Skip to main content
NCBI home page
Medicine logoLink to Medicine
. 2026 Jun 26;105(26):e49266. doi: 10.1097/MD.0000000000049266

The predictive ability of “TyG_CVAI” for incident stroke in individuals with different glycemic metabolic status: A national cohort study

Li Huang a, Zhihua Cao a,*
PMCID: PMC13313657  PMID: 42363562

Abstract

The triglyceride-glucose-Chinese visceral adiposity index (TyG_CVAI) has been validated as an effective predictor of early stroke. However, its predictive efficacy across different glycemic metabolic states remains unexplored. This study utilized data from the China Health and Retirement Longitudinal Study, including 7744 stroke-free participants. Participants were categorized into 3 glycemic groups: diabetes mellitus (DM), prediabetes mellitus (Pre-DM), and normal glucose regulation (NGR). K-means clustering further divided participants into 2 clusters based on TyG_CVAI fluctuations: Cluster 1 (lower TyG_CVAI) and Cluster 2 (higher TyG_CVAI). The primary outcome was the first stroke occurrence. Statistical analyses included Kaplan–Meier survival curves, Cox proportional hazards models, restricted cubic spline analysis, and receiver operating characteristic curve analysis to assess associations between baseline TyG_CVAI and stroke risk. Over 9 years (2011–2020), 540 participants (7.0%) experienced a first stroke. Participants in Cluster 2 (higher TyG_CVAI) exhibited a significantly elevated stroke risk compared to Cluster 1. Kaplan–Meier analysis revealed significant differences in stroke incidence between clusters among Pre-DM, NGR, and the overall population (P < .001), but not in the DM group (P = .071). Cox regression models, adjusted for confounders, confirmed a significant positive correlation between TyG_CVAI and stroke risk in the overall population, Pre-DM, and NGR groups. The hazard ratios and 95% confidence intervals for Cluster 2 versus Cluster 1 were 1.40 (1.15–1.71), 1.42 (1.07–1.88), and 1.68 (1.19–2.36), respectively. In contrast, no significant association was observed in the DM group (P > .1). Restricted cubic spline analysis further supported that higher TyG_CVAI levels correlated with increased stroke risk. Elevated baseline TyG_CVAI is significantly associated with stroke risk in middle-aged and older adults. However, its predictive capacity is attenuated in individuals with diabetes compared to those with normal glycemic metabolism. These findings underscore the importance of considering glycemic status when evaluating TyG_CVAI as a stroke predictor.

Keywords: CHARLS, diabetes mellitus, prediabetes, stroke, TyG_CVAI

1. Introduction

Stroke is a highly prevalent and extremely detrimental cerebrovascular disease worldwide,[1] which is increasingly attracting widespread attention in the medical community.[2] As one of the leading causes of disability and mortality among adults, stroke not only leads to long-term motor, sensory, and cognitive impairments in patients but also imposes a heavy burden on global health systems[3] due to its high recurrence rate and lengthy rehabilitation period.[4] Therefore, there is an urgent need to develop low-cost, highly accurate, and reproducible indicators to more effectively identify high-risk populations for stroke, thereby enabling early intervention.

Currently, numerous studies have clearly identified diabetes,[5] hypertension,[6] and hyperlipidemia[7–9] as important risk factors for stroke. Although metabolic indicators such as the triglyceride-glucose index[10–12] and visceral adiposity index[13–15] have been applied to assess the risk and prognosis of stroke, their predictive power remains insufficient, and their predictive effects in different populations require further in-depth investigation. It is worth noting that the triglyceride-glucose-Chinese visceral adiposity index (TyG_CVAI),[16] as an emerging composite indicator, has shown promising application prospects in stroke prediction.

However, research on the association between TyG_CVAI levels and stroke based on individual glycemic metabolic status remains limited. Therefore, it is necessary to conduct large-scale, prospective cohort studies to further explore the intrinsic relationship between TyG_CVAI levels and stroke risk, thereby providing a more scientific basis for the prevention and treatment of stroke. In this study, we utilized data from the China Health and Retirement Longitudinal Study (CHARLS) database to investigate the association between baseline TyG_CVAI levels and stroke under different glucose metabolic statuses.

2. Materials and methods

2.1. Study participants

This study employed a prospective study design, with all participants recruited from the CHARLS database. As shown in Figure 1, we initially collected data from 13,614 participants in the first wave of the survey. Among them, 5603 individuals were excluded based on the following criteria: age <45 years or missing age data (n = 267), missing data for variables necessary to calculate the TyG_CVAI (n = 3840), missing stroke outcome data or loss to follow-up (n = 1539), and history of stroke at baseline (n = 224). Ultimately, 7744 participants were included and followed up until 2020. The CHARLS study was approved by the Biomedical Ethics Review Committee of Peking University (IRB00001052-11015), and all participants provided written informed consent.

Figure 1.

Figure 1.

Open in a new tab

Flowchart of participant inclusion and exclusion. TyG_CVAI = triglyceride-glucose-Chinese visceral adiposity index.

