Triglyceride-Glucose-Waist-to-Height Ratio Index and Left Ventricular Diastolic Function in Military Athletes Authors: Yen-Chen Lin, Wei-Chieh Huang, Sung-Hao Huang, Wen-Chun Yu, Xuemei Sui, Carl J Lavie, Gen-Min Lin Published: 2026 ABSTRACT AbstractBackgroundThe triglyceride-glucose-waist-to-height ratio (TyG-WHtR) predicts heart failure, but its association with left ventricular diastolic function (LVDF) in physically fit individuals is unclear.ObjectivesThis study aimed to assess whether TyG-WHtR independently associates with echocardiographic LVDF indicators and exercise performance in physically active military personnel.MethodsIn this cross-sectional analysis, 151 male and 25 female military personnel aged 18 to 44 years underwent biochemistry, echocardiography, and fitness testing (3-km run, push-ups, and sit-ups). Men were dichotomized at the median TyG-WHtR (3.954). Kendall correlations with bootstrap 95% CIs and Benjamini–Hochberg false discovery rate adjustment, plus theory-driven multivariable regression adjusted for age, sex, heart rate, atherogenic lipids, and TyG-WHtR, were used; multiplicative sex interaction was tested.ResultsHigher- vs lower-TyG-WHtR men (4.41 [4.13, 4.74] vs 3.62 [3.38, 3.82]) were older (29 [23, 36] vs 25 [21, 30] years), with adverse cardiometabolic profile, lower mitral e′ velocity and E/A, and worse exercise performance. TyG-WHtR was age-independently correlated with e′ velocity (τ −0.382; 95% CI: −0.462 to −0.297) and E/A (−0.265; −0.367 to −0.165), both Benjamini–Hochberg-adjusted P < 0.001. In regression, TyG-WHtR was the leading e′ predictor (standardized β −0.367; −0.491 to −0.242). Sex × TyG-WHtR interaction on e′ velocity was statistically detectable (P = 0.012) but underpowered.ConclusionsTyG-WHtR is age-independently associated with subclinical LVDF decline and lower exercise capacity in young, physically active military personnel. Whether insulin resistance reduction could improve diastolic function requires further study.Key words: insulin resistance, LV diastolic function, military athletes, TyG index, WHtR The triglyceride-glucose-waist-to-height ratio (TyG-WHtR) index1 is the product of the triglyceride-glucose (TyG) index2 and the waist-to-height ratio (WHtR), and predicts incident heart failure (HF) in general and high-risk populations including the elderly.3, 4, 5 TyG represents insulin resistance (IR) and WHtR represents abdominal obesity; both are pathogenic in metabolic syndrome (MetS) and disturbed myocardial metabolism.6,7 IR and MetS have been linked to declined left ventricular (LV) diastolic function (LVDF),8, 9, 10, 11, 12, 13, 14, 15 yet whether TyG-WHtR is associated with LVDF in physically fit individuals is unclear. While the prevalence of overt diastolic dysfunction in physically active young populations is low, whether higher TyG-WHtR scores correlate with subclinical LVDF decline or with loss of training-related supranormal LVDF has not been studied. Additionally, while exercise training reduces IR,15 whether TyG-WHtR scores obtained under long-term physical training predict exercise performance is unknown. We therefore enrolled young military personnel with uniform physical training, diet, and sleep schedules to control these confounders and examined the association of TyG-WHtR with echocardiographic LVDF indicators and exercise performance in this relatively healthy population. Methods This study was approved by the Institutional Review Board/Ethics Committee of Mennonite Christian Hospital, Hualien City, Taiwan (No. 16-05-008). All participants provided written informed consent. Study population Premiddle-aged physically active military personnel were enrolled. Participants underwent uniform physical training, food types, and meal/sleep schedules, with regular examinations including a 3-km field run and 2-minute sit-ups and push-ups.16 Body mass–scaled estimated maximal oxygen consumption (sVO2max) was calculated as: sVO2max = 67.151 − 0.0386 × 3-km running time in seconds, using a formula externally validated against directly measured VO2max in young Asian adults.17,18 Fasting blood samples (12-hour fast) were assayed for fasting plasma glucose (FPG), triglyceride (TG), high- and low-density lipoprotein and total cholesterol (HDLC, LDLC, TC), and liver and renal function tests. Non-HDLC (NHDLC) was calculated as TC − HDLC. Echocardiography and LVDF evaluation Transthoracic M-mode and 2-dimensional echocardiography were performed by the same well-trained echocardiographer using a Philips iE33 machine with a 1 to 5 MHz probe. Speckle-tracking echocardiography in apical 4-chamber views measured left atrial (LA) reservoir strain (LARS) using Philips automated Cardiac Motion Quantification software. Tissue Doppler echocardiography measured lateral mitral annulus e′ velocity. We acknowledge that contemporary guidelines recommend both septal and lateral e′ velocities to determine LVDF; lateral e′ alone has nonetheless been used in clinical and research practice including normal cutoffs.19, 20, 21, 22 The CHIEF (Cardiorespiratory Fitness and Hospitalization Events in Armed Forces) Heart Study used the lateral mitral annulus alone for tissue-Doppler indices in a comparable Asian military cohort.23 LV mass was calculated using the Devereux–Reichek formula, and body surface area (BSA) by the Du Bois formula. BSA-indexed cardiac chamber dimensions, including LV end-diastolic diameter index (LVEDDI), were used. LA