Osteoporosis International

, Volume 20, Issue 1, pp 61–70

Anterior–posterior bending strength at the tibial shaft increases with physical activity in boys: evidence for non-uniform geometric adaptation

Authors

  • H. M. Macdonald
    • Department of Mechanical and Manufacturing EngineeringUniversity of Calgary
  • D. M. L. Cooper
    • Department of Anatomy and Cell BiologyUniversity of Saskatchewan
    • Faculty of MedicineUniversity of British Columbia
    • UBC Department of Orthopaedics, Centre for Hip Health and Musculoskeletal ResearchVancouver Coastal Health Research Institute
Original Article

DOI: 10.1007/s00198-008-0636-9

Cite this article as:
Macdonald, H.M., Cooper, D.M.L. & McKay, H.A. Osteoporos Int (2009) 20: 61. doi:10.1007/s00198-008-0636-9

Abstract

Summary

We investigated bone structural adaptations to a 16-month school-based physical activity intervention in 202 young boys using a novel analytical method for peripheral quantitative computed tomography scans of the tibial mid-shaft. Our intervention effectively increased bone bending strength in the anterior–posterior plane as estimated with the maximum second moment of area (Imax).

Introduction

We previously reported positive effects of a physical activity intervention on peripheral quantitative computed tomography (pQCT)-derived bone strength at the tibial mid-shaft in young boys. The present study further explored structural adaptations to the intervention using a novel method for pQCT analysis.

Methods

Participants were 202 boys (aged 9–11 years) from 10 schools randomly assigned to control (CON, 63 boys) and intervention (INT, 139 boys) groups. INT boys participated in 60 min/week of classroom physical activity, including a bone-loading program. We used ImageJ to process pQCT images of the tibial mid-shaft and determine the second moments of area (Imax, Imin) and cortical area (CoA) and thickness (CTh) by quadrant (anterior, medial, lateral, posterior). We defined quadrants according to pixel coordinates about the centroid. We used mixed linear models to compare change in bone outcomes between groups.

Results

The INT boys had a 3% greater gain in Imax than the CON boys (p = 0.04) and tended to have a greater gain in Imin (∼2%, NS). Associated with the greater gain in Imax was a slightly greater (NS) gain (1–1.4%) in CoA and CTh in the anterior, medial, and posterior (but not lateral) quadrants.

Conclusion

Our results suggest regional variation in bone adaptation consistent with patterns of bone formation induced by anterior–posterior bending loads.

Keywords

BoneChildrenPeripheral QCTPhysical activity

Introduction

It is generally understood that bone strength is increased by formation of new tissue that is targeted to meet the increased mechanical demands of exercise. The capacity for this level of adaptation is greatest during the developmental years [1] and diminishes with increasing age [2]. Tomographic imaging modalities, including peripheral quantitative computed tomography (pQCT), offer the opportunity to directly assess cross-sectional geometric adaptation—a key advantage over bone mass and areal density measures provided by dual energy X-ray absorptiometry (DXA). Common measures of pQCT-derived bone geometry and strength indices including total and cortical bone area and the cross-sectional polar moment of inertia provide some insight into possible surface-specific changes that may occur during growth or with exercise; however, these measures do not account for how bone is distributed about a certain axis [3, 4]. For example, the polar moment of inertia (designated J) provides a measure of bone strength in torsion [3, 4]. The polar moment is calculated relative to the long bone axis, which is oriented perpendicular to the bone cross-section at the centroid. Alternatively, the area moment of inertia (designated I), or second moment of area, characterizes resistance to bending—the primary mode of loading at long bone shafts [5, 6]. The area moment of inertia is typically calculated relative to centroidal axes oriented parallel to the bone cross-section. There are two orthogonal principal bending axes, about which Imax and Imin respectively define the axes along which it is hardest and easiest to bend the long bone. The ratio of Imax/ Imin is a useful measure of cross-sectional shape that indicates the degree to which the bone shape deviates from purely circular [7]. Further, this ratio may provide information on the types of load a bone incurs by how its shape adapts [8]. For example, at the tibial mid-shaft, where the primary mode of bending is in the anterior–posterior (AP) direction [5], the optimal cross-sectional shape is one where the greatest moment area is aligned with the AP bending plane [9].

