Triglyceride–glucose index in the development of heart failure and left ventricular dysfunction: analysis of the ARIC study 

Abstract

Aims

We aimed to investigate whether the triglyceride–glucose (TyG) index, an easy-calculated and reliable surrogate of insulin resistance, was associated with the development of heart failure (HF) and left ventricular (LV) dysfunction.

Methods and results

A total of 12 374 participants (mean age: 54.1 ± 5.7 years, male: 44.7%) free of history of HF and coronary heart disease at baseline from the Atherosclerosis Risk in Communities study were included. The TyG index was calculated as ln[fasting triglyceride (mg/dL) × fasting glucose (mg/dL)/2]. The long-term TyG index was calculated as the updated cumulative average TyG index using all available TyG index from baseline to the events of HF or the end of follow-up. We evaluated the associations of both the baseline and the long-term TyG index with incident HF using Cox regression analysis. We also analysed the effect of the TyG index on LV structure and function among 4889 participants with echocardiographic data using multivariable linear regression analysis. There were 1958 incident HF cases over a median follow-up of 22.5 years. After adjusting for potential confounders, 1-SD (0.60) increase in the baseline TyG index was associated with a 15% higher risk of HF development [hazard ratio (HR): 1.15, 95% confidence interval (CI): 1.10–1.21]. Compared with participants in the lowest quartile of the baseline TyG index, those in the highest quartile had a greater risk of incident HF [HR (95% CI): 1.25 (1.08–1.45)]. In terms of LV structure and function, a greater baseline TyG index was associated with adverse LV remodelling and LV dysfunction. Similar results were found for the long-term TyG index.

Conclusion

In a community-based cohort, we found that a greater TyG index was significantly associated with a higher risk of incident HF and impaired LV structure and function.

Graphical Abstract

Increasing triglyceride-glucose index is associated with higher risk of incident heart failure.

Open in new tabDownload slide

See the editorial comment for this article ‘Heart sweet heart: cardiac long-term effects of sugar kisses’, by Marcello Ricardo Paulista Markus and Marcus Dörr, https://doi.org/10.1093/eurjpc/zwac087.

Introduction

Heart failure (HF) is a life-threatening disease and represents the final common pathway for varieties of cardiovascular disease (CVD). The prevalence of HF is continuously increasing worldwide, but effective treatment is still limited.¹ Therefore, given the poor prognosis and high economic burden, early identification of individuals at high risk for HF and timely intervention have crucial importance. Previous studies have found that insulin resistance, a decrease of insulin efficiency with impaired uptake and utilization of glucose, may be involved in the pathogenesis of HF.^2–4 Other studies also observed a significant relationship between insulin resistance and cardiac remodelling and dysfunction.^5,6 However, as the ‘gold standard’ examination for insulin sensitivity, the hyperinsulinaemic-euglycaemic clamp test is time-consuming, costly, and invasive, leading to its limitation in clinical practice.⁷ The triglyceride–glucose (TyG) index, measured by fasting triglyceride and glucose, is a convenient, reliable, and valid surrogate marker of insulin resistance.^8–10 The TyG index has been proven to have a great correlation with hyperinsulinaemia-euglycaemic clamps and other insulin resistance makers, such as the Homeostatic Model Assessment for Insulin Resistance (HOMA-IR).^8,9 Previous studies indicated that the TyG index was independently associated with arterial stiffness,^11,12 diabetes mellitus,¹³ and cardiovascular outcomes.^14–16 However, few studies have been conducted to analyse the role of insulin resistance assessed by the TyG index on HF, and the relationship between the TyG index and left ventricular (LV) remodelling (high LV mass or high LV wall thickness) and function are still unclear.

In this study, we aimed to evaluate the association between the TyG index and the risk of developing HF in a large community-based cohort, using data from the Atherosclerosis Risk in Communities (ARIC) study. We also analysed the association of the TyG index at middle age with late-life LV structure and function. We hypothesized that a higher level of the TyG index would be associated with an increased risk of incident HF and impaired LV structure and function.

Methods

Study design and population

The ARIC study is an ongoing prospective cohort study that was conducted in four US communities (Forsyth County, NC; Jackson, MS; suburban Minneapolis, MN; and Washington County, MD). A total of 15 792 participants aged 45–64 years were recruited between 1987 and 1989 (Visit 1), and subsequent follow-ups were Visit 2 (1990–92), Visit 3 (1993–95), Visit 4 (1996–98), Visit 5 (2011–13); and other examinations are ongoing. A detailed description of the study design has been published previously.¹⁷ The ARIC study has been approved by institutional review boards and registered at clinicaltrials.gov as NCT00005131. All the participants enrolled in the study have provided their written informed consents.

