Proximal femur bone geometry is appropriately adapted to lean mass in overweight children and adolescents

Bone 36 (2005) 568 – 576 www.elsevier.com/locate/bone

Proximal femur bone geometry is appropriately adapted to lean mass in overweight children and adolescents Moira A. Petita,*, Thomas J. Beckb, Justine Shultsc, Babette S. Zemeld, Bethany J. Fosterd, Mary B. Leonardc,d a

Department of Health Evaluation Sciences, Penn State University College of Medicine, 600 Centerview Drive, Suite 2200, A210, Hershey, PA 17033, USA b Department of Radiology, Johns Hopkins University, Baltimore, MD 21287, USA c Department of Biostatistics and Epidemiology, University of Pennsylvania School of Medicine, Philadelphia, PA 19104, USA d Department of Pediatrics, The Children’s Hospital of Philadelphia, University of Pennsylvania School of Medicine, Philadelphia, PA 19104, USA Received 19 August 2004; revised 6 December 2004; accepted 6 December 2004

Abstract It is unclear if the bones of overweight children are appropriately adapted to increased loads. The objective of this study was to compare bone geometry in 40 overweight (body mass index [BMI] N 85th percentile) and 94 healthy weight (BMI V 85th percentile) subjects, ages 4–20 years. Dual energy X-ray absorptiometry (Hologic QDR 2000) scans were analyzed at the femoral shaft (FS) and narrow neck (NN) by the Hip Structure Analysis program. Subperiosteal width, cortical thickness and indices of bone axial and bending strength (bone cross-sectional area [CSA] and section modulus [Z]) were measured from bone mass profiles. Multivariate regression models were used to compare overweight and healthy weight subjects. Z was 11 (95% CI 5, 19) and 13 (7, 20) percent higher at the FS and NN, respectively, in overweight subjects ( P b 0.001), adjusted for height, maturation and gender. At the NN, higher Z was due to greater subperiosteal width [4% (2, 7)] and bone CSA [10% (5, 16]) and at the FS, to higher bone CSA [10% (5, 16)] and thicker cortices [9% (3, 15)]. When lean mass was added to the models, bone variables did not differ between overweight and healthy weight subjects ( P N 0.22), with the exception of NN subperiosteal width [3% (0, 6), P = 0.04]. Fat mass did not contribute significantly to any model. In summary, proximal femur bone geometric strength in overweight children was appropriately adapted to lean mass and height but greater weight in the form of fat mass did not have an independent effect on bone bending strength. These geometric adaptations are consistent with the mechanostat hypothesis that bone strength adapts primarily to muscle forces, not to static loads represented by body weight. D 2004 Elsevier Inc. All rights reserved. Keywords: Bone strength; Bone development; Pediatrics; Body composition; Obesity

Introduction Despite the epidemic of childhood obesity, the influence of body weight on bone strength during growth is poorly understood. Overweight adults have high areal bone mineral density (BMD, g/cm2) and reduced risk of osteoporotic fracture [1,2]. During growth, increased body weight is associated with increased bone mineral content (BMC, g) * Corresponding author. Present address: University of Minnesota, School of Kinesiology, Cooke Hall, 1990 University Avenue SE, Minneapolis, MN 55455, USA. Fax: 612 626 7700. E-mail address: [email protected] (M.A. Petit). 8756-3282/$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.bone.2004.12.003

and bone geometric strength [3], and several authors report greater bone mass in overweight and obese children [4,5]. Yet, recent work suggests that obese children have low bone mass for their weight and are at increased risk for fracture [6–9]. These seemingly contradictory findings might be explained in part if viewed from a mechanical perspective. It is generally accepted that dynamic, rather than static, loads induce the highest osteogenic stimulus [10]. In vivo, bones function mainly as muscle actuated levers (that is, muscles put bones into action; the bone acts as a lever). Most muscle insertions are located close to the joint, ensuring that the forces they generate in any given activity

M.A. Petit et al. / Bone 36 (2005) 568–576

exceeds the action force [11]; and thus induce large strains on bone that exceed those of body weight alone. Thus, in normal loading conditions, the forces on bone are dominated by muscle action [11,12]. These forces tend to exceed those due to body weight because of the inefficient lever action. Total body lean mass can be measured from DXA with reasonable accuracy and is related to muscle crosssectional area by MRI [13]. Muscle cross-sectional area scales with muscle force [14], thus total body lean mass from DXA can be used as a crude index of skeletal load. Obese children have higher lean mass for height [5]. Therefore, obese individuals should have greater bone strength because of the greater muscle forces required to move the higher body weight. Extra weight from fat mass should have no influence on bone mechanical strength independent of lean mass because it is basically a static load. That is, fat mass and body weight per se are osteogenic bloadsQ only in the sense that it takes more muscle force to move the higher weight. Thus, they should have no effect on bone mechanical strength independent of muscle force. Previous studies of the effects of body composition on bone during growth assessed lumbar spine or total body bone mineral content (BMC) or areal BMD [9,15]. While these bone mass and areal density parameters have been clinically useful to predict fractures in adults, they do not fully capture structurally important differences in bone geometry. Load-bearing long bones are designed to be as stable as necessary, not as heavy as possible [16]. Thus, the strength of a bone depends not only on the amount of material present (the bmassQ component), but also on the way that material is distributed (geometry). Bone material properties also contribute to strength, but in the absence of disease, these properties vary little between normal individuals and are not yet possible to measure in vivo. Although the 2 dimensional nature of DXA is an accepted limitation, bone geometry can be measured in the plane of the image from DXA scans [17–19] and can be used to interpret DXA data in a mechanically meaningful way. The objective of this study, therefore, was to compare proximal femur bone geometry in overweight and healthy weight children. We hypothesized that bone geometry adapts primarily to muscle force (represented by lean mass), and that additional weight in the form of fat mass, is not independently associated with bone geometric strength.

