Comparison of the TyG index, TyG-traditional obesity indices, and TyG-novel obesity indices: does increased complexity help in predicting cardiometabolic multimorbidity?

Abstract

Background

The triglyceride-glucose (TyG) index is an important determinant influencing the incidence of cardiometabolic multimorbidity (CMM). However, it remains unclear whether combining the TyG index with novel obesity indices (CVAI/BRI/ABSI/WWI) can improve the risk stratification of CMM. This study aimed to systematically compare the incremental risk assessment and predictive value of the TyG index, TyG-traditional obesity indices (TyG-WC/TyG-WHtR/TyG-BMI), and TyG-novel obesity indices (TyG-CVAI/TyG-BRI/TyG-ABSI/TyG-WWI) for CMM.

Methods

Trajectory changes and cumulative exposure of TyG-related parameters were quantified using repeated measurements from the CHARLS cohort (n = 3,885). The study endpoint CMM was defined as a comorbid condition encompassing two or more cardiometabolic diseases, namely diabetes, stroke and heart diseases. A multi-model analytical strategy was employed to evaluate the associations between TyG-related parameters and CMM, as well as the contribution of their components. The net reclassification index and integrated discrimination improvement were employed to evaluate the improvement in predictive performance of these indices.

Results

Over a median follow-up period of 8 years, we identified a linear positive association between TyG-related parameters and CMM, with the cumulative effects of glucose and obesity emerging as the key drivers. Compared with the baseline TyG index, the incremental risk assessment value for CMM improved by 10%-17% (baseline) and 12%-20% (cumulative exposure) for TyG-traditional obesity indices, while the improvement for TyG-novel obesity indices ranged from − 1% to 16% and 5%-19%, respectively. In summary, all TyG-traditional obesity indices demonstrated strong associations with CMM, whereas among the TyG-novel obesity indices, only TyG-CVAI showed a similarly strong association. Furthermore, all TyG-related parameters showed significantly increased hazard ratios in their highest-exposure or poor-control status versus the reference (lowest exposure or good control): TyG-index (1.69/2.05), TyG-WC (2.24/2.28), TyG-WHtR (1.92/2.05), TyG-BMI (1.85/2.27), TyG-CVAI (1.89/2.07), TyG-BRI (1.94/2.08), TyG-ABSI (1.70/1.85), and TyG-WWI (1.97/1.95). Predictive analyses showed that, except for TyG index, TyG-ABSI and TyG-WWI, all other TyG-related parameters provided a certain degree of net improvement compared with the baseline risk model.

Conclusion

All eight TyG-related parameters can predict the incidence of CMM. Given their relative simplicity, the TyG-traditional obesity indices demonstrate superior incremental risk assessment and predictive value for CMM compared to the TyG-novel obesity indices and the TyG index, positioning them as promising and more practical tools for clinical practice.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12944-025-02836-8.

Introduction

Multimorbidity is defined as the co-occurrence of at least two chronic conditions in an individual [1]. This phenomenon not only accelerates health deterioration but also has profound, compounding implications for family care, healthcare system capacity, and socioeconomic development. These challenges are particularly acute in resource-limited regions [2, 3]. Driven by population aging, the burden of multimorbidity has emerged as a major global public health challenge, particularly cardiometabolic multimorbidity (CMM) [4]. CMM is defined as the coexistence of two or more cardiometabolic diseases (CMDs, including heart disease, diabetes, and stroke) [5]. Evidence indicates that the global prevalence of CMM among the elderly aged 60 and above is approximately 5% [6]. In China, the prevalence of CMM has risen markedly in recent years, more than doubling within five years [7]. Current data place the prevalence of CMM in China between 5.9% and 16.9%, and this severe disease burden has now emerged as a critical public health challenge [7, 8]. Given that CMM is associated with a higher risk of mortality and shorter life expectancy compared to a single CMD [9, 10], the early identification and intervention of its potential risk factors are particularly crucial.

Obesity and insulin resistance (IR) are two key risk factors affecting vascular health, and they play an important role in the pathogenesis of various CMDs [11–13]; therefore, the effective assessment of these two factors has become a crucial part of CMM risk evaluation. In clinical practice, the assessment of obesity and IR typically relies on body composition analyzers/imaging techniques and the hyperinsulinemic-euglycemic clamp, respectively [14, 15]. However, the operational complexity of these methods limits their applicability in large-scale population studies, thereby hindering their widespread use in epidemiological surveys [14, 16]. To overcome the limitations of complex detection methods in large-scale population studies, a series of relatively simple yet effective assessment tools have been developed. These include surrogate indices for IR, such as the triglyceride-glucose (TyG) index [17], and a range of obesity parameters, encompassing both traditional indicators like waist circumference (WC), waist-to-height ratio (WHtR), and body mass index (BMI), as well as novel parameters such as the Chinese Visceral Adiposity Index (CVAI), Weight-Adjusted Waist Index (WWI), Body Roundness Index (BRI), and A Body Shape Index (ABSI) [18–22]. Subsequent large-scale epidemiological studies have confirmed that these tools can effectively predict the risk of CMD at low cost, demonstrating favorable cost-effectiveness [18, 23–26]. More importantly, research has found that combining the TyG index with various obesity indicators (including both traditional and novel TyG-obesity indices) produces a significant synergistic effect and enhances the incremental predictive value for assessing CMD risk and incident events [27–30]. This integrated strategy further demonstrates broad application potential in CMM risk assessment, providing a powerful tool for population risk stratification and early warning [31–35]. However, the utility of these updated indices, especially the combinations of the TyG index with novel obesity indices (e.g., TyG-CVAI, TyG-BRI, TyG-ABSI, TyG-WWI), for assessing CMM risk remains unclear. Furthermore, the academic community still lacks direct comparative data regarding the incremental value and predictive performance of the TyG index, TyG-traditional obesity indices, and TyG-novel obesity indices in CMM risk assessment. The core objective of this study is to fill this knowledge gap: by leveraging the China Health and Retirement Longitudinal Study (CHARLS) prospective cohort, this study aims to systematically quantify the associations between various TyG-related parameters and CMM and to directly compare their predictive performance.

Methods

Study design and cohort description

This comparative study was conducted using data from the 2012–2020 waves of the CHARLS prospective cohort. The design of this cohort has been described in detail elsewhere [36]. In brief, CHARLS is a nationwide, longitudinal household follow-up survey of the middle-aged and elderly populations in China. The study employed a probability proportional to size sampling method, randomly selecting 150 counties/districts across 28 provinces in China. Three administrative villages or communities were further randomly selected from each county/district, ultimately forming 450 urban and rural community sampling sites (the selection process is illustrated in Supplementary material 1, Figure S1). The longitudinal design of the cohort is as follows: the cohort was initiated in 2012 and has conducted nationwide follow-ups every 2–3 years to update information on sociodemographics, health status, and lifestyle. To date, four rounds of follow-up have been completed (in 2013, 2015, 2018, and 2020). Among all five survey waves, the first (2012) and third (2015) rounds provided hematological and physical measurement data, which served as the core data basis for this study to assess the baseline levels and longitudinal changes in TyG-related parameters. This study included participants from the first three survey waves who had complete TyG-related data and were free of CMM. Incident CMM cases were systematically identified using follow-up data from the fourth and fifth waves (Note: CMD diagnoses were cross-verified using information from multiple survey rounds, with corrections made for any inconsistent reports). A detailed flowchart is provided in Fig. 1.

Fig. 1.

