Derivatives of the triglyceride–glucose index and their association with incident hypertension in prehypertensive individuals: a 4-year cohort study augmented by mendelian randomization

Abstract

Background

The direct association between elevated levels of the triglyceride-glucose index (TyG index) and its derived metrics and the risk of new-onset hypertension in prehypertensive populations remains unclear. The study systematically evaluated the link between the TyG index and its related indicators with new-onset hypertension by integrating cohort study methods with Mendelian randomization (MR) analysis.

Methods

A total of 2,815 prehypertensive participants from the 2011 CHARLS database were included, of whom 877 (31.15%) progressed to new-onset hypertension by 2015. TyG-Waist-to-Height Ratio (TyG-WHtR), TyG-Body Mass Index (TyG-BMI), TyG-Waist Circumference (TyG-WC), and the TyG index were calculated. Logistic regression, restricted cubic spline (RCS) curves, subgroup analyses, and interaction tests were performed to assess the associations. Bayesian weighted MR (BWMR) was further used to validate causal relationships.

Results

Multivariable regression analysis revealed that each unit increase in the TyG index was associated with a 45% higher risk of new-onset hypertension (odds ratio [OR]: 1.45, 95% confidence intervals [CI] 1.24–1.70, P < 0.001), while TyG-WHtR showed a 42% increased risk (OR: 1.42, 95% CI 1.26–1.60, P < 0.001). Categorizing the TyG and its derived metrics by quartiles demonstrated that higher quartiles (Q3 and Q4) were significantly linked to an elevated risk of new-onset hypertension across all models (P < 0.001). RCS models indicated significant positive linear relationships between the TyG index and TyG-WC with new-onset hypertension (P for overall < 0.001, P for nonlinearity = 0.844 and 0.165, respectively), whereas TyG-WHtR and TyG-BMI exhibited significant positive nonlinear relationships (P for overall < 0.001, P for nonlinearity = 0.001 and 0.046, respectively). Subgroup analyses highlighted stronger associations among individuals aged ≥ 70 years, those who were widowed, had cardiovascular disease, or reported a life satisfaction score of 2 (P < 0.05, P for interaction < 0.05). BWMR analysis confirmed a significant causal relationship between genetically elevated TyG index levels and an increased risk of new-onset hypertension (P < 0.05).

Conclusions

Our study reveals a significant link between the TyG index and its related indicators with new-onset hypertension in prehypertensive populations. Causal analysis using BWMR confirmed that genetically elevated TyG index levels increase the risk of new-onset hypertension. These results highlight the importance of monitoring TyG-related indices for early detection and intervention in high-risk individuals, aiding in the prevention of hypertension progression.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-025-02813-6.

Research insights

What is currently known about this topic?

• Hypertension is a major risk factor for cardiovascular disease. The TyG index is a well-established surrogate marker for insulin resistance and metabolic syndrome. However, long-term associations between its derived metrics and incident hypertension remain underexplored, especially in prehypertensive populations.

What is the key research question?

• Which TyG-derived metrics are most strongly associated with hypertension risk?

• Is there genetic evidence supporting a causal link?

What is new?

• TyG, TyG-WHtR, TyG-BMI, and TyG-WC showed significant positive associations with new-onset hypertension over a 4-year follow-up.

• BWMR analysis provided support for a causal relationship between genetically predicted TyG levels and hypertension risk.

How might this study influence clincal practice?

• TyG-related indices offer a simple, cost-effective tool for identifying individuals at high risk of developing hypertension.

• Early monitoring and lifestyle interventions targeting metabolic health may help delay or prevent the progression from prehypertension to hypertension.

Introduction

Prehypertension, defined as a systolic blood pressure (SBP) ranging from 120 to 139 mmHg or a diastolic blood pressure (DBP) between 80 and 89 mmHg, affects approximately 25–50% of the global adult population and significantly increases the risk of progression to hypertension [1, 2]. For example, in the United States, individuals aged 55–65 years who are free of hypertension face a lifetime risk of approximately 90% for developing hypertension [2]. Research has shown that prehypertension not only elevates the risk of new-onset hypertension but also significantly increases the incidence of cardiovascular disease (CVD), coronary heart disease, myocardial infarction, and stroke [2]. In a Chinese study focusing on individuals aged 15 years and older, 34.5% of participants were diagnosed with prehypertension [3]. Similarly, findings from the Korean National Health and Nutrition Examination Survey indicated that 31.6% of individuals had prehypertension, with a greater prevalence noted in females than in males [4]. Based on a synthesis of 18 prospective studies, the estimated global prevalence of prehypertension ranges from 25.2 to 46.0% [5]. Furthermore, in a study examining baseline prehypertension, 44% of participants in the standard care group developed hypertension over a 3-year observation period [6]. In recent years, the strategy of shifting the focus upstream and preventing disease before it occurs has gained global attention, emphasizing early identification and intervention to reduce disease progression and complications [7, 8]. Given the significant occurrence of prehypertension and its related health risks, identifying effective biomarkers for the early prediction and intervention of new-onset hypertension in prehypertensive populations has become particularly important. Furthermore, the healthcare burden and economic costs associated with CVD are projected to rise significantly in the coming years, with hypertension posing a widespread public health challenge globally, severely impacting both individual health and socioeconomic conditions [9]. For example, the total economic cost of CVD in the United States is expected to rise from less than $500 billion in 2010 to $1.1 trillion by 2030 [9]. Therefore, strengthening research in the field of prehypertension is urgently needed.

The Triglyceride-Glucose index (TyG index), first reported in 2008, is calculated using the formula: TyG index = Ln [fasting triglycerides (TG) (mg/dl) × fasting blood glucose (FBG) (mg/dl) / 2] [10]. It is a combined statistical measure that integrates TG and FBG levels. Because of its strong sensitivity and specificity, it functions as an effective proxy biomarker for insulin resistance (IR) [10]. The TyG index is simple to calculate, is not constrained by time or cost limitations, and is suitable for large-scale population studies [11]. In recent years, an increasing number of studies have indicated that an elevated TyG index is independently linked to a higher risk of several chronic diseases, such as type 2 diabetes mellitus (DM), CVD, cerebrovascular disease, fatty liver, kidney disease, and reproductive system disorders [12–16]. However, although evidence supports the TyG index as an independent factor for increased hypertension risk in the general population, previous studies assessing the association between the TyG index and hypertension risk in community-based adult populations have yielded inconsistent results [17–21]. This suggests the need for further exploration of the role of the TyG index and its related indicators (such as TyG-Waist Circumference (TyG-WC), TyG-Waist-to-Height Ratio (TyG-WHtR), and TyG-Body Mass Index (TyG-BMI)) in prehypertensive populations, particularly their potential link to new-onset hypertension.

Mendelian Randomization (MR) uses genetic markers as instrumental variables (IVs) to examine the cause-and-effect relationship between modifiable exposures and health outcomes, taking advantage of the random distribution of alleles during gamete formation to reduce confounding factors and reverse causality [22]. Therefore, we combined MR analysis with the China Health and Retirement Longitudinal Study (CHARLS) database to conduct a four-year cohort study, systematically assessing the connection between the TyG index and its related indicators (such as TyG-WC, TyG-WHtR, and TyG-BMI) and the likelihood of developing new-onset hypertension among prehypertensive individuals. Our goal is to provide a more precise risk stratification tool for prehypertensive individuals, facilitating early identification and intervention to reduce disease progression and the incidence of complications.

Methods

Research design

This study utilized a combined cohort and MR design to thoroughly examine the causal link between the TyG index and its related indicators and new-onset hypertension.

Data source and study population for the cohort study

The data for this cohort study were derived from the CHARLS database, a national longitudinal study initiated in 2011 that specifically targets adults aged 45 years and older in China (http://charls.pku.edu.cn/) [16, 23]. By employing a multi-phase stratified random sampling technique, the study encompassed 150 regions or districts spanning 28 provinces in China, achieving an initial response rate of 80.5% [24]. This approach helped minimize selection bias and improved the sample's representativeness [24]. This study used data from 2011 as the baseline and 2015 data for follow-up outcome analysis. The detailed study flowchart is shown in Fig. 1. During participant screening, we applied the following exclusion criteria: (1) lack of blood biochemical test data from the 2011 baseline assessment; (2) lack of household roster details from the 2011 baseline assessment; (3) not meeting the prehypertension criteria (SBP between 120 and 139 mmHg or DBP between 80 and 89 mmHg); (4) missing SBP and DBP data at the 2015 follow-up; (5) no record of the TyG index and related indicators. Following the implementation of these selection criteria, a total of 2815 individuals were incorporated into the conclusive analysis (Fig. 1). Based on a priori power analysis, the final sample size was sufficient to detect clinically relevant associations with adequate statistical power.

