Association of the triglyceride glucose-Chinese visceral adiposity index with incident cardiometabolic multimorbidity in middle-aged and older adults: a nationwide prospective cohort study

Abstract

Objective

Cardiometabolic multimorbidity (CMM) is a growing global health challenge. Whether the baseline or cumulative triglyceride glucose and Chinese visceral adiposity index product (TyG-CVAI) can predict incident CMM remains unclear.

Methods

We constructed two prospective cohorts from the China Health and Retirement Longitudinal Study (CHARLS): Cohort 1 (n = 8895 patients) to assess the association of the baseline TyG-CVAI with CMM and Cohort 2 (n = 5839 patients) to assess the association of the cumulative TyG-CVAI with CMM. The cumulative TyG-CVAI was calculated as the average TyG-CVAI between baseline and the 2015 wave multiplied by the exposure time. Incident CMM was confirmed via a self-reported physician diagnosis, medication use, and clinical data. Cox regression models were used to estimate hazard ratios (HRs). Nonlinearity was assessed using restricted cubic splines, and predictive performance was evaluated by performing a receiver operating characteristic (ROC) curve analysis.

Results

During follow-up, 875 and 492 incident CMM cases were documented in Cohort 1 and Cohort 2, respectively. Both the baseline and cumulative TyG-CVAI showed graded, positive associations with the CMM risk. Compared with the lowest quartile, the highest quartile was associated with significantly increased risks (baseline: HR = 1.93, 95% CI = 1.46–2.54; cumulative: HR = 1.76, 95% CI = 1.22–2.53). Significant nonlinear relationships with threshold effects were observed for both indices (P for nonlinearity < 0.001). Furthermore, compared with their individual components (TyG or CVAI), both the baseline and cumulative TyG-CVAI demonstrated superior predictive ability for CMM, as indicated by a larger area under the ROC curve.

Conclusions

 Both the baseline and cumulative TyG-CVAI are independent and nonlinear predictors of incident CMM, outperforming TyG or CVAI alone. This easily obtainable metric may enhance risk stratification and help identify high-risk individuals for early preventive intervention.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-026-03091-6.

Research insights

What is currently known about this topic?

• CMM is a major global health burden. While the individual components TyG and CVAI are established metabolic risk markers, whether their composite product, TyG-CVAI, can predict future CMM remains unexplored.

What is the key research question?

• Can single-measurement (baseline) and cumulative TyG-CVAI predict new-onset CMM, and do these associations exhibit nonlinear, threshold-dependent relationships?

What is new?

• This study provides the first evidence that the baseline and cumulative TyG-CVAI are superior, nonlinear predictors of CMM compared to TyG or CVAI alone, with identified risk thresholds.

How might this study influence clinical practice?

• This easily calculable TyG-CVAI metric can serve as a practical tool for the early identification of high-risk individuals. It enables enhanced risk stratification, potentially guiding more intensive monitoring and timely preventive interventions.

Introduction

Cardiometabolic multimorbidity (CMM) [1]—the co-occurrence of at least two conditions among diabetes, coronary heart disease, and stroke—is emerging as a significant clinical and public health challenge worldwide [2–4]. Individuals with CMM face a substantially reduced quality of life, increased healthcare costs, and markedly increased risks of disability and mortality compared with those with a single disease [2]. The increasing prevalence of CMM underscores the urgent need for effective strategies for early identification and prevention, which hinge on the discovery of robust and practical predictive biomarkers.

The pathogenesis of CMM is deeply rooted in two interconnected pathophysiological processes: insulin resistance (IR) and visceral adiposity [5]. The triglyceride glucose index (TyG), a simple and reliable surrogate marker of IR, has been shown to predict the incidence of individual cardiovascular diseases (CVDs) [6, 7] and diabetes [8, 9]. Currently, the Chinese visceral adiposity index (CVAI), a sex-specific index that was developed and validated in the Chinese population [10], effectively captures the burden of visceral fat dysfunction and is associated with cardiometabolic disorders [11]. Given that IR and visceral adiposity synergistically drive metabolic deterioration, a composite index that integrates both aspects may offer a more holistic reflection of an individual’s metabolic status [5]. The TyG-CVAI, which is calculated as the product of the TyG and CVAI [10], has recently been confirmed to be a marker of incident stroke [10, 12] and CVD [13]. However, the longitudinal associations of the baseline and cumulative TyG-CVAI with incident CMM and the potential nonlinearity or threshold effects remain poorly understood in the general middle-aged and older population.

CMM is a progressive condition that results from the sustained, cumulative burden of metabolic dysregulation over time. Therefore, using data from the large, nationally representative China Health and Retirement Longitudinal Study (CHARLS), our objective was to investigate the association between the baseline and cumulative TyG-CVAI (cuTyG-CVAI) and the risk of incident CMM. We hypothesized that the baseline and cumulative TyG-CVAI would be stronger predictors of CMM than its components, including TyG and CVAI, by more accurately reflecting the pathogenic burden of insulin resistance and visceral adiposity, offering a superior tool for risk stratification in the prevention of complex multimorbidity.

Methods

Study design and study population

This study used data from the CHARLS [14], an ongoing nationwide prospective cohort study of Chinese adults aged 45 years and older [15]. The present analysis is a prospective cohort study based on data from the 2011 baseline survey and subsequent follow-up waves until the most recent available wave. The initial study population consisted of 17,705 participants enrolled in the CHARLS at the baseline survey during 2011–2012 (Wave 1). Follow-up activities occurred in 2013 (Wave 2, n = 15,960), 2015 (Wave 3, n = 14,286), 2018 (Wave 4, n = 12,889), and 2020 (Wave 5, n = 11,572) [16].

This study established two longitudinal cohorts. Cohort 1 was constructed to analyse the association between the TyG-CVAI at baseline and incident CMM during the follow-up period. We included subjects aged ≥ 45 years in 2011 (with participants aged 46–92 years), resulting in 8895 eligible individuals after excluding those younger than 45 years (n = 525) and those with missing TyG-CVAI-related components (triglyceride (TG) levels, fasting plasma glucose (FPG) levels, high-density lipoprotein cholesterol (HDL-C) levels, height, weight or waist circumference (WC)) in 2011 (n = 7643), with prevalent CMM in 2011 (n = 418), or with missing follow-up CMM data (n = 224) to assess the association between the baseline TyG-CVAI and CMM risk during the full follow-up. Cohort 2 was constructed to assess the association of the cuTyG-CVAI (2011–2015) and incident CMM after 2015. Cohort 2 was derived from Cohort 1 by further excluding participants with missing TyG-CVAI-related components in 2015 (n = 2860) and those with CMM in 2015 (196), yielding 5,839 subjects for analysis (Fig. 1).

