This cross-sectional observational study adhered to the Declaration of Helsinki and received approval from the Ethics Committee of the First Affiliated Hospital of Dalian Medical University (Approval No: PJ-KS-KY-2025-176). The ethics committee waived the requirement for written informed consent due to the retrospective nature of the study.

Between October 2021 and September 2024, we consecutively enrolled 264 overweight/obese adults (BMI ≥ 24 kg/m²) with essential hypertension, aged 18–75 years. Hypertension was defined according to 2024 hypertension Guidelines (15, 16). Overweight/obesity status was defined based on Chinese national criteria (BMI ≥ 24 kg/m²) (17). Exclusion criteria were as follows: secondary hypertension; coronary heart disease (prior angina pectoris, myocardial infarction, revascularization, or >50% coronary stenosis on CCTA); heart failure [left ventricular ejection fraction (LVEF) < 50%]; moderate or severe valvular disease; atrial fibrillation; severe hepatic or renal dysfunction [estimated glomerular filtration rate (eGFR) < 30 mL/min*1.73 m²]; malignancies; and inadequate echocardiographic image quality precluding reliable 2D speckle-tracking analysis. All participants underwent comprehensive assessment, including physical examination, laboratory tests, echocardiography, and CCTA.

Demographic and hospitalization data were extracted from electronic medical records using standardized case-report forms. Key parameters included: anthropometrics (height, weight, blood pressure and heart rate); metabolic biomarkers fasting plasma glucose [FPG], glycated hemoglobin [HbA1c], total cholesterol [TC], triglyceride [TG], low-density lipoprotein cholesterol [LDL-C], high-density lipoprotein cholesterol [HDL-C], lipoprotein (a) [Lp(a)], apolipoprotein-A 1[Apo-A1], apolipoprotein-B [Apo-B], serum creatinine [Scr], uric acid [UA], and high-sensitivity C-reactive protein [hs-CRP]. The eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula (18). Additionally, we calculated the indices: the TyG using the formula: ln [TG (mg/dL) × FPG (mg/dL)/2], and the TyG-BMI as the product of TyG and BMI.

All patients underwent echocardiographic evaluation using a commercially available system (Vivid E9, GE Vingmed Ultrasound, Horten, Norway) equipped with a 2.5–5.0 MHz phased-array transducer. Parameter measurements followed the guidelines of the American Society of Echocardiography (19). To ensure optimal image quality, particularly in overweight/obese patients, all examinations were conducted by experienced sonographers.

During examinations, patients were instructed to maintain calm breathing in the left lateral decubitus position. Electrocardiograms (ECGs) were recorded simultaneously, and two-dimensional ultrasound was used to measure parameters, including LV end-diastolic diameter (LVEDD), LV posterior wall thickness at end-diastole (LVPWTD), interventricular septum thickness at end-diastole (IVSTD), and LA diameter (LAD). LA maximal volume (LAV) was measured at end-systole using the biplane method of discs (modified Simpson’s rule) in apical 4- and 2-chamber views. LVEF was quantified via the same method. LV mass (LVM, g) was calculated using the Devereux formula: LVM = 0.8 × 1.04 × [(IVSTD + LVEDD + LVPWTD)³ − LVEDD³] + 0.6, and LVM index (LVMI, g/m²) = LVM/Body surface area (BSA). LAV index (LAVI) was calculated as LAV divided by BSA. Pulsed-wave Doppler was used to measure early (E) and late (A) mitral inflow velocities, and the E/A ratio was calculated. The deceleration time of early mitral inflow (EDT) was recorded. Tissue Doppler imaging was performed to obtain the myocardial velocity (e’) of the lateral mitral annulus in the apical 4-chamber view, and the E/e’ ratio was calculated.

Offline 2D-STE analysis was performed using EchoPAC software (version 202; GE Vingmed Ultrasound). Only echocardiographic loops with adequate image quality for accurate speckle-tracking analysis were included; loops with poor endocardial border definition were excluded. LA strain parameters, LA reservoir strain (LAs-s), LA conduit strain (LAs-e), and LA booster strain (LAs-a), were derived from apical 4- and 2-chamber views and average (20) (Figure 1A). The LA stiffness index (LASI) was defined as [E/e’]/LA_S-S (21).