2.2. Data collection

Interviewers used a standardized questionnaire to comprehensively collect participants’ demographic characteristics (such as age, gender, and marital status), health status, and functional status (including height, weight, waist circumference [WC], smoking habits, alcohol consumption, history of hypertension, and diabetes status). According to the standard protocol, staff from the Chinese Center for Disease Control and Prevention collected 3 tubes of venous blood from each participant. Over 92% of the blood donors had fasted overnight. High-sensitivity C-reactive protein (CRP), glycated hemoglobin (HbA1c), lipid profiles (total cholesterol, high-density lipoprotein [HDL], low-density lipoprotein [LDL] cholesterol, and triglycerides [TG]), glucose, blood urea nitrogen, creatinine, uric acid, and cystatin C were measured from frozen plasma or whole blood samples.

2.3. Definitions

Hypertension was diagnosed based on self-reported physician diagnosis and/or the use of any antihypertensive medications, and/or a mean systolic blood pressure/diastolic blood pressure ≥140/90 mm Hg. Diabetes mellitus (DM) was defined as a fasting plasma glucose ≥126 mg/dL or HbA1c ≥6.5%, and/or self-reported physician diagnosis, and/or the use of glucose-lowering medications. Prediabetes mellitus (Pre-DM) was characterized by a fasting plasma glucose of 100 to 125 mg/dL or HbA1c of 5.7% to 6.4%. Individuals without DM or Pre-DM were classified as having normal glucose regulation (NGR). Dyslipidemia was diagnosed based on self-reported physician diagnosis and/or current use of lipid-lowering medications, and/or total cholesterol ≥240 mg/dL, TG ≥150 mg/dL, HDL-C <40 mg/dL, LDL-C ≥160 mg/dL. The body mass index (BMI) was calculated as weight divided by the square of height (kg/m2). The TyG_CVAI was calculated as follows: TyG = ln (TG [mg/dL] × FBG [mg/dL]); for males: CVAI = −67.93 + 0.68 × age (years) + 0.03 × BMI (kg/m2) + 4.00 × WC (cm) + 22.00 × log10 (TG; mmol/L) − 16.32 × HDL-C (mmol/L); for females: CVAI = −187.32 + 1.71 × age (years) + 4.23 × BMI (kg/m2) + 1.12 × WC (cm) + 39.76 × log10 (TG; mmol/L) − 11.66 × HDL-C (mmol/L); TyG_CVAI = TyG × CVAI.

2.4. Follow-up of endpoint events

The primary endpoint of the study was the first occurrence of stroke, including both ischemic stroke and hemorrhagic stroke. Self-reported stroke was assessed through the following questions: “Have you ever been diagnosed with a stroke by a physician?”; “Have you been diagnosed with a stroke by a physician since the last follow-up?”; and “Compared to the last time we interviewed you, has your stroke condition improved, remained about the same, or worsened?” The timing of stroke events was determined by the participants’ responses to the questions: “When was the stroke first diagnosed or known?” and “When was your most recent stroke?” Participants were interviewed in 5 waves from 2011 to 2020 or until the occurrence of stroke, whichever came first.

2.5. Statistical analysis

To classify the trends in the changes of the TyG-CVAI index and observe their impact on stroke, we employed K-means clustering analysis using the “cluster” and “factoextra” packages. The K-means clustering algorithm is an iterative method that clusters data by calculating the within-cluster sum of squares for each K value. The elbow point on the within-cluster sum of squares curve is considered the optimal number of clusters, representing the best partitioning of the dataset. In our study, as shown in Figure 2, the curve stabilizes when K = 2, where the silhouette coefficient plot shows that the silhouette coefficient reaches its maximum value, indicating that the optimal number of clusters is 2. We divided participants into 2 groups based on the changes in the TyG-CVAI index: Cluster 1 (low level, n = 4502) and Cluster 2 (high level, n = 3242).

Figure 2.

Figure 2.

Open in a new tab

K-means analysis for grouping. (A) The elbow method plot shows the greatest slope change at K = 2. (B) The silhouette coefficient plot peaks at K = 2. (C) Histograms of the distributions for Cluster 1 and Cluster 2. TyG_CVAI = triglyceride-glucose-Chinese visceral adiposity index.

For continuous variables, differences between the 2 groups were described using mean ± standard deviation or median (interquartile range), while categorical variables were described using frequency (n) and percentage (%). Normally distributed continuous data were presented as mean ± standard deviation and analyzed for statistical significance using one-way analysis of variance. Non-normally distributed continuous data were presented as median and interquartile range and analyzed using the Kruskal–Wallis test. Categorical data were described using counts and percentages and assessed using the chi-square test.[17]

