Cardiovascular disease (CVD) is the leading cause of morbidity and mortality among chronic noncommunicable diseases in China and worldwide
[1]. According to the study
[2], approximately 330 million Chinese suffer from CVD, resulting in significant socioeconomic burdens. CVD, particularly ischemic heart disease and stroke, were the top 2 causes of death in China
[3]. Furthermore, the dual pressures of an aging population and a rising prevalence of metabolic risk factors are expected to exacerbate the burden of CVD in China. Therefore, identifying individuals at high risk for CVD is critical for delivering cost-effective care and optimizing healthcare service planning.
Obesity, diabetes, hypertension, and dyslipidemia are well-established risk factors for CVD
[4-7]. The study
[8] highlights that metabolic risk factors—such as high systolic blood pressure, low-density lipoprotein cholesterol (LDL-C), fasting blood glucose, and body mass index—are the leading contributors to the burden of ischemic heart disease. Effective control of these metabolic risk factors has been associated with reduced ischemic heart disease mortality. The triglyceride-glucose index (TyGI), which combines triglyceride (TG) levels and fasting plasma glucose (FPG), is a simple marker that reflects both dyslipidemia and insulin resistance (IR)
[9-10]. Moreover, cohort studies
[11-15] have identified elevated TyGI as a risk factor for CVD events. The waist-to-height ratio (WHtR), which incorporates waist circumference (WC) and height, is recognized as an anthropometric indicator of central obesity and is commonly used to predict the risk of chronic diseases
[16-17]. A systematic review and meta-analysis involving over 300 000 adults from various ethnic groups demonstrated that WHtR outperforms WC and BMI in detecting CVD risk factors
[18]. The triglyceride-glucose index-to-waist-to-height ratio (TyGI-WHtR) integrates the TyGI and the waist-to-height ratio (WHtR), both of which can be readily obtained in routine clinical practice. Recent studies
[19-21] have suggested that TyGI-WHtR may outperform TyGI alone in predicting IR, diabetes, and cardiometabolic risk due to the close association between obesity and IR. Thus, this study aims to further clarify the association between TyGI-WHtR and the risk of CVD, as well as its subtypes, using a population-based survey from China.
1 Materials and methods
1.1 Study design and study cohort
We used data from the China Health and Nutrition Survey (CHNS), an open, ongoing cohort study in China that began in 1989. A detailed description of cohort enrollment and methodology has been provided elsewhere
[22-23]. CHNS collects demographic characteristics and basic anthropometric measurements for each follow-up of the participants across 9 Chinese provinces (Liaoning, Heilongjiang, Jiangsu, Shandong, Henan, Hubei, Hunan, Guangxi, and Guizhou). This study analyzed the collected data of CHNS between 2009 and 2015. Notably, blood sample data were collected during the 2009 survey.
Initially, 8 049 participants with biomarker information in 2009 and at least 2 surveys between 2009 and 2015 were considered for inclusion. Given that CVD is more prevalent in middle-aged and elderly adults and that individuals over 80 years old may exhibit different pathophysiological disease characteristics
[24], this study only included patients with age of 40 to 80 years in 2009. In addition, this study included participants with sufficient baseline information for the calculation of TyGI-WHtR. A total of 68 participants with a history of myocardial infarction (MI) and 86 participants with a history of stroke, as reported in the 2009 survey, were excluded to control their confounding effects. Finally, 5 395 participants (2 483 males and
2 912 females) were included in the analysis. A flow chart illustrating the process of creating study cohort from CHNS is presented in
Figure 1.
1.2 Calculation of TyGI-WHtR
As previously described
[11], TyGI was calculated using the formula: ln(fasting TG×FPG/2). WHtR was computed as WC divided by height, TyGI-WHtR was derived by multiplying TyGI by WHtR, as outlined in previously published studies
[21].
1.3 Outcome measures
CVD was defined as a self-reported history of MI or stroke, confirmed by a physician’s diagnosis, with the time of diagnosis also recorded. Data from the 2009 CHNS survey were used as the baseline. Follow-up person-time was calculated for each participant from baseline to the date of the participant’s first new-onset CVD diagnosis, the last survey round prior to loss to follow-up, or the final survey round in 2015, whichever occurred first.