Methods Study subjects The study subjects consisted of 134 individuals who had proximal femur and total body DXA scans done as healthy controls in bone studies in the Nutrition and Growth Laboratory at the Children’s Hospital of Philadelphia [5,20]. These subjects, aged 4 to 20 years, were recruited from general pediatric clinics, and the surrounding com-

569

munity using newspaper advertisements and flyers. Individuals were excluded from participation for chronic medical conditions or medications potentially affecting growth, pubertal development, nutritional status or dietary intake. Overweight was not an exclusion criterion. The protocol was approved by the CHOP Institutional Review Board; all participants and/or parents provided written informed consent. Anthropometry and Tanner staging Weight (kg) (digital electronic stand-on scale) and height (cm) (wall-mounted stadiometer) were measured at the time of the DXA scan. Tanner pubertal stage was assessed by physical examination [21]. All measurements were performed by highly trained research staff. Age and gender specific z scores (standard deviation scores) for height, weight and body mass index (BMI) were calculated using the National Center for Health Statistics (NCHS) 2000 Center for Disease Control growth data [22]. Participants with BMI between the 5th and 85th percentiles were classified as bhealthy weightQ; those with BMI greater than the 85th percentile were classified as boverweightQ, consistent with previous work [6,23]. Body composition and bone measurements Total body and proximal femur (non-dominant side) DXA scans (Hologic QDR 2000, Waltham, MA) were performed by two experienced technicians using a fan beam in the array mode using standard positioning techniques [24]. Estimates of bone mineral free lean mass (kg), fat mass (kg) and percent body fat were obtained from the total body DXA scan excluding the skull. A single operator reanalyzed all whole body DXA scans to standardize the analyses. Daily quality control scans were performed with the manufacturer’s phantom [25]. Hip Structure Analysis Loading forces on bones are distributed over the surface of bone material in cross-sections: the soft tissues within bony spaces do not contribute significantly to the strength of bone. The concentrations of loading forces (stresses) are a function of bending moments and bone cross-sectional geometry (i.e. bone surface area and its distribution about the center of mass). The Hip Structure Analysis (HSA) program uses a principle first described by Martin and Burr [17,19] to derive cross-sectional geometry from images acquired from bone mineral scanners. The main structural parameters derived by HSA are the bone cross-sectional area (bone CSA) and the section modulus (Z, cm3)—which are inversely related to stresses due to axial and bending loads, respectively. The HSA software version (v2.1) used in this study averages geometry measurements for a series of 5 parallel

570

M.A. Petit et al. / Bone 36 (2005) 568–576

pixel mass profiles spaces ~1 mm apart along the bone axis. Analysis locations included the narrow neck (NN) region across the narrowest point of the femoral neck, and the proximal femoral shaft (FS) region located at a distance of 1.5 times the width of the femoral neck distal to the intersection of the NN and FS axes. This program and these regions have been described in detail and illustrated in previous publications [26–29]. The basic principal of the method is that a line of pixels across the bone axis is a projection of the corresponding cross-section and measurements of the pixel distribution can be used to measure geometry. As DXA is a 2dimensional technique, geometry is measured in the image of the plane only. A single DXA pixel value sums all mineral mass (hydroxyapatite in g/cm2) along a linear path through the patient excluding all soft tissue voids within the bone as well as above and below it (conventional BMD is simply the average pixel mass thickness in a region). An equivalent linear thickness of solid bone (with all soft tissue voids collapsed out) is obtained by dividing pixel mass thickness (g/cm2) by average mineral density of bone tissue (1.051 g/cm3) [19]. Once the pixels in the mass profile are converted to thickness, the integral is the bone surface area in the cross-section (CSA). After determining the center of mass, the cross-sectional moment of inertia (CSMI) is the integral of area times the distance from the center of mass. Section modulus (Z, cm3) is calculated as CSMI/dmax where dmax is the maximum distance from the center of mass to the medial or lateral cortical margin. Note that Z calculated from a single projection (as in DXA) only measures stress due to bending in the image plane, not in other directions. The HSA program also measures bone outer diameter (subperiosteal width, cm) directly from the blur-corrected width of the bone mass profile and conventional BMD (g/cm2, the average raw pixel value). Finally, an estimate of cortical thickness was calculated by modeling cortices of the NN and FS cross-sections as concentric circles. Models assume that 60% and 100% of the measured mass is in the cortex for the NN and FS respectively [26]. The fan beam mode of the QDR 2000 scanner produces an error in pixel spacing along the fan beam direction that varies with the height of the bone above the scanner table surface, which could influence outcomes in obese participants. The HSA program version 2.1 was adapted to correct for this problem. A special calibration phantom was used to quantify the height effect with a pair of identical copper foils at different (known) distances above the tabletop. This yielded a correction for pixel spacing as a function of the height of the bone. Based on unpublished observations using computed tomography, the proximal femur is located at approximately the mid-sagittal plane at the level of the hip. DXA scan attenuation can be used to compute total soft tissue thickness. A special calibration phantom was designed to measure how pixel spacing changes with

