Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Assessing somatic maturity in children and adolescents : relevance to pediatric bone health Moore, Sarah Anne 2018

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

Item Metadata

Download

Media
24-ubc_2018_september_moore_sarah.pdf [ 5.39MB ]
Metadata
JSON: 24-1.0370956.json
JSON-LD: 24-1.0370956-ld.json
RDF/XML (Pretty): 24-1.0370956-rdf.xml
RDF/JSON: 24-1.0370956-rdf.json
Turtle: 24-1.0370956-turtle.txt
N-Triples: 24-1.0370956-rdf-ntriples.txt
Original Record: 24-1.0370956-source.json
Full Text
24-1.0370956-fulltext.txt
Citation
24-1.0370956.ris

Full Text

 ASSESSING SOMATIC MATURITY IN CHILDREN AND ADOLESCENTS: RELAVANCE TO PEDIATRIC BONE HEALTH  by  Sarah Anne Moore  MSc, Brock University, 2007  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)  August 2018  ©Sarah Anne Moore, 2018 ii  The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled:  Assessing somatic maturity in children and adolescents: Applicability to pediatric bone health research  submitted by Sarah Anne Moore  in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Experimental Medicine    Examining Committee: Dr. Heather McKay, PhD Supervisor  Dr. Heather Macdonald, PhD Supervisory Committee Member  Dr. Kishore Mulpuri, MBBS Supervisory Committee Member Dr. Tricia Tang, PhD University Examiner Dr. Alexander Scott, PhD University Examiner iii  Abstract All healthy children pass through the same stages of growth, yet they do so at distinct times and tempos. Consequently, there can be large maturational differences between children of the same chronological age. Thus, it is essential to measure maturity as study results may be confounded by biological age. Yet assessing maturity is difficult. As a result, methods to predict maturity exist, however, the validity of these models have been questioned. Further, the influence of maturational timing on bone mass, density, structure, and strength remains unclear.  In this dissertation, I utilized data from healthy children who were participants in the Healthy Bones Study III (HBSIII). Three studies comprise my dissertation. First, given the need for accurate prediction of maturity, in Chapter 5 I assessed the utility of models that estimate maturity offset (MO) and age at peak height velocity (APHV), common indicators of somatic maturity. I discovered errors in the development of the original equations and subsequently, I developed and validated new models. Second, given known growth and maturation differences between ethnicities, in Chapter 6 I examined whether there were differences between Asian and white children who lived in the same community (Metro Vancouver). I observed that growth and maturation differences existed between Asian and white children, despite living in the same neighborhoods. I also assessed the validity of new maturity prediction equations for Asian children, as previous equations were developed in white children. I recalibrated the equations to better predict MO and APHV in this ethnic group. Third, given the unknown influence of maturity on post-pubertal bone health, in Chapter 7 I examined the relationship between maturational timing and bone mass, density, structure, and strength in late adolescence. Though some evidence suggests naturally delayed maturation is deleterious to bone in post-puberty, I found that post-pubertal adolescents who were considered late-maturing youth (mostly) ‘caught-up’ with those that were considered early-maturing, although there were sex- and site-related differences.  iv   Collectively these studies enhance our understanding of maturation, provide maturity prediction models for both white and Asian children, and clarify the complex relationship between maturational timing and bone mass, density, structure, and strength.  v  Lay Summary  Children grow and mature at different rates. In pediatric studies, it is important to assess and control for biological maturity. There are a number of ways to assess a child’s maturity; however, most are considered either invasive, intrusive, and/or logistically challenging. In this dissertation, I 1) developed maturity prediction equations, 2) demonstrated differences in growth and maturation between children of different ethnicities living in the same geographic areas, 3) subsequently developed ethnic-specific maturity prediction equations, and 4) utilized the new maturity equations to assess the role of maturational timing on post-pubertal bone mass, density, structure, and strength in participants from the Healthy Bones Study III. Together, these studies provide tools to non-intrusively assess maturity in Asian and white children, and clarify the maturity-bone relationship.   vi  Preface  This dissertation is an original intellectual product of the author, Sarah A. Moore. This dissertation used data collected as a part of the HBSIII; Dr. Heather McKay conceived and designed HBSIII and received ethics approval from the UBC Behaviour and Research Ethics Board (H15-01194, H07-02013, H2-70537).   Chapter 5: A version of this material was published as Moore SA, McKay HA, Macdonald H, Nettlefold L, Baxter-Jones AD, Cameron N, and Brasher PM (2015). Enhancing a somatic maturity prediction model. Med Sci Sports Exerc, 47(8), 1755-1764. As lead author I was responsible for defining the research question in collaboration with Drs. Heather McKay, Adam Baxter-Jones, Heather Macdonald, and Penny Brasher. I was responsible for HBSIII data collection (2008 to 2012), data inspection, and data cleaning. I requested and received data from the Pediatric Bone Mineral Accrual Study (PBMAS; Dr. Adam Baxter-Jones) and Harpenden Growth Study (HGS; Dr. Noel Cameron). I conducted all statistical analyses with Dr. Penny Brasher and was responsible for drafting the manuscript. All authors reviewed and provided detailed feedback on the manuscript. I presented a version of this manuscript at the North American Society for Pediatric Exercise Medicine (NASPEM) Conference in Minneapolis, Minnesota, on August 22, 2014; I received a student award for my presentation.  Chapter 6: A version of this material is being prepared for submission for publication. As in Chapter 5, it was my role to define the research question in collaboration with Drs. Heather McKay, Heather Macdonald, and Penny Brasher. I was responsible for HBSIII data collection (2008 to 2012), data inspection, and data cleaning. I conducted all statistical analyses with Dr. Penny Brasher and was responsible for drafting the manuscript. All authors reviewed and provided detailed feedback on the manuscript. I presented versions of this manuscript at the NASPEM conference in Minneapolis, Minnesota, August 22, 2014, and at the Pediatric Work Physiology (PWP) conference in Utretch, The Netherlands, on September 10, 2015; I received a student award for my presentation.  Chapter 7: A version of this material is being prepared for submission for publication. As in Chapters 5 and 6, I defined the research question in collaboration with Drs. Heather McKay and Heather Macdonald. vii  I was responsible for HBSIII data collection (2008 to 2012), data inspection, and data cleaning. I conducted all statistical analyses with input from Drs. Heather McKay, Heather Macdonald, and Lindsay Nettlefold. I was responsible for drafting the manuscript. All authors reviewed and provided detailed feedback on the manuscript. A version of this manuscript was accepted for presentation at the NASPEM Conference in Oakland, California (August 22 to 25, 2018) and as a part of the Utrecht Pediatric Summer School (August 27 to 31, 2018).  viii  Table of Contents Abstract ....................................................................................................................................................... iii Lay Summary .............................................................................................................................................. v Preface ......................................................................................................................................................... vi Table of Contents ..................................................................................................................................... viii List of Tables ............................................................................................................................................ xvi List of Figures ........................................................................................................................................... xix List of Abbreviations ............................................................................................................................ xxvii Acknowledgements ................................................................................................................................. xxx Dedication .............................................................................................................................................. xxxii Chapter 1: Introduction ............................................................................................................................. 1 Chapter 2: Literature Review .................................................................................................................... 4 2.1 Growth and Maturation ........................................................................................................... 4 2.1.1 Studies of Growth and Maturation ...................................................................................... 5 2.1.2 From Fetus to Adult State ................................................................................................... 9 2.1.2.1 Scammon’s Growth Curves ....................................................................................... 9 2.1.2.2 Prenatal Growth to Birth .......................................................................................... 10 2.1.2.3 Infancy and Childhood Growth ................................................................................ 13 2.1.2.4 Adolescent Growth and Puberty .............................................................................. 15 2.1.2.5 Cessation of Linear Growth and Attainment of Adult Size ..................................... 19 2.1.2.6 Growth of Body Segments ....................................................................................... 20 2.1.3 Assessing Growth ............................................................................................................. 21 2.1.3.1 Height and its Components ...................................................................................... 21 2.1.3.2 Body Weight and Body Mass Index ........................................................................ 22 2.1.3.3 Other Common Anthropometry in Pediatric Studies ............................................... 23 ix  2.1.4 Assessing Maturity ........................................................................................................... 23 2.1.4.1 Skeletal Maturity ...................................................................................................... 23 2.1.4.2 Sexual Maturity ........................................................................................................ 26 2.1.4.3 Somatic Maturity ...................................................................................................... 29 2.1.4.4 Relationship between Maturity Indicators ............................................................... 56 2.1.5 Secular Changes in Growth and Maturation ..................................................................... 56 2.1.5.1 Secular Changes in Growth ...................................................................................... 57 2.1.5.2 Secular Changes in Maturation ................................................................................ 59 2.1.6 Factors that Influence Growth and Maturation ................................................................. 61 2.1.6.1 Genetics .................................................................................................................... 61 2.1.6.2 Ethnicity ................................................................................................................... 62 2.1.6.3 Sex ............................................................................................................................ 69 2.1.6.4 Hormones ................................................................................................................. 71 2.1.6.5 Nutrition ................................................................................................................... 74 2.1.6.6 Physical Activity ...................................................................................................... 76 2.1.7 Summary of Growth and Maturation Literature ............................................................... 77 2.2 Influence of Growth and Maturation on Bone Accrual ......................................................... 78 2.2.1 Why Assess Bone During the Growing Years? ................................................................ 78 2.2.2 Bone Tissue Anatomy and Physiology ............................................................................. 79 2.2.2.1 Bone Matrix and Cells ............................................................................................. 80 2.2.2.2 Long Bone Geometry ............................................................................................... 83 2.2.2.3 Bone Modeling and Remodeling.............................................................................. 84 2.2.2.4 Endochondral Ossification ....................................................................................... 85 2.2.2.5 Growth Zones of the Epiphysis ................................................................................ 86 2.2.2.6 Epiphysis Closure..................................................................................................... 87 x  2.2.2.7 Peak Bone Mass ....................................................................................................... 87 2.2.3 Assessing Bone in Children and Adolescents ................................................................... 89 2.2.3.1 Dual Energy X-ray Absorptiometry ......................................................................... 89 2.2.3.2 Peripheral Quantitative Computed Tomography ..................................................... 91 2.2.4 Factors that Influence Bone in Childhood and Adolescence ............................................ 93 2.2.4.1 Genetics .................................................................................................................... 94 2.2.4.2 Ethnicity ................................................................................................................... 94 2.2.4.3 Sex ............................................................................................................................ 96 2.2.4.4 Hormones ................................................................................................................. 98 2.2.4.5 Nutrition ................................................................................................................... 99 2.2.4.6 Physical Activity .................................................................................................... 101 2.2.4.7 The Role of Maturity in Bone Accrual................................................................... 104 2.3 Summary of the Growth, Maturation and Bone Accrual Literature .................................... 109 Chapter 3: Rationale, Objectives, Hypotheses, and Contributions .................................................... 110 3.1 STUDY 1: Enhancing a Somatic Maturity Prediction Model ............................................. 111 3.1.1 Rationale ......................................................................................................................... 111 3.1.2 Objectives ....................................................................................................................... 111 3.1.3 Hypotheses ...................................................................................................................... 112 3.1.4 Contribution .................................................................................................................... 112 3.2 STUDY 2: Growth and Maturation of Asian-Canadian Children ....................................... 113 3.2.1 Rationale ......................................................................................................................... 113 3.2.2 Objectives ....................................................................................................................... 113 3.2.3 Hypotheses ...................................................................................................................... 114 3.2.4 Contribution .................................................................................................................... 114 xi  3.3 STUDY 3: Does Maturational Timing Predict Bone Mass, Density, Structure, and Strength in Late-Adolescence? ....................................................................................................................... 115 3.3.1 Rationale ......................................................................................................................... 115 3.3.2 Objectives ....................................................................................................................... 116 3.3.3 Hypotheses ...................................................................................................................... 116 3.3.4 Contribution .................................................................................................................... 116 Chapter 4: Methods ................................................................................................................................ 118 4.1 Healthy Bones Study III ...................................................................................................... 118 4.1.1 Study Design ................................................................................................................... 118 4.1.2 My Role in the Healthy Bones Study III ........................................................................ 121 4.1.3 Participant Recruitment .................................................................................................. 121 4.1.4 Participant Retention and Attrition ................................................................................. 122 4.1.5 Comparison Cohorts ....................................................................................................... 126 4.2 Measurements...................................................................................................................... 126 4.2.1 Planning for Measurement and Participant Flow ............................................................ 126 4.2.2 Anthropometry ................................................................................................................ 127 4.2.2.1 Standing Height, Sitting Height, Leg Length, and Tibial Length .......................... 128 4.2.2.2 Weight and Body Mass Index ................................................................................ 128 4.2.3 Questionnaires ................................................................................................................ 129 4.2.3.1 Health History and Ethnicity .................................................................................. 129 4.2.3.2 Physical Activity and Dietary Calcium Intake ....................................................... 130 4.2.3.3 Maturational Status ................................................................................................ 131 4.2.4 Bone Parameters ............................................................................................................. 131 4.2.4.1 Dual Energy X-Ray Absorptiometry ...................................................................... 131 4.2.4.2 Peripheral Quantitative Computed Tomography ................................................... 133 xii  4.2.5 Data Preparation and Statistical Analyses ...................................................................... 134 4.2.5.1 Medical Exclusions ................................................................................................ 135 4.2.5.2 Missing Data .......................................................................................................... 135 4.2.5.3 Data Cleaning ......................................................................................................... 136 4.2.5.4 Assigning Age at Peak Height Velocity ................................................................. 137 4.2.5.5 Assigning Adult Height in Late-Adolescence ........................................................ 139 4.3 Descriptive Characteristics .................................................................................................. 139 Chapter 5: Enhancing a Somatic Maturity Prediction Model ............................................................ 143 5.1 Introduction ......................................................................................................................... 143 5.2 Methods ............................................................................................................................... 146 5.2.1 Study Participants ........................................................................................................... 146 5.2.2 Anthropometry ................................................................................................................ 148 5.2.3 Observed Age at Peak Height Velocity .......................................................................... 148 5.2.4 Statistical Analyses ......................................................................................................... 150 5.3 Results ................................................................................................................................. 151 5.4 Discussion ........................................................................................................................... 162 5.4.1 Overview of Study Findings ........................................................................................... 162 5.4.2 Assessing the Development of the Mirwald Equations .................................................. 163 5.4.3 Developing New Models to Predict Maturity ................................................................. 163 5.4.4 Application of Maturity Prediction Models .................................................................... 164 5.4.5 Limitations ...................................................................................................................... 165 5.5 Conclusion ........................................................................................................................... 165 Chapter 6: Growth and Maturation of Asian-Canadian Children .................................................... 167 6.1 Introduction ......................................................................................................................... 167 6.2 Methods ............................................................................................................................... 170 xiii  6.2.1 Study Participants ........................................................................................................... 170 6.2.2 Anthropometry ................................................................................................................ 171 6.2.3 Somatic Maturity ............................................................................................................ 171 6.2.4 Sexual Maturity ............................................................................................................... 172 6.2.5 Statistical Analyses ......................................................................................................... 172 6.3 Results ................................................................................................................................. 174 6.3.1 Baseline Characteristics .................................................................................................. 174 6.3.2 Ethnic-related Differences in Growth and Maturation ................................................... 176 6.3.2.1 Growth Differences ................................................................................................ 178 6.3.3 Maturity Differences ....................................................................................................... 178 6.3.3.1 Somatic Maturation ................................................................................................ 178 6.3.3.2 Sexual Maturation .................................................................................................. 179 6.3.3.3 Sequence of Maturational Events ........................................................................... 179 6.3.4 Implications for Predicting Maturity .............................................................................. 181 6.4 Discussion ........................................................................................................................... 184 6.4.1 Overview of Study Findings ........................................................................................... 184 6.4.2 Differences in Growth between Ethnicities .................................................................... 184 6.4.3 Differences in Maturation between Ethnicities ............................................................... 186 6.4.4 Predicting Maturity in Asian Children ............................................................................ 188 6.4.5 Limitations ...................................................................................................................... 188 6.5 Conclusion ........................................................................................................................... 190 Chapter 7: Does Maturational Timing Predict Bone Mass, Density, Structure, and Strength in Late-Adolescence? ................................................................................................................................... 191 7.1 Introduction ......................................................................................................................... 191 7.2 Methods ............................................................................................................................... 193 xiv  7.2.1 Participants...................................................................................................................... 193 7.2.2 Anthropometry and Body Composition .......................................................................... 195 7.2.3 Somatic and Sexual Maturity .......................................................................................... 195 7.2.4 Bone Mineral Content, Density, Structure, and Strength ............................................... 196 7.2.5 Physical Activity and Dietary Calcium ........................................................................... 197 7.2.6 Statistical Analyses ......................................................................................................... 197 7.3 Results ................................................................................................................................. 199 7.3.1 Descriptive Characteristics ............................................................................................. 199 7.3.2 Bone Mineral Content and Areal Bone Mineral Density ................................................ 201 7.3.3 Bone Structure and Cortical Density .............................................................................. 201 7.3.4 Bone Strength ................................................................................................................. 202 7.4 Discussion ........................................................................................................................... 205 7.4.1 Overview of Study Findings ........................................................................................... 205 7.4.2 Is Late-Maturational Timing Deleterious to Adult Mass? .............................................. 205 7.4.3 Is Late-Maturational Timing Deleterious to Adult Bone Structure, Density, and Strength? 208 7.4.4 What are Determinants of Bone Mass, Density, Structure, and Strength? ..................... 209 7.4.5 Limitations ...................................................................................................................... 