Fig. 1.

The flowchart of this study

Conducted in strict accordance with the Declaration of Helsinki, the CHARLS study was granted ethical approval by the Institutional Review Board of Peking University (IRB00001052-11015). All subjects gave their written informed consent. Additionally, this study complies with the STROBE guidelines to ensure the quality and transparency of observational research reporting. These measures collectively safeguard the ethical integrity and scientific rigor of the study.

Assessment of the TyG index and TyG-related indicators

The TyG index and its related indicators were assessed by the Clinical Laboratory at Beijing You'an Hospital, which is operated under Capital Medical University. These assessments utilized frozen plasma or whole blood samples and employed an enzymatic colorimetric technique to measure TG and FBG levels, maintaining a variation coefficient under 2%, thereby ensuring the precision and reliability of the test results [25]. The TyG index, a reliable method for assessing IR, was calculated using the formula TyG index = ln [TG (mg/dL) × FBG (mg/dL)/2]. The TyG-related indicators were further calculated by incorporating anthropometric parameters [26], including TyG-WHtR = TyG × WHtR, TyG-WC = TyG × WC, and TyG-BMI = TyG × BMI. For ease of analysis, the subjects were split into four quartile groups (Q1, Q2, Q3, Q4) based on their TyG index values and associated indicators, with Q1 established as the reference category. This grouping method facilitates a deeper exploration of the link between each group and health outcomes and reveals the potential roles of different levels of the TyG index and its related indicators in the progression of prehypertension.

Assessment of new-onset hypertension events

The primary outcome of this study was to assess the progression to new-onset hypertension among participants who were in the prehypertensive stage in 2011 after a 4-year follow-up period (2015). Specifically, the study focused on individuals who met the prehypertension criteria at baseline (2011)—defined as a SBP between 120 and 139 mmHg or a DBP between 80 and 89 mmHg—and evaluated whether they developed hypertension by the end of the 4-year follow-up in 2015. At the 2015 follow-up, new-onset hypertension was defined as an average SBP ≥ 140 mmHg or DBP ≥ 90 mmHg based on three repeated measurements taken during a single visit using an Omron HEM-7200 digital sphygmomanometer, following standardized protocols from the CHARLS. Participants were seated with their arm supported at heart level during all measurements, and trained field staff conducted the assessments. Prior to measurement, participants were asked whether they had smoked, exercised, eaten, or consumed alcohol within the past 30 min; those who responded affirmatively were excluded from that session to minimize potential confounding effects. Blood pressure values were recorded across three rounds, and implausible readings (e.g., SBP > 300 mmHg or DBP > 200 mmHg) were excluded from analysis to avoid outlier influence. The final SBP and DBP values used in the analysis were calculated as the average of the three valid readings. In addition to measured blood pressure levels, self-reported physician diagnosis of hypertension and antihypertensive medication use were also considered in defining hypertension status. Individuals who did not meet the blood pressure thresholds but reported being diagnosed by a doctor or currently taking antihypertensive medications were still classified as hypertensive. All participants included in the final analysis completed the full 4-year follow-up, ensuring data continuity and enhancing the reliability of the observed outcomes.

Data collection in the cohort study

This study collected comprehensive data on participants, encompassing demographic characteristics such as age, gender, marital status, education level, and residential area; lifestyle habits including alcohol consumption and smoking status; physical examination indicators comprising BMI, WC, baseline SBP, and baseline DBP; laboratory test results included uric acid (mg/dL), white blood cell count (WBC, × 10⁹/L), platelet count (PLT, × 10⁹/L), hemoglobin (HGB, g/dL), TG (mg/dL), FBG (mg/dL; 1 mmol/L = 18 mg/dL for clinical reference), total cholesterol (TC, mg/dL), and low-density lipoprotein cholesterol (LDL-C, mg/dL); comorbidities such as CVD and DM; and mental health assessments including depression and life satisfaction. These integrated data provide a solid foundation for comprehensively analyzing the risk of new-onset hypertension events among participants with prehypertension.

In this study, participant characteristics were categorized based on multiple variables: residential area was divided into urban and rural; age groups were classified into four categories: < 50 years, 50–60 years, 60–70 years, and ≥ 70 years; education level was divided into four categories: illiterate, primary school, middle school, and high school or above; marital status was classified into four categories: divorced, married, unmarried, and widowed; smoking status was divided into three categories: non-smoker, ex-smoker, and smoker; alcohol intake was categorized into three levels: drink but less than once a month, drink more than once a month, and none of these; BMI groups were divided into four categories: < 18.5, 18.5–24, 24–28, and ≥ 28; CVD and DM were each divided into two categories: no and yes; depressive symptoms were also classified into two categories: no and yes; and life satisfaction was further subdivided into five categories: 0, 1, 2, 3, and 4, reflecting varying levels of life satisfaction. These classifications help comprehensively analyze the characteristics of participants across different backgrounds and health statuses and how these factors affect the probability of new hypertension occurrences.

Definition and assessment criteria for comorbidities

In the CHARLS database, comorbidities were determined by analyzing participants' self-reported medical history and biochemical test results. Specifically, CVD was defined as a participant's self-reported history of heart disease or stroke. If the value of variable `da007_8` ((stroke) or `da007_7` (heart disease) was 1, CVD was assigned a value of 1 (indicating presence); otherwise, it was assigned 0. DM was defined based on glycated hemoglobin (HbA1c) levels and self-reported DM diagnosis. If a participant’s HbA1c level was below 6.5%, they did not self-report DM (i.e., variable `da007_3` was not equal to 1), and they did not report using antidiabetic medications (i.e., variables `da014s1` and `da014s2` were not equal to 1), the participant was considered free of DM and assigned a value of 0; otherwise, they were assigned a value of 1 (indicating the presence of DM).

Depression and life satisfaction were assessed through mental health evaluations using the Center for Epidemiologic Studies Depression Scale (CES-D) scale to measure the severity of depressive symptoms. Scores for specific items were reversed where necessary to ensure consistency, and a final CES-D score of 10 or higher was considered indicative of depression. Additionally, life satisfaction was evaluated by reversing the scoring of specific questions, providing a comprehensive reflection of the participants’ mental health status.

MR analysis

This study utilized the UK Biobank cohort dataset to investigate the potential causal links between the TyG index and new-onset hypertension, SBP, and DBP through MR analysis. Previous studies identified (single-nucleotide polymorphisms) SNPs significantly associated with the TyG index from the UK Biobank database as IVs. The inclusion criteria for these IVs were set at P < 5 × 10⁻⁸, and linkage disequilibrium pruning (r² < 0.001) was applied to ensure their independence. A total of 192 SNPs associated with the TyG index were ultimately identified (Supplementary Table S1). Blood pressure-related data were sourced from multiple specific datasets to ensure the stability of the conclusions: hypertension data were derived from ukb-b-14057, ebi-a-GCST90038604, and ukb-a-61; SBP data from ieu-b-38 and ebi-a-GCST90025968; and DBP data from ieu-b-39 and ebi-a-GCST90025981. Additional details regarding the datasets can be found at https://gwas.mrcieu.ac.uk/. To detect pleiotropy and evaluate heterogeneity, we used the MR-PRESSO test along with the inverse-variance weighted (IVW) Cochran's Q statistic. Furthermore, we adopted funnel plots, much like those used to spot publication bias in meta-analyses, to explore the presence of directional pleiotropy [27]. Additionally, sensitivity analyses were conducted using the leave-one-out approach. We sequentially excluded each SNP, recalculated estimates for the other variables, and compared them to the total estimate to evaluate individual SNP effects.

