Accumulation in Visceral Adipose Tissue Over 6 Years Is Associated With Lower Paraspinal Muscle Density

Abstract

Context

Visceral adipose tissue (VAT) has been recognized to be a metabolically active fat depot that may have paracrine effects on surrounding tissues, including muscle. Since many adults accumulate VAT as they age, the effect of changes in VAT on muscle is of interest.

Objective

We determined the association between 6-year changes in VAT and paraspinal muscle density, an indicator of fatty infiltration.

Methods

This study included 1145 participants from the Framingham Study third-generation cohort who had both quantitative computed tomography scans of the spine at baseline and 6-year's follow-up, on whom muscle density was measured along with VAT. We implemented multiple regression to determine the association of muscle density at follow-up as primary outcome measure with changes in VAT (follow-up minus baseline divided by 100), adjusting for VAT at baseline, age, sex, height, menopausal status, presence of diabetes, and physical activity. Analyses were performed in men and women separately.

Results

After adjustment for covariates, individuals with the greatest accumulation of VAT over 6 years had significantly lower paraspinal density at the follow-up with an estimated 0.302 (95% CI, −0.380 to −0.224) and 0.476 (95% CI: −0.598 to −0.354) lower muscle density (HU) per 100-cm³ increase in VAT (both P values < .001) in men and women, respectively.

Conclusion

These results highlight that age-related accumulation of VAT in men and women is associated with lower muscle density. VAT may represent a modifiable risk factor for poor musculoskeletal outcomes with aging.

visceral adipose tissue, muscle density, sarcopenic obesity, musculoskeletal

The epidemic proportions of the obesity problem in the United States and worldwide will lead to an increase in the incidence of a variety of chronic diseases and premature mortality (1, 2). Most obesity research has focused on metabolic and cardiovascular outcomes such as diabetes, hyperlipidemia, hypertension, coronary heart disease, and osteoarthritis. There is considerably less consensus on the role of obesity on the risk for low muscle mass or muscle density. Sarcopenic obesity denotes a condition characterized by excess body adiposity in an individual with reduced muscle mass, and is associated with an excess incidence of mobility impairments and even injuries (3, 4). Equally important is the observation that greater fat mass is associated with declines in muscle strength per unit of lean mass (5) and that muscle density (a reflection of fatty infiltration) predicts fracture independent of bone mineral density (6), and is associated with greater body sway when considering the paraspinal muscles (7). Most studies investigating the associations between adiposity and the musculoskeletal system use body mass index (BMI) (8), which does not specifically address the role of visceral adiposity. Recent trials of weight loss in individuals with obesity have shown that the weight loss is accompanied by dramatic declines in bone density and muscle mass (9, 10). Thus, the links between adiposity, bone, and muscle are extremely important. Since visceral adipose tissue (VAT) is metabolically active, secretes inflammatory mediators (11) and adipokines (12), and serves to metabolize steroid hormones (13) that affect muscle health, we hypothesized that visceral adiposity would be associated with deleterious effects on paraspinal muscle density, a measure of fatty infiltration into muscle. Thus, we determined the association between 6-year changes in VAT and paraspinal muscle density in a community-based cohort study.

Materials and Methods

Study Design and Participants

The study sample for this work consisted of the Framingham Heart Study third-generation cohort (Gen3) who underwent a multidetector computed tomography (CT) scan in 2002 to 2005 and repeated in 2008 to 2011. Between the years 2002 to 2005, the Gen3 Study enrolled 4095 adult children (aged 19-72 years) of participants in the Framingham Offspring Study (14, 15). The Framingham Offspring Study itself was initiated by enrolling 5124 adult children of the original Framingham Study cohort and their spouses (14, 16). The first and second visits of the Gen3 Cohort formed the basis for our proposed study. Scientifically rigorous methods have been used to conduct clinical examinations and collect data in the Gen3 studies. The Framingham Heart Study multidetector computed tomography study recruited 2117 Gen3 participants to obtain trunk CT scans for vascular calcification at the baseline CT visit in 2002 to 2005 (17). Of these participants, 1421 underwent the same CT scan procedure at follow-up visit (∼ 6 years later) in 2008 to 2011. Of those, 1145 participants had validated VAT and trunk muscle density measurements at both examination visits, which comprised the final study sample. The study was approved by Boston University Medical Campus and the Hebrew SeniorLife institutional review boards. All participants in this study had provided written informed consent.

