UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Bone strength accrual across adolescent growth and the influences of physical activity and sedentary… Gabel, Leigh Elizabeth Christine 2017

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2017_may_gabel_leigh.pdf [ 19.78MB ]
Metadata
JSON: 24-1.0343985.json
JSON-LD: 24-1.0343985-ld.json
RDF/XML (Pretty): 24-1.0343985-rdf.xml
RDF/JSON: 24-1.0343985-rdf.json
Turtle: 24-1.0343985-turtle.txt
N-Triples: 24-1.0343985-rdf-ntriples.txt
Original Record: 24-1.0343985-source.json
Full Text
24-1.0343985-fulltext.txt
Citation
24-1.0343985.ris

Full Text

BONE STRENGTH ACCRUAL ACROSS ADOLESCENT GROWTH AND  THE INFLUENCES OF PHYSICAL ACTIVITY AND SEDENTARY TIME  by  Leigh Elizabeth Christine Gabel   B.Sc., The University of Western Ontario, 2007 B.A., The University of Western Ontario, 2008 M.Sc., McMaster University, 2011   A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY  in  THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Experimental Medicine)   THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  April 2017  © Leigh Elizabeth Christine Gabel, 2017 ii  Abstract  With recent advances in imaging technologies, we are acquiring a better understanding of the complex hierarchy of bone and how bone adapts its geometry, microarchitecture and ultimately, its strength to withstand the loads imposed upon it during adolescent growth. Thus, in this thesis, I examine the influence of physical activity (PA), sedentary time, maturity and sex on estimated bone strength and its determinants1 (i.e., microarchitecture, geometry and density) across adolescence. This thesis is based on the UBC Healthy Bones III Study (HBSIII), a mixed longitudinal cohort of healthy girls and boys age 8-12 years at study entry. We assessed bone strength, geometry and density at the tibial shaft using peripheral quantitative computed tomography (pQCT) and bone strength, microarchitecture, geometry and density at the distal tibia and radius using high-resolution pQCT (HR-pQCT). We assessed PA and sedentary time using accelerometry.  Four studies comprise this thesis. First, I investigated cross-sectional associations between sedentary time and bone strength and its determinants at the distal tibia by HR-pQCT. I found no associations between sedentary time and bone parameters. Second, I examined maturity- and sex-related adaptations of bone geometry and strength at the tibial shaft using pQCT. I found that larger bone area in boys provided them a greater bone strength advantage compared with girls across adolescence.  Third, I examined maturity- and sex-related adaptations of bone strength and its determinants by HR-pQCT at the distal tibia and radius. I found greater bone strength in boys across adolescence was underpinned by greater trabecular bone volume and total bone area.  Fourth, I examined prospective associations between PA, sedentary time and bone strength and its determinants at the distal tibia and radius using HR-pQCT. I observed greater bone strength and trabecular bone volume in participants engaging in more PA and lower total bone area in participants engaging in more sedentary time.                                                 1I use bone strength and its determinants to refer to bone microarchitecture, geometry and density throughout this dissertation. iii  Collectively, these studies enhance our understanding of how bone is gained during adolescence and add a unique perspective to the benefits of PA for bone strength and its determinants.   iv  Lay Summary   Bone strength is the bottom line in fracture prevention. However, the intricacies of how bone strength is gained during adolescence are not completely understood. Thus, I used advanced medical imaging tools to study how bone strength is gained across 12 years of adolescent growth in 393 participants from the UBC Healthy Bones III Study. I also examined the influence of maturity, physical activity and sedentary behaviour on bone strength accrual in boys and girls.  The studies that make up this thesis make several novel contributions to the pediatric bone research field. First, they represent the longest studies of bone growth during adolescence using three-dimensional imaging techniques and advanced statistical modelling approaches. Second, they challenge a pre-existing paradigm regarding differences in how bone is accrued between boys and girls. Finally, they highlight adolescence as a critical ‘window’ for bone health and underscore the importance of physical activity for bone strength accrual. v  Preface  This dissertation is an original intellectual product of the author, Leigh Gabel. Chapters 3-6 of this dissertation are versions of stand-alone manuscripts in the peer-reviewed academic literature. The first three chapters are published (Chapters 3-5) and the fourth (Chapter 6) has been submitted for publication and is currently undergoing peer-review. As first author, I led each of these chapters. I provide details of my contributions and those of my collaborators for each publication below.  This dissertation is based on the University of British Columbia (UBC) Healthy Bones III Study (HBSIII). The HBSIII was conceived of and designed by Professor Heather McKay (University of British Columbia) and received ethics approval from the UBC Behavioural Research Ethics Board (H15-01194, H07-02013, H2-70537). I began doctoral studies during the last year of the HBSIII data collection (2012), and I conducted and analyzed all pQCT assessments/scans during the 4-week data collection period. I subsequently became proficient in HR-pQCT scan acquisition and analysis and took a lead role overseeing all aspects of HBSIII HR-pQCT imaging data. I independently reviewed all HR-pQCT scans (over 2000 scans from 5 years of data collection) to quantify motion artifacts, conducted all standard HR-pQCT analyses and applied a customized segmentation algorithm to assess cortical porosity and thickness, including visually inspecting all segmentations to ensure correct differentiation between the cortex and trabeculae. Further, I led statistical analysis of the longitudinal HBSIII data using advanced multilevel modeling techniques. I conducted all of the statistical data analysis in this thesis. Throughout this dissertation, I use ‘I’ to refer to my individual contributions and ‘we’ to refer to contributions of the research team. Chapter 1: A version of this material was published as Gabel L and Macdonald HM. Exercise and the female skeleton. In: Gordon CM and LeBoff MS (Eds.) The female athlete triad: A clinical guide. Springer, New York; 2015: 39-69. As lead author of this book chapter, I was responsible for the literature review and drafting the manuscript. Dr. Macdonald defined the chapter outline and provided detailed feedback and edits on all versions of the chapter. Chapter 3: A version of this material was published as Gabel L, McKay HA, Nettlefold L, Race D, and Macdonald HM. Bone architecture and strength in the growing skeleton: the role of sedentary time. Medicine and Science in Sports and Exercise, 2015, 47(2): 363-72. As lead vi  author I was responsible for defining the research question, conducting the statistical analyses and drafting the manuscript. HR-pQCT data were collected in 2009 by Dr. Danmei Liu (Centre for Hip Health and Mobility) and accelerometry data were cleaned and processed by Douglas Race (HBSIII study coordinator). Dr. Lindsay Nettlefold and Douglas Race provided guidance related to accelerometry and feedback on near final drafts of the manuscript. Drs. Heather McKay and Macdonald provided detailed feedback and edits on all versions of the manuscript. Chapter 4: A version of this material was published as Gabel L, Nettlefold L, Brasher PM, Moore SA, Ahamed Y, Macdonald HM, and McKay HA. Reexamining the surfaces of bone in boys and girls during adolescent growth: a 12-year mixed longitudinal pQCT study. Journal of Bone and Mineral Research. 2015, 30(12): 2158-67. As lead author I was responsible for data collection (in 2012), conducting the statistical analyses and drafting the manuscript. I defined the research question in collaboration with Drs. McKay and Macdonald, whom also helped draft the manuscript and provided detailed feedback on all versions. Dr. Brasher provided statistical guidance and Dr. Nettlefold assisted with statistical analyses and manuscript reviews. Sarah Moore assisted with the estimation of maturity using age at peak height velocity. Drs. Macdonald and Ahamed assisted with data collection between 2001 and 2008.  Chapter 5: A version of this material was published as Gabel L, Macdonald HM, and McKay HA. Sex differences and growth-related adaptations in bone microarchitecture, geometry, density and strength from childhood to early adulthood: a mixed longitudinal HR-pQCT study. Journal of Bone and Mineral Research. 2017, 32(2): 250-63. As lead author I developed the research question, analyzed the HR-pQCT scans, conducted the statistical analyses and drafted the manuscript. Drs. McKay and Macdonald provided detailed feedback on all versions of the manuscript. Chapter 6: A version of this material was published ahead of print as Gabel L, Nettlefold L, Macdonald HM, and McKay HA. Physical activity, sedentary time and bone strength during adolescence: a mixed-longitudinal HR-pQCT study. Journal of Bone and Mineral Research. E-pub ahead of print, DOI: 10.1002/jbmr.3115. As lead author I developed the research question, analyzed the HR-pQCT scans, conducted the statistical analyses and drafted the manuscript. Drs. McKay and Macdonald provided detailed feedback on all versions of the manuscript. Dr. Nettlefold provided guidance regarding accelerometry analyses and feedback on the manuscript. vii  Table of Contents  Abstract .......................................................................................................................................... ii	Lay Summary ............................................................................................................................... iv	Preface .............................................................................................................................................v	Table of Contents ........................................................................................................................ vii	List of Tables .............................................................................................................................. xiv	List of Figures ............................................................................................................................ xvii	List of Abbreviations ............................................................................................................... xxvi	Acknowledgements .................................................................................................................. xxix	Chapter 1: 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 .................................................... 4	1.2.1.2	 Bone modeling and remodeling ...................................................................... 7	1.2.1.3	 Long bone geometry ....................................................................................... 8	1.2.1.4	 Longitudinal growth of long bones ................................................................. 8	1.2.2	 Bone biomechanics ............................................................................................... 10	1.2.2.1	 Material and mechanical properties of bone ................................................. 10	1.2.2.1.1	 Material properties of cortical bone ........................................................ 13	1.2.2.1.2	 Material properties of trabecular bone .................................................... 13	1.2.2.1.3	 Bone strength in bending and compression ............................................ 14	1.2.2.2	 Bone’s response to mechanical stimuli ......................................................... 15	1.2.2.2.1	 Mechanotransduction .............................................................................. 15	1.2.2.2.2	 Mechanostat theory ................................................................................. 17	1.2.2.2.3	 The functional model of bone development ........................................... 19	1.2.2.2.4	 Experimental evidence for bone adaptation to mechanical stimuli ........ 20	1.2.3	 Measuring bone in children and adolescents ......................................................... 22	1.2.3.1	 DXA .............................................................................................................. 23	1.2.3.1.1	 Hip structural analysis ............................................................................. 25	viii  1.2.3.2	 pQCT............................................................................................................. 26	1.2.3.2.1	 Image acquisition and analysis ............................................................... 27	1.2.3.3	 HR-pQCT ...................................................................................................... 29	1.2.3.3.1	 Image acquisition and analysis ............................................................... 30	1.2.4	 Maturity- and sex-related differences in bone strength and its determinants ....... 33	1.2.4.1	 Assessing maturity ........................................................................................ 34	1.2.4.1.1	 Sexual maturation ................................................................................... 34	1.2.4.1.2	 Skeletal maturation ................................................................................. 35	1.2.4.1.3	 Somatic maturation ................................................................................. 35	1.2.4.2	 Maturity- and sex-related differences in bone development ......................... 37	1.2.4.2.1	 Bone strength .......................................................................................... 39	1.2.4.2.2	 Bone geometry ........................................................................................ 41	1.2.4.2.3	 Bone density ............................................................................................ 43	1.2.4.2.4	 Cortical microarchitecture ...................................................................... 43	1.2.4.2.5	 Trabecular microarchitecture .................................................................. 44	1.2.5	 Factors that influence of bone strength during growth ......................................... 46	1.2.5.1	 Genetics ......................................................................................................... 46	1.2.5.2	 Hormones ...................................................................................................... 48	1.2.5.3	 Ethnicity ........................................................................................................ 50	1.2.5.4	 Calcium and vitamin D ................................................................................. 51	1.2.5.5	 Muscle force .................................................................................................. 54	1.2.6	 Physical activity and sedentary time ..................................................................... 56	1.2.6.1	 Measurement of physical activity ................................................................. 56	1.2.6.1.1	 Self-report questionnaires to assess physical activity ............................. 57	1.2.6.1.2	 Accelerometry to assess physical activity ............................................... 58	1.2.6.2	 Measurement of sedentary time .................................................................... 61	1.2.6.2.1	 Self-report questionnaires to assess sedentary time ................................ 61	1.2.6.2.2	 Accelerometry to assess sedentary time ................................................. 61	1.2.6.3	 Sex- and age-related differences in physical activity and sedentary time .... 62	1.2.7	 Influence of physical activity and sedentary time on bone strength development 63	1.2.7.1	 Intervention studies of physical activity ....................................................... 64	ix  1.2.7.2	 Observational studies of physical activity .................................................... 66	1.2.7.2.1	 Athletic populations ................................................................................ 67	1.2.7.2.2	 Habitual physical activity ....................................................................... 71	1.2.7.3	 Long-term effects of physical activity in childhood and adolescence .......... 73	1.2.7.4	 Observational studies of sedentary time ....................................................... 75	1.2.8	 Summary of the literature ...................................................................................... 77	1.3	 Rationale, objectives and hypotheses .................................................................... 77	1.3.1	 Bone strength and microarchitecture in the growing skeleton: the role of sedentary time ........................................................................................................................ 78	1.3.2	 Re-examining the surfaces of bone in boys and girls during adolescent growth: a 12-year mixed longitudinal pQCT study ............................................................... 79	1.3.3	 Sex differences and growth-related adaptations in bone microarchitecture, geometry, density and strength: a mixed longitudinal HR-pQCT study ............... 80	1.3.4	 Physical activity, sedentary time and bone strength from childhood to early adulthood: a mixed longitudinal HR-pQCT study ................................................ 81	Chapter 2: Methods .................................................................................................................... 83	2.1	 Healthy Bones Study overview ............................................................................. 83	2.1.1	 Healthy Bones Study and Bounce at the Bell ....................................................... 83	2.1.2	 Actions Schools! BC ............................................................................................. 85	2.1.3	 New cohort ............................................................................................................ 86	2.1.4	 Recruitment and retention ..................................................................................... 86	2.1.5	 Data collection overview ....................................................................................... 87	2.2	 Heathy Bones III Study protocol ........................................................................... 88	2.2.1	 Anthropometry ...................................................................................................... 88	2.2.2	 Health history questionnaire .................................................................................. 88	2.2.3	 Maturity ................................................................................................................. 89	2.2.3.1	 Sexual maturation ......................................................................................... 89	2.2.3.2	 Age at peak height velocity ........................................................................... 89	2.2.3.3	 Maturity offset equation ................................................................................ 90	2.2.4	 Dietary calcium intake .......................................................................................... 91	2.2.5	 Peak muscle power ................................................................................................ 91	x  2.2.6	 Self-reported screen time and physical activity .................................................... 92	2.2.7	 Objectively measured sedentary time and physical activity ................................. 92	2.2.8	 Bone imaging ........................................................................................................ 93	2.2.8.1	 pQCT............................................................................................................. 93	2.2.8.2	 HR-pQCT ...................................................................................................... 96	2.2.9	 Statistical analysis ............................................................................................... 101	2.2.9.1	 General linear mixed models ...................................................................... 101	Chapter 3: Bone Architecture and Strength in the Growing Skeleton: The Role of Sedentary Time ......................................................................................................................... 103	3.1	 Introduction ......................................................................................................... 103	3.2	 Methods ............................................................................................................... 105	3.2.1	 Study design ........................................................................................................ 105	3.2.2	 Anthropometry, maturity and dietary calcium .................................................... 106	3.2.3	 Sedentary time and physical activity ................................................................... 106	3.2.4	 Bone microarchitecture, geometry, BMD and strength ...................................... 107	3.2.5	 Statistical analysis ............................................................................................... 107	3.3	 Results ................................................................................................................. 108	3.3.1	 Descriptive characteristics ................................................................................... 108	3.3.2	 Screen time and bone parameters ........................................................................ 114	3.3.3	 Objectively measured sedentary time and bone parameters ............................... 114	3.3.4	 Factors that influence bone parameters ............................................................... 114	3.4	 Discussion ........................................................................................................... 117	3.5	 Conclusions ......................................................................................................... 120	Chapter 4: Re-examining the Surfaces of Bone in Boys and Girls During Adolescent Growth: A 12-year Mixed Longitudinal pQCT Study .......................................................... 121	4.1	 Introduction ......................................................................................................... 121	4.2	 Methods ............................................................................................................... 122	4.2.1	 Study design ........................................................................................................ 123	4.2.2	 Anthropometry and age at peak height velocity .................................................. 124	4.2.3	 Healthy history and dietary calcium .................................................................... 125	4.2.4	 Bone geometry, density and strength .................................................................. 125	xi  4.2.5	 Data cleaning ....................................................................................................... 126	4.2.6	 Statistical analysis ............................................................................................... 126	4.3	 Results ................................................................................................................. 128	4.3.1	 Descriptive characteristics ................................................................................... 128	4.3.2	 Comparisons of bone parameters between boys and girls at APHV ................... 129	4.3.3	 Comparison of annual accrual rates for bone parameters between boys and girls pre-and post-APHV ............................................................................................. 130	4.4	 Discussion ........................................................................................................... 134	4.5	 Conclusions ......................................................................................................... 139	Chapter 5: Sex Differences and Growth-Related Adaptations in Bone Microarchitecture, Geometry, Density and Strength from Childhood to Early Adulthood: A Mixed Longitudinal HR-pQCT Study ................................................................................................ 140	5.1	 Introduction ......................................................................................................... 140	5.2	 Methods ............................................................................................................... 141	5.2.1	 Study design ........................................................................................................ 142	5.2.2	 Anthropometry and age at peak height velocity .................................................. 142	5.2.3	 Health history and ethnicity ................................................................................ 143	5.2.4	 Bone microarchitecture, geometry, density and strength .................................... 143	5.2.5	 Statistical analysis ............................................................................................... 144	5.2.5.1	 Model building ............................................................................................ 145	5.3	 Results ................................................................................................................. 148	5.3.1	 Descriptive characteristics ................................................................................... 148	5.3.2	 General growth patterns at the distal tibia and radius ......................................... 157	5.3.3	 Comparisons of model estimates of bone parameters between boys and girls ... 157	5.3.3.1	 Tibia ............................................................................................................ 157	5.3.3.2	 Radius ......................................................................................................... 158	5.4	 Discussion ........................................................................................................... 166	5.4.1	 Trabecular microarchitecture .............................................................................. 166	5.4.2	 Cortical microarchitecture, bone geometry and estimated bone strength ........... 167	5.5	 Conclusions ......................................................................................................... 171	xii  Chapter 6: Physical Activity, Sedentary Time and Bone Strength from Childhood to Early Adulthood: A Mixed Longitudinal HR-pQCT Study ............................................................ 172	6.1	 Introduction ......................................................................................................... 172	6.2	 Methods ............................................................................................................... 173	6.2.1	 Study design ........................................................................................................ 174	6.2.2	 Anthropometry and age at peak height velocity .................................................. 175	6.2.3	 Health history, ethnicity and dietary calcium ...................................................... 175	6.2.4	 Physical activity and sedentary time ................................................................... 175	6.2.5	 Peak muscle power .............................................................................................. 176	6.2.6	 Bone microarchitecture and strength ................................................................... 176	6.2.7	 Statistical analysis ............................................................................................... 177	6.3	 Results ................................................................................................................. 182	6.3.1	 Descriptive characteristics ................................................................................... 182	6.3.2	 Influence of physical activity and sedentary time on bone parameters ............... 184	6.3.2.1	 Moderate to vigorous physical activity ....................................................... 184	6.3.2.2	 Sedentary time ............................................................................................ 185	6.4	 Discussion ........................................................................................................... 192	6.4.1	 Physical activity and bone strength ..................................................................... 192	6.4.2	 Physical activity and cortical bone ...................................................................... 194	6.4.3	 Sedentary time and bone parameters ................................................................... 195	6.5	 Conclusions ......................................................................................................... 198	Chapter 7: Integrated Discussion ............................................................................................ 199	7.1	 Overview of findings ........................................................................................... 199	7.1.1	 Bone strength and microarchitecture in the growing skeleton: the role of sedentary time ...................................................................................................................... 199	7.1.2	 Re-examining the surfaces of bone in boys and girls during adolescent growth 202	7.1.3	 Sex differences and growth-related adaptations in bone microarchitecture, geometry, density and strength ............................................................................ 204	7.1.4	 Physical activity, sedentary time and bone strength from childhood to early adulthood ............................................................................................................. 207	7.2	 Challenges and future directions ......................................................................... 211	xiii  7.2.1	 The use of pQCT and HR-pQCT imaging systems in pediatric bone research .. 211	7.2.2	 Maturity ............................................................................................................... 213	7.2.3	 Assessment of physical activity and sedentary time ........................................... 215	7.3	 Challenges with longitudinal study designs ........................................................ 216	7.4	 Public health implications ................................................................................... 217	7.5	 Future research .................................................................................................... 218	References ...................................................................................................................................221	Appendix A: Information to Participants, Consent and Assent Forms ................................252	Appendix B: Results for Study Participants ...........................................................................268	Appendix C: Questionnaires .....................................................................................................271	Appendix D: Determination of Age at Peak Height Velocity ................................................300	Appendix E: Additional Data for Chapter 4 ...........................................................................304	Appendix F: Additional Data for Chapters 5 and 6 ...............................................................307	 xiv  List of Tables  Table 1.1. Overview of studies that used HR-pQCT to examine sex and maturity-related adaptations in bone strength and its determinants during adolescent growth. .................. 39	Table 3.1. Descriptive characteristics and estimates of sedentary time for boys and girls in the full cohort and in the subsample with accelerometry data. Values are mean (SD) unless otherwise indicated. ........................................................................................................ 109	Table 3.2. Bone parameters at the distal tibia assessed using high-resolution peripheral quantitative computerized tomography (HR-pQCT). Values are mean (SD). ................ 110	Table 3.3. Unstandardized beta coefficients and model variances for multivariable regression analyses of bone parameters in boys. Beta coefficients ± standard error. Values in bold are significant at p < 0.05. ............................................................................................... 111	Table 3.4. Beta coefficients and model variances for multivariable regression analyses of bone parameters in girls. Beta coefficients ± standard error. Values in bold are significant at p < 0.05. ............................................................................................................................. 112	Table 3.5. Beta coefficients and model variances for multivariable regression analyses of bone parameters in boys and girls (Model 3). Beta coefficients ± standard error. .................. 113	Table 4.1. Characteristics of boys and girls at first pQCT measurement. Data are reported as mean (standard deviation) unless otherwise indicated. ................................................... 129	Table 4.2. Estimates of model intercepts. Intercepts represent the average value of the bone parameter at APHV (maturity offset = 0). Numbers in brackets are the standard error of the parameter estimate or the 95% confidence interval for the ratio. ............................. 130	Table 4.3. Estimates of fixed effects slopes and comparison between boys and girls. Slopes represent annual rates of accrual pre- and post-age at peak height velocity (APHV), adjusted for maturity offset and ethnicity. Numbers in brackets are the standard error of the parameter estimate or the 95% confidence interval for the ratio. ............................. 131	Table 5.1. Overview of study participants that comprise the Healthy Bones Study III cohort. . 