Association of vascular aging and insulin resistance with cardiovascular outcomes in China: PURE cohort study

Abstract

Background:

Although vascular aging and insulin resistance (IR) are recognized contributors to cardiovascular disease (CVD) pathophysiology, their independent and joint associations with CVD outcomes have not been fully clarified in large population-based cohorts. Evidence regarding their complementary rather than mediating roles remains limited.

Methods:

We used data from the Prospective Urban Rural Epidemiology (PURE) China study, a large prospective cohort comprising 47 931 individuals from 12 provinces across China. Vascular aging was assessed using estimated pulse wave velocity and categorized into supernormal (SVA), healthy (HVA), and early vascular aging (EVA) based on cohort percentiles. IR was evaluated using the triglyceride–glucose (TyG) index. The primary outcome was major CVD events (myocardial infarction, stroke, and heart failure). Cox frailty models with center-level random effects were used to estimate adjusted hazard ratios (aHRs).

Results:

A total of 40 513 participants with complete information were included in the current study. Over a median follow-up of 11.9 years (interquartile ranges 9.5–12.5), 3615 major CVD events, 829 CVD deaths, and 2344 all-cause deaths occurred. Compared with HVA, EVA was associated with substantially higher risks of major CVD events [aHR = 2.17, 95% confidence interval (CI): 1.99–2.35], CVD mortality (aHR = 4.07, 95% CI: 3.47–4.76), and all-cause mortality (aHR = 2.93, 95% CI: 2.65–3.23), while SVA showed consistently lower risks. Furthermore, higher TyG index levels were independently associated with major CVD event risk in a dose-response manner. Exploratory analyses revealed no significant mediation by the TyG index or interaction between vascular aging and TyG levels. Joint analysis revealed that individuals with both EVA and high TyG had the greatest major CVD risk (aHR = 2.20, 95% CI: 1.75–2.77).

Conclusions:

EVA and IR were independently associated with higher CVD risk, highlighting the need for integrated vascular-metabolic risk assessment in clinical practice, although further validation is warranted.

Introduction

Cardiovascular disease (CVD) remains the leading cause of morbidity and mortality worldwide, posing an immense burden on both healthcare systems and national economies^[1–3^]. As the populations age, the global burden of CVD continues to rise, driven in part by the persistence of traditional risk factors and demographic shifts^[4^]. Identifying modifiable contributors to CVD progression is critical for developing effective prevention and early intervention strategies.

Among these emerging risk factors is vascular aging, which encompasses arterial stiffness and endothelial dysfunction as key physiological components underlying cardiovascular risk^[5,6^]. Although vascular aging is a widely used concept, its precise definition remains debated, and surrogate markers such as pulse wave velocity (PWV) are commonly used to capture vascular aging hemodynamic manifestations^[7^]. Estimated PWV (ePWV), derived from age and mean blood pressure (MBP)^[8,9^], has been validated in Chinese populations and demonstrated strong predictive performance for major cardiovascular outcomes^[10,11^]. Vascular ageing phenotypes are defined using the ePWV percentiles^[12,13^]. Especially, early vascular aging (EVA), which represents arterial stiffness exceeding the expected level for chronological age, has been strongly associated with adverse cardiovascular outcomes^[6^]. Moreover, epidemiological studies have investigated factors influencing vascular aging, revealing a stronger correlation with cardiovascular morbidity and mortality than biological aging^[14,15^], possibly due to its closer association with arterial structure and function, which are strongly influenced by metabolic disorders in the aging population^[16^].

Insulin resistance (IR) is an important and increasingly common pathophysiological driver of CVD^[17^]. It is associated with elevated plasma triglycerides (TG) and decreased high-density lipoprotein (HDL) levels, which may contribute to the pathogenesis of atherosclerosis. However, traditional methods for assessing IR, such as the hyperinsulinemic–euglycemic clamp and homeostasis model assessment of insulin resistance (HOMA-IR), are impractical for large-scale epidemiological studies due to their high cost and technical complexity^[18^]. The triglyceride–glucose (TyG) index has therefore been proposed as a simple and cost-effective surrogate marker for IR^[19,20^], with multiple studies demonstrating its strong associations with incident CVD and all-cause mortality^[21^]. Evidence from the Prospective Urban Rural Epidemiology (PURE) study indicates that the TyG index is associated with CVD mortality, incident CVD, and type 2 diabetes across diverse populations, suggesting that IR promotes the pathogenesis of both cardiovascular and metabolic diseases^[19^]. Recent studies have further identified IR as being related to arterial stiffness and vascular aging^[8,22,23^], suggesting potential interplay between metabolic abnormalities and vascular function. However, whether the TyG index and vascular aging independently or jointly contribute to adverse cardiovascular risk, and to what extent they interact, has not been comprehensively examined. A deeper understanding of this relationship could provide valuable insights for refining cardiovascular risk stratification and guiding targeted prevention strategies.

In this context, our study aims to investigate the associations between vascular aging, the TyG index, and primary outcome-major CVD events, and secondary outcomes including CVD mortality, and all-cause mortality within a large, population-based cohort. We specifically assess the independent and joint associations of vascular aging and the TyG index with those outcomes in the PURE-China cohort, and investigate their potential interaction on both multiplicative and additive scales, and explore whether the TyG index mediates the relationship. This study has been reported in accordance with the STROCSS 2025 criteria^[24^], and the STROBE checklist^[25^] is also provided as a Supplemental Digital Content Material, available at: http://links.lww.com/JS9/H80.

