Triglyceride-glucose index in predicting the risk of new-onset diabetes in the general population aged 45 years and older: a national prospective cohort study

Abstract

Objective

Insulin resistance (IR) is often present in diabetes, which imposes a heavy burden on the prevention and treatment of diabetes. Triglyceride glucose index (TyG) is simple, reliable and reproducible in detecting IR, and has great advantages in predicting the risk of diabetes. The aim of this study was to analyze the potential association between TyG and the risk of diabetes in Chinese middle-aged and older adults using a prospective cohort study design.

Methods

This study used longitudinal data from five waves of the China Health and Retirement Longitudinal Study (CHARLS) conducted in 2011, 2013, 2015, 2018, and 2020, involving 5886 participants. We used Cox proportional risk regression modeling to investigate the association between TyG index and the risk of new-onset diabetes, and decision tree analysis to identify high-risk groups for diabetes. Finally, ROC curves were applied in order to construct a predictive model for diabetes.

Results

A total of 1054 (17.9%) participants developed diabetes throughout the 9-year follow-up. Our study utilized a multivariate Cox proportional risk regression model and found a significant correlation between TyG index and diabetes risk. The analysis also revealed a nonlinear relationship between TyG index and diabetes risk.Receiver Operating Characteristic(ROC) curve analysis showed that the Area under the curve(AUC) area of TyG index in predicting the risk of new-onset diabetes was 0.652 (P < 0.05).

Conclusions

TyG index can be used as a risk factor for predicting new-onset diabetes in the middle-aged and elderly population in China. In addition, there was a nonlinear relationship between TyG index and diabetes. Improving insulin resistance by regulating glucose and lipid metabolism plays an important role in the primary prevention of diabetes.

Introduction

Diabetes, as a worldwide problem and the most prevalent metabolic disease, is a heterogeneous group of metabolic disorders characterized by chronic hyperglycemia due to long-term interaction between genetic and environmental factors [1, 2]. With the improvement of people’s living standards and changes in lifestyle, diabetes has become the third largest non-communicable disease after cardiovascular and oncological diseases [3]. 2021 International Diabetes Federation shows [4, 5]: the number of people suffering from diabetes worldwide is up to 537 million, affecting about 1/10 of adults; among them, the number of people suffering from diabetes in China is about 140 million, which is more than 26% of the total. Despite this high percentage of detected diabetes prevalence, a very large number of hyperglycemic patients may not be aware that they have diabetes. At present, due to the gradual deterioration of the physical condition of middle-aged and elderly people and poor long-term compliance self-management, the incidence of diabetes is increasing year by year in the middle-aged and elderly populations, which seriously jeopardizes the physical and mental health and quality of life of the nation, and increases the economic burden of the family and society [6,7,8]. Therefore, identification and management of risk factors are crucial in preventing the onset of diabetes and alleviating socioeconomic pressure.

Insulin resistance, also known as impaired insulin sensitivity, is the result of reduced insulin signaling in response to blood glucose levels [9]. A growing body of evidence suggests that insulin resistance is not only an important risk factor for the development of diabetes, but also an early sign of diabetes, representing a key component of diabetes progression from multiple perspectives [10, 11]. Therefore, regular assessment of insulin resistance may be excellent for early prevention, development of diabetes and control of disease progression.The TyG index, as a model for homeostatic assessment of insulin resistance, is characterized by simplicity, reproducibility and reliability [12]. Many studies have confirmed that this index shows excellent performance in predicting insulin-related diseases [13, 14]. In addition, further studies have found that the TyG index is significantly associated with diabetes.

Recent cross-sectional studies [15, 16] have shown an association between TyG index and diabetes risk. However, due to the limitations of cross-sectional study design, the relationship between TyG index and diabetes risk remains unclear. Therefore, to investigate the relationship between TyG index and diabetes risk in the middle-aged and elderly Chinese population, we conducted a prospective cohort study using data from the 2011–2020 China Health and Retirement Longitudinal Study (CHARLS).

Methods

Study design

The current study was a cohort study using CHARLS data from 2011 to 2020.TyG index was considered as the main independent variable. The prevalence of diabetes was coded as binary (diabetes = 1, no diabetes = 0) as the outcome of interest.