For this current analysis, participants with prevalent HF or coronary heart disease (n = 1332), missing information regarding HF or coronary heart disease (n = 386), and missing data about covariates of interest (n = 1700) at Visit 1 (baseline) were excluded (Figure 1). A total of 12 374 participants were included for the analysis of the association between the baseline and the long-term TyG index with incident HF. Among these participants, a sample of 4889 people attended the 5th visit and underwent an echocardiographic examination.

Data collection and definitions

Demographic data and lifestyle including age, sex, race, drinking, and smoking status were collected by self-report at Visit 1 (1987–89). Drinking and smoking status were classified as current, former, or never. Specifically, alcohol consumption was ascertained by means of an interviewer-administered questionnaire, in which participants were asked if they currently drank alcoholic beverages and, if not, whether they had done so in the past. Blood pressure and heart rate were measured after a 5 min rest by trained interviewers. Medications were ascertained through self-reported usage during the previous 2 weeks. Plasma total cholesterol (TC), triglyceride, high-density lipoprotein cholesterol (HDL-c), glucose, and insulin were measured by blood sample testing, and low-density lipoprotein cholesterol (LDL-c) was estimated by the Friedewald equation.¹⁸ The body mass index (BMI) was calculated as weight (kg) divided by height in squared (m²). Estimated glomerular filtration rate (eGFR) was calculated by the Chronic Kidney Disease Epidemiology Collaboration creatinine equation.¹⁹ The HOMA-IR index was calculated as [fasting glucose (mg/dL)×fasting insulin (μIU/mL)/405].

Assessment of baseline and the long-term triglyceride–glucose index

The TyG index was calculated as ln[fasting triglyceride (mg/dL)×fasting glucose (mg/dL)/2].⁸ In the ARIC study, fasting triglyceride and fasting glucose were measured at each visit. The baseline TyG index was calculated using data from Visit 1 (baseline). The long-term TyG index was calculated as the updated cumulative average TyG index, using all available TyG index measurements from Visit 1 to the outcome events of HF or the end of follow-up.²⁰

Ascertainment of incident heart failure

Participants in the ARIC study undergo active surveillance for incident cardiovascular events according to International Classification of Diseases codes. Incident HF was defined as the first HF hospitalization or death from HF, which were identified from the hospital discharge list or death certificate that showed an HF code (428.x) in any position.

Echocardiographic measurements of left ventricular structure and function

The echocardiographic assessment was performed using uniform equipment at Visit 5 by certified, dedicated, and trained sonographers according to a study-specific protocol.²¹ Quantitative measures of cardiac structure, systolic, and diastolic function were assessed by a central reading centre according to the American Society of Echocardiography recommendations.²² Left ventricular structure indexes included LV mass (LVM), LV mass index (LVMI), relative wall thickness (RWT), and LV end-diastolic volume (LVEDV). LV hypertrophy (LVH) was defined as LVMI >115 g/m² in men or LVMI >95 g/m² in women. Left ventricular systolic functional indexes included LV ejection fraction (LVEF), global longitudinal strain (GLS), circumferential strain (CS). Left ventricular systolic dysfunction was defined as LVEF <50%. Left ventricular diastolic functional indexes included peak early transmitral velocity (E), peak late transmitral velocity (A), and early peak diastolic mitral annular velocity (e′). Left ventricular diastolic dysfunction was defined as E/e′ ≥13.²³

Statistical analysis

We categorized participants by quartiles of the TyG index. Continuous variables were expressed as mean ± SD for normally distributed data or median (interquartile range) for skewed distributed data. Categorical variables were expressed as a number (percentage). To compare the difference among quartiles of the TyG index, the Wilcoxon or Kruskal–Wallis test was used for continuous variables and the χ² test for categorical variables. We also performed Pearson’s correlation analysis to evaluate the relationship between the TyG index and cardiometabolic risk factors [including HOMA-IR, BMI, TC, LDL-c, HDL-c, systolic blood pressure (SBP), diastolic blood pressure (DBP), and eGFR].