Methods

Study design and participants

The details of the PURE study design and data collection methodology for recruitment and follow-up have been previously published and described^[26^]. Briefly, the PURE-China cohort is a population based, prospective study involving 47 931 participants aged 35–70 years from 115 urban and rural communities across 12 provinces in China during 2005–2009, with follow-up until August 2022. For the current analysis, participants who had complete baseline data on systolic blood pressure (SBP), diastolic blood pressure (DBP), fasting blood glucose (FBG), and TG and at least one follow-up assessment were included. Those with missing data on covariates or lost to follow-up were also excluded.

HIGHLIGHTS

• Early vascular aging (EVA) and high triglyceride–glucose (TyG) index were independently linked to cardiovascular disease (CVD) risk.

• EVA combined with high TyG index showed the highest risk for major CVD events.

• Combined assessment of vascular aging and insulin resistance may enhance CVD risk stratification.

Assessment of vascular aging and IR

ePWV was used as a surrogate of vascular aging. ePWV was calculated using the validated equation derived from the Reference Values for Arterial Stiffness Collaboration^[8,9^] and was validated in Chinese populations^[10,11^]. TyG index and ePWV were derived from baseline assessments as participants entered the study. Fasting blood samples were collected after an overnight fast and processed using standardized protocols. Samples were frozen at −70°C in China and analyzed in central certified laboratories in China using standard enzymatic methods^[19^].

ePWV (m/s) = 9.587 − 0.402 × age + 4.560 × 10⁻³ × age² − 2.621 × 10⁻⁵ × age² × MBP + 3.176 × 10⁻³ × age × MBP − 1.832 × 10⁻² × MBP. Here, age is in years, and MBP is expressed in mmHg and calculated as: MBP = (DBP) + 0.4 (SBP − DBP), with SBP and DBP measured in mmHg.

Blood pressure (BP) was measured by trained medical professionals using a mercury sphygmomanometer with an appropriately sized cuff. Three consecutive results of measurements were obtained after participants rest quietly for 5 min in a temperature-controlled room. Measurements were made in a sitting position and using the right arm unless there were special circumstances. Average BP of three measurements were calculated and used for further analysis.

Then, consistent with prior literature and in the absence of universally accepted clinical cut-offs, vascular aging was classified into EVA, healthy vascular aging (HVA), and super healthy vascular aging (SVA), defined as the highest 10%, the middle 80% (10th–90th percentiles), and the lowest 10% of population-based percentiles ePWV distribution^[13,27^]. IR was assessed by TyG-index^[19,20^], which was calculated as: TyG = ln [fasting TG (mg/dL) × FBG (mg/dL)/2].

Assessment of covariates

Demographic variables (age, sex, residence, marital status, and education), lifestyle behaviors (smoking, alcohol consumption, and physical activity), and medical history were collected by standardized interviewer-administered questionnaires. Trained research staff collected measurements of physical parameters (height, weight, waist circumference, and hip circumference) and collected serum for analyses of laboratory markers [FBG, total cholesterol, low-density lipoprotein (LDL) cholesterol, HDL cholesterol, and TG]. Residence was classified as urban or rural, marital status as currently married or other (never married, living with a partner, widowed, separated, and divorced), and educational attainment as primary or less, secondary, or university/college. Smoking and alcohol consumption were categorized as never or ever (current/former). Physical activity was assessed using the International Physical Activity Questionnaire and categorized. Body mass index (BMI) was calculated as weight (kilogram) divided by height squared (meters²). The waist-to-hip ratio (WHR) was calculated as waist circumference divided by hip circumference.

Outcome ascertainment

The primary outcome was major CVD events, including nonfatal myocardial infarction (MI), stroke, and heart failure (HF). Secondary outcomes included CVD mortality and all-cause mortality. Participants were followed through face-to-face interviews or telephone calls annually for the first 3 years, and then every 3 years thereafter to gather information on clinical outcomes from participants or their family members. Events were documented using information from household interviews, medical records, death certificates, and other relevant sources. The causes of mortality were categorized based on the International Classification of Diseases, Tenth Revision (ICD-10) codes, or study-specific codes when ICD-10 codes were unavailable. All outcomes were adjudicated by professional physicians following standardized procedures and definitions. This analysis included all follow-up data recorded through August 2022.

Statistical analysis

Baseline characteristics were presented as number of participants and frequency for categorical variables, mean ± standard deviation for continuous variables for normally distributed continuous variables, and medians with interquartile ranges (IQRs) for abnormally distributed continuous variables. Baseline characteristics were reported by vascular aging categories (EVA, SVA, and HVA). Differences in characteristics were analyzed using chi-square test or analysis of variance test among three vascular aging groups. We examined the correlation between ePWV and TyG index using Pearson correlation coefficients. The Kaplan–Meier method was used to estimate cumulative incidence of all outcomes. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated for both associations between vascular aging and TyG levels on outcomes, using Cox proportional hazards frailty models with center-level random intercepts. The proportional hazards assumption was tested statistically using Schoenfeld residuals (Supplemental Digital Content Figure S1, available at: http://links.lww.com/JS9/H80) and was supported graphically by log-minus-log plots. A directed acyclic graph (DAG) was constructed to illustrate the assumed causal structure among vascular aging, the TyG index, confounders, and cardiovascular outcomes. The DAG (Supplemental Digital Content Figure S2, available at: http://links.lww.com/JS9/H80) guided the selection of covariates in the multivariable models and helped clarify the roles of potential confounders and mediators.