A number of studies have demonstrated how the bone structure of animals adapts to loading. Changes in Imax, Imin, and Imax/Imin varied based on the pre-existing geometry of the long bone and the mechanical milieu associated with the predominant loading pattern [3, 1014]. For example, at the rat ulnar diaphysis, when bending is imposed in the medio-lateral (ML) direction, the greatest change in strain during loading is on the medial and lateral surfaces [13]. In turn, the ML surfaces experience the greatest bone formation, which leads to significantly greater gains in Imin compared with Imax (which resists bending on the AP axis) [10, 14]. Ultimately, overall bone strength (ultimate force) is enhanced as measured by mechanical testing in this animal model [10]. To date, only one human study has used three-dimensional imaging technology to investigate changes in Imax and Imin in response to an exercise intervention in premenopausal women [15]. These relationships have not been explored in the growing skeleton using pQCT.

In addition to changes in the distribution of bone about a defined axis, recent attention has been given to regional differences in bone adaptation within localized areas of the bone cross-section. A novel application of pQCT was used to investigate regional variation in a nonhuman primate model [16] and in post-menopausal women [17, 18]. Regional variations in bone mineral density (BMD) and geometrical parameters such as cortical thickness were observed, and these adaptations were consistent with expected patterns of remodeling and structural adaptations induced by bending. Used in combination with measures of Imax and Imin, regional analysis of changes in bone geometry may provide a more comprehensive assessment of bone structural adaptation to increased loading.

We recently reported the effects of a 16-month school-based physical activity intervention, Action Schools! BC (AS! BC), on estimated tibial bone strength in boys and girls [19]. Prepubertal boys in the intervention group tended to have greater gains in the pQCT-derived density-weighted polar section modulus (polar strength strain index; SSIp), an estimate of torsional bone strength, compared with same maturity boys in the control group. Thus, our primary objective of the present analysis was to further investigate the structural adaptation at the tibial mid-shaft by comparing changes in Imax, Imin, and the ratio of Imax/Imin between intervention and control boys who participated in the AS! BC trial. Second, we more specifically assessed the adaptation within four bone quadrants to determine whether a regional adaptation in cortical bone area and cortical thickness underpinned any changes in Imax or Imin. We hypothesized that boys in the intervention group would have greater gains in Imax compared with boys in the control group, and that change in Imin would be similar between the groups.

Materials and methods

Study design and participants

We have detailed our methods and the school-based physical activity intervention in previous reports [1922]. Briefly, we conducted a cluster randomized, controlled school-based intervention trial. Ten schools from the Vancouver and Richmond, British Columbia School Districts were randomly assigned to control (CON, 3 schools), Level 1 intervention (4 schools) or Level 2 intervention (3 schools). The intervention arms differed with regard to the amount of facilitation provided to teachers and not in the physical activity delivered to students [22], and thus the two intervention arms were collapsed (INT, 7 schools) for the present analysis.

All children in grades 4 and 5 (at baseline) attending intervention schools participated in AS! BC. However, a total of 514 (47%) children (257 boys, 257 girls) from the 10 volunteer schools received parental consent to participate in the evaluation. We include only the boys in the present analysis. During the study, 30 (12%) boys (17 INT, 13 CON) moved or withdrew. One INT boy was excluded due to a femur fracture sustained during the study period (unrelated to the intervention). A further 2 INT boys were excluded based on conditions that could affect normal physical activity or bone development (fetal alcohol syndrome, childhood leukemia). In addition, 8 boys from one INT school did not have a baseline pQCT scan and pQCT scans from 11 boys (9 INT, 2 CON) could not be analyzed due to movement artefacts or errors in scan acquisition. Thus, the present analysis includes 205 boys (141 INT, 64 CON) who were, on average, 10.2 ± 0.6 years of age at baseline. From a health history questionnaire completed by parents/guardians at baseline, we determined ethnicity according to parents’ or grandparents’ place of birth. The majority of boys were Asian (52%) with both parents or all four grandparents born in Hong Kong or China, India, Philippines, Vietnam, Korea or Taiwan. The remainder of the boys were Caucasian (35%), or were of mixed ethnicity or other ethnic origins (13%).

All measurements were performed at the University of British Columbia Bone Health Research Laboratory. Baseline measurements were performed between February and April 2003. The intervention spanned 1.25 school years (phase I: April to June 2003, phase II: October 2003 to May 2004), and follow-up measurements were performed between April and June 2004.

