BACKGROUND: Metabolic dysfunction-associated steatotic liver disease is increasingly recognized as a precursor to secondary hyperuricemia, significantly exacerbating metabolic burden. However, reliable tools for early risk stratification in MASLD patients remain limited. This study aimed to develop and validate a robust machine learning-based model to identify factors associated with concurrent hyperuricemia and develop a diagnostic nomogram for MASLD patients. METHODS: A multicenter retrospective study was conducted involving 530 participants from two independent hospitals. The Least Absolute Shrinkage and Selection Operator regression was employed to screen characteristic variables. Seven machine learning algorithms, including Logistic Regression, Random Forest, and XGBoost, were constructed and compared. A quantitative nomogram was subsequently developed based on the optimal model. RESULTS: Six independent risk factors were identified: gender, body mass index, gamma-glutamyl transferase, serum creatinine concentration, serum triglycerides concentration, and the TyG index. Among the evaluated algorithms, the logistic regression model exhibited the superior balance of robustness and interpretability, achieving an area under the curve of 0.93 in the internal validation set and 0.67 in the external validation cohort. The constructed nomogram demonstrated satisfactory calibration and significant clinical utility, as confirmed by decision curve analysis, effectively distinguishing high-risk individuals across diverse populations. CONCLUSION: The novel logistic regression-based nomogram provides a practical, accurate, and cost-effective tool for the personalized risk assessment of concurrent hyperuricemia in MASLD patients, facilitating informed clinical decision-making and management.
Wu et al. (Fri,) studied this question.