Five adjusted models were fitted. Model 1: adjusted for sex and area of residence; Model 2: Model 1 added adjustments for BMI, educational attainment, smoking status, alcohol status, and physical activity; Model 3: Model 2 further included history of hypertension, diabetes, stroke, coronary artery disease, cancer, WHR, LDL, HDL, TyG index, and marital status; Model 4: Model 3 excluded TyG index; and associations between TyG index and outcomes were evaluated using Models 1–3 with the aforementioned covariates, and additionally adjusted for age in all models and excluded TyG in Model 3.

We conducted exploratory mediation analyses appropriate for survival outcomes using a counterfactual framework, which implements mediation analysis suitable for time-to-event outcomes, reporting natural direct effects, natural indirect effects, and proportion mediated (PM)^[28,29^]. We performed stratified analyses by vascular aging status to evaluate the associations between TyG index and outcomes across predefined subgroups. Interaction on the multiplicative scale was assessed using product terms of vascular aging status and TyG index levels, with HRs and 95% CIs reported. To evaluate interaction on the additive scale, relative excess risk due to interaction, attributable proportion, and synergy index were calculated based on the model coefficients, standard errors, and covariance matrix of the product term and main effects, as previously documented^[30–32^]. To evaluate joint effects, participants were categorized into 12 groups by vascular aging status (EVA, HVA, and SVA) and TyG index levels (quartiles), using group with HVA and lowest TyG index quartile as the reference group. Missingness in covariates was evaluated using Little’s missing completely at random (MCAR) test, which indicated that the data were not MCAR, P < 0.001. The specific missing rates for all covariates in the initial cohort (N = 47 931) are detailed, all <5.2% as shown in Supplemental Digital Content Table S1, available at: http://links.lww.com/JS9/H80.

Sensitivity and subgroup analyses

Several sensitivity analyses were conducted to strengthen the robustness of our findings. For major CVD events and CVD mortality, we added the competing risk analyses using the Fine–Gray model, considering non-cardiovascular deaths as the competing event^[33^]. To evaluate the impact of missing data in the covariates, multiple imputation using chained equations was performed^[34^]. To evaluate the robustness of the results, a composite endpoint of only MI and stroke (excluding HF) was used, as these outcomes have more reliable adjudication. History of hypertension and diabetes were omitted from the fully adjusted model to examine potential over-adjustment bias. To evaluate whether the ePWV could provide additional information for the risk assessment compared with age and BP alone, the HRs and 95% CIs for these two parameters were also estimated.

To assess whether the associations varied across subgroups, we evaluated potential interaction effects using multiplicative interaction terms. Subgroup analyses were conducted only when a statistically significant interaction was identified.

R statistical software (version 3.6.3) and the R packages, including survival (version 3.8-3), survminer (version 0.5.1), MICE (version 3.18.0), naniar (version 1.1.0), interactionR (version 0.1.7), regmedint (version 1.0.1), and SAS (version 9.4), were used for all analyses; a two-sided P < 0.05 was considered statistically significant.

Results

Population characteristics

Of the 47 931 PURE-China participants at baseline, 651 individuals aged outside the 35–70 years range, 5594 with incomplete data on variates, and 1173 without follow-up data were excluded, resulting in 40 513 participants included in the final analysis. The detailed participant selection process is shown in Figure 1. Over a median follow-up of 11.9 years (IQR 9.5–12.5), there were 3615 major CVD events, 829 CVD deaths, and 2344 all-cause deaths (Supplemental Digital Content Table S2, available at: http://links.lww.com/JS9/H80). This corresponded to unadjusted incidence rates of 8.57 (95% CI: 8.29–8.85), 1.92 (95% CI: 1.79–2.05), and 5.42 (95% CI: 5.20–5.64) per 1000 person-years for major CVD events, CVD mortality, and all-cause mortality, respectively. Kaplan–Meier curves were constructed to illustrate the cumulative incidence of outcomes (major CVD events, CVD mortality, and all-cause mortality) and individual components (stroke, HF, and MI), stratified by vascular aging status and TyG index quartiles (Supplemental Digital Content Figures S4 and S5, available at: http://links.lww.com/JS9/H80). Log-rank tests showed significant differences in the composite outcome across categories of both vascular aging status and TyG index.

Figure 1.

Flowchart of study population.

Baseline characteristics stratified by vascular aging status are presented in Table 1. At baseline, 4110 (10.14%) participants were classified as SVA, 32 426 (80.04%) as HVA, and 3977 (9.82%) as EVA. Compared with the SVA group, participants with EVA were older, predominantly male, and exhibited higher BMI, SBP, DBP, FG, total cholesterol, LDL cholesterol, and TG levels. They also had lower educational attainment, reduced physical activity, and a higher prevalence of cardiometabolic comorbidities, except for cancer. A substantially greater proportion of EVA participants were in the highest TyG quartile (35.23%), compared with HVA (25.70%) and SVA (9.03%). The correlation analysis revealed a significant weak correlation between ePWV and the TyG index (r = 0.24, P < 0.001; Supplemental Digital Content Figure S3, available at: http://links.lww.com/JS9/H80).

