1. Introduction
Hypertension is a leading modifiable risk factor for cardiovascular and cerebrovascular diseases, contributing significantly to the global burden of heart failure, coronary artery disease, and stroke.[1,2] Although hypertension is readily diagnosable, it often progresses silently with underlying pathological alterations in the heart and cerebral vasculature already underway by the time of clinical detection.[3] Consequently, substantial cardiovascular risk may have already accrued, underscoring the urgent need for early identification and intervention. The development of novel biomarkers that can improve risk stratification and elucidate pathophysiological mechanisms is therefore imperative for enhancing primary prevention strategies.
Recent investigations have explored various potential determinants of hypertension, such as gut microbiota-related dietary indices and daily alcohol consumption.[4,5] However, many of these studies have focused on isolated factors, lacking integration of multifaceted physiological influences. The triglyceride glucose (TyG)–weight-adjusted waist index (WWI) (TyG-WWI) emerges as a composite marker that synergistically incorporates elements of glycolipid metabolism and anthropometric measures.[6] Specifically, the WWI serves as an established indicator of obesity-related health risks, with visceral adiposity being closely linked to metabolic disorders that promote cardiometabolic events.[7] Concurrently, the TyG index has been validated as a reliable surrogate marker of insulin resistance, a condition that more than doubles the risk of adverse outcomes in nondiabetic individuals with elevated insulin levels.[8–11] Nevertheless, when used in isolation, the WWI and TyG indices each offer only partial insight into the complex metabolic underpinnings of hypertension.
By integrating triglycerides (TG) and waist-to-weight ratios, TyG-WWI represents a novel approach that may more comprehensively capture metabolic dysregulation associated with hypertension. This integrated measure holds promise for improving the identification of high-risk populations and facilitating targeted preventive interventions. Despite its potential, the association between TyG-WWI and hypertension remains inadequately characterized.
Thus, this cross-sectional study aims to quantitatively evaluate the association between TyG-WWI and the prevalence of self-reported hypertension in a nationally representative sample of United States (US) adults across all body mass index (BMI) categories, and to explore, in a descriptive manner, the potential mediating roles of BMI and uric acid (UA).
2. Methods
2.1. Study population in National Health and Nutrition Examination Survey (NHANES)
The data used in this study were obtained from the NHANES conducted between 2015 and 2018. NHANES is an ongoing cross-sectional surveillance program conducted by the Centers for Disease Control and Prevention. NHANES employs a complex multistage stratified probability sampling design to select a nationally representative sample of noninstitutionalized US residents, with the primary objective of assessing population health trends and nutritional status across all age groups.
The study protocol was approved by the National Center for Health Statistics Research Ethics Review Committee. Written informed consent was obtained from all adult participants and the legal guardians of minors prior to data collection. Detailed documentation regarding the survey methodology and ethical considerations is available on the NHANES official website at www.cdc.gov/nchs/nhanes/irba98.htm.
This cross-sectional study utilized data from the NHANES conducted between 2015 and 2018, with an initial sample size of 19,225 participants. The substantial reduction in sample size reflects the combined effect of restricting the analysis to adults aged ≥ 20 years and the requirement for complete data on all covariates and outcome variables. The primary driver of sample size reduction was the age restriction (exclusion of participants < 20 years: n = 7937), followed by missing data on key variables, including fasting glucose and TG, which were measured only in the NHANES fasting subsample.
Specifically, we excluded participants with missing data on education, hypertension status, waist circumference (WC), weight, glucose, TG, cardiovascular disease history, BMI, smoking status, alcohol consumption, diabetes mellitus (DM), and UA. This complete-case approach was chosen to ensure consistency in the analytic sample across all models and to avoid potential biases introduced by different missing data patterns across variables. The representativeness of the analytic sample was assessed by comparing the distribution of key demographic characteristics with the full adult NHANES population, which showed consistent demographic distributions, suggesting that the exclusion of participants with missing data did not substantially alter the population representativeness (Fig.
1).
To assess the potential for selection bias, we compared the demographic and clinical characteristics of the final analytic sample (N = 2716) with those of participants excluded due to missing data (N = 8572). Participants with missing data were slightly older and had a higher proportion of racial/ethnic minorities compared to the analytic sample. However, the distribution of the core exposure variable (TyG-WWI components) and the outcome (self-reported hypertension) among participants with complete data was consistent with the full NHANES adult population, suggesting that the complete-case analysis did not substantially introduce bias in the exposure-outcome association.