Data sources and study population

CHARLS is a national cohort study assessing the economic, social, and health status of people aged 45 years and older to contribute to the interdisciplinary study of population aging in China [16].The initial survey was conducted in 2011 using multistage probability-proportional-to-size sampling, which included more than 17,000 people in 10,257 households in 450 communities in 150 counties in 28 provinces. Thereafter, at two-year intervals, follow-up interviewers collected data through standardized questionnaires administered by personal interviews, asking questions on demographics, health status, functioning, diagnosed chronic diseases, and health-related behaviors (such as smoking and alcohol consumption status) of the respondents and their families on the line. In addition, these interviewers were equipped with equipment that allowed them to take physical health measurements of the participants to collect their height, weight, blood pressure, and so on. Ethical approval for the study was obtained from the Biomedical Ethics Committee of Peking University, China, and all participants signed a written consent form. The datasets related to this study are publicly accessible on the official website of the CHARLS program. Notably 2011 and 2018 subjects were asked to visit the nearest township hospital or the local Centers for Disease Control and Prevention for a comprehensive health assessment. A certified professional collected 8 mL of fasting blood samples, which were stored at -20 °C and sent to Beijing for detailed analysis at the Chinese Center for Disease Control and Prevention.

Our survey study drew on data from CHARLS surveys conducted in 2011, 2013, 2015, 2018, and 2020. Of the baseline survey participants in 2011, 17,708 completed physical examinations and questionnaire assessments. To refine our study group, we used several exclusion criteria: age < 45 years (648), incomplete information on height, weight, triglycerides, and fasting glucose (8514), BMI extremes (BMI < 15 or > 55) (38), those who had acquired diabetes at the time of the baseline data (1364), and those whose diabetes status was unknown during follow-up (1258). After applying these criteria, our final analysis included 5886 participants. The detailed methodology of our participant selection process is shown (Fig. 1):

Fig. 1

Flowchart of participant selection process for the Charls cohort study from 2011 to 2020 (n = 5886)

Calculation of TyG index

The specific procedure for defining the TyG index in this study was as follows: the TyG index = In[FPG (mg/dl) × TG (mg/dl)/2.

Assessment of new-onset diabetes

The diagnosis of new-onset diabetes was based on self-reported data and laboratory tests. When the interviewer asked, “Have you been diagnosed with diabetes by your doctor?“, the respondents answered “yes” and they were categorized as diabetes. Subjects were also categorized as diabetes if they had a fasting blood glucose level of ≥ 126 mg/dl and/or an HbAlc level of ≥ 6.5% on a blood test, and subjects with diabetes in 2011 were excluded if they had been diagnosed with diabetes after this period until the 2020 follow-up period and were included in the study according to our definition of a diabetes. The interval between assessment and final between diabetes onset and baseline assessment was calculated to determine the time to diabetes. For patients who did not report diabetes during follow-up, we determined the duration of follow-up based on the interval between the baseline assessment and the final investigation.

Covariates

Covariates were screened based on sociodemographic characteristics, lifestyle behaviors, current health status, and clinical expertise. Categorical variables: gender, marriage, place of residence, smoking and drinking status, education, hypertension, dyslipidemia. Continuous variables: age, White blood cell(WBC), Mean Corpuscular Volume(MCV), Hematocrit(HCT), Platelets(PLT), Hemoglobin(Hb), Fasting plasma glucose(FBG), C-Reactive Protein(CRP), Glycated hemoglobin(GHb), Uric acid(UA), Total cholesterol(TC), Triglyceride(TG), High-dendity lipoproteins cholesterol(HDL-C), Low-dendity lipoproteins cholesterol(LDL-C), Systolic blood pressure(SBP), Diastolic blood pressure(DBP), Body mass index(BMI), TyG.

To examine the effect of age, all participants were divided into two groups: <60 years and ≥ 60 years. Educational level was categorized as below primary education, primary education, and secondary and above education. Marital status was categorized as married or other. Residence status was categorized as urban and rural. Smokers were defined as never smokers, past smokers and current smokers. Similarly, alcohol drinkers were defined as never having consumed alcohol, having consumed alcohol in the past and currently consuming alcohol. According to the Chinese Adult Weight Standards, BMI is categorized into the following groups: underweight(BMI < 24), normal weight(BMI:24-27.9), and overweight and obese(BMI > 28).