increasing height above the table. Thus, using phantom data and scan attenuation principals, the HSA program estimates the height of the hip region for each scan as one half of the average soft tissue thickness. Statistical analyses All analyses were conducted with SPSS (version 11.0), using two-sided tests of hypotheses and a significance level of P b 0.05 for all a priori hypotheses. Bone geometric outcome measures were first checked for outliers, normality, and assessed for interactions between gender and overweight group including a multiplicative interaction term in regression models. As there were no significant gender–group interactions, data for both genders were combined. Height, weight and body composition (total body lean and fat mass, percent fat and BMI) were compared between groups using t tests. Categorical variables (gender, Tanner stage and race) were compared using chi-squared tests. Linear regression was used to compare measures of body composition and bone outcomes between overweight and healthy weight participants. Bone variables (such as section modulus and bone CSA), height, weight, lean mass and fat mass were natural log transformed to improve the fit of regression models. The assessment of the bone outcomes was conducted in three stages. First, bone outcomes were compared in overweight and healthy weight subjects, adjusted for gender, Tanner stage and surrogates of moment arm. Height was used as a surrogate of moment arm for the femoral shaft; femoral neck length and neck shaft angle were used as surrogates of moment arm for the narrow neck region. Second, the regression models were also adjusted for total body lean mass. Finally, total body fat mass was also included in the model to determine if fat mass has an independent effect on the bone outcomes. The fit of each model was assessed via the adjusted r 2 value and model assumptions were assessed via graphical checks and application of the Shapiro–Wilk test of normality. The effect of overweight in each multivariate model was calculated as the adjusted ratio of the outcome measure in the overweight participants divided by the outcome measure in the healthy weight group, with 95% confidence intervals. The adjusted ratio and confidence intervals were calculated as the exponentiated estimates of the regression estimates (e h). For example, Model 1 in Table 2 indicates that the ratio of femoral shaft section modulus in overweight compared with normal weight children of the same gender, Tanner stage and moment arm is 1.11, with 95% CI = 1.05, 1.18. This was calculated from the regression coefficient of Beta = 0.108, with 95% CI = 0.048, 0.168. As e 0.108 = 1.11, this is interpreted as overweight children having 1.11 (or ~11%) higher section modulus compared to healthy weight participants accounting for gender, Tanner stage and moment arm. The primary outcomes were section modulus in the femoral shaft and

M.A. Petit et al. / Bone 36 (2005) 568–576

narrow neck. Secondary outcomes, such as bone dimensions, were examined to help explain the differences, or lack thereof, in bone bending strength in overweight compared with healthy weight subjects. We chose to present the ratio of overweight to healthy weight rather than adjusted means or regression coefficients for two reasons: first, the study sample included a broad range of subject age, body size and bone geometry values. The ratio of bone geometry in the overweight compared to healthy weight reflects the difference in the bone geometry values between the groups across the entire range and is not dependent on the absolute value. In contrast, the use of an adjusted mean value from an analysis of covariance would generate a least square mean in the two groups that reflects subjects close to average section modulus. Second, as our data were log transformed, the estimated regression coefficients are difficult to interpret. Thus, the exponentiated outcomes were used to aid in interpretation. This approach has been used in recent studies comparing bone mass between groups of children [20,30].

Results

571

Table 2 Multiple linear regression analysis for section modulus at the femoral shaft and narrow neck regions with gender, maturity, and moment arma in the model (Model 1) and when adding lean mass to the model (Model 2) Ratiob (95% CI) Model 1

Ratiob (95% CI) Model 2

Femoral shaft Ln (shaft Z) Female 0.96 (0.90, 1.01) Tanner2 1.04 (0.95, 1.14) Tanner3 1.15 (1.03, 1.28)* Tanner4 1.18 (1.04, 1.33)* Tanner5 1.28 (1.14, 1.43)** Ln (Height)a 23.2 (17.0, 31.7)** Ln (Lean mass) – Overweight group 1.11 (1.05, 1.18)**

1.05 1.02 1.02 1.03 1.07 3.39 2.33 1.03

(0.99, (0.93, (0.92, (0.91, (0.95, (1.58, (1.71, (0.97,

1.11) 1.11) 1.13) 1.16) 1.20) 7.25)* 3.18)** 1.09)

Narrow neck Ln (neck Z) Female Tanner2 Tanner3 Tanner4 Tanner5 Ln (Height)a Ln (neck shaft angle)a Ln (neck length)a Ln (lean mass) Overweight group

0.95 0.97 0.94 0.99 1.02 3.14 0.21 0.95 2.07 1.05

(0.90, (0.88, (0.85, (0.89, (0.91, (1.28, (0.12, (0.79, (1.54, (0.99,

1.01) 1.05) 1.04) 1.11) 1.14) 6.64)* 0.38)** 1.15) 2.79) 1.12)

0.88 0.99 1.04 1.11 1.19 16.2 0.22 0.98

(0.84, 0.93)** (0.90, 1.08) (0.94, 1.16) (0.99, 1.24) (1.07, 1.32)* (11.7, 22.3)** (0.11, 0.42)** (0.80, 1.20) – 1.13 (1.07, 1.20)**

a

Demographic characteristics Of the 134 subjects, 94 were healthy weight and 40 were overweight. The prevalence of overweight (30%) was representative of national estimates [22]. Participant characteristics are summarized in Table 1. The age, gender and racial distributions were not significantly different between the overweight and healthy weight participants. Tanner stage distributions did not differ across groups; however, the overweight children were slightly younger (1 to 2 years on

Table 1 Descriptive characteristics for healthy weight and overweight subjects

N Agea (year) Gender (M/F) Race (% Black) Tanner (N—1/2/3/4/5) Tanner (%—1/2/3/4/5) Heighta (cm) Weighta (kg) BMIa (kg/m2) Height z scorea Weight z scorea BMI percentilea Lean massa (kg) Fat massa (kg) Lean massa (%) Fat massa (%) a

Mean (SD).

