Insulin resistance (IR) and associated disorders, including a wide array of other metabolic derangements, are major underlying abnormalities that contribute to the development of atherosclerotic cardiovascular disease (ASCVD) in diabetic as well as non-diabetic subjects.^1-3 Triglyceride (TG) and triglyceride glucose (TyG) have been regarded as reliable surrogate biomarkers of IR.² Several studies found that the TyG-derived markers that combined obesity indices and the TyG index for IR, such as the TyG-body mass index (TyG-BMI), might be more efficient for XXXX than the TyG index alone.^4-6 Evidence has suggested the potential usefulness of both the TyG and TyG-BMI in predicting cardiovascular risk.^7-10 Regarding the predictive power of the TyG in ASCVD, especially in coronary artery disease (CAD), mild to moderate values have been reported. Therefore, whether novel TyG-derived indices might be of greater value in cardiovascular risk prediction is an important question.

The gold standard measurement for IR is a euglycemic insulin clamp. Fasting glucose on admission is the indicator of both acute stress condition and chronic glycemic levels. This may limit its utility in identifying the true daily level. The A1C-Derived Average Glucose (ADAG) study demonstrated that a single fasting glucose measurement is a poor measure of the true daily level, whereas A1C can be reliably translated into estimated average glucose (eAG) equivalents for monitoring glycemic control in a mixed population of diabetic and non-diabetic subjects.¹¹ The main purpose of adopting eAG was to provide patients a better understanding of their recent average blood sugar levels. However, while eAG may provide more information for glycemic levels, no studies have focused on the TyG and TyG-related indices after the implementation of eAG.

To fill this gap of knowledge, using a nationally representative data from a Chinese cohort of patients with established ASCVD, we included eAG in the TyG and TyG-BMI indices and analyzed the corresponding parameters, termed TyAG and TyAG-BMI, respectively. Our study aims were as follows: (1) to identify the distribution of the two novel indices, TyAG and TyAG-BMI, among patients with ASCVD; (2) to assess the associations of the two pairs of parameters (TyG and TyAG, TyG-BMI, and TyAG-BMI) with baseline levels of risk factors and coronary severity; and (3) to explore the clinical significance and differences of the four indices in the prediction of cardiovascular events (CVEs).

Ethical approval

This study complied with the Declaration of Helsinki and was approved by the Ethics Committee of Fu Wai Hospital and Cardiovascular Institute, Beijing, China. The Institutional Review Board approval number was IRB2012-BG-006. Informed written consent was obtained from all patients enrolled in this study.

Study population

From April 2011 through July 2018, this observational study with a prospective design enrolled a total of 9,944 adults with established ASCVD as described by our previous study.¹² The severity and extent of coronary stenosis were assessed using the number of diseased coronary vessels and the Gensini scoring (GS) system. The ASCVD risk classification and high-risk factors were identified according to the 2018 American College of Cardiology (ACC)/American Heart Association (AHA) cholesterol guideline.¹³ Patients with high levels of TG (≥5.6 mmol/L), hematologic disorders, infectious or systematic inflammatory disease, thyroid dysfunction, severe liver/ renal insufficiency, and/or malignant disease were excluded from the study.

Clinical assessment

Concentrations of TG, total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C) were measured using an automatic biochemistry analyzer (Hitachi 7150) in an enzymatic assay. The concentrations of fasting glucose at admission were measured by the enzymatic hexokinase method. Glycosylated hemoglobin (HbA1c) was measured using the Tosoh Automated Glycohemoglobin Analyser (HLC-723G8; Tosoh). Other related biochemical and hematological indicators were measured using standard tests. The eAG was calculated by inserting observed A1C into the ADAG linear regression equation: eAG (mg/dL)=[28.7×HbA1c (%)]– 46.7; r=0.92. The TyG index was calculated as ln [TG (mg/dL)× fasting glucose (mg/dL)/2], and the TyG-BMI index was calculated as TyG×BMI. The TyAG index was calculated as ln [TG (mg/dL) ×eAG (mg/dL)/2], and the TyG-BMI index was calculated as TyAG×BMI. The cumulative biological risk (CBR) was evaluated using the following 10 biomarkers as described in a previous study¹⁴: (1) systolic blood pressure (SBP), (2) diastolic blood pressure, (3) BMI, (4) HbA1c, (5) TC, (6) HDL-C, (7) TC/HDL-C ratio, (8) C-reactive protein, (9) albumin, and (10) estimated glomerular filtration rate.

