Fat and bone in children: Differential effects of obesity on bone size and mass according to fracture history

Abstract

Fat mass predicts bone accrual in prepubertal children, but obese children have increased fracture risk. We hypothesised that bone size and mass would vary according to prior fracture in obese children. One hundred and three children (52 obese) underwent dual-energy X-ray absorptiometry (DXA) scanning of the lumbar spine, total body, and radial metaphysis and diaphysis. We derived body size–adjusted bone mineral density (BMD) estimates for each site using commonly employed procedures. Following adjustment for either age, age² and weight, or height and weight based on a reference group of nonobese controls without previous fracture, obese children with prior fracture showed a 0.8 to 1.2 SD reduction in total body areal BMD (aBMD), a 3.0 SD decrease in lumbar (L2–4) aBMD, and a 2.0 SD reduction in radial shaft aBMD. These changes were significant at p < .005. Lumbar volumetric BMD (vBMD) calculated by Carter and Kröger algorithms was significantly reduced in obese children with prior fracture (2.0 to 3.3 SD). Eighteen percent of obese children fulfilled the criteria for osteoporosis. Despite greater lean mass for height in obese children (p < .0001), total body bone mineral content (BMC) for lean mass was reduced (p = .002). Multiple regression models adjusting for height, weight, and gender demonstrated an inverse relationship between total body fat mass and total body, lumbar, and ultradistal radius BMC and aBMD. The data suggest that fat mass substantially inhibits bone accrual in children with prior fracture. These children may require targeted interventions to increase bone mass during adolescence to achieve optimal peak bone mass and reduce the risk of osteoporosis later in life. © 2010 American Society for Bone and Mineral Research

Introduction

Childhood obesity is now a global epidemic. The International Obesity Task Force reported that 1 in 10 children worldwide is overweight (above the 85th centile for body mass index), a total of 155 million. Up to 30 to 45 million of these are classified as obese (above 2.67 SD scores for body mass index), accounting for 2% to 3% of the world's 5 to 17 year olds.1, 2

Fracture rates in childhood appear to be increasing,3 and fractures are now more common in boys than at any age up to 85 years in men. Childhood fractures are associated with short-term morbidity and may result in bone deformity or osteoarthritis in the long term.

In adults, most studies support a positive relationship between fat mass and bone.4, 5 Positive associations of fat mass with bone mass accretion in prepubertal children have been reported from the UK and New Zealand,6, 7 but both case-control and prospective cohort studies have demonstrated reduced bone mass for body size in children who fracture.3, 8–13 In particular, these relationships have been noted in overweight adolescents, but there has been no distinction between those who are overweight and those who are obese.

Dual-energy X-ray absorptiometry (DXA) is a technique that provides a 2D assessment of bone density from a 3D structure and fails to adjust adequately for changing tissue depth and bone size in growing children. Several approaches have been used to determine the relationship between body size and bone mass assessed by DXA in children. Carter and Kröger introduced algorithms to calculate volumetric bone density of the lumbar vertebrae from DXA-derived data based on the assumption that the vertebrae adopt the approximate form of a cube or cylinder, respectively.14, 15 Others have used adjustments for body size, bone area, skeletal and chronologic age, and pubertal stage. Such adjustments often are reported as a single outcome, and it is unclear whether similar results would be obtained using alternative adjustment procedures.

The objective of this study was to determine the effect of obesity on bone mass in children with and without a history of fracture using a comprehensive methodologic approach incorporating eight widely used adjustment processes.

Methods

Participants were divided into two groups according to Body Mass Index (BMI) percentile based on UK BMI reference charts.16 The study sample consisted of 52 obese children (>99.6th percentile) recruited from referrals to an endocrine clinic and by advertisement in local newspapers and 51 normal-weight children (<85th percentile) who were attending Sheffield Children's Hospital for minor surgical procedures of the ears, nose, and throat. The study was given ethical approval by South Sheffield Local Research Ethics Committee. Written informed consent was obtained from all participants.

Subjects underwent a structured history and examination. Subjects were excluded if they had any of the following: acute illness leading to inflammatory changes, metabolic bone disease, chronic illness, endocrine or known chromosomal abnormalities, an eating disorder, or use of any steroid-based or other medications known to alter bone metabolism. Details of fracture history (where present) were recorded. These were then cross-referenced against radiographs and subsequent reports by senior radiologists to verify the region of fracture reported by the subjects. Children also were excluded if their fracture had occurred in the 6 months prior to the study in order to avoid any influence of immobility on outcomes. Tanner pubertal staging was undertaken by direct examination by a trained single observer. Exercise was quantified using a validated exercise questionnaire17 as metabolic equivalents (METs) on the basis of a week's recalled exercise history. METs are multiples of resting metabolic rate representing a value of the energy cost of activities. We generated a METs score for each child by adding the reported weekly frequencies for each of three categories of exercise (strenuous, moderate, or mild) in which the child habitually participated for more that 15 minutes, after multiplying each category by its corresponding MET score (9 = strenuous; 5 = moderate; or 3 = mild, respectively).

