Comparative analysis of 16 baseline obesity and lipid-related indices for cardiovascular disease risk prediction in adults with cardiovascular-kidney-metabolic syndrome stages 0–3: a nationwide prospective cohort study

Abstract

Objective

In the context of the newly defined cardiovascular-kidney-metabolic (CKM) syndrome, this study aimed to systematically compare the predictive value of 16 different obesity- and lipid-related indices for new-onset cardiovascular disease (CVD) in a population with CKM stages 0–3.

Methods

This prospective cohort study utilized data from the China Health and Retirement Longitudinal Study (CHARLS). A total of 5,782 participants aged 45 and older, free of CVD at baseline and classified within CKM stages 0–3, were included. We evaluated 16 indices, including traditional markers (e.g., BMI, WHtR) and novel composite markers (e.g., TyG-WC, CVAI, CTI). The primary outcome was incident CVD over a 6-year follow-up. Cox proportional hazards models were used to assess associations. Predictive performance was compared using the C-index, Akaike Information Criterion (AIC), Integrated Discrimination Improvement (IDI), and Decision Curve Analysis (DCA).

Results

During the 6-year follow-up, 1,134 incident CVD events occurred. In the fully adjusted Cox model, the triglyceride-glucose waist circumference index (TyG-WC) demonstrated the strongest association with CVD risk, with each 1-standard deviation increase corresponding to a 15% higher risk (HR = 1.15, 95% CI: 1.08–1.22, P < 0.001). A prediction model incorporating TyG-WC showed the best performance, with a higher C-index (0.6434), a significant improvement in discrimination (IDI = 0.0035, P < 0.001), and the greatest net benefit in Decision Curve Analysis. The findings remained robust in both landmark and sensitivity analyses.

Conclusion

This study, based on nationally representative CHARLS cohort data, systematically compared the CVD risk prediction ability of 16 obesity and lipid-related indices in adults with CKM stages 0–3. The study found that the TyG-WC index demonstrated the strongest CVD risk prediction ability. These indices provide effective assessment tools for CVD risk stratification in CKM stages 0–3 populations through different pathophysiological mechanisms.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13098-025-01919-x.

Introduction

Cardiovascular disease (CVD) remains the leading cause of death and disability worldwide [1]. In October 2023, the American Heart Association (AHA) formally introduced the concept of cardiovascular-kidney-metabolic (CKM) syndrome. This systemic condition results from pathophysiological interactions between metabolic risk factors, chronic kidney disease (CKD), and CVD [2, 3]. The clinical burden of CKM syndrome is closely associated with CVD risk. This emphasizes the importance of considering metabolic, kidney, and cardiovascular systems as a unified whole [4–6]. The AHA particularly focuses on identifying preclinical stages and emphasizes that research on CKM stages 0–3 populations should prioritize CVD prevention [3].

Insulin resistance (IR) is a metabolic syndrome characterized by reduced or impaired insulin sensitivity in target organs or tissues. It manifests as impaired glucose uptake and oxidation capacity [7–10]. Previous studies have confirmed that patients with persistently high levels of IR have an increased risk of developing CVD, even without type 2 diabetes [9, 10]. The triglyceride glucose index (TyG), initially proposed in 2008 as a marker of IR, has been demonstrated to be associated with CVD occurrence, adverse outcomes, and mortality. These associations include acute coronary syndromes, coronary heart disease, stroke, heart failure, and other cardiac diseases [11–14].

IR is influenced not only by metabolic biomarkers including triglycerides and glucose but is also closely related to body fat content and distribution [15, 16]. Hormones and cytokines from adipocytes can enhance or inhibit glucose response and insulin signaling [15]. Given the close relationship between IR and obesity, increasing attention has been paid to modified TyG indices. These indices combine TyG with body mass index (BMI), waist circumference (WC), and waist-to-height ratio (WHtR) [16]. These modified TyG indices have been demonstrated to have better predictive ability for CVD compared to TyG alone [17].

In addition to TyG-related indices, multiple studies have shown that other obesity and lipid-related indicators are closely associated with CVD risk. Chinese visceral adiposity index (CVAI), as an emerging visceral obesity indicator, has shown excellent performance in predicting metabolic diseases [18, 19]. Lipid accumulation product (LAP), calculated based on waist circumference and triglycerides, is superior to traditional BMI in identifying metabolic syndrome [20–22]. C-reactive protein-triglyceride glucose index (CTI) combines inflammatory and metabolic factors, providing a new perspective for CVD risk assessment [23]. Additionally, lipid-related indicators such as atherogenic index of plasma (AIP), remnant cholesterol (RC), and non-high-density lipoprotein cholesterol (NHDL) have also shown important value in CVD prediction [24–26].

Although previous studies have explored the relationship between some obesity and lipid-related indicators and CVD, most studies have focused only on the general population or specific disease patients. In the context of CKM syndrome, particularly for CKM stages 0–3 populations, systematic comparison of the ability of multiple obesity and lipid-related indicators to predict CVD remains lacking. Furthermore, most existing studies have used cross-sectional designs or short follow-up periods. This makes it difficult to adequately assess the long-term predictive value of these indicators.

Therefore, this study, based on data from the China Health and Retirement Longitudinal Study (CHARLS), aims to systematically compare the ability of 16 baseline obesity and lipid-related indicators to predict CVD risk in adults with CKM stages 0–3. We provide scientific evidence for early identification and intervention through long-term prospective cohort study.

Methods

Study population

This study employed a prospective cohort study design based on CHARLS. CHARLS is a nationally representative longitudinal survey covering adults aged 45 years and older in 28 provinces, 150 counties, and 450 villages across China. The study was conducted by the National School of Development at Peking University. The baseline survey was conducted in 2011–2012 (Wave 1) and subsequent follow-up surveys were completed (Wave 2 in 2013, Wave 3 in 2015, Wave 4 in 2018) [27]. CHARLS was approved by the Institutional Review Board of Peking University (IRB00001052-11015). All participants provided written informed consent before participation. This study strictly followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [28].

The exclusion criteria for this study included: (1) age less than 45 years; (2) existing CVD at baseline or lack of CVD follow-up information; (3) missing essential indicators required for CKM syndrome staging or CKM syndrome stage 4; (4) missing data for any of the 16 obesity- and lipid-related indices; (5) missing data for any other baseline covariates used in the analysis. A total of 5,782 participants were finally included for analysis. The detailed participant selection process is shown in Fig. 1.

Fig. 1.

Flowchart of participant selection

Data assessment

Outcome ascertainment

The primary outcome was the occurrence of CVD during the 6-year follow-up period (Wave 2 to Wave 4), including heart disease and stroke. Follow-up duration was calculated from enrollment to the occurrence of outcome events or last recorded contact. The occurrence of outcome events was assessed through two standardized questions: “Have you been diagnosed by a doctor with myocardial infarction, coronary heart disease, angina, congestive heart failure, or other heart diseases?” and “Have you been diagnosed by a doctor with stroke?” Participants who answered “yes” to either question were considered to have developed CVD.

Calculation formulas for 16 obesity and lipid-related indices

The calculation formulas for these 16 indices are shown in Table 1.

Table 1.

