Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Effectiveness of a school-based physical activity intervention for increasing bone strength in children.. Macdonald, Heather McGillvray 2006

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
[if-you-see-this-DO-NOT-CLICK]
ubc_2006-200353.pdf [ 22.4MB ]
Metadata
JSON: 1.0077027.json
JSON-LD: 1.0077027+ld.json
RDF/XML (Pretty): 1.0077027.xml
RDF/JSON: 1.0077027+rdf.json
Turtle: 1.0077027+rdf-turtle.txt
N-Triples: 1.0077027+rdf-ntriples.txt
Original Record: 1.0077027 +original-record.json
Full Text
1.0077027.txt
Citation
1.0077027.ris

Full Text

EFFECTIVENESS OF A SCHOOL-BASED PHYSICAL ACTIVITY INTERVENTION FOR INCREASING BONE STRENGTH IN CHILDREN: ACTION SCHOOLS! BC by HEATHER MCGILLVRAY MACDONALD BSc, The University of British Columbia, 1999 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Human Kinetics) THE UNIVERSITY OF BRITISH COLUMBIA JULY 2006 © Heather McGillvray Macdonald, 2006 ABSTRACT Introduction: Osteoporosis and related fracture are significant societal health burdens. Physical activity during childhood can result in significant bone health benefits, which may reduce the risk for osteoporosis later in life. Aim: The primary aim was to evaluate the effectiveness of a school-based physical activity model, Action Schools! BC (AS! BC), for enhancing bone mass and strength in boys and girls using novel bone imaging technologies. Methods: Design and Participants: This was a 16-month cluster randomized, controlled, school-based intervention. Ten schools were randomized to Intervention (INT, 7 schools) or Control (CON, 3 schools). The bone-loading component of AS! BC was a daily, high-impact jumping program (Bounce at the Bell) plus 15 minutes/day of classroom physical activity. Participants were 514 children aged 9-11 years at baseline. Bone Measurements: Tibial bone strength was assessed with peripheral quantitative computed tomography. Dual energy x-ray absorptiometry and hip structure analysis were used to assess femoral neck bone mineral content (BMC) and bone strength, respectively. Results Part 1: Cross-sectional Comparisons In pre- and early pubertal boys tibial bone strength was 5-15% greater than in pre- and early pubertal girls. After adjusting for tibial length, muscle cross-sectional area was the primary explanatory variable of bone strength in both sexes. Part 2:16-Month Change - Tibial Bone Strength Intervention boys (n = 147) tended to have greater gains in tibial bone strength than CON boys (n = 64); however, the intervention effect was only significant for prepubertal boys at the distal tibia. Action Schools! BC was not effective for increasing tibial bone strength in girls (n = 137 INT, 65 CON). Part 3:16-Month Change - Femoral Neck Bone Mass and Strength Intervention girls tended to have greater gains in femoral neck bone strength and BMC than CON girls; however, the difference in change between groups was not significant. Action Schools! BC was not effective for increasing femoral neck bone mass or strength in boys. Summary: Skeletal adaptations to the AS! BC intervention were sex-, maturity- and site-specific. Action Schools! BC offers promise as a simple and inexpensive strategy for increasing bone mass and strength in boys and girls. TABLE OF CONTENTS ABSTRACT ii TABLE OF CONTENTS iiLIST OF TABLES viiLIST OF FIGURES x GLOSSARY OF TERMS AND ABBREVIATIONS xiii PREFACE: PUBLICATIONS ARISING FROM THIS THESIS xv ACKNOWLEDGEMENTS xvi1 Introduction, Literature Review, Rationale, Objectives & Hypotheses 1 1.1 Introduction 1 1.2 Literature Review 3 1.2.1 Bone Biology and Bone Growth 3 1.2.1.1 Bone Tissue: Composition and Organization 3 1.2.1.2 Whole Bone Structure 4 1.2.1.3 Physiology of Bone Growth and Bone Turnover 5 1.2.2 Bone Biomechanics 8 1.2.2.1 Mechanical Properties of Bone1.2.2.2 Bone Adaptation to Mechanical Stimuli 12 1.2.2.3 Mechanostat Theory 11.2.3 Measurement of Bone Parameters in Children 19 1.2.3.1 Dual Energy X-Ray Absorptiometry 11.2.3.2 Hip Structure Analysis 20 1.2.3.3 Peripheral Quantitative Computed Tomography 23 1.2.4 The Growing Skeleton: Maturity- and Sex-Related Differences in Bone Geometry, Density and Strength 31 1.2.4.1 Assessing Maturity1.2.4.2 Hormonal Regulation of Bone Development 33 1.2.4.3 Maturity- and Sex-Related Differences in Skeletal Development 36 1.2.5 Determinants of Bone Strength 41 1.2.5.1 Heredity 41.2.5.2 Race and Ethnicity 2 1.2.5.3 Dietary Calcium 4 1.2.5.4 Muscle 6 1.2.6 Physical Activity and Bone Health in Children 53 1.2.6.1 Measurement of Physical Activity in Children 4 1.2.6.2 Cross-sectional Studies in Children - General Physical Activity 55 1.2.6.3 Longitudinal Studies in Children - General Physical Activity 56 1.2.6.4 Cross-sectional Studies in Adults - Racquet-sport Athletes 7 1.2.6.5 Cross-sectional Studies in Children - Athlete Populations 9 1.2.6.6 Exercise Interventions with Children 60 iii 1.2.6.7 Exercise during Growth: Effects on Bone Later in Life 64 1.2.7 Summary of Directions for New Research on Bone Strength in the Growing Skeleton 70 1.2.7.1 Sex Differences, Muscle and Tibial Bone Strength 71.2.7.2 Physical Activity and Bone Strength in the Growing Skeleton 70 1.3 Rationale, Objectives and Hypotheses 71 1.3.1 Part I: Bone Strength and its Determinants in Pre- and Early Pubertal Boys and Girls 71 1.3.2 Part II: Sixteen-Month Longitudinal Study of a School-Based Physical Activity Intervention on Tibial Bone Strength in Boys and Girls 72 1.3.3 Part III: Sixteen Month Longitudinal Study of a School-Based Physical Activity Intervention on Femoral Neck Bone Mass and Strength in Boys and Girls 73 References 74 2 Methods 90 2.1 Action Schools! BC Overview 92.1.1 Development of the Action Schools! BC Model 90 2.1.2 Organizational Structure of Action Schools! BC 1 2.2 AS! BC: Evaluation 3 2.2.1 Study Design 92.2.2 Recruitment - Schools 5 2.2.3 Recruitment - Children 6 2.2.4 Consent, Health History and Ethnicity 92.2.5 Action Schools! BC Implementation2.2.5.1 Creating an Action Plan 97 2.2.5.2 Classroom Action Bin2.2.6 Action Schools! BC Model: Required Components 98 2.2.6.1 Classroom Action 15x5 92.2.6.2 Classroom Action: Bounce at the Bell 92.2.7 Program Compliance 9 2.2.8 Data Collection 92.2.8.1 Data Collection Overview2.2.8.2 Measurement Team Training 100 2.2.8.3 Bone Densitometry 102.2.8.4 Body Composition and Muscle Cross-sectional Area 104 2.2.8.5 Anthropometry 5 2.2.8.6 Physical Performance 102.2.8.7 Questionnaires2.2.9 Statistical Analysis 7 2.2.9.1 Statistical Power 102.2.9.2 Mixed Linear Model 9 References 110 3 Overview of the Cohort 112 iv 3.1 Participants 112 3.2 Participants Lost to Followup 118 References 124 4 Part I: Bone Strength and its Determinants in Pre- and Early Pubertal Boys and Girls 125 4.1 Introduction 125 4.2 Methods 6 4.2.1 Subjects4.2.2 Measurements 127 4.3 Statistical Analysis4.4 Results 128 4.4.1 Comparisons between Sexes - Descriptives 124.4.2 Comparisons between sexes - Bone geometry, density and strength 129 4.4.3 Determinants of Bone Geometry, Density and Strength in Boys and Girls 129 4.4.3.1 Distal Tibia 124.4.3.2 Tibial Shaft 130 4.5 Discussion 144.5.1 Sex Differences in Bone Geometry, Density and Strength 141 4.5.2 Determinants of Bone Geometry, Density and Strength 2 4.5.2.1 Muscle Cross-sectional Area 144.5.2.2 Non-modifiable Factors: Ethnicity and Maturity 143 4.5.2.3 Modifiable Factors: Physical Activity and Dietary Calcium 144 4.5.3 Limitations 144.6 Summary 5 References 146 5 Part II: Effectiveness of a School-Based Physical Activity Intervention for Increasing Tibial Bone Strength in Boys and Girls: A Cluster Randomized Controlled Trial 149 5.1 Introduction 149 5.2 Methods 150 5.2.1 Study Design 155.2.2 Participants5.2.3 Intervention5.2.4 Descriptive Outcomes 151 5.2.5 Primary and Secondary Outcomes 152 5.3 Statistical Analysis 155.4 Results 153 5.4.1 Participants and Compliance 155.4.2 Descriptives 155.4.3 Primary outcomes 4 v 5.4.3.1 Boys 154 5.4.3.2 Girls5.4.4 Secondary Outcomes 155.4.4.1 Boys 155.4.4.2 Girls 5 5.5 Discussion 162 5.5.1 Action Schools! BC: A Unique Intervention 165.5.2 pQCT - A Novel Imaging Modality 163 5.5.3 Sex- and Maturity-specificity of the Bone Response 165 5.5.4 Limitations 167 5.5.5 Implications of the Findings 165.6 Summary 8 References 169 6 Part III: Effectiveness of a School-based Physical Activity Intervention for Increasing Femoral Neck Bone Mass and Strength in Boys and Girls 172 6.1 Introduction 176.2 Methods 3 6.2.1 Study Design 176.2.2 Participants6.2.3 Intervention6.2.4 Measurements 174 6.2.4.1 Descriptive Outcomes6.2.4.2 Primary and Secondary Outcomes 175 6.2.5 Statistical Analysis 176 6.3 Results 177 6.3.1 Participants and Compliance 176.3.2 Descriptives 176.3.3 Primary Objective - Intent-to-Treat 177 6.3.4 Secondary Objective - Intent-to-Treat 8 6.3.5 Primary Objective - Compliant Subgroup 176.3.6 Secondary Objective - Compliant Subgroup 8 6.4 Discussion 186 6.4.1 Study Design Issues 187 6.4.2 Skeletal Adaptations to Increased Physical Activity in Boys 186.4.3 Skeletal Adaptations to Increased Physical Activity in Girls 188 6.4.4 Methodological Limitations 189 6.4.5 Implications of the Findings 190 6.4.6 Summary 19vi References 192 7 Integrated Discussion 195 7.1 Overview of Findings7.1.1 Tibial Bone Strength in Pre- and Early Pubertal Children 195 7.1.2 Is Action Schools! BC an Effective Means to Increase Bone Strength in Elementary School Children? 197.1.3 Bounce at the Bell - A Novel Bone-Loading Program 196 7.2 Utility of pQCT in Pediatric Bone Research 197 7.3 Challenges Associated with School-Based Interventions 199 7.3.1 Study Design 197.3.2 Teacher Compliance 200 7.4 Public Health Implications of Action Schools! BC 201 7.5 Summary and Conclusions 203 7.5.1 Part I: Bone Strength and its Determinants in Pre- and Early Pubertal Boys and Girls 203 7.5.2 Part II: Effectiveness of a School-Based Physical Activity Intervention for Increasing Tibial Bone Strength in Boys and Girls: A 16-month Cluster Randomized Trial 204 7.5.3 Part III: Effectiveness of a School-Based Physical Activity Intervention for Increasing Femoral Neck Bone Mass and Strength in Boys and Girls 20References 206 Appendices 8 Appendix A: Action Schools! BC Advisory Committee Membership 209 Appendix B: Ethics Approval Form, Information Letters, Consent Form, Health History Questionnaires 212 Appendix C: Action Inventory, School Health Inventory, Action Plan Worksheet 223 Appendix D: The Action Schools! BC Manual • Classroom Action 15x5, Classroom Action Bin, Bounce at the Bell 232 Appendix E: Activity Logs 238 Appendix F: Questionnaires 241 Appendix G: Results for Study Participants 249 Appendix H: Intervention Delivery 252 Appendix I: Additional Data for Chapter 5 261 Appendix J: Additional Data for Chapter 6 7 vii LIST OF TABLES Table 1-1. Common bone outcomes in pediatric pQCT studies. The site of analysis, analysis mode and a brief description of each outcome are provided 29 Table 1-2. Physical activity interventions with prepubertal children and mixed maturity cohorts 66 Table 2-1. Bounce at the Bell program for Phase II 9Table 2-2. Transportation schedule for study participants to and from Vancouver General Hospital (VGH) 100 Table 2-3. Analysis modes, thresholds and outcome variables for pQCT measurements at the distal (8%), midshaft (50%) and proximal two-thirds (66%) sites of the tibia 104 Table 3-1. Summary of consent rates across Control (CON) and Intervention (INT) schools at baseline. Intervention schools 1 through 4 are Level 1 intervention schools and schools 5-7 are Level 2 intervention schools 113 Table 3-2. Boys and girls by group and ethnicity at baseline (N = 514: 359 Intervention, 155 Control) and followup (N = 443: 312 Intervention, 131 Control) 115 Table 3-3. Age, maturity, body size and composition, lifestyle variables, physical performance, tibial bone geometry, density and strength indices, bone mineral content (BMC) and femoral neck geometry and strength at baseline for Intervention and Control children 116 Table 3-4. Boys' and girls' age, ethnic distribution, maturity, body size and composition, lifestyle variables, physical performance, tibial bone geometry, density and strength indices, bone mineral content (BMC) and femoral neck bone geometry and strength at baseline 117 Table 3-5. Number of acceptable scans for dual energy X-ray absorptiometry (DXA), hip structure analysis (HSA) and peripheral quantitative computed tomography (pQCT) analysis at baseline and followup for boys and girls in intervention and control groups 121 Table 3-6. Baseline, 16-month absolute and percent (%) change (A) for height, weight, proximal femur bone mineral content (PF BMC), polar strength strain index (SSIP) and narrow neck section modulus (NN-Z) and 16-month average dietary calcium intake, physical activity score (PA Score) and load time for Control (CON) and Intervention (INT) schools. Mean (SD) 122 Table 4-1. Descriptive characteristics of pre- and early pubertal boys (n= 218) and girls (n = 206). Values are mean (SD) unless otherwise indicated 131 Table 4-2. Mean (SD) values for pQCT bone outcomes (unadjusted) at the distal tibia (8% site), tibial midshaft (50%) and proximal two-thirds tibia (66%) for pre- and early pubertal boys and girls 132 Table 4-3. Bivariate correlations (Pearson's R) of pQCT bone outcomes at the distal tibia (8% site), tibial midshaft (50% site) and proximal two-thirds tibia (66% site) with tibial length, muscle cross-sectional area (MCSA), maturity, ethnicity, physical activity, dietary calcium and vertical jump for boys and girls 133 Table 4-4. Multiple regression models for pQCT variables at the distal tibia (8%), tibial midshaft (50%) and proximal two-thirds tibia (66% site) for boys and girls 137 Table 5-1. Baseline and change (where appropriate) in descriptive characteristics for Control and Intervention boys and girls 156 Table 5-2. Tanner stage at baseline and followup for Control and Intervention boys and girls 157 Table 5-3. Baseline, followup and adjusted difference in change in distal tibia and tibial midshaft pQCT outcomes for Control (CON) and Intervention (INT) boys and girls 158 Table 5-4. Adjusted difference in change (Intervention - Control) in pQCT outcomes at the distal tibia (8% site) and tibial midshaft (50% site) for pre- (PRE) and early pubertal (EARLY) boys 160 viii Table 6-1. Baseline age, baseline and final Tanner stages, ethnic distribution, menarcheal status (for girls) average physical activity and calcium outcomes and baseline and change in body size, body composition and jump performance for Control and Intervention boys and girls. Values are mean (SD) for baseline and mean (95% CI) for change (unless otherwise indicated) 180 Table 6-2. Baseline and final Tanner stage for pre- and early pubertal boys and girls in Control and Intervention groups 181 Table 6-3. Baseline, followup and adjusted difference in change between Intervention (INT) and Control (CON) boys for primary and secondary bone outcomes. Results are presented for both the intent-to-treat and efficacy subgroup analysis 182 Table 6-4. Baseline, followup and adjusted difference in change between Intervention (INT) and Control (CON) girls for primary and secondary bone outcomes. Results are presented for both the intent-to-treat and efficacy subgroup analysis 184 ix LIST OF FIGURES Figure 1-1. Illustration of a growing long bone. Adapted from Jee et al. (42) 4 Figure 1-2. Formation (+) and resorption (-) during longitudinal bone growth. During growth from A to B, the funnel like metaphysis is reduced to match the shape of the narrower diaphysis through osteoclastic resorption on the periosteal surface of the metaphysis (a). Thickening of the cortex occurs along the cortical endosteal surface of the metaphysis and enlargement of the marrow cavity occurs by resorption of metaphyseal trabecular bone (b). The diameter of the diaphysis increases by periosteal bone formation (c) and the marrow cavity of the diaphysis expands by bone resorption on the endosteal surface (d). Adapted from Jee (42) and Baron (38) 7 Figure 1-3. A standard stress/strain curve of a bone specimen produced during mechanical testing. This curve can also be used to represent whole bone properties (load/deformation curve). Adapted from Einhom (16) 9 Figure 1-4. The principle loads experienced by bone in nature. Arrows indicate applied forces in (A) compression, (B) bending (tension on the convex side and compression on the concave side), (C) twisting (or torsion) and (D) shear. Adapted from Kontulainen (77) and Pearson and Lieberman (35) 11 Figure 1-5. The cross-sectional moment of inertia (I, mm4) for a bone cross-section describes the distribution of bone material about a defined axis. The cross-section is divided into many square regions of area A that are located at a distance (r) from the neutral axis for bending (x-axis). Adapted from Martin et al. (29) 11 Figure 1-6. Schematic of the mechanostat theory for bone response to mechanical loading. MESm represents the minimum effective strain for bone modeling and MESr represents the minimum effective strain for bone remodeling. In the adapted state between MESr and MESm, bone turnover is minimal as typical strains change. During growth, increasing body weight and muscle loads should increase strains towards MESm and turn modeling on. In contrast, during aging, decreasing muscle strength should reduce the strains towards MESr and turn disuse remodeling on. Adapted from Frost (54,88) 13 Figure 1-7. Maximum bending moments (N • mm) at the fracture test of the femur and tibia in the rats trained with different numbers of jumps per day. Values are means + SD. * significantly (p < 0.01) different from the control group (0 jumps/day); f from 5-jump group and 10-jump group (p < 0.05); % from the 5-, 10-, 20- and 40-jump groups (p < 0.01). Adapted from Umemura et al. (14) 16 Figure 1-8. Schematic of the effect of bone cross-sectional geometry on long bone strength. Changes in distribution of bone mass that influence bone bending strength (i.e. cross-sectional moment of inertia) are not reflected in conventional measures of areal bone mineral density (aBMD) by dual energy x-ray absorptiometry 20 Figure 1-9. Proximal femur image from Hologic DXA scanner showing positions of analysis regions at the narrow neck, intertrochanteric and femoral shaft. To the left of the image are typical bone mass profiles used in measurement of geometric properties 21 Figure 1-10. Illustration of the effect of femur repositioning on the location and orientation of the image plane. (A) Axial view of proximal femur, with the femoral neck axis positioned correctly (co-planar with the DXA scanning plane represented by the dotted line). (B) Repositioning error has placed the femoral neck axis out of the scanning plane and results in a distorted cross-section. Adapted from Khoo et al. (129) with permission from Elsevier 22 Figure 1-11. Schematic of a long bone and its biomechanical properties measured by DXA and pQCT. Whereas DXA calculates bone mineral content (BMC) from planar x-ray attenuation data, pQCT generates a three-dimensional cross-section from which geometric and material properties of the bone are obtained. For example, the bending and torsional cross-sectional moments of inertia (CSMIX, CSMIP) are obtained as the integral sum of the products of the area of each pixel (A) and the squared distance (dx, dy, dz) to the corresponding bending (x, y) or torsion (z) axis. Adapted from Kontulainen (77) and Ferretti et al. (141) 24 Figure 1-12. Schematic of the partial volume effect (PVE) as it pertains to the measurement of cortical bone mineral density (CoD) by pQCT. (A) At sites where the cortical wall is not sufficiently thick (< 2mm), partially filled voxels (assumed in this example to be filled by two-thirds) at the edge of the cortex will reduce the measured CoD. (B) x At sites where the cortical wall is sufficiently thick, partially filled voxels have less of an influence on the measured CoD. Adapted from Schoenau et al. (23) 26 Figure 1-13. Graph illustrating total body bone mineral content (TB BMC) accrual velocity and ages at peak bone mineral content velocity and peak height velocity for boys (blue) and girls (red) according to chronological age. Adapted from Bailey et al. (5) 33 Figure 1-14. Peaks for height velocity and bone mineral content (BMC), amplitudes for growth hormone (GH) and insulin-like growth factor-l and trends for estrogen and testosterone levels in girls relative to chronological age and Tanner stage. Peaks (connected to age by solid lines) for height and BMC velocities and GH and IGF-I indicate the average age at which these peaks occur in girls as well as the corresponding approximate Tanner stage. In boys, peak height velocity and peak BMC velocity occur about 1.5 years later than girls (at 13.4 years (Tanner stage 3) and 14.0 years (Tanner stage 4), respectively). The relationship between peaks for height and BMC velocity and peaks for GH and IGF-I are similar for boys. Adapted from MacKelvie et al. (181) with permission from BMJ Publishing Group 34 Figure 1-15. Total body (A) and femoral neck (B) BMC accrual for boys (solid squares) and girls (open squares) by biological age (years from age at peak height velocity, PHV). Values are means. * p < 0.05 between biological age groups. Adapted from Baxter-Jones et al. (204) with permission from Taylor and Francis 37 Figure 1-16. Schematic diagram depicting 20-month change in cortical bone in early-, peri- and postpubertal girls and boys. Average increases in total bone area and marrow cavity area are drawn to scale. Adapted from Kontulainen et al. (25) 39 Figure 1-17. Velocities of total body lean body mass (LBM) and total body bone mineral content (BMC) velocity during puberty in males and females. From Rauch et al. (334) with permission from Elsevier 51 Figure 1-18. Baseline (A) and 20-month absolute change (B) values for cortical area to muscle area ratio (CoA / MCSA, bone-muscle strength index for compression) for girls (dotted lines) and boys (solid lines) across early pubertal (EARLY), peri-pubertal (PERI) and post-pubertal (POST) maturity groups. Bars indicate 95% confidence intervals, (a) Boys > Girls, P < 0.05. Adapted from Macdonald et al. (24) with permission from Elsevier 52 Figure 1-19. Growth curves for femoral neck (narrow neck region) section modulus (Z) comparing 17 active boys to 17 inactive boys (A) and17 active girls to 17 inactive girls (B). Values for Z represent adjusted means (height, weight) and are plotted against biological age groups (years from age of peak height velocity, APHV). Adapted from Forwood et al. (339) with permission from Elsevier 57 Figure 1-20. Side-to-side differences (%) in cortical area (CoA) and torsional bone strength index (BSIt) at the humeral midshaft in young starters (began training before menarche), old starters (began training after menarche) and controls (no training). Bars are 95% confidence intervals. Adapted from Kontulainen et al. (153). 59 Figure 1-21. Nine-month percent change in femoral neck bone mineral content (BMC) (A) and tibial midshaft bone strength index (BSI) in pre- and postmenarcheal intervention (shaded bars) and control (white bars) girls. Significance values for group comparisons within the pre- and postmenarcheal groups are provided. Bars are 95% confidence intervals. Adapted from Heinonen et al. (11) 63 Figure 2-1. The six Action Zones of Action Schools! BC 91 Figure 2-2. Flowchart outlining the organizational structure of Action Schools! BC (AS! BC) 93 Figure 2-3. Study timeline 94 Figure 2-4. Images of (A) total body, (B) lumbar spine and (C) proximal femur scans acquired with the Hologic 4500W bone densitometer. The lines on each image outline the regions of interest defined in the DXA analysis. 101 Figure 2-5. Peripheral quantitative computed tomography (pQCT) scan acquisition. (A) The child sits comfortably in a chair with adjustable height, places their left leg through the pQCT gantry, rests their leg in the customized leg hold and rests their left foot in the foot rest. A Velcro strap is fastened around the left knee and foot to minimize xi movement during scan acquisition. (B) A scout scan is performed to determine placement of the reference line at the distal surface of the tibial plafond. (C) Single 2.3 mm slice measurements are obtained at the distal (8%), midshaft (50%) and proximal two-thirds (66%) sites. Figures (A) and (C) from Vicky Earle, Medical Illustrator, UBC, The Media Group 103 Figure 3-1. Flow of participants through trial (simplified view) 119 Figure 3-2. Flow of participants through trial (detailed view according to CONSORT guidelines (1)) 120 Figure 4-1. Scatterplots of muscle cross-sectional area (MCSA) and distal tibia (A) total bone cross-sectional area (ToA) and (B) total density (ToD) for boys (solid squares, solid line) and girls (open circles, dashed line) 134 Figure 4-2. Scatterplots of muscle cross-sectional area (MCSA) and (A) distal tibia bone strength index (BSI) and (B) tibial midshaft cortical area (CoA) for boys (solid squares, solid line) and girls (open circles, dashed line) 135 Figure 4-3. Scatterplots of muscle cross-sectional area (MCSA) and (A) cortical density (CoD) and (B) polar strength-strain index (SSIP) at the tibial midshaft for boys (solid squares, solid line) and girls (open circles, dashed line). 136 Figure 4-4. A functional model of bone development based on the mechanostat theory (22) and related approaches (37,38). The central component of the regulation of bone development is the feedback loop between bone deformation (tissue strain) and bone strength. During growth, this homeostatic system must continually adapt to external challenges (increases in bone length and muscle force) to keep tissue strain close to a preset level (sensitivity setpoint). Various modulating factors influence aspects of the regulatory system as indicated by the dashed arrows. Adapted from Rauch et al. (21) and Petit et al. (36) 140 Figure 5-1. Adjusted change in (A) bone strength index (BSI) at the distal tibia and (B) polar strength strain index (SSIp) at the tibial midshaft between prepubertal (PRE) and early pubertal (EARLY) Intervention (INT) and Control (CON) boys. Bars are 95% confidence interval and p value is for the group x maturity interaction. Change in bone strength indices adjusted for baseline bone value, tibial length change, and muscle cross-sectional area change 161 Figure 6-1. Results from the subgroup analysis showing adjusted change in femoral neck cross-sectional area (CSA), subperiosteal width (SPW) and section modulus (Z) for Control and Intervention girls. The INT group includes only those girls (n = 43) whose teacher reported at least 80% compliance with Bounce at the Bell. Bars are 95% confidence intervals and values are adjusted for baseline height, change in height, change in total body lean mass and final Tanner stage 186 xii GLOSSARY OF TERMS AND ABBREVIATIONS Bone Definitions ABBREVIATION DEFINITION ~ pQCT Peripheral quantitative computed tomography. The XCT-2000 model is used in this thesis. ToA Total bone cross-sectional area (mm2) as measured with pQCT. CoA Cortical bone cross-sectional area (mm2) as measured with pQCT. CavA Area of the marrow cavity (mm2) as calculated with pQCT outcomes of ToA and CoA. (CavA = ToA - CoA) TrbA Trabecular cross-sectional area (mm2) as measured with pQCT. CTh Cortical thickness (mm) as measured with pQCT. ToD Total bone mineral density (mg/cm3) as measured with pQCT. CoD Cortical bone mineral density (mg/cm3) as measured with pQCT. TrbD Trabecular bone mineral density (mg/cm3) as measured with pQCT. BSI Bone strength index (mg2/mm4) as calculated with pQCT outcomes of ToA and ToD (BSI = ToA * ToD2). SSIp Polar strength-strain index (mm3) as measured by pQCT. Also known as the density-weighted polar section modulus. SSIx Strength-strain index (mm3) with respect to the x-bending axis as measured by pQCT. SSly Strength-strain index (mm3) with respect to the y-bending axis as measured by pQCT. MCSA Muscle cross-sectional area (mm2) as measured by pQCT. DXA Dual energy x-ray absorptiometry. The Hologic QDR 4500W model is used in this thesis. BA Bone area (cm2) as measured by DXA. BMC Bone mineral content (g) as measured by DXA. aBMD Areal bone mineral density (g/cm2) as measured by DXA. HSA Hip structure analysis. Refers to the computer algorithm applied to proximal femur DXA images to estimate bone structural variables. Version 3.0 is used in this thesis. NN Narrow neck. Refers to the narrowest region of the femoral neck that is analyzed with HSA. CSA Cross-sectional area (cm2) as estimated by HSA. CSMI Cross-sectional moment of inertia (cm4) as measured by HSA Z Section modulus (cm3) as measured by HSA and (mm3) as measured by pQCT TERM DEFINTION ..' Areal bone The ratio of bone mineral content to the projectional area of bone (g/cm2) as measured by mineral density DXA. Bone architecture The size and shape of a whole bone. Can also refer to properties of cortical or trabecular bone (i.e., cortical thickness, trabecular number or thickness). Glossary continued TERM DEFINITION Bone (cross- The surface dimensions of bone that quantify the amount of bone surface and its sectional) distribution about torsion and bending axes (i.e., cross-sectional area, cross-sectional geometry moment of inertia). Bone mass Amount of bone material within a cross-section or region of interest (e.g. BMC by DXA). Bone strength Ultimate failure load of bone. In this thesis bone strength is estimated with pQCT-derived BSI, SSIp, SSIx and SSIy and HSA-derived Z. Bone structure Properties of bone such as size, shape and distribution of material that contribute to bone strength. Volumetric bone The amount of bone mineral averaged over a certain volume as measured by pQCT. Can density be cortical (CoD), trabecular (TrbD) or total bone mineral density (ToD). Maturity Definitions TERM DEFINITION Tanner Staging A self-report method of assessing the stage of reproductive (or sexual) maturity in girls and boys. In this thesis Tanner stage for girls refers to breast stage and Tanner stage for boys refers to pubic hair. Prepuberty (PRE) Tanner stage 1. Early puberty (EARLY) Tanner stage 2 or 3. Peri-puberty (PERI) Refers to a broad category of Tanner stages 2-4 (after prepuberty but before post puberty). Post puberty (POST) Tanner stage 5 Premenarcheal Refers to a girl who has not yet experienced her first menstrual period. Postmenarcheal Refers to a girl who has experienced her first menstrual period. Age at peak height An indicator of somatic maturity in longitudinal studies of child and adolescent velocity (PHV) growth. xiv PREFACE: PUBLICATIONS ARISING FROM THIS THESIS Sections of this thesis have been published as multi-authored manuscripts in peer-reviewed journals and are indicated with a * beside the publication below. Details of the authors' contributions are provided, where relevant. I agree with the seated contributions of the thesis author, as indicated below. Dr. Heather McKay (Thesis supervisor) Published Papers *Macdonald HM, Kontulainen SA, Petit MA, Janssen PA, McKay HA. 2006. Bone strength and its determinants in pre- and