Kaplan–Meier estimates were used to evaluate the incidence rate of HF by TyG index quartiles and the differences in estimates were compared by the log-rank test. Multivariable Cox proportional hazard regressions were constructed to assess the association between the baseline and the long-term TyG index with the risk of incident HF by calculating the hazard ratio (HR) and 95% confidence interval (CI). Three models were used to adjust for potential confounders: Model 1 adjusted for age, sex, and race; Model 2 further adjusted for smoking status, drinking status, hypertension, diabetes, anti-hypertensive medication, and lipid-lowering medication; Model 3 further adjusted for BMI, eGFR, SBP, and LDL-c. In addition, restricted cubic splines with three knots were used to examine the non-linear dose–response association between the baseline and the long-term TyG index with incident HF. To evaluate the robustness of the association, we performed pre-specified subgroup analyses stratifying by age, gender, race, BMI, hypertension, and diabetes, respectively. To determine the optimal cut-off value of the TyG index for predicting incident HF, receiver-operating characteristic (ROC) curve analysis was performed.

Echocardiographic data about LV structure and function were compared and displayed as a number (proportion) for categorical variables and mean ± SD for continuous variables. Using multivariable linear regression models, we evaluated the association between both the baseline and the long-term TyG index with echocardiographic measures of LV structure and function. Statistical differences in the association were adjusted for demographics and CVD risk factors: age, sex, race, smoking status, drinking status, hypertension, diabetes, BMI, LDL-c, a lipid-lowering drug, eGFR, SBP, and anti-hypertensive drug. In the presence of non-linear correlation, restricted cubic splines were performed to analyse the associations.

All analyses were performed using Stata (version 14.0, StataCorp LP, College Station, TX, USA) and R statistical software (version 4.0.2, R Foundation for Statistical Computing, Vienna, Austria). A two-sided P value <0.05 was considered statistically significant.

Results

Baseline characteristics

This study included a total of 12 374 participants (mean age: 54.1 ± 5.7 years, male sex: 44.7%), and the baseline characteristics across quartiles of the baseline TyG index are shown in Table 1. Participants with a higher baseline TyG index were older, more often male and white race; they had higher prevalence of comorbidities (including hypertension and diabetes) and were more prone to anti-hypertensive medication and lipid-lowering drugs; they had higher levels of BMI, SBP, DBP, heart rate, fasting glucose, fasting triglyceride, insulin, TC, and LDL-c, but lower levels of HDL-c, eGFR, and physical activity (all P < 0.001). Results were similar when participants were categorized by the long-term TyG index (see Supplementary material online, Table S1). In the Pearson’s correlation analysis, the baseline TyG index was positively correlated with HOMA-IR, TC, LDL-c, SBP, and DBP, but negatively with HDL-c and eGFR (see Supplementary material online, Table S2).

Associations between the baseline and the long-term triglyceride–glucose index with incident heart failure

During a median follow-up of 22.5 years (range from 0.04 to 25.11 years), 1958 cases (15.8%) of HF occurred. In the model that measured the baseline TyG index as a continuous variable, per 1-SD increase (corresponding to 0.60) in the baseline TyG index was associated with a 15.0% higher rate of incident HF [HR = 1.15 (1.10–1.21), P < 0.001], after adjusting for all the covariates in Model 3 (Table 2). When categorizing participants by quartiles of the baseline TyG index, the highest risk of incident HF was observed in the highest TyG index quartile (Quartile 4) (Table 2, Figure 2). In the full-adjusted model, the HR (95% CI) for incident HF comparing the fourth quartile with the first quartile was 1.25 (1.08–1.45, P = 0.003) (Table 2). Similar results were observed for the long-term TyG index. A higher level of the long-term TyG index was associated with an increased risk of incident HF (Table 2, Figure 2). Restricted cubic splines regression models showed there existed linear relationships between the baseline and the long-term TyG index with incident HF, and the risk of HF increased in participants with a higher TyG index both at baseline and at long-term trend (Figure 3).

Results of subgroup analyses are shown in Figure 4. When participants were stratified by gender (male or female), BMI (<30 or ≥30 kg/m²), hypertension (yes or no), and diabetes (yes or no), consistent associations were found between the baseline TyG index and incident HF. However, there was a significant interaction in the age subgroup (P for interaction <0.001) and race subgroup (P for interaction = 0.015). The HR (95% CI) of a higher baseline TyG index on HF was more prominent in participants aged <60 years [HR = 1.36 (1.23–1.50)] than those aged ≥60 years [HR = 1.09 (0.94–1.27)], and more prominent in the black race [HR = 1.21 (1.11–1.32)] than white race [HR = 1.11 (1.03–1.19)]. Similar results of subgroup analyses were observed in the association between the long-term TyG index and incident HF (see Supplementary material online, Table S3).

