Linear dose-response relationship between TyG-WC index and diastolic dysfunction: insights from restricted cubic spline and machine learning analysis.
While diastolic dysfunction (DD) is a critical heart failure precursor linked to metabolic stress, the specific association between the promising triglyceride-glucose waist circumference (TyG-WC) index and DD remains insufficiently characterized under contemporary diagnostic standards. This study aimed to evaluate the association between the TyG-WC index and the risk of DD in a community-based population of middle-aged and older adults.
In this cross-sectional study of 1,413 participants, DD was adjudicated according to the age-stratified echocardiographic criteria. Multivariable logistic regression and restricted cubic splines (RCS) were employed to assess independent associations and dose-response relationships. Furthermore, machine learning techniques, including LASSO regression for feature selection and Random Forest models interpreted by Shapley Additive exPlanations (SHAP), were utilized to quantify the predictive contribution of TyG-WC.
After full adjustment for demographics, blood pressure, and clinical comorbidities, TyG-WC was independently associated with an increased risk of DD (OR = 1.002, 95% CI: 1.001-1.004, P < 0.001). Individuals in the high TyG-WC group (≥776.5) faced a significantly higher risk of DD compared to the low group (OR = 1.656, 95% CI: 1.266-2.166, P < 0.001). RCS analysis confirmed a significant linear dose-response relationship (P overall<0.001), indicating that the probability of DD escalates with increasing TyG-WC. The full predictive model achieved an AUC of 0.687, demonstrating incremental value beyond traditional risk factors. Subgroup analyses demonstrated the robustness of these findings, with significant interactions observed for hypertension and diabetes status. Machine learning identified TyG-WC as an independent contributor for DD, with SHAP values consistently linking higher index levels to increased risk.
Elevated TyG-WC is independently associated with an increased risk of DD. As a cost-effective and easily accessible marker of cardiometabolic stress, TyG-WC may serve as a practical screening tool for the early identification and risk stratification of subclinical heart failure.
In this cross-sectional study of 1,413 participants, DD was adjudicated according to the age-stratified echocardiographic criteria. Multivariable logistic regression and restricted cubic splines (RCS) were employed to assess independent associations and dose-response relationships. Furthermore, machine learning techniques, including LASSO regression for feature selection and Random Forest models interpreted by Shapley Additive exPlanations (SHAP), were utilized to quantify the predictive contribution of TyG-WC.
After full adjustment for demographics, blood pressure, and clinical comorbidities, TyG-WC was independently associated with an increased risk of DD (OR = 1.002, 95% CI: 1.001-1.004, P < 0.001). Individuals in the high TyG-WC group (≥776.5) faced a significantly higher risk of DD compared to the low group (OR = 1.656, 95% CI: 1.266-2.166, P < 0.001). RCS analysis confirmed a significant linear dose-response relationship (P overall<0.001), indicating that the probability of DD escalates with increasing TyG-WC. The full predictive model achieved an AUC of 0.687, demonstrating incremental value beyond traditional risk factors. Subgroup analyses demonstrated the robustness of these findings, with significant interactions observed for hypertension and diabetes status. Machine learning identified TyG-WC as an independent contributor for DD, with SHAP values consistently linking higher index levels to increased risk.
Elevated TyG-WC is independently associated with an increased risk of DD. As a cost-effective and easily accessible marker of cardiometabolic stress, TyG-WC may serve as a practical screening tool for the early identification and risk stratification of subclinical heart failure.
Authors
Xia Xia, Wang Wang, Huang Huang, Bai Bai, Liu Liu, Cai Cai, He He, Guo Guo, Xie Xie, Wang Wang
View on Pubmed