Bone 74 (2015) 146–152

Contents lists available at ScienceDirect

Bone journal homepage: www.elsevier.com/locate/bone

Original Full Length Article

Associations between body mass index, lean and fat body mass and bone mineral density in middle-aged Australians: The Busselton Healthy Ageing Study Kun Zhu a,b,⁎, Michael Hunter c,d, Alan James b,e, Ee Mun Lim a,f, John P. Walsh a,b a

Department of Endocrinology and Diabetes, Sir Charles Gairdner Hospital, Nedlands, Western Australia, Australia School of Medicine and Pharmacology, University of Western Australia, Crawley, Western Australia, Australia c Busselton Population Medical Research Institute, Busselton, Western Australia, Australia d School of Population Health, University of Western Australia, Crawley, Western Australia, Australia e Department of Pulmonary Physiology and Sleep Medicine, Sir Charles Gairdner Hospital, Nedlands, Western Australia, Australia f Department of Clinical Biochemistry, PathWest Laboratory Medicine, Queen Elizabeth II Medical Centre, Nedlands, Western Australia, Australia b

a r t i c l e

i n f o

Article history: Received 13 October 2014 Revised 22 January 2015 Accepted 23 January 2015 Available online 31 January 2015 Edited by: Kristine Ensrud Keywords: Lean body mass Fat mass Body mass index Bone mineral density Middle-aged adults Busselton Healthy Ageing Study

a b s t r a c t Low BMI is a risk factor for osteoporosis, but it is not clear if relationships between BMI, lean mass (LM), fat mass (FM) and BMD are consistent across different levels of BMI. We studied 1929 Caucasian participants (1014 females) aged 45–66 years in the Busselton Healthy Ageing Study in Western Australia. Body composition and BMD of total body, lumbar spine, total hip and femoral neck were measured using DXA. From generalized additive models, the positive relationships between BMI and BMD were weaker at high BMI, particularly at the spine and in males. In the entire cohort, adjusting for relevant covariates, LM and FM were significant predictors of all BMD measures in both genders. In men, analysis by tertiles of BMI showed that LM and FM (in kg) were positively associated with BMD (in mg/cm2) in tertile 1 except for LM and spine BMD (LM β: 5.18–6.80, FM β: 3.38–9.24, all P b 0.05), but not in the middle or upper tertiles (LM β: −3.12–3.07, FM β: −4.75–1.82, P N 0.05). In women, LM was positively associated with BMD in each tertile of BMI, except for spine BMD in the upper tertile, with regression coefficients lower in the upper tertile (β: 5.16–9.95, 5.76–9.56 and 2.80–5.78, respectively, all P b 0.05). FM was positively associated with total body, spine and total hip BMD in women in BMI tertile 1 (β: 2.86–6.68, P b 0.05); these associations were weaker or absent in the middle and upper tertiles. In conclusion, in middle-aged adults the positive relationships between lean or fat mass with BMD among those with higher BMI are absent in males and weaker in females. © 2015 Elsevier Inc. All rights reserved.

Introduction Body weight is known to be positively correlated with bone mineral density (BMD), and low body weight is a recognized risk factor for osteoporosis. Body mass index (BMI) is routinely used in epidemiological studies and clinical practice to classify adults as underweight, overweight or obese, and studies evaluating the relationship between BMI and BMD using linear regression have shown positive associations [1]. However, it is not clear whether the relationship between BMI and BMD is as strong in overweight and obese people as it is in those with low or normal body weight, which is of particular interest given the substantial increases in the prevalence of obesity worldwide in the past 30 years [2].

Abbreviations: BMD, bone mineral density; BMI, body mass index; DXA, dual-energy X-ray absorptiometry; GAM, generalized additive model; FM, fat mass; LM, lean mass; 25OHD, 25-hydroxyvitamin D. ⁎ Corresponding author at: Department of Endocrinology and Diabetes, Sir Charles Gardiner Hospital, Nedlands, Western Australia 6009, Australia. Fax: +61 8 9346 4109. E-mail address: [email protected] (K. Zhu).

http://dx.doi.org/10.1016/j.bone.2015.01.015 8756-3282/© 2015 Elsevier Inc. All rights reserved.

There is also disagreement in the literature on the relative contributions of the two components of soft tissue, lean and fat body mass, to the relationship between body weight and BMD. Studies in men, and preand post-menopausal women have reported that lean mass is a key determinant of BMD [3–5], whereas some studies of postmenopausal women have suggested that fat mass plays a key role [6,7], and both lean and fat mass have been reported to be significant predictors of BMD in other studies of men and pre- and post-menopausal women [8–10]. This heterogeneity in findings could arise from differences between studies in age, gender and ethnicity of participants (with associated differences in body composition), in turn reflecting the different mechanisms by which lean and fat mass influence bone mass. Lean body mass affects bone density through direct mechanical loading (muscle contraction and gravitational loading) [11]. Adipose tissue influences BMD directly through gravitational loading and has indirect, positive effects by endocrine mechanisms, including aromatization of androgens to estrogens, secretion of leptin by adipocytes and increased pancreatic insulin secretion in individuals with higher fat mass [12,13]. However, adipose tissue can also negatively impact the bone through

