{
  "abstract": "Introduction The prevalence of type 2 diabetes (T2D) has surged, yet body mass index (BMI) fails to explain the 30%–40% of cases that occur in individuals with a healthy weight. Emerging evidence suggests that regional fat distribution differentially impacts glucose metabolism, independent of total adiposity. This study investigated the independent association between regional body composition and T2D risk using BMI-matched National Health and Nutrition Examination Survey (NHANES) data to identify sex-specific effects and the mediating role of insulin resistance.Research design and methods Our study employed data from the 2011–2018 cycles of NHANES. Participants were classified into a high-risk T2D group if they met one or more of the following criteria: fasting blood glucose≥6.1 mmol/L, 2-hour blood glucose≥7.8 mmol/L following an oral glucose tolerance test or self-reported physician’s diagnosis of diabetes or pre-diabetes. Body composition data were assessed via dual-energy X-ray absorptiometry, which provides a precise assessment of regional fat and muscle mass distribution.Results Participants at high T2D risk exhibited significantly reduced lower limb fat mass compared with healthy controls (p <0.001), with higher amounts of lower limb fat serving as a protective factor against both diabetes and insulin resistance. Notably, this protective effect of lower-limb fat (OR 0.86 (0.76–0.97), p=0.01) along with the detrimental impact of visceral fat (OR 7.35 (1.57–34.4), p=0.01) was particularly pronounced in male subjects. Additionally, 36.18% of the protective effect of lower limb fat on diabetes is mediated by improved insulin sensitivity.Conclusions This study delineates a protective role for lower-body fat in diabetes pathogenesis, mediated substantially through ameliorating insulin resistance. The sex-specific associations underscore the protective effect of lower-body fat and the detrimental impact of visceral adiposity in men after controlling for BMI.",
  "authors": [
    {
      "affiliations": [
        "Department of Clinical Nutrition, Peking Union Medical College Hospital, Beijing, China"
      ],
      "name": "Qiong Wang"
    },
    {
      "affiliations": [
        "Department of Clinical Nutrition, Peking Union Medical College Hospital, Beijing, China"
      ],
      "name": "Pei-Pei Chen"
    },
    {
      "affiliations": [
        "Department of Clinical Nutrition, Peking Union Medical College Hospital, Beijing, China"
      ],
      "name": "Wei Wei"
    },
    {
      "affiliations": [
        "Department of Clinical Nutrition, Peking Union Medical College Hospital, Beijing, China"
      ],
      "name": "Jia-Yu Guo"
    },
    {
      "affiliations": [
        "Department of Clinical Nutrition, Peking Union Medical College Hospital, Beijing, China"
      ],
      "name": "Yuan-Yuan Bao"
    },
    {
      "affiliations": [
        "Department of Clinical Nutrition, Peking Union Medical College Hospital, Beijing, China"
      ],
      "name": "Jing Zhang"
    },
    {
      "affiliations": [
        "Department of Clinical Nutrition, Peking Union Medical College Hospital, Beijing, China"
      ],
      "name": "Kang Yu"
    }
  ],
  "full_text": "WHAT IS ALREADY KNOWN ON THIS TOPIC Visceral fat universally increases type 2 diabetes (T2D) risk; lower-body fat’s role is debated.Body composition-T2D associations are confounded by body mass index (BMI)/sex in prior studies.WHAT THIS STUDY ADDS Sex-specific protection: lower-body fat reduces T2D risk only in men, and it is non-significant in women.Quantified mediation: 36% of lower-body fat’s benefit operates through reduced insulin resistance (Homeostasis Model Assessment of Insulin Resistance).HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Clinical practice may shift to sex-specific body composition thresholds.Research focuses on anatomical adiposity distribution>BMI for precision prevention.Introduction The past 30 years have witnessed a dramatic rise in global prevalence of type 2 diabetes mellitus (T2D), 1–3 yet its pathological complexity substantially exceeds the explanatory capacity of conventional obesity metrics such as body mass index (BMI).4 Approximately 30%–40% of individuals with diabetes exhibit normal body weight (BMI<25 kg/m²), while a significant proportion (20%–30%) of obese individuals maintain a metabolically healthy status.5 Emerging evidence indicates that body composition—especially the deposition of adipose and muscle tissues—serves as a pivotal factor in regulating glucose homeostasis.6–8 Notably, individuals with similar BMI may exhibit divergent risks of diabetes, which suggests that regional adipose deposition rather than general obesity could serve as a key determinant of metabolic health.9 10Adipose tissue exhibits diverse characteristics in different depots.11 Visceral adipose tissue (VAT) is well established to compromise