210 7.5 Conclusion ........................................................................................................................... 211 Chapter 8: Integrated Discussion and Conclusion ............................................................................... 212 8.1 Overview of Findings and Contributions ............................................................................ 212 8.1.1 Chapter 5: Enhancing a Somatic Maturity Prediction Model ......................................... 212 8.1.2 Chapter 6: Growth and Maturation in Asian-Canadian Children ................................... 214 8.1.3 Chapter 7: Does Maturational Timing Prediction Bone Mass, Density, Structure, and Strength in Late-Adolescence ...................................................................................................... 217 xv  8.2 Challenges and Limitations of Studying Children’s Growth and Maturation ..................... 220 8.2.1 Longitudinal Studies of Children and Adolescents......................................................... 220 8.2.2 Assessing Ethnicity ......................................................................................................... 222 8.2.3 Assessing Maturation and Maturity Indicators ............................................................... 223 8.2.4 Assessing Pediatric Bone and Bone Accrual into Adulthood ......................................... 224 8.2.5 Assessing Peak Bone Mass ............................................................................................. 225 8.3 Implications and Future Research ....................................................................................... 225 8.3.1 Implications of My Dissertation ..................................................................................... 225 8.3.2 Future Research .............................................................................................................. 226 8.3.3 Conclusion ...................................................................................................................... 229 References ................................................................................................................................................ 231 Appendices ............................................................................................................................................... 259 Appendix A: Detailed Healthy Bones Study III Description .............................................................. 260 Appendix B: Example Information to Participants, Consent and Assent Forms ............................. 265 Appendix C: Example Participants Update and Results Letter ......................................................... 272 Appendix D: Relevant HBSIII Questionnaires .................................................................................... 276 Appendix E: Additional Chapter 7 Data .............................................................................................. 298  xvi  List of Tables Table 2.1 Summary of historical longitudinal studies of children’s growth and maturation. Adapted from Himes [40], with permission from Food and Nutrition Bulletin, Sage Publishing. ...................................... 8 Table 2.2 Description of the stages and growth events in the human lifecycle. Adapted from Cameron & Bogin [20], with permission from Elsevier Publishing. .............................................................................. 11 Table 2.3 Demonstration of the decomposition of the sexual dimorphism in adult size of height, trunk height, shoulder width, and hip width in a British sample of boys and girls. Adapted from Hauspie et al. [77], with permission from Taylor and Francis Publishing. ....................................................................... 17 Table 2.4 Demonstration of a worked sample calculation of height velocity (cm/year) and years from age at peak height velocity (APHV, years) or maturity offset (MO; years) in a girl with idiopathic scoliosis. Adapted from Little et al. [122], with permission from the Journal of Bone and Joint Surgery, Inc., and Rockwater, Inc. ........................................................................................................................................... 30 Table 2.5 Description of the age at peak height velocity (APHV, years), magnitude of PHV (cm/year), and age at take-off of PHV (years) in boys from six selected longitudinal studies. Studies are listed in order by publication date and represent cohorts from Europe, North America, and South Asia. ............... 45 Table 2.6 Description of the age at peak height velocity (APHV; years), magnitude of PHV (cm/year), and age at take-off of PHV (years) in girls from six selected longitudinal studies. Studies are listed in order by publication date and represent cohorts from Europe, North America, and South Asia. ............... 46 Table 2.7 Demonstration of a worked example of maturity offset (MO) prediction in an average-maturing boy. Adapted from Sherar et al [17], with permission from the Journal of Pediatrics, Mosby, Inc. .......... 51 Table 2.8 Demonstration of the difference of predicted age at peak height velocity (APHV) minus observed APHV in a cohort of Polish boys. Age group defined by mid-year (i.e., 9 = 8.5-9.49 years). Reproduced from Malina & Koziel [28], with permission from Journal of Sports Sciences, Taylor & Francis Ltd. ................................................................................................................................................. 52 xvii  Table 2.9 Overview of studies that described the relationship between maturational timing and bone outcomes in young adulthood. .................................................................................................................. 105 Table 4.1 Description of the Healthy Bones Study III (HBSIII) sample by sex (b=boys; g=girls), ethnicity by category (Asian/white/other), and measurement period; where HBSII is the original Healthy Bones Study II cohort, B@B is the Bounce at the Bell study cohort, AS!BC is the Action Schools! BC study cohort, and HBS2009 is the newly recruited HBSIII cohort. Participants were given the opportunity to re-enter the study if they missed a measurement period. Total number of participants measured at each period is listed as well as the total number of observations (obs) for the study. ....................................... 124 Table 4.2 Description of the number of participants at each age by measurement period (1999 to 2012) and age group category (9 to 23 years) in the Healthy Bones Study III. .................................................. 141 Table 4.3 Description of the baseline characteristics by sex (boys and girls) and ethnicity (Asian, white, other/mixed ethnicities) in the Healthy Bones Study III. .......................................................................... 142 Table 5.1 Description of the sample size, number of observations, test occasions, and mean age at peak height velocity (APHV, determined by interpolating cubic spline; in years) in the three studies: the Pediatric Bone Mineral Accrual Study (PBMAS), Healthy Bones Study III (HBSIII), and the Harpenden Growth Study (HGS). ............................................................................................................................... 152 Table 5.2 Illustration of the incremental changes in R2 and the standard error of the estimate (SEE) with step-by-step refitting of the Mirwald equations. ....................................................................................... 153 Table 5.3 Illustration of the results from the forward stepwise regression from the original Pediatric Bone Mineral Accrual Study boys (n=79) and girls (n=72), which employs a cluster-robust variance estimator that accounts for within-subject correlation by estimating robust standard errors (SE) of the coefficients. .................................................................................................................................................................. 154 Table 5.4 Illustration of the results from the 5x7- and 5x6-fold random-splitting analysis with boys and girls from the Pediatric Bone Mineral Accrual Study, respectively. This created 35 subsets in boys and 30 xviii  subsets in girls (with one observation per child; where n≥70). We employed a forward stepwise regression procedure with P ≤ 0.10 for entry and P ≥ 0.11 for removal. ................................................................... 155 Table 5.5 Summary of the results of the calibration curves and descriptive summaries of the prediction residuals (including 25th 50th and 75th percentiles) for the external validation samples: Healthy Bones Study III (HBSIII) and Harpenden Growth Study (HGS), where b0 is the calibration curve intercept (standard error, SE), b1 is the calibration curve slope (SE), and RMSE is the root mean squared error of the prediction. ........................................................................................................................................... 158 Table 5.6 Description of the summary statistics (mean (SD)) for observed maturity offsets (MO), predicted MOs, and prediction errors for the new equations and Mirwald equations (observed minus predicted) by observed MO category, in Healthy Bones Study III (HBSIII, top) and Harpenden Growth Study (HGS, bottom) boys (left) and girls (right). .................................................................................... 160 Table 6.1 Description of the sample characteristics at baseline and adulthood, and maturity indicators in Asian and white boys and girls from the Healthy Bones Study III. Values are mean ± standard deviation unless otherwise indicated. ....................................................................................................................... 175 Table 6.2 Description of the growth and maturity indicators in Asian and white boys from the Healthy Bones Study III. Values are mean ± standard deviation unless otherwise indicated. ............................... 176 Table 6.3 Description of the growth and maturity indicators in Asian and white girls from the Healthy Bones Study III. Values are mean ± standard deviation unless otherwise indicated. ............................... 177 Table 6.4 Summary of the prediction error and calibration coefficients for Asian boys and girls from the Healthy Bones Study III (HBSIII) using the new equations in boys and girls, and the alternate (age * height) equation in boys; where b0 is the calibration curve intercept (standard error, SE), b1 is the calibration curve slope (SE), and RMSE is the root mean squared error of the prediction. ..................... 183 Table 7.1 Description of the characteristics of Asian and white late adolescent boys and girls from the Healthy Bones Study III (HBSIII), presented as mean ± standard deviation unless otherwise specified. 200  xix  List of Figures Figure 2.1 Illustration of substantial differences in biological age when comparing three boys at the same chronological age (14 years). The boy on the left may be considered late-maturing, the boys in the middle may be considered average-maturing, and the boy on the right may be considered early-maturing. Reproduced from Tanner, J.M. [35], with permission from Castlemead Publications. ................................ 5 Figure 2.2 Illustration of height velocity (cm/year) calculated from cross-sectional (X; dashed line), longitudinal (L; dotted line), and mixed longitudinal (M; solid line) height measurements, by age and sex in Southern Chinese children from Hong Kong. Reproduced from Low [36], with permission from Schweizerbart Science Publishers. ................................................................................................................ 6 Figure 2.3 Drawing of Scammon’s growth curve of different parts and tissues of the body. The data shown are as a percentage of growth for each system between birth and 20 years of age. Reproduced from Scammon [53], with permission from the American Journal of Physical Anthropology, John Wiley & Sons, Inc. ....................................................................................................................................................... 9 Figure 2.4 Illustration of the average changes in body proportion from infancy to adulthood. Reproduced from Huelke [72], with permission from the Association for the Advancement of Automotive Medicine. .................................................................................................................................................................... 14 Figure 2.5 A mid-childhood spurt is observed in two-thirds of healthy children between childhood (C) and juvenile (J) years. This illustration also depicts height distance and velocity in infancy (I), adolescence (A), and adulthood (mature state, M) in boys (solid line) and girls (dotted line). Reproduced from Cameron & Bogin [20], with permission from Elsevier Publishing. ................................................. 15 Figure 2.6 Illustration of typical height distance and velocity curves in boys and girls. The left graph shows distance curves and the right graph shows velocity curves in boys (solid) and girls (dotted) from infancy to adulthood in healthy children. Adapted from Malina, Bouchard, & Bar-Or [19], with permission from Human Kinetics Publishing and Archives of Disease in Childhood, BMJ Publishing Group Ltd. ................................................................................................................................................... 16 xx  Figure 2.7 Illustration of distance (left) and velocity (right) curves for weight in boys (solid) and girls (dotted) from infancy to adulthood in healthy children. Reproduced from Malina, Bouchard, & Bar-Or [19], with permission from Human Kinetics Publishing and Archives of Disease in Childhood, BMJ Publishing Group Ltd. ................................................................................................................................. 18 Figure 2.8 Illustration of the sequence of pubertal events in boys (left) and girls (right), where arrow and number represent the mean age for each pubertal event. Reproduced from Haqq, Boston, & LaFranchi [78], with permission from Lippincott Williams & Wilkins....................................................................... 18 Figure 2.9 Demonstration of the measurement of height (left) from bottom of bare feet to vertex of skull in Frankfort plane with participants heels, buttocks, and back against the stadiometer, and measurement of and sitting height (right) using platform of known height with participant’s hands in lap. Adapted from Frisancho [89], with permission from the University of Michigan Press. .................................................. 22 Figure 2.10 Illustration of a hand-wrist radiograph of an eight year old boy. Specific bones are shown in the schematic, excluding the pisiform and adductor sesamoid. Horizontal lines indicate the locations of the epiphysis of each long bone. Reproduced from Malina, Bouchard, & Bar-Or [19], with permission from Human Kinetics Publishing. ............................................................................................................... 24 Figure 2.11 Illustration of the growth in height of Montbeillard's son from birth to 18 years, 1759 to 1777. The distance curve (left panel) indicates height attained at each age, and velocity curve (right panel) indicates the increment of height gained from year to year. Reproduced from Tanner [1] and drawn with data from Scammon [123], with permission from the American Journal of Physical Anthropology, and John Wiley & Sons, Inc. ............................................................................................................................. 31 Figure 2.12 Illustration of the typical individual velocity curves for height in boys and girls. Reproduced from Malina, Bouchard, & Bar-Or [19], with permission from Human Kinetics Publishing and Archives of Disease in Childhood, BMJ Publishing Group Ltd. ............................................................................... 32 Figure 2.13 Illustration of the height velocity (cm/years) aligned to chronological age (left) and years from peak height velocity (right) in early, average, and late-maturing healthy boys from the Saskatchewan xxi  Growth and Development and Pediatric Bone Mineral Accrual Studies. Reproduced from Sherar et al [17], with permission from the Journal of Pediatrics, Mosby, Inc. ............................................................. 33 Figure 2.14 Representation of a distance (solid line) and velocity (dashed line) curves for longitudinal height data for an example boy using Preece-Baines model 1 curve-fitting. Where h1 is adult size; hϴ is the size at age Ɵ; s0 and s1 are growth rate constants related to prepubertal and pubertal velocity. Reproduced from Hauspie, Cameron, & Molinari [25], with permission from Cambridge University Press. ........................................................................................................................................................... 35 Figure 2.15 Illustration of height velocity curves of a girl in the First Zurich Longitudinal study from birth to 20 years, where dots (stars) are the raw mid-point calculated velocities, sold line is the kernel estimation fit, and dashed line is the Preece-Baines model 1 fit. Reproduced from Hauspie, Cameron, & Molinari [25], with permission from Cambridge University Press. ............................................................ 37 Figure 2.16 Demonstration of velocity growth curves of Montebeillard’s son replicated in Stata. The black dots represent the calculated velocities. The black line shows the interpolated cubic spline fit through all knots. The blue line shows a smoothing cubic spline where a roughness penalty has been introduced of 0.000006. The grey horizontal line indicates a roughness penalty of zero. Drawn by Dr. Penny Brasher, PhD, Centre for Clinical Epidemiology and Evaluation, University of British Columbia with data from Scammon [123], with permission from the American Journal of Physical Anthropology, John Wiley & Sons, Inc. ............................................................................................................................. 39 Figure 2.17 Demonstration of the incremental adjustments for a) size (shifting curves vertically), b) tempo (shifting curves horizontally), c) velocity (shrinking or stretching scale), and d) removing extreme outliners, in the SuperImposition by Translation and Rotation (SITAR) model to create average distance curve. The same method can be applied to create velocity curves illustrating mean age at peak height velocity (APHV). Reproduced from Cole, Donaldson, & Ben-Shlomo [153], with permission from the International Journal of Epidemiology, Oxford University Press. .............................................................. 41 xxii  Figure 2.18 Illustration of the limitations associated with prospective studies if a participant is lost to follow-up when determining age at peak height velocity (APHV) using data acquired from a boy in the Healthy Bones Study III (HBSIII). The open circles in the rightmost curve show height (cm; Y-axis) at each measurement, by age (years; X-axis). The connected closed circles in the leftmost curve show velocities (and interpolations) between measures (cm/year; Y-axis). The open square indicates the identified APHV. ........................................................................................................................................ 49 Figure 2.19 Illustration of median values for velocities of height, sitting height (i.e., trunk length), and leg length for boys (left) and girls (right) aligned by age at peak height velocity (APHV), where 0 equals APHV, negative values indicate years pre-APHV and positive values indicate years post-APHV. Reproduced from Buckler [52], with permission from Springer-Verlag Publishing. ................................. 54 Figure 2.20 Illustration of the relationship between and ranges of somatic maturity as measured using age at peak height velocity and sexual maturity in girls (left) and boys (right) Numbers below represent the range for the respective indicator. Reproduced from Danker-Hopfe [211] Drawn from data from Tanner [1], with permission from Progress in Biophysics and Molecular Biology, Pergamon Publishing, and John Wiley & Sons, Inc. ...................................................................................................................................... 56 Figure 2.21 Illustration of the mean height-for-age in Canadian boys (left) and girls (right) in four birth cohorts indicating a positive secular change from 1891 to 1974. Adapted from Hoppa [218], with permission from Annals of Human Biology, Taylor & Francis, Ltd. ......................................................... 59 Figure 2.22 Illustration of the secular changes in age at menarche in various nations from 1960 to 2000. Reproduced from Parent et al. [226], with permission from Endocrine Society. ....................................... 60 Figure 2.23 Illustration of mean height from birth to 18 years of girls (left) and boys (right) from European (Denmark), Asian (Japan), and African (America) origins. Reproduced from Eveleth & Tanner [30], with permissions from Cambridge University Press. ......................................................................... 64 Figure 2.24 Illustration of skeletal maturity scores (by Tanner-Whitehouse 2 method) by chronological age for boys (left) and girls (right) in Asian children from Japan (solid line), white children from the UK xxiii  (solid circle), white children from Belgium (open circle), Asian children from North India (solid diamond), and Asian children from South China (open square). Reproduced from Murata [272], with permissions from Acta Paediatrica, John Wiley & Sons, Inc. .................................................................... 66 Figure 2.25 Illustration of the height change (cm) from age 11 years to age 17 years in boys (left) and from age 10 years to 17 years in girls (right) for selected samples from four regions. Reproduced from Haas & Campirano [280], with permission from Food and Nutrition Bulletin, Sage Publications, Inc. .... 68 Figure 2.26 Illustration of the peaks for height velocity, bone mineral content (BMC) velocity, growth hormone (GH) amplitude, and insulin-like growth factor-1 (IGF-1) amplitude, and trends for estrogen and testosterone levels in girls relative to average age and Tanner stage. In boys, peak height velocity and peak BMC velocity occur approximately 1.5 years later compared with girls. Relations between peaks for height and bone velocities and peaks for GH and IGF-1 are similar for boys and girls. Reproduced from Mackelvie et al. [289], with permission from the British Journal of Sports Medicine, BMJ Publishing Group. ......................................................................................................................................................... 70 Figure 2.27 A schematic representation of the production of growth hormone showing the regulatory influences, hormonal interactions, and effects; where: GHRH, growth hormone-releasing hormone; SRIH, somatotropin release-inhibiting hormone; IGF-1, insulin-like growth factor-1; +ve, amplifies response; -ve, inhibits response. Redrawn from Roche & Sun [142], with permission from Cambridge University Press. ........................................................................................................................................................... 72 Figure 2.28 A schematic representation of the production of testosterone (left) and estradiol (right) showing the regulatory influences, hormonal interactions, and effects; where: GnRh, gonadotropin-releasing hormone; LH, luteinizing hormone; FSH, follicle-stimulating hormone; IGF-1, insulin-like growth factor-1; +ve, amplifies response; -ve, inhibits response; APHV, age at peak height velocity. Redrawn from Roche & Sun [142], with permission from Cambridge University Press. .......................... 74 Figure 2.29 Illustration of the percentage of total body bone mineral content (BMC) gained during adolescent growth in boys (left) and girls (right) from the Pediatric Bone Mineral Accrual Study. xxiv  Approximately 40% of BMC is accrued in the ± two years from age at peak height velocity (APHV) in both sexes. Adapted from Baxter-Jones et al. [342], with permission from the Journal of Bone and Mineral Research, John Wiley & Sons, Inc. ............................................................................................... 79 Figure 2.30 Illustration of the structural elements of a long bone, with cortical bone (outer) comprised of osteons surrounding trabecular bone (inner). Reproduced from Seeley et al. [354], with permission from McGraw-Hill. .............................................................................................................................................. 81 Figure 2.31 Illustration of the proximal femur showing trabeculae oriented along lines of stress. The alignment of trabeculae in this manner resists bending and torsional movements. The trabecular bone is surrounded by cortical bone (outer shell). Reproduced from Seeley et al. [354], with permission from McGraw-Hill. .............................................................................................................................................. 82 Figure 2.32 Illustration of the aspects of human long bone (femur) including the diaphysis, epiphysis, and metaphysis. Reproduced from Kontulainen et al. [360], with permission from Karger. ............................ 83 Figure 2.33 Illustration that depicts stages of endochondral ossification from formation of the bone collar (stage 1) to fully ossified bone in young adulthood (stage 5). Reproduced from Seeley et al. [354], with permission from McGraw-Hill. .................................................................................................................. 85 Figure 2.34 Illustration of the major features in the epiphysis. Reproduced from Scheuer & Black [377], with permission from Elsevier. ................................................................................................................... 87 Figure 2.35 Illustration of bone mass accrual across the lifespan. The dot-dash line represents when peak bone mass (PBM) is compromised during adolescence due to suboptimal lifestyle factors, whereas the solid and dashed lines represent the theoretical increase in PBM with optimal lifestyle factors. The dark shaded area represents a period where bone mass may be low, and the light shaded area where bone mass may be considered osteoporotic and at risk for osteoporosis-related fracture. Reproduced from Weaver et al. [387], with permission from Osteoporosis International, Springer-Verlag London Ltd. ....................... 88 xxv  Figure 2.36 Illustration of the relationship between age at peak height velocity (APHV) and peak bone mineral content velocity by chronological age for boys and girls in the Pediatric Bone Mineral Accrual Study. Reproduced from Bailey et al [85], with permission from John Wiley & Sons, Inc. ...................... 89 Figure 2.37 Illustration of the partial volume effect (PVE), whereby voxels at bone edges (dark grey) contain both soft and bone tissue densities, resulting in a lower density for the dark grey voxels. Smaller bones (e.g., radius) have more voxels close to the bone edge and are at greater risk for PVE. Reprinted from Zemel et al. [406], with permission from Elsevier. ............................................................................ 93 Figure 2.38 Representation of the differences in total area (Tt.Ar), medullary area (Me.Ar), cortical area (Ct.Ar), cortical BMD (Ct.BMD), and strength-strain index (SSIp) between boys and girls by maturity offset (MO) in children from the Healthy Bones Study III. Reproduced from Gabel et al. [458] , with permission from the Journal of Bone and Mineral Research, John Wiley & Sons, Inc. ............................. 97 Figure 4.1 This figure provides an overview of the mixed-longitudinal design of the Healthy Bones Study III (HBSIII). It illustrates sample size by individual study cohorts and total sample size for HBSIII by sex and ethnicity. Light shaded boxes represent when two or more cohorts were measured simultaneously. Dark shaded boxes represent when in school measurements were performed (all other measurements were taken in our bone health laboratory). n(participants) =1071; n(observations; obs) =7943. ...................... 