1.4 Assessment of covariates
The collected demographic information included age, gender, community type (rural or urban), years of education, marital status, smoking habits (never, former, and current smoker), alcohol consumption (non-drinker and drinker), sleep duration (<7, 7 to 9, and >9 hours), annual per capita household income (<5 000, 5 000 to 10 000, and >10 000 yuan/year), and medical history of hypertension, diabetes, MI, stroke, and medical treatment. Education level was classified lower education (no education or elementary school education), middle education (lower middle school or upper middle school education), and higher education (university/college or above). Dietary data included the types and quantities of food consumed at each meal during the previous 24 hours. Energy intake from food and condiments for each meal was calculated using the China Food Composition Database. Physical activity was classified as domestic, occupational, transportation, and leisure
[25]. Metabolic equivalent (MET)-hours-per-week were used to account for both the intensity and duration of activities
[26]. The reported the average number of hours-per-week spent on these activities over the past year was used to assess physical activity according to literature
[27]. The baseline WC, hip circumference, height, weight, blood pressure, BMI, and estimated glomerular filtration rate (eGFR) (based on CKD-EPI equation
[28]) were included as well. The baseline comorbidities, including hypertension, diabetes, and dyslipidemia, were confirmed using their respective diagnosis guidelines
[29-31].
1.5 Laboratory assays
CHNS used standard protocols and guidelines to conduct laboratory assays for the collected blood samples. Detailed information about the biochemical assessments is available at the following website:
https://www.cpc.unc.edu/projects/china/data/datasets/biomarker-data. All biochemical analyses were conducted in a nationally accredited central laboratory in Beijing (Medical Laboratory Accreditation Certificate: ISO 15189꞉2007)
[32]. The performed laboratory assays included hematologic analysis, FPG, HbA1C, insulin, apolipoprotein A (apo-A) and apolipoprotein B (apo-B), total protein (TP), albumin (Alb), alanine aminotransferase (ALT), urea, serum creatinine, serum uric acid (UA), triglycerides (TG), total cholesterol (TC), LDL-C, high-density lipoprotein cholesterol (HDL-C), Hs-CRP, magnesium, ferritin, and transferrin.
1.6 Statistical analysis
All statistical analyses were performed using the R statistical package version 3.4.3 and EmpowerStats software. Participant characteristics were summarized by TyGI-WHtR quartiles as means±standard deviation (SD) for normally distributed continuous variables, medians (25th percentile, 75th percentile) for non-normally distributed continuous variables, or percentages for categorical variables. Differences in group characteristics were compared using the Kruskal-Wallis test or analysis of variance for continuous variables, depending on their distribution, and the weighted χ2 test for categorical variables. Hazard ratio (HR) and 95% confidence interval (CI) for new-onset CVD and its subtypes (MI and stroke) in relation to TyGI-WHtR (continuous and quartile groups) were calculated using Cox proportional hazards models. Covariate screening and selection were conducted following 2 main principles: 1) Multicollinearity assessment. The variance inflation factor (VIF) was calculated to evaluate multicollinearity among variables. Variables with a VIF greater than 10 were considered to have severe multicollinearity and were excluded from the multivariate regression analysis. 2) Impact evaluation. The effect of adding covariates to the basic model or removing them from the full model on the regression coefficient of the primary predictor (X) was assessed. If the impact exceeded 10% and the P-value for the covariate’s regression coefficient on the outcome variable (Y) was less than 0.10, the covariate was deemed significant. Based on these criteria, 23 covariates were considered (age, gender, education level, income, bedtime, total METs, protein intake, smoking status, systolic blood pressure (SBP), diastolic blood pressure (DBP), urea, UA, eGFR, TC, LDL-C, apo-A, apo-B, Hb, platelet count, transferrin, diabetes mellitus, dyslipidemia, and hypertension). However, the use of medications (including lipid-lowering, antihypertensive, and antidiabetic drugs), the amount of alcohol consumption, and dietary pattern scores were not adjusted for, as these data were unavailable. In the regression models that could make no adjustment, adjustment for age and gender only, and adjustment for all 23 covariates. Sensitivity analysis was performed by using E-values to assess robustness to unmeasured confounding.
The dose-response relationship between TyGI-WHtR and CVD was evaluated using weighted generalized additive models and smooth curve fitting. After adjusting for the same covariates in the linear regression models, two-piecewise linear regression models were used to assess the threshold effect of TyGI-WHtR on CVD. To assess the stability of the derived cut-offs, we performed bootstrap resampling with 1 000 iterations. For each iteration, the median was recalculated, and the 95% CI of the bootstrap distribution was obtained. The original cut-off was considered stable if it fell within the corresponding 95% CI.