Healthy weight BMI V 85%

Overweight BMI N 85%

94 11.4 (5.1) 41/53 43% 51/8/11/10/14

40 11.1 (4.0) 13/27 49% 19/9/2/4/6

54/9/12/11/15

48/23/5/10/15

142.9 (21.6) 38.3 (17.1) 17.7 (3.0) 0.27 (1.09) 0.00 (0.79) 45.2 (24.5) 29.9 (13.1) 8.0 (5.5) 80.0 (6.7) 20.0 (6.7)

145.9 (16.6) 54.8 (21.4) 24.7 (5.0) 0.74 (1.08) 1.72 (0.60) 94.2 (4.5) 32.9 (11.2) 21.4 (11.5) 62.6 (8.6) 37.4 (8.6)

P value

0.78 0.23 0.52 0.19

Used as surrogates for moment arm (height for femoral shaft region; height, neck shaft angle and neck length for narrow neck region). b Regression coefficient. * P V 0.01. ** P V 0.001.

average) at each Tanner stage (II–V) than the healthy weight children ( P = 0.075), suggesting advanced maturation for age, as previously described in obese girls [31]. Anthropometric characteristics Overweight participants had significantly greater height, weight and BMI z scores compared with controls (Table 1). Absolute lean mass did not differ between groups and percent lean was significantly lower in the overweight subjects. However, consistent with previous studies [5,32,33], absolute lean mass adjusted for height and gender was significantly greater in the overweight subjects compared with healthy weight controls (P = 0.027). Bone geometry

0.43 b0.001 b0.001 0.023 b0.001 b0.001 0.21 b0.001 b0.001 b0.001

Regression models for the primary outcome, section modulus, are shown in Table 2. The first models (Table 2, Model 1) demonstrated that section modulus was significantly increased in the femoral shaft and narrow neck in the overweight compared with healthy weight subjects, adjusted for gender, maturity and surrogates of moment arm. For example, the ratio for femoral shaft section modulus in overweight compared with healthy weight participants in Model 1 was 1.11 (95% CI 1.05, 1.18), P b 0.001. Thus, on

572

M.A. Petit et al. / Bone 36 (2005) 568–576

average, femoral shaft section modulus was 11% (95% CI 5%, 18%) higher in the overweight group, adjusted for gender, maturity and moment arm. Within these models, Tanner stage and moment arm were significantly and independently associated with section modulus. Female gender was associated with decreased section modulus at the narrow neck only. Model 2 summarizes the effect of including total body lean mass in the model (Table 2, Model 2). Lean mass was significantly and independently associated with section modulus ( P b 0.001) at the femoral shaft and narrow neck. Adjustment for lean mass attenuated the impact of overweight status on section modulus in both regions: section modulus was not significantly increased in overweight compared with healthy weight subjects when adjusted for lean mass. Adding total body lean mass to the models explained much of the variance that was previously accounted for by Tanner stage and sex (which became non-significant). Surrogates of moment arm remained significantly associated with section modulus at both sites. In Model 3, we added total body fat mass to the model. Fat mass was not independently associated with section modulus at the femoral shaft (P = 0.723) or narrow neck (P = 0.458). The addition of fat mass to the model did not affect the ratio in overweight compared with normal weight subjects at either the femoral shaft (ratio = 1.02, 95% CI = 0.92, 1.11) or the narrow neck (ratio = 1.03, 95% CI 0.94, 1.13). Secondary analysis examined structural components of section modulus, adjusted for gender, Tanner stage and moment arm. Table 3 summarizes the effect of overweight compared with healthy weight. The greater Z at the narrow neck was due to greater subperiosteal width (4%, P = 0.002) and bone CSA (10%, P b 0.001). At the femoral shaft, the increased Z was because of greater cortical thickness (9%, P = 0.001) and increased bone CSA (10%, P b 0.001). Cortical thickness at the narrow neck and subperiosteal Table 3 Multiple linear regression analyses: ratio of secondary bone parameters in overweight compared with healthy weight subjects at the femoral shaft and narrow neck regions, adjusted for gender, Tanner stage and moment arma

Femoral shaft BMD CSA Subperiosteal width Cortical thickness Narrow neck BMD CSA Subperiosteal width Cortical thickness a

Ratio in overweight compared with healthy weight

95% CI

P

1.08 1.10 1.02 1.09

1.03, 1.05, 1.00, 1.03,

1.13 1.16 1.04 1.15

0.001 b0.001 0.092 0.001

1.06 1.10 1.04 1.06

1.01, 1.05, 1.02, 1.01,

1.11 1.16 1.07 1.12

0.016 b0.001 0.002 0.022

Surrogate measures of moment arm used were: Height for Femoral Shaft region; and Height, femoral neck length and neck shaft angle for Narrow Neck region.

