Associations between sex hormones and obesity-related indicators: results from the NHANES and Mendelian randomization study Authors: Yong Yuan, Shuhao Cui, Yingqi Hou, Xiaofan Meng, Yanqiu Huang, Feng Xu, Wentao Shi Published: 2025 ABSTRACT ObjectiveThis study sought to explore the relationships between sex hormones and obesity-related markers, while utilizing Mendelian randomization (MR) to evaluate their causal links.MethodsData from 5179 participants of the 2013–2016 National Health and Nutrition Examination Survey (NHANES) were analyzed. Multivariate linear regression and generalized additive models were utilized to examine linear and non-linear associations between sex hormones and obesity-related parameters. Causal inference was explored via a two-sample MR analysis, with inverse-variance weighting (IVW) as the primary approach.ResultsCross-sectional assessments showed that in males, testosterone levels, SHBG, and T/E2 ratio were negatively correlated with obesity, while estradiol displayed a non-linear, inverted U-shaped relationship. In females, estradiol and SHBG had negative associations with adiposity measures. MR analysis revealed that genetically increased Bio-T was associated with lower obesity risk in males, while SHBG showed a negative correlation with visceral and waist obesity. In females, Bio-T and FAI were associated with higher visceral adipose tissue volume and waist circumference, findings consistent with potential causal effects.ConclusionsSex hormones show sex-specific associations with obesity. Testosterone is protective against obesity in males, whereas female hyperandrogenism elevates the risk of obesity, particularly central obesity. These results offer novel perspectives for obesity-related health interventions.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40001-025-03470-x. Introduction The World Health Organization (WHO) characterizes obesity as excess fat storage, which is closely associated with cardiovascular disease (CVD) risk components such as hypertension, type 2 diabetes (T2D), and dyslipidemia [1, 2]. In 2016, WHO documented a global 39% prevalence among overweight adults and 13% among obese adults [3], emphasizing its substantial public health significance. While body mass index (BMI) is commonly employed as a clinical measure for adiposity, it lacks the ability to reflect fat distribution patterns [1]. Specifically, abdominal obesity (waist circumference > 102 cm in males, > 88 cm in females) and android obesity (truncal fat accumulation, “apple” shape) correlate with elevated CVD risk and mortality from lipid metabolism disorders. Visceral adipose tissue (VAT) actively secretes pro-inflammatory cytokines and adipokines that promote insulin resistance and vascular dysfunction, directly driving CVD pathogenesis [4]. In contrast, gynoid obesity (lower-body fat at hips/thighs, “pear” shape) may reduce susceptibility to T2D and CVD [5]. This protective effect is attributed to subcutaneous fat in the gluteofemoral region, which acts as a metabolic “buffer” by storing excess lipids without inducing systemic inflammation [6]. Even regional fat in extremities (e.g., arm/leg fat) exhibits distinct health implications. For instance, reduced limb fat mass in females has been linked to increased central obesity risk and hyperandrogenism-related metabolic disturbances [7], while preserved leg fat in older adults correlates with lower T2D incidence [8]. Low-quality diets (high-fat diet, sugary drinks, etc.) and drug abuse are often the triggers of obesity [9, 10]. Fat accumulation in different parts of the body may have different cardiovascular outcomes, while increasing physical activity and regulating hormone secretion can effectively improve fat distribution, for example, steroid hormones may change central obesity [11]. Research indicates that sex hormones play a crucial role in regulating fat mass and its distribution during puberty in both genders [5]. In women, estrogen works in synergy with adipose tissue genes to enhance the mass of gluteofemoral subcutaneous fat in those of reproductive age, while decreasing central adipose tissue mass, thus alleviating the load of cardiovascular disease (CVD) risk factors [8]. Nevertheless, elevated estrogen levels in women are linked to a higher risk of uterine fibroids or reproductive malignancies [12]. Multiple research studies suggest that hyperandrogenism in females is closely related to polycystic ovary syndrome, frequently accompanied by heightened intra-abdominal fat accumulation [13]. The free androgen index (FAI) serves as a proxy for free testosterone in women and is widely utilized in investigations of hyperandrogenism [14]. Conversely, males with diminished testosterone levels (evidenced by increased upper body fat and visceral adipose tissue mass) [5] face a notably higher risk of type 2 diabetes (T2D) and cardiovascular disease (CVD). Additionally, their overall mortality rate is markedly higher compared to women [15]. During the middle to late stages of puberty, the testes and peripheral tissues might generate relatively more estrogen before testosterone production reaches adult levels, contributing to the development of gynecomastia [16]. Presently, the testosterone-to-estradiol (T/E2) ratio is being established as a biomarker for a balanced hormonal environment [17] and has been identified as a predictor of cardiovascular disease and overall mortality [18]. Furthermore, sex hormone-binding globulin (SHBG) is a factor that modulates the biological effects of steroid hormones [19] and is frequently linked to metabolic abnormalities [17]. Genetic disruption of hypothalamic–pituitary–gonadal signaling underscores the causal link between gonadotropin defects and androgen deficiency. For example, CCDC141 variants are implicated in congenital hypogonadotropic hypogonadism, highlighting how germline architecture can shape circulating sex steroids and downstream metabolic risk [20]. Research has shown that sex hormones play a pivotal role in regulating body fat distribution, giving rise to diverse obesity phenotypes. However, existing observational studies are hampered by small, unrepresentative sample sizes and insufficient consideration of fat distribution patterns and obesity classifications. In light of these limitations, Mendelian randomization (MR) studies, which use genetic data as an instrumental variable to clarify phenotype relationships, offer a valuable supplementary approach to explore causal links. Since genetic alleles are randomly assigned during meiosis and independent of environmental factors, MR-derived genetic associations are less susceptible to confounding bias or reverse causation risks [21]. Notably, some MR studies have only focused on causal connections between sex hormones, BMI, and waist-to-hip ratio (WHR), neglecting crucial indicators such as arm circumference and regional fat masses in the arms, legs, and hips [5, 22]. Recognizing the distinct health implications of pear-shaped (dominated by lower-body fat) and apple-shaped (dominated by upper-body fat) body types, incorporating site-specific fat content is vital for a comprehensive analysis. Therefore, this study aimed to investigate the association between sex hormones and obesity-related markers. By leveraging data from the National Health and Nutrition Examination Survey (NHANES) and applying MR methods in a public genetic database, the