Appendicular Skeletal Muscle Mass-to-Visceral Fat Area Ratio and Bone Mineral Density in Type 2 Diabetes: A Cross-Sectional Study

Summary

This cross-sectional study reports covariate-conditional associations between the appendicular skeletal muscle mass-to-visceral fat area ratio and bone mineral density in adults with type 2 diabetes.

Abstract

Bone fragility is common in type 2 diabetes mellitus (T2DM), but body mass index alone does not capture the balance between appendicular muscle and visceral adiposity. This retrospective cross-sectional study examined the association between the appendicular skeletal muscle mass-to-visceral fat area ratio (SVR) and bone mineral density (BMD) in hospitalized adults with T2DM. A total of 2,820 records were included in descriptive analyses, and 2,798 records with complete BMD endpoints and Model 5 covariates were included in regression analyses. Appendicular skeletal muscle mass and visceral fat area were obtained using multifrequency bioelectrical impedance analysis, whereas lumbar spine, femoral neck, and total hip BMD were measured by dual-energy X-ray absorptiometry. In fully adjusted linear models, each 1-SD increase in SVR was associated with higher lumbar spine BMD (0.017 g/cm²; 95% CI, 0.007–0.027; P = 0.001), femoral neck BMD (0.027 g/cm²; 95% CI, 0.019–0.035; P < 0.001), and total hip BMD (0.028 g/cm²; 95% CI, 0.020–0.036; P < 0.001). Sex-specific tertile plots did not reproduce the monotonic pattern seen with global tertiles, which were strongly sex-imbalanced. Natural cubic-spline models detected nonlinearity at all three sites (all P for nonlinearity < 0.001), with increasing predicted BMD that flattened at higher SVR values. HC3-robust confidence intervals were similar to the primary estimates. These findings describe cross-sectional conditional associations and do not establish causality, bone strength, fracture prediction, or a clinical SVR threshold.

Introduction

Osteoporosis is characterized by low bone mass, deterioration of bone microarchitecture, and increased susceptibility to fragility fracture^1,2. Fracture risk is influenced by age, sex, inherited susceptibility, body composition, and metabolic disease³. In type 2 diabetes mellitus (T2DM), areal bone mineral density (BMD) may be preserved or elevated despite an increased risk of fracture; therefore, BMD in this population should be interpreted as a skeletal phenotype rather than as a direct measure of bone strength or fracture probability⁴.

Muscle, adipose tissue, and bone form an interdependent musculoskeletal-metabolic system. Loss of skeletal muscle may reduce mechanical loading and alter myokine signaling, whereas visceral adiposity may contribute to inflammatory and adipokine pathways relevant to bone remodeling^5,6. Because body mass index (BMI) and waist measurements do not separately quantify appendicular muscle and visceral fat, a composite body-composition ratio may provide complementary information while remaining a statistical exposure rather than a distinct biological mechanism^7,8.

Muscle–bone crosstalk provides additional biological context for evaluating composite body-composition measures in relation to BMD⁹. Previous studies have directly examined SVR in relation to BMD. Liu et al. evaluated the association of appendicular skeletal muscle mass-to-visceral fat area ratio (SVR) with BMD and osteoporosis in a general adult population, whereas Guo et al. studied 422 Chinese patients with T2DM and reported age- and sex-dependent associations with BMD and estimated 10-year fracture probability^10,11.

The novelty of this study lies in its comprehensive evaluation of the SVR in a large cohort of adults with T2DM. By examining bone mineral density at three clinically relevant skeletal sites and incorporating sequential covariate adjustment, sex-standardized and sex-stratified analyses, simultaneous assessment of the muscle and visceral-fat components, nonlinear spline modeling, and robust sensitivity analyses, this study provides a more rigorous and clinically contextualized characterization of the association between body-composition balance and skeletal health in T2DM. The present study extends this literature using a larger single-center endocrinology inpatient dataset (2,820 records), three site-specific BMD outcomes, sequential covariate adjustment, sex-standardized and sex-stratified analyses, simultaneous appendicular skeletal muscle mass (ASM)/visceral fat area (VFA) component models, and prespecified subgroup interaction tests. The authors are not aware of overlap between the present cohort and the datasets reported in those studies. We hypothesized that higher SVR would be positively associated with lumbar spine, femoral neck, and total hip BMD.

Protocol

This retrospective study was approved by the Ethics Review Board of Changzhou Second People's Hospital (Approval No. KY017-01) and conducted in accordance with the Declaration of Helsinki. The requirement for written informed consent was waived because the analysis used anonymized clinical records. The study included records from 2017 through 2022.

Study design and records
This retrospective cross-sectional analysis used inpatient records from the Department of Endocrinology, The Second People's Hospital of Changzhou, The Third Affiliated Hospital of Nanjing Medical University. The screening file contained 3,495 records after preliminary cleaning, and the analysis unit was an inpatient record. T2DM was defined by the documented inpatient diagnosis in the electronic medical record.