Based on the grouping by TyG-CVAI index, the Kaplan–Meier method was used to estimate the cumulative incidence of stroke, and the log-rank test was employed to assess differences between groups. Subsequently, Cox proportional hazards regression models were constructed to explore the association between baseline TyG-CVAI levels and stroke occurrence, with hazard ratios (HRs) and 95% confidence intervals (CIs) calculated. Three models were developed: Model 1 estimated the crude HR without adjustment; Model 2 adjusted for age, sex,[18] marital status, and education level based on Model 1; Model 3 further adjusted for alcohol consumption, smoking, hypertension, dyslipidemia, and diabetes on the basis of Model 2. In all models, Cluster 1 was used as the reference group. In addition, restricted cubic spline (RCS) analysis based on multivariate-adjusted Cox regression was conducted to visually present the linear or nonlinear relationship between baseline TyG-CVAI levels and stroke risk. To clarify the prognostic value of TyG-CVAI for stroke in different glucose metabolic states, analyses were performed separately for participants with DM, Pre-DM, and NGR. Subgroup analyses were stratified by baseline age, sex, education level, marital status, hypertension, dyslipidemia, alcohol consumption, obesity, and glucose metabolic status (DM, Pre-DM, and NGR) to assess the consistency of the adverse effects of TyG-CVAI on incident stroke. Given that the TyG-CVAI is not a static and invariant indicator, to further assess its accuracy and predictive capability, we incorporated not only the baseline TyG-CVAI values but also extracted data from Wave 3 of the study. Subsequently, we computed the dynamic changes in the TyG-CVAI indicator over the period from 2011 to 2015. Employing logistic regression analysis, we examined the correlation between these changes in the indicator and the risk of stroke.

All statistical analyses were conducted using IBM SPSS Statistics (version 25), R (version 4.4.2), and RStudio (2024.04.2 + 764).

3. Results

3.1. General characteristics of participants

The baseline clinical and demographic characteristics of participants in Cluster 1 and Cluster 2, as divided by K-means clustering analysis, are detailed in Figure 3. The average age of the participants was 58.6 years, with 4264 women (55.1%) included. Significant differences were observed across several population characteristics between the 2 clusters. In terms of demographic features, the proportion of individuals aged over 70 years was significantly higher in Cluster 2 compared to Cluster 1, while Cluster 1 had a higher proportion of individuals aged 50 to 69 years. The percentage of women and widowed individuals was higher in Cluster 2 than in Cluster 1, although no significant differences were observed in educational levels between the 2 groups. Regarding health behaviors, Cluster 2 had a lower proportion of alcohol consumers and current smokers compared to Cluster 1. In terms of health status, the prevalence of hypertension, diabetes, and dyslipidemia was higher in Cluster 2 than in Cluster 1, with the most significant and substantial increase observed in the obesity rate. Conversely, the depression rate was slightly lower in Cluster 2 compared to Cluster 1. When stratified by glycemic metabolic status, the majority of individuals in Cluster 1 were categorized as Pre-DM and NGR, with only 8.3% classified as DM. In contrast, over one-fifth (21.8%) of individuals in Cluster 2 were classified as DM, while Pre-DM and NGR accounted for 49.8% and 28.4%, respectively.

Figure 3.

Figure 3.

Open in a new tab

Baseline characteristics of participants classified by changes in TyG-CVAI index. TyG_CVAI = triglyceride-glucose-Chinese visceral adiposity index.

3.2. The predictive value of baseline TyG_CVAI levels for incident stroke

After 9 years of follow-up, a total of 540 participants (7.0%) experienced their first stroke. Comprehensive analysis of individuals across all glycemic metabolic statuses revealed that the incidence rates of stroke in Cluster 1 and Cluster 2, as determined by K-means clustering analysis, were 5.1% and 9.5%, respectively. In Figure 4, the Kaplan–Meier analysis of cumulative incidence rates indicated an increasing trend in the occurrence of stroke events from Cluster 1 to Cluster 2, with statistically significant differences (P < .05).

Figure 4.

Figure 4.

Open in a new tab

Kaplan–Meier analysis of stroke incidence. (A) Based on the overall population. (B) Individuals with diabetes mellitus. (C) Individuals with prediabetes. (D) Individuals with normal glucose regulation. DM = diabetes mellitus, NGR = normal glucose regulation, Pre-DM = prediabetes mellitus.

Cox proportional hazards regression analysis presented in Figure 6 further confirmed that there is a significant association between baseline TyG_CVAI levels and the risk of incident stroke. After fully adjusting for potential confounders, including age, sex, smoking history, alcohol consumption, hypertension, diabetes, and dyslipidemia, the study found that for each 1-unit increase in baseline TyG_CVAI level, the risk of stroke increased by 0.09% in the overall population (HR = 1.0009, 95% CI = 1.0007–1.0011). Further analysis using Cluster 1 as the reference group showed that the risk of stroke was significantly higher in Cluster 2, with a 40% increase compared to Cluster 1 (HR = 1.40, 95% CI = 1.15–1.71).

Figure 6.

Figure 6.

Open in a new tab

Association of TyG_CVAI with stroke risk stratified by glycemic metabolic status. DM = diabetes mellitus, NGR = normal glucose regulation, Pre-DM = prediabetes mellitus, TyG_CVAI = triglyceride-glucose-Chinese visceral adiposity index.