142	Table 5.2. Characteristics of boys and girls at first HR-pQCT measurement. ........................... 149	Table 5.3 Number of HR-pQCT measurements by sex, site and maturity offset. ...................... 150	xv  Table 5.4. Estimates of model intercepts for the effects of maturity, sex and ethnicity as predictors of bone parameters at the distal tibia at age at peak height velocity. Numbers in brackets are the standard error of the parameter estimate. Bold values are p<0.05. ...... 151	Table 5.5. Estimates of model intercepts for the effects of maturity, sex and ethnicity as predictors of bone parameters at the distal radius at age at peak height velocity. Numbers in brackets are the standard error of the parameter estimate. Bold values are p<0.05. .. 153	Table 5.6. Adjusted means for bone parameters at the distal tibia at each maturity offset in boys (B) and girls (G). Maturity offset is years from age at peak height velocity. Data are presented as mean (standard error). Percent change is calculated over 12 years (from a maturity offset of -2 to a maturity offset of +9). ............................................................. 162	Table 5.7. Adjusted means for bone parameters at the distal radius at each maturity offset in boys (B) and girls (G). Maturity offset is years from age at peak height velocity. Data are presented as mean (standard error). Percent change is calculated over 12 years (from a maturity offset of -2 to a maturity offset of +9). ............................................................. 163	Table 6.1. Covariates used in mixed effects models, not including sex, ethnicity, MVPA and sedentary time. Time-varying covariates were retained if p < 0.05. Interactions terms were retained if they significantly improved model fit based on a reduction in the deviance test (-2∆LL) and model parsimony (AIC and BIC) values. ............................. 181	Table 6.2. Characteristics of boys and girls at first HR-pQCT measurement. ........................... 182	Table 6.3. Bone parameters for boys and girls at first HR-pQCT measurement. ....................... 183	Table 6.4. Longitudinal associations of between-person moderate-to-vigorous physical activity (MVPA; per IQR, 30 min) with bone parameters at the distal tibia and radius. Coefficients (95% CI) represent the difference in bone parameter between an individual in the upper quartile for MVPA compared with an individual in the lower quartile MVPA at maturity offset (years from age at peak height velocity) of 0. .................................... 185	Table 6.5. Longitudinal associations of between-person sedentary time (per IQR, 106 min) with bone parameters at the distal tibia and radius. Coefficients (95% CI) represent the difference in bone parameter between an individual in the upper quartile for sedentary time compared with an individual in the lower quartile for sedentary time at maturity offset (years from age at peak height velocity) of 0. ...................................................... 187	xvi  Table E.1. Estimates of model intercepts and fixed effects slopes between boys and girls without interpolation for measurement error. Slopes represent annual rates of accrual pre- and post-age at peak height velocity (APHV), adjusted for maturity offset and ethnicity. Numbers in brackets are the standard error of the parameter estimate. .......................... 305	Table E.2. Estimates of model intercepts between intervention and control group participants. Intercepts represent the average value of the bone parameter at age at peak height velocity (APHV; maturity offset = 0). Numbers in brackets are the standard error of the parameter estimate. ......................................................................................................... 306	Table E.3. Baseline Pearson correlations of age, sex, ethnicity, maturity, and anthropometric variables with bone parameters at the tibial midshaft by peripheral quantitative computed tomography (n=230). Correlations with sex, Tanner stage and ethnicity are Spearman’s rank order correlations. ................................................................................................... 306	Table F.1. Baseline Pearson correlations of age, sex, ethnicity, maturity, anthropometric variables, muscle power, dietary calcium and accelerometry variables with bone parameters at the distal tibia by high-resolution peripheral quantitative computed tomography (n = 393). Correlations with sex, Tanner stage and ethnicity are Spearman’s rank order correlations. ................................................................................................... 308	Table F.2. Baseline Pearson correlations of age, sex, ethnicity, maturity, anthropometric variables, muscle power, dietary calcium and accelerometry variables with bone parameters at the distal radius by high-resolution peripheral quantitative computed tomography (n = 351). Correlations with sex, Tanner stage and ethnicity are Spearman’s rank order correlations. ................................................................................................... 309	 xvii  List of Figures  Figure 1.1 Diagram of structural elements of long bones, illustrating cortical bone, comprised of osteons surrounding blood vessels, and trabecular bone. Reprinted from Martin et al.,[22] with permission from Springer New York. ......................................................................... 5	Figure 1.2 Anterior view of the right human femur with basic anatomy, including diaphysis, epiphysis and metaphysis. Modified from the online edition of the 20th US edition of Gray’s Anatomy of the Human Body, and originally published in 1918 and reprinted from Kontulainan et al.,[27] with permission from Karger. .......................................................... 6	Figure 1.3. The schematic on the left is of a long bone during embryonic development, including the growth plate. The image on the right shows proliferation of chondrocytes at the growth plate. Reprinted from Wallis,[40] with permission from Elsevier. .......................... 9	Figure 1.4 The hierarchical organization of bone, reprinted from Rho et al.,[25] with permission from Elsevier. .................................................................................................................... 10	Figure 1.5. The top image is a stress-strain curve divided into elastic and plastic regions. The bottom image displays the measurement of strength from the stress-strain curve. X marks the stress and strains where failure occurs. Reprinted from Turner and Burr,[45] with permission from Elsevier. ................................................................................................. 12	Figure 1.6. Scale drawings of three cylindrical cross-sections with different outer diameters, fixed length (L), but equal areal bone mineral density (BMD). Corresponding values of volumetric BMD (vBMD), bone mineral content (BMC) or cross-sectional area (bCSA), cross-sectional moment of inertia (CSMI) and section modulus. Reprinted from Beck,[54] with permission. ................................................................................................................ 14	Figure 1.7. Illustration of mechanocoupling. Bending forces cause deformation of osteocytes and create pressure gradients that drive fluid through canaliculae, from regions of compression to tension. The fluid flow generates shear stress on cell membranes. Reprinted from Duncan et al,[56] with permission from Springer. .................................... 16	Figure 1.8. Illustration of the mechanostat theory and influence of mechanical strain on bone modeling and remodeling. Theoretically, bone remodeling occurs in the upper limit of the trivial loading zone (or disuse zone) and in the physiological loading zone; bone modeling occurs in the overload zone; and microdamage repair occurs in the pathological xviii  overload zone. Based on Forwood and Turner[64] and reprinted from Bachrach et al,[65] with permission from Elsevier. ......................................................................................... 18	Figure 1.9. The functional model of bone development based on mechanostat theory. A feedback loop between bone deformation and bone strength is the central component of regulation of bone development and adaptation. 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 value. Factors shown in the bottom box modulate the regulatory system. Reprinted from Schoenau[68] and adapted from Rauch and Schoenau,[57] with permission from Nature Publishing Group. ........................................ 19	Figure 1.10. Three-dimensional images from micro-computed tomography (micro-CT) of exercise-related adaptations in bone microarchitecture at the distal femoral diaphysis in rats. Sedentary controls in A and C, exercised rats in B and D; cortical compartment in the top images, trabecular compartment in the bottom images. Reprinted from Joo et al.,[80] with permission from Elsevier. ............................................................................... 21	Figure 1.11. Illustration of densitometry. Photons are attenuated during transmission, producing an attenuation profile proportional to the mass of mineralized bone in the scanning path. Reprinted from Seeman,[87] with permission from Endocrine Society. ............................ 24	Figure 1.12. Image of peripheral quantitative computed tomography system (pQCT), model XCT 3000 (Stratec Medizintechnick GmbH). An illustration of leg positioning for pQCT tibia scans (by Vicky Earle, Medical Illustrator). ..................................................................... 27	Figure 1.13. Illustration of the partial volume effect (PVE), whereby pixels at bone edges (blue pixels) contain both bone and soft tissue densities, resulting in a lower density for the blue pixels. Smaller bones have more pixels close to the bone edge and may be more affected by PVE. Reprinted from Zemel et al.,[100] with permission from Elsevier. ........ 28	Figure 1.14. Image of high-resolution peripheral quantitative computed tomography (HR-pQCT) XtremeCT system (Scanco Medical) and leg positioning for tibia scan. ......................... 30	Figure 1.15. Illustration of trabecular (top image, green) and cortical (bottom image, grey) regions from a segmented high-resolution peripheral quantitative computed tomography scan. .................................................................................................................................. 32	xix  Figure 1.16 Illustration of stress-strain curve of destructive loading of cadaveric distal radii to determine linear and elastic failure regions. P = platen force. Reprinted from MacNeil et al.,[118] with permission from Elsevier. ............................................................................. 33	Figure 1.17. Total body bone mineral content (BMC TB) accrual velocity and ages at peak BMC and peak height velocity (PHV) for girls (dotted line) and boys (solid line) aligned on chronological age. The lag period between age at PHV and peak BMC is approximately 7-9 months. Reprinted from Bailey et al.,[133] with permission from John Wiley and Sons............................................................................................................................................ 36	Figure 1.18. Illustrations of sex differences in high-resolution peripheral quantitative computed tomography (HR-pQCT) parameters at the distal radius by pubertal group based on the method of Tanner staging: A) cortical density (Ct.BMD), B) cortical porosity (Ct.Po), C) cortical area (Ct.Ar) and D) failure load. a, p < 0.001; b, p < 0.01; c, p < 0.05: significant difference between girls and boys within the same puberty group. d, p < 0.001; e, p < 0.01; significant difference between puberty group and the PRE group within sex. Reprinted from Nishiyama et al.,[4] with permission from John Wiley and Sons. ........... 40	Figure 1.19. Illustration of bone growth over 20 months at the tibia midshaft in early-, peri- and post-pubertal boys and girls using peripheral quantitative computed tomography (pQCT). Numbers show the mean increase (%) in cortical and marrow cavity areas. Adapted from Kontulainen et al.,[153] and reprinted from Daly et al.,[159] with permission from Karger. 42	Figure 1.20. Illustration of sex differences in trabecular microarchitecture at the distal radius and tibia using high-resolution peripheral quantitative computed tomography (pQCT) ........ 46	Figure 1.21. Illustration of peaks for sex steroids, height and BMC velocity, growth hormone and IGF-1 amplitude in relation to age and pubertal stage in girls. Reprinted from MacKelvie et al.,[144] with permission from BMJ Publishing Group Ltd. ........................................... 48	Figure 1.22. Illustration of tissue velocity curves for muscle mass, A) cross-sectional area (CSA) and B) section modulus (Z) at the femoral shaft aligned by maturational age (years from age at peak height velocity). The solid vertical line represents the maturational age when peak tissue velocities occurred. *Indicates significant difference between age of peak muscle velocity and peak CSA velocity. **Indicates a significant difference between age of peak muscle velocity and peak Z velocity. Reprinted from Jackowski et al.,[207] with permission from Elsevier. ................................................................................................. 55	xx  Figure 1.23. Average side-to-side differences in humeral midshaft total bone cross-sectional area (CSA), cortical CSA, cortical bone mineral density (BMD) and bone strength index (BSI) between the playing and nonplaying arm in female racquet sport athletes as measured with peripheral quantitative computed tomography (pQCT). The solid line represents the playing arm (or dominant arm in controls) and the dotted line represents the nonplaying arm. Adapted from Macdonald et al.,[14] with permission from Future Medicine, Ltd. ... 68	Figure 1.24. Illustration of a) bone geometry (total bone area) and b) estimated bone strength (polar strength-strain index, SSIp) at the proximal radius (66% site) measured with peripheral quantitative computed tomography (pQCT) in pre-pubertal girls. Non-gymnasts (Non-Gym), low-training volume gymnasts (Low-Gym) and high-training volume gymnasts (High-Gym). *Indicates significantly different from Non-Gym. Bars represent 95% confidence intervals. Adapted from Burt et al.,[103] with permission from Springer. ............................................................................................................................ 70	Figure 1.25. Illustration of growth curves for section modulus (Z) by hip structural analysis (HSA) of the femoral neck region in the longitudinal subset comparing 17 active girls or boys with 17 inactive girls or boys in relation to biological age, years from age at peak height velocity (APHV). Reprinted from Forwood et al.,[208] with permission from Elsevier. ............................................................................................................................ 73	Figure 2.1. Overview of the University of British Columbia Healthy Bones Study III (HBSIII). 84	Figure 2.2. A) the anatomical reference line defining the distal aspect of the distal cartilage of the tibia and B) a peripheral quantitative computed tomography scan of the tibial midshaft. Bone is indicated in white, muscle in red/purple and subcutaneous fat in blue. .............. 94	Figure 2.3. Set-up for high-resolution peripheral quantitative computed tomography (HR-pQCT) radius (A,C) and tibia (B,D) scans. ................................................................................... 96	Figure 2.4. High-resolution peripheral quantitative computed tomography at the distal tibia. A) scout view image illustrating 8% scan site; B) scout view illustrating position of tibial growth plate. Reprinted from Burrows et al.,[110] with permission from Springer. .......... 97	Figure 2.5. High-resolution peripheral quantitative computed tomography at the distal radius. A) scout view image illustrating 7% scan site; B) scout view illustrating position of ulnar and radial growth plates; C) representative three-dimensional image showing cortical and xxi  trabecular compartments. Reprinted from Burrows et al.,[111] with permission from Elsevier. ............................................................................................................................ 98	Figure 2.6. Distal radius scans illustrating motion artifact grading, ranging from 1 (no motion) on the left to 5 (large discontinuities) on the right. Reprinted from Pauchard et al.,[124] with permission from Elsevier. ................................................................................................. 98	Figure 3.1. Scatterplots of sedentary time (as a % of wear time) regression residuals and bone architecture, BMD and strength regression residuals. Boys are represented by black squares and solid lines; girls are represented by open circles and dashed lines. (A) trabecular bone volume fraction (BV/TV), (B) total bone mineral density (Tt.BMD, mg HA/cm3), (C) total area (Tt.Ar, mm2), (D) failure load (F.Load, N). ............................. 115	Figure 3.2. Contribution of muscle cross-sectional area (MCSA), tibia length, maturity, ethnicity, dietary calcium and accelerometry-derived moderate-to-vigorous physical activity (MVPA) to the prediction of bone architecture, BMD and bone strength in Model 4 in A) boys and B) girls (n = 206). For example, the solid black bar represents the additional variance in bone outcomes explained by maturity when MCSA, tibia length, ethnicity, dietary calcium and MVPA are held constant. F.Load = failure load; BV/TV = trabecular bone volume fraction; Tb.N = trabecular number; Tb.Th = trabecular thickness; Ct.Po = cortical porosity; Ct.Th = cortical thickness; Ct.BMD = cortical bone mineral density; Tt.BMD = total bone mineral density; Tt.Ar = total area. .............................................. 116	Figure 4.1. Number of participants recruited and the number of valid peripheral quantitative computed tomography (pQCT) follow-up scans for boys and girls ............................... 124	Figure 4.2. Individual growth curves (thin, light gray lines), individual growth curves of five randomly selected girls and boys (thin, black lines), a lowess-smoothing curve (thick, dark gray dashed line) and the polynomial mixed model curves (thick, black line) of total area (Tt.Ar), cortical area (Ct.Ar), ratio of cortical to total area (Ct.Ar/Tt.Ar), medullary are (Me.Ar), cortical bone mineral density (Ct.BMD) and polar strength-strain index (SSIp), plotted against maturity offset. The vertical line indicates maturity offset (years from age at peak height velocity) of 0. ........................................................................... 132	Figure 4.3. A schematic representation of differences in total area (Tt.Ar), cortical area (Ct.Ar) and medullary area (Me.Ar) in boys and girls in relation to maturity offset (years from age at peak height velocity). I present maturity offset at -1, 0, 1 and 5 years. Significant xxii  differences between girls and boys are shown for polar strength-strain index (SSIp), where boys’ values exceed girls’ at all time points, and Ct.BMD, where girls’ values exceed boys’ at all time points. (Diagram not to exact scale). ....................................... 133	Figure 4.4. Curves of predicted total area (Tt.Ar), cortical area (Ct.Ar), ratio of cortical to total area (Ct.Ar/Tt.Ar), medullary area (Me.Ar), cortical bone mineral density (Ct.BMD) and polar strength-strain index (SSIp), plotted against maturity offset for boys (solid lines) and girls (dashed lines), Asian (black line), white (blue line) and other (grey line) participants. The vertical line indicates maturity offset (years from age at peak height velocity) of 0. .................................................................................................................. 134	Figure 5.1. Representative high-resolution peripheral quantitative computed tomography images at the distal tibia from a single participant across 4 years acquired at 11- (far left), 12-, 13- and 14- (far right) years of age. Images not to scale. ............................................... 144	Figure 5.2. Distal tibia individual growth curves for boys (thin, blue lines) and girls (thin, grey lines) and the polynomial mixed model growth curves for boys (thick, blue line) and girls (thick, black line) for trabecular bone volume fraction (BV/TV), separation (Tb.Sp), thickness (Tb.Th) and number (Tb.N), cortical BMD (Ct.BMD), area (Ct.Ar), thickness (Ct.Th) and porosity (Ct.Po), and total BMD (Tt.BMD), area (Tt.Ar), failure load (F.Load), and ultimate stress (U.Stress). The vertical line indicates maturity offset (years from age at peak height velocity) of 0. ........................................................................... 155	Figure 5.3. Distal radius individual growth curves for boys (thin, blue lines) and girls (thin, grey lines) and the polynomial mixed model growth curves for boys (thick, blue line) and girls (thick, black line) for trabecular bone volume fraction (BV/TV), separation (Tb.Sp), thickness (Tb.Th) and number (Tb.N), cortical BMD (Ct.BMD), area (Ct.Ar), thickness (Ct.Th) and porosity (Ct.Po), and total BMD (Tt.BMD), area (Tt.Ar), failure load (F.Load), and ultimate stress (U.Stress). The vertical line indicates maturity offset (years from age at peak height velocity) of 0. ........................................................................... 156	Figure 5.4. Sex differences in distal tibia trabecular bone volume fraction (BV/TV), separation (Tb.Sp), thickness (Tb.Th) and number (Tb.N), cortical BMD (Ct.BMD), area (Ct.Ar), thickness (Ct.Th) and porosity (Ct.Po), and total BMD (Tt.BMD), area (Tt.Ar), failure load (F.Load), and ultimate stress (U.Stress) across maturity. The solid black line represents the mean predicted sex difference (boys - girls) accompanied by a shaded 95% xxiii  confidence interval, correcting for multiple comparisons using a Bonferroni adjustment. Estimates above 0 indicate significantly greater values in boys, while estimates below zero indicate significantly greater values in girls. Confidence intervals that cross 0 indicate non-significant sex differences. The vertical line indicates maturity offset (years from age at peak height velocity) of 0. ........................................................................... 159	Figure 5.5. Sex differences in distal radius trabecular bone volume fraction (BV/TV), separation (Tb.Sp), thickness (Tb.Th) and number (Tb.N), cortical BMD (Ct.BMD), area (Ct.Ar), thickness (Ct.Th) and porosity (Ct.Po), and total BMD (Tt.BMD), area (Tt.Ar), failure load (F.Load), and ultimate stress (U.Stress) across maturity. The solid black line represents the mean predicted sex difference (boys - girls) accompanied by a shaded 95% confidence interval, correcting for multiple comparisons using a Bonferroni adjustment. Estimates above 0 indicate significantly greater values in boys, while estimates below zero indicate significantly greater values in girls. Confidence intervals that cross 0 indicate non significant sex differences. The vertical line indicates maturity offset (years from age at peak height velocity) of 0. ........................................................................... 160	Figure 5.6. Load to strength ratio at the distal radius. (A) displays individual data and predicted growth curves for boys (thin black lines and thick black line) and girls (thin grey lines and thick blue line). (B) displays predicted sex differences (boys-girls) across maturity with 95% confidence intervals, correcting for multiple comparisons using a Bonferroni adjustment. Estimates above 0 indicate significantly greater values in boys, while estimates below zero indicate significantly greater values in girls. Confidence intervals that cross 0 indicate non significant sex differences. ...................................................... 161	Figure 6.1. Participant inclusion diagram. .................................................................................. 174	Figure 6.2. Distal tibia individual growth curves (thin, light grey lines) and estimated growth curves from the polynomial mixed model for participants in the upper (~60 min/day; red solid line) and lower quartile of MVPA (~< 30 min/day, black dashed line), and the upper quartile (~11 h/day; blue dashed line) and lower quartile of sedentary time (~< 9 h/day; red solid line) for trabecular bone volume fraction (BV/TV), and thickness (Tb.Th), cortical BMD (Ct.BMD), thickness (Ct.Th) and porosity (Ct.Po), total area (Tt.Ar), and failure load (F.Load). The vertical line indicates maturity offset (years from age at peak height velocity) of 0. Mixed model growth curves are adjusted for maturity, sex, xxiv  ethnicity, lower limb muscle power, limb length and calcium. Growth curves for sedentary models are additionally adjusted for MVPA. ................................................. 188	Figure 6.3. Distal radius individual growth curves (thin, light gray lines) and estimated growth curves from the polynomial mixed model for participants in the upper (~60 min/day; black solid line) and lower quartile of MVPA (~<30 min/day, black dashed line), and the upper quartile (~11 h/day; red dashed line) and lower quartile of sedentary time (~<9 h/day; red solid line) for trabecular bone volume fraction (BV/TV), and thickness (Tb.Th), cortical BMD (Ct.BMD), thickness (Ct.Th) and porosity (Ct.Po), and total area (Tt.Ar), failure load (F.Load) and load-to-strength ratio. The vertical line indicates maturity offset (years from age at peak height velocity) of 0. Mixed model growth curves are adjusted for maturity, sex, ethnicity, lower limb muscle power, limb length and calcium. Growth curves for sedentary models are additionally adjusted for MVPA. .... 189	Figure 6.4. Interaction of MVPA and maturity with bone parameters across growth at the distal tibia and radius. The solid black line represents the coefficient of MVPA accompanied by a shaded 95% confidence interval, correcting for multiple comparisons using a Bonferroni adjustment. Estimates above 0 indicate significant positive relationship with MVPA, while estimates below 0 indicate significant negative relationship with MVPA. Confidence intervals that cross 0 indicate non-significant relationship. The vertical line indicates maturity offset (years from age at peak height velocity) of 0. Bone volume fraction (BV/TV), failure load (F.Load), total area (Tt.Ar), and cortical porosity (Ct.Po).......................................................................................................................................... 190	Figure 6.5. Interaction of sedentary time and maturity with bone parameters across growth at the distal tibia and radius. The solid black line represents the coefficient of sedentary time accompanied by a shaded 95% confidence interval, correcting for multiple comparisons using a Bonferroni adjustment. Estimates above 0 indicate significant positive relationship with sedentary time, while estimates below 0 indicate significant negative relationship with sedentary time. The vertical line indicates maturity offset (years from age at peak height velocity) of 0. Cortical thickness (Ct.Th), trabecular thickness, (Tb.Th), total area (Tt.Ar), cortical porosity (Ct.Po), and cortical BMD (Ct.BMD). ..... 191	xxv  Figure E.1. Illustration of motion artifact from pQCT tibia scans. Scans with streaks in the cortical shell (far right image) are excluded from analysis. Reprinted from Chan et al.,[1] with permission from Elsevier. ....................................................................................... 305	 xxvi  List of Abbreviations  ABBREVIATION TERMS 2D/3D Two-/Three-dimensional aBMD Areal bone mineral density by dual energy X-ray absorptiometry AIC Akaike information criterion APHV Age at peak height velocity BC British Columbia BIC Bayesian information criterion BMC Bone mineral content BMD (volumetric) Bone mineral density BSI Bone strength index BV/TV (trabecular) Bone volume fraction CHMS Canadian Health Measures Survey CI Confidence interval CIHR Canadian Institutes of Health Research cpm Counts per minute CSA Cross-sectional area CSMI Cross-sectional moment of inertia Ct.Ar Cortical bone area Ct.BMD Cortical bone mineral density Ct.Po Cortical porosity Ct.Th Cortical thickness DXA Dual energy X-ray absorptiometry FEA Finite element analysis FFQ Food frequency questionnaire F.Load Failure load GH Growth hormone HBSIII Healthy Bones III Study HHQ Health history questionnaire HR-pQCT High-resolution peripheral quantitative computed tomography xxvii  ABBREVIATION TERMS IBDS Iowa Bone Development Study ICC Intraclass correlation coefficient IGF-1 Insulin-like growth factor-1 IU International units LL Log likelihood LSC Least significant change µSv microSieverts MCSA Muscle cross-sectional area Me.Ar Medullary area MES Minimal effective strain MESm Minimal effective strain for modeling MESr Minimal effective strain for remodeling METs Metabolic equivalents MRI Magnetic resonance imaging MVPA Moderate-to-vigorous physical activity NHANES National Health and Nutrition Examination Survey PA Physical activity PAQ-A Physical Activity Questionnaire for Adolescents PAQ-C Physical Activity Questionnaire for Children PBMAS Pediatric Bone Mineral Accrual Study PHV Peak height velocity pQCT Peripheral quantitative computed tomography PVE Partial volume effect RCT Randomized controlled trial ROI Region of interest SSIp Polar strength-strain index Tb.Ar Trabecular area Tb.BMD Trabecular bone mineral density Tb.N Trabecular number Tb.Th Trabecular thickness xxviii  ABBREVIATION TERMS Tt.Ar Total bone area Tt.BMD Total bone mineral density UBC University of British Columbia U.Stress Ultimate stress Z Section modulus  xxix  Acknowledgements  I would like to express my sincerest thanks to my supervisors, Drs. Heather McKay and Heather Macdonald. I am fortunate to have spent the past five years learning from you – it has been a truly fulfilling journey. Thank you both for challenging my thinking, encouraging me to ask questions and for pushing me out of my comfort zone. To Heather McKay, I admire your dedication and enthusiasm for research, your leadership, vision and writing finesse. I am grateful for your mentorship. To Heather Macdonald, I cannot speak highly enough about my experience under your tutelage; your encouragement, guidance, attention to detail and spot-on feedback have been instrumental to my growth as a researcher. You set high expectations, but provided me the independence and flexibility to develop in my own time and to cultivate my style and voice. I also thank you for encouraging me to pursue research abroad, conference and teaching opportunities – these experiences and the connections I made as a result are invaluable. I am equally appreciative of my committee members. To Dr. Lindsay Nettlefold for your Stata and accelerometry expertise, morning chats, feedback, support and friendship. To Louise Mâsse for discussions about study design and analyses. I could not have asked for a more supportive and talented committee. Your expertise, editing and thought-provoking discussions were critical to this dissertation. I feel privileged to have been a part of the UBC Healthy Bones III Study team. I am indebted to the numerous researchers, staff, investigators and participants of the UBC Healthy Bones III Study. This dissertation would not have been possible without your years of involvement, hard work and dedication. In particular, I would like to thank Douglas Race, study coordinator extraordinaire, for your knowledge on everything Healthy Bones III Study-related and for your assistance with accelerometry processing and database management. To Sarah Moore for recruiting the final Healthy Bones III Study cohort, your work on the maturity offset piece and support throughout my studies. To Danmei Liu for your pQCT and HR-pQCT expertise, availability and guidance. To Mikko Määttä for your assistance with auto-segmentation and finite element analyses. Thank you also to the Canadian Institutes of Heath Research for your generous support of the Healthy Bones III Study.  I gratefully acknowledge the funding sources that made my Ph.D. possible – the Canadian Institutes of Health Research, Australia Endeavor Fellowship Awards and UBC 4-year xxx  Fellowship. To Prof. Jo Salmon and her team at Deakin University in Melbourne, thank you for inviting me into your lab and for being gracious hosts during my research fellowship. To Penny Brasher for providing statistical guidance during the early part of my studies. To Lesa Hoffman for making complex statistical modelling accessible to non-statisticians. To my CHHM family, thank you for a supportive and fun work environment. In particular, thank you to Christa Hoy (and Lindsay) for lunch time conversations where I often thought out loud and sought your opinions. To Anna Chudyk, Amanda Frazer and Christine Voss, my office mates over the years, for your enthusiasm and friendship. Lastly, I would like to thank my wonderful family and friends for your encouragement throughout my studies. To my parents for your love, unwavering support and optimism over years and years of post-secondary studies. Your timely visits over the past few years were helpful beyond measure. To my brother and besties for your humor and for always being just a phone call away. To my Squamish and Vancouver family for providing community, laughter and balance. To Doug, my partner in life and adventure. This journey would not have been nearly as enjoyable without you by my side. You went above and beyond, as super-dad and super-partner, giving me time and support to finish this dissertation, for which I am so appreciative. To Bender, our goofy sheepdog, for keeping me company during days and nights working at home; many of my ‘aha’ moments happened during our walks. Finally, to Aubree, our dear daughter, you bring us joy in ways words cannot describe. Thank you for providing perspective on what truly matters in life. You helped me find balance between perfectionism and efficiency by making me appreciate the value of time. I also thank you for sleeping well – keep it up kiddo! I look forward to our next adventure…1  Chapter 1: Introduction, Literature Review, Rationale, Objectives & Hypotheses  1.1 Introduction  Bone strength2 is irrefutably the most important parameter of skeletal health and is underpinned by bone’s material properties, quantity, dimensions (size and material distribution) and microarchitecture.