CONSORT flow diagram

Assessment of missing data

To evaluate the potential impact of missing data in TyG-related parameters on the study results, we systematically compared the demographic and clinical characteristics between the group with complete data and the group with missing data at the baseline survey. The comparative analysis revealed no significant differences in most baseline characteristics between the two groups (Supplementary material 2: Tables S1 and S2). This distribution pattern suggests a high likelihood that the missing data are consistent with a Missing Completely at Random mechanism, meaning the missingness is not driven by any measured or unmeasured covariates in this study [37].

A total of 3,885 participants from the cohort were finally included in the analysis of this study. All covariates had extremely low rates of missing data (highest: 0.82%), and the specific distribution is shown in Supplementary material 1: Figure S2 and Supplementary material 2: Tables S3. Given the minimal proportion of missing data, all subsequent statistical analyses were conducted directly on the original dataset containing these minimal missing values to preserve full information integrity.

Research ethics

The CHARLS study was approved by the Ethics Review Committee of Peking University (IRB No.: 00001052–11015). Before data collection, the research team fully informed each participant of the study objectives, data usage, and privacy protection measures, and all participants provided voluntary written informed consent. The design and implementation of this study strictly adhered to the ethical principles of the Declaration of Helsinki throughout. The conduct and reporting of the study complied with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines to ensure scientific rigor and ethical compliance.

Assessment of exposure factors

This study defined exposure factors to capture both immediate and cumulative characteristics. For this purpose, it incorporated TyG-related parameters measured at baseline as core indicators of the initial exposure level, and integrated longitudinal data from the baseline and third-wave surveys to quantitatively assess cumulative exposure and trajectory changes. The specific assessment protocol is as follows [17, 29, 30]:

Due to the limitation of having only two repeated measurements for the blood samples in the current study, we were unable to model complex nonlinear trajectories. As an alternative approach, this study employed a K-means clustering algorithm based on Euclidean distance to identify the dynamic evolution patterns (simple trajectories) of TyG-related parameters between 2012 and 2015. This method is renowned for its computational efficiency and strong result visualization capabilities [38]; its core principle involves minimizing the sum of squared distances from data points to their cluster centroid within each cluster through iterative optimization [38, 39]. To determine the optimal number of clusters, we used the elbow method for evaluation. As shown in Supplementary material 1: Figure S3, when the number of clusters (k) was 3, a distinct inflection point occurred in the improvement of within-cluster variance. This indicated that further increasing the number of clusters would yield limited gains in the model’s explanatory power, leading to the selection of three as the optimal number of clusters. Based on this, we identified three distinct trajectory patterns: Class 1 (low-stable, indicating good control), Class 2 (moderate-stable, indicating moderate control), and Class 3 (high-stable, indicating poor control). Compared with traditional methods that rely on predefined classifications, the use of k-means clustering to assess simple trajectory changes offers multiple advantages [40–42]. First, it identifies patterns of biomarker change in a data-driven manner, effectively reducing the subjective bias introduced by researchers when predefining categories. Second, this approach more effectively captures the continuous nature of biomarker changes and accommodates their underlying data distribution. Furthermore, by revealing the inherent structure within the data, it ultimately enhances the interpretability and objectivity of the study results. Detailed classification data and temporal trends are presented in Fig. 2.

Fig. 2.

Clustering of control trajectories in TyG index, TyG-WC, TyG-WHtR, TyG-BMI, TyG-CVAI, TyG-BRI, TyG-ABSI, and TyG-WWI from 2012 to 2015. Three clusters (Class 1, Class 2, and Class 3) were identified using the k-means method with Euclidean distance. Each cluster is represented by a unique color and shape across all figures to distinguish between groups and highlight variations over time

Study outcomes

The endpoint of this study was defined as incident CMM events, which refers to the coexistence of any two or more of the three chronic metabolic diseases: diabetes, heart disease, and stroke [33]. Diabetes was diagnosed based on the following criteria: fasting plasma glucose (FPG) ≥ 7.0 mmol/L, haemoglobin A1c (HbA1c) level ≥ 6.5%, or a self-reported physician diagnosis [43]. The diagnosis of heart disease and stroke was mainly based on self-reported information collected during the baseline survey and follow-ups. The standardized questionnaire included the following questions: “Have you been diagnosed by a doctor with a congestive heart failure, angina, coronary heart disease, heart attack, or other heart problems?” or “Have you been diagnosed by a doctor with a stroke?”

Assessment of covariates

During the health assessment, well-trained staff measured the participants’ height, weight, and WC using standardized methods. Specifically, height and weight were measured by trained staff with participants barefoot and wearing light clothing. WC was measured at the end of a normal expiration, with participants briefly holding their breath; the measurer used a dedicated WC tape to measure the abdomen at the umbilical level and recorded the value. Based on these basic anthropometric measurements, we calculated several obesity indices, including WHtR, BMI, CVAI, BRI, ABSI, and WWI. To ensure data quality, we performed quality control on a small number of physiologically unreasonable outliers present in the baseline and third-wave follow-up data. The specific method involved verifying and correcting these outliers using longitudinal data from the second national survey (2013), while data points that could not be corrected by this method were excluded. The calculation formulas for obesity indices are as follows [18–22]:

As part of the health examination, staff also collected participants’ personal information (sex, age, living place, education, and marital status), health-related behaviors (including smoking and drinking status), and physician-diagnosed medical history (including kidney disease and hypertension) using a standardized questionnaire. In addition, the current study extracted blood test data from the first and third survey waves. For the processing and analysis of blood samples, the CHARLS research team followed a standardized protocol. First, venous blood samples were collected from participants by uniformly trained professionals in the early morning after an overnight fast. To maximize the stability of blood components, all samples were promptly transferred to a -80 °C freezer for storage immediately after collection. Subsequent biochemical tests were conducted in a unified standardized laboratory, with measured indicators including HbA1c, FPG, total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), triglycerides (TG), high-density lipoprotein‐cholesterol, high-sensitivity C-reactive protein, uric acid (UA), and creatinine (Cr). As a core quality control measure, the research team performed weekly quality control monitoring on the blood data from the first and third waves of CHARLS, ensuring all test results remained within an acceptable control range (± 2 standard deviations of the mean control concentration) [36, 44]. All key laboratory methodological parameters have been summarized and listed in Tables S4 and S5 of Supplementary material 2 to ensure the transparency and reproducibility of the study.

Statistical analysis

To systematically reveal the trajectory changes and cumulative exposure status of TyG-related parameters, the study population was grouped according to the outcome. Standardized differences were calculated to compare the baseline characteristics and their longitudinal changes among the participants, thereby evaluating the balance of each covariate, with a value of 0.1 set as the threshold for clinically significant differences.

We used Cox regression to estimate hazard ratios (HRs) for CMM events associated with TyG-related parameters (including baseline values and longitudinal changes). Covariates were selected based on clinical importance and statistical relevance. Furthermore, analysis of the covariate correlation matrix (Supplementary material 1: Figure S4) revealed high collinearity between UA and Cr, as well as between TC and LDL-C. After evaluation, UA, and TC were ultimately excluded from the multivariable-adjusted models. To ensure the comparability of association coefficients between the TyG index, TyG-traditional obesity indices, and TyG-novel obesity indices (assessed at both baseline and longitudinally) and CMM, we standardized the TyG-related parameters (baseline values and cumulative exposure) using Z-scores and calculated the standardized HRs. In addition, we used restricted cubic spline regression to test the dose-response relationship.