EAT quantification utilized dual-source CT (Somatom Definition Flash or Force, Siemens Healthineers, Germany). Prior to CCTA, patients received breath-hold training (>15 s) and antecubital intravenous access for iodinated contrast injection. All patients were administered 0.25 mg sublingual nitroglycerin (AstraZeneca) 5 min pre-scan to optimize coronary vasodilation. All patients maintained heart rate below 80 bpm during CCTA. Those exceeding this threshold received oral metoprolol (25–100 mg) 1 h pre-scan, with continuous blood pressure monitoring to detect hypotension. Patients were positioned supine with both arms elevated, and ECG gating was applied. Scanning extended from the tracheal carina to 1 cm below the diaphragm. Scout images optimized cardiac centering prior to acquisition.

Scanning parameters included: 250 ms rotation time; detector collimation 2 × 64 × 0.6 mm (Somatom Definition Flash) or 2 × 96 × 0.6 mm (Somatom Force); 66 ms temporal resolution; reconstructed slice thickness 0.75 mm with 0.7 mm increment. Tube setting (100–120 kV, 200–400 mAs) employed automatic dose modulation. Contrast administration followed a triphasic protocol: (1) 40–50 mL iodinated contrast (370 mgL/mL) at 5.5–6 mL/s. (2) 20–30 mL contrast at 3.5–4.0 mL/s. (3) 30 mL saline chaser at 4.0 mL/s.

CCTA images (Siemens Somatom Definition Flash/Force) were analyzed on a SyngoVia workstation (VB60, Siemens Healthineers, Germany). DICOM datasets underwent semi-automated segmentation of cardiac structures, followed by EAT quantification. EAT volume (EATV) and density were measured with a standardized attenuation threshold of −190 to −30 Hounsfield units (HU) (22) via the anatomical structure display module (Figure 1B).

Data normality was assessed using the Kolmogorov–Smirnov test. Normally distributed continuous variables are reported as mean ± standard deviation (SD), and non-normally distributed variables as median with interquartile range (IQR; 25th-75th percentiles). Categorical variables are presented as percentages (%). Between-group comparisons were performed using independent Student t-tests for normally distributed continuous variables, while Mann–Whitney U tests for non-normally distributed variables. The chi-square tests (χ² tests) were applied for categorical variables. Correlation analysis employed Pearson’s correlation for normally distributed variables and Spearman’s rank correlation for non-normally distributed variables. Multivariable linear regression models were constructed to evaluate independent associations between EATV and LA strain parameters, with sequential adjustment for clinically relevant covariates. The results are presented as unstandardized regression coefficients (β), and the goodness-of-fit for each model was assessed using the adjusted R² value. Model 1: sex, age, and BMI. Model 2: sex, age, duration of hypertension, history of diabetes mellitus (DM), smoking status, TyG-BMI, HDL, ApoA1, and eGFR. Model 3: Model 2 plus cardiometabolic medications [sodium-dependent glucose transporters 2 inhibitors (SGLT-2i), Glucagon-like peptide-1 receptor agonist (GLP-1RA), and statins]. Model 4: Model 3 plus conventional echocardiographic parameters LVEF and LVMI. Multicollinearity among independent variables was assessed using variance inflation factors (VIF). All VIF values were below 2, indicating no multicollinearity. Mediation analysis was conducted to assess the role of TyG-BMI in the relationship between EATV and LA mechanics. Given the cross-sectional design, mediation analysis was used to explore potential mechanistic pathways but does not establish causality. This was performed using Hayes’s PROCESS macro (version 4.0) in SPSS, with bias-corrected bootstrapping (5,000 resamples) to estimate 95% confidence intervals for indirect effects. All analyses were performed using SPSS (version 25.0; IBM Corp., Armonk NY). Figures were generated using Origin (2021; OriginLab Corp., Northampton, MA) and GraphPad Prism (version 9.5; GraphPad Software, San Diego, CA). A two-sided p-value <0.05 was considered statistically significant.

A total of 264 hypertensive patients with overweight or obesity were stratified into group I and Group II based on the median EATV. The baseline characteristics are detailed in Table 1. Group II had a significantly higher proportion of males, a higher BMI, and a greater prevalence of smokers compared to Group I. The duration of hypertension was also longer in Group II. Metabolic profiling revealed a higher TyG-BMI in Group II. This was accompanied by a more adverse lipid profile, characterized by lower HDL-C and reduced Apo-A1. The use of statins was significantly more frequent in Group II compared with Group I. No significant differences were observed between groups in terms of age, the prevalence of diabetes or obstructive sleep apnea-hypopnea syndrome, renal function, HbA1c, FPG, other lipid parameters, hs-CRP, BNP, or the use of antihypertensive and glucose-lowering medications.