research explored causal relationships, addressing the gaps in previous studies and providing a more nuanced understanding of the hormonal influences on obesity. Beyond endocrine regulation, environmental and thermal stressors can influence testicular function and androgen biosynthesis. Heat stress perturbs spermatogenesis and testicular homeostasis through oxidative stress, ER-stress, and apoptosis-related pathways, potentially lowering circulating testosterone and altering downstream metabolic risk [23]. Methods Overall study design The present study comprised two sequential phases, as illustrated in Figure S1. In the first phase, we leveraged participant data from the NHANES database to conduct generalized additive modeling and linear regression analyses, aiming to explore the relationships between steroid hormones and obesity-associated markers. The second phase involved MR analyses on aggregated Genome-Wide Association Study (GWAS) datasets. This approach was used to evaluate causal links between heritably determined steroid hormone levels and obesity-related indicators, thereby addressing the causal inference gap in observational research. Data sources and study population This study employed two cycles of publicly available NHANES data from 2013 to 2016. Males and premenopausal females (hereafter “females”) were included, with menopausal status evaluated via a reproductive health questionnaire. Specifically, participants were asked, “Have you had at least one menstrual period in the last 12 months?” Those responding “No” were further asked: “What is the reason you have not had a period in the last 12 months? (Options: pregnancy, breastfeeding, hysterectomy, menopause/life change, other)”. Females with no menstrual period in the past 12 months were classified as postmenopausal, and those with a history of sex hormone use were excluded to mitigate bias [19]. After excluding postmenopausal females, participants lacking physical examinations (body measures, X-ray), sex hormone data, or covariates, a total of 5179 subjects were included. All analyses used NHANES public data, approved by the National Center for Health Statistics (NCHS) Research Ethics Review Board, with written informed consent from all participants. A flowchart detailing participant selection is shown in Figure S2. Measurement of sex hormones The isotope dilution liquid chromatography–tandem mass spectrometry, developed by the National Center for Disease Control and Prevention (CDC), enables direct quantification of total testosterone and estradiol. MEC operates 5 days/week (with rotating non-business days). Participants were randomly assigned to morning, afternoon, or evening sessions. Participants aged ≥ 12 in morning sessions were required to fast for 9 h; fasting status was verified by phlebotomists before blood collection. Blood samples were processed, aliquoted into vials, refrigerated or frozen at MEC, then shipped to laboratories across the U.S. The lower limit of detection (LLOD) for total testosterone (ng/dL) is 0.75 ng/mL, and for estradiol (pg/mL) is 2.994 pg/mL. For SHBG assessment, the method leverages the binding interaction between SHBG and an immunizing antibody, with chemiluminescent detection of resultant reaction products. The T/E2 ratio is calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0.1\times total testosterone (ng/dL) / estradiol (pg/mL)$$\end{document}0.1×totaltestosterone(ng/dL)/estradiol(pg/mL)[14]. FAI is calculated using the following formula: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$100\times total testosterone (nmol/L)/ SHBG (nmol/L)$$\end{document}100×totaltestosterone(nmol/L)/SHBG(nmol/L)[24]. Measurement of obesity-related indicators Trained health technicians at the Mobile Examination Center (MEC) obtained anthropometric data (BMI, waist circumference, arm circumference, sagittal mean abdominal diameter) through standardized measurements. Whole-body dual-energy X-ray absorptiometry (DXA) scans were conducted for all participants, quantifying fat distribution in android/gynoid regions (including trunk, visceral adipose tissue [VAT], and subcutaneous adipose tissue [SAT]). The Hologic APEX software, utilized in scan analysis, defined regions for VAT, SAT, and the android-to-gynoid (A/G) ratio. Assessment of covariates Demographic information (age, sex, ethnicity, education, household income) and lifestyle elements (smoking, alcohol consumption, diet, physical activity) were collected through questionnaire interviews. Non-smokers were defined as individuals with a lifetime smoking history of less than 100 cigarettes, while all others were designated as smokers [25]. In the analyses, alcohol consumption was handled as a continuous variable. Dietary intake was assessed using two 24-h dietary surveys, with overall diet quality measured by the healthy eating index 2015 (HEI—2015), which produces scores from 0 to 100 (higher scores indicate healthier diets). Total leisure-time physical activity (LTPA) was calculated as the sum of moderate-intensity recreational activity minutes and double the duration of high-intensity recreational activity. Based on the 2018 Physical Activity Guidelines for Americans, participants were divided into groups: inactive (with no LTPA), underactive (where 0 < LTPA < 150 min per week), or sufficiently active (with LTPA ≥ 150 min per week) [26]. Statistical analysis Data across two survey cycles were integrated with 4-year sampling weights to maintain representativeness. Group comparisons utilized Rao–Scott Chi-square tests and t-tests. Generalized additive models were applied to characterize smooth associations between continuous hormone metrics and obesity markers, accounting for covariates. Sex hormones, stratified by tertiles, underwent multivariate linear regression to assess their impacts on obesity, with β-coefficients and 95% confidence intervals computed. Model 1 controlled for age, sex, and race; Model 2 additionally incorporated education, income, smoking, alcohol intake, physical activity, and dietary patterns. Analyses were performed using SAS 9.4.7 and R 4.2.3, with a two-tailed significance criterion of 0.05. Mendelian randomization Basic concept of MR analysis In MR analysis, valid instrumental variables (IVs) need to follow the following criteria: (1) single-nucleotide polymorphisms (SNPs) are strongly associated with sex hormones; (2) SNPs are not associated with confounders; and (3) SNPs affect obesity-related indicators only through sex hormones (Figure S3). Data sources For two-sample MR, we selected 4 sex hormone indicators (FAI, Bio-T, estradiol, SHBG, stratified by sex) and 16 obesity-related indicators. Genetic data for FAI were retrieved from a meta-analysis of UKB and FinnGen [27]. Bio-T, estradiol, and SHBG genetic data were all obtained from relevant UKB studies [28, 29]. Obesity-related genetic datasets were sourced from GIANT, UKB, HISLA, and consortia of European/African ancestries [30–33]. Included studies were approved by their institutional review boards (IRBs) and ethics committees. Detailed information is provided in Table S1. Selection of SNPs for MR analysis SNPs with genome-wide significant associations to sex hormones (P < 5 × 10–8) underwent initial screening. Subsequently, SNPs within the target gene region and its adjacent > 10,000 kb window were extracted, with linkage disequilibrium (LD) clustering applied in the European 1000 