Table 1.

Baseline characteristics participants according to vascular aging status.

Characteristic	SVA (n = 4110)	HVA (n = 32 426)	EVA (n = 3977)
Age	39.53 ± 4.14	50.99 ± 8.61	63.68 ± 5.27
BMI	22.83 ± 3.11	24.67 ± 3.62	25.49 ± 3.79
Weight	58.66 ± 9.70	64.20 ± 11.59	64.78 ± 11.62
Height	160.15 ± 7.68	161.10 ± 8.27	159.26 ± 8.42
SBP	106.33 ± 8.46	132.33 ± 16.89	170.95 ± 19.26
DBP	67.73 ± 5.88	82.73 ± 10.38	97.96 ± 11.96
FBG	5.11 ± 0.99	5.59 ± 1.53	6.01 ± 1.93
TC	4.24 ± 0.89	4.70 ± 0.98	5.03 ± 1.02
TG	1.16 ± 0.77	1.58 ± 1.12	1.74 ± 1.15
HDL cholesterol	1.31 ± 0.32	1.36 ± 0.33	1.43 ± 0.35
LDL cholesterol	2.40 ± 0.69	2.63 ± 0.77	2.82 ± 0.80
WHR	0.82 ± 0.07	0.86 ± 0.07	0.88 ± 0.07
TyG index
Quantile 1: 4.79–8.23	1859 (45.23)	7670 (23.65)	586 (14.73)
Quantile 2: 8.23–8.42	1151 (28.00)	8150 (25.13)	841 (21.15)
Quantile 3: 8.60–8.79	729 (17.74)	8274 (25.52)	1149 (28.89)
Quantile 4: 9.03–12.55	371 (9.03)	8332 (25.70)	1401 (35.23)
Sex
Female	2962 (72.07)	18 554 (57.22)	2187 (54.99)
Male	1148 (27.93)	13 872 (42.78)	1790 (45.01)
Married	4061 (98.81)	32 236 (99.41)	3960 (99.57)
Education
Primary or less education	829 (20.17)	10 480 (32.32)	2176 (54.71)
Secondary education	2410 (58.64)	17 257 (53.22)	1270 (31.93)
University education	871 (21.19)	4689 (14.46)	531 (13.35)
Smoker
Non-smoker	3333 (81.09)	23 257 (71.72)	2834 (71.26)
Current/former smoker	777 (18.91)	9169 (28.28)	1143 (28.74)
Alcohol
Not drink alcohol	3306 (80.44)	24 212 (74.67)	3029 (76.16)
Current/former drinker	804 (19.56)	8214 (25.33)	948 (23.84)
Location
Urban	2214 (53.87)	16 106 (49.67)	2091 (52.58)
Rural	1896 (46.13)	16 320 (50.33)	1886 (47.42)
Diabetes
No	4073 (99.10)	31 000 (95.60)	3609 (90.75)
Yes	37 (0.90)	1426 (4.40)	368 (9.25)
Hypertension
No	4078 (99.22)	27 181 (83.82)	1905 (47.90)
Yes	32 (0.78)	5245 (16.18)	2072 (52.10)
Physical activity
Low PA	666 (16.20)	4853 (14.97)	668 (16.80)
Moderate PA	1679 (40.85)	13 720 (42.31)	1759 (44.23)
High PA	1765 (42.94)	13 853 (42.72)	1550 (38.97)
CHD
No	4053 (98.61)	30 973 (95.52)	3530 (88.76)
Yes	57 (1.39)	1453 (4.48)	447 (11.24)
Cancer
No	4092 (99.56)	32 208 (99.33)	3952 (99.37)
Yes	18 (0.44)	218 (0.67)	25 (0.63)
Stroke
No	4100 (99.76)	31 903 (98.39)	3746 (94.19)
Yes	10 (0.24)	523 (1.61)	231 (5.81)

BMI, body mass index; DBP, diastolic blood pressure; FBG, fasting blood glucose; HDL cholesterol, high-density lipoprotein cholesterol; LDL cholesterol, low-density lipoprotein cholesterol; SD, standard deviation; SBP, systolic blood pressure; TC, total cholesterol; TG, triglyceride; TyG, triglyceride–glucose; WHR, waist-to-hip ratio; PA, physical activity; CHD, coronary heart disease.

