Association of cumulative obesity- and insulin resistance-related indices with incident cardiovascular disease in Chinese middle-aged and elderly populations: a nationwide prospective cohort study

Abstract

Background

Obesity and insulin resistance (IR) are major contributors to cardiovascular disease (CVD). However, most studies have focused on single indices, and longitudinal evidence on multiple cumulative indices remains limited. This study aimed to systematically assess the associations between 18 cumulative obesity- and IR-related indices (OIRIs) and incident CVD in Chinese middle-aged and elderly adults, identifying the most predictive markers.

Methods

A total of 2,124 participants from the China Health and Retirement Longitudinal Study (CHARLS) were included. Cumulative OIRIs were calculated as time-weighted averages from 2012 to 2015. The primary outcome was incident CVD, including heart disease and stroke, identified via self-report during follow-up from 2015 to 2020. Cox proportional hazards models were used to assess associations, while receiver operating characteristic (ROC) curves and restricted cubic spline (RCS) analyses evaluated predictive performance and dose–response relationships. Incremental predictive value beyond a base model was quantified using net reclassification improvement (NRI) and integrated discrimination improvement (IDI).

Results

During the follow-up period, 378 participants (17.80%) developed CVD. After adjusting for all covariates, eight cumulative OIRIs were significantly associated with CVD. Cumulative estimated glucose disposal rate (eGDR) was inversely associated with CVD risk (HR per SD increase = 0.683, 95% CI: 0.579–0.806). Seven other indices were positively associated with CVD risk: Chinese visceral adiposity index (CVAI), waist-to-height ratio (WHtR), body roundness index (BRI), triglyceride-glucose index (TyG), TyG-waist circumference (TyG-WC), TyG-waist-to-height ratio (TyG-WHtR), and lipid accumulation product (LAP) (all HRs > 1.10, P < 0.05). Cumulative eGDR demonstrated the highest predictive accuracy (AUC = 0.617), outperforming most other indices, followed by cumulative CVAI (AUC = 0.602). RCS models confirmed linear associations for all significant indices except the conicity index. Sex-stratified analyses showed stronger associations for eGDR (P-interaction = 0.024) and CVAI (P-interaction = 0.042) in males. Adding cumulative eGDR to traditional CVD risk models significantly improved prediction (AUC: 0.628→0.654; NRI = 0.285; IDI = 0.016; all P < 0.001).

Conclusions

In this national cohort, cumulative eGDR and CVAI were associated with incident CVD. Incorporating cumulative eGDR into traditional risk models was accompanied by modest improvements in discrimination and risk reclassification. These findings suggest that cumulative eGDR and CVAI may offer additional value for CVD risk stratification, although further studies with clinically adjudicated outcomes are warranted.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s41043-026-01306-w.

Introduction

Cardiovascular disease (CVD) is a major cause of death and illness worldwide, posing a serious health and economic burden [1, 2]. Its development is driven by multiple risk factors and complex clinical features, with obesity and insulin resistance (IR) being key contributors [3, 4]. Growing evidence shows that obesity and IR increase CVD risk through both direct effects (e.g., cardiac hypertrophy) and indirect pathways (e.g., diabetes-related complications), leading to higher mortality and disease severity [5–9]. Consequently, early identification of high-risk individuals with obesity/IR-related metabolic dysregulation is clinically imperative.

However, some commonly used approaches still have clear limitations in CVD risk assessment. Previous studies have shown that risk estimates based only on prediabetes status may vary across different definitions of prediabetes [6]. Other studies suggest that a single measurement of insulin resistance may miss its long-term burden [9], and that BMI alone may not reflect the true cardiometabolic risk of obesity [10]. In addition, the association between insulin resistance and CVD may differ by glucose tolerance status [11]. For insulin resistance, the hyperinsulinemic–euglycemic clamp is considered the gold standard for assessing IR, but its high cost, complexity, and invasiveness limit its use in clinical and epidemiological studies [12]. Body mass index (BMI) offers a simple and convenient measure of overweight and obesity, but it does not accurately reflect visceral fat distribution or lipid metabolism [13]. These issues are especially relevant in middle-aged and elderly adults and in large population screening.

To address these gaps, novel obesity and insulin resistance-related indicators (OIRIs) have emerged as promising alternative markers [14, 15]. Studies have shown that several OIRIs—such as estimated glucose disposal rate (eGDR), Chinese visceral adiposity index (CVAI), triglyceride-glucose index (TyG), conicity index (CI), and lipid accumulation product (LAP)—are significantly associated with CVD risk and may improve cardiometabolic risk stratification [8, 14, 16–21]. However, most existing studies are cross-sectional or focus on single OIRI, lacking longitudinal evidence to assess the cumulative impact of multiple OIRIs on CVD incidence in middle-aged and older populations.

This study aims to systematically assess the associations between 18 OIRIs and incident CVD in Chinese middle-aged and elderly adults, identifying the most predictive markers. To capture long-term metabolic burden, we used cumulative exposure measures based on time-weighted averages across repeated survey waves.

Methods

Study design and population

The China Health and Retirement Longitudinal Study (CHARLS) is an ongoing nationally representative longitudinal survey targeting middle-aged and older adults aged 45 and older in China. The study employs a multi-stage probability sampling strategy to establish its cohort, covering 150 counties/districts and 450 communities (villages) across 28 provinces nationwide. The baseline survey (Wave 1) was conducted between 2011 and 2012, enrolling 17,708 eligible participants, of whom 11,847 completed biomarker collection including blood samples. Participants undergo biennial follow-up assessments conducted by professionally trained interviewers via computer-assisted face-to-face interviews. Subsequent follow-up waves were conducted in 2013 (wave 2), 2015 (wave 3), 2018 (wave 4), and 2020 (wave 5), with blood samples collected again in wave 3. The project received ethical approval from the Institutional Review Board of Peking University (IRB00001052-11015), and all participants provided written informed consent prior to enrollment.

In this study, the baseline was established using data from wave 1 (2011–2012). A total of 11,847 participants with available blood sample data at baseline were included. Consistent with previous research [16, 22, 23], the cumulative OIRIs were evaluated using data from wave 1 (2011–2012) and wave 3 (2015). The cumulative OIRI was calculated as: cumulative OIRI = [(OIRI_2012 + OIRI_2015) / 2] × (2015 − 2012). Participants were sequentially excluded if they met the following criteria: (1) recorded age < 45 years at wave 1; (2) history of CVD up to wave 3 or missing CVD data; (3) missing CVD data during follow-up; (4) missing OIRIs data at both wave 1 and wave 3; or (5) without baseline covariate data. A total of 2,124 participants were included in the final analysis. The exclusion process is illustrated in Fig. 1.

