Increased epicardial adipose tissue area and radiodensity in sarcopenic obesity Authors: Tianwen He, Ruilin Wang, Wenhua Jiang, Rongna Lian, Xiaomei Chen, Ming Yang Published: 2026 ABSTRACT BackgroundSarcopenic obesity (SO) confers substantially greater cardiometabolic risk than either sarcopenia or obesity alone, but the tissue-level structural alterations underlying this risk remain unclear. Epicardial adipose tissue (EAT) characteristics, including both area and radiodensity, may provide clinically relevant insight into the cardiometabolic burden associated with this complex muscle-fat imbalance.MethodsWe conducted a cross-sectional study of 1678 adults undergoing health examinations. Participants were categorized into normal, sarcopenia, obesity, and SO phenotypes based on the European Society for Clinical Nutrition and Metabolism and the European Association for the Study of Obesity (ESPEN/EASO) conceptual framework. Body composition and grip strength were assessed according to the Asian Working Group for Sarcopenia (AWGS) 2025 criteria. EAT area (cm2) and radiodensity (attenuation, measured in Hounsfield units [HU]) were quantified via non-contrast chest computed tomography. Associations were assessed using multivariable linear regression, interaction analysis, and sex-stratified restricted cubic splines.ResultsThe SO phenotype was identified in 45 individuals. Following multivariable adjustment, SO was independently associated with a larger EAT area (beta = 6.93 cm2; 95% CI, 5.84 to 8.03) and a higher (less negative) EAT radiodensity (beta = 9.23 HU; 95% CI, 8.03 to 10.43) compared with the normal phenotype. Effect estimates were larger for SO than for isolated obesity. An interaction between sarcopenia and obesity was observed for EAT area (interaction beta = 1.53 cm2; 95% CI, 0.05 to 3.01; P for interaction = 0.042), whereas no significant interaction was observed for EAT radiodensity. In sex-stratified analyses, grip strength showed a pronounced non-linear association with EAT radiodensity.ConclusionsIn this cross-sectional study, SO was associated with larger EAT area and higher EAT radiodensity. These findings support the presence of a distinct EAT phenotype in SO and suggest that CT-derived EAT measures may be useful imaging markers that warrant prospective validation.Graphical Abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12933-026-03237-6. What is currently known about this topic? Sarcopenic obesity (SO) significantly increases cardiometabolic risk. Epicardial adipose tissue (EAT) is a key depot linked to cardiovascular disease. The structural EAT alterations underlying the muscle-fat imbalance in SO remain unclear. What is the key research question? How does sarcopenic obesity influence the quantitative and qualitative characteristics of EAT? What is new? SO was associated with larger EAT area. SO was also associated with higher EAT radiodensity. Grip strength showed a non-linear association with EAT radiodensity. How might this study influence clinical practice? EAT measurements may help inform imaging-based risk characterization in SO, pending further validationdoes sarcopenic obesity influence the quantitative and qualitative characteristics of EAT. Background Global population aging and the obesity epidemic have converged to create a profound clinical challenge: sarcopenic obesity (SO), a complex body composition phenotype defined by the pathological coexistence of reduced skeletal muscle mass and excess adiposity [1, 2]. This “double burden” confers a substantially greater risk of cardiovascular disease, metabolic dysfunction, and premature mortality than either sarcopenia or obesity alone [1–4]. Despite the growing recognition of SO as a high-risk archetype, the tissue-level structural and functional alterations that mediate its amplified cardiometabolic threat remain incompletely understood. Traditional anthropometric indices, such as body mass index (BMI), inherently fail to capture the nuanced risks associated with this muscle-fat imbalance, necessitating deeper investigation into specific ectopic fat depots. Epicardial adipose tissue (EAT) has emerged as a uniquely informative and metabolically active ectopic fat depot due to its direct anatomic proximity and shared microcirculation with the myocardium and coronary vasculature [5, 6]. While extensive volumetric imaging has established that a greater EAT burden directly correlates with coronary artery disease and adverse cardiovascular events [5, 6], computed tomography (CT)–derived EAT attenuation (radiodensity) has recently been recognized as a critical biomarker for the qualitative aspects of adipose tissue. Lower radiodensity reflects lipid hypertrophy, whereas higher (less negative) radiodensity may indicate reduced lipid content, increased fibrosis, or localized inflammation [7]. Variations in EAT radiodensity independently predict coronary plaque vulnerability and cardiovascular events irrespective of total EAT volume [7–9], underscoring the clinical imperative of assessing both the quantity and quality of epicardial fat. Although SO and EAT have independent prognostic value for cardiometabolic risk, their interrelationship remains a critical knowledge gap [10]. Prior EAT investigations have predominantly focused on generalized obesity or cohorts with established