[1,2] This tenet has guided a paradigm shift away from assessing only two-dimensional (2D) measures of bone mass (measured with dual energy X-ray absorptiometry, DXA) to three-dimensional (3D) measures of bone geometry and microarchitecture. As imaging devices such as peripheral quantitative computed tomography (pQCT) and high-resolution pQCT (HR-pQCT) become more commonly used, we acquire a better understanding of the hierarchical structure of bone. As important, we also gain insight as to how complex bone structures adapt in response to physical activity (PA)3. Despite recent advances in bone imaging, the intricacies of how bone is gained in childhood and lost in later life are still not completely understood. For example, it is unclear whether bone is gained or resorbed at the endocortical surface of the diaphysis during adolescent growth. We also do not fully understand how maturity-related adaptations differ between boys and girls. Adaptations specific to bone microarchitecture during growth are also unclear as few studies have used HR-pQCT to examine maturity- and sex-related differences during childhood and adolescence. Of these, all were cross-sectional or had a short follow-up period[4-6] and scanned different regions of the distal radius and/or tibia. Gaps in knowledge are due in part to inadequate methods used to control for the substantial variation in maturational status among adolescents and the reliance on cross-sectional compared with prospective data. A thorough understanding of bone strength accrual is crucial to appreciate the influence of PA on bone adaptation during growth.                                                 2 I use bone strength to refer to estimated bone strength in studies that used non-invasive imaging. 3 PA defined as any bodily movements expending energy.[3] 2  The ability of bone to adapt to mechanical loads resulting from weight-bearing PA was first described more than a century ago.[7,8] In recent decades, we amassed a substantial body of evidence to support an integral role for PA and weight-bearing PA, specifically, for developing and maintaining a healthy skeleton. In particular, the critical period of adolescence, when more than one quarter of adult bone mass is accrued,[9] and childhood represent windows during which the skeletal benefits of weight-bearing PA may be enhanced.[10-15] In turn, optimal PA during the growing years may prevent adult bone health problems, such as osteoporosis and fragility fractures.  Although we do not know the precise PA prescription (e.g., frequency, intensity, duration, type) for optimal bone strength accrual, evidence from animal studies, school-based interventions and observational studies suggest that “a little goes a long way”. Specifically, short bouts of high-impact PA implemented over relatively short timeframes may be sufficient to enhance bone mass and strength accrual during adolescence.[15] Conversely, we know less about bone microarchitecture adaptations to weight-bearing PA, although they also influence bone strength during childhood and adolescent growth and development.  Despite numerous health benefits associated with PA, today’s youth spend roughly 60% of their waking hours being sedentary (defined as ≤ 1.5 metabolic equivalents (METs).[16] A focus on the consequences of ‘not loading’ a healthy growing skeleton is relatively new. No studies to date have investigated the relationship between sedentary time and bone strength (estimated using pQCT or HR-pQCT) in children and youth. Thus, it is unclear how the potentially deleterious impact of sedentary time interacts with the osteogenic effect of PA in healthy, ambulatory children and adolescents.  Therefore, the primary aim of my thesis is to determine the influence of PA and sedentary time on bone strength accrual and its determinants during childhood and adolescence. I extend previous studies that focused primarily on accrual of bone mass and areal bone mineral density (aBMD; by DXA) as outcomes. My secondary aim is to examine the maturity- and sex-related adaptations in bone strength accrual and its determinants during adolescent growth. To achieve these aims, I employ two novel bone imaging techniques, pQCT and HR-pQCT, and measure PA and sedentary time objectively using accelerometry. In addition, I use general linear mixed models to account for the longitudinal nature of my data, which permits investigation of inter- and intra-individual variation across time. My thesis is divided into four research chapters 3  (Chapters 3-6). In Chapter 3, I describe cross-sectional associations between sedentary time and bone strength and its determinants at the distal tibia using HR-pQCT. In Chapter 4, I examine longitudinal maturity- and sex-related adaptations in bone strength and geometry from childhood to early adulthood at the tibia midshaft using pQCT. In Chapter 5, I investigate longitudinal maturity- and sex-related adaptations in bone strength and its determinants at the distal tibia and radius from childhood to early adulthood using HR-pQCT. Finally, in Chapter 6, I examine the prospective associations between PA, sedentary time and bone strength at the distal tibia and radius during adolescence using HR-pQCT.  1.2 Literature review  In this section, and in six parts, I provide an overview of pertinent literature that informs this thesis: basic bone biology, bone biomechanics, bone imaging in children and adolescents, bone development during childhood and adolescence and factors that influence bone development, with a specific focus on PA and sedentary time. Bone is a complex and dynamic tissue that serves to provide structural support and withstand loads imposed on it by external and internal forces.[17] In addition, bones serve as levers for locomotion, a reservoir for calcium, protector of internal organs, a site for hematopoiesis (formation of blood cells) and attachment sites for muscles, ligaments and tendons.[17] Bones must serve all of these functions while remaining lightweight for locomotion and adapting to substantial changes in morphology imposed by growth.[18]  1.2.1 Bone biology and bone growth  In this section, I briefly describe basic bone biology related to human long bones, the focus of my thesis. I briefly review mechanisms that influence bone development and maintenance.    4  1.2.1.1 Bone tissue: composition and organization  Bone is a composite material of minerals, collagen, water, non-collagenous proteins and lipids.[18] The mineral component makes up approximately 70% of bone by weight and consists of calcium and phosphate arranged in crystals of hydroxyapatite (Ca10(PO4)6(OH)2). Hydroxyapatite provides stiffness, mechanical resistance and a source of minerals (e.g., calcium and phosphate).[18] The non-mineralized component makes up 20-25% of bone by weight and is primarily type 1 collagen and non-collagenous proteins (98%). Collagen is a connective protein that binds hydroxyapatite, providing elasticity and the ability to resist tension.[18] The remainder of bone tissue is comprised of bone cells (osteoblasts, osteoclasts and osteocytes). Together, the mineral and collagen matrix produce a connective tissue with high stiffness and strength[18] that allows bone to withstand stresses in bending, compression and torsion. There are two types of bone tissue: woven and lamellar bone.[19] Woven, or immature, bone is characterized by an irregular pattern of collagen fibre orientation. Woven bone comprises all bone at birth and can be found at ligament and tendon insertions in healthy adult skeletons.[17] Woven bone can be temporarily found during fracture healing as it is formed faster than lamellar bone.[20] Most woven bone is resorbed and replaced by lamellar bone by about four years of age.[21] In contrast, lamellar bone, or mature bone, is characterized by a consistent arrangement of collagen fibres along lines of force. Lamellar bone is the building block of both cortical and trabecular bone, such that the structural subunits, lamellae, are oriented parallel to trabeculae in trabecular bone (also termed cancellous bone) and arranged in osteons in cortical bone (Figure 1.1).[17] Osteons are the major structural units of cortical bone and are cylindrical arrangements of cortical bone (lamellae) around a Haversian canal, channeling a blood vessel for nutrition. Osteons are usually aligned along the long axis of bones and are connected by Volkmann’s canals running at right angles.[19]  5  Figure 1.1 Diagram of structural elements of long bones, illustrating cortical bone, comprised of osteons surrounding blood vessels, and trabecular bone. Reprinted from Martin et al.,[22] with permission from Springer New York.   Although cortical and trabecular bone are made of the same material, their respective bone matrices are arranged differently, and thus serve unique structural and functional roles.[21] Cortical bone provides structure, protects organs and has a porosity of 5-30% of bone volume. Calcium comprises upwards of 80-90% of cortical bone volume.[21,23] Cortical bone forms the diaphyses of long bones and is also located in the thin shells of metaphyses (Figure 1.2). Trabecular bone, on the other hand, is more porous than cortical bone (30-90%) and calcium comprises only 15-25% of bone volume.[21,23] Trabecular bone is found in the epiphyses and metaphyses of long bones and in vertebral bodies. Trabecular bone’s 3D lattice geometry enables transfer of loads through bending moments.[21,24] Trabecular bone is arranged so that bone marrow, blood vessels and connective tissue are in contact with bone. Given its large surface area and proximity to bone marrow, trabecular bone is suited for metabolic activities associated with bone turnover. Trabecular bone is typically ‘younger’ than cortical bone as it is more metabolically active and remodeled frequently.[25] The human skeleton is comprised of 6  predominantly cortical bone (~80%); however, the relative contribution of cortical and trabecular bone to total bone volume varies between and within skeletal sites. For example, the femoral head is 50% cortical bone, while the radial diaphysis is 95% cortical bone.[26] Figure 1.2 Anterior view of the right human femur with basic anatomy, including diaphysis, epiphysis and metaphysis. Modified from the online edition of the 20th US edition of Gray’s Anatomy of the Human Body, and originally published in 1918 and reprinted from Kontulainan et al.,[27] with permission from Karger.  Three cell types found within cortical and trabecular bone regulate bone turnover: osteoblasts, osteoclasts and osteocytes. Osteoblasts, bone-forming cells, lay down extracellular matrix and regulate bone mineralization. Osteoblasts secrete an initial collagen matrix (osteoid, unmineralized protein), which forms the basic framework of bone tissue, and subsequently mineralize the collagen.[18] Osteoclasts, bone-resorbing cells, resorb fully mineralized bone. Osteocytes, the most abundant cell in bone, are derived from former osteoblasts that become 7  embedded within the bone matrix. Osteocytes communicate with other bone cells, detect mechanical loading and coordinate modeling and remodeling activity (described in further detail in Section 1.2.1.2).[18,28]  1.2.1.2 Bone modeling and remodeling  Bone modeling and remodeling allow bone to adapt its size, shape and distribution of microarchitecture throughout the lifespan.[29] These processes optimize bone strength and minimize bone mass by adding bone where it is needed and removing bone from where it is not.[30] Bone modeling involves independent actions of osteoblasts and osteoclasts in response to physiological or mechanical influences, resulting in gradual adaptation of the skeleton to biomechanical forces.[26] Bone modeling increases bone length and size; thus, it predominates during growth and is reduced at skeletal maturity. Modeling enlarges the diaphysis during growth, such that osteoblasts deposit bone onto periosteal surfaces and osteoclasts remove bone from endocortical surfaces.[17] As bone length increases, the wider metaphyses are modelled to match the thinner cross-section of the diaphysis; a process referred to as metaphyseal inwaisting. [31] These processes reshape long bones, preventing the cortex from becoming excessively heavy and thick and positioning the diaphyseal cortex farther from the centre of the bone.[27] In contrast to modeling, bone remodeling couples the action of osteoblasts and osteoclasts (the basic multicellular unit) to preserve bone strength and mineral homeostasis.[26] During remodeling, osteoclasts resorb pockets of old or damaged bone and osteoblasts subsequently fill the pockets with collagen matrix and mineralize the collagen to form new bone. Remodeling occurs throughout the lifespan in response to mechanical loading and to replace dead and damaged bone tissue.[17] Bone resorption takes 2 to 4 weeks, while formation takes 4 to 6 months to complete; thus, one full remodeling cycle takes approximately 5-7 months.[26,32] Modeling and remodeling are integral for strengthening the growing skeleton and I discuss both in greater detail in section 1.2.2.2.2.  8  1.2.1.3 Long bone geometry  The human skeleton contains many different types of bones: long, short, flat, irregular and sesamoid. In this chapter, I focus primarily on long bones, which, during growth, are comprised of the diaphysis, epiphysis and metaphysis (Figure 1.2). The diaphysis is a cylindrical shaft in the middle of the long bone; the outer portion contains cortical bone that encloses the medullary cavity. Cortical bone has two surfaces: the periosteum or periosteal surface, which is the outer surface facing the soft tissue, and endosteum or endocortical surface, which is the inner surface adjacent to bone marrow. Both are active sites of bone modeling and remodeling, lined with osteoblasts and osteocytes. The periosteum contributes to appositional bone growth through modeling-related increases in bone diameter during development.[21] Cortical bone diameter is thinner towards the metaphyses and epiphyses, where the medullary cavity is replaced by trabecular bone. The metaphysis is a transitory region located between the diaphysis and epiphysis, and is comprised of trabecular and cortical bone. A layer of cartilage known as the growth plate separates the metaphysis and epiphysis (Figure 1.2).[21] The flared shape of long bone ends distributes joint forces and reduce stress transmitted from trabecular bone in the metaphysis to cortical bone in the diaphysis.[33]  1.2.1.4 Longitudinal growth of long bones  Skeletal growth (in length and width) occurs between birth and maturity in preparation for a lifetime of loading, and is controlled by systemic, local and mechanical factors.[34] Skeletal growth in length occurs at the growth plates in epiphyseal and metaphyseal regions, where cartilage proliferates.[26] Endochondral ossification is the complex process whereby cartilaginous tissue is replaced by bone (Figure 1.3).[35] The epiphyseal end of the growth plate is predominantly cartilaginous tissue composed of chondrocytes, cartilage producing cells. Approximately 80% of cartilage is resorbed during growth. The remaining cartilage provides a scaffold for osteoblasts to deposit bone matrix, forming primary trabeculae, a mixture of cartilage and bone tissue.[34] During growth, metaphyseal trabeculae thicken and are eventually integrated into the metaphyseal cortex (a process termed trabecular coalescence) and later into the diaphyseal cortex.[34] Metaphyseal trabeculae, located at the centre of long bones, are 9  completely resorbed leaving the diaphysis empty of trabeculae. As a result, the age of bone tissue at a certain distance from the growth plate is directly related to the rate of longitudinal growth. Bone tissue at a given distance from the growth plate will be relatively younger when the rate of longitudinal growth increases, such as during the pubertal growth spurt.[36] Of note, the contribution of distal and proximal growth plate activity to overall long bone growth varies between bone sites during different periods of growth.[37,38] For example, 57% of longitudinal bone growth occurs at the proximal metaphysis of the tibia between 10 and 15 years of age and 43% occurs at the distal metaphysis.[39] After growth plate closure, bone age parallels chronological age.[36]   Figure 1.3. The schematic on the left is of a long bone during embryonic development, including the growth plate. The image on the right shows proliferation of chondrocytes at the growth plate. Reprinted from Wallis,[40] with permission from Elsevier.  10  1.2.2 Bone biomechanics  In this section, I discuss bone’s mechanical properties and mechanisms underlying bone’s adaptation to mechanical loading.  1.2.2.1 Material and mechanical properties of bone  The hierarchical geometry of bone ranges from smaller than the nanoscale to greater than the millimeter scale, with all levels contributing to bone’s mechanical behaviour and function (Figure 1.4).[25] Bone macroarchitecture includes cortical and trabecular bone properties greater than the millimeter scale (i.e., cross-sectional area).[41] Bone microarchitecture refers to individual trabeculae, Haversian systems and osteons, ranging from 10 to 500 micrometers (e.g., cortical thickness and porosity, trabecular number and thickness).[25] Organization of bone on an even smaller scale includes the sub-microarchitecture (1-10 µm; e.g., individual lamella), nanostructure (a few hundred nanometers to 1 µm; e.g., collagen fibres) and sub-nanostructure (less than a few hundred nanometers; e.g., bone mineral crystals).[25] In this thesis, I focus on bone’s organization at the macro- and microarchitecture levels as these are within the measurement capacity of the imaging tools I used for my research.  Figure 1.4 The hierarchical organization of bone, reprinted from Rho et al.,[25] with permission from Elsevier. 11   Bone’s material behaviour reflects its intrinsic material properties and is independent of geometry (size).[42] Material properties include stress, strain, Young’s modulus, ultimate stress and toughness. Structural behaviour, on the other hand, reflects both material and geometric properties. Material properties of bone are typically determined by mechanical tests (destructive testing to determine force required to cause failure) on uniform specimens of intact bone (i.e., part of the bone machined using a saw), whereas structural behaviour is determined by mechanical tests of whole bone specimens (i.e., whole sections of intact bones) where bone geometry is preserved.[2] In this section, I provide a brief overview of bone’s material behaviour, although the concepts I discuss apply to many materials other than bone. Stress is a measure of the intensity of force applied to a material, and is defined as force applied per unit area (N/m2).[43] When normalized to cross-sectional area, direct comparison can be made between specimens of different size and loads of different magnitudes.[42] Strain, on the other hand, is a measure of material deformation and is calculated as relative change in bone dimensions (change in length / initial length; unit-less, but is often expressed as microstrain because it is small in magnitude in bone).[44] Stresses and strains are categorized as normal or shear. Normal stresses act perpendicular to a given plane and the strains either pull apart and elongate the bone (tensile) or compact and shorten the bone (compression). Shear stresses and strains, on the other hand, act parallel to the plane and define the angular change during deformation.[42] Bones experience both normal and shear stresses and strains during normal function.[42] Bones are typically loaded in one of four ways (compression, tension, bending and torsion) or in combination (i.e., long bones are often loaded in bending, but also in compression and torsion).[18] As illustrated in Figure 1.5, the stress-strain curve reflects bone’s material behaviour and describes the amount of stress required to produce a unit of strain.[42] Material properties of stiffness, ultimate stress and Young’s modulus are derived from the stress-strain curve.   12  Figure 1.5. The top image is a stress-strain curve divided into elastic and plastic regions. The bottom image displays the measurement of strength from the stress-strain curve. X marks the stress and strains where failure occurs. Reprinted from Turner and Burr,[45] with permission from Elsevier.  Young’s modulus of elasticity is the amount of force necessary to deform a structure. It is calculated as the slope (stress divided by strain) of the linear segment of the stress-strain curve.[43] The stiffer the material (more force needed to cause deformation), the steeper the slope of the stress-strain curve. Ultimate stress is the maximum stress a bone can sustain without failing, while toughness is the amount of energy a bone can absorb prior to fracture (derived from area under the stress-strain curve).[45] In the elastic portion of the stress-strain curve, a section of the curve prior to a yield point, stress and strain are linearly related. Any strain in the elastic region is only temporary and bone will return to its original shape once a load is removed. The plastic region of the stress-strain curve begins beyond the yield point where the slope decreases. Strain along this section of the curve is permanent and occurs until a fracture point or point of failure is reached.[45] With these mechanical properties in mind, most bones respond by growing large and thick enough to stay within the elastic region of the stress-strain curve.[24] 13  Long bones of the appendicular skeleton primarily experience stresses and strains at the diaphysis in bending and torsion. Cortical bone experiences the highest loads and deformations.[46] In contrast, at metaphyseal sites where trabecular bone predominates, the greatest stress is experienced in compression and bending.[46]  1.2.2.1.1 Material properties of cortical bone  Material properties of cortical bone depend greatly on degree of matrix mineralization and porosity.[47,48] Cortical bone is anisotropic, such that its elastic properties and strength are dependent upon orientation of bone with respect to the applied load.[23] Cortical bone is stronger and stiffer longitudinally versus transversally and is stronger in compression than in tension.[23] It is a viscoelastic (time-dependent) material, such that its mechanical properties are dependent on strain duration and strain rate.[23] The contribution of cortical bone to whole bone strength is also site-specific. For example, cortical bone’s contribution to long bone strength is much greater at the cortical-rich diaphysis compared with the metaphyses, where a combination of cortical and trabecular bone contribute to whole bone strength.[23]  1.2.2.1.2 Material properties of trabecular bone  Trabecular bone is largely responsible for bone’s energy absorbing capacity.[49] As with cortical bone, trabecular bone is anisotropic (material properties are dependent upon the orientation); in contrast, trabecular bone is more porous and highly heterogeneous throughout the body, resulting in varied mechanical properties.[49] The heterogeneity of trabecular bone stems from differences in volume fraction, arrangement of individual trabeculae (i.e., microarchitecture) and tissue properties.[50] Trabecular bone strength and elastic modulus are site-specific, varying with the changing function of bone with respect to location and type of stress. For example, when tested in different directions, there are up to ten-fold differences in the elastic modulus of trabecular bone at the same anatomical site.[51] Mechanical properties of trabecular bone are influenced by changes in the thickness, number, separation and connectivity of trabeculae.[52]   14  1.2.2.1.3 Bone strength in bending and compression   Long bone diaphyses experience loading in tension, compression, bending or torsion; alone or in combination.[46] In theory, a hollow tube provides the greatest strength with least mass in response to torsional or bending loads.[43] Bending strength is proportional to the square of the material’s distance from the cross-section’s center of mass[53] and is influenced by bone’s cross-sectional moment of inertia (CSMI), which quantifies the distribution of material around the bending moment.[42] This principle is illustrated in Figure 1.6, where the bone with the greater outer circumference and larger hollow center is significantly stronger and stiffer than the bone with a smaller outer diameter and hollow centre. Thus, bone added to the periosteal surface contributes more to bending strength than that removed from the endocortical surface.[53] Therefore, the most efficient structure is one where mass is placed furthest from the neutral axis.[18] Figure 1.6. Scale drawings of three cylindrical cross-sections with different outer diameters, fixed length (L), but equal areal bone mineral density (BMD). Corresponding values of volumetric BMD (vBMD), bone mineral content (BMC) or cross-sectional area (bCSA), cross-sectional moment of inertia (CSMI) and section modulus. Reprinted from Beck,[54] with permission.   Metaphyseal sites are primarily loaded in compression, thus resistance to bending is not an appropriate index of strength.[55] In addition, strength indices derived from cortical geometry, 15  such as CSMI, require an accurate measure of the cortical compartment. This may prove challenging in children and at distal sites where the cortical shell is thin. Bone strength index (BSI) is a noninvasive estimate of bone strength in compression that incorporates both bone material properties and its distribution. BSI is the product of total cross-sectional area (CSA) and the square of total BMD. It predicts up to 85% of the variance in failure load at the distal tibia (4% site).[55] Thus, bone strength in compression can be increased by adding bone on the periosteal surface and with increased trabecular BMD. Compressive bone strength can also be estimated using high-resolution images and finite element analysis (FEA). I discuss FEA in greater detail in section 1.2.3.3.1.  1.2.2.2 Bone’s response to mechanical stimuli  Above, I presented that bone is a complex and dynamic tissue. Bone’s primary role is to provide structural support and withstand loads imposed by external and internal forces (e.g., gravitational and muscular forces).[24] Increased bone strength can be achieved in a number of ways – through increased BMD, changes in bone geometry and/or distribution of microarchitecture. The skeleton is continually exposed to a loading environment and bone is deposited and resorbed to achieve an optimum balance between bone strength and mass.[18] In this section, I consider bone’s response to mechanical stimuli in the context of mechanical problems, how bone perceives applied forces, the osteogenic response to applied loads and factors that influence bone’s response to mechanical stimuli.  1.2.2.2.1 Mechanotransduction   As early as the 19th century, Julius Wolff described how bone architecture adapts to mechanical loads applied to it, remodeling over time to better resist similar strains.[8] Bone responds to mechanical loading through mechanotransduction, a process whereby a biophysical force is converted into a cellular response.[56] Briefly, osteocytes sense mechanical strain and initiate a signaling cascade to effector cells (osteoblasts and osteoclasts).[56] The skeleton responds to mechanical strains through modeling and remodeling and adjusts bone mass and geometry to match the demands of the mechanical environment.[18,57]  16  First, in mechanocoupling (Figure 1.7), a mechanical force applied to bone produces fluid shear stresses along the cell membrane that are detected by osteocytes.[58] Deformation of bone creates pressure gradients within osteocyte canaliculi, triggering interstitial fluid flow and communication at gap junctions.[18] Second, biochemical coupling transduces fluid shear stress into a biochemical signal through various biomechanical pathways, including cyclooxygenase (COX) and nitric oxide synthase (NOS).[59] Third, the signal is transmitted from the sensor cell (osteocyte) to effector cell (osteoblasts or osteoclasts) and finally, the effector cell responds to the signals.[59] Once strain is transduced, osteogenic cells initiate one of four possible outcomes: 1) no response, 2) osteoblasts add new bone, 3) osteoclasts resorb bone, or 4) both osteoblasts and osteoclasts are recruited in coordination.[18] Figure 1.7. Illustration of mechanocoupling. Bending forces cause deformation of osteocytes and create pressure gradients that drive fluid through canaliculae, from regions of compression to tension. The fluid flow generates shear stress on cell membranes. Reprinted from Duncan et al,[56] with permission from Springer.     17  1.2.2.2.2  Mechanostat theory   Frost’s mechanostat theory proposes that bone’s mechanical competence is a function of mechanosensory negative feedback loops that sense load-induced strains and respond by adapting bone mass, geometry and strength to maintain bone strain at an optimal level.[60] Osteocytes sense strain and send out signals to initiate bone modeling and remodeling that increase bone strength.[60] Frost describes setpoints, minimum effective strain (MES) thresholds, whereby loads above and below such setpoints stimulate or attenuate bone mineral deposition or resorption (Figure 1.8).[60] When strains exceed the modeling threshold (MESm, ~2000 microstrain), bone modeling is activated, increasing bone strength. Within the remodeling threshold (MESr, ~50-200 microstrain), the amount of bone resorbed and accrued tends to be balanced. Below the remodeling threshold (MESr), (disuse or trivial loading zone), the theory contends that more bone is resorbed than accrued.[61,62] Finally, at the upper end of the spectrum, strains beyond the pathological MES theoretically cause accumulating fatigue damage.[63] Others proposed that non-mechanical factors such as nutrition and hormones alter the MES thresholds.[57] I discuss the influences of non-mechanical factors on bone adaptation in section 1.2.5. During growth, mechanical loads associated with increased bone length and muscle forces increase bone tissue strain above MESm. After skeletal maturity, peak bone tissue strains are reduced, and remodeling enables bone conservation (MESr). In aging adults, PA and muscle strength decreases. Consequently, mechanical loads imposed on bones diminish, and based on Frost’s theory, strains downshift below the remodeling threshold region into the disuse zone.[62]         18   Figure 1.8. Illustration of the mechanostat theory and influence of mechanical strain on bone modeling and remodeling. Theoretically, bone remodeling occurs in the upper limit of the trivial loading zone (or disuse zone) and in the physiological loading zone; bone modeling occurs in the overload zone; and microdamage repair occurs in the pathological overload zone. Based on Forwood and Turner[64] and reprinted from Bachrach et al,[65] with permission from Elsevier.   Frost’s mechanostat theory significantly advanced our understanding of bone’s response to mechanical loading; however, more recent work highlights several inaccuracies.[66] Namely, mechanostat theory cannot account for why bone resorption does not predominate at non-weight bearing sites due to disuse. Cellular accommodation theory attempts to reconcile inconsistencies in mechanostat theory using mathematical principles that assume: 1) bone cells adapt their set point in response to a change in loading environment and 2) set points vary from site to site based on the local strain environment.[66] Therefore, the set point will be higher in weight bearing bones than non-weight bearing bones. Animal studies support the cellular accommodation theory of mechanoadaptation, whereby bone’s response to loading resembled adaptation predicted by cellular accommodation theory, but not adaptation predicted by mechanostat theory.[67] 19  Specifically, bone adaptation in adult rats was proportional to the initial peak load magnitude and bone desensitized to loading after the initial weeks of loading.[67]  1.2.2.2.3 The functional model of bone development  Muscular contractions impose the largest voluntary loads on the body.[62] In contrast, body weight incurs relatively small static loads on bones, which are amplified by muscular contractions.[57] During longitudinal growth, increases in bone length and muscle forces result in greater bone deformation.[57] In their functional model of bone development, an extension of mechanostat theory, Rauch and Schoenau proposed that a negative feedback loop between tissue strain and bone strength is central to bone’s regulation (Figure 1.9). This model also proposes that physiologic loads from muscle forces trigger a cascade of events that allow bones to maintain functional structural integrity and strength.[57]  Figure 1.9. The functional model of bone development based on mechanostat theory. A feedback loop between bone deformation and bone strength is the central component of regulation of bone development and adaptation. 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 value. Factors shown in the bottom box modulate the regulatory system. Reprinted from Schoenau[68] and adapted from Rauch and Schoenau,[57] with permission from Nature Publishing Group.  During growth, bone adapts its strength in response to mechanical stimuli through several mechanisms: 1) periosteal apposition to increase bone CSA; 2) periosteal apposition in 20  conjunction with reduced endocortical resorption to increase cortical thickness; and/or 3) modified cortical and trabecular microarchitecture (i.e., increased trabecular thickness or number or decreased cortical porosity) to increase tissue density.