parameters and composite indices were as previously defined.24 Declined LVDF was determined echocardiographically per contemporary guidelines.19 Mitral e′ velocity reflects early LV relaxation; lateral e′ ≤7 cm/s suggests impaired relaxation. Lateral E/e′ ≥13 suggests probable elevated LA pressure. LARS ≤18% suggests more than mildly declined LVDF given other supportive evidence. E/A shows a V-shaped pattern with progressive dysfunction; E/A ≥2 with elevated LA pressure differentiates grade 3 from grade 2 dysfunction in the general population. Conversely, high E/A and e′ velocity together with normal E/e′ indicates supranormal LVDF, which is sometimes observed in endurance athletes.22,25 No consensus cutoffs currently define supranormal LVDF. Definition of metabolic syndrome and obesity indices MetS was defined per National Cholesterol Education Program Adult Treatment Panel III 200526 as ≥3 of: 1) waist circumference ≥90 cm (men) or ≥80 cm (women); 2) TG ≥150 mg/dL or treatment; 3) HDLC <40 mg/dL (men) or <50 mg/dL (women) or treatment; 4) systolic blood pressure ≥130 mm Hg or diastolic blood pressure ≥85 mm Hg or treatment; 5) FPG ≥100 mg/dL or treatment. Obesity indices used included body mass index (BMI), WHtR, Body Roundness Index, Weight-adjusted Waist Index, Conicity Index, and A Body Shape Index.27, 28, 29, 30 Non–insulin-based insulin resistance indices Insulin levels were not measured, so non–insulin-based insulin resistance (NIIR) indices were used. Group-1 indices use blood glucose, lipid, and liver function: TyG = ln[TG×FPG/2]; cholesterol-HDL-glucose (CHG) index = ln[TC×FPG/(2×HDLC)];31 atherogenic index of plasma = log(TG/HDLC); Castelli Risk Index I = TC/HDLC; Castelli Risk Index II = LDLC/HDLC; alanine aminotransferase to aspartate aminotransferase (ALT/AST) ratio. Group-2 indices combine group-1 with anthropometry: TyG-WHtR = TyG×WHtR; TyG-BMI = TyG×BMI; Metabolic Score for Insulin Resistance (METS-IR) = ln[2×FPG + TG]×BMI/ln(HDLC);32 METS-IR(BMI→WHtR) = ln[2×FPG + TG]×WHtR/ln(HDLC); Zhejiang University index (ZJU) index = BMI + FPG(mmol/L) + TG(mmol/L) + 3×ALT/AST (+2 if female).33 Statistical analysis Analyses were performed in R (RStudio). Two-tailed P < 0.05 defined significance. Continuous variables were tested for normality using Shapiro–Wilk and Anderson–Darling tests; normally distributed variables are reported as mean ± SD, otherwise median [Q1, Q3]. Categorical variables are reported as n/N (%). Welch’s t-test or the Mann-Whitney U test compared 2 independent groups; Fisher exact test was used for low-frequency categorical variables (eg, MetS). Kendall (rather than Spearman or Pearson) correlation was used because of the non-normal distribution of many variables and Kendall’s robustness to outliers. Bootstrap percentile 95% CIs (2,000 resamples) are reported for all Kendall τ coefficients. Partial Kendall correlations controlling age (or other variables as specified) were computed using ppcor::pcor() in R.34 Because the NIIR indices share TG, FPG, lipid, and/or anthropometric components and are therefore highly correlated rather than independent, Bonferroni correction would be inappropriately conservative. Benjamini–Hochberg (BH) false discovery rate adjustment was applied within each LVDF outcome family (e′ velocity treated as the primary outcome; E/A, E/e′, and LARS as secondary outcomes); both raw and BH-adjusted P values are reported. Univariable screening of all candidate predictors for e′ velocity (Supplemental Table 4) and sVO2max (Supplemental Table 5) is provided for transparency. Multivariable linear regression covariates were prespecified a priori on physiological grounds. For e′ velocity, a priori covariates were age (e′ declines with aging), heart rate (affects diastolic filling time), sex, LDLC (atherogenic burden), and TyG-WHtR (primary exposure). For sVO2max, a priori covariates were age, sex, height (allometric scaling of running performance17), LVEDDI (LV size as a marker of training-induced adaptation), estimated glomerular filtration rate (eGFR) (general health), and TyG-WHtR. Variance inflation factors were verified <5 using car::vif().35 Automated stepwise models are reported as sensitivity analyses (Supplemental Table 6). Both raw and standardized β coefficients with 95% CIs are reported because raw coefficients across variables on different scales are not directly comparable. Multiplicative TyG-WHtR × sex interaction was tested; sex-stratified Kendall correlations are reported descriptively. Added-variable plots with 95% confidence and prediction intervals were generated using car::avPlots() with custom prediction-interval overlay. All figures used ggplot2 and patchwork. Results Study population characteristics and TyG-WHtR-grouped comparisons Among 151 male and 25 female physically active military personnel, female participants were younger, had lower NIIR scores (TyG-WHtR 3.43 [3.20, 3.80]), and lower exercise performance; analyses involving sex as a substantive confounder are reported with sex either adjusted or stratified. Male participants were dichotomized at the male median TyG-WHtR (3.954) (Table 1, Supplemental Table 1). Higher- vs lower-TyG-WHtR men were older (29 [23, 36] vs 25 [21, 30] years), with higher systolic blood pressure, diastolic blood pressure, FPG, TC, NHDLC, LDLC, TG, AST, ALT, and obesity and NIIR indices, and lower HDLC; height, resting heart rate, and eGFR were similar. MetS prevalence was 19/75 (25.3%) in