120 Figure 4.2 This figure illustrates dual energy X-ray absorptiometry (DXA) acquired images of the whole body (left), lumbar spine (middle, top), and proximal femur (middle, bottom) from one child who participated in the Healthy Bones Study III (HBSIII). Lines outline regions of interest as defined for standard DXA analysis of each site. On the right is the DXA imaging suite at the Centre for Hip Health and Mobility (CHHM). Images and photo courtesy of CHHM. ............................................................... 132 Figure 4.3 Description of the scans acquired using peripheral quantitative computed tomography (pQCT). On the left, I illustrate positioning of the reference line at the 50% site of the tibia. In the middle (bottom) is a scout view of the reference line position; in the middle (top) is an image at the 50% site of the tibia. xxvi  On the right is a photo of the current pQCT (XCT-3000) imaging suite at the Centre for Hip Health and Mobility (CHHM). Images and photo courtesy of CHHM. ...................................................................... 134 Figure 5.1 Illustration of the calibration curves (observed versus predicted) in HBSIII boys (A), HBSIII girls (B), HGS boys (C), and HGS girls (D); where thick light grey line is y=x, black solid line is calibration line for new equations; black open dots are predicted maturity offset (MO) by new equations, black dashed line is calibration line for Mirwald equations, black dots are predicted MO by Mirwald equations, and dark grey thick dotted line (C only) is new equations to account for significant difference in APHV.  If b0=0 and b1=1, the calibration line would overlay the y=x line; that is, the closer b0 to 0 and b1 to 1, the better the performance of the prediction on average. ............................................................. 157 Figure 5.2 Histogram of the prediction error by observed maturity offset category in Healthy Bones Study-III (HBSIII) boys (A), HBSIII girls (B), Harpenden Growth Study (HGS) boys (C), and HGS girls (D) starting at -3 (top) to +3 (bottom), respectively.................................................................................. 161 Figure 6.1 Average height velocity curves in boys (top) and girls (bottom) by ethnicity (Asian, black line; white, grey line), where age is indicated on the x-axis and height velocity (in cm/year) is indicated on the y-axis. Average age of attainment of pubic hair stage for Asian and white boys and girls, and breast stage for girls, by ethnicity is noted below the velocity curve in the grey shaded boxes. .................................. 180 Figure 6.2 Illustration of scatter plots of observed versus predicted maturity offset and calibration curves in Asian boys and girls. Figures (A) and (B) use equations derived in white children in Asian boys and girls, respectively. Figures (C) and (D) use recalibrated equations for Asian boys and girls, respectively. The thick grey line is y=x and the solid black line is the calibration line. If b0=0 and b1=1, the calibration line would overlay the y=x line; that is, the closer b0 to 0 and b1 to 1, the better the performance of the prediction, on average. The histograms show the distribution of the prediction errors. ........................... 182 Figure 7.1 Illustration of the derivation of the final sample size for this analysis. .................................. 194  xxvii  List of Abbreviations  Abbreviation Description 2D/3D Two-dimensional / Three-dimensional aBMD Areal bone mineral density  APHV Age at peak height velocity AS!BC Action Schools! BC B@B Bounce at the Bell B1-5 Breast (stage 1-5) BA Bone area BC British Columbia BMC Bone mineral content BMD (Volumetric) bone mineral density BMI Body mass index BMU Basic multicellular unit CaMOS Canadian Multicentre Osteoporosis Study CHHM Centre for Hip Health and Mobility CI Confidence interval CSA Cross-sectional area Co.Po Cortical bone porosity Ct.Ar Cortical bone area Ct.BMD Cortical bone mineral density Ct.Th Cortical bone thickness CV% Coefficient of variation DXA Dual energy X-ray absorptiometry DZ Dizygotic fraternal twins FFM Fat free mass FFQ Food frequency questionnaire FN Femoral neck FS Femoral shaft FSH Follicle stimulating hormone G1-5 Genitalia (stage 1-5) GH Growth hormone GHRH Growth hormone-releasing hormone GnHR Gonadotropin-releasing hormone H2 Heritability HBS/ HBSIII Healthy Bones Study / Healthy Bones Study III HGS Harpenden Growth Study HHANES Hispanic Health and Nutrition Examination Study HHQ Health history questionnaire HPG Hypothalamus-pituitary-gonadal  HR-pQCT High resolution peripheral quantitative computed tomography HAS Hip structural analysis IBDS Iowa Bone Development Study IC Infancy-childhood  ICC Intraclass correlation coefficient ICP Infancy-childhood-puberty IGF-1 Insulin-like growth factor 1 IQR Interquartile range xxviii  Abbreviation Description IT Intertrochanteric IU International units IUGR Intrauterine growth restriction LH Luteinizing hormone  LLTS Leuven Longitudinal Twin Study LM Lean mass LMP Last menstrual period LS Lumbar spine µFEA Micro finite element analysis µSV MicroSieverts MC Mid-childhood MCSA Muscle cross-sectional area Me.Ar Medullary area METs Metabolic equivalents MIA Maximal increment age MO Maturity offset MRI Magnetic resonance imaging MSE Mean square error MVPA Moderate to vigorous physical activity MZ Monozygotic identical twins NHANES National Health and Nutritional Examination Survey NHES National Health and Examination Survey NN Narrow neck NORA National Osteoporosis Risk Assessment PA Physical activity PAQ-C/PAQ-A Physical Activity Questionnaire for Children /Adolescents PB1 Preece-Baines model 1 PBM Peak bone mass PBMAS Pediatric Bone Mineral Accrual Study PBMCV Peak bone mineral content velocity PF Proximal femur PH1-5 Pubic hair (stage 1-5) PHV Peak height velocity pQCT Peripheral quantitative computed tomography PVE Partial volume effect R2 Coefficient of determination RCT  Randomized controlled trial RMSE Root mean square error ROI Region of interest SD Standard deviation SE Standard error SEE Standard error of the estimate SES Socioeconomic status SGDS Saskatchewan Growth and Development Study SRIH Somatotropin release-inhibiting hormone SSIp Polar strength strain index Tb.Ar Trabecular bone area Tb.BMD Trabecular bone mineral density Tb.N Trabecular bone number xxix  Abbreviation Description Tb.Th Trabecular bone thickness Tt.Ar Total bone area Tt. BMD Total bone mineral density TW Tanner Whitehouse UBC University of British Columbia UK United Kingdom US/USA United States / United States of America VGH Vancouver General Hospital WB Whole body WBLH Whole body less head WWI/WWII World War I / World War II Z Section modulus    xxx  Acknowledgements  The journey of graduate studies is a rewarding one in which we surround ourselves with a supportive network of family and friends to make it through. I have valued the opportunity to learn and grow broadly during my dissertation and am so thankful for those that were in my corner and cheered me on. To these companions on this journey I would like to express my true gratitude.  I would like to begin, first and foremost, by acknowledging the support of my graduate supervisor, Dr. Heather McKay, who provided a world-class research environment in which I could learn new skills and gain new knowledge. You supported and pushed me throughout this experience and congratulated me on my successes. I have benefited by your passion for research and your guidance, advice, and patience. I am further indebted to my committee members, Drs. Heather Macdonald, and Kishore Mulpuri. Your expertise and ongoing assistance were invaluable to my achievement. It has been an honour to work with you and to learn from your questions, feedback, keen edits, and our discussions. In particular, Dr. Macdonald, I can’t thank you enough for your support and guidance.   Dr. Penny Brasher, I have learned so much from your statistical expertise and guidance. Your passion for and knowledge of statistics amazes me and I am sincerely thankful. Dr. Lindsay Nettlefold, thank you for always lending an ear and continuing to support me throughout this process – you have no idea how much I appreciate it. As well, to the many other researchers and colleagues at the Centre for Hip Health and Mobility (CHHM), many of whom have graduated and/or moved onto new things; I have truly appreciated your encouragement, advice, and friendship. I would like to add a special thank you to Drs. Emily McWalter, Danmei Liu, Yasmin Ahamed, Vina Tan, and Leigh Gabel. As well, my appreciation to Christa Hoy, Douglas Race, and many others at CHHM for your continued optimism. I wish you all success in your future endeavors, academically and personally.    To Drs. Adam Baxter-Jones, Don Bailey, and Stefan Jackowski, and other colleagues and friends from the University of Saskatchewan, I wish to express my sincere appreciation for your mentorship and xxxi  continued support and assistance. You have provided invaluable feedback and guidance and I am grateful. I regard you with the greatest respect and look forward to collaborating in the future.   Of course, a very special thanks to the many participants, parents, teachers, and schools of HBSIII; this dissertation would not be possible without your commitment. I want to acknowledge the countless hours of work by the dedicated HBSIII researchers, students, staff, and volunteers in data collection and management over the entire study duration. I would like to acknowledge that this project was conducted on the unceded territory of the Coast Salish Peoples, including the territories of the xʷməθkwəy̓əm (Musqueam), Skwxwú7mesh (Squamish), and Səl̓ílwətaʔ/Selilwitulh (Tsleil-Waututh) Nations.  Thank you also to those from the HGS, PBMAS, as well as the Saskatchewan Growth and Development Study (SGDS) and Leuven Longitudinal Twin Study (LLTS); I gratefully acknowledge Dr. Adam Baxter-Jones from the PBMAS and SGDS, Dr. Noël Cameron from the HGS, and Dr. Martine Thomis from the LLTS for the contribution of their data for my dissertation.   Teaching and educational development during my dissertation has allowed me to continue to nurture my passion for learning and scholarship in higher education. It has been a privilege to work in partnership with those who are so personally committed to student-centred learning. I am so pleased to have worked with many faculty and staff from a number of institutions across the lower mainland. Importantly, I want to thank all of my students; it has been an honour and I look forward to hearing more about your future successes.  Finally, to my family, friends, and loved ones who were there for me throughout my studies, I owe you the greatest amount of gratitude. To my loving husband, Mike, I would not have made it through this without your ongoing encouragement and patience. To my sons, Ryder and Brody, thank you for giving me a sense of balance and for making me laugh and smile. To my mom, Christina Dusome, and my dad, Alan Braid, thanks for only being a phone call or five hour flight away. You have always supported my academic and personal goals. Thank you for your ongoing and endless support.  xxxii  Dedication  To my little sister, Becca. You taught me to be patient and to persevere. Thank you.  1  Chapter 1: Introduction  It is clear that healthy children pass through the same stages of growth, but at different times and with different tempos [1]. This variability can result in large maturational differences between youth of the same chronological age; these differences are particularly apparent during the adolescent growth spurt [2]. Thus, any investigation of children and adolescents must consider and control for the potential differences associated with maturity, particularly during the pubertal years. Failure to control for maturational differences within and between sexes may result in inaccurate conclusions. That is, the observation may simply be an artifact of maturational differences rather than a function of the independent research variable [3]. To that end, controlling for maturation is necessary to explain independent associations that may otherwise be confounded by biological maturity.  Alignment of boys and girls by biological age accounts for the variation in the timing and tempo of growth in children of the same chronological age. To that end, there are several methods used to assess a child’s maturation status (skeletal [4-7] and dental age [8], secondary sexual characteristics [1, 9-11], percentage of adult height [12-18]) and rate (menarcheal status [19, 20], parameters of the growth curve [19, 20]). However, most elicit concerns about their invasiveness, intrusiveness, and/or are logistically challenging to conduct [19-21]. The most commonly used indicator of maturational timing in longitudinal studies is age at peak height velocity (APHV) [19]. The calculation of APHV requires an individual’s serial longitudinal data spanning the period from late-childhood through adolescence. APHV can be calculated using either parametric methods such as the Preece-Baines model 1 (PB1) [22] or non-parametric methods such as smoothing or interpolating splines [23, 24] in order to visualize and assess parameters on the growth curve [25]. Historically, APHV was estimated with frequent measures using rigid protocols and statistical techniques [2]. APHV has the unique advantage of allowing between and within sex-comparisons when maturity offset (MO) is calculated (where APHV is equivalent to biological zero). However, many investigators simply don’t have the luxury of serial measures and thus, must rely on cross-sectional maturity assessments. 2   Several non-intrusive maturity assessment methods have been proposed. These include percentage of final adult height [12-14] (or predicted adult height [15-18]) and prediction of MO from anthropometry, i.e., Mirwald1 et al. prediction equations [26]. Mirwald equations are one of the most highly utilized tools. However, in recent years the prediction accuracy of the Mirwald equations have been under scrutiny. In three papers [27-29], Professor Malina highlights a systematic problem with the equations in that they are dependent on the age of the child at the time of prediction. Further, predictions made in children who are considered early- or late-maturers yield unfavorable results [27-29]. Results presented in Malina’s papers shows similar prediction errors as noted in the original publication. That is, equations perform best around the time of actual APHV and error increases as predictions are made further from APHV. Despite these critiques, Malina and colleagues [27-29] did not evaluate the development of the Mirwald equations to determine sources of the error. It was however not clear whether Mirwald et al. controlled for the within-subject clustering when developing their equations [26]. If they did not, the standard errors of the estimate (SEE) they presented would be too optimistic. Additionally, given that Mirwald et al. [26] considered each observation as independent (when they were not), sample size was inflated. Thus, true sample size (number of subjects versus number of measures) may have been inadequate to accommodate the number of predictors in the regression models. This may have resulted in overfitting external samples. It seems important to assess these factors and to redevelop the equations as needed, so as to produce better fits in external samples (rather than discounting equations entirely).  One factor that largely influences maturation is a child’s ethnicity [30]. Populations differ in a variety of genotypic and phenotypic characteristics. Groups adopt culturally different behaviour patterns and along with other circumstances, these may influence patterns of growth and maturation [30]. As there are known maturational differences between ethnicities, prediction error in equations developed in white children                                                       1The maturity prediction equations for boys and girls generated by Mirwald et al. will be referred to as ‘Mirwald’ equations for simplicity throughout my dissertation. 3  should be assessed as to their appropriateness for use in non-white children. I contend that if maturational differences exist between ethnicities, ethnic-specific equations would provide more accurate maturity predictions.   Finally, linear growth is irrefutably linked to skeletal development. At the same chronological age during adolescence, late-maturing children have less bone mineral content (BMC) and bone mineral density (aBMD) compared with average- and early-maturing youth [31]. Some content these deficits continued into late-adolescence and young adulthood [32]. In comparison, others suggest late-maturers experienced ‘catch-up’ during post-puberty and more closely resemble their average- and early-maturing peers [33]. I propose to provide further insight into the maturity-bone relationship using data from our HBSIII. As HBSIII is a multi-ethnic cohort I can also explore whether maturity-bone response is ethnic-specific.   Thus, the aims of my dissertation are four-fold. My first aim (Chapter 5) is to evaluate the validity of the Mirwald equations. I will review development of the equations, and if necessary, redevelop the equations to improve applicability. My second aim (Chapter 6) is to assess differences in growth and maturation in a cohort of Asian and white children living in the same geographic area in Metro Vancouver. My third aim (Chapter 6) is to assess the accuracy of prediction equations developed using data acquired from white children when applied to an Asian-Canadian cohort (all collected as part of the HBSIII study), I propose, if necessary, to develop new equations that improve applicability to Asian boys and girls. Lastly, my fourth aim (Chapter 7) is to assess the relationship between maturational timing and post-pubertal bone mass, density, structure, and strength.  4  Chapter 2: Literature Review  In this section I summarize and discuss relevant literature related to growth and maturation in children and adolescents, and factors that influence these phenomena. In the following sections I: 1) define and describe the phases of growth and maturation; 2) identify key factors that influence growth and maturation; 3) describe methods used to assess growth and maturation with a focus on APHV, and 4) describe the role of maturation in bone accrual during growth.  2.1 Growth and Maturation Growth, maturation, and development are often used synonymously in the literature. However, while these processes interact and occur simultaneously, they are distinct. Growth dominates during the first two decades of human life [19], and is defined as an increase in body size, resulting from cellular hypertrophy and hyperplasia and interstitial accretion [34], all of which vary with age and by tissue type [19]. Maturation is the progressive achievement of an adult, mature state and includes various developmental changes, including morphologically [25]. Maturation involves the process of maturing, whereas maturity is a state. We can consider timing and tempo in maturation: timing being when specific events or milestones happen and tempo being how fast an individual passes through the various events or milestones [19]. Development is a broader term that has both behavioural and biological contexts. Behaviourally, children develop cognitive and social competences such as learning in different environments. Biological development refers to the process of differentiation and specialization of cells or tissues that occurs during growth [19]. In my dissertation, I focus primarily on growth and maturation.   Timing of maturation differs between children of the same chronological age [3]. Failure to control for maturational differences may result in inaccurate conclusions, because the observations may simply be artifacts of differences in maturity and/or growth, and not the result of an intervention or the relationship with other independent variables [3]. For example, Figure 2.1 shows three boys who are the same chronological age but differ quite dramatically in biological age. While chronological age is simply 5  calculated by subtracting the current date from a child’s date of birth, biological age is more complex as biological processes (e.g., skeletal maturity) may occur independent of chronological age. When a child’s chronological and biological age differ the child is considered to be an early- or late-maturer. For example, in Figure 2.1, the child on the left could be considered a late-maturer, whereas the boy on the right as an early-maturer. Maturity classifications distinguish children within or outside normal growth limits. I explain how these classifications are utilized in section 2.1.2.4.   Figure 2.1 Illustration of substantial differences in biological age when comparing three boys at the same chronological age (14 years). The boy on the left may be considered late-maturing, the boys in the middle may be considered average-maturing, and the boy on the right may be considered early-maturing. Reproduced from Tanner, J.M. [35], with permission from Castlemead Publications.   2.1.1 Studies of Growth and Maturation Prospective observational studies that span from childhood to adulthood and include measurements of height and weight at regular intervals (e.g., semi-annual) are the gold standard for growth and maturation research [36]. However, such designs are not always feasible due to logistical and financial constraints. Alternatively, researchers may opt for a mixed-longitudinal design whereby data for participants who enter a study at different time points are combined to represent one longer period. For example, groups of children recruited at birth, and at age 4, 8, 12, and 16 years are each followed for 4 years. At the conclusion of the  4-year mixed longitudinal study, the researchers can assess the cohort as a whole from birth to 20 years of 6  age [19]. Results of mixed-longitudinal studies are comparable to those of pure longitudinal designs (Figure 2.2) and thus provide valuable information on the timing and tempo of growth and maturational events. Longitudinal designs are preferable for establishing time-order effects and causality [36].    Figure 2.2 Illustration of height velocity (cm/year) calculated from cross-sectional (X; dashed line), longitudinal (L; dotted line), and mixed longitudinal (M; solid line) height measurements, by age and sex in Southern Chinese children from Hong Kong. Reproduced from Low [36], with permission from Schweizerbart Science Publishers.    In the first half of the 20th century, a number of longitudinal studies of growth and maturation were initiated in the United States (US) and Europe. I have highlighted some of the studies which I later describe in my dissertation in Table 2.1. These studies quantified the rate and magnitude of healthy childhood growth. Impressively, the Fels Longitudinal Study, which began in 1929 to study the growth and maturation of healthy children from birth (pregnant women were recruited) in Ohio, is still running today [37]. The notable British HGS took place from 1948 to 1971. Around the same time, the International Children’s Center in France coordinated studies in Paris, Zurich, Stockholm, London, and Brussels [38, 39]. More recent studies of growth and maturation (i.e., from 1960 to present) focused on adolescent growth, many of 7  which used mixed longitudinal designs and employed newer technology such as DXA to further and more specifically quantify growth and maturity of different tissues (e.g., BMC and aBMD). 8  Table 2.1 Summary of historical longitudinal studies of children’s growth and maturation. Adapted from Himes [40], with permission from Food and Nutrition Bulletin, Sage Publishing.   Study Namea Dates Design Age (years) Sample Size Area and Country  Selected Reference Iowa  1920-1934 Mixed 0-18 2484 Iowa City, USA Meredith [41] Harvard 1922-1934 Longitudinal 6-17 1553 Boston, USA Dearborn et al. [42] Fels 1929-present Longitudinal 0-21 1036 Ohio, USA Roche [43] Harpenden 1948-1971 Mixed 3-18 420 London, UK Tanner [1] Paris 1953-1975 Longitudinal 0-21 542 Paris, France Sempe et al. [44] Zurich 1954-1976 Longitudinal 0-20 413 Zurich, Switzerland Prader et al. [45] West Bengal 1952-1966 Mixed 0-21 562 West Bengal, India Das et al. [46] Stockholm 1955-1978 Longitudinal 0-17 212 Stockholm, Sweden Karlberg & Taranger [47] Wroclaw 1961-1972 Longitudinal 8-18 470 Wroclaw, Poland Bielicki & Waliszko [48]  Saskatchewan 1964-1973 Longitudinal /Mixedb 7-17 305 Saskatchewan, Canada Mirwald [49] Leuven 1968-1974 Longitudinal 12-18 588 Leuven, Belgium Beunen et al [50] Nymegen 1971-1976 Mixed 4-14 467 Nymegen, Holland Prahl-Andersen et al. [51] Leeds 1972-1985 Mixed 9-18 396 Leeds, UK Buckler [52] a Formal name of study shortened; b Longitudinal design for boys and mixed longitudinal design for girls9  2.1.2 From Fetus to Adult State In this section, I describe growth curves of body systems and the stages of growth and maturation.   2.1.2.1 Scammon’s Growth Curves The patterns of growth within body tissues and systems include: a) neurological (e.g., brain), b) genital (e.g., reproductive organs), c) lymphoid (e.g., lymph glands, tonsils, appendix), and d) general (e.g., height, weight, and some tissue/organ growth such as the heart) (Figure 2.3) [53]. Across all four curves, there is individual variability between and within sexes [20].    Figure 2.3 Drawing of Scammon’s growth curve of different parts and tissues of the body. The data shown are as a percentage of growth for each system between birth and 20 years of age. Reproduced from Scammon [53], with permission from the American Journal of Physical Anthropology, John Wiley & Sons, Inc.   The neurological curve (long dash line, Figure 2.3) includes growth of the brain, nervous system, and their associated structures. This curve accelerates early and plateaus by adolescence; the brain reaches adult size between 8 to 10 years of age. The genital curve (dot-dash line, Figure 2.3) includes development of the 10  primary sex organs (e.g., prostate, seminal vesicles in boys; uterus, vagina, fallopian tubes in girls) and secondary sex characteristics (e.g., facial hair in boys, breasts in girls, and axillary and pubic hair in both sexes). This curve illustrates some growth during infancy and rapid development during adolescence due to a steady surge in sex hormones. The lymphoid curve (short dash line, Figure 2.3) describes growth of the lymph and thymus glands, tonsils, appendix and other lymph tissues. Maximal size is attained during adolescence; lymph tissue is greatest during adolescence and is reduced to half by adulthood. That is, lymph tissue (e.g., thymus) shrinks during adulthood given its role in immunity is reduced after puberty [19].  Finally, the general curve (solid line, Figure 2.3) describes the body as a whole; height, weight, and most external dimensions, as well as the pattern of most systems within the body: muscular, respiratory, cardiovascular, digestive, and urinary systems. The ‘S’ shape of the general curve reflects rapid growth during infancy and early childhood, steady growth during mid-childhood, rapid growth during early adolescence, and levelling off in late adolescence. I focus on this general curve in the following sections.  2.1.2.2 Prenatal Growth to Birth  In the next sections, I describe the stages of growth from prenatal life to adulthood (Table 2.2). I first describe influences of the intrauterine environment of growth. I then describe infancy, childhood, and adolescent periods of growth. Finally, I describe the growth of body segments and the cessation of growth at adulthood.               11  Table 2.2 Description of the stages and growth events in the human lifecycle. Adapted from Cameron & Bogin [20], with permission from Elsevier Publishing.   Stage  Growth events/duration  (average and approximate) Fertilization     Prenatal life    First trimester Fertilization to third month: embryogenesis  Second trimester Fourth to sixth month: rapid growth in length  Third trimester Seventh lunar month to birth: rapid growth in weight and organ maturation Birth     Postnatal life    Neonatal   Infancy  Birth to 28 days: extra-uterine adaptations, most rapid rate of postnatal growth and maturation  Second month to end of lactation or usually 36 months: rapid growth velocity, but with steep deceleration in growth rate  Childhood   Juvenile Ages three to seven years: moderate growth rate, possible mid-childhood growth spurt  Ages 7 to 10 for girls, 7 to 12 for boys: slower growth rate  Puberty   Adolescence Occurs at end of juvenile stage, short in duration: dramatic increase in secretion of sex hormones  Occurs after the onset of puberty, lasting 5 to 10 years: growth spurt in height and weight and transition to fully mature state   Adulthood    Young adulthood  to older adulthood Occurs at end of adolescence: plateau in height, loss of ability to growth by hyperplasia but hypertrophic growth still possible  Death     The intrauterine environment often influences postnatal growth and maturation [54]; therefore, I briefly discuss prenatal growth here. Prenatal growth occurs from fertilization to birth. A full-term human pregnancy is defined as 37 to 41 weeks gestation for singleton births [20]. Carrying to term is associated with fewer complications and decreased risk of infant death. However, more recent research defines term pregnancy as 39 to 41 weeks gestation, as infants born at 37 and 38 weeks demonstrate higher mortality rates compared with those born at 39 to 41 weeks gestation [55].  