AS! BC intervention

AS! BC is an active school model that provides tools for schools and teachers to increase physical activity opportunities for children throughout the school day [21, 22]. The bone-loading program within AS! BC consisted of two components: 15 min physical activity 5 days per week (in addition to regular PE), and Bounce at the Bell [23], a simple jumping activity that required students to perform either counter movement jumps or side-to-side jumps three times per day, 4 days per week. For the 15 × 5 component, teachers could choose from a variety of activities including skipping, dancing, playground circuits and simple resistance exercise with exercise bands. All activities required minimal equipment and could be performed in the classroom or hallway or outside on the school grounds. We provided teachers with a Classroom Action Bin that contained equipment and resources to facilitate the 15 × 5 activities. For Bounce at the Bell, we instructed teachers to deliver a progressive program. During phase I, students performed 5 two-foot landing jumps (or 10 one-foot landing jumps) at each session. During phase II, the number of jumps increased (starting from 5 per session) over each month of the school year, until a maximum of 36 jumps per day was achieved. The AS! BC Support Team [22] provided training on the 15 × 5 and Bounce at the Bell programs to intervention teachers (48 across phases I and II). We instructed teachers at control schools to continue with their regular program of physical education, which involved two 40-min PE classes per week, on average.

To monitor compliance we asked teachers at intervention schools to complete activity logs. Each school day teachers recorded the type, frequency and duration of each activity undertaken with their class as well as the number of sessions and jumps per session for Bounce at the Bell. We asked teachers at control schools to complete a modified version of the activity log. Across the study period, INT teachers delivered approximately 60 min more physical activity per week than CON teachers [22]. Median compliance with Bounce at the Bell was 74% (interquartile range: 50−89%) across INT schools. We were unable to assess individual student compliance with Bounce at the Bell. However, intervention teachers delivered Classroom Action activities, including Bounce at the Bell, to all students in their classrooms as part of the regular school program. From school records, we determined that average student attendance was 96% across all schools during the study.

Study outcomes

We used pQCT (XCT-2000) to acquire a 2.3 ± 0.2 mm slice at the mid-shaft of the left tibia using acquisition parameters defined previously [19]. One trained technician who was not blinded to group assignment acquired all scans. A phantom was scanned daily to maintain quality assurance.

We used ImageJ 1.37v (http://rsb.info.nih.gov/ij/) to process the 16-bit grayscale pQCT image data of the tibial mid-shaft. A median filter (1 voxel radius) was applied to suppress image noise prior to analysis. We used a global threshold of 480 mg/cm3 to detect bone surfaces and isolate the tibia from the scan field. Subsequently, we used the 480 mg/cm3 threshold for the analysis of geometric properties (Imax, Imin, cortical area, and cortical thickness). We used this same threshold in our previous analysis [19]. Imax and Imin values and axes (Fig. 1a) were calculated using a modified (batch-automated) version of Moment Macro v1.2 (www.hopkinsmedicine.org/fae/mmacro.htm). This program utilizes Medalia’s [24] approach of equivalent ellipses to calculate mechanical properties and determine the principal orthogonal bending axes originating at the cross-sectional centroid (centroidal axes; CAmax, CAmin, Fig. 1a). We used an anatomical landmark—the anterior tibial crest—to define anatomical axes and assign the cortical pixels into quadrants (anterior, medial, posterior, and lateral) for assessment of cortical area (CoA, mm2). The anterior tibial crest was automatically detected as the pixel of maximal distance anterior to the centroid. In order to assess cortical thickness (CTh, mm), we determined the bone surface (periosteal and endosteal) coordinates at 5° steps about the cross-section. Using this approach it was possible to map out regional changes in bone surfaces between groups (e.g., Fig. 2). We defined CTh as the radial distance between the periosteal and endosteal surfaces for each step. We utilized these 72 measurements to qualitatively assess thickness changes in radial plots (e.g., Fig. 3) and we used quadrant-level averages (18 measurements each) for quantitative comparisons.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-008-0636-9/MediaObjects/198_2008_636_Fig1_HTML.gif
Fig. 1

a Representative baseline (top) and follow-up (bottom) peripheral quantitative computed tomography (pQCT) images of an intervention group boy’s left tibia with the principal centroidal axes superimposed. b Outlines showing bone surfaces and quadrants with the principal centroidal axes (CAmax, CAmin; solid) and anatomical axis (dashed) superimposed. The maximum and minimum second moments of area (Imax, Imin) are calculated about CAmax and CAmin respectively. c Superimposition of the baseline (solid) and follow-up (dashed) bone surfaces

https://static-content.springer.com/image/art%3A10.1007%2Fs00198-008-0636-9/MediaObjects/198_2008_636_Fig2_HTML.gif
Fig. 2

