BACKGROUND: The predictive value of triglyceride-glucose index (TyG) for cardiovascular disease (CVD) in the US elderly diabetic patients is ambiguous. This study aimed to investigate the association between TyG index and the risk of CVD in an older US population with diabetes. METHODS: The study examined data from the 2007-2016 National Health and Nutrition Examination Survey (NHANES). Univariate and multivariate regression analysis models were obtained to explore the association between baseline TyG index and the risk of CVD. Non-linear association were investigated using restricted cubic spline (RCS) regression. Subgroup analyses and interaction tests were constructed and a sensitivity analyses was carried out. The 10 - year CVD risk were evaluated via the Framingham Risk Score (FRS). Mediation analysis explored the mediating role of glycated hemoglobin in the above relationships. RESULTS: A total of 2987 subjects were included (977 CVD patients and 2010 non-CVD persons). CVD patients had higher TyG values (9.01 ± 0.58 vs. 8.94 ± 0.56, P=0.003), and the prevalence of CVD increased with TyG index (P=0.015). In a multifactorial regression model with gradual adjustment for all covariates, the risk of CVD associated with TyG increased by 48.0% in the highest quartile group (OR 1.480, 95% Cl 1.171-1.871, P=0.001). The RCS curves showed a U-shaped association between TyG index and CVD risk (P for overall=0.013, P for nonlinear=0.043). Subgroup analyses showed that in the highest quartile group, individuals with body mass index (BMI) ≥24 kg/m CONCLUSION: In a sample of US elderly diabetic patients, there is the U-shaped association of TyG index with CVD risk. This implies that TyG index can be regarded as an extremely important predictor for screening people at high risk of cardiovascular disease among elderly diabetic patients.