Fig. 1.

Flowchart of participant selection

Calculation of OIRIs

In this study, a total of 18 OIRIS indices were included. These were calculated using the corresponding formulas based on variables such as age, weight, height, waist circumference (WC), triglycerides (TG), total cholesterol (TC), fasting plasma glucose (FPG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), HbA1c, and hypertension-related variables. For eGDR, HTN was defined as hypertension history (0 = no, 1 = yes). All biochemical values are reported in mg/dL, WC was recorded in cm, and height in m. For formulas originally derived using mmol/L, m or cm, coefficients were adjusted accordingly, and all calculations were verified under a unified unit system. All blood biomarkers were collected under fasting conditions. The formulas are as follows:

BMI (body mass index) = weight (kg) /height² (m)

WHtR (waist-to-height ratio) = WC (cm) /height (cm)

WWI (weight-adjusted-waist index) = WC (cm) /weight^1/2 (kg)

ABSI (a body shape index) =WC (m)/ [BMI^2/3 (kg/m²) × height^1/2 (m)]

BRI (body roundness index) =364.2 − 365.5 [1 − π⁻²WC² (m) height⁻² (m)]^1/2

CI (conicity index) =0.109⁻¹WC (m) [weight (kg) /height (m)]^−1/2

TyG (triglyceride-glucose index) = ln [TG (mg/dl) × FPG (mg/dl) /2]

TyG-BMI (triglyceride glucose-body mass index) = ln [TG (mg/dl) × FPG (mg/dl) /2] × BMI (kg/m²)

TyG-WC (triglyceride glucose-waist circumference) = ln [TG (mg/dl) × FPG (mg/dl) /2] × WC (cm)

TyG-WHtR (triglyceride glucose- waist-to-height ratio) = ln [TG (mg/dl) × FPG (mg/dl) /2] × WHtR

CMI (cardio metabolic index) = TG (mmol/L)/HDL-C (mmol/L) × WHtR

AIP (atherogenic index of plasma) = log [TG (mg/dl) / HDL-C (mg/dl)]

METS-IR (metabolic score for insulin resistance) = ln [2 × FPG (mg/dl) + TG (mg/dl)] × BMI (kg/m²) /ln [HDL-C (mg/dl)]

RC (remnant cholesterol) = TC (mg/dl) – HDL-C (mg/dl) – LDL-C (mg/dl)

eGDR (estimated glucose disposal rate) = 21.158 − 0.09 × WC (cm) − (3.407 × HTN) −[0.551 × HbA1c (%)]

eGDR (estimated glucose disposal rate) is expressed in mg/kg/min, and HTN refers to a hypertension history (0 = no, 1 = yes).

For males:

CVAI (Chinese visceral adiposity index) = −267.93 + 0.68 × age (years) + 0.03 × BMI (kg/m²) + 4.00 × WC (cm) + 22.00 × Log₁₀TG (mmol/L) − 16.32 × HDL-C (mmol/L)

VAI (visceral adiposity index) = WC (cm) / [39.68 + 1.88 × BMI (kg/m²)] × [TG (mmol/L) /1.03] × [1.31/HDL-C (mmol/L)]

LAP (lipid accumulation product) = [WC (cm) − 65] × TG (mmol/L)

For females:

CVAI = −187.32 + 1.71 × age (years) + 4.23 × BMI (kg/m²) + 1.12 × WC (cm) + 39.76 × Log₁₀TG (mmol/L) −11.66 × HDL-C (mmol/L)

VAI = WC (cm) / [36.58 + 1.89 × BMI (kg/m²)] × [TG (mmol/L)/0.81] × [1.52/HDL-C (mmol/L)]

LAP = [WC (cm) − 58] × TG (mmol/L)

Definition of CVD and follow-up

The primary outcomes of this study were the incidence of CVD events, heart disease, and stroke. CVD diagnosis was assessed using standardized questions in the survey interviews [24]: “Has a doctor ever told you that you had a heart attack, coronary heart disease, angina, congestive heart failure, or other heart problems?” or “Has a doctor ever told you that you had a stroke?” During follow-up, new- onset CVD events were defined by participants’ self-reports of heart disease or stroke. The event occurrence time was recorded as the interval between the previous interview and the interview reporting the new-onset CVD event. Follow-up began from CHARLS wave 3 (2015). The endpoint was defined as the first occurrence of CVD or the date of the last survey interview, whichever came first. For participants with incident CVD, follow-up time was determined according to the recorded or estimated timing of the event. Participants who did not develop CVD during follow-up were treated as right-censored observations and assigned a follow-up time of 5 years.

Covariates

Covariates were derived from wave 1 and included age, sex, education level, marital status, residence type, smoking status, drinking status, hypertension, diabetes, dyslipidemia, hypertension medications, diabetes medications and dyslipidemia medications. In this study, educational level was categorized into three groups: middle school or below, high school or vocational school, and college or above. Marital status was classified as married or others. Residence type was divided into rural and urban. Smoking status was categorized as never or ever, with ever smokers including both former and current smokers. Similarly, drinking status was classified as never or ever. Hypertension was identified based on a self-reported diagnosis, the use of antihypertensive drugs, or an average systolic/diastolic blood pressure (SBP/DBP) equal to or exceeding 140/90 mmHg [25]. Diabetes was defined by self-reported physician diagnosis, use of glucose-lowering medication, FPG levels ≥ 126 mg/dL, or HbA1c levels ≥ 6.5% [26]. Dyslipidemia was defined as having TC ≥ 240 mg/dL, TG ≥ 150 mg/dL, LDL-C ≥ 160 mg/dL, HDL-C < 40 mg/dL, self-reported diagnosis of dyslipidemia, or use of lipid-lowering medication [16].

Statistical analysis

Continuous variables were expressed as means and standard deviations, while categorical variables were presented as counts and percentages. Between-group comparisons were conducted using independent t-tests, Mann–Whitney U tests, or chi-square tests, as appropriate.