Formulas for 16 baseline obesity and lipid-related indices

Index	Formula with Integrated Units and Logarithmic Base
CVAI [19]	Male: -267.93 + (0.68 * age (years)) + (0.03 * BMI (kg/m^2)) + (4.00 * WC (cm)) + (22.00 * log10(TG (mmol/L))) - (16.32 * HDL-C (mmol/L)) Female: -187.32 + (1.71 * age (years)) + (4.23 * BMI (kg/m^2)) + (1.12 * WC (cm)) + (39.76 * log10(TG (mmol/L))) - (11.66 * HDL-C (mmol/L)). Logarithm is base 10.
CTI [29]	0.412 * ln(CRP (mg/L)) + TyG (unitless). Logarithm is natural (ln).
BMI [30]	Weight (kg) / (Height (m)^2)
WHtR [31]	Waist Circumference (cm) / Height (cm)
ABSI [32]	WC (m) / (BMI (kg/m^2)^(2/3) * Height (m)^(1/2))
BRI [33]	364.2–365.5 * sqrt(1 - (WC (m) / (2*pi))^2 / (0.5 * Height (m))^2)
CI [34]	WC (m) / (0.109 * sqrt(Weight (kg) / Height (m)))
TyG [35]	ln(TG (mg/dL) * FBG (mg/dL) / 2). Logarithm is natural (ln).
TyG-BMI	TyG (unitless) * BMI (kg/m^2)
TyG-WC	TyG (unitless) * WC (cm)
TyG-WHtR	TyG (unitless) * WHtR (unitless)
VAI [36]	Male: (WC (cm) / (39.68 + (1.88 * BMI (kg/m^2)))) * (TG (mmol/L) / 1.03) * (1.31 / HDL-C (mmol/L)) Female: (WC (cm) / (36.58 + (1.89 * BMI (kg/m^2)))) * (TG (mmol/L) / 0.81) * (1.52 / HDL-C (mmol/L))
LAP [20]	Male: (WC (cm) − 65) * TG (mmol/L) Female: (WC (cm) − 58) * TG (mmol/L)
AIP [37]	log10(TG (mmol/L) / HDL-C (mmol/L)). Logarithm is base 10.
RC [25]	TC - (HDL-C + LDL-C). All lipids must be in consistent units (e.g., mmol/L).
NHDL-C [26]	TC - HDL-C. Both lipids must be in consistent units (e.g., mmol/L).

Notes: Abbreviations are used for the index components: BMI, Body Mass Index; WC, Waist Circumference; TG, Triglycerides; HDL-C, High-Density Lipoprotein Cholesterol; CRP, C-Reactive Protein; FBG, Fasting Blood Glucose; TC, Total Cholesterol; LDL-C, Low-Density Lipoprotein Cholesterol. Abbreviations for the indices are as follows: AIP, Atherogenic Index of Plasma; ABSI, A Body Shape Index; BRI, Body Roundness Index; CI, Conicity Index; CTI, C-Reactive Protein to Triglyceride-Glucose Index; CVAI, Chinese Visceral Adiposity Index; LAP, Lipid Accumulation Product; NHDL-C, Non-High-Density Lipoprotein Cholesterol; RC, Remnant Cholesterol; TyG, Triglyceride-Glucose Index; TyG-BMI, TyG-Body Mass Index; TyG-WC, TyG-Waist Circumference; TyG-WHtR, TyG-Waist-to-Height Ratio; VAI, Visceral Adiposity Index; WHtR, Waist-to-Height Ratio. Where applicable, the logarithmic base is specified within the formula description

Definition of CKM syndrome stages 0–3

According to the AHA Presidential Advisory Report, CKM syndrome stages 0–3 were defined as follows [3]:

Stage 0

Individuals without CKM syndrome risk factors.

Stage 1

Individuals with overweight, abdominal obesity, or adipose tissue dysfunction but without CKD.

Stage 2

Individuals with metabolic risk factors (such as hypertriglyceridemia, hypertension, metabolic syndrome, or type 2 diabetes), moderate-to-high risk CKD, or both.

Stage 3

Individuals with high risk of subclinical CVD or subclinical CVD.

Risk equivalents for subclinical CVD were assessed based on high predicted 10-year CVD risk and very high-risk CKD. The Framingham risk score was used to assess 10-year CVD risk, with > 20% considered high predicted risk [38]. CKD staging was classified according to Kidney Disease: Improving Global Outcomes (KDIGO) criteria [3]. High-risk CKD was defined as estimated Glomerular Filtration Rate (eGFR) < 30 ml/min per 1.73 m² [39]. The formula for calculating eGFR was based on the Chinese MDRD equation: eGFR (ml/min per 1.73 m²) = 175 × Scr^(-1.234) × age^(-0.179) × 0.79 (if female) [40].

Data collection

Data collected in this study included the following aspects: Sociodemographic data: age, sex, education level, marital status, and household registration; Health status data: self-reported smoking and drinking status, and self-reported history of hypertension and diabetes; Physical examination data: height, weight, waist circumference, and systolic blood pressure (SBP); Laboratory test data: triglycerides, fasting blood glucose, glycated hemoglobin (HbA1c), total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, blood urea nitrogen (BUN), uric acid (UA), serum creatinine (Cr), and CRP.

In Wave 1 survey, sociodemographic and health status data were collected by trained interviewers using structured questionnaires [27]. All biochemical indicators were tested in quality-controlled laboratories to ensure data accuracy and reliability.

Statistical analysis

Baseline characteristics were presented as mean ± standard deviation or median (interquartile range) for continuous variables based on normality tests, and frequencies and percentages for categorical variables. Between-group differences were compared using one-way ANOVA, Kruskal-Wallis test, or chi-square test.

Survival analysis used Kaplan-Meier method to plot survival curves, with log-rank test to compare CVD incidence differences between quartile groups. Cox proportional hazards regression models were constructed to assess associations between the 16 obesity and lipid-related indices and CVD risk. Three progressively adjusted models were established: Model 1 was unadjusted; Model 2 adjusted for age, sex, marital status, education level, residence, smoking and drinking; Model 3 further adjusted for hypertension, diabetes, dyslipidemia, creatinine, and CRP on the basis of Model 2. Schoenfeld residual tests verified proportional hazards assumptions.

Restricted cubic spline (RCS) regression analysis assessed non-linear relationships between indices and CVD risk. Four knots were set and non-linear P-values were evaluated to determine relationship patterns. Subgroup analyses were stratified by age (< 60 years vs. ≥ 60 years), sex (male vs. female), and CKM stage (stages 0–1 vs. stage 2 vs. stage 3). Interaction term tests evaluated between-subgroup differences.

Based on Cox regression analysis results, obesity and lipid-related indices that performed well in fully adjusted models were selected to construct CVD prediction models. Variable selection criteria were: (1) maintaining statistical significance (P < 0.05); (2) having high hazard ratios; (3) representing different pathophysiological mechanisms. C-index was used to evaluate model discrimination ability, Akaike Information Criterion (AIC) to compare goodness of fit, and likelihood ratio tests to assess model improvement.

For the best-performing model, an in-depth validation was conducted. This included calculating the Integrated Discrimination Improvement (IDI) and Net Reclassification Improvement (NRI) to assess improvement in prediction accuracy over the base model. To assess stability and internal validity, rigorous internal validation was performed using 2,000 bootstrap resamples to generate an optimism-corrected C-index. Model calibration was evaluated by generating a calibration plot and calculating the calibration slope. Furthermore, Decision Curve Analysis (DCA) was performed to assess the clinical utility of the model by determining its net benefit across a range of risk thresholds. Lastly, a risk stratification analysis was conducted by dividing participants into tertiles based on their predicted risk to calculate the 6-year CVD event rates for each risk group.

To test the robustness of our findings against potential reverse causation, two distinct sensitivity analyses were performed. First, we excluded all participants who experienced a CVD event within the first two years of follow-up and re-ran the main analysis on the remaining cohort. Second, we conducted a landmark analysis. A landmark point was set at 2 years, and the analysis was restricted to participants who were alive and free of CVD at this time. Cox models were then refitted for this landmark cohort to assess if baseline predictors were associated with CVD events occurring between 2 and 6 years of follow-up.

