Association between the newly proposed triglyceride glucose-a body shape index (TyG-ABSI) and atherosclerotic cardiovascular disease in the general population: modest diagnostic improvement compared with traditional TyG-related parameters

Abstract

Background

Atherosclerotic cardiovascular disease (ASCVD) is a leading global cause of mortality, and reliable risk assessment tools are critical for prevention. While TyG-related parameters are used for ASCVD risk stratification, the association between the novel TyG-ABSI and ASCVD in the general population requires further characterization, and its potential improvement in discrimination over traditional TyG parameters has not been fully elucidated.

Methods

We analyzed data from the National Health and Nutrition Examination Survey (NHANES, 1999–2018), a nationally representative cross-sectional study including 22,466 participants. ASCVD was defined by self-reported physician diagnosis of coronary heart disease (CHD), angina pectoris (AP), myocardial infarction (MI), or stroke. TyG-ABSI and traditional TyG parameters were calculated using standardized laboratory and anthropometric data. Statistical analyses included weighted logistic regression (to assess associations), restricted cubic spline analysis (to explore dose–response relationships), subgroup analyses (to test effect modification), receiver operating characteristic (ROC) curves (to evaluate discrimination ability), and net reclassification improvement (NRI) and integrated discrimination improvement (IDI) (to quantify improvement in reclassification and risk separation). Sensitivity analyses were conducted to verify robustness.

Results

In fully adjusted models, TyG-ABSI was significantly associated with ASCVD (odds ratio [OR] per standard deviation [SD] increase: 1.15, 95% confidence interval [CI]: 1.09–1.22) and its subtypes (CHD: OR = 1.14, 95%CI: 1.05–1.24; AP: OR = 1.10, 95%CI: 1.01–1.21; MI: OR = 1.15, 95%CI: 1.06–1.25), with a significant linear dose–response relationship observed for ASCVD and all subtypes (all p-trend < 0.05). TyG-ABSI yielded the highest ROC area under the curve (AUC) across all outcomes (ASCVD: 0.69; CHD: 0.70; AP: 0.69; MI: 0.70; stroke: 0.65). Furthermore, TyG-ABSI demonstrated modest but significant improvement in discrimination over traditional TyG parameters for both ASCVD and its subtypes, as evidenced by significant NRI and IDI metrics. Specifically for ASCVD, it correctly reclassified an additional 7.88% (vs TyG-WHtR) to 16.36% (vs TyG-BMI) of patients based on NRI. The association between TyG-ABSI and ASCVD was robust across demographic, lifestyle, and clinical subgroups, as well as in sensitivity analyses.

Conclusion

This study indicates a significant positive correlation between TyG-ABSI and ASCVD in the general population. Moreover, TyG-ABSI demonstrates modest improvement in discrimination compared to traditional TyG parameters, identifying additional high-risk individuals via reclassification. These findings suggest that TyG-ABSI holds promise as a candidate marker for optimizing ASCVD risk stratification in clinical practice, though future prospective validation is warranted.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-025-03069-w.

Introduction

Atherosclerotic cardiovascular disease (ASCVD), encompassing coronary heart disease, myocardial infarction, angina pectoris, and stroke, remains a leading cause of morbidity and mortality worldwide, imposing a substantial healthcare burden across the globe [1, 2]. The complex pathophysiology of ASCVD involves multiple metabolic disturbances, among which insulin resistance (IR) and central obesity are recognized as key drivers of atherosclerotic progression [3, 4]. Accurate risk stratification is essential for implementing targeted preventive strategies; however, current risk assessment tools have limitations in comprehensively capturing the multifaceted nature of metabolic cardiovascular risk.

The triglyceride-glucose (TyG) index has emerged as a reliable and cost-effective surrogate marker of IR, demonstrating strong predictive value for cardiovascular events across diverse populations [5, 6]. The TyG index, calculated as the logarithm of the product of fasting triglyceride and glucose levels, has shown comparable or even superior performance to conventional IR assessment methods such as the homeostasis model assessment of insulin resistance (HOMA-IR) [7–9]. Subsequent developments have led to the creation of composite indices that combine TyG with anthropometric parameters—such as the TyG–body mass index (TyG-BMI), TyG–waist circumference (TyG-WC), and TyG–waist-to-height ratio (TyG-WHtR)—aimed at capturing both metabolic dysfunction and obesity-related risk [10–13]. However, these traditional TyG-derived indices primarily emphasize general or peripheral obesity and may fail to adequately reflect central obesity, a specific pathogenic factor more closely linked to visceral fat accumulation and cardiometabolic risk.

A Body Shape Index (ABSI) is a novel anthropometric indicator that is constructed by normalizing waist circumference against height and BMI, with a greater focus on reflecting central fat distribution [14]. Compared with traditional anthropometric indices, ABSI has been proposed as a superior marker of central obesity and visceral fat accumulation [15, 16]. A cross-sectional study in an obese population found that ABSI was significantly and positively correlated with cardiovascular disease (CVD) risk scores such as the Framingham Risk Score, confirming its potential utility in CVD risk stratification [17]. Furthermore, in settings where traditional CVD risk factors are unavailable, ABSI outperforms conventional anthropometric measures such as BMI in reflecting cardiovascular risk [18–20]. Building on this, researchers have proposed the triglyceride–glucose–body shape index (TyG-ABSI), which integrates the TyG index—a surrogate marker of IR—with ABSI, an indicator of body shape characteristics linked to central obesity. This composite metric represents a promising approach that simultaneously captures IR and central obesity, two interdependent pathological processes that synergistically promote atherosclerosis [21, 22]. Recent studies have provided preliminary evidence supporting the clinical utility of TyG-ABSI in cardiovascular risk assessment. A large-scale analysis based on the U.S. National Health and Nutrition Examination Survey (NHANES) demonstrated that among individuals with metabolic syndrome, TyG-ABSI showed superior predictive power for both CVD prevalence and mortality compared with traditional TyG-related parameters [21, 23, 24]. Another cross-sectional study further confirmed a synergistic effect between TyG and ABSI on cardiovascular mortality, indicating that the risk associated with concurrent high TyG and high ABSI levels is substantially greater than that of either factor alone [22]. Moreover, emerging evidence suggests that this composite index may enhance the prediction of stroke risk in the early stages of cardio–renal–metabolic syndrome [5].

However, prior studies have largely focused on specific subpopulations with pre-existing metabolic disorders, leaving the association between TyG-ABSI and ASCVD in the general population unclear. In addition, its incremental value over traditional TyG-derived indices (TyG, TyG-BMI, TyG-WC, TyG-WHtR) has yet to be systematically evaluated. Therefore, this study aimed to rigorously evaluate the association between TyG-ABSI and ASCVD in the general population using data from the NHANES database, and to compare its incremental diagnostic value relative to traditional TyG-related parameters. We hypothesized that TyG-ABSI would demonstrate superior discriminatory ability and provide significant incremental value in the diagnosis of ASCVD. By achieving these objectives, our study seeks to contribute to the refinement of ASCVD risk stratification and to inform the development of more effective preventive strategies.

Methods

Data sources and study population

This study utilized data from the NHANES, a nationally representative, cross-sectional survey conducted by the U.S. Centers for Disease Control and Prevention (CDC). The NHANES employs a complex, multistage probability sampling design to assess the health and nutritional status of non-institutionalized U.S. civilians, with data collected via standardized physical examinations, laboratory tests, and structured interviews.