The severity and extent of coronary stenosis were assessed using the number of diseased vessels and the GS system. Obstructive CAD was defined as detection of 50% to 99% diameter stenosis in any of the four major epicardial coronary arteries including the left main, left anterior descending, left circumflex, and right coronary artery. Occlusive CAD was defined as 100% occlusion of the above vessels. Multi-vessel CAD was defined as ≥2 diseased vessels as described above or left main-related disease.

Follow-up and endpoints

Follow-up data were described previously.¹² We followed cohort mapping for clinical outcomes until the study end date (February 26, 2019, with a window of 30 days). The primary endpoints were cardiovascular death, nonfatal myocardial infarction, heart failure (HF), and stroke. The secondary end points were unstable angina, unplanned revascularization, or hospitalization in a cardiology ward. Data were obtained from 9,783 patients, and 407 events were documented during an average of 38.5 months of follow-up.

Statistical analysis

Statistical analysis was performed with SPSS version 26.0 software (IBM Corp.). All probability values were 2-sided, and P-values <0.05 were considered statistically significant. Categorical variables are presented as numbers and relative frequencies (percentages), and continuous variables are presented as mean with standard deviation or median with interquartile range (IQR), as appropriate. Categorical variables are presented as number (percentage) and analyzed by chi-square test. Differences between groups were determined using the independent sample t-test or nonparametric test where appropriate. The predictive value for CVEs during the follow-up period was assessed by Cox proportional hazards analysis with hazard ratios (HRs) and 95% confidence intervals. The adjusted factors of age, sex, BMI, SBP, glucose, LDL-C levels, current smoking, and medications at enrollment were entered into the multivariable Cox regression model.

Baseline characteristics of the patients

The baseline biochemical and clinical characteristics of the patients according to recurrent CVEs during follow-up are shown in Table 1. Follow-up data were obtained from 9,783 patients, and the total of 1,651 CVEs comprised 1,244 secondary events and 407 primary events. Patients with CVEs were more likely to be older, women, and non-smokers compared with patients without CVEs. Most metabolic parameters, including SBP, TG, TC, LDL-C, HbA1c, eAG, CBR, and TyAG, were higher in patients with CVEs compared with those without CVEs, but only HbA1c, eAG, and TyAG showed significant differences (P<0.05). However, there was no difference or even lower values of TyG, TyG-BMI, and TyAG-BMI in patients with CVEs compared with those without CVEs.

The distributions of the TyG and TyG-BMI values are demonstrated in Supplementary Fig. 1. After eAG implementation, the overall histogram profile of the TyAG or TyAG-BMI corresponding to the TyG or TyG-BMI, respectively, shifted slightly toward higher values or the right side (TyAG vs. TyG in Supplementary Fig. 1A; TyAG-BMI vs. TyG-BMI in Supplementary Fig. 1B).

Relationships of TyG-derived indices with baseline risk profile

Baseline risk profile was assessed by three parameters: (1) proportion of patients at very high risk (VHR) according to the recent risk stratification by the 2018 ACC/AHA cholesterol guideline,¹³ which was defined as a history of ≥2 major ASCVD events or 1 event and ≥2 high-risk conditions; (2) CBR value, a risk score based on 10 cardiovascular biomarkers; and (3) number of high-risk factors, which was also defined according to the 2018 ACC/AHA cholesterol guideline.

Participants were divided into quartiles of the TyG- and TyAGrelated parameters. The levels of baseline risk profile all tended to increase with higher metabolic parameters (Fig. 1, Supplementary Table 1). The relationships of the TyG- and TyAG-related parameters with baseline risk profile were similar. Notably, patients with TyG-related parameters in the lowest/first quartile (Q1) had higher proportions of VHR (TyG vs. TyAG, 22.7% vs. 21.5%; TyG-BMI vs. TyAG-BMI, 22.9% vs. 22.0%) while patients in the highest/last quartile (Q4) had lower proportions of VHR (TyG vs. TyAG, 28.9% vs. 29.7%; TyG-BMI vs. TyAG-BMI, 29.6% vs. 30.5%) compared with patients with TyAG-related parameters (Fig. 1A). Discrimination of CBR values and number of high-risk factors among quartiles of the TyG or TyAG-related parameters revealed no significant difference (Fig. 1B and C).