Fig. 1.

Flowchart of the population selection process in the study

Demographic characteristics

The baseline characteristics of the included participants were collected through interviews, physical examinations, and laboratory tests. These characteristics included age, sex, marital status, highest educational level, smoking status, and drinking status. Clinical measurements such as blood pressure, lipid profiles (total cholesterol (TC), TG, HDL-C, and low-density lipoprotein cholesterol (LDL-C) levels), and FPG levels were also obtained. The specific methods used were the same as those described previously [12, 17, 18].

Definitions of the main variables

The TyG index was calculated as Ln [fasting TG (mg/dL) × FPG (mg/dL)/2] [19]. CVAI was calculated using a previously validated sex-specific formula based on WC, body mass index (BMI), and TG and HDL-C levels [20]. More precisely, for men, the CVAI = − 267.93 + 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), whereas for women, the 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).

The TyG-CVAI was defined as the arithmetic product of the TyG index and the CVAI [12]. The calculation of cuTyG-CVAI was constrained by data availability in the CHARLS survey, where the complete panel of biomarkers required (TG, FPG, and HDL-C levels) was only consistently collected during the 2011 (baseline) and 2015 waves. The formula (TyG-CVAI at baseline + TyG-CVAI in 2015)/2 × time (2015–2012) was calculated as previously described [10] and was an established method in the CHARLS literature to estimate time-weighted average exposure when serial measurements are limited, providing a pragmatic estimate of the cumulative metabolic burden.

Hypertension was defined as meeting any of the following criteria: a self-reported physician diagnosis, use of an antihypertensive medication, an average systolic blood pressure ≥ 140 mmHg or an average diastolic blood pressure ≥ 90 mmHg during the study examination [21]. Dyslipidaemia was defined as meeting any of the following criteria: a self-reported physician diagnosis, use of a lipid-lowering medication, and objectively measured lipid levels meeting any of the thresholds of a total cholesterol level ≥ 6.22 mmol/L, triglyceride level ≥ 2.26 mmol/L, high-density lipoprotein cholesterol (HDL-C) level < 1.04 mmol/L, or low-density lipoprotein cholesterol (LDL-C) level ≥ 4.14 mmol/L [22]. Kidney disease was defined as meeting any of the following criteria: a self-reported physician diagnosis or objective laboratory calculation of the estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m². The eGFR was calculated using the Modification of Diet in Renal Disease equation based on the from Chinese subjects with chronic kidney disease. Specifically, eGFR = 175 × serum creatinine (mg/dL)^−1.234 × age (years)^−0.179 (× 0.79 for females) [21].

Ascertainment of CMM

Diabetes was diagnosed by a self-reported physician diagnosis, use of an antidiabetic medication, FPG level ≥ 126 mg/dL, or an HbA1c level ≥ 6.5% [23]. Heart disease was diagnosed by a self-reported physician diagnosis of coronary heart disease, heart attack, other heart problems, or the use of cardiovascular drugs. Stroke was defined as a self-reported physician diagnosis or the use of any treatment (traditional Chinese medicine/Western medicine/physical therapy/acupuncture and moxibustion/occupational therapy) to control stroke [10].

CMM was uniformly defined as the confirmed coexistence of at least two of the three diseases listed above (diabetes, heart disease, or stroke) in the same individual during the follow-up period [24, 25]. In this study, individuals with prevalent multimorbidity (≥ 2 diseases) at baseline were excluded. The at-risk population for incident CMM included participants with zero or one confirmed cardiometabolic disease at baseline. Incident CMM was strictly defined as a new diagnosis of a second, distinct cardiometabolic disease during follow-up based on a physician’s diagnosis, medication use, or clinical measurements. The period of onset of CMM was defined as the interval from the last interview to the time a participant was diagnosed with a second distinct cardiometabolic disease (diabetes, heart disease, or stroke) in which the patient transitioned from having 0 or 1 disease to a state of multimorbidity (≥ 2 diseases) [25].

Handling missing data

Approximately 20% of the data points were missing across the collected variables (Supplementary Table 1). Prior to imputation, anthropometric measurements (height, weight, WC, etc.) were screened for outliers using the interquartile ranges (IQRs); values exceeding 1.5 times the IQRs were winsorized to the respective boundaries to prevent the distortion of the imputation process.

The correlations between missing values were assessed to examine the pattern of missingness. Systematic patterns with strong correlations in missingness among several variables suggested that the data were likely missing at random (MAR). Consequently, the multiple imputation by chain equations (MICE) method was employed to handle these missing values. Five independent imputed datasets were generated. The final results were obtained by pooling estimates across these datasets using Rubin’s rules, ensuring valid statistical inferences. Missing data for sex and age exhibited weak correlations with other variables, indicating a pattern consistent with missing completely at random (MCAR). Given their low proportions and independent nature, observations with missing values for these two variables were listwise deleted.

Statistical analysis

All analyses were performed with R software (version 4.2.1). Participants were categorized into four groups based on their baseline and cuTyG-CVAI quartiles (Q1-Q4) or into two groups based on whether they developed CMM during the follow-up period, namely, the “non-CMM” and “CMM” groups. Baseline characteristics were compared across groups. Skewed continuous variables are reported as medians (IQRs); normally distributed continuous data are shown as means ± SDs; and categorical variables are reported as counts (%). For group comparisons, χ² tests were used for categorical variables, and t tests or nonparametric tests (Kruskal‒Wallis/Mann‒Whitney) were used for continuous variables. The cumulative incidence of CMM in the quartiles was visualized using Kaplan‒Meier curves, and the log-rank test was used to assess differences.

The associations between the TyG-CVAI quartiles (both baseline and cumulative) and the risk of incident CMM were evaluated using Cox proportional hazards models, with the lowest quartile (Q1) used as the reference. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated. Four sequential models were constructed: Model 1 was unadjusted; Model 2 was adjusted for demographic factors (sex, age, nationality, marital status and highest education level); Model 3 was further adjusted for smoking, drinking, BMI, and eGFR; and Model 4 was additionally adjusted for clinical confounders (hypertension, dyslipidaemia, hypertension treatment, and dyslipidaemia treatment). The proportional hazards assumption was tested with the Schoenfeld residuals.