cardiovascular disease, often neglecting the modifying impact of skeletal muscle status. Conversely, studies characterizing SO have largely prioritized systemic metabolic outcomes, lacking precise integration with imaging-derived ectopic fat phenotypes [11]. Consequently, it remains unknown whether the profound muscle-fat imbalance in SO distinctively exacerbates epicardial fat remodeling beyond the effects of isolated adiposity, and whether functional measures of muscle health independently associate with EAT quality [12, 13]. To address these gaps, the present study investigated the quantitative and qualitative characteristics of EAT across diverse body composition phenotypes in a large, health-examination cohort. By integrating CT-derived measures of EAT area and radiodensity with comprehensive assessments of muscle mass, muscle strength, and systemic metabolic profiles, we aimed to determine whether SO is associated with distinct adverse EAT patterns, and whether these relationships remain robust across demographic domains. Methods Study population and design This cross-sectional study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines [14]. Participants were drawn from adults undergoing routine health examinations at a single center, including standardized clinical evaluation and non-contrast computed tomography (CT) imaging for quantitative assessment of EAT. During the study period, consecutive adults aged ≥ 18 years who underwent routine health examinations were screened for eligibility. Participants were included if they had complete data on body composition assessment, muscle function evaluation, and CT-based EAT measurements. Participants were excluded if they met any of the following criteria: (1) missing or technically inadequate CT images for reliable EAT quantification; (2) incomplete key variables required for body composition phenotyping, including muscle mass, adiposity, or muscle strength measures; (3) history of major cardiovascular events, active malignancy, or systemic inflammatory or endocrine disorders that could substantially affect body composition or adipose tissue distribution; or (4) conditions precluding accurate assessment of body composition or EAT parameters. Based on predefined criteria for muscle mass, adiposity, and muscle function (detailed below), participants were categorized into four mutually exclusive body composition phenotypes: normal (N), sarcopenia (S), obesity (O), and SO. The detailed participant selection process, including inclusion and exclusion steps, is illustrated in Supplementary Fig. S1. This cross-sectional study was conducted in accordance with the principles of the Declaration of Helsinki. The study protocol was approved by the Biomedical Ethics Committee of West China Hospital, Sichuan University. As this study involved the retrospective analysis of anonymized data routinely collected during health examinations, the requirement for written informed consent was waived by the ethics committee. All data were anonymized prior to analysis to protect participant confidentiality. Assessment of body composition and muscle function Body composition was assessed using bioelectrical impedance analysis (BIA) with an InBody 570 device (InBody Co., Ltd., Seoul, Korea) following standardized procedures. Appendicular skeletal muscle mass and total body fat mass were obtained from the BIA measurements. Appendicular skeletal muscle mass index (ASMI) was calculated as appendicular skeletal muscle mass divided by height squared (kg/m2). Body fat percentage was derived from total body fat mass measured by BIA. Muscle strength was evaluated by handgrip strength using a digital handheld dynamometer (EH101; Xiangshan Inc., Guangdong, China; 0.1 kg accuracy) according to a standardized protocol during the health examination. Grip strength measurements were performed by trained personnel, and the highest value obtained from repeated measurements was used for analysis. Based on predefined operational criteria integrating muscle mass, adiposity, and muscle strength, participants were categorized into four mutually exclusive body composition phenotypes. Sarcopenia was defined according to the Asian Working Group for Sarcopenia (AWGS) 2025 Consensus Update [15] as low muscle mass combined with low muscle strength, employing age- and sex-specific cut-off values. For individuals aged < 65 years, low muscle mass by BIA was defined as ASMI < 7.6 kg/m2 in men and < 5.7 kg/m2 in women, and low muscle strength was defined as handgrip strength < 34.0 kg in men and < 20.0 kg in women. For individuals aged ≥ 65 years, low muscle mass was defined as ASMI < 7.0 kg/m2 in men and < 5.7 kg/m2 in women, and low muscle strength was < 28.0 kg in men and < 18.0 kg in women. Obesity was defined based on BIA-derived body fat percentage using sex-specific thresholds (> 20.21% in men and > 31.7% in women) [16]. According to the conceptual framework of the European Society for Clinical Nutrition and Metabolism and the European Association for the Study of Obesity (ESPEN/EASO) consensus, SO was defined as the concurrent presence of both sarcopenia and obesity. Participants who did not meet criteria for either sarcopenia or obesity were classified as having a normal body composition phenotype. All body composition phenotypes were defined a priori