[11,69] In the following section, I focus on the role of mechanical loading and resultant adaptations in bone strength, geometry and microarchitecture during growth.  1.2.2.2.4 Experimental evidence for bone adaptation to mechanical stimuli  Experimental evidence from animal models advanced our understanding of how bone adapts to mechanical loads.[70-76] Based on this evidence, Charles Turner proposed three fundamental ‘rules’ that predict bone structural adaptations to mechanical stimuli.[77] First, dynamic loading drives bone adaptation, such that the stimulus for bone adaptation increases with load magnitude or frequency. Second, only short bouts of mechanical loading are necessary to elicit an osteogenic response. There is a ceiling effect for bone tissue stimulation (loading frequency or duration), beyond which bone adaptation is subject to diminishing returns. Third, bone cells become accustomed to routine strain; structural change is driven by abnormal strains (‘strain distribution theory’[78]). These ‘rules’ provide insight into how different intensities and modalities of exercise predict human bone adaptation. We cannot directly translate results from animal studies to studies of children and youth; however, animal models shed light on adaptation in bone’s microarchitecture that underpin increased bone strength during growth. For example, pubertal rats (6 weeks old) subjected to 8 weeks of exercise (freefall jumps from 45 cm; strains were similar in magnitude to those observed in human athletes such as triple jumpers) had significantly greater trabecular bone volume fraction (BV/TV) and trabecular thickness (Tb.Th), but not trabecular number (Tb.N; all by micro-CT) at the proximal ulna compared with the control group (no freefall jumps).[79] In the same study, geometric adaptation to loading at the proximal ulna included thicker cortices (Ct.Th) as a result of enhanced endocortical contraction (by pQCT). Compared with controls, exercised rats had significantly greater cortical area (Ct.Ar), periosteal and endocortical circumference at the ulnar shaft and greater trabecular BMD (Tb.BMD) at the distal ulna. The ulnar shaft and distal site were not assessed using micro-CT. Collectively, adaptations in geometry and microarchitecture contributed to 36% greater mechanical strength (ultimate force 21  and energy to failure) in exercised rats compared with controls.[79] Similar bone microarchitecture adaptations were observed in young rats after 10 weeks of treadmill running.[80] As Figure 1.10 illustrates, exercised rats had greater BV/TV, Tb.Th and Tb.N at the distal femur compared with controls. At the femoral shaft, exercised rats had greater Ct.Ar, Ct.Th and maximum load, compared with controls. In line with the adaptations I discussed in section 1.2.2.1.3, these findings confirm that in response to loading, microarchitecture adaptations in trabecular bone predominate at metaphyseal sites whereas changes in cortical bone predominate at diaphyseal sites. Figure 1.10. Three-dimensional images from micro-computed tomography (micro-CT) of exercise-related adaptations in bone microarchitecture at the distal femoral diaphysis in rats. Sedentary controls in A and C, exercised rats in B and D; cortical compartment in the top images, trabecular compartment in the bottom images. Reprinted from Joo et al.,[80] with permission from Elsevier.  Animal studies also provide insight into the prolonged effects of exercise training on the skeleton. For example, rats subjected to axial loads on the right forearm for 7 weeks (approximately the same relative timespan as a human childhood) had significant gains in bone 22  mass (by DXA), Ct.Ar and estimated bone strength (minimum second moment of area; by pQCT) compared with the left forearm.[81] However, only bone geometry and strength gains persisted after 92 weeks of detraining.[81]  The mature skeleton also adapts its strength in response to mechanical loading; however, the adaptive mechanisms may differ to that of growing bone and may be site-specific. For example, in a turkey loading model, young turkeys (1-year olds) experienced significant structural benefits at the ulnar shaft (30% gain in Ct.Ar by microradiography) following an 8-week loading program, while older turkeys (3-years old) reaped no such benefits (-3% change in Ct.Ar).[82] Conversely, a 14-week running program in a rat model elicited similar gains in bone breaking loads (N, compression testing) at the proximal femoral neck in both young and adult rats (30% and 28%, respectively). Gains in bone strength in young rats were attributed to significant gains in total area (Tt.Ar by pQCT; 25%) and no gains in total BMD (Tt.BMD by pQCT), whereas gains in bone strength in adult rats were attributed to significant increases in Tt.BMD (23%) with no increases in Tt.Ar.[83] Collectively, evidence from animal models confirms that the growing skeleton has greater capacity to adapt to mechanical loads at the diaphysis, compared with the mature skeleton. The growing skeleton preferentially adapts to loading at the metaphysis through enhanced bone geometry, whereas the mature skeleton adapts through gains in density. I discuss human bone adaptation to mechanical loading in section 1.2.6.  1.2.3 Measuring bone in children and adolescents  As I described in section 1.2.2.1, direct measurements of whole bone strength can only be acquired through mechanical testing using animal models. However, a number of imaging tools are commonly used to estimate bone’s ability to resist fracture. A number of factors influence choice of imaging tool, including the study aim, site of skeletal assessment and cost. In this section, I discuss three imaging modalities in detail, including strengths and limitations of each: dual energy X-ray absorptiometry (DXA), peripheral quantitative computed tomography (pQCT) and high-resolution pQCT (HR-pQCT).    23  1.2.3.1 DXA  DXA was introduced nearly 30 years ago and is the current clinical gold standard for assessing bone health. Advantages of DXA include its wide availability, relatively low radiation exposure, low cost, short scan time and ability to scan clinically relevant sites such as the lumbar spine and proximal femur in addition to the whole body.[84] The effective dose equivalent for a whole body scan is low; approximately 1.4 to 13 µSv (comparable to effective daily background radiation dose ~ 4 µSv/day) depending on the scan.[85] In addition, availability of pediatric normative values make DXA an attractive tool for clinical and research settings.  DXA quantifies bone mass through attenuation of photons of two different energies, based on known density of different tissues.[84] Attenuation of each pixel of the X-ray beam is measured as it projects from an X-ray source above the participant to one or more X-ray detectors beneath the table. Beam attenuation is greater in mineralized tissues compared with soft tissues; bone is more dense than soft tissue due to heavier calcium and phosphate composition.[86] Figure 1.11 illustrates how photon attenuation is measured within each pixel of a DXA image. DXA measures the mass of hydroxyapatite (in g/cm2) along a straight path from the X-ray source to the detector.[53] X-ray beam attenuation within pixels located in a given region (i.e., whole body, femoral neck) above a certain threshold (set by the manufacturer) are averaged to provide areal BMD (aBMD, g/cm2). Bone area (BA, cm2) is the total area of all pixels that exceeded the bone threshold. Bone mineral content (BMC, g) is the product of aBMD and BA.  24   Figure 1.11. Illustration of densitometry. Photons are attenuated during transmission, producing an attenuation profile proportional to the mass of mineralized bone in the scanning path. Reprinted from Seeman,[87] with permission from Endocrine Society.  Despite widespread use, DXA is not without limitations. First, because DXA derives its output by summing the mineral mass between the X-ray source and detector, it cannot determine 3D cross-sectional geometry, distribution (thickness) of mass[86] or true volumetric density (g/cm3).[88] Given its planar, two dimensional (2D) nature, DXA cannot account for bone depth; thus, measures of bone mass are strongly influenced by body size (e.g., larger bones have greater aBMD compared with smaller bones because of differences in size).[84] For example, aBMD is systematically underestimated in shorter people. Further, catch-up growth results in higher values of aBMD even though actual volumetric BMD has not changed.[88] This issue is problematic when comparing bone mass between children of different sizes or within the same children longitudinally. Strategies to overcome DXA’s limitations include adjusting BMC for bone area, height and age, body weight or muscle mass.[88] Method of adjustment largely depends on the nature of available reference data.[89]  Second, DXA cannot differentiate between cortical and trabecular bone compartments. These parameters must be considered as only quantifying bone mass does not adequately describe bone’s mechanical competence. Further, because bone responds to loading by adding 25  new bone tissue where mechanical demands are greatest, small increases in bone mass can substantially improve bone strength. This was highlighted in several animal studies whereby minimal exercise-induced increases in DXA-derived BMC and aBMD (< 10%) were accompanied by substantial increases (> 60%) in bone strength (by micro-CT).[90,91] The importance of assessing bone geometry is illustrated in Figure 1.6; marked variation in bone bending strength (section modulus) occurs despite bones having the same measured aBMD.[54] Finally, body composition influences DXA outcomes. DXA assumes homogenous distribution of fat around bone;[92] however, soft tissue thickness affects beam magnification, such that non-uniform distribution of fat around bone may lead to inaccurate assessment of BMC and aBMD.[93]   1.2.3.1.1 Hip structural analysis  Application of hip structural analysis (HSA) to proximal femur DXA scans is a strategy commonly used to counter DXA’s 2D limitations. Based on mechanical engineering principles, HSA derives measurements of bone geometry, such as CSMI, and indices of bone strength, such as section modulus (a common measure of bone’s resistance to bending at the femoral neck), from bone mineral data in the image plane.[94] As DXA only projects mineral in the cross-section, excluding soft tissue and voids, a bone thickness profile can be derived by dividing each pixel’s mineral mass (g/cm2) by the average mineral density of fully mineralized bone (1.05 g/cm3).[53] The thickness profile represents the bone cross-section as if it were compressed into solid cortical bone.[53] The thickness profile collapses information regarding the distribution of mass along the X-ray paths, but preserves the distributional information in the image plane. Using HSA, mineral thickness profiles and cross-sections are extracted at three locations of the proximal femur: the narrow neck, the intertrochanteric region and across the shaft, from which CSMI and section modulus can be derived. Limitations of this approach include: 1) the assumption that the femoral neck and shaft are circular; 2) the assumption that tissue mineral density is constant; and 3) the proportion of cortical and trabecular bone is constant within the cross-section, which is often not the case.[84] As children’s bones are less mineralized compared with adult bones, estimates of cross-sectional geometry are often underestimated in pediatric 26  HSA studies.[86] Further, scan dimensions are altered by participant/patient positioning, which makes it difficult to differentiate between actual dimensional changes and positioning errors.[53]  1.2.3.2 pQCT  Quantitative computed tomography (QCT) differs from DXA in that it directly measures bone cross-sectional geometry and volumetric BMD (g/cm3) for a given region of interest (ROI). Central QCT systems are used clinically to image vertebral and femoral geometry and density, but are rarely used in pediatric research due to a higher dose of ionizing radiation (150-300 µSv; ~ one tenth to one fifth of total annual effective background radiation dose) compared with DXA.[41] In contrast, peripheral QCT (pQCT) is primarily a research tool used to assess bone geometry and BMD at distal and shaft sites of the tibia and radius. Like DXA, pQCT is common in pediatric research due to its short scan time (~3 min per scan) and minimal effective radiation dose (< 0.1 µSv for a complete scout view and pQCT scan).[95] pQCT is reliable, with in vivo reproducibility at tibial shaft sites ranging from 0.4-1.9% (coefficient of variation (CV)).[96] Reference data are available for children and young adults for cortical and trabecular BMD, CSA and cortical thickness.[97,98] The Centre for Hip Health and Mobility where I conducted my studies houses the XCT 3000 (Scanco Medical, Basserdorf, Switzerland). Thus, I focus my discussions on this model and on acquisition and analysis of bone images at the tibia shaft, specifically. As with DXA, pQCT quantifies bone by evaluating attenuation of ionizing radiation (from an X-ray beam) through an object, from source to detector. The central pQCT gantry diameter is 300 mm. An X-ray tube produces a narrow beam with a focal spot size of 250 x 250 µm operating at 60 kV (Figure 1.12).[95] The gantry rotates in 12° steps for 15 translations to obtain a single image with a 2.5 mm slice thickness.[95] There is a minimal amount of scatter radiation from pQCT as the beam is tightly collimated (<1 µSv). One pQCT scan results in less than 1/1000 of the recommended yearly radiation exposure (1 mSv), and the effective dose is significantly less than that of a standard chest X-ray (100 µSv).[95] Despite these advantages, given its maximum imaging resolution of 0.2 mm, a limitation of pQCT is that it cannot accurately assess bone microarchitecture or separate cortical and trabecular bone in regions with a thin cortex, such as the distal radius.[99] 27  Figure 1.12. Image of peripheral quantitative computed tomography system (pQCT), model XCT 3000 (Stratec Medizintechnick GmbH). An illustration of leg positioning for pQCT tibia scans (by Vicky Earle, Medical Illustrator).   1.2.3.2.1 Image acquisition and analysis  Protocols for pQCT image acquisition are not standardized; thus, there exists considerable variability in reported scan protocols. The user can alter various parameters, including image resolution and scan speed. Radiation exposure varies with those parameters such that higher resolution scans increase radiation exposure. pQCT pixel sizes range from 0.2 mm (higher resolution) to 0.6 mm (lower resolution). Shorter scan speeds minimize radiation dose and movement artifacts, yet may sacrifice resolution. Use of a lower resolution pixel size increases the possibility of partial volume effects (PVE). The PVE refers to presence of tissues of varying densities (i.e., soft tissue and bone) within the same pixel (Figure 1.13),[100] which could underestimate BMD. Thus, minimizing radiation exposure and PVE should be carefully considered when choosing an acquisition protocol. Common protocol for pQCT image acquisition in pediatric studies is a 0.4 mm pixel size and a 30 mm/s scan speed.[101-103]   28  Figure 1.13. Illustration of the partial volume effect (PVE), whereby pixels at bone edges (blue pixels) contain both bone and soft tissue densities, resulting in a lower density for the blue pixels. Smaller bones have more pixels close to the bone edge and may be more affected by PVE. Reprinted from Zemel et al.,[100] with permission from Elsevier.  Reference line placement is also an issue with pediatric pQCT scans, as the measurement site continually migrates as bone grows. Thus, our research group chose to assess a site that is the same relative position from a fixed bony landmark and is reproducible during growth. Specifically, to assess cortical bone at the tibial midshaft we scan a site 50% of the distance proximal to the distal tibial endplate. We also assess muscle cross-sectional area (MCSA) at this site. This differs slightly from the manufacturer’s recommendation that MCSA be measured at the 66% site as this is where the muscle belly is largest, on average.[104] However, MCSA at the two sites is strongly correlated (r = 0.95, n = 20 girls and boys aged 9-11 years; unpublished data from the Healthy Bones Study (HBS)). Further, acquiring MCSA at the 50% site reduces the number of scans needed in pediatric studies.  As with pQCT image acquisition, image analysis protocols are not standardized. Options include default settings and user-defined thresholds and modes. Prior to image analysis, the user defines a ROI, either automatically or manually. First, pQCT software separates bone from soft tissue by removing pixels below a user-defined threshold, leaving an outer edge of bone. Remaining (bone-filled) pixels are used to calculate total bone outcomes (Contour Mode; Tt.Ar and Tt.BMD). Second, the pQCT software removes all pixels in the ROI with an attenuation 29  coefficient below the defined threshold for cortical bone (Separation Mode; Ct.Ar, Ct.Th and cortical BMD (Ct.BMD)). At shaft sites, the pQCT software provides an estimate of bone strength (SSI) in bending and torsion. SSI is calculated as the integrated product of section modulus and Ct.BMD.[105] The ratio of Ct.BMD to normal physiological density (1200 mg/cm3) provides an estimate of the modulus of elasticity. SSI can be determined with respect to the polar (z) axis (SSIp, measures moment in torsion) or the bending (x,y) axes (SSIx/y, measures moment in bending).[105] SSIp explained over 90% of the variance in long bone torsional mechanical properties in adult cadavers.[106] Bone strength in compression is estimated at distal sites using BSI (described in Section 1.2.2.1.3).[55]   1.2.3.3 HR-pQCT   As with pQCT, high-resolution pQCT (HR-pQCT) uses an X-ray and detector that rotate around the lower leg or forearm (Figure 1.14). The X-ray tube has an 0.08 mm focal point, spanning a 12.6 cm field of view. The system acquires 110 parallel CT slices, stacked to form a 3D image. In contrast to pQCT, the imaging resolution of HR-pQCT ranges from 82 µm in first-generation scanners, to 61 µm in second-generation scanners.[107] In this thesis I focus on the first generation scanner (XtremeCT, Scanco Medical). Resolution of HR-pQCT permits accurate assessment of trabecular microarchitecture, such as trabecular number and thickness.[108] Adult trabecular thickness ranges from 100-300 µm,[41] whereas in children trabecular thickness ranges from ~60-100 µm.[109] Manufacturer’s standard settings include an effective energy of 60 kVp, X-ray tube current of 900 µA and integration time of 100 ms. The < 3 µSv radiation exposure is equivalent to 0.2% of total annual background radiation in Canada; radiation scatter from a standard scan is very low (0.75 µSv).[110]  30   Figure 1.14. Image of high-resolution peripheral quantitative computed tomography (HR-pQCT) XtremeCT system (Scanco Medical) and leg positioning for tibia scan.  1.2.3.3.1 Image acquisition and analysis  To acquire HR-pQCT images, the skeletal site of interest (tibia or radius) is first immobilized in a carbon fiber cast and placed inside the scanner’s gantry. A 2D anterior-posterior scout view scan is performed to identify the region of interest. Manufacturer’s standard protocol uses a ROI of 9.5 mm and 22.5 mm proximal to the radial inclination tuberosity and tibial end plate, respectively. However, as with pediatric studies that use pQCT, a relative ROI is preferable over an absolute ROI for several reasons. First, in growing bone, a fixed site (e.g., 22.5 mm from a bony landmark) is a ‘moving target’ over time, one that is relatively more distal as the participant grows. A relative site (e.g., 8% of bone length from a fixed bony landmark) on the other hand, enables the operator to scan the same ‘relative’ region across growth.[111] Second, cortical and trabecular compartments differ markedly from distal to proximal sites along a long bone. Thus, relative positioning enables comparisons between participants of different sizes and with shorter versus longer limbs lengths. For example, a fixed ROI at 22.5 mm proximal to the tibial end plate is equivalent to a 5% site on a tall participant with a tibia length of 430 mm and a 7% site on a shorter participant with a tibia length of 340 mm. Comparisons between 5% and 7% 31  sites is confounded by a greater proportion of cortical compared with trabecular bone at the more proximal site. Finally, in children we must avoid irradiating the growth plate whenever possible. The 8% site of the distal tibia and 7% site of the distal radius includes both cortical and trabecular bone but excludes the growth plate in most children.[110,111] Once the reference line is placed at the tibial end plate or distal medial edge of the radius, 110 tissue slices are scanned proximal to the 8% or 7% measurement site, respectively. In total, an approximate 9.02 mm region of the tibia or radius is scanned in less than 3 min. HR-pQCT outcomes from a standard morphological analysis include Tt.Ar (mm2), Tt.BMD (mg HA/cm3), Tb.BMD (mg HA/cm3), BV/TV, Tb.N (1/mm), Tb.Th (mm) Tb.Sp (mm), Ct.BMD (mg HA/cm3) and Ct.Th (mm). Of the trabecular measures, Tb.BMD and Tb.N are measured directly, while BV/TV, Tb.Th and Tb.Sp are derived from Tb.BMD and Tb.N[112] HR-pQCT demonstrates good agreement (R2 = 0.59-.96) with micro-CT–derived (~20 µm voxel size) standard outcomes from cadaveric bone specimens.[113] In vivo estimates of reproducibility are less than 1% (root mean squared CV) for density parameters and between 0.7% (BV/TV) - 4.4% (Tb.Sp) for microarchitecture parameters at the distal tibia and radius.[114]  Standard HR-pQCT analysis frequently mistakes thin or porous cortical bone as trabecular bone and vice versa. Therefore, automated segmentation algorithms can be applied to HR-pQCT images using customized software to more accurately separate cortical and trabecular compartments (Figure 1.15).[115] Outcomes from automated segmentation algorithms include: Tt.Ar (mm2), Ct.Ar (mm2), cortical porosity (Ct.Po, %), Ct.BMD (mg HA/cm3) and Ct.Th (mm).[115,116] In vivo estimates of reproducibility (root mean squared CV) for cortical parameters range from 0.6% for Ct.BMD to 13% for Ct.Po.[117]  32  Figure 1.15. Illustration of trabecular (top image, green) and cortical (bottom image, grey) regions from a segmented high-resolution peripheral quantitative computed tomography scan.  Resolution of HR-pQCT is sufficient to obtain finite element analysis (FEA)-derived estimates of compressive bone strength. FEA is a computationally demanding numerical approach that converts 3D image data into FEA meshes voxel by voxel.[118] Conceptually, FEA breaks down a complex structure (i.e., bone) into smaller simpler elements. Computer-generated FEA models simulate applied loads, typically uniaxial compressive forces, onto the smaller elements. FEA outcomes include failure load (force that causes bone to fail; F.Load, N) and stiffness (reaction force using the FEA model at 1% strain divided by the average bone CSA from standard analyses; N/mm).[118] Stiffness is used to estimate ultimate stress (highest stress the bone can withstand per unit area without failing; U.Stress, MPa). F.Load derived from FEA is used to calculate load-to-strength ratio at the distal radius (ratio of estimated fall load applied to the outstretched hand after a fall from standing height; φ), an estimate of forearm fracture risk.[119,120] FEA-derived U.Stress demonstrates good agreement (R2 > 0.94) with experimentally-determined strength from destructive loading in human adult cadaver forearms, suggesting U.Stress is a good surrogate of bone strength (Figure 1.16).[118] Further, strong correlations between bone stiffness measures of the tibia and radius by HR-pQCT (and FEA) and stiffness of the lumbar vertebrae and proximal femur using central QCT (r = 0.69-70), suggest the mechanical competence of the distal radius and distal tibia reflect that of central, clinically 33  relevant sites.[121] However, no studies have investigated the relationship between FEA models and experimentally-determined bone strength in pediatric cadaver bone. Figure 1.16 Illustration of stress-strain curve of destructive loading of cadaveric distal radii to determine linear and elastic failure regions. P = platen force. Reprinted from MacNeil et al.,[118] with permission from Elsevier.  HR-pQCT is limited to scanning the peripheral skeleton. However, it is an attractive imaging tool based on its high-resolution and short scan time (~2.8 min). Nevertheless, HR-pQCT acquisition and analysis protocols are not yet standardized, which limits comparisons across studies. Normative data for HR-pQCT are available in older adolescents (16-19 years)[122] and adults,[123] but not in children or younger adolescents. Finally, movement during high-resolution imaging increases the likelihood of motion artifacts (streaks or discontinuities on the scan), which may require that scans be repeated or excluded from analysis.[124]  1.2.4 Maturity- and sex-related differences in bone strength and its determinants  In the following sections, I outline several methods to assess maturity in children and adolescents. I also briefly review hormones that influence skeletal development and maturity- and sex-related differences in skeletal development. 34  1.2.4.1 Assessing maturity  Growth and maturation refer to a range of developmental processes (i.e., physical, sexual, cognitive). For the purpose of this thesis, I consider growth as somatic; that is, increases in body size or mass. I refer to maturation as the tempo and timing of biological (physical and sexual) changes associated with somatic growth.[125] As there is no constant relationship between maturity and time,[126] chronological age is not equivalent to stage of maturation.   1.2.4.1.1 Sexual maturation  Maturity in children and adolescents is commonly assessed as per the method of Tanner, which is based on development of secondary sex characteristics; breast and pubic hair development in girls and pubic hair and genital development in boys.[127] Tanner divided the continuous process of maturation into five discrete stages: Tanner stage 1 represents pre-puberty, Tanner 2 and 3 early-puberty, Tanner 4 peri-maturity and Tanner 5 post-puberty or reproductive maturity. A physician or nurse may perform the maturity assessment in clinical or research settings; however, if such personnel are unavailable, participants may self-assess using photographs or line drawings depicting stages of breast and genital or pubic hair development. Physician (or nurse)-determined Tanner stages correlate well with testosterone and estrogen levels in boys and girls.[128] However, when self-assessed, younger, less mature participants tend to overestimate their development, while more mature children are prone to underestimate development.[126] Further, the practicality of Tanner stages may be confounded by body composition; obese girls tend to overestimate their Tanner breast stage as adipose tissue can be mistaken for breast development.[129] Nevertheless, maturity assessment using the method of Tanner is attractive for clinical and research purposes since it is cost effective and only requires a one-time measurement. However, maturation is continuous and, thus discrete stages do not account for the large variation between two children of the same Tanner stage. In addition, there is sexual dimorphism in timing and tempo of maturity, such that girls mature at an earlier chronological age and Tanner stage compared with boys, on average.[126,130] For example, boys tend to reach peak height velocity (PHV) after entering pubic hair stage 4, while girls tend to 35  reach PHV after entering into breast or pubic hair stage 3.[130,131] Thus, boys and girls are not comparable at the same chronological age or stage of secondary sex development.  In girls, age at menarche is a common maturity indicator in cross-sectional and longitudinal studies.[125] Menarche (first menstrual period) is a relatively late event in sexual maturation, occurring at approximately Tanner breast stage 4.[132] Menarcheal status is typically self-reported by asking participants if they have experienced menarche. If yes, the participant is queried for a more precise date (month and/or year). Most girls can remember within a month of when their first ‘period’ occurred.[126] For boys, however, there is no such clear maturational event with timing that aligns with menarcheal status.  1.2.4.1.2 Skeletal maturation  Skeletal age, or bone age, is determined through radiography of the hand/wrist and is typically performed by trained clinicians based on one of three methods: Greulich-Pyle, Tanner-Whitehouse and Fels.[126] These methods use the left hand and wrist to estimate skeletal age; however, they differ with respect to bones assessed, scoring method and reference sample.[126] Skeletal age assesses fusion of the epiphyseal plate and is based on the premise that greater bone development and less cartilage will be observed in a more mature individual compared with a less mature individual.[125] While these techniques are used frequently in clinical settings, their broad use is limited by the ionizing radiation associated with radiography.  1.2.4.1.3 Somatic maturation  Maturation can also be assessed based on somatic changes in growth trajectories. Age at PHV (APHV), the age when maximum velocity in stature is attained, is a common indicator of somatic maturation and often termed ‘biological age’.[126] As a continuous measure, APHV overcomes limitations of Tanner staging, as boys and girls can be aligned on a common maturational landmark. APHV occurs at approximately 11.7 years in girls and 13.4 years in boys,[133] with a range of approximately 4 years for each sex. APHV typically occurs when 90% of final adult stature is achieved,[134] which is 7-9 months prior to peak in bone mineral accrual velocity (Figure 1.17)[133] and 5-7 months before peak in femoral neck strength velocity.[135] In 36  addition, APHV is an important relative marker of function in boys and girls. Estimated velocities of many performance tasks in both boys and girls reach a peak at the same time as maximal growth in height.[136-138] Further, APHV approximates peak skeletal age velocity.[139,140]   Figure 1.17. Total body bone mineral content (BMC TB) accrual velocity and ages at peak BMC and peak height velocity (PHV) for girls (dotted line) and boys (solid line) aligned on chronological age. The lag period between age at PHV and peak BMC is approximately 7-9 months. Reprinted from Bailey et al.,[133] with permission from John Wiley and Sons.  Direct assessment of APHV requires frequent (typically annual), serial assessments of height for at least two years surrounding peak growth; from these data growth trajectories can be mapped. However, in cross-sectional or short-term prospective studies, prediction equations can be used to estimate maturity offset (years from APHV).[141,142] Prediction equations use one-time measurements of anthropometric variables, including combinations of height, sitting height, leg length and age to estimate maturity offset. The Mirwald prediction equations are sex-specific and use height, sitting height, leg length, chronological age and their interactions to predict years 37  from APHV.[141] The Mirwald equation was cross-validated in longitudinal datasets (one Canadian sample and one Flemish sample) and predicted 88% and 89% of the variation in APHV in girls and boys, respectively.[141] However, accuracy of maturity offset prediction differs based on timing of maturation, such that the equation underestimates actual APHV in late-maturing boys and girls and overestimates APHV in early-maturing boys and girls.[143] Our research group recently redeveloped the Mirwald equation to address concerns regarding its accuracy.[142] Specifically, longitudinal data (multiple observations per child) were analyzed using cross-sectional techniques, thereby ignoring within-person variation. This underestimated standard errors and p-values, and included spurious variables in the prediction model.[142] The re-developed equation includes fewer predictor variables and is as accurate as its predecessor. For example, the girls’ equation includes age and height as predictors and explained 91% of the variance in APHV, while the re-developed boys’ equation includes age and sitting height and explained 90% of the variance in APHV.[142] As sitting height may not be assessed in all growth studies, an alternate equation was developed for boys. This model used height instead of sitting height and demonstrated comparable accuracy (R2 = 0.90). External validation of the model in two cohorts of children and adolescents demonstrated that 90% of predictions were within ± 1 year of actual APHV.[142] However, both Mirwald and Moore equations were developed and validated in white children only and may be inappropriate for use in ethnically diverse samples. Further, neither equation was robust enough to assess maturity prior to initiation of the growth spurt.   1.2.4.2 Maturity- and sex-related differences in bone development  In this section, I review development of bone strength and its determinants through adolescent growth. I focus on studies that used pQCT to assess maturity- and sex-related adaptations at the diaphysis and studies that used HR-pQCT to assess maturity- and sex-related adaptations at the metaphysis. Finally, I briefly discuss several important determinants of bone development throughout growth. I acknowledge that DXA studies were key to advance our understanding of bone adaptations to PA, and direct the reader to several excellent reviews of DXA-based studies.[10,144,145] In this section, I briefly describe findings from one prospective DXA study; I focus, whenever possible, on studies that employed 3D imaging tools. 