12  Intrauterine prenatal growth is non-linear; growth velocity increases during the first part of pregnancy and decreases in the latter. Throughout, fetal body shape and proportion change drastically. Fetal growth is measured from crown-to-rump and crown-to-heel, before and after 20 weeks, respectively. Prenatal linear growth accelerates rapidly until 14 to 16 weeks of gestation [56], reaches a peak at 20 weeks [35], and diminishes thereafter until term. Sex differences in physical growth are apparent during prenatal growth; at 20 weeks boys are ~3 weeks less mature (skeletally) compared with girls [57]. From weeks 20 to 40, fetal length approximately doubles from 26 cm to 52 cm, whereas weight in this same period increases more than tenfold, from approximately 300 g to 3000 g [58]. Peak growth velocity is faster and occurs earlier for head diameter and circumference (18 weeks) compared with femur length (20 weeks) and abdomen circumference (22 weeks) [59].  A number of factors may lead to growth deficiencies in utero. For example, intrauterine growth restriction (IUGR) is a disorder of placental insufficiency; for infants born in the <10th percentile for weight, there is a significant risk of morbidity [60]. IUGR and associated lower birth weight may be caused by poor maternal nutrition and/or weight gain, use of nicotine or alcohol during pregnancy, teenaged or advanced maternal age pregnancy, placental or uterine malformations, maternal hypoxia or anemia, or genetic or fetal abnormalities, amongst other factors [61]. Other factors that lead to growth deficiencies may include congenital malformations, skeletal dysplasia, malnutrition, chronic diseases, and endocrine disorders. Reference curves for height and weight (and body mass index; BMI) allow researchers and clinicians to better define ‘normal’ growth, as well as diagnose and treat potential growth problems. By definition, normal physiological growth encompasses the 95% confidence intervals (CI) for a specific population [62]; falling below the 3rd or above the 97th percentile (approximately ±2.0 standard deviations (SD)) for sex- and age-specific height, for example, may warrant investigation into potential growth-related issues [20].   13  2.1.2.3 Infancy and Childhood Growth At birth, infant length and weight are 49.5 cm and 3300 grams (g), on average, respectively [63]. Boys tend to weigh slightly more (0.3kg) and be slightly longer (0.6cm) on average, compared with girls [63]. Greater birth weight in boys is typically attributed to more lean tissue mass compared with girls [64]. After birth, most infants experience growth changes in their first two years to reach their genetically determined growth potential [65, 66]. Large variation in growth patterns between infants may be present during the first 18 to 36 months [20]. However, during the third year of life, growth occurs in a more consistent pattern between individuals. Boys tend to reach 50% of adult height at age 2 to 2.5 years, whereas girls tend to reach 50% of adult height by age 1.5 to 2 years [57, 67]. Seasonal variation in growth is evident (moreso after age two years); where growth is slower in the winter compared with summer [68]. The general postnatal growth events are similar between individuals, though there are differences in growth rate (i.e., tempo) as well as overall size attained [19].  Further, proportional differences are apparent at birth. Head size is much nearer to its adult size than is trunk or leg length and represents about one-fourth of the child’s total length, compared with one-eighth of height in an adult (Figure 2.4). Within the limbs there is a distal-proximal gradient; hands and feet mature ahead of lower legs and forearms, which are more advanced compared with thighs or arms [69]. Proportion of trunk length decreases after birth through to adulthood, where ratio of crown-rump to leg length is 1:7 at birth, but is more equal at puberty and into adulthood [67, 70]. Further, the upper limbs are longer than the lower limbs during infancy and childhood. During late childhood and into adolescence the limbs are more representative of a mature adult state with lower limbs longer compared with upper limbs. Thus, childhood growth is dominated by leg length growth compared with trunk growth [71]. I summarize the growth of body segments in Section 2.1.2.6.   14   Figure 2.4 Illustration of the average changes in body proportion from infancy to adulthood. Reproduced from Huelke [72], with permission from the Association for the Advancement of Automotive Medicine.     In two-thirds of healthy children, a small (5 to 7 cm/year) growth spurt during the mid-childhood (MC) years (six to eight years, on average) is observed (Figure 2.5) [20]. A subsequent decline in growth immediately before puberty may indicate this MC growth spurt [71, 73]. In a cohort of 10,253 British children, the MC growth spurt was more apparent in boys as compared with girls. This sex difference was explained by two factors: 1) not all children in the cohort had a MC growth spurt, and 2) the growth modeling technique (i.e., graphic smoothing) used may have masked the growth spurt in some children [74]. During the MC growth spurt, there were also spurts in weight and body girths that occurred at approximately age six to seven years in girls, and a year later in boys [74]. 15   Figure 2.5 A mid-childhood spurt is observed in two-thirds of healthy children between childhood (C) and juvenile (J) years. This illustration also depicts height distance and velocity in infancy (I), adolescence (A), and adulthood (mature state, M) in boys (solid line) and girls (dotted line). Reproduced from Cameron & Bogin [20], with permission from Elsevier Publishing.   2.1.2.4 Adolescent Growth and Puberty Adolescence represents the transitional years between childhood and adulthood; an unpredictable period where physiology changes dramatically. This time is marked by the appearance of secondary sexual characteristics, maturation of the reproductive system, menarche, and the adolescent growth spurt [19, 35, 39]. The terms adolescence and puberty are often used interchangeably; however, there are important differences that distinguish the two terms. Specifically, puberty occurs at the beginning of adolescence and is defined as the dramatic increase in sex hormones [20] that initiates sexual maturation and attainment of reproductive capacity [1]. Puberty is the time of greatest sex differentiation since the early intrauterine months [35]. Some extend puberty from the first appearance of secondary sexual characteristics until the achievement of adult sexual maturity [75], whereas others consider only the short period where sexual maturation is initiated [20].  16  A sudden increase in growth (height) velocity (i.e., take-off) occurs after the onset of puberty and signals the largest change in linear height during adolescence [19]. Adolescent growth is more variable than childhood growth, and unlike the MC growth spurt, the adolescent growth spurt and peak adolescent growth (APHV) is observed in all children. However, timing and magnitude of the growth spurt varies between individuals. The adolescent growth spurt is directly related to physical and hormonal changes that accompany sexual development (i.e., puberty), and this growth spurt is characterized by: 1) minimal height velocity just before the spurt (i.e., pre-pubertal growth lag); 2) maximal growth (i.e., magnitude at peak height velocity; PHV); and 3) decreased height velocity coinciding with fusion of long bone epiphyses [76], as I discuss in Section 2.2.2.6. On average, boys and girls experience maximal growth (i.e., APHV) at approximately 14 and 12 years of age, respectively [20]. Figure 2.6 illustrates the standard distance and velocity curves for height during the growing years. A distance curve plots height at each measurement to allow visualization of height attained at specific ages relative to adult height. In contrast, a velocity curve plots the changing rates of height gain during growth. I summarize key elements of the pubertal growth velocity curve in Section 2.1.4.3.3.   Figure 2.6 Illustration of typical height distance and velocity curves in boys and girls. The left graph shows distance curves and the right graph shows velocity curves in boys (solid) and girls (dotted) from infancy to adulthood in healthy children. Adapted from Malina, Bouchard, & Bar-Or [19], with permission from Human Kinetics Publishing and Archives of Disease in Childhood, BMJ Publishing Group Ltd.  17   Sex differences in adolescent growth contribute significantly to sex differences in adult body size. Once growth has ceased, men have greater linear body dimensions compared with women, particularly for height, trunk length (assessed via sitting height), and leg length [1]. The total difference in adult height due to sex is expressed as the sum of three additive components: D=DP+DT+DA, where D is total difference, DP is the difference in height at take-off in girls, DT is the amount of linear growth achieved by boys between take-off in girls and take-off in boys, and DA is the difference in adolescent height gained between boys and girls [20]. Hauspie et al. summarized sex-differences in height in British children from birth to adulthood (Table 2.3); later onset of pubertal growth in boys (DT=7.9 cm) was the largest contributor to the sex-difference in adult height.    Table 2.3 Demonstration of the decomposition of the sexual dimorphism in adult size of height, trunk height, shoulder width, and hip width in a British sample of boys and girls. Adapted from Hauspie et al. [77], with permission from Taylor and Francis Publishing.    Height (cm) Trunk height (cm) Shoulder width (cm) Hip width (cm) Total difference (D) 12.0 4.5 3.7 -0.6 Difference at take-off in girls (DP) 2.1 0.3 0.3 -0.1 Growth in boys between take-off in girls and take-off in boys (DT) 7.2 3.5 1.7 1.5 Difference in adolescent gain (DA) 2.0 0.7 1.7 -2.0   Weight change from birth to young adulthood follows a similar pattern to height: rapid gain during infancy and childhood, steady gain during middle childhood, rapid gain during adolescence, and slow gain thereafter (Figure 2.7). Peak weight gain follows peak height gain. Unlike height, weight usually continues to increase into adult life, and is much more variable between individuals over the lifespan [19].   18   Figure 2.7 Illustration of distance (left) and velocity (right) curves for weight in boys (solid) and girls (dotted) from infancy to adulthood in healthy children. Reproduced from Malina, Bouchard, & Bar-Or [19], with permission from Human Kinetics Publishing and Archives of Disease in Childhood, BMJ Publishing Group Ltd.     Magnitude of pubertal growth, sequence of maturational events, and duration of pubertal development varies considerably between and within sexes. Figure 2.8 illustrates the typical sequence of pubertal events in boys and girls; age at onset of puberty varies by as much as four to five years in healthy boys or girls [1].       Figure 2.8 Illustration of the sequence of pubertal events in boys (left) and girls (right), where arrow and number represent the mean age for each pubertal event. Reproduced from Haqq, Boston, & LaFranchi [78], with permission from Lippincott Williams & Wilkins.    19  Researchers and clinicians may choose to compare selected factors in children based on whether children are early- (chronological age < biological age), average (chronological age = biological age), or late-maturing (chronological age > biological age). This may be of particular interest when children show signs of clinically precocious or late puberty (i.e., <3rd or >97th percentile for height velocity).  Classification of naturally-occurring late or early maturity is typically done in one of three ways: a) categorizing early-maturers as those who precede average APHV by more than one year and late-maturers as those who follow average APHV by more than one year [33, 79], b) dividing a cohort into tertiles for APHV [80], or c) creating percentile groups (e.g., early-maturers are those in the 1st to 20th percentile to reach APHV, average-maturers are those from the 21st to 79th percentile to reach APHV, and late-maturers are those in the 80th to 99th percentile to reach APHV) [31]. Generally, late-maturers have a prolonged pre-spurt growth period but spend less time in pubertal growth, whereas early-maturers spend more time in pubertal growth and have a prolonged post-pubertal growth period [81]. A longitudinal study of 163 adolescent Spanish girls (aged 9 to 17 years) showed the average time between age of pubertal onset (defined as breast bud development) and advent of menarche was 1.96 years. However, in early-maturing girls the time lapse between pubertal onset and age at menarche was 2.77 years. In late-maturing girls the time lapse was only 0.65 years. Similarly, in Greek girls aged 6.8 to 16.1 years duration of puberty (defined as time of pubertal onset to APHV) was 0.7 years longer, on average, in early-maturing girls compared with late-maturing girls [82].   2.1.2.5 Cessation of Linear Growth and Attainment of Adult Size Completion of the adolescent growth spurt, attainment of adult height, and full sexual (i.e., reproductive) maturity indicate adulthood. Cessation of linear growth can be visualized as a plateau on the height distance curve and is commonly defined as: a) an annual change in height of <1.0 cm/year [83], or b) four successive six-month height measurement increments of <0.5 cm [84]. At APHV, boys and girls are at approximately 90 to 92% of their adult height. Thus, most linear growth occurs before and at APHV 20  [85]. Men may continue to grow in small increments into their early- or mid-twenties, whereas women stop growing in their late teenage years. For example, in the Fels Longitudinal Study men reached adult height approximately eight years after APHV compared with six years after APHV for women [84]. The period between APHV and early-adulthood is more recently been referred to as ‘emerging adulthood’ (i.e., years between adolescence and adulthood) [86].    2.1.2.6 Growth of Body Segments Growth occurs distally to proximally [87]. For example, hands and feet experience accelerated growth first, followed by calves and forearms, hips and chest, and lastly shoulders. Once the adolescent spurt ends, hands and feet are smaller in proportion to arms, legs, and height. Most body segments follow a growth pattern similar to that of height (i.e., acceleration, peak growth, and deceleration). However, the timing and tempo of segmental growth spurts vary widely. For example, from childhood to adolescence the lower extremities (i.e., legs) grow faster than the upper body (i.e., trunk), which results in sitting height contributing less to height as age progresses [1]. In adolescence, leg growth precedes trunk growth [19]. Thus, for a period during early adolescence, an individual may have relatively long legs, but the appearance of long-leggedness disappears with the later increase in trunk length. Sex differences in leg length and trunk length are small during childhood. For a short time during early adolescence, boys, on average, have slightly shorter leg lengths compared with girls. Boys’ leg length begins to exceeds girls’ leg length by about age 12 years, but boys do not catch-up in trunk length until after age 14 years [19]. The longer period of pre-adolescent growth in boys is largely responsible for the fact that men’s legs are longer compared with women’s in relation to trunk length [19]. I discuss the relationship between height, trunk length, and leg length during growth in more detail in Section 2.1.4.4.  21  2.1.3 Assessing Growth  Measurement of human growth is an essential component of both clinical practice and research studies involving children and youth. Anthropometry (from the Greek Anthropos “man” and metron “to measure”) is a set of standardized techniques for the systematic measurement of the body and its parts for the purpose of quantifying dimensions [88]. In this section, I briefly describe anthropometry related to my dissertation.   2.1.3.1 Height and its Components In children over two years of age, height (or stature; cm) is a linear measure of the distance from the bottom of the bare foot to the vertex of the skull and is assessed using a stadiometer in children who are able to stand upright for measurement (heels, buttocks, and upper back in contact with the wall or stadiometer post) (Figure 2.9). The vertex is defined as the highest point on the skull when the head is held in the Frankfort plane. This position is when there is an (imaginary) horizontal line present from the lower border of the orbit to the upper margin of the ear canal [89]. To attain this position, a research assistant places his or her hands along the participant’s jaw and gently applies upward traction through the mastoid process. Whereas in children under two years, length is obtained in the recumbent position using an infant-specific stadiometer [89]. The components of height include standing height, trunk height, and leg length. Sitting height gives an indication of trunk length and is measured from the sitting surface to the skull vertex (Figure 2.9). Sitting height is measured in a similarly to height, where the child is sitting on a measuring box or platform of a known height. The participant sits with his or her hands in their lap, and upper back against the wall or post. The head is in the Frankfort plane as described above. After measurement, the box or platform height is subtracted from the measured height, resulting in a measure of sitting height (i.e., trunk length) [89]. Finally, leg length is most commonly determined by subtracting sitting height from standing height, as leg length itself is difficult to assess given the challenge in locating the trochanteric landmark at the proximal femur [89]. 22  Diurnal variation is commonly noted in height and its components, where height is greater in the morning than the evening [19]. We use stretch stature methods in the HBSIII (where the measurer applies gentle upward traction with the participant in Frankfort plane and a second measurer lowers the measurement block to the participants head) to reduce effects of diurnal variation [89] (see Section 4.2.2.1).   Figure 2.9 Demonstration of the measurement of height (left) from bottom of bare feet to vertex of skull in Frankfort plane with participants heels, buttocks, and back against the stadiometer, and measurement of and sitting height (right) using platform of known height with participant’s hands in lap. Adapted from Frisancho [89], with permission from the University of Michigan Press.     2.1.3.2 Body Weight and Body Mass Index The terms body weight and body mass are often used synonymously; however, they are distinct concepts. Body mass refers the body’s matter, whereas weight refers to the effect that gravity has on body mass [89]. On earth, body mass and weight would be the same. Body weight is measured in kilograms (kg) and is a measure of the body’s mass. I will use the term body weight in my dissertation. Body weight is a composite of independently varying tissues. The relative proportions and distribution of fat and fat-free components depend on a number of factors include: age, sex, and ethnicity, amongst others [89]. As with height, diurnal variation is also commonly noted in weight, where weight is greater in the evening than the morning [19].  23  Body mass index (BMI, kg/m2), also known as the Quetelet index (developed by Belgian statistician Adolphe Quetelet), is a measure of weight per unit of body height. It is expressed as kilograms of weight per square meter of height. BMI allows for a comparison of weight while accounting for height. In general, BMI is correlated to body fatness in adults, though not in those that have a greater proportion of lean mass (e.g., athletes, some ethnic groups) [89]. For children, BMI varies with age, and therefore BMI values are commonly compared with age- and sex-specific reference values. A BMI Z-score or percentile represents a measure of weight, adjusted for height, sex, and age, relative to a reference population [90].   2.1.3.3 Other Common Anthropometry in Pediatric Studies Other anthropometry include skeletal robustness assessed as head circumference and/or breadths (e.g., elbow, wrist, bi-iliac), estimates of muscularity obtained from limb circumference (e.g., thigh, calf, upper arm), and estimates of subcutaneous adipose tissue obtained from skinfold thickness (e.g., triceps, subscapular), as well as hip and waist girth, and hip to waist ratio. [89].   2.1.4 Assessing Maturity Techniques used to assess maturity (i.e., maturity indicators) should meet the following criteria: 1) able to reflect changes in a biological system; 2) be (to some degree) independent of growth; 3) be applicable from birth to adulthood; 4) reach the same adult state in all individuals; and 5) show a continuous increase over the entire maturational process [91]. Methods currently utilized to assess maturity use four biological systems: skeletal, dental, sexual, and somatic/morphologic [21]. I describe these methods and their interrelationships in the following sections.  2.1.4.1 Skeletal Maturity Skeletal maturity is one of the best methods to assess biological maturation in children and adolescents as it meets the aforementioned criteria and is related to endocrine function (e.g., IGF-1) [19, 92]. Beginning 24  in the prenatal period, the skeleton undergoes a series of changes in bone shape and structure that eventually lead to a mature adult skeleton. Maturation status varies between individuals and is reflected by a) the change in bone shape and structure, and b) the degree to which the epiphysis has fused. Skeletal age (also known as bone age or skeletal maturity) is most commonly assessed using radiographs of the hand and wrist (Figure 2.10) and is quantified using one of three popular scoring methods: Greulich-Pyle [4], Tanner-Whitehouse [5], and Fels [6, 7]. Skeletal aging has limited widespread use due to radiation concerns. I describe the scoring methods in detail below.   Figure 2.10 Illustration of a hand-wrist radiograph of an eight year old boy. Specific bones are shown in the schematic, excluding the pisiform and adductor sesamoid. Horizontal lines indicate the locations of the epiphysis of each long bone. Reproduced from Malina, Bouchard, & Bar-Or [19], with permission from Human Kinetics Publishing.   2.1.4.1.1 Gruelich-Pyle  The Greulich and Pyle published Atlas of Skeletal Development (the Radiographic Atlas of Skeletal Development of the Hand and Wrist, 1950), provides images of standard radiographic plates to which hand-wrist radiograph can be compared. The standard plates correspond to successive levels of skeletal maturity at specific chronological ages. A child’s skeletal maturity is determined by visually matching the hand-wrist radiograph to a sex-specific standard hand-wrist radiograph plate [19]. This method is based only on 25  subjective visual matching; no scores are assigned or calculated. The Gruelich-Pyle Atlas is based on approximately 1000 white children of a higher socioeconomic status (SES) in Cleveland, Ohio, from 1931 to 1942. Children were examined every 3 months until 12 months of age, every 6 months until 5 years, and annually thereafter. Each child contributed 2 to 21 left-hand radiographs made at successive examinations. The Atlas’ accuracy has been questioned. For example, Varkkola et al. noted a 9.7 month difference, on average, between skeletal age by Greulich-Pyle and chronological age [93]. Moreover, Loder et al. recommended that the Greulich-Pyle method not be used in black girls, as it systematically overestimated skeletal age [94]. Ontell et al. [95] and Zhang et al. [96] noted that maturity was overestimated when the Gruelich-Pyle method was used to assess Asian and Hispanic children, particularly during puberty. Wenzel et al. [97] and Koc et al. [98] also reported maturity was overestimated using the Greulich-Pyle method in Austrian and Turkish children, respectively, particularly at and after puberty. The Greulich-Pyle technique is influenced by rater skill and accuracy; a rigorous training protocol can minimize inter- and inter-rater differences [99].  2.1.4.1.2 Tanner-Whitehouse  The Tanner-Whitehouse (TW) method is a bone-specific approach [1]. The TW method was developed using a cross-sectional sample of 3000 healthy British children and first introduced in 1962. Since then, two revisions were published (TW2 and TW3); the most recent version included ethnic-specific reference data [19]. Briefly, the TW method involves matching features of 20 individual bones in the wrist and hand (7 carpal bones and 13 long bones – radius, ulna, and metacarpals and phalanges of the first, third, and fifth fingers) to a series of precise, written standards for various stages of maturation [5]. Bones are individually scored and a summed score is converted to a skeletal age. Reference values are available for British, Belgian, Italian, Spanish, Argentinian, American, and Japanese children. Some modifications based on ethnicity were made using converted scores, but scoring individual bones remained constant [19]. The TW2 and TW3 methods are more objective but considerably more complex than the Gruelich-Pyle method and 26  are still subject to inter-rater variability. Efforts are being made to automate ratings using the TW methods [100].   2.1.4.1.3 Fels  Hand-wrist radiographs of 13,823 US children involved in the Fels Longitudinal Study were used to develop Fels method [101]. The X-rays were acquired at scheduled ages (every 3 months up to 1 year old, every 6 months to 18 years old, and every other year thereafter) between 1932 and 1972 [101]. Maturity indicators for each bone in the hand and wrist were defined and then verified on separate sets of radiographs. The Fels method uses the same 20 bones as TW3, plus the pisiform and adductor sesamoid. Unlike other radiograph-based methods, the Fels method provides an estimate of error [101]. Given its complexity, this method may require additional training for accurate assessment [19].   2.1.4.2 Sexual Maturity Sexual maturation extends from embryonic sexual differentiation through puberty to full sexual maturity and fertility. Current methods used to assess sexual maturity include secondary sexual characteristics, menarcheal or spermarcheal status, and serum or salivary sampling of reproductive hormones. I discuss each of these in the following sections. When the aforementioned methods are not feasible, axillary hair in boys and girls [1], facial hair in boys [102], and voice changes in boys [103, 104] may be used to estimate maturation status. However, these indicators are rarely used, as they are less reliable in distinguishing between maturity stages.   2.1.4.2.1 Secondary Sexual Characteristics Sexual maturity is most commonly assessed as per the method of Tanner [1], which involves either physician- or self-assessment of pubic hair development for boys and girls, genital development for boys, and/or breast development for girls. Standard terminology describes pubic hair development in stages PH1-27  PH5 or PH6, genitalia development in stages G1-G5, and breast development in stages B1-B5. Stage 1 corresponds with pre-puberty, Stages 2 and 3 correspond with early- and peri-puberty, Stage 4 corresponds with puberty and sexual maturity is reached at Stages 5 or 6, also known as post-puberty [1].    Precision of the methods as per Tanner is similar to the precision of skeletal aging. Self-assessment correlates well with physician assessment; Pearson correlation coefficients range from 0.8 to 0.9 for breast in girls and pubic stage in boys and girls, respectively [11]. Although this method provides a non-intrusive practical way to assess maturity in growth studies, there are also limitations. Boys and girls tend to overestimate early stages and underestimate later stages of maturity [105]. Stages as per Tanner’s method may also be overestimated in overweight and obese children, particularly breast development in girls, as the breast bud may appear due to adipose tissue in the chest [106]. Further, pubic hair stages do not correspond to the breast stages in girls or genital stages in boys. For example, girls do not enter B2 and PH2 at the same time, instead breast development occurs first. Similarly, most boys enter genitalia development prior to any pubic hair development [20]. In some cases, researchers and/or clinicians will combine pubic hair and genital development in boys, or breast, pubic hair, and/or menarche data in girls to provide a single sexual maturity score [107].   