K. Zhu et al. / Bone 74 (2015) 146–152

production of inflammatory cytokines which increase bone resorption [14]. Importantly, it is not known if the relationships between each of lean and fat mass with BMD differ between overweight and obese people and those with low or normal body weight, as most previous studies have been too small to examine this reliably. In one study of 5025 men and women aged 47–50 and 71–75 years, the association between fat mass and femoral neck BMD appeared to be weaker in participants with higher body fat [15], but BMD at other sites (whole body, lumbar spine and total hip) was not examined. The Busselton Healthy Ageing Study is a large, community-based study of ‘Baby Boomers’, defined as individuals born between 1946 and 1964 and who were aged 45–66 years at the time of baseline survey [16]. This is an important age group for implementing strategies to prevent age-related bone loss and future fracture. The first phase of the study comprises a cross-sectional health survey including measures of bone density and body composition using dual-energy X-ray absorptiometry (DXA). The aims of this analysis were firstly to examine relationships of body weight and BMI with BMD of total body, spine and hip using linear regression as well as generalized additive models, and secondly to examine relationships of lean and fat mass with BMD at different levels of BMI.

147

calibration of the DXA machine using a phantom was performed prior to each scanning session. The precision error was b2.0% for each measured site at standard speed based on repeated scans in a random sample of 30 subjects. Other assessments

Subjects and methods

Standing height and body weight were measured using standard anthropometric techniques with the participants lightly-clothed and shoeless. BMI was calculated as weight (kg) / height (m2). BMI between 18.5 and 24.99 kg/m2 was regarded as normal, b18.5 underweight, 25–29.99 overweight and ≥30 obese according to WHO criteria [17]. Data on health history, medication use, and smoking habit were collected using a questionnaire [16]. The questionnaire did not include detailed dietary history such as calcium and protein intake, but did include a question regarding avoidance of dairy products. Smoking history was collected by questionnaire, and participants classified as current, never or previous smokers. Physical activity level was assessed using the International Physical Activity Questionnaire (IPAQ), and categorized as low, medium and high according to the IPAQ scoring protocol [18]. Fasting blood samples were collected, and the serum 25-hydroxy vitamin D (25OHD) level was measured using the ARCHITECT 25-OH Vitamin D assay (Abbott Laboratories, Abbott Park, Illinois, US).

Subjects

Data analysis

The design and rationale of the Busselton Healthy Ageing Study have been described previously [16]. Busselton is a coastal community in the south-west of Western Australia with a relatively stable population of predominantly European descent. All non-institutionalised baby boomers (defined as born from 1946 to 1964; aged 45–66 years at the time of baseline survey) who currently live in the Shire and listed on the electoral roll are eligible to participate; electoral registration is compulsory in Australia. Phase 1 of the study is a cross-sectional health survey, and will be followed by longitudinal studies of participants. Between May 2010 and June 2012, 2023 participants were recruited to the study (comprising ~80% of those eligible) and of these 1985 had a DXA examination. After excluding 32 participants taking medication for osteoporosis, and 24 from non-Caucasian backgrounds, a total of 1929 subjects (915 males and 1014 females) were included in this analysis. Pilot data from the first 300 subjects showed prevalence values for common risk factors such as obesity and overweight to be almost identical to those in the recent nationally representative Australian National Health Survey [16]. The study has received ethics approval from the University of Western Australia Human Research Ethics Committee (Number RA/4/1/2203) and written informed consent was obtained from each participant.

Variables are presented as means ± standard deviations (SD) by gender. Comparisons between males and females were made by Student's t-test and chi-squared test. Linear regression analysis was used to evaluate the relationships between body weight, BMI, lean and fat mass and bone measures in each gender, with BMD of whole body, spine, total hip and femoral neck as dependent variables; body weight, BMI, or lean and fat mass as predictor variables; and age, height (except for the model for BMI), smoking history (with current and past smoking coded as “Yes” or “No for each subject), serum 25OHD, physical activity level and dairy avoidance as covariates. In addition, generalized additive models (GAM), which offer a greater flexibility to represent the relations between the dependent variable and predictor variables compared with linear regression, were used to generate graphic representations of the dose–response relations of body weight and BMI with BMD in each gender, adjusted for the covariates listed above for linear regression models. Comparisons of different linear and GAM models were made by ANOVA chi-squared test. In further models to assess the influence of lean and fat mass on the bone in relation to body fatness, study participants were sub-grouped according to tertiles of BMI in each gender. Tertiles were used rather than WHO criteria to ensure that groups are of equal size to provide adequate statistical power for the analysis in each subgroup. In addition, because men have higher muscle mass than women, using the WHO BMI cutoff of 25 kg/m2 tends to classify more men into the overweight category [19]. Collinearity or nearcollinearity was not observed in any of the models, based on a variance inflation factor (VIF) value less than 10 [20]. The normality and independence of the residuals and the homogeneity of variance of each model were checked using residual plots (normal probability plot and plot of residuals vs predicted values). Statistical significance level was set at P b 0.05 (two-tailed). All analyses were performed using IBM SPSS (version 21, IBM, Chicago, IL, USA) and R (version 3.0.1, R Foundation for Statistical Computing, Vienna, Austria).