Hypertension was defined as systolic blood pressure ≥ 140 mmHg and diastolic blood pressure ≥ 90 mmHg, taking into account current use of antihypertensive medications or self-reported history of hypertension, according to reference standards set by the World Health Organization. Dyslipidemia was defined as total cholesterol level ≥ 240 mg/dL, triglyceride level ≥ 150 mg/dL, LDL cholesterol level ≥ 160 mg/dL, and HDL cholesterol level < 40 mg/dL, as well as current use of lipid-lowering medications or self-reported history of dyslipidemia.

Treatment of missing data

In our study, missing data on gender status (116, 1.971%), alcohol consumption status (1, 0.017%),smoking status (11, 0.187%), educational status (2, 0.034%), hypertension (19, 0.323%), dyslipidemia (56, 0.951%), SBP (41, 0.697%), DBP (42. 0.714%), WBC (43, 0.731%), MCV (41, 0.697%), HCT (3, 0.051 ¥), LDL-C (1, 0.017%), GHb (42, 0.714%), PLT (42, 0.714%) and Hb (43, 0.731%). In order to minimize the bias due to missing variables, the missing data were estimated by multiple interpolation [17]. (The number of iterations was 5 and the regression model was linear regression)

Statistical analysis

Statistical analysis was performed using R language version 3.4.3, IBM SPSS version 26.0 and Zstats v1.0. Data were described as mean and standard deviation for normally or approximately normally distributed continuous variables, and median and quartiles for non-normally distributed continuous variables. Categorical variables were then described in terms of frequencies and percentages. An analysis of variance or rank sum test was also performed on the baseline data in each data set.

We used multifactor Cox proportional risk regression models to analyze the relationship between TyG index and new-onset diabetes. Three models were used: model 1 (unadjusted for any covariates), model 2 (adjusted for age, sex, education level, place of residence, marital status, smoking and drinking status, and BMI), and model 3 (plus hypertension and dyslipidemia). We then used stratified Cox proportional risk regression analyses for subgroup assessment, with interaction analyses to determine whether sociodemographic and health-related factors influenced the association between TyG index and diabetes. In addition, we used decision tree analysis to identify those at high risk for new-onset diabetes. Finally, we assessed the discriminatory power of the TyG index by Receiver Operating Characteristic(ROC) curves and calculated the Area under the curve(AUC) area to assess the performance of the TyG index.

Results

Baseline characteristics of participants

Our finalized cohort consisted of 5886 participants, 2618 males and 3268 females, with a mean age of 58.29 (8.64 years). Anthropometric and biochemical characteristics of patients stratified according to TyG quartiles showed significant increases in WBC, Hb, FBG, CRP, GHb, ua, tc, TG, SBP, DBP, BMI, female, urban, non-alcohol consumption, hypertension, and dyslipidaemia as triglyceride glucose index increased; and significant decreases in HDL-C, male, rural, and persistent smoking and alcohol consumption(Table 1).

Table 1 Baseline characteristics of study population according to TyG quartiles

The incidence of diabetes

The overall prevalence of diabetes was 17.9%.The prevalence of diabetes among participants in TyG quartiles was Q1: 11.5%; Q2: 14.7%; Q3: 20.3%; and Q4: 25.1%. Participants with lower TyG had a significantly lower incidence of diabetes compared to those with higher TyG (Table 2).

Table 2 Incidence rate of diabetes(%)

The relationship between TyG index and the risk of diabetes

To explore the association between TyG index and diabetes risk, we developed three Cox proportional risk regression models. In model 1, each 10-unit increase in TyG index was associated with a 77.4% increase in the risk of diabetes incidence (HR:1.774, 95% CI: 1.599,1.967); in model 2, each 10-unit increase in TyG index was associated with a 62.2% increase in the risk of diabetes incidence (HR:1.774, 95% CI: 1.599,1.967); and in model 3, a each 10-unit increase in TyG index was associated with a 42.2% increase in the risk of developing diabetes (HR:1.774, 95% CI: 1.599,1.967).