MR analysis relies on three key assumptions [22]: exclusivity of effect, independence, and specificity of association. (1) The IVs must have a strong association with the exposure; (2) tThe IVs should be independent and unrelated to other factors that could affect the "exposure-outcome" link; (3) The IVs' effect on the outcome should be confined to their association with the exposure. The primary analysis was conducted using the IVW method, with weights determined by the reciprocal of the variance of the effect estimates. Additionally, we employed five different MR methods to validate the causal relationships between the TyG index and blood pressure-related outcomes, including the weighted median method, IVW, simple mode method, weighted mode method, and MR-Egger. To control for multiple testing, we applied the False Discovery Rate (FDR) correction using the Benjamini–Hochberg procedure to the primary MR results. To further enhance the stability and reliability of the results, we introduced Bayesian weighted MR (BWMR) analysis alongside the initial IVW analysis [28]. BWMR assigns weights to each genetic instrument based on its pleiotropic effects, thereby reducing potential pleiotropic influences and providing more nuanced uncertainty and sensitivity analyses. This approach enriched our knowledge of the causal relationships involving the TyG index and new-onset hypertension, SBP, and DBP.

Statistical analysis

Descriptive statistics were used to summarize the baseline characteristics of the participants, with records made based on whether they progressed to new-onset hypertension. The normality of the data was assessed using the Shapiro–Wilk test. Continuous variables were presented as means ± standard deviation (SD) or medians (interquartile range, IQR), and categorical variables were expressed as frequencies (percentages). For comparing categorical variables, Fisher’s exact tests or Chi-square tests were conducted, whereas continuous variables were analyzed using the Mann–Whitney U test. For the comparison of three or more groups of continuous variables, the Kruskal–Wallis rank-sum test was conducted. Moreover, Pearson’s chi-square test was applied to analyze associations between categorical variables, ensuring comprehensive and accurate data analysis.

To explore the relationship between the TyG index and its related indicators (TyG-WC, TyG-WHtR, TyG-BMI) and the risk of new-onset hypertension, we constructed univariate and multivariate logistic regression models. Covariates in the multivariate models were selected based on clinical relevance and statistical significance. In our analysis, we assessed the linearity assumption for all continuous covariates using martingale residuals to ensure that the relationships between these variables and the outcome were appropriately modeled. No significant deviations from linearity were observed (Supplementary Figure S1).The TyG index and its related indicators were included in the analysis both as continuous variables and as categorical variables. The study population was divided into four groups based on quartiles (Q1-Q4) of these indicators: for the TyG index, Q1 ranged from 4.960 to 8.238, Q2 from 8.239 to 8.617, Q3 from 8.618 to 9.070, and Q4 > 9.070; for TyG-WC, Q1 ranged from 71.250 to 651.900, Q2 from 651.901 to 727.890, Q3 from 727.891 to 816.300, and Q4 > 816.300; for TyG-WHtR, Q1 ranged from 0.493 to 4.102, Q2 from 4.102 to 4.621, Q3 from 4.622 to 5.191, and Q4 > 5.191; for TyG-BMI, Q1 ranged from 13.690 to 175.537, Q2 from 175.538 to 199.129, Q3 from 199.130 to 229.334, and Q4 > 229.334. This approach ensured a comprehensive evaluation of the connection between these indicators and new-onset hypertension.

To ensure model stability, variance inflation factors (VIFs) were calculated for all predictors. The maximum VIF value across all variables was below 5, indicating no significant multicollinearity. Specifically, VIF values were: TyG index (2.37), triglycerides (2.10), and fasting glucose (1.67). We also fitted an alternative model excluding fasting glucose and triglycerides; in this model, the VIF for TyG decreased to 1.50, and effect estimates remained consistent, supporting robustness. Detailed VIF results are provided in Supplementary Table S2.

We explored the link between the TyG index and its related indicators (TyG-WC, TyG-WHtR, TyG-BMI) and the risk of new-onset hypertension events using a series of logistic regression models. Model 1 was unadjusted; Model 2 adjusted for gender, age, and residential area; Model 3 further included education level, marital status, smoking and alcohol consumption habits, and baseline SBP and DBP; Model 4 additionally incorporated biochemical indicators such as WBC, HGB, PLT, TC, LDL-C, uric acid, creatinine, cystatin C, blood urea nitrogen (BUN), as well as comorbidities such as CVD and DM based on Model 3. Table 1 presents the detailed structure of the multivariable logistic regression models, outlining the progressively adjusted sets of covariates used in the analysis. To provide a more intuitive visualization of the associations between TyG-related indices and new-onset hypertension, we employed restricted cubic spline (RCS) models based on logistic regression. Model selection was guided by the Bayesian Information Criterion (BIC) to achieve an optimal balance between flexibility and parsimony. We evaluated models with 3, 4, and 5 knots for each index using both AIC and BIC, selecting the model with the lowest BIC value for final analysis. As detailed in Supplementary Table S3, the 3-knot model consistently showed the best fit across all four TyG-related indicators. Likelihood ratio tests were used to assess nonlinear relationships in the RCS models, and the adjustment factors in the RCS models were aligned with those in the logistic regression models [29]. To assess the robustness of the logistic regression model, we conducted diagnostic checks including analysis of Pearson and deviance residuals, leverage values, and Cook’s distance. No observations with Cook’s distance > 1 were identified, indicating no highly influential data points that could bias the odds ratio estimates. A plot of fitted values versus Pearson residuals also showed no systematic deviations, suggesting a good model fit. These results support the stability and reliability of the model estimates. Supplementary Fig. 2A and B provide visual representations of the diagnostic analyses.

Table 1.

Summary of logistic regression models

Model	Adjusted variables	Description
Model 1	None	Crude model without adjustment for any covariates
Model 2	Sex, age, residential area	Adjusted for basic demographic factors based on Model 1
Model 3	Sex, age, residential area, education level, marital status, smoking and alcohol consumption habits, baseline SBP and DBP	Further adjusted for socioeconomic and behavioral factors, as well as baseline blood pressure in addition to variables in Model 2
Model 4	Sex, age, residential area, education level, marital status, smoking and alcohol consumption habits, baseline SBP and DBP, WBC, HGB, PLT, TC, LDL, uric acid, creatinine, cystatin C, BUN, CVD, DM	Fully adjusted model including key biochemical markers and comorbidities based on Model 3

Abbreviations: SBP, systolic blood pressure; DBP, diastolic blood pressure; WBC, white blood cell count; HGB, hemoglobin concentration; PLT, platelet count; TC, total cholesterol; LDL-C, low-density lipoprotein cholesterol; BUN, blood urea nitrogen; CVD, cardiovascular disease; DM, diabetes mellitus

To assess multicollinearity among covariates, we calculated variance inflation factors (VIF) for all predictors in the logistic regression model. A VIF value greater than 5 was considered indicative of significant multicollinearity. Given that the TyG index is derived from FBG and TG, we also fitted an alternative model excluding these two variables to evaluate the stability of effect estimates. All statistical analyses were conducted using R software (version 4.4.1). The car package was used to compute VIF values. To further explore potential effect modification, we performed subgroup analyses according to age, gender, marital status, education level, residential area, alcohol consumption, smoking, BMI, DM, CVD, depression, and life satisfaction. Interaction terms between the exposure variable and each stratification factor were included in the models to test for heterogeneity across subgroups. For the MR analysis, we used the TwoSampleMR package (version 0.6.15) in R to evaluate the causal associations between metabolic indicators and incident hypertension. To explore the non-linear relationship between continuous variables (e.g., TyG index) and incident hypertension, we used the rms package (version 6.8.2) along with the splines package (version 4.4.1) for RCS curve fitting. We used the pROC package (version 1.18.5) to calculate the area under the receiver operating characteristic curve (AUC) for evaluating the predictive performance of the TyG index and other biomarkers. Data cleaning and preparation were performed using the tidyverse package (version 2.0.0). Visualization of results was conducted using ggplot2 (version 3.5.1), gtsummary (version 2.0.1), gt (version 0.11.0), and flextable (version 0.9.6). Summary statistics and descriptive tables were generated using the tableone package (version 0.13.2), skimr (version 2.1.5), and arsenal (version 3.6.3). For multivariate regression analyses, we utilized the lm.beta package (version 1.7.2) for standardized regression coefficients, glmnet package (version 4.1.8) for regularized regression, and dplyr (version 1.1.4) and Hmisc (version 5.1.3) for data manipulation and summary statistics. All tests were two-sided, and statistical significance was defined as a P-value < 0.05.