To systematically evaluate and compare the predictive performance of the TyG index, TyG-traditional obesity indices, and TyG-novel obesity indices for incident CMM, this study conducted receiver operating characteristic curve analysis and used the area under the curve to quantify their discriminatory ability. On this basis, we further calculated the net reclassification improvement (NRI) and integrated discrimination improvement (IDI) to measure the incremental predictive value provided by TyG-traditional obesity indices and TyG-novel obesity indices compared with the TyG index alone. Additionally, we further investigated the impact of adding the TyG index, TyG-traditional obesity indices, and TyG-novel obesity indices on the predictive performance of the baseline risk model (which includes traditional risk factors: age, sex, education, living place, kidney disease, hypertension, smoking status, drinking status, high-density lipoprotein‐cholesterol, HDL-C, Cr, HbA1c, and high-sensitivity C-reactive protein). The continuous C-index, NRI, and IDI were calculated as quantitative metrics to assess the improvement in the predictive ability of the model.

After evaluating the associations between TyG-related parameters and CMM, this study further applied weighted quantile sum (WQS) regression with 1,000 bootstrap resamples, based on the final multivariable-adjusted model, to quantify the relative contribution of cumulative exposure to each component of the TyG-related parameters to the risk of incident CMM during the follow-up period. This model assigns a weight between 0 and 1 to each component [45]. The magnitude of the weight directly reflects the relative importance of that variable in predicting CMM events, thereby identifying the key driving components in the mixture exposure.

Several additional analyses were conducted to examine the robustness of the study results: (1) To assess the generalizability of the findings, we further analyzed the association between TyG-related parameters and CMM in subgroups, including males vs. females, middle-aged vs. elderly individuals, participants with vs. without hypertension, and participants with vs. without chronic kidney disease. (2) Missing data were handled by creating dummy variables or using mean/median imputation to form a complete dataset; the association between TyG-related parameters and CMM was then re-analyzed in this dataset. (3) To address potential reverse causality, whereby changes in TyG-related parameters could be a consequence of incident CMM, a lagged analysis was performed by excluding participants diagnosed with CMM during the fourth survey wave. (4) The study comprised middle-aged and elderly participants, where death (a competing event) presents a competing risk that may preclude the observation of CMM incidence. To address this, we applied Fine-Gray competing risk models to the imputed datasets to evaluate associations between exposures and CMM, thereby accounting for this competing risk.

Results

Participant characteristics

During a median follow-up period of 8 years, we identified 279 (7.18%) incident CMM cases, including 115 males and 164 females. Participants diagnosed with CMM were typically older and had lower educational attainment (Table 1). Clinically, these participants had a higher degree of obesity, a higher prevalence of hypertension and kidney disease, along with more severe disorders in glucose, lipid, and UA metabolism (standardized difference > 10%). Notably, the intergroup differences in baseline TyG-related parameters between the CMM and non-CMM groups were the most pronounced (41% ≤ standardized difference ≤ 54%).

Table 1.

Baseline characteristics and cumulative metabolic changes stratified by incident CMM status

	CMM		Standardized difference, % (95% CI)
	No	Yes
No. of subjects	3606	279
Age, years	58.00 (51.00–64.00)	60.00 (55.00–65.00)	0.25 (0.12, 0.37)
Sex, n (%)			0.08 (-0.05, 0.20)
Male	1621 (44.95%)	115 (41.22%)
Female	1985 (55.05%)	164 (58.78%)
Marital status, n (%)			0.07 (-0.05, 0.20)
Married	3238 (89.79%)	244 (87.46%)
Other	368 (10.21%)	35 (12.54%)
Living place, n (%)			0.05 (-0.07, 0.17)
Village	2407 (66.75%)	193 (69.18%)
City	1199 (33.25%)	86 (30.82%)
Education, n (%)			0.27 (0.15, 0.39)
Below primary	1698 (47.10%)	147 (52.69%)
Primary schools	795 (22.05%)	78 (27.96%)
Middle school	751 (20.83%)	36 (12.90%)
High school and above	361 (10.01%)	18 (6.45%)
Kidney disease, n (%)			0.15 (0.03, 0.28)
No	3414 (95.44%)	253 (91.67%)
Yes	163 (4.56%)	23 (8.33%)
Hypertension, n (%)			0.43 (0.31, 0.55)
No	2064 (57.24%)	101 (36.20%)
Yes	1542 (42.76%)	178 (63.80%)
Drinking status, n (%)			0.07 (-0.06, 0.19)
No	2395 (66.47%)	194 (69.53%)
Yes	1208 (33.53%)	85 (30.47%)
Smoking status, n (%)			0.08 (-0.04, 0.20)
No	2524 (70.17%)	205 (73.74%)
Yes	1073 (29.83%)	73 (26.26%)
FPG, mg/dL	101.34 (94.14-109.44)	106.56 (98.46-120.96)	0.38 (0.26, 0.50)
Cr, mg/dL	0.75 (0.64–0.86)	0.76 (0.66–0.87)	0.06 (-0.07, 0.18)
TC, mg/dL	192.93 (38.61)	205.60 (41.80)	0.31 (0.19, 0.44)
TG, mg/dL	101.78 (72.57-146.91)	119.47 (87.61-169.03)	0.23 (0.11, 0.35)
HDL-C, mg/dL	49.87 (40.98–60.31)	45.62 (38.85–55.09)	0.25 (0.13, 0.37)
LDL-C, mg/dL	114.05 (93.94-136.47)	124.10 (101.68-149.61)	0.27 (0.15, 0.40)
hs-CRP, mg/L	0.92 (0.52–1.92)	1.32 (0.68–2.31)	0.08 (-0.04, 0.20)
HbA1c, %	5.21 (0.68)	5.62 (1.32)	0.39 (0.27, 0.51)
UA, mg/dL	4.17 (3.50-5.00)	4.36 (3.51–5.30)	0.16 (0.04, 0.28)
Height, m	1.58 (0.08)	1.58 (0.08)	0.04 (-0.08, 0.17)
Weight, kg	58.77 (10.86)	61.83 (11.50)	0.27 (0.15, 0.40)
BMI, kg/m2	23.45 (3.65)	24.78 (3.77)	0.36 (0.24, 0.48)
WC, cm	84.87 (9.70)	89.19 (9.56)	0.45 (0.33, 0.57)
WHtR	0.54 (0.06)	0.57 (0.06)	0.45 (0.33, 0.57)
WWI	11.12 (0.85)	11.40 (0.84)	0.33 (0.21, 0.45)
ABSI	0.08 (0.01)	0.08 (0.01)	0.22 (0.10, 0.34)
BRI	4.17 (1.32)	4.77 (1.39)	0.44 (0.32, 0.57)
CVAI	92.61 (39.33)	112.87 (38.52)	0.52 (0.40, 0.64)
TyG index	8.62 (0.63)	8.91 (0.75)	0.41 (0.29, 0.54)
TyG-BMI	202.89 (38.56)	221.43 (41.32)	0.46 (0.34, 0.59)
TyG-WC	733.90 (113.71)	796.62 (120.78)	0.53 (0.41, 0.66)
TyG-WHtR	4.65 (0.73)	5.06 (0.80)	0.54 (0.42, 0.66)
TyG-WWI	96.03 (11.03)	101.75 (12.58)	0.48 (0.36, 0.61)
TyG-ABSI	0.71 (0.07)	0.75 (0.08)	0.43 (0.31, 0.56)
TyG-BRI	36.23 (12.53)	42.81 (13.78)	0.50 (0.38, 0.62)
TyG-CVAI	810.54 (375.95)	1018.16 (389.68)	0.54 (0.42, 0.66)
Cumulative TyG index	25.88 (1.62)	26.70 (1.96)	0.45 (0.33, 0.58)
Cumulative TyG-BMI	612.28 (111.01)	664.07 (117.43)	0.45 (0.33, 0.58)
Cumulative TyG-WC	2213.89 (325.79)	2392.85 (350.28)	0.53 (0.41, 0.65)
Cumulative TyG-WHtR	14.05 (2.11)	15.23 (2.34)	0.53 (0.41, 0.65)
Cumulative TyG-WWI	289.31 (30.15)	305.82 (34.63)	0.51 (0.39, 0.63)
Cumulative TyG-ABSI	2.15 (0.19)	2.25 (0.22)	0.48 (0.36, 0.61)
Cumulative TyG-BRI	110.82 (36.89)	129.96 (41.22)	0.49 (0.37, 0.61)
Cumulative TyG-CVAI	2520.89 (1078.46)	3118.30 (1131.00)	0.54 (0.42, 0.66)
Trajectory changes of TyG index			0.48 (0.35, 0.60)
Class 1	1513 (41.96%)	64 (22.94%)
Class 2	1582 (43.87%)	135 (48.39%)
Class 3	511 (14.17%)	80 (28.67%)
Trajectory changes of TyG-BMI			0.46 (0.34, 0.59)
Class 1	1393 (38.63%)	64 (22.94%)
Class 2	1557 (43.18%)	113 (40.50%)
Class 3	656 (18.19%)	102 (36.56%)
Trajectory changes of TyG-WC			0.50 (0.38, 0.62)
Class 1	1362 (37.77%)	51 (18.28%)
Class 2	1495 (41.46%)	123 (44.09%)
Class 3	749 (20.77%)	105 (37.63%)
Trajectory changes of TyG-WHtR			0.49 (0.37, 0.62)
Class 1	1285 (35.64%)	55 (19.71%)
Class 2	1541 (42.73%)	106 (37.99%)
Class 3	780 (21.63%)	118 (42.29%)
Trajectory changes of TyG-WWI			0.47 (0.35, 0.59)
Class 1	1297 (35.97%)	51 (18.28%)
Class 2	1600 (44.37%)	128 (45.88%)
Class 3	709 (19.66%)	100 (35.84%)
Trajectory changes of TyG-ABSI			0.47 (0.35, 0.59)
Class 1	1322 (36.66%)	55 (19.71%)
Class 2	1679 (46.56%)	132 (47.31%)
Class 3	605 (16.78%)	92 (32.97%)
Trajectory changes of TyG-BRI			0.46 (0.33, 0.58)
Class 1	1411 (39.13%)	62 (22.22%)
Class 2	1504 (41.71%)	116 (41.58%)
Class 3	691 (19.16%)	101 (36.20%)
Trajectory changes of TyG-CVAI			0.49 (0.37, 0.61)
Class 1	1248 (34.61%)	50 (17.92%)
Class 2	1599 (44.34%)	117 (41.94%)
Class 3	759 (21.05%)	112 (40.14%)