The association between cumulative OIRIs and incident CVD were evaluated using the Cox proportional hazards regression models. To enable direct comparison of hazard ratios (HR), cumulative OIRIs were standardized as Z-scores. Three models were used in this study. Model 1 was unadjusted. Model 2 was adjusted for age, sex, education, marital status, residence, smoking status and drinking status. Model 3 was further adjusted for hypertension, diabetes, dyslipidemia, hypertension medications, diabetes medications and dyslipidemia medications. Potential multicollinearity among variables in each model was assessed using the variance inflation factor (VIF), with all VIF values below 5, indicating no significant multicollinearity.

The predictive ability of cumulative OIRIs for CVD risk was assessed using receiver operating characteristic (ROC) curves. For each index, the area under the curve (AUC), optimal cutoff value, sensitivity, specificity, and Youden index (sensitivity + specificity – 1) were calculated. The DeLong test was used to compare AUCs among cumulative OIRIs.

Restricted cubic splines (RCS) were applied to investigate the dose–response relationship between cumulative OIRIs and CVD risk. RCS functions with four knots at the 5th, 35th, 65th, and 95th percentiles were used for each cumulative index, and nonlinearity was evaluated by testing the nonlinear spline terms. Additionally, we constructed a base model using variables from CHARLS, referenced from the China-PAR project [27, 28] (a tool for ASCVD risk prediction in China). The incremental predictive value of adding cumulative OIRIs was evaluated using the C-index, net reclassification improvement (NRI), and integrated discrimination improvement (IDI). P < 0.05 indicated statistical significance. All statistical analyses in this study were performed using R software (version 4.4.3).

To assess the robustness of our findings, we conducted two sensitivity analyses. First, to reduce potential misclassification from self-reported outcomes, we applied a more stringent CVD definition by additionally excluding participants who reported CVD at wave 4 but denied the corresponding diagnosis at wave 5. Second, because hypertension status is incorporated into the eGDR formula, we refitted the fully adjusted Model 3 by removing hypertension and further additionally removing antihypertensive medication to reduce the risk of over-adjustment. Results of both sensitivity analyses are reported in Additional file 1.

Results

Baseline characteristics of study population

A total of 2,124 participants (mean age 57.79 ± 8.39 years), including 928 males and 1,196 females, were included in this study. During the follow-up period, the prevalence of CVD was 17.80% (Table 1). Compared to those without CVD, participants with CVD were significantly older, more likely to live in urban areas, and had higher rates of hypertension and dyslipidemia. Additionally, the CVD group exhibited significantly higher levels of LDL-C, FPG, and HbA1c. Among all cumulative OIRIs, participants with CVD showed significantly higher values, except for cumulative eGDR, which was lower (P < 0.05). For individuals affected by either heart disease or stroke, the cumulative OIRIs that were consistently and significantly different included WHtR, WWI, BRI, CI, CVAI, TyG, TyG-WC, TyG-BMI, TyG-WHtR, eGDR, LAP, AIP, and METS-IR (Additional file 1: Table S1 and S2). Comparison of baseline characteristics between participants included and excluded from analysis were presented in Additional file 1 (Table S3).

Table 1.

Baseline characteristics of participants with or without incident CVD

Variable	Overall (n = 2124)	Non-CVD (n = 1746)	CVD (n = 378)	P value
Sex, n (%)				0.005
Male	928 (43.69%)	788 (45.13%)	140 (37.04%)
Female	1196 (56.31%)	958 (54.87%)	238 (62.96%)
Age (years)	57.79 ± 8.39	57.48 ± 8.47	59.23 ± 7.86	< 0.001
Education level, n (%)				0.065
College or above	22 (1.04%)	18 (1.03%)	4 (1.06%)
High school or vocational school	183 (8.62%)	162 (9.28%)	21 (5.56%)
Middle school or below	1919 (90.35%)	1566 (89.69%)	353 (93.39%)
Marital status, n (%)				0.201
Married	1943 (91.48%)	1604 (91.87%)	339 (89.68%)
Others	181 (8.52%)	142 (8.13%)	39 (10.32%)
Residence, n (%)				0.008
Rural	1403 (66.05%)	1176 (67.35%)	227 (60.05%)
Urban	721 (33.95%)	570 (32.65%)	151 (39.95%)
Smoking status, n (%)				0.069
Ever	775 (36.49%)	653 (37.40%)	122 (32.28%)
Never	1349 (63.51%)	1093 (62.60%)	256 (67.72%)
Drinking status, n (%)				0.228
Ever	751 (35.36%)	628 (35.97%)	123 (32.54%)
Never	1373 (64.64%)	1118 (64.03%)	255 (67.46%)
Hypertension, n (%)				< 0.001
Yes	776 (36.53%)	594 (34.02%)	182 (48.15%)
No	1348 (63.47%)	1152 (65.98%)	196 (51.85%)
Diabetes, n (%)				0.119
Yes	329 (15.49%)	260 (14.89%)	69 (18.25%)
No	1795 (84.51%)	1486 (85.11%)	309 (81.75%)
Dyslipidemia, n (%)				0.001
Yes	1021 (48.07%)	809 (46.33%)	212 (56.08%)
No	1103 (51.93%)	937 (53.67%)	166 (43.92%)
SBP (mmHg)	127.38 ± 19.96	126.68 ± 19.88	130.63 ± 20.04	< 0.001
DBP (mmHg)	74.62 ± 11.74	74.21 ± 11.73	76.48 ± 11.60	0.001
Hypertension medications, n (%)				0.007
Yes	307 (14.45%)	235 (13.46%)	72 (19.05%)
No	1817 (85.55%)	1511 (86.54%)	306 (80.95%)
Diabetes medications, n (%)				0.004
Yes	78 (3.67%)	54 (3.09%)	24 (6.35%)
No	2046 (96.33%)	1692 (96.91%)	354 (93.65%)
Dyslipidemia medications, n (%)				< 0.001
Yes	83 (3.91%)	51 (2.92%)	32 (8.47%)
No	2041 (96.09%)	1695 (97.08%)	346 (91.53%)
TG (mg/dl)	130.69 ± 96.51	128.78 ± 96.74	139.50 ± 95.08	0.050
HDL-C (mg/dl)	50.90 ± 14.84	51.07 ± 14.96	50.09 ± 14.25	0.242
LDL-C (mg/dl)	117.91 ± 34.94	116.27 ± 34.17	125.48 ± 37.39	< 0.001
FPG (mg/dl)	110.43 ± 33.66	109.34 ± 31.69	115.50 ± 41.27	0.001
HbA1c (%)	5.29 ± 0.84	5.26 ± 0.77	5.44 ± 1.10	< 0.001
Cumulative OIRIs
BMI	71.66 ± 13.98	71.32 ± 14.60	73.26 ± 10.50	0.014
WHtR	1.62 ± 0.21	1.61 ± 0.20	1.67 ± 0.22	< 0.001
WWI	33.25 ± 2.90	33.14 ± 2.83	33.77 ± 3.15	< 0.001
ABSI	0.25 ± 0.02	0.25 ± 0.02	0.25 ± 0.02	0.008
BRI	12.79 ± 4.36	12.59 ± 4.34	13.75 ± 4.37	< 0.001
CI	3.83 ± 0.31	3.82 ± 0.30	3.88 ± 0.33	< 0.001
CVAI	300.07 ± 122.12	292.95 ± 122.90	332.94 ± 112.96	< 0.001
VAI	6.79 ± 6.62	6.60 ± 6.52	7.67 ± 7.05	0.004
TyG	26.04 ± 1.70	25.96 ± 1.67	26.42 ± 1.78	< 0.001
TyG-WC	2224.51 ± 350.18	2205.13 ± 343.32	2314.02 ± 367.70	< 0.001
TyG-BMI	624.18 ± 136.77	619.21 ± 140.76	647.15 ± 113.99	< 0.001
TyG-WHtR	14.11 ± 2.30	13.98 ± 2.25	14.71 ± 2.41	< 0.001
CMI	2.25 ± 2.28	2.18 ± 2.22	2.53 ± 2.53	0.007
eGDR	26.90 ± 6.37	27.38 ± 6.21	24.72 ± 6.64	< 0.001
LAP	118.94 ± 104.38	113.71 ± 100.00	143.11 ± 119.83	< 0.001
AIP	1.10 ± 0.82	1.07 ± 0.82	1.23 ± 0.82	< 0.001
METS-IR	107.75 ± 25.09	106.95 ± 25.62	111.42 ± 22.15	0.002
RC	83.93 ± 57.28	82.01 ± 55.99	92.80 ± 62.17	0.001