All statistical analyses were completed using R software (version 4.4.2), with two-sided tests and P < 0.05 considered statistically significant.

Results

Baseline characteristics of participants

The final analysis, based on a complete-case approach, included 5,782 participants. Of these, 1,134 (19.6%) developed incident cardiovascular disease during the 6-year follow-up period. The baseline characteristics of the participants, stratified by CKM stage, are presented in Table 2. Significant differences were observed across CKM stages for most baseline characteristics. Specifically, variables such as age, gender, marital status, smoking and drinking habits, prevalence of hypertension, diabetes, and dyslipidemia, and the incidence of future CVD events all showed statistically significant differences (all P < 0.001). Furthermore, all 16 evaluated obesity and lipid-related indices, as well as baseline creatinine and C-reactive protein levels, varied significantly across the CKM stages (all P < 0.001).

Table 2.

Baseline characteristics of participants stratified by CKM syndrome stage

Characteristic	Stage 0 (n = 249)	Stage 1 (n = 669)	Stage 2 (n = 1626)	Stage 3 (n = 3238)	P-value
Age	54.00 [48.00, 59.00]	54.00 [48.00, 58.00]	55.00 [49.00, 60.00]	60.00 [54.00, 68.00]	< 0.001
CVAI	50.48 [38.29, 64.41]	79.50 [62.65, 95.07]	98.32 [78.23, 121.02]	95.03 [64.60, 130.71]	< 0.001
CTI	7.93 [7.59, 8.28]	8.20 [7.86, 8.56]	8.81 [8.32, 9.31]	8.68 [8.16, 9.29]	< 0.001
BMI	20.37 [18.89, 21.38]	23.25 [21.42, 25.03]	24.09 [22.00, 26.58]	22.83 [20.53, 25.50]	< 0.001
WHTR	0.48 [0.46, 0.50]	0.54 [0.50, 0.57]	0.56 [0.52, 0.60]	0.52 [0.48, 0.58]	< 0.001
ABSI	0.08 [0.08, 0.08]	0.08 [0.08, 0.09]	0.08 [0.08, 0.09]	0.08 [0.08, 0.09]	< 0.001
BRI	2.96 [2.56, 3.36]	4.12 [3.43, 4.87]	4.48 [3.66, 5.44]	3.80 [3.02, 4.91]	< 0.001
CI	6.93 [6.73, 7.22]	7.32 [6.95, 7.62]	7.41 [7.09, 7.72]	7.35 [7.03, 7.69]	< 0.001
TyG	8.13 [7.87, 8.41]	8.31 [8.06, 8.54]	8.80 [8.41, 9.14]	8.58 [8.19, 9.05]	< 0.001
TyG-BMI	164.03 [152.15, 175.04]	192.34 [175.65, 209.19]	211.24 [189.71, 238.17]	196.16 [172.12, 227.00]	< 0.001
TyG-WC	598.04 [557.58, 632.44]	687.21 [639.73, 733.32]	750.96 [680.97, 829.35]	727.42 [648.88, 822.70]	< 0.001
TyG-WHtR	3.85 [3.64, 4.10]	4.47 [4.13, 4.77]	4.88 [4.44, 5.35]	4.51 [4.00, 5.12]	< 0.001
VAI	2.23 [1.52, 3.14]	2.64 [1.92, 3.64]	5.06 [3.02, 7.98]	2.94 [1.72, 5.31]	< 0.001
LAP	12.00 [7.93, 17.56]	22.19 [15.40, 30.24]	38.58 [23.39, 60.51]	23.52 [11.90, 44.31]	< 0.001
AIP	0.11 [-0.04, 0.25]	0.16 [0.01, 0.28]	0.43 [0.21, 0.62]	0.32 [0.11, 0.55]	< 0.001
RC	0.30 [0.18, 0.46]	0.34 [0.21, 0.51]	0.63 [0.37, 0.93]	0.50 [0.29, 0.83]	< 0.001
NHDL	17.56 [-5.21, 41.78]	24.14 [2.03, 48.70]	81.10 [34.71, 133.08]	51.26 [15.64, 105.59]	< 0.001
Cr	0.66 [0.60, 0.75]	0.68 [0.60, 0.76]	0.68 [0.61, 0.77]	0.82 [0.71, 0.94]	< 0.001
CRP	0.52 [0.37, 1.01]	0.71 [0.46, 1.43]	0.97 [0.54, 1.94]	1.09 [0.58, 2.29]	< 0.001
Gender					< 0.001
Female	219 (88.0)	609 (91.0)	1530 (94.1)	746 (23.0)
Male	30 (12.0)	60 (9.0)	96 (5.9)	2492 (77.0)
Education					0.085
Less than lower secondary	220 (88.4)	619 (92.5)	1469 (90.3)	2898 (89.5)
Secondary or above	29 (11.6)	50 (7.5)	157 (9.7)	340 (10.5)
Marital Status					< 0.001
Married	231 (92.8)	614 (91.8)	1495 (91.9)	2844 (87.8)
Non-married	18 (7.2)	55 (8.2)	131 (8.1)	394 (12.2)
Residence					0.067
Rural	183 (73.5)	450 (67.3)	1058 (65.1)	2158 (66.6)
Urban	66 (26.5)	219 (32.7)	568 (34.9)	1080 (33.4)
Smoking					< 0.001
No	243 (97.6)	645 (96.4)	1529 (94.0)	1121 (34.6)
Yes	6 (2.4)	24 (3.6)	97 (6.0)	2117 (65.4)
Drinking					< 0.001
No	206 (82.7)	541 (80.9)	1342 (82.5)	1414 (43.7)
Yes	43 (17.3)	128 (19.1)	284 (17.5)	1824 (56.3)
Hypertension					< 0.001
No	249 (100.0)	669 (100.0)	1164 (71.6)	2462 (76.0)
Yes	0 (0.0)	0 (0.0)	462 (28.4)	776 (24.0)
Diabetes Mellitus					< 0.001
No	249 (100.0)	669 (100.0)	1600 (98.4)	2978 (92.0)
Yes	0 (0.0)	0 (0.0)	26 (1.6)	260 (8.0)
Dyslipidemia					< 0.001
No	243 (97.6)	645 (96.4)	1477 (90.8)	2984 (92.2)
Yes	6 (2.4)	24 (3.6)	149 (9.2)	254 (7.8)

Notes: Data are presented as median (interquartile range) for continuous variables and n (%) for categorical variables. CKM, cardiovascular-kidney-metabolic; IQR, interquartile range; BMI, body mass index; WHTR, waist-to-height ratio; CVAI, Chinese visceral adiposity index; VAI, visceral adiposity index; LAP, lipid accumulation product; TyG, triglyceride-glucose index; CRP, C-reactive protein; CVD, cardiovascular disease. P-values were calculated using Kruskal-Wallis test for continuous variables and chi-square test for categorical variables

The relationship between the 16 obesity and lipid-related indices and the incidence of CVD in a population with CKM syndrome stages 0–3

During an average 6-year follow-up period, 1,134 CVD events were observed. To evaluate the predictive ability of baseline values of 16 obesity and lipid-related indices for CVD risk, we grouped each index by quartiles, plotted Kaplan-Meier cumulative incidence curves, and performed log-rank tests. Results showed statistically significant differences in CVD cumulative incidence between quartile groups for all 16 indices (all P < 0.05). Most indices showed clear dose-response relationships, with CVD cumulative incidence gradually increasing as index levels increased (Fig. 2).

Fig. 2.

