Body composition predictors of skeletal integrity in obesity

Abstract

Objective

To determine body composition predictors of skeletal integrity in overweight/obese subjects using DXA. We hypothesized that visceral adiposity would be negatively, and lean mass positively, associated with DXA measures of skeletal integrity in obesity.

Materials and Methods

Our study was IRB approved and HIPAA compliant and written informed consent was obtained. We studied 82 overweight or obese but otherwise healthy premenopausal women and men of similar age who were part of a clinical trial (mean age: 37±10 years, mean BMI: 34±7kg/m²). All subjects underwent DXA of the spine and hip for assessment of bone mineral density (BMD), trabecular bone score (TBS), and hip structural analysis (HSA) and of the whole body for assessment of body composition, including estimated visceral adipose tissue (VAT).

Results

Sixty-three subjects (77%) had normal BMD and 19 subjects (23%) had osteopenia. There were strong age-, sex-, and BMD-independent positive associations between lean mass and HSA parameters (r=0.50 to r=0.81, p<0.0001), while there was no association with TBS. There were strong age-, sex- and BMD-independent inverse association between total fat and VAT mass and TBS (r= −0.60 and r=−0.72, p<0.0001 for both correlations), while there were no associations with HSA parameters.

Conclusion

Lean mass is a positive predictor of hip geometry while fat and VAT mass are negative predictors of trabecular microarchitecture in overweight/obese subjects.

Introduction

Obesity is associated with increased risk for the development of type 2 diabetes, cardiovascular disease, and premature death [1]. Although conventional teaching is that obesity is protective against the development of osteoporosis via mechanical loading and peripheral aromatization of androgens to estrogen [2, 3], recent studies have shown that accumulation of fat, particularly visceral adiposity, is associated with bone loss and fractures [4–10]. Most studies on osteoporosis have focused on the elderly or patients with conditions predisposing for bone loss, such as anorexia nervosa, and body composition predictors of skeletal integrity in young overweight or obese young individuals have not been studied.

Dual energy X-ray absorptiometry (DXA) is usually performed to assess bone mineral density (BMD) and body composition [11, 12] due to its wide availability, minimal radiation dose, and relative low costs. However, BMD measured by DXA is limited in its ability to predict fractures, and patients with fragility fractures often have normal BMD [13]. In addition, the accuracy of DXA has been shown to be decreased in obesity [14, 15]. The limitations of BMD alone to predict fracture risk have led to the development of additional parameters of skeletal integrity. DXA-based hip structural analysis (HSA) is a simple method that can be used to assess proximal femoral geometry [16] and has been shown to improve the prediction of hip fracture risk [17]. Trabecular bone score (TBS), a grey-level texture parameter obtained from DXA images of the lumbar spine, can distinguish patients with and without fragility fractures independent of BMD [18], however this technique has not been validated in obese subjects. Furthermore, a new DXA method for assessment of visceral adipose tissue (VAT) has recently been validated [19].

The purpose of our study therefore was to determine body composition determinants of traditional and novel DXA parameters of skeletal integrity. We hypothesized that visceral adiposity would be negatively, and lean mass positively, associated with HSA parameters and TBS in this population.

Materials and Methods

Our study was IRB approved and complied with HIPAA guidelines. Written informed consent was obtained from all subjects after the nature of the procedure had been fully explained.

Subjects

Data presented here are from subjects who were being screened for eligibility for a clinical research study on obesity and bone and were recruited through advertisements. Inclusion criteria for the study presented here were ages 18–50 years, premenopusal status for women, and BMI ≥ 25 kg/m². Subjects were phone screened prior to the screening visit and subjects who had had hypothalamic or pituitary disorders, diabetes mellitus, rheumatoid arthritis or other chronic illness, chronic glucocorticoid use, or on other medications that are known to affect bone metabolism or bone mineral density (BMD) were not invited to the screening visit.

All subjects were examined during one study visit at our Clinical Research Center. BMI was calculated by using the formula BMI = W/H, where W is weight in kilograms and H is height in square meters. A complete history was recorded and a physical examination was performed.