ABSI, a body shape index; AIP, atherogenic index of plasma; BMI, body mass index; BRI, body roundness index; CI, conicity index; CMI, cardio metabolic index; CVD, cardiovascular disease; CVAI, Chinese visceral adiposity index; DBP, diastolic blood pressure; eGDR, estimated glucose disposal rate; FPG, fasting plasma glucose; HbA1c, glycosylated hemoglobin A1c; HDL-C, high-density lipoprotein cholesterol; LAP, lipid accumulation product; LDL-C, low-density lipoprotein cholesterol; METS-IR, metabolic score for insulin resistance; OIRIs, obesity- and insulin resistance-related indices; RC, remnant cholesterol; SBP, systolic blood pressure; TG, triglyceride; TyG, triglyceride-glucose index; TyG-BMI, triglyceride glucose-body mass index; TyG-WC, triglyceride glucose-waist circumference; TyG-WHtR, triglyceride glucose-waist-to-height ratio; VAI, visceral adiposity index; WHtR, waist-to-height ratio; WWI, weight-adjusted-waist index

Data are presented as mean ± standard deviation or number (%)

P value in bold indicates statistical significance (P < 0.05)

Associations of cumulative OIRIs with CVD

Cox proportional hazards regression analyses were performed to investigate the associations between cumulative OIRIs and CVD (Table 2). The results showed that in both Model 1 and Model 2, cumulative eGDR was negatively associated with CVD risk, whereas seven other OIRIs—WHtR, BRI, TyG, TyG-WC, TyG-WHtR, LAP, and CVAI—were positively associated with an increased risk of CVD. After adjusting for potential confounders in Model 3, all eight OIRIs remained statistically significant. Specifically, a per-standard deviation (SD) increase in cumulative eGDR was associated with a decreased risk of CVD, with an adjusted HR of 0.683 (95% CI: 0.579–0.806). In contrast, per-SD increases in the other seven indices were associated with increased CVD risk, with adjusted HRs (95% CIs) as follows: WHtR, 1.127 (1.010–1.259); BRI, 1.107 (1.010–1.213); TyG, 1.148 (1.011–1.304); TyG-WC, 1.220 (1.083–1.373); TyG-WHtR, 1.167 (1.038–1.313); LAP, 1.142 (1.031–1.264); and CVAI, 1.167 (1.060–1.285).

Table 2.

Association between cumulative OIRIs and CVD: Cox proportional hazards regression models

Variable	Model 1		Model 2		Model 3
-	HR (95%CI)	P value	HR (95%CI)	P value	HR (95%CI)	P value
BMI	1.072 (1.013–1.134)	0.016	1.081 (1.019–1.146)	0.009	1.058 (0.978–1.144)	0.160
WHtR	1.270 (1.156–1.396)	< 0.001	1.206 (1.088–1.338)	< 0.001	1.127 (1.010–1.259)	0.033
WWI	1.246 (1.117–1.391)	< 0.001	1.118 (0.994–1.258)	0.064	1.074 (0.958–1.204)	0.219
ABSI	1.156 (1.040–1.286)	0.007	1.072 (0.961–1.196)	0.210	1.063 (0.955–1.184)	0.261
BRI	1.179 (1.106–1.256)	< 0.001	1.150 (1.066–1.241)	< 0.001	1.107 (1.01–1.213)	0.030
CI	1.256 (1.119–1.410)	< 0.001	1.155 (1.029–1.296)	0.015	1.107 (0.989–1.238)	0.076
VAI	1.121 (1.038–1.210)	0.003	1.095 (1.009–1.188)	0.029	1.035 (0.936–1.145)	0.505
TyG	1.253 (1.141–1.374)	< 0.001	1.221 (1.110–1.344)	< 0.001	1.148 (1.011–1.304)	0.033
TyG-WC	1.322 (1.198–1.459)	< 0.001	1.303 (1.177–1.443)	< 0.001	1.220 (1.083–1.373)	0.001
TyG-BMI	1.105 (1.046–1.166)	< 0.001	1.113 (1.051–1.179)	< 0.001	1.081 (0.998–1.170)	0.055
TyG-WHtR	1.323 (1.203–1.455)	< 0.001	1.262 (1.139–1.397)	< 0.001	1.167 (1.038–1.313)	0.010
CMI	1.115 (1.033–1.204)	0.005	1.113 (1.027–1.206)	0.009	1.047 (0.950–1.154)	0.356
eGDR	0.691 (0.626–0.762)	< 0.001	0.718 (0.650–0.794)	< 0.001	0.683 (0.579–0.806)	< 0.001
LAP	1.233 (1.137–1.338)	< 0.001	1.216 (1.116–1.326)	< 0.001	1.142 (1.031–1.264)	0.011
AIP	1.194 (1.083–1.315)	< 0.001	1.181 (1.069–1.305)	0.001	1.092 (0.963–1.239)	0.170
RC	1.143 (1.058–1.236)	< 0.001	1.129 (1.042–1.223)	0.003	1.062 (0.964–1.170)	0.225
CVAI	1.245 (1.163–1.334)	< 0.001	1.221 (1.128–1.321)	< 0.001	1.167 (1.060–1.285)	0.002
METS-IR	1.109 (1.041–1.181)	0.001	1.120 (1.048–1.196)	< 0.001	1.071 (0.976–1.175)	0.146