2.2. TyG-WWI calculation formula
The TyG-WWI was calculated as follows[12,13]: TyG = ln [TG (mg/dL) × glucose (mg/dL)/ 2]; WWI = WC (cm)/√weight (kg); TyG-WWI = TyG × WWI. Fasting serum TG (LBXTR) and fasting plasma glucose (LBXGLU), both measured in the NHANES fasting subsample, were used to construct TyG-WWI.
2.3. Hypertension
The determination of hypertension cases was made through the following methods. Participants who answered “Yes” to the question “Has a doctor or health professional ever told you that you have high blood pressure?” were classified as having hypertension.
2.4. Covariates
Adjusted covariates include multidimensional parameters: demographic characteristics: gender, age (continuous variable), race/ethnicity; socioeconomic status: family poverty income ratio (PIR), educational attainment, marital status; lifestyle factors: alcohol consumption status, smoking exposure; others: self-reported BMI, UA, total cholesterol, high-density lipoprotein, low-density lipoprotein, diabetes, history of heart failure, coronary heart disease, angina pectoris, myocardial infarction, and stroke. These variables were collected through structured questionnaires, laboratory tests, and physical examinations as part of the NHANES protocol.
2.5. Statistical analysis
We performed statistical analyses using R software (version 4.3.2) and incorporated NHANES complex sampling weights to account for the stratified multistage probability sampling design. Specifically, because TyG-WWI incorporates fasting glucose and TG, which were measured in the NHANES fasting subsample, we used the fasting subsample examination weights (WTSAF2YR) provided by NHANES for the 2015 to 2016 and 2017 to 2018 cycles. To combine the 2 cycles (2015–2016 and 2017–2018), the weights were adjusted by dividing each cycle’s weight by 2, as recommended by the National Center for Health Statistics.
The survey design variables, including stratification (SDMVSTRA), primary sampling units (SDMVPSU), and the masked variance pseudo-PSU codes, were used to account for the complex survey design features. We used the “survey” package in R to analyze the survey-weighted data, with the “svyglm” function for logistic regression models that incorporated the sampling weights, stratification, and clustering. Statistical significance was set at a 2-sided α value of 0.05.
Continuous and categorical variables were compared using weighted Student t tests and weighted chi-square tests, respectively, to account for the complex survey design.
Accounting for the complex survey design (stratification, clustering, and weighting), 3 multivariate logistic regression models were constructed to evaluate the association between TyG-WWI and hypertension: an unadjusted model; Adjust 1, adjusted for age, sex, and race; and Adjust 2, additionally adjusted for BMI, PIR, education levels, marital status, smoking history, alcohol consumption, and DM. Results are presented as odds ratios (OR) with 95% confidence intervals (CI).
The distribution pattern of TyG-WWI was visualized using histograms to assess normality. To investigate potential nonlinear TyG-WWI levels and hypertension associations, restricted cubic spline (RCS) analysis was performed with 3 strategically positioned knots, ensuring accurate capture of underlying data trends. An RCS model was performed to visualize the dose–response associations between TyG-WWI levels and hypertension in the overall population and subgroups of age, gender, race, BMI, PIR, education levels, marital status, smoking history, alcohol consumption, and DM.
To investigate potential effect modification, subgroup analyses were performed across key covariates to examine the consistency of the TyG-WWI–hypertension association within different subgroups. To explore potential pathways, mediation analyses were performed using the “mediation” package in R, and nonparametric bootstrapping with 1000 iterations was employed to estimate the 95% CI for the mediation proportion. These mediation analyses are exploratory and descriptive only: exposure, proposed mediators, and outcome were assessed cross-sectionally, and BMI is mathematically related to components of TyG-WWI; therefore, the estimates should not be interpreted as evidence of biological mediation.
To address the concern that BMI may be overadjusted in models including both TyG-WWI (which incorporates WC and body weight) and BMI as a covariate, we conducted a sensitivity analysis in which BMI was excluded from the fully adjusted model. The association between TyG-WWI and hypertension was reestimated without BMI adjustment to assess the potential impact of collinearity and overadjustment on our estimates.
3. Results
3.1. Baseline characteristics of participants stratified by hypertension status
This study included a total of 2716 participants, of whom 890 (32.8%) had hypertension and 1826 (67.2%) did not. As shown in Table 1, significant differences (P < .05) were observed between the 2 groups for the majority of baseline characteristics. Participants in the hypertension group were significantly older [Median (Q1, Q3): 59.0 (47.0, 67.75) years vs 41.0 (29.25, 56.0) years, P < .001].