In addition, the multivariable-adjusted RCS analysis (Fig. 5) revealed that there is not only a significant dose–response relationship between TyG_CVAI levels and stroke risk, with stroke risk increasing as TyG_CVAI levels rise, but also a nonlinear characteristic of this relationship (P for nonlinear = .0006). This suggests that the rate of increase in stroke risk may vary across different intervals of TyG_CVAI levels.

Figure 5.

Figure 5.

Open in a new tab

Restricted cubic spline analysis of the association between baseline TyG_CVAI levels and stroke risk. (A–C) Represent the overall population, Cluster 1, and Cluster 2, respectively. The numbers 1, 2, 3, and 4 represent the overall population, individuals with diabetes mellitus, individuals with prediabetes, and individuals with normal glucose regulation, respectively.

3.3. Glycemic metabolic status affects the predictive ability of TyG_CVAI

3.3.1. Kaplan–Meier analysis

During the follow-up period, significant differences were observed in the incidence of first-time stroke among individuals with different glycemic metabolic statuses: 101 out of 1081 (9.3%) participants with DM experienced stroke, 264 out of 3516 (7.5%) with prediabetes, and 176 out of 3147 (5.6%) with NGR. Kaplan–Meier curve analysis (Fig. 4) revealed that the cumulative incidence of stroke was significantly different among the overall population, individuals with prediabetes, and those with NGR (P < .0001), while this difference was not statistically significant in the diabetes group (P = .071).

3.3.2. Cox regression analysis

The results of the Cox regression analysis (Fig. 6) further revealed that, except for the diabetes group, higher TyG_CVAI levels were significantly associated with an increased risk of stroke in the other glycemic metabolic groups and the overall population in Model 3. Specifically, in the overall population, the HR for Cluster 2 was 1.40 (95% CI = 1.14–1.71, P < .001); for the prediabetes group, the HR for Cluster 2 was 1.42 (95% CI = 1.07–1.88, P = .014); and for the NGR group, the HR for Cluster 2 was 1.68 (95% CI = 1.19–2.36, P = .003). However, in the 3 Cox models for the diabetes group, no significant differences were found between Cluster 1 and Cluster 2, with all P values being >.05.

3.3.3. RCS analysis

RCS analysis (Fig. 5) further revealed the complex association between TyG_CVAI levels and stroke incidence. The results showed that an increase in baseline TyG_CVAI levels was significantly associated with an increased incidence of stroke in a dose–response manner, with a certain degree of nonlinearity (overall P value < .0001, nonlinearity P value = .0006). Specifically, in the overall population, after stratifying the data into Cluster 1 and Cluster 2, a significant nonlinear relationship was found in Cluster 1 (overall P value < .001, nonlinearity P value < .0001), while a significant linear relationship was observed in Cluster 2 (overall P value = .004, nonlinearity P value = .46). In the normal population, the overall curve had 2 inflection points (both overall and nonlinearity P values < .0001), and in Cluster 1, the dose–response and nonlinear relationships remained significant (both overall and nonlinearity P values < .0001), while in Cluster 2, neither the linear nor the nonlinear relationships were significant (overall P value = .16, nonlinearity P value = .11). In the prediabetes population, an overall significant linear relationship was observed (overall P value < .0001, nonlinearity P value = .26), but after further stratification into the 2 clusters, neither the linear nor the nonlinear relationships were significant. In the diabetes population, although significant dose–response and nonlinearity were observed overall (overall P value = .036, nonlinearity P value = .048), neither relationship reached significance in either Cluster 1 or Cluster 2 (all P values > .05).

3.3.4. Subgroup analysis

To further explore the potential association between baseline TyG_CVAI levels and incident stroke, this study conducted stratified subgroup analyses based on key risk factors such as age, sex, education level, hypertension, diabetes, dyslipidemia, alcohol consumption, and obesity. During the analysis, Cluster 1 was used as the reference group. The results (Fig. 7) showed that an increase in TyG_CVAI levels was positively correlated with an increased incidence of stroke in all subgroups. Specifically, in individuals with prediabetes and NGR, an increase in TyG_CVAI levels significantly increased the risk of stroke. However, in the diabetes group, this association did not reach statistical significance (HR = 1.53, 95% CI = 0.96–2.44, P > .05). In addition, a significant interaction was found between TyG_CVAI levels and age (interaction P value = .036), while no similar interaction effects were observed in the other variables examined (all P values > .05).

Figure 7.

Figure 7.

Open in a new tab

Subgroup analysis and interaction effects of TyG_CVAI levels with stroke. DM = diabetes mellitus, NGR = normal glucose regulation, Pre-DM = prediabetes mellitus, TyG_CVAI = triglyceride-glucose-Chinese visceral adiposity index.