Predictive values of various potential predictors [TyGI-WHtR, BMI, WHtR, TyGI, TyGI-BMI, and TyGI-waist-to-hip ratio (WHR)] for endpoints (CVD, MI, and stroke) were estimated using time-dependent receiver operating characteristic (ROC) curves. Kaplan-Meier curves and log-rank tests were employed to analyze unadjusted cumulative hazards across TyGI-WHtR quartiles. The threshold for statistical significance was set at P<0.05.
2 Results
2.1 Baseline characteristics
A total of 5 395 eligible participants with a age of (55.738±10.129) years were included in this study, of whom 46.0% were male. Baseline characteristics of the participants by TyGI-WHtR quartiles are presented in
Table 1. Participants in the higher quartile groups were more likely to have a higher prevalence of hypertension, diabetes, and dyslipidemia; as well as higher levels of BMI, TyGI, WHtR, TC, TG, FPG, hemoglobin A1c, UA, serum ferritin, transferrin, and blood pressure. Conversely, they exhibited lower levels of HDL-C, eGFR, and METs compared to the TyGI-WHtR quartile 1 (Q1) group (all
P<0.05).
2.2 Association between TyGI-WHtR and risk of CVD and its subtypes
The median follow-up period was 72 months (interquartile range: 48 to 72), during which 210 (3.89%) newly diagnosed CVD cases were identified, including 126 (2.34%) strokes and 92 (1.70%) MIs (8 patients had both MI and stroke). Overall, TyGI-WHtR was positively correlated with the risk of new-onset CVD. As shown in
Table 2, in the fully adjusted regression model (Model 3), the
HR per SD increase in TyGI-WHtR for CVD was 1.397 (95%
CI 1.111 to 1.758). The fully adjusted
HRs (95%
CI) for CVD by quartile were as follows: 1.679 (1.004 to 2.807) for quartile 2 (Q2), 2.170 (1.305 to 3.610) for quartile 3 (Q3), and 2.314 (1.335 to 4.011) for quartile 4 (Q4) (
P for trend <0.001). Additionally, the fully adjusted regression model demonstrated a significant association between TyGI-WHtR and MI. For each SD increase in TyGI-WHtR, the
HR for MI was 1.747 (95%
CI 1.241 to 2.460). The fully adjusted
HR for MI in Q4 was 3.137 (95%
CI 1.359 to 7.240). MI risk increased significantly across quartiles (
P for trend =0.003). In contrast, the association between TyGI-WHtR and stroke was not statistically significant in the fully adjusted regression model.
In addition, the results of the smooth curve fitting and generalized additive models, used to characterize the dose-response relationship between TyGI-WHtR and CVD, stroke, or MI, are shown in
Figure 2. Nonlinear associations were observed between TyGI-WHtR and CVD or stroke. For TyGI-WHtR<4.804, the effect size of
HR for CVD was 2.121 (95%
CI 1.350 to 3.333), whereas for TyGI-WHtR >4.804, the effect size of
HR was 1.066 (95%
CI 0.754 to 1.507). The ln(likelihood ratio) was 0.029. Similarly, the inflection point for TyGI-WHtR in relation to stroke was 4.477 (
Table 3). However, the relationship between TyGI-WHtR and the risk of MI was linear, as indicated by a ln(likelihood ratio) of 0.212. Overall, when TyGI-WHtR exceeded the inflection point, the strength of the association between TyGI-WHtR and CVD, MI, and stroke weakened. Bootstrap validation with 1 000 iterations confirmed the stability of the original cut-offs: The bootstrap-derived 95%
CI were 4.607 to 5.011 for CVD, 5.261 to 5.954 for MI, and 4.333 to 4.640 for stroke, with all original cut-offs falling within these intervals.
2.3 Time-dependent ROC curve analysis of different markers for the outcome of CVD and its subtypes
The time-dependent ROC curve results for TyGI-WHtR and other predictors are presented in
Table 4. TyGI-WHtR was the strongest predictor of CVD, with an area under the curve (AUC) of 0.629, sensitivity of 0.773, specificity of 0.432, and a cut-off value of 4.367, outperforming BMI, WHtR, TyGI, TyGI-BMI, and TyGI-WHR (
Table 4). For predicting MI, the AUC of TyGI-WHtR was 0.639, which was lower than that of WHtR but higher than other predictors. In the prediction of stroke, the AUC of TyGI-WHtR was 0.622, ranking second, exceeding the AUC of other indicators except for TyGI.