and applied consistently across the study population. EAT measurement EAT measurement was carried out using CT at West China Hospital, Sichuan University, employing a 16-slice spiral CT scanner (Brilliance; Philips Healthcare, Ohio, USA) with a slice thickness of 5 mm. EAT was precisely defined as the adipose tissue situated between the visceral and parietal pericardial layers, manifesting as low-density regions on CT images with attenuation values ranging from − 190 to − 30 Hounsfield units (HU). The measurements were conducted utilizing 3D Slicer software (version 5.8.1, https://www.slicer.org/) at the axial level of the left main coronary artery ostium. The procedure entailed: (1) manual delineation of the pericardial boundary; (2) threshold-based segmentation (HU range: − 190 to − 30) for fat identification; (3) automatic calculation of the EAT area (cm2) and mean radiodensity (HU). All measurements were executed by a researcher who was blinded to the clinical data.Fig. 1Phenotype distribution and EAT parameters. A Distribution of study participants across body composition phenotypes stratified by age group. B Standardized comparison of EAT area (cm2) across the four body composition phenotypes. C Standardized comparison of EAT radiodensity (attenuation, Hounsfield units [HU]) across the four body composition phenotypes. D Correlation between EAT area and radiodensity among the total study sample and stratified by sex Phenotype distribution and EAT parameters. A Distribution of study participants across body composition phenotypes stratified by age group. B Standardized comparison of EAT area (cm2) across the four body composition phenotypes. C Standardized comparison of EAT radiodensity (attenuation, Hounsfield units [HU]) across the four body composition phenotypes. D Correlation between EAT area and radiodensity among the total study sample and stratified by sex Assessment of covariates and metabolic indices Comprehensive demographic, clinical, anthropometric, and laboratory data were obtained from routine health examinations. Clinical assessments and imaging examinations were performed within a defined period as part of standard health screening. Blood samples were collected after an overnight fast of at least 8 h and analyzed using standardized laboratory methods. Baseline demographic and lifestyle covariates included age, sex, smoking status, and alcohol consumption. Age was treated as a continuous variable in primary models and categorized into prespecified groups (young [< 40 years], middle-aged [40–59 years], and older [≥ 60 years]) for stratified analyses. Smoking status and alcohol consumption were recorded as binary self-reported variables (current use vs. non-use). Anthropometric measurements, including height, weight, waist circumference, and hip circumference, were obtained using standardized protocols. Body mass index (BMI) and waist-to-hip ratio (WHR) were calculated accordingly. Seated blood pressure was measured by trained personnel after a resting period. Cardiometabolic comorbidities, which served as critical covariates in adjusted models, were rigorously defined. Hypertension was identified by a self-reported history, current use of antihypertensive medication, or elevated baseline blood pressure. Diabetes mellitus was defined by a self-reported history, use of glucose-lowering medication, or elevated fasting plasma glucose (> = 7.0 mmol/L) or glycated hemoglobin (HbA1c > = 6.5%). Dyslipidemia was determined based on self-reported diagnosis, lipid-lowering therapy, or abnormal lipid panels (e.g., hypertriglyceridemia or low high-density lipoprotein cholesterol [HDL-C]). Fatty liver disease was diagnosed based on abdominal ultrasound imaging findings evaluated by experienced radiologists, rather than by broader contemporary steatotic liver disease classifications. Laboratory parameters included fasting blood glucose (FBG), HbA1c, fasting insulin, triglycerides (TG), total cholesterol, HDL-C, low-density lipoprotein cholesterol, creatinine, uric acid, aspartate aminotransferase, and alanine aminotransferase. To further characterize systemic metabolic dysfunction, several composite indices were calculated using established formulas. The triglyceride-glucose (TyG) index was calculated as ln[TG (mg/dL) × FBG (mg/dL)/2]. The atherogenic index of plasma (AIP) was calculated as log10[TG (mmol/L)/HDL-C (mmol/L)]. The estimated glucose disposal rate (eGDR, mg kg−1 min−1) was calculated as 21.158 − 0.09 × waist circumference (cm) − 3.407 × hypertension (1 = yes, 0 = no) − 0.551 × HbA1c (%). The weight-adjusted waist index (WWI) was calculated as waist circumference (cm) / √weight (kg). Statistical analysis Continuous variables are presented as mean ± standard deviation (SD) or median (interquartile range, IQR), as appropriate, while categorical variables are expressed as counts and percentages. Baseline characteristics across body composition phenotypes were compared using one-way analysis of variance (ANOVA) or the Kruskal–Wallis test for continuous data, and the chi-square test for categorical data. Covariates for the primary multivariable models were selected a priori based on clinical relevance and their plausibility as broad demographic, lifestyle, and cardiometabolic confounders of the association between body composition phenotype and