Radial plot of bone surface changes (follow-up/baseline) for the periosteal (outer lines) and endosteal (inner lines) surfaces for Intervention (solid line) and control (dashed line) boys. Lines were smoothed by a three-point running average

https://static-content.springer.com/image/art%3A10.1007%2Fs00198-008-0636-9/MediaObjects/198_2008_636_Fig3_HTML.gif
Fig. 3

Radial plot of change in cortical thickness for Intervention (solid line) and control (dashed line) boys. Lines smoothed by a three-point running average

While the anatomical axes did not directly correspond with the principal centroidal axes (e.g., the axis of Imin and the AP anatomical axis), they were in close agreement (mean difference of 3.8° in the baseline scans, Fig. 1b). The reproducibility of the quadrant-based measurements of cortical area and the second moments of area was assessed on a sample of 10 young adults (20–40 years of age), who were scanned twice with repositioning (0.5 mm in-plane pixel size). The coefficients of variation (CV%) for the cortical area measurements were 0.4%, 0.5%, 0.8%, and 0.9% for the anterior, medial, lateral, and posterior quadrants respectively. The CV values for Imax and Imin were 0.6% and 0.5% respectively. As reproducibility is resolution-dependent, these values represent a conservative estimate of the reproducibility associated with the higher resolution (0.4 mm in-plane pixel size) used in the current study.

Descriptive and independent variables

At baseline and follow-up we measured sitting and standing height to the nearest 0.1 cm, and body weight to the nearest 0.1 kg [19]. We measured the length of the left tibia as the distance from the distal edge of the medial malleolus to the tibial plateau (to the nearest 0.1 cm) using an anthropometric tape. For each variable, we report the mean of two measurements. Using the mean value for follow-up measurements of sitting and standing height and body weight we calculated a maturity offset value for each boy according to the equation developed by Mirwald et al. [25]. We also assessed maturity status at baseline and follow-up using self-rated Tanner staging [26]. We assessed muscle cross-sectional area (MCSA, mm2) at the proximal two-thirds site (66% of the total tibial length) of the left tibia with pQCT. We estimated lower limb dynamic power with maximal height (cm) for vertical jump and maximal distance (cm) for standing long jump, as previously reported [19]. We used a modified version of the Physical Activity Questionnaire for Older Children (PAQ-C) [27, 28] to assess self-reported leisure-time physical activity. We calculated a general physical activity score (PA Score) as an average of the nine PAQ-C items in a continuous range between 1 (low activity) and 5 (high activity) and estimated the time (h/week) spent participating in common sports and activities designated as loaded (impact > walking, load time) from item 1. We used a validated food frequency questionnaire (FFQ) [29] to determine dietary calcium (mg/day). We administered the PAQ-C and FFQ at baseline and follow-up, plus three additional times during the study period (June 2003, September 2003, and January 2004). We report the average across the five reports for PA Score, load time, and dietary calcium.