Dual energy X-ray absorptiometry (DXA) DXA scans were undertaken to assess BMD of whole body, anterior– posterior spine (L1–L4), total hip and femoral neck using a GE Lunar Prodigy Pro densitometer (Madison, WI, USA). The scans were analyzed using enCORE Version 13 (GE Health) software with manual inspection of regions of interest and adjustment where necessary by two independent reviewers (MH and KZ). BMD is normally reported in g/cm2, but that results in very low numerical values for regression coefficients in this study. For clarity, therefore we present BMD data in mg/cm2; to convert to g/cm2, divide by 1000. Body composition estimates including whole body fat mass (g) and lean mass (bone free) (g) were also obtained from the total body DXA scan. Percentage body fat mass was calculated as (fat mass / total mass) × 100 and percentage trunk fat mass was calculated as (trunk fat mass / total mass) × 100. Fat mass index was calculated as fat mass (kg) / height (m2) and lean mass index was calculated as lean mass (kg) / height (m2). Annual servicing and calibration according to manufacturer's specifications were carried out and

Results Descriptive statistics The mean age (±SD) of participants was 56.6 ± 5.6 years for males and 56.0 ± 5.5 years for females. Males were taller and heavier compared with females and had slightly higher BMI (Table 1). The percentage of

148

K. Zhu et al. / Bone 74 (2015) 146–152

Associations of body weight and BMI with BMD

Table 1 Anthropometric, body composition and bone measures in participants. Male (n = 915)

Female (n = 1014)

Pa

56.5 ± 5.6 175.8 ± 6.7 87.8 ± 13.9 28.4 ± 4.0

56.0 ± 5.5 163.0 ± 6.0 73.8 ± 14.3 27.8 ± 5.4

0.025 b0.001 b0.001 0.014

0/18.0/52.5/29.5

0.4/33.3/36.5/29.8

b0.001

44.8/43.0/12.2

51.1/39.9/9.0

18.6/31.7/49.7 2.1 83.5 ± 24.2

23.0/42.8/34.2 2.2 76.1 ± 22.9

b0.001 0.887 b0.001

Body composition measures Lean mass (kg) Lean mass index (kg/m2) Fat mass (kg) Fat mass index (kg/m2) Percentage body fat (%) Percentage trunk fat (%)

59.7 ± 6.7 19.3 ± 1.7 24.9 ± 9.3 8.0 ± 2.9 27.5 ± 6.8 17.2 ± 4.4

40.5 ± 4.8 15.3 ± 1.6 30.4 ± 11.0 11.5 ± 4.2 40.2 ± 7.9 20.8 ± 4.9

b0.001 b0.001 b0.001 b0.001 b0.001 b0.001

Bone mineral density (mg/cm2) Total body Spine Total hip Femoral neck

1310.5 ± 96.6 1264.9 ± 186.7 1105.9 ± 139.6 1015.3 ± 133.1

1193.5 ± 96.5 1180.7 ± 172.2 997.8 ± 139.9 948.3 ± 132.4

b0.001 b0.001 b0.001 b0.001

Age (year) Height (cm) Weight (kg) BMI (kg/m2)b BMI category (%) Low/normal/overweight/obese Smoking (%) Never/past/current Physical activity level (%) Low/moderate/high Avoidance of dairy products (%) Serum 25OHD (nmol/L)

Regression coefficients from the linear regression models with body weight (Model 1) or BMI (Model 2) as the predictor variables are presented in Table 2. Both body weight and BMI were significant predictors of BMD of total body, spine, total hip and femoral neck in men and women. However, body weight appeared to be a stronger predictor than BMI, with Model 1 explaining a higher percentage of the variation in BMD measures compared with model 2 (adjusted R2: 0.076–0.177 vs 0.041–0.120 in males; 0.147–0.279 vs 0.124–0.238 in females, all P b 0.001). Fig. 1 depicts the dose–response relations of body weight and BMI with BMD in each gender using the generalized additive models. The associations appear weaker at higher values of body weight and BMI, particularly for the models with BMI as the predictor variable, for spine BMD compared with other sites and in males compared with females. The percentage of variation in BMD measures explained by the GAM was slightly but significantly higher than those of the linear regression models (except for the model for body weight and femoral neck BMD in females) (Table 2), indicating the relationships are better represented by GAM.

