Sex-specific performance of composite metabolic-adiposity indices in identifying hypertension and carotid atherosclerosis: a cross-sectional study of 154,873 Chinese adults Authors: Xiaoqian Dong, Xue He, Yinglong Duan, Sha Wang, Xingxing Wang, Nandan Chen, Zitong Lu, Ying Li, Lijun Li, Honglian Zhang Published: 2026 ABSTRACT The comparative superiority among the triglyceride-glucose (TyG) index and its obesity-related composite indicators in identifying hypertension and carotid atherosclerosis (CAS) remains unknown. This study aims to evaluate and compare the discriminative ability of these indices in a large-scale population to identify the optimal marker for cardiovascular risk assessment. This cross-sectional study included 154,873 participants who underwent routine physical examinations and carotid ultrasonography. Binary logistic regression, restricted cubic spline (RCS), receiver operating characteristic (ROC) analyses, and mediation analyses were used to evaluate the associations of TyG, TyG-waist circumference (TyG-WC), TyG-waist-to-height ratio (TyG-WHtR), and TyG-body mass index (TyG-BMI) with hypertension and CAS. The mean age of participants was 45.7 ± 12.5 years, with 42.1% being female and 91.0% being married. A total of 16,235(10.5%) participants had hypertension, and 81,337 (52.5%) had CAS. Comparing the highest to the lowest quartile, TyG-WHtR demonstrated the strongest association with hypertension (OR = 8.644, 95% CI: 7.927–9.426) and CAS (OR = 2.173, 95% CI: 2.086–2.263). RCS analyses revealed monotonic increases for hypertension and complex nonlinear patterns for CAS. Stratified analyses revealed distinct phenotypic signatures. For hypertension, risk magnitudes were markedly amplified in younger adults (≤ 60 years) (P for interaction < 0.001 for all indices). For CAS, a significant interaction between sex and the indices was observed (P for interaction < 0.05 for TyG, TyG-WHtR, and TyG-BMI): TyG-WHtR exhibited the strongest association in women (OR = 1.504), whereas in men, the isolated TyG index was inversely associated with CAS, and only TyG-BMI showed a significant positive association (OR = 1.091). Mediation analyses revealed that obesity indicators (WHtR, BMI) demonstrated substantial direct associations with on outcomes, with TyG providing negligible or partial mediation, which may account for the stronger associations observed with composite indices. TyG-obesity composite indices are strongly associated with hypertension and CAS, particularly in younger adults. TyG-WHtR is optimal for women and the general population, while TyG-BMI is superior for men. These easily accessible and cost-effective metrics can serve as practical tools for early cardiovascular risk screening in primary care. Implementing such age-targeted and sex-specific risk stratification could facilitate more precise and timely preventive interventions.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-50379-4. Introduction Cardiovascular disease (CVD) remains the leading cause of mortality and global disease burden, with its prevalence and associated deaths rising steadily over the past three decades1–3. Globally, the CVD burden has risen significantly over the past 30 years and is projected to continue increasing in the coming decades4. As the country with the highest number of CVD deaths, China faces escalating cardiometabolic risks driven by rapid population aging1, making CVD the foremost cause of mortality in both rural and urban areas5. Among CVDs, stroke, ischemic heart disease, and hypertensive heart disease remain the top contributors to global and Chinese CVD mortality3. Carotid atherosclerosis (CAS), the leading cause of ischemic stroke, is highly prevalent among adults and accounts for a substantial proportion of stroke-related deaths6–10. In addition, hypertension can increase the burden of the heart, accelerate the process of arteriosclerosis, and lead to functional or organic damage in the brain, heart, and other vital organs, significantly increasing the risk of adverse cardiovascular events11–13. Despite imposing a substantial economic burden14–16, adequate blood pressure control remains persistently low globally17. Collectively, these data highlight the considerable CAS and hypertension burden and the imperative for early assessment and preventive strategies to slow CVD progression. Abnormal glucose and lipid metabolism are highly prevalent in CVD patients, while insulin resistance (IR) plays a vital role in this biological process18,19. IR is closely associated with endothelial dysfunction, oxidative stress, and arterial stiffness, making it a key contributing factor for the onset and progression of hypertension and CAS20–22. However, the hyperinsulinemic-euglycemic clamp, the gold standard for assessing IR, is invasive, time-consuming, and resource-intensive, limiting its clinical application. Surrogate indices like the homeostasis model assessment of IR (HOMA-IR) and the quantitative insulin sensitivity check index (QUICKI) require fasting insulin measurements, which are not routinely performed in large-scale screenings23. Recently, the triglyceride-glucose (TyG) index has emerged as a simple, reliable, and cost-effective surrogate marker for IR24,25. Mechanistically, IR in adipose tissue increases free fatty acid flux to the liver, promoting triglyceride (TG) synthesis, while hepatic IR impairs glycogen synthesis and increases fasting blood glucose (FBG) output26. By combining these two easily accessible biomarkers, the TyG index effectively reflects this metabolic dysfunction. Clinical studies have demonstrated that the TyG index correlates highly with both the hyperinsulinemic-euglycemic clamp and HOMA-IR, showing comparable or even superior diagnostic accuracy for IR27,28. Further, accumulating evidence, including recent comprehensive reviews and meta-analyses, has highlighted that the TyG index is not only associated with the risk and severity of CVD but also serves as a reliable marker for subclinical atherosclerosis, such as increased carotid intima-media thickness (CIMT) and arterial stiffness29–33. This association has been consistently observed across various populations, including patients with ischemic stroke34. Obesity is closely linked to impaired glucose tolerance, IR, and other metabolic disorders, all of which contribute to the onset, progression, and prognosis of CVD35–37. To enhance risk prediction, derivatives of the TyG index incorporating obesity-related metrics, such as body mass index (BMI), waist circumference (WC), and the waist-to-height ratio (WHtR), have been proposed (TyG-BMI, TyG-WC, TyG-WHtR). The rationale for these combinations lies in the fact that visceral adiposity acts as a metabolically active endocrine organ. It secretes pro-inflammatory adipokines and free fatty acids that may exacerbate endothelial dysfunction and arterial stiffness via pathways distinct from or synergistic with systemic IR captured by TyG alone38,39. Specifically regarding the choice of adiposity metrics, while the waist-to-hip ratio has been traditionally used to assess central obesity, WHtR has emerged as a superior and more universal screening tool. Unlike WHR, which can be confounded by proportional increases in both waist and hip circumferences, WHtR adjusts central adiposity for stature, providing a more accurate reflection of visceral fat accumulation across different ages, sexes, and ethnicities40,41. Prior studies suggest that TyG-WC may show the strongest association with CVD risk42, whereas moderate or high TyG-WHtR is linked to elevated CVD risk and to all-cause and cardiovascular mortality43,44. However, current research has largely focused on single outcomes or limited sample sizes, and comprehensive head-to-head comparisons of these composite indices against both hemodynamic (hypertension) and structural (CAS) endpoints are lacking38,43. More importantly, the novelty of how these indices perform across different age and sex groups remains underexplored. Further, while composite indices are presumed to capture a