The nonlinear relationship between the cuTyG-CVAI and CMM risk was examined using restricted cubic splines (RCSs) with four knots. The performance of the TyG, CVAI, baseline TyG-CVAI, and cuTyG-CVAI for predicting incident CMM was compared by analysing receiver operating characteristic (ROC) curves, and the areas under the curve (AUCs) were calculated. DeLong’s test was employed to statistically compare the differences between the AUCs of these indicators.

We further evaluated the robustness of our findings by performing comprehensive subgroup analyses of patients stratified by key demographic and clinical characteristics, including sex, age, nationality, marital status, highest educational level, smoking status, drinking status, BMI category, and the presence of hypertension, dyslipidaemia, diabetes, heart disease, stroke, or kidney disease at baseline.

Additionally, we conducted a series of sensitivity analyses to assess the stability of our primary results. First, we performed analyses using multiple imputation to address the potential bias from missing data. Second, given the strong correlations among baseline hypertension, diabetes, and dyslipidaemia, we repeated the Cox regression analysis after excluding participants with any of these conditions at baseline to minimize concerns regarding multicollinearity. Third, we excluded participants who were diagnosed with CMM within the first year of follow-up to reduce the potential for reverse causality. Fourth, since lipid-lowering or glucose-lowering medications can directly affect FPG, TG, and HDL-C levels (components of TyG and CVAI), we conducted an analysis in which users of these medications at baseline or during follow-up were excluded. Fifth, we excluded participants with baseline or cumulative TyG-CVAI values in the top and bottom 1% of the distribution to ensure that our findings were not driven by extreme value. Finally, we performed a competing risk regression analysis using the Fine and Gray subdistribution hazards model to account for the potential influence of non-CMM death.

A two-sided P value < 0.05 was considered to indicate statistical significance in all analyses.

Results

Demographic characteristics of the subjects studied at baseline

We established two cohorts using CHARLS data to evaluate the relationships between baseline TyG-CVAI and cuTyG-CVAI with incident CMM. A total of 8,895 and 5,839 participants were included in Cohort 1 and Cohort 2, respectively. The subjects were stratified into four groups according to the baseline TyG-CVAI and cuTyG-CVAI. The baseline characteristics of the participants are summarized in Table 1 and Supplementary Table 2.

Table 1.