the higher-TyG-WHtR group vs 0/76 (0%) in the lower group (Fisher exact P < 0.001). Echocardiographic LV indices showed lower LVEDDI, lower mitral E/A and e′ velocity, higher mitral A-wave velocity and E/e′, and similar E-wave velocity, LV mass index, LV ejection fraction, and LV global longitudinal strain in higher- vs lower-TyG-WHtR men. LA size differences (LAD, LAVmax) disappeared after BSA indexing, while myocardial-deformation–based LA stiffness index (ie, (E/e′)/LARS) was higher in higher-TyG-WHtR men. Higher TyG-WHtR was associated with worse 3-km run time, push-ups, and sit-ups. Table 1Features of TyG-WHtR–Grouped Male Physically Active Military Personnel, With Female Participants Shown for ContextG1: Men (n = 75)Higher TyG-WHtRG2: Men (n = 76)Lower TyG-WHtRG1 vs G2P ValueG3: Women (n = 25)Essential characteristics and biochemical blood tests (see Supplemental Table 1) Age (y)29 [23, 36]25 [21, 30]0.00323 [21, 28] Weight (kg)79.6 ± 11.368.0 ± 7.92<0.00159.8 ± 7.60 BMI (kg/m2)26.3 [24.9, 28.4]22.7 [21.4, 24.2]<0.00123.0 [21.0, 24.9] WC (cm)89.5 ± 6.4377.2 ± 5.82<0.00173.8 ± 6.84 WHtR0.52 ± 0.0320.45 ± 0.03<0.0010.457 ± 0.043 SBP (mm Hg)124 ± 10.2118 ± 11.0<0.001113 ± 9.32 DBP (mm Hg)76.2 ± 9.4270.6 ± 8.53<0.00168.1 ± 8.08 FPG (mg/dL)92 [87, 97]89 [84, 93]0.00688 [81, 91] TC (mg/dL)180 [160, 204]165 [154, 182]0.003162 [152, 188] HDLC (mg/dL)47.5 ± 9.0354.9 ± 9.38<0.00160.3 ± 8.07 NHDLC (mg/dL)132 ± 27.8112 ± 25.1<0.001110 ± 30.6 LDLC (mg/dL)118 ± 26.7106 ± 23.90.005103 ± 30.6 TG (mg/dL)109 [78.5, 172]62 [52.8, 83.2]<0.00153 [43, 74] AST (U/L)20 [18, 26]18 [16, 23.2]0.00416 [14, 18] ALT (U/L)23 [17.5, 42.5]16 [13, 23.2]<0.00111 [9, 13] MetS, n/N (%)19/75 (25.3%)0/76 (0%)<0.0010/25 (0%)NIIR indices TyGa8.59 ± 0.5357.98 ± 0.354<0.0017.80 ± 0.369 TyG × WHtR4.41 [4.13, 4.74]3.62 [3.38, 3.82]<0.0013.43 [3.20, 3.80] METS-IRb39.1 [36.9, 42.7]31.2 [29.1, 34.3]<0.00130.7 [27.5, 33.9] ZJUc36.7 [34.8, 39.2]31.8 [29.9, 33.3]<0.00132.9 [30.3, 34.3] CHGd5.17 ± 0.2514.9 ± 0.181<0.0014.78 ± 0.224 TC/HDLC3.89 [3.17, 4.36]3.08 [2.68, 3.39]<0.0012.62 [2.46, 2.97] log(TG/HDLC)0.391 [0.229, 0.528]0.0726 [−0.0607, 0.193]<0.001−0.0134 [−0.151, 0.114] ALT/AST1.16 [0.939, 1.57]0.906 [0.736, 1.21]<0.0010.7 [0.529, 0.842]Sonographic LV indices (see Supplemental Table 1) LVEDDI (mm/m2)25.7 ± 2.0827.3 ± 2.08<0.00128.1 ± 2.09 Mitral A-wave velocity (cm/s)54.9 [49.3, 60.4]47.9 [43.7, 56.0]<0.00153.1 [48.3, 62.5] Mitral E/A1.59 [1.30, 1.88]1.78 [1.54, 2.09]<0.0011.79 [1.67, 2.03] Mitral e' (cm/s)15.6 ± 3.3418.7 ± 3.42<0.00118.6 ± 2.79 Mitral E/e'5.52 [4.83, 6.45]4.88 [4.26, 5.29]<0.0015.46 [4.89, 5.94]Sonographic LA indices (see Supplemental Table 1) LAD (mm)35 [33, 36]32 [29, 34]<0.00130 [27, 33] LAVmax (mL)37.7 ± 16.332.0 ± 14.40.04727.0 ± 8.07 LASI_me0.152 [0.127, 0.196]0.143 [0.115, 0.162]0.0290.132 [0.119, 0.171]Exercise performance 3-km running time (s)925 [860, 991]845 [789, 876]<0.0011,020 [960, 1,054] sVO2max (mL/min/kg)f31.4 [28.9, 34.0]34.5 [33.4, 36.7]<0.00127.8 [26.5, 30.1] Push-ups (count)42 [40, 49.2]46.5 [42, 55]<0.00125 [21.5, 35] Sit-ups (count)42 [40, 45]45 [44.5, 50]<0.00135 [32, 41]ALT = alanine transaminase; AST = aspartate aminotransferase; BMI = body mass index; BSA = body surface area; CHG = cholesterol-HDL-glucose; DBP = diastolic blood pressure; FPG = fasting plasma glucose; HDLC = high-density lipoprotein cholesterol; LA = left atrium; LAD = LA anteroposterior diameter; LASI_m = myocardial deformation–based LA stiffness index; LAVmax = LA maximum volume; LDLC = low-density lipoprotein cholesterol; LV = left ventricle; LVEDDI = body surface area-indexed LV end-diastolic diameter; MetS = metabolic syndrome; METS-IR = Metabolic Score for Insulin Resistance; NHDLC = non-HDL cholesterol; NIIR = non–insulin-based insulin resistance; SBP = systolic blood pressure; sVO2max = body mass-scaled estimated maximal oxygen consumption; TC = total cholesterol; TG = triglyceride; TyG = triglyceride-glucose; TyG-WHtR = triglyceride-glucose-waist-to-height ratio; WC = waist circumference; WHtR = waist-to-height ratio; ZJU = Zhejiang University index.aTyG = ln[TG(mg/dL) × FPG(mg/dL)/2].bMETS-IR = ln[2×FPG(mg/dL) + TG(mg/dL)] × BMI/ln[HDLC(mg/dL)].cZJU = BMI(kg/m2) + FPG(mmol/L) + TG(mmol/L) + 3 × ALT/AST (+2 if female).dCHG = ln[TC(mg/dL) × FPG(mg/dL)/(2 × HDLC(mg/dL))].eLASI_m = (E/e′)/LA reservoir strain.fsVO2max = 67.151 − 0.0386 × 3-km running time (sec). Features of TyG-WHtR–Grouped Male Physically Active Military Personnel, With Female Participants Shown for Context ALT = alanine transaminase; AST = aspartate aminotransferase; BMI = body mass index; BSA = body surface area; CHG = cholesterol-HDL-glucose; DBP = diastolic blood pressure; FPG = fasting plasma glucose; HDLC = high-density lipoprotein cholesterol; LA = left atrium; LAD = LA anteroposterior diameter; LASI_m = myocardial deformation–based LA stiffness index; LAVmax = LA maximum volume; LDLC = low-density lipoprotein cholesterol; LV = left ventricle; LVEDDI = body surface area-indexed LV end-diastolic diameter; MetS = metabolic syndrome; METS-IR = Metabolic Score for Insulin Resistance; NHDLC = non-HDL cholesterol; NIIR = non–insulin-based insulin resistance; SBP = systolic blood pressure; sVO2max = body mass-scaled estimated maximal oxygen consumption; TC = total cholesterol; TG = triglyceride; TyG = triglyceride-glucose; TyG-WHtR = triglyceride-glucose-waist-to-height ratio; WC = waist circumference; WHtR = waist-to-height