Anthropometry was undertaken with the subjects wearing light clothing. Height was measured using a portable stadiometer (SECA 214 portable stadiometer, Birmingham, UK) to the nearest completed 1 mm and weight to the nearest 0.1 kg using electronic balance scales (SECA 770 digital weighing scales, Birmingham, UK). Body Mass Index (BMI) was calculated as weight (kg)/height (m²). BMI SD score was calculated using the UK reference values produced by the Child Growth Foundation.18 Waist measurements were taken with a waist circumference measure (SECA 200 waist circumference measure, Birmingham, UK) to the nearest 1 mm.

Bone and body composition measurements were derived by DXA using a Lunar Prodigy instrument and software Version 4.0 (Madison WI, USA). We measured total body, lumbar spine (L2–4), ultradistal radius, and 33% radius (distal third of the radial diaphyseal shaft) bone mineral content (BMC, g), bone area (BA, cm²), and areal bone mineral density (aBMD, g/cm²).19 The DXA instrument also provided information on total body fat (g) and lean mass (g). Bone age was assessed by a single trained rater using a radiograph of the left wrist using the Tanner-Whitehouse 3 method.20

Statistical analysis

The study was powered to detect a difference of 0.85 SD score in body size–adjusted bone mass between the lean and obese groups based on our previously collected data13 with 90% power at the 5% significance level. Where outcome data were skewed, appropriate transformations to create normal distributions were undertaken prior to analysis. Nonobese children without a prior fracture were used as the reference group. We further divided the groups according to the prior history of fracture because of the known association of fracture with reduced bone size and mass after adjusting for body size.3, 13

Intergroup comparisons of obese and nonobese children were made using Student's two-sample t test. Analyses undertaken when the groups were divided by obesity and prior fracture were performed using a linear model with group as a categorical variable and analysis of variance of the group means. Alternatively, analysis of variance of the group means was used to determine differences in measures between obese and nonobese children divided by a history of fracture. A significance level of p < .05 was used for all analyses.

A series of methodologies to determine differences in total body and regional bone mass relative to body size between obese and nonobese children was employed. The Carter and Kröger algorithms were applied to the lumbar spine data to generate “bone mineral apparent density” (BMAD, Carter) or BMD_vol (Kröger). Calculation of BMAD from the Carter model is made using the algorithm BMAD = BMC/BA^1.5. Using the Kröger model, BMD_vol was calculated as BMD_vol = BMC/volume = BMD_areal × [4/(π × width)], where width = mean width of vertebral body.

Using the nonobese children without previous fracture as a reference group, height- and weight-adjusted Z-scores were created based on the methods previously adopted by Clark (for total body less head) and Manias (for the spine and total body).3, 13, 21 In addition, Z-scores were created using the same reference group and correcting for age, age squared, and weight based on regression models created in a previous study by Goulding and colleagues.22 The Goulding and Manias methods included DXA head measurements in the analysis of total body bone analyses.

In concordance with the two-staged approach used by Crabtree and colleagues,23Z-scores based on controls were used to compare lean body mass for height and BMC for lean mass in obese and nonobese children divided according to prior fracture. Finally, relationships between fat mass and total and regional bone measures were determined by using multiple regression including height, weight, gender, and skeletal maturity as covariates. Linearity and normality of residuals for regression models were assessed by histogram. All analyses were undertaken in DataDesk Version 6.1 (Data Description, Inc., Ithaca, NY, USA).

Results

Basic demographic and anthropometric data are shown in Table 1. In addition to the expected differences in body size parameters based on the selection criteria, obese children were older than nonobese children (p = .02) and had more mature skeletons by TW3 bone age assessment (13.1 versus 10.9 years, p = .0006).