Statistical analysis

As in our previous analysis [19], we used a linear mixed effects model (xtmixed in Stata, Version 9.2) to compare change in our primary (Imax, Imin, Imax/Imin) and secondary (CoA and CTh by quadrant) outcomes between intervention and control boys. We used this approach in order to account for both the within- and between-school variance associated with the clustered design. For the analysis of Imax, Imin, and Imax/Imin we included the baseline bone value (to account for bone size at baseline), baseline body weight (to account for the baseline imbalance between groups), change in tibial length (to account for rate of linear growth) and maturity offset at follow-up (to account for maturity status) as covariates and we designated group as the fixed effect and school as the random effect. For the quadrant analysis of CoA and CTh, we used relative change (absolute change/baseline value) as the dependent variable in order to account for the differences in CoA and CTh across quadrants at baseline. We included baseline weight, change in tibial length and maturity offset as covariates. In addition, we included both group and quadrant as fixed factors in order to investigate potential group by quadrant interactions. In the case of a significant main effect of quadrant we conducted post-hoc pairwise comparisons with Bonferroni correction. We performed all analyses according to randomization, used standard residual plots to assess normality, linearity, and homoscedasticity, and identified outliers using Cook’s distance values. We calculated the intracluster correlation coefficient (ICC) as \({\text{ICC}} = {{s_c^2 } \mathord{\left/ {\vphantom {{s_c^2 } {\left( {s_c^2 + s_w^2 } \right)}}} \right. \kern-\nulldelimiterspace} {\left( {s_c^2 + s_w^2 } \right)}}\) where \(s_c^2 \) equals the variance between clusters (schools) and \(s_w^2 \) equals the variance within clusters [30]. We used Stata, Version 9.2 (StataCorp, College Station, TX, USA) for all analyses.

Results

Descriptive and independent variables

Baseline and change in descriptive and independent variables are presented (Table 1). Despite randomization, we noted a slight imbalance in body weight and MCSA between INT and CON boys at baseline. CON boys tended to be heavier and have a larger MCSA than INT boys. The majority of INT and CON boys were prepubertal (Tanner stage 1) at baseline. At follow-up, CON boys tended to have a greater gain in body weight than INT boys, whereas INT boys tended to have greater improvements in vertical and long jump performance.
Table 1

Baseline and 16-month change in descriptive characteristics of control and intervention boys

 

Control (n = 63)

Intervention (n = 139)

Baseline

Change (95% CI)

Baseline

Change (95% CI)

Baseline age (years)

10.3 (0.6)

10.2 (0.6)

No. Asian/Caucasian/other

30/21/12

75/50/14

Baseline Tanner stage (1/2/3/4/5)

44/17/2/0/0

85/47/7/0/0

Final Tanner stage (1/2/3/4/5)

26/24/12/1/0

45/55/32/6/1

Height (cm)

141.2 (6.8)

6.6 (6.2, 7.1)

141.1 (6.9)

6.9 (6.5, 7.2)

Tibial length (cm)

32.8 (2.2)

1.6 (1.4, 1.8)

32.8 (2.3)

1.7 (1.6, 1.8)

Weight (kg)

39.6 (9.6)

6.1 (5.4, 6.9)

36.5 (8.2)

5.5 (5.1, 6.0)

Maturity offset (at follow-up)

−1.99 (0.04)

 

−2.14 (0.03)

 

MCSA (cm2)

35.2 (5.5)

4.0 (3.5, 4.5)

33.5 (5.0)

4.0 (3.6, 4.4)

Long jump (cm)

133.2 (20.9)

6.3 (2.4, 10.1)

133.6 (17.5)

9.2 (6.7, 11.7)

Vertical jump (cm)

28.3 (5.2)

2.4 (1.2, 3.7)

27.7 (5.8)

4.2 (3.3, 5.2)

Average PA score (/5)

2.7 (0.4)

2.7 (0.4)

Average load time (h/week)

6.8 (4.3)

6.6 (4.2)

Average dietary calcium (mg/day)

935 (453)

903 (418)

SD = standard deviation, CI = confidence interval, MCSA = muscle cross-sectional area, PA = physical activity

Baseline values are mean (SD) unless otherwise stated and 16-month change values are mean (95% CI)

Primary and secondary outcomes

We provide baseline, follow-up, and adjusted difference in change in primary and secondary pQCT outcomes for boys (Table 2). We excluded 3 boys (2 INT, 1 CON) from the analysis because Cook’s distance values for their bone outcomes were 2.5 times greater than the next closest value.
Table 2

Baseline and follow-up values for peripheral quantitative computed tomography (pQCT) bone outcomes for control (CON) and intervention (INT) boys and the difference in change in pQCT bone outcomes between groups

Outcome

Group

Baseline

Follow-up

Adjusted difference in change (95% CI)a

p

ICC

Mean (SD)

Mean (SD)

Imax (mm4)

CON

10,814.9 (3,250.7)

13,805.9 (4,318.1)

338.6 (16.4, 660.9)

0.04

0.01

INT

10,241.1 (3,301.0)

13,431.0 (4,524.9)

Imin (mm4)

CON

5,396.1 (1,589.5)