2.1.4.2.2 Age at Menarche and Spermarche Age of menarche (i.e., first menstrual cycle) in girls is widely used to assess maturity as it is non-invasive, non-intrusive, and doesn’t require specialized equipment [19]. Three methods are used to determine age at menarche: prospective, status quo, and recall. Prospective assessment is most accurate, as girls are followed at regular intervals starting in pre-puberty (ideally every three months) and asked at each study visit if they have started to menstruate. In the status quo method girls are asked their date of birth and if they have experienced their first menstrual period at the time of the study visit. If yes, the date of first menstrual cycle is estimated and recorded. Finally, in the recall method post-menarcheal girls are asked to 28  recall when the date of their first menstrual period [108, 109]. Guiding questions regarding other referenced events may increase recall accuracy; however, this method is still limited by recall bias [110].  Age of spermarche may offer a comparable level of sexual maturity in boys. This method involves detection of spermatozoa in ejaculate or in urine to assess functional state of the gonads; however, age of spermarche has seen limited use to date [111]. Although some believe age of spermarche is an equivalent level of sexual maturation to age at menarche in girls these stages are not equivalent [25]; in boys spermarche occurs at the early to middle stage of puberty, whereas in girls menarche tends to occur towards the later stage of puberty [20]. Importantly, both age at menarche in girls and age at spermarche in boys can only be assessed once an individual has experienced the event.   2.1.4.2.3 Circulating Hormones  Hormone levels are a less common method used to assess maturity. Estradiol, testosterone, LH, FSH, and IGF-1 can be assessed via serum, urine, or salivary samples [39, 112, 113] and can distinguish between pre- and post-pubertal boys and girls [92, 112, 114]. However, accurate assessment of hormone levels is challenging due to diurnal and seasonal variability in hormone levels and variability in dietary intake and activity levels that affect hormones [115]. Further, a single sample does not account for episodic hormone secretion. Although more difficult, 24-hour monitoring (where samples are collected every 20 minutes and capture hormone pulses) provide a more accurate assessment of hormone levels [116]. Lastly, samples must be collected in very controlled environments and specific biochemical assays are used to assess maturity. However, hormone levels can typically only distinguish pre- from post-pubertal children. For example, increasing testosterone levels during adolescence initiates genital development in boys, and increasing estrogen levels in girls during adolescence initiates breast development, though testosterone level in boys, nor estrogen level in girls cannot distinguish a child’s Tanner stage [117-120]. However, LH, FSH, and estrogen levels can distinguish pre- from post-menarcheal girls. Serum levels of IGF-1 correlate with skeletal maturity, but cannot distinguish skeletal age from chronological age [92, 114]. Consequently, 29  circulating hormones from blood, urine, or saliva samples are more common in clinical settings when investigate whether or not a child has experienced puberty [19].  2.1.4.3 Somatic Maturity Somatic maturity refers to the morphological changes in the body over time [121]. Thus, longitudinal data are needed to plot a growth curve on which growth parameters can be determined. APHV, a focus of my dissertation, is the most commonly used indicator of somatic maturity [20]. APHV is ideally determined using serial measures of height; though APHV can also be predicted with one-time anthropometry data. In the following sections I discuss methods used to determine APHV, the relationship between APHV and peak sitting height and trunk length velocities, and development of equations used to predict APHV.  2.1.4.3.1 Age at Peak Height Velocity and Maturity Offset Assessment of APHV uses the somatic milestone of peak linear growth to identify maturational status. As an indicator of biological maturity, APHV identifies the age at which maximal velocity (cm/year) for gain in height occurs during adolescence. In boys and girls, take-off (the period of early accelerated growth) occurs at approximately 80% of adult height and APHV occurs at approximately 90 to 92% of adult height [106].  As described in Section 2.1.2.4, serial height measures can be plotted in two ways. In the first approach, height (cm) is plotted against age (years) to generate a distance curve from which maximum height can be identified (i.e., the age at height’s plateau). In the second approach, height gain (cm/year) is plotted against age to generate a velocity curve from which APHV can be identified. Table 2.4 illustrates a simple example of calculating height gain (i.e., height velocities) using raw values [122].    30  Table 2.4 Demonstration of a worked sample calculation of height velocity (cm/year) and years from age at peak height velocity (APHV, years) or maturity offset (MO; years) in a girl with idiopathic scoliosis. Adapted from Little et al. [122], with permission from the Journal of Bone and Joint Surgery, Inc., and Rockwater, Inc.     Date  Interval (years) Interval used for velocity (years) Height of patient (cm) Height velocity (cm/year) Years from APHV (MO) (years) May 2, 1988 -- -- 148.0 -- -1.8 January 30, 1989 0.75 0.75 152.0 5.3 -1.0 February 12, 1990 1.04 1.04 162.5 10.1 0.0 April 12, 1990 0.16 1.20 164.0 10.0 0.2 August 13, 1990 0.34 0.50 165.5 6.0 0.5 October 15, 1990 0.17 0.51 166.0 3.9 0.7 February 11, 1991 0.33 0.50 168.0 5.0 1.0 September 30, 1991 0.63 0.63 169.0 0.0 1.6 April 27, 1992 0.58 0.58 160.0 0.0 2.2 May 10, 1993 1.04 1.04 169.0 0,0 3.2   In a seminal case study, Montebeillard [123] measured his son every 6 months from birth to age 18 years and calculated height distance and velocity using a similar technique as just outlined in Table 2.4. He then plotted the data to create curves. Although these distance (left) and velocity (right) curves (Figure 2.11) only represent growth of one child, they show the general pattern of growth that is common across all children and are historically important in growth and maturation literature.   31   Figure 2.11 Illustration of the growth in height of Montbeillard's son from birth to 18 years, 1759 to 1777. The distance curve (left panel) indicates height attained at each age, and velocity curve (right panel) indicates the increment of height gained from year to year. Reproduced from Tanner [1] and drawn with data from Scammon [123], with permission from the American Journal of Physical Anthropology, and John Wiley & Sons, Inc.    The velocity curve of Montbeillard’s son, and height velocity curves from a larger sample of British children from the HGS (Figure 2.12), illustrate five distinct periods of growth: a) rapid deceleration during infancy, b) steady acceleration during childhood, c) initiation (take-off) of the adolescent growth spurt, d) deceleration after the adolescent growth spurt, and e) cessation of growth in young adulthood. I describe average timing of these notable growth curve parameters in Section 2.1.4.3.3.  32   Figure 2.12 Illustration of the typical individual velocity curves for height in boys and girls. Reproduced from Malina, Bouchard, & Bar-Or [19], with permission from Human Kinetics Publishing and Archives of Disease in Childhood, BMJ Publishing Group Ltd.    Lastly, to compare a child with others in a cohort, children can be aligned based on their maturity represented by APHV and MO (i.e., all children would be aligned at MO=0, or biological zero) rather than their chronological age. MO represents maturational time-points based on distance (years) from APHV. Negative values are assigned for years preceding APHV, and positive values for years following APHV (one year before APHV, MO= -1; at APHV, MO= 0; and one year after APHV, MO= +1, so on).  Aligning children by biological age can account for maturational differences in children of the same chronological age. To illustrate, Figure 2.13 shows similarly shaped height velocity curves (although peaks were of different magnitudes) of early-, average-, and late-maturing children from the PBMAS when curves are aligned by MO (right) as compared with variability in take-off, APHV between the curves when aligned on chronological age (left) [17]. Before discussing these maturational events (take-off, APHV, and magnitude), I will first review the methods used to assess APHV in the next section.  33   Figure 2.13 Illustration of the height velocity (cm/years) aligned to chronological age (left) and years from peak height velocity (right) in early, average, and late-maturing healthy boys from the Saskatchewan Growth and Development and Pediatric Bone Mineral Accrual Studies. Reproduced from Sherar et al [17], with permission from the Journal of Pediatrics, Mosby, Inc.   2.1.4.3.2 Assessing Age at Peak Height Velocity Prior to the advent of sophisticated statistical software there was first the incremental approach, where the adolescent growth spurt was quantified by plotting change between two measurements (e.g., Table 2.4) and connecting the two points with a straight line [124, 125]. However, this simple incremental method fails to account for the curvilinear nature of growth. In addition, measurement errors are more likely to influence shape of the curve and thus, APHV may be over- or underestimated [126]. Second, a graphic approach was subsequently developed to quantify the adolescent growth spurt [127]. In the graphic approach, height at each observation point is plotted and then a smoothed curve is fitted using the plotted velocities to calculate a velocity curve for each individual. From the velocity curve, one can visually assess age at take-off, magnitude of PHV, and APHV [126]. Graphic approaches were used in several longitudinal studies to estimate APHV [127-130]. Both incremental and graphic methods are limited by their inability to estimate values between measurement points, and curves created with incremental and graphic methods 34  can be difficult to re-create given the subjective nature of the curve-fitting procedure; subsequently, both tend to overestimate magnitude of PHV [25, 131].  The third and fourth approaches used to quantify the adolescent spurt are mathematical in nature. The five main goals when modeling longitudinal growth data using mathematical functions are to: a) estimate continuous growth from a set of discontinuous measures to obtain a smooth graphical representation of growth, b) estimate between measurement occasions (i.e., an interpolation technique), c) summarize the data using a limited number of function parameters (i.e., a data reduction technique), d) estimate a smooth velocity curve representing instantaneous velocities, and e) assess the ‘average’ curve for a population [126]. Of these models, there are options for structural (or parametric) and non-structural (or non-parametric) modeling of height data to determine APHV. I discuss these approaches in more detail below.  Structural Models Structural (parametric) approaches apply a basic functional form to the growth model. I will provide a brief description of one parametric model: the Preece-Baines model 1 (PB1). Other examples of structural models include logistic (e.g., double, triple), and the Gompertz, Infancy-Childhood-Puberty (ICP), amongst others [25].   Preece-Baines Model 1 The PB1 is one of the more frequently used structural models used to in pediatric growth and development studies compared with other structural curve-fitting procedures [22, 87, 132, 133]. PB1 fits height data from childhood to adolescence; though, it was not designed for modeling growth before two years of age. There are five parameters: parameter θ locates the adolescent growth spurt along the time axis and identifies APHV, hϴ is size at age θ, and so and s1 represent parameters of growth rate constants, related to pre-pubertal and pubertal velocity, respectively [22]. As illustrated in Figure 2.14, distance and velocity 35  curves derived from the PB1 fits raw data for this example fairly well; though, magnitude of PHV seems to be blunted and no mid-childhood spurt is indicated.    Figure 2.14 Representation of a distance (solid line) and velocity (dashed line) curves for longitudinal height data for an example boy using Preece-Baines model 1 curve-fitting. Where h1 is adult size; hϴ is the size at age Ɵ; s0 and s1 are growth rate constants related to prepubertal and pubertal velocity. Reproduced from Hauspie, Cameron, & Molinari [25], with permission from Cambridge University Press.    Limitations of Structural Models Structural models have some noted limitations. First, structural models tend to impose a rigid shape on the growth pattern, as opposed to letting the data ‘speak for themselves’. This makes it difficult to determine sufficient individual variability. For instance, PB1 may have difficulty defining the correct curve during the pubertal growth period when growth is more variable over the population and cannot be predefined [134]. Second, curve magnitude may be affected by the data introduced to the model. For example, using PB1, the mid-childhood and adolescent height curve may appear to be blunted or not present at all if limited data 36  were used from the pre-pubertal years [131]. Third, some structural models require data across the entire growth period (e.g., ICP) to estimate the curve parameters; many studies simply do not have the amount of data required to perform such functions [25].   Non-Structural Models Non-structural (non-parametric) growth models were developed to overcome the limitations of structural models described above. Non-structural models do not assign a particular form to the growth curve [135]; these models work without an a priori fixed model and involve smoothing that suppresses measurement error and short-term variation. Curves developed with non-structural models are normally short-term functions (e.g., the adolescent growth spurt) that smooth data during specific growth periods instead of fitting functions from birth to adulthood [24]. Examples of non-structural models include polynomials, kernel estimators, splines, and shape-invariant models (e.g., SuperImposition by Translation and Rotation (SITAR)), amongst others. I will briefly describe some of these below.  Polynomials The polynomial model is a nonlinear system of data representation for longitudinal data [136]. Polynomials are generally easy to fit and describe growth over short-term intervals (e.g., adolescent growth spurt) [136, 137]. Polynomials were used by the LLTS and Leuven Longitudinal Growth Study to estimate APHV and magnitude of PHV in Belgian children [138-140]. When assessing a longer periods of growth (e.g., from pre- to late-puberty), polynomials are limited by several factors including the large number of parameters that must be estimated, a lack of functional interpretation for the parameters estimated, an inability to assess sudden changes in growth rate such as the adolescent growth spurt, instability in extremities of the growth curve, and no asymptote to a final value. More sophisticated methods such as kernel estimations and spline functions are more appropriate in these cases [137].    37  Kernel Estimation Kernel estimation (kernel regression) is a non-parametric procedure that provides good curve-fitting to individual serial data and can be applied to various ages [141]. Like other non-structured models, kernel estimation does not assume a specific shape to restrict the curve and as a result of this flexibility, kernel estimation can be applied to variables other than height (e.g., weight). The principle of this method is that the estimates are weighted averages of the observations [142]. To apply this model, one must choose either a weight function or kernel k as well as the smoothing parameter or bandwidth h [25]. Kernel estimation was used to determine APHV in the Zurich Longitudinal Study [131, 133], amongst others. Overall, kernel estimates represent the growth curve fairly well compared with other models (Figure 2.15). However, kernel estimates are more variable at tail ends (e.g., closer to birth and 18 years of age) and do not perform as well when height measurements are not equally spaced. For this reason, the Zurich Longitudinal Study group moved to using smoothing splines that fit the growth curve better when data are missed or mistimed (i.e., sporadic) and found the models to be relatively more stable at the curve ends [143].   Figure 2.15 Illustration of height velocity curves of a girl in the First Zurich Longitudinal study from birth to 20 years, where dots (stars) are the raw mid-point calculated velocities, sold line is the kernel estimation fit, and dashed line is the Preece-Baines model 1 fit. Reproduced from Hauspie, Cameron, & Molinari [25], with permission from Cambridge University Press.    38  Smoothing and Interpolating Splines Splines functions are defined by polynomials and may either be smoothing or interpolating. The mostly commonly used spline is cubic (of the third power) because this spline stays true to the form of the original data but estimates points between measures through interpolation [144]. The cubic spline was used to model growth parameters in the Zurich Longitudinal Study [24], the SGSD, and the PBMAS [145].  A smoothing cubic spline forms a smooth line between data points that do not necessarily intersect the individual knots (i.e., measurement points). This allows for interpolation at APHV and reduces variability caused by measurement error. However, with more smoothing there is greater likelihood of bias. Specifically, a smoothing parameter introduces a penalty for ‘roughness’, which represents rate of exchange between residual error and local variation or between smoothing and goodness-of-fit [23]. Thus, as the penalty increases, the degree of smoothness also increases. In comparison, interpolating cubic spline interpolates polynomials from the information of neighboring knots (or data points), and directs the curve through each knot to ensure all measurements are included in the curve. As a result, the line is constrained to pass through the given points but is otherwise free to fall into any shape [23]. Interpolation preserves all observed measurements. Overall, interpolating cubic splines are simple to construct and tend to estimate the growth curve well regardless of spacing of height measurements [146]. Figure 2.16 illustrates the difference between using a smoothing and interpolating spline to model the data from Montebeillard’s son [1], which I first described in Section 2.1.4.3.1. The smoothing cubic spline reduces noise (i.e., variability) in the curve, whereas the interpolating cubic spline passes through each data point. Thus, there is a tradeoff between bias (with smoothing curves) and variability (with interpolating curves) depending on the type of spline function used. 39   Figure 2.16 Demonstration of velocity growth curves of Montebeillard’s son replicated in Stata. The black dots represent the calculated velocities. The black line shows the interpolated cubic spline fit through all knots. The blue line shows a smoothing cubic spline where a roughness penalty has been introduced of 0.000006. The grey horizontal line indicates a roughness penalty of zero. Drawn by Dr. Penny Brasher, PhD, Centre for Clinical Epidemiology and Evaluation, University of British Columbia with data from Scammon [123], with permission from the American Journal of Physical Anthropology, John Wiley & Sons, Inc.     The Zurich Longitudinal Study was the first to use smoothing cubic spline functions to describe the adolescent growth curve [24], following the work of Reinsch [144] who estimated APHV for boys and girls. More recently, researchers from the SGDS and the PBMAS used interpolating cubic splines to fit growth data [3, 85, 147, 148] instead of PB1 as in their earlier analyses [149, 150] because splines were less rigid, used observed height data, allowed for missed or mistimed measures, performed well at the tail ends, and required fewer parameters compared with the PB1. Interpolating cubic splines (that stay true to the original data) are best used when there is confidence that height measures are accurate. Interpolating splines also better estimated mid-childhood growth (and spurt), take-off, and height gain compared with other functions (e.g., PB1 or smoothing splines) [151].     40  SuperImposition by Translation and Rotation  SuperImposition by Translation and Rotation (SITAR) models average distance and velocity height curves during pubertal growth (and other variables such as weight). SITAR models were developed based on infant growth models proposed by Beath [152]. SITAR models average growth curves using a shape invariant model with random effects [153] and were recently used to model growth for the National Survey of Health and Development and the Avon Longitudinal Study [154]. SITAR fits average curves for height very well (explaining over 95% of the variance) based on the size, timing, and intensity of individual curves. That is, the model aims to make individual curves as similar as possible (i.e., to create a mean curve). The model incrementally adjusts for each of the SITAR’s parameters [153]. First, the individual curves are shifted up or down to adjust for random height offset (i.e., the size parameter). Second, the curve is shifted left-right to adjust for random age offset (i.e., the tempo parameter). Third, curves are compressed or stretched to adjust for different velocities (i.e. the velocity parameter). Fourth, the residuals are ‘trimmed’ and outliers removed (Figure 2.17) [153].   41   Figure 2.17 Demonstration of the incremental adjustments for a) size (shifting curves vertically), b) tempo (shifting curves horizontally), c) velocity (shrinking or stretching scale), and d) removing extreme outliners, in the SuperImposition by Translation and Rotation (SITAR) model to create average distance curve. The same method can be applied to create velocity curves illustrating mean age at peak height velocity (APHV). Reproduced from Cole, Donaldson, & Ben-Shlomo [153], with permission from the International Journal of Epidemiology, Oxford University Press.    Summary of Modelling Techniques Overall, the choice of model (i.e., structural or non-structural) depends on which part of the growth curve is being estimated, whether you are creating individual or average curves, availability and regularity of height measures, available software, among other considerations. Structural models describe the growth curve of most children adequately, but are limited when the pattern of growth is not typical (e.g., early- or late-maturers). In contrast, non-structural models may be more appropriate when estimating individual growth curves in children with different patterns of growth, as they do not assume a specific shape to the curve and instead let the data determine the pattern of the curve. After careful consideration, I used 42  interpolating cubic splines to create individual curves and SITAR to create average curves in the studies in my dissertation.  2.1.4.3.3 Average Timing and Magnitude of Take-off and Peak Height Velocity  Now that I have described the techniques used to model the growth curve, I will summarize some of the key features of the growth curve and give examples from historical longitudinal studies. In girls, the adolescent growth spurt starts between age 9 and 10 years, on average, peaks at approximately age 12 years, and stops at age 16 years. Whereas in boys, the growth spurt starts one to two years later than in girls, on average, at age 10 to 11 years, peaks just prior to age 14 years, and stops at approximately age 18 years [19]. However, these values vary slightly with geography, ethnicity, and method of curve fitting. In Table 2.5 and Table 2.6, I list studies in which APHV, magnitude at PHV, and age at take-off of PHV were available for boys and girls, respectively. From these key growth studies, I provide means and standard deviations for APHV, PHV, and age at take-off and I subsequently determined percentiles (5th, 10th, 25th, 50th, 75th, 90th, and 95th), where I assumed normal distribution for each variable.  In Study 1, Roy and colleagues assessed height velocity in healthy boys and girls in Paris, France between 1953 and 1975. Children were recruited between 1953 and 1955 and followed longitudinally until 1975. Findings were presented separately for 68 boys [128] and 80 girls [129]. In both sexes, Roy and colleagues used a graphic approach to determine APHV; they plotted each observation and fit a smooth curve (described in Section 2.10.4.1.2). Age and height at take-off, APHV, and PHV were determined from the individual growth curves [126].  Study 2, the HGS, assessed height in healthy British children from 1948 to 1971. Pre-pubertal children were assessed through to adulthood. One technician (R.W. Whitehouse) measured each child in their home for the duration of the study [2]. Measurements were obtained every six months before the expected time of peak height gains, three months during the time of expected peak height gain, and annually thereafter until cessation of growth [2]. Findings are presented for 55 boys and 35 girls [2]. Authors used Gompertz 43  curves and logistic curves fit to individual height measurements for those whose growth was followed until adult height. Curves were also fit for sitting height, leg length, bi-acromial diameter, and bi-iliac diameter. In later publications, the Harpenden investigators transitioned to the PB1 to model growth data [57].  Study 3, the Zurich Longitudinal Study (1955 to 1976), included 222 healthy Swiss children (112 boys and 110 girls) and used smoothing spline functions to model growth data to estimate age at take-off, APHV, and PHV. The Zurich researchers excluded data from children who were missing more than three measures or two consecutive measurements between ages 4 and 18 years, those with illnesses known to influence growth and those undergoing hormone treatments [24]. Children were measured annually before the age of 9 years in girls and 10 years in boys, within 2 weeks of their birthdays. Thereafter, children were measured every six months until their height change was no more than 0.5 cm per year. Measurements were discontinued after adult height was attained and a plateau was observed for two years (no more than 0.5 cm growth over two years). Study 4 included 212 randomly selected children for the International Children’s Centre study (122 boys and 90 girls) from Stockholm, Sweden (1955 to 1978) who were measured between the ages of 9 and 23 years (boys) and 6 and 20 years (girls). After age 10, height was recorded every three months. Graphic curves (without smoothing) were created using incremental age (adjusted to full year) and height measures, with the purpose of comparing maturational timing and magnitude between children in medical clinics [155]. Study 5, the first Indian Longitudinal Growth Survey (1952 to 1966), included 303 boys and 260 girls (age 0.5 to 20 years), from middle class families in semi-urban south Calcutta (Kolkata, capital of the Indian state of West Bengal). Table 2.5 and Table 2.6 includes growth parameters for a subset of 63 boys and 42 girls for whom APHV could be determined [156]. Children were measured every 6 months between the ages of 6 months and 5 years, annually from age 5 to 10 years, every 6 months from age 10 to 14 years, and yearly thereafter. Average length of observation was 7.5 years for boys and 7.1 years for girls. Following protocols from the HGS, one technician assessed height exclusively throughout the study. Children under 44  three years were assessed supine, and stretch stature (height) was obtained thereafter. Study investigators used the PB1 to fit individual growth curves; where the maximum gap for interpolation between height measures was 13 months [156]. Study 6, the SGDS (1964 to 1973), randomly selected children on a stratified socio-economic basis from elementary schools in Saskatoon, Canada. The study employed a longitudinal design for boys (measuring 207 boys at baseline and annually thereafter) and a mixed longitudinal design for girls (recruiting and measuring 7 to 10 year old girls annually for 5 years until they had recruited 148 girls). Table 2.5 and Table 2.6 includes data from a subset of 75 boys and 22 girls for whom APHV could be determined using the PB1 [157]. 