Dual-energy X-ray absorptiometry

All subjects underwent DXA (Discovery A; Hologic Inc., Bedford, MA, USA) for assessment of BMD, hip structural analysis (HSA), trabecular bone score (TBS), and body composition. Examinations of lateral lumbar spine were excluded from analysis if the thickness score was ≥ 11 as recommended by the manufacturer.

Bone mineral density

BMD of the posterior-anterior (PA) and lateral lumbar spine and hip was determined. Coefficient of variation of DXA has been reported as less than 1% for BMD [20]. Osteopenia was defined by a T-score of ≤ −1.0 ≥ −2.5 and osteoporosis was defined by a T-score of < −2.5 [21].

Hip structural analysis

Hip structural analysis (HSA) was performed from properties of hip DXA images at the narrow neck (NN; narrowest part of the femoral neck), the intertrochanteric region (IT; along the bisector of the angle of the axes of NN and FS), and the femoral shaft (FS; across the shaft 1.5 cm from the NN to the intersection of the neck and shaft axes) to estimate hip geometry and strength indices [22]. The following HSA parameters were obtained: cross-sectional area (CSA; index of resistance to axial forces); cross-sectional moment of inertia (CSMI; estimate of resistance to bending forces in a cross-section); section modulus (Z; index of bending strength); cortical thickness; subperiosteal width; endocortical width; and buckling ratio (BR; index of susceptibility to cortical buckling under compressive loads). Higher values are associated with greater predicted femoral strength for all HSA-derived parameters except for BR, for which higher values are predictive of inferior strength.

Trabecular bone score

Trabecular bone score (TBS) was calculated as a mean value of the measurements for vertebrae L1–L4 at the same ROI as spine BMD using TBS iNsight software (Medimaps SA, France). TBS was evaluated by determining the variogram of the trabecular bone projected image as previously described [23]. The mean value of the individual measurements for L1–L4 represents the lumbar spine TBS (unitless) [23]. Reported in vivo precision for TBS ranges from 1.1% to 2.05% [24–26].

Body composition

Whole-body DXA was used to assess total lean mass and total fat mass. Coefficients of variation of DXA have been reported as 4.1% for body fat mass, and 1.0% for lean body mass [27, 28]. Estimated visceral adipose tissue (VAT) was assessed using Hologic APEX 4.0 software (Hologic Inc., Bedford, MA) as previously described [19, 29].

Statistical Analysis

Statistical analysis was performed using JMP software (version 11, SAS Institute, Cary, NC). Data are presented as mean ± standard deviation (SD). Nonparametric linear regression analyses were performed to determine body composition determinants of BMD, TBS, HSA and Spearman rho values are reported. Analyses were adjusted for age, sex, and BMD using multivariate standard least squares regression modeling. P values for univariate regression were corrected for multiple comparisons by the Bonferroni method. P<0.05 indicated statistical significance.

Results

Subject characteristics and body composition are shown in Table 1. Our study group was comprised of 82 overweight/obese but otherwise healthy subjects with a mean age of 37±10 years and a mean BMI of 34±7 kg/m². Twenty-four subjects (29%) were overweight (BMI ≥ 25 kg/m² ≤ 30 kg/m²) and 58 (71%) were obese (BMI ≥ 30 kg/m²). There were 41 men and 41 women.

Table 1.

Clinical characteristics and body composition of study subjects, mean±SD and (range)

	Subjects (n=82)
Age (years)	37 ±10 (18 – 50)
Women/men (n)	41/41
Weight (kg)	101±21 (64 – 157)
BMI (kg/m2)	34±7 (25 – 50)
Total lean mass (kg)	60±13 (39 – 98)
Total fat mass (kg)	39±14 (13 – 73)
Visceral adipose tissue mass (kg)	0.7±0.3 (0.2 – 1.9)

Sixty-three subjects (77%) had normal BMD (T-score > −1.0) and 19 subjects (23%) had osteopenia (T-score ≤ −1.0 ≥− 2.5) based on hip or spine BMD, which is higher than what would be expected in a normal population (T-score < −1SD = 15.8%). None of the subjects had osteoporosis. Within the women, 29 (71%) had normal BMD and 12 (29%) had osteopenia. Within the men, 34 (83%) had normal BMD and 7 (17%) had osteopenia.