Characteristics of the participants at baseline by quartiles of TyG-CVAI in 2011

Variables	Total (n = 8895)	Q1 (n = 2223)	Q2 (n = 2224)	Q3 (n = 2224)	Q4 (n = 2224)	P
Age, year	58 (52, 65)	56 (49,63)	57 (51,64)	58 (52,65)	60 (54,67)	< 0.001
Age category, n (%)						< 0.001
< 60	5043 (56.69)	1455 (65.45)	1358 (61.06)	1212 (54.50)	1018 (45.77)
≥ 60	3852 (43.31)	768 (34.55)	866 (38.94)	1012 (45.50)	1206 (54.23)
Sex, n (%)						< 0.001
Male	4149 (46.64)	1298 (58.39)	970 (43.62)	864 (38.85)	1017 (45.73)
Female	4746 (53.36)	925 (41.61)	1254 (56.38)	1360 (61.15)	1207 (54.27)
Nationality, n (%)						0.097
Han nationality	7757 (87.21)	1905 (85.70)	1945 (87.46)	1956 (87.95)	1951 (87.72)
National minority	1138 (12.79)	318 (14.30)	279 (12.54)	268 (12.05)	273 (12.28)
Married status, n (%)						0.002
Married	7784 (87.51)	1985 (89.29)	1961 (88.17)	1933 (86.92)	1905 (85.66)
Others	1111 (12.49)	238 (10.71)	263 (11.83)	291 (13.08)	319 (14.34)
Highest education level, n (%)						< 0.001
Illiteracy	2605 (29.29)	561 (25.24)	673 (30.27)	713 (32.06)	658 (29.59)
Primary school	3667 (41.23)	988 (44.44)	932 (41.93)	864 (38.85)	883 (39.70)
Middle school or above	2622 (29.48)	674 (30.32)	618 (27.80)	647 (29.09)	683 (30.71)
BUN, mg/dL	15.74 ± 4.50	15.59 ± 4.51	15.49 ± 4.38	15.72 ± 4.38	16.15 ± 4.69	< 0.001
Creatinine, mg/dL	0.78 ± 0.22	0.76 ± 0.19	0.77 ± 0.29	0.81 ± 0.21	0.78 ± 0.18	< 0.001
TC, mg/dL	190.59 (167.40, 215.34)	182.48 (160.83,206.25)	188.66 (165.08,211.57)	192.53 (169.33,217.27)	199.49 (175.13,225.39)	< 0.001
TG, mg/dL	105.32 (74.34, 152.22)	73.46 (57.53,95.14)	92.04 (71.68,120.58)	117.71 (87.61,155.98)	164.61 (120.36,242.49)	< 0.001
HDL-C, mg/dL	49.48 (40.59, 60.31)	59.54 (50.26,70.75)	53.35 (45.62,62.63)	47.55 (40.21,55.77)	40.21 (33.63,47.17)	< 0.001
LDL-C, mg/dL	114.24 (93.17, 137.24)	107.09 (88.14,127.19)	114.82 (94.72,135.02)	118.69 (98.20,141.50)	118.30 (93.56,143.43)	< 0.001
Serum UA, mg/dL	4.44 ± 1.24	4.28 ± 1.18	4.22 ± 1.16	4.39 ± 1.21	4.87 ± 1.32	< 0.001
FPG, mg/dL	102.24 (94.32, 112.50)	98.10 (91.08,106.74)	100.26 (92.88,108.18)	102.96 (95.40,113.09)	108.72 (99.36,123.30)	< 0.001
CRP, mg/dL	1.03 (0.55, 2.14)	0.72 (0.42,1.53)	0.79 (0.48,1.75)	1.10 (0.62,2.15)	1.54 (0.85,3.01)	< 0.001
HbA1c, %	5.1 (4.9, 5.4)	5.1 (4.8,5.3)	5.1 (4.8,5.4)	5.1 (4.9,5.4)	5.3 (5.0,5.6)	< 0.001
eGFR, ml/min/1.73m2	108.35 ± 28.77	111.42 ± 26.77	110.75 ± 28.82	108.36 ± 27.95	102.87 ± 30.62	< 0.001
eGFR stage						< 0.001
≥ 90 ml/min/1.73m2	6669 (75.00)	1779 (80.03)	1732 (77.88)	1687 (75.85)	1471 (66.23)
60–89 ml/min/1.73m2	2066 (23.23)	416 (18.71)	463 (20.82)	501 (22.53)	686 (30.89)
< 60 ml/min/1.73m2	157 (1.77)	28 (1.26)	29 (1.30)	36 (1.62)	64 (2.88)
Cystatin C, mg/dL	0.99 (0.86, 1.13)	0.99 (0.88,1.14)	0.97 (0.85,1.11)	0.98 (0.85,1.11)	1.01 (0.86,1.16)	< 0.001
Smoking, n (%)						< 0.001
No	5412 (60.89)	1107 (49.86)	1407 (63.29)	1491 (67.13)	1407 (63.26)
Yes	3476 (39.11)	1113 (50.14)	816 (36.71)	730 (32.87)	817 (36.74)
Drinking, n (%)						< 0.001
No	6642 (74.72)	1494 (67.27)	1693 (76.16)	1756 (79.06)	1699 (76.39)
Yes	2247 (25.28)	727 (32.73)	530 (23.84)	465 (20.94)	525 (23.61)
Height, cm	157.9 ± 8.6	158.3 ± 7.9	157.4 ± 8.4	157.0 ± 8.7	158.9 ± 9.0	< 0.001
Weight, kg	58.5 ± 11.1	51.1 ± 8.1	55.1 ± 8.3	59.4 ± 8.9	68.2 ± 10.9	< 0.001
BMI, kg/m2	23.39 ± 3.67	20.32 ± 2.43	22.20 ± 2.46	24.04 ± 2.63	26.96 ± 3.31	< 0.001
WC, cm	84.2 (77.5, 91.5)	74.2 (70.1,78.4)	81.5 (78.0,84.5)	87.5 (84.0,91.0)	96.1 (92.0,100.0)	< 0.001
TyG	8.67 ± 0.66	8.22 ± 0.47	8.47 ± 0.46	8.76 ± 0.51	9.23 ± 0.69	< 0.001
CVAI	91.51 (65.21, 121.26)	47.98 (32.87,57.92)	78.91 (71.73,85.38)	105.04 (97.64,113.05)	142.20 (130.42,157.43)	< 0.001
SBP, mmHg	127 (115, 143)	120 (110,134)	124(113,138)	129(117,145)	136 (123,150)	< 0.001
DBP, mmHg	76 ± 12	73 ± 11	74 ± 12	77 ± 12	80 ± 12	< 0.001
HR, beats/min	72± 10	71± 11	72±10	73±10	74 ± 10	< 0.001
Hypertension, n (%)						< 0.001
No	4926 (55.96)	1598 (72.47)	1409 (63.96)	1120 (50.95)	799 (36.37)
Diagnosed before 2011	2389 (27.14)	297 (13.47)	443 (20.11)	633 (28.80)	1016 (46.24)
New onset after 2011	1488 (16.90)	310 (14.06)	351 (15.93)	445 (20.25)	382 (17.39)
Dyslipidemia, n (%)						< 0.001
No	6516 (74.33)	1871 (84.82)	1776 (80.91)	1558 (71.21)	1311 (60.19)
Diagnosed before 2011	933 (10.64)	109 (4.94)	160 (7.29)	240 (10.97)	424 (19.47)
New onset after 2011	1318 (15.03)	226 (10.24)	259 (11.80)	390 (17.82)	443 (20.34)
Diabetes, n (%)						< 0.001
No	6409 (72.48)	1835 (82.88)	1782 (80.52)	1549 (70.09)	1243 (56.38)
Diagnosed before 2011	1299 (14.69)	178 (8.04)	212 (9.58)	357 (16.15)	552 (25.03)
New onset after 2011	1134 (12.83)	201 (9.08)	219 (9.90)	304 (13.76)	410 (18.59)
Heart problems, n (%)						< 0.001
No	6839 (77.75)	1832 (82.97)	1737 (78.88)	1681 (76.37)	1589 (72.72)
Diagnosed before 2011	934 (10.62)	182 (8.24)	225 (10.22)	255 (11.59)	272 (12.45)
New onset after 2011	1023 (11.63)	194 (8.79)	240 (10.90)	265 (12.04)	324 (14.83)
Stroke, n (%)						< 0.001
No	7992 (90.87)	2078 (94.07)	2046 (92.95)	1975 (89.85)	1893 (86.56)
Diagnosed before 2011	204 (2.32)	43 (1.95)	36 (1.64)	60 (2.73)	65 (2.97)
New onset after 2011	599 (6.81)	88 (3.98)	119 (5.41)	163 (7.42)	229 (10.47)
Kidney disease, n (%)						< 0.001
No	7577 (86.14)	1868 (84.76)	1939 (88.14)	1900 (86.36)	1870 (85.31)
Diagnosed before 2011	699 (7.95)	215 (9.75)	149 (6.77)	171 (7.78)	164 (7.48)
New onset after 2011	520 (5.91)	121 (5.49)	112 (5.09)	129 (5.86)	158 (7.21)

BUN, blood urea nitrogen; FPG, fasting plasma glucose; TC, Total cholesterol; TG, Triglycerides; LDL, low-density lipoproteins cholesterol; HDL, high-density lipoproteins cholesterol; CRP, C-reactive protein; HbA1c, glycosylated hemoglobin; eGFR, Estimated Glomerular Filtration Rate; UA, uric acid; TyG, triglyceride-glucose index; CVAI, Chinese visceral adiposity index; TyG-CVAI, triglyceride glucose Chinese visceral adiposity index; SBP, systolic blood pressure; DBP, diastolic blood pressure; HR, Heart rate; BMI, body mass index; WC, waist circumference.

Table 2.