For the present analysis, we included NHANES participants surveyed between 1999 and 2018. The initial study pool comprised 101,316 participants. We applied the following exclusion criteria to ensure data quality and analytical validity:(1) Participants with missing diagnostic information on ASCVD (n = 46,516); (2) Participants with missing data on TyG index components (i.e., fasting triglycerides [TG] or fasting blood glucose [FBG]) (n = 31,311); (3) Participants with missing anthropometric data (height, weight, or waist circumference [WC]) (n = 1,023). After exclusions, a total of 22,466 participants were included in the final analysis. A detailed flowchart of participant inclusion is provided in Fig. 1.

Fig. 1.

Study population screening flowchart. The NHANES data were screened on the basis of the study design to select eligible participants

The NHANES protocol was approved by the CDC’s National Center for Health Statistics (NCHS) Institutional Review Board, and all participants provided written informed consent prior to data collection. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for cross-sectional studies.

Definition of ASCVD

On the basis of the 2013 American College of Cardiology (ACC) and American Heart Association (AHA) guideline on the treatment of blood cholesterol to reduce atherosclerotic cardiovascular risk in adults, we defined ASCVD as the presence of at least one of the following conditions: CHD, angina pectoris (AP), myocardial infarction (MI), or stroke [25, 26]. All diagnostic information on ASCVD was obtained through NHANES’ standardized questionnaire interviews, where trained interviewers asked participants about past medical diagnoses confirmed by a healthcare provider. This method of ASCVD ascertainment is consistent with previous NHANES-based cardiovascular research [27, 28].

Definitions of TyG-related parameters

All TyG-related parameters were calculated using laboratory and anthropometric data collected during NHANES physical examinations. Fasting blood samples (≥ 8 h of fasting) were analyzed by the NCHS-certified laboratory to measure TG (mg/dL) and FBG (mg/dL). Anthropometric measurements (height, weight, WC) were obtained by trained technicians following standardized protocols.

The definitions and formulas for each TyG-related parameter are as follows:

[29];

[30];

[30];

[30];

[30];

[21];

Covariates

On the basis of established literature regarding cardiovascular disease risk factors, we selected and defined the following covariates for adjustment in our statistical models to account for potential confounding.

Demographic and socioeconomic characteristics

The demographic characteristics included in the analysis were age (in years), sex (male or female), and race/ethnicity, which was categorized as non-hispanic white, non-hispanic black, Mexican American, or other races. Educational attainment was classified into three levels: less than high school, high school graduate or General educational development (GED) equivalent, and more than high school [31]. Marital status was grouped as married/living with partner, widowed/divorced/separated, or never married. The poverty-income ratio (PIR), a measure of family income relative to the federal poverty threshold, was categorized into three levels: low (PIR < 1.35), medium (1.35 ≤ PIR < 3.0), and high (PIR ≥ 3.0).

Lifestyle factors

Lifestyle factors included smoking status, alcohol consumption, and physical activity. Smoking status was categorized as never, former, or current smoker [32]. Alcohol consumption was classified into five categories: never, former, mild, moderate, or heavy drinker. The specific criteria for these classifications were based on standard NHANES definitions. Physical activity (PA) was assessed based on the World Health Organization (WHO) analytical guidelines. We calculated the total volume of PA by converting the time spent on activities into metabolic equivalent of task (MET)-minutes per week. The MET score was computed using the formula: MET value × weekly frequency × duration per activity, summing across vigorous and moderate-intensity recreational and work-related activities, with reference MET values provided by NHANES [33].

Comorbid conditions and medication use

Comorbid conditions were defined using a combination of questionnaire data, physical examination, and laboratory tests. Hypertension was defined based on any of the following criteria: (1) a prior physician diagnosis; (2) current use of antihypertensive medications; or (3) an average systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg from two separate readings during the examination [34]. Hyperlipidemia was defined according to the National Cholesterol Education Program (NCEP) Adult Treatment Panel III (ATP III) criteria as meeting any of the following laboratory thresholds: total cholesterol ≥ 200 mg/dL, triglycerides ≥ 150 mg/dL, high-density lipoprotein cholesterol (HDL-C) < 40 mg/dL in men or < 50 mg/dL in women, or low-density lipoprotein cholesterol (LDL-C) ≥ 130 mg/dL [35]. Alternatively, individuals who reported current use of cholesterol-lowering medications were also classified as having hyperlipidemia [36]. Diabetes mellitus (DM) was defined if participants met any of the following criteria: (1) self-reported physician diagnosis of diabetes; (2) current use of insulin or glucose-lowering medications; (3) glycated hemoglobin (HbA1c) ≥ 6.5%; (4) fasting plasma glucose ≥ 7.0 mmol/L; (5) random plasma glucose ≥ 11.1 mmol/L; or (6) 2-h oral glucose tolerance test (OGTT) blood glucose ≥ 11.1 mmol/L [37]. Information on specific medication use was derived from the NHANES prescription medication questionnaire; participants reporting the use of relevant drugs were identified as users of antihypertensive, lipid-lowering, or glucose-lowering medications, respectively.

Statistical analysis

Given NHANES’s complex, multistage probability sampling design, all analyses incorporated the appropriate survey weights, strata (SDMVSTRA), and primary sampling units (SDMVPSU) to obtain nationally representative estimates and valid standard errors. Because our exposure variables (TyG-ABSI and other TyG-related indices) required fasting triglycerides and fasting glucose, we used the fasting subsample laboratory weights as recommended by NCHS/NHANES guidance. Specifically, for the combined 1999–2018 dataset, we constructed a 20-year fasting weight by rescaling the cycle-specific fasting weights: for the 1999–2002 4-year cycle, ; for each 2-year cycle from 2003–2004 through 2017–2018, . All statistical analyses were performed in R 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria), and reporting followed the STROBE statement for observational studies. All statistical analyses were performed using R 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria), adhering to the STROBE statement for reporting observational studies.

First, baseline characteristics were compared between participants with and without ASCVD. For continuous variables (e.g., age, TyG-ABSI), differences between groups were tested using survey-weighted linear regression, and results were presented as median (interquartile range, IQR). For categorical variables (e.g., sex, smoking status), between-group differences were assessed using the survey-weighted chi-square test, with results reported as frequencies and percentages. All P-values for baseline comparisons were two-sided, with a significance level set at P < 0.05.

Second, weighted logistic regression models were used to analyze the association between TyG-ABSI and ASCVD, with odds ratios (ORs) and 95% confidence intervals (CIs) as effect measures. Three sequential models were constructed to control for potential confounders: (1) Crude Model: no covariates adjusted; (2) Model I: adjusted for age (continuous variable) and sex (male/female); (3) Model II (fully adjusted model): further adjusted for race/ethnicity (non-Hispanic White/non-Hispanic Black/Mexican American/other races), educational attainment (less than high school/high school or equivalent/more than high school), marital status (married/living with a partner/widowed/divorced/separated/never married), Poverty Income Ratio (PIR: low [< 1.35]/middle [1.35 ≤ PIR < 3.0]/high [≥ 3.0]), alcohol consumption (never/former/mild/moderate/heavy), smoking status (never/former/current), hypertension (yes/no), and hyperlipidemia (yes/no) on the basis of Model I. To assess potential multicollinearity among covariates in the fully adjusted models, particularly given that TyG-ABSI integrates both metabolic and anthropometric components, we calculated the variance inflation factor (VIF) for all independent variables. A VIF value of less than 5 was considered to indicate the absence of problematic multicollinearity [38]. To minimize sample loss and retain the integrity of original data, dummy variables were created for covariates with missing values for reclassification and included in the regression models. Prior to logistic regression, TyG-ABSI was standardized using z-score to quantify the change in ASCVD risk per 1-standard deviation (SD) increase in TyG-ABSI. Additionally, TyG-ABSI was categorized into quartiles (Q1: lowest, Q4: highest) in all three models, and the Wald test was used to analyze the linear trend of ASCVD risk with increasing TyG-ABSI quartiles.

