Triglyceride-glucose–body mass index as an integrated marker for ultrasound-defined diastolic dysfunction: variable importance via machine learning and SHAP analysis in a Chinese community cohort

Introduction The triglyceride-glucose–body mass index, which integrates the triglyceride-glucose index with body mass index, has emerged as a promising marker of metabolic risk for heart failure with preserved ejection fraction; however, its association with subclinical diastolic dysfunction remains underexplored. Methods This cross-sectional study enrolled 1,467 participants from a Chinese community-based population who underwent echocardiographic evaluation and metabolic profiling. The triglyceride-glucose–body mass index was calculated as the natural logarithm of triglycerides multiplied by fasting blood glucose divided by 2, then multiplied by body mass index. Ultrasound-defined diastolic dysfunction was diagnosed according to the 2025 American Society of Echocardiography criteria. Multivariable logistic regression, restricted cubic spline analysis, and machine learning were employed to evaluate the association. Results The prevalence of ultrasound-defined diastolic dysfunction was 33.6%. The optimal Youden-derived triglyceride-glucose–body mass index cutoff was 220.000, categorizing participants into low ( n = 543 ) and high ( n = 924 ) groups. In the fully adjusted model, each 1-standard deviation increase was not significantly associated with ultrasound-defined diastolic dysfunction in continuous analysis (odds ratio=1.35; 95% confidence interval: 0.42–4.40; P = 0.616 ); however, the binary comparison showed a significant association (odds ratio=1.60; 95% confidence interval: 1.11–2.31; P = 0.013 ). Restricted cubic spline analysis revealed a significant overall association ( P overall < 0.001 ) without significant nonlinearity ( P nonlinear = 0.194 ). Discussion In this cohort, the triglyceride-glucose–body mass index demonstrated a significant independent association with ultrasound-defined diastolic dysfunction in the binary analysis, supporting its potential as a screening index for subclinical diastolic dysfunction, although external validation in independent populations is needed.