Threshold effect analysis of TyG-CVAI at baseline, cuTyG-CVAI, TyG or CVAI at baseline on new-onset CMM using a two-piecewise linear regression model

Variable	Model 1	P	Model 2	P	Model 3	P
TyG-CVAI at baseline
Standard regression	1.58 (1.40–1.79)	< 0.001	1.51 (1.33–1.69)	< 0.001	1.32 (1.14–1.52)	< 0.001
Fitting by two-piecewise linear regression
Inflection point	787.81		787.81		787.81
TyG-CVAI < 787.81	1.29 (1.14–1.43)	< 0.001	1.13(0.99–1.28)	0.132	0.98(0.86–1.11)	0.625
TyG-CVAI ≥ 787.81	2.66 (2.25–3.13)	< 0.001	2.51 (2.12–2.96)	< 0.001	1.58 (1.29–1.92)	< 0.001
P for Log-likelihood ratio	< 0.001		< 0.001		< 0.001
cuTyG-CVAI
Standard regression	1.49 (1.27–1.69)	< 0.001	1.40 (1.22–1.59)	< 0.001	1.25 (1.09–1.43)	< 0.001
Fitting by two-piecewise linear regression
Inflection point	2470.49		2470.49		2470.49
cuTyG-CVAI < 2470.49	1.19(1.08–1.32)	< 0.001	1.11(0.96–1.27)	0.233	0.92 (0.81–1.05)	0.743
cuTyG-CVAI ≥ 2470.49	2.52 (2.15–2.96)	< 0.001	2.45 (2.09–2.87)	< 0.001	1.56 (1.30–1.88)	< 0.001
P for Log-likelihood ratio	< 0.001		< 0.001		< 0.001
TyG at baseline
Standard regression	1.38 (1.22–1.43)	< 0.001	1.25 (1.15–1.36)	< 0.001	1.17 (1.09–1.27)	0.021
Fitting by two-piecewise linear regression
Inflection point	8.59		8.59		8.59
TyG < 8.59	1.17(1.04–1.32)	0.032	1.08(0.92–1.25)	0.403	0.91(0.78–1.06)	0.751
TyG ≥ 8.59	1.97 (1.68–2.30)	< 0.001	1.95 (1.67–2.28)	< 0.001	1.44 (1.23–1.70)	< 0.001
P for Log-likelihood ratio	< 0.001		< 0.001		< 0.001
CVAI at baseline
Standard regression	1.47 (1.28–1.67)	< 0.001	1.30 (1.15–1.46)	< 0.001	1.15 (1.02–1.30)	0.027
Fitting by two-piecewise linear regression
Inflection point	91.51		91.51		91.51
CVAI < 91.51	1.23(1.12–1.35)	< 0.001	1.12(1.01–1.34)	0.045	0.94(0.76–1.13)	0.816
CVAI ≥ 91.51	2.48 (2.11–2.91)	< 0.001	2.33 (1.97–2.74)	< 0.001	1.41 (1.16–1.72)	< 0.001
P for Log-likelihood ratio	< 0.001		< 0.001		< 0.001

HR, Hazard ratio; CI, confidence interval

Model 1: Unadjusted;

Model 2: Adjust sex, marital status, educational level, nationality, age, smoking status and drinking status;

Model 3: Adjusted sex, nationality, married status, highest education level, age, smoking, drinking, BMI, eGFR, Cystatin C, hypertension, dyslipidemia, hypertension treatment, and dyslipidemia treatment

The participants in the higher TyG-CVAI quartiles were generally older, with a greater proportion aged ≥ 60 years. The sex distribution varied significantly between the quartiles, with a greater proportion of women in Q2 and Q3. Educational attainment and marital status also differed significantly. No significant difference in nationality was observed between the groups. Key metabolic parameters, including BMI, WC, TyG, and CVAI, increased consistently across the quartiles (all P < 0.001). Similarly, TG, TC, FPG, and C-reactive protein (CRP) levels were significantly elevated in the high TyG-CVAI group. Moreover, the prevalence of nonexistent hypertension, dyslipidaemia, diabetes, heart problems, and stroke gradually decreased from Q1 to Q4 (all P < 0.001) (Table 1).

In addition, participants in Cohort 1 and Cohort 2 were grouped based on whether they developed CMM, and their demographic characteristics are presented in Supplementary Tables 3 and 4, respectively.

These findings indicate that higher baseline TyG-CVAI and cuTyG-CVAI values were associated with an older age, unfavourable metabolic profiles, and a greater burden of cardiometabolic risk factors and diseases at baseline.

The incidence rate of CMM

During the 2011–2020 follow-up period, a total of 875 incident CMM events were identified among the 8895 participants without CMM at baseline in Cohort 1. In addition, 492 incident CMM events were diagnosed from 2015 to 2020 in Cohort 2. These values corresponded to overall incidence densities of 197.78 and 189.46 cases per 10,000 person-years, respectively (Supplementary Table 5). When patients were stratified into quartiles by the TyG-CVAI at baseline, a clear gradient for the CMM risk was observed. The incidence rate was lowest in the first quartile (4.86% in Q1) and increased progressively across the higher quartiles (6.65% in Q2; 10.57% in Q3; and 17.27% in Q4). Similarly, compared with participants in the lowest quartile (Q1), participants in the highest cuTyG-CVAI quartile (Q4) exhibited a significantly increased risk of developing CMM (P for trend < 0.001) (Supplementary Table 5).

Nonlinear relationship between the TyG-CVAI and CMM risk

We performed RCS analyses to further determine the type of association between the TyG-CVAI and the risk of CMM. As illustrated in Fig. 2, a nonlinear relationship was observed between the TyG-CVAI and the incidence of CMM. A significant threshold effect was identified for both the baseline and cumulative TyG-CVAI. The risk increased modestly below specific inflection points but increased sharply above them. For the TyG-CVAI at baseline, the inflection point was 787.81 (Table 2). In the model adjusted for multiple factors (Table 2, Model 3), the hazard ratio was 0.98 (95% CI 0.86–1.11) below this threshold but increased significantly to 1.58 (95% CI 1.29–1.92) per unit increase above it (P for log-likelihood ratio < 0.001; Table 2). Similar J-shaped patterns and significant inflection points were observed for the baseline cuTyG-CVAI (2470.49), TyG (8.59), and CVAI (91.51), with hazard ratios increasing more steeply above these thresholds. Consistent nonlinear associations were also observed when the individual components of CMM (diabetes, heart disease, and stroke) were analysed in relation to the TyG-CVAI at baseline (Supplementary Fig. 1) and cuTyG-CVAI (Supplementary Fig. 2). These findings suggest the existence of critical thresholds for metabolic health, beyond which the risk of developing cardiometabolic diseases accelerates disproportionately.

Fig. 2.

Restricted cubic splines (RCS) analysis exploring the relationship between TyG-CVAI at baseline (A), cuTyG-CVAI, TyG-CVAI (B), TyG at baseline (C), CVAI at baseline (D) and CMM

Analysis of Kaplan–Meier curves

Both the TyG-CVAI at baseline and the cuTyG-CVAI were significantly associated with an increased risk of CMM during the follow-up period. The Kaplan‒Meier curves revealed clear separation among the quartiles of the TyG-CVAI at baseline, with the highest quartile (Q4) exhibiting the highest cumulative incidence of CMM (log-rank P < 0.001; Fig. 3). Similar results were also observed between cuTyG-CVAI and the risk of CMM (log-rank P < 0.001; Supplementary Fig. 3).