Visceral Fat Measurements

Participants underwent volumetric CT scan of the spine at baseline and the follow-up visits, as previously described (18-20). At baseline, scans were conducted using 8-slice multidetector tomography (Lightspeed Ultra; General Electric Medical Systems) and were acquired through a series of contiguous 5-mm-thick slices from the upper edge of the S1-vertebrate and 125 mm superiorly. Scans at the follow-up visit used a Discovery VCT 64-slice positron emission tomography/CT scanner (GE Healthcare). They acquired 30 contiguous 5-mm-thick slices (120 kVP; 100-300 mA, depending on BMI) beginning 2 cm above the S1 vertebra. VAT volumes were assessed manually to separate VAT and subcutaneous adipose tissue (SAT) with the abdominal muscular wall at baseline and follow-up scans (6 years later) using the same analysis protocol (Aquarius 3D Workstation, TeraRecon Inc) (20). The fat in Hounsfield units (HUs) was identified between −195 and −45 (18). High correlations were noted for interreader comparisons as well intrareader comparisons (0.99) (18).

Paraspinal Muscle Density

We measured density of the paraspinal muscles using the same scans as the VAT, as described previously (21). The density of the trunk muscles was measured using a customized commercial image processing platform (Analyze, Biomedical Imaging Resource). Briefly, each muscle was contoured using a semiautomated approach at the midvertebral slice. Muscle attenuation was calculated as the mean voxel attenuation in HUs, averaging the right and left sides. Attenuation values were standardized based on a hydroxyapatite phantom scanned with each patient. We derived muscle area-weighted mean density of 4 bilateral paraspinal muscles in a mid-L3 vertebral slice (erector spinae, transversospinalis, psoas major, and quadratus lumborum) on scans from a visit occurring between 2008 and 2011 (Fig. 1). In this study, we considered the muscle area-weighted average muscle density across the 4 muscles combined at the follow-up visit in 2008 to 2011.

Figure 1.

Computed tomography scan. We derived muscle area-weighted mean density of 4 bilateral paraspinal muscles in a mid-L3 vertebral slice, erector spinae (ES), transversospinalis (TS), psoas major (PM), and quadratus lumborum (QL).

Covariates

We obtained covariate information from the Gen3's first research visit (2002-2005). Covariates we considered included age (years), body height (inches), physical activity (Physical Activity Score for Elderly [PASE]) (22), diabetes status, and in women, menopausal status and estrogen use. Body height was measured without shoes using a stadiometer to the nearest quarter inch (25.4 mm). Menopause in women was determined if periods stopped for at least 12 months. Estrogenic status in women was defined as “yes” for premenopausal women or women currently taking estrogen and “no” for postmenopausal women without taking estrogen. Diabetes status was defined as affirmative for those with fasting glucose greater than or equal to 7 mmol/L, nonfasting blood glucose greater than or equal to 11 mmol/L, or the use of glucose-lowering medications.