Significant differences were noted in racial distribution (P < .001), with a higher proportion of Non-Hispanic Black individuals in the hypertension group (25.3% vs 19.3%) and a lower proportion of Mexican Americans (10.3% vs 17.0%). Marital status also differed between groups (P < .001). However, no statistically significant differences were found in gender (P = .090), higher education attainment (P = .295), or family PIR (P = .820).
Regarding cardiovascular comorbidities, the prevalence of all recorded medical histories was significantly higher in the hypertension group than in the non-hypertension group, including heart failure (6.3% vs 0.7%, P < .001), coronary heart disease (7.3% vs 1.5%, P < .001), angina (4.4% vs 0.7%, P < .001), heart attack (7.1% vs 1.7%, P < .001), and stroke (5.6% vs 1.4%, P < .001). Furthermore, the prevalence of diabetes was markedly higher in the hypertension group (27.4% vs 6.5%, P < .001). Concerning lifestyle factors, a higher proportion of participants in the hypertension group had a history of smoking (55.3% vs 44.3%, P < .001), and the proportion of current smokers differed between groups (19.1% vs 16.4%, P = .005), while a lower proportion consumed alcohol (58.2% vs 65.0%, P < .001).
Critically, the core index of this study, TyG-WWI, was significantly higher in the hypertension group compared to the non-hypertension group [98.36 (91.59, 106.37) vs 91.11 (83.34, 99.33), P < .001]. Regarding metabolic parameters, the hypertension group exhibited a more adverse metabolic profile, including higher BMI, WC, weight, glycated hemoglobin, fasting glucose, UA, and TG levels (all P < .001). No significant differences were found between groups for high-density lipoprotein (P = .078) and total cholesterol (P = .501), while low-density lipoprotein was slightly lower in the hypertension group (P = .010).
3.2. Logistic regression analysis and nonlinear relationship between TyG-WWI and hypertension
To investigate the independent association between TyG-WWI and hypertension, we performed multivariable logistic regression analyses (Table 2). In the unadjusted model, each unit increase in TyG-WWI was associated with a significant 6% increase in the odds of hypertension (OR = 1.06, 95% CI: 1.05–1.06, P < .001).
This positive association, although attenuated, remained highly significant after adjustment for age, gender, and race (Adjust 1) (OR = 1.04, 95% CI: 1.04–1.05, P < .001). After further adjustment for BMI, PIR, education levels, marital status, smoking history, alcohol consumption, and DM (Adjust 2), the positive correlation between TyG-WWI and hypertension remained statistically significant. In the fully adjusted model, a 1-unit increase in TyG-WWI was associated with a 2% increase in the odds of hypertension (OR = 1.02, 95% CI: 1.01–1.03, P < .001), indicating a modest but statistically significant independent association.
When participants were categorized into quartiles (Q1–Q4) of TyG-WWI, with Q1 as the reference, the risk of hypertension showed a significant increasing trend across ascending quartiles (P for trend < .001). In the fully adjusted model (Adjust 2), compared to the Q1 group, the odds of hypertension were significantly higher in the Q3 group (OR = 1.49, 95% CI: 1.08–2.07, P = .017) and the Q4 group (OR = 1.72, 95% CI: 1.21–2.46, P = .003), by 49% and 72%, respectively. The risk increase in the Q2 group was not statistically significant (OR = 1.27, 95% CI: 0.92–1.74, P = .142).
In addition, RCS were used to flexibly assess the distribution (Fig. 2A) and dose–response relationship (Fig. 2B) between TyG-WWI and hypertension.
After fully adjusting for all covariates, a significant nonlinear positive correlation was observed between TyG-WWI and the risk of hypertension (nonlinear test P value < .05). The curve indicated that the risk increased relatively gradually at lower levels of TyG-WWI but exhibited a steeper increase at higher levels of TyG-WWI.
3.3. Subgroup analyses of the association between TYG-WWI and hypertension
We performed subgroup analyses to examine the consistency and potential effect modification of the association between TyG-WWI and hypertension across various population characteristics. All models were adjusted for age, sex, race, education, income, marital status, smoking, alcohol consumption, BMI, and diabetes (Fig. 3).
In the overall population, each unit increase in TyG-WWI was associated with a significant 2% increase in the odds of hypertension (OR = 1.02, 95% CI: 1.01–1.03). The subgroup analyses revealed heterogeneity in this association. The interaction with sex was of borderline statistical significance (P = .068), with the association being more pronounced and significant in females (OR = 1.028, 95% CI: 1.012–1.045) than in males (OR = 1.012, 95% CI: 0.997–1.028).