6,622.9 (2,143.8)

99.2 (−157.7, 356.1)

0.45

0.12

INT

5,133.0 (1,580.2)

6,398.7 (2,110.8)

Imax/Imin (mm4)

CON

2.02 (0.04)

2.11 (0.04)

0.018 (−0.021, 0.056)

0.38

0.04

INT

2.01 (0.03)

2.12 (0.03)

aAdjusted for baseline value, baseline body weight, change in tibial length, and maturity offset at follow-up

ICC = intraclass correlation

At baseline, we observed a slight imbalance in Imax and Imin with values being 5–6% greater among CON boys than INT boys. After adjusting for select covariates, change in Imax was significantly greater (∼3%) for INT boys (+3,229.9 mm4; 95% CI: 3,051.9, 3,408.0) than CON boys (+2,891.3 mm4; 95% CI: 2,625.4, 3,157.1). Change in Imin also tended to be greater among INT boys (∼2%); however, the difference in change between groups was not statistically significant. Similarly, change in the ratio of Imax/Imin was not significantly different between the groups.

Similar to baseline values for Imax and Imin, CON boys tended to have greater CoA (1–3%) than INT boys across all four quadrants and greater (2–3%) CTh in the anterior and medial quadrants (Table 3). Qualitative investigation of the radial plots revealed that INT boys appeared to add slightly more bone on the periosteal surface in the anterior, posterior, and lateral quadrants (Fig. 2). In both groups CTh increased primarily along the anterior–posterior axis while the medial–lateral axis was relatively static (Fig. 3). Quantitatively, the percentage change in CoA and CTh across all quadrants was not significantly different between INT and CON boys after adjusting for covariates and multiple comparisons. However, INT boys tended to have a greater gain (approximately 1–2%, NS) in CoA and CTh than CON boys in the medial and anterior quadrants. Regardless of group, at baseline CoA and CTh were greatest in the anterior quadrant and smallest in the medial quadrant. Change in CoA in the anterior and posterior quadrants was similar (∼14%) and both were significantly greater than change in CoA in the medial (9.5%) and lateral (11.0%) quadrants after adjusting for multiple comparisons (both p < 0.001, data not shown). The anterior quadrant experienced the greatest change in CTh (9.6%) and this was significantly different from change in CTh in the medial quadrant only (7.2%, p < 0.001) after adjusting for multiple comparisons. The ICCs for change in CoA and CTh across sectors were 0.04 and 0.03 respectively.
Table 3

Baseline, follow-up, and percentage change for cortical area (CoA) and cortical thickness (CTh) by quadrant for control (CON) and intervention (INT) boys

Quadrant

Group

Baseline

Follow-up

Percentage change (95% CI)

p*

Mean (SD)

Mean (SD)

CoA-Ant (mm2)

CON

85.2 (12.5)

96.3 (14.2)

13.1 (11.1, 15.2)

0.35

INT

82.9 (12.6)

95.0 (15.5)

14.3 (12.9, 15.7)

CoA-Med (mm2)

CON

47.2 (6.6)

51.2 (7.7)

8.7 (6.6, 10.7)

0.38

INT

45.6 (6.6)

50.2 (8.2)

9.8 (8.4, 11.1)

CoA-Post (mm2)

CON

72.8 (11.0)

82.3 (12.7)

13.2 (11.2, 15.3)

0.53

INT

71.8 (11.9)

81.9 (14.2)

14.0 (12.6, 15.4)

CoA-Lat (mm2)

CON

48.8 (6.6)

54.0 (8.5)

10.6 (8.5, 12.6)

0.74

INT

48.3 (7.5)

53.7 (9.2)

11.0 (9.6, 12.4)

CTh-Ant (mm)

CON

5.64 (0.63)

6.12 (0.63)

8.9 (6.8, 11.1)

0.47

INT

5.52 (0.80)

6.07 (0.80)

10.0 (8.4, 11.4)

CTh-Med (mm)

CON

3.17 (0.47)

3.35 (0.47)

6.3 (4.1, 8.5)

0.33

INT

3.07 (0.54)

3.31 (0.54)

7.6 (6.1, 9.0)

CTh-Post (mm)

CON

4.65 (0.63)

5.04 (0.63)

8.8 (6.6, 11.0)

0.93

INT

4.68 (0.81)