3.3.5. Sensitivity analysis

The sensitivity of TyG_CVAI levels in predicting stroke events was evaluated by plotting receiver operating characteristic curves for the overall population and individuals with different glycemic metabolic statuses (Fig. 8). First, receiver operating characteristic curves were plotted separately for the overall population, individuals with diabetes, prediabetes, and NGR. The results showed that the area under the curve (AUC) for the overall population was 0.620, while the AUC values for individuals with diabetes (GR1), prediabetes (GR2), and NGR (GR3) were 0.595, 0.602, and 0.626, respectively. This indicates that the predictive ability of TyG_CVAI is relatively weak in the diabetes group.

Figure 8.

Figure 8.

Open in a new tab

Comparison of the sensitivity of TyG_CVAI in different populations. (A) ROC curves for the overall population with AUC values: overall-AUC = 0.620, DM-AUC = 0.595, Pre-DM-AUC = 0.602, and NGR-AUC = 0.626. (B) ROC curves for Cluster 1 in each group with AUC values: overall-AUC = 0.582, DM-AUC = 0.543, Pre-DM-AUC = 0.577, and NGR-AUC = 0.584. (C) ROC curves for Cluster 2 in each group with AUC values: overall-AUC = 0.569, DM-AUC = 0.576, Pre-DM-AUC = 0.567, and NGR-AUC = 0.555. DM = diabetes mellitus, NGR = normal glucose regulation, Pre-DM = prediabetes mellitus, ROC = receiver operating characteristic, TyG_CVAI = triglyceride-glucose-Chinese visceral adiposity index.

Further, the study conducted a longitudinal comparison of Cluster 1 and Cluster 2 within the overall population and the 3 glycemic metabolic status groups. In Cluster 1, where TyG_CVAI levels were lower, the AUC for the overall population was 0.582, while the AUC values for GR1, GR2, and GR3 were 0.543, 0.577, and 0.584, respectively. Although the overall AUC in this subgroup was lower than that of the overall population, the trend of increasing AUC from GR1 to GR3 was maintained, indicating that the predictive ability of TyG_CVAI remained the lowest in the diabetes group within this subgroup. However, in Cluster 2, where TyG_CVAI levels were higher, the trend of AUC values was reversed. The AUC for the overall population was 0.569, while the AUC values for GR1, GR2, and GR3 were 0.576, 0.567, and 0.555, respectively, showing a decreasing trend.

3.3.6. Dynamic analysis

The results of the dynamic analysis are presented in Figure 9 in the form of a forest plot. Three models were constructed in this study: Model 1 is the original model without adjusting for any variables; Model 2 adjusted for 4 covariates: age, gender, education, and marital status; Model 3 further adjusted for drinking, smoking, hypertension, dyslipidemia, and diabetes. In each model, Cluster 1 serves as a reference. The results showed that for the total population (total) and the NGR group, the dynamic changes in TyG_CVAI were significantly associated with stroke risk in all 3 models. In contrast, the results for DM were not significant in any of the 3 models, and the results for Pre-DM were not significant in Model 3. This analysis is consistent with previous findings, indicating that in the diabetic population, neither the baseline TyG_CVAI nor its dynamic changes over 4 years were effective in predicting stroke risk. The dynamic analysis suggests that patients’ physiological status, lifestyle habits, and disease progression collectively influence TyG_CVAI levels and stroke occurrence. Moreover, the dynamic changes in TyG_CVAI themselves may contain important predictive information, which is of great significance for constructing a more robust stroke prediction model.

Figure 9.

Figure 9.

Open in a new tab

Forest map of TyG_CVAI dynamic analysis. (A) Model 1: no variables have been adjusted. (B) Model 2: adjusted for age, gender, education, and marital status. (C) Model 3: adjusted for age, gender, education, marital status, as well as drinking, smoking, hypertension, dyslipidemia, and diabetes. DM = diabetes mellitus, NGR = normal glucose regulation, Pre-DM = prediabetes mellitus, TyG_CVAI = triglyceride-glucose-Chinese visceral adiposity index.

4. Discussion

This study utilized the CHARLS database to fill the gap in the differential predictive ability of the novel stroke predictor TyG_CVAI across different glycemic metabolic populations. In this large national longitudinal survey cohort targeting middle-aged and older adults, the study elucidated the significant correlation between higher TyG_CVAI levels and increased risk of incident stroke. This association was particularly pronounced in individuals with NGR and those with Pre-DM, but not in individuals with DM. The findings suggest that TyG_CVAI levels may serve as a reliable biomarker for stratifying stroke risk in populations without diabetes, and maintaining lower TyG_CVAI levels may be beneficial for primary stroke prevention in individuals without diabetes.[19]

Hyperglycemia leads to increased oxidative stress and activation of inflammatory responses,[3,20,21] which damage vascular endothelial cells and promote the development of atherosclerosis.[22] Insulin resistance[23,24] not only causes dyslipidemia (such as hypertriglyceridemia and low HDL-C levels)[25,26] but also results in endothelial dysfunction, further exacerbating atherosclerosis.[27] Moreover, individuals with diabetes often exhibit a hypercoagulable state, characterized by increased platelet aggregation and impaired fibrinolytic system function, significantly increasing the risk of thrombus formation.[28,29] Microvascular complications lead to cerebral hypoperfusion and increase the susceptibility of brain tissue to ischemia.[21,25,30] The combined effects of obesity,[7,31] hypertension,[32,33] hypertriglyceridemia,[10,34] low HDL-C,[35] and hyperglycemia[36] in metabolic syndrome significantly increase the risk of stroke. Chronic inflammatory states continuously damage vascular endothelial cells, promote the development of atherosclerosis, and lead to platelet activation and aggregation, further increasing the risk of thrombus formation.[37] These mechanisms interact with each other, significantly increasing the incidence of stroke.