Values were expressed as mean (standard deviation) or medians (quartile interval) or n (%)

Abbreviations:HbA1c Haemoglobin A1c, HDL-C High‐density lipoprotein‐cholesterol, hs‐CRP High‐sensitivity C‐reactive protein, LDL‐C Low‐density lipoprotein‐cholesterol, TC Total cholesterol, TG Triglycerides, UA Uric acid, Cr Creatinine, CI Confidence interval, TyG Triglyceride glucose, BMI Body mass index, WC Waist circumference, WHtR Waist-to-height ratio, ABSI A body shape index, WWI Weight adjusted waist index, CVAI Chinese Visceral Adiposity Index, BRI Body Roundness Index, CMM Cardiometabolic multimorbidity

In terms of cumulative exposure characteristics and trajectory changes, participants who developed CMM during follow-up had significantly higher cumulative exposure to TyG-related parameters and significantly poorer control. Notably, the largest standardized differences for both cumulative exposure and trajectory control between groups were observed for the TyG-WC, TyG-WHtR, and TyG-CVAI indices (Table 1).

TyG-Related parameters (baseline values and cumulative exposure) in association with CMM risk

Table 2 shows the HRs for the association between Z-score standardized TyG-related parameters (baseline values and cumulative exposure) and CMM. In the unadjusted model, all TyG-related parameters showed strong positive associations with CMM, with the strongest associations observed for TyG-WC, TyG-WHtR, and TyG-CVAI. These associations were slightly attenuated in the subsequent multivariable Cox regression analysis. In the final model, all TyG-traditional obesity indices were strongly associated with CMM. In contrast, among the TyG-novel obesity indices, only TyG-CVAI demonstrated a strong association. Notably, compared to the TyG-traditional obesity indices, the novel indices TyG-WWI and TyG-ABSI demonstrated weaker associations with CMM. Furthermore, in the current study, we identified another important finding: compared with static baseline values, the longitudinal assessment of cumulative exposure to TyG-related parameters exhibited a significant incremental effect on CMM risk assessment, with an increase of approximately 2–7% (Fig. 3).

Table 2.

Association between TyG-related parameters (baseline value and cumulative exposure value) with CMM events

	HR Per SD increase (95% CI)
	Non-adjusted Model	Model I	Model II	Model III
TyG index	1.42 (1.29, 1.56)	1.43 (1.30, 1.58)	1.37 (1.24, 1.51)	1.19 (1.02, 1.38)
TyG-BMI	1.45 (1.32, 1.60)	1.51 (1.37, 1.66)	1.45 (1.30, 1.61)	1.29 (1.14, 1.47)
TyG-WC	1.59 (1.43, 1.76)	1.60 (1.44, 1.78)	1.54 (1.38, 1.72)	1.36 (1.19, 1.57)
TyG-WHtR	1.61 (1.45, 1.78)	1.62 (1.45, 1.81)	1.54 (1.37, 1.73)	1.34 (1.16, 1.54)
TyG-WWI	1.54 (1.39, 1.71)	1.52 (1.37, 1.70)	1.44 (1.29, 1.62)	1.24 (1.07, 1.43)
TyG-ABSI	1.47 (1.33, 1.63)	1.43 (1.29, 1.60)	1.37 (1.22, 1.52)	1.18 (1.03, 1.35)
TyG-BRI	1.52 (1.37, 1.67)	1.53 (1.38, 1.70)	1.44 (1.29, 1.61)	1.27 (1.12, 1.44)
TyG-CVAI	1.61 (1.44, 1.79)	1.57 (1.41, 1.75)	1.50 (1.33, 1.68)	1.35 (1.17, 1.55)
Cumulative TyG index	1.49 (1.35, 1.64)	1.51 (1.36, 1.67)	1.44 (1.30, 1.60)	1.25 (1.08, 1.44)
Cumulative TyG-BMI	1.47 (1.33, 1.64)	1.54 (1.38, 1.70)	1.46 (1.31, 1.63)	1.31 (1.15, 1.49)
Cumulative TyG-WC	1.61 (1.44, 1.79)	1.63 (1.46, 1.82)	1.56 (1.39, 1.76)	1.39 (1.21, 1.59)
Cumulative TyG-WHtR	1.62 (1.46, 1.81)	1.65 (1.48, 1.85)	1.56 (1.39, 1.76)	1.37 (1.19, 1.57)
Cumulative TyG-WWI	1.59 (1.43, 1.76)	1.59 (1.42, 1.79)	1.51 (1.34, 1.70)	1.29 (1.12, 1.49)
Cumulative TyG-ABSI	1.54 (1.39, 1.71)	1.50 (1.35, 1.68)	1.44 (1.29, 1.61)	1.24 (1.08, 1.42)
Cumulative TyG-BRI	1.52 (1.38, 1.69)	1.55 (1.39, 1.72)	1.45 (1.30, 1.63)	1.30 (1.14, 1.47)
Cumulative TyG-CVAI	1.62 (1.45, 1.81)	1.58 (1.41, 1.77)	1.51 (1.34, 1.70)	1.38 (1.19, 1.58)

Abbreviations:HR Hazard ratios, CI Confidence interval, SD Standard deviations, CMM Cardiometabolic multimorbidity; other abbreviations as in Table 1

Model I adjusted for age, sex;

Model II adjusted for age, sex, education, living place, kidney disease, hypertension, smoking status, drinking status;

Model III adjusted for age, sex, education, living place, kidney disease, hypertension, smoking status, drinking status, LDL-C, HDL-C, Cr, HbA1c, hs-CRP

Fig. 3.