2.4 Cumulative hazards for CVD and its subtypes stratified by the quartile of the TyGI-WHtR
Cumulative hazard curves for cardiovascular events across TyGI-WHtR quartiles are illustrated in
Figure 3. The incidence of CVD and MI was significantly higher in the Q4 group compared to the other groups. However, for stroke, the highest incidence was observed in the Q3 group.
3 Discussion
Concurrently, the incidence and prevalence of CVD have risen due to population aging and urbanization, particularly among the elderly
[33]. Despite considerable efforts to improve clinical outcomes, CVD remains the leading cause of mortality and morbidity in China
[34-35]. In this study, we identified nonlinear associations between TyGI-WHtR and CVD or stroke but a linear association with MI. Each one-SD increase in TyGI-WHtR corresponded to a 39.7% higher CVD risk and a 74.7% higher MI risk, with a non-significant increase in stroke risk. Higher TyGI-WHtR levels were linked to increased CVD and MI risks. TyGI-WHtR outperformed other indices in predicting CVD, but ranked second for MI and stroke, where WHtR and TyGI performed best, respectively.
These findings align with previous studies, which have established that elevated TyGI and its related indices levels are closely linked to chronic diseases
[36]. A recent cross-sectional study of 11 937 adults from the National Health and Nutrition Examination Survey (NHANES) demonstrated that TyGI and its related indices, including TyGI-WHtR, were significantly associated with CVD mortality, total CVD, congestive heart failure, MI, angina pectoris, and coronary heart disease
[37]. Ye, et al
[38] investigated TyGI and CVD risk in middle-aged and elderly Chinese populations using the China Health and Retirement Longitudinal Study dataset, reporting a J-shaped relationship. However, a systematic review and meta-analysis of 12 cohort studies revealed a linear association between TyGI and the risk of coronary artery disease (CAD) and CVD
[39]. More studies are needed to clarify these relationships.
TyGI-WHtR, which integrates triglycerides, fasting glucose, waist circumference, and height, serves as a surrogate marker for visceral fat and IR, likely explaining its association with CVD. Visceral fat contributes to dyslipidemia, type 2 diabetes, and hypertension-key CVD risk factors. Compared to BMI, TyGI-WHtR may better estimate metabolic health and fat distribution
[40]. Visceral fat and IR also promote free fatty acid release, endothelial dysfunction, oxidative stress, and chronic inflammation
[41-44], and elevate plasminogen activator inhibitor-1, which facilitates thrombosis during atherosclerosis
[45]. The non-linear relationship between TyGI-WHtR and CVD may reflect saturation of lipotoxicity or glucotoxicity effects, desensitization of metabolic pathways, or competing risks, though these hypotheses require experimental validation.
This study has several strengths, including its prospective follow-up design within a diverse population-based cohort, providing robust evidence on the association between TyGI-WHtR and CVD risk in middle-aged and elderly Chinese. The use of standardized questionnaires and detailed hematologic analyses of CHNS enhanced the reliability and accuracy of the collected data. However, several limitations should be noted. First, physician-diagnosed MI or stroke was self-reported without verification against clinical records, which may introduce recall bias. Underreporting of mild events is more common in older adults and likely biases effect estimates toward the null, making our risk estimates conservative. Second, residual confounding cannot be ruled out due to the lack of data on some covariates, including the use of antihypertensives, antiplatelet agents, and lipid-lowering therapies, as well as dietary pattern scores and the quantity of alcohol consumption. These unmeasured factors may have biased the observed associations. Third, the follow-up period of 6 years, while sufficient to observe medium-term outcomes, may not capture the long-term predictive value of TyGI-WHtR. Specifically, we note that the small number of events increases the width of confidence intervals and reduces statistical power for subgroup and secondary analyses. Finally, the findings of this study are limited by the inclusion of a racially homogeneous population, which may restrict generalizability to other ethnicities. Additionally, the inclusion of participants aged 40-80 years excludes younger and older populations, in whom the relationship between TyGI-WHtR and CVD may differ. Extended prospective studies with larger samples and longer follow-up are needed to validate these findings.
In conclusion, there is a positive and nonlinear association between TyGI-WHtR and the risk of CVD among middle-aged and elderly Chinese adults, with an inflection point at 4.804. Combining TyGI and WHtR, TyGI-WHtR offers a simple, accessible, and effective marker for predicting and potentially preventing CVD. However, further research is needed to clarify its long-term predictive value and underlying mechanisms.
,
京公网安备11010202008535号
PDF
(1158KB)