Adiponectin circulates in plasma as 3 oligomeric isoforms. High-molecular-weight adiponectin is thought to be the most biologically active form of adiponectin in terms of glucose homeostasis. In the Gen3 cohort, adiponectin was measured at the time of the baseline VAT and muscle density measures using the Quantikine Adiponectin Immunoassay (Antibody ID: AB_829250) from ALPCO, a 4.5-hour solid-phase enzyme-linked immunosorbent assay (ELISA) designed to measure total (low, middle, and high-molecular-weight) human adiponectin. The minimum detectable dose (MDD) of adiponectin ranges from 0.079 to 0.891 ng/mL. The mean MDD is 0.246 ng/mL. Intra-assay precision for the assay ranges from a coefficient of variation percentage (CV%) of 2.5 to 4.7, while interassay precisions are CV% of 5.8 to 6.9. Interleukin-6 (IL-6) was measured using the high-sensitivity, solid-phase ELISA Quantikine HS Human IL-6 Immunoassay from R&D Systems (Antibody ID: AB_2893335). Intra-assay precision for the assay ranges from a CV% of 6.9 to 7.4 while interassay precisions range from CV% of 6.5 to 9.6. The MDD of IL-6 ranges from 0.016 to 0.110 pg/mL. The mean MDD is 0.039 pg/mL.

Statistical Analysis

We conducted statistical analyses using statistical software R 4.0.2. We calculated visceral fat change by subtracting the visceral fat volume at baseline (2002-2005) from the visceral fat volume at follow-up (2008-2011), then scaled by 100. To quantify the association between trunk muscle density and visceral fat change, we performed multiple linear regression analysis with trunk muscle density at follow-up as the dependent variable and visceral fat change as the primary independent variable. Sex-stratified analysis was our primary analysis. We considered 3 models. The first model adjusted for VAT at baseline, age, height, and estrogenic status (for women), the second additionally adjusted for diabetes status, and the third further adjusted for physical activity. We carried out the same analysis as model 3, additionally adjusted for the adiponectin (model 4), interleukin-6 (model 5), or both (model 6), as a sensitive analysis to determine if the strength of the association was attenuated by the addition of these potential factors that might represent mediators of the VAT and muscle density associations. As a secondary analysis, we analyzed women stratified by menopause status. We calculated the change of trunk muscle density by subtracting trunk muscle density at baseline from trunk muscle density at follow-up. We then substituted trunk muscle density at follow-up with the change in trunk muscle density as the outcome variable, and conducted an exploratory analysis with the same strategy described earlier.

Results

Characteristics of Study Samples

Table 1 presents individuals’ sex-specific clinical characteristics. In total, there were 1145 participants, including 635 (55%) men and 510 (45%) women who had both muscle density and VAT measurements required in this study (Fig. 2). At the baseline research visit, the mean ages of participants were 44.1 ± 6.3 years and 46.9 ± 5.9 years for men and women, respectively. A total of 2% and 2% of men and women participants had diabetes, and 33% and 30% of men and women participants had obesity with BMI greater than or equal to 30. The changes in VAT from baseline to follow-up were 961.1 ± 800.3 cm³ for men and 382.5 ± 541.2 cm³ for women. Our outcome variable, the paraspinal muscle density (HU) at follow-up, on average was 44.9 ± 8.7 and 39.4 ± 9.9 for men and women, respectively. Thirty percent of women were postmenopausal, and among them 35% took estrogen supplements (Supplementary Table S1 (23)).

Figure 2.

Sample size flowchart.

Table 1.