0.008

Association of lean and fat mass with BMD

Values are mean ± SD unless otherwise stated. To convert BMD from mg/cm2 to BMD in g/cm2, divide by 1000. a Student's t-test or chi-squared test.

participants with normal BMI was lower in males than females (18.0% vs 33.3%), a higher proportion of males were overweight (52.5% vs 36.5%), and the prevalence of obesity was similar in males and females (29.5% vs 29.8%). More men than women were current smokers and had high physical activity levels, but the percentage of males and females avoided dairy products were similar. Men had higher serum 25OHD levels, and fewer men than women had serum 25OHD levels below 50 nmol/L (5.1% vs 11%, P b 0.001). Men had significantly higher lean body mass and lean mass index, whereas women had significantly higher fat mass, fat mass index, and percentage body and trunk fat (Table 1). BMD at total body, spine, total hip and femoral neck were all significantly higher in males compared with females (Table 1).

In the linear regression models with BMI (in kg/m2) as the dependent variable, lean and fat mass (in kg) as the predictor variables and age as the covariate, the regression coefficients (SE) were 0.128 (0.009) and 0.350 (0.007) for lean and fat mass, respectively, in males; and 0.145 (0.013) and 0.428 (0.006), respectively, in females. These indicate that each additional 1 kg of fat mass is associated with a two- to three-fold greater increase in BMI than is each additional 1 kg of lean mass. Tables 3 and 4 show the regression coefficients of lean and fat mass in males and females for total body, spine, total hip and femoral neck BMD for all participants and according to tertile of BMI, adjusted for age, height, current and past smoking, 25OHD, physical activity and dairy avoidance. In the cohort as a whole, lean and fat mass were significant positive predictors of all BMD measures in both genders. Lean body mass was a stronger predictor than fat mass of BMD at each site in both genders, with the exception of lumbar spine BMD, where the results were similar for lean and fat mass. In men, each additional kilogram of lean mass was associated with an expected BMD increase of 2.48–5.90 mg/cm2 across different sites, compared with 1.48–

Table 2 Regression coefficients of models with body weight (kg) or BMI (kg/m2) as the predictor variable for total body, spine, total hip and femoral neck BMD (mg/cm2). Male

Female a

Linear regression model β Total body BMD Model 1: Weight Model 2: BMI Spine BMD Model 1: Weight Model 2: BMI Total hip BMD Model 1: Weight Model 2: BMI Femoral neck BMD Model 1: Weight Model 2: BMI

2

GAM Adjusted R2b

Linear regression modela

GAM Adjusted R2b

SE

P

Adjusted R2

2.87 6.99

0.20 0.54

b0.001 b0.001

0.279 0.238

0.283* 0.244**

0.089** 0.055***

3.61 8.66

0.38 1.01

b0.001 b0.001

0.147 0.124

0.154** 0.132**

0.135 0.120

0.146** 0.132**

4.41 11.03

0.29 0.76

b0.001 b0.001

0.246 0.231

0.251* 0.235*

0.155 0.113

0.161* 0.120**

3.39 8.07

0.27 0.73

b0.001 b0.001

0.255 0.211

0.257 0.216*

SE

P

Adjusted R

2.59 7.74

0.25 0.80

b0.001 b0.001

0.177 0.103

0.188*** 0.112***

3.21 9.59

0.51 1.60

b0.001 b0.001

0.076 0.041

3.96 12.06

0.36 1.13

b0.001 b0.001

3.33 10.01

0.34 1.08

b0.001 b0.001

β

To convert BMD from mg/cm2 to BMD in g/cm2, divide by 1000. *P b 0.05, **P b 0.01, ***P b 0.001 compared with the corresponding linear regression model. Covariates adjusted in both linear regression and GAM include age, height (for the model for body weight only), smoking, serum 25OHD, physical level and dairy avoidance. a Linear regression analysis with BMD as dependent variable; body weight (Model 1) or BMI (Model 2) as the predictor variable. b Adjusted R2 obtained from generalized additive models (GAM) with BMD as dependent variable; body weight or BMI as the predictor variables (graphic representations see Fig. 1).

K. Zhu et al. / Bone 74 (2015) 146–152

149

A: Male Total body BMD

Spine BMD

Total hip BMD

Femoral neck BMD

Total hip BMD

Femoral neck BMD

Model 1: Body weight

Model 2: BMI

B: Female Total body BMD

Spine BMD

Model 1: Body weight

Model 2: BMI

Fig. 1. Graphic presentation of the dose–response relationship between body weight (Model 1) or BMI (Model 2) in male (A) and female (B) obtained by generalized additive regression models. Models adjusted for age, height (for the models for body weight only), smoking, serum 25OHD, physical level and dairy avoidance as covariates. Dotted lines represent 95% confidence intervals. The reference value for BMD is the value associated with the mean body weight or BMI for all subjects in each gender. The rug plot along the bottom of each graph depicts each observation.

3.29 mg/cm2 for each kilogram of fat mass. In women, each additional kilogram of lean mass was associated with an additional 3.54– 7.81 mg/cm2 of BMD across different sites, compared with an expected BMD increase of 2.04–3.51 mg/cm2 per each kilogram of fat mass.

between either lean mass or fat mass and any measure of BMD (Table 3).