Table 1. Comparison of Basic Anthropometry in the Lean and Obese Groups With and Without Subdivision by Prior Fracture

	Age (years)	Weight SDS	Height SDS	BMI SDS
Values are mean (SD). a Two-sample t test. b ANOVA. c Indicates the most divergent group mean(s).
Lean (n = 51)	10.6 (3.2)	0.36 (1.09)	0.42 (1.30)	0.2 (1.0)
Obese (n = 52)	12.1 (2.9)	3.46 (0.66)	1.34 (1.09)	3.3 (0.6)
p valuea	.02	<.0001	.0002	<.0001
Lean, no fracture (n = 38): Group 1	10.0 (3.2)c	0.32 (1.17)	0.34 (1.36)c	0.24 (1.12)
Lean, prior fracture (n = 13): Group 2	12.3 (2.7)	0.48 (0.85)	0.64 (1.12)	0.25 (0.89)
Obese, no fracture (n = 39): Group 3	11.7 (3.0)	3.48 (0.68)c	1.33 (1.04)	3.30 (0.61)c
Obese, prior fracture (n = 13): Group 4	12.8 (2.3)	3.33 (0.66)c	1.22 (1.31)	3.21 (0.38)c
p valueb	.008	<.0001	.004	<.0001

We were able to verify reported fractures in 20 of 26 children by radiographs and subsequent reports by senior radiologists. All verbal reports matched with the available radiographs. Of the 26 children who reported fractures, 27% (7/26) reported more than one fracture. Fracture of the distal radius was reported most frequently (54%), followed by fingers (18%), toes (9%), scaphoid (6%), tibia (3%), clavicle (3%), calcaneus (3%), and supracondylar fracture of humerus (3%).

Unadjusted bone area, BMC, and areal BMD (aBMD) for the total body including/less head, lumbar spine, and radial sites and unadjusted data for volumetric BMD (vBMD) of L2–4 calculated by the Carter and Kröger methods are shown by group in Table 2.

Table 2. Unadjusted Data for Total Body (Less Head), Lumbar Spine, 33% Radius, and Ultradistal Radius BMC, BA, and aBMD and Lumbar Spine BMAD as per the Carter and Kröger Models

	Nonobese		Obese		p Value
	Group 1, no fracture (n = 38)	Group 2, prior fracture (n = 13)	Group 3, no fracture (n = 39)	Group 4, prior fracture (n = 13)
All data are expressed as mean (SD). Significance was derived from ANOVA (F test) of the means (significance is defined as p ≤ .05). * Indicates the most divergent group mean(s).
Age	10.0 (3.2)*	12.3 (2.7)	11.7 (3.0)	12.8 (2.3)	.008
Total body (including head)
BA (cm^2)	1431 (436)*	1710(421)	1912 (424)	1941 (230)	<.0001
BMC (g)	1378 (589)*	1757 (651)	2106 (686)	2053 (361)	<.0001
aBMD (g/cm^2)	0.93 (0.13)*	1.00 (0.14)	1.08 (0.13)	1.05 (0.08)	<.0001
Total body (less head)
BA (cm^2)	1228 (432)*	1501 (417)	1696 (428)	1733 (236)	<.0001
BMC (g)	1050 (531)*	1412 (608)	1721 (644)	1679 (355)	<.0001
aBMD (g/cm^2)	0.81 (0.14)*	0.90 (0.16)	0.98 (0.14)	0.96 (0.10)	<.0001
Lumbar spine
BA (cm^2)	26.9 (7.2)*	33.2 (8.6)	34.6 (8.4)	35.6 (5.5)	.0003
BMC (g)	24.2 (12.1)*	31.8 (16.3)	37.0 (14.9)	34.3 (8.8)	.002
aBMD (g/cm^2)	0.85 (0.21)*	0.91 (0.22)	1.04 (0.20)	0.94 (0.12)	.001
BMAD (g/cm^3) (Carter)	0.16 (0.02)	0.16 (0.02)	0.18 (0.03)*	0.16 (0.01)	.007
BMDvol (g/cm^3) (Kröger)	0.32 (0.05)	0.32 (0.05)	0.36 (0.05)*	0.31 (0.03)	.005
Radius
UDRBA (cm^2)	2.51 (0.43)*	2.84 (0.38)	2.84 (0.51)	3.09 (0.34)	.002
UDRBMC(g)	0.93 (0.34)*	1.07 (0.43)	1.32 (0.41)	1.32 (0.24)	.0005
UDRaBMD (g/cm^2)	0.36 (0.08)*	*0.37 (0.10)	0.45 (0.07)	0.43 (0.07)	<.0001
R33%BA (cm^2)	2.26 (0.37)*	2.27 (028)	2.43 (0.36)	2.55 (0.23)	.06
R33%BMC(g)	1.40 (0.44)*	1.63 (0.48)	1.90 (0.47)	1.84 (0.30)	.0003
R33%aBMD (g/cm^2)	0.62 (0.14)*	0.70 (0.15)	0.77 (0.10)	0.72 (0.07)	<0.0001

Preliminary analyses showed, as expected, that bone area, mineral content, and aBMD estimates were strongly influenced by age, age squared, height, and weight, as were the calculated volumetric measures using the Carter and Kröger procedures. All the estimates thus are shown as raw data and then as Z-scores following the adjustment procedures described previously.