EAT characteristics. Multivariable linear regression models were constructed to evaluate the independent associations between body composition phenotypes (modeled as a categorical exposure, with the normal phenotype serving as the reference) and EAT characteristics (area and radiodensity, modeled as continuous outcomes). To assess the robustness of these associations, covariates were entered into sequentially adjusted models: Model 1 was unadjusted; Model 2 adjusted for fundamental demographic factors (age and sex); and Model 3 further adjusted for lifestyle factors and key cardiometabolic comorbidities that could influence ectopic fat distribution (smoking status, alcohol consumption, diabetes, hypertension, and fatty liver disease). Interaction analyses between sarcopenia and obesity on EAT parameters were evaluated using formal interaction testing within the regression framework, and sex-stratified interaction models were additionally fitted in men and women separately. Furthermore, sex-stratified restricted cubic splines with three knots were utilized to explore potential non-linear, dose–response relationships between continuous body composition metrics (ASMI, body fat percentage, and grip strength) and EAT outcomes. A post hoc sensitivity power analysis confirmed that the sample size achieved 80% power to detect a minimum effect size (Cohen’s d) of 0.43 for the primary normal vs SO comparison. Sensitivity and diagnostic checks (including residual analysis, variance inflation factors [VIF] to rule out multicollinearity, and Cook’s distance for influential observations) were conducted to ensure model validity. Two-sided P values < 0.05 were considered statistically significant. All statistical analyses and graphical visualizations were performed using R (version 4.2) and Python (version 3.9). Results Baseline characteristics of the study population The participant selection process, including detailed inclusion and exclusion steps, is illustrated in Supplementary Fig. S1. Table 1 summarizes the baseline clinical and demographic characteristics across the four body composition phenotypes (normal, sarcopenia, obesity, and SO). Participants with SO were older and more frequently male than participants in the normal group, while lifestyle factors, including smoking and alcohol consumption, differed across phenotypes and were more prevalent in the obesity-related groups.Table 1Baseline characteristics of the four study groupsVariableOverall (n = 1678)N (Normal) (n = 786)S (Sarcopenia) (n = 50)O (Obesity) (n = 797)SO (Sarcopenic Obesity) (n = 45)Overall PAge (years)44.0 (34.0, 53.0)40.0 (33.0, 50.0)49.5 (32.0, 59.0)47.0 (36.0, 55.0)59.0 (52.0, 69.0) < 0.001Sex < 0.001Male944 (56.3%)340 (43.3%)24 (48.0%)550 (69.0%)30 (66.7%)Female734 (43.7%)446 (56.7%)26 (52.0%)247 (31.0%)15 (33.3%)Smoking434 (26.1%)166 (21.2%)8 (16.0%)247 (31.4%)13 (28.9%) < 0.001Alcohol consumption731 (43.6%)277 (35.2%)14 (28.0%)419 (52.6%)21 (46.7%) < 0.001BMI (kg/m2)23.5 (21.2, 25.8)21.7 (20.1, 23.2)18.4 (17.4, 19.7)25.7 (24.2, 27.5)22.1 (20.4, 23.4) < 0.001FG (mmol/L)5.1 (4.6, 5.8)5.2 (4.6, 5.8)4.9 (4.1, 5.7)5.2 (4.5, 5.9)5.1 (4.2, 5.5)0.204TG (mmol/L)1.5 (1.0, 2.0)1.4 (0.9, 1.9)1.4 (0.8, 1.9)1.6 (1.1, 2.2)1.4 (0.8, 1.8) < 0.001HDL-C (mmol/L)1.3 ± 0.31.4 ± 0.31.4 ± 0.31.3 ± 0.31.3 ± 0.30.208LDL-C (mmol/L)2.5 (2.0, 2.9)2.4 (2.0, 2.9)2.5 (2.0, 2.7)2.6 (2.2, 3.0)2.2 (1.9, 2.7) < 0.001SBP (mmHg)119.0 (112.0, 132.0)117.0 (109.0, 125.0)118.0 (108.0, 131.0)124.0 (116.0, 136.0)125.0 (116.0, 142.0) < 0.001DBP (mmHg)72.0 (65.0, 80.0)69.0 (63.0, 76.0)69.0 (61.0, 74.0)77.0 (68.0, 84.0)74.0 (68.0, 81.0) < 0.001Diabetes295 (17.6%)119 (15.1%)8 (16.0%)159 (19.9%)9 (20.0%)0.086Hypertension319 (19.0%)77 (9.8%)8 (16.0%)217 (27.2%)17 (37.8%) < 0.001Fatty liver32 (1.9%)9 (1.2%)0 (0.0%)23 (2.9%)0 (0.0%)0.038Hyperlipidemia559 (33.3%)258 (32.8%)17 (34.7%)272 (34.1%)12 (26.7%)0.739MetS269 (16.0%)29 (3.7%)0 (0.0%)234 (29.4%)6 (13.3%) < 0.001ASMI (kg/m2), Male7.92 (7.33, 8.51)7.86 (7.42, 8.36)5.50 (5.50, 5.85)8.06 (7.49, 8.71)5.50 (5.50, 6.41) < 0.001ASMI (kg/m2), Female6.29 (5.84, 6.71)6.30 (5.89, 6.67)4.00 (4.00, 4.20)6.45 (5.97, 6.94)4.00 (4.00, 5.43) < 0.001Grip strength (kg), Male37.8 (32.3, 42.9)39.1 (34.1, 44.1)26.1 (25.6, 26.8)38.0 (33.0, 42.8)26.4 (26.3, 26.9) < 0.001Grip strength (kg), Female23.2 (19.9, 26.2)23.9 (20.8, 26.6)16.2 (15.5, 16.9)23.0 (20.0, 25.8)16.8 (16.3, 16.9) < 0.001WC (cm), Male86.0 (80.0, 91.0)81.0 (76.0, 85.0)72.5 (66.5, 75.5)89.0 (85.0, 94.0)82.0 (79.0, 86.0) < 0.001WC (cm), Female73.0 (68.0, 78.0)70.0 (67.0, 74.0)65.5 (63.0, 70.0)80.0 (74.0, 85.0)76.0 (72.0, 80.0) < 0.001BFP (%), Male25.7 (21.9, 29.1)20.8 (18.4, 22.7)16.2 (12.6, 21.0)28.2 (26.3, 31.0)27.1 (25.0, 30.8) < 0.001BFP (%), Female30.4 (26.3, 34.9)27.4 (24.2, 30.0)28.1 (25.6, 30.5)36.0 (34.5, 38.6)35.9 (34.7, 38.5) < 0.001EAT area (cm2)10.5 ± 4.08.6 ± 3.310.0 ± 3.212.1 ± 3.815.0 ± 3.7 < 0.001EAT radiodensity (HU)− 66.2 ± 4.4− 68.2 ± 4.0− 62.7 ± 4.5− 65.0 ± 3.8− 59.1 ± 4.1 < 0.0011Abbreviations: ASMI, Appendicular Skeletal Muscle Mass Index; BFP, Body Fat Percentage; BMI, Body Mass Index; BP, Blood Pressure; CHD, Coronary Heart Disease; CKD, Chronic Kidney Disease; CVD, Cerebrovascular Disease; DBP, Diastolic Blood Pressure; FG, Fasting Glucose; HDL-C, High-Density Lipoprotein Cholesterol; LDL-C, Low-Density Lipoprotein