Results

Baseline characteristics by hypertension status

Table 2 presents the demographic and clinical characteristics of prehypertensive participants from 2011 to 2015, classified by whether they progressed to new-onset hypertension. The mean age of the study cohort was 59.72 years, with males accounting for 55.57%. Among the 2,815 participants, 877 (31.15%) progressed to new-onset hypertension by the end of the follow-up period. Compared to the non-progressors, the new-onset hypertension group exhibited several significant characteristics: older age, particularly a higher proportion in the 60–70-year age group; a higher proportion of males; higher BMI, especially in the 24–28 range; a higher prevalence of being married; and a greater likelihood of residing in rural areas. Additionally, the new-onset hypertension group demonstrated higher levels in waist circumference, SBP, DBP, uric acid, HGB, TG, FBG, BUN, creatinine, as well as the TyG index and its related indicators. The prevalence of CVD and DM was also higher in this group. However, no significant differences were observed between the two groups in terms of alcohol consumption, education level, LDL-C, TC, cystatin C, depression, and life satisfaction (all P > 0.05).

Table 2.

Demographic and clinical characteristics of prehypertensive participants progressing to developing new-onset hypertension

Characteristic	Overall N = 2,8151	Non-hypertension N = 1,9381	Hypertension N = 8771	p-value2
Age, years	59.72 (9.46)	59.21 (9.36)	60.86 (9.60)	< 0.001
Age_group, N(%)				< 0.001
< 50	567 (20.14%)	413 (21.31%)	154 (17.56%)
50–60	998 (35.45%)	729 (37.62%)	269 (30.67%)
60–70	877 (31.15%)	559 (28.84%)	318 (36.26%)
≥ 70	373 (13.25%)	237 (12.23%)	136 (15.51%)
Gender, N(%)				0.006
Female	1,249 (44.43%)	893 A(46.17%)	356 (40.59%)
Male	1,562 (55.57%)	1,041 (53.83%)	521 (59.41%)
Education, N(%)				0.700
High school +	317 (11.26%)	214 (11.04%)	103 (11.74%)
Illiterate	603 (21.42%)	426 (21.98%)	177 (20.18%)
Middle school	632 (22.45%)	433 (22.34%)	199 (22.69%)
Primary	1,263 (44.87%)	865 (44.63%)	398 (45.38%)
Marital, N(%)				0.001
Divorced	54 (1.92%)	42 (2.17%)	12 (1.37%)
Married	2,269 (80.60%)	1,593 (82.20%)	676 (77.08%)
Unmarried	24 (0.85%)	15 (0.77%)	9 (1.03%)
Widowed	468 (16.63%)	288 (14.86%)	180 (20.52%)
Residential area, N(%)				0.010
Urban	230 (8.17%)	141 (7.28%)	89 (10.15%)
Rural	2,585 (91.83%)	1,797 (92.72%)	788 (89.85%)
Smoking, N(%)				< 0.001
Ex-smoker	307 (10.93%)	181 (9.37%)	126 (14.37%)
Non-smoker	1,569 (55.86%)	1,093 (56.57%)	476 (54.28%)
Smoker	933 (33.21%)	658 (34.06%)	275 (31.36%)
Drinking, N(%)				0.500
Drink but less than once a month	214 (7.61%)	147 (7.60%)	67 (7.64%)
Drink more than once a month	779 (27.71%)	524 (27.09%)	255 (29.08%)
None of these	1,818 (64.67%)	1,263 (65.31%)	555 (63.28%)
BMI, kg/m2	23.71 (3.94)	23.44 (3.99)	24.29 (3.78)	< 0.001
BMI_group, kg/m2 N(%)				< 0.001
< 18.5	180 (6.39%)	140 (7.22%)	40 (4.56%)
≥ 28	349 (12.40%)	212 (10.94%)	137 (15.62%)
18.5–24	1,438 (51.08%)	1,045 (53.92%)	393 (44.81%)
24–28	848 (30.12%)	541 (27.92%)	307 (35.01%)
Waist circumference,cm	85.00 (12.37)	84.03 (12.44)	87.15 (11.93)	< 0.001
SBP_baseline,mmHg	128.95 (6.13)	128.18 (6.02)	130.63 (6.05)	< 0.001
DBP_baseline,mmHg	75.97 (7.14)	75.58 (7.03)	76.83 (7.30)	< 0.001
Uric acid, mg/dL	270.06 (77.16)	266.49 (73.32)	277.95 (84.53)	0.002
WBC, × 10⁹/L	6.40 (1.88)	6.44 (1.91)	6.30 (1.82)	0.025
HGB, g/dL	14.61 (2.19)	14.46 (2.05)	14.95 (2.44)	< 0.001
PLT, × 10⁹/L	213.80 (75.27)	214.71 (78.68)	211.80 (67.13)	0.700
TG, mg/dL	136.25 (101.50)	127.95 (88.80)	154.60 (123.15)	< 0.001
FBG, mg/dL	112.30 (36.01)	110.10 (32.63)	117.18 (42.15)	< 0.001
LDL, mg/dL	116.84 (35.56)	117.37 (35.04)	115.65 (36.68)	0.700
TC, mg/dL	194.20 (37.46)	193.65 (37.33)	195.40 (37.74)	0.091
CVD, N(%)				< 0.001
No	2,425 (86.15%)	1,702 (87.82%)	723 (82.44%)
Yes	390 (13.85%)	236 (12.18%)	154 (17.56%)
DM, N(%)				< 0.001
No	2,551 (90.62%)	1,787 (92.21%)	764 (87.12%)
Yes	264 (9.38%)	151 (7.79%)	113 (12.88%)
BUN, mg/dL	15.64 (4.19)	15.40 (4.01)	16.19 (4.51)	< 0.001
Creatinine, mg/dL	0.79 (0.18)	0.78 (0.17)	0.82 (0.20)	< 0.001
Cystatin_C, mg/dL	1.00 (0.25)	1.00 (0.24)	1.00 (0.28)	0.400
Depression, N(%)				0.400
No	1,797 (63.84%)	1,248 (64.40%)	549 (62.60%)
Yes	1,018 (36.16%)	690 (35.60%)	328 (37.40%)
Life Satisfaction, N(%)				0.110
0	67 (2.38%)	54 (2.79%)	13 (1.48%)
1	380 (13.50%)	273 (14.09%)	107 (12.20%)
2	1,776 (63.09%)	1,217 (62.80%)	559 (63.74%)
3	563 (20.00%)	375 (19.35%)	188 (21.44%)
4	29 (1.03%)	19 (0.98%)	10 (1.14%)
TyG index	8.74 (0.67)	8.68 (0.64)	8.86 (0.72)	< 0.001
TyG-WC	744.87 (137.15)	731.26 (134.05)	774.96 (139.20)	< 0.001
TyG-WHtR	4.69 (0.86)	4.61 (0.84)	4.87 (0.88)	< 0.001
TyG-BMI	207.84 (42.06)	204.12 (41.46)	216.07 (42.22)	< 0.001

¹Mean (SD); n (%)

²Wilcoxon rank sum test; Pearson's Chi-squared test

Abbreviation: SBP, systolic blood pressure; DBP, diastolic blood pressure; WBC, white blood cell count; HGB, hemoglobin concentration; PLT, platelet count; TC, total cholesterol; LDL-C, low-density lipoprotein cholesterol; BUN, blood urea nitrogen; BMI, body mass index; HDL-C, high-density lipoprotein cholesterol; TG, triglycerides; FBG, fasting blood glucose; CVD, cardiovascular disease; DM, diabetes mellitus; TyG index, Triglyceride-glucose index; TyG-WHtR, TyG-Waist-to-Height Ratio; TyG-BMI, TyG-Body Mass Index; TyG-WC, TyG-Waist Circumference

Baseline characteristics by TyG and derivatives quartiles

Table 3, along with Tables S4, S5, and S6 in the supplementary materials, presents the baseline characteristics of participants classified by quartiles of the TyG, TyG-WHtR, TyG-WC, and TyG-BMI indices. In the Q4 group of the TyG index, key characteristics included: an age distribution concentrated between 50–60 years, a higher proportion of males, predominantly married marital status, and education levels primarily at the primary school level. Participants in this group were more likely to reside in rural areas. Additionally, this group exhibited higher BMI, particularly in the 24–28 range; greater waist circumference; higher DBP; elevated levels of uric acid, WBC, HGB, PLT, TG, FBG, TC, BUN, and creatinine; and a higher prevalence of CVD and DM. However, no significant differences were observed across groups in terms of alcohol consumption, SBP, depression, and life satisfaction (all P > 0.05). Similar trends were identified in the quartile groups of TyG-WC, TyG-WHtR, and TyG-BMI, indicating that these composite indices are strongly linked to various health risk factors.