Since nonobese children without prior fracture had been chosen as the reference group for this study, age-adjusted values derived from our study population were compared with the quoted manufacturer age-adjusted Z-scores for the total body and spine (L2–4) from the same group. There are currently no age-adjusted Z-scores for the radius quoted by the manufacturer. Total body and lumbar spine manufacturer age-adjusted Z-scores for nonfractured controls were 0.22 (0.99) and 0.21 (1.36), respectively. Therefore, we felt that our control population was truly representative of healthy controls.

Table 3 shows the total body Z-scores by group for each measurement following adjustment for (1) age, (2) age, age squared, and weight (Goulding method), (3) weight and height (Manias method), and (4) height and weight on total body less head (Clark method). Adjusting for body size using methods previously devised by Manias, Clark, and Goulding suggested a reduction in total body BMC, size, and aBMD in obese children with a prior history of fracture compared with other children. These results all reached statistical significance apart from the value for total body aBMD derived from the Goulding method.

Table 3. Mean (SD) Z-Scores for BA, BMC, and aBMD of the Total Body Following Adjustment for (1) Age, (2) Age, Age Squared, and Weight (Goulding Method), (3) Weight and Height (Manias Method), and (4) Height and Weight for Total Body Bone Values less Head (Clark Method) by Group

Z-scores	Nonobese		Obese		p Value
	Group 1, no fracture (n = 38)	Group 2, prior fracture (n = 13)	Group 3, no fracture (n = 39)	Group 4, prior fracture (n = 13)
Significance was derived from ANOVA (F test) of the means (significance is defined as p ≤ .05). * Indicates the most divergent group mean(s).
Total body adjusted for age
BA	0 (1.00)	0.10 (0.71)	1.32 (0.95)*	0.82 (0.88)	<.0001
BMC	0 (1.00)	0.09 (0.68)	1.54 (0.92)*	0.84 (0.94)	<.0001
aBMD	0 (1.00)	0 (0.87)	1.50 (1.00)*	0.62 (1.01)	<.0001
Goulding method
BA	0 (1.00)	−0.06 (0.91)	−2.78 (2.83)	−3.49 (2.27)*	<.0001
BMC	0 (1.00)	−0.05 (1.07)	−1.73 (2.53)	−2.83 (1.90)*	<.0001
aBMD	0 (1.00)	−0.06 (1.15)	0.31 (1.73)	−0.82 (1.34)	.12
Manias method
BA	0 (1.00)	−0.56 (1.00)	−1.46 (3.16)	−3.17 (2.74)*	.0007
BMC	0 (1.00)	−0.38 (1.22)	−0.77 (2.31)	−2.50 (2.07)*	.0009
aBMD	0 (1.00)	−0.15 (1.23)	0.07 (1.46)	−1.24 (1.39)*	.02
Clark method
BA	0 (1.00)	−0.44 (1.18)	−1.15 (2.51)	−2.98 (2.24)*	.0002
BMC	0 (1.00)	−0.65 (0.93)	−1.65 (3.38)	−3.20 (2.71)*	.0009
BMD	0 (1.00)	−0.08 (1.15)	−0.29 (1.41)	−1.83 (1.51)*	.0005

Table 4 shows the lumbar Z-scores by group for each measurement following adjustment for (1) age, (2) age, age squared, and weight (Goulding method), and (3) weight and height (Manias method). Figure 1 demonstrates the differences in BMAD among the four groups using each adjustment. Body size–adjusted areal and volumetric lumbar bone using the Manias and Goulding methods was reduced in obese children, particularly those with a history of prior fracture.

Table 4. Mean (SD) Z-Scores for Lumbar (L2–4) Spine Following Adjustment for (1) Age, (2) Age, Age Squared, and Weight (Goulding Method), and (3) Weight and Height (Manias Method)