insulin sensitivity via increased systemic free fatty acid levels and upregulated proinflammatory mediator secretion,12 while the metabolic function of lower-body adipose depots exerts a protective effect against these metabolic disorders.5 13 14 These studies are often limited by several methodological issues: (1) inadequate control for potential confounding by BMI and lean mass and (2) a nearly universal neglect of important sex-specific role.15 16 Crucially, adipose tissue distribution exhibits profound sexual dimorphism—estrogen promotes subcutaneous fat storage preferentially in women, whereas androgens drive visceral adiposity predominance in men.17–19 Preliminary evidence indicates that sex influences the associations between adipose tissue characteristics and metabolic risk profiles.20 21 However, no study has rigorously tested this hypothesis in BMI-matched cohorts.While earlier investigations have mainly concentrated on adipose distribution, the reduction of muscle mass has long been considered critical to the pathogenesis of diabetes.22 23 However, emerging evidence suggests that adipose distribution may exert stronger predictive power over insulin sensitivity compared with muscle mass.24 25 This controversy highlights the necessity for multidimensional body composition assessment.Recently, propensity score matching has emerged as a prominent methodological approach to address confounding variables in observational research.26 This investigation employed data from the National Health and Nutrition Examination Survey (NHANES) for cross-sectional evaluation, with rigorous matching of age, sex, and BMI. The objectives of this study are: (1) to comprehensively evaluate body composition and determine whether regional adipose and muscle distribution are independently associated with the risk of T2D; (2) to explore sex-specific differences in this association; and (3) to determine the mediating role of insulin resistance in the pathway linking body composition to diabetes.Subjects and methods Study design This research incorporated publicly available NHANES data for comprehensive analysis, an ongoing population-based surveillance program administered by the Centers for Disease Control and Prevention (CDC). 27 Employing a multistage probability cluster sampling strategy with stratification, NHANES systematically recruits nationally representative cohorts of non-institutionalized US residents. The survey operates through biennial cycles, combining in-home questionnaire administration with subsequent standardized physical evaluations conducted at specialized mobile assessment units to collect comprehensive health metrics. This open-access epidemiological resource annually incorporates approximately 5000 individuals into its continuously updated database, with all participants providing documented consent prior to inclusion. Complete survey protocols and publicly available datasets can be accessed through the official CDC portal.Study participants The study population comprised participants from all available NHANES cycles between 2011 and 2018 that included dual-energy X-ray absorptiometry (DXA) measurements. To enhance the validity and reliability, specific exclusion criteria were implemented. Individuals aged<18 or >60 years were excluded from the analysis because older adults have experienced age-related muscle loss and develop T2D at a relatively lower BMI. Participants with specific diseases (liver or kidney disease, stroke, angina pectoris, myocardial infarction, or cancer) and those receiving hormone-modulating medications (including testosterone, somatropin, adrenal cortical steroids, and thyroid hormones) were excluded, given the known effects of these conditions and agents on body composition and/or T2D risk. We also excluded participants if they had missing information on independent variables, dependent variables, or covariates. Following these exclusion criteria, our final analytical sample included 3567 eligible participants ( online supplemental figure 1).SP110.1136/bmjdrc-2025-005397.supp1Supplementary dataBody composition variables Body composition assessment in the study was conducted using whole-body DXA technology (Hologic, Bedford, Massachusetts, USA). 