Table 4 Multiple linear regression analyses: ratio of secondary bone parameters in overweight compared with healthy weight subjects at the femoral shaft and narrow neck regions, adjusted for gender, tanner stage moment arma and total body lean mass

Femoral shaft BMD CSA Subperiosteal width Cortical thickness Narrow neck BMD CSA Subperiosteal width Cortical thickness

Ratio in overweight compared with healthy weight

95% CI

P

1.02 1.02 1.00 1.02

0.97, 0.97, 0.98, 0.97,

1.06 1.07 1.03 1.07

0.455 0.420 0.843 0.477

1.01 1.04 1.03 1.00

0.96, 0.99, 1.00, 0.95,

1.06 1.08 1.06 1.06

0.833 0.158 0.040 0.893

a

Surrogate measures of moment arm used were: Height for Femoral Shaft region; and Height, femoral neck length and neck shaft angle for Narrow Neck region.

width at the femoral shaft also contributed to increased Z, but were less significant. The models were subsequently adjusted for lean mass (Table 4). Each of these bone measures was comparable in overweight and healthy weight subjects when adjusted for lean mass, with the exception of subperiosteal width which remained slightly higher in the overweight group at the narrow neck (+3%, P = 0.04). These models were subsequently adjusted for fat mass. Fat mass was not independently associated with any of the bone parameters (P N 0.22 for all). As seen in Fig. 1, section modulus values were similar between groups relative to lean mass (Fig. 1A), and appear lower relative to fat mass alone (Fig. 1B). However, these apparently blowQ values were not significant as shown by the overlapping confidence intervals for the slope shown in the figure.

Discussion Using a DXA-based method of assessing bone geometry, we showed that section modulus (Z), an index of bone bending strength, was greater in overweight compared with healthy weight children and adolescents. The higher Z appeared to be appropriate for the higher lean mass observed in overweight children—that is, for the same gender, stage of maturity, height and lean mass, proximal femur Z was equivalent in overweight and healthy weight children. The extra weight from fat mass was not independently related to bone geometry. These data are consistent with the theory that bone adapts its strength primarily to muscle force. We assessed mechanically meaningful measures of bone geometry, which, when interpreted from a functional perspective, help to explain some discrepancies in the current literature.

M.A. Petit et al. / Bone 36 (2005) 568–576

573

Fig. 1. The relationship between femoral shaft section modulus and total body lean mass (A) or fat mass (B) for healthy weight (5) and overweight (n) participants. Regression equations, r values and slopes (h, 95% CI) are listed on the figure.

Obesity and the muscle–bone unit Previous studies exploring the influence of overweight or obesity on bone in children have mainly reported total body or lumbar spine BMC. Researchers concluded that overweight or obese children have normal [4] or increased BMC relative to healthy weight peers [5,34–36], while Goulding et al. concluded that obese children have decreased BMC relative to bone size and body weight [9,30]. Our results are consistent with both findings in that overweight children had higher bone bending strength than healthy weight children, but Z was appropriate for their lean mass. The extra weight from fat mass had no additional influence. Thus, in two children of the same weight and height, the child with higher percent body fat would likely have a lower bone bending strength—because they would have less lean mass. A close association between various surrogates of muscle force (dynamometry, total body lean mass and muscle crosssectional area) with indices of bone mass or geometry has been shown both during growth [37–40] and in adults [41]. It has been known for more than three decades that muscle mass and bone mass are closely associated [42,43]. Thus, our finding that bone bending strength is closely related to total body lean mass is not surprising. Nonetheless, our results are important in that they suggest the need to interpret bone strength (or indices of) in the context of an individual’s lean mass (or indices of muscle force), rather than body weight alone—particularly in cases where body composition may differ from the average healthy child (i.e. obese children and other clinical groups). Various approaches for assessing bone strength relative to muscle mass in clinical pediatric populations have been recently published [20,38,40,44–47]. That others report a strong association with body weight and various measures of bone

bmassQ or bstrengthQ in healthy children or adults does not conflict with our findings. In healthy children, body weight may be a reasonable surrogate for lean mass or muscle force. However, in overweight children, our data clearly show that the components of body weight need to be considered separately when assessing a child’s bone health. Our results are consistent with the functional model of bone development [16,38,40,44,48]. Bone adapts its strength primarily to dynamic, rather than static loads [10,49,50]. There is no doubt that muscle force causes a larger dynamic strain on bone than body weight—it takes more than 2 kg of muscle force on bone to move 1 kg of body weight [11,12,51]. Although these relationships do not prove cause and effect, it is not surprising that in an absolute sense, overweight individuals (both children and adults) have higher bone mass and geometric strength; they need greater muscle force to move their body weight. Our observation that fat mass had no independent association to bone geometry supports known concepts that body weight itself is not the largest osteogenic stimulus; rather, bone appears to adapt to the highest dynamic loads by altering bone geometry dimensions to maintain keeping average daily bone strain within certain limits [44]. Other factors Other factors may modify the relationship between muscle and bone [52]. For example, bone strength relative to lean mass appears to increase around puberty in girls— perhaps due to increased mechanosensitivity of bone from sex steroids [53–55]. Mice with genetically induced increases in muscle hypertrophy do not have expected increases in bone strength in the absence of muscle use [56]. Genetic, hormonal and nutritional factors may mediate the muscle–bone relationship by influencing mechanosensitiv-

574

M.A. Petit et al. / Bone 36 (2005) 568–576

ity of bone, or may have a direct effect on muscle mass or force. However, none of these factors are likely to provide the central regulatory information to which bone adapts [16,57,58]. As muscle takes a shorter time to adapt to physical activity than does bone, it is possible that recent history of physical activity could have influenced the muscle–bone relationship. This did not appear to be the case with our data as there was a very tight relationship between lean mass and section modulus in both groups (Fig. 1A). However, we did not assess past physical activity in this cross-sectional study. Clinical significance The clinical significance of altered bone strength in children lies in current or future fracture risk. Recent studies suggest an increased risk of fracture in obese children [6,8,23] attributed to either low bone strength or reduced balance and postural sway [7,59]. That obese children have higher bone strength in an absolute sense suggests that they should be relatively protected from fragility fractures. This seems to be true in older individuals where body weight is seen as protective against osteoporosis and fractures [1]. However, in cases of traumatic fractures (such as a fall from swing-set), which are more common during growth, the forces on bone come from body weight and the height of the fall. Thus, traumatic fracture may be more likely in overweight children who have greater forces on the bone from added body weight in fat mass, i.e. weight that does not appear to provide an additional osteogenic stimulus (above that of the muscle force required to move the extra weight). Interestingly, reports of increased fracture risk in childhood obesity were limited to forearm fractures. While we demonstrated that bone strength was appropriate for loads in the weightbearing femoral neck, the forearm may not manifest comparable increases in bone strength.