Z-score	Nonobese		Obese		p Value
	Group 1, no fracture (n = 38)	Group 2, prior fracture (n = 13)	Group 3, no fracture (n = 39)	Group 4, prior fracture (n = 13)
Z-scores for lumbar volumetric density (BMAD) derived from the Carter and Kröger equations are given for each method. Significance was derived from ANOVA (F test) of the means (significance is defined as p ≤ .05). * Indicates the most divergent group mean(s).
Lumbar (L2–4) adjusted for age
BA	0 (1.00)	0.56 (1.20)	1.49 (1.17)*	1.11 (1.01)	<.0001
BMC	0 (1.00)	0.04 (0.99)	1.32 (1.02)*	0.44 (0.95)	<.0001
aBMD	0 (1.00)	−0.34 (0.80)	1.02 (1.01)*	−0.14 (0.87)	<.0001
BMAD (Carter)	0 (1.00)	−0.66 (0.73)	0.46 (1.46)	−0.67 (0.81)*	.004
BMDvol (Kröger)	0 (1.00)	−0.50 (0.79)	0.56 (1.15)*	−0.59 (0.74)	.0008
Goulding method
BA	0 (1.00)	0.57 (1.34)	−0.37 (1.38)	−0.80 (1.41)*	.04
BMC	0 (1.00)	−0.08 (1.23)	−1.39 (1.97)	−2.54 (1.52)*	<.0001
aBMD	0 (1.00)	−0.56 (0.97)	−1.64 (2.32)	−3.14 (1.45)*	<.0001
BMAD (Carter)	0 (1.00)	−0.84 (0.81)	−1.75 (2.95)	−3.05 (1.32)*	<.0001
BMDvol (Kröger)	0 (1.00)	−0.64 (0.72)	−1.30 (2.33)	−2.70 (1.18)*	<.0001
Manias method
BA	0 (1.00)	0.23 (1.22)	0.63 (1.13)	0.28 (0.95)	.12
BMC	0 (1.00)	−0.30 (1.34)	−0.42 (1.86)	−1.82 (1.34)*	<.0001
BMD	0 (1.00)	−0.61 (1.21)	−1.08 (1.96)	−3.02 (1.64)*	<.0001
BMAD (Carter)	0 (1.00)	−0.84 (0.90)	−0.73 (1.99)	−1.98 (0.97)*	.0016
BMDvol (Kröger)	0 (1.00)	−0.53 (1.02)	−1.48 (2.17)	−3.31 (1.62)*	<.0001

Figure 1. Four group boxplot comparison of BMAD (Carter) Z-scores following adjustment for (1) age, (2) age, age squared, and weight (Goulding method), and (3) weight and height (Manias method). Bars show medians; box edges, interquartile ranges; and whiskers, the main body of data. O = outliers. *Denotes extreme outliers. Gray area is median ± 95% confidence interval. Groups are (1) nonobese, no previous fracture; (2) nonobese, history of prior fracture; (3) obese, no previous fracture; and (4) obese, history of prior fracture.

Table 5 shows the ultradistal and 33% radius Z-scores by group for each measurement following adjustment for (1) age, (2) age, age squared, and weight (Goulding method), and (3) weight and height (Manias method). Size adjustment using the Goulding or Manias methods showed a similar reduction in radial diaphyseal BMC and aBMD in obese children, which was more profound in obese children with prior fracture. Despite a significant increase in weight- and height-adjusted radial metaphyseal bone area, aBMD was significantly reduced in obese children with prior fracture compared with the other groups.

Table 5. Mean (SD) of Ultradistal and 33% Radius Adjusted for (1) Age, (2) Age, Age Squared, and Weight (Goulding Method), and (3) Weight and Height (Manias Method)

Z-score	Nonobese		Obese		p Value
	Group 1, no fracture (n = 38)	Group 2, prior fracture (n = 13)	Group 3, no fracture (n = 39)	Group 4, prior fracture (n = 13)
Significance was derived from ANOVA (F test) of the means (significance is defined as p ≤ .05). * Indicates the most divergent group mean(s).
Ultradistal radius and 33% radius adjusted for age
UDBA	0 (1.00)	0.32 (0.80)	0.79 (1.14)	1.15 (1.09)*	.006
UDBMC	0 (1.00)	−0.39 (1.24)	1.40 (1.06)*	0.96 (1.39)	<.0001
UDaBMD	0 (1.00)	−0.59 (1.03)	1.26 (0.91)*	0.45 (1.47)	<.0001
R33%BA	0 (1.00)	−0.35 (0.69)	0.23 (0.96)	0.53 (0.73)	.15
R33%BMC	0 (1.00)	−0.30 (0.95)	1.06 (1.12)*	0.70 (1.30)	.0004
R33%BMD	0 (1.00)	−0.25 (0.99)	1.18 (0.81)*	0.29 (1.07)	<.0001
Goulding method
UDBA	0 (1.00)	0.30 (0.74)	−0.02 (1.30)	0.59 (1.44)	.42
UDBMC	0 (1.00)	−0.52 (1.42)	−0.48 (1.81)	−0.75 (2.06)	.49
UDaBMD	0 (1.00)	−0.71 (1.20)	−0.46 (1.74)	−1.28 (2.07)	.11
R33%BA	0 (1.00)	−0.38 (0.77)	−0.35 (1.16)	0.12 (0.86)	.41
R33%BMC	0 (1.00)	−0.51 (1.33)	−1.22 (1.65)	−1.47 (1.87)*	.004
R33%BMD	0 (1.00)	−0.43 (1.37)	−0.91 (1.51)	−2.01 (1.55)*	.0006
Manias method
UDBA	0 (1.00)	−0.19 (0.61)	1.12 (1.94)	1.63 (1.09)*	.0009
UDBMC	0 (1.00)	−0.77 (1.49)	−0.38 (1.79)	−0.67 (1.77)	.44
UDaBMD	0 (1.00)	−0.66 (1.34)	−1.29 (1.71)	−2.02 (2.39)*	.003
R33%BA	0 (1.00)	−0.62 (0.73)	0.05 (1.15)	0.51 (0.68)	.09
R33%BMC	0 (1.00)	−0.97 (1.40)	−0.91 (1.50)	−1.27 (1.90)*	.02
R33%BMD	0 (1.00)	−0.68 (1.38)	−1.12 (1.55)	−2.20 (1.60)*	.0001