Cholesterol; MetS, Metabolic Syndrome; SBP, Systolic Blood Pressure; TG, Triglycerides; WC, Waist Circumference. Continuous variables are presented as mean ± SD for normally distributed variables (HDL-C) or median (IQR) for non-normally distributed variables. Categorical variables are presented as n (%). P values are overall tests, not post hoc pairwise comparisons Table 1Baseline characteristics of the four study groupsVariableOverall (n = 1678)N (Normal) (n = 786)S (Sarcopenia) (n = 50)O (Obesity) (n = 797)SO (Sarcopenic Obesity) (n = 45)Overall PAge (years)44.0 (34.0, 53.0)40.0 (33.0, 50.0)49.5 (32.0, 59.0)47.0 (36.0, 55.0)59.0 (52.0, 69.0) < 0.001Sex < 0.001Male944 (56.3%)340 (43.3%)24 (48.0%)550 (69.0%)30 (66.7%)Female734 (43.7%)446 (56.7%)26 (52.0%)247 (31.0%)15 (33.3%)Smoking434 (26.1%)166 (21.2%)8 (16.0%)247 (31.4%)13 (28.9%) < 0.001Alcohol consumption731 (43.6%)277 (35.2%)14 (28.0%)419 (52.6%)21 (46.7%) < 0.001BMI (kg/m2)23.5 (21.2, 25.8)21.7 (20.1, 23.2)18.4 (17.4, 19.7)25.7 (24.2, 27.5)22.1 (20.4, 23.4) < 0.001FG (mmol/L)5.1 (4.6, 5.8)5.2 (4.6, 5.8)4.9 (4.1, 5.7)5.2 (4.5, 5.9)5.1 (4.2, 5.5)0.204TG (mmol/L)1.5 (1.0, 2.0)1.4 (0.9, 1.9)1.4 (0.8, 1.9)1.6 (1.1, 2.2)1.4 (0.8, 1.8) < 0.001HDL-C (mmol/L)1.3 ± 0.31.4 ± 0.31.4 ± 0.31.3 ± 0.31.3 ± 0.30.208LDL-C (mmol/L)2.5 (2.0, 2.9)2.4 (2.0, 2.9)2.5 (2.0, 2.7)2.6 (2.2, 3.0)2.2 (1.9, 2.7) < 0.001SBP (mmHg)119.0 (112.0, 132.0)117.0 (109.0, 125.0)118.0 (108.0, 131.0)124.0 (116.0, 136.0)125.0 (116.0, 142.0) < 0.001DBP (mmHg)72.0 (65.0, 80.0)69.0 (63.0, 76.0)69.0 (61.0, 74.0)77.0 (68.0, 84.0)74.0 (68.0, 81.0) < 0.001Diabetes295 (17.6%)119 (15.1%)8 (16.0%)159 (19.9%)9 (20.0%)0.086Hypertension319 (19.0%)77 (9.8%)8 (16.0%)217 (27.2%)17 (37.8%) < 0.001Fatty liver32 (1.9%)9 (1.2%)0 (0.0%)23 (2.9%)0 (0.0%)0.038Hyperlipidemia559 (33.3%)258 (32.8%)17 (34.7%)272 (34.1%)12 (26.7%)0.739MetS269 (16.0%)29 (3.7%)0 (0.0%)234 (29.4%)6 (13.3%) < 0.001ASMI (kg/m2), Male7.92 (7.33, 8.51)7.86 (7.42, 8.36)5.50 (5.50, 5.85)8.06 (7.49, 8.71)5.50 (5.50, 6.41) < 0.001ASMI (kg/m2), Female6.29 (5.84, 6.71)6.30 (5.89, 6.67)4.00 (4.00, 4.20)6.45 (5.97, 6.94)4.00 (4.00, 5.43) < 0.001Grip strength (kg), Male37.8 (32.3, 42.9)39.1 (34.1, 44.1)26.1 (25.6, 26.8)38.0 (33.0, 42.8)26.4 (26.3, 26.9) < 0.001Grip strength (kg), Female23.2 (19.9, 26.2)23.9 (20.8, 26.6)16.2 (15.5, 16.9)23.0 (20.0, 25.8)16.8 (16.3, 16.9) < 0.001WC (cm), Male86.0 (80.0, 91.0)81.0 (76.0, 85.0)72.5 (66.5, 75.5)89.0 (85.0, 94.0)82.0 (79.0, 86.0) < 0.001WC (cm), Female73.0 (68.0, 78.0)70.0 (67.0, 74.0)65.5 (63.0, 70.0)80.0 (74.0, 85.0)76.0 (72.0, 80.0) < 0.001BFP (%), Male25.7 (21.9, 29.1)20.8 (18.4, 22.7)16.2 (12.6, 21.0)28.2 (26.3, 31.0)27.1 (25.0, 30.8) < 0.001BFP (%), Female30.4 (26.3, 34.9)27.4 (24.2, 30.0)28.1 (25.6, 30.5)36.0 (34.5, 38.6)35.9 (34.7, 38.5) < 0.001EAT area (cm2)10.5 ± 4.08.6 ± 3.310.0 ± 3.212.1 ± 3.815.0 ± 3.7 < 0.001EAT radiodensity (HU)− 66.2 ± 4.4− 68.2 ± 4.0− 62.7 ± 4.5− 65.0 ± 3.8− 59.1 ± 4.1 < 0.0011Abbreviations: ASMI, Appendicular Skeletal Muscle Mass Index; BFP, Body Fat Percentage; BMI, Body Mass Index; BP, Blood Pressure; CHD, Coronary Heart Disease; CKD, Chronic Kidney Disease; CVD, Cerebrovascular Disease; DBP, Diastolic Blood Pressure; FG, Fasting Glucose; HDL-C, High-Density Lipoprotein Cholesterol; LDL-C, Low-Density Lipoprotein Cholesterol; MetS, Metabolic Syndrome; SBP, Systolic Blood Pressure; TG, Triglycerides; WC, Waist Circumference. Continuous variables are presented as mean ± SD for normally distributed variables (HDL-C) or median (IQR) for non-normally distributed variables. Categorical variables are presented as n (%). P values are overall tests, not post hoc pairwise comparisons Baseline characteristics of the four study groups 1Abbreviations: ASMI, Appendicular Skeletal Muscle Mass Index; BFP, Body Fat Percentage; BMI, Body Mass Index; BP, Blood Pressure; CHD, Coronary Heart Disease; CKD, Chronic Kidney Disease; CVD, Cerebrovascular Disease; DBP, Diastolic Blood Pressure; FG, Fasting Glucose; HDL-C, High-Density Lipoprotein Cholesterol; LDL-C, Low-Density Lipoprotein Cholesterol; MetS, Metabolic Syndrome; SBP, Systolic Blood Pressure; TG, Triglycerides; WC, Waist Circumference. Continuous variables are presented as mean ± SD for normally distributed variables (HDL-C) or median (IQR) for non-normally distributed variables. Categorical variables are presented as n (%). P values are overall tests, not post hoc pairwise comparisons Regarding cardiometabolic features, the SO group had the highest prevalence of hypertension, whereas several other metabolic abnormalities were comparable to or more pronounced in the obesity group. Supplementary Fig. S2 provides an expanded metabolic overview and highlights a clear phenotype separation for composite metabolic indices: WWI increased in a stepwise manner (normal approximately sarcopenia < obesity < SO), eGDR was generally lower in obesity-related phenotypes, and the TyG index and AIP were predominantly elevated in the obesity group, with SO typically showing intermediate-to-high levels. Distribution and unadjusted patterns of EAT measures Figure 1 summarizes the distribution of body composition phenotypes by sex and provides a visual representation of the unadjusted EAT area and radiodensity values. Supplementary Fig. S3 presents representative CT images illustrating greater EAT area and