLD, NSS, and NAFLD Ridge Score in the MASLD cohort. (B) Comparison of the diagnostic performance of the models in the non-obese MASLD population. (C) Comparison of the diagnostic performance of the models in the obese MASLD population.