Model 1 was unadjusted

Model 2 was adjusted for age, sex, education, marital status, residence, smoking status and drinking status

Model 3 was adjusted for age, sex, education, marital status, residence, smoking status, drinking status, hypertension, diabetes, dyslipidemia, hypertension medications, diabetes medications and dyslipidemia medications

ABSI, a body shape index; AIP, atherogenic index of plasma; BMI, body mass index; BRI, body roundness index; CI, conicity index; Cl, confidence interval; CMI, cardio metabolic index; CVD, cardiovascular disease; CVAI, Chinese visceral adiposity index; eGDR, estimated glucose disposal rate; HR, hazard ratio; LAP, lipid accumulation product; METS-IR, metabolic score for insulin resistance; OIRIs, obesity- and insulin resistance-related indices; RC, remnant cholesterol; TyG, triglyceride-glucose index; TyG-BMI, triglyceride glucose-body mass index; TyG-WC, triglyceride glucose-waist circumference; TyG-WHtR, triglyceride glucose-waist-to-height ratio; VAI, visceral adiposity index; WHtR, waist-to-height ratio; WWI, weight-adjusted-waist index

P value in bold indicates statistical significance (P < 0.05)

Predictive performance of each cumulative OIRI for CVD

AUCs and ROC curves were used to evaluate and compare the predictive performance of each cumulative OIRI for CVD (Table 3; Fig. 2). The top six indices based on AUC values were eGDR, CVAI, TyG-WC, CI, TyG-WHtR, and LAP. Among the studied indices, the absolute AUC values of cumulative OIRIs were modest, cumulative eGDR showed the highest discriminatory ability for CVD risk, with an AUC of 0.617 (95% CI: 0.585–0.649), followed by cumulative CVAI (AUC: 0.602, 95% CI: 0.571–0.634). Furthermore, cumulative eGDR also showed the highest AUC values for predicting heart disease and stroke among all indices, with AUCs of 0.606 (95% CI: 0.570–0.642) for heart disease and 0.618 (95% CI: 0.566–0.669) for stroke (Table S4 and Table S5).

Table 3.

Comparative predictive performance of each cumulative OIRI for CVD

Variable	Cutoff	Sensitivity (%)	Specificity (%)	Youden	AUC	95%CI	P value
eGDR	27.300	63.00	57.10	0.201	0.617	0.585–0.649	Ref.
CVAI	308.235	58.50	57.30	0.157	0.602	0.571–0.634	0.295
TyG-WC	2130.172	69.80	44.60	0.144	0.589	0.557–0.621	0.029
TyG-WHtR	14.241	56.60	57.70	0.143	0.589	0.557–0.621	0.047
CI	3.954	46.30	69.40	0.157	0.585	0.552–0.618	0.040
LAP	78.538	68.30	45.80	0.141	0.584	0.553–0.616	0.031
WHtR	1.798	29.60	84.30	0.139	0.581	0.548–0.614	0.013
BRI	16.239	29.60	84.30	0.139	0.581	0.548–0.614	0.014
TyG-BMI	605.956	61.40	50.80	0.122	0.579	0.547–0.611	0.016
WWI	34.650	39.90	73.90	0.138	0.576	0.543–0.609	0.018
TyG	25.409	70.60	41.90	0.126	0.575	0.544–0.606	0.023
VAI	4.068	72.20	40.80	0.131	0.571	0.54–0.601	0.014
ABSI	0.252	49.50	64.80	0.143	0.568	0.535–0.601	0.007
BMI	74.820	44.70	68.10	0.128	0.567	0.535-0.6	0.002
METS-IR	103.006	62.20	48.60	0.108	0.566	0.534–0.598	0.001
CMI	1.270	75.10	37.50	0.126	0.565	0.534–0.595	0.003
RC	77.076	51.60	60.50	0.121	0.564	0.533–0.596	0.008
AIP	0.959	62.70	48.40	0.111	0.559	0.528–0.59	0.002

ABSI, a body shape index; AIP, atherogenic index of plasma; AUC, areas under the curve; BMI, body mass index; BRI, body roundness index; CI, conicity index; Cl, confidence interval; CMI, cardio metabolic index; CVD, cardiovascular disease; CVAI, Chinese visceral adiposity index; eGDR, estimated glucose disposal rate; LAP, lipid accumulation product; METS-IR, metabolic score for insulin resistance; OIRI, obesity- and insulin resistance-related indice; RC, remnant cholesterol; TyG, triglyceride-glucose index; TyG-BMI, triglyceride glucose-body mass index; TyG-WC, triglyceride glucose-waist circumference; TyG-WHtR, triglyceride glucose-waist-to-height ratio; VAI, visceral adiposity index; WHtR, waist-to-height ratio; WWI, weight-adjusted-waist index

P value in bold indicates statistical significance (P < 0.05)

Fig. 2.