Kaplan-Meier survival curves for CVD risk by quartiles of 16 obesity and lipid-related indices in CKM syndrome stages 0–3. Notes: Kaplan-Meier curves show cumulative incidence of cardiovascular disease (CVD) events over 6-year follow-up period stratified by quartiles (Q1-Q4) of each obesity and lipid-related index. Q1 represents the lowest quartile (reference group) and Q4 represents the highest quartile. Log-rank test P-values are shown for each index, indicating significant differences in CVD cumulative incidence across quartiles for all 16 indices (all P < 0.05). Most indices demonstrate a clear dose-response relationship with increasing CVD risk from Q1 to Q4. CVAI, Chinese visceral adiposity index; CTI, C-reactive protein-triglyceride glucose index; BMI, body mass index; WHTR, waist-to-height ratio; ABSI, A body shape index; BRI, body roundness index; CI, conicity index; TyG, triglyceride glucose index; TyG_BMI, triglyceride glucose-body mass index; TyG_WC, triglyceride glucose-waist circumference index; TyG_WHTR, triglyceride glucose-waist-to-height ratio index; VAI, visceral adiposity index; LAP, lipid accumulation product; AIP, atherogenic index of plasma; RC, remnant cholesterol; NHDL, non-high-density lipoprotein cholesterol; CKM, cardiovascular-kidney-metabolic

To quantify the association between each baseline index and the risk of incident CVD, we performed Cox proportional hazards regression analysis. We constructed three models with progressive levels of adjustment, and the results for each index, treated as a continuous variable per 1-standard deviation (SD) increase, are detailed in Table 3. In the fully adjusted model (Model 3), TyG-WC demonstrated the strongest and most significant association with CVD risk, with each 1-SD increase corresponding to a 15% rise in risk (HR = 1.15, 95% CI: 1.08–1.22, P < 0.001). Several other indices also remained significant predictors after full adjustment, including CTI (HR = 1.10, P = 0.004), TyG-WHtR (HR = 1.09, P = 0.002), WHtR (HR = 1.07, P = 0.005), LAP (HR = 1.07, P = 0.020), AIP (HR = 1.07, P = 0.021), CVAI (HR = 1.05, P = 0.008), and TyG (HR = 1.06, P = 0.046). Conversely, the associations for the remaining eight indices—BMI, ABSI, BRI, CI, TyG-BMI, VAI, RC, and NHDL—were attenuated and lost statistical significance after adjusting for all covariates in Model 3.

Table 3.

Cox proportional hazards regression analysis for CVD risk: continuous variables

Index	Model 1		Model 2		Model 3
	HR (95% CI)	P-value	HR (95% CI)	P-value	HR (95% CI)	P-value
CVAI	1.08 (1.05–1.11)	< 0.001	1.07 (1.04–1.09)	< 0.001	1.05 (1.01–1.09)	0.008
CTI	1.20 (1.13–1.27)	< 0.001	1.18 (1.12–1.25)	< 0.001	1.10 (1.03–1.17)	0.004
BMI	1.03 (1.00-1.06)	0.072	1.03 (1.00-1.06)	0.051	1.02 (0.98–1.06)	0.353
WHtR	1.12 (1.09–1.16)	< 0.001	1.11 (1.06–1.15)	< 0.001	1.07 (1.02–1.12)	0.005
ABSI	1.11 (1.04–1.18)	0.002	1.04 (0.97–1.10)	0.257	1.03 (0.97–1.09)	0.357
BRI	1.04 (1.02–1.07)	< 0.001	1.04 (1.01–1.06)	0.012	1.03 (0.99–1.06)	0.169
CI	1.18 (1.10–1.27)	< 0.001	1.10 (1.03–1.18)	0.004	1.06 (1.00-1.13)	0.061
TyG	1.14 (1.08–1.21)	< 0.001	1.14 (1.08–1.21)	< 0.001	1.06 (1.00-1.13)	0.046
TyG-BMI	1.04 (1.01–1.06)	0.009	1.03 (1.01–1.06)	0.008	1.02 (0.99–1.06)	0.205
TyG-WC	1.26 (1.18–1.33)	< 0.001	1.27 (1.19–1.35)	< 0.001	1.15 (1.08–1.22)	< 0.001
TyG-WHtR	1.16 (1.12–1.20)	< 0.001	1.14 (1.10–1.19)	< 0.001	1.09 (1.03–1.15)	0.002
VAI	1.07 (1.03–1.12)	0.002	1.08 (1.03–1.13)	0.002	1.03 (0.98–1.09)	0.221
LAP	1.13 (1.08–1.19)	< 0.001	1.14 (1.09–1.20)	< 0.001	1.07 (1.01–1.12)	0.020
AIP	1.14 (1.08–1.20)	< 0.001	1.15 (1.09–1.22)	< 0.001	1.07 (1.01–1.14)	0.021
RC	1.07 (1.01–1.12)	0.012	1.08 (1.02–1.14)	0.004	1.03 (0.98–1.09)	0.268
NHDL	1.08 (1.02–1.13)	0.004	1.09 (1.04–1.15)	< 0.001	1.03 (0.98–1.09)	0.238

Notes: Model 1 = unadjusted; Model 2 = adjusted for age, gender, marital status, education, residence, smoking and drinking; Model 3 = Model 2 + hypertension, diabetes mellitus, dyslipidemia, creatinine, and CRP. HR, hazard ratio; CI, confidence interval. CVAI, Chinese visceral adiposity index; CTI, C-reactive protein-triglyceride glucose index; BMI, body mass index; WHTR, waist-to-height ratio; ABSI, A body shape index; BRI, body roundness index; CI, conicity index; TyG, triglyceride glucose index; VAI, visceral adiposity index; LAP, lipid accumulation product; AIP, atherogenic index of plasma; RC, remnant cholesterol; NHDL, non-high-density lipoprotein cholesterol

To further explore the associations and assess for potential non-linear relationships, we also performed Cox regression analysis by categorizing participants into quartiles for each index, using the lowest quartile (Q1) as the reference group. The fully adjusted results are presented in Table 4.

Table 4.

Cox proportional hazards regression analysis for CVD risk: quartile analysis

Index	Quartile	HR (95% CI)	P-value
CVAI	Q1	Ref.
	Q2	1.23 (1.02–1.48)	0.031
	Q3	1.35 (1.12–1.62)	0.002
	Q4	1.41 (1.17–1.70)	< 0.001
CTI	Q1	Ref.
	Q2	1.16 (0.97–1.39)	0.108
	Q3	1.26 (1.06–1.51)	0.010
	Q4	1.34 (1.12–1.61)	0.002
BMI	Q1	Ref.
	Q2	1.04 (0.87–1.24)	0.692
	Q3	1.03 (0.86–1.23)	0.770
	Q4	1.41 (1.19–1.69)	< 0.001
WHtR	Q1	Ref.
	Q2	1.14 (0.95–1.36)	0.162
	Q3	1.15 (0.96–1.37)	0.143
	Q4	1.39 (1.16–1.67)	< 0.001
ABSI	Q1	Ref.
	Q2	1.26 (1.06–1.51)	0.010
	Q3	1.22 (1.02–1.45)	0.030
	Q4	1.21 (1.01–1.45)	0.043
BRI	Q1	Ref.
	Q2	1.14 (0.95–1.36)	0.162
	Q3	1.15 (0.96–1.37)	0.143
	Q4	1.39 (1.16–1.67)	< 0.001
CI	Q1	Ref.
	Q2	1.09 (0.91–1.31)	0.334
	Q3	1.12 (0.94–1.34)	0.222
	Q4	1.29 (1.09–1.54)	0.004
TyG	Q1	Ref.
	Q2	1.14 (0.96–1.36)	0.137
	Q3	1.20 (1.01–1.43)	0.039
	Q4	1.13 (0.95–1.35)	0.168
TyG-BMI	Q1	Ref.
	Q2	1.22 (1.02–1.46)	0.029
	Q3	1.27 (1.06–1.52)	0.010
	Q4	1.50 (1.25–1.80)	< 0.001
TyG-WC	Q1	Ref.
	Q2	1.18 (0.98–1.41)	0.076
	Q3	1.33 (1.12–1.59)	0.001
	Q4	1.42 (1.19–1.70)	< 0.001
TyG-WHtR	Q1	Ref.
	Q2	1.19 (0.99–1.42)	0.059
	Q3	1.11 (0.93–1.34)	0.247
	Q4	1.36 (1.14–1.64)	0.001
VAI	Q1	Ref.
	Q2	1.11 (0.93–1.33)	0.247
	Q3	1.36 (1.14–1.63)	0.001
	Q4	1.21 (1.01–1.45)	0.044
LAP	Q1	Ref.
	Q2	1.09 (0.91–1.31)	0.363
	Q3	1.37 (1.15–1.64)	0.001
	Q4	1.36 (1.13–1.64)	0.001
AIP	Q1	Ref.
	Q2	1.05 (0.88–1.25)	0.614
	Q3	1.28 (1.08–1.52)	0.004
	Q4	1.13 (0.95–1.35)	0.156
RC	Q1	Ref.
	Q2	1.02 (0.85–1.21)	0.850
	Q3	1.17 (0.99–1.39)	0.065
	Q4	1.12 (0.95–1.33)	0.183
NHDL	Q1	Ref.
	Q2	1.10 (0.92–1.31)	0.298
	Q3	1.21 (1.02–1.43)	0.032
	Q4	1.20 (1.01–1.43)	0.039