Records were excluded for age younger than 18 years or older than 80 years; alanine aminotransferase or aspartate aminotransferase of at least 120 U/L; eGFR below 30 mL/min/1.73 m²; hemoglobin below 90 g/L; thyroid-stimulating hormone below 0.3 µIU/L or at least 10 µIU/L; or documented active chronic infection, inflammatory disease, malignancy, extreme physical immobility, or another condition expected to substantially distort body-composition or BMD assessment. A total of 366 records were removed because key laboratory or body-composition information was incomplete, and 309 were removed by the prespecified clinical or data-quality criteria, leaving 2,820 records for the descriptive analyses. Twenty-two records lacked all three BMD endpoints and were excluded from regression, leaving 2,798 complete records for each skeletal-site model.

Structured variables were exported from the electronic medical record and curated. Variable definitions and data-quality summaries were independently reviewed, and any discrepancies were resolved by consensus before statistical analysis.

Clinical and laboratory measurements
Age, sex, diabetes duration, smoking status, alcohol consumption, blood pressure, and anthropometric measurements were extracted from electronic medical records. Body weight and height were measured with calibrated equipment while participants wore light clothing and no shoes. BMI was calculated as weight in kilograms divided by height in meters squared. Waist circumference was measured at the umbilical level, hip circumference at the maximal gluteal protrusion, and waist-to-hip ratio as waist circumference divided by hip circumference. Resting blood pressure was obtained after 10 min of seated rest using an automated sphygmomanometer.

All clinical, laboratory, body-composition, and BMD measurements were obtained during the same index hospitalization. Venous blood was collected after an overnight fast within 24 h of admission.

Blood assays
Venous blood samples were collected after an overnight fast within 24 h of admission. Hemoglobin was measured using an automated hematology analyzer. Alanine aminotransferase, aspartate aminotransferase, gamma-glutamyl transferase, serum uric acid, creatinine, triglycerides, total cholesterol, high-density lipoprotein cholesterol, and low-density lipoprotein cholesterol were measured using an automated chemistry analyzer. Glycated hemoglobin (HbA1c) was measured by high-performance liquid chromatography. Thyroid-stimulating hormone, free triiodothyronine, and free thyroxine were quantified using chemiluminescent immunoassay. Estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) 2021 creatinine equation.

Bone mineral density and body composition
Areal bone mineral density was measured at the lumbar spine, femoral neck, and left total hip using a dual-energy X-ray absorptiometry system and was expressed in g/cm². Standard anteroposterior lumbar spine and proximal femur acquisition protocols were used. Before scanning, participants removed metallic objects and were positioned supine by trained radiologic technologists. For lumbar spine acquisition, the lower legs were supported to reduce lumbar lordosis, and the lumbar vertebrae were included within the predefined analysis region. For proximal femur acquisition, the left leg was internally rotated and stabilized using the manufacturer-provided positioning device to obtain standardized femoral-neck and total-hip regions.

All scans were reviewed by trained operators for patient motion, metallic interference, incomplete acquisition, anatomical positioning, and region-of-interest placement. Scans with substantial motion, positioning errors, or artifacts that could invalidate the BMD estimate were repeated when clinically feasible or excluded from the analytical dataset. Daily phantom calibration and routine quality control checks were performed before clinical scanning in accordance with the manufacturer's instructions and the hospital's standard operating procedures. Images were analyzed using the manufacturer-provided analysis software.

Body composition was assessed using a multi-frequency bioelectrical impedance analyzer. The device uses direct segmental multifrequency bioelectrical impedance analysis with an eight-point tactile-electrode system and impedance measurements at 1, 5, 50, 250, 500, and 1,000 kHz. Measurements were performed in the morning after an overnight fast and after voiding, before breakfast and before strenuous physical activity. Participants were measured barefoot and wearing light clothing. The palms and soles were cleaned before measurement to ensure adequate electrode contact. Participants stood upright with their heels correctly positioned on the foot electrodes, held the hand electrodes with their thumbs placed on the designated contacts, maintained their arms extended away from the trunk, and remained still and silent throughout the measurement.

Appendicular skeletal muscle mass and visceral fat area were obtained from the device-generated body-composition report. Visceral fat area was a proprietary algorithm-derived estimate based on the bioelectrical impedance measurements and participant characteristics; it was not directly quantified by computed tomography, magnetic resonance imaging, or dual-energy X-ray absorptiometry. SVR was calculated as appendicular skeletal muscle mass (kg) divided by bioelectrical impedance analysis (BIA)- estimated VFA (cm²), with units of kg/cm². In the 2,820-record descriptive cohort, raw SVR was 0.207 ± 0.090 kg/cm² (median, 0.195; range, 0.063–0.766), and the global tertile cut points were 0.1518 and 0.2361 kg/cm². VFA ranged from 18.9 to 270.6 cm²; no zero or negative VFA values were present. No prespecified threshold for very-low-VFA, winsorization rule, or additional extreme-ratio exclusion criterion was applied; all finite values were analyzed. SVR was interpreted as a composite ratio exposure reflecting muscle relative to visceral adiposity, not as an independent biological mechanism. Cohort-standardized SVR was calculated as (SVR − cohort mean)/cohort SD.