38  One of the most widely cited studies, the University of Saskatchewan Pediatric Bone and Mineral Accrual Study (PBMAS), followed approximately 200 healthy children (age 8 to 15 years at study entry) through early adulthood. Researchers aligned children on APHV to control for maturation. Bone mass accrual peaked approximately 1 year after PHV, while approximately 35% of total body and lumbar spine BMC and more than 27% of femoral neck BMC was accrued during the 4-years around PHV.[9,133,134] Bone accrued in this short period represents more than will eventually be lost across 50 years of adulthood.[146] Sex differences in timing and magnitude of bone accrual were observed, such that total body BMC accrual occurred 1.4 years earlier in girls and was smaller in magnitude compared with boys.[133] Further, boys had significantly greater total body, femoral neck and lumbar spine BMC at all biological ages (-3 to +4 years from APHV).[147] Less is known about sex differences in bone strength and its determinants during growth. A better understanding could provide insight into the higher incidence of low-energy fractures in boys compared with girls during the adolescent growth spurt.[148] Despite recent advances in high-resolution imaging technologies, only three studies (one from our group) used HR-pQCT to examine maturity- and sex-related adaptations in bone geometry and microarchitecture during adolescent growth.[4-6] Only two of these studies (one from our group) used FEA to estimate bone strength.[4,5] I briefly summarize these studies in Table 1.1 and refer to each by study name throughout this section.  39   Table 1.1. Overview of studies that used HR-pQCT to examine sex and maturity-related adaptations in bone strength and its determinants during adolescent growth. Cohort Mayo Clinic[5] Australian[6] HBSIII[4]  Study design Cross-sectional Cross-sectional 2 years at radius,  3 years at tibia; 1-year between measures  N  N = 127  N = 129  N = 398  Sex 66 girls / 61 boys 60 girls / 69 boys 212 girls / 186 boys  Ethnicity 96% white, 1% Asian, 3% other 100% white 47% white, 46% Asian, 7% other  Age 6-21 years 5-18 years 9-22 years   Maturity 14% bone age 6-8 years  26% bone age 9-11 years 23% bone age 12-14 years  23% bone age 15-17 years  14% bone age 18-21 years 51% Tanner 1 12% Tanner 2 9% Tanner 3 13% Tanner 4 15% Tanner 5 13% Tanner 1  24% Tanner 2/3  32% Tanner 4  31% Tanner 5   Site scanned Radius (1 mm proximal to the epiphyseal growth plate of radius) Radius (4% site); Tibia (7% site) Radius (7% site); Tibia (8% site)  Bone outcomes Standard analysis: BV/TV, Tb.Th, Tb.Sp, Tb.N Gaussian filter and threshold: Ct.BMD, Ct.Th, periosteal and endosteal circumference, Ct.Po index Finite element analysis: F.Load, Fall force, load-to-strength ratio, % load cortical bone Standard analysis: Tt.BMD, BV/TV, Tb.Th, Tb.Sp, Tb.N, Tt.Ar, Ct.Ar, Ct.Th, Ct.BMD  Standard analysis: Tt.BMD, BV/TV, Tb.Th, Tb.Sp, Tb.N Automated segmentation: Ct.Po, Ct.Th, Ct.BMD, Ct.Ar, Tb.Ar, Tt.Ar   Finite element analysis: U.Stress, F.Load, load-to-strength ratio   1.2.4.2.1 Bone strength  As discussed in section 1.2.2.2, changes in geometry, BMD and microarchitecture influence gains in bone strength during growth. Of the three cohorts assessed using HR-pQCT, all demonstrated substantial increases in bone strength throughout adolescent growth. Figure 1.18 illustrates that girls and boys in the HBSIII cohort had approximately 100% and 200% greater compressive bone strength (F.Load), respectively, at the distal radius during post-puberty compared with pre-puberty.[4] There were smaller differences (approximately 35% in girls and 50% in boys) at the distal tibia.[4] In concert with increased bone strength, load-to-strength ratio 40  (an indicator of distal radius fracture risk) decreased by half from pre- to post-puberty in girls and boys.[4,5] Similar maturity-related increases in bone bending strength (SSIp or section modulus using pQCT) were observed at shaft sites. Specifically, there were gains of approximately 100% in girls and 200% in boys at the radius from pre- to post-maturity[149] and 24% in girls and 44% in boys at the tibia from early to post-puberty.[150] Maturity-related gains in bone strength appear larger at the radius compared with the tibia; however, this is an artifact of expressing change as a percentage. Absolute bone strength and gains in absolute bone strength were smaller at the radius than the tibia (likely due to the non-weight-bearing nature of the site). Despite this, the radius demonstrated greater percentage gains across growth due to smaller baseline (pre-puberty) values compared with the tibia.  Figure 1.18. Illustrations of sex differences in high-resolution peripheral quantitative computed tomography (HR-pQCT) parameters at the distal radius by pubertal group based on the method of Tanner staging: A) cortical density (Ct.BMD), B) cortical porosity (Ct.Po), C) cortical area (Ct.Ar) and D) failure load. a, p < 0.001; b, p < 0.01; c, p < 0.05: significant difference between girls and boys within the same puberty group. d, p < 0.001; e, p < 0.01; significant difference between puberty group and the PRE group within sex. Reprinted from Nishiyama et al.,[4] with permission from John Wiley and Sons.  41  Sex differences in magnitude of bone strength accrual result in consistently stronger bones in boys compared with girls throughout adolescent growth at the weight-bearing tibia and from mid-puberty onwards at the radius. In the HBSIII cohort, F.Load was 16-25% greater in boys at the tibial metaphysis from pre- (Tanner 1) to post- (Tanner 5) puberty.[4] SSIp (by pQCT) was 6% greater at the diaphysis in pre- and early-pubertal boys compared with girls.[151] At the radial metaphysis, F.Load was 27-39% greater in boys compared with girls from peri-(Tanner 4) to post-puberty in the HBSIII cohort (Figure 1.18),[4] and significantly greater in boys compared with girls from 12-14 years (bone-age) onwards in the Mayo Clinic cohort.[5] Similarly, SSIp (by pQCT) was 18-32% greater in boys compared with girls at all pubertal stages at the radial diaphysis, except for a non-significant 14% difference at Tanner stage 4.[149] However, these comparisons are limited by their cross-sectional design and by use of Tanner staging to assess maturation. Prospective studies that use pQCT and HR-pQCT are needed to confirm the trajectory of bone strength accrual during adolescence and to illustrate how this may differ between sexes and skeletal sites.  1.2.4.2.2 Bone geometry  As described in section 1.2.2.1.3, bone CSA is a key determinant of overall bone strength, as bone’s resistance to bending is proportional to its CSA to the third power.[152] Bone’s resistance to compression is also proportional to CSA.[55] Boys’ greater Tt.Ar confers them a strength advantage throughout growth compared with girls.[4-6,151] Substantial increases in bone strength during adolescent growth are underpinned by increases in Tt.Ar from pre- to peri-puberty; Tt.Ar plateaus thereafter.[4-6] For example, in the Australian cohort, Tt.Ar at the distal tibia was 35% and 55% greater in peri-pubertal girls and boys, respectively, compared with same sex pre-pubertal children.[6] The difference in distal radius Tt.Ar was 70% greater in peri- compared with pre-pubertal children.[6] Slightly smaller gains in distal tibia (10-20%) and radius (20-40%) Tt.Ar were observed in the HBSIII cohort from pre- (mean age 11 years) to peri- (mean age 16 years) puberty.[4]  Diaphyseal sites also demonstrated gains in Tt.Ar through periosteal expansion (Figure 1.19).[153,154] However, there is some discrepancy regarding sex-specific adaptation at the endocortical surface. Early cross-sectional radiographic studies of the second metacarpal 42  concluded that boys and girls experienced periosteal expansion and endocortical contraction at the diaphysis of the second metacarpal during growth. However, girls experienced more endocortical contraction compared with boys.[155-157] These findings are supported by a 2-year longitudinal study of pubertal girls (age 10-13 years at baseline) where narrowing of the tibial shaft (60% site) marrow cavity was observed in girls after menarche.[158] However, these results are discordant with reports of increased marrow cavity area (Tt.Ar – Ct.Ar) in both boys and girls throughout puberty (Figure 1.19).[153,154] Mechanistically, an increase in the CSA of bone and marrow would enhance bone strength by placing the neutral axis farther from the centre of mass. Contradictory findings may be due to differences in study design, imaging modality and/or method used to control for maturity. Importantly, none of the aforementioned studies examined changes at the endocortical surface in girls and boys relative to biological age.   Figure 1.19. Illustration of bone growth over 20 months at the tibia midshaft in early-, peri- and post-pubertal boys and girls using peripheral quantitative computed tomography (pQCT). Numbers show the mean increase (%) in cortical and marrow cavity areas. Adapted from Kontulainen et al.,[153] and reprinted from Daly et al.,[159] with permission from Karger.  43  1.2.4.2.3 Bone density  Measures of Ct.BMD using 3D imaging techniques typically reflect mass of mineral per unit volume of the cortical compartment, including intracortical pores.[160] Bone modeling and remodeling activity is inversely related to BMD. During periods of active modeling or remodeling, more young bone matrix with lower mineral density is present compared with older, denser bone matrix.[160] Ct.BMD increases approximately 15-30% throughout adolescence at distal sites.[4] However, mid-puberty tends to be characterized by a transient decrease or plateau in Ct.BMD, followed by considerable increases in Ct.BMD throughout later maturity (i.e., 4% increase from pre-to early-puberty and 16% increase from early- to post-puberty in girls at the distal tibia) (Figure 1.18).[4,5] Similar maturity-related gains in Ct.BMD were observed at diaphyseal sites (by pQCT) in pubertal girls, such that greater consolidation of cortical bone (approximately 10% increase in Ct.BMD) occurred following menarche.[158,161] However, no transient decreases in Ct.BMD were observed in studies of the tibial shaft[158,161,162] or distal radius (by pQCT).[31] Nonetheless, the limited resolution of pQCT and subsequent PVE may preclude accurate assessment of Ct.BMD in bone with relatively thin cortices (< 2.5 mm) such as in children at the tibia and at all ages at the distal radius. Consistent sex differences in Ct.BMD were observed at distal and shaft sites in later adolescence; girls demonstrated 2-10% greater Ct.BMD compared with boys from peri-puberty onwards.[4,5,162] Sexual dimorphism in Ct.BMD may arise in response to increased calcium demands during rapid adolescent growth,[163] and likely reflects boys’ greater magnitude of growth and prolonged growth period compared with girls.[164] Future prospective studies should confirm these findings by aligning girls and boys on biological age.   1.2.4.2.4 Cortical microarchitecture  Growth-related increases in Ct.BMD throughout adolescence are underpinned by thickened cortices and decreased Ct.Po. Specifically, at the distal tibia and radius, Ct.Th increased by 50-70% from pre- to post-maturity in each of the three pediatric studies that use HR-pQCT.[4-6] Ct.Po decreased by 20-50% across the same period in the HBSIII cohort.[4] Further, transient decreases in Ct.BMD during mid-puberty were mirrored by transient decreases 44  in Ct.Th at the radius in boys and girls in Mayo Clinic and Australian cohorts [5,6] and increases in Ct.Po at the tibia and radius in HBSIII boys.[4] These maturity-related deficits at the cortex (i.e., a thin cortical shell and increased porosity during accelerated growth) may contribute to the heightened risk of fracture during the pubertal growth spurt, when growth outpaces consolidation of cortical bone.[36] Prospective studies are needed to confirm this hypothesis. Sex differences in Ct.BMD are underpinned by sexual dimorphism in Ct.Po, such that boys have 25-175% greater Ct.Po compared with girls from early-puberty (Figure 1.18).[4] The timing of sex differences in Ct.Th is less clear. For example, in our HBSIII cohort, Ct.Th at the distal radius and tibia was 12-16% greater in boys compared with girls, but only during post-puberty.[4] In the Mayo Clinic cohort, boys demonstrated thicker cortices at the distal radius earlier in adolescence (early- and mid-puberty), but not in late- or post-puberty.[5] Finally, in the Australian cohort, girls demonstrated an advantage in Ct.Th compared with boys at both sites during peri-puberty. However, at post-puberty, cortices were thicker in boys compared with girls.[6] Discrepancies between studies likely reflect differences in methods used to segment the cortex, regions of interest and measures of maturity. Collectively, these findings suggest that greater Ct.BMD in girls compared with boys during peri- and post-puberty is largely a function of lower Ct.Po and is related to the increased intracortical bone turnover that boys experience as a result of greater magnitude of longitudinal growth.[163] Studies that align participants on biological age and incorporate automated segmentation algorithms are needed to clarify sexual dimorphism of the cortical shell.  1.2.4.2.5 Trabecular microarchitecture  As discussed in section 1.2.2.1.2, trabecular bone volume fraction (BV/TV; synonymous with Tb.BMD) is a function of the number, thickness and separation of trabeculae. Increases in BV/TV throughout growth may function to more efficiently transfer compressive loads from joint surfaces and increase bone’s mechanical competence.[165] Histomorphometric study of the iliac crest (age 2-23 years; 33 females, 25 males) suggested that Tb.N varied little with age, while increased Tb.Th contributed to gains in BV/TV during growth.[166] This may occur due to remodeling with a positive balance, such that osteoblasts add more bone than was previously resorbed during each remodeling cycle (or through modeling, where new bone is added without 45  prior resorption), resulting in a gradual increase in BV/TV.[167] However, recent studies that used HR-pQCT to examine maturity-related changes in BV/TV were inconsistent. Figure 1.20 illustrates that BV/TV did not change significantly in girls at the distal radius or tibia throughout adolescent growth in all three pediatric cohorts mentioned previously.[4-6] In contrast, BV/TV at both sites were approximately 20% greater in peri- compared with pre-pubertal boys in the Mayo Clinic and Australian cohorts.[5,6] In the HBSIII cohort, boys’ BV/TV did not differ between pre- to post-puberty.[4] Despite inconsistencies across these cohorts in growth-related adaptations in BV/TV, Tb.N did not vary with maturation in girls or boys at either bone site.[4-6] Therefore, as observed in the histomorphometric study, growth-related adaptations in BV/TV observed in boys were underpinned by 10-30% gains in Tb.Th across maturity,[4-6] potentially in response to increased serum testosterone.[5] It is difficult to explain why girls’ trabecular parameters did not differ with stage of maturation. One hypothesis is that trabecular bone volume and microarchitecture are programmed early in life in girls.[5] Thus, prospective studies that span a longer period prior to adolescent growth might clarify maturity-related changes in trabecular microarchitecture. Consistently greater BV/TV at the distal radius in boys (approximately 5-20%) compared with girls from peri-puberty onwards (Figure 1.20) was underpinned by adaptations in Tb.Th.[4-6] However, discrepancy exists regarding sexual dimorphism in trabecular microarchitecture at the tibia. In the Australian cohort, greater BV/TV in boys during peri- and post-puberty was related to significantly greater Tb.Th.[6] In contrast, boys’ greater BV/TV at the tibia in the HBSIII cohort was a function of a more substantial network of trabeculae, as indicated by significantly greater Tb.N, but not Tb.Th, in boys compared with girls.[4] Additional study is warranted to clarify sex-related differences in trabecular microarchitecture throughout adolescent growth and how such differences influence bone strength. 46  Figure 1.20. Illustration of sex differences in trabecular microarchitecture at the distal radius and tibia using high-resolution peripheral quantitative computed tomography (pQCT) across pubertal groups based on Tanner staging. ‡p<0.05 for sex difference. Reprinted from Wang et al.,[6] with permission from John Wiley and Sons.  1.2.5 Factors that influence of bone strength during growth  In this section, I briefly review intrinsic (i.e., genetics, hormones, ethnicity and muscle) and extrinsic (i.e., calcium, vitamin D, PA and sedentary time) factors that influence bone strength and its determinants in children and adolescents. I discuss the influence of PA and sedentary time on bone development in section 1.2.6 as they are primary variables of interest in this thesis.  1.2.5.1 Genetics  Total population variance for a given trait is explained by genetic and environmental factors and measurement error. Heritability is defined as the proportion of total population variance attributed to genetic factors.[168] Many genes regulate bone strength and its 47  determinants, such as those encoding receptors for steroid and calciotrophic hormones, local regulators of bone metabolism including growth factors and cytokines, bone matrix proteins and transporting factors.[169,170] Those implicated in bone remodeling include, but are not limited to, vitamin D, estrogen, calcitonin and parathyroid hormone receptors.[170] Classic assessment of heritability is based on twin models, under the assumption that monozygotic and dizygotic twins experience similar environmental factors that may influence a trait. If monozygotic twins are more similar to one another than dizygotic twins, the twin model assumes this must be attributed to shared genes.[168] Monozygotic twins share 100% of their genes, whereas dizygotic twins share 50%. Thus, the correlation for a given trait for a monozygotic twin should be double that of the dizygotic twin if that trait is 100% genetically-determined.[169] However, the twin model cannot prove that genetic factors are the sole cause of any correlation, as there may be non-genetic explanations for a stronger relationship between monozygotic twins (i.e., gene-environment interactions related to lifestyle or preferential loading of limbs).[168] Heritability estimates for BMC and aBMD (by DXA) at the lumbar spine and proximal femur range from 40-60% in family studies and 70-80% in twin studies. [169,171,172] However, much of this may be explained by body size, as heritability estimates for stature range from 60-80%.[169] Further, adjusting for body size attenuates heritability estimates of aBMD in twin and familial studies.[173]  Evidence from heritability studies of bone geometry and strength (by pQCT) suggest the influence of genetics may vary across skeletal sites. For example, heritability estimates of CSA at the distal radius within several Hutterite colonies in the United States ranged from 27% at the 4% site to 75% at the 20% site, after adjustment for age, sex, height and weight.[174] Site-specific differences in heritability may be partly explained by greater measurement error at the 4% site compared with the 20% site.[174] Heritability of compressive bone strength (BSI) in elderly female twins was greater at the distal radius (83%) compared with the distal tibia (61%).[175] Unlike the tibia, the distal radius is not subjected to compressive loads from body weight. Thus, the complex interaction between genetic and environmental influences is likely site-specific, such that the weight-bearing tibia may be more sensitive to environmental factors compared with the non-weight bearing radius.   48  1.2.5.2 Hormones  Significant alterations in the hormonal environment drive dramatic increases in linear growth and bone strength during maturation. For example, growth hormone (GH) and insulin-like growth factor (IGF-1) regulate longitudinal bone growth and influence bone modeling by stimulating osteoblasts and chondrocytes.[176] GH deficiency during childhood significantly reduces (50%) longitudinal bone growth resulting in smaller bone size and less bone mass accrual. [176] GH secretion peaks in concert with PHV and decreases thereafter (Figure 1.21), while IGF-1 peaks slightly later.[177] GH and IGF-1 continue to influence bone remodeling following cessation of linear growth.[176]  Figure 1.21. Illustration of peaks for sex steroids, height and BMC velocity, growth hormone and IGF-1 amplitude in relation to age and pubertal stage in girls. Reprinted from MacKelvie et al.,[144] with permission from BMJ Publishing Group Ltd.  Sex steroids, estrogen and testosterone, influence bone regulation throughout growth.[178] Estrogen’s influence on bone development is biphasic in girls and boys, such that low concentrations during early-puberty stimulate skeletal growth through increased secretion of GH and IGF-1, while elevated concentrations during late-puberty limit growth by stimulating growth 49  plate closure.[179] Testosterone, on the other hand, directly encourages bone formation in girls and boys by inhibiting osteoblast apoptosis, promoting osteoblast formation at the growth plate[180] and stimulating GH and IGF-1.[178] In addition, testosterone indirectly affects bone formation through its anabolic effect on muscle mass which increases bending moments.[178] As I discussed in section 1.2.4.2.2, growth-related adaptations in bone geometry enhance bone strength. Historically, estrogen was believed to inhibit periosteal apposition, such that sexual dimorphism in bone size (larger bones in men compared with women) was due to greater testosterone and less estrogen exposure in men compared with women.[178] However, case studies of boys with aromatase-deficiency (estrogen insensitivity) highlighted the critical influence of estrogen on normal skeletal growth in boys.[178,181] For example, a 16-year old boy had a bone age of just 12 years, despite normal levels of testosterone and full pubertal development. Three years of estrogen treatment increased his radius total CSA by 46%, Ct.Th by 12% (by pQCT) and increased bone age to 17 years.[181] Thus, estrogen and testosterone are essential for normal periosteal expansion. Males’ greater bone size is attributed to greater periosteal apposition during adolescent growth, due to extended pubertal growth and later epiphyseal fusion, compared with girls.[178,179,182] Calcitropic hormones (parathyroid hormone, vitamin D and calcitonin) also influence bone modeling and remodeling and regulate serum calcium. Parathyroid hormone regulates calcium homeostasis by interacting with bone, kidney and intestine.[183] Parathyroid hormone increases serum calcium by stimulating bone resorption and enhancing calcium absorption from the kidneys and intestine. Parathyroid hormone responds in accordance to serum calcium concentrations, such that large increases in serum calcium suppress parathyroid hormone secretion. Drops in serum calcium increase parathyroid hormone secretion.[183] Parathyroid hormone exerts anabolic and catabolic influences on bone, depending on its release pattern (intermittent or continuous) and subsequent stimulation of growth factors and cytokines.[183] In contrast, calcitonin influences calcium homeostasis by reducing serum calcium through inhibiting bone resorption and attenuating renal calcium resorption.[183] I discuss the influence of calcium and vitamin D on bone strength accrual in detail in section 1.2.5.4.     50  1.2.5.3 Ethnicity  Race is traditionally used to define biologic (genetic) differences in a person’s appearance whereas ethnicity is commonly used to describe sociological and cultural factors such as nationality, ancestry and language. There is no consensus within the bone research literature as to which term is most appropriate to describe biological differences in bone development. Thus, I use the ethnicity to refer to biological and environmental factors that contribute to differences in bone strength and its determinants between people of different ancestries (i.e., white versus Asian). I use ‘white’ to describe those of European descent, ‘Asian’ to describe those of Asian descent, ‘black’ to describe those of African descent and ‘other’ to describe those of mixed ethnicity. Evidence suggests that fracture incidence is lower among Asian children and adults compared with white children and adults.[184,185] However, we know little about how ethnic differences in bone strength and microarchitecture contribute to ethnic differences in fracture incidence. For example, in the multi-ethnic HBSIII cohort, Tt.Ar (by HR-pQCT) at the distal radius was smaller in Asian males compared with their same age white peers, independent of muscle mass and limb length.[186] However, Asian males and females had thicker and denser cortices compared with their white peers, while Asian males also had less porous cortices, contributing to similar estimates of bone strength (F.Load and load to strength ratio) between Asian and white participants. These data suggest that despite smaller bone geometry in Asian youth, bone adapts other parameters to maintain bone strength. With the exception of 11% greater trabecular separation in Asian females, there were no significant differences in trabecular microarchitecture between Asian and white adolescents and young adults.[186] Although data at weight-bearing sites are limited, one previous study reported similar bone outcomes between Asian and white boys and girls at the distal tibia, but smaller Ct.Ar and greater Ct.BMD at the tibial shaft (by pQCT) in Asian girls and smaller Ct.Ar in Asian boys compared with their white peers.[151] Genetics likely drives these ethnic-specific phenotypes; however, ethnic differences in timing of maturation may also explain the smaller bone geometry of Asian children who tend to mature earlier than their white peers.[187] Differences in modifiable lifestyle factors such as lower calcium intake and lower participation in PA in Asian compared with white children may also contribute to ethnic differences in bone accrual and geometry.[188]  51  Ethnic differences in bone strength and its determinants are also apparent in black compared with white children. These differences may contribute to a lower fracture incidence in blacks (half that of their white peers).[184] For example, after adjusting for tibia length and leg muscle MCSA, early- and peri-pubertal black children had 2-8% greater SSIp, Ct.BMD and Tt.Ar (by pQCT at the tibia shaft) compared with white children.[189] This study and others[190,191] reported greater bone strength at the tibial diaphysis and metaphysis during childhood and adolescence in blacks compared with their white peers and suggest that these differences are already present in the early stages of puberty. Ethnic differences in markers of bone turnover reflect greater bone strength in black children who have greater levels of bone formation (osteocalcin and bone-specific alkaline phosphatase) and lower bone resorption (N-terminal telopeptide) markers, despite lower indices of modifiable factors such as vitamin D, dietary calcium and PA, compared with white children.[189] Studies that specifically investigate the influence of lifestyle and environmental factors on bone strength accrual among different ethnic groups are warranted.   1.2.5.4 Calcium and vitamin D  The dietary nutrients, calcium and vitamin D, impact development and maintenance of skeletal health. Calcium is the most abundant mineral in the human body and is stored primarily (99%) in bones and teeth. Calcium supports structural integrity of the skeleton and regulates metabolic function.[192] Vitamin D stimulates bone matrix formation and regulates calcium metabolism and absorption in concert with parathyroid hormone, such that intestinal calcium absorption doubles in the presence of adequate vitamin D.[192] Both nutrients are critical for skeletal health; inadequate intake or absorption of calcium and vitamin D during growth can result in rickets.[192] Current North American dietary standards recommend 1000 to 1300 mg/day of calcium and 600 IU daily of vitamin D for children and adolescents.[193]   Although calcium is the main building block of bone, whether or not calcium supplements are effective for bone accrual is equivocal. For example, a meta-analysis of 21 randomized controlled trials (RCTs) concluded that among children with normal baseline dietary calcium, supplemental calcium had little impact on total body BMC (by DXA). However, in children with low baseline intakes of calcium, a regimen of supplemental calcium significantly 52  increased total body and lumbar spine BMC.[194] In one of the longest trials of calcium supplementation to date, 3-years of calcium supplementation resulted in significantly greater increases in aBMD at the radius and lumbar spine in pre-pubertal, but not pubertal twins, compared with non-supplemented twin controls.[195] Thus, there may be a window of opportunity during pre-puberty when bone more positively responds to supplemental calcium.  We know less about the effects of calcium supplementation on bone strength and geometry. Few trials used pQCT to evaluate the bone strength response to supplemental calcium during growth. To my knowledge, no studies examined calcium supplementation independent of other lifestyle or dietary interventions (i.e., PA or vitamin D). A trial of pre-pubertal children (mean age 10 years) combined calcium supplementation and PA in an RCT across 12 months. Importantly, children in the study consumed recommended daily intakes for calcium at baseline.[196] Following daily calcium supplements (500 mg), Tb.BMD at the distal tibia (by pQCT) increased 5% more in non-gymnasts (defined as low-PA group) compared with non-supplemented non-gymnast controls, whereas no differences were observed in gymnasts (high-PA group).[196] No differences in bone density, geometry or strength were observed between calcium supplemented and non-supplemented groups at the distal radius or at the radial or tibial shaft for either non-gymnasts or gymnasts. Thus, supplemental calcium may interact with weight-bearing PA in pre-pubertal children, such that supplementation may not benefit those already engaging in high-impact PA. The bones of these athletes may have already adapted their density to high mechanical demands imposed through gymnastics.[196] No study has examined the interaction between calcium supplementation and PA in the adolescent skeleton. As adequate vitamin D is necessary for optimal calcium absorption, several trials examined effectiveness of combined supplemental calcium and vitamin D on bone mineral (trabecular BMC and BMD) and strength accrual. In one trial, after 6-months of supplemental calcium (800 mg/day) and vitamin D (400 IU/day), peri-pubertal female identical twins (9-13 years) demonstrated 5% greater gains in Tb.BMD and Tb.Ar at the distal tibia and radius (by pQCT) and 6% greater increases in Ct.Ar at the tibial shaft compared with the non-supplemented control twin group; however, the latter did not translate into greater gains in SSIp at the tibial shaft.[197] Similarly, a 12-month calcium (800 mg/day) and vitamin D (400 IU/day) supplementation trial in early and peri-pubertal girls (age 12 years) girls found a significant intervention effect for trabecular BMC and BMD at the distal tibia (by pQCT; no other sites were 53  evaluated).[198] Thus, combined supplementation of calcium and vitamin D may enhance trabecular volume at metaphyses and bone geometry at diaphysis. It is unclear how these gains influence bone strength or whether gains persist once supplementation ceases. Vitamin D stimulates bone matrix formation, calcium and phosphate absorption in the small intestine, reabsorption of renal calcium and mobilization of calcium in bones.[199] Upwards of 90% of vitamin D is derived from the photo conversion of 7-dehydrocholesterol in the skin by solar UVB radiation, hence its termed the ‘sunshine’ vitamin.[200] Daily synthesis of 400 IU vitamin D in children and adolescents is possible through casual exposure of the face and hands to sunshine at midday in all latitudes during part of the year, but not in higher latitudes for the entire year.[201] Thus, factors such as living in northern or southern latitudes with low UVB production in winter months and clothing that covers most of the skin surface area can contribute to low vitamin D levels. Low serum vitamin D is prevalent in children and adolescents throughout the world.[202] The most recent estimates from the Canadian Health Measures Survey (CHMS; a representative sample of 5600 Canadians from 15 sites around the country) suggested that 11-29% of Canadian children and adolescents have serum vitamin D below recommended levels (< 50 nmol/L). [203]  While vitamin D deficiency increases risk for conditions such as rickets, supplements for children with adequate serum vitamin D may not enhance their bone mineral accrual. In a recent Cochrane systematic review of 6 placebo-controlled RCTs, which varied in geographical location and included higher latitude countries, supplemental vitamin D did not increase whole body BMC or aBMD of the forearm, hip and spine (by DXA).[204] However, when children who were deficient in serum vitamin D at baseline (< 35 nmol/l, with > 50 nmol/L considered adequate) received supplemental vitamin D, whole body BMC and lumbar spine aBMD increased significantly compared with a placebo group.[202,204] In contrast, a recent RCT supplemented vitamin D deficient post-menarcheal girls (< 37.5 nmol/liter) and found no effect on bone density, geometry or strength (by pQCT) or muscle force or power (by jumping mechanography) after 1-year.[205] Future trials using 3D imaging techniques would help to clarify the maturity-specific influence of supplemental vitamin D on bone strength.    