BMI was a significant effect modifier (P for interaction = 0.003). Notably, the association was strongest in the normal-weight group (OR = 1.060, 95% CI: 1.033–1.088) and was attenuated and became nonsignificant in the overweight and obese groups. UA level also showed a significant interaction effect (P = .028).
The strength of the association was highest in the first quartile (Q1) of UA (OR = 1.026, 95% CI: 1.002–1.051) and diminished across higher quartiles.
Furthermore, significant associations persisted within specific subgroups, including Non-Hispanic White individuals (OR = 1.027), never smokers (OR = 1.016), former smokers (OR = 1.024), and nondrinkers (OR = 1.029). In other subgroups, such as different age strata and some racial groups, the point estimates suggested a positive trend, but the CI included 1, indicating nonsignificance.
3.4. Mediation analysis of the role of BMI and UA in the association between TyG-WWI and hypertension
To investigate the potential pathways through which TyG-WWI might be associated with hypertension, we conducted exploratory mediation analyses to estimate the proportions of the association potentially accounted for by BMI and UA (Fig. 4). Models were adjusted for age, sex, ethnicity, marital status, education, smoking, and alcohol consumption.
In these exploratory analyses, the average causal mediating effect of BMI was 0.000166 (95% CI: 0.000058–0.000432), and its average causal direct effect was 0.000374 (95% CI: 0.000197–0.000625); BMI accounted for 30.49% of the total effect (Fig. 4A, 4C, and 4D). UA exhibited an average causal mediating effect of 0.000078 (95% CI: 0.000020–0.000215) and an average causal direct effect of 0.000897 (95% CI: 0.000457–0.001285); UA accounted for 7.95% of the total effect (Fig.
4B, 4C, and 4D). In a sensitivity analysis excluding BMI from the fully adjusted model (Adjust 2 without BMI), the association between TyG-WWI and hypertension remained significant, with a 1-unit increase in TyG-WWI associated with a 3.6% increase in the odds of hypertension (OR = 1.036, 95% CI: 1.026–1.046, P < .001). This estimate was slightly larger than that of the fully adjusted model including BMI (OR = 1.02, 95% CI: 1.01–1.03, P < .001), suggesting that adjustment for BMI partially attenuates but does not negate the independent association of TyG-WWI with hypertension (Fig.
S1, Supplemental Digital Content 1). We emphasize that these mediation estimates are descriptive only: exposure, mediators, and outcome were assessed cross-sectionally, and BMI is mathematically related to the WC and weight components of TyG-WWI. These analyses should not be interpreted as evidence that BMI or UA mediates a biological pathway to hypertension; they require confirmation in prospective studies with an appropriate temporal design.
4. Discussion
In this cross-sectional study of a nationally representative US sample, we found that a novel integrated metric, TyG-WWI, was significantly and nonlinearly associated with a higher prevalence of self-reported hypertension, even after comprehensive adjustment for a multitude of established confounders. Our findings extend the current understanding of metabolic health assessment by introducing and validating a novel composite index that synergistically integrates markers of TyG and WWI. The observed nonlinear dose–response relationship, wherein hypertension risk escalates more sharply at higher TyG-WWI levels, highlights the potential value of TyG-WWI for risk stratification; exploratory mediation analyses further suggested that BMI and UA may account for substantial proportions of the association, although these estimates are descriptive only.
The robustness of the TyG-WWI-hypertension association, which persisted after adjusting for demographic, socioeconomic, lifestyle, and clinical comorbidities, underscores the index’s value as an independent risk indicator. This suggests that TyG-WWI captures a distinct facet of metabolic dysregulation beyond traditional risk factors. The quartile analysis revealed a clear gradient effect, with individuals in the highest quartile facing a 72% increased odds of hypertension compared to the lowest, reinforcing its potential for risk stratification.
We acknowledge the concern regarding potential overadjustment and collinearity, as TyG-WWI is constructed from WC and body weight, which are components strongly correlated with BMI. To address this, we conducted a sensitivity analysis excluding BMI from the adjusted model, which yielded similar results. This suggests that while BMI partially contributes to the TyG-WWI-hypertension association, TyG-WWI retains an independent association beyond its correlation with BMI.
This finding supports the utility of TyG-WWI as a composite marker that captures metabolic dysregulation beyond general adiposity.