ratio; ZJU = Zhejiang University index. TyG = ln[TG(mg/dL) × FPG(mg/dL)/2]. METS-IR = ln[2×FPG(mg/dL) + TG(mg/dL)] × BMI/ln[HDLC(mg/dL)]. ZJU = BMI(kg/m2) + FPG(mmol/L) + TG(mmol/L) + 3 × ALT/AST (+2 if female). CHG = ln[TC(mg/dL) × FPG(mg/dL)/(2 × HDLC(mg/dL))]. LASI_m = (E/e′)/LA reservoir strain. sVO2max = 67.151 − 0.0386 × 3-km running time (sec). Identification of supranormal LVDF before further analysis Some participants had E/A ≥2 (Figure 1A), which can indicate either severe diastolic dysfunction or supranormal LV relaxation; clarification was therefore required before subsequent analyses. Participants with high E/A also had higher e′ velocity (Figure 1A), normal E/e′ (<13) (Figure 1B), and normal LARS (>18%) (Figure 1C) without other evidence of LVDF decline, consistent with supranormal LV relaxation rather than dysfunction. Presence of Supranormal Left Ventricular Diastolic Function in Some Study Participants Participants with high mitral E/A have (A) normal to high e′ velocity and normal (B) E/e′ and (C) LARS and are unlikely to have declined LVDF. E/A = mitral E-wave to A-wave velocity ratio; LARS = left atrial reservoir strain; LVDF = left ventricular diastolic function; τ = Kendall’s correlation coefficient. NIIR indices and LVDF parameters: correlations with BH-false discovery rate adjustment Kendall correlations between TyG-WHtR-related and other NIIR indices and echocardiographic LV systolic and diastolic parameters were computed for the whole study population (Table 2, Figure 2). After BH-false discovery rate adjustment within each LVDF indicator family, all 12 NIIR indices retained statistically significant negative correlations with mitral e′ velocity (smallest BH-adjusted P = 8.4 × 10−13 for TyG-WHtR, largest = 2.6 × 10−6 for ALT/AST) and with E/A (smallest BH-adjusted P = 2.9 × 10−8 for CHG, largest = 0.002 for ALT/AST). TyG-WHtR-related indices correlated more strongly with mitral e′ velocity than with E/A. Their positive correlations with E/e′ were largely abolished after controlling for e′ velocity (partial Kendall τ near zero with 95% CIs crossing zero for 11 of 12 NIIR indices) (Supplemental Table 2); the exception, ZJU, retained a small residual partial τ of +0.122 [95% CI + 0.030 to +0.215]), and NIIR indices were either weakly negatively correlated or uncorrelated with the E-wave velocity (the numerator of E/e′), supporting the interpretation that the NIIR–E/e′ association is largely mediated through reduced e′ velocity rather than reflecting elevated LV filling pressure. Kendall correlations between all 12 NIIR indices and additional echocardiographic indices—LARS, LAVI, LV ejection fraction, and LV global longitudinal strain—were all insignificant (all P > 0.05 and |τ| < 0.10) and are not shown. Replacing BMI with WHtR in METS-IR (ie, METS-IR[BMI→WHtR]) improved correlation strength vs the original METS-IR (Supplemental Figure 1). Multiplying group-1 NIIR indices (LDLC/HDLC, TC/HDLC, NHDLC, log[TG/HDLC], CHG, ALT/AST) by WHtR or BMI—especially the former—generally improved correlations with e′ velocity (Supplemental Figure 2). In summary, higher NIIR scores were associated with lower e′ velocity, lower E/A, and lower probability of training-related supranormal LVDF. Table 2Kendall Correlations Between NIIR Indices and Sonographic LV Functional ParametersLVDF ParameterNIIR Indexnτ [95% CI]P (Raw)P (BH-FDR)e′ velocityTyG-WHtR176−0.382 [−0.462, −0.297]7.0 × 10−148.4 × 10−13e′ velocityMETS-IR(BMI→WHtR)176−0.361 [−0.444, −0.271]1.4 × 10−128.5 × 10−12e′ velocityCHG176−0.348 [−0.430, −0.261]9.1 × 10−123.6 × 10−11e′ velocityTyG-BMI176−0.344 [−0.426, −0.259]1.4 × 10−113.9 × 10−11e′ velocitylog(TG/HDL)176−0.343 [−0.430, −0.255]1.6 × 10−113.9 × 10−11e′ velocityTC/HDL176−0.337 [−0.421, −0.248]4.0 × 10−117.4 × 10−11e′ velocityTyG176−0.336 [−0.428, −0.243]4.3 × 10−117.4 × 10−11e′ velocityMETS-IR176−0.329 [−0.414, −0.239]1.1 × 10−101.7 × 10−10e′ velocityZJU176−0.322 [−0.403, −0.234]2.7 × 10−103.5 × 10−10e′ velocityNHDL176−0.318 [−0.408, −0.218]5.1 × 10−106.1 × 10−10e′ velocityLDL/HDL176−0.307 [−0.397, −0.214]1.7 × 10−91.8 × 10−9e′ velocityALT/AST176−0.240 [−0.335, −0.145]2.6 × 10−62.6 × 10−6E/ACHG176−0.304 [−0.396, −0.210]2.4 × 10−92.9 × 10−8E/ATC/HDL176−0.270 [−0.359, −0.175]1.1 × 10−76.8 × 10−7E/ATyG-WHtR176−0.265 [−0.367, −0.165]1.9 × 10−77.4 × 10−7E/Alog(TG/HDL)176−0.262 [−0.360, −0.161]2.7 × 10−77.4 × 10−7E/ALDL/HDL176−0.261 [−0.350, −0.162]3.1 × 10−77.4 × 10−7E/AMETS-IR(BMI→WHtR)176−0.257 [−0.355, −0.155]4.5 × 10−77.7 × 10−7E/ANHDL176−0.257 [−0.343, −0.161]5.1 × 10−77.7 × 10−7E/ATyG176−0.256 [−0.359, −0.146]5.1 × 10−77.7 × 10−7E/ATyG-BMI176−0.240 [−0.345, −0.136]2.4 × 10−63.2 × 10−6E/AMETS-IR176−0.232 [−0.335, −0.123]5.2 × 10−66.2 × 10−6E/AZJU176−0.209 [−0.307, −0.109]4.2 × 10−54.5 × 10−5E/AALT/AST176−0.157 [−0.251, −0.063]0.00210.0021HDL = high-density lipoprotein; LARS = left atrial reservoir strain; LAVI = body surface area-indexed left atrial maximum volume; LDL = low-density lipoprotein; LVDF = left ventricular diastolic function; LVEF = left ventricular ejection fraction; LVGLS = left ventricular global longitudinal strain; other abbreviations as in Table 1.