54  1.2.5.5 Muscle force  In section 1.2.2.2.2, I highlighted the influence of muscular forces on bone strength development and maintenance. Muscle contractions impose the greatest mechanical challenge on bone (stresses several fold greater than body weight alone) and drive bone adaptation.[57,60] The functional model of bone development contends that bone continually adapts to mechanical loads induced by muscular strain by adjusting bone strength and its determinants (up or down) to maintain strains within safe limits.[57,60] Given the strong influence of muscle on bone development, growth-related changes in bone parameters should be considered in the context of the functional muscle-bone unit. If the central tenet of the functional model of bone development is true, muscle development should precede bone development. This was observed in the 14-year University of Saskatchewan PBMAS. For example, peak total body muscle mass accrual (surrogate of muscle force; by DXA) preceded peak BMC accrual (by DXA) by approximately 6 months in girls and 4 months in boys.[206] In a subsequent analysis, peak total body muscle mass velocity occurred 2 to 4 months ahead of peak bone CSA and estimated bone strength (section modulus) velocity at the narrow neck and femoral shaft (by HSA; Figure 1.22)[207] Together these findings suggest that enhanced muscle mass promotes bone adaptation. In contrast, in a 7-year longitudinal study of Finnish girls, MCSA (surrogate of muscle force; by pQCT) peaked prior to BMC and BMD at the tibia shaft, but lagged behind total and cortical bone CSA (all bone measures by pQCT).[161] Thus, the muscle-bone relationship might not function uniformly across skeletal sites. In section 1.2.7.2.2, I provide further evidence from longitudinal pediatric bone studies that suggests muscle forces mediate the PA-bone relationship.[208,209] 55  Figure 1.22. Illustration of tissue velocity curves for muscle mass, A) cross-sectional area (CSA) and B) section modulus (Z) at the femoral shaft aligned by maturational age (years from age at peak height velocity). The solid vertical line represents the maturational age when peak tissue velocities occurred. *Indicates significant difference between age of peak muscle velocity and peak CSA velocity. **Indicates a significant difference between age of peak muscle velocity and peak Z velocity. Reprinted from Jackowski et al.,[207] with permission from Elsevier.  Direct assessment of muscle force is not possible using non-invasive techniques. However, dynamometry and mechanography provide reliable estimates of muscle force. Hand-held dynamometers are an easy and reliable approach to measure maximal isometric grip force,[210] which is strongly associated (r = 0.80-0.90) with CSA and bone strength (BSI) at the distal radius (by pQCT). [211] In the lower limbs, jumping mechanography is commonly used to assess peak muscle force (N) and peak muscle power (W) during single- and two-legged jumps, respectively. However, to my knowledge no study has examined associations between lower limb muscle force or power (by mechanography) and bone strength in healthy children or adolescents (by pQCT or HR-pQCT). Given the high cost of the force platform, mechanography may not be feasible for field-based measures. Vertical jump height, on the other hand, is a simple 56  test used to estimate peak muscle power in conjunction with validated prediction equations.[212] Estimated muscle power was significantly associated (r = 0.54-0.78) with bone strength at the distal (4%; BSI by pQCT) and midshaft site (66%; SSIp by pQCT) of the tibia in adolescents (mean age 17 years).[213] Thus, grip strength and vertical jump are easy and cost-effective tools to assess muscle force and power in the laboratory or in the field. When functional measures of muscle force are not available, muscle mass (g) and MCSA (mm2) are frequently used as surrogates. Muscle mass is derived from DXA whole body scans based on attenuation of X-rays through muscle tissue that is assumed to be of fixed density[214] Measures of muscle mass were highly correlated (r = 0.77) with leg muscle power in adolescent girls.[215] As with bone analyses using pQCT (described in Section 1.2.3.2.1), MCSA is derived using density thresholds that separate muscle from bone and fat. MCSA is highly correlated with estimated muscle power (using vertical jump height and prediction equation; r = 0.70) and bone strength at the distal tibia (r = 0.56-0.66; BSI by pQCT) and tibial shaft (r = 0.68; SSIp by pQCT) in adolescents.[213]  1.2.6 Physical activity and sedentary time  In this section, I describe how PA and sedentary time are assessed and summarize the current literature regarding the influence of PA and sedentary time on bone strength and its determinants in children and adolescents.  1.2.6.1 Measurement of physical activity  PA is defined as any bodily movements expending energy.[3] Current Canadian PA guidelines recommend that children and adolescents (5 to 17 years) engage in 60 min/day of moderate-to-vigorous PA (MVPA) to achieve health benefits, while adults (18 to 64 years) should achieve 150 min of MVPA every week.[216] Guidelines for children and adolescents were based on a systematic review of the health benefits of PA in children and adolescents (including the influence of PA on aBMD by DXA).[217] Guidelines recommend that youth engage in muscle and bone-strengthening activities that use major muscle groups at least 3 days/week.[216]  57  Various tools are used to assess PA in children and youth. Measurement techniques include subjective administered or self-report questionnaires and direct monitoring devices, such as pedometers or accelerometers. Questionnaires are often the tool of choice as they are cost-effective, easy to administer and are have low participant burden.[218] However, self-report questionnaires are subject to recall bias. Thus, while they provide behavioural information regarding PA (setting and type of PA), they do not adequately capture PA intensity and duration. In contrast, objective tools (e.g., accelerometers) measure PA intensity, frequency and duration and can be time-stamped for time-of-day analyses. Further, ground reaction forces were strongly correlated with raw acceleration output in adults (r = 0.85)[219] and in children and adolescents (healthy children and those with osteogenesis imperfecta type 1, age 6-21 years; r = 0.96).[220] This suggests accelerometers are an appropriate tool with which to estimate mechanical loads associated with weight-bearing PA. However, high cost (i.e., $200-400 per unit), low wear compliance and inability to capture certain types of activity (i.e., swimming and biking) may limit use of accelerometry. Thus, investigators should consider the study aims and feasibility when choosing a PA measurement tool.  1.2.6.1.1 Self-report questionnaires to assess physical activity  Self-report questionnaires most often rely on a participant’s ability to recall or report their PA. Participants tend to overestimate their PA in self-report questionnaires, compared with direct measures.[218] Structured activities may be easy to recall, but unstructured activities that make up most of daily PA are difficult to quantify. A recent systematic review concluded that no currently available PA questionnaire for children and adolescents (61 reviewed) were of both acceptable validity and reliability (based on an intraclass correlation coefficient (ICC) of > 0.70).[221] The Physical Activity Questionnaires for Children (PAQ-C; 8-14 year olds or grades 4-8) and Adolescents (PAQ-A; 14-18 year olds or high school students) were designed for the Saskatchewan PBMAS and are widely used to estimate MVPA.[222,223] In both questionnaires, children/youth recall their participation in activities during the past 7 days. The PAQ-C is a nine-item questionnaire with the first question providing a checklist of common sport or leisure PA; the remaining questions are segmented by time-of-day (e.g., at lunch, after school) or day-of-the-week (e.g., last weekend). Items 1-9 are scored on a 1 (low PA) – 5 (high PA) scale; the 58  summary score is the average sum of nine questions.[222] Our research group modified the PAQ-C for HBSIII to include an estimate of dose (added time spent per activity session (item 1), involvement in extracurricular activities and number of nights of organized sport PA per week)[188] and perception of PA involvement on each day of the week (5-point scale ranging from none to very often). The PAQ-C demonstrated good test-retest reliability in 9-15 year old boys (r = 0.75) and girls (r = 0.82) using the summary score across seasons (r = 0.80 for average of two or three responses in fall, winter and spring).[222] However, the PAQ-C summary score demonstrated weak (r = 0.25; mean age 11 years)[224] to moderate (r = 0.39; participants in grades 4-8)[225] agreement with MVPA and activity counts respectively, by accelerometry. The PAQ-A is almost identical to the PAQ-C, with the exception that it does not include a question regarding PA during morning recess. The PAQ-A summary score was moderately correlated with accelerometry-derived MVPA (r = 0.49) in 14-year olds.[226] Despite limitations of self-reported PA, questionnaire-based assessment remains prevalent due to its ease of administration and ability to provide contextual information.  1.2.6.1.2 Accelerometry to assess physical activity  Accelerometers are non-invasive devices that record frequency, duration and intensity of everyday activities. Validation studies in children and adolescents demonstrate high reproducibility, validity and feasibility.[227] Thus, accelerometry is the preferred method to assess PA in children and adolescents.[228,229] Accelerometers are small (approximately the size of a matchbox), light (< 30g) and robust devices typically attached to a band worn around the hip, although new models can be worn on the wrist. Some devices (triaxial accelerometers) measure motion in three planes: vertical, horizontal and perpendicular. For my thesis, I focus on ActiGraph GT1M accelerometer, as this model was available at the Centre for Hip Health and Mobility. The GT1M is worn at the hip and assesses motions in the vertical plane only. The sensors inside accelerometers detect acceleration of the body and produce an analog voltage proportional to the magnitude of acceleration. The analog signal is digitized (sample rate of 30 Hz) and filtered (bandwidth of 0.25 to 2.5 Hz; excludes non-human movement) to produce values known as 'counts'.[230] Counts are summed over user-specified intervals known as ‘epochs’ and stored in the unit’s memory.  59  Accelerometers assess PA in short measurement intervals (e.g., every 3 sec) over long periods of time (e.g., months). Early accelerometer models had a limited memory; thus, an epoch of 1-min was common. Newer devices have greater storage capacity and permit shorter epochs (e.g., 1-sec, 3-sec, 15-sec), which more accurately assess the intermittent nature of children’s PA.[231] The potential for misclassification of PA increases in concert with epoch length. High-intensity PA may be underestimated when averaged over longer epochs, such that short bouts of high-intensity PA may be combined with bouts of low-intensity PA within the same epoch. To minimize misclassification, as short an epoch length as possible should be used. Given the importance of high-intensity PA for bone adaptation, a short epoch is particularly relevant for examining the relationship between PA and bone parameters. Further, accelerometry data can be easily re-integrated into longer epochs during post-acquisition data analysis, as needed.[228] The number of days and h/day an accelerometer must be worn to accurately assess habitual PA is an important consideration. In children and adults, three to five days of monitoring are recommended to achieve reasonable reliability (r = 0.70-0.80).[232,233] Further, a minimum of 10 h/day of wear time is recommended to define a valid day, as this length minimized the effects of varying day length on PA outcomes in a study of over 6000, 11-year olds.[232] Following the monitoring period, accelerometer data must first be screened for non-wear time. Much debate has concerned periods of non-wear time versus true sedentary time (e.g., reading, watching TV). This classification is crucial to accurately assess PA and sedentary time. Non-wear time can be defined visually during data analysis. This is conducted in combination with a PA diary filled out by the participant or parent that notes times when the accelerometer was worn (on-off times). However, this process is time consuming, subjective and inaccurate. Alternatively, some assumed that a certain number of minutes (e.g., 60-min) of consecutive ‘0’ counts represents non-wear time. Given the sensitive nature of accelerometers, even small motions create values greater than zero.[234] A non-wear criteria defined as 20-min or longer of motionless data was recommended for youth, as 17.5 min, on average was the longest bout of motionless data over 7 days of monitoring in 115 youth (8-13 years).[235] Other commonly used non-wear criteria in youth ranged from 10-180 min of consecutive 0 counts, with the option of allowing for 1 to 2 min of counts between 0 and 100 during that period.[236] Although consensus has not been reached, a criteria of 60-min of consecutive 0 counts without interruptions was 60  recently suggested for use in children and youth, based on the most realistic number of non-wear periods per day observed in a more robust sample of 1000, 9-13 year olds.[237] The 60-min criteria resulted in a maximum of 4 non-wear periods per day, compared with an unrealistically high number of non-wear periods for the 20-min criteria (maximum of 10).[237] However, in future, non-wear criteria must be validated against direct measures such as video or direct observation. Raw accelerometer counts are unit-less. Thus, they must be converted into a value that has meaning for the user. Accelerometer calibration studies used energy expenditure (indirect calorimetry, METs) or direct observation to develop count thresholds, known as cut points. Cut points are age- (and often study-) specific and correspond with PA intensity.[238] Thus, PA estimates are not comparable across studies that used different cut points. For example, five commonly used PA cut points for youth ranged from approximately ≥ 2200 cpm to ≥ 3600 cpm for moderate PA, to ≥ 4000 to ≥ 8200 cpm for vigorous PA. Accurate classification of MVPA against energy expenditure (using indirect calorimetry) was poor for 3 of the 5 cut points (those with higher cut points). This was based on low sensitivity (high false-negative rate) of MVPA that was misclassified as low-intensity PA.[239] Both Freedson[240] and Evenson[238]cut points demonstrated excellent classification accuracy for MVPA (receiver operating characteristic – area under the curve = 0.90) and fair accuracy for light PA (receiver operating characteristic – area under the curve: 0.69-0.70, respectively).[239] However, Evenson cut points are recommended because the MVPA cut point performed equally well among children and youth at all ages and all levels of PA intensity demonstrated acceptable classification accuracy.[239] Expressing PA as min/day is a convenient way to convey recommendations to the public and practitioners and to assess compliance with PA recommendations. However, this is not without bias as expression of PA is influenced by duration of accelerometer wear time. Several approaches eliminate wear-time biases and facilitate comparisons between participants who had different wear times. For example, wear time can be accounted for: 1) by expressing PA relative to wear time,[234] 2) in regression-based analyses by including wear time as a predictor alongside PA, or 3) by using a residuals approach, which obtains the residuals from regressing the PA variable of interest on wear time.[241] While accelerometers effectively capture frequency, duration and intensity of PA and eliminate reporting bias, they have limitations. Accelerometers cannot account for increases in 61  energy expenditure associated with walking up an incline or carrying a load.[242] Further, they do not accurately measure activities that occur mainly in the horizontal plane (e.g., skating). Finally, most accelerometers are not waterproof so cannot measure energy expenditure associated with swimming or other water sports.  1.2.6.2 Measurement of sedentary time  Sedentary time is activity defined by a low energy expenditure < 1.5 METs in a seated or reclined posture.[243] Parental or self-reported TV viewing, computer use, video games, phone use and reading are traditionally used to assess sedentary time.[244] However, more recently accelerometers have been used to objectively assess sedentary time in children and adolescents.  1.2.6.2.1 Self-report questionnaires to assess sedentary time  Unlike objective methods, self-report measures of sedentary time provide type and context to the behaviour.[244] However, few self-report measures of sedentary time demonstrated acceptable reliability and validity in children and adolescents (based on intraclass correlation coefficient > 0.70).[245] While many questionnaires demonstrated acceptable reliability (similar test-retest scores), most demonstrated poor construct validity compared with objective measures.[245] For example, self-reported sedentary/screen time in 6-11 year olds (n = 878) in the CHMS was weakly correlated (r = 0.17) with accelerometry-derived sedentary time measured over 4 to 7 days.[16] The low construct validity may be partially due to a mismatch between constructs addressed by questionnaires and comparison measures; questionnaires ask about specific leisure-time behaviours (e.g., time spent watching TV or playing video games), while accelerometers typically assess sedentary time over the entire day.[245]   1.2.6.2.2 Accelerometry to assess sedentary time  Accelerometery-derived sedentary time is more reliable and valid compared with self-reported sedentary time.[244] A cut point of < 100 cpm is commonly used to determine total sedentary time.[238,239] However, accelerometers cannot differentiate between sitting and standing 62  with minimal movement (standing by definition is not sedentary time). As mentioned previously, accelerometers do not provide context for a sedentary activity. Considering strengths and limitations of approaches to measure sedentary time, a combination of self-report questionnaires and objective monitors may best describe children’s sedentary behaviours.   1.2.6.3 Sex- and age-related differences in physical activity and sedentary time  Boys are more active than girls during childhood and adolescence and PA declines from childhood into adulthood.[246,247] Two large nationwide surveys conducted in Canada and the United States examined prevalence of PA and sedentary time in children and youth. From 2007 to 2009, CHMS collected PA and sedentary time data in over 1600 children and adolescents at 15 measurement sites across Canada using parent-reports and accelerometry (Actical).[248] From 2003 to 2006, National Health and Nutrition Examination Survey (NHANES) collected accelerometer (ActiGraph) data from almost 1800 children and adolescents in the United States.[249] In both nationwide surveys, boys accumulated more MVPA compared with girls throughout childhood and adolescence.[248,249] MVPA decreased with age in both sexes to a similar extent.[248,249] From CHMS, children (6-10 years) engaged in approximately 1 h/day of MVPA (69 and 58 min/day for boys and girls, respectively), while adolescent boys and girls (11-14 years) accumulated less MVPA (59 and 47 min/day, respectively).[248] Compared with CHMS, NHANES reported slightly greater values for MVPA during childhood (6-11 years; 95 and 75 min/day for boys and girls, respectively), but lower values of MVPA during adolescence (12-15 years; 45 and 25 min/day respectively).[249] Based on estimates of MVPA, only 22% of Canadian boys and 24% of American boys (age 6-19 years) met MVPA recommendations of 60 min/day (using a 5 of 7 days criterion).[248,249] Just 11% of Canadian girls and 15% of American girls (age 6-19 years) achieved the recommended 60 min/day of MVPA (using a 5 of 7 days criterion).[248,249] Both studies also assessed sedentary time using accelerometry. Those in the Canadian cohort (6-19 years) engaged in 8.6 h/day of sedentary time, on average. Sedentary time increased with age from approximately 7 h/day during childhood to 9 h/day during late adolescence.[248] Sedentary time was slightly lower in the American cohort (7.2 h/day; 6-19 years), on average, but increased by approximately 2 h/day (6 h/day to 8 h/day) from childhood to late 63  adolescence.[250] Sedentary time did not differ between Canadian boys and girls between 6 and 14 years of age, but was higher in girls thereafter (by 30 min/day).[248] In contrast, American girls were significantly more sedentary, compared with boys, throughout childhood and adolescence (by approximately 12 min/day).[250] Of note, the Canadian study did not control for differences in accelerometer wear time between participants, while the American study did. Thus, differences in wear time may confound comparisons of PA and sedentary time between studies. However, CHMS data confirm that boys are more active than girls across adolescence, MVPA decreases across adolescence in both sexes and sedentary time increases in both sexes from childhood to early adulthood.  1.2.7 Influence of physical activity and sedentary time on bone strength development  In this section, I briefly review current literature regarding how PA and sedentary time influence bone strength and its determinants during adolescent growth. I first highlight intervention studies, followed by observational studies in athletic and non-athletic cohorts and finish by presenting the link between PA during childhood and adolescence and adult bone outcomes. The positive influence of PA on bone development is summarized in several excellent reviews.[12-15,251] As discussed in section 1.2.2.2.3, bone can adapt its strength in response to mechanical stimuli during growth through several mechanisms: 1) periosteal apposition can increase bone CSA; 2) periosteal apposition and reduced endocortical resorption can increase Ct.Th; 3) modifications to cortical and trabecular microarchitecture (i.e., increased Tb.Th or Tb.N or decreased Ct.Po) can increase tissue density.[11,69] Specifically, there is strong evidence to suggest that pre- and early-puberty may provide a ‘window of opportunity’ during which the skeleton is particularly responsive to loads associated with weight-bearing PA.[144,145] In contrast, we know less about the mechanisms underpinning bone’s adaptation to PA in later adolescence.[10,15,252-255] This may be due, in part, to the use of imaging systems such as DXA, which cannot capture subtle adaptations in bone strength and its determinants and to the complex and extremely variable nature of adolescent growth.   64  1.2.7.1 Intervention studies of physical activity   Targeted bone-loading interventions were traditionally implemented in elementary schools as schools reach large numbers of children from diverse backgrounds. Effective PA interventions incorporated dynamic, high-impact activities that were of short duration, elicited ‘unusual’ strains and were separated by rest periods, thus mirroring the principles derived from the animal literature.[44] Length of PA intervention across studies ranged from 28 weeks to 2 years.[15] Most used DXA (11 of 14 RCTs) to monitor exercise-related gains in bone mass.[10,15] Importantly, children assigned to exercise intervention groups gained significantly more bone mass (1-6%) at the spine and hip compared with children in control groups, on average.[10] One of the longest school-based RCTs conducted to date is the UBC HBS. Children (aged 9-11 years) who attended schools randomized to the exercise group participated in 10-12 min of high-impact jumping activities, 3 times per week for 7-months each year for two school years.[256,257] After the first school year, both groups demonstrated gains in BMC, but girls and boys who attended intervention schools demonstrated significantly greater gains in BMC and aBMD at the femoral neck and lumbar spine compared with children attending control schools. However, in girls, the intervention effect was only apparent in those who were early pubertal (Tanner stage 2 or 3) at baseline.[258,259] After two school years, significantly greater gains in femoral neck (5%) and lumbar spine (4%) BMC were observed in intervention girls[256] and in femoral neck (4%) BMC in intervention boys.[257] The HBS and similar interventions[254,260-262] highlighted that a simple exercise program, which requires very little time in the school day may enhance bone mass accrual.  Animal models demonstrated that the skeleton adapts to mechanical loading by adding bone to the periosteal surface of long bone shaft sites where strains are the greatest.[44,263] Although small in magnitude, these subtle structural adaptations confer dramatic increases in experimentally measured bone strength.[90,91] As mentioned previously, DXA cannot capture these bone adaptations to PA. Only 11 intervention trials conducted in the last decade used imaging tools (e.g., HSA, pQCT or MRI) able to assess exercise-induced adaptations in bone geometry and strength.[15] I review key findings from these studies below.  In the HBS, the greater gain in femoral neck BMC in early pubertal girls in the intervention group was associated with a 4% greater increase in femoral neck bone strength 65  (section modulus, HSA) compared with controls.[69] This strength gain was attributed to greater increase in CSA and reduced endocortical expansion, leading to a thicker cortex in the intervention group. In contrast, intervention-related gains in femoral neck bone strength were only observed in boys after the second year of the trial.[257] The apparent sex difference in the timing of structural adaptations to the HBS intervention may be related to maturity status. At baseline, 60% of girls were early-pubertal, whereas most boys were pre-pubertal. A later adaptation in bone strength in boys may be a result of advanced maturity (77% advanced to early- or peri-puberty) over the second year of the study and/or the prolonged intervention. These findings suggest that early-puberty may be a window of opportunity for femoral neck bone strength adaptations. As there were no differences in strength gains at the total proximal femur between groups during pre-puberty, a more intense exercise intervention may be necessary to confer beneficial structural adaptations at this larger site.  The influence of maturity status on bone structure adaptations to PA may also vary with skeletal site. To illustrate, Action Schools! BC involved short bouts of classroom-based exercise (including ~3 min/day of jumping) during a 16-month intervention period. Girls attending intervention schools that reported high compliance (≥ 80%) demonstrated 5% greater gains in femoral neck bone strength (section modulus, HSA) compared with girls attending control schools.[264] Conversely, an intervention effect was not observed at the distal or shaft site of the tibia (by pQCT) in girls, most of whom were early-pubertal at baseline.[265] Despite significantly greater gains in lumbar spine and total body BMC in boys attending intervention schools, there were no structural differences at follow-up compared with boys (pre- and early-pubertal boys pooled) attending control schools.[264] However, there was a significant group by maturity interaction, such that pre-pubertal boys at baseline in the intervention group demonstrated a 4% greater gain in distal tibia bone strength (BSI) compared with controls. The group by maturity interaction suggested the appendicular skeleton may be more responsive to loading during pre-puberty.[265] Alternatively, maturity-specific findings may be explained by lower participation in self-reported leisure time weight-bearing PA in pre-pubertal (6 h/week) compared with early-pubertal boys (8 h/week).[265] Thus, consistent with the cellular accommodation theory of mechanoadaptation, PA interventions may be less beneficial to those already engaging in high levels of weight-bearing PA. A more frequent and intense intervention may also be necessary to elicit an osteogenic response in the more active, mature group. 66  Not all intervention studies observed differences in bone outcomes between intervention and control groups. For example, pre-pubertal girls who completed seven months of a drop-jumping program using their non-dominant leg (3 times/week) did not demonstrate greater gains in bone strength at the mid-femur (by MRI) compared with girls in the control group.[266] As the jumps performed were unidirectional, of low magnitude (14-28 cm), over a short time period (28 weeks) and in a small sample (n = 13 in each of control, low- and high-impact groups), a more robust sample size with greater dynamic loads may be necessary to detect an osteogenic response. School-based intervention trials are complex and challenging. Success depends largely upon participant and teacher compliance in intervention schools, activities conducted within control schools during the study period and activity levels of all participants at baseline – all are beyond a researcher’s control. These and other factors (e.g., type, intensity, frequency and duration of the intervention, imaging tools used, scan acquisition and analysis procedure) represent significant heterogeneity across school-based intervention trials.[15] Nevertheless, the bone response to skeletal loading appears to be sex- and maturity-specific. Convincing evidence supports the role of high-impact exercise for augmenting bone mass (but less so structure and strength) in pre- and early-pubertal children. Further work is needed to better understand bone structural and microarchitectural adaptations to loads associated with weight-bearing PA and the optimal dose necessary to elicit meaningful bone health benefits.  1.2.7.2 Observational studies of physical activity  Observational studies of habitual PA and athlete groups subjected to different loading conditions, represent a large body of evidence supporting a positive association between PA and bone development during growth. These studies traditionally relied on DXA to image bone; all demonstrated significant bone mass benefits in children who participated in weight-bearing sports such as gymnastics, tennis and running compared with non-athlete groups.[267-270] Similarly, leisure-time PA was a significant predictor of bone mass accrual in girls and boys across maturity.[133,271,272] With increased accessibility to 3D imaging tools, researchers are now gaining insight into bone structural advantages associated with weight-bearing PA. Thus, in this section I focus on observational studies that employed 3D imaging tools. 67  1.2.7.2.1 Athletic populations   Racquet Sports: Athletes in racquet sports such as tennis and badminton provide a unique model for investigating bone adaptation to loading. Within-subject comparisons of the playing versus non-playing arms controls for confounding factors such as genetics, hormones and diet. The seminal cross-sectional DXA study by Kannus and colleagues[267] paved the way for studies that used more sophisticated imaging tools. They reported side-to-side differences in BMC in the playing versus non-playing arms of female racquet sport players were significantly greater compared with controls (9-16% vs. 3-5%). However, of greater interest, players who initiated training prior to menarche had side-to-side differences nearly twice that of players who initiated training after menarche. This finding raised the possibility that skeletal benefits of weight-bearing PA are maximized during pre-menarcheal years.[267] The bone mass advantage in the playing arm of female racquet sport athletes was also associated with significant bone strength benefits. To illustrate, bone strength (polar second moment of area) at distal and shaft sites of the humerus (by MRI) in young female tennis players was 11-23% greater in the playing arm compared with the non-playing arm.[273] Similarly, side-to-side differences in BMC, Ct.Ar, Tt.Ar and strength (BSI; all by pQCT) at distal and shaft sites of the radius and humerus were 8-22% greater in female racquet sport athletes compared with controls.[274] As in the Kannus et al. study, side-to-side differences in bone geometry and strength were double in magnitude in women who began racquet sport training prior to menarche compared with women who began training after menarche (Figure 1.23).[274] Finally, in the only prospective study of racquet sport athletes conducted to date, 12-month changes in bone geometry (Tt.Ar and Ct.Ar by MRI) were significantly greater among pre- and peri-pubertal female competitive tennis players compared with post-pubertal players.[275] These findings provide further support for a ‘window of opportunity’ during pre- and early-puberty when the skeleton is most responsive to mechanical stimuli. 68  Figure 1.23. Average side-to-side differences in humeral midshaft total bone cross-sectional area (CSA), cortical CSA, cortical bone mineral density (BMD) and bone strength index (BSI) between the playing and nonplaying arm in female racquet sport athletes as measured with peripheral quantitative computed tomography (pQCT). The solid line represents the playing arm (or dominant arm in controls) and the dotted line represents the nonplaying arm. Adapted from Macdonald et al.,[14] with permission from Future Medicine, Ltd.  In racquet sport athletes, adaptations in bone geometry at shaft and distal sites (Tt.Ar and Ct.Ar) during pre- and early-puberty were attributed to bone accrual on the outer bone surface (periosteal expansion) as opposed to endocortical expansion or contraction.[273-275] In contrast, training initiated after puberty was associated with greater bone contraction or decreased expansion on the endocortical surface, conferring little benefit to bone bending strength.[273] Similar sport-related gains in bone geometry and strength were observed in the playing-arms of young adult men, all of whom began training during childhood.