To evaluate the dose–response relationship between TyG-ABSI and ASCVD, restricted cubic splines (RCSs) with 4 knots were applied in the fully adjusted logistic regression model (Model II). The knots were placed at the 5th, 35th, 65th, and 95th percentiles of TyG-ABSI, in line with standard practices for RCS analysis. A spline curve was plotted to visualize potential non-linear associations, with the reference value set at the median of TyG-ABSI.

Subgroup analyses were conducted to examine potential effect modification in the association between TyG‑ABSI and ASCVD. Participants were stratified by age (< 60/ ≥ 60 years), sex (male/female), race/ethnicity (non-Hispanic White/other), educational attainment (< high school/high school or equivalent/ > high school), marital status (married or living with a partner/widowed, divorced, or separated/never married), poverty income ratio (PIR: low [< 1.35]/middle [1.35 ≤ PIR < 3.0]/high [≥ 3.0]), alcohol consumption (never or former/current), smoking status (never or former/current), hypertension (yes/no), and hyperlipidemia (yes/no). Alcohol consumption and smoking status were dichotomized to improve statistical power for subgroup comparisons. Within each subgroup, logistic regression models were fitted using the same covariate adjustment strategy as Model II, and interaction terms (TyG‑ABSI × subgroup variable) were evaluated using likelihood ratio tests. To account for multiplicity across subgroup/interaction testing, we additionally applied the Benjamini–Hochberg false discovery rate (FDR) procedure and reported the corresponding q-values in Table 3.

Table 3.

Stratified associations between TyG-ABSI and ASCVD by age, sex, ethnicity, education, marital status, PIR, drinking status, smoking status, hypertension and hyperlipidemia

Subgroup	No. of participants	Adjusted OR (95%CI)	P-interaction	q-value
Age			0.810	0.900
< 60 years	11,508	1.22 (1.07, 1.39)
≥ 60 years	10,958	1.20 (1.13, 1.27)
Sex			0.275	0.688
Female	11,606	1.12 (1.03, 1.21)
Male	10,860	1.18 (1.10, 1.28)
Ethnicity			0.199	0.663
Non-Hispanic White	10,046	1.15 (1.07, 1.25)
Other races	12,420	1.08 (1.00, 1.16)
Education			0.407	0.581
Below high school	6045	1.09 (1.00, 1.20)
High school or equivalent	5140	1.17 (1.04, 1.32)
Above high school	11,254	1.19 (1.09, 1.30)
Marital status			0.969	0.969
Married/cohabiting	13,731	1.15 (1.07, 1.24)
Widowed/divorced/separated	4752	1.16 (1.06, 1.27)
Never married	3767	1.18 (0.97, 1.42)
PIR			0.615	0.769
Low	6424	1.10 (1.00, 1.21)
Medium	6294	1.16 (1.04, 1.28)
High	7822	1.18 (1.05, 1.33)
Drinking status			0.085	0.850
Current	6511	1.09 (1.00, 1.19)
Never/past	14,045	1.20 (1.11, 1.30)
Smoking status			0.280	0.560
Current	17,779	1.18 (1.11, 1.26)
Never/past	4667	1.10 (0.99, 1.23)
Hypertension			0.344	0.573
No	13,301	1.20 (1.07, 1.33)
Yes	9157	1.13 (1.06, 1.20)
Hyperlipidemia			0.173	0.865
No	5969	1.29 (1.08, 1.55)
Yes	6497	1.14 (1.07, 1.21)

All abbreviations as in Table 1. Note: (1) Counts are unweighted numbers of participants. (2) Because some stratifying variables have missing/unknown values in NHANES, participants with missing values cannot be assigned to the displayed categories; therefore, subgroup counts may not sum to the overall included sample size (n = 22,466), and denominators may differ across subgroup variables

To assess the diagnostic value of TyG-ABSI for ASCVD, receiver operating characteristic (ROC) curves were generated and the area under the curve (AUC) was calculated. AUCs of TyG-ABSI and traditional TyG-related parameters (TyG, TyG-BMI, TyG-WC, TyG-WHtR) were compared, and the DeLong test was used to test for significant differences between TyG-ABSI and other parameters. To quantify the incremental predictive value of TyG-ABSI over traditional TyG parameters, both the net reclassification improvement (NRI) and the integrated discrimination improvement (IDI) were calculated, along with their 95% confidence intervals (CIs). NRI quantifies the correctness of reclassification, while IDI measures the improvement in the mean separation of predicted risks between cases and non-cases. Finally, decision curves were plotted to compare the clinical utility of TyG-ABSI and traditional TyG parameters by calculating the net benefit across a range of threshold probabilities for ASCVD.

Several sensitivity analyses were conducted to verify the robustness of the results. First, Sensitivity-1 (complete-case analysis) was performed by excluding participants with any missing covariate data, and weighted logistic regression (Model II) was repeated. Second, Sensitivity-2 was conducted to evaluate the impact of additional potential confounders; on the basis of Model II, we further adjusted for physical activity, diabetes status, and medication use (including antihypertensive, glucose-lowering, and lipid-lowering medications). Third, Sensitivity-3 to Sensitivity-7 addressed missing covariate data using multiple imputation. The extent of missingness for each variable is summarized in Supplementary Table 2 (n and %). We generated four imputed datasets using multivariate imputation by chained equations. Analyses were performed on each of the four imputed datasets separately (Sensitivity-3 to Sensitivity-6), and the results were pooled using Rubin’s rules to obtain combined estimates (Sensitivity-7).

To evaluate the potential impact of unmeasured confounding on the observed associations, we calculated E-values for both the point estimates (ORs) and the lower limits of the 95% CIs. The E-value quantifies the minimum strength of association that an unmeasured confounder would need to have with both the exposure and the outcome, conditional on the measured covariates, to explain away the observed exposure–outcome association.

Results

Baseline characteristics of the study participants

Table 1 presents the baseline characteristics of participants stratified by ASCVD. Among the 22,466 participants, 2,186 (9.73%) had ASCVD. As expected, there were significant differences in baseline characteristics between the ASCVD and non-ASCVD groups. Specifically, participants with ASCVD had higher values for key anthropometric and metabolic indicators: compared with the non-ASCVD group, they had a greater age (64.41 vs. 45.62 years), higher weight (84.47 vs. 81.92 kg), higher BMI (29.87 vs. 28.57 kg/m²), larger WC (104.87 vs. 97.84 cm), higher TyG index (8.88 vs. 8.61), and higher TyG-ABSI (0.75 vs. 0.70). In contrast, the ASCVD group had a lower height (167.80 vs. 169.07 cm), and there was also a significant difference in ABSI between the two groups.