The area under ROC curves of the TyG index for predicting the occurrence of HF was 0.598 (95% CI: 0.567–0.630; P < 0.001), and the cut-off value of the TyG index was 8.36 (see Supplementary material online, Figure S1).

Associations between the baseline and the long-term triglyceride–glucose index with left ventricular structure and function

A total of 4889 participants with available echocardiographic data were included for the analysis. Echocardiographic characteristics are shown in Table 3. After adjusting for potential covariates, a greater baseline TyG index was significantly associated with altered LV structure with higher RWT (coefficient: 0.005 cm; P = 0.027) and LVM/LVEDV ratio (coefficient: 0.134%; P < 0.001). For LV systolic function, the greater baseline TyG index had more favourable LVEF (coefficient: 0.457%; P = 0.021). For LV diastolic function, the greater baseline TyG index was independently associated with decreasing E/A ratio (coefficient: −0.029%; P = 0.001) and increasing E/e′ ratio (coefficient: 0.510%; P < 0.001). These relationships were stronger for the long-term TyG index, which was further associated with LVMI (coefficient: 1.319 g/m²; P = 0.040) and GLS (coefficient: 0.315%; P < 0.001) (Table 4). However, the relationship between the TyG index and LV structure and function were modified by sex group, with a more significant effect observed in female than male (see Supplementary material online, Tables S4 andS5).

Discussion

In this large community-based cohort study, we found that the incidence of HF was higher in participants with a greater baseline and a long-term TyG index during a median follow-up of 22.5 years. Besides, we found that a greater TyG index was significantly related to adverse LV remodelling and impaired LV function. The results of our study for the first time showed that insulin resistance assessed by the TyG index was associated with the development of HF and impaired cardiac structure and function.

Several prior studies have evaluated the relationship between insulin resistance and incident HF. A prospective cohort performed in Sweden have found that insulin resistance, estimated by the hyperinsulinaemic-euglycaemic clamp technique, could predict HF independently of other risk factors.²⁴ The hyperinsulinaemic-euglycaemic clamp technique is regarded as the gold standard for evaluating insulin sensitivity, but its invasive procedure and expansive cost have limited its use in daily clinical practice. Therefore, a variety of surrogates for insulin resistance have been proposed and compared with the hyperinsulinaemic-euglycaemic clamp technique. Homeostatic Model Assessment for Insulin Resistance, calculated by fasting glucose and insulin, is commonly used for evaluating insulin resistance.²⁵ Several studies have examined the relationship between HOMA-IR and incident HF, but the results have varied in different researches. An analysis of the ARIC study found that the relationship between HOMA-IR and incident HF was non-linear, which was no longer significant at HOMA-IR values >2.5.² In several other studies, no significant association between HOMA-IR and incident HF was observed.^24,26 These studies indicated that HOMA-IR might not be applicable to show the effect of insulin resistance in the development of HF. Recently, the TyG index has been proved to be well related to the hyperinsulinaemic-euglycaemic clamp test and HOMA-IR^8–10 and could be used as a convenient, cost-effective, and reliable surrogate of insulin resistance. Former studies found that an elevated level of the TyG index was significantly associated with a higher risk of CVD.^14,15,27 However, data about the association between the TyG index and the development of HF is still lacking, and whether the TyG index is related to altered LV structure and function is unknown. We tried to figure out the effect of insulin resistance assessed by the TyG index on the development of incident HF and impaired LV structure and function, making it convenient for clinicians to identify patients at high risk of HF.

In the present study, we found that insulin resistance assessed by the TyG index was independently associated with the risk of HF, highlighting the potential utility of using the TyG index as a potential predictor for HF. Furthermore, taking the fluctuation of long-period exposure into account, we calculated the updated mean TyG index at multiple visits as the long-term TyG index, to provide relatively robust results. We further explored the relationship between the TyG index and LV structure and function among the participants with available echocardiographic data at ARIC Visit 5 (n = 4889). We found that, over a median follow-up of 22.5 years, both the baseline and the long-term TyG index were associated with adverse LV remodelling and impaired LV function, and long-standing exposure to a high level of the TyG index had an even stronger impact. Our findings were consistent with previous studies which found that higher insulin resistance was related to impaired cardiac structure and function even though LVEF was preserved.²⁸

A number of mechanisms may explain the association between insulin resistance assessed by the TyG index and the development of HF. First, higher level of the TyG index reflects the status of insulin resistance that is related to metabolic imbalance, including hyperglycaemia and dyslipidaemia, which have a direct effect on cardiomyocyte hypertrophy,²⁹ myocardial contractility,³⁰ and cardiac stiffness.³¹ Second, insulin resistance facilitated the progression of chronic inflammation³² and endothelium dysfunction,³³ promoted the formation of vulnerable plaque,³⁴ and weakened the compensatory mechanisms of the myocardium, making it vulnerable to ischaemia and pressure overload.³⁵ Third, insulin resistance was found to have an effect on cardiac structure and function, correlated with increased LV mass and adverse LV remodelling,^36,37 which was consistent with the findings in our study. However, the mechanism underlying the association between the TyG index and the risk of HF has not been exactly understood. Further studies are required to figure out the implicated mechanism.