[276,277] Gymnastics: Artistic gymnastics imposes an extremely high mechanical stimulus on the skeleton (ground reaction forces greater than 10 times body weight).[278] Thus, gymnasts represent a unique population within whom to examine the effects of intense loading on bone. However, no longitudinal studies of gymnasts utilized pQCT or HR-pQCT to examine bone strength and geometry. A 4-year DXA follow-up study of recreational gymnasts (aged 4-9 years at baseline) demonstrated 3% greater total body BMC, 7% greater femoral neck BMC[268] and 3-6% greater CSA at all three femoral neck sites compared with their non-gymnast peers.[279] Gymnasts also 69  showed 6-7% greater estimated bone strength (section modulus by HSA) at the narrow neck and intertrochanter compared with their non-gymnast peers.[279] Conversely, ex-recreational gymnasts (most of whom ceased participation between first and second measurement) did not demonstrate a bone advantage compared with non-gymnasts.[268,279] This suggests that continued participation is required to maintain benefits associated with recreational gymnastics during pre-puberty. Gymnasts consistently demonstrated greater bone strength (by pQCT) in the upper and lower limbs compared with their non-gymnast peers.[103,165,279-281] As with racquet sport athletes, this bone strength advantage at shaft sites was due in part to enhanced bone geometry that conferred strength to long bones. For example, pre-pubertal elite gymnasts (aged 5-11 years) demonstrated 9% greater estimated bone strength (SSIp by pQCT) at the radial shaft compared with non-gymnast controls. This advantage was likely driven by reported 5-7% greater Tt.Ar and Ct.Ar in the gymnasts.[165] Similar advantages were reported in 6-11 year old non-elite gymnasts (<16 h/week gymnastics; Figure 1.24)[103] and in 4-9 year old recreational current and ex-gymnasts (at least 45 min/week of gymnastics).[280] The distal site also demonstrated greater bone strength; however, adaptations were due to increased bone density. To illustrate, recreational gymnasts and ex-gymnasts had similar Tt.Ar. However, the 6-8% greater Tt.BMD in recreational gymnasts contributed to 22-25% greater bone strength (BSI) at the distal radius compared with non-gymnast controls.[280] Collectively, these findings suggest that participating in gymnastics at a recreational level confers bone health benefits during pre-puberty. 70  Figure 1.24. Illustration of a) bone geometry (total bone area) and b) estimated bone strength (polar strength-strain index, SSIp) at the proximal radius (66% site) measured with peripheral quantitative computed tomography (pQCT) in pre-pubertal girls. Non-gymnasts (Non-Gym), low-training volume gymnasts (Low-Gym) and high-training volume gymnasts (High-Gym). *Indicates significantly different from Non-Gym. Bars represent 95% confidence intervals. Adapted from Burt et al.,[103] with permission from Springer.  Bone geometry and strength benefits associated with gymnastics training may persist into late adolescence. For example, post-menarcheal girls (n = 16, mean age 17 years) who participated in gymnastics during early-puberty (> 5 h/week for at least 2 years) but stopped training 1-year post-menarche, on average, had 19% greater Tb.BMD and 25-26% greater Ct.Ar and Tt.Ar. This conferred 34% greater estimated bone strength (BSI) at the distal radius compared with non-gymnasts.[281] Similarly, Tt.Ar and Ct.Ar were 22-33% greater in ex-gymnasts at the radial shaft. The larger bone area was associated with 46% greater bone strength (SSIp) compared with non-gymnasts.[281] Longitudinal studies are needed to confirm these findings. Other Sports: Studies in adolescents and young adults that used HR-pQCT support a benefit of high-impact PA on bone strength and microarchitecture. For example, late adolescent and young adult female athletes (age 14-21 years; at least 20 miles of running or 4 h/week of aerobic weight bearing exercise) demonstrated greater Tb.Ar, Tt.Ar and F.Load at the distal tibia compared with non-athletes.[282,283] Similarly, late adolescent and young adult male and female athletes who participated in high-impact sports (skiers and soccer players) demonstrated 71  significantly greater distal tibia Tb.BMD and F.Load compared with swimmers.[284] Further, female skiers and soccer players had greater Ct.Th and lower trabecular separation (Tb.Sp) compared with swimmers, while male soccer players had greater Tb.N compared with swimmers.[284] Thus, athletes who participated in high-impact sports had greater metaphyseal bone strength conferred by adaptations in trabecular microarchitecture and/or bone geometry.   1.2.7.2.2 Habitual physical activity  Leisure-time PA: Many children do not engage in structured PA such as gymnastics or racquet sports. Therefore, it is important to consider and better understand the influence of general, leisure-time PA on bone parameters. Observational studies, both cross-sectional and longitudinal, consistently demonstrated that more active children and adolescents accrued more bone mass and strength compared with their less active peers.[133,208,209] For example, vigorous PA (> 6 METs; using accelerometry) predicted 3-7% of femoral neck strength (by HSA) after adjusting for age, weight and height in pre-[285,286] and early-pubertal boys and girls.[286] Similarly, 9-13 year old girls in the highest PA quintile (via self-report questionnaire) demonstrated 7-9% greater estimated bone strength (SSIp and BSI) and 3-4% greater periosteal circumference at the tibial diaphysis and metaphysis (by pQCT) compared with peers in the lowest PA quintile.[102] Finally, pre-pubertal girls who participated in high-impact PA (by questionnaire) demonstrated 3% greater CSA, 7% greater estimated bone strength (CSMI) and 6% greater Ct.Th at the tibial shaft (by pQCT) compared with girls who engaged in low-impact PA. However, no differences were observed in peri-pubertal girls.[287] While accepting known limitations of cross-sectional studies, these findings suggest a positive influence of weight-bearing PA on bone strength during pre- and early-puberty in boys and girls. The strength of associations are less clear in later adolescence, as few studies examined this age group. In those that did, results were equivocal. For example, self-reported weight-bearing PA was not associated with SSIp at the tibial shaft in 11-14 year old peri-pubertal girls.[101] There was also no relationship between self-reported PA and radius strength (breaking bending resistance index by DXA; n = 1116 girls, Tanner 1-5).[288] In contrast, in adolescents and young adults (15-20 years old), impact PA (by questionnaire, impact > walking) was significantly positively associated with Tt.BMD, Tb.BMD and Tb.N at the distal tibia in girls and Tt.Ar and bone strength 72  (minimum and maximum moment of inertia) at the distal tibia in boys.[109] Differences in PA-bone associations are not surprising given the variation in methods between studies (i.e., imaging modalities, imaging of weight bearing and non-weight bearing sites, population and method used to assess PA). Higher quality prospective or intervention studies that use objective measures of PA are needed to clarify the relationship between habitual PA and bone strength in later adolescence. Prospective studies such as the University of Saskatchewan PBMAS examined the influence of PA on normal bone accrual, while controlling for maturation using APHV. In their seminal study, Bailey and colleagues demonstrated that boys and girls in the highest quartile of PA (via self-report questionnaire) gained 7-18% more BMC at the femoral neck, lumbar spine and total body over 7 years compared with boys and girls in the lower quartile of PA.[133] In a subsequent analysis of the PBMAS cohort, self-reported PA positively predicted bone CSA and estimated bone strength (section modulus, HSA) at the femoral neck across maturity (Figure 1.25).[208] PA no longer predicted bone CSA or section modulus once muscle mass (surrogate of muscle force) was included in the multilevel model. This suggested a mediating role of muscle forces in the relationship between PA and bone geometry.[208] Similar findings were demonstrated in the Iowa Bone Development Study (IBDS). This ongoing longitudinal study of bone health during childhood, adolescence and young adulthood showed that MVPA (by accelerometry) positively predicted bone CSA and strength (section modulus by HSA) from age 5-11 years.[209] However, PA did not predict bone outcomes in girls when muscle mass was included in the multilevel model, adding more support for the mediating role of muscle force in the PA and bone health relationship. Despite the known influence of muscle forces on bone adaptation and development,[57,60] few observational studies examined the PA-bone relationship in the context of muscle.[15] Thus, the influence of muscle on associations between PA and bone parameters should be considered in studies of children and youth, in future. 73  Figure 1.25. Illustration of growth curves for section modulus (Z) by hip structural analysis (HSA) of the femoral neck region in the longitudinal subset comparing 17 active girls or boys with 17 inactive girls or boys in relation to biological age, years from age at peak height velocity (APHV). Reprinted from Forwood et al.,[208] with permission from Elsevier.  In a subsequent analysis of the IBDS cohort, boys and girls who engaged in the highest trajectory of MVPA (by group-based trajectory modelling; no adjustment for muscle mass) throughout growth had significantly greater estimated bone strength at the distal (BSI) and 38% tibial site (polar moment of inertia by pQCT) at age 17 years compared with peers in the lowest MVPA trajectory.[289] Whether MVPA positively influences bone microarchitecture similarly is not yet known. That the association between PA and estimated bone strength did not vary with maturity in either longitudinal study[208,289] contrasts intervention studies that demonstrated greater benefits during pre- and early-puberty.[257,265] Such discrepancies highlight the complexity of bone adaptation to loading during growth and suggest that maturity-specific responses to PA may only be observed with more intense PA (i.e., jumping activities performed in RCTs).   1.2.7.3 Long-term effects of physical activity in childhood and adolescence  Benefits of PA on bone accrual are irrefutable. However, we do not have a clear understanding of whether PA-related gains in bone parameters are maintained into adulthood and associated with reduced fracture risk later in life. As discussed in section 1.2.2.2.4, elegantly designed studies that used animal models support lifelong benefits of exercise during growth on bone geometry, strength and fracture resistance.[81] A similar prospective study has not yet been 74  conducted in humans due to prohibitive methodological challenges. However, a recent cross-sectional study of former major league baseball pitchers and catchers demonstrated lifelong benefits of PA participation during youth (participants started throwing at mean age 6 years, ceased habitual throwing at mean age 31 years).[290] Side-to-side differences between throwing and non-throwing arms at the humeral midshaft were observed after 20 years of detraining for Ct.Th (by pQCT) and after 40 years of detraining for cortical BMC and Ct.Ar.[290] Tt.Ar and estimated bone strength (polar moment of inertia) were 0.56 mm2 and 0.62 mm4 greater, respectively, in the throwing compared with non-throwing arm after more than 50 years of detraining.[290] Compared with currently active professional baseball players, these values represent a 56% and 34% throwing-derived benefit in Tt.Ar and estimated bone strength, respectively, 50 years post-training.[290] Detraining and aging in later life, are characterized by a decline in bone strength and its determinants. Thus, support for preservation of bone geometry and strength across the lifespan in baseball pitchers and catchers suggests that PA during childhood and youth may have enduring benefits, despite reduced PA in adulthood.[290] Prospective observational studies[291-293] and studies of athletes[294,295] demonstrated that PA during childhood and adolescence predicted bone parameters in young adulthood. Specifically, in the Penn State Young Women’s Health Study (YWHS) and the University of Saskatchewan PBMAS, individuals who were most active during childhood and early adolescence maintained bone mass and strength advantages over their less active peers in later adolescence and young adulthood. Pre-menarcheal girls in the most active tertile (self-report questionnaire; mean age 12 years) at baseline in the YWHS had 10-11% greater estimated femoral neck strength (HSA) at age 17 compared with less active girls.[291] Similarly, when participants in the PBMAS cohort were followed up 9-11 years after baseline, young women and men 23-30 years of age who were most active as adolescents were still more active as adults compared with their peers.[292] Further, women who were in the upper quartile for PA during adolescence had 9-10% greater total hip and femoral neck BMC, 10-12% greater Ct.Ar and BMC at the tibia diaphysis and 3% greater trabecular content at the distal tibia in adulthood compared with their inactive peers (after adjusting for adult height, muscle area and adult PA). Men who were most active during adolescence demonstrated 10% greater Tt.Ar and 13% greater estimated bone strength (SSIp) at the tibial diaphysis compared with their inactive peers.[293] However, bone parameters of women and men who reported average PA levels during 75  adolescence did not differ from inactive or active peers.[293] Thus, higher levels of PA during adolescence may be required to retain long-term benefits. Skeletal benefits from gymnastics participation during childhood may also persist into later adolescence and young adulthood. For example, women aged 18-36 years who participated in high-level gymnastics during childhood and adolescence had 13-32% greater Ct.Ar and Tt.Ar and 16-25% greater BMC at radial and humeral shafts (pQCT) compared with same-aged women with no previous history of gymnastics participation.[294] Greater bone geometry in former gymnasts contributed to 36-38% greater estimated bone strength (SSIp) at shaft sites of the radius and humerus (by pQCT).[294] Former gymnasts also had 15-18% greater MCSA at radial and humeral shafts compared with non-gymnasts.[294] Benefits from gymnastics were also observed at shaft sites of the tibia and femur and at the distal tibia, but were smaller in magnitude than those observed in the upper limbs.[294] Similarly, retired elite female gymnasts (10 years post-retirement, aged 22-30 years) had 10-50% greater estimated bone strength (BSI and SSIp by pQCT) at distal and shaft sites of the radius and tibia compared with non-gymnast controls.[295] Bone strength adaptations in former gymnasts were associated with 15-28% greater Tt.Ar and BMC (total, cortical and trabecular) at the radius and 9-15% greater BMC (total, cortical and trabecular) and Tb.BMD at the tibia compared with non-gymnasts. Thus, very high-impact PA during adolescence may confer long-term benefits for bone geometry and strength. Although these retrospective studies suggest that PA during childhood and adolescence may enhance bone strength later in life, a lifelong commitment to weight-bearing PA is recommended. However, we may never know whether such benefits in bone geometry and strength translate to reduced fracture risk later in life given the challenges of conducting such a study (i.e., long-term follow-up and associated costs).  1.2.7.4 Observational studies of sedentary time  Today’s youth spend the majority of their waking hours in sedentary activities, yet few studies investigated the relationship between sedentary time and bone health (mass, geometry, or strength) in children and adolescents.[296-301] Too much sedentary time may negatively impact bone health by disrupting the balance between bone resorption and formation.[302] In an extreme example, prolonged sedentary time in a bed rest study of healthy adults increased bone 76  resorption rates without changes in bone formation rates.[303] In healthy children and youth, however, bone formation predominates. Thus, it is unclear how sedentary time interacts with the osteogenic effects of PA during growth and development. A recent systematic review suggested there was insufficient evidence to support an association between sedentary time (by accelerometry) and bone parameters (predominantly DXA-based studies).[304] Studies that examined sedentary time-bone relationships primarily relied on DXA to assess BMC or aBMD. With exception of recent reports,[296,300,301] previous studies also relied on self-reported screen time to quantify sedentary time.[297-299,305] In brief, these studies suggested an inverse association between whole body BMC and internet use for non-academic purposes in adolescent boys;[297] a negative association between TV viewing and proximal femur aBMD in pre-pubertal girls;[306] and a negative association between TV viewing during childhood and adolescence (age 5, 8, 10, 14 and 17 years) and whole body BMC at age 20 years,[305] adjusting for PA. In contrast, three studies reported no association between self-reported sedentary time and whole body BMC or aBMD either with[298] or without[307,308] adjusting for PA. Results were equally mixed in the three studies that examined accelerometry-derived sedentary time. For example, sedentary time was positively associated with lumbar spine and proximal femur aBMD in adolescents and young adults, independent of MVPA.[296] Based on these limited findings, authors speculated that extended periods of sedentary time between bouts of PA might be required for optimal adaptation of bone to mechanical loading. Similarly, a 2-year follow-up study of 10-14 year old girls and boys found that increased sedentary time, when substituted for time spent in light PA, was positively associated with whole body BMC and aBMD.[301] In contrast, in peri-pubertal boys (age 11-13 years), a 5% increase in sedentary time over 12 months was negatively related to change in femoral neck aBMD (adjusted for vigorous PA).[300] Discrepancies clearly exist in the literature, and thus, well-designed prospective studies (appropriately powered for different rates and timing of maturity) are needed to clarify the bone strength-sedentary time relationship (measured objectively). To supplement the paucity of information regarding the influence of sedentary time on bone microarchitecture, I briefly describe the Women International Space Simulation for Exploration (WISE) bed rest study, which provided an extreme example of unloading in twenty-four women aged 25-40 years. After 60 days of bed rest, BV/TV (by HR-pQCT) decreased by 0.1-0.3%, Tb.N decreased by 1-2% and Tb.Sp increased by 1-2% at both the distal tibia and 77  radius. At the distal tibia, Ct.Th also decreased by 1%.[309] Trabecular microarchitecture deficits persisted at one-year following cessation of bed rest, [309] suggesting that sustained periods of unloading may have long-term consequences for trabecular bone.  1.2.8 Summary of the literature   Bone strength is a function of bone geometry, density and microarchitecture, which all continually adapt to variable mechanical loads during growth. Maturity- and sex-related adaptations in bone strength and its determinants during childhood and adolescence are unclear. This is in part due to the large sample size needed to appropriately assess change across maturation given the substantial variation in its magnitude and timing. It is also due to few prospective studies that used 3D imaging to characterize bone. Although time and labour intensive, it is through longitudinal studies that we will better understand nuanced adaptations of bone over time. However, few such studies have been conducted during adolescent growth.  Therefore, in this thesis I overcome recognized gaps in the literature. That is, I examine bone strength and its determinants prospectively using pQCT and HR-pQCT. I also align boys and girls on biological age, APHV, and extend the scant literature that assessed the role of sedentary time on bone during growth using accelerometry. Further, I employ advanced statistical modelling approaches to maximize value of the HBSIII longitudinal dataset and to adequately distinguish within-person from between-person differences across adolescent growth. Collectively, these novel components represent the first time that maturity- and sex-related adaptations in bone strength and its determinants were examined prospectively across 12-years of adolescent growth, with participants aligned on biological age to control for maturational differences. Finally, it is also the first prospective study to investigate the influence of both PA and sedentary time on bone strength using HR-pQCT.  1.3 Rationale, objectives and hypotheses  In this section, I outline the rationale and specific aims and hypotheses for each of the four studies that comprise this dissertation.  78  1.3.1 Bone strength and microarchitecture in the growing skeleton: the role of sedentary time  Rationale: The benefits of PA for bone strength and parameters that underpin bone strength during childhood and adolescence are well established.[15] However, we know little about the potentially deleterious effects of sedentary time on bone during these key periods of growth.  Objectives: 1. To examine associations between self-reported screen time and bone strength and its determinants (bone parameters), independent of self-reported PA. 2. To examine associations between objectively measured volume4 and patterns of sedentary time and bone parameters, independent of objectively-measured MVPA. 3. To assess the contribution of muscle force and modulator variables (i.e., maturity, ethnicity, dietary calcium and PA) to bone parameters.  Hypotheses: 1. Self-reported screen time will be negatively associated with bone parameters, independent of PA. 2. Objectively measured volume and patterns of sedentary time will be negatively associated with bone parameters, independent of MVPA. 3. Maturity and MCSA (surrogate of muscle force) will be the primary explanatory variables of tibial bone parameters.  Contribution: This cross-sectional study[310] is the first to examine the relationship between sedentary time and bone strength and its determinants using 3D bone imaging (HR-pQCT), which permits evaluation of Ct.Po and estimates of bone strength. Further, this study objectively measures the volume and patterns of sedentary time accumulation in addition to self-reported screen time. Current PA guidelines recommend limiting sedentary time for optimal health in children and youth;[311] however, there is a relatively limited body of evidence regarding how                                                 4 Volume refers to total duration of sedentary time (min/day) 79  unloading the skeleton may be detrimental to bone parameters in healthy children and adolescents. Thus, these findings may inform future public health recommendations regarding sedentary time of children and youth.  1.3.2 Re-examining the surfaces of bone in boys and girls during adolescent growth: a 12-year mixed longitudinal pQCT study  Rationale: A plethora of research supports childhood and adolescence as critical periods for bone mineral accrual.[9,312] However, the intricacies of how bone is gained during childhood is not completely understood. In the 1970s, a landmark study by Garn and colleagues examined surface-specific differences in bone growth and development. Specifically, this cross-sectional study examined radiographs of the second metacarpal and concluded that boys and girls experience periosteal expansion and endocortical contraction during adolescent growth. However, boys exhibit more periosteal expansion while girls exhibit more endocortical contraction.[155-157] The current study adopts a 12-year mixed longitudinal design and examines the tibial midshaft of boys and girls who are aligned on biological age (years from APHV) to revisit Stanley Garn’s theory related to sex differences in periosteal expansion and endocortical contraction. Findings extend those from our previous study that used pQCT to assess bone development across 20 months.[153]  Objectives: 1. To compare rates of bone expansion and/or contraction at the periosteal and endocortical surfaces of the tibial midshaft between boys and girls pre- and post-APHV. 2. To compare rates of Ct.BMD and bone strength (SSIp) accrual at the tibial midshaft between boys and girls pre-and post-APHV.  Hypotheses: 1. Both boys and girls will demonstrate expansion at the periosteal and endocortical surface. Boys will demonstrate a greater magnitude of change at both surfaces pre- and post-APHV. 80  2. Boys will demonstrate greater bone strength, but lower Ct.BMD compared with girls pre- and post-APHV.  Contribution: This 12-year study of adolescent bone growth using 3D imaging techniques (pQCT) is the longest to date.[313] Longitudinal studies are difficult to conduct, time consuming and relatively rare. Thus, the few existing studies that examined changes on bone surfaces during growth were cross-sectional or short term prospective. My study is uniquely able to account for the tremendous variability that accompanies bone adaptation throughout adolescence; I control for the potentially profound influence of maturity by aligning boys and girls on a common maturational landmark, APHV. Findings may challenge commonly held notions regarding sex differences in how bone is gained at the mid-tibia during growth and may improve our understanding of factors that influence fracture risk during adolescence.  1.3.3 Sex differences and growth-related adaptations in bone microarchitecture, geometry, density and strength: a mixed longitudinal HR-pQCT study  Rationale: Sex differences in adult bone strength and fracture risk are well-documented. However, we know less about adaptations in bone microarchitecture, geometry and density that accompany gains in bone strength during growth. Only three studies used HR-pQCT to evaluate sex differences in bone strength and its determinants during adolescence. Two of these were cross-sectional and one had a short follow-up period. Prospective studies of longer duration are key to evaluate the nuances of bone development over time and to further our understanding of factors that might contribute to the elevated fracture risk during adolescence, and ultimately, in later life.   Objectives: 1. To describe growth-related adaptations in bone strength and its determinants (parameters) at the distal tibia and radius in boys and girls. 2. To compare differences in growth-related adaptations in bone parameters between boys and girls.  81  Hypotheses: 1. Boys and girls will demonstrate increases in bone parameters throughout adolescence, with the exception that Ct.Po and load-to-strength ratio will decline during adolescence. 2. Boys will demonstrate greater bone strength, geometry and cortical porosity, but lower Ct.BMD throughout adolescence compared with girls.  Contribution: This longitudinal study of boys and girls across adolescent growth using HR-pQCT to evaluate bone, is the longest to date.[314] Uniquely, I use advanced mixed modelling approaches and align boys and girls on a common maturational landmark (APHV) to more clearly characterize changes in 3D aspects of bone microarchitecture, geometry, density and strength that accompany adolescent growth. This study provides new insight into sex differences in bone parameters and factors that may contribute to greater skeletal fragility during adolescence and, ultimately later in life.  1.3.4 Physical activity, sedentary time and bone strength from childhood to early adulthood: a mixed longitudinal HR-pQCT study  Rationale: Bone strength and its determinants continually adapt to increased mechanical loads during growth and PA is essential for optimal bone strength accrual. However, given the relatively recent evolution of bone imaging technologies, less is known about how bone microarchitecture adapts to PA and whether sedentary time independently influences bone parameters. A recent systematic review suggested there is insufficient evidence to ascertain an association between sedentary time and bone health in children and youth, independent of PA.[304] Thus, it remains unclear how the potentially deleterious impacts of sedentary time interact with the positive effects of PA to influence skeletal growth and development in healthy, children and youth. Prospective studies are poised to clarify adaptations in bone microarchitecture associated with independent effects of PA and sedentary time during growth.    82  Objectives:  1. To evaluate prospective associations between PA, sedentary time and growth-related adaptations in bone parameters at the distal tibia and radius in boys and girls across adolescence.  Hypotheses: 1. PA will positively predict adaptations in bone parameters. Sedentary time will be negatively associated with bone parameters, independent of PA.  Contribution: Previous cross-sectional studies evaluated the association between PA, sedentary time and bone parameters. These early studies used DXA to image bone and subjective measures to assess PA and sedentary time. I extend this body of literature in three distinct ways: 1) I use longitudinal data acquired across 4-years at the tibia and 3-years at the radius in boys and girls across adolescent growth; 2) I evaluate bone using more advanced 3D imaging techniques; 3) I assess PA and sedentary time objectively (using accelerometry). This is the first prospective study to examine how trabecular and cortical bone microarchitecture adapts to PA and sedentary time during adolescence. Findings will further clarify the consequences of positive health behaviours such as PA versus negative health behaviours such as sedentary time on bone parameters during adolescent growth. Outcomes might inform PA and sedentary time public health guidelines for youth, in future.   83  Chapter 2: Methods  In this chapter, I present the research methods used to address my research aims. I first provide an overview of the study cohort in section 2.1 and then present specific methods used in section 2.2.  2.1 Healthy Bones Study overview  Participants were healthy girls (n=556) and boys (n=515) aged 8 to 23 years who participated in the mixed longitudinal University of British Columbia Healthy Bones III Study (HBSIII; Figure 1). The HBSIII cohort includes participants from three school-based studies: the Healthy Bones Study (HBS; Healthy Bones and Bounce at the Bell), which began in 1999; the Action Schools! BC (AS!BC) project, which began in 2003; and the most recent cohort, recruited in 2009. The three cohorts are collectively referred to as HBSIII and I describe this cohort in detail below.  2.1.1 Healthy Bones Study and Bounce at the Bell  HBS was a cluster randomized controlled school-based intervention that investigated effects of a 20-month exercise intervention on bone mass, as measured by DXA. We recruited participants (n = 383) in the fall of 1999 from grade 4, 5 and 6 classes in 14 schools in Richmond, BC, described in detail elsewhere.[257-259] We implemented the intervention over two academic years (2, 7-month intervention periods); the intervention consisted of brief (10-12 min) high-impact, weight-bearing exercise twice per week during physical education class and once per week in the classroom or outside. Participants in intervention and control schools took part in 40 min of physical education, twice per week, as mandated by the school board. We invited participants from intervention and control schools to attend annual assessments in the spring of each year following the intervention (until 2011). I describe the substantial efforts taken to retain participants for up to 12 years in section 2.1.4.   84   Figure 2.1. Overview of the University of British Columbia Healthy Bones Study III (HBSIII). 85  The HBS companion study, Bounce at the Bell, investigated the effect of frequent bouts of jumping exercises on bone mass over 8-months (intervention period not indicated in Figure 1.1). We recruited participants (n = 51) in the fall of 2000 from grades 4 and 5 classes in 3 schools in Richmond, BC.[315] Participants performed 10 counter-movement jumps, 3 times/day (morning, noon and home bell; ~ 3 min/day of jumping) in addition to twice weekly physical education class. We invited participants to attend annual assessments each spring following the intervention (until 2011).  2.1.2 Actions Schools! BC  The Actions Schools! BC (AS!BC) trial was a 16-month cluster randomized controlled school-based intervention that evaluated the effectiveness of the AS!BC model for increasing bone mass and strength (as measured with DXA and pQCT). The AS!BC model is an active school model designed to promote PA in elementary schools. The model helps schools develop individualized action plans to promote healthy living based on evidence and best practice. The model is flexible and based on principles of health promotion.[316] We recruited participants (n = 515) in early 2003 from grade 4 and 5 classes at 10 elementary schools in Vancouver and Richmond, BC. [265] In phase one (3-months prior to summer holiday), we oriented participants to the program. In phase two (resumption of school in fall 2003), participants at intervention schools completed an 8-month active intervention. As with the HBS cohort, we invited AS!BC participants to attend annual follow-up measurements each spring through 2011. We merged the HBS, Bounce at the Bell and AS!BC cohorts in 2006 because they employed nearly identical protocols and because participation in the exercise intervention was not associated with sustained benefits at the tibial shaft.[317] However, we excluded observations from children actively participating in the AS!BC intervention (n = 451, spring 2004) because we previously demonstrated a positive effect of a PA intervention on bone accrual.