Fig. 3.

Kaplan–Meier curves for new-onset CMM across quartiles of TyG-CVAI at baseline

Associations between the TyG-CVAI and CMM or component diseases

Cox proportional hazards models revealed significant positive associations between the TyG-CVAI and the risk of incident CMM, indicating a clear dose‒response relationship (Table 3). In the model adjusted for multiple variables (Model 4), compared with participants in the lowest quartile (Q1), participants in the highest quartile (Q4) of the TyG-CVAI at baseline had a significantly increased risk of CMM (HR = 1.93; 95% CI 1.46–2.54; P < 0.001). Similarly, a strong association was observed for the highest quartile of the cuTyG-CVAI (HR = 1.76; 95% CI 1.22–2.53; P < 0.001). The risk increased progressively across quartiles for both indices, with all P values for the trend being statistically significant (P < 0.05). The strength of the association progressively increased with each quartile, and the results remained robust after sequential adjustments for demographic, lifestyle and clinical covariates (Models 1–4; Table 3). We formally assessed the proportional hazards assumption for all covariates in our primary models using the Schoenfeld residual test. This test examines whether the hazard ratios for each variable are constant over time. A statistically nonsignificant result (P value > 0.05) provided evidence that the PH assumption was not violated (Supplementary Table 6).

Table 3.

Cox regression analysis of TyG-CVAI and the risk of new-onset CMM

Variables	Model1		Model2		Model3		Model4
	HR (95%CI)	P	HR (95%CI)	P	HR (95%CI)	P	HR (95%CI)	P
Quartiles of TyG-CVAI at baseline
Q1	1.00 (reference)		1.00 (reference)		1.00 (reference)		1.00 (reference)
Q2	1.37 (1.17–1.57)	< 0.001	1.31 (1.12–1.58)	< 0.001	1.20 (1.08–1.35)	0.006	1.15 (1.05–1.28)	0.017
Q3	2.27 (1.80–2.85)	< 0.001	2.12 (1.68–2.67)	< 0.001	1.78 (1.39–2.28)	< 0.001	1.46 (1.14–1.88)	< 0.001
Q4	3.93 (3.18–4.87)	< 0.001	3.61 (2.90–4.48)	< 0.001	2.63 (2.01–3.44)	< 0.001	1.93 (1.46–2.54)	< 0.001
Quartiles of cuTyG-CVAI
Q1	1.00 (reference)		1.00 (reference)		1.00 (reference)		1.00 (reference)
Q2	1.49 (1.18–1.96)	< 0.001	1.46 (1.15–1.92)	< 0.001	1.38 (1.12–1.65)	< 0.001	1.31 (1.09–1.53)	0.015
Q3	1.96 (1.44–2.66)	< 0.001	1.87 (1.37–2.55)	< 0.001	1.65 (1.18–2.29)	< 0.001	1.41 (1.11–1.87)	< 0.001
Q4	3.18 (2.39–4.24)	< 0.001	2.96 (2.21–3.96)	< 0.001	2.29 (1.60–3.27)	< 0.001	1.76 (1.22–2.53)	< 0.001

HR, Hazard ratio; CI, confidence interval

Model 1: unadjusted;

Model 2: Adjusted sex, age, nationality, married status and highest education level;

Model 3: Adjusted sex, age, nationality, married status, highest education level, smoking, drinking, BMI, and eGFR;

Model 4: Adjusted sex, age, nationality, married status, highest education level, smoking, drinking, BMI, eGFR, hypertension, dyslipidemia, hypertension treatment, and dyslipidemia treatment

Analyses of individual CMM component diseases yielded largely consistent patterns. An elevated TyG-CVAI at baseline or cuTyG-CVAI were significantly associated with increased risks of new-onset diabetes (Supplementary Table 7), heart disease (Supplementary Table 8), and stroke (Supplementary Table 9). The strength of the association varied across outcomes, and the point estimates were generally highest for diabetes and stroke in relation to the TyG-CVAI at baseline and the cuTyG-CVAI. These findings suggest that the adverse metabolic profile captured by the TyG-CVAI contributes to the risk of developing multiple cardiometabolic conditions.

The utility of the TyG-CVAI in predicting new-onset CMM

In addition, we evaluated the predictive performance of different indices for new-onset CMM by analysing receiver operating characteristic (ROC) curves. The area under the curve (AUC) was 0.652 (95% CI 0.631–0.669) for the TyG-CVAI at baseline, compared with 0.641 (95% CI 0.615–0.667) for the cuTyG-CVAI, 0.623 (95% CI 0.604–0.642) for the baseline CVAI, and 0.598 (95% CI 0.584–0.613) for the baseline TyG (Fig. 4). We performed DeLong’s test for paired ROC curves to statistically compare these models (Supplementary Table 10). The results indicated that the AUCs of the TyG-CVAI at baseline and the cuTyG-CVAI were significantly higher than those of the baseline CVAI and baseline TyG (P < 0.05). While the absolute increases in the AUC were modest, these statistically significant comparisons suggest that the TyG-CVAI provides a discernible improvement in discriminative ability over its individual components.

Fig. 4.

The ROC curve of TyG, CVAI, Tyg-CVAI at baseline and cuTyG-CVAI for predicting new-onset CMM

Subgroup analysis

We performed prespecified subgroup analyses of patients stratified according to sex, age category, nationality, marital status, highest education level, BMI category, smoking status, drinking status, and the status of hypertension, dyslipidaemia, diabetes, heart problems, stroke, and kidney disease to assess the robustness of the primary association and explore potential effect modifications. The results are summarized in a forest plot (Fig. 5). A positive association between elevated cumulative TyG-CVAI levels and an increased risk of incident CMM was consistently observed across all predefined subgroups. The point estimates of the HRs and their 95% CIs demonstrated a generally homogeneous effect.

Fig. 5.

Subgroup risk forest plot

Formal tests for interactions revealed no statistically significant effect modification for most factors, such as sex (P for interaction = 0.301), age (P = 0.505), or BMI category (P = 0.162). However, a statistically significant interaction was observed with diabetes, heart problems, and stroke (P for interaction < 0.05), suggesting that the strength of the association might vary among the three components of CMM. No significant interactions were observed for the status of hypertension, dyslipidaemia, or kidney disease, indicating that the association between the TyG-CVAI and CMM risk is largely independent of these prevalent metabolic conditions. These findings suggest that the identified association is robust across a wide range of clinically relevant patient characteristics while highlighting the components of CMM, including diabetes, heart problems, and stroke, as potential modifiers of the risk relationship.