ROC curves of the top six cumulative OIRIs for predicting CVD, heart disease and stroke. AUC, area under the curve; CI, conicity index; CMI, cardiometabolic index; CVAI, Chinese visceral adiposity index; CVD, cardiovascular disease; eGDR, estimated glucose disposal rate; LAP, lipid accumulation product; ROC, receiver operating characteristic; TyG-BMI, triglyceride-glucose body mass index; TyG-WC, triglyceride-glucose waist circumference; TyG-WHtR, triglyceride-glucose waist-to-height ratio

Dose–response relationship between cumulative OIRIs and CVD

RCS analyses were conducted to examine the potential non-linear dose–response relationships between the top six cumulative OIRIs (based on AUC values) and CVD risk. As illustrated in Fig. 3, after adjusting for confounding factors, cumulative eGDR exhibited a linear and inverse association with CVD risk (P for nonlinear > 0.05). Similarly, cumulative CVAI, TyG-WC, TyG-WHtR, and LAP were linearly and positively associated with increased CVD risk (P for nonlinear > 0.05). In contrast, cumulative CI demonstrated a significant non-linear positive association with CVD risk (P for nonlinear < 0.05). The results from the unadjusted models were consistent with these findings (Figure S1).

Fig. 3.

Dose–response relationship between top six cumulative OIRIs and CVD risk. Model was adjusted for age, sex, education, marital status, residence, smoking status, drinking status, hypertension, diabetes, dyslipidemia, hypertension medications, diabetes medications and dyslipidemia medications. CI, conicity index; CVAI, Chinese visceral adiposity index; CVD, cardiovascular disease; eGDR, estimated glucose disposal rate; LAP, lipid accumulation product; TyG-WC, triglyceride-glucose waist circumference; TyG-WHtR, triglyceride-glucose waist-to-height ratio

Subgroup analyses to determine the associations between cumulative OIRIs and CVD

Subgroup analyses were conducted to examine the associations between the top six cumulative OIRIs and CVD risk across different sex and age groups (Fig. 4). Significant interactions by sex were observed for cumulative eGDR and CVAI (P for interaction < 0.05), with stronger protective effects of eGDR and stronger harmful effects of CVAI found in males. Additionally, cumulative TyG-WC, TyG-WHtR, and LAP were primarily associated with increased CVD risk among men or elderly participants (age ≥ 60 years).

Fig. 4.

Subgroup analyses of the associations between the top six cumulative OIRIs and CVD risk. Each subgroup analysis was adjusted for age, sex, education, marital status, residence, smoking status, drinking status, hypertension, diabetes, dyslipidemia, hypertension medications, diabetes medications and dyslipidemia medications. HRs are expressed per one SD increase in the cumulative OIRIs with corresponding 95% confidence interval (CI). CI, conicity index; CVAI, Chinese visceral adiposity index; CVD, cardiovascular disease; eGDR, estimated glucose disposal rate; HR, hazard ratio; LAP, lipid accumulation product; TyG-WC, triglyceride-glucose waist circumference; TyG-WHtR, triglyceride-glucose waist-to-height ratio

Assessment of cumulative eGDR or CVAI’s incremental predictive performance for CVD

To assess whether cumulative eGDR or CVAI provided additional prognostic value beyond traditional clinical risk factors, we constructed a basic prediction model based on the China-PAR project, which included age, sex, systolic blood pressure, total cholesterol, high-density lipoprotein cholesterol, residence, smoking status, and diabetes. As shown in Fig. 5; Table 4, further analysis demonstrated that the inclusion of cumulative eGDR significantly enhanced the predictive performance of the base model for new-onset CVD and heart disease (P < 0.05; CVD: AUC = 0.654, 95% CI: 0.625–0.684; heart disease: AUC = 0.645, 95% CI: 0.611–0.678). In contrast, adding cumulative CVAI yielded a relatively smaller improvement in predictive ability compared to cumulative eGDR. Moreover, the addition of cumulative eGDR significantly improved both the NRI and IDI for predicting new-onset CVD, heart disease, and stroke, as presented in Table 4.

Fig. 5.

ROC curves of base model, plus cumulative eGDR or CVAI for predicting CVD, heart disease and stroke. AUC, area under the curve; CVAI, Chinese visceral adiposity index; CVD, cardiovascular disease; eGDR, estimated glucose disposal rate

Table 4.

Improvement in discrimination and risk reclassification for CVD after adding cumulative eGDR or cumulative CVAI

Model	AUC (95%CI)	P value	NRI (95%CI)	P value	IDI (95%CI)	P value
CVD
Base model	0.628 (0.598–0.658)	Ref	Ref		Ref
+cumulative eGDR	0.654 (0.625–0.684)	0.011	0.285 (0.175–0.396)	< 0.001	0.016 (0.001–0.022)	< 0.001
+cumulative CVAI	0.639 (0.609–0.669)	0.068	0.149 (0.038–0.260)	0.009	0.005 (0.002–0.008)	0.002
Heart disease
Base model	0.620 (0.586–0.653)	Ref	Ref		Ref
+cumulative eGDR	0.645 (0.611–0.678)	0.031	0.275 (0.151–0.399)	< 0.001	0.012 (0.006–0.017)	< 0.001
+cumulative CVAI	0.633 (0.599–0.667)	0.049	0.152 (0.028–0.276)	0.017	0.003 (0.001–0.006)	0.001
Stroke
Base model	0.644 (0.593–0.695)	Ref	Ref		Ref
+cumulative eGDR	0.666 (0.617–0.715)	0.103	0.232 (0.045–0.419)	0.015	0.006 (0.001–0.010)	0.010
+cumulative CVAI	0.649 (0.599–0.698)	0.439	0.127 (-0.060-0.314)	0.184	0.001 (-0.001-0.002)	0.314

Base model adjusted for age, sex, systolic blood pressure, total cholesterol, high-density lipoprotein cholesterol, residence, smoking status and diabetes

AUC, area under the curve; CVD, cardiovascular diseases; CI, confidence intervals; eGDR, estimated glucose disposal rate; CVAI, Chinese visceral adiposity index; NRI, net reclassification improvement; IDI, integrated discrimination improvement. P value in bold indicates statistical significance (P < 0.05)

Sensitivity analyses

In sensitivity analyses excluding participants with “yes-to-no” reversals (n = 52), the restricted analytic sample included 2,072 participants. The number of incident CVD cases decreased from 378 to 326, while the direction and magnitude of associations for all 18 cumulative indices were materially unchanged (Table S6). Predictive performance analyses showed consistent results under the stringent outcome definition (Table S7). The inverse association between cumulative eGDR and incident CVD remained statistically significant after refitting Model 3 without hypertension and after additionally excluding antihypertensive medication (Table S8).