28 Certified radiologic technologists performed all scans following standardized protocols to ensure measurement precision. The detailed examination procedures are published in the official NHANES technical documentation. Upper limb lean mass and upper limb fat mass were considered as the total of the lean mass and fat mass of both the left arm and the right arm. Lower limb lean mass and lower limb fat mass were calculated as the sum of the lean mass and fat mass of both the left arm and the right leg. Appendicular lean muscle mass (ALM) was defined as the sum of lean mass from the limbs.29 To evaluate muscle mass, we employed the appendicular lean mass index (ALMI), derived by normalizing ALM (kg) to height (m²).30Diabetes variables All procedures conducted in the NHANES followed the standardized protocols established by the American Diabetes Association. Refer to the Laboratory Method Files section on the NHANES website for a detailed description of the laboratory methods used. Venous blood sampling for fasting glucose and serum insulin measurement was performed following a 9-hour fast, after which a 75 g oral glucose tolerance test (OGTT) was performed. Subsequently, insulin resistance was estimated through the Homeostasis Model Assessment of Insulin Resistance (HOMA-IR), which is calculated by the formula: HOMA-IR=(glucose in mmol/L × insulin in mIU/mL)/22.5.To capture a broad spectrum of metabolic risk, we defined our exposure group as individuals at high risk for diabetes, which includes those with either established diabetes or pre-diabetes. Participants were defined as high-risk T2D group if they met one or more of the following conditions: (1) fasting blood glucose≥6.1 mmol/L; (2) 2-hour blood glucose≥7.8 mmol/L during an OGTT; and (3) self-reported physician’s diagnosis of diabetes or pre-diabetes.Covariates The analysis incorporated multiple established covariates based on their documented associations with T2D risk: demographic data 31–33 (ethnicity, marital status, smoking status, education), anthropometric data34 35 (weight, waist circumference, height), and health-related factors31 36–38 (hypertension, triglyceride, low-density lipoprotein). Given substantial missing data, additional potential confounders were excluded from analytical models. All included covariates were presumed to influence the body composition-outcome association. This selective approach enhanced analytical validity by prioritizing variables with complete data.Statistical analyses This study used sampling weight adjustments recommended from NHANES for planned oversampling. All statistical analyses incorporated sample weighting procedures, accounting for the complex hierarchical, multistage, clustered sampling design. Propensity scores were estimated using multivariable logistic regression incorporating sex, age, and BMI as covariates, followed by 1:1 nearest-neighbor matching without replacement by using the MatchIt package to establish case-control pairs with balanced baseline characteristics. To assess the accuracy of the matching, absolute standardized differences≤10% are considered to denote negligible imbalances between the two groups.In descriptive statistics for the high-risk T2D group and healthy controls (HC) group, all continuous variables are represented as the weighted median and IQR, and categorical variables are represented as frequencies and weighted percentages. The distribution difference between the high-risk T2D group and the HC group was analyzed by a weighted Wilcoxon rank-sum test.The relationship between regional body composition and diabetes was explored using weighted univariate logistic regression and multivariate logistic regression (svyglm). Weighted linear regression was used to explore the relationship between regional body composition and insulin resistance indicators (HOMA-IR). The model adjusted for demographic data (ethnicity, marital status, smoking status, education), anthropometric data (weight, waist circumference, height), hypertension prevalence, plasma triglycerides, and plasma low-density lipoprotein. Restricted cubic splines with four knots were used to examine the non-linear relationship between lower limb fat mass and diabetes incidence.Following the establishment of these analyses, we conducted a mediation analysis to ascertain the extent to which insulin resistance mediates the relationship between regional body composition and diabetes. The proportion of the effect mediated by insulin resistance was computed using the formula (mediated effect/total effect) × 100%. The mediating role of insulin resistance (HOMA-IR) was analyzed using the mediation package. All statistical analyses were implemented using RStudio software (V.1.4.1717), with statistical significance defined as p<0.05 in bilateral testing frameworks.Results Characteristics of the participants The characteristics of 3567 eligible subjects before matching in our study are listed in online supplemental table 1). Patients with high T2D risk were older, had higher BMI, and body fat percentage than HC.The characteristics of 2092 eligible participants after matching by sex, age, and BMI in our study are listed in online supplemental table 2. As measured using BMI, both the high-risk T2D group and HC group had obesity on average. Fasting blood glucose, insulin, glycosylated hemoglobin, 2-hour glucose, and HOMA-IR were all significantly greater in the high-risk T2D group.Regional lean and fat mass After matching by sex, age, and BMI, lower limb lean mass was 0.26 kg lower in the high-risk T2D group (p =0.02, table 