broader metabolic burden, the independent contributions of adiposity versus IR to vascular damage lack empirical quantification. To address these gaps, we utilized a large-scale Chinese study to quantify and compare the associations of TyG, TyG-WC, TyG-WHtR, and TyG-BMI with hypertension and CAS. Specifically, we sought to: (1) evaluate their dose-response relationships and discriminative capacities; (2) identify age- and sex-specific risk patterns to optimize subgroup screening; and (3) employ mediation analyses to delineate the mechanistic pathways underlying the superiority of adiposity-integrated indices over the isolated TyG index. Methods Study design and setting We conducted a cross-sectional study of adults attending the Physical Examination Management Center of the Third Xiangya Hospital, which primarily serves residents of Changsha (Hunan, China) and neighboring cities. The study population comprised individuals attending either employer-sponsored annual health examinations or individual routine physical check-ups. Importantly, the protocols for laboratory measurements and ultrasound examinations, including the equipment, assay methodologies, and quality control procedures, remained strictly consistent and standardized throughout the entire study period from 2017 to 2024. Using convenience sampling, from January 1, 2017 to December 31, 2024, participants were invited to complete an 86-item self-reported health questionnaire online prior to their routine check-up through the website (https://new.selfhealth.com.cn/#/login) (details in Supplementary Material 1). We acknowledge that convenience sampling may introduce selection bias, as participants attending health check-ups might possess higher health awareness or differ socioeconomically from the general population; however, the inclusion of both mandatory employer-sponsored screenings and voluntary individual visits helps diversify the cohort and partially mitigate this bias. Participation was voluntary without incentives, and electronic informed consent was obtained before enrollment. We recorded the physical examination ID of participants to prevent duplication, and, where necessary, used concordant name, sex, and birth year for deduplication. When multiple visits from the same individual were identified, the earliest record was retained for the main analysis. The study protocol was approved by the Ethics Committee of the Third Xiangya Hospital, Central South University (No. 2020-S587). All methods were performed in accordance with the relevant guidelines and regulations. Reporting followed the STROBE guidelines (Supplementary Material 2). Participants In the end, the cross-sectional study included 154,873 participants. The study procedure is illustrated in in Fig. 1. To handle missing data, we explicitly employed a complete-case analysis approach, excluding individuals with incomplete records for the variables of interest. Exclusion criteria include: (1) participants lacking the hypertension and CAS outcomes; (2) participants lacking the covariates (sociological characteristics, lifestyle, blood indicators and disease history); and (3) participants with invalid or missing data for the TyG index and its combinations with obesity indicators (including missing values for specific components or current use of lipid-lowering or glucose-lowering medications, which render the TyG index unreliable). Fig. 1Flowchart depicting the participants’ selection. Flowchart depicting the participants’ selection. Definitions of TyG, TyG-WC, TyG-WHtR, and TyG-BMI Standardized protocols were strictly followed for all physical and laboratory measurements. The physical examination included measurements of weight, height and blood pressure (BP). Specifically, height and weight were measured to the nearest 0.1 cm and 0.1 kg, respectively, with participants wearing light clothing and no shoes. WC was measured to the nearest 0.1 cm at the midpoint between the lower margin of the least palpable rib and the top of the iliac crest at the end of normal expiration. BP was measured using a validated automated electronic sphygmomanometer after the participant had rested in a seated position for at least 5 min. Based on these physical measurements, the BMI along with the WHtR were calculated. For laboratory analyses, blood samples were drawn in the morning. Participants were instructed to fast strictly after midnight (12:00 AM) the night before the examination. FBG and TG were measured at baseline using standardized automated analyzers. Subsequently, the TyG index, which quantifies IR by combining fasting glucose with triglyceride levels, was calculated. The participants were classified into four groups (Q1, Q2, Q3, Q4) by the quartiles of the TyG index, TyG-WC, TyG-WHtR, and TyG-BMI, respectively, and the Q1 group was used as the reference group. The specific quartile cut-offs (data-driven boundaries for the 25th, 50th, and 75th percentiles) for each index were as follows: TyG (Q1: < 8.25, Q2: 8.25–8.68, Q3: 8.68–9.16, Q4: ≥ 9.16); TyG-BMI (Q1: < 183.27, Q2: 183.27–209.14, Q3: 209.14–237.02, Q4: ≥ 237.02); TyG-WC (Q1: < 623.56, Q2: 623.56–718.01, Q3: 718.01–810.83, Q4: ≥ 810.83); and TyG-WHtR (Q1: < 3.86, Q2: 3.86–4.37, Q3: 4.37–4.89, Q4: ≥ 4.89). TyG, TyG-WC, TyG-WHtR, and TyG-BMI were calculated according to the following formulas: TyG = ln[fasting TG [mg/dL] × FBG [mg/dL]/2]; BMI=body mass (kg)/height2(m2); WHtR = WC/height; TyG-WC = TyG×WC; TyG-WHtR = TyG×WHtR; TyG-BMI = TyG×BMI. Definition of hypertension BP was measured with participants seated, after at least 10 min of rest, using an appropriately sized cuff and an automated oscillometric device. BP was measured in both arms at the first assessment, and subsequent measurements were taken on the arm with the higher average value25. Three readings were obtained at 1-minute intervals, and the average of the three was used for analyses. Hypertension was defined as systolic blood pressure ≥ 140 mmHg, and/or diastolic blood pressure ≥ 90 mmHg, and/or self-reported physician-diagnosed hypertension, and/or current use of antihypertensive medication45. Definition of carotid atherosclerosis Carotid ultrasonography examinations were performed bilaterally by two certified sonographers who had received unified training. They were unaware of the characteristics and laboratory results of the participants. Participants were examined in the supine position with the head slightly extended and rotated contralaterally. Using a high-frequency linear transducer, longitudinal and transverse scans were obtained of the common carotid artery, carotid bulb, and proximal internal carotid artery. Discrepancies between their evaluations were resolved by consensus, and the final consistent carotid artery imaging data between the two certified sonographers were recorded. Increased CIMT was defined as CIMT ≥ 0.9 mm in either the left or right carotid artery. Carotid artery plaque (CAP) was defined as a focal structure encroaching into the lumen that met at least one of the following, maximum thickness ≥ 1.5 mm, focal wall thickening ≥ 50% relative to the adjacent CIMT, or a distinct focal lesion causing lumen irregularity, when identifiable, plaques were further described as calcified, ulcerated, intraplaque hemorrhage like, or associated with mural thrombus46,47. Further, as part of the routine health checkup protocol, arterial stenosis or occlusion was recorded when sonographic evidence showed marked luminal narrowing or absent flow. To reflect the overall burden of carotid artery disease and evaluate the general performance of metabolic-adiposity indices, these structural abnormalities (increased CIMT, CAP, or arterial stenosis/occlusion) were collectively categorized as a composite CAS positive outcome. Assessment of covariates Trained staff conducted face-to-face interviews with participants to investigate (1) Demographic characteristics, including age, sex and marital-status; (2) Lifestyle factors; (3) Blood samples; and (4) Disease history. The details were as follows: About lifestyle factors, coffee