Relationships of TyG-derived indices with coronary severity

Coronary severity was assessed by three parameters: (1) proportion of patients with multi-vessel CAD; (2) proportion of patients with occlusive CAD; and (3) Gensini score. The quartiles of the TyG- and TyAG-related parameters were significantly and positively associated with coronary severity (Fig. 2, Supplementary Table 2). Discrimination of the three parameters including proportion of patients with multi-vessel CAD, proportion of patients with occlusive CAD, and Gensini score among quartiles of the TyG- and TyAGrelated parameters revealed no significant difference (Fig. 2A). Patients with TyG-related parameters in the lowest/first quartile (Q1) tended to have higher proportions of multi-vessel or occlusive CAD (TyG vs. TyAG, 65.7% vs. 64.6% or 15.0% vs. 14.7%; TyG-BMI vs. TyAG-BMI, 66.0% vs. 65.7% or 15.4% vs. 14.8%), while patients in the highest/lowest quartile (Q4) had lower proportions (TyG vs. TyAG, 69.6% vs. 70.7% or 21.6% vs. 21.7%; TyG-BMI vs. TyAGBMI, 70.2% vs. 70.9% or 22.1% vs. 22.4%) compared with patients with TyAG-related parameters (Fig. 2A and B), illustrating the potential better discrimination of coronary severity by quartile of Ty- AG-related parameters.

Predictive value of TyG-derived indices for CVEs

Univariate and multivariate Cox regression analyses were performed to determine the predictive values of the TyG- and TyAGrelated parameters for primary CVEs in the overall population (Table 2). The medians of TyG, TyAG, TyG-BMI, and TyAG-BMI for patients with primary, secondary, and no events were 8.84 (IQR, 8.38 to 9.26) vs. 8.82 (IQR, 8.46 to 9.19) vs. 8.83 (IQR, 8.46 to 9.23), 9.13 (IQR, 8.77 to 9.53) vs. 9.08 (IQR, 8.76 to 9.47) vs. 9.08 (IQR, 8.72 to 9.45), 228.01 (IQR, 203.99 to 253.21) vs. 226.67 (IQR, 204.01 to 250.46) vs. 229.49 (IQR, 207.14 to 252.70), and 236.07 (IQR, 212.68 to 260.61) vs. 234.06 (IQR, 211.09 to 257.19) vs. 235.82 (IQR, 213.22 to 259.30), respectively. In the univariate Cox regression analysis, there was no positive association of quartiles of the TyG- and TyAG-related parameters with recurrence of CVEs, except a significant inverse correlation of the TyG in the second quartile (vs. the reference Q1) with CVE risk. In the multivariate Cox regression analysis, the increasing trends of HRs with growth of the eAG-included measure (TyAG and TyAG-BMI) quartiles were apparent upon examining the data as a whole in model 1. Both TyAG-related parameters in the highest quartile (Q4) showed significant HRs for CVEs (TyAG 1.42 [IQR, 1.08 to 1.86], TyAGBMI 1.31 [IQR, 1.01 to 1.73]) while the two TyG-related parameters were not significantly related to CVEs (TyG 1.15 [IQR, 0.89 to 1.50], TyG-BMI 1.25 [IQR, 0.98 to 1.64]) compared with the reference quartile (Q1). Furthermore, in the multivariate Cox regression analysis of model 2 including metabolic risk factors and other important participants as confounding factors, only TyGBMI and TyAG-BMI (Q4 vs. Q1) showed significant HRs, and the latter exhibited a better value for CVE prediction (TyG-BMI 1.59 [IQR, 1.01 to 2.58], TyAG-BMI 1.91 [IQR, 1.16 to 3.15]).