early pubertal boys and girls. Bone. In Press. Authors' contributions: Heather Macdonald was responsible for the original ideas behind the paper, analysis and presentation of findings and writing and editing of the original paper. Saija Kontulainen and Moira Petit stimulated discussion of results and provided editorial assistance. Patti Janssen provided statistical consultation and editorial assistance. Heather McKay guided all aspects of the research and was the key editor of this manuscript. Rhodes RE, Macdonald HM, McKay HA. 2006. Predicting physical activity motivation and behaviour among children in a longitudinal sample. Social Science and Medicine; 62:3146-56. *Naylor PJ, Macdonald HM, Reed KE, McKay HA. 2006. Action Schools! BC: A socio-ecological approach to modifying chronic disease risk factors in elementary school children. Preventing Chronic Disease [serial online]. Available from: URL: http://www.cdc.qov/pcd/issues/2006/apr/05 0090.htm. *Naylor PJ, Macdonald HM, Zebedee JA, Reed KE, McKay HA. 2006. Lessons learned from Action Schools! BC -An 'active school' model to promote physical activity in elementary schools. Journal of Science and Medicine in Sport. In Press. McKay HA, Reed KE, Macdonald H, Khan KM. 2003. Exercise interventions for health: time to focus on dimensions, delivery and dollars. British Journal of Sports Medicine; 37:98-99. Papers submitted Macdonald HM, Kontulainen SA, Khan KM, McKay HA. Is a school-based physical activity intervention effective for increasing tibial bone strength in boys and girls? Submitted to the Journal of Bone and Mineral Research, June 2006. Authors' contributions: Heather Macdonald was responsible for the original ideas behind the paper, analysis and presentation of findings and writing and editing of the original paper. Saija Kontulainen and Karim Khan stimulated discussion of results and provided editorial assistance. Heather McKay guided all aspects of the research and was the key editor on this paper. Macdonald HM, Kontulainen SA, Beck TJ, McKay HA. Effectiveness of a school-based physical activity model for increasing femoral neck bone mass and strength in boys and girls: The Action Schools! BC model. Submitted to Bone, May 2006. Authors' contributions: Heather Macdonald was responsible for the original ideas behind the paper, analysis and presentation of findings and writing and editing of the original paper. Saija Kontulainen stimulated discussion of results and provided editorial assistance. Tom Beck provided training and guidance with HSA analysis, stimulated discussion of results and provided editorial assistance. Heather McKay guided all aspects of the research and was the key editor on this paper. Petit MA, Macdonald HM, McKay HA. Growing bones: How important is exercise? Submitted to Current Opinion in Orthopaedics, June 2006. xv Ahamed YA, Macdonald HM, Reed KE, Naylor PJ, McKay HA. Physical activity does not compromise academic performance of school children. Submitted to Medicine and Science in Sports & Exercise, February 2006. Abstracts Ahamed Y, Macdonald HM, Reed KE, Naylor PJ, McKay HA. 2006. Time devoted to physical activity does not compromise academic performance of elementary school children. Medicine and Science in Sports and Exercise; 38 (S5). Macdonald HM, Manske SA, Reed KE, Khan KM, McKay HA. 2005. Action Schools! BC: Daily physical activity increases bone strength in prepubertal boys. Journal of Science and Medicine in Sport 8;4(Suppl):153. Presented orally at the Australian Conference of Science and Medicine in Sport, Fifth National Physical Activity Conference. Recipient of Asics Medal for Best Paper Overall. Macdonald HM, Kontulainen SA, Petit MA, McKay HA. 2005. Determinants of bone geometry, density and strength in girls and boys. Journal of Bone and Mineral Research; 20 (S1): S33. Presented orally at the American Society for Bone and Mineral Research Annual Meeting, September 2006, Nashville TN. Young Investigator Award recipient. Kontulainen S, Liu D, Jamieson M, Macdonald H, Manske S, Oxland T, McKay H. 2005. Defining cortical bone by pQCT: In vitro and in vivo accuracy of bone geometry. Journal of Bone and Mineral Research; 20 (S1): S338. Macdonald HM, Kontulainen SA, Petit MA, McKay HA. 2005. Action Schools! BC: Effects of a 16-month school-based physical activity intervention on bone geometry, strength and the muscle-bone unit in pre- and early pubertal girls. Bone; 36:S56. Presented in poster format at the 3rd International Conference on Children's Bone Health, Sorrento, Italy, May 2005. Recognized as Selected Poster. xvi ACKNOWLEDGEMENTS This thesis would not have been possible without the generous support of many people. First, I would like to gratefully acknowledge my supervisor, Dr. Heather McKay. Thank you Heather, for your expert guidance and continual encouragement and for always challenging me to reach outside my comfort zone. Your dedication to research has inspired me and your contribution to my growth both as a researcher and as an individual are immeasurable. Thank you to my other committee members, Dr. Karim Khan and Dr. Patti Janssen, for sharing your insight and knowledge and for helping me to see the big picture. Dr. Penny Brasher, thank you for your statistical help and for introducing me to the wonderful world of STATA. Thank you to Dr. Tom Beck and Lisa Semanick for your excellent tutelage and hospitality during my time in Baltimore. I could not have asked for a better project to work on for the past 4 years. Action Schools! BC represents the efforts of an incredible group of people and I feel privileged to have had the chance to work closely with them. Kate Reed, thanks for coming to Vancouver when you did! You were a great companion on the many school visits and it was a pleasure to discuss stats with you over many tall Americanos. Bryna Kopelow, Jennifer Fenton and Sydney Millar at JW Sporta, the passion and enthusiasm you bring to your work are inspiring. Thank you for welcoming me into the mix and for developing such an amazing program. Debbie Keel and Judy Howard, thanks for making sure the classes did their jumps properly! PJ Naylor, thank you for the great data discussions. Josie McKay and Connie Waterman, without you this study could not have run so smoothly. Thank you for your time and efforts and your superb scheduling abilities! Thank you to the UBC Graduate Fellowship for supporting me throughout my studies. And of course, a huge thanks to the 514 children and their parents as well as the teachers and administrators who participated in Action Schools! BC. I was fortunate to have three amazing women as mentors and friends throughout this journey. Dr. Kerry MacKelvie O'Brien, thank you for providing great footsteps for me to follow in and for being a trusted friend in and out of the lab. Dr. Moira Petit, thank you for helping me to keep everything in perspective, for being an excellent reviewer and a wonderful traveling companion. And Dr. Saija Kontulainen, I cannot imagine what these last few years would have been like without you! Your enthusiasm for research is incredible and I thank you for sharing it with me. Thank you for all that you have contributed to this thesis - from the many discussions of pQCT analysis to your careful review of the manuscripts. Kiitos! To all of my amazing friends and colleagues in the Bone Health Research Group and the Division of Orthopaedic Engineering Research - to name you all would be a thesis in itself, but thanks to everyone for the great discussions, many laughs and of course, the coffee breaks! Special thanks to my officemates in Room 590 for always providing a listening ear and for sharing my sweet tooth. Thank you to my family - Mom, Dad, Roy, Jayne and Polly Ann - for your unconditional love and support and most of all, your patience. Thank you to the girls - Emily, Hayley and Nicole for always being just a phone call away. I treasure our friendship and hope that one day the calls won't be long distance! And last, but certainly not least, Jamie. Thank you for your unwavering support and endless encouragement and for making me laugh every day. I look forward to starting our post-grad school life together. xvii Chapter 1 - Introduction 1 Introduction, Literature Review, Rationale, Objectives & Hypotheses 1.1 Introduction Osteoporosis and related fractures are serious societal health burdens. In particular, hip fracture is a significant cause of morbidity and mortality worldwide (1). Annual direct health care costs of hip fracture in Canada are approximately 650 million dollars and in the absence of effective preventive strategies, this number is expected to rise to 2.4 billion dollars by 2041 (2). Based on the current level of evidence, regular physical activity may be the most cost-effective, safe and readily available method to prevent both osteoporosis and low-trauma falls that lead to osteoporotic fracture (3). More specifically, physical activity during childhood can result in substantial bone health benefits, which in turn may reduce the risk of osteoporosis later in life (3-6). Unfortunately, recent trends suggest that more than 50% of Canadian children are not active enough for optimal growth and development (7). Thus, there is an immediate need for effective interventions to promote physical activity and bone health among youth. Schools, by nature, have a captive audience and are therefore an excellent arena in which to encourage positive physical activity behaviours. To date, the longest school-based exercise intervention (20 months) resulted in significant gains (+4.3-4.6%) in bone mineral content (BMC) as measured by dual energy x-ray absorptiometry (DXA) at the clinically significant femoral neck in boys (8) and girls (9). Despite the effectiveness of this program, and others like it (10-12), it is not likely to be sustainable in the elementary school setting due to the time required (10-20 minutes, 3 times per week) and the need for equipment and specially trained teachers. For a school program to gain acceptance on a population-wide basis, it must be effective, simple to administer, short in duration and possible to perform in the classroom (13). Recently, a school-based bone-loading program, Bounce at the Bell, was developed to meet these criteria (13). Bounce at the Bell was designed based on promising results from animal studies that have demonstrated significant gains in bone strength with short bouts of loading (14) and an improved osteogenic response to exercise sessions separated by rest periods (15). This program was effective for increasing proximal femur bone mass in boys and girls; however, its effects on bone structure and strength are not known. In addition to changes in bone mass that occur in response to exercise, it is equally important to characterize bone structural changes as both properties contribute to ultimate bone strength (16). Currently, DXA is the most widely used method to evaluate bone mass in clinical and research settings (17). However, due to its planar technology, DXA is unable to assess bone structure or separate cortical and trabecular bone compartments which may respond differently to loading. A computer software algorithm, hip structure analysis (HSA), can be used to supplement conventional DXA measures of BMC with estimates of proximal femur bone geometry and strength (18), but it to is limited by DXA's planar technology. Peripheral quantitative computed tomography (pQCT) provides a means with which to directly evaluate cross-sectional geometry and (volumetric) bone mineral density of the appendicular skeleton, which together can be used to estimate bone strength (17,19). A number of cross-sectional (20-23) and longitudinal (24-26) studies have used pQCT to describe sex- and maturity-related differences in long bone development in the upper and lower extremities, but few pediatric intervention studies have used pQCT to determine the effect of high-impact physical activity on bone strength (11,27). 1 Chapter 1 - Introduction An additional advantage of pQCT is its ability to assess the muscle-bone relationship. Muscle contraction incurs the largest physiological load on the skeleton (28,29) and during growth, bones must continually adapt their geometry and mass to withstand loads from increases in both stature and muscle forces (30,31). Thus, there is a need for pediatric bone densitometric data to be interpreted in the context of these mechanical challenges (31). Muscle cross-sectional area as measured with pQCT provides a surrogate for muscle force (32) and together with pQCT-derived bone geometry and strength can be used to gain further insight into the functional muscle-bone unit. Thus, the primary purpose of my thesis is to determine the effectiveness of a school-based physical activity intervention, which includes a high-impact jumping program, for increasing bone strength in boys and girls. The secondary focus is to define the muscle-bone relationship in the weight-bearing tibia in boys and girls. To provide a comprehensive evaluation of the growing skeleton I employ two novel bone imaging techniques, pQCT and HSA, in addition to DXA. My thesis is outlined in three parts. Part I describes sex differences in pQCT-derived tibial bone strength in pre- and early pubertal children and identifies determinants of bone strength in both boys and girls. Part II defines the site- and sex-specific bone structural response to a 16-month, school-based physical activity intervention using pQCT. Part III investigated the sex-specific effect of the 16-month physical activity intervention on bone structural adaptations and bone mineral accrual at the clinically relevant proximal femur using both HSA and DXA. I provide relevant background literature in Chapter 1 including bone biology and biomechanics, measurement of bone mass and strength in children, sex- and maturity-related differences in skeletal development, determinants of bone strength in children with a focus on the role of muscle and the influence of physical activity on bone health in children. I conclude Chapter 1 with the rationale, specific objectives and hypotheses for the three studies that comprise this thesis and in Chapter 21 provide a detailed description of the physical activity model and the methods employed. Chapter 3 describes the cohort of children who participated in the study. Chapters 4-6 provide the results of the three studies that are In Press (Part I) or have been submitted for publication (Parts II and III). Finally, in Chapter 7 I discuss the three studies as an integrated whole and propose future directions for pediatric bone research. In addition, I summarize the results of this thesis and present conclusions. 2 Chapter 1 - Literature Review 1.2 Literature Review In this chapter I discuss relevant background information in 6 key areas: bone biology, bone biomechanics, measurement of bone health in children, maturity- and sex-related differences in skeletal development, determinants of bone strength in children and finally, the influence of physical activity on pediatric bone health. 1.2.1 Bone Biology and Bone Growth Bone is a dynamic tissue with multiple functions. The primary function of bone is to be stiff and strong in order to resist deformation resulting from internal (mainly muscular) and external loads (33). In addition, bones serve as levers for locomotion, attachment sites for muscles, ligaments and tendons and as a central reservoir for calcium while also providing a site for haematopoiesis (formation of blood cells) and protecting organs (34,35). During growth, the skeleton must maintain these functions while dramatic changes in size and shape occur. Within the skeleton, the structure of bone tissue, and of whole bones, is complex and ultimately influences bone's mechanical properties. 1.2.1.1 Bone Tissue: Composition and Organization Bone is a composite material comprised of an organic and inorganic form (36). The inorganic, or mineral, phase is composed mainly of a specific crystalline hydroxyapatite (calcium and phosphorus) while the organic phase is composed mainly of type I collagen. Collagen is organized into fibres, which are the major structural component of the bone matrix and give bone its flexibility and tensile strength. Collagen also provides a location for the deposition of the inorganic mineral crystals, which give bone rigidity and compressive strength (36). Two types of bone tissue exist: woven and lamellar bone (36). Woven bone is a quickly formed, poorly organized tissue in which collagen fibres and mineral content demonstrate a random distribution. Woven bone is considered immature and is the only form of bone present in the embryonic and newborn skeleton. It can also be found in fracture callus, certain metaphyseal regions of the growing skeleton, certain bone tumours and patients with osteogenesis imperfecta (36). In contrast, lamellar bone is a slowly formed, highly organized material that results from remodeling of woven or pre-existing bone. The collagen and associated mineral are arranged in sheets (lamellae) which are organized according to the stress orientation of the collagen fibres (36,37). Differences in the structural properties of woven and lamellar bone are reflected in their mechanical characteristics. Woven bone is isotropic, which means the mechanical behaviour is similar regardless of the orientation of the force. Lamellar bone is anisotropic, which means the mechanical behaviour differs according to the orientation of the force (36). Within the appendicular and axial skeleton, woven and lamellar bone are organized into cortical (or compact) and trabecular (spongy or cancellous) compartments (36). Although cortical and trabecular bone are made of the same material, there are structural and functional differences between them (38). Trabecular bone exists as a three-dimensional lattice structure composed of individual trabeculae and is found at the ends of long bones and in the vertebral bodies. The lattice organization determines the porosity of trabecular bone (75-95%), provides a vast surface area where metabolic activities such as bone turnover occur and houses bone marrow which functions in haematopoiesis. 3 Chapter 1 - Literature Review Conversely, cortical bone is arranged in cylinders, is much less porous (5-10%) than trabecular bone and forms the diaphysis of long bones (39). Haversian bone is the most complex form of cortical bone, and is arranged in osteons, which consist of vascular channels circumferentially surrounded by lamellae of bone. The dense arrangement of osteons within cortical bone confers a higher degree of calcification when compared to trabecular bone (80-90% for cortical bone vs. 15-25% for trabecular bone). As a result cortical bone fulfills mainly, but not only, mechanical and protective functions (38). I discuss the structural and mechanical properties of trabecular and cortical bone in more detail in Section 1.2.2. 1.2.1.2 Whole Bone Structure Within the human skeleton, bones can be roughly grouped as either long (e.g. tibia), short (e.g. metacarpal), flat (e.g. skull or scapula), irregular (e.g. vertebrae) or sesamoid (e.g. patella) (29). In this thesis, I focus on long bones. As illustrated in Figure 1-1 a growing long bone consists of a tubular diaphysis that flares into two wide ends (the epiphyses). The epiphyses are separated from the funnel-shaped metaphyses by a layer of cartilage known as the growth plate which is the site of endochondral ossification (Section 1.2.1.3.1). The outer portion of the diaphysis contains cortical bone, whereas the inside contains bone marrow and is known as the medullary or marrow cavity. In contrast, the epiphyses are filled with moderately thick low-turnover trabeculae and the metaphysis is a transitional region of trabecular and cortical bone (40). The broad shape of bone ends serves to better distribute joint forces and reduce stress (force per unit area) that is transmitted by trabecular bone in the metaphysis to cortical bone in the diaphysis (41). Cancellous Bone Diaphysis 1 Fused Growth Plate Figure 1-1. Illustration of a growing long bone. Adapted from Jee et al. (42). 4 Chapter 1 - Literature Review The organization of the cortical and trabecular components within a long bone results in two bone surfaces that are in contact with soft tissue (Figure 1-1) (38). The external (periosteal surface or periosteum) and internal (endosteal surface or endosteum) surfaces of cortical bone are lined with osteogenic cells, and are therefore sites of bone tissue turnover. There are three cell types found in bone; osteoblasts, osteoclasts and osteocytes, each with specific roles in regulating bone turnover (36). Osteoblasts, or bone-forming cells, secrete osteoid, the unmineralized protein component of the bone matrix that forms the basic framework of bone tissue. Once an osteoblast has secreted and mineralized the osteoid, it becomes an osteocyte, the most abundant cell in fully formed bone (42). The osteocytes remain in contact with osteoblasts through gap junctions, and this linkage is thought to be important in transducing mechanical signals into biological activity (36). Osteoclasts, or bone resorbing cells, are usually found in cavities on bone surfaces called resorption pits. Osteoclasts secrete lysosomal enzymes and hydrogen ions that together, work to dissolve the bone matrix. During growth, osteoblasts and osteoclasts may function independently to modify the size and shape of bones during bone modeling or their cellular activities may be combined in the basic multicellular unit (BMU) which is responsible for bone remodeling (43). I discuss bone modeling and remodeling in further detail in Sections 1.2.1.3.2 and 1.2.1.3.3. 1.2.1.3 Physiology of Bone Growth and Bone Turnover In this section I discuss the three processes that determine and maintain the architectural structure of bone: growth, modeling and remodeling. 1.2.1.3.1 Longitudinal Bone Growth Growth refers to the enlargement of bones in width and length due to an increase in cell number, and occurs from birth through maturity (42). Longitudinal bone growth in the appendicular skeleton occurs via endochondral ossification, which is defined as the process of bone formation from pre-existing calcified cartilage (42). During bone development, endochondral ossification occurs in the epiphyseal growth plate (Figure 1-1) (44). Initially, the cartilage is calcified (primary spongiosum or primary trabeculae) and this allows for the deposition of more calcified material in the form of woven bone (secondary spongiosum or secondary trabeculae). Chondrocytes (cartilage producing cells) within the growth plate are organized according to their stage of maturation, with the most mature cells in the calcifying zone being incorporated into metaphyseal bone (44). With increasing distance to the growth plate, metaphyseal trabeculae located in the centre of the bone are thinned out and eventually resorbed leaving the diaphysis devoid of trabeculae (45). In contrast, metaphyseal trabeculae on the periphery serve to transfer loads from the growth plate to the metaphyseal cortex. Eventually the peripheral trabeculae coalesce and become part of the metaphyseal cortex (46). Growth plate activity varies with age and the contribution of the distal and proximal growth plate to overall longitudinal growth varies between bones (47,48). For example, for the tibia, the proportion of growth at the proximal growth plate in girls varies from 50% at age 7 to 80% at age 14 (48). For boys, the proportion of growth at the proximal growth plate varies from 50% at age 7 to 80% at age 16. Similar variation with age occurs in the femur; 5 Chapter 1 - Literature Review however, approximately 70% of femoral growth occurs at the distal growth plate (48). In the upper extremities, the proximal growth plate accounts for approximately 80% of growth in the humerus, whereas the distal growth plate accounts for 80% and 85% of growth in the radius and ulna, respectively (47). Regulation of longitudinal growth is thought to occur on three levels: systemic, local and mechanical (45). At the systemic level growth hormone (GH), insulin-like growth factor-l (IGF-1), thyroid hormones and glucocorticoids regulate longitudinal growth during childhood, whereas sex hormones (estrogen and testosterone) are more influential during puberty (49). In boys and girls, estrogen is pivotal for epiphyseal fusion which, until recently, was commonly believed to be the central determinant of cessation of linear growth (50). More recent evidence suggests that epiphyseal fusion is only a marker of growth cessation that occurs after a decline in chondrocyte proliferation and subsequent 'growth-plate senescence' (51). Regulation of chondrocyte proliferation also occurs at the local level (within the growth plate) and involves a number of different growth factors (e.g., fibroblast growth factors, Indian hedgehog) (49). Little is known about mechanical regulation of longitudinal growth; however, according to Harold Frost's chondral growth response curve (45,52), mild tension and compression are thought to increase longitudinal growth while severe compression is thought to inhibit growth. 12.13.2 Bone Modeling During growth, the size and shape of bones is modified through modeling, which involves the independent actions of osteoblasts and osteoclasts on bone surfaces (29). Continuous addition of bone to the periosteal surface by osteoblasts and simultaneous endosteal resorption by osteoclasts contributes to diaphyseal enlargement. This motion of surfaces in tissue space is known as drift, and may increase or decrease bone curvature according to the specific mechanical needs of the bone (29,53). Bone modeling during growth is thought to be regulated by mechanical strain (54). I discuss this in more detail in Section 1.2.2.2. At metaphyseal sites, modeling also serves to decrease the diameter of newly formed metaphyseal bone (metaphyseal inwaisting) to match the cross-sectional size of the diaphysis (Figure 1-2) (53,55). At the distal radius (4% site), the rate of periosteal resorption is estimated to be 8 um/day, while the rate of endocortical apposition is about 9.5-10 um/day. The high rate of endosteal apposition is necessary to maintain the cortical thickness at this site (55). Once skeletal maturity is reached, the modeling rate is greatly reduced. 6 Chapter 1 - Literature Review © © c a Figure 1-2. Formation (+) and resorption (-) during longitudinal bone growth. During growth from A to B, the funnel like metaphysis is reduced to match the shape of the narrower diaphysis through osteoclastic resorption on the periosteal surface of the metaphysis (a). Thickening of the cortex occurs along the cortical endosteal surface of the metaphysis and enlargement of the marrow cavity occurs by resorption of metaphyseal trabecular bone (b). The diameter of the diaphysis increases by periosteal bone formation (c) and the marrow cavity of the diaphysis expands by bone resorption on the endosteal surface (d). Adapted from Jee (42) and Baron (38). 12.13.3 Bone Remodeling Unlike modeling, bone remodeling continues throughout life. The primary function of remodeling is to maintain load bearing capacity, and this is accomplished by preventing and/or repairing fatigue damage through the replacement of old bone with newly formed bone (34). Remodeling occurs on bone surfaces (periosteal, endosteal, trabecular) and within cortical bone, and involves the tightly coupled activity of osteoblasts and osteoclasts in one unit, the basic multicellular unit (BMU). The BMU follows a regulated sequence of activation, resorption and formation during which time (~4 months: 3 weeks for resorption, 3 months for formation) a volume of damaged bone is removed by osteoclasts and replaced by osteoblasts (34). Activation occurs in the presence of chemical (hormones) or mechanical signals. In cortical bone, initiation of remodeling may also be related to the emergence of microcracks under conditions of fatigue loading. Specifically, osteocyte-mediated signalling pathways (56) and interstitial fluid flow (57) are thought to be two mechanisms associated with the activation of repair-damage remodeling. During growth there is a positive balance between bone formation and resorption due to high activation frequency (58). In contrast, during ageing there is a negative imbalance between osteoclast and osteoblast activity arises such that with each progressive remodeling cycle bone resorption exceeds formation. This uncoupling of bone formation and resorption may result from the absence of estrogen at menopause (29) or immobilization/disuse (59). Within cortical bone, the BMU proceeds in the form of a tunnel at a rate of approximately 40 um/day and in adults is responsible for replacing about 5% of cortical bone each year (29). Cortical bone remodeling (intracortical or osteonal remodeling) results in the formation of secondary osteons with new Haversian canals that are bounded by cement lines. The activation frequency of cortical bone formation is highest during childhood as the primary osteons produced during modeling processes are quickly converted to secondary osteons (29). The greater activation 7 Chapter 1 - Literature Review frequency during growth may be associated with increased levels of growth hormone, or age-related variations in the material properties of the bone matrix. Trabecular remodeling occurs in a similar fashion; however, BMUs work on the surface of trabeculae by digging and refilling trenches known as Howship lacunae (29). The rate of trabecular bone turnover (-25% per year in adults) is considerably higher than that of cortical bone; however, the rate is known to vary throughout the skeleton (29). In the growing skeleton a given location of trabecular bone surface undergoes approximately 1.04 remodeling cycles per year (60). During remodeling, a number of BMUs are in the resorption phase, while others are in the formation phase. Therefore, at sites where remodeling is occurring there is a temporary loss of bone, or undermineralization (61). The immature skeleton tends to be more undermineralized than the mature skeleton due to the high rate of BMU activation associated with rapid longitudinal growth. This temporary deficit results in a low elastic modulus (stiffness) and higher strains for a given load (29). The increased strains may in turn increase fatigue damage and lead to increased activation frequency. 12.2 Bone Biomechanics In this section I discuss whole bone mechanical properties as well as mechanical properties of cortical and trabecular bone. I then discuss theories of bone adaptation to mechanical stimuli and evidence from animal research that has furthered our understanding of the skeletal response to weight-bearing activity. 