status, sugary drinks, smoking status were determined by whether individuals reported them or not48. Coffee consumption and sugary drink intake were originally assessed using a four-point frequency scale: “Never”, “Occasionally (1–2 times/week)”, “Regularly (3–5 times/week)”, and “Every day (> 5 times/week)”. During data analysis, participants who answered “Never” were categorized as “No”, while those reporting any frequency of consumption (≥ 1–2 times/week) were categorized as “Yes”49. Smoking status was also determined by whether individuals reported current smoking or not49. According to the Dietary Guidelines for Chinese Residents (2022), alcohol consumption was self-evaluated as ≥ 20 g, 10–20 g and < 10 g based on daily intake. We defined drinking as consuming more than 20 g of alcohol per day. The World Health Organization and the Dietary Guidelines for Chinese Residents (2022) now recommends getting at least 150 min per week of moderate moderate-intensity aerobic physical activity to prevent chronic illness48,50. Based on the number and length of physical activity per week, if the exercise meet this standard, we defined it as qualified physical exercise. Sleep situation was assessed with the question “During the past month, how is your sleep?” Responses were categorised as good, fair, or poor. Sleep duration was assessed with the question “During the past month, how many hours of actual sleep did you get?”, which was adapted from the Pittsburgh Sleep Quality Index51. Under the National Sleep Foundation guidelines, adults should have 7 to 9 h of sleep every night. Therefore, the sleep duration was further divided into < 7 h, 7–9 h and > 9 h49. Finally, the Patient Health Questionnaire-9 (PHQ-9) is a well-validated self-report instrument that evaluates the frequency of depressive symptoms (i.e., sadness, trouble sleeping, fatigue, problems concentrating) over the two weeks and has moderate concordance with clinical psychiatric interviews52. It consists of nine items that are scored on a four-point Likert scale, and the total score ranges from 0 to 27. The sensitivity, specificity and Cronbach’α coefficient of the scale were 0.87, 0.81 and 0.89, respectively52. We used the standard PHQ-9 cutoff point of ≥ 10 to define depressive symptoms, which agrees with the previous literature48,49. This division has been shown to have a sensitivity of 88% and a specificity of 88%53. Blood samples were collected for high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), FBG, TG, total cholesterol, hemoglobin (HB), alanine aminotransferase (ALT), total bilirubin (TBil), and albumin-globulin ratio (A/G). Blood urea nitrogen (BUN), serum creatinine (SCR), and uric acid (UA) were also collected. Ascertainment of disease history, self-reported diabetes, self-reported cancer, and self-reported hypercholesterolemia were established by self-reported physician diagnoses obtained during an individual interview using a standardized medical condition questionnaire. The participants were asked: “Have you ever been told by a physician that you have diabetes?”or“Have you ever been told by a physician that you have cancer or a malignancy?”or “Have you ever been told by a physician that you have hypercholesterolemia?” and answered by yes or no. Statistical analysis All collected data were analyzed using IBM SPSS Statistics software, version 25.0 (IBM Corp., Armonk, NY, USA; https://www.ibm.com/products/spss-statistics) and R software, version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria; https://www.R-project.org/). Continuous variables were assessed for normality using skewness and kurtosis. Normally distributed continuous variables are expressed as mean ± standard deviation (SD) and were compared across TyG quartiles using One-way analysis of variance (ANOVA). Non-normally distributed continuous variables are presented as medians (25th–75th percentiles) and were compared using the Kruskal-Wallis H test. Categorical variables are presented as numbers (percentages) and were compared using Chi-square tests. Baseline characteristics were stratified by quartiles of the TyG index, TyG-WC, TyG-WHtR, and TyG-BMI. Binary logistic regression was employed to evaluate the associations of TyG and its composite indices with hypertension and CAS, calculating odds ratios (ORs) and 95% confidence intervals (CIs). Four progressively adjusted models were constructed: Model 1 included demographics (age, sex, marital status); Model 2 added lifestyle factors (coffee/sugary drinks, smoking, alcohol, physical exercise, sleep metrics, and depressive symptoms); Model 3 further incorporated laboratory parameters (HDL-C, LDL-C, HB, ALT, TBil, A/G, BUN, SCR, UA); and Model 4 (the fully adjusted model) additionally controlled for self-reported comorbidities (diabetes, cancer, and hypercholesterolemia). To explore potential non-linear dose-response relationships, restricted cubic spline (RCS) analyses (based on Model 4) were fitted. Following standard statistical recommendations to balance curve flexibility and avoid overfitting, we specified three knots located at the 5th, 50th, and 95th percentiles of the exposure distributions. Likelihood ratio tests were utilized to assess the significance of non-linearity54. ROC curves and the area under the curve (AUC) were used to compare performance of the indices. Stratified analyses were conducted by age (≤ 60 vs. >60 years) and sex. To compare the predictive performance and model fit of different TyG-related indices, the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) were calculated. Lower AIC and BIC values indicate a better model fit. Crucially, mediation analyses were performed to quantify the direct cardiovascular effects of adiposity indicators and the indirect effects mediated by the TyG index. A two-sided P-value < 0.05 was considered statistically significant. Results Baseline characteristics of participants according to the quartile of TyG, TyG-WC, TyG-WHtR, and TyG-BMI indicators Data were obtained from 154,873 participants who underwent routine physical examinations and carotid ultrasonography. The baseline characteristics, stratified by TyG quartiles (Q1 = lowest), are presented in Table 1. Similar stratifications for TyG-WC, TyG-WHtR, and TyG-BMI are provided in Supplementary Material 3 (Tables S1–S3). The mean age of participants was 45.7 ± 12.5 years, with 42.1% being female and 91.0% being married. Across higher quartiles of TyG and its related indices, participants were older, and more likely to be male and married. Lifestyle patterns also varied significantly. Participants in higher quartiles reported higher rates of current smoking, alcohol consumption, and sugary drink intake, but lower coffee consumption. They also experienced shorter and poorer-quality sleep. Conversely, depressive symptoms were less common, and self-reported physical activity was modestly higher in the upper quartiles. Regarding biochemical indices, higher quartiles were associated with an unfavorable metabolic profile: HDL-C levels decreased, whereas LDL-C, HB, liver enzymes (ALT), and renal markers (BUN, SCR, UA) increased. TBil and the A/G showed minimal changes. For medical history, the prevalences of diabetes and hypercholesterolemia increased significantly across higher quartiles (e.g., hypercholesterolemia increased from 3.3% in Q1 to 21.4% in Q4). In contrast, the prevalence of cancer fluctuated slightly across quartiles (ranging from 0.9% to 1.2%). Furthermore, hypertension and CAS were more frequent in the higher quartiles. Notably, the overall prevalence of CAS was relatively high (52.5%), which likely reflects the broad composite definition utilized in our study. All observed trends across the quartiles were statistically significant (P < 0.001). Table 1Baseline characteristics according to triglyceride-glucose (TyG) quartiles (n = 154,873 ).Variablesn = 154,873TyG_quartileF/X2/H P Q1(Lowest)(n = 38,723)Q2(n = 38,723)Q3(n = 38,723)Q4(Highest)(n = 38,723)Age (years)45.7 ± 12.5140.96 ± 12.3746.11 ± 12.8147.93 ± 12.2847.80 ± 11.21117.112 a< 