This study builds upon the foundation of many previous studies. The current research hotspot focuses on predicting stroke using combined indicators,[38–40] but there are fewer studies on the differences in these indicators across different populations. Wang et al first investigated the predictive role of the TyG_CVAI index for stroke and compared its predictive power with other combined indicators (TyG-waist-to-height ratio, TyG-WC, TyG-BMI, and TyG-weight-adjusted waist index).[16] Their study showed that the dynamic changes, baseline, and cumulative values of the TyG_CVAI index are independent risk factors for incident stroke.

Similarly, Qu et al examined the relationship between the plasma atherogenic index and incident stroke in individuals with different glycemic metabolic statuses.[41] Huo et al explored the correlation between the CRP-triglyceride-glucose index and stroke risk in different glucose metabolic states.[36] Both studies were based on data from the CHARLS and investigated the ability of novel combined indicators to predict stroke risk, particularly in populations with different glucose metabolic statuses.

The first study found that the CRP-triglyceride-glucose index had a significant positive linear relationship with stroke risk, which was significant in individuals with NGR and Pre-DM, but not in those with DM. The second study revealed that the plasma atherogenic index was also significantly positively correlated with stroke risk, particularly in Pre-DM and DM populations, but not in the NGR group.

These findings suggest that the utility of predictive indicators varies across different populations, and there is a need to identify a more stable, practical, and accurate indicator for use in individuals with diabetes.

4.1. Strengths and limitations

This study represents the first and largest investigation to date on the association between TyG_CVAI levels and stroke risk in middle-aged and older adults across different glycemic metabolic statuses. The study leverages a high-quality, nationally representative longitudinal survey database that covers urban and rural areas across China, focusing on the middle-aged and older population. Unlike traditional indicators such as BMI and WC, TyG_CVAI integrates blood markers with the Chinese visceral adiposity index, providing deeper insight into the impact of visceral obesity on stroke risk beyond mere surface adiposity. To ensure the robustness and reliability of the findings, the study adjusted for numerous potential confounders to minimize the influence of extraneous factors on the results. After nearly a decade of follow-up, the analysis revealed that TyG_CVAI levels can serve as a reliable indicator for predicting stroke risk in middle-aged and older adults with NGR and Pre-DM. Given the common use and ease of obtaining fasting TG, fasting glucose, BMI, WC, TG, and HDL-C in clinical practice, TyG_CVAI is recommended as an effective and convenient indicator for assessing stroke risk, holding significant scientific and practical value. This study holds practical significance, especially considering that the predictive power of TyG_CVAI is weaker in individuals with DM compared to other groups. Therefore, there is a need to identify more stable and sensitive indicators for predicting stroke risk in the DM population.

However, this study also has several limitations. First, due to the strict exclusion criteria of the CHARLS cohort, some participants were excluded from this study, resulting in a relatively limited sample size and potentially introducing attrition bias. Second, the effective sample size was further reduced due to missing values and outliers in the outcome variables, time variables, and variables necessary for calculating the predictive indicators in the database. Third, based on the Chinese middle-aged and older population, the applicability of the study’s conclusions to other ethnic groups and age groups needs further validation. Fourth, this study mainly focused on the impact of baseline TyG_CVAI levels and did not conduct an in-depth exploration of the longitudinal changes in TyG_CVAI during the follow-up period. Finally, although this study has tried its best to control for confounding variables, there may still be confounding factors that have not been fully considered. Therefore, it is necessary to carry out further investigations to verify the universality and reliability of the study’s results in other large cohort studies.

5. Conclusions

In the search for primary prevention strategies for stroke, this longitudinal prospective study found that among individuals with normal glucose metabolism and those with prediabetes, a high baseline TyG_CVAI level indicates a higher risk of stroke. However, among middle-aged and older adults with diabetes, the TyG_CVAI level is not effective in predicting the occurrence of stroke.

Author contributions

Data curation: Li Huang.

Formal analysis: Li Huang.

Investigation: Li Huang.

Methodology: Li Huang.

Project administration: Li Huang.

Resources: Li Huang.

Software: Li Huang.

Validation: Li Huang.

Writing – original draft: Li Huang.

Conceptualization: Zhihua Cao.

Funding acquisition: Zhihua Cao.

Writing – review & editing: Zhihua Cao.