Table 3.

The baseline characteristics stratified by TyG index quartiles

Characteristic	Overall N = 2,8151	Q1 N = 6521	Q2 N = 6761	Q3 N = 7511	Q4 N = 7361	p-value2
Age, years	59.72 (9.46)	60.13 (10.15)	60.32 (8.98)	59.68 (9.83)	58.85 (8.80)	0.012
Age_group, N(%)						< 0.001
< 50	567 (20.14%)	156 (23.93%)	108 (15.98%)	150 (19.97%)	153 (20.79%)
≥ 70	373 (13.25%)	101 (15.49%)	78 (11.54%)	109 (14.51%)	85 (11.55%)
50–60	998 (35.45%)	191 (29.29%)	238 (35.21%)	279 (37.15%)	290 (39.40%)
60–70	877 (31.15%)	204 (31.29%)	252 (37.28%)	213 (28.36%)	208 (28.26%)
Gender, N(%)						< 0.001
Female	1,249 (44.43%)	230 (35.28%)	285 (42.41%)	371 (49.40%)	363 (49.32%)
Male	1,562 (55.57%)	422 (64.72%)	387 (57.59%)	380 (50.60%)	373 (50.68%)
Education, N(%)						0.024
High school +	317 (11.26%)	86 (13.19%)	75 (11.09%)	83 (11.05%)	73 (9.92%)
Illiterate	603 (21.42%)	134 (20.55%)	148 (21.89%)	178 (23.70%)	143 (19.43%)
Middle school	632 (22.45%)	149 (22.85%)	129 (19.08%)	157 (20.91%)	197 (26.77%)
Primary	1,263 (44.87%)	283 (43.40%)	324 (47.93%)	333 (44.34%)	323 (43.89%)
Marital, N(%)						< 0.001
Divorced	54 (1.92%)	20 (3.07%)	19 (2.81%)	3 (0.40%)	12 (1.63%)
Married	2,269 (80.60%)	514 (78.83%)	549 (81.21%)	597 (79.49%)	609 (82.74%)
Unmarried	24 (0.85%)	11 (1.69%)	3 (0.44%)	5 (0.67%)	5 (0.68%)
Widowed	468 (16.63%)	107 (16.41%)	105 (15.53%)	146 (19.44%)	110 (14.95%)
Residential area, N(%)						< 0.001
Urban	230 (8.17%)	43 (6.60%)	36 (5.33%)	74 (9.85%)	77 (10.46%)
Rural	2,585 (91.83%)	609 (93.40%)	640 (94.67%)	677 (90.15%)	659 (89.54%)
Smoking, N(%)						< 0.001
Ex-smoker	307 (10.93%)	73 (11.20%)	58 (8.58%)	91 (12.12%)	85 (11.64%)
Non-smoker	1,569 (55.86%)	318 (48.77%)	366 (54.14%)	423 (56.32%)	462 (63.29%)
Smoker	933 (33.21%)	261 (40.03%)	252 (37.28%)	237 (31.56%)	183 (25.07%)
Drinking, N(%)						0.200
Drink but less than once a month	214 (7.61%)	58 (8.90%)	52 (7.69%)	48 (6.39%)	56 (7.65%)
Drink more than once a month	779 (27.71%)	198 (30.37%)	189 (27.96%)	198 (26.36%)	194 (26.50%)
None of these	1,818 (64.67%)	396 (60.74%)	435 (64.35%)	505 (67.24%)	482 (65.85%)
BMI, kg/m2	23.71 (3.94)	22.49 (4.19)	23.07 (3.77)	23.81 (3.73)	25.27 (3.55)	< 0.001
BMI_group, kg/m2 N(%)						< 0.001
< 18.5	180 (6.39%)	63 (9.66%)	58 (8.58%)	39 (5.19%)	20 (2.72%)
≥ 28	349 (12.40%)	32 (4.91%)	58 (8.58%)	108 (14.38%)	151 (20.52%)
18.5–24	1,438 (51.08%)	421 (64.57%)	379 (56.07%)	385 (51.26%)	253 (34.38%)
24–28	848 (30.12%)	136 (20.86%)	181 (26.78%)	219 (29.16%)	312 (42.39%)
Waist circumference,cm	85.00 (12.37)	81.09 (11.09)	82.77 (13.29)	85.58 (11.67)	89.93 (11.51)	< 0.001
SBP_baseline,mmHg	128.95 (6.13)	128.34 (6.32)	129.04 (6.13)	129.13 (5.99)	129.21 (6.09)	0.110
DBP_baseline,mmHg	75.97 (7.14)	74.61 (7.02)	76.53 (7.50)	75.66 (6.95)	76.98 (6.87)	< 0.001
Uric acid, mg/dL	270.06 (77.16)	257.62 (69.98)	263.32 (76.58)	270.31 (76.59)	287.01 (81.34)	< 0.001
WBC, × 10⁹/L	6.40 (1.88)	6.17 (1.81)	6.18 (1.73)	6.45 (2.07)	6.74 (1.83)	< 0.001
HGB, g/dL	14.61 (2.19)	14.31 (1.94)	14.54 (2.13)	14.60 (2.37)	14.97 (2.20)	< 0.001
PLT, × 10⁹/L	213.80 (75.27)	206.46 (66.00)	210.55 (76.87)	213.76 (79.62)	223.34 (76.10)	< 0.001
TG, mg/dL	136.25 (101.50)	62.56 (15.11)	90.39 (14.42)	130.58 (25.03)	249.45 (137.08)	< 0.001
FBG, mg/dL	112.30 (36.01)	95.25 (14.01)	103.10 (14.16)	109.63 (23.04)	138.60 (55.04)	< 0.001
LDL, mg/dL	116.84 (35.56)	110.04 (32.01)	120.75 (31.54)	120.73 (34.35)	115.30 (41.75)	< 0.001
TC, mg/dL	194.20 (37.46)	179.01 (36.18)	192.99 (32.95)	193.82 (35.62)	209.14 (38.61)	< 0.001
CVD, N(%)						< 0.001
No	2,425 (86.15%)	590 (90.49%)	587 (86.83%)	646 (86.02%)	602 (81.79%)
Yes	390 (13.85%)	62 (9.51%)	89 (13.17%)	105 (13.98%)	134 (18.21%)
DM, N(%)						< 0.001
No	2,551 (90.62%)	636 (97.55%)	640 (94.67%)	693 (92.28%)	582 (79.08%)
Yes	264 (9.38%)	16 (2.45%)	36 (5.33%)	58 (7.72%)	154 (20.92%)
BUN, mg/dL	15.64 (4.19)	16.56 (4.60)	15.42 (4.25)	15.10 (3.96)	15.59 (3.84)	< 0.001
Creatinine, mg/dL	0.79 (0.18)	0.80 (0.17)	0.77 (0.17)	0.79 (0.20)	0.80 (0.18)	0.016
Cystatin_C, mg/dL	1.00 (0.25)	1.04 (0.25)	1.04 (0.23)	1.01 (0.26)	0.92 (0.25)	< 0.001
Depression, N(%)						0.500
No	1,797 (63.84%)	419 (64.26%)	416 (61.54%)	489 (65.11%)	473 (64.27%)
Yes	1,018 (36.16%)	233 (35.74%)	260 (38.46%)	262 (34.89%)	263 (35.73%)
Life Satisfaction, N(%)						0.065
0	67 (2.38%)	16 (2.45%)	18 (2.66%)	15 (2.00%)	18 (2.45%)
1	380 (13.50%)	87 (13.34%)	92 (13.61%)	110 (14.65%)	91 (12.36%)
2	1,776 (63.09%)	396 (60.74%)	424 (62.72%)	489 (65.11%)	467 (63.45%)
3	563 (20.00%)	151 (23.16%)	129 (19.08%)	128 (17.04%)	155 (21.06%)
4	29 (1.03%)	2 (0.31%)	13 (1.92%)	9 (1.20%)	5 (0.68%)