1.2.2.1 Mechanical Properties of Bone Functionally, the most relevant property of bone is its strength. From an evolutionary standpoint, selection favours mechanisms that maintain bone's mechanical integrity by whatever means possible (31). Therefore, the goal during development is to create a strong skeleton that will withstand functional mechanical loads and prevent fracture later in life. In fashioning a stronger skeleton, bone development is controlled by the functional requirements of bone as an organ (31). The mechanical properties of long bones (stiffness, flexibility, strength and lightness) balance conflicting demands on the skeleton. For example, bone must be stiff to ensure efficient muscle action and yet compliant (less stiff) to absorb energy and avoid fracture (29). The mechanical behaviour of whole bones under conditions of experimental or physiological loads is dependent not only on the mass and material properties, but also on bone geometry and architecture (16). Whereas the material properties of bone are classically defined by performing standardized mechanical tests on uniform specimens of bone, structural properties of bone are determined by whole bone structural tests (16,62). Such mechanical tests involve applying different types of loads, such as compression, bending and/or torsion, to whole bones from different skeletal regions including the lumbar vertebrae and long bones (62). Data generated from a mechanical test is used to generate a load-deformation curve, which defines the extrinsic (structural) properties of the bone including whole bone strength, stiffness and work (or energy) to failure. When the extrinsic properties are normalized for bone size (cross-sectional area, CSA), the load-deformation curve is converted to a stress-strain 8 Chapter 1 - Literature Review curve, which defines the intrinsic (or material) properties of bone (63). The intrinsic properties include the ultimate stress (strength), elastic modulus and toughness. 1.2.2.1.1 Material Properties The material properties of bone are the characteristics at the tissue level that contribute to overall bone strength. To better understand these properties it is important to discuss the underlying fundamental biomechanics. Under conditions of loading, bone will experience deformation from its original dimensions. This phenomenon is known as strain, and is equivalent to the change in length of the bone divided by the original length (16). The intensity of the load applied is referred to as stress, and is measured by force divided by the area of bone over which it acts (64). Bone can experience three types of stress: tension, compression and shear, which may occur independently or in combination according to specific loading configurations (65). The material properties of stiffness, strength and toughness can be derived from the stress-strain curve (Figure 1-3). The slope of the linear portion of the curve represents the material stiffness, or the modulus of elasticity (elastic region). Before the yield point, any deformation experienced by the bone will be temporary meaning it will return to its original shape once the load is removed. After the yield point, the load will cause permanent deformation (plastic region) up to the point of maximum stress which ultimately results in bone failure. The area under the curve represents the material toughness; a tougher bone will be more resistant to fracture (62). The design of long bones is an appropriate combination of stiffness and toughness which, in the healthy skeleton, allows bones to bear the loads imposed upon them (41). Strain Figure 1-3. A standard stress/strain curve of a bone specimen produced during mechanical testing. This curve can also be used to represent whole bone properties (load/deformation curve). Adapted from Einhorn (16). 9 Chapter 1 - Literature Review The mechanical properties of cortical and trabecular bone are influenced by the material properties of each bone compartment. Within cortical bone, characteristics of individual secondary osteons including the collagen fibre orientation as well as the overall cortical porosity and mineralization are known to be key determinants of mechanical integrity (29). In addition, microdamage due to fatigue may also alter cortical bone material properties. As cortical bone is a viscoelastic, or time-dependent, material its mechanical properties are highly dependent on strain rate (29). The early cadaveric work of Currey and Butler (66) demonstrated age-related changes in the mechanical properties of bone tissue. The femoral midshaft of 18 cadavers aged 2 to 48 years was assessed under static loading conditions. Bending strength and modulus of elasticity increased with age, whereas the material toughness decreased. Further, significant correlations were found between ash content and bending strength, elastic modulus and material toughness. These results highlight the important relationship between the degree of mineralization and the mechanical properties of bone. In general, the elastic properties of bone are influenced solely by bone's mineral phase. However, the ultimate strength of bone is related to both the content and distribution of bone mineral within the matrix (16,33). Trabecular bone is unique from other biological tissues due to its substantial heterogeneity across sites, ages and species (67). This heterogeneity contributes to differences in mechanical properties, and is a function of underlying variations in porosity (or apparent density), material properties of individual trabeculae (thickness) and orientation (anisotropy) of trabecular architecture (67,68). In the spine and at articulating joints, trabecular bone experiences mainly axial compression; however, axial torsion and horizontal shear stress also contribute to forces incurred on the surface of trabeculae in the axial skeleton (69). The compressive strength of trabecular bone is related to the square of apparent density, whereas elastic modulus demonstrates between a squared and cubic relationship with apparent density (70,71). The architecture of trabecular bone provides the requirements for optimal load transfer by combining appropriate strength and stiffness with minimal weight according to rules of mathematical design proposed by Wolff (72,73). 1.2.2.1.2 Bone Cross-sectional Geometry Within the appendicular skeleton, long bone cross-sectional geometry is complex and varies along the length of the bone. The principle forces experienced at the diaphysis of long bones include axial compression, bending, shear and torsion or twisting (Figure 1-4) (16). Often these loads occur in combination, for example most long bone shafts experience mainly bending, but are also compressed and twisted to varying extents. Resistance to such forces is more a product of the distribution of cortical bone (cross-sectional properties) than the mass or density of the mineralized tissue (16,71). For a long bone such as the tibia, the most efficient cross-sectional shape is one in which the mineralized tissue is placed as far from the neutral axis of the load as possible. This geometric arrangement maximizes the greatest strength/lightness ratio, and is best described by the cross-sectional moment of inertia (/ or CSMI) (Figure 1-5). The polar moment of inertia [J] is calculated as the sum of any two perpendicular measures of I (e.g., Ix + ly) (74). The cross-sectional moment of inertia can also be used to calculate the section modulus (Z) as I / (D/2) where D is the cross-section's diameter in the bending plane (74,75). Age-related changes in 10 Chapter 1 - Literature Review bone formation/resorption on the periosteal and endosteal surfaces contribute to the gain in bone strength during growth and strength maintenance during aging (76). A B CD Figure 1-4. The principle loads experienced by bone in nature. Arrows indicate applied forces in (A) compression, (B) bending (tension on the convex side and compression on the concave side), (C) twisting (or torsion) and (D) shear. Adapted from Kontulainen (77) and Pearson and Lieberman (35). r A\ A-X Xr2A Figure 1-5. The cross-sectional moment of inertia (I, mm4) for a bone cross-section describes the distribution of bone material about a defined axis. The cross-section is divided into many square regions of area A that are located at a distance (r) from the neutral axis for bending (x-axis). Adapted from Martin et al. (29). 11 Chapter 1 - Literature Review 1.2.2.2 Bone Adaptation to Mechanical Stimuli The primary mechanical function of the skeleton is to provide rigid levers for muscles to act against as they work to hold the body upright in the presence of gravitational forces (78). Consequently, the skeleton is continually exposed to a loading environment. The mechanical stimuli encountered throughout life serve to sculpt the skeleton's genetic blueprint to match the loading requirements. This relationship between physical loads and bone structure was theorized over a century ago by Roux (79) (summarized by Roesler (80)) and is commonly referred to as bone functional adaptation (81-84). In a general sense, bone functional adaptation involves feedback loops that serve to maintain an "equilibrium" or "customary" strain level in the presence of bone strain (84). An increase in bone strain (e.g., through an increase in physical activity) results in bone formation, which in turn reduces bone strain to its original customary level. In contrast, a decrease in bone strain (e.g., through physical inactivity) results in bone resorption which again restores strain to the customary level. The customary strain level likely varies by skeletal location (33,82) and as proposed by Frost (30) may be altered by both mechanical and nonmechanical factors. In addition, characteristics of bone strain such as frequency and distribution, as well as the loading history of bone cells also influence the magnitude of the bone response (85,86). 1.2.2.2.1 Mechanotransduction Although the specific cellular mechanisms underpinning bone adaptation are poorly understood, it is known that some form of mechanotransduction is required (78). Mechanotransduction is the conversion of a biophysical force into a cellular response. In bone it is proposed that four phases are involved: mechanocoupling, biochemical coupling, transmission of signal and the effector cell response. Mechanical loads induce strains in bone tissue that are detected by osteocytes, a process that is thought to be mediated by strain-induced interstitial fluid flow (87). Signals are then transmitted from the osteocytes to mechanoreceptors within the cell membrane and cytoskeleton where a signalling cascade is initiated. Once the signal reaches the effector cells (osteoblasts and osteoclasts), bone remodeling begins and ultimately results in architectural changes that adjust bone structure to match the requirements of the mechanical environment (78). 1.2.2.3 Mechanostat Theory As discussed, it is now recognized that the customary strain level and regulation of bone adaptation involves more than simply mechanical strain. The mechanostat theory, proposed by Frost, suggests that the control of skeletal physiology involves the actions of many interlocking, and usually negative feedback loops that are influenced by both mechanical and nonmechanical factors (30,54). Rather than one customary strain level, Frost describes a threshold range that includes a minimum effective strain (MES) for both remodeling (MESr) and modeling (MESm) (Figure 1-6). During growth, the main mechanical challenges of increasing body weight and muscle loads function to increase bone strains towards the MESm and turn modeling on. Once skeletal maturity is reached, peak bone strains are reduced to the level needed to initiate "conservation-mode" remodeling. With decreased activity and with aging, decreasing muscle strength reduces the loads on bone and shifts strains below the MESr. This results in disuse-12 Chapter 1 - Literature Review mode remodeling, which causes a slow loss of bone next to marrow (88). The setpoints of the mechanostat may be altered by nonmechanical agents such as hormones and nutrition (31). It is suggested that estrogen may lower the MESm and MESr (on the endosteal surface, next to bone marrow) so that smaller strains are required to turn modeling and remodeling on (89). Mechanostat theory has been verified in several animal experiments (90-93). For example, in growing dogs, disuse osteopenia in the casted forelimb was the result of both reduced modeling on the periosteal surface (decrease in periosteal expansion) and increased remodeling on the endosteal surface (increased endosteal expansion) (92). Remobilization of the dogs reversed both of these trends such that periosteal apposition increased and endosteal apposition was also restored. Overload Bone Formation Bone Resorption 100 (MESr) 1000 (MESm) 2000 Strain (|je) Figure 1-6. Schematic of the mechanostat theory for bone response to mechanical loading. MESm represents the minimum effective strain for bone modeling and MESr represents the minimum effective strain for bone remodeling. In the adapted state between MESr and MESm, bone turnover is minimal as typical strains change. During growth, increasing body weight and muscle loads should increase strains towards MESm and turn modeling on. In contrast, during aging, decreasing muscle strength should reduce the strains towards MESr and turn disuse remodeling on. Adapted from Frost (54,88). 12.2.3.1 Rules for Bone Adaptation to Mechanical Stimuli The structural changes that result from bone adaptation can be predicted by three fundamental rules outlined by Turner (85): 1) adaptation is driven by dynamic, rather than static, loading; 2) extending the loading duration has a negative effect on bone adaptation, 3) adaptation is "error-driven", meaning abnormal strains drive structural change. These rules are important to consider when designing bone loading interventions. Results from 13 Chapter 1 - Literature Review animal studies provide insight into how such interventions can be optimized to further the anabolic effect of mechanical loading. Rule 1: Dynamic Loading. Studies in both growing (94) and mature (95) rats have demonstrated that bone adapts only in response to dynamic loads. Further, it appears that static loading may actually suppress normal longitudinal growth. Robling and colleagues (94) measured ulnar periosteal bone formation rates in three groups of growing male rats who received either a static (8.5 N or 17 N at 1 N sec1) or dynamic loading (17 N at a frequency of 2 Hz) protocol for 10 min/day for 2 weeks. Periosteal bone formation was suppressed by 28-41% in those rats who received static loading, and at the end of the 2-week period the loaded ulnae was 2 to 4% shorter than the contralateral control ulna in those animal who received the 17 N static load. In contrast, an osteogenic effect of dynamic loading was observed on both the periosteal and endosteal surfaces as bone formation rate increased by 78% and 300%, respectively, compared with the control limb (94). However, similar to the static load group, the dynamic load group also demonstrated significant growth suppression of the loaded limb. This was the result of a reduction in the number of proliferating chondrocyte lacunae in the distal growth plate and an increase in growth plate height. The suppression of longitudinal growth observed in this study may reflect tissue damage due to the high peak strains used in this loading model. In the male rat ulna, 17 N elicits a compressive strain on the medial surface of approximately 3500 ue. This is considerably higher than peak strains recorded during running (1200 UE) and jumping from a 30 cm height (2500 ue) in growing rats (96) and peak strains at the human tibial midshaft during vigorous activity such as downhill running (< 2000 ue) (97). Dynamic loading conditions may involve variation in strain magnitude, strain frequency and/or strain rate; however, evidence suggests that bone adaptation is more responsive to rate-related phenomena (98). Mosley and Lanyon (99) investigated the effects of strain rate on the adaptive modeling response in the ulna of growing male rats subjected to 2 weeks of axial compressive loading. The loading protocol involved three strain rates (low, moderate, high) at a constant frequency of 2 Hz, and similar peak strain magnitudes across the three groups. The strain rate and frequency used were similar to those recorded by strain gauges (implanted on the ulna) during normal activity. At study completion, the high-strain-rate group demonstrated a 67% greater adaptive modeling response (as measured by change in bone volume) than the low-strain-rate group. It is suggested that strain rate may influence the magnitude of the load induced fluid flow, thus impacting the mechanotransduction process (99). A more recent study in growing turkeys suggests that the effects of strain rate under certain loading conditions may be surface-specific (100). Judex and Zernicke subjected 10 young roosters to a 3 week drop jumping program. Roosters performed 200 drop jumps (from 50-60cm) daily. When compared to strain rates associated with baseline walking, the drop jumping protocol resulted in larger peak strain rates in the cortex of the middiaphyseal tarsometatarsus (TMT) (+740%). Strain magnitude was increased to a lesser extent (+30%) and strain distribution was unchanged. Although bone formation on the periosteal surface increased significantly (+40%), the increase on the endocortical surface was much more dramatic (+370%) despite smaller mechanical stimuli at this surface. Under normal loading conditions, the bone formation rate on the endocortical surface is lower than on the periosteal 14 Chapter 1 - Literature Review surface. Therefore, these results suggest that the endocortical surface may have a greater potential for change in the presence of increased strain rates associated with axial compressive loading. Although strain rate may be more osteogenic than strain magnitude, it is important to consider the changes in bone structure that are associated with varying dynamic loads. In growing rats, the adaptive modeling response has been shown to be greater in those animals receiving loads of greater magnitude (4000 ue) than those experienced during normal locomotion (1000 to 2000 ue) (96). However, it should be noted that in reference to the mechanostat hypothesis 4000 UE represents the pathological overload zone for human bone, and may result in microdamage. In adults, strain magnitude at the tibial midshaft during vigorous activity (uphill and downhill zigzag running) is approximately 2000 ue, which is nearly three times higher than strains recorded during walking (97). Rule 2: Short Duration of Loading is More Osteogenic. In a now classic experiment, Rubin and Lanyon (95) used the avian ulnar loading model to demonstrate that the cellular response to mechanical loading saturates quickly. The results of this study showed that 36 cycles/day at physiological strain magnitudes (2000 UE) were just as effective for eliciting an osteogenic response as 1800 cycles/day at the same strain magnitude. Further, beyond 36 cycles, the bone response was not enhanced. The saturation of the bone response to mechanical loads has also been described in growing bone. Umemura and colleagues (14) assigned immature female rats to one of five jump-trained groups (5, 10,20,40, or 100 drop jumps) or a control group. The jump training began at a jump height of 25 cm and progressed to 40 cm by the fourth week. After 8 weeks of jump training (5 days/week), the 5-jump group showed significant gains in bending rigidity at the femur and tibia (Figure 1-7) and in tibia cortical area compared with the control group. Although there was a trend towards an increased cortical area and rigidity with an increased number of jumps, the differences between the 10-, 20- and 40-jumps/day groups were small. These results suggest that short loading bouts were just as effective in initiating a bone response as prolonged loading bouts. 15 Chapter 1 - Literature Review Number of Jumps per Day Figure 1 -7. Maximum bending moments (N • mm) at the fracture test of the femur and tibia in the rats trained with different numbers of jumps per day. Values are means + SD. * significantly (p < 0.01) different from the control group (0 jumps/day); t from 5-jump group and 10-jump group (p < 0.05); t from the 5-, 10-, 20- and 40-jump groups (p < 0.01). Adapted from Umemura et al. (14). An additional feature of bone cell saturation is that recovery periods either within a single loading bout, or between individual loading bouts, can re-establish mechanosensitivity. In adult rats, partitioning a daily loading protocol into brief sessions separated by recovery periods produced greater gains in bone mass, geometry and strength than one single loading bout (15). After 16-weeks, rats who received 4 bouts of 90 cycles/bout (90 x 4) with 3 hours of recovery between bouts showed a 70% greater BMC, 37% greater CSA and 46% greater minimum second moment of area at the tibial diaphysis than rats who received one uninterrupted bout (360 x 1). Similarly, Robling et al. (101) found that 14 seconds of rest between load cycles resulted in 66-190% higher relative bone formation rates on the loaded tibia of adult female rats. It is suggested that short-term recovery sessions may enhance the recruitment and/or activation of osteoblasts via fluid-flow mechanisms, while long-term recovery sessions may allow reorganization of the actin cytoskeleton (101). It is not known whether a similar loading protocol is effective in growing animals. Rule 3: Abnormal strains drive bone adaptation. During mechanical loading, bone adaptation is dominated not by a large number of cycles of 'normal' strain distribution but rather by fewer cycles of relatively 'abnormal' strain distribution (102). Thus, adaptation is 'error-driven' in that bone responds to unusual strain distribution by making architectural changes to eliminate or reduce the perceived deviations from normal loading patterns (85). Using the avian ulnar model, Rubin and Lanyon (91) noted a linear relationship between strain magnitude and change in ulnar 16 Chapter 1 - Literature Review cross-sectional area (CSA) such that loads less than 1000 UE were associated with a reduction in CSA while loads between 1000 and 4000 UE conferred a significant increase in CSA. Due to the artificial nature of the loading environment, the osteogenic response observed was attributed to an unusual strain distribution. The strain distribution error was sufficient to elicit a bone response even at strains of low magnitude. More specifically, regional differences in strain distribution influence bone adaptation. At the tibial shaft, strain gauge data from animal (103,104) and human cadaver (105) studies indicate that the primary mode of loading is bending in the anterior-posterior direction. The strain distribution associated with bending loads results in tensile strain on the anterior cortex and compressive strain on the posterior cortex, the sites furthest away from the neutral axis of bending (medial-lateral direction) (71,105). Generally speaking, bone is preferentially added in regions of highest strain in order to return strains at these sites to the "customary" level (81). This has been demonstrated in several experiments with both young (96,99) and adult (106,107) animals. There may also be regional differences in cortical bone properties such as mineral content and porosity that are related to loading patterns (93). For example, Skedros and colleagues (93) reported increased mineralization in the compression cortex compared with the tension cortex in the immature mule deer calcaneus. This is likely related to the influence of loading on Haversian bone remodeling (35,108). Although the function of Haversian bone remodeling is not well understood, one hypothesis is that it prevents or repairs fatigue damage (microcracks) that results from high strain magnitudes and/or frequencies (33,35,98). Based on results from these animal studies, Turner and Robling (109) generated an equation that can be used to estimate the osteogenic potential of an exercise protocol. Three parameters are required to calculate the osteogenic index (Ol): intensity (peak ground reaction force, GRF), number of jumps (N), and time betweens sessions. The equation predicts that the Ol for weekly exercise protocols is higher if the number of sessions per week is increased rather than the duration of the individual sessions. For example, 300 jumps per day (GRF = 3 x body weight), 2 times per week produces an Ol of 33. However, if the 600 jumps are performed over 5 days (120 jumps per day) the Ol is more than doubled (109). A similar increase in the Ol is observed if one daily session is divided into two sessions separated by a recovery period. Although the Ol may prove useful for identifying effective exercise programs, this equation has not been validated for use in animal or human studies. 1.2.2.3.2 Adaptive Differences between Cortical and Trabecular Bone The aforementioned studies used animal models to evaluate adaptation of the cortical diaphysis. It is clear that the adaptive response to loading is complex, and is influenced by a number of properties within the loading environment. Less is known about loading characteristics that influence trabecular bone adaptation, possibly because of difficulties associated with applying and controlling loads at trabecular sites such as the metaphysis (110,111). Previous studies in growing rats have demonstrated osteogenic effects of endurance exercise on trabecular bone formation and microarchitecture (112,113); however, variations in loading protocols were not assessed in either study. Recently, van der Meulen and colleagues (111) developed a novel device to apply controlled loads to the distal lateral femoral condyle of the rabbit. They then used a similar loading protocol to that of 17 Chapter 1 - Literature Review Robling et al. (15) to investigate the role of the number of loading cycles on trabecular bone volume fraction and architecture at the distal femur in 14 skeletally mature rabbits. Compared with loading cycles of 10 cycles/day and 25 cycles/day, the loading cycle of 50 cycles/day for 4 weeks increased bone volume fraction and trabecular thickness as measured by microCT (uCT). However, due to the small sample size, no conclusions could be made regarding the effects of load cycles on trabecular bone formation (mineral apposition rate) as measured with histomorphometry. Future investigations with this animal model will help to define other loading characteristics such as load magnitude and duration that influence trabecular bone functional adaptation. The adaptive response of cortical bone in the growing skeleton involves mainly changes in periosteal bone formation and cross-sectional geometry, whereas the adaptive response of trabecular bone involves changes in trabecular microarchitecture. Joo et al. (113) studied the effects of 10 weeks of endurance exercise (treadmill running) on trabecular aBMD (by DXA) and microarchitecture (with uCT) at the distal femoral metaphysis in skeletally immature male rats. Compared with controls, exercised rats demonstrated a significant increase in trabecular aBMD which was the result of significant increases in trabecular thickness, number and connectivity and a significant decrease in trabecular separation. Although not tested directly, the authors interpreted the structural changes as an increase in trabecular bone strength based on established relationships between bone strength and microarchitectural parameters (114,115). Similar to the studies described previously, the adaptive response to treadmill running at the femoral diaphysis in these same animals involved an increase in periosteal bone formation (113). These results highlight how the adaptive response of bone to loading differs between regions of the same bone. 1.2.2.3.3 Adaptive Differences between Immature and Mature Bone Evidence from animal studies suggests that the ability of bone to adapt to mechanical loading is much greater in the growing than in the non-growing skeleton (108,116-118). The age-related differences may be due to changes in the strain-related thresholds necessary to activate the (re)modeling response. Rubin and colleagues (116) found that following an 8-week loading program the increase in ulnar cortical area was significantly greater in young (1 year old) turkeys when compared to old (3-year old) turkeys (30% vs. -3%, respectively). Although this study suggests that the aging skeleton may not be responsive to loading, Turner et al. (117) found that the response to loading in older rats was dependent on a higher level of strain than was required in the younger animals. This may reflect a reduction in osteoblast or osteocyte function with aging (108,117). More recently, Jarvinen et al. (118) examined the adaptive mechanisms of the growing and aged rat femoral neck in response to a 14-week running program. When compared to controls, both young and adult rats showed similar gains in the breaking load (+30% and +28%, respectively); however the mechanism underpinning this gain was age-dependent. Whereas young rats demonstrated a significant increase in pQCT-derived ToA (+25%) and non significant gain in ToD (+11%), the adult rats had significant increases in ToD (+23%) and non-significant increases in ToA (+10%). Thus, it appears that the growing skeleton may preferentially adapt to mechanical loading through geometric changes while the mature skeleton responds mainly through increases in bone density (118). 18 Chapter 1 - Literature Review 1.2.3 Measurement of Bone Parameters in Children As discussed, the amount of bone within a cross-sectional area affects bone strength. In general, more bone equals a stronger bone. However, the architectural distribution of the bone mass (e.g., the diameter of the bone) also affects whole bone strength. Therefore, in order to understand the changes in bone strength that occur during growth, it is essential that both the material and architectural properties of the skeleton be assessed. This thesis will involve the use of three technologies that together, will provide a comprehensive evaluation of the growing skeleton. I discuss the strengths and limitations of each technology in this section. 