higher radiodensity in an SO participant than in a normal control with a similar BMI. Figure 2 presents effect size-based comparisons of EAT parameters, along with interaction and age-stratified analyses. Relative to the normal group, standardized differences in EAT area increased progressively across sarcopenia and obesity, with the largest effect size observed in SO (Cohen’s d for SO vs normal = 1.95, 95% CI 1.63 to 2.26; Fig. 2A). Similar patterns emerged for EAT radiodensity, where SO showed the greatest between-group separation, reflecting higher (less negative) attenuation values (SO vs normal: d = 2.28, 95% CI 1.96 to 2.60; Fig. 2B). SO also differed from isolated sarcopenia and isolated obesity for both EAT area and radiodensity.Fig. 2Effect sizes and interactions of sarcopenia and obesity on EAT. A Standardized effect sizes (Cohen’s d with 95% CI) for the differences in EAT area relative to the normal phenotype. B Standardized effect sizes for the differences in EAT radiodensity relative to the normal phenotype. C Interaction plot illustrating the interaction between sarcopenia and obesity on EAT area. D Interaction plot for EAT radiodensity. E Age-stratified subgroup analyses presenting standardized differences for EAT area comparing SO vs normal phenotype. F Age-stratified subgroup analyses presenting standardized differences for EAT radiodensity comparing SO vs normal phenotype Effect sizes and interactions of sarcopenia and obesity on EAT. A Standardized effect sizes (Cohen’s d with 95% CI) for the differences in EAT area relative to the normal phenotype. B Standardized effect sizes for the differences in EAT radiodensity relative to the normal phenotype. C Interaction plot illustrating the interaction between sarcopenia and obesity on EAT area. D Interaction plot for EAT radiodensity. E Age-stratified subgroup analyses presenting standardized differences for EAT area comparing SO vs normal phenotype. F Age-stratified subgroup analyses presenting standardized differences for EAT radiodensity comparing SO vs normal phenotype Interaction analyses revealed a significant interaction between sarcopenia and obesity for EAT area (interaction beta = 1.53 cm2, 95% CI 0.05 to 3.01, P = 0.042; Fig. 2C). In contrast, no significant interaction was observed for EAT radiodensity (interaction beta = 0.46 HU, 95% CI − 1.16 to 2.08, P = 0.581; Fig. 2D). Complete overall interaction-model coefficients are presented in Supplementary Tables S1 and S2. In age-stratified analyses, the SO-normal contrast for both EAT area and radiodensity remained evident across young, middle-aged, and older subgroups, with no statistically significant effect modification by age (Fig. 2E-F). In sex-stratified interaction models, the interaction term for EAT area remained directionally positive in both men and women but was not statistically significant in either sex; no significant sex-specific interaction was observed for EAT radiodensity (Supplementary Tables S3, S4). Sex-stratified analyses further confirmed the phenotype-related differences in both men and women, with SO consistently showing the highest EAT area and the least negative EAT attenuation (Supplementary Fig. S4). Independent associations between body composition phenotypes and EAT parameters Table 2 presents the independent associations between body composition phenotypes and EAT parameters across sequentially adjusted regression models. Compared with the normal group, all three non-normal phenotypes were independently associated with a higher EAT area in the fully adjusted model (Model 3), demonstrating a graded pattern with the largest difference in SO: sarcopenia (β = 1.59 cm2, 95% CI 0.57–2.61), obesity (β = 3.81 cm2, 95% CI 3.44–4.19), and SO (β = 6.93 cm2, 95% CI 5.84–8.03).Table 2Association between study groups and EAT parametersComparisonModel 1Model 2Model 3EAT Area (cm2)S vs N1.49 (0.47, 2.51)**1.63 (0.61, 2.64)**1.59 (0.57, 2.61)**O vs N3.58 (3.23, 3.93)***3.73 (3.36, 4.10)***3.81 (3.44, 4.19)***SO vs N6.45 (5.38, 7.52)***6.90 (5.80, 7.99)***6.93 (5.84, 8.03)***EAT Density (HU)S vs N5.51 (4.40, 6.62)***5.57 (4.45, 6.69)***5.54 (4.42, 6.66)***O vs N3.22 (2.84, 3.60)***3.25 (2.85, 3.65)***3.23 (2.82, 3.64)***SO vs N9.07 (7.90, 10.24)***9.25 (8.05, 10.45)***9.23 (8.03, 10.43)***Values are β coefficient (95% confidence interval). Model 1: Crude model (unadjusted). Model 2: Adjusted for age and sex. Model 3: Adjusted for age, sex, smoking, alcohol consumption, diabetes, hypertension, and fatty liver. Reference group: N (Normal). P-value symbols: *P < 0.05, **P < 0.01, ***P < 0.001. Abbreviations: EAT, Epicardial Adipose Tissue; CI, Confidence Interval; N, Normal; S, Sarcopenia; O, Obesity; SO, Sarcopenic Obesity Table 2Association between study groups and EAT parametersComparisonModel 1Model 2Model 3EAT Area (cm2)S vs N1.49 (0.47, 2.51)**1.63 (0.61, 2.64)**1.59 (0.57, 2.61)**O vs N3.58 (3.23, 3.93)***3.73 (3.36, 4.10)***3.81 (3.44, 4.19)***SO vs N6.45 (5.38, 7.52)***6.90 (5.80, 7.99)***6.93 (5.84, 8.03)***EAT Density (HU)S vs N5.51 (4.40, 6.62)***5.57 (4.45, 6.69)***5.54 (4.42, 6.66)***O vs N3.22 (2.84, 3.60)***3.25 (2.85, 3.65)***3.23 (2.82, 3.64)***SO vs N9.07 (7.90, 10.24)***9.25 (8.05, 10.45)***9.23 (8.03, 10.43)***Values are β coefficient (95% confidence interval). Model 1: Crude model (unadjusted). Model 