Associations in males in relation to BMI

When women were analyzed by tertile of BMI, lean and fat mass were positively associated with all BMD measures in the lower tertile, except for fat mass and femoral neck BMD which had no significant association. In the middle BMI tertile, lean mass was positively associated with all BMD measures with similar expected increases in BMD per kilogram of lean mass to the lower tertile, but fat mass was not a

When males were analyzed by tertiles of BMI, the significant positive relationships between lean and fat mass with BMD observed in the entire cohort were present only in lower BMI tertile. For participants in the middle and upper tertiles of BMI, there were no significant associations

Associations in females in relation to BMI

150

K. Zhu et al. / Bone 74 (2015) 146–152

Table 3 Regression coefficients of lean and fat body mass for total body, spine, total hip and femoral neck BMD (mg/cm2) in male participants according to BMI tertile. All

Tertile 1 (≤26.5 kg/m2) n = 306

Tertile 2 (26.6–29.4 kg/m2) n = 306

Tertile 3 (≥29.5 kg/m2) n = 303

β

SE

P

β

SE

P

β

SE

P

Total body BMD Lean mass (kg) Fat mass (kg)

4.65 1.48

0.60 0.37

b0.001 b0.001

6.08 3.38

1.51 1.23

b0.001 0.006

2.67 −3.02

2.05 2.06

0.194 0.145

Spine BMD Lean mass (kg) Fat mass (kg)

2.48 3.29

1.25 0.76

0.047 b0.001

4.76 9.24

2.84 2.31

0.095 b0.001

0.60 −3.50

4.32 4.38

Total hip BMD Lean mass (kg) Fat mass (kg)

5.90 2.97

0.89 0.54

b0.001 b0.001

6.80 5.62

2.18 1.77

0.002 0.002

3.07 −4.75

Femoral neck BMD Lean mass (kg) Fat mass (kg)

4.77 2.57

0.84 0.51

b0.001 b0.001

5.18 4.04

2.04 1.66

0.012 0.016

2.91 −4.22

β

SE

P

1.62 0.14

1.08 0.69

0.136 0.841

0.890 0.424

−3.12 0.26

2.35 1.48

0.187 0.862

3.04 3.08

0.314 0.124

1.75 1.76

1.66 1.04

0.293 0.090

2.75 2.78

0.290 0.130

0.36 1.82

1.63 1.02

0.824 0.076

To convert BMD in mg/cm2 to BMD in g/cm2, divide by 1000. Linear regression analysis with BMD as dependent variable; lean and fat body mass as the predictor variables; and age, height, smoking, serum 25OHD, physical level and dairy avoidance as covariates. P values reached statistical significance level (Pb0.05, two tailed) are indicated in bold.

significant predictor of BMD at any site. In women in the upper tertile of BMI, there were significant positive associations between lean mass and BMD of total body, total hip and femoral neck, and significant positive associations between fat mass and BMD of spine and total hip, but the regression coefficients were lower than the corresponding values in the lower tertile (Table 4).

1 kg of fat mass is associated with a two- to three-fold greater increase in BMI than is each additional 1 kg of lean mass, and lean mass was a stronger predictor than fat mass of BMD, as shown in the present study and a recent meta-analysis [21]. Our study is the first to evaluate in detail the associations between lean mass, fat mass and BMD in relation to BMI. Our results are consistent with limited data from one previous study in which there was a stronger association between fat mass and femoral neck BMD in participants with lower levels of body fat [15], but that study did not examine whole body, spine or total hip BMD, and included only participants aged 47–50 and 71–75 years. The mechanisms resulting in weaker associations between lean and fat mass with BMD at higher BMI are not known, and given the complexities of bone physiology, several possibilities exist as detailed below.

Discussion In this study of middle-aged Australians, we found that body weight was a stronger predictor than BMI for BMD measures, and the relationships between BMI and BMD were weaker in individuals with higher BMI, particularly for BMD measured at the lumbar spine, and in males compared with females. In the cohort as a whole, lean mass and fat mass each had a positive relationship with all BMD measures in both genders but when analyzed by tertiles of BMI, the positive relationships between lean body mass and fat mass with BMD are largely lost in males and weaker in females at higher BMI. In the present study, we found that body weight, as a measure of body mass, was a stronger predictor than BMI, a measure of body fatness, for BMD measures. The models with body weight as the predictor variable explained a higher percentage of the variation in BMD than the models with BMI as the predictor variable. This could be because BMI is more closely correlated with fat mass, in that each additional

Lean mass Lean mass influences bone density through direct mechanical effects of muscle including both muscle contraction and gravitational loading on the bone that produces a positive osteogenic response [11]. It is possible that the response of the mechanical loading effect on increasing bone density decreases with increased BMI, as there is relatively greater increase in fat than muscle. An interesting finding of our study is that in the second and third tertiles of BMI, lean mass had significant