τ, Kendall's tau rank correlation coefficient with bootstrap percentile 95% CIs (2,000 resamples; set.seed(42)). BH-FDR, Benjamini-Hochberg false discovery rate adjustment applied within each LVDF outcome family (12 NIIR exposures per outcome). Kendall correlations between the 12 NIIR indices and additional echocardiographic indices (LARS, LAVI, LVEF, LVGLS) were not statistically significant (all P > 0.05 and |τ| < 0.10) and are not shown. Kendall Correlations Between NIIR Indices and Sonographic LV Functional Parameters HDL = high-density lipoprotein; LARS = left atrial reservoir strain; LAVI = body surface area-indexed left atrial maximum volume; LDL = low-density lipoprotein; LVDF = left ventricular diastolic function; LVEF = left ventricular ejection fraction; LVGLS = left ventricular global longitudinal strain; other abbreviations as in Table 1. τ, Kendall's tau rank correlation coefficient with bootstrap percentile 95% CIs (2,000 resamples; set.seed(42)). BH-FDR, Benjamini-Hochberg false discovery rate adjustment applied within each LVDF outcome family (12 NIIR exposures per outcome). Kendall correlations between the 12 NIIR indices and additional echocardiographic indices (LARS, LAVI, LVEF, LVGLS) were not statistically significant (all P > 0.05 and |τ| < 0.10) and are not shown. Higher Triglyceride-Glucose-Waist-to-Height Ratio Scores Are Associated With Unavailability of Supranormal Left Ventricular Diastolic Function Higher TyG-WHtR scores negatively correlate with LVDF indicators including (A) E/A, (B) e′ velocity, and (C) E/e′ and are (D) unrelated to LARS in our study population. Correlations between TyG-BMI and either (E) E/A or (F) e′ velocity are provided for comparison. BMI = body mass index; TyG = triglyceride-glucose; TyG-WHtR = triglyceride-glucose-waist-to-height ratio; other abbreviations as in Figure 1. Age- and sex-independent association of TyG-WHtR with LVDF Age, like TyG-WHtR, was negatively correlated with E/A and e′ velocity, positively correlated with E/e′, and unrelated to LARS (Supplemental Figure 3); TyG-WHtR was positively correlated with age (Supplemental Figure 3A). Partial Kendall correlations (controlling for age) confirmed independent negative associations of TyG-WHtR with e′ velocity (τ −0.321; 95% CI: −0.400 to −0.239) and with E/A (τ −0.204; 95% CI: −0.302 to −0.101) (Supplemental Table 3). In a theory-driven multivariable model for e′ velocity (predictors specified a priori on physiological grounds: TyG-WHtR, age, heart rate, LDLC, sex), all predictors except sex contributed independently (Table 3, Supplemental Figure 4). TyG-WHtR was the leading predictor (standardized β −0.367; 95% CI: −0.491 to −0.242, P < 0.001), followed by age (−0.359, −0.481 to −0.238), heart rate (−0.205, −0.315 to −0.094), and LDLC (−0.125, −0.241 to −0.008). The model explained 47.2% of e′ velocity variance (adjusted R2 = 0.456); all variance inflation factors were <1.3. Replacing TyG-WHtR with TyG-BMI yielded a smaller adjusted R2. The automated stepwise model (Supplemental Table 6) reached the same conclusions, supporting robustness. Table 3Theory-Driven Multivariable Linear Regression for Mitral e′ VelocityPredictorβ [95% CI]Standardized β [95% CI]P ValueVIFTyG-WHtR−2.563 [−3.432, −1.694]−0.367 [−0.491, −0.242]<0.0011.28Age−0.212 [−0.284, −0.140]−0.359 [−0.481, −0.238]<0.0011.22Heart rate−0.074 [−0.114, −0.034]−0.205 [−0.315, −0.094]<0.0011.02LDLC−0.019 [−0.036, −0.001]−0.125 [−0.241, −0.008]0.0361.12Sex (female)−0.586 [−1.895, +0.722]−0.052 [−0.167, +0.064]0.3781.10Model: n = 176, adjusted R2 = 0.456, P < 0.001β = beta coefficient; VIF = variance inflation factor; other abbreviations as in Table 1.Theory-driven model with covariates prespecified a priori on physiological grounds: TyG-WHtR (primary exposure); age (e′ velocity declines with aging); heart rate (affects diastolic filling time); LDLC (atherogenic burden); sex (known sex differences in LV diastolic function). All variance inflation factors <1.30. Standardized β allows direct comparison across predictors on different scales. Theory-Driven Multivariable Linear Regression for Mitral e′ Velocity β = beta coefficient; VIF = variance inflation factor; other abbreviations as in Table 1. Theory-driven model with covariates prespecified a priori on physiological grounds: TyG-WHtR (primary exposure); age (e′ velocity declines with aging); heart rate (affects diastolic filling time); LDLC (atherogenic burden); sex (known sex differences in LV diastolic function). All variance inflation factors <1.30. Standardized β allows direct comparison across predictors on different scales. TyG-WHtR and exercise performance Higher- vs lower-TyG-WHtR men achieved lower sVO2max during 3-km field runs, with a substantial sex difference favoring men (Table 1). TyG-WHtR and other NIIR indices were negatively correlated with sVO2max, with women having observably lower sVO2max than men (Figure 3). In the theory-driven multivariable model for sVO2max, all 6 prespecified covariates contributed independently (Table 4, Figure 4): sex (standardized β −0.437; 95% CI: −0.575 to −0.299), LVEDDI (+0.382, 0.230-0.534), eGFR (−0.247, −0.384 to −0.110), height (+0.233, 0.076-0.389), TyG-WHtR (−0.211, −0.353 to −0.069), and age (−0.150, −0.297 to −0.004). Adjusted R2 = 0.410; all VIFs <1.85. Adding LAVI/(E/e′) to the model further increased adjusted R2 to 0.469; TyG-WHtR remained significant (standardized β −0.167, P = 0.033). Splitting TyG-WHtR into TyG and WHtR yielded a similar but slightly weaker model in which at least one component lost significance; replacing WHtR with BMI consistently reduced model performance. Non–Insulin-Based Insulin Resistance Index Scores Negatively Correlate With Body Mass–Scaled Estimated Maximal Oxygen