Sex-specific characteristics for study participants^a

	Men (N = 635)	Women (N = 510)
Age at baseline, mean (SD), y	44.1 (6.3)	46.9 (5.9)
BMI at baseline, mean (SD)	27.7 (4.1)	26.6 (5.8)
Obesity (BMI ≥ 30) at baseline, N (%)	211 (33)	153 (30)
Height at baseline, mean (SD), cm	177.6 (6.6)	164.2 (6.4)
Physical Activity Index at baseline, mean (SD)	38.1 (8.9)	36.4 (6.3)
Weight at baseline, mean (SD), kg	87.5 (14.5)	71.9 (16.4)
Diabetes status at baseline, N (%)	15 (2)	11 (2)
Estrogenic at baseline, N (%)	NA	411 (81)
Estrogen replacement therapy at baseline, N (%)	NA	56 (11)
Underwent menopausal transition, N (%)	NA	178 (35)
Adiponectin at baseline, mean (SD)a, ng/mL	6.2 (4.0)	11.5 (5.8)
Interleukin-6 at baseline, mean (SD)a, pg/mL	1.8 (2.0)	2.1 (4.8)
VAT at baseline, mean (SD), cm3	1915.3 (836.7)	1157.9 (732.2)
VAT at follow-up, mean (SD), cm3	2876.4 (1347.9)	1540.4 (1044.3)
Difference in VAT: follow-up—baseline, mean (SD), cm3	961.1 (800.3)	382.5 (541.2)
Paraspinal muscle density (HU) at baseline, mean (SD)	41.6 (9.8)	36.8 (9.6)
Paraspinal muscle density (HU) at follow-up, mean (SD)	44.9 (8.7)	39.4 (9.9)
Difference in paraspinal muscle density (HU), mean (SD)	3.2 (6.7)	2.6 (6.8)

The numbers in the cell are the mean (SD) for quantitative variables and count (%) for qualitative variable.

Abbreviations: BMI, body mass index; HU, Hounsfield unit; NA, not applicable; VAT, visceral adipose tissue.

Measured approximately 6 years before initial VAT measurement.

Association of Difference in Visceral Adipose Tissue With Paraspinal Muscle Density

Table 2 presents sex-stratified association results adjusted for VAT at baseline, age, height, estrogenic status, diabetes status, and physical activity, that is, model 3. Only results from model 3 are presented here because we observed similar associations across all models. Both in men and women, we observed that greater increases in VAT were associated with lower paraspinal muscle density. Specifically, every 100-cm³ increase in the change in VAT was significantly associated with 0.302 HU (95% CI, −0.380 ∼ −0.224) and 0.476 HU (95% CI, −0.598 ∼ −0.354) lower paraspinal muscle density in men and women, respectively (both P values < .001). We also observed the same direction of association results in women stratified by menopausal status from model 3 with association coefficients of −0.515 and −0.348 in premenopausal and menopausal women, respectively (Supplementary Table S2 (23)). The results of associations for various subsamples across all 6 models are shown in Supplementary Tables S3 and S4 (23). In the exploratory analysis where we used the change of trunk muscle density as the outcome variable, we observed a statistically significant association between VAT and muscle density in women, but not in men (Supplementary Table S5 (23)). For most individuals, as VAT increased over time, muscle density decreased. Additionally, we observed a statistically significant association between the change of trunk muscle density and VAT change in premenopausal women, but not in postmenopausal women (Supplementary Table S6 (23)). We provide sex-stratified scatterplots to demonstrate the association between VAT change and muscle density (Fig. 3) and change in muscle density (Supplementary Fig. S1 (23)).

Figure 3.

Relationship between 6-year changes in visceral adipose tissue (VAT) and paraspinal muscle density (HU) at follow-up in men and women. The dashed line represents the association in women, and the dash-dotted line represents the association in men. Triangle and cycle dots represent data points for men and women sample, respectively. The negative slopes depict inverse associations between changes in VAT and muscle density; on average, those experiencing lesser increases (or, in limited cases, decreases) in VAT are more likely to exhibit increases in muscle density with time, and there is limited evidence that this relationship is more pronounced in women than in men.

Table 2.

Association results with paraspinal muscle density in all samples stratified by sex for model 3

	Men					Women
	β	SE	95% CI	P		β	SE	95% CI	P
VAT difference, scaled by 100 cm3	−0.302	0.040	−0.380 to –0.224	< .001		−0.476	0.062	−0.598 to −0.354	< .001
VAT at baseline, scaled by 100 cm3	−0.222	0.039	−0.298 to −0.146	< .001		−0.644	0.048	−0.739 to −0.550	< .001
Age at baseline, y	−0.356	0.049	−0.452 to −0.259	< .001		−0.420	0.062	−0.543 to −0.298	< .001
Height at baseline, cm	−0.021	0.045	−0.110 to 0.067	.632		0.028	0.050	−0.071 to 0.127	.58
Diabetes status at baseline	−2.003	1.952	−5.836 to 1.830	.305		−3.307	2.215	−7.659 to 1.046	.136
Physical Activity Index at baseline	−0.075	0.033	−0.141 to −0.010	.024		0.004	0.051	−0.096 to 0.103	.945
Estrogenic status at baseline						−0.583	0.911	−2.372 to 1.206	.522

Abbreviation: VAT, visceral adipose tissue.