Table 4 Regression coefficients of lean and fat body mass for total body, spine, total hip and femoral neck BMD (mg/cm2) in female participants according to BMI tertile. All

Tertile 1 (≤24.9 kg/m2) n = 342

β

SE

P

β

SE

Total body BMD Lean mass (kg) Fat mass (kg)

5.42 2.04

0.73 0.30

b0.001 b0.001

6.40 2.86

1.89 1.20

Spine BMD Lean mass (kg) Fat mass (kg)

3.54 3.51

1.42 0.57

0.013 b0.001

8.41 6.68

Total hip BMD Lean mass (kg) Fat mass (kg)

7.81 3.38

1.07 0.43

b0.001 b0.001

Femoral neck BMD Lean mass (kg) Fat mass (kg)

5.32 2.79

1.01 0.41

b0.001 b0.001

Tertile 2 (25.0–29.3 kg/m2) n = 339

Tertile 3 (≥29.4 kg/m2) n = 333

β

SE

P

0.001 0.018

7.28 1.06

1.79 1.34

b0.001 0.429

3.38 2.12

0.013 0.002

9.56 3.35

3.58 2.69

9.95 4.62

2.59 1.64

b0.001 0.005

8.27 2.26

5.16 2.34

2.43 1.54

0.035 0.131

5.76 1.56

P

β

SE

P

2.80 1.27

1.13 0.67

0.014 0.058

0.008 0.213

−2.67 3.10

2.38 1.35

0.264 0.023

2.64 1.98

0.002 0.254

5.78 2.12

1.83 1.04

0.002 0.043

2.49 1.87

0.021 0.402

3.55 1.63

1.73 0.99

0.041 0.099

To convert BMD in mg/cm2 to BMD in g/cm2, divide by 1000. Linear regression analysis with BMD as dependent variable; lean and fat body mass as the predictor variables; and age, height, smoking, serum 25OHD, physical level and dairy avoidance as covariates. P values reached statistical significance level (Pb0.05, two tailed) are indicated in bold.

K. Zhu et al. / Bone 74 (2015) 146–152

associations with BMD measures in females but not in males, and one possible explanation could be related to the role of estrogen receptors and their different actions on male and female skeleton during loading, as it has been reported that estrogen receptor-a (ERa) plays a critical role in the adaptive response of bone to loading in female mice [22], but not in male mice [23].

Fat mass Adipose tissue can directly influence bone through gravitational loading. In addition, adipose tissue may also indirectly influence bone metabolism by endocrine mechanisms. Sex steroids play a role in the regulation of body composition, and in turn, sex hormone levels are influenced by change in body composition. Adipose tissue enhances estrogen production through the aromatization of androgens [13], whereas serum testosterone has been reported to be inversely related to BMI and total body fat mass in older men [24,25]. Therefore, high fat mass may be associated with increased estrogen production in women, but reduced testosterone level in men (which in turn results in reduced aromatization of testosterone to estradiol in men). As both estrogen and testosterone are important for the maintenance of bone mass, this may explain that in those with BMI at the upper tertile, fat mass remained to be a significant predictor of spine and total hip BMD in female but not in male. It is not entirely clear why fat mass was not a significant predictor of BMD in the second tertile of BMI although may be due to the strong influence of lean mass on BMD in this tertile. Further complexity is added by reports that high fat mass, especially high visceral fat, may have deleterious effects on bone due to increased inflammatory cytokines [14], reduced circulating 25OHD levels [26], increased parathyroid hormone concentration [27], and dysregulation of the growth hormone (GH)/insulin-like growth factor-1 (IGF-1) axis [28], an important determinant of BMD [29]. While women generally have a higher body fat percentage compared to men, women store more fats in the gluteal–femoral region, and men tend to store more metabolically active, visceral fat [30]. Further research is required to elucidate the mechanisms underlying our findings. In a meta-analysis of population-based cohort studies, low BMI was a risk factor for total, osteoporotic and hip fracture in both men and women [31]. Although the prevalence of obesity is increasing, a recent population-based historical cohort study from Canada concluded that this did not account for the reduction in major fracture incidence which occurred over the same period [32]. This in turn suggests that low body weight should be considered as a risk factor for osteoporosis and related fracture, rather than obesity being a protective factor. Our study showed that the associations of BMI with BMD measures were attenuated in those with high BMI, particularly in men, and at the lumbar spine. Since lean mass is more strongly associated with BMD than is fat mass, lifestyle factors such as nutrition including adequate intake of protein and dairy products [33,34] and physical activity that benefit both bone and lean body mass [35,36] may be beneficial for the maintenance of bone mass, especially in those with relatively low BMI. The strengths of our study include its large sample size, which allowed us to evaluate the associations of lean and fat mass with BMD in people at different levels of BMI; the narrow age range of participants and restriction of the analysis to Caucasians which minimizes potential confounding effects of age and ethnicity [13,37]; the use of generalized additive models to evaluate the dose–response relations of body weight and BMI with BMD, and the fact that study participants are representative of middle-aged Australians in general [16]. Our study also has limitations. The age restriction and ethnic homogeneity of the cohort mean that the results may not apply to other age groups or non-Caucasians. In addition, the study is observational and cross-sectional, and it cannot be assumed that the relationships demonstrated are causal in nature. Longitudinal analysis of intra-individual changes in body composition and bone density will be of interest when the next phase of the study