Conventional echocardiographic parameters, including LVEF, E/A ratio, E/e’ ratio, LAVI, and LVMI, did not differ significantly between the two groups (Table 2). However, a comprehensive assessment of LA mechanics using 2D-STE revealed significant impairments in Group II. Patients in the high-EATV group demonstrated significantly reduced reservoir strain, conduit strain, and booster pump strain. Furthermore, LASI was significantly elevated in Group II compared to Group I (Figure 2).

As detailed in Table 3, EATV demonstrated significant, albeit weak-to-moderate, positive correlations with several metrics of adiposity and insulin resistance, including BMI, FPG, TyG, TyG-BMI, and TG level. Conversely, EATV was negatively correlated with HDL-C. No significant correlations were observed between EATV and other lipid parameters, renal function markers, or inflammatory biomarkers. Crucially, EATV was significantly correlated with impaired LA function. It exhibited negative correlations with all LA strain parameters, including LAs-s, LAs-e, and LAs-a. Furthermore, a positive correlation was found between EATV and LASI. No significant correlation was observed between EATV and LAVI. Similarly, TyG-BMI showed significant negative correlations with LAs-s and LAs-e, as well as a positive correlation with LASI and LAVI. However, no significant correlation was observed between TyG-BMI and LAs-a (Table 4, Figures 3, 4).

As shown in Table 5, Figure 5, univariate and multivariate linear regression analyses were performed to assess the independent associations between EATV and LA function. The univariate analysis revealed significant inverse associations of EATV with all LA strain parameters (LAs-s, LAs-e, LAs-a) and a positive association with LASI. After sequential adjustment for demographic factors, cardiometabolic risk factors, medication use, and LV structural and functional parameters (Models 1–4), the inverse associations with LAs-s and LAs-a, as well as the positive association with LASI, remained statistically significant. In contrast, the association between EATV and conduit strain (LAs-e) was attenuated and became non-significant following multivariable adjustment.

Subgroup analyses were conducted to further explore these associations. When stratified by sex (Supplementary Table S1), the inverse associations of EATV with LAs-s and LAs-a remained robust and significant in males across all adjusted models. Conversely, these associations were generally attenuated and lost statistical significance in females. The positive association between EATV and LASI was significant in both sexes but appeared more pronounced in females.

When stratified by age using the median of 53 years (Supplementary Table S2), the associations were consistently stronger in the older subgroup (≥53 years). In these older patients, EATV was independently associated with worse reservoir and booster pump strain, as well as significantly higher LASI, across all multivariate models. In the younger subgroup (<53 years), these associations were largely attenuated and non-significant.

Based on our previous results showing that both higher TyG-BMI and EATV were independently associated with impaired LA function, we conducted a mediation analysis to explore whether insulin resistance, assessed by TyG-BMI, partially accounts for the association between EATV and LA mechanics.

After adjusting for age, sex, smoking, hypertension duration, diabetes, HDL-C, Apo-A1, eGFR, SGLT-2i, GLP-1RA, statins, LVEF, and LVMI, mediation analysis showed that TyG-BMI accounted for 12.87% of the association between EATV and LAs-s (indirect effect: −0.0101, 95% CI: −0.0228 to −0.0018). Furthermore, it explained 10.00% of the association between EATV and LASI (indirect effect: 0.0002, 95% CI: 0.0001–0.0005). However, no significant mediation effect of TyG-BMI was observed in the relationships between EATV and LAs-e or LAs-a (Figure 6).

Our study has several limitations. First, the cross-sectional, single-center design precludes causal inference. Second, residual confounding from unmeasured factors, particularly abdominal visceral fat, cannot be excluded despite multivariate adjustment. Third, measuring total rather than region-specific (e.g., peri-atrial) EATV may dilute the mechanistic specificity of our findings, as peri-atrial fat may exert more direct paracrine and mechanical effects on the adjacent atrial myocardium and be more relevant to LA mechanics. Fourth, the use of the TyG-BMI, which incorporates BMI, means its mediation effect may reflect general adiposity in addition to insulin resistance. Future prospective studies combining high-resolution CT for peri-atrial EAT quantification, direct abdominal fat imaging, and direct insulin resistance assays could help disentangle local from systemic effects on atrial dysfunction.