TyG-WHtR predicts incident type 2 diabetes mellitus in NAFLD: a 12-year prospective cohort study

Background Metabolic composite indices effectively predict type 2 diabetes mellitus (T2DM) in the general population. However, whether these indices retain predictive value in patients with nonalcoholic fatty liver disease (NAFLD)—a population characterized by profound insulin resistance and metabolic heterogeneity—remains unclear. This study aimed to compare the predictive ability of twelve metabolic composite indices for the onset of T2DM in the NAFLD population. Methods We conducted a secondary analysis of 2,370 NAFLD patients from a prospective Japanese cohort. Multivariate Cox proportional hazards models, subgroup analyses, and restricted cubic spline (RCS) regression were used to evaluate the associations between the twelve metabolic composite indices and incident T2DM. Predictive performance was assessed using receiver operating characteristic (ROC) and area under the curve (AUC) analyses, with optimal cut-offs determined using Youden’s index and validated by Kaplan–Meier curves. Results In the fully adjusted model, all twelve indices were significantly associated with new-onset T2DM. The triglyceride–glucose–waist circumference (TyG-WC) index exhibited the greatest increase in the hazard ratio (HR) per 1-standard deviation (SD) [HR: 1.80, 95% confidence interval (CI): 1.56–2.07], followed by the triglyceride–glucose–waist–height ratio (TyG-WHtR) (HR: 1.77, 95% CI: 1.54–2.04) and the triglyceride–glucose–body–mass index (TyG-BMI) (HR: 1.70, 95% CI: 1.47–1.96). Conversely, the visceral adiposity index (VAI) showed the weakest association (HR per 1-SD increase: 1.29; 95% CI: 1.15–1.42). All associations were linear ( P for nonlinearity >0.05 for all indices), and subgroup analyses revealed no significant interactions ( P for interactions >0.05 for all). Notably, the TyG-WHtR exhibited the highest predictive accuracy (AUC = 0.680; optimal cut-off: 4.54), although its discriminative ability was not statistically superior to that of the TyG-WC index ( P = 0.492). Its discriminatory ability remained stable throughout the follow-up period, and AUC values ranged from 0.634 to 0.719. Conclusions All the assessed metabolic composite indices are linearly associated with incident T2DM in NAFLD patients. The TyG-WHtR may serve as a practical tool for T2DM risk stratification in patients with NAFLD, given its moderate yet temporally consistent discriminatory ability and easily obtainable components.