Genome dataset (r2 < 0.001). Moreover, SNPs showing genome-wide significance for obesity indicators (P < 1.0 × 10–5) were excluded. The F-statistic was calculated to remove SNPs with F < 10. MR and sensitivity analysis Inverse variance weighting (IVW) served as the core analytical technique to examine causal links between sex hormones and obesity-related indicators under a random-effects framework. MR-Egger and weighted median methods were employed for result validation. Our two-sample MR workflow (harmonization, clumping, IVW as primary, with weighted median and MR-Egger sensitivity) follows current practice applied across complex traits, including recent analyses of circulating immune cell traits and cancer risk [34]. Heterogeneity was gauged using Cochran’s Q test. Horizontal pleiotropy was identified via MR-Egger regression (intercept indicating directional pleiotropy) and funnel plot symmetry assessment. Leave-one-out sensitivity analysis was utilized to verify result stability. Multidimensional pleiotropy was scrutinized through the MR-PRESSO test. All procedures were executed with R 4.2.3 and the TwoSampleMR software package. Results Population characteristics of NHANES A total of 5179 individuals were finally included in the analysis during the two cycles 2013–2016, of which 2870 (55.4%) were males and 2309 (44.6%) were females, whose main characteristics are shown in Table 1. Compared to males, females were younger, consumed less alcohol, and had lower smoking prevalence, but had higher HEI scores. These two groups differed significantly in racial composition, household poverty ratio, and marital status, while education and history of diabetes did not differ significantly. There were significant differences between the two groups in the comparison of 16 obesity-related indicators and sex hormone levels. Table 1Baseline characteristics: NHANES (2013–2016)CharacteristicsTotal (N = 5179)Male (N = 2870)Female (N = 2309)P-valueDemographic information Age, years, mean (SE)31.6 (0.2)32.5 (0.3)30.5 (0.3) < 0.001 Alcohol consumption, gram, mean(SE)24.7 (0.5)30.8 (0.7)17.1 (0.5) < 0.001 HEI-2015, mean (SE)48.5 (0.2)47.7 (0.3)49.5 (0.3) < 0.001Ethnicity, n (%) Mexican American1041 (20.1)543 (18.9)498 (21.6)0.005 Other Hispanic588 (11.4)298 (10.4)290 (12.6) Non-Hispanic White1711 (33.0)992 (34.6)719 (31.1) Non-Hispanic Black986 (19.0)558 (19.4)428 (18.5) Other race—including multi-racial853 (16.5)479 (16.7)374 (16.2)Education levels, n (%) Less than high s chool1936 (37.4)1064 (37.1)872 (37.8)0.609 High school or above3243 (62.6)1806 (63.0)1437 (62.2)Marriage status, n (%) Have a partner or be married2277 (44.0)1335 (46.5)942 (40.8) < 0.001 Single or widowed2902 (56.0)1535 (53.5)1367 (59.2)Poverty status, n (%) ≤ 1.301989 (39.2)1048 (37.2)941 (41.7)0.002 1.30–3.50764 (15.1)423 (15.0)341 (15.1) > 3.502324 (45.8)1347 (47.8)977 (43.3)Recent tobacco use, n (%) Yes1124 (21.7)745 (26.0)379 (16.4) < 0.001 No4055 (78.3)2125 (74.0)1930 (83.6)Leisure time physical activity, n (%) 02009 (38.8)1049 (36.6)960 (41.6) < 0.001 0–150779 (15.0)361 (12.6)418 (18.1) ≥ 1502391 (46.2)1460 (50.9)931 (40.3)History of diabetes, n (%) Yes251 (4.9)149 (5.2)102 (4.4)0.197 No4928 (95.2)2721 (94.8)2207 (95.6)Sex hormones and indicators of obesity and adiposity Sex hormones and related indices Total testosterone, ng/dL, mean(SE)248.5 (3.4)427.3 (3.5)26.2 (0.4) < 0.001 Estradiol, pg/mL, mean(SE)52.1 (0.9)23.8 (0.2)87.4 (1.7) < 0.001 SHBG, nmol/L, mean (SE)50.0 (0.5)36.9 (0.4)66.3 (1.0) < 0.001 T/E2 ratio, mean (SE)–1.9 (0.02)– < 0.001 FAI, mean (SE)––1.9 (0.04)–Body measurements Body mass index, kg/m2, mean(SE)27.7 (0.1)27.3 (0.1)28.1 (0.2) < 0.001 Waist circumference, cm, mean(SE)93.5 (0.2)94.4 (0.3)92.3 (0.4) < 0.001 Arm circumference, cm, mean(SE)32.3 (0.1)33.2 (0.1)31.2 (0.1) < 0.001 Average sagittal abdominal diameter, cm, mean(SE)21.2 (0.1)21.6 (0.1)20.8 (0.1) < 0.001Android/gynoid measurements Android fat mass, gram, mean(SE)2145.4 (18.2)2066.5 (24.6)2243.4 (26.9) < 0.001 Gynoid fat mass, gram, mean(SE)4291.3 (26.1)3696.5 (30.7)5030.6 (39.5) < 0.001 Subcutaneous fat mass, gram, mean(SE)1468.3 (11.5)1178.4 (13.7)1828.7 (16.7) < 0.001 Total abdominal fat mass, gram, mean(SE)1896.7 (14.0)1630.6 (17.4)2227.5 (20.9) < 0.001 Visceral adipose tissue mass, gram, mean(SE)428.4 (3.7)452.2 (4.9)398.8 (5.5) < 0.001Whole body measurements Head fat, gram, Mean (SE)1166.6 (2.3)1238.2 (2.9)1077.7 (2.7) < 0.001 Left arm fat, gram, mean(SE)1559.1 (11.2)1369.9 (13.3)1794.4 (17.6) < 0.001 Left leg fat, gram, mean(SE)4570.4 (29.2)3902.2 (33.7)5401.0 (44.8) < 0.001 Right arm fat, gram, mean(SE)1549.5 (11.2)720.5 (13.4)1782.8 (17.7) < 0.001 Right leg fat, gram, mean(SE)4669.0 (29.84)3979.0 (34.3)5526.8 (45.7) < 0.001 Trunk fat, gram, mean (SE)11,834.5 (87.7)10,966.0 (113.7)12,914.0 (133.3) < 0.001 Total fat, gram, mean (SE)25,349.3 (162.1)22,817.1 (201.3)28,496.6 (248.6) < 0.001Descriptive data were shown as mean (SE) while categorical variables were reported as n (%). P-values less than 0.05 (P-value < 0.05) were considered significant. SE standard error of mean, HEI healthy eating index, CVD cardiovascular disease, N number, SHBG sex hormone-binding globulin, T/E2 testosterone-to-estradiol, FAI free androgen index Table 1Baseline characteristics: NHANES (2013–2016)CharacteristicsTotal (N = 5179)Male (N = 2870)Female (N = 2309)P-valueDemographic information Age, years, mean (SE)31.6 (0.2)32.5 (0.3)30.5 (0.3) < 0.001 Alcohol consumption, gram, mean(SE)24.7 (0.5)30.8 (0.7)17.1 (0.5) < 0.001 HEI-2015, mean (SE)48.5 (0.2)47.7 (0.3)49.5 (0.3) < 0.001Ethnicity, n (%) Mexican American1041 (20.1)543 (18.9)498 (21.6)0.005 Other Hispanic588 (11.4)298 (10.4)290 (12.6) Non-Hispanic White1711 (33.0)992 (34.6)719 (31.1) Non-Hispanic Black986 (19.0)558 (19.4)428 (18.5) Other race—including multi-racial853 (16.5)479 (16.7)374 (16.2)Education levels, n (%) Less than high s chool1936 (37.4)1064 (37.1)872 (37.8)0.609 High school or above3243 (62.6)1806 (63.0)1437 (62.2)Marriage status, n (%) Have a partner or be married2277 (44.0)1335 (46.5)942 (40.8) < 0.001 Single or widowed2902 (56.0)1535 (53.5)1367 (59.2)Poverty status, n (%) ≤ 1.301989 (39.2)1048 (37.2)941 (41.7)0.002 1.30–3.50764 (15.1)423 (15.0)341 (15.1) > 3.502324 (45.8)1347 (47.8)977 (43.3)Recent tobacco use, n (%) Yes1124 (21.7)745 (26.0)379 (16.4) < 0.001 No4055 (78.3)2125 (74.0)1930 (83.6)Leisure time physical activity, n (%) 02009 (38.8)1049 (36.6)960 (41.6) < 0.001 0–150779 (15.0)361 (12.6)418 (18.1) ≥ 1502391 (46.2)1460 (50.9)931 (40.3)History of diabetes, n (%) Yes251 (4.9)149 (5.2)102 (4.4)0.197 No4928 (95.2)2721 (94.8)2207 (95.6)Sex hormones and indicators of obesity and adiposity Sex hormones and related indices Total testosterone, ng/dL, mean(SE)248.5 (3.4)427.3 (3.5)26.2 (0.4) < 0.001 Estradiol, pg/mL, mean(SE)52.1 (0.9)23.8 (0.2)87.4 (1.7) < 0.001 SHBG, nmol/L, mean (SE)50.0 (0.5)36.9 (0.4)66.3 (1.0) < 0.001 T/E2 ratio, mean (SE)–1.9 (0.02)– < 0.001 FAI, mean (SE)––1.9 (0.04)–Body measurements Body mass index, kg/m2, mean(SE)27.7 (0.1)27.3 (0.1)28.1 (0.2) < 0.001 Waist circumference, cm, mean(SE)93.5 (0.2)94.4 (0.3)92.3 (0.4) < 0.001 Arm circumference, cm, mean(SE)32.3 (0.1)33.2 (0.1)31.2 (0.1) < 0.001 Average sagittal abdominal diameter, cm, mean(SE)21.2 (0.1)21.6 (0.1)20.8 (0.1) < 0.001Android/gynoid measurements Android fat mass, gram, mean(SE)2145.4 (18.2)2066.5 (24.6)2243.4 (26.9) < 0.001 Gynoid fat