Anthropometry-Adjusted TyG Indices Predict Diabetes, Graft Failure, and Mortality in Kidney Transplant Recipients

Figures

• 

  Figure 1. Overview and distribution of TyG-derived indices in kidney transplant recipients. (A) Schematic classification of the 15 TyG-derived indices into 3 domains: traditional/metabolic indices, fat-distribution/body-shape-adjusted indices, and variability-based indices. (B) Ridgeline diagrams illustrating the standardized (z-score) distributions of the 15 TyG-derived indices by category. For TyG-HTGW, only the positive component among participants meeting the HTGW definition is displayed in the ridgeline density plot; participants not meeting the HTGW definition were assigned a value of 0 in the analyses. Abbreviations: ABSI, a body-shape index; HTGW, hypertriglyceridemic waist; NODAT, new-onset diabetes after transplantation; TyG, triglyceride-glucose index.

• 

  Figure 2. Multimethod selection and time-dependent discrimination of TyG-derived indices. (A) Variable importance of 15 TyG-derived indices for death, graft failure, and NODAT derived from random survival forest (RSF) models. (B) Variable importance rankings from XGBoost survival models across the 3 outcomes. (C-E) Time-dependent AUC curves (with 95% CI) for the top 8 TyG-derived indices predicting NODAT (C), graft failure (D), and all-cause mortality (E), demonstrating outcome-specific and time-varying discriminative performance, with TyG-WWI and TyG-WHR showing consistently superior discrimination.

• 

  Figure 3. Kaplan–Meier survival analyses of selected TyG-derived indices across outcomes. (A) Kaplan–Meier curves for NODAT stratified by optimal cutoff values of TyG-WWI, TyG-WHR, TyG-BMI, and TyG_mean. (B) Kaplan–Meier curves for graft failure stratified by optimal cutoff values of TyG-WWI, TyG-WHR, TyG-BMI, and TyG_mean. (C) Kaplan–Meier curves for all-cause mortality stratified by optimal cutoff values of TyG_WWI, TyG-WHR, TyG-BMI, and TyG_mean. P values were derived from log-rank tests, and numbers at risk are shown below each plot.

• 

  Figure 4. Combined modeling and comparative discrimination of TyG-derived indices across outcomes. (A) Summary of TyG-derived indices selected by SHAPBoost across NODAT, graft failure and all-cause mortality, with bubble size indicating C-index and red dots denoting selection by SHAPBoost. (B-D) Cross-validated time-dependent AUC curves comparing XGBoost-based combined models and individual TyG indices for NODAT (B), graft failure (C), and all-cause mortality (D).