Table 1.

Baseline Characteristics of Participants Stratified by Presence of ASCVD

	Non-ASCVD	ASCVD	P-value
Participants	20,280	2186
Age, years	45.62 (45.20, 46.04)	64.41 (63.66, 65.16)	< 0.001
Weight, kg	81.92 (81.46, 82.38)	84.47 (83.23, 85.71)	< 0.001
Height, cm	169.07 (168.88, 169.25)	167.8 (167.26, 168.33)	< 0.001
BMI, kg/m2	28.57 (28.42, 28.73)	29.87 (29.48, 30.27)	< 0.001
WC, cm	97.84 (97.44, 98.24)	104.87 (103.9, 105.83)	< 0.001
ABSI	0.08 (0.08, 0.08)	0.08 (0.08, 0.08)	< 0.001
TyG	8.61 (8.59, 8.62)	8.88 (8.84, 8.93)	< 0.001
TyG-ABSI	0.70 (0.70, 0.70)	0.75 (0.75, 0.76)	< 0.001
Sex			< 0.001
Female	10,682 (52.67%)	924 (42.27%)
Male	9598 (47.33%)	1262 (57.73%)
Ethnicity			< 0.001
Non-Hispanic White	8827 (43.53%)	1219 (55.76%)
Non-Hispanic Black	4001 (19.73%)	430 (19.67%)
Mexican American	3781 (18.64%)	255 (11.67%)
Other races	3671 (18.10%)	282 (12.90%)
Education			< 0.001
Below high school	5256 (25.94%)	789 (36.19%)
High school or equivalent	4628 (22.84%)	512 (23.49%)
Above high school	10,375 (51.21%)	879 (40.32%)
Marital status			< 0.001
Married/cohabiting	12,440 (61.96%)	1291 (59.41%)
Widowed/divorced/separated	3995 (19.90%)	757 (34.84%)
Never married	3642 (18.14%)	125 (5.75%)
PIR			< 0.001
Low	5675 (30.61%)	749 (37.43%)
Medium	5614 (30.28%)	680 (33.98%)
High	7250 (39.11%)	572 (28.59%)
Smoking status			< 0.001
Never	11,268 (55.62%)	828 (37.88%)
Former	4803 (23.71%)	880 (40.26%)
Now	4189 (20.68%)	478 (21.87%)
Drinking status			< 0.001
Never	2615 (14.07%)	273 (13.82%)
Former	2988 (16.08%)	635 (32.14%)
Mild	6249 (33.63%)	683 (34.56%)
Moderate	2839 (15.28%)	181 (9.16%)
Heavy	3889 (20.93%)	204 (10.32%)
Hypertension	7490 (36.95%)	1667 (76.26%)	< 0.001
Hyperlipidemia	14,556 (71.78%)	1941 (88.79%)	< 0.001

ASCVD: Atherosclerotic cardiovascular disease; BMI: Body mass index; WC: Waist circumference; ABSI: A body shape index; TyG-ABSI: Triglyceride-Glucose-a body shape index; PIR: Poverty income ratio

Regarding demographic and lifestyle factors, the ASCVD group had a higher proportion of males (57.73% vs. 47.33%) and non-Hispanic White individuals (55.76% vs. 43.53%), while a larger share of them had educational attainment below high school (36.19% vs. 25.94%) and were in the widowed/divorced/separated marital status (34.84% vs. 19.90%). They also had a higher proportion of participants with low-to-middle Poverty Income Ratio (PIR: 37.43% for low PIR + 33.98% for middle PIR vs. 30.61% + 30.28% in the non-ASCVD group). For lifestyle behaviors, the ASCVD group had a higher proportion of former smokers (40.26% vs. 23.71%) and former drinkers (32.14% vs. 16.08%), but lower proportions of never smokers (37.88% vs. 55.62%), moderate drinkers (9.16% vs. 15.28%), and heavy drinkers (10.32% vs. 20.93%). Clinically, the prevalence of comorbidities was markedly higher in the ASCVD group: 76.26% of participants with ASCVD had hypertension (vs. 36.95% in the non-ASCVD group), and 88.79% had hyperlipidemia (vs. 71.78% in the non-ASCVD group).

Association between TyG-ABSI and ASCVD

Before interpreting the regression estimates, we evaluated multicollinearity in the fully adjusted models. The results showed that all VIF values for the covariates were less than 5, indicating that there was no substantial multicollinearity among the variables included in the models (Supplementary Table 1). Table 2 presents the weighted logistic regression results for the association between TyG-ABSI and ASCVD and its subtypes (CHD, AP, MI, and stroke) across three models with increasing covariate adjustment. For overall ASCVD, each 1-SD increase in TyG-ABSI was significantly associated with a higher risk in the Crude Model (OR = 1.85, 95%CI:1.77–1.93), which remained significant but attenuated after adjusting for age and sex (Model I: OR = 1.32, 95%CI:1.26–1.40) and further adjustment for ethnicity, education, marital status, PIR, drinking, smoking, hypertension, and hyperlipidemia (Model II: OR = 1.15, 95%CI:1.09–1.22). A clear dose–response relationship was observed: compared with the lowest quartile (Q1), the highest quartile (Q4) of TyG-ABSI was associated with significantly elevated ASCVD risk in all models (Model II: OR = 1.56, 95%CI:1.28–1.91), with all p-trend < 0.01.

Table 2.

Logistic regression analyses for the association between TyG-ABSI and ASCVD in different models

Exposure	OR (95%CI)
	Crude Model	Model I	Model II
ASCVD
TyG-ABSI (per SD)	1.85 (1.77, 1.93)	1.32 (1.26, 1.40)	1.15 (1.09, 1.22)
TyG-ABSI (quartiles)
Q1	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
Q2	2.66 (2.20, 3.21)	1.46 (1.20, 1.78)	1.25 (1.02, 1.53)
Q3	4.33 (3.62, 5.17)	1.69 (1.40, 2.04)	1.32 (1.08, 1.61)
Q4	7.50 (6.31, 8.91)	2.29 (1.90, 2.75)	1.56 (1.28, 1.91)
p-trend	< 0.01	< 0.01	< 0.01
CHD
TyG-ABSI (per SD)	1.91 (1.79, 2.03)	1.35 (1.26, 1.46)	1.14 (1.05, 1.24)
TyG-ABSI (quartiles)
Q1	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
Q2	2.71 (1.97, 3.72)	1.32 (0.95, 1.83)	1.03 (0.74, 1.44)
Q3	4.98 (3.70, 6.72)	1.63 (1.20, 2.23)	1.07 (0.78, 1.48)
Q4	9.43 (7.08, 12.57)	2.35 (1.73, 3.17)	1.33 (0.97, 1.83)
p-trend	< 0.01	< 0.01	< 0.01
AP
TyG-ABSI (per SD)	1.77 (1.65, 1.91)	1.33 (1.22, 1.45)	1.10 (1.01, 1.21)
TyG-ABSI (quartiles)
Q1	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
Q2	3.10 (2.09, 4.60)	1.81 (1.21, 2.70)	1.42 (0.95, 2.13)
Q3	6.06 (4.18, 8.78)	2.60 (1.78, 3.81)	1.76 (1.19, 2.60)
Q4	9.31 (6.48, 13.36)	3.16 (2.16, 4.60)	1.82 (1.23, 2.70)
p-trend	< 0.01	< 0.01	< 0.01
MI
TyG-ABSI (per SD)	1.85 (1.74, 1.96)	1.34 (1.25, 1.45)	1.15 (1.06, 1.25)
TyG-ABSI (quartiles)
Q1	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
Q2	3.00 (2.21, 4.09)	1.62 (1.18, 2.21)	1.36 (0.98, 1.87)
Q3	5.27 (3.93, 7.05)	1.99 (1.47, 2.69)	1.51 (1.10, 2.06)
Q4	9.33 (7.04, 12.37)	2.74 (2.04, 3.68)	1.79 (1.31, 2.45)
p-trend	< 0.01	< 0.01	< 0.01
Stroke
TyG-ABSI (per SD)	1.60 (1.50, 1.71)	1.18 (1.09, 1.27)	1.07 (0.98, 1.16)
TyG-ABSI (quartiles)
Q1	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
Q2	2.34 (1.78, 3.09)	1.38 (1.04, 1.84)	1.25 (0.94, 1.68)
Q3	3.15 (2.42, 4.12)	1.38 (1.04, 1.82)	1.20 (0.90, 1.61)
Q4	4.76 (3.68, 6.15)	1.64 (1.25, 2.16)	1.28 (0.95, 1.72)
p-trend	< 0.01	< 0.01	0.221