Consumption, and Sex Difference Is an Even Stronger Discriminator of Exercise Performance In general, scores of NIIR indices including (A) TyG-WHtR, (B) TyG-BMI, (C) METS-IR, and (D) METS-IR(BMI→WHtR) fairly correlate with sVO2max, and women vs men largely have observably lower performance of 3-km field run. IR = Metabolic Score for Insulin Resistance; NIIR = non-insulin-based insulin resistance; sVO2max = body mass–scaled estimated maximal oxygen consumption; METS-other abbreviations as in Figures 1 and 2. Table 4Theory-Driven Multivariable Linear Regression for sVO2maxPredictorβ [95% CI]Standardized β [95% CI]P ValueVIFSex (female)−6.175 [−8.122, −4.229]−0.437 [−0.575, −0.299]<0.0011.40LVEDDI+0.774 [+0.466, +1.082]+0.382 [+0.230, +0.534]<0.0011.70eGFR−0.070 [−0.110, −0.031]−0.247 [−0.384, −0.110]<0.0011.39Height+0.151 [+0.049, +0.253]+0.233 [+0.076, +0.389]0.0041.81TyG-WHtR−1.776 [−2.968, −0.584]−0.211 [−0.353, −0.069]0.0041.48Age−0.104 [−0.206, −0.002]−0.150 [−0.297, −0.004]0.0451.59Model: n = 171, adjusted R2 = 0.410, P < 0.001eGFR = estimated glomerular filtration rate; LVEDDI = body surface area-indexed left ventricular end-diastolic diameter; other abbreviations as in Tables 1 and 3.Theory-driven model with covariates prespecified a priori on physiological grounds: TyG-WHtR (primary exposure); sex (well-known sex difference in VO2max); height (allometric scaling of running performance); LVEDDI (LV cavity size as marker of training-induced adaptation); age (fitness declines with age); eGFR (general health). All variance inflation factors <1.85. Theory-Driven Multivariable Linear Regression for sVO2max eGFR = estimated glomerular filtration rate; LVEDDI = body surface area-indexed left ventricular end-diastolic diameter; other abbreviations as in Tables 1 and 3. Theory-driven model with covariates prespecified a priori on physiological grounds: TyG-WHtR (primary exposure); sex (well-known sex difference in VO2max); height (allometric scaling of running performance); LVEDDI (LV cavity size as marker of training-induced adaptation); age (fitness declines with age); eGFR (general health). All variance inflation factors <1.85. Added-Variable Plots Showing Results of Partial Linear Regression Predicting Body Mass–Scaled Estimated Maximal Oxygen Consumption Six-panel display showing the partial association of each prespecified covariate with body mass–scaled estimated maximal oxygen consumption (sVO2max) in the model: sVO2max ∼ sex + LVEDDI + eGFR + height + TyG-WHtR + Age. (A) Sex (female = 1, male = 0). (B) LVEDDI. (C) eGFR. (D) Height. (E) TyG-WHtR. (F) Age. Within each panel, both the focal predictor and the outcome are residualized on all other covariates, and the residuals are scattered against each other; the slope of the resulting regression line equals the raw β for that predictor in the full model. Each panel shows the OLS fit with a 95% confidence band (inner, darker navy) and a 95% prediction band (outer, lighter, much wider). Men are plotted in navy circles and women in coral triangles; MetS-positive participants (n = 19) are highlighted with solid black markers. The full-model adjusted R2 is reported under the panel grid. The visual partial slopes are consistent with the standardized β estimates reported in Table 4. eGFR = estimated glomerular filtration rate; std β = standardized beta coefficient; “others,” given other predictors controlled; other abbreviations as in Figures 2 and 3. Sex × TyG-WHtR interaction Multiplicative TyG-WHtR × sex interaction was statistically detectable for e′ velocity (interaction β +3.54; 95% CI: 0.79-6.28; P = 0.012) (Supplemental Figure 5) but not for E/A (P = 0.760) or sVO2max (P = 0.213). Sex-stratified Kendall correlations were stronger in men (e′: τ −0.415, P < 0.001) than women (e′: τ −0.043, P = 0.760). Two methodological caveats limit interpretation: 1) only 25 women were enrolled, providing limited power; and 2) women’s TyG-WHtR distribution was range-restricted (median 3.43 [Q1 3.20, Q3 3.80], maximum 4.77) compared with men (median 3.95, maximum 5.67), mechanically attenuating apparent effect sizes even if the underlying relationships were identical. We therefore present the interaction as hypothesis-generating, requiring confirmation in larger samples of women. Discussion We showed (Central Illustration) that TyG-WHtR is age-independently, monotonically, and negatively correlated with LVDF indicators including mitral e′ velocity and E/A in physically active military personnel. While supranormal early LV relaxation—frequently observed in elite endurance athletes—facilitates LV filling during tachycardia-shortened diastoles to maintain stroke volume, its loss can be expected as TyG-WHtR gradually increases. Atherogenic cholesterols (LDLC and NHDLC) were also negatively correlated with e′ velocity, with stronger correlations after multiplication by WHtR or BMI. TyG-WHtR was the leading predictor of achievable sVO2max in men in multivariable regression. Echocardiographic LA parameters were not directly correlated with TyG-WHtR; however, LAVI/(E/e′) helped predict sVO2max alongside TyG-WHtR. Replacing WHtR with BMI consistently weakened the associations, supporting WHtR as the more informative anthropometric component. Sex was a stronger determinant of sVO2max than TyG-WHtR. Insulin Resistance Surrogates (Triglyceride-Glucose-Waist-to-Height Ratio) and Subclinical Left Ventricular Diastolic Function Decline in Military Athletes