Abbreviations:

AUC
area under the curve
BMI
body mass index
CHARLS
China Health and Retirement Longitudinal Study
CI
confidence interval
DM
diabetes mellitus
HbA1c
glycated hemoglobin
HR
hazard ratio
NGR
normal glucose regulation
Pre-DM
prediabetes mellitus
RCS
restricted cubic spline
TyG_CVAI
triglyceride-glucose-Chinese visceral adiposity index
WC
waist circumference

The authors have no funding and conflicts of interest to declare.

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

How to cite this article: Huang L, Cao Z. The predictive ability of “TyG_CVAI” for incident stroke in individuals with different glycemic metabolic status: A national cohort study. Medicine 2026;105:26(e49266).

During the preparation of this work the authors used DeepSeek in order to improve language. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

References

  • [1].GBD 2019 Stroke Collaborators. Global, regional, and national burden of stroke and its risk factors, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Neurol. 2021;20:795–820. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [2].Tu W-J, Zhao Z, Yin P, et al. Estimated burden of stroke in China in 2020. JAMA Netw Open. 2023;6:e231455. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [3].Potter TBH, Tannous J, Vahidy FS. A contemporary review of epidemiology, risk factors, etiology, and outcomes of premature stroke. Curr Atheroscler Rep. 2022;24:939–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [4].Marsh JD, Keyrouz SG. Stroke prevention and treatment. J Am Coll Cardiol. 2010;56:683–91. [DOI] [PubMed] [Google Scholar]
  • [5].Zuloaga KL, Krasnow SM, Zhu X, et al. Mechanism of protection by soluble epoxide hydrolase inhibition in type 2 diabetic stroke. PLoS One. 2014;9:e97529. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [6].Buonacera A, Stancanelli B, Malatino L. Stroke and hypertension: an appraisal from pathophysiology to clinical practice. Curr Vasc Pharmacol. 2019;17:72–84. [DOI] [PubMed] [Google Scholar]
  • [7].Després J-P, Lemieux I. Abdominal obesity and metabolic syndrome. Nature. 2006;444:881–7. [DOI] [PubMed] [Google Scholar]
  • [8].Lin M, Zhou S, Gu S. Association between dynamic change patterns of body mass or fat mass and incident stroke: results from the China Health and Retirement Longitudinal Study (CHARLS). Front Cardiovasc Med. 2023;10:1269358. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [9].Pillay P, Lewington S, Taylor H, Lacey B, Carter J. Adiposity, body fat distribution, and risk of major stroke types among adults in the United Kingdom. JAMA Netw Open. 2022;5:e2246613. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [10].Cui C, Liu L, Zhang T, et al. Triglyceride-glucose index, renal function and cardiovascular disease: a national cohort study. Cardiovasc Diabetol. 2023;22:325. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [11].Huo R-R, Zhai L, Liao Q, You X-M. Changes in the triglyceride glucose-body mass index estimate the risk of stroke in middle-aged and older Chinese adults: a nationwide prospective cohort study. Cardiovasc Diabetol. 2023;22:254. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [12].Wu Y, Yang Y, Zhang J, Liu S, Zhuang W. The change of triglyceride-glucose index may predict incidence of stroke in the general population over 45 years old. Cardiovasc Diabetol. 2023;22:132. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [13].Cui C, He C, Sun Q, et al. Association between visceral adiposity index and incident stroke: data from the China Health and Retirement Longitudinal Study. Nutr Metab Cardiovasc Dis. 2022;32:1202–9. [DOI] [PubMed] [Google Scholar]
  • [14].Ren Y, Hu Q, Li Z, Zhang X, Yang L, Kong L. Dose-response association between Chinese visceral adiposity index and cardiovascular disease: a national prospective cohort study. Front Endocrinol (Lausanne). 2024;15:1284144. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [15].Zhou C, Peng Y, Jiang W, Yuan J, Zha Y. Comparison of novel visceral obesity indexes with traditional obesity measurements in predicting of metabolically unhealthy nonobese phenotype in hemodialysis patients. BMC Endocr Disord. 2021;21:244. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [16].Wang M, Gao B, Huang F. Association between the triglyceride glucose-Chinese visceral adiposity index and new-onset stroke risk: a national cohort study. Cardiovasc Diabetol. 2025;24:119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [17].Mbotwa J, Singini I, Mukaka M. Discrepancy between statistical analysis method and study design in medical research: examples, implications, and potential solutions. Malawi Med J. 2017;29:63–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [18].Zuo W, Zhang W, Chen N-H. Sexual dimorphism in cerebral ischemia injury. Eur J Pharmacol. 2013;711:73–9. [DOI] [PubMed] [Google Scholar]
  • [19].Yan F, Yan S, Wang J, et al. Association between triglyceride glucose index and risk of cerebrovascular disease: systematic review and meta-analysis. Cardiovasc Diabetol. 2022;21:226. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [20].Hotamisligil GS. Inflammation and metabolic disorders. Nature. 2006;444:860–7. [DOI] [PubMed] [Google Scholar]