In the overweight group (n=24), 8 (33%) had osteopenia and 16 (67%) had normal BMD, while in the obese group (n=58) 11 (19%) had osteopenia and 47 (81%) had normal BMD.

TBS correlated positively with PA lumbar spine BMD (r=0.40, p=0.005) and lateral lumbar spine BMD (r=0.31, p=0.01).

Hip BMD correlated positively with CSA, CSMI, Z and cortical thickness (r=0.32 to r=73, p=0.003 to p<0.0001) and inversely with BR (r= −0.36 to r= −0.44, p=0.0008 to p<0.0001) while there were no associations between hip BMD and subperiosteal or endocortical width (p>0.07).

In order to assess BMD-independent predictors of skeletal integrity, associations between measures of body composition and TBS and HSA were controlled for lumbar spine and hip BMD, respectively.

There was a positive age- and sex-independent association between BMI and total hip BMD and an inverse association with TBS, which remained significant after controlling for lateral and AP lumbar spine BMD (p<0.0001). There were positive associations between BMI and HSA parameter cortical thickness and inverse association with BR however, the associations lost significance after controlling for multiple comparisons, age, sex and BMD. There were no significant associations between BMI and lumbar spine BMD (Table 2). There was no association between TBS and hip BMD (p=0.3)

Table 2.

Associations between measures of body composition and skeletal integrity in overweight/obese subjects.

	BMI		Lean mass		Total fat mass		VAT mass
	r	p	r	p	r	p	r	p
TBS (L1–L4)	−0.55	<0.0001*†‡	−0.17	0.2	−0.60	<0.0001*†‡	−0.72	<0.0001*†‡
HSA NN CSA	0.17	0.1	0.66	<0.0001*†‡	0.06	0.6	0.07	0.5
HSA IT CSA	0.16	0.1	0.71	<0.0001*†	0.05	0.7	0.16	0.2
HSA FS CSA	0.15	0.2	0.74	<0.0001*†‡	0.03	0.8	0.14	0.2
HSA NN CSMI	0.01	0.9	0.79	<0.0001*†‡	−0.08	0.5	0.06	0.6
HSA IT CSMI	0.06	0.6	0.81	<0.0001*†‡	−0.05	0.7	0.12	0.3
HSA FS CSMI	0.03	0.8	0.70	<0.0001*†‡	−0.11	0.3	0.04	0.7
HSA NN Z	0.005	1.0	0.76	<0.0001*†‡	−0.03	0.8	0.05	0.7
HSA IT Z	0.10	0.4	0.78	<0.0001*†‡	−0.06	0.6	0.18	0.1
HSA FS Z	0.02	0.8	0.69	<0.0001*†‡	−0.06	0.6	0.08	0.5
HSA NN cortical thickness	0.22	0.05	0.29	0.007	0.12	0.3	0.10	0.4
HSA IT cortical thickness	0.30	0.006	0.43	<0.0001*†	0.19	0.09	0.29	0.01
HSA FS cortical thickness	0.16	0.2	0.35	0.001*	0.04	0.7	0.19	0.1
HSA NN subperiost. width	0.04	0.7	0.55	<0.0001*†‡	0.02	0.8	0.01	0.9
HSA IT subperiosteal width	−0.04	0.7	0.70	<0.0001*†‡	−0.09	0.4	0.04	0.7
HSA FS subperiosteal width	0.03	0.8	0.62	<0.0001*†‡	−0.02	0.9	0.02	0.8
HSA NN endocortical width	0.01	0.9	0.46	0.004	0.02	0.8	0.008	1.0
HSA IT endocortical width	−0.08	0.5	0.65	<0.0001*†‡	−0.11	0.3	0.03	0.8
HSA FS endocortical width	−0.11	0.3	0.25	0.02	−0.05	0.7	−0.12	0.3
HSA NN BR	−0.14	0.2	−0.04	0.7	−0.06	0.6	−0.08	0.5
HSA IT BR	−0.33	0.002	0.07	0.5	−0.25	0.02	−0.29	0.02
HSA FS BR	−0.12	0.3	0.09	0.4	−0.04	0.7	−0.12	0.3
L-spine BMD AP	0.22	0.05	0.31	0.005	0.09	0.4	0.10	0.4
L-spine BMD Lat	0.15	0.2	0.28	0.02	0.03	0.8	−0.13	0.3
Total hip BMD	0.38	0.0004*†	0.50	<0.0001*†	0.24	0.03	0.31	0.008