Given that the TyG- and TyAG-related indices can be affected by diabetes mellitus (DM) status and the BMI spectrum, we also analyzed the associations between the indices and CVEs in these subgroups separately. There were no positive association of quartiles of the TyG- and TyAG-related parameters with recurrence of CVEs in any subgroup by DM status or BMI in univariate analysis (Supplementary Table 3). In the multivariate analysis (Supplementary Table 4), the trends of increased HRs with increasing TyGBMI and TyAG-BMI quartiles were apparent upon examining the data as whole in both DM and non-DM subgroups, but significant HRs were only observed in TyAG-BMI among the non-DM subgroups (Q4 vs. Q1 2.66 [IQR, 1.37 to 5.17], Q3 vs. Q1 1.78 [IQR, 1.06 to 2.97]). When the analysis was performed by BMI, the increasing trend of HRs was only found in eAG-adopted measures (TyAG and TyAG-BMI) among subgroups with higher BMI. Significant HRs were observed among the subgroup of BMI >25 kg/m² for TyAG-BMI (Q4 vs. Q1 1.39 [IQR, 1.00 to 1.99], Q3 vs. Q1 1.45 [IQR, 1.01 to 2.11]) and in TyG-BMI (Q4 vs. Q1 1.31 [IQR, 1.00 to 1.97], Q3 vs. Q1 1.45 [IQR, 1.01 to 2.08]).

In the present study, we investigated the TyG-derived indices and developed novel indices, named the TyAG and TyAG-BMI, by replacing fasting glucose with eAG in the TyG and TyG-BMI indices, respectively. Accumulating evidence has indicated the association of the TyG and TyG-related indices with cardiovascular risk.^7-10 Here, for the first time, we systematically evaluated the relationship of the TyG, TyAG, and their related indices with cardiovascular risk in a cohort of Chinese patients with ASCVD. We observed that individuals with high TyG or TyAG including their related indices were more likely to experience unfavorable risk profiles and severe disease extent. Moreover, the TyAG, TyAG-BMI, and TyGBMI indices showed a positive association with future CVEs. Importantly, among these indices, the TyAG-BMI index showed the strongest value in predicting adverse cardiovascular outcomes, independent of conventional cardiovascular risk factors. Patients with a TyAG-BMI in the highest quartile were 1.97 times more likely to experience recurrent events after adjustment for confounding variables compared with those with a TyAG-BMI in the lowest quartile. Importantly, our analyses revealed that the TyAG and its related indices exhibit more information regarding disease progression in ASCVD compared with TyG and its related indices, and the Ty- AG-BMI may provide better predictive performance for future adverse cardiovascular outcomes than other indices.

As a useful surrogate for IR, the TyG index is linked to the development and outcome of ASCVD. In a large study by Sánchez-Íñigo et al.¹⁵ in 2016, in the Vascular Metabolic Clinica Universidad de Navarra (CUN) cohort with a median follow-up of 10 years, positive associations between the TyG index and CVEs including coronary HF, cerebrovascular disease, and peripheral arterial disease, independent of confounding factors, were demonstrated. Since then, the relationship between TyG index and type or severity of ASCVD has continued to be reported.^2,16 The TyG index may help refine cardiovascular risk stratification and enable the administration of more targeted therapeutics or prevention measures. However, contrary to the above studies, several other studies failed to support any association between the TyG index and CVEs.^2,17 Notably, in most studies, TG and fasting glucose were examined only at baseline, regardless of their changes over time, which may affect the predictive value of the TyG index in cardiovascular disease (CVD) risk. Recently, Cui et al.¹⁸ showed that the cumulative TyG index (defined as the summation of the average TyG index for each pair of consecutive evaluations multiplied by the time between these two consecutive visits in years) was independent and better than the TyG index at baseline in predicting CVD. Some investigators proposed that the clinical significance of a postprandial TyG index should be explored.² Because increased postprandial levels of TG and glucose are metabolically abnormal responses to IR, an elevated postprandial TyG index may be associated with a higher risk of DM or CVEs, and this remains to be clarified. In the present study, while a graded association of the TyG index to baseline risk profile and coronary severity was presented, we failed to find an independent association between the TyG index and CVEs. In contrast, the TyAG, a readily available index, was independent and better than the TyG index in predicting CVEs.