mass, gram, mean(SE)4291.3 (26.1)3696.5 (30.7)5030.6 (39.5) < 0.001 Subcutaneous fat mass, gram, mean(SE)1468.3 (11.5)1178.4 (13.7)1828.7 (16.7) < 0.001 Total abdominal fat mass, gram, mean(SE)1896.7 (14.0)1630.6 (17.4)2227.5 (20.9) < 0.001 Visceral adipose tissue mass, gram, mean(SE)428.4 (3.7)452.2 (4.9)398.8 (5.5) < 0.001Whole body measurements Head fat, gram, Mean (SE)1166.6 (2.3)1238.2 (2.9)1077.7 (2.7) < 0.001 Left arm fat, gram, mean(SE)1559.1 (11.2)1369.9 (13.3)1794.4 (17.6) < 0.001 Left leg fat, gram, mean(SE)4570.4 (29.2)3902.2 (33.7)5401.0 (44.8) < 0.001 Right arm fat, gram, mean(SE)1549.5 (11.2)720.5 (13.4)1782.8 (17.7) < 0.001 Right leg fat, gram, mean(SE)4669.0 (29.84)3979.0 (34.3)5526.8 (45.7) < 0.001 Trunk fat, gram, mean (SE)11,834.5 (87.7)10,966.0 (113.7)12,914.0 (133.3) < 0.001 Total fat, gram, mean (SE)25,349.3 (162.1)22,817.1 (201.3)28,496.6 (248.6) < 0.001Descriptive data were shown as mean (SE) while categorical variables were reported as n (%). P-values less than 0.05 (P-value < 0.05) were considered significant. SE standard error of mean, HEI healthy eating index, CVD cardiovascular disease, N number, SHBG sex hormone-binding globulin, T/E2 testosterone-to-estradiol, FAI free androgen index Baseline characteristics: NHANES (2013–2016) Descriptive data were shown as mean (SE) while categorical variables were reported as n (%). P-values less than 0.05 (P-value < 0.05) were considered significant. SE standard error of mean, HEI healthy eating index, CVD cardiovascular disease, N number, SHBG sex hormone-binding globulin, T/E2 testosterone-to-estradiol, FAI free androgen index Associations of sex hormones, SHBG, or T/E2 ratio with obesity-related indicators in males Using tertile groupings of total testosterone, estradiol, and SHBG, we examined relationships with 16 obesity-associated markers in male subjects (Table 2, Table S2, Table S3). Smoothed curves (solid line) and 95% confidence intervals (dashed line) for associations between these hormones, the T/E2 ratio, and the 16 indicators—adjusted for covariates—are presented in Fig. 1, Figure S4–S6. Total testosterone, SHBG, and the T/E2 ratio exhibited negative correlations with all 16 obesity-related metrics (P < 0.05). Specifically, SHBG showed significant negative associations with 14 indicators across three tertiles in both models (P < 0.05), with only non-significant links to visceral and head adiposity in the second tertile (P > 0.05). In contrast, total testosterone demonstrated consistent significant negative relationships with all 16 indicators in the upper two tertiles (P < 0.05). For estradiol, smoothed curves for 14 indicators (excluding visceral and head fat mass) revealed an inverted U-shaped relationship. At a mean estradiol level of 23.8 pg/mL in males, these 14 indicators showed significant positive correlations with estradiol (P < 0.05). Table 2Association of total testosterone with 16 obesity-related indicators in males from NHANES 2013–2016 (n = 2870)Male—total testosterone, ng/dLQ1 (< 336.0 ng/dL)Q2 (336.0–480.0 ng/dL)Q3 (≥ 480.0 ng/dL)β (95% CI)P-valueβ (95% CI)P-valueβ (95% CI)P-valueBody measures Body mass index, kg/m2Model 1−0.009 (−0.017, −0.0002)0.046−0.016 (−0.028,−0.005)0.007−0.007 (−0.010,−0.004) < 0.001Model 2−0.013 (−0.021, −0.004)0.005−0.017 (−0.027,−0.006)0.003−0.007 (−0.010,−0.004) < 0.001 Waist circumference, cmModel 1−0.020 (−0.0410, 0.000)0.054−0.047 (−0.076,−0.018)0.002−0.020 (−0.029,−0.011) < 0.001Model 2−0.030 (−0.051,−0.009)0.006−0.048 (−0.074,−0.022) < 0.001−0.020 (−0.029,−0.010) < 0.001 Arm circumference, cmModel 1−0.001 (−0.007,0.006)0.834−0.009 (−0.018,0.001)0.080−0.006 (−0.008,−0.003) < 0.001Model 2−0.005 (−0.011,0.001)0.099−0.009 (−0.018,−0.000)0.049−0.006 (−0.008,−0.003) < 0.001 Average sagittal abdominal diameter, cmModel 1−0.007 (−0.013,−0.002)0.014−0.014 (−0.022,−0.005)0.002−0.005 (−0.008,−0.002) < 0.001Model 2−0.010 (−0.015,−0.004) < 0.001−0.014 (−0.021,−0.006) < 0.001−0.005 (−0.008,−0.002)0.001Android/gynoid measurements Android fat mass, gramModel 1−2.446 (−4.257,−0.634)0.010−4.769 (−7.101,−2.437) < 0.001−1.396 (−2.088,−0.703) < 0.001Model 2−3.202 (−5.108,−1.296)0.002−4.855 (−6.944,−2.766) < 0.001−1.280 (−2.010,−0.549)0.001 Gynoid fat mass, gramModel 1−1.705 (−4.205,0.794)0.174−4.536 (−7.302,−1.771)0.002−2.065 (−2.907,−1.224) < 0.001Model 2−2.854 (−5.329,−0.379)0.025−4.683 (−7.179,−2.187) < 0.001−1.956 (−2.808,−1.105) < 0.001 Subcutaneous fat mass, gramModel 1−1.017 (−2.012,−0.021)0.046−2.678 (−3.962,−1.394) < 0.001−0.891 (−1.280,−0.502) < 0.001Model 2−1.470 (−2.467,−0.4730.005−2.725 (−3.856,−1.594) < 0.001−0.830 (−1.234,−0.425) < 0.001 Total abdominal fat mass, gramModel 1−1.699 (−2.895,−0.504)0.007−3.379 (−5.063,−1.696) < 0.001−1.143 (−1.646,−0.639) < 0.001Model 2−2.165 (−3.380,−0.949)0.001−3.432 (−4.912,−1.952) < 0.001−1.058 (−1.585,−0.531) < 0.001 Visceral adipose tissue mass, gramModel 1−0.683 (−0.949,−0.416) < 0.001−0.701 (−1.208,−0.195)0.008−0.252 (−0.383,−0.120) < 0.001Model 2−0.695 (−0.972,−0.418) < 0.001−0.707 (−1.179,−0.234)0.005−0.228 (−0.367,−0.089)0.002Whole body measurements Head fat, gramModel 10.042 (−0.230,0.315)0.752−0.349 (−0.621,−0.077)0.014−0.157 (−0.263,−0.052)0.005Model 2−0.062 (−0.326,0.203)0.637−0.350 (−0.619,−0.081)0.013−0.142 (−0.247,−0.036)0.010 Left arm fat, gramModel 1−1.555 (−2.653,−0.457)0.007−1.952 (−3.282,−0.621)0.006−0.706 (−1.036,−0.375) < 0.001Model 2−1.994 (−3.151,−0.837)0.001−1.997 (−3.231,−0.764)0.003−0.669 (−1.007,−0.331) < 0.001 Left leg fat, gramModel 1−3.300 (−6.111,−0.489)0.023−4.444 (−7.566,−1.321)0.007−2.180 (−2.942,−1.417) < 0.001Model 2−4.272 (−7.065,−1.478)0.004−4.574 (−7.553,−1.594)0.004−2.064 (−2.819,−1.308) < 0.001 Right arm fat, gramModel 1−1.700 (−2.790,−0.609)0.003−2.305 (−3.467,−1.142) < 0.001−0.748 (−1.080,−0.417) < 0.001Model 2−2.140 (−3.289,−0.991) < 0.001−2.362 (−3.496,−1.227) < 0.001−0.703 (−1.043,−0.364) < 0.001 Right leg fat, gramModel 1−3.229 (−5.962,−0.496)0.022−4.638 (−7.717,−1.566)0.004−2.222 (−2.970,−1.475) < 0.001Model 2−4.186 (−6.925,−1.446)0.004−4.756 (−7.687,−1.825)0.002−2.096 (−2.836,−1.355) < 0.001 Trunk fat, gramModel 1−10.462 (−19.686,−1.238)0.028−21.927 (−32.526,−11.327) < 0.001−6.896 (−10.192,−3.599) < 0.001Model 2−14.106 (−23.310,−4.902)0.004−22.382 (−31.912,−12.852) < 0.001−6.310 (−9.753,−2.868) < 0.001Total fat, gramModel 1−20.203 (−36.704,−3.702)0.018−35.614 (−54.092,−17.137) < 0.001−12.909 (−18.354,−7.464) < 0.001Model 2−26.760 (−43.311,−10.209)0.003−36.421 (−53.389,−19.453) < 0.001−11.983 (−17.577,−6.390) < 0.001Model 1 was adjusted for age and ethnicity. Model 2 was adjusted for age, ethnicity, education level, poverty status, smoking status, HEI-2015, alcohol consumption and leisure-time physical activity. CI confidence interval, P-values less than 0.05 (P-value < 0.05) were considered significantFig. 