^*: significant after controlling for multiple comparisons;

^†: significant after controlling for multiple comparisons, age and sex;

^‡: significant after controlling for multiple comparisons, age, sex and lumbar spine BMD (for TBS) and hip BMD for HSA parameters

HSA: hip structural analysis; NN: narrow neck; IT: intertrochanteric; NS: femoral shaft; CSA: cross-sectional area; CSMI: cross-sectional moment of inertia; Z: section modulus; BR: buckling ratio

There were age- sex- and BMD-independent positive associations between lean mass and hip BMD and most HSA parameters (Table 2, Fig. 1),while there were no age- and sex-independent associations between lean mass and TBS or lumbar spine BMD. There were strong age-, sex- and BMD-independent inverse associations between total fat and VAT mass and TBS (Table 2, Fig. 2), while there were no associations between total fat and VAT mass and HSA parameters or BMD.

Fig. 1.

Regression analysis between lean mass and section modulus (Z, index of bending strength) at the narrowest part of the femoral neck. There is a positive association between lean mass and Z, which remained significant after controlling for multiple comparisons, age, sex, and hip BMD (p=0.02).

Fig. 2.

Regression analysis between visceral adipose tissue (VAT) mass and trabecular bone score (TBS). There is an inverse association between VAT mass and TBS, which remained significant after controlling for age, sex, and lumbar spine BMD (p<0.0001).

As the accuracy of TBS and HSA may decrease with increasing BMI, we performed a secondary analysis between TBS and HSA and body composition measures after excluding subjects with severe (class II and III) obesity (BMI >35 kg/m²) (n=29). There were inverse associations between TBS and VAT (r= −0.51, p=0.0005) and total fat mass (r= −0.30, p=0.05) and there was a trend of an inverse association between VAT and lumbar spine BMD (r= −0.22, p=0.1), while there was no association between lean mass and TBS (p=0.6). Strong positive associations between lean mass and hip BMD and HSA parameters (r=0.25 to r=0.85, p<0.04 to p<0.0001) persisted, while there were no associations between hip BMD and VAT or total fat mass (p>0.5).

Discussion

Recent data demonstrate that the epidemic of obesity is complicated by bone loss and fractures [6, 9, 30, 31]. Studies have suggested that excess fat mass, particularly visceral fat, may be a risk factor for bone loss and increased fracture risk in obesity [4–10]. Advances in DXA techniques allow assessment of proximal hip geometry by HSA [16] and surrogates of lumbar spine microarchitecture by TBS to better assess fracture risk [17, 18]. Furthermore, quantification of VAT mass by DXA has been shown to correlate well with values from computed tomography (CT) [19]. While most studies on DXA and fracture risk have focused on the elderly or populations at risk for bone loss, we focused on overweight/obese but otherwise healthy premenopausal women and men of similar age, a group which was thought to be protected from bone loss. We aimed to assess body composition predictors of traditional and novel measures of skeletal integrity by DXA.

In our study, BMI and fat mass were inversely associated with TBS, The inverse association between BMI and TBS is consistent with a large population study examining TBS in 29,407 women [25] and a study in healthy men and women across a wide age (30 to 80 years) and BMI range (15 to 47 kg/m²) [26]. We found strong inverse associations between VAT and fat mass and TBS, independent of age, sex, and lumbar spine BMD, which remained significant after excluding subjects with BMI >35 kg/m². However, there was no significant association between VAT mass and lumbar spine BMD, suggesting that high BMI and VAT mass disproportionally affect trabecular bone. Our findings of impaired microarchitecture in abdominal adiposity are consistent with a transiliac bone biopsy study, which found impaired bone microarchitecture and strength with lower bone formation in premenopausal women with abdominal adiposity [32]. Potential mechanisms for the detrimental effects of excess fat, and particularly VAT, on bone in obesity include dysregulation of the GH/IGF-1 axis with low GH and IGF-1 levels, which are important regulators of bone homeostasis [33–35], and the release of proinflammatory cytokines by VAT, which stimulate osteoclast activity [36, 37]. Additionally, low vitamin D levels are often found in obese individuals [38, 39].

