Construction and Validation of a TyG-ALT-Based Diagnostic Risk-Stratification Model for Metabolic-Associated Fatty Liver Disease in Patients with Obstructive Sleep Apnea

Objective: The objective of this study is to investigate the clinical value of the triglyceride-glucose (TyG) index combined with alanine aminotransferase (ALT) in identifying prevalent metabolic-associated fatty liver disease (MAFLD) in patients with obstructive sleep apnea (OSA), and to construct and validate a diagnostic risk-stratification model. Methods: Clinical data of OSA patients were retrospectively collected from two centers: the Department of Geriatrics, Peking University People’s Hospital (August 2021 to December 2025) and the Department of Geriatrics, Shijiazhuang People’s Hospital (June 2023 to December 2025). MAFLD was diagnosed by abdominal ultrasonography performed by experienced radiologists blinded to laboratory results. Candidate predictors were selected using univariate logistic regression, LASSO regression, and bootstrap stability testing. Model performance was assessed by discrimination (area under the receiver operating characteristic curve, AUC), calibration (Hosmer–Lemeshow test, calibration curves), and clinical utility (decision curve analysis). Internal validation was performed using 10-fold cross-validation and bootstrap resampling with optimism correction. External validation was conducted in an independent cohort. Sensitivity analyses included subgroup analyses stratified by sex, age, BMI, and OSA severity. Results: A total of 962 patients were included in the development set and 116 in the external validation set. Multivariate analysis identified TyG index (OR = 1.95, 95% CI: 1.51–2.53), LDL-C (OR = 1.24, 95% CI: 1.02–1.49), BMI (OR = 1.21, 95% CI: 1.16–1.26), and ALT (OR = 1.04, 95% CI: 1.02–1.05) as variables independently associated with prevalent MAFLD, while platelet-to-lymphocyte ratio (PLR) was protective (OR = 0.996, 95% CI: 0.993–0.999). The simplified TyG-ALT model achieved an AUC of 0.714 (95% CI: 0.685–0.744) in the development set, with an optimism-corrected AUC of 0.712, and an AUC of 0.783 (95% CI: 0.701–0.866) in external validation. The model demonstrated good calibration and favorable clinical net benefit within the threshold range of 0.30–0.70. The optimal cutoff was 0.564, with sensitivity of 68.7% and specificity of 79.6%. Conclusions: The TyG-ALT model demonstrates good discriminative ability, calibration, and clinical utility for case-finding and risk stratification of prevalent MAFLD in OSA patients, particularly for identifying high-risk individuals requiring confirmatory imaging.