Discussion

CVD remains a leading cause of morbidity and mortality worldwide, particularly among aging populations. Accumulating evidence suggests that obesity and IR play central roles in the development and progression of CVD through mechanisms involving metabolic dysregulation, chronic inflammation, and vascular dysfunction [29–31]. In this context, various OIRIs have been developed to capture different aspects of metabolic health, including fat distribution and glucose–lipid metabolism. The main findings are as follows: (1) Among all included indices, eight cumulative OIRIs—eGDR, CVAI, WHtR, BRI, TyG, TyG-WC, TyG-WHtR and LAP—were independently associated with CVD incidence after adjusting for potential confounders. (2) Cumulative eGDR was negatively associated with CVD risk and showed the highest AUC for CVD prediction, followed by cumulative CVAI. In both subgroup and RCS analyses, the inverse association between cumulative eGDR and CVD remained consistent and independent. (3) The addition of cumulative eGDR significantly improved the performance of predictive models for CVD and heart disease, while the inclusion of cumulative CVAI significantly enhanced the model for predicting heart disease.

Previous studies have reported the relationship of OIRIs with CVD [14, 32, 33]. For instance, Li et al. studied multiple IR-related indices, including eGDR, TyG, and METS-IR, and found that all were significantly associated with current and future CVD risk in middle-aged and elderly populations, with eGDR demonstrating the strongest predictive power [34]. Jiang et al. studied the predictive utility of six insulin resistance surrogate indices (eGDR, CVAI, TyG, TyG-BMI, METS-IR, AIP) for stroke risk, ultimately identifying ​eGDR as the most robust predictor [35]. Notwithstanding these findings, prior investigations predominantly relied on single-time-point OIRI assessments, which have limitations. As they fail to capture temporal changes and cannot accurately reflect the longitudinal relationship between cumulative conditions—such as obesity, hypertension, and dyslipidemia—and incident CVD. This may lead to biased estimates of cardiovascular risk. ​To mitigate such bias, this study explored and compared the associations between a range of commonly used cumulative OIRIs and the incidence of CVD in middle-aged and elderly populations.

Cox proportional hazards regression analyses were employed in this study, to examine the associations between cumulative OIRIs and incident CVD. Eight cumulative OIRIs remained significantly associated with CVD incidence after adjustment for all potential covariates in the fully adjusted model (Model 3). These cumulative OIRIs reflect different aspects of individuals’ physiological status over long-term observation: WHtR and BRI indicate central obesity [36]; CVAI, developed specifically for the Chinese population, captures visceral fat distribution [21]; the TyG-related indices assess insulin resistance linked to hyperglycemia and hypertriglyceridemia [37]; LAP primarily reflect hepatic fat accumulation [38]. eGDR provides a comprehensive assessment by incorporating central obesity, hyperglycemia, and hypertension [39]. Therefore, our findings highlight that central/visceral obesity and metabolic disturbances—rather than elevated BMI alone—are the primary drivers of increased CVD risk.

These associations between cumulative OIRIs and CVD risk are biologically plausible. Visceral and ectopic fat accumulation can promote insulin resistance, dyslipidemia, and chronic inflammation, thereby accelerating atherosclerosis. Even after controlling for overall adiposity, individuals with higher visceral fat still tend to exhibit hyperinsulinemia, impaired glucose tolerance, elevated triglycerides, hypertension, and increased levels of inflammatory markers [40, 41]. Piché et al. emphasized that unlike subcutaneous fat, visceral and ectopic fat are the key determinants of cardiovascular risk in individuals with obesity [8]. They proposed the concept of “obesities,” highlighting that individuals with similar BMIs can have vastly different metabolic risks depending on fat distribution. In our study, indices reflecting visceral fat (e.g., CVAI) or insulin resistance (e.g., eGDR) were most strongly associated with CVD risk, consistent with the current pathophysiological model linking insulin resistance, atherosclerosis, and CVD.

eGDR is a simple surrogate marker for estimating whole-body insulin sensitivity using routine clinical parameters [42]. It has been validated against the hyperinsulinemic–euglycemic clamp technique and has recently gained attention due to its ease of use and strong predictive value for CVD [43]. Tao et al. examined the association between single-time-point eGDR and incident CVD over an 8-year follow-up, reporting a 31.4% lower CVD risk in the highest eGDR quartile (Q4) compared to the lowest (Q1) [44]. RCS analysis revealed a linear dose–response relationship between eGDR and CVD risk. Similarly, Yan et al. investigated longitudinal changes in eGDR and found that individuals with persistently low eGDR had a 2.51-fold higher risk of developing CVD compared to those with persistently high eGDR [45]. However, existing studies have largely focused on single indices and lack comprehensive cross-comparisons. It remains unclear whether cumulative eGDR outperforms other commonly used OIRIs in predicting CVD. Our study addressed this gap by systematically comparing multiple cumulative OIRIs and found that cumulative eGDR exhibited the highest predictive accuracy for incident CVD, outperforming other indices such as cumulative TyG. This may be attributed to eGDR’s integration of both clinical measures (hypertension and waist circumference) and a laboratory marker (HbA1c), allowing for a more comprehensive assessment of IR. Importantly, in the eGDR formula, HTN refers to hypertension history rather than a single blood pressure measurement, while HbA1c reflects glycemic control over a recent period. Therefore, compared with single-time-point measures such as SBP/DBP or FPG, eGDR may better capture broader metabolic and clinical information relevant to CVD risk. The inclusion of such time-averaged or chronic-state variables may enhance the sensitivity of these indices. These findings suggest that future development of similar risk markers should integrate both anthropometric and biochemical variables—particularly those reflecting chronic disease states—to improve diagnostic and predictive performance.

To assess the consistency of our findings, we conducted subgroup analyses stratified by age and sex. The results revealed a significant sex interaction for both cumulative eGDR and CVAI (P < 0.05). This may be attributed to particularly the influence of estrogen and androgen levels, whereby men are more prone to visceral fat accumulation and reduced skeletal muscle insulin sensitivity [46, 47]. These physiological differences may render males more susceptible to vascular endothelial dysfunction driven by obesity and IR [48]. However, it is worth noting that other similar studies have not consistently reported significant sex-specific associations, which could be due to differences in study populations or the definitions of outcome events [44, 45, 49]. Therefore, future research should take these factors into account to better understand the role of OIRIs across different demographic groups and to further elucidate the potential impact of sex on CVD risk.