1.2.3.1 Dual Energy X-Ray Absorptiometry Currently, dual energy x-ray absorptiometry (DXA) is the most commonly used modality to assess bone mineral status of the growing skeleton in both clinical practice and research (119). DXA is a relatively inexpensive, noninvasive technology that requires a short scan time and is associated with low radiation exposure (119). The outcome variable of bone mineral content (BMC) in grams represents the attenuation values of photons that pass from an X-ray tube (source) through the region of interest. The common regions of interest in pediatric studies are the total body, lumbar spine and proximal femur. For each region, the projected, 2-dimensional area of bone (bone area, cm2) analyzed is used to calculate the areal bone mineral density (aBMD, g/cm2). In our laboratory (UBC Bone Health Research Group), the in vivo precision values (%CV) with the Hologic QDR 4500W are less than 2% in adults for all regions of interest (unpublished data). Due to ethical concerns associated with repeated scans, precision studies with children are uncommon. However, Litaker et al. (120) recently reported high reproducibility (intraclass correlation, ICC = 0.997) for total body BMC with the QDR 4500W in a large cohort (n = 219) of 13 to 18 year olds. Although DXA-derived aBMD is a reasonable predictor of bone strength, and ultimately fracture risk (17,121-123), the planar nature of the measurement is a considerable limitation (17). Of particular relevance to longitudinal pediatric studies, DXA is unable to account for changes in bone size and geometry that occur during growth (119) (Figure 1-8). As a result, BMC and aBMD measures in a child with short stature (e.g., with smaller bones) will likely be underestimated while the opposite is true for a child with tall stature. To correct for the third dimension, a mathematical equation can be applied to the DXA outcomes to generate bone mineral apparent density (BMAD), or the amount of BMC per total bone volume (124). Underlying this correction is the assumption that bone cross-sectional shape is geometrically similar between subjects, and that bone thickness scales linearly with the measured projectional area (121,124). Although this assumption may hold true for skeletal sites that are considered cylindrical (femoral neck), it is likely not appropriate for more complex geometries such as those of the lumbar vertebrae (121). As will be discussed in more detail in Section 2.2.4.3 a prediction algorithm has been developed that can be applied to proximal femur scans to generate three-dimensional approximations of bone geometry and bone bending strength. 19 Chapter 1 - Literature Review aBMD 1 1 1 Bending Strength 1 4 8 Figure 1-8. Schematic of the effect of bone cross-sectional geometry on long bone strength. Changes in distribution of bone mass that influence bone bending strength (i.e. cross-sectional moment of inertia) are not reflected in conventional measures of areal bone mineral density (aBMD) by dual energy x-ray absorptiometry. Additional limitations of DXA include its inability to distinguish between cortical and trabecular bone, and inaccuracies associated with surrounding soft tissue (119). Within the DXA 2-component model, the composition and distribution of soft tissue in the region of interest is assumed to be absorptiometrically homogeneous (125). Corrections for soft tissue are based on a uniform distribution of fat around the bone (119) and thus, longitudinal DXA values in children may reflect changes in body size and composition associated with growth rather than true changes in BMC (119). More specifically, a nonuniform distribution of fat around the bone can result in inaccuracies in DXA measurements of aBMD in the range of 20-50% (126). Despite these limitations, DXA remains the most commonly used methodology to assess pediatric bone, especially in clinical settings (127). 1.2.3.2 Hip Structure Analysis Structural parameters of the proximal femur can be estimated by applying the hip structure analysis (HSA) program to DXA images. HSA is a predictive computer algorithm that incorporates mechanical engineering principles into a software-specific analysis of bone mineral data (18). The principle used in the HSA program is that a line of pixels along the bone axis in a bone mass image is a projection of the corresponding cross-section (128). The dimensions of this projection are used to estimate bone cross-sectional geometry (18,128). Bone mass profiles are generated at the femoral neck across its narrowest point ("narrow neck"; NN), the intertrochanteric (IT) region along the bisector of the shaft and femoral neck axes and the femoral shaft (FS) at a distance 1.5 times the femoral neck width (Figure 1-9) (18,129). At each site, five contiguous bone mass profiles are created and are spaced 1 pixel width apart. Thus, the total cross-section is approximately 5 mm thick. The profile integral is equivalent to bone surface area (CSA, cm2) after removal of soft tissue voids and assuming a fixed average mineralization of 1.051 g/cm3 (value for fully mineralized adult cortical bone). As such, CSA measures the amount of bone within the cross-section (like DXA-derived BMC) but the quantity is expressed in terms of cortical equivalent surface area rather than mineral mass (130). Subperiosteal width (SPW, cm) is the blur-corrected profile 20 Chapter 1 - Literature Review width and cross-sectional moment of inertia (CSMI, cm4) in the bending plane is derived as CSMI = "f adi2 where a, is the pixel thickness x pixel spacing along the bone mass profile and di is the distance of the pixel from the centre of mass (131). The section modulus (Z, cm3), an indicator of bone bending strength, is computed as Z = CSMI/dmax where dmax is the maximum distance from the centre of mass medial or lateral cortical margin (131). In modeling these regions, HSA assumes that the NN and shaft regions are circular and that the IT region is an asymmetric ellipse. In addition, each region is assumed to contain different proportions of cortical and trabecular bone: IT region, 50/50; NN region, 60/40; shaft region, 100% cortical bone (132). The HSA program was validated using 22 cadaver specimens (18). Estimated strength from HSA correlated well with measured bone breaking strength (r = 0.66-0.89). HSA-computed cross-sectional properties were also compared to those obtained from CT images. Agreement between the two methods was within 10% (18). Narrow-Neck Distance (cm) Figure 1-9. Proximal femur image from Hologic DXA scanner showing positions of analysis regions at the narrow neck, intertrochanteric and femoral shaft. To the left of the image are typical bone mass profiles used in measurement of geometric properties. Bone structural outcomes from HSA can be used to supplement standard DXA outcomes of aBMD and BMC and have recently been shown to have similar predictive power to aBMD in determining fracture risk (133). To date, HSA has been used in several pediatric studies and has proven useful in evaluating sex differences in femoral neck bone geometry and strength (130) as well as in understanding the bone structural response to physical activity (8,134,135). However, there are clearly limitations in estimating three-dimensional properties using two-dimensional imaging techniques. In particular, CSMI and Z are only relevant for bending in the image plane and cannot be used to 21 Chapter 1 - Literature Review estimate bending in other directions (18). Further, the mineralization density of pediatric bone is lower than that of adults. The assumption of adult mineralization in the HSA program therefore results in an underestimation of CSA, CSMI and Z in children (130). The precision of HSA varies according to the manufacturer or model of the DXA instrument. Khoo et al. (129) recently assessed the in vivo short-term precision of key HSA variables from each of three regions of interest within the proximal femur using paired scans from two large clinical trials involving post-menopausal women. Precision error (CV%) for Z, CSMI, CSA and SPW for the Hologic QDR 4500 ranged from 2.3 - 5.1% at the narrow neck, 1.9-4.6% at the intertrochanteric region and 0.9 - 3.2% at the femoral shaft (129). Precision error tended to be poorer for CSMI and Z than CSA and SPW. This difference may be explained by the fact that similar to conventional DXA measures of BMC and BMC, CSA and SPW are relatively insensitive to patient positioning (129). In contrast, CSMI and Z estimate bending strength in the image plane only and in asymmetric cross-sections such as the IT region scans with differing rotations about the axis perpendicular to the cross-section will produce different values. Errors in positioning would affect CSMI and Z at the narrow neck and femoral shaft to a lesser degree due to the more symmetrical cross-sections. In addition, changes in femur position (hip rotation) may also alter the location and orientation of the image planes at the NN and IT regions (Figure 1-10). Regarding the precision of HSA analyses, results from our lab have shown intraoperator precision is <1.5% CV for all structural variables within the three regions (134). Figure 1-10. Illustration of the effect of femur repositioning on the location and orientation of the image plane. (A) Axial view of proximal femur, with the femoral neck axis positioned correctly (co-planar with the DXA scanning plane represented by the dotted line). (B) Repositioning error has placed the femoral neck axis out of the scanning plane and results in a distorted cross-section. Adapted from Khoo et al. (129) with permission from Elsevier. 22 Chapter 1 - Literature Review 1.2.3.3 Peripheral Quantitative Computed Tomography Given the limitations of DXA, 3-dimensional imaging modalities such as pQCT are being employed more frequently in pediatric bone research. Peripheral QCT was developed specifically as an extension of the larger QCT systems, which are able to measure (volumetric) bone density in the axial and appendicular skeleton. Although both instruments are unique in their ability to separate cortical and trabecular bone, pQCT has several advantages over QCT including higher resolution, higher precision, lower radiation and lower cost (136-138). The two most common pQCT machines on the current market are the XCT 2000 (Stratec Medizintechnik GmbH, Pforzheim, Germany) and the Densiscan 1000 (Scanco Medical, Basserdorf, Switzerland). The UBC Bone Health Research Group owns one of the three XCT 2000 machines in Canada, and thus I focus my discussion of pQCT on this model. 1.2.3.3.1 pQCT Technology Similar to DXA, pQCT measures the attenuation of radiation as it passes from the source to the detector through the object of interest. However, unlike DXA, pQCT scans a single tomographic slice and is used to assess bone geometry and apparent density of the axial skeleton (radius, tibia) (Figure 1-11). Peripheral QCT scans are performed using a translate/rotate mechanism, which involves a series of transverse measurements following successive, partial rotation displacements (12°) of the source-detector couple (19). This process continues until a 180° excursion is complete (15 rotations). The absorption of x-rays by the object of interest at each angular position creates multiple absorption profiles. Using a technique called filtered backprojection, the absorption profiles are mathematically combined to create a cross-sectional image which represents the original object (139). Each individual unit (voxel) within the image corresponds to an attenuation coefficient. These coefficients are transformed into volumetric mineral content and density by comparing the values to a reference hydroxyapatite phantom (19). The result is a value for apparent (volumetric) bone mineral density in the bone slice. It is important to note that due to limited resolution, pQCT is unable to measure the degree of mineralization within the cortex. Thus, the measure of apparent density (or BMDcompartment) is a volume that includes porous spaces such as osteonal canals and is related to cortical porosity (140). 23 Chapter 1 - Literature Review Figure 1-11. Schematic of a long bone and its biomechanical properties measured by DXA and pQCT. Whereas DXA calculates bone mineral content (BMC) from planar x-ray attenuation data, pQCT generates a three-dimensional cross-section from which geometric and material properties of the bone are obtained. For example, the bending and torsional cross-sectional moments of inertia (CSMIX, CSMIP) are obtained as the integral sum of the products of the area of each pixel (A) and the squared distance (dx, dy, dz) to the corresponding bending (x, y) or torsion (z) axis. Adapted from Kontulainen (77) and Ferretti et al. (141). 1.2.3.3.2 Data Acquisition Unlike DXA, pQCT offers the operator a number of choices related to scan acquisition parameters. These include the resolution, scan time, reference line placement and scan site. Although these parameters influence outcomes of interest, it is uncommon for researchers to report their acquisition protocol(s) and there are currently no standardized protocols for pediatric pQCT studies. This poses significant challenges when comparing results across pQCT studies. Peripheral QCT image quality is dependent on spatial resolution or voxel size. The number of voxels per scan field is fixed for each machine; however, the operator can change the field size. A smaller field size, and thus, a smaller voxel size, results in an improved resolution (19). Unfortunately, a small voxel size (e.g. 0.2 mm) requires a longer scan time to minimize the signal to noise ratio and therefore increases effective radiation dose. Thus, a voxel size of 0.4 mm or 0.5 mm is often used in pediatric pQCT studies. Although this resolution decreases radiation exposure, it does increase the potential for the partial volume effect (PVE) to influence bone outcomes. The PVE occurs when a voxel contains two or more tissues of different densities and is particularly relevant when assessing cortical bone (Figure 1 -12). If the average density of the voxel is within the limit of the user-defined analysis threshold the partially filled voxel is included in the analysis and may contribute to an underestimation of the true apparent density. At the periosteal surface where the boundary between bone and soft tissue is well-defined, a partially filled voxel is not likely to be included in the analysis. However, at the endosteal surface the boundary between cortical bone, "subcortical bone", trabecular bone and marrow is less clear due to smaller differences in tissue densities. As a result, there is a greater chance that partially filled voxels would be included in the analysis. The PVE can be 24 Chapter 1 - Literature Review minimized by choosing appropriate analysis modes and thresholds. Further, correction factors have been proposed; however, these may only be necessary when absolute rather than relative values are required (142). In children, the PVE is also a concern when assessing skeletal sites with a thin cortical shell such as the distal radius or tibia (143,144). If the cortex is not sufficiently thick (<2-2.5 mm) to allow at least one voxel to lie fully within the periosteal and endosteal border the volume averaging errors increase (144). Thus, cortical bone should be evaluated with pQCT at shaft sites only. An additional concern when assessing pediatric bone with pQCT is placement of the reference line. In order to determine the exact measurement site, a reference line must be defined at an anatomical landmark (139). This landmark is located using a scout scan that is performed prior to the real pQCT scan. Placement of the reference line is not standardized for pediatric pQCT studies and in many cases, researchers do not report the landmark used (11). The manufacturer recommends the reference line to be placed in the middle of the epiphyseal plate (139). However, at the distal radius, others place the reference line through the most distal portion of the growth plate in children with open growth plates or through the middle of the ulnar border of the articular cartilage in children whose growth plate is no longer visible (20). In my opinion, neither of these approaches is appropriate for prospective pediatric studies since reproducibility of the landmark is limited due to reliance on visual estimation and inevitable changes in the growth plate. The need for reproducible landmarks as well as standardization of other pQCT protocols was recently discussed by Ashe and colleagues (145) with reference to analysis of low density bone. A similar report has yet to be provided for pediatric pQCT studies. 25 Chapter 1 - Literature Review i voxel pQCT result for CoD Mean: 4x710 mg/cm3 + 2 x 469 mg/cm3 630 mg/cm3 8x710 mg/cm3 + 2 x 469 mg/cm3 662 mg/cm3 Figure 1-12. Schematic of the partial volume effect (PVE) as it pertains to the measurement of cortical bone mineral density (CoD) by pQCT. (A) At sites where the cortical wall is not sufficiently thick (< 2mm), partially filled voxels (assumed in this example to be filled by two-thirds) at the edge of the cortex will reduce the measured CoD. (B) At sites where the cortical wall is sufficiently thick, partially filled voxels have less of an influence on the measured CoD. Adapted from Schoenau et al. (23). 1.2.3.3.3 Data Analysis The Stratec software that accompanies the XCT 2000 (Version 5.5 is used in this thesis) provides numerous analysis modes and thresholds from which the operator must determine the most appropriate combination for the particular region of interest (ROI). Similar to pQCT data acquisition, standardized pQCT analysis protocols for pediatric bone do not exist. As a result, default protocols provided by the manufacturer (139) are commonly used. Total and trabecular bone areas and densities are obtained in two steps with iterative detection algorithms in the CALCBD function. In the first step soft tissue is separated from the outer bone edge to derive total bone outcomes (Contour Mode). In the second step trabecular bone outcomes are determined by peeling away the cortical and subcortical bone (bone that falls between the cortical and trabecular bone thresholds) (Peel Mode). Cortical area and density are obtained in the CORTBD function by removing all voxels in the ROI that have an attenuation 26 Chapter 1 - Literature Review coefficient below the defined threshold (Separation Mode). The CALCBD and CORTBD functions produce a large number of bone outcomes; however, only a few are commonly reported in the current literature. These select variables are described in Table 1-1. Bone areas and densities can be obtained accurately with the XCT 2000 (137,146) although the precision is generally poorer at trabecular sites such as the distal radius and tibia compared with cortical bone sites such as the radial and tibial shafts (146). The various modes and thresholds recommended by the manufacturer have not be validated for the assessment of cortical and trabecular bone properties in children. The in vivo precision of pQCT (Norland XCT 2000) at the tibia in children between 6 and 14 years is less than 2% (147). The major sources of imprecision associated with pQCT scanning are gross movements during scanning, subject positioning and limb alignment (148) and placement of the reference line. Imprecision can be minimized by having one trained operator perform the scans according to standardized procedures. 1.2.3.3.4 Measurement of Bone Strength with pQCT A significant advantage of pQCT compared with DXA is that pQCT provides estimates of bone strength. As discussed, the distribution of cortical bone at diaphyseal sites such as the tibial midshaft influences both bone torsional and bending strength (16,62,71). With pQCT, the distribution of each pixel from the reference axis (x, y or z) is used to estimate CSMI (Figure 1-11). The polar moment of inertia is the sum of the principal moments of inertia (Lax and Lin or lx and ly) and is important in determining stress in torsion (74). However, pQCT-derived CSMI is a also a reliable indicator of bone bending strength in animal studies (149,150). Section modulus, an additional indicator of long bone bending strength (75), is derived by dividing CSMI by the maximum distance from the bending axis to the outer surface (d max)-Bone bending strength is also dependent on the material stiffness of cortical bone (i.e., elastic modulus) (62). Material stiffness can only be determined with mechanical tests; however, it can be estimated with pQCT-derived bone mineral apparent density (151). The strength strain index (SSI or density-weighted section modulus) provided by the XCT-2000 combines architectural properties of the cross-section (section modulus) with apparent cortical density (normalized to the physiological density of 1200 mg/cm3). This bone strength index has been validated against measurements of whole bone strength in animal studies (152). At the metaphyses of long bones, compressive strength is dependent on the material and architectural properties of trabecular bone. Since the limited resolution of pQCT prevents assessment of the architectural properties of trabeculae, estimation of compressive strength must rely on measures of trabecular mass (trabecular BMC) or apparent density (141). Further, combinations of apparent density with an indicator of bone size such as cross-sectional area may also provide a reasonable assessment of compressive strength. Kontulainen and colleagues (153) calculated a compressive bone strength index (BSI) for the distal radius as the product of the square of pQCT-derived total density and total cross-sectional area (the load bearing area). At distal sites, total density is more a reflection of trabecular density than of cortical density and the square of the density is used based on relationships established in compression tests of trabecular bone (70,71,154). 27 Chapter 1 - Literature Review 1.2.3.3.5 Strengths and Limitations ofpQCT As discussed, pQCT offers several advantages over conventional DXA measures of BMC in the assessment of pediatric bone. In addition to bone areas, densities and estimated strength, pQCT can also estimate muscle cross-sectional area (MCSA) using similar user-defined modes and thresholds. I discuss this feature of pQCT in Section 1.2.5.4.1.2. With the exception of a lack of standardized acquisition and analysis protocols there are few limitations of pQCT. One concern for longitudinal studies is the long-term precision of pQCT measurements (155). As discussed, the contribution of the proximal and distal growth plates to overall long bone growth varies with age (47,48). Thus, it is not possible to reproduce the same exact location over time. However, the same relative location along the bone length can be determined using a fixed anatomical reference line. Perhaps the most obvious limitation of pQCT is that this technology is unable to measure the clinically relevant proximal femur. Initially, it was anticipated that pQCT measurements of the upper and lower limbs might provide a means to estimate hip fracture risk. However, experimental pQCT data indicates that geometric and densitometric properties of the upper and lower limbs explain only 30-45% of the variability in femoral strength (156). This is less than the 50-60% explained by site-specific measurements of the femoral neck and proximal femoral shaft (156). Further, correlations between parameters of bone geometry and (volumetric) density display only moderate correlations across skeletal sites (156). Therefore, a comprehensive evaluation of both peripheral and axial regions of the growing skeleton can only be achieved with the use of multiple imaging techniques. 28 Table 1-1. Common bone outcomes in pediatric pQCT studies. The site of analysis, analysis mode and a brief description of each outcome are provided. Outcome (units) Bone geometry Site of Analysis Analysis mode Bone mass Bone mineral content (BMC, mg/mm) Distal, shaft Contour mode Total bone cross-sectional area (ToA, mm2) Distal, shaft Trabecular bone cross-sectional Distal area (TrbA, mm2) Cortical bone cross-sectional area (CoA, mm2) Average cortical thickness (CTh, mm) Volumetric Total density Density (ToD, mg/cm3) Trabecular density (TrbD, mg/cm3) Cortical density (CoD, mg/cm3) Shaft Shaft Distal Distal Shaft Contour mode Contour mode, Peel mode Separation mode Separation mode Contour mode Peel mode Separation mode The amount of bone mineral within a cross-sectional slice of 1-mm thickness. Can be multiplied by the slice thickness to obtain total BMC in a 2.5 mm slice. The surface area of the entire bone cross-section including the cortex and marrow cavity. ToA directly reflects changes in bone size resulting from periosteal apposition (157). The surface area of the trabecular bone cross-section. This area is influenced by endocortical apposition or resorption. The surface area of cortical bone within the cross-section. CoA is influence by periosteal apposition and endocortical apposition and resorption. The distance between the outer and inner border of the cortical shell. CTh is determined with the circular ring model which assumes a circular cross-section. ToD is the volumetric density averaged over the entire cross-section. It is influenced by the relative contributions and densities within both the cortical and trabecular bone compartments. TrbD is the volumetric density averaged over the trabecular area of the cross-section. TrbD is influenced by trabecular number and thickness and the degree of mineralization at the material level. CoD is the volumetric density averaged over the cortical area of the cross-section. CoD is influenced by cortical porosity and the degree of mineralization at the material level. ro Table 1-1 continued Outcome (units, «&, Description Strength Indices Cross-sectional moment of inertia (CSMI, mm4) Shaft Separation mode CSMI = I (A x di2) CSMI is proportional to the distribution of bone mass about the neutral axis and is an indicator of bone strength in bending or torsion (depending on which axis is used as a reference). It is calculated as the integral sum of the products of area (A) of each voxel and the squared distance (d2) of the corresponding voxel to the bending (x, y) or torsion (z) axes (Figure 1-11) (19). Section modulus (Z, mm3) Shaft Separation mode Z = CSMI/dmax where dmax is the maximum distance from the bending axis to the outer surface, in the plane of bending. Thus, Z approximates a cross-section's resistance to bending in a given plane (75) Strength-Strain Index (SSI, mm3) Shaft Separation mode SSI = 7 [(axdi2)(CoD/ND) ' dmax SSI is calculated as the integrated product of Z and CoD. The ratio of CoD and normal physiological density (ND = 1200 mg/cm3) provides an estimate of the modulus of elasticity (158). Similar to CSMI, SSI can be determined with respect to the polar (z) axis or the bending (x, y) axes. Compressive Bone strength index (BSI, mg2/mm4) Distal Contour mode BSI = ToA* ToD2 At distal sites, compressive strength is estimated as the square of the total density and the total cross-sectional area (the load-bearing area) (70,71,153,154) Chapter 1 - Literature Review 1.2.4 The Growing Skeleton: Maturity- and Sex-Related Differences in Bone Geometry, Density and Strength In this section I first provide a description of the methods commonly used to assess physical maturity in children and provide a brief overview of the hormones that influence skeletal development. I then go on to discuss the maturity- and sex-related differences in skeletal development focusing on pediatric pQCT studies. 1.2.4.1 Assessing Maturity While growth is defined as the increase in the size of the body as a whole or in parts, maturation refers to the tempo and timing of the physical changes associated with growth (159). Maturity, or biological age, does not follow the same temporal pattern as chronological age thereby conferring a large amount of variation in maturation among children of the same chronological age. Measures of maturity vary according to the biological system being assessed. Historically, skeletal maturation is the most common method. The development of the skeleton spans the entire period of growth, and is fairly uniform with each bone progressing through standard changes from childhood to early adulthood (159). The progression of bone from an immature to a mature state can be characterized through specific changes observed on a hand-wrist X-ray that occur in a definite, irreversible order. Skeletal age is determined based on ratings provided by one of three methods: Greulich-Pyle, Tanner-Whitehouse and Fels (159). These methods are often used in clinical settings, but are of limited practicality in research settings due to ionizing radiation associated with radiography. Given the limitations associated with measuring skeletal age, assessment of sexual maturation is more common in pediatric research. Originally described by Tanner (160), this method relies on criteria associated with the development of secondary sex characteristics, specifically breast and pubic hair development in girls and pubic hair and genitalia development in boys. The criteria are incorporated into a five-stage scale, which is then used to rate a child's sexual maturation. Tanner stage 1 represents prepuberty, Tanner stages 2 and 3 represent early puberty, Tanner stage 4 indicates the late stages of maturity and Tanner stage 5 represents reproductive maturity. These stages correlate well with testosterone and estrogen levels (r=0.5-0.8) in girls and boys (161). However, similar to skeletal development, the development of secondary sex characteristics is continuous. Therefore, a five-stage scale does not allow one to account for the large degree of variation that may be present between two children rated at the same Tanner stage (159). There are also known differences in the timing of sexual maturation between boys and girls that confound comparisons between sexes at the same Tanner stage (162,163). Ratings in clinical settings are usually made by direct observation; however, due to the invasive nature of observation self-assessment methods are more common in research settings. Adolescents' self-assessment of maturity using line drawings of the 5 Tanner stages demonstrates favourable agreement with physician assessment (164,165), although there may be a tendency for children, mainly boys, to overestimate their development at early stages of maturation (165). In addition, the use of Tanner staging may be limited in overweight children. A recent study of 244 children between the ages of 6 and 12 years, 41% of whom were obese (BMI > 95th percentile), found 31 Chapter 1 - Literature Review that Tanner breast stage was overestimated in 38% of obese girls (166). This discrepancy may be due to the difficulties in distinguishing breast tissue from adipose tissue. In order to assess the timing of growth, somatic methods such as age of peak height velocity (PHV) are commonly used in longitudinal studies (159). Age of PHV requires serial measurements of height (or sitting height, leg length etc.) from which individual growth trajectories can be identified. This approach can account for the wide variation amongst children's growth parameters at any chronological age, and in the rate of change of these parameters. In boys and girls, PHV occurs at a maturational point equivalent to 92% of adult stature (167) and the characteristics of the growth spurt in boys and girls are largely under genetic control (168). On average, PHV occurs earlier in girls (11.8 years, Tanner stage 3) than in boys (13.4 years, Tanner stage 4). In both sexes PHV occurs approximately 9 months before the peak in bone mineral accrual velocity (5,160) and between 5-7 months before the peak in estimated femoral neck strength velocity (Figure 1-13) (130). Mirwald and colleagues (169) recently developed a prediction equation to estimate age of PHV from one-time measurements of height, sitting height and leg length. Cross-validation with longitudinal growth data indicated that age at PHV could be predicted within 1 year of the actual value. Although this equation offers promise for estimating biological age in cross-sectional studies it has yet to be validated in ethnically diverse samples. In addition, this technique is only able to classify children as either pre- or post-PHV and thus, cannot be used to assess maturity prior to takeoff (initiation of the growth spurt). As such, maturity offset would be similar to the pre- and postmenarcheal categorization that is used for girls (169). 