151

has been completed. Detailed dietary histories were not available, and intake of protein and calcium could not be included as covariates. In conclusion, in this study of middle-aged Australians, we found body weight, as a measure of body mass, appeared to be a stronger predictor than BMI, as a measure of body fatness, for BMD measures. We also found that the positive relationships between lean and fat mass and bone density among those with higher BMI are absent in males and weaker in females. These findings suggest that at higher BMI, there are limited additional benefits of increased body mass on BMD, especially in males. Further research is required to explore the mechanisms underlying these findings, which may have public health implications for fracture prevention strategies in the setting of the global epidemic of obesity. Funding support The baseline survey of the Busselton Healthy Ageing Study was funded by grants from the Office of Science and Department of Health (grant number G05911) of the Government of Western Australia, the City of Busselton, a bequest of the late Dr Janet Elder and private donors. The DXA machine was funded by a NHMRC equipment grant. None of the funding agencies had any role in the conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript. Author's disclosure Kun Zhu, Michael Hunter, Alan James, Ee Mun Lim, and John P Walsh declare that they have no conflict of interest. Acknowledgments We thank the Western Australian Country Health Service—South West for core infrastructure support, the operational team in Busselton of Elspeth Inglis, Aida Embling, Darcy Bennett, Shelley Cheetham, Jessica Storey, Debra Burwood, Dianne Toovey, Stephanie Murphy and Jenifer George for participant recruitment and data collection, and the community of Busselton for their ongoing support and participation. We thank BD Biosciences for donating blood collection kits and Abbott Australasia Pty Ltd for donating assay kits for 25-OH Vitamin D. References [1] Lloyd JT, Alley DE, Hawkes WG, Hochberg MC, Waldstein SR, Orwig DL. Body mass index is positively associated with bone mineral density in US older adults. Arch Osteoporos 2014;9:175. [2] Ng M, Fleming T, Robinson M, Thomson B, Graetz N, Margono C, et al. Global, regional, and national prevalence of overweight and obesity in children and adults during 1980–2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet 2014;384(9945):766–81. [3] Douchi T, Kuwahata R, Matsuo T, Uto H, Oki T, Nagata Y. Relative contribution of lean and fat mass component to bone mineral density in males. J Bone Miner Metab 2003;21(1):17–21. [4] Douchi T, Matsuo T, Uto H, Kuwahata T, Oki T, Nagata Y. Lean body mass and bone mineral density in physically exercising postmenopausal women. Maturitas 2003; 45(3):185–90. [5] Liu JM, Zhao HY, Ning G, Zhao YJ, Zhang LZ, Sun LH, et al. Relationship between body composition and bone mineral density in healthy young and premenopausal Chinese women. Osteoporos Int 2004;15(3):238–42. [6] Reid IR, Evans MC, Ames RW. Volumetric bone density of the lumbar spine is related to fat mass but not lean mass in normal postmenopausal women. Osteoporos Int 1994;4(6):362–7. [7] Reid IR, Ames R, Evans MC, Sharpe S, Gamble G, France JT, et al. Determinants of total body and regional bone mineral density in normal postmenopausal women—a key role for fat mass. J Clin Endocrinol Metab 1992;75(1):45–51. [8] Gnudi S, Sitta E, Fiumi N. Relationship between body composition and bone mineral density in women with and without osteoporosis: relative contribution of lean and fat mass. J Bone Miner Metab 2007;25(5):326–32. [9] Ijuin M, Douchi T, Matsuo T, Yamamoto S, Uto H, Nagata Y. Difference in the effects of body composition on bone mineral density between pre- and postmenopausal women. Maturitas 2002;43(4):239–44.