Study overview and key findings in 176 physically active military personnel (151 men, 25 women, age 18-44 y) undergoing uniform training, diet, and sleep schedules. Header strip: study population, primary exposure (TyG-WHtR, a non–insulin-based insulin resistance index defined as the product of the triglyceride-glucose index and waist-to-height ratio), and outcomes assessed. (A) Marginal correlation: TyG-WHtR was negatively correlated with mitral e′ velocity in the whole study population (Kendall τ = −0.382; 95% CI: −0.46 to −0.30). (B) Marginal correlation: TyG-WHtR was also negatively correlated with sVO2max; both sexes are plotted, but the Kendall tau is computed on men only (τ = −0.353; 95% CI: −0.44 to −0.26). (C) Adjusted associations: forest plot of standardized β coefficients with 95% CIs from the two prespecified theory-driven multivariable models. TyG-WHtR (highlighted in navy bold) remained significantly associated with both lower e′ velocity (standardized β = −0.367; 95% CI: −0.491 to −0.242; full model adjusted R2 = 0.456) and lower sVO2max (standardized β = −0.211; 95% CI: −0.353 to −0.069; full model adjusted R2 = 0.410), independently of all other prespecified covariates. ∗ indicates P < 0.05. MetS-positive participants (solid black markers in panels A and B; all 19 are men) cluster at high TyG-WHtR with low e′ velocity and low sVO2max. e′ = mitral lateral early diastolic velocity; LDLC = low-density lipoprotein cholesterol; LV = left ventricular; LVEDDI = body surface area-indexed LV end-diastolic diameter; MetS = metabolic syndrome; other abbreviations as in Figure 1, Figure 2, Figure 3, Figure 4. Both MetS and IR have been linked to declined LVDF in general or high-risk populations.8, 9, 10, 11, 12, 13, 14, 15 Few earlier studies adjusted for the well-known negative effect of aging on e′ velocity and E/A. Von Bibra et al showed that MetS was associated with lower e′ even after age correction in echocardiography-indicated patients.10 Kishi et al found that higher IR groups had lower mean e′ velocity in a group-based trajectory model adjusted for age, sex, race, and BMI category.9 Zhou et al showed TyG predicted e′ and E/e′ in age-adjusted multivariable regression.36 Here we show, with both partial Kendall correlation and theory-driven multivariable regression controlling for age, that TyG-WHtR was independently and negatively correlated with e′ velocity and E/A in young, physically active military personnel; the availability of training-related supranormal LVDF decreased as TyG-WHtR increased. e′ velocity correlated more strongly with NIIR indices than did E/A and may be a more sensitive marker of subclinical IR-related diastolic decline in this population. Although current guidelines recommend the use of both septal and lateral e′ velocities,19 the lateral mitral annulus alone has been used in several clinical and research contexts. Park et al directly compared septal and lateral tissue Doppler imaging in 276 outpatients and found that septal e′ tended to overestimate the severity of dysfunction relative to lateral e′.20 The CHIEF Heart Study used lateral e′ alone in a comparable Asian military cohort to characterize the athlete’s heart phenotype.23 The Sharifov et al meta-analysis showed that lateral and average E/e′ had similar sensitivity for elevated LV filling pressure (30% and 37%, respectively), while septal had higher specificity (98% vs 91% and 92%) but lower sensitivity (24%).21 Lateral e′ is also less affected by right ventricle tethering and septal regional remodeling than septal e′.22 Although having only lateral e′ is a limitation, our findings remain interpretable within the existing evidence base; future studies in similar populations will incorporate both septal and lateral e′. IR and abdominal obesity are the cornerstones of MetS, and NIIR indices incorporating both may improve risk stratification. Group-2 NIIR indices have been shown to outperform group-1 in predicting prediabetes in normal-weight adults,37 long-term type 2 diabetes,38 hepatic steatosis in diabetic patients,39 and all-cause mortality in elderly diabetics.40 We extend these observations by showing that multiplying group-1 NIIR indices by obesity indices—with WHtR generally outperforming BMI—improved correlations with e′ velocity, E/A, and exercise performance. We acknowledge, however, that no published study has formally validated TyG-WHtR against gold standard IR measures (Homeostatic Model Assessment of Insulin Resistance] or euglycemic-hyperinsulinemic clamp) specifically in physically active young populations. The closest precedent is the CHIEF cohort41 in which TyG and related NIIR indices were associated with prevalent hypertension in young to middle-aged military personnel; TyG-WHtR has shown clinical associations in National Health and Nutrition Examination Survey adults aged 18 to 60,42 in normal-weight Qatari adults,37 and in Japanese longitudinal cohorts38 but has not been formally validated in athletes or military personnel against direct measures of IR. Exercise training itself reduces fasting TG and may compress the dynamic range of TyG-derived indices, potentially biasing associations toward the null; therefore, the findings we observed should be interpreted as conservative estimates of the true association. Echocardiographic LA size, pressure, and reservoir function indicators are recommended for differentiating mild from severe LVDF.19 We did not observe direct