32 Chapter 1 - Literature Review Age PHV 13.44 yrs Age PHV 11.77 yrs Boys Age of Peak 14.05 yrs Peak Value 409 g Size Adjusted 394 g Girls Age of Peak 12.54 yrs Peak Value 325 g Size Adjusted 342 g 9 10 11 12 13 14 15 16 17 18 19 Age in Years Figure 1-13. Graph illustrating total body bone mineral content (TB BMC) accrual velocity and ages at peak bone mineral content velocity and peak height velocity for boys (blue) and girls (red) according to chronological age. Adapted from Bailey et al. (5). 1.2.4.2 Hormonal Regulation of Bone Development During puberty, dramatic alterations in linear growth and body composition occur as a result of the interactions of gonadal steroid hormones and the growth hormone (GH)/insulin-like growth factor (IGF-1) axis (170). In this section I discuss these interactions, with a specific focus on how they influence bone size and structure. GH and IGF-1. Pubertal growth relies heavily on the activity of the growth hormone (GH)/insulin-like growth factor 1 (IGF-1) axis. GH is secreted by the pituitary gland, and is necessary for the proliferation of cartilage cells at the epiphyseal plate (170). During prepuberty, GH works in concert with thyroid hormones to promote cartilage development and bone formation. GH also promotes the production and secretion of IGF-1 from the liver, as well as other sources (171). Similar to GH, IGF-1 plays an important role in cartilage and bone development as well as muscle tissue growth (171). The GH/IGF-1 axis undergoes a dramatic activation at the time of puberty due to the rising levels of sex hormones, with estrogen likely the dominant mediator (170). Increased physical activity may also influence activation of the GH/IGF-I axis in girls (172) and boys (173) during puberty. In girls, GH levels increase 33 Chapter 1 - Literature Review around Tanner stage 2 and peak around the Tanner stage 3-4 transition (174) while in boys, the increase in GH occurs slightly later and peaks at Tanner stage 4 (Figure 1-14) (174). This temporal difference in GH secretion between girls and boys follows the pattern of change in height velocity (170). The rise in IGF-1 also occurs around Tanner stage 2 (in girls and boys) and peaks around Tanner stage 4-5 (175). Several DXA studies have documented decreased aBMD in children with GH deficiency (176,177). In contrast, a recent pQCT study of GH deficient children (mean age 7.5yrs) showed that proximal radius CoD was normal compared to healthy, age-matched children while values for CoA and CTh were significantly lower (178). After one year of GH therapy, CoD decreased, which was proposed to be a result of increased bone turnover and catch-up growth. Similar to GH, the effects of IGF-1 are likely limited to the cross-sectional properties of bone rather than the material properties. Mora and colleagues (179) examined the relationship between IGF-1 and measures of bone ToA, CoA and CoD with QCT at the femoral shaft in 197 healthy children (94 girls, ages 7 to 18 years). Serum levels of IGF-1 were significantly correlated with ToA and CoA, but not with CoD. Thus, the GH/IGF-1 axis likely influences periosteal bone formation. This has been confirmed in studies of growth hormone deficient male and female rats (180). Figure 1-14. Peaks for height velocity and bone mineral content (BMC), amplitudes for growth hormone (GH) and insulin-like growth factor-l and trends for estrogen and testosterone levels in girls relative to chronological age and Tanner stage. Peaks (connected to age by solid lines) for height and BMC velocities and GH and IGF-I indicate the average age at which these peaks occur in girls as well as the corresponding approximate Tanner stage. In boys, peak height velocity and peak BMC velocity occur about 1.5 years later than girls (at 13.4 years (Tanner stage 3) and 14.0 years (Tanner stage 4), respectively). The relationship between peaks for height and BMC velocity and peaks forGH and IGF-I are similar for boys. Adapted from MacKelvie et al. (181) with permission from BMJ Publishing Group. 34 Chapter 1 - Literature Review Sex steroids. Sex steroids play an essential role in skeletal development and in the maintenance of bone health throughout life (182). During puberty, estrogen and testosterone influence the secretion of GH and IGF-1, and are therefore key contributors to linear growth (171). In addition, estrogen plays a key role in epiphyseal fusion in girls and boys (49,183), and may inhibit periosteal bone formation and promote endosteal apposition in girls (184-186). In contrast, animal studies have demonstrated positive effects of androgens on periosteal apposition (184). The differential effects of estrogen and testosterone on the periosteal surface are thought to contribute to sexual dimorphism in bone size that begins to emerge during puberty (187). Estrogen is also thought to influence bone adaptation to loading during growth in girls by lowering the theoretical mechanostat setpoint on the endosteal surface (88,89). As a result, exercise may have a more beneficial effect on bone formation in the presence rather than the absence of estrogen. In addition, signalling through the estrogen receptors, which are found in osteoblasts, osteoclasts and osteocytes, may also influence the bone surface-specific effects of exercise (188,189). In particular, absence of the a-form of the estrogen receptor (ER- a), which is expressed in osteoblasts and osteoclasts, limits bone's adaptive response in adult female mice (190,191). Polymorphisms in the ER-a gene may also modulate the effect of weight-bearing physical activity on aBMD at the lumbar spine and femur and ToD at the tibia in pubertal girls as was recently reported by Suuriniemi et al. (192); however, the mechanisms underlying this association remain unclear. Signalling through the second estrogen receptor, ERB, is thought to act as an "antimechanostat" by suppressing osteoblastic activity in the presence of increased loading (189). Genotyping studies in humans are needed to confirm these relationships. Results from a 4-year prospective study of 27 girls ages 8 to 18 years documented the correlations between changes in sex hormones and pubertal development (193). Estrogen levels begin to rise with the onset of breast development (Tanner stage 2), continue to increase until menarche is reached and remain relatively constant after menarche (Figure 1-14) (193). Testosterone levels show a similar progression in girls; however, the magnitude of the increase at Tanner 2 is less than for estrogen. The increase in estrogen levels during puberty is associated with the timing of peak bone mineral accrual velocity (Tanner 3) (5). Further, at the time of PHV (~11.8 yrs), estrogen levels in girls represent 72% of the concentration achieved at 15-16 years. In boys, estrogen levels increase with advancing Tanner stage, although they remain considerably lower than in girls (194). In prepuberty, estrogen levels are already 8-fold higher in girls than in boys (195). The largest absolute increase in boys' estrogen levels occurs between Tanner stage 4 and 5 (Figure 1-14) (195). This rise correlates with the timing of PHV, and is influenced by the testosterone concentration which peaks one year before PHV. As growth velocity decreases in boys and epiphyseal fusion begins, estrogen levels remain elevated. Together with the pattern of change in estrogen levels for girls described previously, these data support a biphasic effect of estrogen on linear growth: at low levels estrogen stimulates onset of the pubertal growth spurt and at high levels estrogen influences growth plate senescence and epiphyseal fusion (183,188,193,196). The overall greater growth of boys is due to two additional years of prepubertal growth compared with girls and a greater magnitude of the pubertal growth spurt which is likely influenced by the pubertal increase in testosterone (188,197). 35 Chapter 1 - Literature Review Few studies have investigated the effects of estrogen and testosterone on bone geometry or (volumetric) density in the growing skeleton (198,199). Most recently, Wang et al. (199) investigated the influence of estrogen and testosterone on cortical bone properties at the tibial shaft (60% site) in 258 girls aged 10 to 13 years over a 2-year period. To control for biological age, estrogen and testosterone concentrations were introduced into hierarchical linear models that controlled for time relative-to-menarche. Estrogen, which increased progressively to menarche and gradually decreased thereafter, was a positive predictor of ToD and CTh and a negative predictor of endosteal circumference. In contrast, testosterone was a negative predictor of ToD and a positive predictor of ToA and periosteal and endosteal circumferences. When combined with results of their previous study which demonstrated a gradual increase in marrow cavity area up to menarche and a decrease thereafter (26), these results suggest that estrogen-mediated endosteal apposition (or decreased resorption) is limited to the time after menarche. These results must be interpreted with caution due to the use of single measurements of serum hormone levels. In addition, the circular ring model was used to estimate periosteal and endosteal circumferences and CTh at the tibial shaft assuming the bone to be a perfect cylinder. This method may not be appropriate given the triangular shape of the tibial shaft. 12.4.3 Maturity- and Sex-Related Differences in Skeletal Development To date, the growing skeleton has been characterized using mainly DXA-based bone outcomes. The cross-sectional (200,201), short-term (202) and longer-term prospective (203) studies conducted found that total body bone mass is similar for boys and girls before entering puberty (9 to 11 years of age). As they approach puberty, boys achieve greater bone mass at all measurable sites (204). An analysis of data from the University of Saskatchewan Pediatric Bone Mineral Accrual Study (PBMAS) that controlled for biological age, body size and body composition within a multilevel statistical model (204) showed that boys had statistically significantly higher total body (TB) and femoral neck (FN) BMC at all maturity levels (Figure 1-15). Less is known about changes in bone geometry and strength that occur during growth in boys and girls. An understanding of sex-specific developmental differences in these parameters may provide insight into the higher incidence of fragility fractures among women than men (187,205). 36 Chapter 1 - Literature Review 325CH oJ o.o-J i 1 1 1 1 1 1 1 1 i 1 1 1 1 1 1 1 1 •4 -3 -2 -1 0 1 2 3 4 -4 -3 -2 .1 o 1 i 3 4 Biological Age (PHV=0), yrs Biological Age (PHV=0), yrs Figure 1-15. Total body (A) and femoral neck (B) BMC accrual for boys (solid squares) and girls (open squares) by biological age (years from age at peak height velocity, PHV). Values are means. * p < 0.05 between biological age groups. Adapted from Baxter-Jones et al. (204) with permission from Taylor and Francis. 1.2.4.3.1 Bone Geometry A number of cross-sectional pQCT studies have described maturity- and/or sex-related developmental patterns associated with long bone growth in the upper and lower extremities (20,21,26,147). At the distal and proximal radius Neu et al. (20,21) reported an age-related difference in total bone area during growth in both sexes, although the magnitude of the difference was larger in boys. The sex difference in bone cross-sectional geometry (size) is thought to become evident during puberty when periosteal diameter expands to a greater extent in boys compared with girls (24,25,187,206). In turn, the larger bone size in boys is thought to confer greater bone bending strength in boys. However, there is disagreement in the literature as to when this sex difference emerges. Specifically, it is not clear if measures of bone geometry are greater in boys than girls as early as prepuberty. At the femoral diaphysis total and cortical bone areas (by QCT or MRI) are similar between prepubertal boys and girls (207,208), whereas radiographic data from the second metacarpal (186) and pQCT data from the radius (20,21) suggest that total and cortical bone areas are greater in prepubertal boys than girls. Of these investigations, the QCT and MRI studies of the femur (207,208) and the pQCT studies of the radius (20,21) provide the most accurate measurements of bone cross-sectional geometry. Although the discrepancy in the findings of these studies suggests possible differences between weight-bearing and non-weight-bearing bones, it is important to acknowledge that the 37 Chapter 1 - Literature Review data for the proximal (21) and distal (20) radius were not expressed relative to body size (i.e., height, limb length, weight or muscle cross-sectional area) as they were for the mid-femur (207,208). Thus, it is possible that the sex difference in bone areas at the radius may not be apparent after accounting for the confounding effects of body size. Further studies of the upper and lower limbs are needed to confirm whether sex differences in bone cross-sectional geometry are evident in prepuberty. In particular, longitudinal studies that use an appropriate indicator of biological age, rather than maturity categories, are needed to determine when sexual dimorphism in bone geometry emerges. At diaphyseal sites, changes in cortical bone area during growth are a function of bone formation on the endosteal surface, which also determines the area of the marrow cavity (or medullary area). In early radiographic studies of the metacarpals, Garn and colleagues (186,209,210) noted a sex difference in the magnitude and duration of endosteal bone formation that resulted in a narrowing of the marrow cavity in girls. They proposed that the observed endosteal apposition in girls resulted from the pubertal estrogen surge and served to create a calcium store for reproduction (186,209,210). However, the evidence regarding this surface-specific pattern of circumferential bone growth remains controversial. Cross-sectional comparisons of cortical bone structure by QCT (207), MRI (208) and pQCT (21) indicate that area of the marrow cavity increases with advancing maturation and age in both sexes. In contrast, a recent longitudinal HSA study of the femoral neck found that once biological age, height and lean mass were controlled for in a multilevel model, girls had greater CSA than boys after PHV (130). Since boys had significantly greater subperiosteal width, the authors proposed that greater CSA in girls must reflect greater bone gain on the endocortical surface in accordance with Garn's theory. Discrepancies in the aforementioned studies may reflect true site-specific differences, differences in imaging techniques (2-dimensional vs. 3-dimensional) or differences in estimation algorithms. To date, only one pQCT study has attempted to substantiate Garn's theory using prospective data. Kontulainen et al. (25) compared 20-month changes in ToA, CoA, marrow cavity area (CavA) and cortical proportion (CoA/ToA) at the tibial midshaft in 128 girls and boys across early-, peri- and postpuberty. According to Garn's theory, endosteal apposition in peri- and postpubertal girls (those who reached menarche over the 20-months or were postmenarcheal at baseline) would be reflected by a decrease in CavA and an increase in CoA/ToA (an indicator of cortical thickness). However, the pQCT findings did not support this hypothesis. Similar to boys, both peri- and postpubertal girls experienced periosteal apposition and endosteal resorption as evidenced by increases in both ToA and CavA (Figure 1-16). Although this study was limited by cross-sectional comparisons across pubertal groups that were defined differently for boys and girls, these findings do suggest that the commonly accepted paradigm that endosteal apposition is a hallmark of bone development in postmenarcheal girls be re-evaluated. Prospective studies with a more appropriate indicator of biological age are needed to confirm these results. 38 Chapter 1 - Literature Review A) Girls Baseline Followup EARLY PERI POST Figure 1-16. Schematic diagram depicting 20-month change in cortical bone in early-, peri- and postpubertal girls and boys. Average increases in total bone area and marrow cavity area are drawn to scale. Adapted from Kontulainen et al. (25). 1.2.4.3.2 Bone Density When discussing the developmental changes in bone density (by pQCT) it is important to recognize that BMD (mass of mineral per unit volume) can be analyzed at three levels: bone material (BMDmateriai), trabecular and cortical compartments (BMDcompartment) and the entire bone (BMDtotai) (140). Unlike DXA, pQCT is able to assess both BMDcompartment and BMDtotai and can therefore provide insight into underlying biological processes. Based on current evidence, some discrepancy exists regarding sexual dimorphism in cortical density. In a cross-sectional study of the proximal radius, Schoenau et al. (23) observed similar values for pQCT-derived CoD for prepubertal boys and girls, but after Tanner stage 3 and through to adulthood, CoD was 3-4% higher among women than men (23). Cortical density is determined by two physical properties, cortical porosity and mean material density, which themselves are influenced by the rate of intracortical remodeling (140). It is suggested that higher CoD at the proximal radius in females is due to a lower rate of intracortical remodeling that is perhaps controlled by estrogen, but this has yet to be confirmed with histomorphometric analysis. Conversely, there was no sex difference in CoD at the femoral shaft as measured by QCT in a longitudinal study of girls and boys 8-18 years of age (211). The discrepancy in these findings may be related to differences in imaging resolution or detection algorithms or may reflect site-specific variation between the upper and lower limbs. 39 Chapter 1 - Literature Review To date, two prospective pQCT studies have described growth-related changes in CoD at the tibia (26,212) and only the study from our group compared the change in CoD between boys and girls (212). Kontulainen et al. (212) used pQCT to assess maturity- and sex-specific differences in 20-month change of CoD at the tibial midshaft (50% site) in 127 early-, peri- and postpubertal boys and girls. Similar to the findings of Wang et al. (26) who reported a gradual increase in CoD at the tibial shaft (60% site) over 2 years in pubertal girls, CoD increased (2-4%) significantly across maturity groups of girls. The greatest increase was observed for peripubertal girls (those who reached menarche during the followup). In contrast, CoD increased in peripubertal boys only but the change was not significantly different from that of the other maturity groups. When CoD change was compared between sexes (within each maturity group), girls in each maturity group demonstrated a significantly greater increase (2-3%) in CoD than boys (212). These findings do not agree with the QCT results of the mid-femur from Loro et al. (213) discussed previously. However, it is important to note that the Loro et al. (213) study was limited by direct comparisons of boys and girls across Tanner stages 2-5. As highlighted previously for studies of bone geometry, future investigation of sexual dimorphism in cortical bone development would benefit from aligning boys and girls on biological age. In a secondary analysis, Kontulainen et al. (212) investigated the change in CoD distribution using the radial distribution function in the Bonalyse software (Bonalyse 2.1, BonAlyse Oy, Jyvaskyla). For peripubertal girls, CoD change was greatest in the subcortical region, whereas for postpubertal girls (post-menarcheal at baseline) CoD change was greatest in the high density mid-cortical region. Together with earlier findings from this cohort that showed expansion of the marrow cavity (rather than endosteal apposition) (25), these results suggest that pubertal girls appear to consolidate bone in the sub- and mid-cortical regions more than pubertal boys. Although pQCT has insufficient resolution to image cortical porosity or material density, regional differences in cortical density distribution are shown to be related to tissue porosity (214). In addition to the possible influence of estrogen on intracortical remodeling discussed previously, there is also the possibility that higher mechanical demands (i.e. larger body size and greater muscle force) in boys than girls may have caused more microdamage in boys' cortical bone, which in turn may have increased intracortical remodeling and in turn, cortical porosity (35). This has yet to be confirmed with histomorphometric analysis. Measurements of trabecular density (TrbD) by pQCT represent the trabecular number, trabecular thickness and mean material density of the trabeculae (20). Values for TrbD are generally lower than those for CoD due to the differences in porosity between the two types of bone. Cross-sectional data from the distal (4% site) radius suggest that TrbD does not increase with age in girls, and may increase slightly in boys after Tanner stage 3 (20). As a result of these age-related differences, boys have 13% and 23% greater TrbD compared with girls at Tanner stage 5 and in adulthood, respectively. In contrast, sex differences in TrbD (by QCT) were not observed at the lumbar spine, although TrbD increased by approximately 18% between Tanner stages 2 and 5 in both sexes (213). The differences in TrbD at the axial and appendicular skeleton may reflect the more transient nature of the trabeculae within the metaphyseal region of the distal radius (20). To my knowledge, there have been no prospective pQCT studies of trabecular bone properties in children. 40 Chapter 1 - Literature Review Recently, the development and validation of high-resolution three-dimensional pQCT systems have allowed trabecular microstructure to be evaluated non-invasively at the distal radius (215-217). In young adults (aged 20-29 yrs), trabecular bone volume/tissue volume and trabecular thickness at the distal radius were greater in men than women, whereas there was no sex difference in trabecular number or separation (217). This technology has not been used to evaluate trabecular microstructure in the growing skeleton. 1.2.4.3.3 Bone Strength The relative rate of bone formation on the periosteal surface and bone formation and resorption on the endosteal surface together determine bone size, the amount of bone material within the bone envelope, and ultimately, whole bone strength (186,218). The magnitude of the growth-related increase in bone strength is substantial; when comparing section modulus (by pQCT) at the proximal radius between children (6 years of age) and adults (40 years of age) there is a difference of about 300-400% (22). Although both sexes experience age-related gains in radial bone strength, the larger bone size in boys confers a strength advantage that can be observed after Tanner stage 2 (22). Similarly at the tibial midshaft, we found that 20-month changes in section modulus were 14-16% greater in early-, peri- and post-pubertal boys compared with girls of similar maturity status and that the sex difference in bone strength change mirrored the sex difference in CoA change (24). Although larger body size likely explains part of this sex difference in bone strength, it is also thought that due to smaller musculature in women than men, the lower bone strength in women may result from adaptation to smaller forces and bending moments than in men (219). However, few studies have investigated sexual dimorphism in bone strength relative to sex differences in muscle mass or muscle forces (24,130,219). I discuss these studies in more detail in Section 1.2.5.4.4. At the distal radius, the most common site for fracture in children (220,221), it appears that although bone size increases with age in both boys and girls, development of bone strength (as estimated with pQCT-derived strength-strain index) lags behind the increases in the product of forearm length and body weight, an indicator of a fall-related mechanical challenge (55). Based on estimated rates of endocortical apposition and periosteal resorption at the metaphyseal cortex, Rauch et al. (22) hypothesized that the lag in bone strength is due to an insufficient increase in metaphyseal cortical thickness relative to the increase in mechanical challenges during growth. 12.5 Determinants of Bone Strength In this section I discuss several key factors that are known to influence bone strength during growth. These factors fall into two-categories: non-modifiable and modifiable. The non-modifiable factors include heredity (genetics) and ethnicity (or race) and the modifiable factors include dietary calcium and muscle forces. Physical activity is also a key determinant of bone mass and strength and is the focus of this thesis. I discuss the relationship between physical activity and bone strength in detail in Section 1.2.6. 1.2.5.1 Heredity Numerous family (222-224) and twin (225-227) studies have shown that genetic factors may account for 60-70% of the interindividual variation in aBMD (by DXA). Further, familial resemblance for aBMD is present before 41 Chapter 1 - Literature Review puberty (222). The heritability of aBMD is complex and does not demonstrate classic Mendelian recessive or dominant inheritance patterns that can be attributed to a single gene. Instead, aBMD is a polygenic trait explained by a collection of candidate genes including the vitamin D receptor (VDR) gene (228), the estrogen receptor gene (229), the parathyroid hormone receptor gene (230) and the COL1A1 gene (encodes the alpha I chain of type I collagen) (231) among others. Twin studies have also highlighted the presence of shared genetic determinants of aBMD, BMC and other aspects of body size such as lean mass (226,232,233). Heritability of lean mass is thought to be between 50 and 80% (232,234), and when lean mass is controlled for the heritability estimate for bone mass is reduced by 5-20% (226,232,235). Thus, some of the genetic variability in bone mass reflects genetically-determined variability in body size (i.e. height, muscle mass) (235). The remaining 30-40% of the variation in bone mass is likely explained by environmental factors such as physical activity and nutrition as well as measurement error (235,236). Unfortunately heritability estimates of bone mass do not provide estimates of the genetic variability in mechanically relevant parameters such as cross-sectional geometry. Few genetic studies of bone size or geometry have been conducted in humans (237-242), and most have used DXA estimates of bone size, HSA or radiographic analysis. At the femoral neck, heritability of parameters of cross-sectional geometry such as CSA and Z (estimated with HSA) range from 37% to 62% (241). Polymorphisms in several candidate genes responsible for variation in femoral neck bone geometry have been identified including the IGF-I gene (243), the tumour necrosis factor alpha (TNF-a) gene (244) and the ER-a gene (245). However, compared with aBMD, the genetics of bone geometry and strength are not well understood. Further, there have been no heritability or linkage studies of bone cross-sectional properties as measured with pQCT. Recently, Volkman et al. (246,247) assessed the influence of genetic determinants of cortical bone geometry (assessed by uCT) in the mouse femur using quantitative trait loci (QTL) analysis. In contrast to the high heritability estimates of BMC and aBMD, genetic markers accounted for only 3-22% of trait variances in cross-sectional geometry (246,247). The authors discussed several ways that genes may influence bone structure: 1) directed activity of osteoblasts and osteoclasts, which in turn influence bone size and shape; 2) indirect action on factors such as muscle strength/force, which in turn alter mechanical load and thus bone structure; and 3) alter the mechanostat setpoint, which in turn would affect bone adaptation to loading (246). With the exception of the first proposed mechanism, these proposals highlight the interaction between genetic and environmental effects, specifically mechanical loading (84). Thus, although genetics plays a key role in shaping skeletal morphology, appropriate mechanical loading is necessary for the development of a healthy skeleton (31,84). 