1, and figure 1), and lower limb fat mass was 1.00 kg lower in the high-risk T2D group compared with the HC group (p<0.001), while visceral fat mass tended to be 0.03 kg higher in the high-risk T2D group compared with the HC group (p<0.001). There were no other between-group differences in regional lean mass, ALM, and ALMI.Table 1Regional lean and fat massVariablesHC group (n=1046)T2D group (n=1046)W/FP valueLean mass Upper limb (kg)6.23 (4.57, 7.91)6.10 (4.49, 8.26)549 275.000.87 Lower limb (kg)16.81 (13.72, 19.66)16.55 (13.14, 20.64)579 902.000.02 Trunk (kg)25.90 (22.79, 30.30)26.17 (22.05, 30.46)554 836.000.57 ALM (kg)22.98 (18.30, 27.25)22.58 (17.86, 27.10)570 608.000.09 ALMI (kg/m2)8.06 (6.97, 9.20)8.08 (6.85, 9.29)553 419.000.65Fat mass Upper limb (kg)3.37 (2.61, 4.44)3.42 (2.58, 4.47)543 551.000.80 Lower limb (kg)9.77 (7.30, 13.51)8.77 (6.83, 11.91)608 283.00<0.001 Trunk (kg)13.59 (10.03, 17.46)13.86 (10.63, 17.81)520 622.000.06 Abdominal (kg)2.36 (1.61, 3.03)2.37 (1.71, 2.88)529 084.000.19 Subcutaneous (kg)1.73 (1.19, 2.37)1.67 (1.20, 2.19)554 132.000.61 Visceral (kg)0.53 (0.37, 0.73)0.56 (0.40, 0.80)452 750.00<0.001ALMI, appendicular lean mass index.ALM, appendicular lean mass; HC, healthy controls; T2D, type 2 diabetes; W/F, F-test or Wilcoxon rank-sum test.Figure 1(a) Regional lean mass in HC and T2D group. (b) Regional fat mass in HC and T2D group. ALM, appendicular lean muscle mass; ALMI, appendicular lean mass index; HC, healthy controls; T2D, type 2 diabetes.Association between body composition and diabetes prevalence Restricted cubic spline analysis revealed a significant non-linear, inverse association between lower limb fat mass and the odds of high T2D risk (p for non-linearity=0.0019). After adjustment for demographic data (ethnicity, marital status, smoking status, education), anthropometric data (weight, waist circumference, height), hypertension prevalence, plasma triglycerides, and plasma low-density lipoprotein, the relationship was consistently inverse, with higher levels of lower limb fat mass conferring a protective effect across most of its distribution ( online supplemental figure 2 and figure 2).Figure 2Non-linear relationship between lower limb fat mass and diabetes. LDL-C: Low-Density Lipoprotein Cholesterol.Subsequent subgroup analyses based on sex indicated that regional lean and fat mass distribution exhibited different trends in their association with diabetes prevalence in different sexes (online supplemental table 4 and figure 3). The protective effect of lower limb fat mass against diabetes remained significant in men (OR 0.86 (0.76–0.97), p=0.01) but did not reach significance in women. The odds of high T2D risk were 7.35 times greater in man (OR 7.35 (1.57–34.4), p=0.01) per per cent increase in visceral fat mass after controlling for all covariates. This relationship, however, was not statistically significant in the female participants. Additionally, the odds of high T2D risk were 1.50 times greater in woman (OR 1.50 (1.04–2.15), p=0.03) per per cent increase in terms of upper limb lean mass.Figure 3The odds of diabetes in subgroup analyses based on sex. ALM, appendicular lean muscle mass; ALMI, appendicular lean mass index.Association between regional body composition and insulin resistance Insulin resistance was estimated using HOMA-IR. Weighted linear regression analysis was employed to examine the associations between regional body composition and HOMA-IR. Model adjusted for all covariates revealed a significant association between higher lower limb fat mass and lower HOMA-IR (β=−0.11, p<0.001). Higher lower limb fat mass was associated with a lower risk of insulin resistance. Additionally, we also indicated a significant linear negative association between subcutaneous fat mass and HOMA-IR (β=−0.53, p=0.04). The complete results are presented in online supplemental table 5.The mediating role of insulin resistance between lower limb fat mass and diabetes Mediation analysis was further used to explore the role of insulin resistance represented by HOMA-IR in mediating the relationship between lower limb fat mass and diabetes. Insulin resistance had a significant indirect effect (mediation effect) with a mediation ratio of 36.18% ( figure 4).Figure 4The mediating role of insulin resistance (Homeostasis Model Assessment of Insulin Resistance) in the association between lower limb fat mass and diabetes. ACME, Average Causal Mediation Effect; ADE, Average Direct Effect.Discussion By meticulously matching participants for age, sex, and BMI, we minimized the confounding effects of these variables, thereby isolating the unique contribution of regional fat and muscle to diabetes pathogenesis. Our study provides novel insights into the sex-specific associations between body composition and diabetes risk.The negative association between lower limb fat mass and diabetes risk, particularly in men, aligns with