  • [21].Jia G, Bai H, Mather B, Hill MA, Jia G, Sowers JR. Diabetic vasculopathy: molecular mechanisms and clinical insights. Int J Mol Sci. 2024;25:804. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [22].Zhong P, Tan S, Zhu Z, et al. Brain and cognition signature fingerprinting vascular health in diabetic individuals: an international multi-cohort study. Am J Geriatr Psychiatry. 2023;31:570–82. [DOI] [PubMed] [Google Scholar]
  • [23].Guo C, He L, Tu Y, et al. Insulin resistance and sarcopenia: a prognostic longitudinal link to stroke risk in middle-aged and elderly Chinese population. BMC Public Health. 2024;24:2757. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [24].Jiang L, Zhu T, Song W, et al. Assessment of six insulin resistance surrogate indexes for predicting stroke incidence in Chinese middle-aged and elderly populations with abnormal glucose metabolism: a nationwide prospective cohort study. Cardiovasc Diabetol. 2025;24:56. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [25].Zunker P, Schick A, Buschmann HC, et al. Hyperinsulinism and cerebral microangiopathy. Stroke. 1996;27:219–23. [DOI] [PubMed] [Google Scholar]
  • [26].Lee S-H, Park S-Y, Choi CS. Insulin resistance: from mechanisms to therapeutic strategies. Diabetes Metab J. 2022;46:15–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [27].Garg A. The role of body fat distribution in insulin resistance. In: Reaven GM, Laws A, eds. Insulin resistance: the metabolic syndrome X. Humana Press; 1999:83–96. [Google Scholar]
  • [28].Williams KJ, Wu X. Imbalanced insulin action in chronic over nutrition: clinical harm, molecular mechanisms, and a way forward. Atherosclerosis. 2016;247:225–82. [DOI] [PubMed] [Google Scholar]
  • [29].Wieczór R, Wieczór AM, Kulwas A, Rość D. Type 2 diabetes and cardiovascular factors contrasted with fibrinolysis disorders in the blood of patients with peripheral arterial disease. Medicina (Kaunas). 2019;55:395. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [30].Zuliani G, Cherubini A, Ranzini M, Ruggiero C, Atti AR, Fellin R. Risk factors for short-term mortality in older subjects with acute ischemic stroke. Gerontology. 2006;52:231–6. [DOI] [PubMed] [Google Scholar]
  • [31].Wu W, Zheng X. Weight change over 4 years and risk of cardiovascular diseases in China: the China Health and Retirement Longitudinal Study. Obes Facts. 2022;15:694–702. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [32].Jin Q, Chen Y. Risk of heart disease and stroke among individuals with prehypertension or blood pressure progression: a national population-based cohort study. Blood Press Monit. 2024;29:173–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [33].Zheng X, Han L, Shen S. Hypertension, remnant cholesterol and cardiovascular disease: evidence from the China Health and Retirement Longitudinal Study. J Hypertens. 2022;40:2292–8. [DOI] [PubMed] [Google Scholar]
  • [34].Li X, Hu J-G, Liao Q, Wu Y, Huo R-R. Triglyceride-glucose index mediates the association between residual cholesterol and stroke among middle-aged and older adults in China: a prospective, nationwide, population-based study. Front Cardiovasc Med. 2024;11:1429993. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [35].Lv S, Zhang H, Chen J, et al. The effect of triglycerides to high-density lipoprotein cholesterol ratio on the reduction of renal function: findings from China Health and Retirement Longitudinal Study (CHARLS). Lipids Health Dis. 2021;20:110. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [36].Huo G, Tang Y, Liu Z, Cao J, Yao Z, Zhou D. Association between C-reactive protein-triglyceride glucose index and stroke risk in different glycemic status: insights from the China Health and Retirement Longitudinal Study (CHARLS). Cardiovasc Diabetol. 2025;24:142. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [37].Zuurbier CJ, Demirci C, Koeman A, Vink H, Ince C. Short-term hyperglycemia increases endothelial glycocalyx permeability and acutely decreases lineal density of capillaries with flowing red blood cells. J Appl Physiol (1985). 2005;99:1471–6. [DOI] [PubMed] [Google Scholar]
  • [38].Chen X, Liu S, Chu J, Hu W, Sun N, Shen Y. Joint effect of elevated-c-reactive protein level and hypertension on new-onset stroke: a nationwide prospective cohort study of CHARLS. Front Public Health. 2022;10:919506. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [39].Huo R-R, Liao Q, Zhai L, You X-M, Zuo Y-L. Interacting and joint effects of triglyceride-glucose index (TyG) and body mass index on stroke risk and the mediating role of TyG in middle-aged and older Chinese adults: a nationwide prospective cohort study. Cardiovasc Diabetol. 2024;23:30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [40].Jiang S, Zhang X, Song C, Wu G, Yang A. Joint association of METS-IR and uric acid with stoke, mediated by C-reactive protein. Front Endocrinol (Lausanne). 2024;15:1448021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [41].Qu L, Fang S, Lan Z, et al. Association between atherogenic index of plasma and new-onset stroke in individuals with different glucose metabolism status: insights from CHARLS. Cardiovasc Diabetol. 2024;23:215. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Medicine are provided here courtesy of Wolters Kluwer Health

RESOURCES