Discussion

To our knowledge, this is the first large, longitudinal study of the association between changes in VAT and a radiologic measure of muscle density, an indicator of fatty infiltration of muscle. Many adults accumulate VAT as they age and hence the association of changes in VAT with muscle density is of considerable interest. In this study, we conducted association analyses to investigate the relationship between 6-year changes in VAT, and paraspinal muscle density in a community-based study. After adjustment for covariates, men and women with the greatest accumulation of VAT over 6 years had statistically significantly lower paraspinal density at follow-up. Even though the distributions of VAT and muscle density were different between men and women, the same inverse association was observed in men and in women, and in premenopausal and postmenopausal women as well.

The use of CT scans as a method for measuring VAT and muscle density has been documented as valid, reliable, and accurate (24). CT imaging measures the density of the abdominal adipose VAT and SAT compartments by radiographic pixels that are denoted in HUs (25), identifying those between −195 and −45 HU with center attenuation of −120 HU. The SAT and VAT measurements are highly reproducible with interreader and intrareader reliability both greater than 0.99 (17, 25). The measurement of abdominal CT adipose tissue deposition is highly reproducible and as it is noninvasive and less expensive than magnetic resonance imaging scans, is useful in a large, community-based sample. In addition, using quantitative CT scans to assess muscle volume and density is an established technique (26).

Previous studies investigating the association between visceral adiposity and muscle density have been cross-sectional. In the Multi-Ethnic Study of Atherosclerosis (MESA), a cross-sectional analysis of inflammatory markers (IL-6, resistin, and C-reactive protein) found that the deleterious association between higher inflammation and lower paraspinal muscle density was either attenuated or became insignificant after adjusting for visceral adiposity, implicating visceral adiposity as a potential mediator (27). In contrast to our longitudinal assessment of visceral adiposity over 6 years, MESA had a single measure of adiposity and muscle density and did not specifically test for associations between VAT and muscle density.

The Nishimura Health Survey, an ongoing cohort investigation of risk factors for chronic diseases including hypertension, metabolic syndrome, and diabetes mellitus, as well as atrial fibrillation, studied the association between diabetes and low attenuation muscle in 621 Japanese men and women in their early 50s. Over 3 years of follow-up, those individuals who developed diabetes had greater visceral fat area and lower attenuation paraspinal muscle than did those individuals who did not develop diabetes (28). There were no direct investigations of the VAT and muscle density association. Nevertheless, the greater visceral fat that accompanies type 2 diabetes mellitus is accompanied by lower density skeletal muscle.

The mechanisms underlying the deleterious associations between VAT and muscle have not been clearly established in human studies. In the Framingham Offspring Cohort, VAT has been shown to be negatively correlated with adiponectin, an anti-inflammatory adipokine. Thus lower adiponectin concentrations in the face of visceral adiposity may lead to increased inflammation with subsequent fatty infiltration of muscle (29). Furthermore, skeletal muscle is an important peripheral target tissue for adiponectin (30), and muscle has 2 adiponectin receptors. When these receptors are knocked out in mice, adiponectin binding does not occur and muscle accumulates more triglycerides (31). In our study, adiponectin had an unexpected negative association with muscle density showing that higher concentrations were associated with lower density muscle. When we adjusted for adiponectin, the VAT–muscle density association was unchanged (see Supplementary Table S3 (23)), suggesting that adiponectin does not explain the deleterious association. IL-6 concentrations were not associated with muscle density and when we adjusted for IL-6 concentrations, we did not observe any attenuation of the inverse association between VAT and muscle density. These findings are limited by the fact that adiponectin and IL-6 concentrations were available only at the time of the first measures of VAT and muscle density. Nevertheless, our findings do not support the hypothesis that adiponectin and IL-6 explain that association between VAT and muscle density. There may be other mediators of the association that were not assessed in this study.