[258,259]    86  2.1.3 New cohort  In 2008, our research group acquired a first generation HR-pQCT and incorporated this into the HBSIII measurement protocol to assess bone microarchitecture. Since the youngest children from the HBS and AS!BC cohorts were in grade 10 (approximately 15 years old) by the time of first HR-pQCT assessment, we recruited a younger cohort (pre- and early-pubertal children) in order to examine changes in bone microarchitecture earlier and through the period of adolescent growth. We recruited the new cohort of younger participants (n=120; mean age 10.5 ± 0.6 years) in 2009 from grade 4 and 5 classes in 5 schools in Vancouver and Richmond, BC. We invited participants to attend annual assessments from spring 2009 through 2012.  In this thesis, I include bone data from HBSIII annual measurements conducted between spring 2001 and 2012. We obtained written informed consent from the parents or legal guardians, written assent from participants younger than 18 years of age and informed consent from participants 18 years of age and older. The University of British Columbia’s Clinical Research Ethics Board approved all procedures (H15-01194, H07-02013, H2-70537).  2.1.4 Recruitment and retention  We employed similar recruitment methods for HBS and AS!BC studies. In brief, principals volunteered their schools to participate after the recruitment team made presentations to school principals at district meetings. Next, the recruitment team presented the project to grade 4, 5, and 6 teachers in volunteer schools. Finally, the recruitment team presented the study to grade 4, 5 and 6 students whose teachers had volunteered. We gave information letters and consent forms to classroom teachers for students to take home to parents (Appendix A). We obtained consent and assent for the follow-up studies for HBS participants in 2001, 2003, 2006 and 2009 and for AS!BC participants in 2004, 2006, 2007 and 2009. We used similar recruitment methods to recruit the new cohort of grade 4 and 5 students. We distributed information letters and consent forms (Appendix A) in the classroom and obtained consent and assent in 2009. We used several incentives over the 12-year study period to retain participants, including distributing items at assessments (i.e., snacks, stickers, pencils, socks, Frisbees, $20). We mailed 87  detailed individual and group (data de-identified, collapsed and reported by age and sex) results to each participant (Appendix B) in advance of the next years’ data collection. Results included a picture of their whole body skeleton from DXA along with a note reminding them of upcoming data collection.  2.1.5 Data collection overview  For HBSIII participants attending elementary schools, the research coordinator contacted their teachers and arranged for participants from each class to be picked up at the school door to travel to the lab in groups of 5 (plus a research assistant as chaperone). The driver and research assistant accompanied participants during their trip from the elementary schools to the measurement site. The research coordinator contacted HBSIII study participants who attended secondary schools by telephone. Whenever possible, we booked group measurement sessions (up to 6 participants/session) for students attending the same secondary school (participants attended 41 different secondary schools across the study period), and we transported participants to the Bone Health Research Lab at Vancouver General Hospital (VGH) (and in 2012 to the Centre for Hip Health and Mobility at VGH) by minivan. For participants who already graduated secondary school, the research coordinator contacted participants individually to schedule assessment. At the lab, participants rotated through 6 stations: anthropometry (5 min), questionnaires (30 min), mechanography (15 min), DXA (20 min), HR-pQCT (20 min) and pQCT (10 min). All members of the research team attended a full-day training session conducted by the research coordinator to learn measurement protocols prior to data collection. All research team members were trained to correctly conduct anthropometry and administer questionnaires. Team members practiced measurements during the training session to maintain quality assurance and were also trained on the ethics of data collection. Trained technicians/measurers conducted bone imaging and mechanography procedures.     88  2.2 Heathy Bones III Study protocol  In this section I discuss the specific measurements conducted for HBSIII. I personally acquired and analyzed all pQCT scans in 2012 and analyzed all HR-pQCT scans.  2.2.1 Anthropometry  Anthropometry included height (cm), sitting height (cm), body mass (kg), tibial and ulnar length (mm). We measured height and sitting height to the nearest 0.1 cm with a wall-mounted digital stadiometer (Seca Model 242, Hanover, MD) using stretch stature techniques. We assessed height with the participant’s head positioned in the Frankfort plane, heels flat on the floor and shoes off and applied gentle traction to the participant’s mastoid process.[318] We assessed body mass to the nearest 0.1 kg using an electronic scale (Seca Model 840, Hanover, MD). Participants removed any heavy clothing and shoes prior to stepping onto the scale. We assessed limb length to the nearest mm as the distance from the distal edge of the medial malleolus to the tibial plateau for the tibia and the distance from the olecranon to the ulnar styloid process for the ulna. Trained research assistants took all measurements in duplicate, unless differences were > 0.4 cm or 0.2 kg when they obtained a third measure. We used the mean of two values or the median of three for all analyses. In our laboratory, reproducibility (CV%) is < 0.3% for measures of stature and < 3.5% for tibia length.  2.2.2 Health history questionnaire   Parents completed a health history questionnaire for their child at baseline (HBS: 1999, Bounce at the Bell: 2000, AS!BC: 2003, new cohort: 2009) and participants completed a shorter version at subsequent annual visits (Appendix C). We excluded participants with diseases known to affect bone metabolism (e.g, osteogenesis imperfecta, fetal alcohol syndrome, Type 1 diabetes) or participants taking medication known to influence bone metabolism. We determined each participant’s ethnicity based on their parents’ and/or grandparents’ place of birth as reported on the health history questionnaire at baseline. Parents were asked to classify their own, and their child’s ethnicity. We classified participants as “Asian” if both parents or three of four 89  grandparents were born in Hong Kong, China, Japan, Taiwan, Philippines, Korea or India; “white” if both parents or three of four grandparents were born in North America or Europe; and “other” if the participant had parents of other or mixed ethnicities. We also considered parental self-report of ethnicity to ensure correct classification of each participant.  2.2.3 Maturity   2.2.3.1 Sexual maturation  We assessed maturity using the method of Tanner: self-reported pubic hair stage in boys and self-reported breast and pubic hair stage in girls;[127] however, we only used breast stage in girls as it showed better alignment with timing of menarche.[132] We gave participants a set of line drawings that depicted the 5 stages of sexual development and asked participants to select the drawing most similar to his/her own physical appearance. A brief description of the visual appearance at each stage accompanied drawings (Appendix C). Participants completed the questionnaire in private after receiving instructions from a research assistant and returned the questionnaire in a sealed envelope once completed. Participants who had reached maturity (Tanner stage 5) based on a previous year’s data collection were not required to complete the questionnaire. I considered participants who were in Tanner stage 1 as pre-pubertal, Tanner stage 2 and 3 as early-/peri-pubertal and Tanner stage 4 and 5 as late-/post-pubertal.[4] We also assessed maturity in female participants using self-reported menarcheal status. A research assistant asked female participants if they had experienced their first menstrual period. If “yes”, they were asked to recall the approximate date. Participants who reported reaching menarche at a previous year’s data collection were not asked this question in subsequent years.  2.2.3.2 Age at peak height velocity  To control for well-known maturational differences between adolescent boys and girls of the same chronological age, we calculated age at peak height velocity (APHV; years) as an estimate of biological maturity. I provide a detailed description of this process in Appendix D. In brief, we fit an interpolating cubic spline to each participant’s height velocity data.[142] The 90  magnitude of PHV was identified as growth per year (cm/year) that occurred at APHV. We used APHV to calculate a biological maturity offset (in years) by subtracting the APHV from chronological age at time of measurement. Thus, we generated a continuous measure of biological maturity offset (e.g., -1 year is equivalent to 1 year prior to attainment of APHV; +1 to one year after APHV). Due to missing and mistimed (e.g., 3 to 12 months between height measurements) measurements surrounding APHV, we were able to identify APHV for 235 of 1071 participants (112 boys, 123 girls).  2.2.3.3 Maturity offset equation  As we were unable to calculate APHV for all participants, I also estimated maturity offset (years from APHV) using a recalibrated version of the Mirwald prediction equation.[141,142] The recalibrated equation is a simplified version of the Mirwald equation that uses age and height for girls and age and sitting height for boys.[142] In the calibration sample from the Saskatchewan PBMAS (79 boys and 72 girls; 7.5-17.5 years), predicted APHV explained approximately 90% of variance in actual APHV.[142] Of note, the published equations by Moore et al., were based on data from white participants only. However, our research group also developed equations for Asian boys and girls (unpublished data). Therefore, I used the published equations[142] to predict maturity offset in white participants and in participants of other/mixed ethnicities. I used ethnic-specific equations to predict maturity offset in Asian participants.  Maturity offset estimation equations (age in years and height and sitting height in cm): 1. White/other boys: (-8.128741 + (0.0070346 x age x sitting height) 2. Asian boys: (-8.128741 + 0.7482624) + (0.0070346 x age x sitting height) 3. White/other girls: -7.709133 + (0.0042232 x age x height) 4. Asian girls: (-7.709133 + 0.7303442) + (0.0042232 x age x height) For all participants I used anthropometry data from the measurement occasion closest to a reported average APHV (approximately 11.6 years in girls and 13.5 years in boys) to estimate maturity offset.[142]     91  2.2.4 Dietary calcium intake  All participants completed a validated food frequency questionnaire to estimate dietary intake of calcium (mg/day; Appendix C). Validity of the FFQ was assessed against a 1-day food recall (r = 0.98) and reliability was assessed on two occasions separated by 3 months (r = 0.76).[319] Participants reported how often they consumed 20 calcium-rich foods items (times per week, times per month) and how much they consumed each time (number of servings as per serving size described in the food frequency questionnaire).  2.2.5 Peak muscle power  We used the Leonardo Mechanograph Ground Reaction Force Plate (GRFP; Novotec, Germany) to assess peak leg muscle power from 2008 onwards, the mechanics of which are described in detail elsewhere. [320] Briefly, the GRFP is divided into two sections, which allows for simultaneous measurement of forces (vertical component only) applied to the right and left legs separately. The sample rate is set to 800 Hz (800 measurements/s for each force sensor). We used the manufacturer’s software (Leonardo Mechanography v4.3) to detect, store and calculate mechanography outcomes. The software uses force and time data to calculate velocity of the movement (m/s), power (Watts, W) and jump height (m).  All participants performed a single two-legged countermovement vertical jump on the GRFP with their hands held static at their waist and their feet hip width apart. The research assistant explained the jumping protocol to all participants in a standardized manner. We asked participants to perform the countermovement jump after hearing a tone (from the computer). The research assistant instructed each participant to initiate a downwards movement and then immediately jump up as high as possible using both legs. We instructed participants to land with both feet on the platform (with each foot on the appropriate side of the middle line) and to remain still until after hearing a second tone from the computer signaling the end of the trial. Each participant performed one practice jump and three trial jumps. We used peak power during lift off phase (kW) from the jump associated with the maximum height for analysis.  92  2.2.6 Self-reported screen time and physical activity  We estimated screen time using a self-report questionnaire that inquired about h/day spent watching television and/or playing video or computer games during the previous week (Appendix C). There were five response options ranging from “none at all, or less than 1 hour per day” to “more than 4 hours per day”. In addition to examining these data using the 5 response groups, I collapsed these responses into two groups (< 2 h/day and ≥ 2 h/day) in order to examine whether meeting current sedentary time guidelines for youth, which recommend “limiting recreational screen time to no more than 2 hours per day”,[321] is associated with bone outcomes. We assessed self-reported PA time over the previous week using the previously validated self-report Physical Activity Questionnaire for Children (PAQ-C) in elementary school participants and Physical Activity Questionnaire for Adolescents (PAQ-A) in participants in high school or older (Appendix C).[222,223] We calculated a general PA score as an average of the PAQ items in a continuous range between 1 (low activity) and 5 (high activity). Based on participants’ estimates of time spent in common sports and activities in Item 1, we also estimated time spent in MVPA (min/day) and time spent in activities designated as loaded (impactPA in h/week; impact > walking).[109]  2.2.7 Objectively measured sedentary time and physical activity  In 2008, our research group acquired accelerometers to estimate PA and sedentary time (ActiGraph GT1M; Pensacola, FL). The GT1M is a small, uniaxial accelerometer that detects vertical accelerations of 0.05 – 2.00 g. The signal is band filtered to the frequency range of 0.25-2.50 Hz to exclude non-human movement. We attached each accelerometer to an elastic belt and instructed participants to wear the belt around the waist with the accelerometer positioned at the iliac crest. We asked participants to wear the device during all waking hours for seven consecutive days, except during water-based activities (e.g., swimming and showering). Participants received a log sheet to record accelerometer on and off times each day. We set the accelerometers to record in 15-sec epochs and analyzed all data using KineSoft software (v3.3.75; KineSoft, Loughborough, UK). 93  We included participants who recorded at least 10 h/day of data on three or more days[232] and defined non-wear time as 60-min of consecutive zero counts. As non-wear criteria are inconsistent within the literature and significantly alter accelerometer output, I determined my own non-wear criteria using HBSIII data based on suggestions from Mâsse and colleagues to examine the number of wearing interruptions observed in the data, as a high number of interruptions (e.g., 10 per day) are unlikely.[322] I examined several different non-wear criteria ranging from 10- to 60-min of consecutive zeros with or without a 2-min interruption. The 60-min non-wear criteria without interruptions resulted in an average of 1-2 wearing interruptions per day, compared with an average of 7 interruptions per day using the 10-min criteria without interruptions. Further, the longer criteria allowed me to include more participants (more valid wear days), while demonstrating strong rank order correlations with the other criterion (10-, 20-, 30-min without interruptions and 60-min with interruptions; r = 0.92-0.99). I presented these findings at the 2013 ICAMPAM conference in Amherst, Massachusetts (Gabel et al., Relationships between physical activity and adiposity: Does accelerometer non-wear criteria matter? ICAMPAM Poster Presentation June 2013). I used a cut point of < 100 cpm to classify sedentary time[239] and the Evenson cut points[238,239] to determine intensities of PA: light ≥ 100 cpm, moderate ≥ 2296 cpm and vigorous ≥ 4012 cpm. Thus, I defined MVPA using an accelerometer cut point ≥ 2296 cpm.  2.2.8 Bone imaging  2.2.8.1 pQCT  Acquisition: We used pQCT to assess bone structure and strength at the tibial midshaft. One of 8 trained operators acquired pQCT scans over the 12-year period. Each operator conducted inter- and intra-rater reliability training to ensure consistency across technicians (e.g., scanning five participants with repositioning). We acquired a 2.3 mm slice at the midshaft (ROI: 50% site; proximal to the distal tibial endplate) of the left tibia using the XCT-2000 (Norland/Stratec Medizintechnic GmbH, Pforzheim, Germany) from 2001-2007 and the XCT-3000 (Norland/Stratec Medizintechnic GmbH, Pforzheim, Germany) from 2008-2012. We previously reported excellent agreement between XCT-2000 and XCT-3000 (root mean squared 94  coefficient of variation, 0.6-1.5% for tibial midshaft bone parameters).[323] We used a scan speed of 30 mm/sec and a resolution (pixel size) of 0.5 mm in participants recruited prior to 2003 and a resolution of 0.4 mm thereafter. Previous work confirmed no significant differences in Tt.Ar or Tt.BMD between 0.4 mm and 0.5 mm resolution pQCT scans at the distal radius.[324] We acquired a 30 mm planar scout view over the joint line to define the anatomic reference line, located at the distal aspect of the distal cartilage of the tibia (Figure 2.2). We used the same anatomical reference line to assess the same relative site each year. The effective dose equivalent (risk of exposure from a single tissue in terms of whole body exposure risk) for pQCT is negligible at 0.22 µSv (~ 5% of daily effective background radiation dose).  Figure 2.2. A) the anatomical reference line defining the distal aspect of the distal cartilage of the tibia and B) a peripheral quantitative computed tomography scan of the tibial midshaft. Bone is indicated in white, muscle in red/purple and subcutaneous fat in blue.  Prior to imaging, the pQCT technician screened participants to rule out pregnancy, ascertain prior exposure to ionizing radiation and to determine each participant’s fracture history. If participants reported a prior fracture of the limb of interest in the past six months we scanned the contralateral side. Additionally, to maintain consistency with previous assessments, we scanned the contralateral side if it had been scanned in a previous year’s assessment due to fracture of the limb of interest. A B 95  The pQCT technician briefly explained how pQCT worked and asked participants to extend their left leg into the pQCT gantry, resting on the supported platform (Figure 1.12). The operator secured the limb firmly with Velcro straps to minimize movement during the scan. The technician conducted a second scan if movement occurred during the first scan. Each scan took approximately 3 min to complete. I acquired and analyzed all scans in 2012 (n=59). Analysis: We analyzed all scans using Stratec software version 6.0 as per our standard protocol.[151,153] An automatic ROI was generated after placing the cursor at the center of the tibia marrow cavity. The algorithm uses modes and thresholds set by the operator to determine numerous bone variables. I list modes, thresholds and outcome variables used in this thesis in Table 2.1. As discussed in section 1.2.3.2.1, pQCT protocols are not standardized; thus, I used modes and thresholds similar to those used in previous studies by our group and based on the manufacturer’s recommendations.[151,153]  Table 2.1. Analysis modes, thresholds and outcome variables for pQCT measurements at the tibial midshaft (50% site). Variable Analysis Mode (Threshold, mg/cm3) Total bone cross-sectional area (Tt.Ar, mm2) Contour mode 1 (711 mg/cm3)  Cortical Area (Ct.Ar, mm2) Peel mode 2 (540 mg/cm3) Cortical bone mineral density (Ct.BMD, mg/cm3) Separation mode 1 (711 mg/cm3)   Polar strength-strain index (SSIp, mm3) Contour mode 1 (711 mg/cm3) Peel mode 2 (540 mg/cm3) Separation mode 2 (480 mg/cm3)   Muscle cross-sectional area (MCSA, mm2) Contour mode 1 (-100 mg/cm3) Peel mode 2 (40 mg/cm3) Separation mode 2 (711 mg/cm3)    We determined in-vivo precision with repositioning at the 50% site using the XCT-2000 in 14 participants (12-27 years); the CV was less than 2% for all bone variables. [151] We maintained quality assurance by scanning a pQCT anthropomorphic phantom daily during measurement periods.    96  2.2.8.2 HR-pQCT  Acquisition: We assessed bone strength, microarchitecture, BMD and geometry at the non-dominant tibia and radius using HR-pQCT (XtremeCT; Scanco Medical AG, Brüttisellen, Switzerland.), unless the participant sustained a previous fracture of the tibia or radius, in which case we scanned the opposite limb.[4] We identified the preferred leg for kicking (i.e., “which leg would you use to kick a soccer ball”) as the dominant tibia. We used a standard ROI to assess the same relative site from year to year. The ROI included both cortical and trabecular bone and excluded the growth plate in most children.[110]  Prior to each scan the HR-pQCT technician immobilized the limb in a carbon fibre cast shaped for the leg or forearm (Figure 2.3 A-B). The technician placed the limb into the gantry and adjusted the chair to ensure the participant was as comfortable as possible (Figure 2.3 C-D). The technician explained the importance of being still during the scans and dimmed the lights to create a relaxing environment.  Figure 2.3. Set-up for high-resolution peripheral quantitative computed tomography (HR-pQCT) radius (A,C) and tibia (B,D) scans. 97  The technician first performed a 2D scout view over the joint line to identify the ROI. The technician placed a reference line at the distal tibia end plate (Figure 2.4) or medial edge of the distal radius (Figure 2.5) and defined the ROI as proximal to the reference line and equivalent to 8% of the total tibia length (Figure 2.4) or 7% of total ulnar length (Figure 2.5). We scanned all participants using the manufacturer’s standard protocol of 60 kVp effective energy, 900 µA, matrix size of 1536 × 1536, 100 ms integration time and 82 µm nominal isotropic resolution. We acquired 110 slices (approximately 9.02 mm) scanned proximally toward the 8% or 7% site of the tibia and radius, respectively. The effective dose equivalent for the tibia scan is < 3 µSv per measurement (~ three-quarters of daily effective background radiation dose), with a measurement time of 2.8 min. A second scan was acquired if there were significant motion artifacts (> grade 3; Figure 2.6) on the first.[124] We conducted daily quality control procedures to assess density fluctuation and weekly procedures to standardize geometry using a calibration phantom provided by the manufacturer. Figure 2.4. High-resolution peripheral quantitative computed tomography at the distal tibia. A) scout view image illustrating 8% scan site; B) scout view illustrating position of tibial growth plate. Reprinted from Burrows et al.,[110] with permission from Springer. A B 98  Figure 2.5. High-resolution peripheral quantitative computed tomography at the distal radius. A) scout view image illustrating 7% scan site; B) scout view illustrating position of ulnar and radial growth plates; C) representative three-dimensional image showing cortical and trabecular compartments. Reprinted from Burrows et al.,[111] with permission from Elsevier.  Analysis: I assessed all HR-pQCT images for motion artifacts using a grading scale from 1 (no motion) to 5 (medium-large streaks/discontinuities) (Figure 2.6).[124] I excluded scans with motion artifacts > 3.[124] Following motion grading, we performed three separate analyses to assess bone microarchitecture, geometry, BMD and estimate bone strength.  Figure 2.6. Distal radius scans illustrating motion artifact grading, ranging from 1 (no motion) on the left to 5 (large discontinuities) on the right. Reprinted from Pauchard et al.,[124] with permission from Elsevier.  99  Standard Protocol: The manufacturer’s (Scanco) standard protocol separates cortical from trabecular bone using a semi-automated threshold based algorithm equivalent to 1/3 the apparent density of cortical bone.[325] This step requires hand drawn contours of the periosteal surface of the bone. I contoured all 110 slices for each scan and analyzed all HR-pQCT scans. I excluded the first three and last three slices from the analysis as per the manufacturer’s protocol; thus, final values are based on 104 slices. The following parameters are directly measured from the standard analysis: Tt.BMD (mgHA/cm3), Tb.BMD (mgHA/cm3) and Tb.N (1/mm). Tb.N, the mean number of trabeculae per mm, is a truly 3D measure, and is calculated as the inverse of the mean spacing between the mid-axes of trabeculae. The following variables are derived from the standard analysis: BV/TV, Tb.Th (mm) and Tb.Sp (mm). BV/TV is calculated as: BV/TV = !".$%&		 ()*+,(-./00		 ()*+,(        (Equation 1) It is not possible to directly measure Tb.Th because the HR-pQCT voxel size approximates the average thickness of individual trabeculae; thus, trabeculae would not be resolved at their actual thickness due to partial volume effects. Therefore, standard histomorphometry techniques are used to measure trabecular thickness as:[112] Tb.Th (mm) = $1/!1!".3 4((        (Equation 2) Trabecular separation is also derived from the same measures as: Tb.Sp (mm) =  .5$1/!1!".3         (Equation 3) Standard evaluation parameters of Tb.N, BV/TV and Tb.Sp are highly correlated with micro-CT measures of human cadaver bone at the distal tibia, r = 0.64-0.91[121] and radius, r = 0.59-0.96 in adults.[118] However, validity in the growing skeleton is currently unknown. Reproducibility in our lab is 0.2% (Tb.BMD) to 1.2% (Tb.Th) for all HR-pQCT acquired standard analysis measures at the tibia and radius (University of British Columbia Bone Health Research Group, unpublished data). Automated Segmentation: The standard manufacturer’s protocol performs well for quantifying trabecular microstructure; however, its utility is limited by its inability to accurately classify cortical bone. The standard threshold-based algorithm frequently mistakes thin or porous 100  cortical bone as trabecular bone and thick trabeculae for cortex.[112] A sophisticated dual-threshold automated-segmentation algorithm was developed to more accurately separate the cortical from trabecular compartment.[115] The automated-segmentation is a two-step protocol that first extracts the periosteal surface of the cortical shell (~170 mgHA/cm3 threshold) and secondly extracts the endocortical surface of the cortical shell (~540 mgHA/cm3) (Figure 2.5). Where the manufacturer’s standard protocol would often ‘clip’ the endocortical surface of the bone by mistaking thin cortex for trabeculae or including thick trabeculae in the cortical shell, the automated segmentation algorithm more accurately identifies the endocortical surface.[115]  I applied the auto-segmentation algorithm and visually inspected all segmentations to ensure correct differentiation between cortex and trabeculae. The following variables are measured with the automated segmentation algorithm: Tt.Ar (total compartment CSA; mm2), Ct.Po (as the number of void voxels within the cortex; %), Ct.BMD (apparent density of the cortex including all pore space; mgHA/cm3), and Ct.Th (directly measured after removing intracortical pores; mm). Auto-segmentation parameters of Ct.Th and Ct.Po are highly correlated with micro-CT parameters in bone cadavers (r = 0.80 and r = 0.98, respectively).[116] The manufacturer now provides the automated segmentation algorithm as part of their standard HR-pQCT analysis software. As with the standard analysis, the validity of this algorithm for use in children and adolescents has yet to be confirmed. Finite Element Analysis: Lastly, we applied a validated FEA to HR-pQCT images to estimate bone strength.[118] Post-doctoral fellow, Dr. Mikko Määttä, conducted the FEA, in consultation with me. We generated FE meshes from 3D HR-pQCT images using the voxel conversion approach.[326,327] We simulated uniaxial compression on each tibia section up to 1% strain using a single homogenous tissue modulus of 6829 MPa and a Poission’s ratio of 0.3.[114] We used a custom FE solver (FAIM, Version 4.0, Numerics88Solutions, Calgary, Canada) on a desktop workstation (Mac Pro, OSX, Version 10.5.6, Apple Inc., Cupertino, CA, USA; 2 × 2.8 GHz Quad-Core Intel Xenon) to solve the FE models. FE outcomes were failure load (F.Load, N) and ultimate stress (U.Stress, MPa). We also calculated load-to-strength ratio of estimated fall load applied to the outstretched hand after a fall from standing height (simulation model that includes participant’s height):[119,328] Fall Load (N) = 670 /∗9.:.∗;<=>;?@.0 		     (Equation 4) 101  670 is the damping coefficient (Ns/m), height = participant height (cm) and 9.81 = gravitational constant (m/s2).  Load-to-strength ratio (φ) = ABCC	DEBF	(3)ABICJKL	DEBF	 3     (Equation 5)  2.2.9 Statistical analysis  I this section I provide an overview of the statistical analyses performed in this thesis. I performed all analyses in Stata Version 12.1 (StataCorp, College Station, TX, USA). I visually inspected all data using histograms for continuous variables and dotplots for categorical variables. For cross-sectional analyses, I examined scatterplots to assess relationships between descriptive and predictor variables against bone parameters. For longitudinal analyses, I examined scatter plots of bone parameters against maturity offset for each participant. I used these plots to identify potential measurement errors or outliers. For each statistical model (cross-sectional and longitudinal), I examined model adequacy using histograms of residuals, residual vs. fit plots and residual vs. covariate plots.  In Chapter 3, I used multivariable linear regression to investigate cross-sectional associations between sedentary time and bone parameters during adolescence. In Chapters 4 and 5, I used general linear mixed models to evaluate maturity- and sex-related differences in bone parameters at the tibial midshaft and distal tibia and radius, respectively, across adolescence. In Chapter 6, I used general linear mixed models to examine longitudinal associations between PA, sedentary time and bone parameters across adolescence. In section 1.2.9.1, I provide an overview of the general linear mixed models used in Chapters 4-6. I provide a detailed description of the statistical analysis specific to each research aim within each research chapter.  2.2.9.1 General linear mixed models  Our longitudinal data are comprised of repeated measures of bone parameters that are unique to that individual and are related to each other. Thus, I used general linear mixed models, also known as multilevel models or random coefficients models, to allow each individual to have his or her own slope and intercept, just as in a summary measures approach. However, unlike the 102  summary measures approach, mixed models imposes conditions on the distribution of the coefficients.[329] Using mixed models allows for any pattern of repeated measures and estimates the coefficient for an individual as a weighted average of the average for the sample and the individual’s ordinary least squares estimate. This means that the slope and intercept values calculated from the population data are used to pull or ‘shrink’, towards the grand mean, the slope and intercept values of those participants who had fewer data points and/or shorter time ranges between measurements.[329] By allowing for variation in both intercepts and slopes within individuals, growth velocities may vary between individuals. I used maturity offset centered at 0 as the time indicator. I provide an equation for a mixed model below:  Random linear maturity model, including the fixed effects of sex and ethnicity predicting the intercept and sex predicting the linear slope Level 1: yti = β0i + β1iMOti + εti Level 2: Intercept: β0i = γ00 +γ01Boysi + γ02Ethnicityi + µ0i Linear time: β1i = γ10 + γ11Boysi + µ1i Composite: yti = [γ00 + γ10MOti + γ01Boysi + γ11MOti*Boysi + γ02Ethnicityi] + [µ0i + µ1iMOti + εti]  MO is maturity offset (centered at 0, APHV); Boys = 0, girl; 1, boy Ethnicity= 0, Asian; 1, white; 2, other where yti is the bone parameter on measurement occasion t in the ith individual,  (µ0i, µ1i ) ~ N(0,Σ) is the vector of random effects for the ith individual and  εij ~ N(0,σ 2) is the within-subject residual error. Thus, the intercepts γ00, γ01Boysi and γ02Ethnicityi represent the mean value of the bone parameter and the fixed effect of sex and ethnicity on the mean intercept of the bone parameter when maturity offset is zero, respectively, while µ0i is the person-specific deviation from the mean intercept, assumed to follow a normal distribution with a mean of zero and variance of σ2. The slopes γ10 and γ11Boys represent the fixed linear effect of maturity and the fixed effect of sex on linear maturity at APHV, respectively, while µ1i is the person-specific deviation from the fixed linear effect of time. The slope for a given individual is γ10 + γ11Boys + µ1i, such that it will be higher or lower than the overall slope, γ10 + γ11Boys, by and amount µ1i. I specified the covariance structure as unstructured to allow the random intercepts and slopes to covary.  103  Chapter 3: Bone Architecture and Strength in the Growing Skeleton: The Role of Sedentary Time  SYNOPSIS: We k