Data are presented as mean ± SD or n (%);

Association between vascular aging and outcomes

The association between vascular aging status and outcomes is summarized in Figure 2 and Supplemental Digital Content Table S3, available at: http://links.lww.com/JS9/H80. After multivariate adjustment for TyG index, sex, residence area, BMI, educational level, smoking and drinking status, physical activity, WHR, LDL and HDL cholesterol, marital status, and history of comorbidities (Model 3), EVA participants had higher risks of major CVD events (aHR = 2.17, 95% CI: 1.99–2.35), CVD mortality (aHR = 4.07, 95% CI: 3.47–4.76), and all-cause mortality (aHR = 2.93, 95% CI: 2.65–3.23). Conversely, the SVA participants had a reduced risk for major CVD events (aHR = 0.33, 95% CI: 0.26–0.41), CVD mortality (aHR = 0.40, 95% CI: 0.24–0.67), and all-cause mortality (aHR = 0.38, 95% CI: 0.29–0.49). In sensitivity analyses, exclusion of the TyG index from the fully adjusted model (Model 4) yielded similar results. Compared with participants with HVA, those with the SVA participants had a lower risk of major cardiovascular events (aHR = 0.32, 95% CI: 0.25–0.40), whereas EVA participants had a higher risk (aHR = 2.18, 95% CI: 2.01–2.36). Similar patterns were observed for cardiovascular mortality and all-cause mortality (Supplemental Digital Content Table S3, available at: http://links.lww.com/JS9/H80).

Figure 2.

Associations of vascular aging status with outcomes. CVD, cardiovascular disease; CI, confidence interval. Model 3 included sex, residence area, BMI, educational level, smoking status, drinking status, physical activity, history of hypertension, diabetes, stroke, coronary artery disease, cancer, waist-to-hip ratio, LDL and HDL cholesterol, TyG, and marital status; model 4, TyG was excluded from the model.

Association between TyG index and outcomes

The association between TyG index levels and outcomes is summarized in Supplemental Digital Content Table S4, available at: http://links.lww.com/JS9/H80. After multivariate adjustment for age, sex, residence area, BMI, educational level, smoking and drinking status, physical activity, WHR, LDL and HDL cholesterol, marital status, and history of comorbidities, higher TyG index levels demonstrated a dose-dependent association with increased risks of major CVD events. Using the lowest quartile (TyG quartile 1) as reference, major CVD events progressively increased across TyG quartiles: TyG quartile 2 = aHR 1.11 (95% CI: 1.00–1.23), TyG quartile 3 = aHR 1.14 (95% CI: 1.03–1.27), and TyG quartile 4 = aHR 1.25 (95% CI: 1.12–1.39; P-trend < 0.001). When analyzed as a continuous variable, each 1-unit increase in the TyG index was associated with a higher risk of CVD (Model 1 aHR = 1.44, 95% CI: 1.35–1.54; Model 2 aHR = 1.25, 95% CI: 1.18–1.32; and Model 3 aHR = 1.15, 95% CI: 1.09–1.22). No significant associations were observed for CVD mortality and all-cause mortality in fully adjusted models.

Expletory mediating role of TyG index in the relationship between vascular aging and major CVD events

Mediation analysis showed that the TyG index did not significantly mediate the association between vascular aging status and outcomes. For all outcomes, total natural indirect effect was close to zero, indicating no significant mediation through TyG index. The PM also supported no statistically significant mediation role of TyG in the association (Table 2).

Table 2.

Mediation analysis decomposing the association of vascular aging with outcomes via the TyG index.

Outcomes	Group	TE	TNIE	TNDE	PM
Major CVD events	EVA	0.843 (0.760, 0.925)	−0.0002 (−0.006, 0.006)	0.832 (0.749, 0.914)	−0.026% (−1.083%, 1.030%)
	SVA	−0.540 (−0.668, − 0.411)	−0.013 (−0.034, 0.009)	−0.525 (−0.653, − 0.397)	1.790% (−1.473%, 5.053%)
All-cause mortality	EVA	1.090 (0.991, 1.189)	0.001 (−0.006, 0.008)	1.084 (0.986, 1.183)	0.113% (−0.912%, 1.138%)
	SVA	−0.467 (−0.614, − 0.321)	−0.001 (−0.026, 0.024)	−0.459 (−0.605, − 0.313)	0.118% (−4.094%, 4.330%)
CVD mortality	EVA	1.444 (1.288, 1.599)	−0.001 (−0.011, 0.009)	1.431 (1.275, 1.587)	−0.092% (−1.390%, 1.206%)
	SVA	−0.349 (−0.612, − 0.086)	−0.030 (−0.074, 0.013)	−0.332 (−0.594, − 0.069)	7.363% (−6.526%, 21.252%)

TE, total effect; TNIE, total natural indirect effect; TNDE, total natural direct effect; PM, proportion mediated; EVA, early vascular aging; SVA, supernormal vascular aging; CVD, cardiovascular disease. HVA served as the reference category.

Interaction and joint effects of vascular aging status and TyG index levels

The interaction effects between vascular aging status and TyG index levels on the studied outcomes were evaluated using both multiplicative and additive scales, as detailed in Table 3. Neither the additive nor multiplicative scales provided statistically robust evidence between vascular aging status and the TyG index with cardiovascular outcomes, although negative interaction was observed in the EVA group. Stratified analyses by vascular aging categories showed the association between the TyG index and CVD in the HVA group, whereas the associations were less pronounced in the SVA and EVA groups (Fig. 3 and Supplemental Digital Content Figure S5, available at: http://links.lww.com/JS9/H80). In the HVA group, compared with TyG quartile 1, the HRs for major CVD events were 1.16 (95% CI: 1.03–1.31) for TyG quartile 2, 1.20 (95% CI: 1.06–1.35) for TyG quartile 3, and 1.26 (95% CI: 1.12–1.43) for TyG quartile 4.