1.2.5.2 Race and Ethnicity The terms race and ethnicity are often used interchangeably in the current literature. Further, researchers often fail to differentiate between these terms. This poses significant challenges when interpreting and comparing results across studies. Within bone research, there is currently no consensus regarding the definitions of race and 42 Chapter 1 - Literature Review ethnicity. For the purposes of this review I will use the terminology as outlined by the authors and provide definitions if available. Epidemiological studies have documented that the incidence of hip fracture across races is lower in non-Caucasians compared with Caucasians (248). However, among Asians there has been a dramatic increase in hip fracture rates over the past 3 decades. The highest rates were observed in urbanized countries (249). In addition, the rate of vertebral fracture was reported to be higher in Asians compared with Caucasians living in North America (250). The differences in fracture rates across races has been attributed primarily to variation in aBMD, with Blacks > Caucasians > Asians. Although this trend has also been partially explained by differences in bone size (251), very little is known about the contribution of bone geometry to fracture rates across races (252). An apparent discrepancy is noted in the lower aBMD and lower fracture risk in Asians compared with Caucasians. DXA studies suggest that lower fracture risk may be related to shorter hip (or femoral neck) axis length (HAL or FNAL) (253) and a smaller femoral neck angle (254) in Asians. Recently, Wang et al. (255) demonstrated that racial differences in femoral neck size, cortical thickness and indices of bone strength (estimated with DXA) may be related to growth- and age-related differences in periosteal apposition and endocortical resorption. For example, young adult (Australian) Chinese women (-30 years of age) had a significantly shorter FNAL, narrower femoral neck (smaller periosteal diameter) and thinner cortex than young adult Caucasian women after adjusting for differences in body size. However, from young to old age (-70 yrs) similar increments in femoral neck periosteal and endosteal diameter occurred in Chinese and Caucasian women. As a result, in old age, Chinese women maintained a smaller femoral neck periosteal diameter, section modulus and buckling ratio (255). The authors proposed that racial-differences in the onset of puberty and peak growth velocity (i.e., Chinese earlier than Caucasians) may explain the structural differences in femoral neck dimensions such that earlier exposure to estrogen in Chinese children may result in a shorter leg length and FNAL compared with Caucasians (255). Several studies have shown that age of onset of puberty may be earlier in (mainland or Hong Kong) Chinese than Caucasian girls (256,257); however, this trend may not apply to Chinese children that live in westernized countries (258). Of the few studies that examined race/ethnic differences in bone parameters during growth, most have focused on DXA measured BMC and aBMD. Results from cross-sectional and prospective studies suggest that after adjusting for body size, there is little to no difference in BMC or aBMD, or the gain in these parameters, between Asian and Caucasian (defined by country of origin) prepubertal children (259-264). However, ethnic differences in bone mineral may become more apparent in the later stages of puberty. MacKelvie et al. (261) compared Tanner stage 2 Asian and Caucasian girls and found substantial differences in BMC and aBMD across several skeletal sites in favour of Caucasian girls. This finding is in agreement with previous cross-sectional studies that suggested ethnic differences in bone mass may become apparent with advancing maturity (211,265,266). To date, one cross-sectional study has examined racial differences in bone geometry during growth. Gilsanz and colleagues (211) used QCT to measure properties of the axial and appendicular skeleton in 80 Black and 80 Caucasian boys and girls between 8 and 18 years of age (definition for race not provided). The children were matched for age, gender, height, weight and Tanner Stage. Race did not influence the cross-sectional area of the 43 Chapter 1 - Literature Review vertebral bodies. However, race did influence vertebral BMD as evidenced by higher BMD among Black children at puberty (211). In contrast, there was no difference for CoD at the femoral shaft between groups. There was, however, a significant difference between races for femoral shaft cross-sectional area and femoral length (211). When children across Tanner stages were collapsed, values for cross-sectional area were, on average, 3% and 8.4% greater in Black girls and boys, respectively compared with same sex Caucasian children (211). It is not known if similar differences in bone cross-sectional geometry and density exist between Asian and Caucasian children. Genetics is likely a significant factor underlying race/ethnic differences in fracture rates and bone phenotypes (i.e., bone size or density). Asian and Caucasian populations have different patterns of variation in polymorphisms of candidate genes for osteoporosis such as the vitamin D receptor gene and the collagen type 1 alpha 1 (COL1A1) gene, among others (267-269). For example, the high-fracture risk allele "s" of the COL1A1 gene was found to be absent in East Asian populations and more frequent in Caucasian populations (0.15 to 0.32) (267). It is not known how genetic differences may interact with documented differences in environmental and lifestyle factors between races/ethnicities (260,261). 1.2.5.3 Dietary Calcium Calcium is a major constituent of bone mineral and is thus an essential nutritional factor for optimal skeletal development. However, controversy exists over the amount of calcium that is necessary for bone health during growth (270,271) and whether calcium intake during childhood is related to fracture risk later in life (272,273). The majority of total body calcium forms hydroxyapatite crystals that comprise the inorganic phase of the bone matrix. The amount of calcium incorporated into the skeleton is determined by both calcium absorption (ingested - fecal) and retention (absorbed - (renal + dermal + endogenous secretion)) (274). Calcium retention is regulated by a complex homeostatic control mechanism that involves the actions of several calciotropic hormones (parathyroid hormone, vitamin D, calcitonin) (61) as well as possible genetic influences (275,276). From birth to the end of adolescence, 150 mg/day, on average, of calcium must be retained from the diet to meet the needs of the growing skeleton (277). The highest calcium retention is required during puberty, specifically for the 2 years around peak bone mineral accrual velocity (5). In the Saskatchewan PBMAS study, Bailey et al. (274) calculated peak calcium accretion rates in 130 children (60 boys) from longitudinal BMC measurements assuming a 32.2% calcium fraction in bone (278). Peak calcium accretion rates were 359 ± 81 mg/day at age 14 ± 1 yrs for boys and 282 ± 58 mg/day at age 12.5 yrs for girls. Slightly lower estimates of accretion rates (201) were used to determine the current adequate intake (Al) for calcium for children and adolescents (9-18 yrs: 1300 mg/day) (271). Although it is possible that higher intakes may be required during the adolescent growth spurt to optimize bone mineral accretion (275), calcium does exhibit threshold behaviour meaning that at a certain point there is little increase in retention with increased intake (279). The current Al for calcium during growth is based, in part, on results from several randomized, double-blind, placebo-controlled trials that have shown positive effects for calcium supplementation on BMC and aBMD in children (280-283). However, only one of these trials reported maintenance of bone gains in the supplemented group several years after the end of the intervention (284). Bonjour and colleagues (284) also noted larger gains in height in the 44 Chapter 1 - Literature Review previously supplemented group. They attributed the differences in longitudinal growth to the calcium intervention; however, maturity status was not controlled for within the analysis. Failure to adequately control for maturation and other confounding factors is one of several limitations of calcium supplementation trials that were recently discussed by Lanou and colleagues (270). An additional limitation of the current calcium literature is that few pQCT studies have investigated the effects of dietary calcium on bone geometry, density and strength during growth (285-289). Moyer-Mileur et al. (287) conducted a 12-month randomized controlled trial (RCT) in which 71 early adolescent girls (aged 12 yrs, Tanner stage 2) were randomized to either treatment (daily supplement = 800 mg calcium carbonate, 400 IU Vitamin D) or control (placebo). Peripheral QCT was used to assess changes in trabecular bone properties at the distal tibia (10% site). After adjusting for body size and menarcheal status, girls in the treatment group gained significantly more trabecular BMC (+4.1% vs. -1.6%) and TrbD (+1.0% vs. -2.0%) than controls. There was no difference between groups for 12-month change in TrbA (+3.1% vs. +2.0%). In subsequent regression analyses, it was determined that supplementation accounted for 5.2% and 10.4% of the variability in trabecular BMC and TrbD change, respectively. Two other supplementation trials that used pQCT were unable to determine a causal relationship between calcium and ToD of the radial shaft (289) or cortical bone properties at the tibial shaft (288) due to a lack of pQCT data at baseline. There is a need for well-designed RCTs that span adolescent growth in boys and girls to more closely investigate the relationship between calcium and pQCT-derived measures of bone density, geometry and strength. Calcium may have a more significant influence on skeletal development when in a deficient state (290); however, ethical concerns prevent such studies in children. Animal models have therefore been used to investigate the effects of low-calcium diets on the growing skeleton (291-293). In a recent histomorphometric study of growing female rats, Iwamoto et al. (293) found that rats fed a mild calcium deficient diet (0.1%) for 10 weeks had increased bone resorption and suppressed bone mineralization in trabecular bone at the proximal tibia compared with rats fed a normal calcium diet (0.5%). In addition, cortical area was significantly smaller in the low-calcium animals. Similar findings have been reported for the rat femur (291,292) with the ultimate result being a decrease in bone breaking strength (291). The negative effects of low dietary intake on skeletal development are likely mediated by the actions of parathyroid hormone (PTH) as levels of the hormone are significantly elevated in low calcium animals compared to controls (293). 1.2.5.3.1 Ethnic Differences in Calcium Intake An additional limitation of the current Al for calcium during growth is that it has limited use across ethnic populations because the calculations were based on data from Caucasian children (271). Ethnic differences in calcium intake are well established (260,261,294,295). Previous studies from our lab (260,261) found that pre- and early pubertal Caucasian children consumed 35-41% more calcium than their Asian peers. It also appears that the sources of dietary calcium vary by ethnicity. In Asian diets, a large percentage of the calcium intake may come from non-dairy sources such as breads, cereals, vegetables and legumes (296). In contrast, Caucasians are more likely to obtain a large percentage of dietary calcium from dairy products. The bioavailability of calcium from non-dairy 45 Chapter 1 - Literature Review sources is generally less than from dairy sources (297). The ethnic difference in calcium intake is also confounded by a higher prevalence of lactose intolerance among Asians (298). The dietary intake of calcium in Asian youths may have particular relevance to the incidence of hip fracture later in life as low calcium intakes among older Asian populations has been identified as a risk factor for hip fracture (299). However, it is not known if ethnic differences in dietary calcium are related to ethnic differences in bone strength at sites such as the proximal femur. 1.2.5.3.2 Calcium and Exercise Interactions As will be discussed in detail in Section 1.2.6, physical activity is a significant determinant of bone strength in the growing skeleton. Whether dietary calcium and physical activity have a synergistic effect on bone during childhood and adolescence is controversial (300-302). Few randomized trials in children have investigated the combined effects of calcium and exercise (288,303-305) and evidence from these trials in support of an interaction is weak. Of these trials, one used pQCT to investigate the influence of a possible calcium and exercise interaction on cortical bone properties at the tibia (20% site) in young children (3-5 yrs) (288). Unfortunately due to the exclusion of a large number of baseline pQCT scans, 12-month change in pQCT outcomes could not be assessed. Thus, it remains unclear if calcium and physical activity together provide an osteogenic stimulus more beneficial than physical activity alone. 1.2.5.4 Muscle The relationship between muscle and bone was acknowledged decades ago (306); however, it has only recently gained significant attention in pediatric bone research (31,307). As discussed, muscle forces incur the largest voluntary loads on the skeleton (30,88). During growth, the skeleton continually adapts to these loads to keep bone deformation within safe limits and in the absence of normal muscle forces, long bones fail to develop normal width, mass and longitudinal curvature (308). Due to the strong relationship between muscle and bone, an understanding of muscle development is important for pediatric studies of bone health. In this section I discuss measurement of muscle mass and force during growth as well as the few pediatric studies that have investigated the muscle-bone relationship in children. 1.2.5.4.1 Measurement of Muscle Mass There are several methods available for determining skeletal muscle mass or muscle size in children including limb circumferences, creatinine excretion and imaging techniques such as DXA, (p)QCT and magnetic resonance imaging (MRI) (197,309). For the purpose of this thesis I discuss only DXA and pQCT. 1.2.5.4.1.1 DXA DXA can safely and easily evaluate body composition in pediatric studies. The DXA body composition approach assumes that the human body consists of three components - fat, bone mineral and residual or "lean soft tissue" - that are distinguishable by their X-ray attenuation properties (310). Theoretically, three different photon energies are needed to separate the three different components. The two-energy DXA system circumvents this by 46 Chapter 1 - Literature Review first separating pixels with only soft tissue (fat + lean) from pixels containing soft tissue + bone mineral based on the ratio, or R value, of attenuation characteristics within each pixel. Soft tissue pixels are then further separated into fat or lean soft tissue pixels according to assumptions of stable attenuation ratios for both tissues (311). Total body lean mass can be used as a surrogate for skeletal muscle mass (312) and it is estimated that during childhood and adolescence lean body mass accounts for 70% of total body mass (197). The appendicular skeleton, which is approximately 75% skeletal muscle mass, represents the largest contribution to total body lean mass. DXA-derived lean mass has been validated against chemical analysis (313,314) and precision (in adults) with repositioning (%CV) in our laboratory with the Hologic QDR 4500W is less than 0.5% (Bone Health Research Group, unpublished data). Similarly, a recent study with 13 to 18 year olds found high reproducibility (ICC = 0.997) of the QDR 4500W for repeated measurements of total body lean mass (120). 1.2.5.4.1.2 pQCT In addition to characterizing bone geometry and density, pQCT can quantify muscle cross-sectional area (MCSA) in the appendicular skeleton. For the lower leg, the most common measurement site is the proximal two-thirds site, or 66% of total tibial length proximal to the distal end. In adults, this site correlates with the largest muscle belly (32); however, this relationship has not been defined in children. At the radius, MCSA is measured at 65% of total ulnar length proximal to the radial endplate because forearm circumference is largest at this site in adults (309). Similar to pQCT analysis for bone outcomes, MCSA is obtained with user-defined threshold driven algorithms. The analysis involves two steps: 1) separate muscle and bone from fat and 2) separate muscle from bone. If a small voxel size is used (0.4 mm), an additional step may be required to remove the skin from the total muscle cross-sectional area. There is good agreement between pQCT-derived MCSA and spiral CT-derived MCSA in adults (R2 = 0.9) (315); however accuracy of this method has not been determined in children. Similarly, precision of pQCT-derived MCSA has not been determined in children, but Neu et al. (309) reported a precision error of 1.93% for MCSA in adult women. 1.2.5.4.2 Measurement of Muscle Force According to mechanostat theory, it is muscle force that drives bone development, not muscle size (88). Direct measurements of muscle force can only be determined invasively with force transducers and thus, indirect methods such as dynamometry are used to estimate muscle force in children and adults (316,317). In addition, pQCT-derived MCSA can be used as a surrogate for potential force development of muscles (32,318). As proposed by Rittwegeret al. (32), the bending moment that a muscle exerts may be estimated by the product of MCSA and the length of the lever arm (i.e., tibial length). The relationship between MCSA and muscle force is based on the well established association between physiological muscle cross-sectional area (PCSA) and maximum isometric muscle force (319). Physiological CSA is calculated as the ratio of muscle volume to muscle length and as such, represents muscle area assuming a constant area along the entire muscle length. In parallel-fibred muscles (e.g., biceps 47 Chapter 1 - Literature Review brachii), maximum force increases linearly with PCSA. However, for pennate muscles (e.g. soleus), the force per anatomical cross-sectional area depends on pennation angle (319). Both grip force by dynamometry and MCSA by pQCT are used to estimate muscle force. However, to my knowledge there are no published reports of the relationship between these measures in healthy children, nor are there data to support the relationship between pQCT-derived MCSA and maximal force production in the lower leg. There are, however, clinical pediatric pQCT data showing a strong association (r = 0.91) between forearm grip force and MCSA in children with juvenile rheumatoid arthritis (320). In the legs, a moderate to strong association (r = 0.5-0.7) was reported for thigh MCSA determined with MRI and isokinetic strength (maximum force under dynamic conditions) at the knee in 10-14 yr old children (321). However, in the MRI study, once stature and weight were accounted for in the multilevel regression model, the influence of MCSA was non-significant. This indicates that MCSA is not the sole determinant of muscle force. Additional factors that may influence muscle force production in children include muscle pennation angle and neuromuscular characteristics such as contractile properties (197). Field measures of muscle performance include vertical and standing long jump (197). Jumping tests are used as indicators of muscle coordination and explosive power. In boys and girls, jump performance increases linearly with age and there are consistent sex differences in jump performance throughout childhood and adolescence (197). Force platforms such as the Leonardo Jumping Platform (Novotec GmbH, Pforzheim, Germany) measure both stationary forces (body weight) and variation in forces during vertical jumping (i.e., ground reaction forces) and can be used to assess muscle power (force x velocity) in children (322). Force platforms are a valuable research tool because they provide an indirect measure of the magnitude and rate of external load on the legs during weight-bearing activity; however, their use in pediatric research has been limited to date. 1.2.5.4.3 Sex Differences in Muscle Development Early radiographic studies of muscle widths helped to characterize the sex-specific patterns of muscle development (323). A small sex difference in muscle widths of the arm and calf exists during childhood, with boys having slightly wider muscles. This difference is magnified when testosterone levels increase in boys and they begin their adolescent growth spurt in musculature and it persists into adulthood, especially in the upper limbs (197,323). Similarly, cross-sectional pQCT data from the forearm indicate that both MCSA and grip force are greater in boys than girls at all stages of maturity (309). However, specific grip force (grip force per MCSA) was similar between sexes at all stages of development and thus appears to be independent of sex hormones. 1.2.5.4.4 The Muscle-Bone Relationship during Growth In this section I discuss the cross-sectional and longitudinal studies of the muscle-bone relationship during growth with a focus on studies that used pQCT. 1.2.5.4.4.1 Cross-sectional Studies Eckhard Schoenau's group at the University of Cologne in Germany may be considered the pioneers of pediatric pQCT research. Since their first pQCT study in 1996 (324) this group has contributed a great deal to our 48 Chapter 1 - Literature Review understanding of the functional muscle-bone unit in children and have highlighted the value of pQCT as a research tool. Initially, Schoenau (324) used the XCT-900 to evaluate the influence of muscle force on bone strength at the distal radius (20% site) in 168 males and females aged 3 to 62 years. Bone strength was estimated with a bone strength index (BSI) as the product of Z and CoD. Both BSI and isometric muscle force (as assessed by dynamometry) demonstrated an age-dependent course and appeared to peak between 25 and 30 years of age, especially in males. In addition, there was a strong association between BSI and muscle force (r = 0.87) at all ages. As a followup to the aforementioned study, Schoenau et al. (325) examined the influence of muscle force on properties of both trabecular and cortical bone at the distal radius in 14 healthy children aged 6 to 13 years. Whereas grip force correlated significantly with ToA, CoA and BSI (r = 0.8-0.9), there were no significant associations between grip force and CoD or TrbD. Based on these findings, the authors proposed that adaptation to muscle loads in the growing skeleton is dependent on changes in bone geometry rather than changes in density. Similar to the previous study, these results are limited by the questionable accuracy of cortical bone measurements at the distal radius due to the thin cortical shell (143). Despite this limitation, these early pQCT studies provided valuable insight to the functional muscle-bone unit and mapped future directions for pediatric pQCT trials. There is a need for comparable both cross-sectional and prospective exploratory studies in the lower limbs. A subsequent study assessed a much larger sample of children and adults (318 children, aged 6-22 yrs; 336 parents) who were participants in the DONALD study (Dortmund Nutritional and Anthropometric Longitudinally Designed Study) (326). In order to examine the influence of puberty on MCSA and CoA at the radial shaft (65% site), boys and girls were compared based on Tanner stage. Again, a strong correlation (r = 0.77) was observed between CoA and MCSA in all children, adolescents and adults. However, the authors noted that at Tanner stage 3, the ratio of CoA to MCSA was significantly greater in girls. Similar results were obtained when the ratio of radial shaft BMC (by pQCT) to MCSA was compared across Tanner stages and between sexes in the same cohort (327). These pQCT findings complement earlier DXA studies that also described a sex difference in the muscle-bone relationship during puberty. Originally described by Ferretti and colleagues (200), and repeated by Schiessl et al. (328) using cross-sectional data from Zanchetta et al. (329), the ratio of total body BMC to lean body mass has been shown to be greater for females after puberty. One explanation put forward to explain this difference is that sex hormones, specifically estrogen, may affect the mechanostat setpoint (54). This theory, and evidence surrounding it, was recently discussed in detail by Jarvinen et al. (330). In brief, rising estrogen levels during puberty may lower the (re)modeling threshold and in turn sensitize the bone adjacent to marrow to mechanical loading. This would result in increased endocortical apposition (89). From an evolutionary perspective, the estrogen-induced "packing" of bone into the female skeleton during puberty may serve to fill a calcium reservoir that is needed for reproduction (330). As discussed, the postulated endocortical (or endosteal) apposition may in fact be consolidation of cortical bone that cannot be adequately described by planar imaging techniques (212). Evidence also suggests that the muscle-bone relationship may be region and site-specific. Recent advances in magnetic resonance imaging (MRI) technology have allowed regional muscle mass and whole bone cross-sectional area to be quantified (331). Heinonen and colleagues (332) employed this technique to assess the muscle-49 Chapter 1 - Literature Review bone relationship at the tibial midshaft in growing girls. A unique aspect of this study was the division of the tibial cross-section into three anatomical sectors (posterior, anteromedial, anterolateral). Although ToA and total MCSA were strongly correlated, the only significant correlation between CoA and MCSA was observed in the anterolateral sector. Further, CoA and MCSA in this sector were significantly correlated with ground reaction forces during a side-to-side jump. These results highlight the region-specificity of the muscle-bone relationship in the lower limb; however, it is not clear how this relationship may change during growth or following an exercise intervention. Finally, a recent HSA study in 40 overweight (body mass index [BMI] > 85th percentile) and 94 healthy weight (BMI ^ 85th percentile) children and adolescents aged 4-20 years demonstrated the importance of assessing indices of muscle force separately from total body weight or fat mass (333). At the femoral shaft and narrow neck regions of the proximal femur, overweight subjects had 11% and 13%, respectively, higher estimated bone strength (section modulus) than healthy weight subjects when adjusted for height, maturation and sex. However, when lean mass was added to the regression model, femoral shaft and narrow neck bone strength were similar between overweight and healthy weight subjects. Total body fat mass did not contribute significantly to these models. Thus, in accordance with mechanostat theory (30,31), it is important to interpret estimated bone strength in the context of dynamic loads (indices of muscle force) rather than static loads represented by body weight. 1.2.5.4.4.2 Longitudinal Studies Few longitudinal studies have attempted to describe the changes in the muscle-bone relationship during growth (24,130,219,334). Ruff (219) used upper and lower limb radiographs from a sample of 20 subjects who were measured on average 34.5 times from near birth through late adolescence. Section modulus was estimated from radiographic humeral and femoral diaphyseal breadth measurements assuming a cylindrical^ shaped section. Muscle breadths were also measured radiographically at the mid femur and at the maximum width of the humerus. In this sample, there was a marked sex difference in the muscle-bone relationship over the entire age range which was more pronounced in the upper limb (219). In males, growth in humeral muscle size was highly correlated with change in humeral strength, while in females the two were less closely related. Significant sex differences were also documented for the femur; however, they became nonsignificant after controlling for body size (body weight x bone length). These findings suggest that muscular strength has a less pronounced effect on femoral bone strength which may be explained by the significant influence of body size on the weight-bearing lower limb (219). Although these findings provide strong support for the importance of mechanical factors in the development of bone strength, they are limited by the use of planar techniques. Recently, Rauch and colleagues (334) used longitudinal data from the University of Saskatchewan PBMAS to test the mechanostat hypothesis that increasing muscle force drives the development of bone strength during growth. Total body lean mass (LBM) and total body BMC were used as surrogates of muscle force and bone strength, respectively. As illustrated in Figure 1-17, PHV preceded the peak in LBM by an average of 0.30 years in boys and by 0.39 years in girls. In turn, the peak in LBM preceded the peak in BMC accrual velocity by an average of 0.36 years in boys and by 0.51 years in girls. Similar relationships were observed in the upper and lower extremities 50 Chapter 1 - Literature Review in both sexes. Additional regression analyses revealed that of sex, PHV and peak velocity for LBM (PVLBM), PVLBM was the only independent predictor of peak velocity for BMC accrual accounting for 40 and 60% of the variance. Although these results provide only an approximation of the muscle4Done relationship, they are in agreement with the mechanostat theory that muscle development precedes bone development during puberty. Males 5 5 CO 900O 8000-7O0O-6O0O-500& 400O 3000-200O 1000-Age PHV 13.45 y -BMC Age Peak 14.11 y Peak Value 404 g }--—L8M " Age Peak 13.75 y Peak Value 8550 g r5O0 -450 400 350 •300 •250 •200 150 100 h50 0 2L .E S CO 9000-j 8000-7000' 6000' 5000-40O0-3000-2000-1000' Females j Age PHV 11.80 y . LBM Ago Peak 12,19 y Peak Value 6050 g . BMC Ago Peak 12,69 y Peak Value 318 g i i i _ 10 11 12 13 14 i i i i i 15 16 17 18 19 500 450 Moo 350 300 250 200 •150 •100 •50 0 O 8. Age in y i i i i i i i i i i i 9 10 11 12 13 14 15' 16 17 18 19 Age in y Figure 1-17. Velocities of total body lean body mass (LBM) and total body bone mineral content (BMC) velocity during puberty in males and females. From Rauch et al. (334) with permission from Elsevier. The PBMAS data was also used to address the mechanostat hypothesis (54,335) that based on greater muscle mass and force in boys before PHV (336) bone strength should also be greater in boys than girls before PHV (130). As discussed, femoral neck CSA per unit lean mass was greater in girls than boys 2 and 3 years after PHV which would suggest greater axial bone strength in girls than boys (130). However, the authors argue that this sex difference has little mechanical relevance for the femoral neck where bending forces dominate the loading history (130). In contrast to the CSA findings, Z per unit lean mass was significantly greater in boys than girls at all biological ages due to boys' greater subperiosteal width (130). Although this finding agrees with mechanostat theory that estrogen may only alter the mechanostat setpoint on the endosteal surface next to bone marrow and not on the periosteal surface (89), the authors suggest that because the sex difference in Z was small and close to the error of measurement it was not biologically significant (130). It is interesting to note, however, that the sex difference was consistent across all biological ages. It is likely that an accurate picture of sexual dimorphism in femoral neck bone strength cannot be obtained with HSA due to the underlying limitations of planar DXA technology. 51 Chapter 1 - Literature Review To date, only one prospective pQCT study has evaluated the muscle-bone relationship during growth (24). Using 20-month followup data from the Healthy Bones Study II (HBSII) (8,9,261-263), we assessed changes in two bone-muscle strength indices (BMSI) for the tibial midshaft in 128 early, peri- and post pubertal boys and girls. Bone-muscle strength indices represent the strength of bone relative to its mechanical environment and provide a means to compare bone strength in groups where body size and muscle mass differ. As discussed, the tibia is subject to both compressive and bending stresses. Thus, we calculated a BMSI for compression as the ratio of CoA to MCSA (CoA/MCSA) and a BMSI for bending as the ratio of lever arm-adjusted section modulus [27(tibial length/2)] to MCSA. Based on the results of Schoenau and colleagues (326,327), girls would be expected to demonstrate an increase in both BMSIs over the 20-months and have higher values than boys for both ratios. However, our results did not provide support for the theory of estrogen-mediated changes in the mechanostat threshold. Both early and peripubertal girls experienced a slight decrease in BMSIs, while BMSIs were maintained in postpubertal girls (Figure 1-18). Further, comparisons between sexes within each maturity group showed a significantly greater increase in BMSIs in early- and peripubertal boys compared with early- and peripubertal girls. Although our findings were limited by the cross-sectional comparisons within maturity groups and different criteria used to assess maturity in boys and girls, they do provide insight to the muscle-bone relationship in the lower limbs during puberty. Differences between these results at the tibial midshaft and those of Schoenau et al. (326) at the radius may reflect site-specificity of the muscle-bone relationship or may be a function of differences in study design. There is a need for prospective pQCT studies similar in design to that of the Saskatchewan PBMAS to clarify the maturity- and sex-specific development of the muscle-bone relationship in the upper and lower limbs. 