Statistical analysis
Continuous variables were summarized as mean ± standard deviation (SD) or median (interquartile range), and categorical variables as n (%). Across SVR tertiles, approximately normal continuous variables were compared by analysis of variance, skewed variables by the Kruskal-Wallis test, and categorical variables by the chi-square test. The primary linear model for each BMD endpoint was: BMD = β₀ + β₁(SVR z score) + β₂(age) + β₃(sex) + β₄(BMI) + β₅(waist-to-hip ratio) + β₆(diabetes duration) + β₇(hemoglobin) + β₈(eGFR) + β₉(alanine aminotransferase [ALT]) + β₁₀(aspartate aminotransferase [AST]) + ε. Sequential Models 1-5 retained the definitions reported in Table 2. Age and sex were treated as demographic confounders. BMI and waist-to-hip ratio were included as body-size and central-adiposity covariates but may partly lie on the pathway linking body composition with BMD. Diabetes duration, hemoglobin, eGFR, ALT, and AST were included as clinical-status covariates and may also be downstream correlates. Accordingly, Model 5 estimates a covariate-conditional association rather than a total causal effect.

A VIF below 5 was considered acceptable. Because the three primary models used the same 2,798 records and the same design matrix, the VIFs were identical across BMD endpoints: diabetes duration 1.25, hemoglobin 1.57, eGFR 1.64, BMI 2.01, age 2.13, waist-to-hip ratio 2.50, sex 2.51, AST 2.80, ALT 3.05, and standardized SVR 3.48.

Subgroup models used the Model 5 covariate set after removing the stratifying variable: sex was removed from sex-stratified models, age from age-stratified models, BMI from BMI-stratified models, diabetes duration from duration-stratified models, and eGFR from renal-function-stratified models. Glycemic-control category was not a Model 5 covariate, so the Model 5 covariates were retained in that analysis. Sex-stratified models and the sex interaction used sex-specific SVR z scores; all other subgroup analyses used the cohort-standardized SVR. Each interaction model added subgroup indicators and SVR-by-subgroup product terms and was compared with the corresponding main-effects model by a nested F test. Benjamini-Hochberg correction was applied to the complete family of 18 interaction tests (three BMD endpoints × six subgroup factors).

No a priori power calculation was performed because all eligible records from the fixed retrospective period were analyzed. Precision was evaluated using two-sided 95% confidence intervals. Linear-model assumptions were assessed using residual-versus-fitted, quantile–quantile (Q–Q) plots, scale-location, Cook’s distance, Breusch–Pagan, and studentized-residual diagnostics. HC3 heteroscedasticity-consistent confidence intervals were calculated as a sensitivity analysis. Nonlinearity was tested by comparing Model 5 with a natural cubic-spline exposure term 3 degrees of freedom [df]) using a partial F test; 4-df and central 2^nd–98^th percentile analyses assessed shape sensitivity. Population-average predictions were displayed over the 2^nd–98^th percentiles of raw SVR.

Results

Record selection and analytic samples
Among 3,495 inpatient T2DM records available after preliminary cleaning, 366 were excluded because key laboratory or body-composition fields were incomplete and 309 were excluded based on prespecified clinical or data quality criteria, leaving 2,820 records for the descriptive analyses. Within this cohort, 22 records lacked all three BMD endpoints and were excluded from regression. The same 2,798 records contributed to each skeletal-site Model 5 analysis. Missi...

Discussion

In this retrospective cross-sectional analysis, higher SVR was associated with higher areal BMD at the lumbar spine, femoral neck, and total hip after covariate adjustment. The apparent global-tertile pattern coincided with marked sex imbalance and was not reproduced in sex-specific unadjusted plots, underscoring the importance of continuous adjusted analyses. Spline analyses suggested increasing associations that flattened at higher SVR values, and HC3-robust estimates were similar to the primary results. These estimate...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Automated chemistry analyzer	Siemens	ADVIA 2400	Biochemical assays
Automated hematology analyzer	Sysmex	XN-2800	Hemoglobin and complete blood count measurements
Automated sphygmomanometer	Omron	U703	Resting blood pressure measurement
Chemiluminescent immunoassay analyzer	Siemens	ADVIA Centaur XPT	Thyroid hormone measurements
Dual-energy X-ray absorptiometry scanner	Hologic	EXPLORER	Lumbar spine, femoral neck, and total hip BMD measurements
Height and weight scale	Hengqi Inc.	RGZ-120-RT	Anthropometric measurements
High-performance liquid chromatography analyzer	TOSOH	G8-90SL	HbA1c measurement
Multifrequency bioelectrical impedance analyzer	Biospace	InBody 720	Appendicular skeletal muscle mass and visceral fat area measurements
R statistical software	R Foundation for Statistical Computing	Version 4.6.0	Statistical analysis and figure generation