¹Mean (SD); n (%)

²Kruskal-Wallis rank sum test; Pearson's Chi-squared test

Abbreviation: SBP, systolic blood pressure; DBP, diastolic blood pressure; WBC, white blood cell count; HGB, hemoglobin concentration; PLT, platelet count; TC, total cholesterol; LDL-C, low-density lipoprotein cholesterol; BUN, blood urea nitrogen; BMI, body mass index; HDL-C, high-density lipoprotein cholesterol; TG, triglycerides; FBG, fasting blood glucose; CVD, cardiovascular disease; DM, diabetes mellitus

Logistic regression results for TyG indicators

Table 4 shows the link between the TyG index, its related indicators, and new-onset hypertension. Multivariable logistic regression analysis revealed significant positive associations between these continuous variables and new-onset hypertension. In the fully adjusted Model 4, the TyG index showed the strongest association, with a 45% increased risk of new-onset hypertension for each unit increase (odds ratio [OR]: 1.45, 95% confidence intervals [CI] 1.24–1.70, P < 0.001), followed by the TyG-WHtR index (OR: 1.42, 95% CI 1.26–1.60, P < 0.001). Further categorization of these indicators by quartiles revealed that higher quartiles of the TyG index (Q2, Q3, and Q4) were linked to a higher risk of new-onset hypertension across all four models. However, for TyG-WHtR, TyG-WC, and TyG-BMI indices, only Q3 and Q4 showed significant associations with new-onset hypertension across all four models (Table 4). Specifically, in Model 4, the highest risk of new-onset hypertension was observed in the Q4 groups: TyG Q4 (OR: 1.62, 95% CI 1.23–2.16, P < 0.001), TyG-WC Q4 (OR: 2.31, 95% CI 1.76–3.04, P < 0.001), TyG-WHtR Q4 (OR: 1.99, 95% CI 1.52–2.61, P < 0.001), and TyG-BMI Q4 (OR: 1.90, 95% CI 1.45–2.49, P < 0.001), compared to the Q1 group. In addition, in both unadjusted and fully adjusted models, the TyG index demonstrated modestly higher AUC values compared to FBG and TG, suggesting its potentially superior value in predicting hypertension risk (Supplementary Figure S3). These findings indicate that as levels of the TyG index and its related indicators increase, so does the risk of new-onset hypertension, suggesting that they could be important tools for assessing and predicting the risk of new-onset hypertension.

Table 4.

TyG index its related indicators and new-onset hypertension multivariate logistic regression models

Characteristic	Model 1			Model 2			Model 3			Model 4
	OR	95%CI	P	OR	95%CI	P	OR	95%CI	P	OR	95%CI	P
TyG index
Continuous	1.5	1.33, 1.69	< 0.001	1.54	1.37, 1.74	< 0.001	1.47	1.29, 1.66	< 0.001	1.45	1.24, 1.70	< 0.001
Quartile
Q1	1 (Reference)			1 (Reference)			1 (Reference)			1 (Reference)
Q2	1.42	1.11, 1.82	0.005	1.42	1.11, 1.82	0.005	1.36	1.06, 1.76	0.016	1.47	1.13, 1.90	0.004
Q3	1.66	1.31, 2.12	< 0.001	1.66	1.31, 2.12	< 0.001	1.56	1.22, 1.99	< 0.001	1.63	1.26, 2.11	< 0.001
Q4	1.95	1.54, 2.48	< 0.001	1.95	1.54, 2.48	< 0.001	1.81	1.42, 2.32	< 0.001	1.62	1.23, 2.16	< 0.001
TyG-WC
Continuous	1	1.00, 1.00	< 0.001	1	1.00, 1.00	< 0.001	1	1.00, 1.00	< 0.001	1	1.00, 1.00	< 0.001
Quartile
Q1	1 (Reference)			1 (Reference)			1 (Reference)			1 (Reference)
Q2	1.42	1.10, 1.84	0.007	1.41	1.09, 1.83	0.01	1.24	0.95, 1.62	0.11	1.21	0.92, 1.59	0.2
Q3	2.01	1.58, 2.57	< 0.001	2.13	1.67, 2.74	< 0.001	1.9	1.47, 2.45	< 0.001	1.89	1.46, 2.47	< 0.001
Q4	2.53	2.00, 3.22	< 0.001	2.8	2.19, 3.58	< 0.001	2.52	1.95, 3.25	< 0.001	2.31	1.76, 3.04	< 0.001
TyG-WHtR
Continuous	1.43	1.30, 1.58	< 0.001	1.55	1.39, 1.72	< 0.001	1.47	1.32, 1.63	< 0.001	1.42	1.26, 1.60	< 0.001
Quartile
Q1	1 (Reference)			1 (Reference)			1 (Reference)			1 (Reference)
Q2	1.47	1.16, 1.88	0.002	1.49	1.17, 1.91	0.002	1.28	1.0, 1.65	0.055	1.24	0.96, 1.61	0.11
Q3	1.58	1.24, 2.02	< 0.001	1.72	1.34, 2.20	< 0.001	1.51	1.17, 1.96	0.002	1.47	1.13, 1.92	0.005
Q4	2.13	1.69, 2.69	< 0.001	2.51	1.97, 3.21	< 0.001	2.21	1.72, 2.85	< 0.001	1.99	1.52, 2.61	< 0.001
TyG-BMI
Continuous	1.01	1.00, 1.01	< 0.001	1.01	1.01, 1.01	< 0.001	1.01	1.01, 1.01	< 0.001	1.01	1.00, 1.01	< 0.001
Quartile
Q1	1 (Reference)			1 (Reference)			1 (Reference)			1 (Reference)
Q2	0.86	0.67, 1.11	0.2	0.94	0.73, 1.21	0.6	0.89	0.68, 1.16	0.4	0.83	0.64, 1.09	0.2
Q3	1.66	1.31, 2.11	< 0.001	1.87	1.47, 2.40	< 0.001	1.71	1.33, 2.21	< 0.001	1.65	1.27, 2.16	< 0.001
Q4	1.98	1.57, 2.49	< 0.001	2.41	1.89, 3.07	< 0.001	2.14	1.66, 2.76	< 0.001	1.90	1.45, 2.49	< 0.001

Model 1: Unadjusted

Model 2: Adjusted for sex, age, and residential area

Model 3: Further adjusted for education level, marital status, smoking and alcohol consumption habits, baseline systolic blood pressure (SBP), and diastolic blood pressure (DBP)

Model 4: Additionally adjusted for white blood cell count (WBC), hemoglobin (HGB), platelet count (PLT), total cholesterol (TC), low-density lipoprotein (LDL), uric acid, creatinine, cystatin C, blood urea nitrogen (BUN), as well as comorbidities including cardiovascular disease (CVD) and diabetes mellitus (DM)

RCS models: TyG indicators and new-onset hypertension

The unadjusted RCS models showed a significant positive linear connection between the TyG index (Fig. 2A) and new-onset hypertension (P for overall < 0.001, P for nonlinearity = 0.804), while the TyG-WC index (Fig. 2B) also exhibited a similar positive linear relationship (P for overall < 0.001, P for nonlinearity = 0.283). However, the TyG-WHtR index (Fig. 2C) and TyG-BMI index (Fig. 2D) demonstrated significant positive nonlinear relationships (P for overall < 0.001, P for nonlinearity = 0.040 and P for overall < 0.001, P for nonlinearity = 0.032, respectively).

Fig. 2.

Restricted cubic spline models

After adjusting for variables such as gender, age, residential area, marital status, education level, smoking and alcohol consumption habits, baseline SBP and DBP, WBC, HGB, PLT, TC, LDL-C, uric acid, creatinine, cystatin C, and BUN, the conclusions remained consistent. Specifically, the TyG index (Fig. 2E) and TyG-WC index (Fig. 2F) still exhibited positive linear relationships (P for overall < 0.001, P for nonlinearity = 0.844 and P for overall < 0.001, P for nonlinearity = 0.165, respectively), while the TyG-WHtR index (Fig. 2G) and TyG-BMI index (Fig. 2H) continued to show significant positive nonlinear relationships (P for overall < 0.001, P for nonlinearity = 0.001 and P for overall < 0.001, P for nonlinearity = 0.046, respectively). This nonlinear but monotonically increasing relationship suggests that individuals with higher TyG-WHtR and TyG-BMI indices require special attention in the assessment and management of prehypertensive populations.