2: Adjusted for age and sex. Model 3: Adjusted for age, sex, smoking, alcohol consumption, diabetes, hypertension, and fatty liver. Reference group: N (Normal). P-value symbols: *P < 0.05, **P < 0.01, ***P < 0.001. Abbreviations: EAT, Epicardial Adipose Tissue; CI, Confidence Interval; N, Normal; S, Sarcopenia; O, Obesity; SO, Sarcopenic Obesity Association between study groups and EAT parameters Values are β coefficient (95% confidence interval). Model 1: Crude model (unadjusted). Model 2: Adjusted for age and sex. Model 3: Adjusted for age, sex, smoking, alcohol consumption, diabetes, hypertension, and fatty liver. Reference group: N (Normal). P-value symbols: *P < 0.05, **P < 0.01, ***P < 0.001. Abbreviations: EAT, Epicardial Adipose Tissue; CI, Confidence Interval; N, Normal; S, Sarcopenia; O, Obesity; SO, Sarcopenic Obesity A highly consistent pattern was observed for EAT radiodensity, where significantly higher values (i.e., less negative attenuation, indicating qualitative deterioration) were seen in sarcopenia (beta = 5.54 HU, 95% CI 4.42–6.66), obesity (beta = 3.23HU, 95% CI 2.82–3.64), and most profoundly in SO (beta = 9.23 HU, 95% CI 8.03–10.43) relative to the normal group. Effect estimates remained directionally consistent and statistically significant after stepwise adjustment for demographics, lifestyle, and cardiometabolic comorbidities. Supplementary Fig. S5 further shows that residual and influence diagnostics (e.g., Cook’s distances and VIFs) were within acceptable thresholds. Dose–response relationships between continuous body composition indices and EAT Figure 3 evaluates the sex-stratified dose–response relationships between continuous body composition indices and EAT measures using restricted cubic splines. ASMI showed no evidence of a non-linear association with EAT area in either men or women (Fig. 3C), while a non-linear association with EAT radiodensity was evident in men but not in women (Fig. 3D). By contrast, body fat percentage demonstrated clear, strong non-linear dose–response relationships with both EAT area and EAT radiodensity in both sexes (all non-linear P < 0.001) (Fig. 3E–F).Fig. 3Dose–response and interaction relationships between body composition indices and EAT measurements. A and B Forest plots showing the interaction effects between continuous body composition metrics and EAT parameters. C and D Restricted cubic spline analyses visualizing the dose–response relationship between Appendicular Skeletal Muscle Mass Index (ASMI) and EAT area/radiodensity in men and women. E and F Restricted cubic spline analyses visualizing the dose–response relationship between body fat percentage and EAT area/radiodensity in men and women. G and H Restricted cubic spline analyses visualizing the dose–response relationship between grip strength and EAT area/radiodensity in men and women Dose–response and interaction relationships between body composition indices and EAT measurements. A and B Forest plots showing the interaction effects between continuous body composition metrics and EAT parameters. C and D Restricted cubic spline analyses visualizing the dose–response relationship between Appendicular Skeletal Muscle Mass Index (ASMI) and EAT area/radiodensity in men and women. E and F Restricted cubic spline analyses visualizing the dose–response relationship between body fat percentage and EAT area/radiodensity in men and women. G and H Restricted cubic spline analyses visualizing the dose–response relationship between grip strength and EAT area/radiodensity in men and women Notably, grip strength (a functional marker of muscle quality) was not non-linearly associated with EAT area in either sex (Fig. 3G), but it showed highly significant non-linear associations with EAT radiodensity in both men and women (both non-linear P < 0.001). As grip strength declined, EAT density values became progressively less negative across the observed range in both sexes (Fig. 3H). Discussion In this large, cross-sectional analysis of a health-examination cohort, SO was associated with larger EAT area and higher (less negative) EAT radiodensity than the normal phenotype, isolated sarcopenia, or isolated obesity. An interaction between sarcopenia and obesity was observed for EAT area, although the corresponding 95% CI was close to the null value and this estimate should therefore be interpreted cautiously, whereas sex-stratified interaction terms were not statistically significant in either men or women. Grip strength also showed a pronounced non-linear association with EAT radiodensity in sex-stratified analyses. Collectively, these findings suggest that SO is associated with a distinct EAT phenotype. Our findings extend the current understanding of SO, a uniquely high-risk phenotype increasingly recognized for its disproportionate cardiometabolic burden [17, 18]. While previous literature has largely emphasized the systemic consequences of SO—such as insulin resistance, systemic inflammation, and impaired glucose whole-body disposal—studies integrating precise, imaging-derived ectopic fat phenotypes have been scarce [19–22]. Epicardial fat, given its shared microcirculation with the myocardium, acts as a highly sensitive transducer of systemic metabolic stress [23]. CT-derived EAT radiodensity has recently emerged as a critical qualitative biomarker