35 o < o o 4.0H 2.0-H O.OH -2.0-H -4.0-H -8.0H Dashed = Girls Solid = Boys •r ! i ! i i ! I ! I I I I 1 EARLY PERI POST EARLY PERI POST Figure 1-18. Baseline (A) and 20-month absolute change (B) values for cortical area to muscle area ratio (CoA / MCSA, bone-muscle strength index for compression) for girls (dotted lines) and boys (solid lines) across early 52 Chapter 1 - Literature Review pubertal (EARLY), peri-pubertal (PERI) and post-pubertal (POST) maturity groups. Bars indicate 95% confidence intervals, (a) Boys > Girls, P < 0.05. Adapted from Macdonald et al. (24) with permission from Elsevier. 1.2.5.4.5 Effects of Exercise on the Functional Muscle-Bone Unit during Growth The strong biomechanical link between muscle and bone suggests that increased muscle size, strength or force due to physical activity would lead to greater strain on bone and a resultant increase in bone mass, size and strength (307,337). However, there is little evidence to support the hypothesis that an exercise-induced increase in muscle size, strength or force directly influences bone adaptation to loading during growth. As I will discuss in more detail in Section 1.2.6, several HSA studies in children have shown how the relationship between physical activity and bone strength during growth is mediated by estimated muscle force (i.e., total body lean mass) (135,338,339), but they do not provide evidence of a cause-effect relationship. Similarly, in a recent cross-sectional MRI study, Daly et al. (340) investigated differences in the muscle-bone relationship between the playing and non-playing arms of 47 competitive female tennis players (aged 8 to 17 years). Similar to other studies of racquet sport players, muscle size and bone mass, size and strength were greater in the playing arm. Further, percent side-to-side differences in muscle area were positively associated with the percent side-to-side differences in BMC, CoA, ToA and polar moment of inertia. However, in regression analysis side-to-side differences in muscle area accounted for only 12-16% of the variance in side-to-side differences in bone parameters. This suggests that other factors associated with loading may contribute to skeletal adaptations to exercise. The effect of exercise on the muscle-bone relationship in children has yet to be studied prospectively. 1.2.6 Physical Activity and Bone Health in Children As discussed in Section 1.2.2.2 mechanical loading provides a significant osteogenic stimulus to the immature (animal) skeleton. The relationship between physical activity and bone health during human growth has been reviewed extensively (6,181,341,342) and it is clear that the same principles of adaptation observed in animal studies hold true in the human skeleton. To date, the majority of studies that evaluated the effects of physical activity on the growing skeleton used DXA and thus, much of the existing literature focuses on the relationship between physical activity and peak bone mass. Although the use of imaging modalities such as pQCT and programs such as HSA are becoming more widespread, there are still only a handful of studies that assessed the relationship between physical activity and bone geometry, density and estimated strength indices in children using cross-sectional (135,338,343-345) or longitudinal (339) data or controlled intervention trials (8,11,13,27,134,346,347). In this section I first briefly discuss the measurement of physical activity in children, in particular the Physical Activity Questionnaire for Children (PAQ-C). Next I review cross-sectional and prospective studies that either used pQCT to compare bone geometry, density and strength between athlete and non-athlete populations or to evaluate the role of physical activity as a determinant of cortical and trabecular bone properties in the growing skeleton. Finally, I discuss 6 physical activity intervention trials that evaluated the bone structural response to increased activity using either pQCT or HSA. 53 Chapter 1 - Literature Review 1.2.6.1 Measurement of Physical Activity in Children Accurate methods assess children's physical activity are essential to determine current levels of activity, describe the relationship between physical activity and health outcomes and assess the effectiveness of interventions designed to increase physical activity (348,349). Physical activity is "bodily movement produced by the contraction of skeletal muscle that increases energy expenditure above the resting level" (350). Based on this definition, the criterion standards for physical activity assessment are direct observation, doubly labelled water (DLW) and indirect calorimetry (348). Unfortunately, these methods are not practical for large studies as they require expensive equipment (DLW, calorimetry) and are associated with high researcher burden (direct observation) (351). Therefore, a wide range of methods have been developed to estimate a child's energy expenditure. These include objective measures such as heart rate monitoring, accelerometry and pedometry and subjective measures such as self-report questionnaires (348). Many of the objective measures are also inappropriate for large scale field research projects due to high costs and potential for subject reactivity. Further, it is difficult to determine the most appropriate method for research as many of these methods have not been validated against a gold standard (351). In addition, one must consider the strengths and weaknesses of each method, the size and characteristics of the study population and the specific research objectives. Ultimately, the method chosen should be valid, reliable, practical and nonreactive (352). Self-report instruments are most commonly used to assess children's physical activity levels (353). This is likely due to the low costs, low researcher burden, relatively quick administration time and thus, the ability to measure large numbers of children in a short time period (354). This thesis will focus on one self-report questionnaire, the Physical Activity Questionnaire for Children (PAQ-C), which has been used previously in several prospective studies including the University of Saskatchewan Bone Mineral Accrual Study and the UBC Healthy Bones Studies (5,10,262,263). 1.2.6.1.1 Physical Activity Questionnaire for Children The PAQ-C is a self-administered 7-day recall questionnaire that was designed to assess habitual MVPA in children aged 8-14 years participating in the 6-year University of Saskatchewan Bone Mineral Accrual Study (5,354). It was developed through a multi-step process that included item modification based on feedback from students, research assistants and item analysis. The final questionnaire consisted of ten items, nine of which are used to calculate a summary activity score. The other question assesses whether the child was sick in the previous week, or was prevented from normal activity as a result of other events. Within each PAQ-C item, physical activity is described as "sports, games, or dance that make your legs feel tired, or make you sweat", and all items are scored on a 5-pont scale (1 = low activity to 5 = high activity, continuous scale). The PAQ-C has since been modified to include an estimate of time per activity session (item 1) as well as involvement in extracurricular activities (sports, music lessons, tutoring, language lessons, etc.), the number of nights per week spent in organized sporting activities, and the number of hours of television watched and/or video/computer games played per day (355). The PAQ-C has been validated against other self-administered questionnaires, teacher rating, uniaxial accelerometer counts (Caltrac), fitness test (step test) and interview-assisted recall (r = 0.28-0.63) (356). The 54 Chapter 1 - Literature Review moderate correlation between the Caltrac and the PAQ-C (r = 0.39) is similar to correlations found with other self-and interview-administered questionnaires (357). The lack of strong association in validation studies of self-report is attributed to limitations in memory and recall skills, particularly in young children (less than 10 yrs), overestimation of physical activity and bias due to social desirability (348,357,358). A further limitation of the PAQ-C is that it does not discriminate between specific activity intensities (i.e., moderate and vigorous). Despite these limitations, the PAQ-C is a cost- and time-effective instrument that is useful in studies of large-scale populations. However, the PAQ-C was a reliable means to assess physical activity levels when administered several times during one school year. Crocker and colleagues (354) found acceptable levels of test-retest reliability for both girls (r = 0.82) and boys (r = 0.75) aged 9 to 14 years after one week. Further, when the questionnaire was administered during the fall, winter and spring seasons, correlations were greater than 0.80 for the average of two or three responses. Similarly, across the 6 PAQ-C measurements for the Healthy Bones Study (355) the reliability improved for both girls and boys when the average of multiple assessments was used (single assessment: r = 0.34-0.59; multiple assessments: r = 0.75-0.90). Crocker et al. (354) also noted that the PAQ-C was a sensitive means to detect physical activity differences between boys and girls, and differences across seasons. This feature is especially relevant as a substantial body of literature supports both seasonal (359) and sex (360,361) differences in physical activity. Currently, the PAQ-C is unable to discriminate between group activity levels; however, this has been identified as an area to examine in future (356). 1.2.6.2 Cross-sectional Studies in Children - General Physical Activity Few studies have investigated the relationship between general physical activity and bone geometry or strength during growth using HSA or pQCT. In the Iowa Bone Development Study, Janz and colleagues (135) used both accelerometry and parental report to determine the relationship between physical activity and bone geometry (CSA, Z) as estimated with HSA in 467 young children (mean age 5.2 yrs). This study represents the first time accelerometers were used to determine the relationship between physical activity intensity and bone structural variables in children. After adjusting for age, weight and height, participation in vigorous (5: 2818 accelerometer counts) activity was positively correlated (r = 0.19 to 0.32) with CSA and Z at each of the three regions of the proximal femur in boys and girls. Moderate (^ 527 counts) activity was also significantly associated with CSA and Z but the relationship varied between sexes and sites. In contrast, sedentary activity (< 152 counts) was negatively associated with CSA and Z at each region in girls only. When the relationship between vigorous activity and bone geometry was explored further in linear regression models, vigorous activity was found to explain, on average, 7% of the variance in CSA and Z at each region in boys and girls. When lean mass was included in the regression model as an estimate of muscle force the amount of variance explained by vigorous activity decreased to approximately 4%. Thus, lean mass explains some, but not all, of the relationship between physical activity and bone geometry (135). Given that current physical activity guidelines for children include both moderate and vigorous physical activity (362), it would be interesting to know how much, if any, of the variance in CSA and Z is accounted for by moderate activity. 55 Chapter 1 - Literature Review Recently, Wang et al. (344) used pQCT to evaluate the influence of leisure-time physical activity on cortical bone properties at the tibial diaphysis (60% site) in 242 pre- and early pubertal Finnish girls. A physical activity score was calculated for each participant from a self-report questionnaire. The score incorporated the frequency, estimated intensity and duration of weekly physical activities as well as an indication of whether the activity was weight-bearing or not. Girls were classified as low-, moderate- or high-active based on their score and were also categorized into low- or high-impact physical activity groups according to whether their "favourite" activity was weight-bearing. In the prepubertal girls, there was no consistent trend in the differences in tibial bone properties between the three activity groups. High-active girls had a 7% larger ToA than moderate-active girls and a 2% greater CoD and 4% greater CTh than low-active girls, but low-active girls had larger ToA and CTh than moderate-active girls. When the prepubertal girls were compared based on participation in weight-bearing activity, all bone outcomes were significantly greater in the high-impact group. There were no significant differences in any cortical bone properties between activity groups of early pubertal girls. Based on these findings, the authors suggest that prepuberty may be the most beneficial time for physical activity to effect bone development. It is interesting to note, however, that there was a trend for lean mass to be greater in the low-active group of Tanner II girls. Therefore, an association between physical activity and cortical bone properties may have been observed if lean mass had been controlled for in the analysis. Alternatively, differences between groups may have been confounded by inaccuracies associated with the questions and calculations used to determine the activity score since these had not been used in previous studies, nor had they been validated. Similar associations between physical activity and cross-sectional bone properties have not been reported in boys. However, a recent population-based study of 1068 Swedish men (mean age 19 yrs) found that men who began participating in regular physical activity before age 13 had significantly greater CoA at the tibial shaft (25% site) and trabecular density at the distal tibia (4%) site than boys who began participating in physical activity at age 13 or later. This result is similar to those reported from pQCT studies of male (363) and female racquet-sport athletes (153). I discuss these studies in more detail in Section 1.2.6.4. 1.2.6.3 Longitudinal Studies in Children - General Physical Activity With the exception of the few intervention studies discussed in Section 1.2.6.6, there have been no prospective pQCT studies of the relationship between physical activity and bone geometry or strength in healthy children. Results from the Saskatchewan PBMAS provided conclusive evidence that active children (with PAQ-C activity scores in the highest quartile) have greater absolutes values for DXA-derived bone mineral content and greater bone mineral accrual than their less-active peers (5). To followup the work of Bailey et al. (5), Forwood et al. (339) applied HSA to proximal femur scans of the PBMAS cohort to investigate the influence of physical activity on femoral neck bone strength during adolescence. Children were classified as physically inactive if their age-sex-specific z score for the PAQ-C fell below the lowest quartile, physically active if their z score fell in the highest quartile and of average activity if their z score fell between the lowest and highest quartile. To account for repeated measures within individuals and individual growth 56 Chapter 1 - Literature Review characteristics sex-specific hierarchical random-effects models were created using a multi-level modeling approach. Within these models maturation was controlled for using age at peak height velocity. In both boys and girls, physical activity was a significant predictor of narrow neck CSA and Z but not SPW. Differences in Z between inactive and active boys and girls are illustrated in Figure 1-19. It is not clear whether differences between activity groups were significant for all biological age groups. Importantly, when leg length and leg lean mass were entered into the random-effects models (instead of height and weight), the significant effects of physical activity were no longer apparent. Similar to the results of Janz et al. (135) discussed above, the effects of physical activity on bone strength are likely mediated by the relationship between physical activity and lean mass, which provides an estimate of muscle force. -2 -1 0 +1 +2 -2 -1 0 +1 +2 Years from APHV Years from APHV Figure 1-19. Growth curves for femoral neck (narrow neck region) section modulus (Z) comparing 17 active boys to 17 inactive boys (A) and17 active girls to 17 inactive girls (B). Values for Z represent adjusted means (height, weight) and are plotted against biological age groups (years from age of peak height velocity, APHV). Adapted from Forwood et al. (339) with permission from Elsevier. 1.2.6.4 Cross-sectional Studies in Adults - Racquet-sport Athletes Racquet-sport athletes provide an excellent model to study the relationship between physical activity and bone health. Loaded-to-unloaded arm comparisons reduce the influence of confounding factors such as genetics, hormones and nutrition. Several unilateral control studies (153,363-365) conducted at the UKK Institute in Finland contributed significantly to the bone and physical activity literature and advanced our understanding of how the skeleton adapts to mechanical loading. Initially, DXA was used to compare BMC and estimated bone strength (CSMI, Z) at the humerus and radius in players who had started their training in childhood (young starters) or before menarche and those who began their careers later in adulthood (old starters) or after menarche (364,365). Kannus et al. (364) divided players into groups according to the number of years before or after menarche that their training began. Among all players, the side-to-side difference in BMC at all measured sites was significantly greater than controls. However, women who began training before or at menarche had a two to four times greater side-to-side difference in BMC at the humerus and Chapter 1 - Literature Review radius than women who began their training 15 years after menarche. As a followup to this study, Haapasalo et al. (365) evaluated side-to-side differences in BMC and estimated CTh, CSMI and Z in 67 male and female tennis players and 57 sedentary controls. The female players were either young or old starters whereas all male players were young starters. Compared with controls, the players' relative side-to-side differences in BMC and estimated geometry and strength were significantly larger (+6 to +45%). Further, the side-to-side differences were much larger in young starters (+12 to +45%) compared with old starters (+3 to +12%). Together these findings suggest that intense physical activity, if begun during childhood or adolescence, has a pronounced positive effect on the growing skeleton. With the advent of pQCT, the UKK group was able to further investigate skeletal adaptations in adult racquet-sport athletes by focusing on side-to-side differences in bone geometry and estimated strength indices. In their first pQCT study, Haapasalo et al. (363) compared 12 former national-level male tennis players (25-35 yrs) who began their training during childhood with 12 age-matched controls. They measured trabecular and cortical bone properties at the distal, shaft and proximal sites of the radius and humerus and found that the significantly greater side-to-side difference in BMC and estimated bone strength (BSI, minimum and maximum moments of inertia: Lin, Lax) in the players was due to enlarged bone area (ToA, CoA) and not greater (volumetric) bone density (CoD, TrbD). The side-to-side differences in bone geometry (12-32%) and strength (23-67%) in the players were significantly greater than those for controls (geometry: 0.5-6%; strength: 5-16%). In the players, a side-to-side difference in CoD was only observed at the distal humerus where CoD was slightly greater (2%) in the non-playing arm. Similar findings were reported for female racquet-sport athletes who began training before or at menarche (young starters) (153). Compared with athletes who started training after menarche (old starters), young starters demonstrated greater side-to-side differences in humeral shaft ToA, CoA and CTh (13-20%), which in turn resulted in greater estimated bone strength of the playing arm (+26%) as measured by the torsional bone strength index (BSIt, density-weighted polar section modulus) (Figure 1-20). There were no side-to-side differences for CoD in young starters. Similar adaptations were observed in the old starters; however, the magnitude of the side-to-side difference in BSIt was significantly smaller (+11%) due to less periosteal expansion (+2.6%). At the distal radius, a significant side-to-side difference was found for total BMC (9%) and TrbD (+5%) in the young starters compared with controls, but not in ToA. Although these findings from the upper limbs do not represent the entire skeleton, they do suggest that during growth the immature skeleton adapts to increased loading mainly through periosteal expansion at shaft sites and through increased TrbD at distal sites. As discussed, these adaptations serve to improve bone bending/torsional and compressive strength at the shaft and distal sites, respectively, according to the type of load incurred at each site (62,64,71). 58 Chapter 1 - Literature Review 35 30 H 25 CD i 20 Q m 15 10 H • CoA BBSIt Young Starters Old Starters Controls Figure 1-20. Side-to-side differences (%) in cortical area (CoA) and torsional bone strength index (BSIt) at the humeral midshaft in young starters (began training before menarche), old starters (began training after menarche) and controls (no training). Bars are 95% confidence intervals. Adapted from Kontulainen et al. (153). 1.2.6.5 Cross-sectional Studies in Children - Athlete Populations The literature provides few cross-sectional comparisons of bone structural adaptations of loading between child athletes and non-athlete controls. Most recently, Ward and colleagues (345) conducted a study with 44 young male and female pre-pubertal gymnasts and 42 controls of a wide age range (5-12 yrs). Peripheral QCT was used to assess bone geometry, density and estimated strength (SSI) at the distal radius (4% of forearm length), radial midshaft (50% of forearm length), distal tibia (10 mm from the distal metaphysis) and proximal two-thirds tibia (65% of tibial length). At both shaft sites, gymnasts had 5-13% greater SSI than controls and this strength advantage was a result of a 5-8% greater CoA with no differences in CoD at either shaft site. Gymnasts also had a significantly greater MCSA (4%) than controls at the proximal two-thirds tibia. In contrast, at distal sites of the radius and tibia, gymnasts had 5-21% greater ToD and TrbD than controls, but bone areas were not significantly different between the two groups. These findings are consistent with the aforementioned studies of racquet-sport players and again highlight the site-specific differences in bone adaptation (within and between bones) according to the loads experienced at the distal and shaft sites. Similar site-specific differences in bone structure were observed by Faulkner et al. (338) in an HSA study of the proximal femur in 30 premenarcheal gymnasts and 20 age-matched controls. At both the narrow neck and shaft sites of the proximal femur, size-adjusted bone bending strength (Z) was significantly greater for gymnasts than controls. However, geometric adaptations that contributed to the greater strength varied between the regions of interest at the proximal femur. At the narrow neck region, gymnasts' CSA was larger due to a smaller endocortical 59 Chapter 1 - Literature Review diameter. Despite a smaller subperiosteal width in gymnasts, the distribution of mass was sufficient to confer greater bending strength at this site. In contrast, the gymnasts' greater bending strength at the femoral shaft was a result of a greater subperiosteal width and moment of inertia. This site-specificity may reflect differences in loading conditions between the narrow neck and femoral shaft, but it must be noted that HSA is unable to evaluate bone geometry outside of the image plane. The strength advantages observed in the gymnasts disappeared after lean mass was controlled for. Consistent with HSA findings discussed previously (135,339), these results provide further support for the strong biomechanical link between physical activity, muscle forces and bone strength. In addition to pQCT and HSA, magnetic resonance imaging (MRI) has also been used to evaluate the relationship between physical activity and bone cross-sectional geometry. Bass et al. (343) used MRI to evaluate side-to-side differences in ToA, CoA and CavA at the distal and midshaft sites of the humerus in 47 competitive female tennis players aged 8-17 yrs. To determine if maturity status influenced the bone response to loading, the players were classified as either pre-, peri- or postpuberal according to Tanner stage. In the prepubertal group side-to-side differences in CoA were associated with greater periosteal expansion (larger ToA) whereas in the postpubertal group side-to-side differences were associated with greater endocortical contraction (smaller CavA). The side-to-side differences in bone strength (measured by the polar second moment of inertia) appeared to be greater among the prepubertal girls, but were similar between peri- and postpubertal girls. The authors suggest that because endocortical expansion predominated during the later stages of puberty that no additional gains in bone bending strength were achieved. Unfortunately, due to the cross-sectional design of this study and the lack of comparison between side-to-side differences in control participants it is difficult to accurately assess changes in bone geometry that occurred as a result of growth and those that occurred as a result of physical activity. 1.2.6.6 Exercise Interventions with Children To date, most intervention trials with children have used conventional DXA measures of BMC and aBMD to evaluate the bone response to increased physical activity. Results from the UBC Healthy Bones Studies (HBSI and II) (8-10,262,263) represent a significant contribution to the pediatric bone and exercise literature. In addition to being the first (and longest, 20-months) school-based intervention to implement a bone-loading program within the physical education curriculum, HBS II identified a "window of opportunity" during puberty when bone is highly responsive to loading and demonstrated the effectiveness of a simple exercise program on gains in bone mass in Asian and Caucasian girls and boys during growth (8-10,262,263). Most pediatric physical activity intervention studies are conducted in schools and there are a number of reasons why this is so. First, it is suggested that since children spend a significant portion of their formative years in school the school setting may provide the best opportunity to positively influence childhood physical activity behaviours (366). Second, schools cater to large and diverse numbers of children. This is advantageous from a public health perspective as it increases intervention reach (367). Finally, schools typically have facilities and resources required to provide physical activity opportunities (368). Therefore, activity programs and/or exercises that 60 Chapter 1 - Literature Review are part of an intervention study can be implemented within physical education (PE) classes or can use existing activity space and equipment within the school. In randomized trials of school-based interventions, the school (cluster) is most often the unit of randomization, whereas the children are the unit of analysis. This design is advantageous as it prevents contamination that would occur if intervention and control children attended the same school. However, in this situation, children within a school cannot be regarded as independent observations, and as a result the effective sample size is less than the total number of individual participants (369,370). The reduction in sample size depends on the average cluster size and the degree of correlation between clusters (369). The correlation between clusters is known as the intracluster (or intraclass) correlation coefficient (ICC or p) and is the proportion of the total variance of the outcome that can be explained by the variation between clusters (371). Statistical software and methods are available to adjust for the ICC in an analysis (372). However, of the school-based studies discussed below, those that randomized by school (8,134,346) did not account use appropriate statistical techniques. As a result, the variability of the intervention effect may be underestimated (372). Only 7 intervention studies have used pQCT, HSA or estimates from DXA scans to describe changes in bone geometry, density or estimated strength that occurred in response to an exercise regime or physical activity program (8,11,13,27,134,346,347). The results of these studies are summarized in Table 1-2. In addition, one 12-month randomized controlled trial used pQCT to evaluate the effects of calcium supplementation and physical activity on cortical bone properties at the tibial shaft (20% site) in 3-5 year old children (288). Unfortunately, 12-month change in pQCT outcomes could not be evaluated in that study due to poor quality (movement artefacts) of pQCT scans at baseline. 1.2.6.6.1 Exercise Interventions with Prepubertal Children Two of the 6 studies described in Table 1-2 involved boys who were prepubertal at baseline (8,346). It is difficult to compare results between these studies due to differences in the: 1) change in maturational status of the cohort, 2) type and length of the intervention and 3) techniques used to assess changes in bone structure. These factors may help to explain why the observed structural adaptation to increased loading differed between the two studies. In the 8-month study of Bradney et al. (346) intervention boys demonstrated a greater gain in cortical thickness compared with control boys. However, this was a result of a greater decrease in endocortical (medullary) diameter rather than a greater increase in periosteal diameter as was observed in intervention boys in the HBSII study (8). At diaphyseal sites such as the mid-femur, resistance to bending forces is achieved through increased periosteal bone formation (76). In addition, increasing testosterone levels in boys during puberty may stimulate periosteal expansion (373). Thus, the surface-specific response to increased loading may be related to maturational stage and the associated growth velocity of each bone surface. It is possible that the advanced maturational stage of the boys in the 20-month HBS II study (77% advanced to early puberty) may have conferred a readiness for adaptation at the femoral neck, whereas boys in the Australian study remained prepubertal at followup and had not yet experienced rapid growth-related changes at the periosteal surface. 