152

K. Zhu et al. / Bone 74 (2015) 146–152

[10] Zhu K, Briffa K, Smith A, Mountain J, Briggs AM, Lye S, et al. Gender differences in the relationships between lean body mass, fat mass and peak bone mass in young adults. Osteoporos Int 2014;25(5):1563–70. [11] Kohrt WM, Barry DW, Schwartz RS. Muscle forces or gravity: what predominates mechanical loading on bone? Med Sci Sports Exerc 2009;41(11):2050–5. [12] Hamrick MW, Ferrari SL. Leptin and the sympathetic connection of fat to bone. Osteoporos Int 2008;19(7):905–12. [13] Reid IR. Fat and bone. Arch Biochem Biophys 2010;503(1):20–7. [14] Braun T, Schett G. Pathways for bone loss in inflammatory disease. Curr Osteoporos Rep 2012;10(2):101–8. [15] Gjesdal CG, Halse JI, Eide GE, Brun JG, Tell GS. Impact of lean mass and fat mass on bone mineral density: The Hordaland Health Study. Maturitas 2008;59(2):191–200. [16] James A, Hunter M, Straker L, Beilby J, Bucks R, Davis T, et al. Rationale, design and methods for a community-based study of clustering and cumulative effects of chronic disease processes and their effects on ageing: The Busselton Healthy Ageing Study. BMC Public Health 2013;13:936. [17] WHO. Obesity: preventing and managing the global epidemic. Report of a WHO Consultation. WHO Technical Report Series, 894. Geneva: World Health Organization; 2000. [18] IPAQ. http://www.ipaq.ki.se/scoring.htm. [Last accessed 21st December 2014]. [19] Price RA, Sorensen TI, Stunkard AJ. Component distributions of body mass index defining moderate and extreme overweight in Danish women and men. Am J Epidemiol 1989;130(1):193–201. [20] Kleinbaum DG, Kupper LL, Muller KE, Nizam A. Applied Regression Analysis and Other Multivariable Methods. Pacific Grove, USA: Duxbury Press; 1998. [21] Ho-Pham LT, Nguyen UD, Nguyen TV. Association between lean mass, fat mass, and bone mineral density: a meta-analysis. J Clin Endocrinol Metab 2014;99(1):30–8. [22] Lee K, Jessop H, Suswillo R, Zaman G, Lanyon L. Endocrinology: bone adaptation requires oestrogen receptor-alpha. Nature 2003;424(6947):389. [23] Callewaert F, Bakker A, Schrooten J, Van Meerbeek B, Verhoeven G, Boonen S, et al. Androgen receptor disruption increases the osteogenic response to mechanical loading in male mice. J Bone Miner Res 2010;25(1):124–31. [24] Muller M, den Tonkelaar I, Thijssen JH, Grobbee DE, van der Schouw YT. Endogenous sex hormones in men aged 40–80 years. Eur J Endocrinol 2003;149(6):583–9. [25] Vandenput L, Mellstrom D, Karlsson MK, Orwoll E, Labrie F, Ljunggren O, et al. Serum estradiol is associated with lean mass in elderly Swedish men. Eur J Endocrinol 2010;162(4):737–45.

[26] Bolland MJ, Grey AB, Ames RW, Mason BH, Horne AM, Gamble GD, et al. Determinants of vitamin D status in older men living in a subtropical climate. Osteoporos Int 2006;17(12):1742–8. [27] Bolland MJ, Grey AB, Ames RW, Horne AM, Gamble GD, Reid IR. Fat mass is an important predictor of parathyroid hormone levels in postmenopausal women. Bone 2006;38(3):317–21. [28] Pijl H, Langendonk JG, Burggraaf J, Frolich M, Cohen AF, Veldhuis JD, et al. Altered neuroregulation of GH secretion in viscerally obese premenopausal women. J Clin Endocrinol Metab 2001;86(11):5509–15. [29] Giustina A, Mazziotti G, Canalis E. Growth hormone, insulin-like growth factors, and the skeleton. Endocr Rev 2008;29(5):535–59. [30] Blaak E. Gender differences in fat metabolism. Curr Opin Clin Nutr Metab Care 2001; 4(6):499–502. [31] De Laet C, Kanis JA, Oden A, Johanson H, Johnell O, Delmas P, et al. Body mass index as a predictor of fracture risk: a meta-analysis. Osteoporos Int 2005;16(11):1330–8. [32] Leslie WD, Lix LM, Yogendran MS, Morin SN, Metge CJ, Majumdar SR. Temporal trends in obesity, osteoporosis treatment, bone mineral density, and fracture rates: a population-based historical cohort study. J Bone Miner Res 2014;29(4): 952–9. [33] Radavelli-Bagatini S, Zhu K, Lewis JR, Prince RL. Dairy food intake, peripheral bone structure and muscle mass in elderly ambulatory women. J Bone Miner Res 2014;29(7):1691–700. [34] Radavelli-Bagatini S, Zhu K, Lewis JR, Dhaliwal SS, Prince RL. Association of dairy intake with body composition and physical function in older community-dwelling women. J Acad Nutr Diet 2013;113(12):1669–74. [35] Gomez-Cabello A, Ara I, Gonzalez-Aguero A, Casajus JA, Vicente-Rodriguez G. Effects of training on bone mass in older adults: a systematic review. Sports Med 2012; 42(4):301–25. [36] Bolam KA, van Uffelen JG, Taaffe DR. The effect of physical exercise on bone density in middle-aged and older men: a systematic review. Osteoporos Int 2013;24(11): 2749–62. [37] Dimitri P, Bishop N, Walsh JS, Eastell R. Obesity is a risk factor for fracture in children but is protective against fracture in adults: a paradox. Bone 2012;50(2):457–66.