OR: Odds ratios; SD: standard deviation. CHD: Coronary heart disease; AP: Angina pectoris; MI: Myocardial infarction

Crude Model adjust for: None

Model I model adjust for: Age and Sex

Model II model adjust for: Age, Sex, ethnicity, education, marital status, PIR, drinking status, smoking status, hypertension and hyperlipidemia

For ASCVD subtypes: AP and MI retained significant associations in the fully adjusted model (1-SD increase: AP OR = 1.10, 95%CI:1.01–1.21; MI OR = 1.15, 95%CI:1.06–1.25; Q4 vs Q1: AP OR = 1.82, 95%CI:1.23–2.70; MI OR = 1.79, 95%CI:1.31–2.45). CHD showed a significant per-SD association in Model II (OR = 1.14, 95%CI:1.05–1.24), but the Q4 vs Q1 comparison was non-significant (OR = 1.33, 95%CI:0.97–1.83). Stroke associations were attenuated to non-significance after full adjustment (1-SD: OR = 1.07, 95%CI:0.98–1.16; Q4: OR = 1.28, 95%CI:0.95–1.72), with p-trend = 0.22.

Dose‒response relationships between TyG-ABSI and ASCVD

Figure 2 illustrates the dose–response relationship between TyG-ABSI and ASCVD and its subtypes (CHD, MI, AP, stroke) using RCS analysis, adjusted for all covariates in Model II. For CHD, the RCS plot showed a significant positive association (P-overall = 0.003) with a linear trend (P-non-linear = 0.655), indicating that the OR increased continuously as TyG-ABSI elevated. Similarly, MI exhibited a significant positive linear association (P-overall = 0.010, P-non-linear = 0.156), with ORs rising steadily with increasing TyG-ABSI. For overall ASCVD, the RCS analysis demonstrated a significant positive linear relationship (P-overall = 0.002, P-non-linear = 0.958), confirming a consistent increase in ASCVD risk with higher TyG-ABSI. Angina pectoris showed a significant association (P-overall = 0.021) with a predominantly linear trend (P-non-linear = 0.115), though the curve exhibited a slight plateau after initial elevation. In contrast, stroke had no significant dose–response relationship (P-overall = 0.463, P-non-linear = 0.291), as neither linear nor non-linear associations reached statistical significance.

Fig. 2.

Dose‒response relationships between the TyGABSI and ASCVD. All abbreviations are listed in Table 1. Restricted cubic splines were adjusted for Age, Sex, ethnicity, education, marital status, PIR, drinking status, smoking status, hypertension and hyperlipidemia

Subgroup analysis

Table 3 delineates the stratified associations between TyG-ABSI and ASCVD across various subgroups, with all models adjusted for the same covariates as Model II. The positive association between TyG-ABSI and ASCVD was generally consistent across strata. Notably, we found no evidence of significant effect modification by any stratifying variable (all P-interaction > 0.05). This lack of interaction was further confirmed after controlling for the false discovery rate, as all q-values were > 0.05. Specifically, significant positive associations were observed in most subgroups, including different age groups (ORs 1.20–1.22), sexes (ORs 1.12–1.18), and ethnicities (non-Hispanic White OR = 1.15; Other races OR = 1.08 [borderline]). While positive trends were maintained, statistical significance was not reached in some smaller subgroups (e.g., ‘Never married’ or ‘Never/past smokers’), likely due to reduced sample sizes. Regarding alcohol consumption, although the raw P-interaction was 0.085, the FDR-adjusted q-value was 0.850, indicating no robust difference between current and non-drinkers. Collectively, these findings underscore that the positive association between TyG-ABSI and ASCVD is robust across diverse demographic, lifestyle, and clinical profiles.

Comparison of the diagnostic value of the TyG-ABSI and TyG-related parameters for ASCVD

Table 4 and Fig. 3 present the ROC analysis and incremental value metrics comparing TyG-ABSI with conventional TyG-related parameters (TyG, TyG-WC, TyG-BMI, TyG-WHtR) for diagnosing ASCVD and its subtypes. For overall ASCVD, TyG-ABSI demonstrated the highest AUC (0.69, 95%CI: 0.68–0.70), significantly outperforming TyG (0.61), TyG-WC (0.64), TyG-BMI (0.58), and TyG-WHtR (0.64) (all DeLong p < 0.001). Beyond AUC, TyG-ABSI showed significant incremental value: it yielded significant NRI ranging from 7.88% (vs TyG-WHtR) to 16.36% (vs TyG-BMI), and significant IDI ranging from 1.52% (vs TyG-WHtR) to 2.95% (vs TyG-BMI) (all p < 0.001). This indicates that adding TyG-ABSI not only improved classification accuracy but also significantly increased the separation of predicted risks.

Table 4.

ROC analysis comparing TyG-ABSI and conventional TyG-related parameters for the diagnosis of ASCVD