that may provide prognostic information beyond simple volumetric fat measurements [24, 25]. Lower (more negative) attenuation typically reflects healthy, lipid-rich adipocytes, whereas higher (less negative) attenuation may indicate pathological tissue remodeling, including impaired lipid-storage capacity, adipocyte dedifferentiation, fibrosis, and localized inflammation [7, 26]. In the present study, SO was associated not only with greater EAT burden but also with higher EAT radiodensity. Although the cross-sectional design precludes causal inference, this pattern is consistent with prior literature suggesting that adverse ectopic fat remodeling may accompany muscle-fat imbalance. From a mechanistic perspective, reduced skeletal muscle mass may limit peripheral glucose disposal and lipid buffering capacity, potentially favoring lipid deposition in ectopic depots such as EAT [27, 28]. These mechanisms remain hypothetical in the current study and require prospective and translational confirmation. Future studies incorporating volumetric and radiomic analyses may further characterize EAT heterogeneity. A notable finding of our study is the non-linear association between handgrip strength and EAT radiodensity, which was observed in both sexes [29]. This pattern suggests that lower muscle function may accompany a less favorable EAT phenotype. Grip strength appeared to provide information not fully captured by muscle mass (ASMI) alone [30]. However, the clinical thresholds suggested by these curves should be interpreted cautiously and require prospective confirmation [31]. From a clinical perspective, these findings suggest that anthropometric measures alone may not fully capture body-composition-related variation in epicardial fat phenotype [32]. Opportunistic assessment of EAT radiodensity on chest CT may help identify higher-risk imaging patterns in individuals with SO, although its clinical utility and thresholds require external and prospective validation. SO management remains centered on lifestyle and exercise-based strategies, and prior work suggests that exercise may favorably influence EAT characteristics [33, 34]. The age-stratified analyses further suggest that these phenotype-related differences are not restricted to older adults [35–37]. Several strengths of this study merit consideration. First, the analysis utilized a large cohort with comprehensive phenotyping, enabling the simultaneous evaluation of EAT characteristics, body composition, and muscle function within a unified framework. Second, we employed sequential statistical modeling, including interaction testing and non-linear dose–response modeling via restricted cubic splines, which allowed examination of these relationships across clinically relevant age and sex subgroups. Finally, the use of routinely acquired, non-contrast chest CT for EAT assessment enhances the translational relevance of our findings, as these imaging features can be opportunistically derived without additional radiation exposure or specialized protocols. Several limitations should be acknowledged. First, the cross-sectional design precludes causal inference, and the observed associations may reflect bidirectional or confounded relationships. Second, the SO subgroup was relatively small, which may limit precision and the stability of subgroup and interaction estimates. This is particularly relevant for the statistically significant EAT-area interaction term, whose 95% CI ranged from 0.05 to 3.01, with a lower bound close to 0. Third, although we adjusted for several clinically relevant covariates, residual confounding from unmeasured lifestyle, physical-activity, dietary, behavioral, inflammatory, and medication-related factors cannot be excluded. In addition, fatty liver status was based on recorded abdominal ultrasound findings rather than broader contemporary steatotic liver disease classifications, which may have led to conservative ascertainment. Fourth, body composition was assessed using BIA rather than DXA or MRI; although practical for large-scale screening, BIA is less precise and may be influenced by hydration status and recent activity. Fifth, richer measures of muscle quality and physical performance were not available. Sixth, EAT measurements were derived from 2D single-slice assessments (axial level of the left main coronary artery ostium) rather than complete 3D volumetric segmentation, and formal inter-reader and intra-reader reproducibility analyses were not available. Finally, as our cohort was drawn from a single-center health-examination setting, external validity may be limited, and multicenter prospective validation is required. Large population-based imaging cohorts, such as UK Biobank, may provide valuable opportunities for external validation and to confirm the observed interaction finding. Conclusions In this cross-sectional study, SO was associated with a distinct EAT phenotype characterized by larger area and higher radiodensity. Lower grip strength was also associated with less favorable EAT radiodensity patterns. These findings support the potential value of CT-derived EAT characterization in SO, while prospective studies are needed to establish temporality, prognostic significance, and clinical utility. Supplementary Information Supplementary Material 1. Supplementary Material 1.