body weight. Future studies using QCT methods may help to resolve this issue. Fan beams introduce magnification errors in measures of bone geometry as well as traditional DXA outcomes of BMC and BMD that are relevant when measuring overweight children. Scanners with an anterior–posteroior projection (as Hologic has) would tend to underestimate BMC with increased elevation off the table (as occurs with obesity) [61,62]. New versions of the Hip Structure Analysis software used in this study incorporate a calibration phantom that corrects for this fan beam projection error (described in Methods section). The cross-sectional nature of our data is a limitation—we cannot prove cause and effect, only show associations that are consistent with theory. Longitudinal data show that bone accrual lags behind increases in lean body mass in boys and girls [39]. Overweight children in our study were approximately 1 year younger at each Tanner stage than healthy weight children. As our data were cross-sectional, it is not clear if this represents early maturation or if this would influence the adaptation of bone to muscle load. A longitudinal assessment of the muscle–bone unit in obese and non-obese adolescents would be useful to further understand these relationships during growth. A final limitation is our use of DXA total body lean mass as a crude surrogate of muscle force. DXA uses a 2 component model for measuring body composition, the limitations of which are widely addressed in the literature [63]. However, lean mass can be accurately measured by DXA [13] and is strongly correlated with muscle crosssectional area by pQCT in children aged 9–12 years (r = 0.821, P b 0.01, unpublished observation). The relationships between muscle force and moment arm in their action on bone at any particular cross-section is complex. Importantly, we used necessarily crude and imperfect indices of moment arm (height) and muscle strength (total body lean mass). Future studies are needed to determine the best measures to describe the forces on bone.

Limitations Summary and conclusions As with all DXA-based studies, there are fundamental problems with trying to measure the strength of a complex three-dimensional structure like the proximal femur using two-dimensional DXA data, particularly during growth. DXA scanners were designed with the idea that the amount, or bmassQ of bone is the primary measure of bone strength. We used the HSA program to interpret the DXA data in a mechanically meaningful way. The technical details, strengths and limitations of the Hip Structure Analysis program have been thoroughly reviewed recently [60]. One assumption in the HSA algorithm is that bones are fully mineralized which may not be the case in 4–20 year olds. The effect of an overestimate of mineralization is that bone CSA and section modulus values would be underestimated. It is not likely that the error would influence the relationships we observed between bone geometry and lean mass or

In summary, we showed that overweight children had wider and stronger bones at the proximal femur in the absolute sense and that higher strength was appropriate for their higher lean mass and height. Body weight in the form of fat mass did not contribute to bone strength. Our data are consistent with the theory that bone adapts its strength primarily to dynamic loads from muscle force, not static loads (i.e. body weight). Our findings suggest that overweight individuals could be protected from fragility fractures (as evidence suggests), but the increased bone strength may not be sufficient to overcome the greater forces (from high body weight) generated when an obese child falls. Our study was not designed to address this question directly—and the influence of overweight or obesity on lifelong fracture risk is not yet clear.

M.A. Petit et al. / Bone 36 (2005) 568–576

Acknowledgments We appreciate the dedication and enthusiasm of the children and their families who participated in this study and we thank Lisa Semanick for her careful scan analysis. This work was supported by National Institutes of Health (NIH) research grants K23AR49040-01A1 (MAP), K08-DK02523 (MBL) and 1-R03-DK058200 (MBL), the General Clinical Research Center (M01RR00240) and the Nutrition Center, The Children’s Hospital of Philadelphia, University of Pennsylvania School of Medicine.

