Association of triglyceride–glucose index and traditional risk factors with cardiovascular disease among non-diabetic population: a 10-year prospective cohort study

From 2005 to 2006, we collected data of 12 communities aged 35 to 70 years who had lived in Eastern China for at least five years through a random, multi-stage and cluster sampling scheme. A total of 7082 participants without a history of CVD or diabetes mellitus (DM) completed the questionnaires. 987 individuals were lost to follow-up, and a total of 6095 individuals involving 2990 men (49.1%) and 3105 women (50.9%) were included in this study (Fig. 1).

This study was approved by the Ethics Review Committee of the Shandong Academy of Medical Sciences (Approval No. 202111120194) and complied with the Declaration of Helsinki. Written informed consent were provided from all participants.

Standardized questionnaire by trained staffs were used to collect information of participants about clinical and demographic characteristics including age, sex, medication, lifestyle and education. Education level was classified as primary or less, secondary, and trade, college or university. Tobacco and alcohol use were classified as never, former and current. Energy intake from fat and carbohydrate were split into tertiles (Energy intake from fat: tertile 1, ≤ 13.41%; tertile 2, 13.41% to 17.74%; tertile 3, > 17.74%. Energy intake from carbohydrate: tertile 1, ≤ 59.72%; tertile 2, 59.72% to 68.43%; tertile 3, > 68.43%). Physical activity was assessed by the International Physical Activity Questionnaire (IPAQ)[10] and low physical activity was defined as < 600 metabolic equivalent task (MET)\(\times\) minutes per week or < 150 min per week of moderate intensity physical activity. Body mass index (BMI) and waist hip ratio (WHR) were calculated respectively by the 2 formulas: weight(kg)/height(m)² and waist circumference(cm)/hip circumference(cm). Hypertension was defined as SBP ≥ 140 mmHg or DBP ≥ 90 mmHg. Obesity was defined as BMI ≥ 30 kg/m² as previous described [11]. Before blood collection, participants were forbidden to drink and exercise violently within 24 h and were forbidden to change their eating habits. EDTA anticoagulation tubes were used to collect blood sample from venous blood of the anterior elbow after at least eight-hour fasting in the morning at 7 to 9 o’clock. Then these blood samples were sent to the laboratory for testing including fasting plasma glucose (FPG), total cholesterol (TC), triglyceride (TG) and high-density lipoprotein cholesterol (HDL-C) using an autoanalyzer. Low-density lipoprotein cholesterol (LDL-C) was estimated using Friedewald’s formula [12]. Blood pressure (BP) was measured by the sphygmomanometer according to standard steps with no smoking and no coffee for half an hour and resting in a quiet environment for at least 5 min.

The TyG index was calculated by the formula [ln (fasting triglyceride [mg/dL] × fasting glucose [mg/dL]/2)] [13]. The primary endpoint was CVD and the secondary endpoints were coronary heart disease (CHD) and stroke. CHD was defined as myocardial infarction, angina pectoris and angiography-proven coronary heart disease, and stroke included ischemic, hemorrhagic, or unspecified stroke. CVD was defined as the composite of CHD and stroke.

SPSS version 25.0 (SPSS, Chicago, IL) and R software version 4.2.0 (R Foundation for Statistical Computing) were used for statistical analysis. Continuous variables were described by mean and standard deviation (SD) and categorical variables were described by frequency and percentage. One-way ANOVA for continuous variables and the Pearson chi-square test for categorical variables were used for the comparison of baseline data of different TyG quartile groups. Pearson or Spearman correlation analysis was used to evaluate the association between the TyG index and cardiovascular risk factors. During the total follow-up period, the incident rates of the three outcomes (CVD, CHD, stroke) for each TyG quartile group were calculated. Kaplan–Meier analysis was used to draw the incidence curves by R language and the Log-rank test was used for the p value. In order to clarify the independent predictive value of the TyG index, we developed multivariate Cox proportional hazards regression and controlled the confounding factors by building 6 regression models. Model 1 was unadjusted; Model 2 was adjusted for age and gender; Model 4 was fully adjusted for age, gender, WHR, tobacco use, alcohol use, education, physical activity, hypertension, BMI, LDL-C, intake of fat and carbohydrates, use of antihypertensive drugs and antilipemic drugs. Moreover, we used these fully adjusted variables to develop multivariate Cox regression analyses for the incidence of CVD, CHD and stroke. And Model 3, 5, 6 were respectively partially adjusted for the variables statistically associated with the incidence of CVD, CHD and stroke. P value less than 0.05 was considered to be statistically significant.

Table 1 showed the baseline characteristics of the 6095 participants in the cohort study according to the TyG quartiles. Among them, the average age of the participants was 48.69 ± 9.95 years, and 49.1% were men. Individuals in the higher TyG index quartile were more likely to be with current alcohol use, obesity, hypertension, use of antihypertensive drugs and to have a higher age, BMI, WHR, SBP, DBP, FPG, TC and TG (all p ≤ 0.001). And the differences of LDL, HDL, sex, education levels, the intake of fat and carbohydrates in the four groups were significant (all p < 0.01). In terms of endpoints, significantly increased prevalence of CVD and CHD was found in the Q4 group than in the other groups (all p < 0.001) (Table 1).

The correlation between the TyG index and cardiovascular risk factors examined by Spearman or Pearson correlation analysis was displayed in Table 2. Among all participants, the TyG index was positively correlated to age (r = 0.198, p < 0.001), BMI (r = 0.342, p < 0.001), WHR (r = 0.178, p < 0.001), SBP (r = 0.176, p < 0.001), DBP (r = 0.211, p < 0.001) and TC (r = 0.420, p < 0.001). No significant correlation between the TyG index and LDL-C or HDL-C was observed (Table 2).