1Smoothed curves derived from a generalized additive model illustrating the relationship between total testosterone (continuous variable) and 16 obesity-related indicators (continuous variable). Model was adjusted for age, ethnicity, education level, poverty status, smoking status, HEI-2015, alcohol consumption and leisure-time physical activity. gm gram Table 2Association of total testosterone with 16 obesity-related indicators in males from NHANES 2013–2016 (n = 2870)Male—total testosterone, ng/dLQ1 (< 336.0 ng/dL)Q2 (336.0–480.0 ng/dL)Q3 (≥ 480.0 ng/dL)β (95% CI)P-valueβ (95% CI)P-valueβ (95% CI)P-valueBody measures Body mass index, kg/m2Model 1−0.009 (−0.017, −0.0002)0.046−0.016 (−0.028,−0.005)0.007−0.007 (−0.010,−0.004) < 0.001Model 2−0.013 (−0.021, −0.004)0.005−0.017 (−0.027,−0.006)0.003−0.007 (−0.010,−0.004) < 0.001 Waist circumference, cmModel 1−0.020 (−0.0410, 0.000)0.054−0.047 (−0.076,−0.018)0.002−0.020 (−0.029,−0.011) < 0.001Model 2−0.030 (−0.051,−0.009)0.006−0.048 (−0.074,−0.022) < 0.001−0.020 (−0.029,−0.010) < 0.001 Arm circumference, cmModel 1−0.001 (−0.007,0.006)0.834−0.009 (−0.018,0.001)0.080−0.006 (−0.008,−0.003) < 0.001Model 2−0.005 (−0.011,0.001)0.099−0.009 (−0.018,−0.000)0.049−0.006 (−0.008,−0.003) < 0.001 Average sagittal abdominal diameter, cmModel 1−0.007 (−0.013,−0.002)0.014−0.014 (−0.022,−0.005)0.002−0.005 (−0.008,−0.002) < 0.001Model 2−0.010 (−0.015,−0.004) < 0.001−0.014 (−0.021,−0.006) < 0.001−0.005 (−0.008,−0.002)0.001Android/gynoid measurements Android fat mass, gramModel 1−2.446 (−4.257,−0.634)0.010−4.769 (−7.101,−2.437) < 0.001−1.396 (−2.088,−0.703) < 0.001Model 2−3.202 (−5.108,−1.296)0.002−4.855 (−6.944,−2.766) < 0.001−1.280 (−2.010,−0.549)0.001 Gynoid fat mass, gramModel 1−1.705 (−4.205,0.794)0.174−4.536 (−7.302,−1.771)0.002−2.065 (−2.907,−1.224) < 0.001Model 2−2.854 (−5.329,−0.379)0.025−4.683 (−7.179,−2.187) < 0.001−1.956 (−2.808,−1.105) < 0.001 Subcutaneous fat mass, gramModel 1−1.017 (−2.012,−0.021)0.046−2.678 (−3.962,−1.394) < 0.001−0.891 (−1.280,−0.502) < 0.001Model 2−1.470 (−2.467,−0.4730.005−2.725 (−3.856,−1.594) < 0.001−0.830 (−1.234,−0.425) < 0.001 Total abdominal fat mass, gramModel 1−1.699 (−2.895,−0.504)0.007−3.379 (−5.063,−1.696) < 0.001−1.143 (−1.646,−0.639) < 0.001Model 2−2.165 (−3.380,−0.949)0.001−3.432 (−4.912,−1.952) < 0.001−1.058 (−1.585,−0.531) < 0.001 Visceral adipose tissue mass, gramModel 1−0.683 (−0.949,−0.416) < 0.001−0.701 (−1.208,−0.195)0.008−0.252 (−0.383,−0.120) < 0.001Model 2−0.695 (−0.972,−0.418) < 0.001−0.707 (−1.179,−0.234)0.005−0.228 (−0.367,−0.089)0.002Whole body measurements Head fat, gramModel 10.042 (−0.230,0.315)0.752−0.349 (−0.621,−0.077)0.014−0.157 (−0.263,−0.052)0.005Model 2−0.062 (−0.326,0.203)0.637−0.350 (−0.619,−0.081)0.013−0.142 (−0.247,−0.036)0.010 Left arm fat, gramModel 1−1.555 (−2.653,−0.457)0.007−1.952 (−3.282,−0.621)0.006−0.706 (−1.036,−0.375) < 0.001Model 2−1.994 (−3.151,−0.837)0.001−1.997 (−3.231,−0.764)0.003−0.669 (−1.007,−0.331) < 0.001 Left leg fat, gramModel 1−3.300 (−6.111,−0.489)0.023−4.444 (−7.566,−1.321)0.007−2.180 (−2.942,−1.417) < 0.001Model 2−4.272 (−7.065,−1.478)0.004−4.574 (−7.553,−1.594)0.004−2.064 (−2.819,−1.308) < 0.001 Right arm fat, gramModel 1−1.700 (−2.790,−0.609)0.003−2.305 (−3.467,−1.142) < 0.001−0.748 (−1.080,−0.417) < 0.001Model 2−2.140 (−3.289,−0.991) < 0.001−2.362 (−3.496,−1.227) < 0.001−0.703 (−1.043,−0.364) < 0.001 Right leg fat, gramModel 1−3.229 (−5.962,−0.496)0.022−4.638 (−7.717,−1.566)0.004−2.222 (−2.970,−1.475) < 0.001Model 2−4.186 (−6.925,−1.446)0.004−4.756 (−7.687,−1.825)0.002−2.096 (−2.836,−1.355) < 0.001 Trunk fat, gramModel 1−10.462 (−19.686,−1.238)0.028−21.927 (−32.526,−11.327) < 0.001−6.896 (−10.192,−3.599) < 0.001Model 2−14.106 (−23.310,−4.902)0.004−22.382 (−31.912,−12.852) < 0.001−6.310 (−9.753,−2.868) < 0.001Total fat, gramModel 1−20.203 (−36.704,−3.702)0.018−35.614 (−54.092,−17.137) < 0.001−12.909 (−18.354,−7.464) < 0.001Model 2−26.760 (−43.311,−10.209)0.003−36.421 (−53.389,−19.453) < 0.001−11.983 (−17.577,−6.390) < 0.001Model 1 was adjusted for age and ethnicity. Model 2 was adjusted for age, ethnicity, education level, poverty status, smoking status, HEI-2015, alcohol consumption and leisure-time physical activity. CI confidence interval, P-values less than 0.05 (P-value < 0.05) were considered significant Association of total testosterone with 16 obesity-related indicators in males from NHANES 2013–2016 (n = 2870) Model 1 was adjusted for age and ethnicity. Model 2 was adjusted for age, ethnicity, education level, poverty status, smoking status, HEI-2015, alcohol consumption and leisure-time physical activity. CI confidence interval, P-values less than 0.05 (P-value < 0.05) were considered significant Smoothed curves derived from a generalized additive model illustrating the relationship between total testosterone (continuous variable) and 16 obesity-related indicators (continuous variable). Model was adjusted for age, ethnicity, education level, poverty status, smoking status, HEI-2015, alcohol consumption and leisure-time physical activity. gm gram Associations of sex hormones, SHBG, or FAI with obesity-related indicators in females Compared to males, in addition to total testosterone, estradiol, and SHBG (Figure S7, S8, S9), we additionally added smoothed curves between FAI and 16 indicators (Fig. 2). Estradiol was consistently and significantly negatively correlated with the 16 indicators in the higher two tertiles (P < 0.05) (Table 3), which is consistent with the trend of the smoothed curves. In contrast, there was no significant association between total testosterone levels and whole-body adiposity measures and body measurements (P > 0.05), and the trends were inconsistent across tertile levels. In addition, inverted U-shaped associations of both total testosterone and FAI with obesity-related indicators were observed, which were consistent with the results of the two models. SHBG showed an inverse trend in the smoothed curves with the obesity-related indicators, and was significantly negatively correlated with 16 indicators at the 1 st, and 2nd tertile levels in both models (P < 0.05).Fig. 2Smoothed curves derived from a generalized additive model illustrating the relationship between free androgen index (continuous variable) and 16 obesity-related indicators (continuous variable) in women. Model was adjusted for age, ethnicity, education level, poverty status, smoking status, HEI-2015, alcohol consumption and leisure-time physical activity. gm gramTable 3Association of estradiol with 16 obesity-related indicators in females from NHANES 2013–2016 (n = 2309)Female—estradiol, pg/mLQ1 (< 37.0 pg/mL)Q2 (37.0 −98.0 pg/mL)Q3 (≥ 98.0 pg/mL)β (95%CI)P-valueβ (95%CI)P-valueβ (95%CI)P-valueBody measures Body mass index, kg/m2Model 10.114 (0.059,0.169) < 0.001−0.055 (−0.079,−0.030) < 0.001−0.010 (−0.017,−0.003)0.005Model 20.112 (0.057,0.167) < 0.001−0.054 (−0.079,−0.028) < 0.001−0.010 (−0.018,−0.003)0.006 Waist circumference, cmModel 10.234 (0.097,0.370)0.002−0.108 (−0.172,−0.044)0.002−0.022 (−0.040,−0.003)0.027Model 20.228 (0.098,0.359)0.001−0.102 (−0.167,−0.038)0.003−0.021 (−0.004,−0.003)0.026 Arm circumference, cmModel 10.093 (0.057,0.130) < 0.001−0.037 (−0.056,−0.018) < 0.001−0.008 (−0.014,−0.003)0.003Model 20.093 (0.057,0.129) < 0.001−0.036 (−0.055,−0.016)0.001−0.008 (−0.014,−0.003)0.003 Average sagittal abdominal diameter, cmModel 10.049 (0.013,0.085)0.009−0.030 (−0.046,−0.014) < 0.001−0.007 (−0.011,−0.003)0.001Model 20.047 (0.011,0.082)0.011−0.029 (−0.045,−0.012)0.002−0.007 (−0.011,−0.003)0.002Android/gynoid measurements Android fat mass, gramModel 117.013 (4.765,29.262)0.008−10.318 (−16.318,−4.319)0.001−1.944 (−3.358,−0.531)0.009Model 216.693 (4.577,28.809)0.009−10.001 (−16.272,−3.729)0.003−1.923 (−3.302,−0.545)0.008 Gynoid fat mass, gramModel 123.618 (9.541,37.695)0.002−10.788 (−19.403,−2.172)0.016−2.709 (−4.714,−0.703)0.010Model 223.537 (8.938,38.135)0.003−11.055 (−20.437,−1.674)0.023−2.674 (−4.772,−0.575)0.014 Subcutaneous fat mass, gramModel 112.093 (5.133,19.052)0.001−6.614 (−9.875,−3.354) < 0.001−1.248 (−2.166,−0.329)0.009Model 211.948 (5.005,18.892)0.001−6.576 (−10.027,−3.124) < 0.001−1.236 (−2.156,−0.316)0.010 Total abdominal fat mass, gramModel 113.030 (4.154,21.906)0.005−8.020 (−12.118,−3.922) < 0.001−1.497 (−2.696,−0.298)0.016Model 212.686 (3.873,21.499)0.006−7.922 (−12.201,−3.645) < 0.001−1.469 (−2.646,−0.293)0.016 Visceral adipose tissue mass, gramModel 10.937 (−1.523,3.398)0.443−1.406 (−2.530,−0.282)0.016−0.249 (−0.563,0.065)0.116Model 20.738 (−1.680,3.156)0.538−1.347 (−2.521,−0.173)0.026−0.234 (−0.526,0.058)0.113Whole body measurements Head fat, gramModel 11.681 (0.623,2.734)0.003−0.455 (−0.869,−0.041)0.032−0.086 (−0.235,0.063)0.247Model 21.675 (0.631,2.718)0.003−0.436 (−0.837,−0.034)0.034−0.086 (−0.236,0.063)0.248 Left arm fat, gramModel 19.735 (3.395,16.076)0.004−6.056 (−9.341,−2.779) < 0.001−1.216 (−2.038,−0.395)0.005Model 29.653 (3.187,16.119)0.005−5.984 (−9.448,−2.520)0.001−1.207 (−2.050,−0.365)0.007 Left leg fat, gramModel 128.436 (12.726,44.146) < 0.001−11.623 (−21.829,−1.418)0.027−3.045 (−5.356,−0.733)0.012Model 228.144 (11.683,44.605)0..002−12.027 (−22.969,−1.085)0.032−3.052 (−5.535,−0.569)0.018 Right arm fat, gramModel 19.981 (3.512,16.451)0.004−6.190 (−9.590,−2.790) < 0.001−1.036 (−2.006,−0.067)0.037Model 29.945 (3.517,16.373)0.004−6.150 (−9.771,−2.529)0.002−1.020 (−1.994,−0.045)0.041 Right leg fat, gramModel 126.848 (10.414,43.282)0.002−11.358 (−21.650,−1.067)0.032−3.255 (−5.521,−0.990)0.006Model 226.652 (9.535,43.769)0.003−11.914 (−23.036,−0.793)0.037−3.241 (−5.633,−0.848)0.010 Trunk fat, gramModel 184.536 (25.935,143.138)0.006−49.928 (−75.771,−24.086) < 0.001−9.032 (−15.969,−2.096)0.012Model 283.035 (25.356,140.715)0.006−49.344 (−76.466,−22.222) < 0.001−8.890 (−15.740,−2.041)0.013Total fat, gramModel 1161.218 (72.813,249.623) < 0.001−85.615 (−135.281,−35.949)0.001−17.671 (−30.405,−4.937)0.008Model 2159.104 (69.491,248.717)0.001−85.855 (−138.681,−33.029)0.002−17.496 (−30.505,−4.488)0.010Model 1 was adjusted for age and ethnicity. Model 2 was adjusted for age, ethnicity, education level, poverty status, smoking status, HEI-2015, alcohol consumption and leisure-time physical activity. CI confidence interval, P-values less than 0.05 (p < 0.05) were considered significant Smoothed curves derived from a generalized additive model illustrating the relationship between free androgen index (continuous variable) and 16 obesity-related indicators (continuous variable) in women. Model was adjusted for age, ethnicity, education level, poverty status, smoking status, HEI-2015, alcohol consumption and leisure-time physical activity. gm gram Table 3Association of estradiol with 16 obesity-related indicators in females from NHANES 2013–2016 (n = 2309)Female—estradiol, pg/mLQ1 (< 37.0 pg/mL)Q2 (37.0 −98.0 pg/mL)Q3 (≥ 98.0 pg/mL)β (95%CI)P-valueβ (95%CI)P-valueβ (95%CI)P-valueBody measures Body mass index, kg/m2Model 10.114 (0.059,0.169) < 0.001−0.055 (−0.079,−0.030) < 0.001−0.010 (−0.017,−0.003)0.005Model 20.112 (0.057,0.167) < 0.001−0.054 (−0.079,−0.028) < 0.001−0.010 (−0.018,−0.003)0.006 Waist circumference, cmModel 10.234 (0.097,0.370)0.002−0.108 (−0.172,−0.044)0.002−0.022 (−0.040,−0.003)0.027Model 20.228 (0.098,0.359)0.001−0.102 (−0.167,−0.038)0.003−0.021 (−0.004,−0.003)0.026 Arm circumference, cmModel 10.093 (0.057,0.130) < 0.001−0.037 (−0.056,−0.018) < 0.001−0.008 (−0.014,−0.003)0.003Model 20.093 (0.057,0.129) < 0.001−0.036 (−0.055,−0.016)0.001−0.008 (−0.014,−0.003)0.003 Average sagittal abdominal diameter, cmModel 10.049 (0.013,0.085)0.009−0.030 (−0.046,−0.014) < 0.001−0.007 (−0.011,−0.003)0.001Model 20.047 (0.011,0.082)0.011−0.029 (−0.045,−0.012)0.002−0.007 (−0.011,−0.003)0.002Android/gynoid measurements Android fat mass, gramModel 117.013 (4.765,29.262)0.008−10.318 (−16.318,−4.319)0.001−1.944 (−3.358,−0.531)0.009Model 216.693 (4.577,28.809)0.009−10.001 (−16.272,−3.729)0.003−1.923 (−3.302,−0.545)0.008 Gynoid fat mass, gramModel 123.618 (9.541,37.695)0.002−10.788 (−19.403,−2.172)0.016−2.709 (−4.714,−0.703)0.010Model 223.537 (8.938,38.135)0.003−11.055 (−20.437,−1.674)0.023−2.674 (−4.772,−0.575)0.014 Subcutaneous fat mass, gramModel 112.093 (5.133,19.052)0.001−6.614 (−9.875,−3.354) < 0.001−1.248 (−2.166,−0.329)0.009Model 211.948 (5.005,18.892)0.001−6.576 (−10.027,−3.124) < 0.001−1.236 (−2.156,−0.316)0.010 Total abdominal fat mass, gramModel 113.030 (4.154,21.906)0.005−8.020 (−12.118,−3.922) < 0.001−1.497 (−2.696,−0.298)0.016Model 212.686 (3.873,21.499)0.006−7.922 (−12.201,−3.645) < 0.001−1.469 (−2.646,−0.293)0.016 Visceral adipose tissue mass, gramModel 10.937 (−1.523,3.398)0.443−1.406 (−2.530,−0.282)0.016−0.249 (−0.563,0.065)0.116Model 20.738 (−1.680,3.156)0.538−1.347 (−2.521,−0.173)0.026−0.234 (−0.526,0.058)0.113Whole body measurements Head fat, gramModel 11.681 (0.623,2.734)0.003−0.455 (−0.869,−0.041)0.032−0.086 (−0.235,0.063)0.247Model 21.675 (0.631,2.718)0.003−0.436 (−0.837,−0.034)0.034−0.086 (−0.236,0.063)0.248 Left arm fat, gramModel 19.735 (3.395,16.076)0.004−6.056 (−9.341,−2.779) < 0.001−1.216 (−2.038,−0.395)0.005Model 29.653 (3.187,16.119)0.005−5.984 (−9.448,−2.520)0.001−1.207 (−2.050,−0.365)0.007 Left leg fat, gramModel 128.436 (12.726,44.146) < 0.001−11.623 (−21.829,−1.418)0.027−3.045 (−5.356,−0.733)0.012Model 228.144 (11.683,44.605)0..002−12.027 (−22.969,−1.085)0.032−3.052 (−5.535,−0.569)0.018 Right arm fat, gramModel 19.981 (3.512,16.451)0.004−6.190 (−9.590,−2.790) < 0.001−1.036 (−2.006,−0.067)0.037Model 29.945 (3.517,16.373)0.004−6.150 (−9.771,−2.529)0.002−1.020 (−1.994,−0.045)0.041 Right leg fat, gramModel 126.848 (10.414,43.282)0.002−11.358 (−21.650,−1.067)0.032−3.255 (−5.521,−0.990)0.006Model 226.652 (9.535,43.769)0.003−11.914 (−23.036,−0.793)0.037−3.241 (−5.633,−0.848)0.010 Trunk fat, gramModel 184.536 (25.935,143.138)0.006−49.928 (−75.771,−24.086) < 0.001−9.032 (−15.969,−2.096)0.012Model 283.035 (25.356,140.715)0.006−49.344 (−76.466,−22.222) < 0.001−8.890 (−15.740,−2.041)0.013Total fat, gramModel 1161.218 (72.813,249.623) < 0.001−85.615 (−135.281,−35.949)0.001−17.671 (−30.405,−4.937)0.008Model 2159.104 (69.491,248.717)0.001−85.855 (−138.681,−33.029)0.002−17.496 (−30.505,−4.488)0.010Model 1 was adjusted for age and ethnicity. Model 2 was adjusted for age, ethnicity, education level, poverty status, smoking status, HEI-2015, alcohol consumption and leisure-time physical activity. CI confidence interval, P-values less than 0.05 (p < 0.05) were considered significant Association of estradiol with 16 obesity-related indicators in females from NHANES 2013–2016 (n = 2309) Model 1 was adjusted for age and ethnicity. Model 2 was adjusted for age, ethnicity, education level, poverty status, smoking status, HEI-2015, alcohol consumption and leisure-time physical activity. CI confidence interval, P-values less than 0.05 (p < 0.05) were considered significant MR of sex hormones and obesity-related indicators in males IVW analyses using a random-effects model revealed associations between sex hormones and obesity indicators in males, consistent with significant potential causal effects (Fig. 3). A higher FAI for gene compensation was associated with reduced total body adiposity but was associated with increased WHR [β (95%CI) 0.045 (0.006, 0.084)] and BMI-adjusted WHR [0.067 (0.004, 0.129)]. Bio-T was also associated with decreased total body adiposity, and showed an inverse association with waist circumference—findings consistent with potential causal effects [−0.073 (−0.141, −0.005)]. Estradiol was genetically linked to a lower risk of WHR [−0.637 (−1.092, −0.183)]. SHBG was associated with lower visceral adipose tissue volume [−0.056 (−0.096, −0.016)], WHR [−0.057 (−0.089, −0.025)] and BMI-adjusted WHR [−0.057 (−0.103, −0.011)].Fig. 3Forest plot of the Mendelian randomization study investigating the effect of sex hormones on obesity-related indicators in males. BioT bioavailable testosterone, BMI body mass index, FAI free androgen index, SHBG sex hormone-binding globulin. P-values less than 0.05 (p < 0.05) were considered significant Forest plot of the Mendelian randomization study investigating the effect of sex hormones on obesity-related indicators in males. BioT bioavailable testosterone, BMI body mass index, FAI free androgen index, SHBG sex hormone-binding globulin. P-values less than 0.05 (p < 0.05) were considered significant Analysis of obesity-related indicators stratified by sex showed that Bio-T was significantly and negatively associated with BMI in males [−0.068 (−0.134, −0.002)], SHBG was significantly and negatively associated with BMI-adjusted waist circumference [−0.065 (−0.104, −0.025)] and BMI-adjusted WHR [−0.066 (−0.106, −0.027)]. MR of sex hormones and obesity-related indicators in females Figure 4 illustrates associations between sex hormones and obesity indicators in females, consistent with potential causal relationships. Similar to males, results showed that genetic susceptibility to FAI was significantly associated with lower total body adiposity, but higher visceral adipose tissue volume [0.059 (0.013, 0.104)] and BMI-adjusted WHR [0.123 (0.040, 0.207)] showed significant associations consistent with potential causal effects (Fig. 4). In contrast to males, Bio-T was significantly associated with abdominal subcutaneous adipose tissue volume [0.114 (0.049, 0.178)], visceral adipose tissue volume [0.113 (0. 054, 0.171)], BMI [0.102 (0.058, 0.147)] and waist circumference [0.062 (0.002, 0.122)], which showed a significant positive association consistent with a potential causal effect. SHBG was significantly and positively associated with BMI-adjusted WHR [0.067 (0.004, 0.129)].Fig. 4Forest plot of the Mendelian randomization study investigating the effect of sex hormones on obesity-related indicators in females. BioT bioavailable testosterone, BMI body mass index, FAI free androgen index, SHBG sex hormone-binding globulin. P-values less than 0.05 (p < 0.05) were considered significant Forest plot of the Mendelian randomization study investigating the effect of sex hormones on obesity-related indicators in females. BioT bioavailable testosterone, BMI body mass index, FAI free androgen index, SHBG sex hormone-binding globulin. P-values less than 0.05 (p < 0.05) were considered significant In sex-stratified analyses, FAI was significantly positively associated with BMI-adjusted WHR [0.088 (0.033, 0.143)] and BMI-adjusted waist circumference [0.055 (0.002, 0.108)] and significantly positively associated with BMI [−0.054 (−0.102, −0.005)] while BMI-adjusted hip circumference was significantly negatively associated [−0.083 (−0.138, −0.028)]. Similarly, Bio-T was significantly positively associated with BMI [0.060 (0.002, 0.119)]. Reverse MR of sex hormones and obesity-related indicators in females and males Reverse MR results in females Reverse MR analysis revealed sex-specific causal associations between obesity-related indicators and sex hormones in females. The key outcomes from the inverse-variance weighting (IVW) analysis are as follows (Figure S12): BMI exerted a significantly positive causal effect on female BioT [0.245 (0.225, 0.265)]. In contrast, BMI exhibited a causal negative association with female FAI [−0.054 (−0.083, −0.026)], findings consistent with a potential causal effect. Left leg fat mass had a strong positive effect on female BioT [0.271 (0.234, 0.308)]. Trunk fat mass showed a positive effect on BioT [0.167 (0.136, 0.198)] and a negative effect on FAI [−0.111 (−0.155, −0.068)] simultaneously. Waist circumference exerted a positive effect on BioT [0.165 (0.091, 0.239)] and a positive effect on sex hormone-binding globulin (SHBG) [4.415 (2.028, 6.802)]. Meanwhile, it showed negative effects on estradiol [−0.254 (−0.426, −0.082)] and FAI [−0.135 (−0.247, −0.022)]. Waist-to-hip ratio (WHR) had a negative effect on estradiol [−0.102 (−0.190, −0.013)] and positive effects on FAI [0.263 (0.189, 0.337)] and SHBG [1.615 (0.152, 3.078)]. Reverse MR results in males Reverse MR analysis in males showed an opposite pattern to that in females: most obesity-related indicators exerted negative effects on androgens (BioT, FAI) and positive effects on estradiol. The specific IVW results are as follows (Figure S13). Left arm fat mass had a significantly negative association with male BioT [−0.175 (−0.206, −0.144)] and a positive association with estradiol [0.018 (0.011, 0.024)]—both consistent with potential causal effects. Right arm fat mass, left leg fat mass, and right leg fat mass all exhibited consistent trends: negative effects on BioT and FAI, and positive effects on estradiol. Specifically, left leg fat mass had a negative effect on FAI [−0.106 (−0.161, −0.051)] and a positive effect on estradiol [0.021 (0.013, 0.029)]. IVW analysis showed that BMI had a strong negative effect on male BioT [−0.135 (−0.156, −0.114)], along with a positive effect on estradiol [0.014 (0.009, 0.018)] and a negative effect on FAI [−0.051 (−0.082, −0.021)]. Trunk fat mass exerted a negative effect on male BioT [−0.179 (−0.211, −0.147)], a negative effect on FAI [−0.119 (−0.160, −0.077)], and a positive effect on estradiol [0.015 (0.008, 0.021)]. Waist circumference had a negative effect on BioT [−0.121 (−0.214, −0.027)] and a positive effect on estradiol [0.014 (0.003, 0.025)]. WHR showed a negative effect on estradiol [−0.016 (−0.026, −0.006)] and a positive effect on FAI [0.166 (0.079, 0.253)]. Sensitivity analysis For detecting potential bias in horizontal pleiotropy, MR-Egger and weighted median were utilized as sensitivity assessments to explore the linkage between sex hormones and –obesity-associated indicators (Figures S10–S11). The results of these methods were analogous to the estimates generated by the IVW approach. Discussion To our knowledge, this research constitutes the first comprehensive examination of the association between sex hormones and obesity-related parameters, combining data from a large observational study and MR analysis of large-scale genetic datasets. In males, higher testosterone, SHBG, and T/E2 ratios were associated with a decreased obesity risk, while estradiol and obesity showed an inverted U-shaped relationship. In females, estradiol and SHBG mitigated obesity-related markers, whereas testosterone and FAI displayed inconsistent or inverted U-shaped trends. MR analyses validated that elevated SHBG and testosterone potentially decreased the risk of obesity in males, while hyperandrogenism increased central obesity risk in females. Multiple sensitivity analyses, employing diverse methodologies, confirmed the robustness of the findings, with consistent results across different analytical strategies. Total testosterone denotes the entire quantity of testosterone detected in plasma. Roughly 60%–70% of testosterone associates with binding proteins (mainly sex hormone-binding globulin and albumin), forming bound testosterone. These bound