Test	AUC (95%CI)	Best threshold	Specificity	Sensitivity	IDI (95%CI), %	NRI (95%CI), %	Delong-P value
ASCVD
TyG-ABSI	0.69 (0.68, 0.70)	0.72	0.59	0.69	–	–	–
TyG	0.61 (0.60, 0.62)	8.56	0.48	0.68	2.39 (2.21, 2.58)	12.25 (10.37, 14.14)	< 0.001
TyG-WC	0.64 (0.63, 0.65)	887.78	0.61	0.59	1.62 (1.38, 1.87)	8.38 (6.09, 10.67)	< 0.001
TyG-BMI	0.58 (0.57, 0.59)	227.30	0.41	0.71	2.95 (2.67, 3.24)	16.36 (13.86, 18.85)	< 0.001
TyG-WHtR	0.64 (0.63, 0.65)	5.13	0.54	0.66	1.52 (1/27, 1.77)	7.88 (5.67, 10.10)	< 0.001
CHD
TyG-ABSI	0.70 (0.69, 0.72)	0.72	0.58	0.74	–	–	–
TyG	0.62 (0.60, 0.64)	8.64	0.52	0.66	1.29 (1.14, 1.45)	13.32 (10.51, 16.14)	< 0.001
TyG-WC	0.64 (0.62, 0.66)	891.88	0.61	0.59	1.07 (0.87, 1.27)	11.03 (7.83, 14.23)	< 0.001
TyG-BMI	0.57 (0.55, 0.59)	216.32	0.33	0.78	1.76 (1.53, 1.99)	19.89 (16.52, 23.25)	< 0.001
TyG-WHtR	0.63 (0.61, 0.65)	4.84	0.42	0.77	1.20 (1.00, 1.39)	11.68 (8.73, 14.62)	< 0.001
AP
TyG-ABSI	0.69 (0.67, 0.71)	0.71	0.53	0.77	–	–	–
TyG	0.63 (0.61, 0.65)	8.56	0.47	0.75	0.56 (0.46, 0.67)	8.22 (4.82, 11.62)	< 0.001
TyG-WC	0.65 (0.63, 0.67)	888.22	0.60	0.64	0.31 (0.16, 0.46)	5.73 (1.48, 9.98)	< 0.001
TyG-BMI	0.60 (0.58, 0.62)	245.42	0.53	0.64	0.71 (0.54, 0.87)	13.16 (8.60, 17.72)	< 0.001
TyG-WHtR	0.66 (0.64, 0.68)	5.18	0.55	0.71	0.29 (0.14, 0.44)	4.96 (1.02, 8.90)	< 0.001
MI
TyG-ABSI	0.70 (0.68, 0.71)	0.72	0.57	0.74	–	–	–
TyG	0.61 (0.59, 0.63)	8.69	0.56	0.60	1.20 (1.06, 1.34)	15.28 (12.44, 18.12)	< 0.001
TyG-WC	0.64 (0.62, 0.66)	884.32	0.60	0.61	0.85 (0.67, 1.04)	10.73 (7.45, 14.02)	< 0.001
TyG-BMI	0.57 (0.55, 0.59)	215.28	0.32	0.79	1.53 (1.32, 1.74)	19.92 (16.61, 23.23)	< 0.001
TyG-WHtR	0.64 (0.62, 0.65)	5.15	0.54	0.66	0.95 (0.76, 1.13)	10.57 (7.34, 13.80)	< 0.001
Stroke
TyG-ABSI	0.65 (0.63, 0.66)	0.72	0.57	0.64	–	–	–
TyG	0.57 (0.55, 0.59)	8.66	0.53	0.57	0.64 (0.54, 0.74)	10.85 (7.62, 14.07)	< 0.001
TyG-WC	0.61 (0.59, 0.62)	810.52	0.42	0.74	0.41 (0.28, 0.53)	4.82 (1.37, 8.27)	< 0.001
TyG-BMI	0.56 (0.54, 0.58)	227.59	0.41	0.70	0.70 (0.55, 0.84)	10.54 (6.41, 14.68)	< 0.001
TyG-WHtR	0.62 (0.60, 0.64)	4.97	0.47	0.71	0.25 (0.12,0.39)	3.26 (0.00, 6.79)	0.005

Delong’s test p-value, less than 0.05 was considered significantly different

Fig. 3.

ROC curves of the TyG-ABSI in the diagnosis of ASCVD. The optimal threshold for the ROC curve is determined by the Youden index. Abbreviations ROC: Receiver operating characteristic. Other abbreviations are listed in Table 1

Similar superiority was observed for all ASCVD subtypes. For CHD, TyG-ABSI (AUC = 0.70) outperformed all comparators (AUCs: 0.57–0.64), with NRI from 11.03% to 19.89% and IDI from 1.07% to 1.76% (all p < 0.001). For AP, TyG-ABSI (AUC = 0.69) surpassed others (AUCs: 0.60–0.66), showing NRI from 4.96% to 13.16% and IDI from 0.29% to 0.71% (all p < 0.001). For MI, TyG-ABSI achieved an AUC of 0.70 (vs 0.57–0.64), with NRI from 10.57% to 19.92% and IDI from 0.85% to 1.53% (all p < 0.001). For stroke, despite a lower overall AUC (0.65), TyG-ABSI still significantly outperformed traditional parameters (AUCs: 0.56–0.62), providing NRI from 3.26% (vs TyG-WHtR, p = 0.005) to 10.85% (vs TyG) and IDI from 0.25% (vs TyG-WHtR) to 0.70% (vs TyG-BMI) (all p < 0.01). Figure 3 visually confirms these findings, showing the ROC curve for TyG-ABSI consistently above those of traditional indices.

Figure 4 presents the decision curve analyses. For ASCVD and all subtypes, TyG-ABSI consistently demonstrated a higher net benefit than conventional TyG parameters across a wide range of clinically relevant threshold probabilities, suggesting superior clinical utility for identifying high-risk individuals.

Fig. 4.

Decision curve for comparing TyG-ABSI and TyG-related parameters. Abbreviations are listed in Table 1

Sensitivity analysis

Table 5 presents the results of sensitivity analyses to assess the robustness of the association between TyG-ABSI and ASCVD. Across all sensitivity scenarios, the positive association remained consistent and statistically significant. In analyses addressing missing data (Sensitivity-1 and Sensitivity-3 to Sensitivity-7), the ORs for per-SD increase in TyG-ABSI remained stable, ranging from 1.15 (95% CI: 1.08–1.22) in the complete-case analysis to 1.16 (95% CI: 1.09–1.23) in the multiple imputation analyses. When further adjusted for physical activity, diabetes status, and medication use (Sensitivity-2), the association was slightly attenuated (OR per SD = 1.10, 95% CI: 1.01–1.20) but remained statistically significant. Consistently, categorization of TyG-ABSI into quartiles showed that the highest quartile (Q4) was associated with a significantly elevated risk compared to the lowest quartile (Q1) across all scenarios (ORs ranging from 1.42 to 1.57), with all linear trends remaining significant (p-trend < 0.05). These results indicate that the positive association between TyG-ABSI and ASCVD is robust to different approaches for handling missing data and additional confounding control.

Table 5.

Sensitivity analysis of the association between TyG-ABSI and ASCVD

OR (95% CI)
	Sensitivity‐1	Sensitivity‐2	Sensitivity‐3	Sensitivity‐4	Sensitivity‐5	Sensitivity‐6	Sensitivity‐7
TyG-ABSI (per SD)	1.15 (1.08, 1.22)	1.10 (1.01, 1.20)	1.16 (1.09, 1.22)	1.16 (1.09, 1.22)	1.16 (1.09, 1.22)	1.16 (1.09, 1.23)	1.15 (1.08, 1.22)
TyG-ABSI (quartiles)
Q1	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)	1.0 (Ref)
Q2	1.33 (1.06, 1.66)	1.35 (1.05, 1.73)	1.25 (1.02, 1.53)	1.24 (1.02, 1.52)	1.25 (1.02, 1.53)	1.25 (1.02, 1.53)	1.33 (1.06, 1.66)
Q3	1.32 (1.05, 1.64)	1.26 (0.98, 1.63)	1.33 (1.09, 1.62)	1.32 (1.08, 1.61)	1.32 (1.08, 1.62)	1.33 (1.09, 1.62)	1.32 (1.05, 1.64)
Q4	1.56 (1.25, 1.95)	1.42 (1.09, 1.84)	1.57 (1.28, 1.91)	1.56 (1.28, 1.91)	1.57 (1.28, 1.91)	1.57 (1.29, 1.92)	1.56 (1.25, 1.95)
p-trend	< 0.001	0.047	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001