Sensitivity analysis

We assessed the robustness of our findings by conducting multiple sensitivity analyses. First, we evaluated the potential impact of missing data (Supplementary Table 1) on the primary findings by conducting a sensitivity analysis using multiple imputation. The association between the TyG-CVAI at baseline or the cuTyG-CVAI and CMM risk was then reanalysed in this imputed cohort. The results indicated that the significant association persisted, as shown in Supplementary Table 11. The effect estimates for both the baseline and cumulative TyG-CVAI across all quartiles except for Q2 of the cuTyG-CVAI remained directionally consistent and statistically significant compared with those of the primary analysis, confirming the robustness of our main findings to different assumptions about the missing data.

Second, we excluded participants with baseline hypertension (Supplementary Table 12), diabetes (Supplementary Table 13), or dyslipidaemia (Supplementary Table 14). The positive association, particularly for the higher quartiles of the baseline and cumulative TyG-CVAI, remained robust in these disease-free subgroups (Supplementary Tables 12, 13, and 14), underscoring its predictive value for incident CMM even in the absence of individual component diseases.

Third, we excluded participants who were diagnosed with CMM within the first year of follow-up to further address potential reverse causality. The associations between both the baseline and cumulative TyG-CVAI and CMM risk persisted in this analysis (Supplementary Table 15). These results reinforce that the observed associations are not merely driven by early, potentially prevalent cases.

Fourth, we excluded participants who used lipid-lowering or glucose-lowering drugs to minimize potential confounding by medications that directly influence TyG-CVAI components. The association remained robust in this medication-naive subcohort (Supplementary Table 16), with a significant hazard for the highest quartile of the baseline TyG-CVAI (HR = 1.99, 95% CI 1.43–2.75; P < 0.001 in Model 4), indicating that the observed risk is not primarily driven by pharmacological treatment.

Fifth, we excluded participants with extreme metabolic values (the top and bottom 1% of the baseline and cumulative TyG-CVAI) to ensure that our findings were not unduly influenced by outliers. The significant association remained robust in this analysis (Supplementary Table 17), with the highest quartile of the baseline TyG-CVAI maintaining a strong association with the CMM risk (HR = 2.04; 95% CI 1.49–2.78; P < 0.001).

Finally, we applied Fine-Gray competing risk regression analysis to address potential bias from competing risks (non-CMM death). The results remained consistent even after we adjusted for demographic, lifestyle, and clinical factors (Supplementary Table 18). This finding supports the robustness of the primary findings.

Discussion

In this large prospective nationwide cohort of middle-aged and older Chinese adults, we observed that both a single baseline measurement and the cumulative TyG-CVAI were strongly and independently associated with an increased risk of developing CMM. We observed a clear dose‒response relationship between the baseline or cumulative TyG-CVAI and new-onset CMM, even after extensive multivariable adjustment for a wide range of potential confounders, including demographic, lifestyle, and clinical factors. Furthermore, our analysis revealed a nonlinear relationship with a threshold effect between the baseline and cumulative TyG-CVAI and CMM risk. Crucially, the baseline or cumulative TyG-CVAI demonstrated superior predictive performance for incident CMM compared with their individual components (TyG index and CVAI), underscoring the critical importance of evaluating the composite index and long-term metabolic health.

Our findings are consistent with and significantly extend the literature on metabolic indices and cardiovascular disease. The TyG, a reliable surrogate marker of insulin resistance, has previously been linked to incident heart disease [26, 27], stroke [28, 29] and diabetes [30, 31]. Similarly, the CVAI, a validated indicator of visceral adiposity in Chinese populations, has been shown to predict cardiovascular disease [32, 33], stroke [34, 35], and diabetes [36, 37]. In addition, the TyG and CVAI play complex roles in determining the cardiovascular disease risk [11].

However, these studies focused primarily on the relationships between single TyG or CVAI and single disease endpoints such as cardiovascular disease, stroke, and diabetes. In the most recent 8 months, a total of 5 public new studies have integrated the TyG and CVAI, proposed a new concept of the TyG-CVAI, and confirmed that the TyG-CVAI can predict new strokes [10, 12], CVD [13] and hypertension [38]. To date, few studies have investigated the new indicator TyG-CVAI and geriatric comorbidities. Diseases in elderly individuals are predominantly characterized by multimorbidity, particularly the co-occurrence of conditions such as cardiovascular disease, stroke, and diabetes. Furthermore, the pathogenesis of these age-related disorders is largely driven by cumulative exposure to various risk factors over time rather than single acute causes. Considering all of the above factors, we investigated the relationships between the baseline and cumulative TyG-CVAI and the occurrence of new CMM. We found that the baseline and cumulative measurements of the comprehensive indicator of insulin resistance and abdominal obesity, the TyG-CVAI, were strongly associated with incident CMM. Moreover, for the first time, we confirmed that both the baseline and cumulative TyG-CVAI were nonlinearly related to new-onset CMM with threshold effects.

Consistent with the prior CHARLS study [39], we confirm that the TyG-CVAI is a significant predictor of CMM. However, while that analysis revealed a linear association for all eight TyG-related indices [39], we specifically identified a threshold-effect nonlinear relationship for the TyG-CVAI. A potential reason for this discrepancy is attributed to the methodology: the requirement for complete data across all eight TyG-related indices in the previous study may have introduced selection bias through substantial participant exclusion (from 17,705 to 3885) [39], potentially obscuring nuanced relationships. A key strength of our analysis is the application of more sophisticated nonlinear models to a more robust sample, thereby identifying the critical inflection points for both the baseline and cumulative TyG-CVAI, beyond which the synergistic risk of insulin resistance and visceral adiposity for CMM increases nonlinearly. This result highlights a pivotal biological transition and provides a potential target for clinical intervention.