5.09 (0.81)

8.9 (7.5, 10.4)

CTh-Lat (mm)

CON

3.24 (0.46)

3.50 (0.46)

8.2 (6.0, 10.4)

0.85

INT

3.26 (0.56)

3.52 (0.56)

7.9 (6.5, 9.4)

Ant = anterior quadrant, Med = medial quadrant, Lat = lateral quadrant, Post = posterior quadrant

* p value for difference in percentage change between groups after adjusting for baseline body weight, change in tibial length, and maturity offset

Discussion

We aimed to advance the current literature by examining the effects of a physical activity intervention on the biomechanically relevant parameters Imax and Imin in the growing human skeleton. In addition, we used a novel method to analyze pQCT images to acquire quadrant-specific results for cortical area and thickness. Using this approach we were able to more closely examine the effect of the intervention relative to the estimated strain patterns associated with AP bending loads, which are known to predominate at the tibial diaphysis [5]. The trend toward greater gains in torsional bone strength in our previous analysis [19] became a significantly greater gain when we examined bone bending strength as estimated with Imax. The slightly larger increase in Imax compared with Imin is in agreement with expected strain patterns associated with AP bending loads at the tibial diaphysis.

A novel analysis of pediatric pQCT scans

The second moments of area, Imax, and Imin, describe bone’s resistance to bending in the AP and ML planes respectively. Currently, these measures and their orientations cannot be estimated with the manufacturer-designed pQCT software. The Stratec software provides second moments of area about the anatomical x and y axes, which approximate the principal centroid axes about which Imin and Imax are calculated (Fig. 1a); however, these parameters have not been reported in any pediatric pQCT studies to date. Our findings are in agreement with several animal studies [10, 12, 3133] and one intervention study with pre-menopausal women [15]. Lieberman and colleagues [31] found that after approximately 3 months of treadmill running, juvenile male sheep had 21–29% greater Imax and 20–21% greater Imin (relative to body mass) at the diaphysis of the tibia and metatarsal than non-exercising sheep. Similar to our findings, there was no difference in the ratio of Imax/Imin between the exercise and control groups. The authors suggested that stability in this ratio indicates that during regular activity, long bones in the growing skeleton are bent in both the AP and ML directions and that further increases in bending loads such as those associated with exercise intervention remain proportional in the AP and ML planes [31].

In the one previous human study that investigated change in Imax and Imin, Vainionpaa et al. [15] used spiral QCT to determine the effects of a 12-month high-impact exercise program on bone geometry at the femur and tibia in 120 healthy pre-menopausal women. They compared change in Imax and Imin between exercisers and controls and found that those women who were most compliant with the exercise program had a 2.5% greater increase in Imax than controls. Further, they reported a significant relationship between the number of daily impacts (by accelerometer) and the change in the ratio of Imax/Imin at both the mid-femur and the proximal tibia. The authors suggested that this indicates a change in the cross-sectional shape due to increased periosteal apposition in the AP plane. This result differs from that of the present analysis and that of Lieberman et al. [31], which may reflect the well-documented differences in the adaptive response to loading between the adolescent and adult skeleton [1, 34]. It is possible that in mature long bones, bending loads are not proportional in the AP and ML directions, whereas they are in growing bone. In addition, the multidirectional nature of the jumps in the Bounce at the Bell program may have contributed to increased strain on the medial and lateral surfaces as well as the anterior and posterior surfaces, whereas the jumps in the Vainionpaa et al. [15] study were mainly unidirectional drop jumps.

Although the difference in change in Imax between intervention and control boys was relatively small (∼3%) compared with the overall growth-related increase (∼27%), evidence from animal studies suggests that the second moment of area has an exponential relationship with overall bone strength. At the rat ulna, Imin has been shown to contribute to 92% of the variance in ultimate failure [10] and small changes (∼2-fold) in Imin in response to loading can result in a substantial 100-fold increase in fatigue resistance [35]. Further, the magnitude of the intervention effect on Imax is similar to that observed in previous school-based trials that have investigated changes in DXA-derived bone mineral content and estimated bone strength at various skeletal sites in children of similar age and physical activity level [36].