45  Table 2.5 Description of the age at peak height velocity (APHV, years), magnitude of PHV (cm/year), and age at take-off of PHV (years) in boys from six selected longitudinal studies. Studies are listed in order by publication date and represent cohorts from Europe, North America, and South Asia.   Percentiles Study Details Variable N Mean (SD) 5th 10th 25th 75th 90th 95th Study 1 [128] APHV (years) 68 13.8 (0.9) 12.3 12.6 13.2 14.4 15.0 15.3 Paris, France  Magnitude at PHV (cm/year)  9.7 (1.1) 7.9 9.3 9.0 10.4 11.1 11.5  Age at take-off (years)  11.0 (1.3) 8.8 9.3 10.1 11.9 12.7 13.2           Study 2 [2] APHV (years) 55 13.9 (0.8) 12.5 12.8 13.4 14.4 15.0 15.3 Harpenden, England Magnitude at PHV (cm/year)  8.8 (1.1) 8.1 7.5 8.1 9.5 10.1 10.5  Age at take-off (years)  12.1 (0.9) 10.7 11.0 11.5 12.7 13.1 13.4           Study 3 [24] APHV (years) 112 13.9 (0.8) 12.6 12.9 13.4 14.4 14.9 15.2 Zurich, Switzerland Magnitude at PHV (cm/year)  9.0 (1.1) 7.2 7.6 8.3 9.7 10.4 10.8  Age at take-off (years)  11.0 (1.2) 9.0 9.5 10.2 11.8 12.5 13.0           Study 4 [155] APHV (years) 122 14.1 (1.1) 12.3 12.7 13.4 15.1 15.5 15.9 Stockholm, Sweden Magnitude at PHV (cm/year)  9.9 (1.1) 8.2 8.5 9.2 10.6 11.3 11.7  Age at take-off (years)  12.1 (1.2) 10.1 10.6 11.3 12.9 13.6 14.1           Study 5 [156] APHV (years) 63 14.3 (1.0) 12.6 13.0 13.6 15.0 15.5 15.9 West Bengali, India Magnitude at PHV (cm/year)  8.7 (1.3) 6.6 7.1 7.8 9.6 10.4 10.8  Age at take-off (years)  10.5 (1.5) 8.1 8.8 9.5 11.5 12.4 12.9           Study 6 [49] APHV (years) 75 14.3 (1.0) 12.7 13.0 13.6 15.0 15.6 15.9 Saskatchewan, Canada Magnitude at PHV (cm/year)  9.4 (1.5) 6.9 7.5 8.4 10.4 11.3 11.9   Age at take-off (years)  11.1 (1.0) 9.5 9.8 10.4 11.8 12.4 12.7 46  Table 2.6 Description of the age at peak height velocity (APHV; years), magnitude of PHV (cm/year), and age at take-off of PHV (years) in girls from six selected longitudinal studies. Studies are listed in order by publication date and represent cohorts from Europe, North America, and South Asia.    Percentiles Study Details Variable N Mean (SD) 5th 10th 25th 75th 90th 95th Study 1 [129] APHV (years) 80 12.0 (0.9) 10.5 10.8 11.4 12.6 13.2 13.5 Paris, France  Magnitude at PHV (cm/year)  8.4 (0.9) 6.9 7.2 7.8 9.0 9.6 9.9  Age at take-off (years)  9.3 (1.1) 7.5 7.9 8.6 10.0 10.7 11.1           Study 2  [2] APHV (years) 35 11.9 (0.9) 10.4 10.7 11.3 12.5 13.0 13.4 Harpenden, England Magnitude at PHV (cm/year)  8.1 (0.8) 7.0 7.3 7.6 8.6 9.1 9.4  Age at take-off (years)  10.3 (1.0) 8.7 9.1 9.6 11.0 11.5 11.9           Study 3 [24] APHV (years) 110 12.2 (1.0) 10.6 10.9 11.5 12.9 13.5 13.8 Zurich, Switzerland Magnitude at PHV (cm/year)  7.1 (1.0) 5.5 5.8 6.4 7.8 8.4 8.7  Age at take-off (years)  9.6 (1.1) 7.8 8.2 8.9 10.3 11.0 11.4           Study 4 [155] APHV (years) 90 12.0 (1.0) 10.4 10.7 11.3 12.7 13.3 13.6 Stockholm, Sweden Magnitude at PHV (cm/year)  8.6 (1.1) 6.8 7.2 7.9 9.3 10.0 10.4  Age at take-off (years)  10.0 (1.3) 7.8 8.3 9.1 10.9 11.7 12.2           Study 5 [156] APHV (years) 42 12.4 (1.0) 10.7 11.1 11.7 13.1 13.6 14.0 West Bengali, India Magnitude at PHV (cm/year)  7.2 (1.1) 5.3 6.1 6.5 7.9 8.6 9.1  Age at take-off (years)  9.3 (1.1) 7.5 7.9 8.6 10.0 10.7 11.1           Study 6 [49] APHV (years) 22 11.8 (0.7) 10.6 10.9 11.3 12.3 12.7 13.1 Saskatchewan, Canada Magnitude at PHV (cm/year)  8.0 (1.7) 5.2 5.8 6.9 9.1 10.2 10.8   Age at take-off (years)  9.0 (1.0) 7.4 7.7 9.3 9.7 10.3 10.6 47  2.1.4.3.4 Methodological Considerations for Estimating Age at Peak Height Velocity When assessing APHV, investigators must consider the following questions: a) how much missing or mistimed data are acceptable when calculating APHV? b) how much data pre- and post-APHV are needed to accurately calculate APHV?, and c) if a peak is observed at a participant’s first or last measurement, does it represent APHV?     Missing and Mistimed Data Few studies consider the optimal time between measures, or amount of missingness, considered acceptable to accurately assess APHV. Tanner and Davies [158] suggested using whole-year height velocities to reduce both measurement noise and the likelihood of seasonality being captured in the measurement. Further, Tanner and Davies [158] suggested a minimum distance of 0.85 years and a maximum of 1.15 years between height measures; too many measures (i.e., minimum distance) may increase error and variability, and too few measures (i.e., maximum distance) may blunt the curve [158]. The Fels group also argued that no more than three years between measures should be tolerated, as larger gaps could potentially ‘smooth out’ (i.e., blunt) observed height velocity peaks [159]. Hauspie et al. [156] also argued a maximal measure interval (13 months). Whereas other groups adopted a minimum interval of six months between measures [122, 160]. Increments calculated over a short interval can reflect seasonal variation and are more affected by measurement error [158], while increments over intervals larger than one year smooth out the curve [20]. It is crucial to consider the time between measures and be aware of the limitations of measurements that are too close (potential for noise, measurement error, and inaccurate peaks) or too sporadic (potential for blunting peaks) during the growth period.   Ages of Entry and Exit Many of the longitudinal studies in Table 2.5 and Table 2.6 extended from infancy into adulthood and were therefore able to capture all aspects of the growth spurt (age at take-off, APHV, and cessation of 48  growth). However, pediatric studies need to recruit children early enough and retain them late enough so that APHV is captured within the study window. When measuring across a shorter time span during puberty, the first or last measurement may appear as the peak value. This may or may not be accurate and can depend on the timing and magnitude of the peak during the first or last measurement. In these cases, APHV can only be ascertained if the measurement period is extended so that actual peak magnitude (if different) can be identified. This also speaks to the important task (although time consuming) of plotting and visually reviewing each individual’s data – as I did for this dissertation. I illustrate some of the interpretation challenges in Figure 2.18 using HBSIII data.  Figure 2.18 presents a boy who was 10 years of age at study entry. He left the study (lost to follow-up) at age 13.5 years. If I had only reviewed the interpolation data and not plotted and inspected these data visually, I may have incorrectly assigned an APHV of 12.7 years with a magnitude (PHV) of 7.5 cm. However, it is clear from the plot that APHV may occur after the measurement window. I excluded this boy from my analysis after visually inspecting the plot. Thus, to accurately determine APHV, it behooves researchers to design longitudinal studies that capture (and retain participants through) the years before, during, and after the adolescent growth spurt.       49   Figure 2.18 Illustration of the limitations associated with prospective studies if a participant is lost to follow-up when determining age at peak height velocity (APHV) using data acquired from a boy in the Healthy Bones Study III (HBSIII). The open circles in the rightmost curve show height (cm; Y-axis) at each measurement, by age (years; X-axis). The connected closed circles in the leftmost curve show velocities (and interpolations) between measures (cm/year; Y-axis). The open square indicates the identified APHV.     First or Last Measurement  Recommendations on whether or not to accept a first or last measure (velocity value) as peak, are inconsistent. Some investigators recommend including these height velocity data  [122], whereas others do not accept these values as peak and subsequently exclude participants [138]. Using retrospective clinical data (five years of height data, on average), Little et al. [122] assessed growth curve parameters in 371 girls with idiopathic scoliosis. Authors suggested that the first height velocity value be identified as peak if the velocity is greater than 9 cm/year and no other values appear to be peak. The authors based their suggestion on the observation that 9 cm/year was two standard deviations from the median value for magnitude of PHV six months before and after peak growth [52]. In contrast, investigators with the Ghent Youth Soccer Project, a five year longitudinal study of youth soccer players aged 10.4 to 13.7 years at entry, did not accept the first or last point as APHV as they were not convinced this value was actually APHV [138]. Given the widespread use of APHV as an indicator of somatic maturity, there is a need for researchers and clinicians to standardize their approaches by collectively selecting minimum acceptable values for APHV for first or last data points.  50  2.1.4.3.5 Predicting Somatic Maturity In some cases sufficient serial measures of height are not available, height measurements are mistimed or sporadic and study duration is too short to capture the full growth curve from acceleration to deceleration. In these cases prediction equations are often used to estimate APHV. Mirwald and colleagues [26] published the first sex-specific prediction regression equations that use one-time morphologic measures (age, standing height, sitting height, weight, and leg length) to predict MO and APHV.   In boys, MO is calculated with the formula:   MO= -9.236+(0.0002708 x LL*SH)–(0.001663 x A*LL)+(0.007216 x A*SH)+(0.02292 x W/H*100)  In girls, MO is calculated with the formula:  MO= -9.376+(0.0001882 x LL*SH)–(0.0022 x A*LL)+(0.005841 x A&SH)+(0.07693 x W/H*100) Where MO = maturity offset, LL = leg length, SH = sitting height, A = age; W = weight; and H = height. The authors note R = 0.94, R2 = 0.891 and SEE = 0.592 for boys and R = 0.94, R2 = 0.890 and SEE = 0.569 in girls [26]. Mirwald and colleagues developed these prediction equations using data from the PBMAS. This mixed longitudinal study, conducted from 1991 to 1997, included 113 boys and 115 girls from Saskatchewan, Canada [145]. Participant height, sitting height, and weight were assessed semiannually using standard protocols. To derive the equations, data were included for those participants with height measures from four years pre-APHV to three years post-APHV. Sex-specific multiple regression equations were developed using hierarchical entry that considered both statistical and biological significance (i.e., they kept variables such as leg length in the model even if it was not statistically significant to account for the differential timing of peak trunk length, peak height, and peak leg length). Bland Altman plots were used to determine ability of the equations to estimate APHV and MO and to compare predicted versus actual values of APHV. The equations were refit using samples from the SGDS and the LLTS. I use data acquired from an average-maturing boy to illustrate how one estimates MO using the Mirwald equation (Table 2.7). . 51  Table 2.7 Demonstration of a worked example of maturity offset (MO) prediction in an average-maturing boy. Adapted from Sherar et al [17], with permission from the Journal of Pediatrics, Mosby, Inc.   Age 11.3 years Height 149.4 cm Weight 40.0 kg Leg length 70.4 cm Sitting height 79.0 cm Leg length * sitting height  70.4 x 79.0 = 5561.6   Age * leg length  11.3 x 70.4 = 795.5 Age * sitting height  11.3 x 79.0 =  892.7 Weight / height  (40.0/149.4) x 100 =  26.8 MO= -9.236(0.0002708 x 5561.6) – (0.001663 x 795.5) + (0.007216 x 892.7) + (0.02292 x 26.8) MO = -2.0 years from age at peak height velocity Age at peak height velocity = age – MO 11.3 years – (-2 years) =  13.3 years This boy is -2.0 years away from his age at peak height velocity, which is expected to occur at 13.3 years (categorizing the boy as an average-maturer as the mean APHV in this cohort is 13.6 years)    The Mirwald equation has demonstrated utility for predicting APHV and MO in cross-sectional and short-term prospective studies. However, the equation’s validity is limited if applied to children whose age falls outside the age range of participants in the data source (PBMAS; 9.8 to 16.8 years; estimated MO within error of ±1 year in 95% of cases). Authors recommended that users a) utilize the equations in children ages 9.8 to 16.8 years, and b) estimate a categorical MO (e.g., pre-APHV and post-APHV) with the predicted MO (as opposed to calculating a continuous biological age) [26].   2.1.4.3.6 Critiques of Prediction Equations Malina’s Critiques  Malina and colleagues critically analyze the Mirwald equations in three papers: in a cohort of 13 elite female gymnasts [29] aged 10 years at baseline and followed annually for 7 years, and in 193 Polish boys [28] and 198 Polish girls [27] aged 8 years at baseline, measured annually for 10 years from the Wroclaw Growth Study (1961 to 1972) [27, 28]. Collectively, Malina et al. [27-29] suggested that the Mirwald equations systematically predicted APHV earlier than observed APHV at younger chronological ages and later than observed APHV at older chronological ages (e.g., Table 2.8).  52  Table 2.8 Demonstration of the difference of predicted age at peak height velocity (APHV) minus observed APHV in a cohort of Polish boys. Age group defined by mid-year (i.e., 9 = 8.5-9.49 years). Reproduced from Malina & Koziel [28], with permission from Journal of Sports Sciences, Taylor & Francis Ltd.   Difference of predicted APHV minus observed APHV, years Age Group n Mean SD Min Max 8 186 -1.47 1.09 -4.89 1.43 9 186 -1.09 1.09 -4.49 1.81 10 184 -0.73 1.08 -4.11 2.19 11 185 -0.35 1.03 -3.80 2.36 12 184 -0.01 1.02 -3.40 2.41 13 183 0.18 0.87 -3.03 2.50 14 191 0.25 0.78 -2.66 2.78 15 184 0.30 0.75 -1.67 3.07 16 185 0.38 0.89 -2.00 3.39 17 179 0.57 1.01 -2.02 3.89 18 173 0.82 1.09 -2.20 4.20   Differences between observed and predicted APHV for children in these external populations may reflect a systematic error in the Mirwald equations failure to recognize the dependent nature of observations in the original sample may have led to conservative estimates of the SEE due to within-subject correlation. It is important to note, that the Mirwald equations performed best when used to predict MO for average-maturing boys aged 12 to 15 years and average-maturing girls aged 8 to 11 years [27, 28]. Importantly, these findings are consistent with statements made in the original Mirwald et al. paper [26].   An Analysis of Mirwald Citations The Mirwald equations [26] are commonly used in pediatric studies to predict MO and categorize children and adolescents by maturity status or to align participants on a common maturational landmark. While a number of studies adhere to the guidelines suggested by Mirwald et al. [26], a number of others have used the equations ‘inappropriately’. For example, given the known ethnic differences in tempo and timing of growth and maturation (Section 2.1.6.2), the Mirwald equations [26] may misclassify maturity in non-white children. Despite this, Mirwald equations were applied to data acquired from Middle Eastern 53  [161], Brazilian [162], Portuguese [163], Japanese [164], and Aboriginal boys and girls [165]. Equations were also applied to mixed ethnic-samples [166], including three HBSIII cohorts (42% white, 47% Asian, 11% other or mixed ethnicity) [167-170]. In our HBSIII publications, our authors acknowledged that although equations were not validated for these ethnicities; however, no ethnic-specific alternatives were available at the time of publication [167-170]. This query forms one central tenet of my dissertation and in Chapter 6, I sought to determine whether they are accurate for use in other ethnicities.  Mirwald equations [26] were also used extensively to assess maturity in athletes, for talent identification, or to determine functional changes following activity-related interventions [171-188]. They were also applied to assess maturity status in clinical populations (e.g., Crohn’s disease, inflammatory bowel disease, cystic fibrosis, Osgood Schlatter Disease, juvenile dermatomyositis, obesity) [189-197]. However, once again, because the equations were not validated for use in these populations they may provide inaccurate estimates. Finally, although Mirwald equations [26] were validated in children and adolescents aged 9.8 to 16.8 years [26] some investigators used the equations to provide estimates in children as young as five years [198-205]. Others used the equation to predict (backwards) APHV in young adults [206], many of whom had already reached adult height. It is unclear how the Mirwald equations perform in these age groups.  2.1.4.3.7 Future Directions for Prediction Equations Thus, I recognize an opportunity to further validate the Mirwald equations [26] and to address some of the aforementioned limitations and concerns. First, the standard error of the estimate (SEE; which measures the accuracy of the prediction) for these equations assumed data points were independent. However, data used to develop the equations were not independent as they represented serial measures from the same individual. I contend that the model will be improved (i.e., greater prediction accuracy) when intra-person correlation is considered. Second, if prediction error is high in cohorts used to validate equations, I suggest the equations require modifications (i.e., calibration) to improve accuracy. Third, PBMAS participants were 54  healthy white boys and girls; equations were not validated for use in other ethnic groups. I contend that equations require further validation in other samples of children (e.g., different ethnic groups, clinical groups, athletes who are known to be early or late maturing). Depending on outcomes, it may be necessary to develop new prediction equations for these groups. Thus, this query forms a central tenet of my dissertation (Chapters 5 and 6).  2.1.4.3.8 Other Indicators of Somatic Maturation Leg and trunk lengths and body circumferences and breadths can be tracked to assess somatic maturity. Maximal gains in leg length, thigh, and calf circumference and bicondylar breadth occur before APHV. Conversely, trunk length, biacromial breadth, chest breadth, arm circumference, and weight attain maximal gains after APHV [19]. In boys and girls, leg length velocity precedes PHV, which precedes trunk length velocity (Figure 2.19) [52].   Figure 2.19 Illustration of median values for velocities of height, sitting height (i.e., trunk length), and leg length for boys (left) and girls (right) aligned by age at peak height velocity (APHV), where 0 equals APHV, negative values indicate years pre-APHV and positive values indicate years post-APHV. Reproduced from Buckler [52], with permission from Springer-Verlag Publishing.  55  Ratios of leg and trunk length to standing height are used as indicators of maturation. For example, the ratio of sitting height to standing height is highest in infancy and declines throughout growth. The ratio is lowest in the year around APHV, and then increases slightly into adulthood. The ratio is similar in boys and girls before puberty and slightly higher in girls into adulthood [20]. Ethnic differences in length velocities and ratios have also been examined, though they tend to follow the same pattern as in white children. In boys and girls from NHANES and NHANES II  [207], black children were taller and had longer leg lengths compared with white and Mexican American children at all ages and in both sexes, whereas black children had shorter trunk lengths. Despite the differences in absolute lengths, the pattern of growth was similar; leg length velocity preceded PHV, which preceded trunk length velocity. Trunk length to height ratios were highest in infancy and decreased throughout childhood to puberty, with a slight increase after APHV. The pattern of distributional growth was found to be the same in Asian boys, where leg length velocity preceded APHV, which preceded trunk length velocity [30].  2.1.4.3.9 Percent of Adult Height Percent of adult height is also an indicator of somatic maturity; it increases with age and reaches the same endpoint in all individuals, where adult height is 100% [18]. When adult height is known, percentage of adult height during growth is a retrospective maturity indicator. When adult height is unknown, prediction equations for adult height may be employed, some of which rely on skeletal age [15], whereas others rely on morphological and menarche data (e.g., height, leg length, trunk length, forearm circumference, menarcheal status) [17, 18]. Errors in prediction of adult height across models ranges from 3 to 7 cm [19]. Two simpler models (with error as much as 10 cm or more) include doubling a child’s height at age two [208] and using maternal and paternal height measurements to estimate adult height as per the equations below: 1. Adult height for boys = (Father’s height (cm) + Mother’s height (cm) +13) /2 2. Adult height for girls = (Father’s height (cm) + Mother’s height (cm) -13) /2 56  2.1.4.4 Relationship between Maturity Indicators Associations between maturity indicators are moderate to high [130, 209, 210]; though there is sufficient variability to suggest that no single maturity indicator provides a complete description of the tempo of maturation. For example, although sexual maturation and skeletal development are associated, one cannot assume that an individual in one stage of secondary sexual development is in a relative stage of corresponding skeletal development [130]. Figure 2.20 illustrates the relationship between various maturity indicators, and highlights the tremendous range in the normal timing of these events. The apparent discord across indicators reflects individual variation in timing and tempo of sexual and somatic maturity [1].    Figure 2.20 Illustration of the relationship between and ranges of somatic maturity as measured using age at peak height velocity and sexual maturity in girls (left) and boys (right) Numbers below represent the range for the respective indicator. Reproduced from Danker-Hopfe [211] Drawn from data from Tanner [1], with permission from Progress in Biophysics and Molecular Biology, Pergamon Publishing, and John Wiley & Sons, Inc.   2.1.5 Secular Changes in Growth and Maturation Tanner [212] suggested that “growth is a mirror of conditions in society”. Thus, secular trends may evaluate a population’s changing health status. A ‘secular trend’ describes population-wide change of a 57  characteristic over a relatively long period (i.e., generations, decades, centuries, or even longer) [213]. A positive (or upward) secular trend represents an increase in body size (e.g., height, weight) or earlier onset of maturational events over time, whereas a negative (or downward) secular trend indicates a decrease in body size or later onset of maturity over time [142]. Positive secular trends may reflect improved nutrition, control of infectious diseases, increased access to health care, available immunizations, sanitary living conditions, and/or greater population mobility (geographic, social, and/or economic). However, positive secular trends are not necessarily health promoting. For example, a positive secular change in weight and BMI (toward an overweight or obese state) may be attributed to a decline in physical activity and/or an increase in caloric consumption [214]. In contrast, periods of deprivation, such as wartime, may lead to negative secular trends in growth and maturation [30].   2.1.5.1 Secular Changes in Growth Over the past 100 years children have become taller and heavier compared with their peers from the early 20th century [213]. This is particularly true in industrialized countries, but has also been observed in developing nations. This trend continued into the 1990s, although to a lesser extent and with more variability in more recent years [19]. Children are now noticeably larger from birth, considered a factor of increased SES leading to better nutritional intake, maternal size (i.e., taller and heavier mothers), and reduced maternal cigarette smoking [215]. The largest secular gains in heights and weights have been observed from mid-childhood growth through adulthood, considered due to better nutritional conditions during pubertal growth [213]. While increased height is positively associated with health [216], the continuing increase in children’s weight is of concern [217]. Developed nations report an increase in height of at least 0.5 cm per decade, to as much as 2 cm or more per decade (from 1900 to present) [142]. Further, weight has increased at least 0.5 kg per decade to as much as 3kg or more per decade [142]. These trends were observed globally and across ethnicities, with most changes occurring in post-war industrialized countries. For example, in a Canadian study of four 58  cohorts of primarily (95%) white children, boys and girls were 10 cm taller (1.2 cm per decade), on average, in 1974, compared with children in 1891 (Figure 2.21) [218]. Further, Hughes et al. [219] assessed secular changes in British white children and reported a 0.5 cm/decade increase in height and 0.3 kg/decade increase in weight in 5 to 7 year olds, and a 0.8 cm/decade increase in height and 1 kg/decade increase in weight for 8 to 10 year olds. In the US, NHANES data indicate an increase in height of approximately 1 cm/decade for almost all age groups and ethnicities [220, 221]. Similarly, in Japan height increased in both boys and girls after age three by more than 2 cm/decade [222]. The exception was during World War II (WWII; 1940 to 1945) when conditions during the war negatively affected growth. Improved conditions post-war demonstrated the same if not a greater positive secular trend of increased height in Japanese boys and girls [222]. Recently, Cole & Mori [223] assessed 50 years of height data obtained from Japanese and South Korean boys and girls. Authors described similar secular trends in Japanese populations. South Koreans demonstrated the largest secular increase in height compared with other Asian countries (e.g., South Korean women are now 20 cm taller than they were a century ago). Most importantly, Cole and Mori determined that increased height gains from birth to age two years was the largest contributor (compared to height gains during other periods of growth) to the secular increase in height [223]. In this study, the strongest positive influence on height gain was daily protein (r=0.79), while rice diets had strong negative effects (r=-0.74) [223]. Conceivably, secular changes in height in developing nations could be due to the adoption of western-influenced diets.    59    Figure 2.21 Illustration of the mean height-for-age in Canadian boys (left) and girls (right) in four birth cohorts indicating a positive secular change from 1891 to 1974. Adapted from Hoppa [218], with permission from Annals of Human Biology, Taylor & Francis, Ltd.     2.1.5.2 Secular Changes in Maturation Limited prospective data suggest that onset of puberty occurs earlier in today’s children compared with previous generations [224, 225]. However, others noted no (or minimal) secular change in timing and tempo of maturity [142]. Specifically, sexual maturation is occurring earlier in today’s boys and girls compared with previous decades; however, there is less evidence that somatic maturity occurs earlier.  For example, in developed countries, onset of sexual maturation (e.g., menarche) occurs earlier now than the last several decades (Figure 2.22). The most notable decline in age at menarche were observed from the mid-19th to mid-20th centuries, where age at menarche declined from 17 years of age to 14 years of age or younger [226]. Mean age at menarche in white European girls declined about 0.3 years per decade between 1880 and 1960 [227]. In the US, average age at menarche declined from 14.7 years in the 1870s to 14 years in 1900 and 12.8 years in the 1950s [228]. Some suggest these changes are due to increased rates of childhood overweight and obesity, where higher levels of pre-pubertal body fat is thought to be related to earlier sexual maturation [229, 230]. Similar secular trends were observed in Asian girls: in South Korea, age at menarche declined from 16.8 to 12.7 years between 1920 and 1986 [231]. From 1963 to 1993 age at menarche declined 0.5 years in urban Chinese girls [232], and declined by 2.8 years over 40 years in 60  a second cohort of Chinese girls from rural areas [233]. Similar trends were observed in Japanese girls, where age at menarche declined from 0.4 years per decade from 1884 to 1980 [234]. Roche and Sun suggested that much of the secular difference in age at menarche can be explained by reduced variability in reported age at menarche with fewer girls experiencing late puberty [142]. Secondary sexual characteristics also appear to be emerging at younger ages. For example, in the Netherlands, (physician assessed) breast bud development in girls and pubic hair development in boys and girls were reached 0.3 to 0.6 years per decade earlier in 1997 compared with 1965 [235]. The authors suggested that the secular trend was due to a number of factors, including increase in parental education, increase in SES, improvement in food quantity and quality, decreased energy expenditure, increased dairy in diet, and overall better levels of child health (i.e., hygiene, vaccines, etc.) [235].   