Notes: Results are from Model 3 adjusted for age, gender, marital status, education, residence, smoking, drinking, hypertension, diabetes mellitus, dyslipidemia, creatinine, and CRP. Q1 serves as the reference group. HR, hazard ratio; CI, confidence interval. CVAI, Chinese visceral adiposity index; CTI, C-reactive protein-triglyceride glucose index; BMI, body mass index; WHTR, waist-to-height ratio; ABSI, A body shape index; BRI, body roundness index; CI, conicity index; TyG, triglyceride glucose index; VAI, visceral adiposity index; LAP, lipid accumulation product; AIP, atherogenic index of plasma; RC, remnant cholesterol; NHDL, non-high-density lipoprotein cholesterol

This analysis reinforced the findings from the continuous variable models. A clear, graded relationship was observed for several indices. For instance, compared to the lowest quartile, the risk of CVD in the highest quartile (Q4) was significantly increased for TyG-BMI (HR = 1.50, 95% CI: 1.25–1.80), TyG-WC (HR = 1.42, 95% CI: 1.19–1.70), BMI (HR = 1.41, 95% CI: 1.19–1.69), and CVAI (HR = 1.41, 95% CI: 1.17–1.70). For these indices, the risk progressively increased with each successive quartile, demonstrating a strong dose-response pattern.

Other indices, including CTI, WHtR, BRI, CI, and TyG-WHtR, also showed a significantly elevated risk in their highest quartiles. In contrast, the association for RC was not statistically significant in any quartile.

To further explore dose-response relationships between the 16 obesity and lipid-related indices and CVD risk, we used RCS regression analysis with 3 knots to assess non-linear associations between indices and CVD risk. Figure 3 shows RCS curves for each index. Five indices showed non-linear relationships: TyG-BMI, VAI, BMI, RC, and NHDL. The remaining indices showed no non-linear relationships (all non-linear P > 0.05).

Fig. 3.

Restricted cubic spline analysis of 16 obesity and lipid-related indices for CVD risk in CKM syndrome stages 0–3. Notes: Restricted cubic spline curves show the association between each obesity and lipid-related index and cardiovascular disease (CVD) risk over 6-year follow-up period. Each subplot displays hazard ratios (HR) with 95% confidence intervals (gray shaded areas) adjusted for age, gender, marital status, education, residence, smoking, drinking, hypertension, diabetes mellitus, dyslipidemia, creatinine, and CRP. The red dashed horizontal line represents HR = 1.0 (reference). Knots were placed at the 10th, 50th, and 90th percentiles. Non-linear p-values are shown in subplot titles, with p < 0.05 indicating significant non-linear relationships. Five indices demonstrated significant non-linear associations: TyG_BMI, VAI, BMI, RC, and NHDL. CVAI, Chinese visceral adiposity index; CTI, C-reactive protein-triglyceride glucose index; BMI, body mass index; WHTR, waist-to-height ratio; ABSI, A body shape index; BRI, body roundness index; CI, conicity index; TyG, triglyceride glucose index; TyG_BMI, triglyceride glucose-body mass index; TyG_WC, triglyceride glucose-waist circumference index; TyG_WHTR, triglyceride glucose-waist-to-height ratio index; VAI, visceral adiposity index; LAP, lipid accumulation product; AIP, atherogenic index of plasma; RC, remnant cholesterol; NHDL, non-high-density lipoprotein cholesterol; CKM, cardiovascular-kidney-metabolic

To assess differences in predictive efficacy of the 16 obesity and lipid-related indices across different population characteristics, we conducted subgroup analyses stratified by age (< 60 years vs. ≥ 60 years), sex (male vs. female), and CKM syndrome stage (stages 0–1 vs. stages 2–3) (Table 5; Fig. 4). Age-stratified analysis showed that most indices demonstrated stronger predictive ability in the younger group. For example, TyG-WC had HR 1.18 (95% CI: 1.08–1.29) in the younger group and HR 1.11 (95% CI: 1.01–1.21) in the older group. Sex-stratified analysis revealed clear sex differences. BMI, TyG-BMI, and BRI showed significant associations only in males, where BMI had HR 1.36 (95% CI: 1.10–1.68) in males but no significant association in females (HR = 1.00). Interaction tests showed significant sex interactions for CVAI, BMI, BRI, TyG-BMI, and AIP (all P < 0.05). CKM stage-stratified analysis showed that most indices demonstrated stronger predictive ability in CKM stages 2–3 population. For example, CTI showed significant association only in CKM stages 2–3 (HR = 1.11, P = 0.011) but no significant association in CKM stages 0–1. All indices had age and CKM stage interaction P-values > 0.05.

Table 5.

Subgroup analysis of 16 obesity and lipid-related indices for CVD risk prediction