(1) Sensitivity‐1: Repeated analyses were conducted after deleting the missing covariates, adjusting for all variables in Model II presented in Table 2; (2) Sensitivity‐2: On the basis of Model II, further adjustments were made for physical activity, diabetes status and the use of antihypertensive, lipid-lowering, and hypoglycemic medications. (3) Sensitivity‐3 to Sensitivity‐6 were repeated analyses using data after multiple imputations of missing covariates; (4) Sensitivity‐7: Results of Rubin’s rules-based merging of four sets of multiple imputation data

We calculated E-values to assess the robustness of the associations to potential unmeasured confounding (Supplementary Table 3). For overall ASCVD, the E-value for the point estimate was 1.56, with a lower CI limit E-value of 1.40. This implies that an unmeasured confounder would need to be associated with both TyG-ABSI and the outcome by an odds ratio of at least 1.56 to explain away the observed effect, whereas weaker confounding could not. Similarly, the E-values for CHD (1.54), angina pectoris (1.43), and MI (1.57) suggested reasonable robustness. For stroke, although the point estimate E-value was 1.34, the lower CI limit E-value was 1.00, consistent with the non-significant association observed in the primary analysis.

Discussion

In this nationally representative cross-sectional study based on NHANES data, we were the first to identify that elevated TyG-ABSI levels were significantly associated with an increased risk of ASCVD and its subtypes, including coronary heart disease, angina pectoris, myocardial infarction, and stroke, in the general population. This association was independent of traditional cardiovascular risk factors. Even after comprehensive adjustment for demographic, socioeconomic, and metabolic confounders, the relationship remained robust. Moreover, restricted cubic spline analysis revealed a significant linear dose–response relationship between TyG-ABSI and ASCVD risk, further strengthening the reliability of this association. The stratified analyses demonstrated consistent results across various demographic and clinical subgroups, with no significant interaction effects observed. Overall, these findings suggest that TyG-ABSI may serve as a comprehensive indicator reflecting both metabolic status and adiposity distribution, offering superior predictive value for ASCVD risk assessment and stratification compared with traditional TyG-related parameters.

Growing evidence suggests that IR and related disorders contribute to the development of CVD in both diabetic and non-diabetic individuals [39]. It is well established that individuals with IR are prone to a range of metabolic abnormalities, including hyperglycemia, dyslipidemia, and hypertension—all of which are strongly associated with adverse cardiovascular outcomes [40]. Therefore, IR is recognized not only as a causal factor but also as a powerful predictor of CVD in both the general population and diabetic patients. As a reliable and easily obtainable surrogate marker of IR, TyG index has been increasingly linked to the development and adverse outcomes of CVD [41–43]. However, most studies have assessed the TyG index alone without incorporating anthropometric parameters [44], which may compromise predictive accuracy. Obesity—particularly central obesity—is strongly associated with increased IR, activation of systemic inflammatory pathways, and heightened oxidative stress [45, 46], all of which substantially elevate the risk of CVD and overall mortality [47]. Consequently, recent research has focused on integrating the TyG index with obesity-related indicators such as BMI, waist circumference, and waist-to-height ratio to develop composite indices, especially for assessing CVD and mortality risks [10, 48]. Compared with traditional obesity-related metrics, ABSI stands out as the only anthropometric parameter that is not affected by the “obesity paradox” [47, 49–51], and it demonstrates superior predictive performance. A meta-analysis reported that each 1-SD increase in ABSI was associated with a 13% higher risk of hypertension, a 35% higher risk of type 2 diabetes, and a 21% higher risk of CVD [52]. Notably, previous studies have shown limitations when evaluating TyG or ABSI independently, whereas combining these two indices exhibited a synergistic effect and provided a more accurate stratification of cardiovascular risk [22]. Consistent with these findings, the present study demonstrated that TyG-ABSI outperformed the TyG index and other TyG-derived indices in predicting ASCVD risk in the general population, suggesting that the combined evaluation of TyG and ABSI offers unique advantages over TyG-BMI, TyG-WC, and TyG-WHtR. Furthermore, the association between TyG-ABSI and ASCVD risk remained significant across subgroups stratified by sex, age, and metabolic status, highlighting its robustness and generalizability.

Notably, our findings regarding the association between TyG‑ABSI and stroke differ from several prior reports in Chinese populations, which generally suggested a positive relationship between TyG‑ABSI and stroke risk in both the general population and CKM-related populations [5, 53, 54]. In contrast, in this NHANES-based analysis of the general U.S. population, TyG-ABSI did not exhibit a statistically significant association with stroke in the fully adjusted model (OR = 1.07, 95% CI: 0.98–1.16), and the dose–response relationship was not statistically significant in restricted cubic spline analyses (p = 0.22), indicating a weaker and less consistent pattern than that observed for ASCVD overall and other ASCVD subtypes. Several factors may plausibly explain this discrepancy. First, stroke is a highly heterogeneous clinical entity, encompassing both ischemic and hemorrhagic subtypes that differ substantially in their underlying pathophysiology [55, 56]. This heterogeneity, combined with the self-reported definition of stroke in NHANES, may have attenuated the observed association. Second, the relatively low prevalence of stroke in the general population, along with the cross-sectional design of NHANES, may have limited the statistical power to detect a significant relationship. Third, stroke risk is affected by multiple non-metabolic factors, such as cardioembolic sources and blood pressure variability, which could reduce the independent contribution of metabolic indices after multivariable adjustment. Nevertheless, it is important to emphasize that despite the lack of a significant independent association in the regression models, TyG-ABSI still demonstrated acceptable discriminatory ability (AUC = 0.65) and significant incremental predictive value (via NRI) over traditional TyG parameters. Therefore, while this finding suggests TyG-ABSI is less robust for stroke compared to CHD or MI, it does not entirely negate its utility for risk classification. These nuances highlight the need to consider specific pathological and statistical characteristics when interpreting the association with stroke, and future prospective studies with adjudicated stroke subtypes are warranted.

From a pathophysiological perspective, the strong association between TyG‑ABSI and ASCVD risk may reflect the synergistic effects of IR and central obesity in the development and progression of atherosclerosis [57]. Unlike TyG or ABSI alone, TyG‑ABSI, as a composite index, integrates metabolic dysfunction (TyG) with central and visceral adiposity (ABSI) at the individual level, thereby providing a more comprehensive representation of their interactive effects on the cardiometabolic environment [24]. Specifically, TyG primarily reflects hepatic and systemic IR, which can promote atherogenic dyslipidemia, endothelial dysfunction, and a prothrombotic state [58–61]. In contrast, ABSI is more sensitive to central and visceral fat accumulation and adverse body shape, closely associated with abnormal adipokine secretion, low-grade inflammation, and perivascular fat dysfunction [62, 63]. In the coexistence of obesity and IR, elevated plasma insulin and aldosterone levels may activate endothelial mineralocorticoid receptors, enhancing sodium influx through epithelial sodium channels, which in turn induces cytoskeletal stiffening and drives vascular and cardiac fibrosis and structural remodeling [64–66]. Furthermore, endothelial stiffening is accompanied by reduced endothelial nitric oxide synthase activity, decreased nitric oxide production, and impaired NO bioavailability, leading to vascular stiffness and diastolic dysfunction [67–69]. Based on these mechanisms, we propose a working hypothesis: high TyG‑ABSI may represent a “highly adverse phenotype” characterized by pronounced IR and maladaptive fat distribution, corresponding to so-called “metabolically unhealthy central obesity.” This phenotype may identify high-risk individuals who cannot be fully captured by TyG combined with BMI, waist circumference, or waist-to-height ratio alone, potentially explaining the stronger association and improved reclassification performance of TyG‑ABSI observed in our study. It should be noted, however, that our study is cross-sectional and NHANES lacks detailed imaging and biomarker data, precluding direct elucidation of the specific biological pathways linking TyG-ABSI to ASCVD. The mechanistic hypothesis integrating IR and central obesity remains exploratory and warrants further validation in mechanistic and prospective studies incorporating ectopic fat imaging, inflammatory mediators, and vascular function assessments.