Adjusting for age only, total body and regional BMC and aBMD and lumbar BMAD were significantly greater in obese children who had not previously fractured compared with nonobese children. Lumbar vBMD (Carter and Kröger) adjusted for age was reduced in both nonobese and obese children with a prior history of fracture (p = .004 and p = .0008. respectively).

We divided the four groups by gender to identify gender-specific variations. This rendered the group sizes too small to determine valid statistical differences. However, groups divided by gender followed the same trends as those observed for total-body and site-specific bone measurements presented using the Manias and Clark methods. Ultradistal radius BMC and aBMD were lower in female obese patients with a prior history of fracture using the Goulding method. We did not observe the same trend at the ultradistal radius for boys.

Lean mass was significantly greater in obese compared with nonobese children (41.6 ± 12.5 kg versus 28.1 ± 9.8 kg, p < .0001). However, there was no significant difference in total body lean mass between the fracture and nonfracture obese children (42.8 ± 9.6 kg versus 40.0 ± 13.3 kg, respectively, p = .42). Using Crabtree's method, log-transformed lean mass relative to height was significantly greater in obese children (p < .0001), although, again, there was no significant difference in lean mass for height between the two obese groups (p = .87; Fig. 2A). In contrast, despite a greater lean mass relative to height, obese children, particularly those with a prior history of fracture, had a lower log-transformed total body (less head) BMC for lean mass compared with nonobese children without a prior history of fracture (p = .004; see Fig. 2B). Obese children who had not previously fractured had the lower log-transformed METs scores compared with lean children and obese children who had previously fractured (p = .0002).

Figure 2. (a) Log-transformed lean mass relative to height. (b) Log-transformed BMC relative to lean mass for the four groups. Bars show medians; box edges, interquartile ranges; and whiskers, the main body of data. Gray area is median ± 95% confidence interval. Groups are (1) nonobese, no previous fracture; (2) nonobese, history of prior fracture; (3) obese, no previous fracture; and (4) obese, history of prior fracture.

Finally, following correction for height, weight, and gender, multiple-regression models analyzing the relationship between total body and regional bone measurements and body composition revealed a negative relationship between total body fat mass and total body BMC and aBMD; lumbar BA, BMC, and aBMD; and ultradistal radius BMC, BA, and aBMD. Lean mass was positively related to all bone measurements except total body BA, lumbar vBMD, and 33% radius aBMD (Table 6).

Table 6. Relationship Between Log-Transformed Total Body Fat Mass, Truncal Fat Mass, and Lean Mass and Log-Transformed Total and Regional Bone Measures in Multiple Regression Models Accounting for Gender and Log-Transformed Height and Weight at Baseline

	Total body fat mass (loge)	Lean mass (loge)
The strength of the relationship is expressed as the coefficient (adjusted r^2 value, p value). The null hypothesis is rejected at p ≤ .05.
Total body (less head)
BA (loge)	−0.05 (0.90, 0.16)	0.15 (0.90, 0.21)
BMC (loge)	−0.11 (0.90, 0.05)	0.39 (0.90, 0.04)
aBMD (loge)	−0.06 (0.78, 0.02)	0.26 (0.80, 0.002)
Lumbar spine
BA (loge)	−0.18 (0.94, <0.0001)	0.62 (0.95, <0.0001)
BMC (loge)	−0.27(0.86, <0.0001)	0.94 (0.86, <0.0001)
aBMD (loge)	−0.09 (0.64, 0.04)	0.36 (0.65, 0.02)
BMAD (loge) (Carter)	−0.001(0.12, 0.98)	0.01 (0.12, 0.94)
BMDvol (loge) (Kröger)	−0.004 (0.21, 0.93)	0.04 (0.21, 0.80)
Radius
UDRBA (loge)	−0.10 (0.76, 0.002)	0.49 (0.79, <0.0001)
UDRBMC (loge)	−0.22 (0.70, 0.003)	1.21 (0.77, <0.0001)
UDRaBMD (loge)	−0.11 (0.50, 0.05)	0.73 (0.55, 0.0003)
R33%BA (loge)	−0.06 (0.44, 0.14)	0.50 (0.50, 0.0009)
R33%BMC (loge)	−0.08 (0.81, 0.11)	0.67 (0.84, 0.0002)
R33%aBMD (loge)	0.02 (0.76, 0.67)	0.17 (0.77, 0.21)