Table 3.

Interaction analysis of vascular aging status and TyG index levels with cardiovascular outcomes.

Outcomes	Group	Multiplicative effect	RERI	AP	SI
Major CVD events	EVA	0.77 (0.66, 0.90)	−0.4 (−0.76, −0.05)	−0.17 (−0.32, −0.01)	0.78 (0.63, 0.96)
	SVA	1.00 (0.60, 1.66)	0.28 (−0.30, 0.86)	0.08 (−0.10, 0.27)	1.14 (0.84, 1.54)
All-cause mortality	EVA	0.84 (0.70, 1.00)	−0.38 (−0.86, 0.11)	−0.13 (−0.31, 0.04)	0.83 (0.66, 1.05)
	SVA	0.89 (0.49, 1.61)	−0.22 (−0.96, 0.52)	− 0.09 (−0.36, 0.18)	0.87 (0.60, 1.25)
CVD mortality	EVA	0.64 (0.48, 0.86)	−1.06 (−2.19, 0.08)	− 0.24 (−0.5, 0.03)	0.77 (0.58, 1.01)
	SVA	0.83 (0.30, 2.32)	0.03 (−1.41, 1.47)	0.01 (−0.53, 0.55)	1.02 (0.42, 2.48)

For the interaction analysis, vascular aging status was categorized into three levels: EVA, SVA, and HVA, with HVA as the reference group. TyG index levels were categorized into two levels: low (Q1–Q2) and high (Q3–Q4). CVD, cardiovascular disease; CI, confidence interval; RERI, relative excess risk due to interaction; AP, attributable proportion; SI, synergy index. Model was adjusted for sex, residence area, BMI, educational level, smoking status, drinking status, physical activity, history of hypertension, diabetes, stroke, coronary artery disease, cancer, waist-to-hip ratio, LDL and HDL cholesterol, and marital status.

Figure 3.

Associations of vascular aging status with outcomes by TyG index. (A) Major CVD events. (B) CVD mortality. (C) All-cause mortality. CVD: cardiovascular disease; CI: confidence interval. Hazard ratios were adjusted for sex, residence area, BMI, educational level, smoking status, drinking status, physical activity, history of hypertension, diabetes, stroke, coronary artery disease, cancer, waist-to-hip ratio, LDL and HDL cholesterol, TyG, and marital status (Model 3).

Joint analysis demonstrated the combined effects of vascular aging status and TyG index levels on the primary outcome (Fig. 4 and Supplemental Digital Content Table S6, available at: http://links.lww.com/JS9/H80). In joint analyses, compared with participants in the HVA group and TyG quartile 1 (reference group), those in the SVA group and TyG quartile 1 had HRs of 0.34 (95% CI: 0.24–0.50) for major CVD events. Conversely, individuals in the EVA group with TyG index quartile 4 exhibited markedly higher risks: HRs were 2.20 (95% CI: 1.75–2.77) for major CVD events, 4.64 (95% CI: 2.87–7.49) for CVD mortality, and 2.60 (95% CI: 1.94–3.48) for all-cause mortality.

Figure 4.

Joint associations of vascular aging status and TyG index levels with incident of outcomes. (A) Major CVD events. (B) CVD mortality. (C) All-cause mortality. CVD, cardiovascular disease; CI, confidence interval. Hazard ratios were adjusted for sex, residence area, BMI, educational level, smoking status, drinking status, physical activity, history of hypertension, diabetes, stroke, coronary artery disease, cancer, waist-to-hip ratio, LDL and HDL cholesterol, TyG, and marital status (Model 3).

Sensitivity and subgroup analyses

Several sensitivity analyses were assessed to confirm the robustness of our findings. In Fine-Gray models that accounted for competing events, the associations of major CVD events and CVD mortality were very similar with main results (Supplemental Digital Content Table S7–S10, available at: http://links.lww.com/JS9/H80). After performing multiple imputations for missing covariate variables, the analysis showed similar findings (Supplemental Digital Content Tables S11 and S12, available at: http://links.lww.com/JS9/H80). Additionally, both redefining the composite endpoint to exclude HF (Supplemental Digital Content Tables S13–S15, available at: http://links.lww.com/JS9/H80) and excluding hypertension and diabetes from primary model (Supplemental Digital Content Tables S3 and S4, available at: http://links.lww.com/JS9/H80) produced materially unchanged effect estimates. Moreover, vascular aging showed a greater association with cardiovascular outcomes than age and SBP alone (Supplemental Digital Content Table S16, available at: http://links.lww.com/JS9/H80). In the subgroup analysis, the associations of vascular aging and the TyG index with cardiovascular outcomes were consistent between male and female, with no meaningful heterogeneity observed (Supplemental Digital Content Tables S17–S21, available at: http://links.lww.com/JS9/H80).