In terms of clinical applicability, TyG-ABSI is calculated from routine fasting triglyceride and glucose measurements combined with simple anthropometric parameters, making it both easy to compute and cost-effective, and thereby offering substantial translational potential. Based on our findings, TyG-ABSI may be applied in several scenarios. First, it could serve as an initial screening tool in primary care or community settings to identify individuals in the general population at elevated risk of ASCVD who may require further evaluation. Second, among metabolically high-risk populations, such as those with insulin resistance or obesity-related phenotypes, it could be used for risk stratification to guide the prioritization of lifestyle interventions, metabolic management, and follow-up intensity. Third, it may provide a reference for further testing or referral, indicating the need for more comprehensive cardiovascular risk assessment and complementing existing risk evaluation strategies. Importantly, TyG-ABSI should primarily be considered as a risk marker and stratification tool, with its value residing in the identification of high-risk phenotypes rather than replacing imaging studies or the gold-standard clinical diagnosis of ASCVD.

Despite the large sample size and strong population representativeness of this study, several limitations should be acknowledged. First, due to its cross-sectional design, causal inferences between TyG-ABSI and ASCVD cannot be established. Future prospective cohort studies are needed to confirm the longitudinal association between TyG-ABSI and the development of ASCVD. Second, the present analysis was based on single-time measurements, which may be subject to short-term metabolic fluctuations. Third, although multiple potential confounders were carefully adjusted for, the possibility of residual confounding cannot be completely excluded. Fourth, ASCVD outcomes were determined based on self-reported information and medical history, which may introduce recall or misclassification bias. Finally, this study was conducted based on the US NHANES population, and external validation in independent cohorts, other ethnic groups, and clinical populations is still lacking. Therefore, TyG‑ABSI serves as a risk stratification marker and cannot currently replace gold-standard diagnostic assessments for ASCVD; before it is adopted as a routine tool, its optimal thresholds and integration into clinical decision-making pathways require rigorous validation in independent external cohorts and prospective studies. Future investigations should further explore the temporal dynamics between TyG-ABSI and various cardiovascular events—such as coronary artery calcification, myocardial infarction, and stroke—and integrate imaging and biomarker data to validate its generalizability and predictive performance across different ethnicities, sexes, and metabolic phenotypes.

In summary, TyG-ABSI, as a composite indicator integrating insulin resistance and central adiposity, provides a more comprehensive reflection of the intrinsic link between cardiometabolic dysfunction and atherosclerosis. Our findings suggest that TyG-ABSI exhibits superior predictive and stratification capability compared with conventional TyG-related indices and may serve as a novel and practical tool for ASCVD risk assessment, offering valuable insights for early intervention and personalized prevention strategies.

Conclusion

In this nationally representative cross-sectional study, TyG-ABSI—a composite index integrating insulin resistance (assessed via the TyG index) and body shape evaluation (via A Body Shape Index)—was significantly and positively associated with the risk of atherosclerotic cardiovascular disease (ASCVD) and its major subtypes (coronary heart disease, angina pectoris, myocardial infarction), independent of traditional cardiovascular risk factors. Notably, TyG-ABSI exhibited superior diagnostic performance compared with conventional TyG-related indices (TyG, TyG-BMI, TyG-WC, TyG-WHtR), with the highest area under the ROC curve (AUC = 0.65–0.70) across all outcomes. More importantly, it provided substantial incremental value in ASCVD risk stratification: via net reclassification improvement (NRI), TyG-ABSI correctly reclassified an additional 7.88%–16.36% of ASCVD patients, 11.03%–19.89% of coronary heart disease patients, and up to 19.92% of myocardial infarction patients, compared with traditional TyG parameters. These findings suggest that TyG-ABSI serves as a practical and robust tool for the early identification of ASCVD high-risk individuals in the general population, which can facilitate the development of targeted prevention strategies and personalized intervention measures.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ABSI

A body shape index

ACA

American College of Cardiology

AHA

American Heart Association

AP

Angina pectoris

ASCVD

Atherosclerotic cardiovascular disease

ATP III

Adult treatment panel III

AUC

Area under the curve

CDC

Centers for disease control and prevention

CHD

Coronary heart disease

CI

Confidence interval

CVD

Cardiovascular disease

DM

Diabetes mellitus

FBG

Fasting blood glucose

GED

General educational development

HbA1c

Glycated hemoglobin

HDL-C

High-density lipoprotein cholesterol

HOMA-IR

Homeostasis model assessment of insulin resistance

IDI

Integrated discrimination improvement

IQR

Interquartile range

IR

Insulin resistance

LDL-C

Low-density lipoprotein cholesterol

MET

Metabolic equivalent of task

MI

Myocardial infarction

NCHS

National Center for Health Statistics

NCEP

National Cholesterol Education Program

NHANES

National Health and Nutrition Examination Survey

NRI

Net reclassification improvement

OGTT

Oral glucose tolerance test

OR

Odds ratio

PA

Physical activity

PIR

Poverty-income ratio

RCS

Restricted cubic splines

ROC

Receiver operating characteristic

SD

Standard deviation

STROBE

Strengthening the reporting of observational studies in epidemiology

TG

Triglyceride

TyG

Triglyceride glucose

TyG-ABSI

Triglyceride glucose-a body shape index

TyG-BMI

Triglyceride glucose–body mass index

TyG-WC

Triglyceride glucose–waist circumference

TyG-WHtR

Triglyceride glucose–waist-to-height ratio

VIF

Variance inflation factor

WC

Waist circumference

WHO

World Health Organization

Author contributions

JP-L: Conceptualization, methodology, supervision, and project administration. JJ-Q, JE-L, SX: writing—original draft preparation. JJ-Q, JE-L, SX and YZ: writing—reviewing and editing. JJ-Q and JE-L: Software. JJ-Q, LX-F and JE-L: formal analysis and validation. JJ-Q, HT-Z, YJ-Z, JY-F, YL and YZ: data curation. All the authors read and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (Grant Nos. 82160162 and 81760150), the Key Research, Development Program of Jiangxi Province (20243BBI91008), the Project of the Second Affiliated Hospital of Nanchang University (2022efyA04) and the Jiangxi Province Key Laboratory of Molecular Medicine (No. 2024SSY06231).

Data availability

The NHANES datasets are available for download from the NHANES home website (https://www.cdc.gov/nchs/nhanes/index.htm).

Declarations

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.