Discussion

It seems self-evident that bone size in children should be influenced both by growth intrinsically and by mechanical forces extrinsically. Adjustments for body size are likely to be imperfect in capturing both elements. Nevertheless, such procedures in growing children are used frequently when assessing skeletal size and mass outcomes. By applying all those in common use, we hoped to demonstrate consistent patterns of difference in bone size and mass between children who were or were not obese and with and without prior fracture. Molgaard's approach (which uses age-banded normative data for each variable) could not be applied to our data set because we do not have sufficient reference values across the age range of the older obese children using the Lunar Prodigy instrument.24

We found a consistent pattern of differences in bone mass at different skeletal sites according to both obesity and prior fracture. In essence, children who fracture have narrower bones. Thus fracture is a marker for altered bone architecture. The novel finding in this study is that this difference is substantially increased in children who have a history of fracture and are obese. This may explain the consistent finding across many studies that obese children are over represented in fracture studies.10, 13, 25

We have clearly demonstrated that estimates of total body, lumbar, and radial bone mass when adjusted for body size were reduced in children with prior fracture compared with those who were fracture-free, and the presence of obesity increased the size of this difference. In addition, body size–adjusted lean mass was positively and fat mass negatively related to bone mass and size at the radius, spine, and total body, although not to lumbar spine vBMD.

The recent pediatric position statements from the International Society of Clinical Densitometry define children as being osteoporotic when their body size–adjusted bone mass falls more than 2 SD scores below the mean and when there is a history of one or more lower limb or spine fractures or two or more upper limb fractures.19 According to this definition, nine (18%) of the obese children were osteoporotic.

Lean mass was similar between the obese children with and without fracture, and reported physical activity was increased in those with fracture, including obese children. Muscle mass relative to height also was significantly greater in obese compared with nonobese children, yet BMC relative to lean mass was significantly lower in obese children. This reduction was more pronounced in obese children with prior fracture. According to the two-staged approach proposed by Crabtree, a normal total body lean mass for body size but significantly reduced BMC for lean mass suggests a primary bone abnormality.26 The mechanostat model proposed by Harold Frost in the 1960s suggests that the growing skeleton is sensitive to mechanical strain and responds by increasing periosteal apposition. This results in wider bones and increased trabecular bone mass.27 With a reduction in total body BA relative to body size and BMC relative to lean mass in obese children despite relatively normal or increased levels of physical activity, we speculate that obesity impairs the normal response of the growing skeleton to mechanical loading, effectively resulting in an intrinsic bone abnormality, and that this effect is greater in obese children with prior fracture. In the mechanostat system, this would equate to increasing the threshold at which bone responded to mechanical stimulation.

Others have reported that children who fracture, particularly those with recurrent fracture, are more likely to be obese. Manias and colleagues indicated that children with recurrent fractures had a higher BMI than children with only one or no fractures.13 In a number of studies, Goulding and colleagues have demonstrated similar findings, suggesting that obese children are more likely to fracture owing to a reduction in total body and regional bone mass.8, 11, 25, 28, 29 Furthermore, prospective analysis has shown that the detrimental effect of obesity on bone during a critical phase of bone mass accrual may persist up to 4 years from initial observations.30 Using densitometry, Afghani and colleagues also have demonstrated a reduction in total body BMC and BA relative to truncal adiposity.31

The response of the spine to loading is of particular importance in children because spinal bone maturation relative to the lower limbs is delayed during rapid periods of growth.32, 33 Following correction of lumbar vBMD for age, we demonstrated a significant reduction in estimated BMD in both fracture groups. We also showed a significant reduction in lumbar BMC and aBMD and vBMD (Goulding and Manias methods) relative to body size in obese children, particularly those with prior fracture. The reduction in vBMD observed in obese children in our study is supported by other studies using densitometry.12, 25