Quadrant-specific changes

In addition to the change in Imax and Imin, we further explored regional variation in bone structural adaptation to our intervention using a quadrant-based approach to measuring cortical area and cortical thickness. Region-specific analysis of cortical bone adaptation has recently been investigated in both a nonhuman primate model [16] and in postmenopausal women [17]. In both studies, regional variation in volumetric BMD was consistent with expected patterns of strain associated with the predominant bending loads at the tibial shaft. Although neither study assessed cortical area, Lai and colleagues [17] reported values for cortical thickness that were 21% greater in the anterior than in the posterior cortex and 8.4% greater in the medial than in the lateral cortex.

In the present study, intervention boys did not demonstrate a significantly greater gain in either cortical area or thickness than control boys in any quadrant; however, a trend toward increased cortical area and thickness was evident in 3 of the 4 quadrants, with the greatest difference in change observed in the anterior quadrant (∼1.4%). Taken together, these results suggest that the anterior cortex, which incurs tensile strains as a result of AP bending [5], likely adapts to increased strain via periosteal apposition. In turn, the slightly greater gain in cortical area may contribute exponentially to a greater gain in Imax and function to reduce tensile strains in this region. Although the difference in change did not achieve statistical significance, it is worth noting that intervention boys had a slightly greater increase in CoA in the medial quadrant. Greater changes in the medial quadrant may reflect either a shift in the neutral axis toward the tensile surface such that the tensile strains are distributed across the anterior and medial surfaces, or the diversity of the weight-bearing activities involved in AS! BC, such that the multidirectional jumps varied the strain distribution. Importantly, we are only able to infer characteristics of the strain environment associated with AS! BC activities based on findings from strain gauge studies in animals and humans (intact and cadavers). Further, we and others [3, 9] recommend caution when interpreting bone cross-sectional geometry in terms of loading environments.

Limitations

The limitations of the study design and the school-based intervention were discussed at length in our previous publications [19, 22]. In the present analysis we note limitations specific to the pQCT analysis. First, the anatomical and principal axes determined by the ImageJ algorithm may not represent those actually experienced during loading. Several animal studies have shown that the combination of bending and axial compressive forces shifts the neutral axis away from the centroid toward the cortex under tension [3, 37]. Thus, the quadrants we assessed in the present study approximate the regions where the largest strains occur during loading. Second, differences in the quadrant-based measurement of area and cortical thickness could potentially be affected by a shift in the location of the centroid with growth. For example, if growth is asymmetric (e.g., greater anteriorly than posteriorly) then the centroid would be pulled forward. This would affect the delineation of quadrants and the measurement of area and thicknesses within them. With pQCT data it is difficult to overcome this limitation since identifying consistent “landmarks,” such as fluorochrome labels in histological analysis, is not possible. One potential avenue for future studies might be registration of baseline and follow-up images by minimizing the absolute differences between the two images and determining a single shared center point. Third, although we used a standard pixel size (0.4 mm) for our pQCT analyses, this resolution may have limited the accuracy with which we could detect small differences in the change in cortical area between groups.

In contrast to our previous analysis [19], in which we categorized participants as either pre- or early pubertal based on self-report Tanner staging, in the present study we calculated a maturity offset value that represents years from peak height velocity (PHV) [25]. As none of the boys had attained PHV by the end of the study, it was not possible to apply a categorical maturity offset value (pre- or post-PHV) as recommended by Mirwald and colleagues [25].

Conclusion

We identified an enhanced bone strength adaptation consistent with the predominant anterior–posterior bending experienced in the tibial diaphysis in boys undertaking a school-based physical activity program. The specificity of the response we observed highlights the need for more region-specific analyses of bone structural adaptation than are routinely employed in pQCT-based studies.

Acknowledgements

We gratefully acknowledge the participation of the principals, teachers, children, and their parents from the ten volunteer schools. Thanks also to Bryna Kopelow, Jennifer Fenton, and the JW Sporta team for their considerable input and expertise on design and delivery of the intervention model. We also appreciate the invaluable assistance of Dr. David Thomas in the development of the quadrant-based analysis macro and we thank Dr. Christopher Ruff for making the MomentMacro available. We acknowledge funding support from the BC Ministry of Health, 2010 Legacies Now, Provincial Health Services Authority, The National Science and Engineering Research Council of Canada and the Canadian Institutes for Health Research. Professor McKay is a Michael Smith Foundation for Health Research Senior Scholar.

Conflicts of interest

None.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2008