0.001Gender(n%)Male89,626(57.9%)12,156(31.4%)20,294(52.4%)25,704(66.4%)31,472(81.3%)21473.604 b< 0.001Female65,247(42.1%)26,567(68.6%)18,425(47.6%)13,014(33.6%)7241(18.7%)Marital status(n%)unmarried13,965(9%)6155(15.9%)3558(9.2%)2411(6.2%)1841(4.8%)3458.552 b< 0.001Married140,908(91%)32,568(84.1%)35,161(90.8%)36,307(93.8%)36,872(95.2%)Coffee status(n%)No113,513(73.3%)26,925(69.5%)28,485(73.6%)28,968(74.8%)29,135(75.3%)403.727 b< 0.001Yes41,360(26.7%)11,798(30.5%)10,234(26.4%)9750(25.2%)9578(24.7%)Sugary drinks(n%)No87,050(56.2%)20,614(53.2%)22,064(57%)22,199(57.3%)22,173(57.3%)186.488 b< 0.001Yes67,823(43.8%)18,109(46.8%)16,655(43%)16,519(42.7%)16,540(42.7%)Smoking status(n%)No107,895(69.7%)32,842(84.8%)29,065(75.1%)25,876(66.8%)20,112(52%)10634.164 b< 0.001Yes46,978(30.3%)5881(15.2%)9654(24.9%)12,842(33.2%)18,601(48%)Drink status(n%)No105,635(68.2%)32,067(82.8%)28,372(73.3%)25,230(65.2%)19,966(51.6%)9371.73 b< 0.001Yes49,238(31.8%)6656(17.2%)10,347(26.7%)13,488(34.8%)18,747(48.4%)Physical exercise(n%)No52,372(33.8%)13,741(35.5%)12,461(32.2%)12,473(32.2%)13,697(35.4%)181.043 b< 0.001Yes102,501(66.2%)24,982(64.5%)26,258(67.8%)26,245(67.8%)25,016(64.6%)Sleepsituation(n%)Good64,349(41.5%)15,479(40%)15,898(41.1%)16,256(42%)16,716(43.2%)93.44 b< 0.001Fair77,811(50.2%)20,054(51.8%)19,578(50.6%)19,239(49.7%)18,940(48.9%)Poor12,713(8.2%)3190(8.2%)3243(8.4%)3223(8.3%)3057(7.9%)Sleepduration(n%)≤ 7 h112,842(72.9%)27,174(70.2%)28,184(72.8%)28,659(74%)28,825(74.5%)244.835 b< 0.0017–9 h40,836(26.4%)11,289(29.2%)10,237(26.4%)9758(25.2%)9552(24.7%)>9 h1195(0.8%)260(0.7%)298(0.8%)301(0.8%)336(0.9%)Depressive symptoms(n%)No143,917(92.9%)35,111(90.7%)35,988(92.9%)36,271(93.7%)36,547(94.4%)461.535 b< 0.001Yes10,956(7.1%)3612(9.3%)2731(7.1%)2447(6.3%)2166(5.6%)LDL_C(mmol/L)2.92 ± 0.832.74 ± 0.693.02 ± 0.763.14 ± 0.822.77 ± 0.952228.118 a< 0.001HB(g/L)145.55 ± 15.72136.84 ± 14.76143.50 ± 15.11148.37 ± 14.47153.48 ± 13.499296.858 a< 0.001UA(µmol/L)345.1 ± 88.52296.58 ± 73.45330.48 ± 80.43360.25 ± 82.63393.10 ± 86.8810037.009 a< 0.001A/G1.76 ± 0.281.76 ± 0.271.76 ± 0.291.76 ± 0.281.75 ± 0.2818.912 a< 0.001HDL_C(mmol/L)1.28(1.11, 1.48)1.47(1.29, 1.67)1.34(1.18, 1.53)1.23(1.09, 1.39)1.11(0.98, 1.25)36090.158 c< 0.001ALT(U/L)20.00(14.00, 31.00)15.00(11.00, 21.00)18.00(14.00,26.00)23.00(16.00, 32.00)29.00(20.00, 43.00)28631.184 c< 0.001TBil(µmol/L)12.40(9.80, 15.70)12.50(9.90, 16.00)12.50(10.00, 15.90)12.50(9.90, 15.70)12.00(9.50, 15.30)428.757 c< 0.001BUN(mmol/L)4.65(3.92, 5.51)4.46(3.74, 5.32)4.63(3.89, 5.51)4.71(3.99, 5.55)4.81(4.08, 5.65)1707.879 c< 0.001SCR(µmol/L)72.00(60.00, 84.00)63.00(56.00, 75.00)70.00(59.00, 83.00)75.00(63.00, 86.00)78.00(68.00, 88.00)13041.048 c< 0.001Diabetes(n%)No143,869(92.9%)38,428(99.2%)37,779(97.6%)36,491(94.2%)31,171(80.5%)12736.124 b< 0.001Yes11,004(7.1%)295(0.8%)940(2.4%)2227(5.8%)7542(19.5%)Hypercholester-olemia(n%)No137,319(88.7%)37,443(96.7%)35,755(92.3%)33,709(87.1%)30,412(78.6%)7040.135 b< 0.001Yes17,554(11.3%)1280(3.3%)2964(7.7%)5009(12.9%)8301(21.4%)Cancer(n%)No153,201(98.9%)38,323(99%)38,255(98.8%)38,266(98.8%)38,357(99.1%)17.981 b< 0.001Yes1672(1.1%)400(1%)464(1.2%)452(1.2%)356(0.9%)Hypertension(n%)No138,638(89.5%)36,974(95.5%)35,099(90.7%)33,740(87.1%)32,825(84.8%)2676.054 b< 0.001Yes16,235(10.5%)1749(4.5%)3620(9.3%)4978(12.9%)5888(15.2%)Carotid Atherosclerosis(n%)No73,536(47.5%)21,860(56.5%)18,626(48.1%)16,781(43.3%)16,269(42%)1984.085 b< 0.001Yes81,337(52.5%)16,863(43.5%)20,093(51.9%)21,937(56.7%)22,444(58%)Continuous variables are expressed as mean ± standard deviation (SD) if normally distributed, or as median (25th–75th percentiles) if not. Categorical variables are presented as numbers (percentages). One-way analysis of variance (a), Chi-square tests (b), and Kruskal-Wallis H tests (c) were conducted to compare normally distributed continuous, categorical, and non-normally distributed continuous baseline characteristics, respectivelyHDL-C: high density lipoprotein cholesterol; LDL-C: low density lipoprotein cholesterol; HB: hemoglobin; ALT: alanine aminotransferase; TBil: total bilirubin; A/G: albumin-globulin ratio; BUN: blood urea nitrogen; SCR: serum creatinine; UA: uric acid. Table 1Baseline characteristics according to triglyceride-glucose (TyG) quartiles (n = 154,873 ).Variablesn = 154,873TyG_quartileF/X2/H P Q1(Lowest)(n = 38,723)Q2(n = 38,723)Q3(n = 38,723)Q4(Highest)(n = 38,723)Age (years)45.7 ± 12.5140.96 ± 12.3746.11 ± 12.8147.93 ± 12.2847.80 ± 11.21117.112 a< 0.001Gender(n%)Male89,626(57.9%)12,156(31.4%)20,294(52.4%)25,704(66.4%)31,472(81.3%)21473.604 b< 0.001Female65,247(42.1%)26,567(68.6%)18,425(47.6%)13,014(33.6%)7241(18.7%)Marital status(n%)unmarried13,965(9%)6155(15.9%)3558(9.2%)2411(6.2%)1841(4.8%)3458.552 b< 0.001Married140,908(91%)32,568(84.1%)35,161(90.8%)36,307(93.8%)36,872(95.2%)Coffee status(n%)No113,513(73.3%)26,925(69.5%)28,485(73.6%)28,968(74.8%)29,135(75.3%)403.727 b< 0.001Yes41,360(26.7%)11,798(30.5%)10,234(26.4%)9750(25.2%)9578(24.7%)Sugary drinks(n%)No87,050(56.2%)20,614(53.2%)22,064(57%)22,199(57.3%)22,173(57.3%)186.488 b< 0.001Yes67,823(43.8%)18,109(46.8%)16,655(43%)16,519(42.7%)16,540(42.7%)Smoking status(n%)No107,895(69.7%)32,842(84.8%)29,065(75.1%)25,876(66.8%)20,112(52%)10634.164 b< 0.001Yes46,978(30.3%)5881(15.2%)9654(24.9%)12,842(33.2%)18,601(48%)Drink status(n%)No105,635(68.2%)32,067(82.8%)28,372(73.3%)25,230(65.2%)19,966(51.6%)9371.73 b< 0.001Yes49,238(31.8%)6656(17.2%)10,347(26.7%)13,488(34.8%)18,747(48.4%)Physical exercise(n%)No52,372(33.8%)13,741(35.5%)12,461(32.2%)12,473(32.2%)13,697(35.4%)181.043 b< 0.001Yes102,501(66.2%)24,982(64.5%)26,258(67.8%)26,245(67.8%)25,016(64.6%)Sleepsituation(n%)Good64,349(41.5%)15,479(40%)15,898(41.1%)16,256(42%)16,716(43.2%)93.44 b< 0.001Fair77,811(50.2%)20,054(51.8%)19,578(50.6%)19,239(49.7%)18,940(48.9%)Poor12,713(8.2%)3190(8.2%)3243(8.4%)3223(8.3%)3057(7.9%)Sleepduration(n%)≤ 7 h112,842(72.9%)27,174(70.2%)28,184(72.8%)28,659(74%)28,825(74.5%)244.835 b< 0.0017–9 h40,836(26.4%)11,289(29.2%)10,237(26.4%)9758(25.2%)9552(24.7%)>9 h1195(0.8%)260(0.7%)298(0.8%)301(0.8%)336(0.9%)Depressive symptoms(n%)No143,917(92.9%)35,111(90.7%)35,988(92.9%)36,271(93.7%)36,547(94.4%)461.535 b< 0.001Yes10,956(7.1%)3612(9.3%)2731(7.1%)2447(6.3%)2166(5.6%)LDL_C(mmol/L)2.92 ± 0.832.74 ± 0.693.02 ± 0.763.14 ± 0.822.77 ± 0.952228.118 a< 0.001HB(g/L)145.55 ± 15.72136.84 ± 14.76143.50 ± 15.11148.37 ± 14.47153.48 ± 13.499296.858 a< 0.001UA(µmol/L)345.1 ± 88.52296.58 ± 73.45330.48 ± 80.43360.25 ± 82.63393.10 ± 86.8810037.009 a< 0.001A/G1.76 ± 0.281.76 ± 0.271.76 ± 0.291.76 ± 0.281.75 ± 0.2818.912 a< 0.001HDL_C(mmol/L)1.28(1.11, 1.48)1.47(1.29, 1.67)1.34(1.18, 1.53)1.23(1.09, 1.39)1.11(0.98, 1.25)36090.158 c< 0.001ALT(U/L)20.00(14.00, 31.00)15.00(11.00, 21.00)18.00(14.00,26.00)23.00(16.00, 32.00)29.00(20.00, 43.00)28631.184 c< 0.001TBil(µmol/L)12.40(9.80, 15.70)12.50(9.90, 16.00)12.50(10.00, 15.90)12.50(9.90, 15.70)12.00(9.50, 15.30)428.757 c< 0.001BUN(mmol/L)4.65(3.92, 5.51)4.46(3.74, 5.32)4.63(3.89, 5.51)4.71(3.99, 5.55)4.81(4.08, 5.65)1707.879 c< 0.001SCR(µmol/L)72.00(60.00, 84.00)63.00(56.00, 75.00)70.00(59.00, 83.00)75.00(63.00, 86.00)78.00(68.00, 88.00)13041.048 c< 0.001Diabetes(n%)No143,869(92.9%)38,428(99.2%)37,779(97.6%)36,491(94.2%)31,171(80.5%)12736.124 b< 0.001Yes11,004(7.1%)295(0.8%)940(2.4%)2227(5.8%)7542(19.5%)Hypercholester-olemia(n%)No137,319(88.7%)37,443(96.7%)35,755(92.3%)33,709(87.1%)30,412(78.6%)7040.135 b< 0.001Yes17,554(11.3%)1280(3.3%)2964(7.7%)5009(12.9%)8301(21.4%)Cancer(n%)No153,201(98.9%)38,323(99%)38,255(98.8%)38,266(98.8%)38,357(99.1%)17.981 b< 0.001Yes1672(1.1%)400(1%)464(1.2%)452(1.2%)356(0.9%)Hypertension(n%)No138,638(89.5%)36,974(95.5%)35,099(90.7%)33,740(87.1%)32,825(84.8%)2676.054 b< 0.001Yes16,235(10.5%)1749(4.5%)3620(9.3%)4978(12.9%)5888(15.2%)Carotid Atherosclerosis(n%)No73,536(47.5%)21,860(56.5%)18,626(48.1%)16,781(43.3%)16,269(42%)1984.085 b< 0.001Yes81,337(52.5%)16,863(43.5%)20,093(51.9%)21,937(56.7%)22,444(58%)Continuous variables are expressed as mean ± standard deviation (SD) if normally distributed, or as median (25th–75th percentiles) if