Bone area at both the shaft and ultradistal radius was increased in the obese children, more so in obese children who have previously fractured both before and after adjusting for body size. Despite the increase in radial bone size, the BMC in the body size–adjusted models was lower at both sites in the obese children compared with the control group. Areal BMD thus was inevitably lower at both sites. It is quite possible that some of the differences at the radial sites between obese and nonobese children are in part a reflection in differences in skeletal maturity. The obese children were skeletally more mature, as shown by their significantly greater BAs. As puberty proceeds, children grow in the appendicular and subsequently axial skeleton. Nevertheless, skeletal maturity was similar in the two groups of obese children, yet the size-adjusted aBMD was lower in those with prior fracture at both the midshaft and ultradistal radial sites. This suggests that some of the difference may be attributable to the interactive effect of obesity on bones that are in some way predisposed to fracture.

It is impossible to accurately determine from densitometry whether the cortical or trabecular compartment is affected by an increase in fat mass. The reduction in bone mass in both the lumbar spine and radial ultradistal site in the obese children suggests a lower trabecular bone mass. The reduction in total-body and radial shaft bone mass observed in obese children likewise suggests a reduction in the cortical fraction of bone.34 Bone area in the total body was substantially reduced in relation to body size in the obese children, but size-adjusted bone area in the radius was increased in both obese groups, more so in those with a prior fracture. This could reflect a site-specific variation in the effect of obesity and prior fracture on bone size or an adaptive response of bone to perceived weakness. The discrepancy in increased radial bone area relative to a reduced radial BMC following adjustment may reflect thinner cortices that are more likely to fracture under load.

Nagasaki and colleagues demonstrated a negative correlation between percentage fat mass and BMD and observed a reduction in total body aBMD in obese boys over 12 years of age and in girls aged 12 to 15 years.35 Goulding has shown recently that fat mass is an independent predictor of bone size and mass in preschool children,6 despite earlier studies reporting a reduction in body size–adjusted bone mass in older obese children.8, 10, 11, 22, 25, 36 One study using quantitative computed tomography (QCT) has demonstrated a reduction in vertebral cancellous density, but only in obese adolescent males,37 although, in contrast, Wetzsteon and colleagues recently have demonstrated that tibial cortical density measured by peripheral QCT is significantly greater in obese children aged 9 to 11 years.38

The detrimental effect of obesity on bone mass in our study and other studies may be limited to specific growth phases in children. The highest incidence of fracture in childhood is during the pubertal period.39–41 In our study, the mean skeletal age of the obese children was 13 years. Fractures of the distal radius are reported most frequently at this age, as demonstrated by our study and others.42, 43 Peak bone density at the lumbar spine occurs between 11 and 15 years of age in females,32, 44–49 with a significant reduction in bone mass accumulation after this stage.32, 47, 49 In males, it reaches a peak at 16 years and onward,32, 45, 48, 49 with significant gains continuing between 17 and 20 years.32, 49, 50 Only 14 of the 52 obese children recruited were assessed as prepubertal. Obesity appears to act as a significant risk factor for the attainment of peak bone mass during a critical stage of growth and development. However, since our study was cross-sectional, we were not able to assess the effects of increasing age and maturity within individuals directly. The effect on peak bone mass accrual could only be addressed in the context of longitudinal design.

A strength of this study is that both total and regional bone size and mass were considered in evaluating fracture risk in addition to the effects of increased fat mass. However, we recognize that there are limitations in the use of DXA. Peripheral QCT (pQCT) would afford more definitive information on skeletal geometry and the effect of fat mass on the individual cortical and trabecular compartments. A further limitation is that we could obtain previous X-rays and reports for only 20 of the 26 children reporting fractures. However, in all 20, the self-report tallied precisely with the radiologist's report and available radiographs. We thus have confidence that those reporting fractures are doing so appropriately.

In summary, we have demonstrated that total body and regional bone mass relative to body size is reduced in obese children. The size of this effect was greater in obese children with a prior history of fracture. This was so despite having greater lean mass relative to height and having similar physical activity scores to nonobese children. Importantly, the use of several established methods to examine relationships between fat and bone mass in children have yielded the same pattern of reduced bone size and mass in the four groups studied.

Fat mass may limit the response of total body and regional bone mass to the mechanical force induced by muscle, resulting in a reduced skeletal mass relative to body size. In some obese children, this may result in osteoporosis at an age when fracture incidence is highest. Achieving an optimal peak bone mass during adolescence is crucial. The possible detrimental effect of fat on bone accrual in the face of an increasing incidence of obesity has significant implications for future skeletal health.

Disclosures

The authors state that they have no conflicts of interest.

Acknowledgements

This study was funded by the Sheffield Children's Charity, Sheffield, UK, and the British Society for Paediatric Endocrinology and Diabetes.