In addition, we further converted the TyG index from a continuous variable to a categorical variable based on quartiles. The multivariate adjusted model showed the following risk ratios (HR) associated with diabetes risk: 1.226 (95% CI: 1.001,1.501) for Q2; 1.611 (95% CI: 1.324,1.960) for Q3; and 1.782 (95% CI: 1.435,2.212) for Q4 This suggests that compared to Q1 participants, the risk of diabetes increased by 22.6% for Q2 participants, 61.1% for Q3 participants, and 78.2% for Q4 participants compared to Q1 participants (Table 3; Fig. 2).

Table 3 Prospective relationship between baseline TyG and incident diabetes at follow-up in CHARLS

Fig. 2

Kaplan-Meier curve and cumulative incidence function of new-onset diabetes

Restricted Cubic Spline Curves to Study the Association Between TyG Index and New-Onset Diabetes

We used restricted cubic spline to flexibly model and visualize the association between TyG index and the risk of new-onset diabetes, and observed a nonlinear relationship between TyG index and the risk of new-onset diabetes (p for overall < 0.001, p for nonlinear = 0.018), identifying an inflexion point of 8.516 for TyG. a pre-inflexion point HR of 1.927 was identified (95% CI:1.313, 2.829); post-inflection point HR was 1.453 (95% CI:1.211, 1.744) (Table 4; Fig. 3).

Table 4 The result of two-piecewise linear regression model

Fig. 3

Nonlinear relationship between TyG index and new onset diabetes

Hierarchical analysis

To assess the stability of the positive association between TyG index and new-onset diabetes, participants were divided into subgroups based on sociodemographic variables and disease history, and associations were analyzed within each group. Further interaction analyses showed that the association between TyG index and new-onset diabetes was not affected by age, gender, marriage, place of residence, educational status, smoking and drinking status, BMI status, hypertension and dyslipidemia. This means that the interaction between these variables and triglyceride glucose index was not statistically significant (P > 0.05 for interaction) (Table 5; Fig. 4).

Table 5 Stratified analysis of TyG index and diabetes

Fig. 4

Forest plot of the stratified analysis of the correlation between TyG and the risk of diabetes

Decision tree analysis

The root node of the model indicates that the development of diabetes is influenced by factors such as TyG index and hypertension. From our analysis, the accuracy of the DT model was 82.1%. The results of the decision tree model showed that TyG index was the most important risk factor for the development of diabetes, followed by hypertension in the middle-aged and older age group of 45 years and above (Fig. 5).

Fig. 5

Decision tree analysis of TyG

The TyG index has predictive value for diabetes events

The ROC curve showed that the TyG index had a specificity of 0.652, a sensitivity of 0.521, and an AUC area of 0.652.The cut-off point was 8.687 (Fig. 6).

Fig. 6

ROC curve of TyG

Discussion

With the development of society and the progress of medical care, the proportion of middle-aged and elderly people has gradually increased, and the incidence of chronic diseases, such as diabetes, hypertension, coronary heart disease and other chronic diseases has also increased with each passing year [18, 19]. Reducing the incidence of chronic diseases and carrying out a healthy life have become the pursuit goals of middle-aged and elderly people [20, 21]. As the country with the largest number of diabetes in China, identification of diabetes risk factors to reduce morbidity and social burden is crucial from the perspective of preventing diabetes occurrence [4, 5]. Our longitudinal study found that elevated TyG index increases the incidence of diabetes in individuals aged 45 years and older in our population; a nonlinear relationship was observed between TyG index and new-onset diabetes. In addition, subgroup analyses provided additional evidence for this finding, suggesting that TyG index and the onset of diabetes are not influenced by other factors. Our findings reveal an association between TyG index and the risk of diabetes onset, and it may be an effective test for preventing the onset of diabetes.