not. Categorical variables are presented as numbers (percentages). One-way analysis of variance (a), Chi-square tests (b), and Kruskal-Wallis H tests (c) were conducted to compare normally distributed continuous, categorical, and non-normally distributed continuous baseline characteristics, respectivelyHDL-C: high density lipoprotein cholesterol; LDL-C: low density lipoprotein cholesterol; HB: hemoglobin; ALT: alanine aminotransferase; TBil: total bilirubin; A/G: albumin-globulin ratio; BUN: blood urea nitrogen; SCR: serum creatinine; UA: uric acid. Baseline characteristics according to triglyceride-glucose (TyG) quartiles (n = 154,873 ). Sleep situation(n%) Sleep duration(n%) 1.28 (1.11, 1.48) 1.47 (1.29, 1.67) 1.34 (1.18, 1.53) 1.23 (1.09, 1.39) 1.11 (0.98, 1.25) 20.00 (14.00, 31.00) 15.00 (11.00, 21.00) 18.00 (14.00,26.00) 23.00 (16.00, 32.00) 29.00 (20.00, 43.00) 12.40 (9.80, 15.70) 12.50 (9.90, 16.00) 12.50 (10.00, 15.90) 12.50 (9.90, 15.70) 12.00 (9.50, 15.30) 4.65 (3.92, 5.51) 4.46 (3.74, 5.32) 4.63 (3.89, 5.51) 4.71 (3.99, 5.55) 4.81 (4.08, 5.65) 72.00 (60.00, 84.00) 63.00 (56.00, 75.00) 70.00 (59.00, 83.00) 75.00 (63.00, 86.00) 78.00 (68.00, 88.00) Hypertension (n%) Continuous variables are expressed as mean ± standard deviation (SD) if normally distributed, or as median (25th–75th percentiles) if not. Categorical variables are presented as numbers (percentages). One-way analysis of variance (a), Chi-square tests (b), and Kruskal-Wallis H tests (c) were conducted to compare normally distributed continuous, categorical, and non-normally distributed continuous baseline characteristics, respectively HDL-C: high density lipoprotein cholesterol; LDL-C: low density lipoprotein cholesterol; HB: hemoglobin; ALT: alanine aminotransferase; TBil: total bilirubin; A/G: albumin-globulin ratio; BUN: blood urea nitrogen; SCR: serum creatinine; UA: uric acid. Associations of TyG, TyG-WC, TyG-WHtR, and TyG-BMI with hypertension and carotid atherosclerosis Tables 2 and 3 present multivariable-adjusted associations (Models 1–4). After full adjustment, higher quartiles of TyG, TyG-WC, TyG-WHtR, and TyG-BMI were each positively associated with hypertension and CAS (all P < 0.001; Q4 vs. Q1 as reference contrast). In the fully adjusted model (Q4 vs. Q1), TyG was associated with 2.21-fold higher odds of hypertension (OR = 2.212, 95% CI: 2.065–2.368) and 1.43-fold higher odds of CAS (OR = 1.426, 95% CI: 1.372–1.482). Table 2Associations between TyG, TyG-WC, TyG-WHtR, and TyG-BMI with hypertension (n = 154, 873).HypertensionOR (95%CI) P P for interactionQ1(Lowest)Q2Q3Q4(Highest)AgeGenderTyGquartilesModel 1Ref.1.734(1.631,1.842)2.313(2.181,2.453)2.872(2.708,3.045)< 0.001< 0.001< 0.001Model 2Ref.1.713(1.611,1.821)2.298(2.166,2.437)2.871(2.707,3.046)< 0.001< 0.001< 0.001Model 3Ref.1.705(1.602,1.815)2.222(2.086,2.365)2.483(2.325,2.651)< 0.001< 0.001< 0.001Model 4Ref.1.685(1.583,1.794)2.142(2.011,2.281)2.212(2.065,2.368)< 0.001< 0.001< 0.001TyG_WC quartilesModel 1Ref.2.800(2.592,3.025)5.059(4.688,5.459)7.743(7.168,8.364)< 0.001< 0.0010.178Model 2Ref.2.714(2.512,2.933)4.964(4.599,5.358)7.902(7.312,8.54)< 0.001< 0.0010.129Model 3Ref.2.901(2.681,3.140)5.470(5.044,5.932)8.538(7.835,9.303)< 0.001< 0.0010.263Model 4Ref.2.880(2.661,3.118)5.357(4.939,5.811)8.097(7.420,8.837)< 0.001< 0.0010.264TyG_WHtRquartilesModel 1Ref.2.798(2.578,3.037)5.039(4.657,5.453)7.959(7.362,8.604)< 0.001< 0.001< 0.001Model 2Ref.2.702(2.489,2.933)4.882(4.511,5.283)7.997(7.395,8.648)< 0.001< 0.001< 0.001Model 3Ref.2.935(2.700,3.191)5.517(5.078,5.994)8.973(8.238,9.774)< 0.001< 0.001< 0.001Model 4Ref.2.924(2.690,3.179)5.456(5.021,5.929)8.644(7.927,9.426)< 0.001< 0.001< 0.001TyG_BMIquartilesModel 1Ref.2.233(2.082,2.396)3.837(3.587,4.105)5.779(5.403,6.182)< 0.001< 0.0010.224Model 2Ref.2.181(2.032,2.340)3.769(3.522,4.033)5.912(5.524,6.328)< 0.001< 0.0010.157Model 3Ref.2.303(2.142,2.475)4.051(3.769,4.353)6.230(5.774,6.721)< 0.001< 0.0010.108Model 4Ref.2.285(2.126,2.456)3.970(3.693,4.267)5.909(5.472,6.382)< 0.001< 0.0010.100Model 1 only contained demographic characteristics, including age, sex and marital-status. Model 2 included demographic characteristics and lifestyle factors, including age, sex, marital-status, coffee status, sugary drinks, smoking status, alcohol consumption, physical exercise, sleep situation, sleep duration and depressive symptoms; Model 3 adjusted for the same variables as Model 2 and for HDL-C, LDL-C, HB, ALT, TBil, A/G, BUN, SCR and UA; Model 4 was a complete model, and adjusted for the same variables as Model 3: self-reported diabetes, self-reported cancer, and self-reported hypercholesterolemia. TyG: triglyceride-glucose; BMI: body mass index; WC: waist circumference; WHtR: the waist-to-height ratio; OR: odds ratios ; 95%CI: 95%confidence intervals; Ref.: Reference. Table 2Associations between TyG, TyG-WC, TyG-WHtR, and TyG-BMI with hypertension (n = 154, 873).HypertensionOR (95%CI) P P for interactionQ1(Lowest)Q2Q3Q4(Highest)AgeGenderTyGquartilesModel 1Ref.1.734(1.631,1.842)2.313(2.181,2.453)2.872(2.708,3.045)< 0.001< 0.001< 0.001Model 2Ref.1.713(1.611,1.821)2.298(2.166,2.437)2.871(2.707,3.046)< 0.001< 0.001< 0.001Model 3Ref.1.705(1.602,1.815)2.222(2.086,2.365)2.483(2.325,2.651)< 0.001< 0.001< 0.001Model 4Ref.1.685(1.583,1.794)2.142(2.011,2.281)2.212(2.065,2.368)< 0.001< 0.001< 0.001TyG_WC quartilesModel 1Ref.2.800(2.592,3.025)5.059(4.688,5.459)7.743(7.168,8.364)< 0.001< 0.0010.178Model 2Ref.2.714(2.512,2.933)4.964(4.599,5.358)7.902(7.312,8.54)< 0.001< 0.0010.129Model 3Ref.2.901(2.681,3.140)5.470(5.044,5.932)8.538(7.835,9.303)< 0.001< 0.0010.263Model 4Ref.2.880(2.661,3.118)5.357(4.939,5.811)8.097(7.420,8.837)< 0.001< 0.0010.264TyG_WHtRquartilesModel 1Ref.2.798(2.578,3.037)5.039(4.657,5.453)7.959(7.362,8.604)< 0.001< 0.001< 0.001Model 2Ref.2.702(2.489,2.933)4.882(4.511,5.283)7.997(7.395,8.648)< 0.001< 0.001< 0.001Model 3Ref.2.935(2.700,3.191)5.517(5.078,5.994)8.973(8.238,9.774)< 0.001< 0.001< 0.001Model 4Ref.2.924(2.690,3.179)5.456(5.021,5.929)8.644(7.927,9.426)< 0.001< 0.001< 0.001TyG_BMIquartilesModel 1Ref.2.233(2.082,2.396)3.837(3.587,4.105)5.779(5.403,6.182)< 0.001< 0.0010.224Model 2Ref.2.181(2.032,2.340)3.769(3.522,4.033)5.912(5.524,6.328)< 0.001< 0.0010.157Model 3Ref.2.303(2.142,2.475)4.051(3.769,4.353)6.230(5.774,6.721)< 0.001< 0.0010.108Model 4Ref.2.285(2.126,2.456)3.970(3.693,4.267)5.909(5.472,6.382)< 0.001< 0.0010.100Model 1 only contained demographic characteristics, including age, sex and marital-status. Model 2 included demographic characteristics and lifestyle factors, including age, sex, marital-status, coffee status, sugary drinks, smoking status, alcohol consumption, physical exercise, sleep situation, sleep duration and depressive symptoms; Model 3 adjusted for the same variables as Model 2 and for HDL-C, LDL-C, HB, ALT, TBil, A/G, BUN, SCR and UA; Model 4 was a complete model, and adjusted for the same variables as Model 3: self-reported diabetes, self-reported cancer, and self-reported hypercholesterolemia. TyG: triglyceride-glucose; BMI: body mass index; WC: waist circumference; WHtR: the waist-to-height ratio; OR: odds ratios ; 95%CI: 95%confidence intervals; Ref.: Reference. Associations between TyG, TyG-WC, TyG-WHtR, and TyG-BMI with hypertension (n = 154, 873). TyG quartiles TyG_WHtR quartiles TyG_BMI quartiles Model 1 only contained demographic characteristics, including age, sex and marital-status. Model 2 included demographic characteristics and lifestyle factors, including age, sex, marital-status, coffee status, sugary drinks, smoking status, alcohol consumption, physical exercise, sleep situation, sleep duration and depressive symptoms; Model 3 adjusted for the same variables as Model 2 and for HDL-C, LDL-C, HB, ALT, TBil, A/G, BUN, SCR and UA; Model 4 was a complete model, and adjusted for the same variables as Model 3: self-reported diabetes, self-reported cancer, and self-reported hypercholesterolemia. TyG: triglyceride-glucose; BMI: body mass index; WC: waist circumference; WHtR: the waist-to-height ratio; OR: odds ratios ; 95%CI: 95%confidence intervals; Ref.: Reference. Table 3Associations between TyG, TyG-WC, TyG-WHtR, and TyG-BMI with carotid atherosclerosis (n = 154, 873).Carotid AtherosclerosisOR (95%CI) P P for interactionQ1(Lowest)Q2Q3Q4(Highest)AgeGenderTyGquartilesModel 1Ref.1.204(1.169,1.241)1.379(1.338,1.422)1.464(1.418,1.511)< 0.0010.2080.092Model 2Ref.1.183(1.148,1.220)1.350(1.308,1.392)1.414(1.369,1.460)< 0.0010.2640.069Model 3Ref.1.211(1.174,1.249)1.426(1.379,1.475)1.585(1.528,1.645)< 0.0010.2070.056Model 4Ref.1.203(1.166,1.241)1.389(1.343,1.437)1.426(1.372,1.482)< 0.0010.051 0.048 TyG_WCquartilesModel 1Ref.1.365(1.323,1.407)1.641(1.587,1.698)1.804(1.740,1.869)< 0.0010.0060.086Model 2Ref.1.328(1.287,1.370)1.596(1.543,1.652)1.758(1.695,1.823)< 0.0010.0100.046Model 3Ref.1.414(1.369,1.461)1.808(1.741,1.878)2.155(2.065,2.250)< 0.0010.0070.080Model 4Ref.1.399(1.354,1.445)1.752(1.686,1.820)1.965(1.880,2.054)< 0.0010.0040.154TyG_WHtRquartilesModel 1Ref.1.315(1.276,1.355)1.695(1.642,1.749)1.923(1.860,1.987)< 0.0010.0120.004Model 2Ref.1.275(1.237,1.314)1.633(1.581,1.686)1.860(1.799,1.923)< 0.0010.0140.002Model 3Ref.1.374(1.331,1.418)1.885(1.820,1.953)2.343(2.252,2.437)< 0.0010.0110.003Model 4Ref.1.366(1.323,1.410)1.848(1.783,1.915)2.173(2.086,2.263)< 0.0010.008 0.002 TyG_BMIquartilesModel 1Ref.1.264(1.227,1.303)1.441(1.396,1.487)1.505(1.456,1.555)< 0.0010.4600.005Model 2Ref.1.235(1.198,1.273)1.401(1.357,1.447)1.477(1.428,1.527)< 0.0010.4740.006Model 3Ref.1.276(1.236,1.317)1.505(1.453,1.559)1.687(1.622,1.754)< 0.0010.5140.011Model 4Ref.1.267(1.228,1.308)1.467(1.416,1.520)1.553(1.492,1.617)< 0.0010.502 0.014 Model 1 only contained demographic characteristics, including age, sex and marital-status. Model 2 included demographic characteristics and lifestyle factors, including age, sex, marital-status, coffee status, sugary drinks, smoking status, alcohol consumption, physical exercise, sleep situation, sleep duration and depressive symptoms; Model 3 adjusted for the same variables as Model 2 and for HDL-C, LDL-C, HB, ALT, TBil, A/G, BUN, SCR and UA; Model 4 was a complete model, and adjusted for the same variables as Model 3: self-reported diabetes, self-reported cancer, and self-reported hypercholesterolemia.TyG: triglyceride-glucose; WC: waist circumference; WHtR: the waist-to-height ratio; BMI: body mass index; OR: odds ratios ; 95%CI: 95%confidence intervals; Ref.: Reference. Table 3Associations between TyG, TyG-WC, TyG-WHtR, and TyG-BMI with carotid atherosclerosis (n = 154, 873).Carotid AtherosclerosisOR (95%CI) P P for interactionQ1(Lowest)Q2Q3Q4(Highest)AgeGenderTyGquartilesModel 1Ref.1.204(1.169,1.241)1.379(1.338,1.422)1.464(1.418,1.511)< 0.0010.2080.092Model 2Ref.1.183(1.148,1.220)1.350(1.308,1.392)1.414(1.369,1.460)< 0.0010.2640.069Model 3Ref.1.211(1.174,1.249)1.426(1.379,1.475)1.585(1.528,1.645)< 0.0010.2070.056Model 4Ref.1.203(1.166,1.241)1.389(1.343,1.437)1.426(1.372,1.482)< 0.0010.051 0.048 TyG_WCquartilesModel 1Ref.1.365(1.323,1.407)1.641(1.587,1.698)1.804(1.740,1.869)< 0.0010.0060.086Model 2Ref.1.328(1.287,1.370)1.596(1.543,1.652)1.758(1.695,1.823)< 0.0010.0100.046Model 3Ref.1.414(1.369,1.461)1.808(1.741,1.878)2.155(2.065,2.250)< 0.0010.0070.080Model 4Ref.1.399(1.354,1.445)1.752(1.686,1.820)1.965(1.880,2.054)< 0.0010.0040.154TyG_WHtRquartilesModel 1Ref.1.315(1.276,1.355)1.695(1.642,1.749)1.923(1.860,1.987)< 0.0010.0120.004Model 2Ref.1.275(1.237,1.314)1.633(1.581,1.686)1.860(1.799,1.923)< 0.0010.0140.002Model 3Ref.1.374(1.331,1.418)1.885(1.820,1.953)2.343(2.252,2.437)< 0.0010.0110.003Model 4Ref.1.366(1.323,1.410)1.848(1.783,1.915)2.173(2.086,2.263)< 0.0010.008 0.002 TyG_BMIquartilesModel 1Ref.1.264(1.227,1.303)1.441(1.396,1.487)1.505(1.456,1.555)< 0.0010.4600.005Model 2Ref.1.235(1.198,1.273)1.401(1.357,1.447)1.477(1.428,1.527)< 0.0010.4740.006Model 3Ref.1.276(1.236,1.317)1.505(1.453,1.559)1.687(1.622,1.754)< 0.0010.5140.011Model 4Ref.1.267(1.228,1.308)1.467(1.416,1.520)1.553(1.492,1.617)< 0.0010.502 0.014 Model 1 only contained demographic characteristics, including age, sex and marital-status. Model 2 included demographic characteristics and lifestyle factors, including age, sex, marital-status, coffee status, sugary drinks, smoking status, alcohol consumption, physical exercise, sleep situation, sleep duration and depressive symptoms; Model 3 adjusted for the same variables as Model 2 and for HDL-C, LDL-C, HB, ALT, TBil, A/G, BUN, SCR and UA; Model 4 was a complete model, and adjusted for the same variables as Model 3: self-reported diabetes, self-reported cancer, and self-reported hypercholesterolemia.TyG: triglyceride-glucose; WC: waist circumference; WHtR: the waist-to-height ratio; BMI: body mass index; OR: odds ratios ; 95%CI: 95%confidence intervals; Ref.: Reference. Associations between TyG, TyG-WC, TyG-WHtR, and TyG-BMI with carotid atherosclerosis (n = 154, 873). TyG quartiles TyG_WC quartiles TyG_WHtR quartiles TyG_BMI quartiles Model 1 only contained demographic characteristics, including age, sex and marital-status. Model 2 included demographic characteristics and lifestyle factors, including age, sex, marital-status, coffee status, sugary drinks, smoking status, alcohol consumption, physical exercise, sleep situation, sleep duration and depressive symptoms; Model 3 adjusted for the same variables as Model 2 and for HDL-C, LDL-C, HB, ALT, TBil, A/G, BUN, SCR and UA; Model 4 was a complete model, and adjusted for the same variables as Model 3: self-reported diabetes, self-reported cancer, and self-reported hypercholesterolemia. TyG: triglyceride-glucose; WC: waist circumference; WHtR: the waist-to-height ratio; BMI: body mass index; OR: odds ratios ; 95%CI: 95%confidence intervals; Ref.: Reference. The magnitude of the associations was larger for TyG-WHtR and TyG-WC: odds of hypertension were 8.64-fold (OR = 8.644, 95% CI: 7.927–9.426) and 8.10-fold (OR = 8.097, 95% CI: 7.420–8.837) higher, respectively; odds of CAS were 2.17-fold (OR = 2.173, 95% CI: 2.086–2.263) and 1.97-fold (OR = 1.965, 95% CI: 1.880–2.054) higher, respectively. TyG-BMI was also significantly associated with these outcomes (hypertension: OR = 5.909, 95% CI: 5.472–6.382; CAS: OR = 1.553, 95% CI: 1.492–1.617). Overall, elevated TyG and its adiposity-integrated indices were independently associated with higher odds of both hypertension and CAS. Further, as shown in Table S4, TyG_WHtR exhibited the lowest AIC and BIC values in relation to CAS, indicating the best model fit. For hypertension, TyG_BMI showed the lowest AIC and BIC values, followed closely by TyG_WHtR, both of which were superior to TyG and TyG_WC. Restricted cubic splines analysis investigating the relationship between TyG, TyG-WC, TyG-WHtR, TyG-BMI and hypertension and carotid atherosclerosis RCS models (Fig. 2), adjusted for prespecified covariates and referenced to each exposure’s median, supported nonlinearity for all four markers with both outcomes (hypertension and CAS: P-overall < 0.0001; P-nonlinear < 0.0001). For hypertension, the risks generally increased with higher levels of TyG, TyG-WC, TyG-WHtR, and TyG-BMI, with the curves appearing to flatten at the upper extremes. For CAS, the observed shapes were more complex, displaying non-linear variations (such as apparent U-shaped or inverse-U patterns) across the range of the indices. However, these complex patterns, particularly at the extreme ends of the distributions, should be interpreted with caution. Fig. 2Associations between TyG, TyG-WC, TyG-WHtR, and TyG-BMI with hypertension and carotid atherosclerosis were evaluated by RCS after adjustment for the covariables.The solid black lines correspond to the central estimates, and the gray-shaded regions indicate the 95% confdence intervals. TyG: triglyceride-glucose; BMI: body mass index; WC: waist circumference; WHtR: the waist-to-height ratio. Associations between TyG, TyG-WC, TyG-WHtR, and TyG-BMI with hypertension and carotid atherosclerosis were evaluated by RCS after adjustment for the covariables.The solid black lines correspond to the central estimates, and the gray-shaded regions indicate the 95% confdence intervals. TyG: triglyceride-glucose; BMI: body mass index; WC: waist circumference; WHtR: the waist-to-height ratio. Receiver operating characteristic curves of TyG, TyG-WC, TyG-WHtR, and TyG-BMI in relation to hypertension and carotid atherosclerosis ROC curves (Fig. 3) showed that TyG-WHtR provided the highest discrimination for hypertension (AUC = 0.708, 95% CI: 0.704–0.711), followed by TyG-WC (AUC = 0.685, 95% CI: 0.681–0.689), TyG-BMI (AUC = 0.672, 95% CI: 0.668–0.676), and TyG (AUC = 0.627, 95% CI: 0.623–0.631). For CAS, TyG-WHtR again ranked highest (AUC = 0.600, 95% CI: 0.598–0.603), followed by TyG-WC (AUC = 0.583, 95% CI: 0.580–0.586), while TyG (AUC = 0.565, 95% CI: 0.562–0.568) and TyG-BMI (AUC = 0.564, 95% CI: 0.561–0.567) were similar and lower. However, it is important to note that the overall discriminative ability of all tested indices for CAS was poor (AUC ≤ 0.600), indicating limited predictive and clinical utility for this specific outcome. Stratification of TyG, TyG‑WC, TyG‑WHtR, and TyG‑BMI in relation to Hypertension and carotid atherosclerosis Stratified analyses by age and sex (Supplementary Tables S5–S12) and interaction tests (Tables 2 and 3) showed that the associations between the indices and the two outcomes varied across subgroups. Fig. 3Receiver operating characteristic (ROC) curves of TyG, TyG-WC, TyG-WHtR, TyG-BMI in relation to hypertension and carotid atherosclerosis. Area Under Curve: AUC; TyG: triglyceride-glucose; BMI: body mass index; WC: waist circumference; WHtR: the waist-to-height ratio; OR: odds ratios ; 95%CI: 95% confidence intervals. Receiver operating characteristic (ROC) curves of TyG, TyG-WC, TyG-WHtR, TyG-BMI in relation to hypertension and carotid atherosclerosis. Area Under Curve: AUC; TyG: triglyceride-glucose; BMI: body mass index; WC: waist circumference; WHtR: the waist-to-height ratio; OR: odds ratios ; 95%CI: 95% confidence intervals. Age-stratified associations For hypertension, all indices showed positive associations in both age groups, with significant age interactions observed across all indices (P for interaction < 0.001 for all). In participants aged ≤ 60 years, TyG-WC (OR = 5.511, 95% CI: 4.931–6.160) and TyG-BMI (OR = 5.432, 95% CI: 4.898–6.024) yielded the highest odds ratios. In participants aged > 60 years, the ORs were generally lower than in the younger group, with TyG-WC (OR = 2.849, 95% CI: 2.452–3.310) and TyG-WHtR (OR = 2.736, 95% CI: 2.346–3.191) showing the highest estimates. For CAS, significant age interactions were only observed for TyG-WC (P for interaction = 0.004) and TyG-WHtR (P for interaction = 0.008). TyG showed an inverse association in participants aged ≤ 60 years (OR = 0.949, 95% CI: 0.909–0.990) but was not significant in those > 60 years (P = 0.525). In contrast, the composite indices (TyG-WHtR, TyG-WC, and TyG-BMI) showed positive associations with CAS in both age groups. For example, the OR for TyG-WHtR was 1.127 (95% CI: 1.076–1.180) in the younger group and 1.447 (95% CI: 1.218–1.718) in the older group. Sex-stratified associations For hypertension, all indices were positively associated with the outcome in both sexes. Significant sex interactions were observed for TyG and TyG-WHtR (P for interaction < 0.001 for both). In men, the highest OR was observed for TyG-BMI (OR = 4.772, 95% CI: 4.278–5.323). In women, the highest OR was observed for TyG-WHtR (OR = 4.375, 95% CI: 3.807–5.027). For CAS, sex interactions were significant for TyG (P for interaction = 0.048), TyG-WHtR (P for interaction = 0.002), and TyG-BMI (P for interaction = 0.014). In men, TyG was inversely associated with the outcome (OR = 0.932, 95% CI: 0.881–0.985), TyG-BMI showed a positive association (OR = 1.091, 95% CI: 1.027–1.159), and TyG-WC and TyG-WHtR were not statistically significant (P > 0.05). In women, all indices were positively associated with CAS, with ORs of 1.504 (95% CI: 1.402–1.614) for TyG-WHtR, 1.347 for TyG-WC, 1.128 for TyG, and 1.122 for TyG-BMI. Mediation role of TyG index in the associations of obesity indices with hypertension and carotid atherosclerosis Building upon the sex-stratified findings, mediation analyses were conducted to evaluate the potential mediating role of the TyG index in the associations between specific obesity indices (WHtR in females and BMI in males) and the outcomes (Fig. 4). In females, the indirect association of WHtR with CAS through the TyG index was 0.017 (P = 0.015; Fig. 4b), while the indirect association for hypertension was not statistically significant (P = 0.082; Fig. 4a). In males, the indirect estimates of BMI through TyG were statistically significant but close to zero for both outcomes (indirect estimates=-0.000, P < 0.01; Fig. 4c and d). Across all four models, the direct associations of the obesity indicators (WHtR in females, BMI in males) with the outcomes remained statistically significant (all P < 0.001). Fig. 4Mediation Role of TyG Index in the Associations of Obesity Indices with Hypertension and Carotid Atherosclerosis. The terms “Direct” and “Indirect” refer to the standard statistical outputs of the mediation analysis models and do not imply definitive causal pathways, given the cross-sectional nature of the data. Mediation Role of TyG Index in the Associations of Obesity Indices with Hypertension and Carotid Atherosclerosis. The terms “Direct” and “Indirect” refer to the standard statistical outputs of the mediation analysis models and do not imply definitive causal pathways, given the cross-sectional nature of the data. Discussion This study demonstrates that composite metabolic indices, particularly TyG-WHtR, outperform the isolated TyG index in identifying hypertension and CAS. We identified distinct dose-response patterns: a monotonic, age-dependent relationship for hypertension, and a U-shaped pattern for CAS. Additionally, notable sex differences were observed in CAS assessment, with TyG-WHtR being the primary indicator in women and TyG-BMI in men. Finally, mediation analyses suggest that adiposity is directly associated with these outcomes, largely independent of TyG-mediated IR. The TyG index is a well-established surrogate for IR24,30,31, yet emerging evidence suggests its predictive power is amplified when integrated with adiposity metrics. Previous studies indicate that TyG-related composites (e.g., TyG-WHtR, TyG-BMI) are superior predictors of major adverse cardiovascular events55, cardiovascular mortality56, and CVD onset44 compared to TyG alone. Consistent with this literature30,42,44,56, our study demonstrates that adiposity-integrated indices, particularly TyG-WHtR and TyG-WC, outperform isolated TyG in discriminating both hypertension and CAS. However, it is crucial to explicitly acknowledge the discrepancy in performance between the two outcomes. While the discriminative ability for hypertension was moderate (AUC ~ 0.70), the AUC values for CAS were notably lower (~ 0.56–0.60), indicating limited correlation performance for CAS. Therefore, these outcomes should not be viewed as having equally robust associations with the indices. A key strength of our analysis is the use of RCS to explore potential non-linear associations. The relationship between the indices and hypertension was largely monotonic, characterized by a steady increase in risk that appeared to plateau at the upper extremes. However, given the notably widened 95% confidence intervals at these upper tails, this apparent plateau is highly susceptible to data sparsity or residual confounding, rather than representing a definitive biological “saturation effect.” Biologically, while cumulative hemodynamic and neurohormonal stress generally correlates with worsening IR57 and increasing hypertension risk, our data cannot confirm whether a true risk threshold exists at extreme index levels. Although some studies report linear associations58, such discrepancies likely stem from differences in population characteristics and modeling choices (e.g., knot placement). In contrast, the associations with CAS presented more complex, multi-phasic non-linear patterns59, generally