correlations between TyG-WHtR and sonographic LA parameters. Differences in LA diameter and maximum volume disappeared after BSA indexing, and LARS and left atrial emptying fraction were not correlated with TyG-WHtR. Differences from prior studies (eg, Ayalon et al,12 Nyman et al,43 Fang et al44) likely reflect our study population’s relatively low cardiometabolic risk. Nevertheless, both TyG-WHtR and LA parameters such as LAVI/(E/e′) were independently associated with sVO2max. TyG-WHtR was independently and negatively correlated with sVO2max. We previously found that higher mitral e′ velocity, lower WHtR, and larger LVEDDI predict better 3-km running, push-ups, and sit-ups.18 Here, we show that TyG-WHtR is negatively correlated with e′ velocity and LVEDDI and positively correlated with WHtR; it is therefore unsurprising that TyG-WHtR negatively correlates with exercise performance. TyG-WHtR remained significant in stepwise regressions and precluded TyG, WHtR, and e′ velocity—but not LVEDDI—from significance during model construction. From a preventive cardiology standpoint, the central observation is that even within a young, physically active, low-risk population whose individual echocardiographic parameters typically remain in the normal range, gradual increases in TyG-WHtR are accompanied by measurable declines in early LV relaxation and exercise capacity. This pattern is consistent with subclinical insulin-resistance–related cardiometabolic strain that begins well before overt disease and supports the notion that risk stratification with simple, widely available NIIR indices could complement traditional risk factors in young populations. Whether targeted reduction of IR and abdominal obesity in physically active populations preserves diastolic function and exercise capacity over time should be addressed in prospective intervention studies. Study limitations First, this cross-sectional design cannot demonstrate causality or examine effects of intervention on the TyG-WHtR–LVDF relationship over time. Second, the female participant count (n = 25) is small relative to men (n = 151), reflecting the sex composition of military personnel. Sex-stratified estimates and the sex × TyG-WHtR interaction test are accordingly underpowered, and the apparent attenuation of effects in women may reflect both reduced power and a range-restricted exposure distribution rather than a true biological difference. Third, e′ velocity was measured only at the lateral mitral annulus; we could not compare septal vs lateral correlations. Fourth, no consensus cutoff for supranormal LVDF currently exists, so we did not provide a TyG-WHtR threshold for its loss. Fifth, serum insulin was not measured and Homeostatic Model Assessment of Insulin Resistance could not be calculated for direct validation of TyG-WHtR; no published study has formally validated TyG-WHtR against gold standard IR measures specifically in physically active young populations. Sixth, individual medication histories (lipid-lowering and antihypertensive therapy) were not collected as part of the routine annual military health examination on which this study is based; given participants’ youth, fitness, and absence of established cardiovascular disease, baseline pharmacological treatment is expected to be uncommon, but this remains a limitation that future studies should record. Seventh, findings from this young, physically active, low-cardiometabolic-risk population may not generalize to other groups. Conclusions TyG-WHtR scores correlate with echocardiographic LVDF indicators including mitral e′ velocity and E/A in a negative, monotonic, and age-independent fashion in young, physically active military personnel. Higher TyG-WHtR scores were associated with progressive loss of training-related supranormal LVDF. TyG-WHtR did not correlate with LV systolic function or LA structural and functional indices but was negatively correlated with achievable sVO2max in 3-km runs. Whether reducing IR and abdominal obesity preserves diastolic function, supports exercise capacity, and lowers future HF risk in physically active individuals warrants prospective study. COMPETENCY IN MEDICAL KNOWLEDGE: Simple NIIR indices (TyG-WHtR, METS-IR, ZJU) calculable from routine fasting biochemistry and waist measurement are negatively correlated with mitral e′ velocity and exercise capacity even in young, physically active individuals whose individual echocardiographic parameters typically remain in the normal range. TRANSLATIONAL OUTLOOK: Whether targeted lifestyle and pharmacological reduction of IR and abdominal obesity in physically active young populations preserves LVDF and exercise capacity over time and reduces future HF risk should be tested in prospective intervention studies. Funding support and author disclosures This study was supported by the Medical Affairs Bureau, Ministry of National Defense (MND-MAB-D-115254), and the Hualien Armed Forces General Hospital (HAFGH-D-115004), Taiwan. The funders had a role in study design, data collection, analysis, and writing. The authors have reported that they have no relationships relevant to the contents of this paper to disclose. Footnotes The authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the Author Center. Appendix For supplemental tables and figures, please see the online version of this paper. Appendix Supplemental Material References Associated Data Supplementary Materials Supplemental Material