Our findings indicated that both the baseline and cumulative TyG-CVAI outperform its individual components, TyG and CVAI, in predicting CMM. This superior predictive capacity underscores the interconnected roles of insulin resistance and abdominal adiposity in driving CMM development. Furthermore, the strong association observed with the cumulative TyG-CVAI highlights the significance of chronic exposure to these metabolic disturbances, which is consistent with the progressive nature of cardiometabolic diseases. This approach aligns with the growing recognition that chronic disease risk accrues over the course of life. These results support the use of integrated, long-term monitoring of the TyG-CVAI in clinical practice for early risk stratification. In this study, the calculation of the cuTyG-CVAI was constrained by the availability of the necessary biomarkers, which were measured only during the 2011 and 2015 waves of CHARLS [10]. We acknowledge this methodological limitation, and we have addressed it by performing sensitivity analyses such as excluding early cases, which support the robustness of our primary association.

The robust association between the TyG-CVAI and CMM can be explained by the interrelated pathophysiological mechanisms of insulin resistance and visceral adiposity. Visceral adipose tissue is highly metabolically active and releases a cascade of proinflammatory cytokines (such as IL-6, IL-1β, and tumour necrosis factor-α) [40, 41] and free fatty acids [42] into the portal circulation. The release of these molecules promotes systemic inflammation and hepatic insulin resistance, which in turn exacerbates hyperinsulinaemia and dyslipidaemia [43, 44], which are the core components of the TyG. This vicious cycle leads to endothelial dysfunction, accelerated atherosclerosis [45], and β-cell dysfunction, creating a shared environment for the development of coronary heart disease, stroke, and diabetes [46, 47]. The cuTyG-CVAI effectively quantifies the burden of this maladaptive process over time. The observed nonlinear relationship suggests a critical metabolic threshold where the cumulative, synergistic burden of insulin resistance and visceral adiposity transitions from a state of subclinical, potentially compensable stress to one that triggers overt, disproportionate cardiometabolic damage. Beyond this level of accumulated exposure, we hypothesize that pathophysiological processes such as systemic inflammation, endothelial dysfunction, and impaired cellular insulin signalling may accelerate nonlinearly, leading to a sharp increase in the CMM risk. These findings are consistent with the established biological concept that chronic diseases often have a latency period before they manifest clinically.

Strengths and limitations

The major strengths of our study include its prospective design, the use of a nationally representative cohort, and long-term follow-up with repeated measurements, which facilitated the calculation of cumulative exposure. The rigorous adjudication of outcomes and the comprehensive adjustment for a wide range of confounders increase the validity of our findings.

However, several limitations should be acknowledged. First, despite our extensive adjustments for known confounders, the observational nature of this study means that residual confounding from unmeasured (such as detailed dietary patterns and genetic predisposition) or imperfectly measured (such as physical activity) factors cannot be completely excluded. Second, CMM component diagnoses were partially based on self-reports, although this diagnosis was supplemented with biomedical data and medication use to improve precision. Third, this study was conducted within the CHARLS cohort of adults aged ≥ 45 years, which may limit the generalizability of our findings to other ethnicities or younger populations. Fourth, the calculation of the cuTyG-CVAI relied on only two measurements (baseline and 2015). While this method of estimating the time-weighted average burden is an established, pragmatic approach in CHARLS-based research, it remains a simplification that does not capture dynamic, nonlinear metabolic trajectories between time points. This analysis was constrained by the CHARLS design, where the complete panel of required biomarkers was only systematically collected during these two waves, a limitation that was further compounded by the suspension of blood collection in the 2020 wave due to the COVID-19 pandemic. Finally, although our analyses demonstrated strong, graded, and temporally coherent associations, the observational design necessitates caution in inferring causality, as unmeasured confounding remains a possibility.

Future work

Future research should aim to validate the TyG-CVAI in other independent cohorts, including those of different ethnicities and age groups, to establish its broad applicability. Investigating whether targeting people with a high TyG-CVAI for more intensive lifestyle or pharmacological interventions can effectively prevent the onset of CMM would be a critical next step from a clinical and public health perspective. Furthermore, exploring the genetic and molecular mechanisms of this index could provide deeper insights into the biological pathways linking long-term metabolic dysregulation to multimorbidity.

Conclusions

In conclusion, this study introduced the baseline and cumulative TyG-CVAI as useful predictors of incident CMM. Our findings underscore that the burden of insulin resistance and visceral adiposity, rather than its single component, is a key determinant of future disease complexity. The TyG-CVAI is a simple, low-cost tool that can be easily derived from routine clinical measurements. It holds significant promise for improving risk stratification and identifying high-risk individuals who may benefit from early and aggressive preventive strategies to curb the growing global burden of multimorbidity.

Abbreviations

BUN

Blood urea nitrogen

HbA1c

Glycosylated hemoglobin

FPG

Fasting plasma glucose

HDL-C

High-density lipoproteins cholesterol

LDL-C

Low-density lipoproteins cholesterol

CMM

Cardiometabolic multimorbidity

TyG

Triglyceride-glucose index

CVAI

Chinese visceral adiposity index

TyG-CVAI

Triglyceride glucose-Chinese visceral adiposity index

cuTyG-CVAI

Cumulative TyG-CVAI

CI

Confidence intervals

RCS

Restricted cubic splines

ROC

Receiver operating characteristic

eGFR

Estimated glomerular filtration rate

TG

Triglycerides

UA

Uric acid

TC

Total cholesterol

CRP

C-reactive protein

BP

Blood pressure

SBP

Systolic blood pressure

DBP

Diastolic blood pressure

HR

Heart rate

BMI

Body mass index

WC

Waist circumference

HRs

Hazard ratios

AUC

Area under the curve

CVD

Cardiovascular disease

IR

Insulin resistance

Author contributions

Conceptualization: Wenling Zheng, Ziyue Man and Yanping Ren; methodology: Wenling Zheng, Ziyue Man, Yu Li and Xiaohong Zhu; software: Ziyue Man, Lan Wang and Xi Zhang; validation: Ziyue Man, Xi Zhang, Guilin Hu and Yu Cao; formal analysis: Wenling Zheng, Ziyue Man, Lan Wang, Xi Zhang, Guilin Hu and Yu Cao; resources: Wenling Zheng and Yanping Ren; writing—original draft preparation: Wenling Zheng and Ziyue Man; writing—review and editing: Wenling Zheng and Yanping Ren; visualization: Ziyue Man; supervision: Wenling Zheng and Yanping Ren; funding acquisition: Wenling Zheng. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by the National Natural Science Foundation of China No. 82200472(W.-L.Z.).

Data availability

The CHARLS datasets used in this study are available online. The specific data can be found at website at [http://charls.pku.edu.cn//]. The specific analysis code used for this study is available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

CHARLS was approved by the Biomedical Ethics Review Board of Peking University, and approval numbers were IRB00001052-11015 and IRB00001052-11014.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.