Index	Age < 60 years	Age ≥ 60 years	P for interaction	Male	Female	P for interaction	CKM Stages 0–1	CKM Stages 2–3	P for interaction
CVAI	1.07 (1.02–1.12)	1.03 (0.97–1.09)	0.106	1.18 (1.06–1.32)	1.03 (0.97–1.08)	0.010	1.03 (0.95–1.12)	1.05 (1.00-1.10)	0.288
CTI	1.13 (1.03–1.23)	1.06 (0.96–1.16)	0.065	1.15 (1.04–1.27)	1.06 (0.97–1.16)	0.192	1.01 (0.90–1.14)	1.11 (1.02–1.20)	0.453
BMI	1.03 (0.97–1.08)	1.01 (0.95–1.07)	0.483	1.36 (1.10–1.68)	1.00 (0.96–1.06)	0.005	1.01 (0.94–1.09)	1.02 (0.97–1.07)	0.216
WHTR	1.10 (1.03–1.17)	1.05 (0.97–1.12)	0.094	1.11 (0.98–1.25)	1.06 (1.00-1.12)	0.362	1.04 (0.95–1.14)	1.07 (1.00-1.14)	0.318
ABSI	1.05 (0.97–1.15)	1.01 (0.93–1.09)	0.284	1.00 (0.91–1.10)	1.05 (0.97–1.13)	0.518	1.00 (0.90–1.11)	1.03 (0.95–1.12)	0.205
BRI	1.03 (0.99–1.08)	1.02 (0.96–1.07)	0.395	1.25 (1.02–1.53)	1.02 (0.97–1.06)	0.023	1.01 (0.94–1.09)	1.03 (0.98–1.08)	0.235
CI	1.10 (1.00-1.21)	1.03 (0.95–1.12)	0.120	1.03 (0.94–1.14)	1.08 (1.00-1.18)	0.615	1.02 (0.92–1.14)	1.06 (0.98–1.16)	0.240
TyG	1.06 (0.98–1.15)	1.06 (0.97–1.16)	0.433	1.10 (1.01–1.20)	1.04 (0.95–1.12)	0.286	0.97 (0.86–1.08)	1.07 (1.00-1.16)	0.391
TyG_BMI	1.03 (0.98–1.09)	1.01 (0.96–1.07)	0.378	1.37 (1.14–1.65)	1.01 (0.96–1.06)	0.002	1.01 (0.94–1.09)	1.02 (0.98–1.07)	0.232
TyG_WC	1.18 (1.08–1.29)	1.11 (1.01–1.21)	0.053	1.16 (1.05–1.28)	1.14 (1.05–1.24)	0.616	1.07 (0.96–1.20)	1.16 (1.07–1.25)	0.612
TyG_WHTR	1.12 (1.04–1.20)	1.06 (0.98–1.14)	0.064	1.14 (1.02–1.27)	1.07 (1.00-1.14)	0.237	1.03 (0.93–1.14)	1.09 (1.02–1.17)	0.457
VAI	1.05 (0.98–1.11)	1.00 (0.91–1.10)	0.257	1.04 (0.95–1.14)	1.03 (0.97–1.10)	0.746	0.94 (0.83–1.06)	1.06 (1.00-1.12)	0.327
LAP	1.08 (1.01–1.15)	1.04 (0.95–1.14)	0.214	1.07 (0.99–1.17)	1.06 (0.98–1.14)	0.628	1.00 (0.90–1.11)	1.08 (1.01–1.15)	0.405
AIP	1.07 (0.99–1.16)	1.08 (0.98–1.18)	0.5380	1.15 (1.06–1.26)	1.01 (0.93–1.10)	0.022	0.95 (0.86–1.06)	1.11 (1.03–1.20)	0.427
RC	1.03 (0.96–1.11)	1.03 (0.94–1.12)	0.5847	1.07 (0.99–1.16)	1.00 (0.92–1.08)	0.162	0.94 (0.84–1.06)	1.05 (0.98–1.12)	0.277
NHDL	1.04 (0.96–1.11)	1.03 (0.94–1.12)	0.5660	1.06 (0.99–1.14)	1.00 (0.92–1.09)	0.248	0.95 (0.84–1.06)	1.05 (0.98–1.12)	0.289

Notes: Significant association (P < 0.05). Data are presented as hazard ratio (95% confidence interval). All models were adjusted for age, gender, marital status, education, residence, smoking, drinking, hypertension, diabetes mellitus, dyslipidemia, creatinine, and CRP (excluding the stratifying variable in each subgroup). P for interaction was calculated using interaction terms in Cox regression models. CVAI, Chinese visceral adiposity index; CTI, C-reactive protein-triglyceride glucose index; BMI, body mass index; WHTR, waist-to-height ratio; ABSI, A body shape index; BRI, body roundness index; CI, conicity index; TyG, triglyceride glucose index; TyG_BMI, triglyceride glucose-body mass index; TyG_WC, triglyceride glucose-waist circumference index; TyG_WHTR, triglyceride glucose-waist-to-height ratio index; VAI, visceral adiposity index; LAP, lipid accumulation product; AIP, atherogenic index of plasma; RC, remnant cholesterol; NHDL, non-high-density lipoprotein cholesterol; CKM, cardiovascular-kidney-metabolic

Fig. 4.

Subgroup analysis of 16 obesity and lipid-related indices for CVD risk prediction by age, sex, and CKM syndrome stage. Notes: Forest plots showing hazard ratios (HR) with 95% confidence intervals (CI) for cardiovascular disease risk across different subgroups. Each panel represents one of the 16 obesity and lipid-related indices: CVAI (Chinese visceral adiposity index), CTI (C-reactive protein-triglyceride glucose index), BMI (body mass index), WHTR (waist-to-height ratio), ABSI (A body shape index), BRI (body roundness index), CI (conicity index), TyG (triglyceride glucose index), TyG_BMI (triglyceride glucose-body mass index), TyG_WC (triglyceride glucose-waist circumference index), TyG_WHTR (triglyceride glucose-waist-to-height ratio index), VAI (visceral adiposity index), LAP (lipid accumulation product), AIP (atherogenic index of plasma), RC (remnant cholesterol), and NHDL (non-high-density lipoprotein cholesterol). Subgroup analyses were stratified by age (< 60 years vs. ≥ 60 years), sex (male vs. female), and CKM syndrome stage (stages 0–1 vs. stages 2–3). All models were adjusted for age, gender, marital status, education, residence, smoking, drinking, hypertension, diabetes mellitus, dyslipidemia, creatinine, and C-reactive protein (excluding the stratifying variable in each subgroup). Filled circles indicate statistically significant associations (P < 0.05), while empty circles represent non-significant associations (P ≥ 0.05). The dashed vertical line represents the null effect (HR = 1.0). Different colors distinguish the 16 indices. CKM, cardiovascular-kidney-metabolic; CVD, cardiovascular disease

To further evaluate the CVD predictive value of different obesity and lipid-related indices in CKM stages 0–3 population, based on previous Cox regression analysis results, we selected the best-performing indices and constructed multiple prediction models based on Model 3. Variable selection criteria were: (1) maintaining statistical significance in fully adjusted models (P < 0.05); (2) having high hazard ratios; (3) representing different pathophysiological mechanisms.

Based on these criteria, we selected four indices: CTI (HR = 1.10, P = 0.004), TyG-WC (HR = 1.15, P < 0.001), AIP (HR = 1.07, P = 0.021), and CVAI (HR = 1.05, P = 0.008), representing inflammation-metabolism integration, insulin resistance-abdominal obesity, atherosclerotic risk, and Chinese population visceral obesity characteristics, respectively. The 4 constructed prediction models were: Model 4: Model 3 + CTI; Model 5: Model 3 + TyG-WC; Model 6: Model 3 + AIP; Model 7: Model 3 + CVAI.

Table 6 shows performance comparison results for the 4 prediction models: C-indices ranged from 0.6396 to 0.6434. The insulin resistance-obesity model performed best, with C-index 0.6434 (95% CI: 0.6267–0.6602). Compared to the basic model, all models with added biomarkers showed AIC improvement, with the insulin resistance-obesity model showing the most significant AIC reduction (ΔAIC = -16.64). Likelihood ratio test results showed that all models including biomarkers had statistically significant improvements compared to the basic model.

Table 6.