Notably, subgroup analyses unveiled important effect modifications. The stronger association observed in females, albeit of borderline significance, may hint at sex-specific pathophysiological pathways, potentially involving hormonal influences on adiposity distribution and insulin sensitivity. The most striking modification was observed across BMI strata.
The strongest association was observed in individuals with normal BMI, suggesting that TyG-WWI is exceptionally adept at identifying the “metabolically obese normal-weight” phenotype.[14] This subgroup, characterized by normal weight but adverse metabolic profiles, is at elevated cardiometabolic risk yet often remains undetected by conventional screening based on BMI alone. The TyG-WWI may therefore be particularly informative for risk stratification in this population, which is at elevated cardiometabolic risk yet is not readily identified by BMI-based screening. The attenuation of the association in overweight/obese groups implies that at higher BMI levels, the additional risk conveyed by TyG-WWI’s specific metabolic components might be overshadowed by the overwhelming risk burden of general adiposity itself.
Exploratory mediation analyses suggested that BMI and UA may account for substantial proportions of the TyG-WWI–hypertension association (30.49% and 7.95%, respectively). We emphasize, however, that these estimates are descriptive: exposure, proposed mediators, and outcome were measured cross-sectionally, and BMI is mathematically related to the WC and weight components of TyG-WWI. These analyses should not be interpreted as evidence that BMI or UA mediates a biological pathway to hypertension.
Causal mediation hypotheses would require evaluation in prospective studies designed specifically to test them. Collectively, the current study, along with emerging evidence, underscores the significant role of TyG-WWI in cardiometabolic risk assessment. Our findings reveal its independent association with hypertension.
This is further supported by research linking TyG-WWI to an increased risk of stroke, with 1 study reporting an 11% higher risk per standard deviation increase.[12] Another large-scale analysis not only confirmed this association but also highlighted its predictive value when compared with other indices, noting a nonlinear relationship that may aid in identifying critical risk thresholds.[15] By concurrently reflecting insulin resistance and central obesity, TyG-WWI demonstrates superior utility as a composite marker for early risk stratification.
However, several limitations of this study should be considered. The cross-sectional design precludes inference of causality; longitudinal studies are warranted to establish whether TyG-WWI predicts incident hypertension. Another limitation is the reliance on self-reported hypertension, rather than direct blood pressure measurements, which may have led to outcome misclassification.
Consequently, this study does not capture undiagnosed or previously unrecognized (“hidden”) hypertension. Specifically, individuals with undiagnosed hypertension would be misclassified as non-hypertensive, potentially biasing our estimates toward the null and resulting in an underestimation of the true association. Additionally, the use of self-report may differentially affect individuals with lower healthcare access, further contributing to the conservativeness of our findings.
The use of complete-case analysis may introduce selection bias, particularly if participants with missing data differ systematically from those included in the analysis. However, our comparison of baseline characteristics between the analytic sample and excluded participants suggested that the exposure-outcome association was not substantially affected by the exclusion of participants with missing data.
Despite these limitations, our study possesses notable strengths, including the use of a large, nationally representative sample with standardized data collection, comprehensive adjustment for confounders, rigorous investigation of nonlinear relationships and effect modification, and formal mediation analysis to explore biological pathways.
5. Conclusion
In conclusion, this study establishes TyG-WWI as a robust and independent biomarker associated with the prevalence of self-reported hypertension among US adults across all BMI categories. Its ability to integrate information on insulin resistance and central adiposity into a single metric offers a more holistic assessment of metabolic health, and the stronger association observed in normal-weight individuals suggests that TyG-WWI may complement BMI-based screening. The observed nonlinear association and the exploratory mediation estimates warrant confirmation in future prospective studies designed to validate TyG-WWI as a predictive tool for incident hypertension and to test causal mediation hypotheses.
Acknowledgments
We sincerely acknowledge the contributors of the publicly available datasets NHANES (2015–2018) from the National Health and Nutrition Examination Survey (https://wwwn.cdc.gov/nchs/nhanes/Default.aspx), which were indispensable for this study. This study received no grants, contracts, or other forms of financial support, and the authors declare no competing interests.
Author contributions
Conceptualization: Ling Feng.
Data curation: Shiyan Long.
Formal analysis: Jianyu Peng.
Investigation: Jianyu Peng.
Methodology: Jianyu Peng, Ling Feng.
Project administration: Ling Feng.
Software: Huan Liu, Shiyan Long.
Supervision: Ling Feng.
Validation: Dan Wei.
Visualization: Jianyu Peng, Huan Liu.
Writing – original draft: Jianyu Peng.
Writing – review & editing: Dan Wei, Ling Feng.