The design of our study is consistent with recent reports that note muscle function and composition should be a focus rather than muscle area or muscle quantity. In the Health Aging and Body Composition Study, lower paraspinal muscle density, but not area, was associated with worse physical performance and greater report of back pain (32). In the MESA cohort, greater abdominal muscle density, but not muscle area, was associated with markedly lower risk of all-cause mortality over a 10-year follow-up period, highlighting that low-density muscle may be a biomarker of visceral adiposity, which confers an increased risk of death (33). Finally, in a group of very old men and women, muscle density, but not muscle area, was associated with reduced postural sway, particularly sway velocities, and better Short Physical Performance Battery score in women (7). The fact that our study included older adults as well as middle-aged men and women may also be important. If interventions to reduce visceral fat were undertaken in middle age, it may be possible to increase paraspinal muscle density to improve function, reduce back pain, and reduce mortality risk with aging.

There are a few limitations worth noting. While the Framingham Study is a well-designed study with rich, high-quality research information, the measures necessary for our study were available only for a sample of European ancestry. To generalize our results to other ethnic groups may not be warranted and requires further evaluation. Additionally, while the negative association of VAT depot with surrounding muscle tissue is observed from our results, a real causal-effect relationship is not guaranteed from our association study design. Therefore, further exploration to investigate this causal-effect relationship is needed. In addition, if we had a larger sample size and shorter interval between the 2 visits to have a more precise menopausal transition date, it would be interesting to consider the analysis of women who had undergone the menopausal transition during the follow-up.

This study also has a number of strengths, including a focus on longitudinal visceral fat measures from CT scans, as well as a wide age range of middle-aged and older adults from a well-characterized cohort. Our study also included large numbers of men and women, and the ability to examine premenopausal and postmenopausal women. The associations we observed were fairly robust across subgroups along with consideration of important confounders.

Studying the association between VAT and muscle density remains of considerable interest in the community. Our results highlight that increased accumulation of VAT in men and women is associated with lower muscle density. As lower paraspinal muscle tissue density has been associated with increases in falls, VAT may serve as a modifiable risk factor in prevention treatment for poor musculoskeletal outcomes.

Funding

Research reported in this publication was supported by the National Institutes of Health (NIH) (National Institute of Arthritis Musculoskeletal and Skin Diseases [NIAMS], R01 AR041398; NIH Heart, Lung, and Blood Institute's Framingham Heart Study (contract no. N01-HC-25195 and HHSN268201500001I), HL076784, AG028321, HL070100, HL060040, HL080124, HL071039, HL077447, HL107385). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Disclosures

DPK as receives research grants from Amgen and Solarea Bio. He serves on a DSMB for Agnovos, and on scientific advisory boards for Pfizer, Reneo and Solarea Bio. He has received royalties for publication by Wolters-Kluwer for UpToDate. MTH has received an unrelated research grant from Amgen.

Data Availability

Restrictions apply to the availability of all data analyzed during this study to preserve participants’ confidentiality.

Abbreviations

 
• BMI

  body mass index

 
• CT

  computed tomography

 
• CV

  coefficient of variation

 
• ELISA

  enzyme-linked immunosorbent assay

 
• Gen3

  Framingham Heart Study third-generation cohort

 
• HU

  Hounsfield unit

 
• IL

  interleukin

 
• MDD

  minimum detectable dose

 
• MESA

  Multi-Ethnic Study of Atherosclerosis

 
• SAT

  subcutaneous adipose tissue

 
• VAT

  visceral adipose tissue