displaying an initial U-shaped dip at the lower extremes followed by a rise and a subsequent flattening at the higher extremes. This is highly consistent with a recent prospective study, which also reported a U-shaped association between the TyG index and elevated CIMT, particularly among non-obese adults60. While previous literature has sometimes discussed U-shaped patterns at the lower extreme in the context of malnutrition or frailty (often termed the “obesity paradox”), our descriptive findings cannot substantiate these specific biological mechanisms. Given the wide confidence intervals at both ends of the distributions, these observed fluctuations at the tails might equally reflect model instability, boundary artefacts inherent to spline modeling, or unmeasured confounding due to fewer participants in these extreme ranges. Indeed, these fluctuating trajectories align with the generally modest effect sizes and lower AUC values observed for CAS. Ultimately, while the observed patterns support a potential performance hierarchy (TyG-WHtR > TyG-WC > TyG-BMI > TyG) for identifying hypertension56, their utility as associative markers for CAS57 risk appears much more limited and should be interpreted with caution. Subgroup analyses unveiled critical heterogeneity that informs clinical application. A striking age-dependent gradient was observed for hypertension, where the discriminative strength of TyG-related indices was nearly four-fold higher in younger adults (< 60 years) compared to the elderly. This highlights the “cumulative burden” hypothesis: early-onset metabolic dysregulation may be more profoundly associated with vascular damage before traditional risk factors (e.g., arterial stiffness due to aging) become dominant confounding drivers in older populations56,61. In contrast, the associations with CAS were highly inconsistent across subgroups. For instance, in younger participants, the isolated TyG index showed inverse or null associations with CAS, while composite indices showed only modest positive associations. This aligns with recent findings in early adulthood cohorts, which suggest that the overall relationship between the TyG index and early structural atherosclerosis (e.g., CIMT) in younger populations is relatively modest and complex62. Even when statistically significant, the effect sizes for CAS were generally modest and often close to null after full adjustment. Further, sex stratification provided additional insights for vascular screening56,58. While TyG-WHtR appeared to be the most consistent marker for hypertension and for CAS in women, the associations in men were more variable and generally weaker. Specifically for male CAS, indices such as TyG, TyG-WC, and TyG-WHtR yielded inconsistent or non-significant results, whereas TyG-BMI demonstrated a modest association. Therefore, rather than serving as a definitive indicator, TyG-BMI may simply be relatively more informative among the evaluated indices for men. These sex-specific variations highlight the complexity of metabolic risk across genders and warrant cautious interpretation. Mediation analyses provide potential insights into why composite indices might outperform the isolated TyG index. We found that adiposity metrics (WHtR and BMI) maintain substantial associations with cardiovascular outcomes that are statistically independent of the TyG index. Further, the relatively small magnitude of the indirect associations observed in our analysis implies that the TyG index only partially accounts for the relationship between adiposity and vascular outcomes. This suggests that while TyG captures the metabolic toxicity of IR, it may not fully account for the full spectrum of obesity-related cardiovascular risk. When integrated with central (WHtR or WC) or general (BMI) adiposity, these composites may better index increased hemodynamic load, ectopic fat deposition, low-grade inflammation, endothelial dysfunction, oxidative stress, and pro-thrombotic activation, all of which contribute to the development of hypertension and the acceleration of carotid atherogenesis63–66. These processes have also been shown to promote arterial stiffening and intimal lipid deposition, which could theoretically explain why the combined metrics offer better risk stratification than single indicators. However, given the cross-sectional design of our study, temporal relationships between exposure, mediator, and outcome cannot be established. Therefore, these mediation findings must be interpreted cautiously as strictly exploratory and hypothesis-generating, rather than providing definitive mechanistic or causal evidence. Strengths and limitations The present study has several strengths. First, it represents one of the largest cross-sectional investigations to date, including 154,873 Chinese adults, which provides robust statistical power and enhances the representativeness of the findings. Second, we comprehensively evaluated the TyG index and its composite measures with obesity indicators (TyG-WC, TyG-WHtR, and TyG-BMI) in relation to both hypertension and CAS, allowing for a direct comparison of their discriminative ability. Third, the use of RCS modeling, ROC analysis, and stratified analyses by age and sex enabled us to examine nonlinear dose–response relationships, assess discriminative ability, and explore effect modification, thereby yielding nuanced insights into population subgroups. However, several limitations should be acknowledged. First, the cross-sectional design precludes establishing causal relationships between TyG-related indices and the risk of hypertension or CAS. Second, reliance on a single-center health check-up population may introduce selection bias and limit generalizability, although our stratified analyses yielded directionally consistent results to partly mitigate this. Third, despite extensive covariate adjustments, residual confounding remains possible. Notably, data on glycated hemoglobin and specific medications (e.g., antihypertensive, lipid-lowering, or antidiabetic drugs) were unavailable. The potential underrepresentation of participants actively taking lipid- or glucose-lowering medications may have introduced selection bias by removing higher-risk individuals with established metabolic severity. As a result, the associations observed in our relatively healthier data might be conservative, potentially underestimating the true magnitude of these relationships. Fourth, self-reported diagnoses for covariates like diabetes and hypercholesterolemia may introduce recall bias. Fifth, the discriminative ability of the TyG-related indices for CAS was modest (AUCs: 0.56–0.60). While valuable for population-level epidemiological risk stratification, they lack sufficient predictive accuracy to serve as standalone clinical screening tools. Future work should include multi-center longitudinal cohorts to externally validate these findings across diverse settings and ethnic groups, as well as to verify the causal mechanisms underlying these associations. Conclusions This study demonstrates that adiposity-integrated TyG indices, particularly TyG-WHtR, are more strongly associated with hypertension and CAS compared to the isolated TyG index. Crucially, we observed distinct phenotypic variations in these associations based on age and sex. While TyG-WHtR demonstrates the most robust association with hypertension, especially in younger adults where the magnitude of association is amplified, the relationship with CAS exhibits a sex-specific pattern: TyG-WHtR shows a stronger association in women, whereas TyG-BMI appears to be a more relevant associative marker for men. These findings suggest that these low-cost, accessible metrics could potentially inform a more tailored approach to metabolic assessment, prioritizing early screening in high-risk younger populations and adapting vascular risk evaluation based on sex. However, as associative markers, their clinical utility as causal predictors requires further validation in longitudinal studies. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2