References [1] Bainbridge KE, Sowers M, Lin X, Harlow SD. Risk factors for low bone mineral density and the 6-year rate of bone loss among premenopausal and perimenopausal women. Osteoporos Int 2004; 15:439 – 46. [2] Margolis KL, Ensrud KE, Schreiner PJ, Tabor HK. Body size and risk for clinical fractures in older women. Study of Osteoporotic Fractures Research Group. Ann Intern Med 2000;133:123 – 7. [3] Moro M, van der Meulen MC, Kiratli BJ, Marcus R, Bachrach LK, Carter DR. Body mass is the primary determinant of midfemoral bone acquisition during adolescent growth. Bone 1996;19:519 – 26. [4] Ellis KJ, Shypailo RJ, Wong WW, Abrams SA. Bone mineral mass in overweight and obese children: diminished or enhanced? Acta Diabetol 2003;40:S274–7. [5] Leonard M, Shults J, Wilson B, Stallings V, Zemel B. Obesity during childhood and adolescence augments bone mass and bone dimensions. Am J Clin Nutr 2004;80:514 – 23. [6] Goulding A, Jones IE, Taylor RW, Williams SM, Manning PJ. Bone mineral density and body composition in boys with distal forearm fractures: a dual-energy X-ray absorptiometry study. J Pediatr 2001;139:509 – 15. [7] Goulding A, Jones IE, Taylor RW, Manning PJ, Williams SM. More broken bones: a 4-year double cohort study of young girls with and without distal forearm fractures. J Bone Miner Res 2000; 15:2011 – 8. [8] Skaggs DL, Loro ML, Pitukcheewanont P, Tolo V, Gilsanz V. Increased body weight and decreased radial cross-sectional dimensions in girls with forearm fractures. J Bone Miner Res 2001; 16:1337 – 42. [9] Goulding A, Taylor RW, Jones IE, McAuley KA, Manning PJ, Williams SM. Overweight and obese children have low bone mass and area for their weight. Int J Obes Relat Metab Disord 2000;24: 627 – 32. [10] Lanyon LE, Rubin CT. Static vs. dynamic loads as an influence on bone remodelling. J Biomech 1984;17:897 – 905. [11] Martin RB, Burr DB, Sharkey NA. Skeletal tissue mechanics. New York7 Springer Verlag; 1998. [12] Burr DR. Muscle strength, bone mass, and age-related bone loss. J Bone Miner Res 1997;12:1547 – 51. [13] Shih R, Wang Z, Moonseong H, Wang W, Heymsfield SB. Lower limb skeletal muscle mass: development of dual-energy Xray absorptimetry prediction model. J Appl Physiol 2000;89: 1380 – 1386. [14] Bamman MM, Newcomer BR, Larson-Meyer DE, Weinsier RL, Hunter GR. Evaluation of the strength–size relationship in vivo using various muscle size indices. Med Sci Sports Exerc 2000;32:1307 – 13. [15] Manzoni P, Brambilla P, Pietrobelli A, Beccaria L, Bianchessi A, Mora S, et al. Influence of body composition on bone mineral content in children and adolescents. Am J Clin Nutr 1996;64:603 – 7.

575

[16] Rauch F, Schoenau E. The developing bone: salve or master of its cells and molecules? Pediatr Res 2001;50:309 – 14. [17] Beck TJ, Ruff CB, Warden KE, Scott Jr WW, Rao GU. Predicting femoral neck strength from bone mineral data: a structural approach. Invest Radiol 1990;25:6 – 18. [18] van der Meulen MC, Ashford MW, Kiratli BJ, Bachrach LK, Carter DR. Determinants of femoral geometry and structure during adolescent growth. J Orthop Res 1996;14:22 – 9. [19] Martin RB, Burr DB. Non-invasive measurement of long bone crosssectional moment of inertia by photon absorptiometry. J Biomech 1984;17:195 – 201. [20] Leonard MB, Feldman HI, Shults J, Zemel BS, Foster B, Stallings VA. Long-term, high-dose glucocorticoids and bone mineral content in childhood glucocorticoid-sensitive nephrotic syndrome. N Engl J Med 2004;351:868 – 75. [21] Tanner JM. Growth at adolescence. Oxford7 Blackwell Scientific Pub; 1955. [22] Ogden CL, Flegal KM, Carroll MD, Johnson CL. Prevalence and trends in overweight among US children and adolescents, 1999–2000. JAMA 2002;288:1728 – 32. [23] Goulding A, Cannan R, Williams SM, Gold EJ, Taylor RW, LewisBarned NJ. Bone mineral density in girls with forearm fractures. J Bone Miner Res 1998;13:143 – 8. [24] Hologic model QDR-4500 user’s guide. Waltham, MA7 Hologic, Inc; 1996. [25] Leonard MB, Feldman HI, Zemel BS, Berlin JA, Barden EM, Stallings VA. Evaluation of low density spine software for the assessment of bone mineral density in children. J Bone Miner Res 1998;13:1687 – 90. [26] Beck T, Looker A, Ruff C, Sievanen H, Wahner H. Structural trends in the aging femoral neck and proximal shaft: analysis of the Third National Health and Nutrition Examination Survey (NHANES) dualenergy X-ray absorptiometry data. J Bone Miner Res 2000;15: 2297 – 2304. [27] Petit MA, McKay HM, MacKelvie KJ, Heinonen A, Khan KM, Beck TJ. A randomized school-based jumping intervention confers site and maturity specific benefits on bone structural properties in girls: a Hip Structural Analysis (HSA) study. J Bone Miner Res 2002;17:363 – 72. [28] Looker AC, Beck TJ, Orwoll ES. Does body size account for gender differences in femur bone density and geometry? J Bone Miner Res 2001;16:1291 – 9. [29] Duan Y, Beck TJ, Wang XF, Seeman E. Structural and biomechanical basis of sexual dimorphism in femoral neck fragility has its origins in growth and aging. J Bone Miner Res 2003;18:1766 – 74. [30] Goulding A, Taylor R, Jones I, Manning P, Williams S. Spinal overload: a concern for obese children and adolescents? Osteoporos Int 2002;13:835 – 40. [31] Adair LS, Gordon-Larsen P. Maturational timing and overweight prevalence in US adolescent girls. Am J Public Health 2001; 91:642 – 4. [32] Forbes GB, Welle SL. Lean body mass in obesity. Int J Obes 1983; 7:99 – 107. [33] Foster BJ, Shults J, Zemel BS, Leonard MB. Interactions between growth and body composition in children treated with high-dose chronic glucocorticoids. Am J Clin Nutr 2004;80:1334 – 41. [34] Fischer S, Milinarsky A, Giadrosich V, Dib G, Arriagada M, Arinoviche R. X-ray absorptiometry of bone in obese and eutrophic children from Valparaiso, Chile. J Rheumatol 2000;27:1294 – 6. [35] DeSchepper J, VandenBroeck M, Jonckheer M. Study of lumbar spine bone mineral density in obese children. Acta Paediatr 1995; 84:313 – 5. [36] Hasanoglu A, Bideci A, Cinaz P, Tumer L, Unal S. Bone mineral density in childhood obesity. J Pediatr Endocrinol Metab 2000; 13:307 – 11. [37] Heinonen A, McKay HA, Whittall K, Forster B, Khan K. Muscle cross sectional area is associated with specific site of bone in