Subgroup analysis: TyG indicators and new-onset hypertension

Subgroup analyses were performed according to gender, age, education level, residential area, marital status, alcohol consumption, smoking status, BMI, CVD, DM, depression, and life satisfaction, with covariates adjusted. Figure 3 presents the subgroup analysis results of the association between the TyG index and new-onset hypertension. The findings revealed that the TyG index exhibited a stronger association with new-onset hypertension in individuals aged ≥ 70 years, widowed status, those with CVD, and those with a life satisfaction score of 2 (P < 0.05, P for interaction < 0.05). However, the relationship between the TyG index and new-onset hypertension was not influenced by gender, education level, residential area, alcohol consumption, BMI, DM, or depression status, as the interactions between these variables and the TyG index were not statistically significant (P for interaction > 0.05). These results indicate that the impact of the TyG index on new-onset hypertension is more pronounced in specific high-risk populations, while the association remains consistent across other differing factors.

Fig. 3.

TyG index and new-onset hypertension subgroup analysis

Supplementary Tables S7, S8, and S9 detail the subgroup analysis results of the associations between the TyG-WC, TyG-WHtR, and TyG-BMI indices and new-onset hypertension, revealing consistent findings. The TyG-WC index's association with new-onset hypertension was not significantly influenced by gender, marital status, residential area, smoking, alcohol consumption, BMI, or DM (P for interaction > 0.05), but a stronger correlation was observed among individuals aged ≥ 70 years, those with CVD, depression, and life satisfaction scores of 2 and 4 (P < 0.05, P for interaction < 0.05). Similarly, the TyG-WHtR index's relationship with new-onset hypertension remained unaffected by gender, residential area, marital status, smoking, alcohol consumption, BMI, or DM (P for interaction > 0.05), yet it exhibited a stronger association in individuals aged ≥ 70 years, those who were illiterate, had CVD, depression, and a life satisfaction score of 4 (P < 0.05, P for interaction < 0.05). For the TyG-BMI index, its association with new-onset hypertension was not impacted by age, residential area, education level, marital status, alcohol consumption, DM, or depression status (P for interaction > 0.05), but it showed a stronger correlation in individuals with a BMI in the range of 18.5–24 (P < 0.05, P for interaction < 0.05). Taken together, these findings suggest that the impact of TyG-related indices on new-onset hypertension is particularly significant among certain high-risk populations, while remaining relatively stable across different levels of other variables. This highlights their potential value in identifying and managing individuals who are at an increased risk of developing hypertension within specific subgroups.

Genetic evidence: TyG and hypertension

Figures 4 and 5 present robust genetic evidence of significant associations between the TyG index, SBP, DBP, and hypertension. In Fig. 5, IVW P-values for ukb-b-14057, ebi-a-GCST90038604, and ukb-a-61 were all below 0.001. Similarly, Fig. 5 shows that the corresponding IVW P-values for SBP (ieu-b-38 and ebi-a-GCST90025968) and DBP (ieu-b-39 and ebi-a-GCST90025981) were also less than 0.001. To account for the increased risk of false positives due to multiple testing across various models, we applied the FDR correction using the Benjamini–Hochberg method to all primary associations derived from MR analyses. After adjustment, most of the significant associations remained statistically robust, indicating the reliability of our findings. A summary of the original and FDR-adjusted p-values is provided in Supplementary Table S10. Supplementary Table S11 summarizes the SNP-level metrics, including effect sizes, standard errors, F-statistics, and R² values, for the TyG-blood pressure associations derived from seven independent GWAS datasets. To further validate these findings, we conducted additional analyses using MR-PRESSO and MR-Egger on the involved SNPs (Supplementary Table S12). The funnel plot showed no evidence of bias in this study (Supplementary Figure S4). To examine the contribution of each SNP to the overall causal relationship in detail, we conducted a leave-one-out sensitivity analysis (Supplementary Figure S5), systematically excluding each SNP to assess its impact. The results demonstrated that even after excluding individual SNPs, the causal pattern remained stable, indicating that the overall effect was not driven by any single genetic variant.

Fig. 4.

The scatter plot for the Mendelian randomization analyses of causal associations

Fig. 5.

Forest Plot of Genetic Association Models

To enhance the robustness of our findings, we extended the IVW analysis to the BWMR method. BWMR provides a more detailed assessment of uncertainty and sensitivity by reducing pleiotropy and integrating multiple sources of evidence. The BWMR analysis confirmed a significant association (all P-values < 0.05, Fig. 5), further supporting our conclusions. These findings suggest that higher TyG index levels are significantly associated with an increased risk of new-onset hypertension, highlighting the potential importance of the TyG index in hypertension prevention.

Discussion

Based on the CHARLS database, this study conducted a four-year cohort study and MR analysis to thoroughly evaluate the link between the TyG index and its related measures (TyG-WHtR, TyG-WC, TyG-BMI) and new-onset hypertension in individuals predisposed to hypertension. The results showed that in multivariable logistic regression models, for each unit increase in the TyG index, the probability of developing new-onset hypertension increased by 45% (OR: 1.45, 95% CI 1.24–1.70, P < 0.001), while for the TyG-WHtR index, this increase was 42% (OR: 1.42, 95% CI 1.26–1.60, P < 0.001). There was a significant positive correlation between the TyG index and its related indicators with new-onset hypertension. The RCS analysis revealed a significant positive linear connection between the TyG index, TyG-WC, and new-onset hypertension. In contrast, TyG-WHtR and TyG-BMI showed significant positive nonlinear relationships. These results highlight the importance of considering the TyG index and its related indicators in assessing new-onset hypertension risk, particularly in high-risk populations.

Previous research has revealed that people with prehypertension face a two- to three-fold increased risk of developing hypertension compared to those with normal blood pressure. [30–32]. In our study of 2815 participants with a mean age of 59.72 years, 877 individuals (31.15%) were identified with new hypertension after four years. This observation corroborates the Framingham Study's findings, where 30.3% of prehypertensive adults developed hypertension within the same period [33]. These results emphasize the importance of monitoring prehypertensive status across different populations and ethnicities. Previous studies exploring the link between the TyG index and hypertension has yielded inconsistent results. For instance, Song et al. (2017) demonstrated a significant positive connection between the TyG index and the risk of hypertension through a large cross-sectional study [34], whereas Liu et al. (2019) found no significant link between the TyG index and hypertension in normal-weight Chinese adults; instead, they identified metabolic score IR as an effective indicator associated with hypertension [18]. Overall, despite ongoing debate about the TyG index, most existing studies support its positive association with new-onset hypertension [19, 20, 34, 35]. These variations may arise from differences in study design, sample characteristics, and statistical methods. Notably, most previous studies relied primarily on traditional observational designs, which may introduce biases due to reverse causality or unmeasured confounding factors. In contrast, we employed BWMR analysis, which effectively addresses reverse causality issues and reduces the influence of unmeasured confounders, providing additional support for our conclusions [28]. Furthermore, using multiple models and RCS analysis, our findings are consistent with prior research on the relationship between prehypertension and new-onset hypertension. Our study demonstrates that, after extensive adjustment for multiple confounding factors, both the TyG index and its related indicators significantly predicted new-onset hypertension. Subgroup analyses revealed that, in most cases, the relationship between TyG and its related indices with new-onset hypertension was not influenced by gender, education level, residential area, alcohol consumption, BMI, DM, or depression status. This indicates that the TyG index and its related indicators exhibit broad applicability and robustness as tools for predicting hypertension risk. Importantly, our study demonstrates that multiple TyG-derived metrics, including TyG-WHtR, TyG-WC, and TyG-BMI, also have significant predictive power. This association is particularly pronounced in specific high-risk subgroups. These results suggest that the TyG index and its related indicators can serve as effective tools for risk stratification in prehypertensive individuals, especially among the elderly, widowed, or those with CVD. Specifically, stronger associations were observed in subgroups such as individuals aged ≥ 70 years, widowed individuals, and those with CVD. Therefore, for elderly individuals or patients with a history of CVD who fall into the highest quartile (Q4) of the TyG index, healthcare providers could monitor TyG and its derived metrics to implement more proactive lifestyle interventions, such as adjustments to diet, increased exercise levels, and weight management strategies, and even consider pharmacological treatment when necessary to prevent the onset and progression of hypertension [36–38]. Through this comprehensive approach, we not only validated the link between the TyG index and its related indicators with hypertension risk but also provided new insights into how these biomarkers influence the development of hypertension.