Standardized hazard ratios representing the association of TyG-related parameters (baseline value and cumulative exposure value) with cardiometabolic multimorbidity

When we further examined the dose-response relationship curves between TyG-related parameters (baseline values and cumulative exposure) and CMM using a 4-knot restricted cubic spline, no significant nonlinear associations were observed (Fig. 4; all P for nonlinearity > 0.05).

Fig. 4.

Using restricted cubic splines to analyze the shape of the association between TyG-related parameters (baseline/cumulative) and cardiometabolic multimorbidity

TyG-Related parameters (cumulative exposure status and trajectory changes) in association with CMM risk

Table 3 shows the associations of cumulative exposure status and trajectory changes in TyG-related parameters with CMM risk. Across all models, a significantly increased CMM risk was observed in individuals with high exposure to TyG-related parameters and those with poor control of these parameters during follow-up. Based on the final model, the groups with the highest cumulative exposure or a poor-control trajectory exhibited significantly higher risks of CMM compared to those with the lowest exposure or a good-control trajectory. The HRs were as follows: TyG index, 1.69/2.05; TyG-WC, 2.24/2.28; TyG-WHtR, 1.92/2.05; TyG-BMI, 1.85/2.27; TyG-CVAI, 1.89/2.07; TyG-BRI, 1.94/2.08; TyG-ABSI, 1.70/1.85; and TyG-WWI, 1.97/1.95. In contrast, the longitudinal assessment of trajectory changes provided a more pronounced incremental effect on CMM risk stratification. Overall, among all TyG-related parameters, evaluating the longitudinal changes of TyG-traditional obesity indices showed considerable incremental value for predicting CMM risk.

Table 3.

Association of TyG-related parameters (cumulative exposure status and trajectory changes) with CMM risk

	HR (95% CI)
	Non-adjusted model	Model I	Model II	Model III
Cumulative TyG index tertile groups
T1	1.0	1.0	1.0	1.0
T2	1.78 (1.26, 2.51)	1.77 (1.25, 2.50)	1.68 (1.19, 2.39)	1.45 (1.01, 2.07)
T3	2.77 (2.01, 3.81)	2.80 (2.02, 3.87)	2.49 (1.79, 3.48)	1.69 (1.16, 2.45)
Cumulative TyG-BMI tertile groups
T1	1.0	1.0	1.0	1.0
T2	1.35 (0.96, 1.91)	1.45 (1.03, 2.05)	1.38 (0.97, 1.96)	1.18 (0.82, 1.69)
T3	2.60 (1.91, 3.53)	2.92 (2.13, 4.00)	2.58 (1.85, 3.60)	1.85 (1.29, 2.67)
Cumulative TyG-WC tertile groups
T1	1.0	1.0	1.0	1.0
T2	1.79 (1.25, 2.56)	1.83 (1.28, 2.61)	1.80 (1.25, 2.59)	1.53 (1.05, 2.23)
T3	3.23 (2.32, 4.48)	3.30 (2.37, 4.60)	3.02 (2.13, 4.27)	2.24 (1.53, 3.29)
Cumulative TyG-WHtR tertile groups
T1	1.0	1.0	1.0	1.0
T2	1.42 (1.00, 2.02)	1.48 (1.03, 2.11)	1.37 (0.96, 1.98)	1.19 (0.82, 1.72)
T3	2.92 (2.13, 3.99)	3.04 (2.19, 4.22)	2.66 (1.88, 3.76)	1.92 (1.32, 2.80)
Cumulative TyG-WWI tertile groups
T1	1.0	1.0	1.0	1.0
T2	1.79 (1.25, 2.54)	1.83 (1.28, 2.61)	1.71 (1.19, 2.47)	1.46 (1.01, 2.11)
T3	3.11 (2.25, 4.31)	3.16 (2.24, 4.47)	2.77 (1.94, 3.96)	1.97 (1.34, 2.90)
Cumulative TyG-ABSI tertile groups
T1	1.0	1.0	1.0	1.0
T2	1.61 (1.14, 2.27)	1.56 (1.10, 2.22)	1.45 (1.01, 2.06)	1.24 (0.87, 1.78)
T3	2.82 (2.06, 3.88)	2.65 (1.91, 3.67)	2.38 (1.70, 3.33)	1.70 (1.18, 2.44)
Cumulative TyG-BRI tertile groups
T1	1.0	1.0	1.0	1.0
T2	1.25 (0.88, 1.78)	1.31 (0.92, 1.87)	1.23 (0.86, 1.78)	1.12 (0.77, 1.61)
T3	2.79 (2.06, 3.80)	2.92 (2.12, 4.04)	2.56 (1.82, 3.59)	1.94 (1.35, 2.78)
Cumulative TyG-CVAI tertile groups
T1	1.0	1.0	1.0	1.0
T2	1.65 (1.16, 2.35)	1.60 (1.12, 2.28)	1.59 (1.11, 2.28)	1.37 (0.94, 1.98)
T3	3.01 (2.18, 4.14)	2.81 (2.03, 3.89)	2.52 (1.79, 3.55)	1.89 (1.28, 2.78)
Trajectory changes of TyG index
Class 1	1.0	1.0	1.0	1.0
Class 2	1.97 (1.46, 2.65)	1.98 (1.47, 2.67)	1.83 (1.35, 2.49)	1.53 (1.12, 2.11)
Class 3	3.43 (2.47, 4.77)	3.54 (2.54, 4.94)	3.14 (2.24, 4.42)	2.05 (1.36, 3.07)
Trajectory changes of TyG-BMI
Class 1	1.0	1.0	1.0	1.0
Class 2	1.55 (1.14, 2.11)	1.68 (1.23, 2.30)	1.63 (1.19, 2.25)	1.35 (0.97, 1.88)
Class 3	3.15 (2.30, 4.31)	3.60 (2.61, 4.97)	3.22 (2.29, 4.53)	2.27 (1.57, 3.31)
Trajectory changes of TyG-WC
Class 1	1.0	1.0	1.0	1.0
Class 2	2.12 (1.53, 2.95)	2.17 (1.56, 3.02)	2.12 (1.52, 2.97)	1.77 (1.25, 2.52)
Class 3	3.50 (2.50, 4.89)	3.60 (2.57, 5.05)	3.21 (2.24, 4.58)	2.28 (1.54, 3.39)
Trajectory changes of TyG-WHtR
Class 1	1.0	1.0	1.0	1.0
Class 2	1.58 (1.14, 2.19)	1.65 (1.18, 2.30)	1.55 (1.10, 2.17)	1.31 (0.93, 1.86)
Class 3	3.29 (2.39, 4.54)	3.47 (2.47, 4.86)	2.97 (2.08, 4.23)	2.05 (1.39, 3.04)
Trajectory changes of TyG-WWI
Class 1	1.0	1.0	1.0	1.0
Class 2	1.98 (1.43, 2.74)	2.03 (1.46, 2.82)	1.89 (1.35, 2.64)	1.55 (1.10, 2.20)
Class 3	3.35 (2.39, 4.70)	3.37 (2.35, 4.82)	2.91 (2.00, 4.22)	1.95 (1.29, 2.94)
Trajectory changes of TyG-ABSI
Class 1	1.0	1.0	1.0	1.0
Class 2	1.84 (1.35, 2.52)	1.79 (1.30, 2.46)	1.67 (1.21, 2.31)	1.41 (1.01, 1.97)
Class 3	3.39 (2.43, 4.73)	3.15 (2.23, 4.45)	2.78 (1.95, 3.95)	1.85 (1.24, 2.74)
Trajectory changes of TyG-BRI
Class 1	1.0	1.0	1.0	1.0
Class 2	1.71 (1.26, 2.33)	1.80 (1.31, 2.46)	1.73 (1.25, 2.39)	1.49 (1.07, 2.07) 0
Class 3	3.11 (2.27, 4.27)	3.28 (2.34, 4.60)	2.85 (2.00, 4.07)	2.08 (1.43, 3.05)
Trajectory changes of TyG-CVAI
Class 1	1.0	1.0	1.0	1.0
Class 2	1.78 (1.28, 2.48)	1.72 (1.23, 2.40)	1.67 (1.19, 2.35)	1.43 (1.00, 2.05)
Class 3	3.44 (2.46, 4.80)	3.19 (2.28, 4.47)	2.82 (1.98, 4.03)	2.07 (1.37, 3.10)