61 Chapter 1 - Literature Review Alternatively, the type of loading associated with each intervention may explain the differences in structural adaptation. The activities performed by the Australian boys involved essentially normal loading but increased magnitude (346). The authors suggest that these stresses may not have been sufficiently unusual in distribution to increase strains at the periosteal surface and that opposing muscle contraction may have generated axial stress that in turn would increase endocortical apposition (346). In contrast, the high-impact jumping circuit in the HBS II study was designed to provide dynamic strains to the growing skeleton in order to maximize the osteogenic potential of the intervention (8). This exercise program may have been more effective for increasing strains at the periosteal surface. Finally, it is possible that the ruler function used by Bradney et al. (346) may lack the precision necessary to detect small changes in periosteal dimensions. In addition, neither the DXA ruler function nor HSA are able to evaluate structural changes in other planes that may be important to overall bone strength. This highlights the need for 3-dimensional imaging technologies to assess bone structural and surface-specific adaptations that contribute to bone strength. Most recently, Valdimarsson et al. (347) investigated the effects of increased general physical activity on bone mineral accrual and estimated bone width in grade 1 and 2 girls aged 7 to 9 yrs. Unlike previous studies, this 12 month intervention did not involve any activities designed to be osteogenic. Instead, the one intervention school provided daily PE (40 minutes/day) which included general activities as part of the Swedish school curriculum (i.e., ball games, running and jumping). The control participants continued with their regular PE curriculum of 60 minutes/week. Although the intervention was effective for increasing lumbar spine BMC and estimated bone width of the third lumbar vertebrae, it had no effect on change in femoral neck BMC or bone width. Further, the lumbar spine results are questionable due to the significantly greater gain in fat mass in intervention girls. This may have resulted in an overestimation of changes in bone mass by DXA (126). Interestingly, Valdimarsson et al. (347) suggest that because no additional resources were required and that the intervention could be led by classroom teachers, this type of intervention could "instantly be organized in all schools". In addition, they suggest that by continuing with regular PE, children would be more motivated to participate than if they had to perform repetitive jumping activities (347). These points are debatable for several reasons. First, daily PE may not be feasible in all schools due to limited access to the school gymnasium. Second, studies of children's activity levels in physical education suggest that less than 50% of class time is spent in moderate-to-vigorous activity (MVPA) (374,375). Factors contributing to low levels of MVPA in PE include low student motivation, poor teaching skills or lack of adequate resources, space, training, and administrative support for PE (374). Thus, if this type of intervention is to be implemented on a wide-scale it is likely that additional resources and teacher training would be required to ensure delivery of high-quality physical activity. 1.2.6.6.2 Exercise Interventions with Mixed-Maturity Cohorts Exercise intervention studies that included children in distinct maturational categories at baseline identified the early pubertal years as a window of opportunity when osteogenic effects of exercise are magnified. In particular, exercise-induced gains in BMC and aBMD observed in the HBS II were specific to early pubertal girls while bone 62 Chapter 1 - Literature Review mineral changes in prepubertal intervention girls did not differ from same-maturity controls (263). Few studies (11,27,134) have investigated maturity-specific changes in bone geometry or strength in response to increased loading. As a followup to the work of MacKelvie et al. (263), Petit and colleagues (134) investigated whether structural adaptations (estimated with HSA) in response to the HBS II intervention would also be maturity-specific. Similar to the DXA findings, there was no significant difference in change between prepubertal intervention and control girls for CSA, CTh or Z at any of the three proximal femur regions. However, early pubertal intervention girls demonstrated greater gains in CSA, CTh and Z at the narrow neck than early pubertal controls. Significant bone structural adaptations in the early pubertal girls may reflect an interaction between increased estrogen levels and the bending forces associated with the specific jumps. Estrogen may promote bone formation on the endosteal surface in the presence of increased loads by lowering the theoretical mechanostat threshold (88). In contrast, estrogen may counteract the positive effects of exercise on the periosteal surface via activation of the ERB receptor (189). It is not known if a similar maturity-specific window of opportunity exists at other skeletal sites. Based on the results from studies of racquet sport athletes that demonstrated a 2-fold greater skeletal benefit of loading in women who began their training before menarche (364,376), Heinonen and colleagues (11) compared the bone response to a 9-month high-impact exercise intervention between pre- (Tanner stages 1-3) and postmenarcheal (Tanner stages 2-5) girls. Consistent with their previous investigations (364,376), high-impact exercise had a significant effect on BMC accrual in premenarcheal girls only. This finding provided further support for the concept that the premenarcheal period is a critical time during which the osteogenic potential of exercise is maximized. However, despite significantly greater gains in BMC at the lumbar spine and femoral neck (3 and 4%, respectively) in the premenarcheal exercising girls, they observed no group differences in CoD, CoA or BSI (density-weighted section modulus) at the tibial shaft (Figure 1-21) (11). Given the wide range of Tanner stages within the premenarcheal group, it is possible that the growth-related changes in bone size may have been large enough to mask the effects of the exercise stimulus in the lower limbs (11). B A 15 15 O CO cu 10 CD co sz o CD 03 CL p = 0.01 p = 0.57 co m & 10 C3> O 03 03 Q_ p = 0.28 p = 0.58 Premenarcheal Postmenarcheal Premenarcheal Postmenarcheal Figure 1-21. Nine-month percent change in femoral neck bone mineral content (BMC) (A) and tibial midshaft bone strength index (BSI) in pre- and postmenarcheal intervention (shaded bars) and control (white bars) girls. Significance values for group comparisons within the pre- and postmenarcheal groups are provided. Bars are 95% confidence intervals. Adapted from Heinonen et al. (11) 63 Chapter 1 - Literature Review Johannsen and colleagues (27) attempted to identify a maturity-specific response to high-impact activity in both boys and girls. Compared with the aforementioned interventions, this study was unique by way of the novel jumping program employed. Intervention children performed 25 drop jumps per day from a 45 cm box, 5 days per week for 12 weeks. This program was designed based on results from animal studies that demonstrated significant gains in bone strength ingrowing rats with a minimum of 5 jumps per day (14). Despite the unique bone-loading program, this study is fraught with a number of methodological limitations. First, the authors aimed to compare the bone response to the intervention between pre-, peri- and pubertal children. However, due to the small overall sample size (n = 54) of both boys and girls, exclusion of 28% of baseline pQCT scans and the large variability in the age and maturity status of participants it would seem that this study was underpowered to detect a maturity-specific effect. Second, the intervention lasted only 12-weeks. Given that a complete bone remodeling cycle in children may take up to 4 months (29) and that an intervention period of at least 10 months is recommended to account for the bone remodeling transient (61), it would seem that a 3-month intervention period is insufficient. In light of these limitations, the noted gains in BMC and ToD (by pQCT) at the distal tibia (4% site) among pubertal children (Tanner stage 4 and 5) in the intervention group are difficult to interpret. A longer study with participants in a narrower maturational range is needed to determine the effectiveness of this jumping program. The efficacy of a similar jumping program, Bounce at the Bell, was recently evaluated by McKay et al. (13). Children at intervention schools performed 10 high-impact countermovement jumps 3 times daily at the time of the morning, noon and afternoon bell for 8-months. Although this study was limited by the non-randomized study design, the results suggest that short bouts of high-impact activity separated by rest periods offer promise as a simple and cost-effective strategy to increase proximal femur bone mass in pre- and early pubertal children. The effects of Bounce at the Bell on bone structure in the pilot study were less clear possibly due to the small sample size of boys and girls. There is a need for a larger randomized controlled trial of Bounce at the Bell to determine the effectiveness of this program for increasing bone strength in boys and girls. 1.2.6.7 Exercise during Growth: Effects on Bone Later in Life It is apparent that increased mechanical loading during growth has beneficial effects on both the material and structural properties of the skeleton. Whether these exercise-induced adaptations are preserved in the absence of ongoing intervention is currently a contentious issue (377). In addition, due to the logistics of conducting a large, long-term randomized controlled trial, it is not known (and may never be) whether enhanced bone structure during childhood contributes to reduced fracture risk (378). Several studies of former athletes suggest that vigorous activity during childhood contributes to the maintenance of skeletal health in both men (379,380) and women (381,382). In addition, sports-exercise history during adolescence predicts 16% to 22% of the variance in HSA-derived femoral neck and shaft bone strength in young adult women (383) although this relationship may be explained, in part, by lean mass (384). Finally, epidemiological data support the association between physical activity during childhood and adolescence and decreased fracture risk (380,385,386). 64 Chapter 1 - Literature Review Of the pediatric intervention trials discussed, only the trial conducted by Heinonen et al. (11) assessed maintenance of the bone gains in a followup study (387). One-year after the end of the intervention, trainees demonstrated a 5% advantage in lumbar spine BMC accrual compared with controls and tended to have a greater gain in proximal femur BMC (2-3%). The tibial midshaft was not assessed in the followup study so it is not known if the slightly greater gains in CoA and BSI observed in the trainees after the 9-month intervention was maintained. Several deconditioning studies in growing animals provide evidence that skeletal benefits achieved during training are lost in the absence of the exercise stimulus (118,388,389). However, when discussing the effects of "detraining" on skeletal health, it is important to consider the physical activity behaviours of children in the absence of intervention. There is a considerable body of evidence that supports the tracking of physical activity patterns from childhood through adolescence (390) and, although the correlations are generally low, also into adulthood (391,392). Therefore, active children who participate in a physical activity intervention will likely continue to be active once the program is stopped. Unfortunately, it is also a reality that activity levels among adolescents tend to decline, especially among girls (393). In order to curb this decline, it is imperative that physical activity promotion strategies be tailored towards children and adolescents so that we may attempt to instil positive, long-lasting health behaviours in this population. In addition, there is a need for simple, targeted bone-loading programs such as Bounce at the Bell, to become part of the regular physical activity program in elementary schools in order to optimize development of the growing skeleton. 65 Table 1-2. Physical activity interventions with prepubertal children and mixed maturity cohorts First Author Subjects and Design Intervention Statistical approach Results Bradney (346) Subjects: BOYS, Caucasian n = 20 INT (10.4 yrs at baseline, all prepubertal) n = 20 CON (age, ht and baseline aBMD matched to INT boys) 2 schools randomly allocated to INT or CON Program: Extra weight-bearing physical activity (B-ball, weight training, aerobics, soccer etc) in addition to regular PE class Frequency & duration. 30 min, 3x/week, 8 months Progression: none stated INT and CONT boys matched for age, sitting height, height and baseline aBMD Unpaired t-tests: 8-month changes different between groups Did not account for within school variance. Femoral midshaft: Peri diam: NS Endo diam: INT - 4.9%, CON - +2.3% CTh: +6.4% CSMI: +2.4% (CON > INT) Z: +1.6% (CON > INT) DXA: Lunar DPX-L MacKelvie Subjects: BOYS, 44% Asian, Tanner (8) stage 1 at baseline n = 31 INT n = 33 CON Randomized by school (stratified by number of participants) DXA: Hologic QDR 4500W with HSA analysis Program: Classroom-based high impact jumping program Frequency & Duration. 10-12 min, 3x/week, 2 school years Progression: # jumps & height of jump (progressed through levels) advanced every 8-10 weeks Yr 1: 50 (baseline) to 100 (final) jumps Yr 2:55 jumps (baseline) to 132 (final) jumps Independent t-tests: baseline comparisons ANCOVA: 20-month change in DXA and HSA outcomes - covariates: baseline bone, change in height, final Tanner stage Bivariate correlations: body composition and bone bending strength Did not account for within school variance. INT > CON: NN CSA: +2.5%, NS NN SPW: +2.6%, NS NNCSMI: 12.4%, p< 0.05 NN Z: +7.4%, p < 0.05 NN ED: + 2.9%, NS NN CTh: NS IT CSA: NS (p = 0.07) IT SPW, CSMI, Z, ED: NS IT CTh: NS (p = 0.07) FS: NS cn Table 1-2 continued. First Author twills -t Subjects and Design Intervention Statistical approach - Results Valdimarsson Subjects: GIRLS, Caucasian, Tanner Proqram: Increased number of Univariate and multivariate INT > CON (347) stage 1 at baseline physical education classes. Classes ANCOVA: baseline age and annual TB BMC: NS included both indoor and outdoor increments in height and weight as LS BMC: +4.7% n = 53 INT general physical activity (i.e., ball covariates FN BMC: NS n = 50 CON games, running, jumping) Leg BMC: NS NOT randomized: 1 school assigned to Frequency & Duration: 40 min/day Bone width: INT, controls were volunteers from 3 (200 min/wk) L3: +2.9% neighbouring schools FN: NS Progression: none stated DXA: Lunar DPX-L version 1.3z Heinonen Subjects: GIRLS, Caucasian Proqram: Jump training sessions, ANCOVA: baseline values and age Pre-menarcheal (INT>CON): (11) two- and one-foot jumps, used jump as covariates (within each maturity LS BMC: + 3.3% N = 58 Premenarcheal (25 INT, 33 CON, boxes, also did aerobic exercises group) FN BMC: + 4.0% ~ 11 yrs at baseline) Troch BMC: NS N = 68 Postmenarcheal (39 INT, 33 Frequency & Duration: 50 min (20 Individual BMC values obtained CON, -14 yrs at baseline) min jumping), 2x week, 9 months from the ROI were normalized by Tibia: 50% site the length of the ROI CoD, CoA, BSI: INT CON NOT randomized: schools self-selected Progression: into INT group Month 1:100 2-foot jumps, no box Post-menarcheal: Months 7-9:150 two-foot and 50 one- INT «-> CON (at all sites) DXA: Norland XR-26; oQCT: Norland leg box jumps (multidirectional) XCT 3000 Table 1-2 continued. First Author Subjects and Design Intervention Statistical approach Results Petit (134) Subjects: GIRLS, 34% Hong Kong Program: Classroom-based high Separate analyses for each Prepubertal: INT «-» CON Chinese, 57% white impact jumping program maturity group: ANCOVA (baseline wt, change in ht, Tanner Breast Early-pubertal (INT > CON): N = 68 Prepubertal (43 INT, 25 CON, 10 Frequency & Duration : 10-12 min, stage and sport nights) yrs at baseline) 3x/week, 7 months NN BMD: + 2.6% IT BMD:+ 1.7% N = 106 Early-pubertal (43 INT, 63 CON, Did not account for within school NN CSA:+ 2.3% ITCSA:NS 10.5 yrs at baseline) Progression: # iumps and height of variance. NN CTh: + 3.2% IT CTh: NS jump (progressed through levels) NN ED: NS IT ED:-1.4% Randomized by school (stratified by NNZ: + 4.0% ITZ:NS ethnicity) - started with 50 jumps per session and progressed to 100 jumps per FS: NS DXA: Hologic QDR 4500 with HSA session analysis Johannsen Subjects: GIRLS & BOYS, ethnicity not Program: Hiqh-impact jumping t-tests: baseline differences Main effects: (27) stated program conducted in schools and between INT and CON DXA: childcare centres TB BMC: INT > CON (-1%) N = 26 CON (14 girls, 12 boys, 10.0 + 5.1 ANCOVA: baseline weight, sex, Leg BMC: INT > CON (-1.5%) yrs at baseline, 13 Pre, 7 Peri, 6 Post) Freguencv & Duration : 25 Tanner stage pQCT: NS jumps/day, 5x/week, 12 weeks N = 28 INT (17 girls, 11 boys, 10.3 + 5.3 Least-square means & Tukey- Interaction effects (group x yrs at baseline, 13 Pre, 5 Peri, 10 Post) Progression: None stated Kramer HSD test: interaction of maturity): (children jumped off 45 cm - high box, pubertal stage and intervention Tib BMC (4% site): p = 0.04 Randomized (blocks of 2) by gender and GRFs 4-5x BW group (baseline wt, calcium, sex as LS BMC: p = 0.10 age group covariates) ToD (4% site): p = 0.03 Intervention effect in pubertal group DXA: Hologic 4500A; pQCT: Stratec only (Tanner stage 4 or 5) XCT 2000 CD CO Table 1-2 continued. First author Subjects and Design Intervention Statistical approach Results McKay (13) Subjects: Girls & Boys, 48% Asian N = 51 INT N = 73 CON (from HBS II study) Proqram: "Bounce at the Bell" - simple ANOVA: to compare baseline bone jumping program (countermovement and descriptive variables and 8-jumps) Frequency & Duration: 10 jumps, 3x/day, 8 months 65% Tanner I at baseline, -10 yrs of age 3 schools volunteered for the intervention Progression: none DXA: Hologic QDR 4500W with HSA analysis month change in descriptive variables ANCOVA: 8-month change in bone variables - covariates: baseline bone, weight, change in height, final Tanner stage, PA load time INT > CON: TB BMC: +0.9% (p = 0.004) TB BA: +1.4% (p = 0.036) PF BMC: +2.1% (p = 0.019) IT BMC: +2.7% (p = 0.017) NN Z: +3.3% (NS) NN CSA: +2% (NS) NNCTh:+1.2%(NS) No sex x group interactions INT = intervention; CON = control; yrs = years; ht = height; aBMD = areal bone mineral density (g/cm3); DXA = dual energy x-ray absorptiometry; Peri diam = periosteal diameter (mm); Endo diam = endosteal diameter (mm); CTh = cortical wall thickness (mm); CSMI = cross-sectional moment of inertia (mm4); Z = section modulus (mm3); HSA = Hip Structural Analysis; Peri circ = periosteal circumference (cm); Endo circ = endosteal circumference (cm); CoA = cortical bone cross-sectional area (mm2); CSA = cross-sectional area (cm2); NN = narrow neck; IT = intertrochanteric region; FS = femoral shaft; SPW = subperiosteal width (cm); ED = endosteal diameter (cm); NS = non-signficant; LS = lumbar spine; BMC = bone mineral content (g); FN = femoral neck; L3 = third lumbar vertebrae; Troch = trochanteric region; <-+ = no difference; BSI = bone strength index; GRF = ground reaction force; BW = body weight CD Chapter 1 - Literature Review 1.2.7 Summary of Directions for New Research on Bone Strength in the Growing Skeleton The preceding literature review highlighted the need for further studies of bone strength in the growing skeleton using imaging techniques such as pQCT and HSA. In particular, there is a paucity of data characterizing the bone structural response to physical activity in boys and girls. 1.2.7.1 Sex Differences, Muscle and Tibial Bone Strength There is disagreement in the current literature as to whether sex differences in long bone cross-sectional geometry, (volumetric) density and strength are evident as early as prepuberty. Results from pQCT studies of the distal and proximal radius suggest that a strength advantage in favour of boys exists during prepuberty (20,21), whereas at weight-bearing sites (femur) bone geometry is similar between prepubertal boys and girls (207,208,213). Sex differences in tibial bone strength have yet to be investigated in pre- and early pubertal children using pQCT. Further, given the significant mechanical challenge posed by increasing muscle forces on the growing skeleton, there is a need for pQCT data to be interpreted in the context of a functional model of bone development (31,88). In addition to the influence of muscle, there is also need to understand the role of other biological (maturity, ethnicity) and lifestyle (physical activity, dietary calcium) factors on tibial bone strength in pre- and early pubertal children. 1.2.7.2 Physical Activity and Bone Strength in the Growing Skeleton Peripheral QCT is a novel imaging technique that evaluates skeletal adaptations to a physical activity intervention in children. To date, only two controlled intervention studies have used pQCT to investigate the effects on high-impact exercise on tibial bone geometry, density and strength and results from these trials were inconclusive (11,27). There remain unanswered questions regarding the type and duration of exercise required to elicit a bone structural response at the tibia and whether such a response may differ between metaphyseal and diaphyseal sites within the same bone. Results from animal studies suggest that short bouts of high-impact activity separated by rest periods may provide a significant osteogenic stimulus (14,15); however the effectiveness of a similarly designed bone-loading program (13) for increasing tibial bone strength in boys and girls has yet to be determined. In addition to evaluating the appendicular skeleton, there is a need to determine the effectiveness of a novel bone-loading program on bone strength at the clinically relevant proximal femur. Results from the HBS II trial (8,134) indicated that HSA is a valuable tool that can be used to supplement conventional measures of proximal femur bone mass with estimates of bone cross-sectional geometry and strength. From a public health perspective, effective school-based physical activity programs have the potential to positively influence pediatric bone health on a population-wide basis. However, to be feasible in the school and classroom setting bone-loading programs must be inexpensive and easily administered by generalist teachers. As few of the previous school-based interventions have met these criteria there is a need for development of an innovative physical activity model and the subsequent evaluation of this model in a large cohort of children using novel bone imaging techniques. 70 Chapter 1 - Rationale, Objectives and Hypotheses 1.3 Rationale, Objectives and Hypotheses In this chapter, I outline the rationale, objectives and hypotheses for each of the three studies that make up this thesis. In addition, I provide the scientific contribution that each study will make to the pediatric bone health field. 1.3.1 Part I: Bone Strength and its Determinants in Pre- and Early Pubertal Boys and Girls Rationale. Higher fracture rates in women than men may be related to a sex difference in bone strength that may emerge during growth (187). Despite a number of pediatric DXA studies that address this (5,203,204,232,261,394), there is still confusion about when the sex difference in bone strength appears and which factors (i.e., body weight, muscle mass or force) have the greatest influence on bone strength during growth. This confusion is due, in part, to the use of BMC or aBMD as surrogates of bone strength as DXA outcomes ignore the contribution of bone geometry to bone strength. Peripheral QCT-derived bone strength indices which combine geometry and (volumetric) density provide a more accurate means to investigate sexual dimorphism in long bone strength. Peripheral QCT can also evaluate muscle cross-sectional area (MCSA) which is a surrogate for muscle force. Mechanostat theory postulates that the greatest direct mechanical challenges on the skeleton during growth come from increasing bone length and muscle forces, whereas additional modulating factors such as maturity, physical activity and nutrition may affect the mechanostat indirectly by influencing either longitudinal bone growth or muscle force (31). Although a number of pQCT studies evaluated bone strength in the growing skeleton (20,21,285,286), few interpreted their data in the context of the mechanostat model (24,326). Further, no pQCT study has evaluated the influence of ethnicity on tibial bone strength in boys and girls. Objectives. The primary objective is to determine if there is a sex difference in tibial bone strength and its components (geometry and density) in pre- and early puberty. The secondary objective is to evaluate the contribution of MCSA, a surrogate of muscle force, and modulating factors including maturity, ethnicity, physical activity, dietary calcium and vertical jump height to tibial bone strength. Primary Hypothesis. After adjusting for tibial length and MCSA, tibial bone strength will be similar between pre- and early pubertal boys and girls. Secondary Hypothesis. After adjusting for tibial length, MCSA will be the primary explanatory variable of tibial bone strength in boys and girls. Contribution. This will be the first study to evaluate sex differences in tibial bone strength in pre- and early pubertal children using pQCT. In addition, this study will provide further insight into the muscle-bone relationship during growth and the relationships between non-modifiable (maturity, ethnicity) and modifiable factors (physical activity, dietary calcium, physical performance) and tibial bone strength. Results from this cross-sectional study will identify factors that contribute to bone strength in children. Based on these findings, researchers can develop intervention trials that address these factors. 71 Chapter 1 - Rationale, Objectives and Hypotheses 1.3.2 Part II: Sixteen-Month Longitudinal Study of a School-Based Physical Activity Intervention on Tibial Bone Strength in Boys and Girls. Rationale. The osteogenic effects of exercise on the growing skeleton are well recognized (4,6,341). A previous 20-month, school-based, bone-loading intervention in an ethnically diverse cohort of pre- and early pubertal boys and girls reported significant gains (+4.3-4.6%) in BMC at the clinically significant femoral neck (8,9,262,263). Few studies have assessed the bone structural adaptations to physical activity in children using pQCT (11,27,288). Only one of these trials (11) evaluated tibial bone strength and none of the studies evaluated bone strength at both distal and midshaft sites. Thus, it is not known if the bone response to physical activity differs between primarily trabecular (distal) and cortical (midshaft) sites of the tibia. Further, it is not known if adaptations to increased loading within a bone cross-section occur in a site-specific manner as no pQCT studies have investigated changes in tibial bone strength in the x- and y-bending planes. Finally, many school-based interventions have involved modifications to existing physical education curriculum (8,9,134,262,346,355). In light of current curricular demands this type of intervention may not be feasible or sustainable. Thus, there is a need for a novel and effective physical activity program for elementary schools that incorporates a simple bone-loading component. Objectives. The primary objective is to compare changes in bone strength at the distal tibia and tibial midshaft as estimated with pQCT-derived bone strength indices between children who participate in a 16-month school-based physical activity program, Action Schools! BC, and same-sex, non-participating controls. The secondary objective is to determine if, at the tibial midshaft, structural adaptations to increased loading differ in the x-and y-bending planes. Primary Hypothesis. Changes in bone strength indices will be greater in both boys and girls participating in Action Schools! BC compared with sex-matched, non-participating controls. Differences in change between intervention and control children will be similar at the distal tibia and tibial midshaft. Secondary Hypothesis. Changes in bone strength at the tibial midshaft will be similar in the x- and y-bending planes. Contribution. This is the largest school-based physical activity intervention undertaken to date in elementary schools. I use a novel assessment tool (pQCT) to determine the effects of physical activity on bone strength at the distal and midshaft tibia in boys and girls. This has not been achieved previously. Beyond what has been done before, this study will characterize the bone structural response to a novel bone loading program implemented within a larger school-based physical activity model, Action Schools! BC. If effective, this program will offer a simple and inexpensive means to improve bone strength in elementary school children. 72 Chapter 1 - Rationale, Objectives and Hypotheses 1.3.3 Partlli. Sixteen Month Longitudinal Study of a School-Based Physical Activity Intervention on Femoral Neck Bone Mass and Strength in Boys and Girls. Rationale. Our research group previously reported enhanced bone mass and structure compared with controls in both boys (8,262) and girls (9,134,263) who participated in a school-based, high-impact circuit training intervention for 20-months. Despite the effectiveness of HBS II, a model that modifies the physical education curriculum is not likely to be sustainable in elementary schools. Based on findings from animal studies (14,15) that demonstrated osteogenic effects of short bouts of high-impact loading separated by rest periods, our group designed the Bounce at the Bell program. Bounce at the Bell is a simple, inexpensive and feasible bone-loading model that can be easily implemented by generalist teachers. Action Schools! BC is a school-based physical activity model that incorporates Bounce at the Bell as part of the required classroom-based physical activity component. Bounce at the Bell was effective for increasing proximal femur bone mass in a combined sample of boys and girls. The effects of this program on boys' or girls' femoral neck bone strength as estimated with HSA were less clear. Hip structure analysis provides a means to estimate bone structure and strength in response to this novel intervention at the clinically relevant proximal femur. Objectives. The