Figure 2.22 Illustration of the secular changes in age at menarche in various nations from 1960 to 2000. Reproduced from Parent et al. [226], with permission from Endocrine Society.     Interestingly, although height has increased secularly in most nations, little evidence supports a significant secular change in APHV in white European or North American children. For example, in Belgian girls, APHV occurred only slightly earlier (less than 3 months, on average) in 1980 than in 1960 61  [236]; however, APHV was estimated from cross-sectional data using maximum increment age (MIA) of the means as an estimate of mean APHV [236]. Similarly, in a sex-specific parent-child comparison from participants in the Fels Longitudinal Study [237], APHV did not differ between parents and their children. In Asian countries (Japan, Taiwan and China), APHV (assessed by MIA) occurred 0.2 years per decade earlier in Asian boys and 0.1 years per decade earlier in Asian girls between 1943 and 1978, with the exception of years during WWII [238]. Similarly, Cole & Mori [223] did not observe a difference in APHV (assessed by MIA) in Korean and Japanese boys and girls over a 60 year period (1950 to 2010). Though estimates of APHV using MIA are suggested to be less accurate [236].   2.1.6 Factors that Influence Growth and Maturation In the following sections I describe key factors that influence growth and maturation, including non-modifiable (i.e., heredity, ethnicity, sex, and hormones) and modifiable (i.e., nutrition and physical activity) factors.   2.1.6.1 Genetics Genetic factors account for 60 to 90% of the total variation in adult height [239]. Heritability (h2) of a trait provides an understanding of genetic influence on a trait. The h2 observed in a trait can be ascribed to genetic factors ranging from 0% (or 0.0, no genetic effect) to 100% (or 1.0, completed genetic control) [240]. In general, height is highly heritable, as h2 estimates range from 0.6 to 0.9 [240]. Heritability estimates are highest in developed countries, in twins and siblings, and in families [241]. Conversely, estimates of h2 tend to be systematically lower in developing countries [241]. This may be explained by poorer environmental conditions, which increase environmental variability, thus reducing estimates of h2.  In a study of 4071 white male adult twin pairs (monozygotic, identical (MZ); and dizygotic, fraternal (DZ)) from the US, h2 for height was 0.80 [242]. Similarly, in a study of 3375 white Australian adolescent and adult twin (MZ and DZ) and sibling pairs, h2 of height was 0.80 [243]. In comparison, in 365 Chinese 62  families (living in China) h2 of height was 0.65 [244]. Although it is difficult to directly compare heritability estimates across studies, lower heritability in Asian families were explained, in part, by differences in SES or environmental conditions as compared with American or Australian cohorts [241].  Maturational events such as APHV are also heritable, although h2 of height velocities at different stages of growth are not well established. In the Leuven Longitudinal Twin Study h2 for APHV was 0.92 (range of growth curve characteristics: h2 =0.89 to 0.93) in 99 white adolescent boy and girl twin pairs (MZ and DZ) [245]. Heritability was slightly less in the Wroclaw Growth Study, as Hauspie et al. reported h2=0.74 for APHV in 86 male twins twin pairs (MZ and DZ) [246]. Differences in h2 between studies may be due to environmental factors (e.g., low SES, nutritional deficits) that may lead to underestimation of h2 [245].  Heritability for menarche is slightly lower than that of APHV. In a sample of 371 white girls and women (representing 112 nuclear and extended families) from the Fels Longitudinal Growth Study, h2 of menarche was 0.49 [247]. Similarly, in white Australian female twin pairs (MZ and DZ) h2 of menarche was 0.5 [248]. Whereas in a cohort of 176 Asian twin pairs (MZ and DZ) h2 of menarche was slightly higher, 0.66 [249]. Lower estimates and variation in h2 may be due to measurement error, recall bias, or environmental interactions. These aforementioned studies all used retrospective self-report to determine age at menarche.   2.1.6.2 Ethnicity The terms ‘ethnicity’ and ‘race’ are often used synonymously in the literature [250]; however, there are important distinctions. ‘Ethnicity’ is multi-faceted and refers to a group an individual belongs to that shares characteristics such as cultural traditions, diet, religion, language, and geographical and ancestral origin. ‘Race’ more specifically refers to the genetic make-up of a population or group leading to similar physical characteristics [251]. In health research, there has been a movement to categorize individuals by ‘ethnicity’ to recognize qualities and traditions of groups. ‘Race’ has become increasingly imprecise, ambiguous, and difficult to categorize [251, 252]. Thus, I use the term ‘ethnicity’ throughout my dissertation. I describe ethnicity classification specific to the HBSIII cohort in Section 4.2.3.1.  63  In their comprehensive compendia of international studies, Eveleth and Tanner [30] noted considerable geographic variation in growth and maturation parameters across countries. This variability was likely due to a number of factors including genotypic and phenotypic characteristics, cultural-specific behaviours, and environmental factors [30]. Some note that ethnic variation in growth may partly be driven by the ecological rules of Bergmann [253] and Allen [254], which suggest that individuals who inhabit colder regions have greater body mass and less exposed portions of the body (i.e., limb length) compared with those living in warmer climates. The decreased ratio of surface area to weight and decreased limb length results in enhanced thermoregulation. I highlight key ethnic-differences in growth (e.g., height and weight) and maturation (e.g., skeletal age, APHV, secondary sexual characteristics, and menarche) below.   2.1.6.2.1 Ethnic Differences in Growth  The most extensive data on ethnic-related differences in growth were collected as part of national surveys. In the US, height and weight data were gathered from the National Health Examination Survey (NHES), National Health and Nutrition Examination Surveys (NHANES), and Hispanic Health and Nutrition Examination Study (HHANES)) in white (European Ancestry), black (African ancestry), and Hispanic (Mexican ancestry) children; these data indicate significant differences in growth by ethnicity [207, 255-258]. Pre-pubertal black boys and girls were taller (1 to 3 cm, on average), compared with white and Hispanic boys and girls. Post-pubertal white boys and girls were taller (1 to 2 cm, on average) compared with white and Hispanic boys and girls. Hispanic boys and girls had the greatest trunk length, and shortest leg length, compared with white and black children, and black boys and girls had the greatest leg length, on average, compared with Hispanic and white children, at all ages [207]. In Canada, mean height, trunk length, and leg length in Canadian white children tend was comparable to white children from the US and Europe [30, 157, 259, 260].   Further, Eveleth and Tanner plotted height and weight data from three samples (birth to 18 years) selected to represent ‘average’ black, white, and Asian ethnic groups (Figure 2.23). The three samples 64  included black American boys and girls from the NHANES II survey [207], white European (Danish) boys and girls [30], and Asian (Japanese) boys and girls from the Japanese National Health and Nutrition Surveys [261]. Asian children were shorter (between 1 and 5 cm over childhood and adolescent growth) compared with white and black children. Black children were slightly shorter than white children between ages three and eight years, but caught up at puberty. In a cohort of Asian children living in the US, Asian children were shorter and had shorter leg lengths compared with black, white, and Hispanic peers in both sexes and  at all ages [30]. In this cohort, height was 8cm greater in white men and women compared with Asian men and women by adulthood [30].   Figure 2.23 Illustration of mean height from birth to 18 years of girls (left) and boys (right) from European (Denmark), Asian (Japan), and African (America) origins. Reproduced from Eveleth & Tanner [30], with permissions from Cambridge University Press.   Birth weight was also less in Asian children (living in Asia) compared with white American and Canadian children (115 to 235 g, on average) at birth [262-265]. Growth of Asian children born in North America also differs from white children. For example, Wen et al. assessed birth weight in white (n=2006) and Chinese-Canadian (n=192) children born in Montreal and Chinese children (n=283) from eastern China. Mean birth weights were 3369 g, 3195 g, and 3171 g, respectively; these differences remained after sex, gestational age, and type of delivery were considered [265]. After birth, Asian children continue to weigh less (between 1 and 8 kg over childhood and adolescent growth) than their white European and North American counterparts at all ages [261]. 65  2.1.6.2.2 Do Ethnic Differences in Growth Persist when a Child Relocates?   Relocation or immigration may affect a child’s growth due to the known influence of environmental factors (e.g., better nutrition, access to health care). Bogin and Loucky [266] illustrate how immigration may influence growth in children. The study assessed two groups of boys and girls aged 5 to 14 years who were children of Maya adults: children of those who immigrated to the US (Indiantown, Florida and Los Angeles, California) between the 1970s and 1990s, and children of those who stayed in Guatemala. Children of Maya immigrants were significantly taller compared with their counterparts living in Guatemala. Although body proportions were not measured, the authors suggested that height differences were likely due to greater leg lengths in US cohorts, which could be attributed to better living conditions [266]. Refugee children or children adopted from poor nations with adverse living conditions will experience an even larger benefit (in growth) with relocation [267, 268].   2.1.6.2.3 Ethnic Differences in Maturation  Similarly to ethnic differences in growth, most studies that assessed ethnic differences in maturation used data from national surveys (e.g., NHES) in the US and included black, white, and Hispanic children [269, 270]. Black children tend to have advanced skeletal maturity (Gruelich-Pyle; 3 to 8 months, on average) before puberty and then white children tend to have advanced skeletal maturity (1 to 6 months, on average) after puberty, and Hispanic children were the least skeletally mature at all ages (1 to 3 months, on average) [19, 94, 271].  Differences in skeletal maturation between Asian cohorts become apparent after puberty. For example, Murata compared skeletal age (TW2) of three Asian cohorts (ages 3 to 19 years) from Japan, China, and India, to two white cohorts from the UK and Belgium [272]. Skeletal maturity was similar across ethnicities before puberty; however, at and after puberty Japanese children were the most skeletally mature, followed by Chinese, Indian, Belgian, and British children (Figure 2.24) [272]. Similar findings were observed in a cross-sectional study of 572 Chinese children and 600 white British children aged 7 to 17 years; Chinese 66  children were of comparable skeletal maturity (TW2) to white children until age 9 years in girls and 10 years in boys. Thereafter, Chinese children matured faster (up to one year) compared with their British counterparts [273]. Murata suggested that earlier ossification (i.e., earlier maturation) of long bones in the hand (i.e., metacarpals and phalanges) in Asian children may explain their overall shorter heights, on average, as compared with other ethnic groups such as white and black children [274].    Figure 2.24 Illustration of skeletal maturity scores (by Tanner-Whitehouse 2 method) by chronological age for boys (left) and girls (right) in Asian children from Japan (solid line), white children from the UK (solid circle), white children from Belgium (open circle), Asian children from North India (solid diamond), and Asian children from South China (open square). Reproduced from Murata [272], with permissions from Acta Paediatrica, John Wiley & Sons, Inc.   Growth during adolescence also demonstrates ethnic-related variability. Berkey et al. compared APHV in black and white US children; black boys (mean age: 13.3 years) and girls (mean age: 10.8 years) reached APHV six and three months earlier than white boys (mean age: 13.6 years) and girls (mean age: 13.3 years), respectively [275]. Magnitude of PHV was not different between these cohorts (girls: 9.5 cm/year and boys: 8.3 cm/year)  In Asian children (living in Asia), APHV occurred earlier than the previously reported groups [19, 30]. For example, Chae et al. reported APHV as 12.5 years in boys (8.6 cm/year), and 10.5 years in girls, (7.1 cm/year) in a cohort of 783 Korean children [276]. This was similar to two cohorts of Chinese boys and girls, where APHV was 13.0 years in boys (7.9 cm/year) and 10.6 years in girls (6.9 cm/year) 67  [232, 277]. Similarly, in Taiwanese children APHV was 12.5 years (9 cm/year) in boys and 10.5 years (7 cm/year) in girls [278]. Whereas in Japanese children, APHV occurred at 13.1 years (9.9 cm/year) in boys and 11.1 years (8.3 cm/year) in girls [279]. Thus, these cohorts of Asian children tend to experience APHV earlier than the white and black children described previously.  Further, height at take-off may influence magnitude of growth at APHV. In their review on international variation in height, Haas and Campirano found that when height gain from age 7 to 18 years was plotted against height at take-off, children from populations with the lowest height at take-off (e.g., white children from European and North American countries) had the greatest growth during puberty. Values were up to a 34.5 cm increase in height in boys (from 11 to 17 years) and a 27.5 cm increase in height in girls (from 10 to 17 years) [280]. Interestingly, Haas and Campirano plotted height velocity in children from African, Asian, and European countries (Figure 2.25). On average, pubertal growth velocity (calculated as height changes from ages 11 to 17 years) was greatest in white children (19 to 34 cm), followed by black children (21 to 30 cm), and least in Asian (particularly East Asian) children (17 to 29 cm) [280]. The authors noted that the amount of growth during puberty was greater for populations living in favourable conditions (i.e., higher SES, better availability to health care, and adequate nutrition) [280].    68    Figure 2.25 Illustration of the height change (cm) from age 11 years to age 17 years in boys (left) and from age 10 years to 17 years in girls (right) for selected samples from four regions. Reproduced from Haas & Campirano [280], with permission from Food and Nutrition Bulletin, Sage Publications, Inc.    Finally, timing of secondary sexual characteristic development and menarche also varies across ethnicities. From US surveys, PH development occurred first in black children (boys: age 11.2 years and girls: 9.5 years, on average), followed by white children (boys: age 12.0 years and girls: 10.5 years, on average), and lastly Hispanic children (boys: 12.3 years and girls: 9.8 years, on average). Whereas breast development was earliest in black girls (age 9.5 years, on average), followed by Hispanic girls (age 9.8 years, on average), and latest in white girls (age 10.3 years, on average) [281]. There are noted differences in hair-related characteristics observed in Asians [282] which may influence the accuracy of PH assessment in Asian boys and girls. Using the method as per Tanner [1], Asian (primarily Chinese) boys (n=18,807) reported later PH development (12.8 years, on average) [283], and similarly, Asian (primarily Chinese) girls (n=20,654) reported later PH development (11.2 years, on average) [284]. The same Asian girls reported earlier breast development (9.2 years, on average) [282] compared with the aforementioned cohorts from US national surveys [281]. It remains unclear whether assessing PH development as per the 69  methods of Tanner [1] (which were developed in white children) is appropriate in ethnic cohorts with dissimilar hair growth.  Lastly, in girls, age at menarche also differs across ethnic groups. For example, black girls reached menarche first (age 12.1 years, on average) [285] followed by Japanese (age 12.3 years, on average) [286] and Chinese (age 12.3 years, on average) [284], white North American (12.6 to 12.9 years, on average) and European girls (12.4 to 13.5 years, on average) [19]. Many variables (nutrition, body weight, SES, sleep, physical activity) influence age at menarche.   2.1.6.3 Sex Sexual dimorphism in body size is apparent prenatally [287], but is most pronounced during and after  adolescence. Sexual dimorphism is attributed to the action of sex (steroid or gonadal) hormones [288], which I describe in Section 2.1.4.2.3. In this section, I will summarize sex differences in growth and maturation. In boys, the adolescent growth spurt in height starts relatively later in puberty (PH and genitalia stage 4, approximately 14 years of age) and approximately 2 years later than the girls’ growth spurt. The additional two years of pre-pubertal growth in boys can result in as much as a 10 cm difference in the overall height between boys and girls  [69]. In boys, adult height is attained approximately eight years after APHV. At APHV, boys and girls have reached approximately 90 to 92% of their adult height [145]. In girls, the adolescent growth sport in height starts relatively early in puberty (PH and breast stage 3 or 4, approximately 12 years of age). APHV typically occurs before menarche, which occurs later during PH stage 4 or 5 [73]. In girls, adult height is attained approximately six years after APHV, or three to five years after menarche [52]. As it relates to skeletal development, boy experience peak bone mineral content velocity (PBMCV) after APHV and approximately 1.5 years later compared with girls  (Figure 2.26) [289].  70   Figure 2.26 Illustration of the peaks for height velocity, bone mineral content (BMC) velocity, growth hormone (GH) amplitude, and insulin-like growth factor-1 (IGF-1) amplitude, and trends for estrogen and testosterone levels in girls relative to average age and Tanner stage. In boys, peak height velocity and peak BMC velocity occur approximately 1.5 years later compared with girls. Relations between peaks for height and bone velocities and peaks for GH and IGF-1 are similar for boys and girls. Reproduced from Mackelvie et al. [289], with permission from the British Journal of Sports Medicine, BMJ Publishing Group.    Sex differences in growth are most apparent in adults. On average, adult men are taller compared with women in all linear body dimensions, height in particular [77]. Men have longer legs and trunks compared with women. Further, shoulder breadth is also wider and limb circumference larger in men compared with women; hip width does not differ between sexes (at least in absolute terms) [77]. Interestingly, sex differences are greater between men in women in some ethnicities (up to 8 cm, on average) compared with other ethnicities (as little as 4 cm, on average) [290]. For example, the largest difference in adult height was observed in white men and women (8 cm differences between men and women, on average) whereas differences in height between the sexes are less noticeable in black, Hispanic, and Asian men and women (4 to 6 cm between men and women, on average) [290]. I outlined the factors that contribute to the overall difference in height between men and women in Table 2.3. 71  Finally, boys and girls experience the same maturational events during growth but at differ times and with differ tempos (described previously, see Figure 2.8 and Figure 2.20). For example, in boys, genitalia stage 2 signals the onset of puberty at approximately 11 years of age, followed by PH development. In boys, APHV more consistently occurs during PH stage 4, along with axillary and facial hair development. Whereas in girls, breast stage 2 signals onset of puberty between ages 8 and 10 years of age, followed by PH development approximately 6 to 12 months thereafter. In girls, APHV typically occurs during PH stage 3 or 4, and menarche follows after APHV. Importantly, a boy and girl at PH stage 4 will have different degrees of somatic maturity, making sex comparisons by PH stages limited [291].  2.1.6.4 Hormones Growth hormone (GH) drives the initial phases of postnatal growth. GH is regulated by growth hormone-releasing hormone (GHRH; which causes synthesis and secretion) and somatostatin or somatotropin release-inhibiting hormone (SRIH; which inhibits GH release) (Figure 2.27) [292]. Somatotropic cells in the anterior pituitary synthesize and secrete GH in a pulsatile manner, in response to stimuli by the hypothalamus. This pulsatile pattern regulates growth in children and adolescents; the largest surges of GH are released during the first hours of deep sleep [116]. When GH binds to the GH receptor at the tissue level, this leads to a complex series of events in which the receptor induces various responses leading to cell differentiation. In turn, this results in bone or soft-tissue growth, fat metabolism, and/or insulin action.  72   Figure 2.27 A schematic representation of the production of growth hormone showing the regulatory influences, hormonal interactions, and effects; where: GHRH, growth hormone-releasing hormone; SRIH, somatotropin release-inhibiting hormone; IGF-1, insulin-like growth factor-1; +ve, amplifies response; -ve, inhibits response. Redrawn from Roche & Sun [142], with permission from Cambridge University Press.     Other hormones play crucial roles in human growth. Specifically, IGF-1 is released in response to GH binding, which leads to further tissue expansion such as in the skeletal growth plate [292]. In addition, thyroid hormones respond to increases in GH and IGF-1 and stimulate maturation of osteoblasts and ossification of bone tissue during growth (Section 2.2.4.4) [293]. Further, insulin is required for full expression of GH as it ensures that tissues receive the metabolic fuel required for growth. Insulin also influences IGF-1 levels, which stimulate bone and cartilage growth [142]. During adolescence, increasing concentrations of sex hormones (Figure 2.28) interact with GH and play an important role to initiate the growth spurt and its tempo [292]. First, there is a two- to three-fold increase in the amplitude of GH secretory bursts, which coincides with gains in height during puberty. IGF-1 also regulates linear growth as it acts to stimulate protein synthesis and increased cell proliferation. Serum levels of IGF-1 peak after APHV [294]. During puberty, activation of the HPG axis leads to production and 73  secretion of GnRH, which stimulates production and circulation of LH, FSH, and sex steroids (testosterone and estrogens) [142].   Testosterone increases IGF-1 production by enhancing the response to GHRH and increases the amplitude of GH release [292]. Testosterone levels are related to height gain and sexual maturity from childhood to APHV in boys [39]. In pre-puberty, serum testosterone concentrations are similar between the sexes (average 9 ng/dl). Sex differences are observed after the onset of puberty, where at APHV, serum testosterone concentration in boys is 900 ng/dl – a ten-fold increase, yet remain similar to pre-pubertal levels in girls [292]. Testosterone also acts to increase lean mass (or fat free mass, FFM) by enlarging muscles through increased protein synthesis, and to decrease fat mass [295].  In girls, estrogens comprise one of the main categories of sex hormones, the most potent of which is estradiol. During adolescence, GH and IGF-1 stimulate the ovaries to increase estradiol secretion. Estradiol stimulates chondrocyte and osteoblast proliferation, which in turn prompts epiphyseal closure [142]. In boys, the lack of estradiol results in later epiphyseal fusion, allowing boys to continue bone growth into young adulthood (Section 2.2.4.4). This is one mechanism that partially explains men’s greater heights compared with women, on average [296]. Estradiol also plays a crucial role in sexual maturation, promotes breast and pubic hair development, widens the pelvic area, and increases fat mass.   74   Figure 2.28 A schematic representation of the production of testosterone (left) and estradiol (right) showing the regulatory influences, hormonal interactions, and effects; where: GnRh, gonadotropin-releasing hormone; LH, luteinizing hormone; FSH, follicle-stimulating hormone; IGF-1, insulin-like growth factor-1; +ve, amplifies response; -ve, inhibits response; APHV, age at peak height velocity. Redrawn from Roche & Sun [142], with permission from Cambridge University Press.    2.1.6.5 Nutrition The influence of nutrition on growth and maturation is primarily observed when nutritional intake deviates from recommended levels (i.e., deficient or excessive).  Although ‘malnutrition’ generally implies suboptimal nutritional intake, malnutrition actually refers to any deviation from recommended levels (i.e., under-nutrition and over-nutrition [297]). Dietary intake is considered optimal when intake supports all metabolic and physiological functions (including growth and maturation). Energy is derived from chemical energy bound in food and its macronutrient elements. During growth, energy cost is related to the development of growing tissue. Thus, during periods of rapid growth, the energy cost of growth demands a significantly greater proportion of total growth requirements. Specifically, during the first year of life, 35 to 40% of nutrient intake is allocated to growth, which is then reduced to 5% at 12-months, 3% at 2-years, and 1 to 2% until pubertal growth when it reaches 50%. Energy costs of growth are negligible when linear growth ceases [298].  75  When nutritional intake is inadequate, stunted or diminished growth may result [35]. Stunting of growth that occurs before age 2 years may be irreversible [299-301], although catch-up growth may be possible if conditions improve and proper nutrition is initiated [268, 302-306]. For example, catch-up growth may occur when children are relocated from poorer geographic regions to more prosperous areas through migration or adoption [267, 268], when children were born prematurely or are small for gestational age due to IUGR [307], or when children are treated for a chronic condition that negatively impacts growth [308]. For example, catch-up growth was evident in children during post-war years when nutrition and living conditions improved [213, 222, 309]. Further, there may also be ‘growth adjustment factors’ to consider in situations where a malnourished child experiences reduced growth and to compensate grows for an extended period of time into their mid- or late-20s [310, 311]. However, it is still unclear whether individuals who experience catch-up growth achieve an adult height similar to those who did not experience nutritional deprivation. Small changes in diet appear to make large contributions to catch-up growth in malnourished infants, particularly in the first two years of life. For example, in a cohort of 245 boys and 215 girls from rural Guatemala, adding 100 kcal/day to daily dietary intake was associated with additional 9 mm of growth in the first year, 5 mm in year two, 4 mm in year three; no effects were noted after age three [312]. Authors speculated that adult height would not be compromised in children who ‘caught up’ to their well-nourished peers by age three, suggesting that diet in the first two years of growth is crucial  [312]. Undernourishment may also delay menarche, APHV, and skeletal maturation [21]. This is particularly apparent in children living in developing countries that lack nutritional resources [313]. However, undernourishment may also occur in athletes if energy intake is substantially less than what is needed to sustain an activity (e.g., elite gymnasts; Section 2.1.6.6). Interestingly, over-nourishment also affects maturational timing. Excess adiposity was associated with accelerated maturation in a 15-year longitudinal study of 177 overweight and obese (by CDC BMI cutoffs of 85th and 95th percentiles) boys and girls age 13 years at baseline; higher BMI was associated with earlier peak growth in overweight and obese boys and 76  girls [314]. There is clear evidence that menarche and early-maturity are interconnected. Overweight and obese girls experienced earlier menarche and breast development compared with normal weight girls [315, 316]. The causal direction and what drives the association between adiposity and maturation is unknown. Leptin (an adipocyte-derived hormone) may have an important role in the timing of puberty [317]. The ‘critical weight (fat) hypothesis’ suggested that puberty was initiated when fat stores (and thus circulating leptin levels) reached a specific level [318, 319]; though this hypothesis was criticized and later dispelled [320]. It has been suggested, particularly in girls, that the presence of endocrine factors during sexual maturation enforces the accumulation of body fat. [314]. Studies, predominantly in girls, suggest that obesity and early sexual maturation share a number of predisposing factors such as genetics, nutritional intake, SES, climate, general health, and physical activity [315, 316, 321].  