Comparison of CVD prediction models performance

Model	Variables	C-index	95% CI	ΔAIC	LRT P-value
Model 3	Basic covariates	0.6396	(0.6228–0.6565)	0.00	Ref
Model 4	Basic + CTI	0.6426	(0.6259–0.6593)	-6.04	< 0.001
Model 5	Basic + TyG-WC	0.6434	(0.6267–0.6602)	-16.64	< 0.001
Model 6	Basic + AIP	0.6428	(0.6261–0.6596)	-3.28	< 0.001
Model 7	Basic + CVAI	0.6419	(0.6252–0.6587)	-2.94	< 0.001

Notes: Basic covariates include age, gender, marital status, education, residence, smoking, drinking, MET, hypertension, diabetes mellitus, dyslipidemia, creatinine, and CRP. C-index, concordance index; CI, confidence interval; ΔAIC, difference in AIC compared to base model; LRT, likelihood ratio test; CTI, C-reactive protein-triglyceride glucose index; TyG-WC, triglyceride glucose-waist circumference index; AIP, atherogenic index of plasma; CVAI, Chinese visceral adiposity index

To evaluate the predictive performance and clinical utility of the best-performing model (the Insulin Resistance-Obesity model, which incorporates the TyG-WC index), we conducted a comprehensive validation analysis.Further analysis of prediction accuracy showed a statistically significant Integrated Discrimination Improvement (IDI) of 0.0035 (95% CI: 0.0018–0.0052, P < 0.001), indicating that the new model more accurately assigns higher risk to individuals who develop CVD and lower risk to those who do not. The Net Reclassification Improvement (NRI) was not statistically significant (Total NRI = 0.0130, P = 0.331).To assess model stability and guard against overfitting, we performed rigorous internal validation using 2,000 bootstrap resamples. The results confirmed the model’s robustness, showing minimal optimism (-0.0031) and yielding an optimism-corrected C-index of 0.6465. The model also exhibited excellent calibration, as shown in the calibration plot (Fig. 5), with a calibration slope of 0.980, which is extremely close to the ideal value of 1.0.To directly assess its clinical value, we performed a Decision Curve Analysis (DCA). As shown in Fig. 6, the Insulin Resistance-Obesity model provided a greater net benefit than both the “treat-all” and “treat-none” strategies across a wide range of clinically relevant risk thresholds. Crucially, it was also superior to the base model, demonstrating that using this model to guide clinical decisions could lead to better patient outcomes. A summary of the validation metrics for the Insulin Resistance-Obesity model is presented in Table 7.

Fig. 5.

Calibration plot for the insulin resistance-obesity model in predicting 6-year CVD risk. Notes: The x-axis represents the model-predicted probability of a cardiovascular disease (CVD) event within 6 years. The y-axis represents the actual observed probability of CVD, estimated using the Kaplan-Meier method. The plot is generated by grouping subjects into deciles based on their predicted risk. The dashed red line (“Ideal”) represents a state of perfect calibration where predicted risk equals observed risk. The solid blue line (“Fitted”) represents the performance of the Insulin Resistance-Obesity model

Fig. 6.

Decision curve analysis (DCA) for the competing models in predicting 6-year CVD risk. Notes: The y-axis represents the net benefit, and the x-axis represents the threshold probability. The net benefit is a measure of the clinical utility of a model, calculated by weighing the benefits of true-positive predictions against the harms of false-positive predictions. The grey horizontal line at y = 0 represents the “Treat None” strategy, while the other grey line represents the “Treat All” strategy. The colored lines represent the net benefit of using each respective prediction model to make clinical decisions

Table 7.

Model performance validation and calibration analysis

Validation Metric	Value	95% CI / P-value
Discrimination vs. Base Model
IDI	0.0035	(0.0018–0.0052), P < 0.001
Total NRI	0.0130	P = 0.331
Internal Validation & Calibration
Optimism-Corrected C-index	0.6465	(Bootstrap, n = 2000)
Calibration Slope	0.980	(Ideal = 1.0)

Notes: IDI, integrated discrimination improvement; NRI, net reclassification improvement; C-index, concordance index; CI, confidence interval. Bootstrap validation was performed with 2000 resamples to assess internal validity

Risk stratification analysis based on the insulin resistance-obesity model showed good clinical utility. The 6-year CVD event rates for low, intermediate, and high-risk groups divided by tertiles were 11.3%, 17.7%, and 29.8%, respectively, showing clear risk gradients. The risk gradient between high-risk and low-risk groups was 18.5% points, and between intermediate-risk and low-risk groups was 6.4% points, indicating that this model can effectively identify populations with different risk levels, providing valuable stratification information for clinical decision-making. Table 8 shows detailed risk stratification results.

Table 8.

Risk stratification analysis based on insulin resistance-obesity model

Risk Group	N (Patients)	N (Events)	6-Year Event Rate (%)	Risk Gradient (%)
Low Risk	1,928	218	11.3	-
Intermediate Risk	1,927	341	17.7	6.4
High Risk	1,927	575	29.8	18.5

Notes: Risk groups were defined by tertiles of predicted 6-year CVD risk based on the insulin resistance-obesity model (TyG-WC index combined with basic covariates). Risk gradient represents the absolute difference in event rates compared to the low-risk group

To verify the robustness of our main findings, we conducted two sensitivity analyses addressing potential reverse causation. First, we performed an analysis on a subset of the cohort that excluded participants who experienced a CVD event within the first two years of follow-up (n = 5,374). The results showed that the associations remained stable; key indices such as TyG-WC (HR = 1.143, 95% CI: 1.051–1.243, P = 0.002) and TyG-WHtR (HR = 1.090, 95% CI: 1.017–1.169, P = 0.015) retained their significance. Furthermore, a landmark analysis at the 2-year mark (n = 5,590) confirmed these findings. In this analysis, both TyG-WC (HR = 1.116, 95% CI: 1.042–1.195, P = 0.0016) and TyG-WHtR (HR = 1.066, 95% CI: 1.004–1.133, P = 0.038) remained significant predictors of subsequent CVD events. Taken together, these consistent results from two different analytical approaches strongly support the robustness of our main conclusions (Supplementary Table 2).

Discussion

This study, based on the nationally representative CHARLS cohort, systematically compared the ability of 16 obesity and lipid-related indices to predict CVD risk in adults with CKM stages 0–3. Main findings include: In fully adjusted models, TyG-WC index demonstrated the strongest CVD risk prediction ability (HR = 1.15, 95% CI: 1.08–1.22). This was followed by TyG-WHtR, CTI, WHtR, CVAI, LAP, AIP, and TyG, all showing significant CVD risk prediction value. These findings indicate that different obesity and lipid-related indices participate in CVD development through different pathophysiological mechanisms. They provide diverse assessment tools for risk stratification in CKM stages 0–3 populations.

TyG-related indices (TyG, TyG-WC, TyG-WHtR) all showed significant CVD risk prediction ability in this study. This is consistent with recent international studies [16, 17, 41–43]. TyG index, as a surrogate marker of IR, reflects individual glucose and lipid metabolism status. IR is a key pathophysiological link connecting obesity and CVD. It not only directly affects glucose metabolism but also increases CVD risk through multiple mechanisms. These include activating inflammatory pathways, promoting atherosclerotic plaque formation, and affecting vascular endothelial function [44, 45]. TyG-WC and TyG-WHtR indices further incorporate information about abdominal obesity. They provide a more comprehensive reflection of the synergistic effects of IR and visceral fat accumulation. Waist circumference and waist-to-height ratio, as important indicators of abdominal obesity, can more accurately reflect visceral fat content [46]. Visceral adipocytes have stronger metabolic activity and can secrete more pro-inflammatory cytokines (such as TNF-α, IL-6) and adipokines (such as leptin, adiponectin). These directly participate in atherosclerosis development [47, 48].

CTI index, as a composite indicator combining inflammatory and metabolic factors, also showed important CVD risk prediction value in this study. CRP, as a classic marker of systemic inflammation, reflects the body’s chronic inflammatory state. Chronic inflammation is an important driving factor for atherosclerosis [49]. When CRP is combined with TyG index, it can simultaneously capture information from both inflammatory and metabolic abnormality dimensions. This provides a more comprehensive perspective for CVD risk assessment. Notably, CTI index showed significant association only in CKM stages 2–3 populations in subgroup analysis. This may be related to the pathophysiological characteristics of CKM syndrome. In early CKM stages, inflammatory response is relatively mild. However, in CKM stages 2–3, chronic inflammatory state is more obvious, and CTI index can better capture the association between inflammation-metabolism-cardiovascular risk [50, 51].