Discussion

In this large, population-based prospective cohort study, we investigated the associations of vascular aging status and IR, assessed by the TyG index, with long-term cardiovascular outcomes, including major CVD events, CVD mortality, and all-cause mortality over nearly 12 years of follow-up. Our findings exploratively found that vascular aging is independently associated with major CVD events, CVD mortality, and all-cause mortality. These associations persist across sensitivity analyses. Importantly, although the TyG index is independently related to major CVD events, it does not significantly mediate the relationship between vascular aging status and outcomes, indicating that vascular aging and IR represent largely independent pathways. No significant interactions between vascular aging status and TyG index levels were observed, while joint analyses further demonstrate that individuals with both EVA and high TyG index levels experience the highest risks of major CVD events and all-cause mortality.

Our findings are in line with evidence highlighting the clinical relevance of vascular aging as a determinant of cardiovascular risk. Previous studies have shown strong associations between arterial stiffness or ePWV and mortality in diverse populations, consistent with our observation that EVA identifies individuals at markedly elevated risk^[35,36^]. For instance, Liu et al have reported similar associations between arterial stiffness and all-cause mortality, supporting our findings that link EVA to an increased risk of cardiovascular events and mortality^[37^]. Vascular aging plays a central role in the morbidity and mortality of older people^[38^], as microcirculation exists in nearly all tissues, and their deterioration can influence multiple organ systems, contributing to CVD. The strong association between EVA and increased cardiovascular risk aligns with earlier studies identifying arterial stiffness as a key predictor of CVD^[27,39^]. Several pathophysiological mechanisms underlying vascular aging have also been implicated in the progression of CVD, including mitochondrial dysfunction^[40^], impaired resistance to molecular stressors^[41^], genomic instability^[42^], telomere attrition and cellular senescence^[43^], epigenetic alterations^[44^], loss of protein homeostasis^[45^], stem cell exhaustion^[46^], and altered intercellular communication in the vascular system. These mechanisms collectively accelerate arterial remodeling and reduce vascular resilience, predisposing individuals to CVD events. Furthermore, the present findings support the utility of ePWV as a scalable, non-invasive surrogate of vascular aging in large epidemiological studies^[36,47,48^].

Building on prior research, our study extends these findings by integrating the TyG index as a surrogate marker of IR, providing a more comprehensive framework to evaluate the metabolic contribution to vascular aging and CVD outcomes. Consistent with previous studies, higher TyG levels showed a clear dose–response association with CVD risk^[21,49^]. The null association with mortality after full adjustment may be partly explained by competing risk effects and the comparatively shorter latency of non-fatal CVD events, which is in line with the finding from PURE study^[19^]. However, the mediation analysis revealed that the IR explained only a very small proportion of the association between EVA and major CVD events. This suggests that although IR contributes to CVD, it is not the predominant mechanism linking vascular aging to clinical outcomes. In other words, vascular aging status and the IR appear to be independently and jointly associated with adverse cardiovascular outcomes. This finding contrasts with some studies that emphasize IR as a central mechanism linking metabolic dysfunction to vascular damage^[42,43^]. Several explanations are possible. First, the TyG index captures only part of the complex spectrum of IR and does not fully represent tissue-specific metabolic dysfunction^[50^]. Second, vascular aging is influenced by multiple mechanisms beyond metabolic impairment, including oxidative stress, extracellular matrix remodeling, and immune dysregulation^[51^]. Third, measurement error and residual confounding, particularly from unmeasured lifestyle, dietary, sleep, or genetic factors, might attenuate observed mediation effects^[52^]. Overall, our results indicate that EVA and IR contribute independently to cardiovascular risk rather than following a simple causal chain, corroborating previous evidence that IR plays a significant role in cardiovascular pathophysiology^[21,53,54^].

We also observed no significant multiplicative or additive interaction between vascular aging status and TyG index levels. It suggests the absence of detectable synergistic effects, although a true biological interaction cannot be excluded due to measurement constraints, shared upstream determinants, and limited precision in some subgroups. This finding might reflect the multifactorial nature of cardiovascular risk, where vascular aging and metabolic dysfunction contribute via distinct, albeit occasionally overlapping mechanisms. For example, vascular aging might primarily affect arterial stiffness and endothelial function^[5^], while IR might drive systemic inflammation and lipid metabolism abnormalities^[55^]. Importantly, the lack of interaction does not imply pathway independence, but rather highlights that joint effects observed likely reflect the accumulation of separate risk processes. The joint exposure analysis nonetheless demonstrated substantial gradient in absolute and relative cardiovascular risk, with individuals exhibiting both EVA and elevated TyG index facing the highest burden of adverse outcomes. These findings support the clinical value of jointly assessing vascular and metabolic status for risk stratification. To our knowledge, this is the first large-scale nationwide study in China to comprehensively examine the association among vascular aging, the TyG index, and cardiovascular outcomes.

Our finding advances the understanding of vascular–metabolic interplay in cardiovascular outcomes^[56^], bridging the gap between vascular biology and metabolic research, and providing a basis for integrated risk assessment frameworks. Moreover, it leverages large-scale cohort data and robust statistical modeling to provide comprehensive insights, setting a foundation for future studies. From a clinical perspective, the combined assessment of vascular aging and IR may help identify Chinese individuals at elevated cardiometabolic risk who could benefit from closer monitoring, early intervention, and optimization of modifiable risk factors. Because both ePWV and the TyG index are simple, inexpensive, and feasible to obtain in routine practice, integrating these markers into existing cardiovascular risk assessment frameworks may improve early detection of high-risk Chinese individuals and support more personalized prevention strategies. In individuals, especially surgical populations, exhibiting both EVA and elevated TyG index levels, attention to BP control, metabolic health, lifestyle modification, and timely cardiovascular evaluation may warrant closer perioperative monitoring and optimization of cardiometabolic status, and may be particularly important to mitigate long-term adverse outcomes.