Abbreviations:HR Hazard ratios, CI Confidence interval, SD Standard deviations, CMM Cardiometabolic multimorbidity; other abbreviations as in Table 1. Model I adjusted for age, sex; Model II adjusted for age, sex, education, living place, kidney disease, hypertension, smoking status, drinking status; Model III adjusted for age, sex, education, living place, kidney disease, hypertension, smoking status, drinking status, LDL-C, HDL-C, Cr, HbA1c, hs-CRP

Predictive value of TyG-Related parameters (baseline values and cumulative exposure) for CMM

The receiver operating characteristic curve analysis showed that TyG-WC, TyG-WHtR, and TyG-CVAI had the highest accuracy in predicting CMM events, including both baseline measurements and cumulative exposure values (Fig. 5). Continuous NRI and IDI analyses showed that, except for TyG-WWI and TyG-ABSI, all other TyG-related parameters provided a certain degree of net improvement compared with the TyG index model. Among these, TyG-WHtR, TyG-WC, TyG-BRI, and TyG-CVAI exhibited the largest increases in NRI and IDI (Supplementary material 2: Table S6).

Fig. 5.

Receiver operating characteristic curve analysis comparing the predictive value of TyG-obesity indices and the TyG index for cardiometabolic multimorbidity. Cum: Cumulative

Incremental predictive performance of TyG-Related parameters for CMM events

Table 4 presents the results of comparisons between different models to evaluate the incremental predictive value of different parameter combinations. Adding the TyG index to the baseline risk model did not improve the C-index. However, the further incorporation of TyG-obesity indices yielded a small but significant 0.01 increase in the C-index. The NRI and IDI (except for TyG-WWI and TyG-ABSI) were also generally statistically significant. In contrast, the TyG-traditional obesity indices consistently enhanced the predictive value of the baseline risk model for CMM.

Table 4.

Evaluating the incremental prognostic value of adding the TyG-related parameters to the baseline risk model for predicting CMM events

Model Comparison	C-index (95% CI)	ΔC-index	Continuous NRI (95% CI)	IDI (95% CI)
Model 1: Baseline risk model	0.70 (0.67–0.73)	-	-	-
Model 2: + TyG index	0.70 (0.67–0.73)	+ 0.00	0.01 (-0.05-0.10)#	0.00 (-0.00-0.01)#
Model 3: + TyG-BMI	0.71 (0.68–0.74)	+ 0.01	0.15 (0.05–0.24)*	0.01 (0.00-0.02)*
Model 4: + TyG-WC	0.71 (0.68–0.74)	+ 0.01	0.19 (0.09–0.27)*	0.01 (0.00-0.03)*
Model 5: + TyG-WHtR	0.71 (0.68–0.74)	+ 0.01	0.14 (0.05–0.24)*	0.01 (0.00-0.02)*
Model 6: + TyG-WWI	0.71 (0.68–0.74)	+ 0.01	0.10 (0.02–0.18)*	0.01 (-0.01-0.02)#
Model 7: + TyG-ABSI	0.71 (0.68–0.74)	+ 0.01	0.07 (-0.01-0.15)#	0.01 (-0.00-0.01)#
Model 8: + TyG-BRI	0.71 (0.68–0.74)	+ 0.01	0.13 (0.05–0.23)*	0.01 (0.00-0.02)*
Model 9: + TyG-CVAI	0.71 (0.68–0.75)	+ 0.01	0.18 (0.09–0.26)*	0.01 (0.00-0.02)*
Model 10: + Cumulative TyG index	0.70 (0.67–0.74)	+ 0.00	0.09 (-0.01-0.17)#	0.00 (-0.00-0.02)#
Model 11: + Cumulative TyG-BMI	0.71 (0.68–0.74)	+ 0.01	0.15 (0.05–0.24)*	0.01 (0.00-0.02)*
Model 12: + Cumulative TyG-WC	0.71 (0.68–0.74)	+ 0.01	0.21 (0.10–0.31)*	0.01 (0.00-0.03)*
Model 13: + Cumulative TyG-WHtR	0.71 (0.68–0.74)	+ 0.01	0.16 (0.08–0.25)*	0.01 (0.00-0.02)*
Model 14: + Cumulative TyG-WWI	0.71 (0.68–0.74)	+ 0.01	0.10 (0.01–0.23)*	0.01 (-0.01-0.02)#
Model 15: + Cumulative TyG-ABSI	0.71 (0.68–0.74)	+ 0.01	0.12 (0.01–0.20)#	0.01 (-0.00-0.02)#
Model 16: + Cumulative TyG-BRI	0.71 (0.68–0.74)	+ 0.01	0.15 (0.05–0.24)*	0.01 (0.00-0.02)*
Model 17: + Cumulative TyG-CVAI	0.71 (0.68–0.75)	+ 0.01	0.18 (0.09–0.28)*	0.01 (0.00-0.02)*

Abbreviations:NRI Net reclassification index, IDI Integrated discrimination improvement; other abbreviations as in Table 1

^#P > 0.05, *P < 0.05

WQS analysis

The WQS regression model was used to quantify the relative contribution of cumulative exposure to each component of the TyG-related parameters to CMM risk. The analysis revealed that in the majority of models, the weights for cumulative FPG and cumulative obesity index were significantly higher than those of other metabolic parameters (Fig. 6), indicating that long-term dysglycemia and obesity are pivotal drivers of CMM pathogenesis. Furthermore, it is worth noting that in the WQS analysis of the components of Cumulative TyG-ABSI, Cumulative FPG had the highest weight, followed by Cumulative TG, and finally Cumulative ABSI. This finding is consistent with the results of association and prediction analyses, suggesting that ABSI may have limited utility as an obesity index in the Chinese population.

Fig. 6.

Relative weights of cumulative obesity index, cumulative FPG and cumulative TG contributing to cardiometabolic multimorbidity risk via weighted quantile sum regression. Cum: Cumulative

Sensitivity analysis

Sensitivity analyses across different subgroups yielded findings consistent with the overall population, indicating that TyG-traditional obesity indices were the superior parameter combination for estimating CMM risk in the vast majority of subgroups (Supplementary material 2: Tables S7-10). In the second sensitivity analysis, A complete-data analysis further confirmed the robustness of the primary findings (Supplementary material 2: Table S11). Thirdly, the results remained consistent when accounting for a lag in event occurrence (Supplementary material 2: Table S12). Finally, the results of the competing risk analysis were consistent with the primary findings presented in Table 2, both in the direction of the risk effects and their statistical significance. This confirms the robustness of our main conclusions and indicates that they were not substantially influenced by competing risks (Supplementary material 2: Table S13).