The findings of our study are of great clinical importance for preventing the development of HF among the general population. First, estimating insulin resistance by the TyG index, which has been validated against hyperinsulinaemic-euglycaemic clamp measures and HOMA-IR,^8,9 only requires paired concentrations of fasting triglyceride and glucose. Our study found that higher levels of both single-point and the long-term TyG index were associated with an elevated risk of incident HF and worse LV structure and function. Thus, using the TyG index for risk stratification may facilitate targeting those at high risk of HF, highlighting the utility and relatively low cost of using the TyG index as an indicator of incident HF in clinical practice. Second, our study demonstrated that the higher the TyG index is, the higher is the likelihood of incident HF, both in diabetic and non-diabetic participants. This indicated that insulin resistance assessed by the TyG index could be considered for routine use no matter what the diabetic status is. Third, insulin resistance is not irreversible and considered as the early-stage in the development of diabetes.³⁸ Some therapeutic interventions have been proven to improve the prognosis and reduce cardiac remodelling or dysfunction even for those without diabetes. The Insulin Resistance Intervention after Stroke trial found that pioglitazone, an insulin-sensitizing agent, could reduce the risk for myocardial infarction and stroke in patients with insulin resistance but without diabetes.³⁹ Metformin treatment was also found to significantly reduce LVH in patients without diabetes who had insulin resistance or pre-diabetes.⁴⁰ Another study showed that SGLT2 inhibitors may improve the abnormal energy metabolism caused by insulin resistance, along with the improvement of cardiac remodelling.⁴¹ Therefore, using the TyG index to identify insulin resistance may allow early identify of high risk individuals who could benefit from specific interventions for preventing incident HF.

There exists several limitations in this study. First, due to the observational nature of the ARIC trial, the results of our study could not establish a causal association between the TyG index and the incident HF. We have adjusted for a variety of potential risk factors in Cox regression model, including demography, lifestyle, comorbidities, medications, BMI, blood pressure, and renal function, but residual confounders may still remain of concern. Second, the types of incident HF were not recorded, and therefore we cannot examine the association between the TyG index and different types of HF, such as HF with reduced, mildly reduced, or preserved ejection fraction. Third, since a middle-age (45–64 years old) community and only white and black race were recruited in the ARIC study, additional research studies are required in younger participants, older adults, and other ethnicities. Fourth, although the TyG index has been demonstrated to have a great correlation with the gold standard of the hyperinsulinaemic-euglycaemic clamp test and considered as a reliable indicator of insulin resistance in previous studies, we were unable to evaluate the correlation between the TyG index and the hyperinsulinaemic-euglycaemic clamp test, which was not performed among participants in the ARIC study.

In conclusion, the results of our study showed that elevated levels of both the baseline and the long-term TyG index were significantly associated with adverse LV remodelling and LV dysfunction, as well as an increased risk of incident HF. Further research is warranted to evaluate the role of insulin resistance assessed by the TyG index in the development of HF.

Authors’ contributions

X.L., X.Z., and R.H. contributed to the conception or design of the work. X.L., X.Z., R.H., Y.L., X.Y., X.Z., P.X., and M.L. were responsible for the acquisition, analysis, and interpretation of data. R.H., Y.L., and X.Y. drafted the manuscript. A critical revision of the manuscript for important intellectual content was performed by all authors. All authors agreed with the content of the article to be submitted.

Supplementary material

Supplementary material is available at the European Journal of Preventive Cardiology.

Acknowledgements

We thank the staff and participants of the ARIC study for their contributions.

Funding

This work was supported by the National Natural Science Foundation of China [81600206 to Z.X.D. and 81870195 to L.X.X.] and Natural Science Foundation of Guangdong Province [2016A030310140 to Z.X.D.; 2016A020220007 and 2019A1515011582 to L.X.X.]. The supporting organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Availability

Data from the ARIC study are available to all researchers upon application. No additional data are available.

References

Author notes