576

[38]

[39]

[40] [41]

[42] [43] [44]

[45]

[46]

[47]

[48] [49]

[50]

M.A. Petit et al. / Bone 36 (2005) 568–576 prepubertal girls: a quantitative magnetic resonance imaging study. Bone 2001;29:388 – 92. Schoenau E, Neu C, Beck B, Manz F, Rauch F. Bone mineral content per muscle cross-sectional area as an index of the functional muscle– bone unit. J Bone Miner Res 2002;17:1095 – 101. Rauch F, Bailey D, Baxter-Jones A, Mirwald R, Faulkner R. The dmuscle–bone unitT during the pubertal growth spurt. Bone 2004; 34:771 – 5. Ruff C. Growth in bone strength, body size and muscle size in a juvenile longitudinal sample. Bone 2003;33:317 – 29. Rittweger J, Beller G, Ehrig J, Jung C, Koch U, Ramolla J, et al. Bone–muscle strength indices for the human lower leg. Bone 2000;27:319 – 26. Doyle F, Brown J, LaChance C. Relation between bone mass and muscle weight. Lancet 1970;1:391 – 3. Thompson D. On growth and form. Cambridge, UK7 Cambridge University Press; 1961. Frost HM, Schoenau E. The bmuscle–bone unitQ in children and adolescents: a 2000 overview. J Pediatr Endocrinol Metab 2000;13: 571 – 90. Hogler W, Briody J, Woodhead HJ, Chan A, Cowell CT. Importance of lean mass in the interpretation of total body densitometry in children and adolescents. J Pediatr 2003;143:81 – 8. Bechtold S, Rauch F, Noelle V, Donhauser S, Neu C, Schoenau E, et al. Musculoskeletal analyses of the forearm in young women with Turner syndrome: a study using peripheral quantitative computed tomography. J Clin Endocrinol Metab 2001;86:5819 – 23. Crabtree N.J., Kibirige M.S., Fordham J.N., Banks L.M., Mutoni F., Chinn D., et al. The relationship between lean body mass and bone mineral content in paediatric health and disease. Bone 2004;35: 965 – 72. Frost HM. Bone bmassQ and the bmechanostatQ: a proposal. Anat Rec 1987;219:1 – 9. Mosley JR, Lanyon LE. Strain rate as a controlling influence on adaptive modeling in response to dynamic loading of the ulna in growing male rats. Bone 1998;23:313 – 8. Mosley JR. Osteoporosis and bone functional adaptation: mechanobiological regulation of bone architecture in growing and adult bone, a review. J Rehabil Res Dev 2000;37:189 – 99.

[51] Frost HM. Obesity, and bone strength and bmassQ: a tutorial based on insights from a new paradigm. Bone 1997;21:211 – 4. [52] Daly R, Saxon L, Turner C, Robling A, Bass S. The relationship between muscle size and bone geometry during growth and in response to exercise. Bone 2004;34:281 – 7. [53] Ferretti JL, Capozza RF, Cointry GR, Garcia SL, Plotkin H, Alvarez Filgueira ML, et al. Gender-related differences in the relationship between densitometric values of whole-body bone mineral content and lean body mass in humans between 2 and 87 years of age. Bone 1998;22:683 – 90. [54] Schiessl H, Frost HM, Jee WS. Estrogen and bone–muscle strength and mass relationships. Bone 1998;22:1 – 6. [55] Jarvinen TLN, Kannus P, Sievanen H. Estrogen and bone: a reproductive and locomotive perspective. J Bone Miner Res 2003;18:921 – 31. [56] Turner C. Muscle–bone interactions, revisited. Bone 2000;27: 339 – 340. [57] Frost HM. Muscle, bone, and the Utah paradigm: a 1999 overview. Med Sci Sports Exerc 2000;32:911 – 7. [58] Ferretti J, Cointry G, Capozza R, Frost H. Bone mass, bone strength, muscle–bone interactions, osteopenias and osteoporoses. Mech Ageing Dev 2003;124:269 – 79. [59] Goulding A, Jones I, Taylor R, Piggot J, Taylor D. Dynamic and static tests of balance and postural sway in boys: effects of xprevious wrist bone fractures and high adiposity. Gait Posture 2003;17:36 – 141. [60] Beck T. Measuring the structural strength of bones with dual-energy X-ray absorptiometry: principles, technical limitations, and future possibilities. Osteoporos Int 2003;14:S81–8. [61] Pocock NA, Noakes KA, Majerovic Y, Griffiths MR. Magnification error of femoral geometry using fan beam densitometers. Calcif Tissue Int 1997;60:8 – 10. [62] Bolotin HH, Sievanen H. Inaccuracies inherent in dual-energy X-ray absorptiometry in vivo bone mineral density can seriously mislead diagnostic/prognostic interpretations of patient-specific bone fragility. J Bone Miner Res 2001;16:799 – 805. [63] Wells JCK, Fuller NJ, Dewit O, Fewtrell MS, Elia M, Cole TJ. Fourcomponent model of body composition in children: density and hydration of fat-free mass and comparison with simpler models. Am J Clin Nutr 1999;69:904 – 12.