We found no association between lean mass and TBS or lumbar spine BMD, suggesting that lean mass has no significant effects on lumbar spine microarchitecture or lumbar spine BMD, possibly because there is less mechanical traction of muscle on the lumbar spine compared to extremities. Our data are consistent with a recent study in a healthy Italian population across a wide age-spectrum that found no association between TBS and lean mass [40].

BMI was positively associated with hip BMD, independent of age and sex, while there were no significant independent associations with HSA parameters. This is consistent with studies that have shown adaptive increase in hip BMD in response to increasing BMI [41, 42], while relative to BMI, obese subjects had impaired hip structure compared to normal-weight subjects [41–43], suggesting that hip structure is insufficient to compensate for increasing BMI. However, in a study of older (>65 years) individuals this did not translate into increased risk of developing a major osteoporotic fracture over a follow-up time of 6.2 ± 3.7 years in women and 4.7 ± 3.4 6.2 ± 3.7 years in women and 4.7 ± 3.4 years [43].

Lean mass was positively associated with hip BMD and HSA parameters, independent of age and sex. Most of the associations between lean mass and HSA parameters remained significant after controlling for hip BMD and after excluding subjects with BMI > 35 kg/m², suggesting that HSA predicts skeletal fragility independent of BMD. Our results are consistent with a recent study by Hu et al. that showed lean mass, but not fat mass, to be positively associated with BMD and HSA parameters in Chinese men and women across a wide age spectrum [44]. Beck et al. found higher hip BMD, CSA, and section modulus (Z, an engineering index of bending strength), in postmenopausal women with greater BMI, and these effects were proportionate to total body lean mass but not total fat or total body mass [41]. Positive effects of lean mass on bone include skeletal loading at weight-bearing sites, stimulating adaptive increases in BMD and alteration of hip geometry [45]. In a different study, Beck et al. demonstrated that despite declines in BMD, subjects with constant skeletal loads maintained section moduli at the femoral neck and shaft, suggesting that mechanical homeostasis in the hip is better characterized by section modulus than bone density [45]. Furthermore, recent studies suggest that several genes and molecular pathways exert a pleiotropic effect on both muscle and bone [46]. For example, bone morphogenic proteins (BMP) are regulators of both bone and muscle formation and homeostasis [47]. In-vitro studies have demonstrated that myocytes secrete factors that can protect osteocytes from glucocorticoid-induced cell death through activation of the Wnt/β-catenin pathway, a pathway known to be important for the maintenance of normal bone [48, 49].

Interestingly, 23% of our overweight/obese but otherwise healthy premenopausal women and men of similar age had osteopenia, which is higher than what would be expected in a normal population (t-score < −1SD = 15.8%). Within this group, the prevalence of osteopenia was higher in women compared to men (29% vs 17%) and in overweight subjects compared to obese (33% vs 19%), suggesting that women and overweight subjects are at highest risk of osteopenia.

We did not report the World Health Organization fracture risk assessment tool (FRAX) as our female subjects were premenopausal and the male subjects of similar age and none were taking glucocorticoids, had secondary osteoporosis or rheumatoid arthritis.

Our study had several limitations. First is the cross-sectional study design, which limits our ability to prove causality. Second, we used DXA-derived parameters of lumbar spine microarchitecture (TBS) and hip geometry (HSA) to assess skeletal integrity as well as estimated VAT by DXA instead of CT or MRI. However, studies have shown strong correlations between these variables and equivalent measurements from QCT with less radiation dose [19, 50, 51]. Third, obesity may affect the accuracy of DXA-derived parameters of skeletal integrity and body composition [14, 15]. We therefore excluded examinations with a thickness score of >11. Strengths of our study include the unique population of young, overweight/obese but otherwise healthy men and women and detailed assessment of skeletal integrity and body composition.

In conclusion, lean mass is a positive predictor of hip geometry while fat and VAT mass are negative predictors of trabecular microarchitecture in overweight/obese subjects.