From a public health perspective, these findings underscore the importance of integrating vascular and metabolic health metrics into routine screenings, paving the way for precision prevention strategies, and interventions targeting metabolic health and arterial stiffness may help mitigate CVD risk, particularly in high-risk subgroups with EVA among Chinese population. Furthermore, healthcare systems could prioritize the development of integrated care models that combine cardiovascular and metabolic risk assessments, thereby optimizing preventive strategies and improving overall population health.

Despite its strengths, our study has several limitations. First, while the prospective cohort design provides valuable insights, residual confounding cannot be entirely ruled out, particularly from unmeasured lifestyle and genetic factors. Second, the observational nature of our study restricts our ability to establish causal relationships. Future research using Mendelian randomization or intervention studies could help explore the causal pathways linking vascular aging, IR, and cardiovascular outcomes. Third, although both ePWV and the TyG index are validated surrogate markers, they remain indirect assessments and may not fully reflect the complexity of vascular aging or IR. Although those makers have been validated in various populations, including Chinese adults, it is an indirect measure that may introduce residual measurement bias. Although BP was measured in triplicate using standardized procedures, random measurement errors is inevitable and may slightly attenuate true associations. Fourth, vascular aging and TyG were measured at baseline only, limiting the ability to assess temporal ordering, and the mediation analysis should therefore be interpreted as exploratory. This also limits causal interpretation of mediation analyses, which should be viewed as exploratory decompositions rather than evidence of temporal causality. Fifth, outcome misclassification is possible, although event adjudication by a central Clinical Event Committee reduces this risk. Sixth, despite the large total sample size, in the joint analysis, the cardiovascular death events in SVA with TyG Q4 was relatively low (n < 10). This affected the accuracy of protective risk estimation for these subgroups, although it did not influence the main conclusion of this study that EVA significantly increases cardiovascular risk. The observations regarding SVA require validation in larger cohorts. Finally, our study population was limited to individuals aged 35–70 years in China, and the exclusion of participants with missing data may have introduced selection bias, potentially limiting the generalizability of our results. Future studies should aim to include more diverse populations to validate the broader applicability of our findings.

Conclusions

In conclusion, this study provides important evidence that vascular aging and IR independently contribute to the long-term risk of major cardiovascular events and mortality. Combined assessment of these two factors enhances risk stratification and may help identify individuals at particularly high cardiovascular risk in China. While we explored the potential mediating role of the TyG index as a secondary objective, the observed associations between vascular aging and cardiovascular outcomes may involve pathways beyond IR, which may not be fully captured by the TyG index. Notably, ePWV and TyG are non-invasive and accessible markers, but their routine use for integrated vascular–metabolic assessment requires validation in clinical trials. Future research should explore causal pathways using longitudinal metabolic and vascular measurements and evaluate whether interventions targeting metabolic health and vascular function can jointly mitigate CVD risk.

Ethical approval

Ethical approval was obtained from the Institutional Review Board at Fuwai Hospital of the Chinese Academy of Medical Sciences and the Beijing Hypertension League Institute.

Consent

All participants provided written informed consent.

Sources of funding

The PURE study is an investigator-initiated study that is supported by the Population Health Research Institute (PHRI), Hamilton Health Sciences Research Institute (HHSRI), the Canadian Institutes of Health Research, and Heart and Stroke Foundation of Ontario. PURE-China study is partially supported by the National Science and Technology Major Project of the Ministry of Science and Technology of China (Grant Nos. 2024ZD0528106; 2024-KJCX37), the National Center for Cardiovascular Diseases, the National Clinical Research Center for Cardiovascular Diseases, Fuwai Hospital, Chinese Academy of Medical Sciences, National High-level Hospital Clinical Research Funding, and Shenzhen High-level Hospital Construction Fund (Grant Nos. NCRCSZ-2023-012; 2022-GSP-PT-10; 2023-GSP-GG-1; 2023-GSP-GG-36).

Author contributions

Conceptualization: Q.C., B.W., Z.L., L.A.T., S.Y., S.R., B.H., and W.L. Data curation: Q.C., B.W., M.Y., and X.L. Formal analysis: Q.C. and B.W. Funding acquisition: B.H. and W.L. Investigation: Q.C., B.W., M.Y., and X.L. Methodology: Q.C., B.W., and W.L. Project administration: Z.L. and B.H. Software: Q.C. and B.W. Supervision: W.L. Validation: Q.C. and B.W. Writing – original draft: Q.C., B.W., and M.Y. Writing – review and editing: E.A.v.R., P.J., S.R., S.Y., and W.L.

Conflicts of interest disclosure

All authors declared no competing interests.

Research registration unique identifying number (UIN)

ClinicalTrials.gov ID: NCT03225586.

Guarantor

Bo Hu and Wei Li.

Provenance and peer review

Not commissioned, externally peer-reviewed.

Data availability statement

The datasets generated and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Presentation

None.