Insulin resistance is one of the main pathologic mechanisms leading to the development of diabetes; the assessment of individual insulin resistance can help to avoid the risk of diabetes [22]. However, the high cost and limited availability of serum insulin test make it difficult to use it as a homeostatic model of insulin resistance to evaluate the level of insulin resistance in the clinic.The validity and homeostatic modeling of TyG index, as a readily available and easy-to-calculate metric, has an advantageous position over other metrics, such as HOMA-IR, in assessing the level of insulin resistance in individuals [23, 24]. Further studies have found [14] that the TyG index consistently demonstrates strong sensitivity and specificity regardless of diabetes status or insulin treatment. Therefore, the association between TyG index and the risk of developing diabetes has attracted much attention from scholars at home and abroad. A 9-year prospective study showed [25]: participants in the highest and second highest quartiles of TyG index had a significantly higher risk of new-onset diabetes than those in the lowest quartile.The DRYAD data study showed [26]: TyG index was positively correlated with the risk of diabetes onset in both men and women; the predictive power of this relationship was more prominent in women. In addition, higher TyG index may independently predict the risk of developing gestational diabetes in Asian women [27]. Our study differed from previous studies in that continuous and categorical variables associated with the risk of diabetes were analyzed in terms of TyG index showed a nonlinear relationship between TyG index and new-onset diabetes. This is consistent with the results of Xuan et al. [28, 29]. The difference is that our study found that the nonlinear relationship between TyG index and the risk of new-onset diabetes may show an inverted “U”-shaped relationship, i.e., the risk of diabetes onset showed an increase and then a decrease with the increase of the TyG index, and the cutoff point of the change between TyG index and the risk of new-onset diabetes was 8.516; and with the influence of the TyG index and the risk of new-onset diabetes, the change in the risk of new-onset diabetes was 8.516. 8.516; and the risk of diabetes incidence gradually decreased with the increase of influencing factors such as age, gender, education level, living environment, marital status, smoking and drinking status.

As an independent predictor of the risk of diabetes, the TyG index is used to assess insulin resistance through two risk factors: glucose metabolism and lipid metabolism. Impaired glucose metabolism is directly related to the development of diabetes; healthy lifestyle, high-intensity intermittent exercise, and theaflavins can effectively regulate glucose metabolism, which is promising in preventing the risk of diabetes [30,31,32,33,34]. Lipid metabolism is another important mediator of insulin resistance. Excessive accumulation of fat in the body accumulates large amounts of adipokines, which increase the source of inflammatory factors, chemokines and metabolically active substances, which in turn affects the development of diabetes [35,36,37,38,39]. Therefore, managing lipid metabolism and adjusting dietary patterns in middle-aged and elderly people are conducive to reducing the accumulation of body fat, improving lipid metabolism, reducing the damage of inflammatory responses to insulin cells, and preventing and mitigating the disease process. In addition, glucose metabolism and lipid metabolism tend to interact with each other, and regulation of glucose-lipid metabolism and timely reduction of insulin resistance may delay or even avoid the onset and development of diabetes.

Our study has some noteworthy limitations. First, our study was directed at the demographic concerns of middle-aged and elderly people in China, and is likely not applicable to younger groups as well as other groups of different ethnicities. Second, certain indicators related to diabetes, such as the presence of family history, medication use, etc., were missing from the raw data, which may have had some impact on the observations. Third, as with all observational studies, unmeasured and uncontrolled confounding above anger may still exist despite adjustment for potential confounders. Fourth, the data may be subject to recall bias and measurement error, which may lead to sampling bias. Fifth, our study could not determine a causal relationship between TyG index and diabetes, only an association between them. Finally, due to the lack of documentation of physical activity in the questions of the included data, it failed to take into account whether physical activity has an effect on the occurrence of diabetes, which is something we need to focus on in the next studies. Despite these limitations, the study also has a number of key strengths. First, the findings are based on a national study design of middle-aged and older adults in China, which provides us with a large sample size and good representation. Second, quality control and convenient measurement were strictly carried out throughout our study, thus ensuring the quality of the current study.

Conclusion

In conclusion, our study demonstrated that the TyG index has some clinical value in predicting the risk of developing diabetes. The present study emphasizes the importance of reducing insulin resistance and addressing associated metabolic problems in delaying the onset of diabetes, which may contribute to timely diagnosis, early prevention, and optimal treatment of diabetes, as well as slowing its progression. In addition, we found that the relationship between TyG index and diabetes onset may belong to a nonlinear relationship, identifying an inflexion point of 8.516 for TyG; for every 10-unit increase in TyG index on the left side of this inflection point. The risk of diabetes onset increased by 92.7%, and on the right side of this inflection point, for every 10-unit increase in TyG index. the risk of developing diabetes increased by 45.3%. The findings of this study may provide an additional reference to facilitate clinical consultation and optimize diabetes prevention decisions.