The China-PAR project is a CVD risk prediction model developed specifically for the Chinese population. Due to the absence of certain variables in CHARLS—such as family history of CVD and region of residence (South/North)—we constructed a base model using the available variables. Notably, adding cumulative eGDR significantly improved the model’s predictive performance (AUC = 0.654, 95% CI: 0.625–0.684, P < 0.05). Similarly, cumulative CVAI enhanced the prediction of heart disease (AUC = 0.645, 95% CI: 0.611–0.678, P < 0.05). Although the discriminative ability of individual OIRIs alone was modest, these results suggest that selected cumulative OIRIs may provide incremental predictive information when incorporated into multivariable risk prediction frameworks. Future studies should externally validate these findings and evaluate whether adding high-performing OIRIs (e.g., cumulative eGDR or CVAI) can meaningfully improve established tools such as China-PAR, the ASCVD Risk Estimator Plus [50], or the Framingham Risk Score [51].

Our study has several strengths. First, it utilized data from the nationally representative CHARLS dataset, and its prospective cohort design strengthens the causal inference of our findings. Second, we evaluated the predictive performance of 18 cumulative OIRIs for CVD risk among middle-aged and elderly Chinese adults, incorporating time-to-event analyses to more comprehensively assess their impact on CVD risk. Third, we carefully adjusted for potential confounders and conducted subgroup analyses to ensure the reliability and robustness of our results.

However, this study has several limitations. (1) Similar to other studies, CVD data were based on self-reports from participants, which may have led to outcome misclassification through both under- and over-reporting. To partially mitigate this concern, we conducted a sensitivity analysis excluding “yes-to-no” reversals across follow-up waves, and the results remained robust, although clinical endpoint validation is still warranted in future studies. (2) Only a small proportion of participants with baseline blood data were included in the final analysis (2,124/11,847) because of missing repeated measures and complete-data requirements, which may lead to selection bias and limit generalizability (Additional file 1: Table S3). (3) Although we adjusted for known major confounders, the influence of unmeasured factors such as dietary habits, physical activity, and environmental changes— cannot be entirely ruled out, given their potential importance in affecting glucose-lipid metabolism and cardiovascular risk. (4) Since this study focused on middle-aged and elderly individuals in China, the generalizability of the findings to other populations requires further investigation and validation.

Our findings suggest that cumulative eGDR may provide a pragmatic way to incorporate longitudinal metabolic information into CVD risk stratification, although its clinical utility requires further validation. Next steps should include external validation in independent, multicenter cohorts across diverse regions, age groups, and clinically relevant subgroups such as individuals with prediabetes or chronic kidney disease. Future studies should also evaluate model calibration, clinical benefit, and potential decision thresholds. In addition, real world studies using electronic health records or insurance databases could assess feasibility, longitudinal stability [52], and the incremental predictive value of cumulative eGDR when added to established risk models.

From an implementation perspective, eGDR can be calculated using three routinely available parameters, waist circumference, hypertension status, and HbA1c, which makes it suitable for automation. Cumulative eGDR could be generated within electronic health record systems or via simple web or mobile calculators by storing serial measurements and computing a time weighted average across visits. If validated, periodic tracking of cumulative eGDR in primary care may help identify individuals whose cardiometabolic risk is not fully captured by BMI alone and support more individualized preventive counseling. However, the optimal measurement frequency and clinically actionable cut points should be determined in future prospective studies.

Conclusion

In conclusion, among the 18 cumulative obesity and insulin resistance related indices evaluated, eight were associated with incident CVD in this national cohort of Chinese middle aged and older adults. Cumulative eGDR and CVAI showed consistent associations with CVD, and adding cumulative eGDR to multivariable risk models was accompanied by modest improvements in discrimination and risk reclassification. These findings suggest that cumulative eGDR and CVAI may provide complementary information for CVD risk stratification. Further studies with clinically adjudicated outcomes and external validation are warranted before broader clinical implementation.

Abbreviations

ABSI

A body shape index

AIP

Atherogenic index of plasma

AUC

Area under the curve

BMI

Body mass index

BRI

Body roundness index

CHARLS

China Health and Retirement Longitudinal Study

China-PAR

China Prediction for Atherosclerotic Cardiovascular Disease Risk

CI

Confidence interval (statistical term)

CI

Conicity index (anthropometric index)

CMI

Cardio metabolic index

CVD

Cardiovascular disease

CVAI

Chinese visceral adiposity index

DBP

Diastolic blood pressure

eGDR

Estimated glucose disposal rate

FPG

Fasting plasma glucose

HbA1c

Glycosylated hemoglobin A1c

HDL-C

High-density lipoprotein cholesterol

HR

Hazard ratio

HTN

hypertension status

IDI

Integrated discrimination improvement

IR

Insulin resistance

LAP

Lipid accumulation product

LDL-C

Low-density lipoprotein cholesterol

METS-IR

Metabolic score for insulin resistance

NRI

Net reclassification improvement

OIRIs

Obesity- and insulin resistance-related indices

RC

Remnant cholesterol

RCS

Restricted cubic spline

ROC

Receiver operating characteristic

SBP

Systolic blood pressure

SD

Standard deviation

TC

Total cholesterol

TG

Triglycerides

TyG

Triglyceride-glucose index

TyG-BMI

Triglyceride glucose-body mass index

TyG-WC

Triglyceride glucose-waist circumference

TyG-WHtR

Triglyceride glucose-waist-to-height ratio

VIF

Variance inflation factor

VAI

Visceral adiposity index

WC

Waist circumference

WHtR

Waist-to-height ratio

WWI

Weight-adjusted-waist index

Author contributions

LL and JrZ designed the study. JrZ and YZ analyzed the data, and drafted the original manuscript. XZ contributed to data collection and figure mapping. JZ, CC and YL participated in interpreting the results. JW carried out literature search. SL and BL supervised the study. All authors critically revised the manuscript for intellectual content and approved the final version.

Funding

This study was funded by Beijing Natural Science Foundation (7232270), China National Natural Science Foundation (82374575), Capital’s Funds for Health Improvement and Research (CFH2024-2–2235), Out-standing Young Talents Program of Capital Medial University (B2207), and Beijing Hospital Management Center “peak” talent training plan team (DFL20241001).

Data availability

The dataset was based on the CHARLS, which is publicly available and can be obtained by visiting their website (https://charls.pku.edu.cn/).

Declarations

Ethics approval and consent to participate

This study protocol was reviewed and approved by the Ethical Review Committee of Peking University (IRB00001052-11015), and all participants provided written informed consent at the time of participation. The study was conducted in accordance with the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Clinical trial number

Not applicable.