According to the findings of this study, we speculate that the association between the TyG index and its related indicators with new-onset hypertension may be explained by multiple biological mechanisms. This is particularly relevant given that the TyG index serves as a marker of IR [39]. As a key driver of metabolic syndrome, IR not only induces dysfunction in glucose and lipid metabolic pathways but also triggers endothelial impairment, promotes inflammatory reactions, and activates the renin-angiotensin system (RAS). These mechanisms collectively play crucial roles in promoting the development of hypertension [39, 40]. IR, as reflected by the TyG index, contributes to hypertension through several pathways. IR promotes dyslipidemia and hyperglycemia, leading to increased oxidative stress and inflammation. These metabolic disturbances impair endothelial function by reducing nitric oxide availability, a key mediator of vascular relaxation, and by promoting vasoconstriction and arterial stiffness. Endothelial dysfunction is central to elevated blood pressure pathogenesis. Therefore, the association between higher TyG values and increased hypertension risk may be explained by these underlying metabolic and vascular mechanisms, reinforcing TyG's importance as a marker of both metabolic and cardiovascular health. Specifically, IR can elevate blood pressure by increasing sympathetic nervous system activity, reducing sodium excretion, and regulating intracellular calcium ion concentrations [39]. Additionally, studies have shown that IR can induce oxidative stress, further impairing endothelial function and promoting the progression of arteriosclerosis [41]. Central obesity, a key component of TyG-WHtR and TyG-WC, is closely related to the development of systemic inflammation, oxidative stress, and IR, all of which collectively accelerate the transition from prehypertension to hypertension [41, 42]. Jin et al. (2025) found that WC, BMI, lipid accumulation product (LAP), Chinese visceral adiposity index (CVAI), and TyG were all significantly linked to type 2 diabetes risk in elderly Chinese patients with hypertension, with TyG showing the strongest predictive value [43]. Cui et al. (2025) found that both the TyG index and atherogenic index of plasma (AIP) were significantly associated with coronary microvascular dysfunction (CMD), and their combined use improves clinical decision-making [44]. Wan et al. (2025) showed that elevated hsCRP and TyG levels were significantly linked to higher risk of cardiometabolic multimorbidity in middle-aged and older adults, suggesting their combined use may improve early prevention [45]. Notably, the TROPHY study demonstrated that a two-year treatment with moderate-dose renin-angiotensin system inhibitors led to a 66% reduction in the relative risk of new-onset hypertension compared to the placebo group. This protective effect persisted even two years post-treatment (total follow-up of four years), reducing the relative risk by approximately 15% [46, 47]. These results support the theory that IR and the RAS are crucial in the development of hypertension. Consequently, interventions targeting IR and related metabolic disorders—such as lifestyle modifications (dietary modifications, increased physical activity), weight control, and pharmacological treatment when necessary—are effective strategies for preventing progression from prehypertension to hypertension [38, 48]. These measures can not only improve insulin sensitivity but also mitigate the impact of other cardiovascular risk factors, thereby helping to reduce the incidence of hypertension. Notably, the stronger associations observed in specific subgroups may reflect complex physiological mechanisms and psychosocial factors within different population contexts [49, 50]. For example, elderly individuals and those with a history of CVD may be more susceptible to the effects of IR and other metabolic disorders, thereby increasing their risk of hypertension [51, 52]. Psychosocial factors such as widowhood or depression may also indirectly influence blood pressure control by affecting lifestyle choices and mental health status [50, 53]. Therefore, conducting more detailed risk assessments and implementing personalized interventions for these high-risk subgroups is particularly important.

This study is the first to systematically investigate the association between the TyG index and its related indicators (including TyG-WC, TyG-WHtR, and TyG-BMI) with new-onset hypertension by integrating cohort study methods with MR. This comprehensive causal analysis provides a novel perspective on assessing new-onset hypertension. We not only validated the effectiveness of the TyG index as a risk prediction tool but also extended its application to derived metrics. Our research employed multiple statistical methods, including logistic regression, RCS curves, subgroup analyses, and interaction assessments, constructing a thorough evaluation model. These varied approaches strengthen the reliability and thoroughness of our findings. Furthermore, using the CHARLS database for a large-scale cohort study helped minimize selection bias and ensured our sample was representative. Meanwhile, the MR method, using large datasets from the UK Biobank and GWAS databases, helped address issues of reverse causality and potential false associations due to unmeasured confounders. In the MR analysis, all genetic variants used had F-statistics greater than 10, indicating strong instrument variables, which further increases the credibility of causal inference. Moreover, validation was further conducted using BWMR, consolidating the robustness of our conclusions. Given the widespread prevalence and high incidence of prehypertension globally, this study holds significant clinical implications. The TyG index's simplicity in calculation and freedom from time or cost constraints make it highly suitable for large-scale population screenings. It serves as an accurate risk stratification tool for individuals with prehypertension, aiding in the early identification and intervention of new-onset hypertension. By monitoring the TyG index and related metrics, high-risk individuals can be identified early, enabling preventive measures that reduce disease progression and complications, improve patient quality of life, and alleviate socioeconomic burdens. This approach offers valuable insights and practical strategies for managing and preventing hypertension at both individual and public health levels.

This study has several limitations that should be considered when interpreting the results. First, although the data from the CHARLS are relatively representative of middle-aged and older adults in China, they may not fully reflect the variability in risk factors or demographic characteristics associated with new-onset hypertension in other populations or countries. Second, the study only included participants aged 45 years or older, which limits the generalizability of our findings to younger age groups. Additionally, some data were collected through self-reported questionnaires, which may introduce potential recall bias and affect the accuracy of the information obtained. Third, due to limitations in the CHARLS database, we were unable to incorporate certain important behavioral and lifestyle factors—particularly detailed dietary sodium intake—into our analyses. Furthermore, fasting insulin measurements were not available, preventing us from calculating the Homeostatic Model Assessment of IR (HOMA-IR). This limited our ability to directly compare its predictive performance with that of the TyG index and its components. Moreover, while we focused on the overall association between the TyG index and incident hypertension, further research is needed to explore its relationship across different stages of prehypertension (such as stage 1 vs. stage 2), which could provide more refined stratification and clinical guidance for early intervention. Finally, although we discussed plausible biological mechanisms linking insulin resistance to hypertension, further basic and translational research is necessary to better understand the underlying pathophysiological pathways, particularly those relevant to the prevention and management of prehypertension. By addressing these limitations and expanding on these areas in future studies, more comprehensive insights and practical recommendations can be provided for the early detection, prevention, and treatment of hypertension.

Conclusion

This study underscores the role of the TyG index and related indicators in the development of new-onset hypertension. The TyG index and TyG-WC showed a significant positive linear association, while TyG-BMI and TyG-WHtR demonstrated significant positive nonlinear associations. In high-risk subgroups such as individuals aged ≥ 70 years, those who are widowed, or those with CVD, the association with the TyG index is even stronger. Notably, causal analysis using Bayesian weighted Mendelian randomization confirmed that genetically elevated TyG index levels are causally linked to an increased risk of new-onset hypertension.

Author contributions

M.W. had full data access and ensured the accuracy and integrity of the analysis. M.W., T.T., and N.Z. (co-first authors) equally contributed to the study's execution, data collection, analysis, interpretation, and initial drafting. J.X., Z.D., and Q.J. aided in data collection and analysis. H.Y., N.Z., and M.W. conceived the study and reviewed the manuscript. All authors approved the final manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (Grant ID: 82200401) and by the Qingdao Municipal Science and Technology Bureau through its Science and Technology Beneficial Demonstration Guidance Special Fund project (Award Number: 20-3-4-54-nsh), with HaiChu Yu as the project leader for the latter.

Data availability

The datasets generated and analyzed in this study are accessible through the CHARLS database (http://charls.pku.edu.cn/) and the GWAS summary data repository at https://gwas.mrcieu.ac.uk/.

Declarations

Ethical approval

Not applicable.

Conflict of interest

The authors declare no competing interests.