The association of incidence CVD, CHD and stroke with covariates was shown in Additional file 1: Table S1. The covariates including age, tobacco use, alcohol use, education, physical activity, hypertension, BMI, intake of carbohydrates, use of antihypertensive drugs, and use of antilipemic drugs showed statistically significant association with the incidence CVD(p < 0.05). Likewise, age, tobacco use, physical activity, hypertension, BMI, LDL-C, intake of carbohydrates, use of antihypertensive drugs, and use of antilipemic drugs increased the risk of CHD incidence. Moreover, age, gender, tobacco use, alcohol use, education, hypertension, use of antihypertensive drugs, and use of antilipemic drugs were statistically significant related to the incidence stroke(p < 0.05). (Additional file 1: Table S1).

Over a median follow-up of 10.58 years (interquartile range: 9.92 to 10.75), 357 (5.9%) CVD, 224 (3.7%) CHD and 151 (2.5%) stroke were identified. The Kaplan–Meier curves for the cumulative incidences of CVD, CHD and stroke grouped by the TyG index quartiles were shown in Fig. 2A–C. The probability of cumulative incidences of CVD and CHD was significantly higher in patients with a higher TyG index than in those with a lower TyG index (p < 0.001). Although the incidence of stroke was likely to increase with increasing quintiles of the TyG index, no statistical significance between the TyG index and incident stroke was found (Fig. 2).

As shown in Table 3, the hazard ratios (HR) (95% CI) calculated respectively in Model 1, 2, 3 and 4 for the incidence of CVD with per SD increase in the TyG index was 1.276 (1.159–1.405), 1.204 (1.087–1.334), 1.137 (1.019–1.269) and 1.145 (1.025–1.279) while with per unit increase in the TyG index was respectively 1.463 (1.259–1.702), 1.337 (1.140–1.568), 1.222 (1.030–1.450) and 1.236 (1.040–1.468). Compared with Group Q1, population in Group Q4 had a twofold higher risk of CVD in Model 1 [2.174 (1.607–2.941)]. The age- and sex-adjusted HR for the incidence of CVD increased by 70.7% (95%CI 1.260–2.311) in Group Q4. In partially adjusted Model 3 and fully adjusted Model 4, with the highest TyG index in Group Q4, the possibility of the incidence of CVD risk respectively increased by 43.5%[1.435 (1.040–1.981)] and 48.4%[1.484 (1.074–2.051)]. The increased risk of CVD from Group Q1 to Group Q4 in Model 1, 2, 3 and 4 was statistically significant (p for trend < 0.05).

In Tables 4 and 5, with per SD increase in the TyG index, the fully adjusted HR (95%CI) for risk of CHD and stroke was respectively 1.188 (1.033–1.366) and 1.122 (0.946–1.330). Population in Group Q4 had an increased risk of the incidence for CHD [unadjusted HR 2.768 (1.856 to 4.127)] and stroke [unadjusted HR 1.706 (1.076 to 2.705)] compared to Group Q1. After adjustment for age and sex, population in Group Q4 had 2.2 times the risk of incidence of CHD compared to the population in Group Q1 (95%CI 1.453–3.243). After adjustment for partial covariates in Model 5 and 6, the risk in Group Q4 for the incidence of CHD and stroke was respectively 1.663 (1.091–2.534) and 1.298 (0.811–2.078). After adjustment for all covariates, population in Group Q4 had 1.687 times the risk of CHD incidence compared to the population in Q1 (95% CI 1.105–2.575). However, the increased risk of stroke from Group Q1 to Group Q4 in Model 1,2,3 and 4 showed no statistical significance (all p for trend > 0.05) (Tables 4, 5).

To explore the association between the TyG index and CVD, CHD and stroke incidence in detail, subgroup analyses stratified by sex, age, hypertension and BMI were performed. The association of the TyG index with the risk of CVD, CHD and stroke showed consistence and stability across subgroups.

In the light of baseline characteristics and cardiovascular risk factors, we classified the participants and calculated the HRs (95%CI) and p values for interaction of each group for CVD, CHD and stroke with fully adjusted in Model 4 in Table 6. We found that when cardiovascular risk factors not presented, the relationship between the TyG index and CVD, CHD, stroke was generally consistent. Subgroup analysis demonstrated higher predictive values of the TyG index for CVD, CHD and stroke in participants younger than 60 years, female, without hypertension or obesity. There was no significant interaction between the TyG index and sex, age, hypertension or BMI (all p > 0.05) (Table 6).

As shown in Table 7, C-statistic, NRI and IDI were used to calculate the incremental predictive value of the TyG index for CVD, CHD and stroke and so as to evaluate the influence of the TyG index. We concluded that the addition of the TyG index into the baseline risk model, which contained age, gender, WHR, tobacco use, alcohol use, education, physical activity, hypertension, BMI, LDL-C, intake of fat and carbohydrates, use of antihypertensive drugs and antilipemic drugs, improved significantly the prediction probability for CVD, CHD and stroke. In light of C-statistic, after adding the TyG index, the prediction probability of the the baseline risk mode was improved significantly with C-statistic increased from 0.730 to 0.731 (p = 0.016) for CVD and C-statistic increased from 0.731 to 0.733 (p = 0.017) for CHD. Likewise, according to continuous NRI, the TyG index significantly improved the prediction for CVD [continuous NRI (95% CI) 0.1587(0.0518–0.2655), p = 0.004], CHD [continuous NRI (95%CI) 0.1637(0.0304–0.2971), p = 0.016] and stroke [continuous NRI (95%CI) 0.1818(0.0203–0.3432), p = 0.027]. However, IDI did not show statistically significance in improving the prediction of CVD, CHD and stroke. (Table 7).