emerging evidence.11 39 40 Unlike visceral fat, which releases proinflammatory cytokines and free fatty acids that impair insulin signaling, lower body adiposity possibly acts as a ‘metabolic sink’, sequestering excess lipids and reducing ectopic fat deposition in organs such as the liver and pancreas.12 Our mediation analysis further supports this hypothesis, demonstrating that 36.18% of the protective effect of lower limb fat mass on diabetes is mediated by improved insulin sensitivity (HOMA-IR).Our subgroup sex-specific results showed the robust protective effect of lower limb fat mass against diabetes in men and its non-significance in women, coupled with the male-specific detrimental impact of visceral adiposity. This may be attributed to the significant effects of sex hormones on fat mass and distribution. Estrogen promotes the storage of subcutaneous adiposity instead of visceral adiposity, primarily in the gluteal and femoral regions, resulting in greater lower-body fat accumulation in women compared with men.41 Androgens promote VAT accumulation, predisposing to central obesity and impairment of insulin sensitivity.42 43 VAT in men exhibits heightened secretion of proinflammatory cytokines (interleukin 6, tumor necrosis factor-α), while the ‘metabolic buffering’ capacity of lower-body fat assumes greater physiological significance in male individuals.44 45To our knowledge, this study provides the first evidence that lower-body fat influences diabetes risk through the mediating effect of insulin resistance in BMI-matched individuals. Current reliance on BMI obscures critical variations in fat distribution, lower-body fat distribution, particularly for men with normal BMI but adverse fat distribution (higher visceral adiposity combined with lower limb fat mass), may serve as a novel protective marker against T2D, partially mediated by ameliorating insulin resistance. For men, interventions specifically targeting visceral fat reduction—such as high-intensity interval training, resistance exercise, and dietary modifications focused on reducing refined carbohydrates—may yield greater metabolic benefits than general weight loss approaches alone. Clinical interventions targeting fat redistribution, particularly those combining aerobic exercise, resistance training, and nutritional strategies, warrant further exploration in sex-specific diabetes prevention strategies. While challenging, such targeted approaches are becoming increasingly feasible through personalized exercise prescriptions and nutritional interventions that account for sex-specific metabolic responses.Our investigation did not find a significant association between muscle mass and T2D risk, challenging the conventional view of muscle mass as uniformly protective. It may suggest that assessing muscle mass in combination with functional parameters (such as handgrip strength) or fat infiltration provides a more comprehensive evaluation of the muscle’s impact on glucose metabolism than using individual metrics alone.46–48 In addition, participants in our study are younger than 60 years of age, who have not totally experienced age-related muscle mass declines. Furthermore, myosteatosis (pathological fat infiltration within muscle tissue) may counteract the beneficial effects of muscle mass, necessitating assessment via CT/MRI imaging modalities to quantify muscle fat infiltration.49 50Moreover, several limitations warrant acknowledgment. Our sample’s demographic scope may limit generalizability, and unmeasured confounders like dietary patterns could influence associations. In addition, we prioritized reporting unadjusted p values in this exploratory analysis to avoid overcorrection and a potential loss of statistical power for identifying novel associations. The findings regarding the sex-specific protective role of lower limb fat require confirmation in future confirmatory studies. Future research should combine advanced imaging (CT, DXA/MRI) with omics profiling to unravel molecular pathways linking regional adiposity to insulin sensitivity.Conclusion In conclusion, this study delineates a protective role for lower-body fat in diabetes pathogenesis, mediated substantially through ameliorating insulin resistance. The sex-specific associations underscore the protective effect of lower-body fat and the detrimental impact of visceral adiposity in men after controlling for BMI. Further work is needed to confirm whether lower limb fat itself is a causal factor in the development of T2D through mechanisms such as enhanced lipid buffering capacity, secretion of beneficial adipokines, or reduced systemic inflammation, and to uncover the mechanisms underpinning these sex-specific associations.",
  "title": "Sex-specific protective role of lower-body fat in type 2 diabetes: mediation through insulin resistance in a BMI-matched population",
  "uid": "2965ce44-2efc-599e-98d4-d7a2fe493bca"
}