Discussion

We investigated the associations of the TyG index, TyG-traditional obesity indices, and TyG-novel obesity indices with CMM in the national CHARLS cohort. Compared with the TyG index, most TyG-related parameters, except for TyG-ABSI, demonstrated an improved ability to assess the risk of CMM. Additionally, prediction analyses indicated that, except for TyG index, TyG-ABSI and TyG-WWI, all other TyG-related parameters enhanced the predictive performance for CMM based on the baseline risk model. In contrast, TyG-traditional obesity indices are more practical tools for CMM risk assessment and event prediction.

It is well-established that metabolic dysregulation serves as a primary driver of CMD pathogenesis [11–13, 46]. Against the backdrop of a continuously rising global obesity rate and an aging population, the burden of metabolic disorder-related CMD is exhibiting a rapid upward trend [46, 47]. From a metabolic perspective, obesity and IR exert the most prominent threats to the vascular system. They directly contribute to CMD development by triggering multiple metabolic disorders (e.g., dyslipidemia, glucose dysregulation), and exacerbating associated risk factors [11–13, 48, 49]. More critically, IR and obesity interact synergistically to form a vicious cycle, thereby substantially elevating the risk of CMD incidence [11, 48–51].

From epidemiology to clinical practice, substantial evidence has confirmed that the combined assessment of IR and obesity indices exerts a significant incremental effect on the risk assessment of CMD. The application of such composite indices has effectively optimized the efficacy of early risk stratification and predictive accuracy for CMD [27–30, 52–56], providing a useful tool for advancing precise public health intervention strategies. In recent years, with the progression of global aging, a growing number of researchers have begun to focus on the substantial disease burden of comorbidities, especially CMM [2–4]. Given that the concept of CMM was introduced relatively recently, empirical research on this health issue remains limited. In this study, we comprehensively evaluated the incremental risk assessment and predictive value of the TyG index, TyG-traditional obesity indices, and TyG-novel obesity indices for CMM. Consistent with previous similar studies [31–35], our results confirmed the associations of the TyG index and TyG-traditional obesity indices with CMM. Furthermore, we established that integrating the TyG index with traditional obesity indices is a superior tool for CMM risk assessment. In addition, building on similar studies [31–35], we further expanded the evidence for the application of the TyG index combined with more complex novel obesity parameters in CMM prediction and risk assessment, identifying the optimal TyG-related parameters for CMM assessment through systematic comparisons. Based on the results of association and prediction analyses, our findings indicate that among the TyG-novel obesity indices, TyG-CVAI holds the greatest application potential. However, TyG-WWI and TyG-ABSI demonstrate limited value in the risk assessment and event prediction of CMM. We postulate that the discrepancy in these results may be attributable to the following reasons: (1) The significant incremental effect of visceral adipose accumulation on CMM risk: Recent advances in body composition research consistently indicate that visceral fat accumulation is a more critical health risk factor than subcutaneous fat, with associated metabolic hazards being substantially greater [57–59]. From a pathophysiological perspective, the TyG-CVAI incorporates two core, mutually reinforcing components critical to the pathophysiology of CMM: IR (represented by the TyG index) and visceral obesity (represented by the CVAI) [11, 12]. As a visceral adiposity assessment tool specifically developed for the Chinese population [19], the CVAI, when combined with the TyG index, provides more comprehensive information regarding metabolic health and abnormal fat distribution. In contrast, the ABSI and WWI focus more on evaluating body shape and physique [20, 21]. Although they are also associated with health risks, their ability to reflect visceral fat accumulation and direct metabolic disorders may be less accurate than that of the CVAI [60–62]. (2) The additive effect of risk factors: Methodologically, the CVAI itself is a composite formula that incorporates multiple variables [19], thereby providing richer information content that may have enhanced the discriminatory power of the TyG-CVAI model. In addition, according to further WQS analysis evidence in the current study, the obesity components in TyG-WWI and TyG-ABSI contributed less to the pathogenesis of CMM. This finding suggests that the synergistic effect between the obesity indices (WWI and ABSI) and the TyG index may be relatively weak in middle-aged and elderly Chinese populations. Overall, in terms of computational simplicity, predictive value, and risk assessment value, TyG-traditional obesity indices have significant advantages for primary prevention of CMM.

Given that CMM development is a lengthy dynamic process, relying on a single-timepoint measurement of TyG-related parameters limits accurate risk stratification. Therefore, many studies have focused on their long-term trajectories and cumulative exposure to capture the dynamic accumulation of metabolic disorders over time. This strategy has been proven to significantly optimize the predictive efficacy for CMD [30, 52, 53, 56, 63, 64]. It is worth noting that in several recent studies on CMM, researchers have also evaluated the association between changes in TyG-related parameters and CMM. In an analysis by Zhou et al. based on the Coronary Artery Risk Development in Young Adults study cohort, they found that a persistently increasing trajectory of the TyG index during follow-up was associated with an elevated risk of CMM [65]. Additionally, using the CHARLS cohort, Lv et al. identified that cumulative TyG-WHtR could independently predict CMM risk; individuals with high TyG-WHtR exposure had a 228% increased risk of CMM during follow-up [34]. In the current study, we evaluated the cumulative exposure to TyG-related parameters in the study population during follow-up and further analyzed their control status over this period. Our results showed that, compared with the group with the lowest cumulative exposure or good control during follow-up, the HRs for the groups with the highest cumulative exposure or poor control were as follows: 1.69/2.05 for TyG index, 2.24/2.28 for TyG-WC, 1.92/2.05 for TyG-WHtR, 1.85/2.27 for TyG-BMI, 1.89/2.07 for TyG-CVAI, 1.94/2.08 for TyG-BRI, 1.70/1.85 for TyG-ABSI, and 1.97/1.95 for TyG-WWI. Compared with findings from previous similar studies [34, 65], the current study comprehensively considered the longitudinal changes of all TyG-related parameters and conducted a systematic comparative analysis. In summary, after accounting for the longitudinal changes of TyG-related parameters, TyG-traditional obesity indices remain the most practical combined indices for estimating CMM risk.

In addition to establishing the incremental risk assessment and predictive value of TyG-related parameters for CMM, our analysis of baseline measurements and longitudinal changes revealed some interesting findings. Specifically, the study showed that compared with a single time-point baseline measurement, assessing the cumulative exposure to TyG-related parameters provided a 2–7% incremental value for CMM risk assessment (Fig. 3). Based on this discovery, we have organized several lines of thinking to provide useful references for future related studies: (i) Metabolism exhibits daily fluctuations, seasonal variations, or long-term trends. A single measurement may coincidentally capture an individual’s “trough” or “peak” state, which can potentially lead to a misleading assessment. Additionally, alterations in lifestyle/dietary patterns, other metabolic factors, disease progression, genetic variations, or environmental changes can all exert significant impacts on IR and obesity [66–70]. (ii) The risk of developing many diseases is closely associated with the duration and cumulative dose of risk factors, and a single baseline measurement cannot quantify this crucial dimension. Therefore, longitudinal assessment provides more comprehensive information of participants [71, 72], allows for the quantification of cumulative effects during follow-up, and thus enables a superior estimation of CMM risk. (iii) Population heterogeneity means that risk factors can follow different trajectories (e.g., rapid increase, stability, slow decline) among individuals with the same baseline levels, offering key prognostic insights [73, 74]. For instance, two individuals with the same baseline BMI will have vastly different CMM risks if one remains stable while the other gains weight consistently. Repeated measurements can identify these high-risk trajectories, thereby enabling the precise distinction of these high-risk individuals from the general population with similar baseline levels and enabling risk re-stratification. (iv) Compared with a single measurement, repeated measurements can better correct data bias, effectively reduce research measurement errors, and improve the reliability of study results [75–77]. A study on repeated-measures regression mixed models demonstrated that utilizing repeated measurements significantly enhances model classification accuracy and parameter estimation reliability. Even with limited sample sizes, repeated measurements yield more robust results [78]. In the current study, we identified and corrected anthropometric measurements that fell outside physiologically plausible ranges by utilizing repeated measurement data, thereby effectively mitigating bias arising from measurement errors. (v) A prolonged preclinical phase often precedes the clinical diagnosis of a disease. During this phase, relevant biomarkers may already exhibit subtle yet persistent changes, which can be more effectively captured through repeated measurements [79].