testosterone function as stabilizers and reservoirs within the circulatory system. Bio-T represents the portion of testosterone capable of binding to albumin, thereby entering cells via the cell membrane and exerting physiological effects. In males, Bio-T exhibits a negative correlation with nearly all obesity-related markers. Earlier research has indicated that testosterone promotes the decrease of body fat in males [35], a finding that matches our results. Multiple studies have suggested that the impact of testosterone on obesity is multifaceted. Investigations into adipose precursor cells in male rats revealed that testosterone augments catecholamine-induced lipolysis [36]. Additionally, it has been established that androgens suppress adipogenesis in human adipose progenitor cells [37]. Our research also identified a significant association between higher SHBG levels in males and lower visceral fat mass as well as reduced waist circumference. A study on transgenic male mice with human SHBG showed that SHBG can directly stimulate lipolysis in white adipose tissue [38]. Studies have shown that SHBG can bind to LRP1 on adipocyte membranes, activate the cAMP–PKA pathway, and promote triglyceride breakdown [39, 40]. Besides its role in regulating sex hormone levels, SHBG indirectly influences fat metabolism. Cross-sectional analysis further uncovered an inverse U-shaped relationship between estradiol and obesity markers in men. Prior studies have documented a positive correlation between estradiol and BMI, although this association diminishes at higher BMI percentiles [41]. While the exact mechanisms remain unclear, they might be related to the negative feedback inhibition exerted by high estrogen levels on the hypothalamic–pituitary–testicular axis [42]. Aligned with a Nigerian investigation, our analysis identified an inverse relationship between the T/E2 ratio and obesity-associated markers [43]. A key limitation is the non-execution of a reverse MR study. Given that Lee’s research indicated that testosterone undergoes enzymatic conversion to estradiol in adipose tissue, and that increased fat mass may exert a reversing effect on sex hormone levels, future research should prioritize exploring bidirectional biological impacts [44, 45]. In comparison to males, females display lower testosterone concentrations in their blood; however, free androgens hold crucial significance for females, especially those with polycystic ovary syndrome (PCOS). Our findings illustrate that in females, FAI is strongly linked to reduced limb fat mass, higher visceral adipose tissue volume, and an elevated WHR. This also suggests that elevated levels of free androgens tend to induce central obesity in females. Epidemiological investigations have shown that females with PCOS have more severe hyperandrogenism and central obesity [13]. Testosterone can facilitate fat accumulation and insulin resistance in females with PCOS [46]. Moreover, our study discovered that estradiol in females could boost adiposity in the limbs and trunk while decreasing the waist-to-hip ratio. Although this result was not significant in the MR analysis, a significant correlation was found in the cross-sectional analysis. A previous observational study also reported a negative association between estradiol and WHR [47]. Similar to the male results, SHBG had a significant negative correlation with the fat content in the limbs and trunk of females. SHBG is closely associated with lipid metabolism. A recent MR study has indicated that genetically predicted higher SHBG levels may improve lipid profiles by increasing HDL cholesterol levels and decreasing triglyceride levels, which in turn safeguards against coronary atherosclerotic disease [48]. Adiposity is tightly coupled to immunometabolic remodeling. In particular, macrophage glutamine metabolism modulates inflammatory tone within adipose tissue and contributes to insulin resistance and obesity phenotypes. If androgen or SHBG signaling reshapes adipose macrophage programs, this could partially explain the observed associations between sex steroids and central fat depots [49]. The study boasts several notable strengths. Firstly, we integrated NHANES observational datasets with MR methods to probe into the associations between steroid hormones and obesity phenotypes. Secondly, NHANES is a large-scale, population-based study in the US. We made use of 16 obesity-related metrics as outcomes. Besides the traditional waist circumference and BMI, we zeroed in on indicators like android/gynoid obesity and also looked into the linear and non-linear relationships between steroid hormones and obesity-related indicators. Thirdly, in MR analyses, using genetic variation as an instrumental variable makes the study less prone to reverse causality and confounders, and the sample size is fairly objective. There are also some limitations worth noting. Firstly, due to our inability to obtain repeated hormone level measurements from the participants, we relied solely on questionnaires and self-reports to determine menopausal status, which introduces a certain degree of recall bias. Secondly, there is a certain amount of heterogeneity in the outcome data. Fortunately, the random-effects IVW model we selected can overlook the precision loss caused by this heterogeneity. Second, owing to variations in allele frequencies and disease/exposure prevalence across distinct ancestral groups, IVs originating from European populations might be absent in other cohorts. This circumstance confines the scope of the genetic analysis to individuals of European descent. Going forward, these results should be validated through a prospective cohort study conducted within multi-ethnic groups to address the limitations imposed by ancestral disparities in IV applicability. Thirdly, we clarify GAM’s role in our study as a visualization tool for potential non-linear trends, acknowledging its inability to formalize non-linearity testing, and suggesting that future studies could complement RCS analysis to validate these observed U-shaped associations. Fourth, defining “premenopausal women” solely by self-reported “having menstrual periods in the past 12 months” may misclassify perimenopausal women (aged 45–55 years). Their fluctuating sex hormone levels could dilute true associations between sex hormones and obesity indicators, reducing the precision of female subgroup results. Future studies should integrate biochemical markers (e.g., FSH) with menstrual history for accurate classification. Conclusions To summarize, the present study uncovered sex-specific connections between sex hormones and obesity-related indicators. Among males, androgens and SHBG showed a significant negative correlation with overall adiposity, while estradiol had a non-linear correlation with obesity. In females, hyperandrogenism was closely linked to obesity, particularly central obesity, often leading to an apple-shaped physique. Based on these results, we propose sex-specific interventions: males prioritize SHBG modulation to reduce visceral fat, and testosterone replacement for low Bio-T. Females (especially PCOS) use androgen-lowering therapies to cut central obesity risk, with DXA limb fat monitoring as an early warning. Further research is required to delve deeper into the underlying biological mechanisms. Supplementary Information Supplementary material 1. Supplementary material 1.