In the present study, the TyG-BMI index, which is a combined measurement of TyG and obesity indices, demonstrated a graded and significant association with CVEs. The TyG-BMI index also showed good correlation with IR, and studies reported that the TyG-BMI index was linearly related to ischemic stroke¹⁹ and cardiovascular outcomes.⁷ Obesity has been shown to be positively associated with ASCVD. Accordingly, we hypothesized that the performance of the TyAG-BMI index to predict ASCVD progression would be better than the performance of the TyG-BMI, which was supported by our study findings. Remarkably, our results showed the TyAG-BMI was the best biomarker in predicting CVEs in our population, with the highest HRs for CVEs in the third and fourth quartiles. Kim et al.²⁰ and Cho et al.¹⁰ showed that the combined measurements of TyG and obesity indices predict coronary artery calcification progression, a surrogate marker of the activity of atherosclerosis and a good predictor for potential CVEs, better than other indices of IR such as the TyG index. Similarly, we found better performance of the TyG-BMI index to predict ASCVD progression than the TyG alone. Moreover, our analysis suggested that the TyAG-BMI index appeared superior to the TyG-BMI in predicting cardiovascular risk in a cohort of Chinese patients with ASCVD.

To the best of our knowledge, this is the first study to explore the influence of the TyAG and its related indices on the incidence of CVEs in patients with ASCVD. The current research revealed that the TyAG-BMI index had a graded connection with CVEs in this population, better than other indices of IR such as the TyG, TyAG, and TyG-BMI indices. However, this study has several limitations. First, as these results were derived from cross-sectional examinations, a causal relationship between the parameters and ASCVD progression could not be derived. The mean values of the three parameters (TyG, TyG-BMI, and TyAG-BMI) but not TyAG tended to be lower in the group with primary outcome events compared with the group without events. We also found no significant association of the four parameters as quartiles with the primary outcomes without any adjustment for confounders. Only TyG-BMI and TyAGBMI showed significant positive values in multivariable models. Beyond the statistical significance, these findings might suggest potential clinical applications and utility. Our present study provided a novel perspective supporting the use of TG and glucose reflecting IR for predicting CVEs in a Chinese cohort study. Second, the eAG was calculated from HbA1c but did not consider the average glucose from continuous glucose monitoring data. While mean glucose is potentially more useful than HbA1c in understanding an individual patient’s glycemic control, mean glucose is an average, and different degrees of glycemic variability and many different glycemic patterns could produce similar mean glucose concentrations and similar HbA1c levels. Third, we could not include other IR measurements such as TyG-waist circumference or homeostasis model assessment of IR because of database restrictions. We used the TyG and TyGBMI, two simple, convenient, low-cost, and well-established reliable alternative biomarkers of IR, in the present study. Further studies for detecting other parameters should be explored. Finally, the population in this study comes from a single center of China, limiting the applicability of the results to other communities. More research is required to verify our findings.

In conclusion, the TyG, TyAG, and their related indices were significantly correlated with unfavorite cardiovascular risk profile and coronary severity at baseline in patients with ASCVD. Among those indices, the TyAG-BMI was the best for predicting CVEs in this population. These findings suggest that the TyAG and its related parameters provide more information than TyG and its related indices in disease progression among patients with ASCVD, and that the TyAG-BMI index may be more predictive of future coronary disease and patient prognosis than other indices. We recommend the application of eAG into IR-related biomarkers in cardiovascular risk estimation in real clinical practice and epidemiologic surveys. Future studies should explore novel indices based on the TyG and its related indices to investigate their associations with ASCVD progression.

Supplementary materials can be found online at https://doi.org/10.7570/jomes23055.

The authors declare no conflict of interest.

This work was partially supported by the Capital Health Development Fund (grant number 201614035), the CAMS Major Collaborative Innovation Project (grant number 2016-I2M-1-011), and the CAMS Innovation Fund for Medical Science (grant number 2021-I2M-1-008) awarded to Jian-Jun Li. The authors thank all the staff and participants of this study for their important contributions.

Study concept and design: SL and JJL; acquisition of data: HHL, YZ, MZ, HWZ, CGZ, YLG, NQW, RXX, QD, KFD, and JQ; analysis and interpretation of data: HHL, YZ, MZ, HWZ, CGZ, YLG, NQW, RXX, QD, KFD, and JQ; drafting of the manuscript: SL and JJL; critical revision of the manuscript: SL and JJL; statistical analysis: SL and JJL; obtained funding: JJL; administrative, technical, or material support: JQ and JJL; and study supervision: JQ and JJL.