Our WQS regression results reveal an important phenomenon: the development of CMM is collectively driven by a shared metabolic pathway characterized by sustained exposure to chronic hyperglycemia and obesity. In the WQS model, FPG and obesity indices are assigned higher weights, indicating that these components are not merely independent risk factors but more likely manifestations of core physiological dysregulation. This pathway is characterized by a vicious cycle between elevated blood glucose levels and dysfunctional adipose tissue (particularly visceral adipose tissue) [11, 48–51]: Under IR pathological conditions, circulating free fatty acids exhibit abnormally elevated levels, which on one hand stimulate the activation of lipid synthesis pathways, and on the other hand suppress lipid catabolism and mitochondrial oxidative function, leading to lipid metabolic imbalance [80]. Regarding vascular function, IR impairs vasomotor function by affecting nitric oxide bioavailability, resulting in decreased arterial elasticity and increased stiffness, thereby promoting the development of CMD [81, 82]. Under obese conditions, adipose tissue lipolytic activity is enhanced, and the excessive release of free fatty acids inhibits skeletal muscle glucose uptake and impairs insulin signaling transduction in skeletal muscle [50]. Furthermore, obesity-associated adipose tissue macrophage infiltration and inflammatory cytokine secretion not only promote the development of IR [83] but also lead to hyperlipidemia by activating lipolytic enzymes and hepatic lipid synthesis pathways [50, 51, 84]. Consequently, our analysis indicates that CMM risk is driven predominantly by a state of global metabolic dysregulation, rather than by any single biomarker. For clinical practice, these findings underscore the necessity of a comprehensive risk assessment strategy. This necessitates a shift from a focus on individual metrics (such as blood glucose or body weight) toward integrated tools that can simultaneously evaluate glycemic homeostasis and body fat distribution. Accordingly, based on these findings, we propose that individuals with extreme mixed exposure (i.e., those with both severe IR and significant obesity) should undergo comprehensive CMM risk assessment more frequently. At the public health level, we call for interventions targeting the root causes of this shared pathway, such as promoting strategies to reduce the consumption of high-glycemic-index and ultra-processed foods, as these are key environmental factors that jointly drive the epidemics of chronic hyperglycemia and obesity [85–87].

Study strengths and limitations

This study provides the first comparison of the TyG index, TyG-traditional obesity indices, and TyG-novel obesity indices for incremental risk assessment and prediction of CMM. A systematic analysis of these modifiable risk factors reveals that the TyG-traditional obesity indices serve as practical indicators for CMM risk stratification and event prediction. These findings align with the existing evidence chain [31–35, 65], underscoring their considerable external validity and offering important guidance for CMM prevention and management practices. Their key public health value lies in the following: for middle-aged and elderly populations in China, focusing on the TyG-traditional obesity indices enables effective CMM risk stratification across different physical examination frequencies. This offers a feasible strategy to streamline screening and achieve broad risk coverage in resource-limited settings.

Several limitations of the current study should be considered: (1) Although the current analysis estimated the cumulative exposure and control of TyG-related parameters through repeated measurements, the limited number of these measurements precluded us from analyzing their long-term trajectories in relation to CMM. Future studies should aim to incorporate detailed data from more follow-up time points to better clarify the association between more trajectory subtypes and CMM. (2) The assessment of CMM status in this study was primarily based on self-reported information, which is susceptible to recall bias. This could have introduced non-differential misclassification, as the errors in classifying cases and non-cases are likely random, potentially leading to an underestimation of the true effect size [88]. However, according to recent evidence from the re-examination, misreporting of CMD is generally non-systematic, and the potential for classification bias is minimal [89–91]. (3) Due to the broad and complex range of influencing factors for CMM, we were unable to account for all potential confounding factors with the current sample size, which is a common challenge in observational studies. (4) Although patients with diagnosed CMM were excluded at baseline, potential reverse causality introduced by subclinical cases cannot be completely ruled out. This inherent limitation of observational studies suggests that the current findings should be interpreted as proposing a strong associative hypothesis, and the exact causal direction awaits confirmation from future prospective studies targeting earlier stages of the disease. (5) Although the CHARLS study has the advantage of a prospective design, it remains essentially an observational study [36]. Therefore, the results of this study can only reveal statistical associations between variables and cannot establish causality. Furthermore, the lack of an active intervention in this study precludes an assessment of the impact of specific interventions on the outcomes, which leaves room for exploration in future randomized controlled trials or interventional studies. (6) The conclusions of this study are mainly derived from middle-aged and elderly populations in China, and their generalizability to other ethnic groups remains to be confirmed. Therefore, caution is warranted when applying these findings. Furthermore, given the marked trend of younger onset ages for diabetes, stroke, and heart diseases [92–94], future studies urgently need to validate our findings in younger populations to assess the generalizability of the results and facilitate the timely identification of early risks.

Conclusion

The TyG index, TyG-traditional obesity indices, and TyG-novel obesity indices all effectively predict CMM incidence. Given their relative simplicity, the TyG-traditional obesity indices demonstrate superior incremental risk assessment and predictive value for CMM compared to the TyG-novel obesity indices and the TyG index, positioning them as promising and more practical tools for clinical practice.

Abbreviations

TyG

Triglyceride glucose

CMM

Cardiometabolic multimorbidity

CMD

Cardiometabolic diseases; IR: Insulin resistance

CHARLS

China Health and Retirement Longitudinal Study

HbA1c

Haemoglobin A1c

FPG

Fasting plasma glucose

TC

Total cholesterol

TG

Triglycerides

UA

Uric acid

HRs

Hazard ratios

WQS

Weighted quantile sum

Authors’ contributions

CW, YZ and WW: Conceptualization, methodology, and project administration.GB-X: supervision.FZ-Z and YZ: writing-original draft preparation.BZ, XF-H, ZY-W, JW, HH-L, GB-X, WW, and CW: writing-reviewing and editing.GB-X and YZ: Software.YZ, WW and GB-X: formal analysis and validation.YZ and FZ-Z: data curation.All authors read and approved the final manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (81670370, 82360073), the Natural Science Foundation of Jiangxi Province (20232BAB216004, 20224ACB206004), and the Science and Technology Project of Jiangxi Provincial Health Commission (202510148).

Data availability

CHARLS datasets are available for download at the CHARLS home website (http://charls.pku.edu.cn/en).

Declarations

Ethics approval and consent to participate

The CHARLS study was approved by the Ethics Review Committee of Peking University (IRB No.: 00001052–11015). All participants provided voluntary written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.