CVAI, as a visceral obesity assessment indicator developed specifically for Chinese population characteristics, performed well in this study with important clinical significance. Compared to traditional visceral fat assessment methods, CVAI considers the body type characteristics and metabolic features of Chinese populations. It provides more accurate reflection of visceral fat distribution characteristics in Asian populations [19]. Visceral fat accumulation is not only closely related to IR but also affects systemic metabolic homeostasis through the liver-fat axis. This thereby increases CVD risk [52, 53]. The advantage of CVAI index lies in its comprehensive consideration of multiple factors including age, sex, BMI, waist circumference, triglycerides, and HDL cholesterol. It provides more comprehensive evaluation of individual metabolic risk status.

LAP and AIP, as traditional lipid-related indicators, also showed significant CVD risk prediction value in this study. LAP, calculated based on waist circumference and triglycerides, reflects abnormal lipid accumulation in the abdomen. It is superior to traditional BMI in identifying metabolic syndrome [20]. AIP reflects the ratio relationship between small dense low-density lipoprotein and HDL cholesterol. It provides good evaluation of atherosclerotic risk [37]. The significance of these indicators suggests that even when TyG-related indices perform excellently, traditional lipid-related indicators still have independent predictive value. This possibly reflects different pathophysiological pathways in CVD development.

Age-stratified analysis showed that most obesity and lipid-related indices demonstrated stronger CVD risk prediction ability in the younger group (< 60 years). This is consistent with previous study results [54]. In younger populations, metabolic abnormalities contribute relatively more to CVD risk. However, in elderly populations, age itself becomes the main CVD risk factor. This possibly weakens the predictive value of other risk factors [55]. Additionally, metabolic abnormalities in younger populations often reflect more severe genetic susceptibility or lifestyle issues. Thus, they have stronger predictive significance [56]. From a clinical prevention perspective, this finding emphasizes the importance of early identification and intervention of metabolic abnormalities in middle-aged populations.

CKM stage-stratified analysis showed that most indices demonstrated stronger CVD risk prediction ability in CKM stages 2–3 populations. This is consistent with the pathophysiological characteristics of CKM syndrome. CKM stages 2–3 patients already have clear metabolic risk factors or subclinical CVD. Their pathophysiological states are closer to the critical point of CVD occurrence. This makes the predictive value of obesity and lipid-related indices more prominent [3]. This finding supports the rationality of adopting stratified risk assessment strategies in CKM stage management.

The four prediction models constructed in this study all showed similar predictive ability. Although the insulin resistance-obesity model (based on TyG-WC index) performed best, model performance was relatively similar. This finding suggests that different pathophysiological mechanisms have similar importance in CVD risk prediction. These mechanisms include inflammation-metabolism integration, insulin resistance-abdominal obesity, atherosclerotic risk, and Chinese population visceral obesity characteristics. The similar predictive abilities across different mechanistic approaches support a multi-pathway hypothesis of CVD development in CKM syndrome, where metabolic, inflammatory, and adiposity-related processes operate synergistically rather than independently. This finding has significant clinical implications for precision medicine approaches, suggesting that biomarker selection should be tailored to individual patient phenotypes and clinical contexts. For instance, CTI may be preferentially employed in patients with evident systemic inflammation (elevated CRP), while TyG-WC or CVAI would be more appropriate for patients presenting with central obesity patterns.

In-depth validation analysis of the insulin resistance-obesity model (based on TyG-WC index) further confirmed its clinical application value in CKM stages 0–3 populations. IDI analysis showed this model had significant discrimination improvement ability compared to the basic model (P < 0.001). Bootstrap internal validation results showed good model stability and reproducibility. The corrected C-index (0.6465) was close to the original C-index (0.6434). This suggests no obvious overfitting risk. Calibration curve analysis showed excellent calibration, indicating high agreement between predicted probabilities and actual incidence rates. This is important for clinical risk stratification.

Risk stratification analysis showed that the insulin resistance-obesity model based on TyG-WC index can effectively distinguish CKM stages 0–3 populations with different risk levels. The 6-year CVD event rate in the high-risk group (29.8%) increased by 18.5% points compared to the low-risk group (11.3%). This risk gradient has important guiding significance in clinical practice. More aggressive lifestyle interventions and pharmacological treatments should be considered for high-risk populations, while routine preventive measures can be used for low-risk populations. TyG-WC index, as a simple and readily available indicator, requires only fasting blood glucose, triglycerides, and waist circumference measurements. It does not require expensive imaging examinations or complex laboratory tests. This makes it particularly suitable for promotion in resource-limited primary healthcare institutions. This simplicity gives it unique advantages in early screening and risk stratification for CKM syndrome. It helps achieve precision prevention and individualized management goals.

This study is the first to systematically compare 16 different obesity and lipid-related indices within the CKM syndrome framework. It provides more comprehensive evidence for risk assessment in CKM stages 0–3 populations. From a clinical practice guidance perspective, study results provide diverse choices for risk stratification in CKM stages 0–3 populations. Besides TyG, CTI index, although requiring additional CRP testing, has unique predictive value in CKM stages 2–3 patients. It is particularly suitable for identifying high-risk patients with high inflammatory states. Indices such as CVAI, LAP, and AIP each have unique characteristics and can be selected based on patient-specific conditions and testing circumstances.

Study limitations mainly include: First, the ascertainment of CVD events relied on self-reported physician diagnoses, which is subject to recall bias. However, this potential misclassification is likely non-differential, as inaccuracies in reporting are probably unrelated to baseline metabolic index values. Such misclassification typically biases hazard ratios towards the null (i.e., towards 1.0). Consequently, our reported associations may be conservative, and the true predictive strength of indices like TyG-WC could be even stronger than what we observed. Second, as an observational study, causal relationships cannot be established, and unmeasured confounding factors may exist. Third, the study population mainly comprised Chinese adults aged 45 and older, and our prediction models were validated internally but not in an independent external cohort. Therefore, the generalizability of our findings to other populations and ethnicities requires further investigation. Fourth, our study design is based on a single measurement of all predictor indices at baseline. Consequently, our findings reflect the predictive value of an individual’s initial risk status and do not capture the impact of changes in these metabolic indices over the follow-up period.

Conclusions

This study, based on nationally representative CHARLS cohort data, systematically compared the CVD risk prediction ability of 16 obesity and lipid-related indices in adults with CKM stages 0–3. The study found that TyG-WC index demonstrated the strongest CVD risk prediction ability. These indices provide effective assessment tools for CVD risk stratification in CKM stages 0–3 populations through different pathophysiological mechanisms.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Yilin Pan: Conceptualization, Methodology, Formal analysis, Data curation, Writing - original draft, Visualization. Beibei Du: Methodology, Formal analysis, Funding acquisition, Writing - review & editing. Long Feng: Methodology, Formal analysis, Data curation, Writing - original draft, Visualization. Jingru Bi: Conceptualization, Supervision, Project administration, Writing - review & editing.

Funding

This study was supported by National Natural Science Foundation of China (No. 82470327, 82100337), Jilin Provincial Natural Science Foundation (No. YDZJ202201ZYTS097) and the Project of Jilin Provincial Department of Finance (No. 2022SCZ40).

Data availability

The data that support the findings of this study are available from the CHARLS repository (http://charls.pku.edu.cn/index/en.html).

Declarations

Ethics approval and consent to participate

The CHARLS study was approved by the Biomedical Ethics Committee of Peking University (approval number: IRB00001052-11015). All participants provided written informed consent before participating in the survey.

Competing interests

The authors declare no competing interests.