2.1.6.6 Physical Activity  Generally, participation in recreational physical activity during childhood and adolescence does not influence pubertal growth or final adult height [20]. There were earlier concerns that sport participation negatively affected height (and body size in general); though now it is believed that a child’s natural body size influences their choice and success in sport [322, 323]. Thus, continued success in elite sports may be related to an individual’s likelihood (i.e., genetic potential) to be taller or shorter, heavier or lighter, early- or late-maturing, rather than sports participation causing deficient growth or late maturation. Some adolescent athletes are more mature, taller, and heavier than their non-active counterparts [324]. For instance, basketball players, ice hockey players, and swimmers of both sexes are taller and heavier compared with their non- and minimally-active peers [324]. As size and maturational status influences athletic performance, many coaches use growth and maturity indicators as a means of talent identification in elite sport [325, 326]. Gymnasts and figure skaters of both sexes, and female ballet dancers, tend to be shorter and lighter compared with non-athletes of the same age and sex [107, 327]. In some cases, lack of growth and delayed 77  puberty may be due to substantial nutritional deficits in these athletes, as previously discussed (Section 2.1.6.5). For example, female elite gymnasts aged 12 to 14 years had inadequate caloric intake for their activity level, and failed to meet estimated average daily requirements for important micronutrients (e.g., calcium, vitamin D) [328, 329]. These gymnasts experienced delayed skeletal maturity (1.7 years, on average); this delay was attributed to energy availability imbalance and high training frequencies [328, 329]. However, catch-up growth was observed in some gymnasts when training volume was reduced or the athlete retired in late puberty, when energy intake is more balanced [330, 331].    2.1.7 Summary of Growth and Maturation Literature  Assessing maturity remains important as significant physiological changes occur at adolescence. However, there is wide variation in the timing of normal puberty, with onset varying by as much as four or five years among healthy children. Thus, children of the same chronological age can differ considerably in their degree of biological maturation – resulting in potentially inaccurate conclusions in studies that only align children by chronological age [3]. APHV is a commonly used indicator of somatic maturity that permits investigators to align children on biological age (i.e., MO). In the event that sufficient serial height measures are not available, prediction equations may be used to predict MO and APHV. The use of the Mirwald equations [26] is frequently reported in the literature, although some have questioned accuracy of these equations [27-29]. My dissertation is a response to the need for accurate measures of maturity, operationalized by my review of methods used to derive the Mirwald equations, validate use of the equations in other populations, and assess whether factors such as ethnicity may restrict their use. Should there be a need to redevelop the Mirwald equations, this might serve to increase their usability in pediatric studies, including studies of children of different ethnicities, in future.    78  2.2 Influence of Growth and Maturation on Bone Accrual  Bone is a complex and dynamic tissue [332]. Linear growth in height and sexual maturation are inextricability linked to bone accrual [333]. In the last research chapter of my dissertation (Chapter 7), I explore the relationship between maturational timing and young adult bone mass, density, structure, and strength. Therefore, in this section of my literature review I provide an overview of bone biology, growth, and assessment, as well as factors (with a focus on maturational timing) that influence bone development from adolescence into young adulthood. As my dissertation primarily focuses on maturation, I will not review bone biomechanics, though will describe some factors (such as physical activity) that demonstrate bone’s response to mechanical stimuli during the growing years.   2.2.1 Why Assess Bone During the Growing Years?  Characterized by low bone mass and deterioration of bone structure and strength [334], osteoporosis is a major cause of morbidity and mortality in older adults [335, 336]. In Canada, approximately 6.6% of men and 15.8% of women aged 50 years and older have osteoporosis [337]. Despite osteoporosis being characterized as a condition of aging, the seeds of this condition are sown during the growing years [338, 339]. During growth, bone is laid down to increase its mass in response to increasing length [340, 341]. Adolescence represents a ‘critical period’ of rapid growth where more than one-quarter of total adult bone mass is accrued (Figure 2.29) [342]. After the adolescent growth spurt, bone mass peaks and slowly declines with age [343]. Conceivably, children, adolescents, and young adults with low bone mass, density, structure, and strength may be at higher risk for osteoporosis and related fractures in adulthood [344]. Therefore, it is critical to assess factors that can optimize bone accrual during these years [345, 346].  79   Figure 2.29 Illustration of the percentage of total body bone mineral content (BMC) gained during adolescent growth in boys (left) and girls (right) from the Pediatric Bone Mineral Accrual Study. Approximately 40% of BMC is accrued in the ± two years from age at peak height velocity (APHV) in both sexes. Adapted from Baxter-Jones et al. [342], with permission from the Journal of Bone and Mineral Research, John Wiley & Sons, Inc.    2.2.2 Bone Tissue Anatomy and Physiology  Bone is a highly specialized and dynamic tissue [347]. Functions of the skeletal system include structural support, storage of minerals and lipids, blood cell production, protection of organs, and leverage and movement [347]. The human skeleton begins to develop in utero. At birth, the human skeleton has nearly 300 bones and cartilage elements. These later fuse to form a compliment of 206 bones that vary in shape and size (e.g., long, flat, sutural, irregular, short, and sesamoid) [348] in the adult skeleton. In the following sections, I describe bone composition and organization, as well as bone modeling and remodeling.  80  2.2.2.1 Bone Matrix and Cells  Bone is continually subject to forces such as compression, tension, bending, and torsion [349]. Briefly, normal stresses act perpendicular to a given plane and the bone experiences either compacting/shortening (compression) or elongation/pulling (tension), whereas shear stresses act parallel to the plane and the bone experiences bending or twisting (torsion). As a result, bone adapts it’s mass, structure, and strength to withstand these forces and mitigate fracture risk.   Calcium phosphate, Ca3(PO4)2 contributes approximately two-thirds of the weight of bone. Calcium phosphate interacts with calcium hydroxide, Ca(OH)2, to form crystals of hydroxyapatite, Ca10(PO4)6(OH)2 [332]. As they form, these crystals incorporate other calcium salts, such as calcium carbonate (CaC03), and ions such as sodium, magnesium, and fluoride. These crystals are relatively brittle and inflexible, though handle compressive forces well [332]. Collagen fibers contribute the remaining one-third of the bone’s weight [350]. These, by contrast, are flexible, yet strong, and tolerate bending and torsional forces well [350]. Osteoblasts, osteoclasts, and osteocytes are bone cells that comprise bone tissue [347]. The combination of the rigid hydroxyapatite along with the pliable collagen gives bone its unique viscoelastic properties that enable it to withstand compressive and tensile stresses and bending and torsional movements [347].   Bone cells are classified by origin and function. Osteoblasts release matrix vesicles that bud off to form needlelike hydroxyapatite crystals; this release stimulates further hydroxyapatite formation and mineralization of the matrix [348]. In contrast, osteoclasts resorb bone matrix by releasing hydrogen ions from their ruffled border. This produces an acidic environment within the bone that decalcifies the matrix, and release enzymes that digest protein components of the matrix. Finally, osteocytes are relatively inactive cells that form when osteoblasts become entrapped in bone matrix [348]. Osteocytes are mechanosensor cells and stimulate remodeling by transmitting deformation to osteoclasts and osteoblasts [351]. Bone remodeling is the process of replacing old bone matrix with new bone matrix (Section 2.2.2.3). 81   Bone tissue is either woven (immature) or lamellar (mature); these tissues differ by organization of their collagen fibers within the bone matrix [347]. Woven bone is formed during fetal development and comprises all bone at birth (replaced by age 4 years, on average). It is also found in ligament and tendon insertions in healthy adult skeletons, or at sites of fracture repair. Woven bone fibers are oriented randomly and in different directions. Osteoclasts resorb the woven bone and osteoblasts build new matrix. Woven bone is remodeled to form lamellar bone [352], which is organized into thin layers (3-7µm) called lamellae. Lamellar bone provides the structural basis for spongy (trabecular) and compact (cortical) bone [353], structural subunits, lamellae. Osteocytes lay within the lacunae between the lamellae. Lamellae are oriented parallel to trabeculae in trabecular bone and arranged in osteons in cortical bone. Osteons are the structural units of cortical bone and are cylindrical arrangements of cortical bone around a Haversian system (Figure 2.30) [354].   Figure 2.30 Illustration of the structural elements of a long bone, with cortical bone (outer) comprised of osteons surrounding trabecular bone (inner). Reproduced from Seeley et al. [354], with permission from McGraw-Hill.     Bone with a volume fraction of solids less than 70% is classified as spongy (trabecular) bone, whereas over 70% is classified as compact (cortical) bone [355]. Trabecular bone consists of interconnecting rods 82  or plates of bone called trabeculae. Most trabeculae are thin (50-400µm) and consist of several lamellae with osteocytes located in lacunae between lamellae; these trabeculae are primarily oriented along the bone lines of stress (Figure 2.31) [356]. Trabecular bone is largely responsible for bone’s energy absorbing capacity, though its mechanical properties differ from site to site as it is highly heterogeneous [357]. Cortical bone is much less porous (e.g., 5 to 10% porosity) compared with trabecular bone (e.g., 50 to 90% porosity) [358]. Cortical bone is stronger and stiffer longitudinally moreso than transversally due to its parallel orientation [359]. Cortical bone houses vessels within central canals that run parallel to bone’s long axis. Concentric lamellae are circular layers of bone matrix with a common central canal. The osteon is the functional unit of cortical bone and consists of a single central canal and its concentric lamellae and osteocytes [352]. In a long bone, the diaphysis is primarily comprised of cortical bone and cortical bones serves as a hard outer shell to protect the metaphysis. Trabecular bone is found primarily in the epiphyses and metaphyses of long bones and within other bones such as (irregularly shaped) vertebrae  [352].    Figure 2.31 Illustration of the proximal femur showing trabeculae oriented along lines of stress. The alignment of trabeculae in this manner resists bending and torsional movements. The trabecular bone is surrounded by cortical bone (outer shell). Reproduced from Seeley et al. [354], with permission from McGraw-Hill.   83  2.2.2.2 Long Bone Geometry   Briefly, long bones are comprised of the diaphysis, epiphysis, and metaphysis (Figure 2.32) [360]. The diaphysis is the middle portion of the bone and consists of a cylindrical shaft, mostly comprised of cortical bone. The metaphysis is a transitional region between the diaphysis and epiphysis, comprised of cortical and trabecular bone [360]. The epiphyseal plate (or growth plate), a hyaline cartilage disk, separates the metaphysis from the diaphysis, and plays an important role in lengthening the long bone during growth (Section 2.2.2.4) [361]. The epiphyses are at long bone ends and are comprised primarily of trabecular bone. The wider shape at the epiphysis serves to distribute and reduce compressive forces transmitted to the diaphysis [362]. The bone’s outer shell (periosteum or periosteal surface) and inner shell (endosteum or endocortical surface) are both active sites of bone modeling and remodeling [363].     Figure 2.32 Illustration of the aspects of human long bone (femur) including the diaphysis, epiphysis, and metaphysis. Reproduced from Kontulainen et al. [360], with permission from Karger.   84  2.2.2.3 Bone Modeling and Remodeling   Bone is a dynamic tissue that undergoes significant turnover to adapt its size, shape, and distribution for the purpose of optimizing the skeleton to withstand forces [364]. Modeling occurs primarily during growth and in response to physiological and mechanical forces [365]. During modeling, the bone widens and lengthens. More specifically, the diaphysis expands with bone being laid down on the outer periosteal surface. Simultaneously, osteoclasts resorb bone from the inner endocortical surface. Skeletal growth in length occurs at the growth plates in the epiphyseal and metaphysical regions, where cartilage proliferates (Section 2.2.2.4). This process increases long bone length (i.e., linear height) until full skeletal maturity [347]. The metaphysis is modeled through a process called metaphyseal inwasting [348, 366]. During rapid growth (such as APHV), long bones are lengthened and the bone is reshaped to increase linear height [360]. At times of rapid growth, there appears to be transient cortical weakness (due to metaphyseal inwasting) that can lead to an increase in fracture risk [367-369]. For example, some studies demonstrate a decrease in bone density and strength (particularly at the distal radius) during early puberty, where bone at the metaphysis requires approximately five to seven months to ‘catch up’ with the increase in bone length [370-373].  In contrast, bone remodeling involves a basic multicellular unit (BMU), in which osteoclasts and osteoblasts work together to replace old bone matrix with new bone matrix [348]. The BMU lifespan is approximately 6 months and the skeleton is replaced approximately every 10 years. [348] Osteoclasts of a BMU resorb pockets of old or damaged bone and form tunnels within the matrix. Osteoblasts of a BMU then lay down new bone along the tunnel wall, forming a concentric lamella [347]. Remodeling continues throughout the lifespan to repair and replace damaged bone, and in response to mechanical loads placed on the body (e.g., physical activity); where (similar to above) one full cycle takes five to seven months to complete [374]. Both modeling and remodeling are critical for developing the growing skeleton’s strength [375]. In the next section I describe endochondral ossification, the process responsible for formation of most of the axial and appendicular skeleton during growth.   85  2.2.2.4 Endochondral Ossification  Most bones develop through an essential process called endochondral ossification  (Figure 2.33) where cartilaginous tissue is replaced by bone and long bones are lengthened [376]. The cellular events that comprise endochondral ossification include chondrocyte proliferation, chondrocyte hypertrophy, matrix mineralization, apoptosis, vascular invasion, ossification, and remodeling of lamellar bone. Endochondral ossification begins with primary ossification centres that appear mostly during fetal development. This is followed by development of secondary ossification centres at the epiphyses [377].  Shortly after formation of the primary ossification centre, an organized region of rapid growth develops between the epiphysis and diaphysis (i.e., the growth plate), which is primarily responsible for the increase in length of the developing diaphysis [378, 379]. Distinct growth zones of the epiphysis are defined by each zone’s unique histological appearance [379].   Figure 2.33 Illustration that depicts stages of endochondral ossification from formation of the bone collar (stage 1) to fully ossified bone in young adulthood (stage 5). Reproduced from Seeley et al. [354], with permission from McGraw-Hill.  86  2.2.2.5 Growth Zones of the Epiphysis  Figure 2.34 illustrates growth zones of the epiphysis. Of note, though lengthening of bone occurs both at the proximal and distal ends, growth does not necessarily occur proportionately. For example, approximately 90% of the increase in radius length occurs at the distal end [380], whereas only approximately 30% of the increase in tibia length occurs at the distal end [380, 381]. Although the tempo of growth differs between the proximal and distal growth plates, the structure is similar. I describe these growth zones below.  First, the germinal zone (also known as resting or reserve zone) is furthest away from the diaphysis and closest to the metaphysis. In this area, chondrocytes are small, quiescent and randomly distributed. These cells receive vascular supply from epiphyseal vessels that penetrate the terminate plate [361]. Second, in the proliferating zone, adjacent to the germinal zone, chondrocytes increase in size as they accumulate glycogen. These cells exhibit mitotic division and become arranged in longitudinal pillars. These pillars make up approximately half the height of the growth plate and are the site of new cartilage formation that will ensure the continued longitudinal expansion of the diaphysis. Following mitosis the daughter cells lie side by side. The cells migrate towards the metaphysis and align longitudinally [361]. Third, in the zone of cartilage transformation (also known as hypertrophic or calcification zone), cell division ceases and chondrocytes start to hypertrophy in preparation for their replacement by bone. There is progressive degeneration of the cellular component, with crystalline hydroxyapatite being deposited primarily. The metaphyseal arteries form loops and advance into this zone [361]. Fourth, in the zone of ossification (also known as metaphysis) osteoblasts form a layer of bone on the remnants of the cartilage from the preceding zone. Here, bone is reorganized and subsequently, new bone called spongiosa, or immature bone, is deposited [137, 361, 379]. These zones combined are responsible for lengthening long bones. Conversely, the zone of Ranvier is responsible for bone widening [382]. Transverse expansion occurs by cell division and matrix expansion within the growth plate and by the cellular addition from the periphery at the zone of Ranvier [361]. 87    Figure 2.34 Illustration of the major features in the epiphysis. Reproduced from Scheuer & Black [377], with permission from Elsevier.     2.2.2.6 Epiphysis Closure  As skeletal maturity approaches, bone growth diminishes as growth plate chondrocytes senesce. This decreased growth rate is associated with structural changes in the epiphysis, including a gradual decline in thinning of the growth plate. Fusion occurs when rate of chondrocyte proliferation approaches zero. Fusion is triggered when estrogen accelerates programmed senescence of growth plate [383]. The growth plate is completely resorbed following puberty, resulting in epiphyseal fusion in young adulthood [19, 379]. Timing of epiphyseal fusion varies by skeletal site, from proximal to distal epiphysis, and across individuals [87]; I previously described differential timing of growth in Sections 2.1.2.6 and 2.2.2.5.   2.2.2.7 Peak Bone Mass  The maximal amount of bone gained by the end of skeletal maturation is called peak bone mass (PBM) [384]. Some consider PBM to more broadly capture peak bone strength, which is characterized by mass, 88  density, and structural properties that provide bone its structural strength [385-387]. Like growth and maturation, timing of PBM varies by site, bone compartment, and across individuals. Peak bone mass accrual is influenced by several non-modifiable (e.g., genetics, ethnicity, sex, maturity) and modifiable (e.g., nutrition, physical activity) factors. In later life, PBM is considered a determinant of osteoporosis and osteoporosis-related fracture risk [385], where higher PBM theoretically lowers risk of fracture in later life (Figure 2.35).    Figure 2.35 Illustration of bone mass accrual across the lifespan. The dot-dash line represents when peak bone mass (PBM) is compromised during adolescence due to suboptimal lifestyle factors, whereas the solid and dashed lines represent the theoretical increase in PBM with optimal lifestyle factors. The dark shaded area represents a period where bone mass may be low, and the light shaded area where bone mass may be considered osteoporotic and at risk for osteoporosis-related fracture. Reproduced from Weaver et al. [387], with permission from Osteoporosis International, Springer-Verlag London Ltd.     PBM occurs after APHV and cessation of linear growth. For example, in the PMBAS, 228 healthy boys and girls were assessed annually for 6 years (ages 8 to 15 years at baseline) [145]. Peak bone mineral content velocity (PBMCV) occurred in girls at 12.5 years of age compared with boys at 14.1 years of age, 89  with an interval of 0.7 years between APHV and PBMCV (Figure 2.36). At four years post-APHV, 95% of adult bone mass was achieved in both sexes [85].     Figure 2.36 Illustration of the relationship between age at peak height velocity (APHV) and peak bone mineral content velocity by chronological age for boys and girls in the Pediatric Bone Mineral Accrual Study. Reproduced from Bailey et al [85], with permission from John Wiley & Sons, Inc.    2.2.3 Assessing Bone in Children and Adolescents  A number of imaging tools are used to assess bone mass, density, structure, and strength accrual, each with their unique advantages and disadvantages [388]. Below, I briefly describe DXA and pQCT, two of the most widely used bone imaging tools.   2.2.3.1 Dual Energy X-ray Absorptiometry  Introduced nearly 30 years ago, DXA is the most commonly used imaging technique to assess bone (total body (TB), lumbar spine (LS), proximal femur (PF), femoral neck (FN)) in research and clinical practice [389]. DXA is a preferred method based on its rapid scan time (3-7 minutes per scan), low radiation 90  (effective dose of approximately 1.4-13 µSv depending on scan site, equivalent to 1/10 of a chest X-ray), high precision (CV<2% at most sites), and wide availability. DXA attenuates ionizing radiation from the X-ray source to a detector. Beam attenuation is greater in mineralized tissue compared with soft tissues and bone is denser compared with soft tissues due to heavier calcium and phosphate composition [390]. There are pediatric normative DXA data available for clinicians and researchers at all measured sites. DXA assumes the body is comprised of two components: soft tissue and bone. DXA measures whole body or segmental aBMD (g/cm2) and BMC (g), which is calculated by multiplying aBMD by the projected bone area (BA, cm2). DXA further measures whole body lean mass (g) and fat mass (g).   Despite widespread use and good precision, DXA has some limitations. For example, DXA cannot discern length along the X-ray beam path due to its 2-dimensional (2D) projection. As a result, this areal rather than volumetric measure may generate a size bias (i.e., aBMD lacks a depth dimension, which differentiates it from true, 3-dimensional (3D) volumetric BMD (BMD, g/cm3)). For example, aBMD was systematically underestimated in shorter people [391]. Corrections to resolve this challenge include cubic, cylindrical, or other structural approximations to account for missing dimensions and shapes [392]; however, corrections may be further confounded by changes in body size. In children, assumptions of homogeneity are particularly problematic, as children grow and mature at different rates between measurement periods. Thus, changes in aBMD by DXA in longitudinal studies may reflect growth-related body size or composition changes rather than true changes in bone mineral [393]. Strategies to correct for these differences include adjusting for bone area, height for age, body weight, or muscle mass. DXA is also limited by its inability to separate cortical and trabecular bone compartments, and cannot formally assess bone strength [394]. Thus, DXA cannot ascertain the independent contributions of cortical and trabecular bone components to bone strength. However, aBMD by DXA is a strong independent predictor of bone failure [395]. DXA remains the most commonly used method to assess bone in children, youth, and young adults, although more advanced technologies have emerged. 91   Derived from DXA, hip structural analysis (HSA) was developed to counter DXA’s 2D limitation. This technique estimates bone structure and strength (e.g., narrow neck (NN), intertrochanteric (IT), femoral shaft (FS) cross-sectional area (CSA, cm2) and section modulus (Z, cm3)) from bone mineral data in the image plane [396]. This technique also has its limits. For example, measurements are attained based on the assumption that the mineral density is constant. As children’s bones are less mineralized compared with adults, estimates of cross-sectional structure can be can be underestimated [390].   2.2.3.2 Peripheral Quantitative Computed Tomography  Peripheral QCT (pQCT) provides volumetric measurements of the appendicular skeleton for a given region of interest (ROI), such as the distal or midshaft tibia, or the distal or midshaft radius. Unlike DXA, pQCT has the ability to visualize the volumetric morphology of bone and discriminate between cortical and trabecular bone [397, 398]. Further, pQCT has relatively low radiation (one pQCT scan is <1.5 µSv), quick scan time (3-minute), and high precision (CV<2%). There are normative reference values for children and young adults for cortical and trabecular BMD (Ct.BMD, mg/mm3, Tb.BMD (mg/mm3), CSA (mm2), and cortical thickness (Ct.Th, mm) [399-401]. Similar to DXA, pQCT attenuates ionizing radiation from the X-ray source to a detector. However, unlike DXA, pQCT employs a rotation mechanism that conducts a transverse scan following successive partial displacements every 12° for a total of 180° or 15 rotations. Multiple absorption profiles are mathematically combined to produce an image through filtered back projection [402]. Using defined analysis modes and thresholds (see Section 4.2.4.2), this technology provides measures of total area (Tt.Ar, mm2), cortical area (Ct.Ar, mm2), cortical thickness (Ct.Th, mm), trabecular area (Tb.Ar, mm2), medullary area (Me.Ar, mm2), as well as cortical volumetric BMD (Ct.BMD, mg/mm3), and estimates of bone strength: polar strength-strain index (SSIp, mm3) [403], amongst others. Moreover, pQCT can simultaneously quantify muscle cross-sectional area (MSCA, cm2), important when assessing the bone-muscle unit (Section 2.2.4.6). Common pQCT instruments are the Stratec XCT 2000 (pediatric scanner) and the XCT 3000 (adult scanner). We [404] and others [405] reported strong 92  associations between the machines, allowing for comparison of results (e.g., correlations ranged from 0.90 to 0.99) for across all sites and measures.   Peripheral QCT is not without its limitations. First, pQCT offers choices in scan acquisition parameters, such as scan time, resolution, placement and scan site. This is an advantage in that a specific protocol can be determined for each study and region of interest but also a disadvantage, in that there are no established guidelines for acquisition or analysis in pediatric studies [398]. Second, lower resolution (larger pixel size; 0.4 to 0.6 mm) increases the likelihood of partial volume effects (PVE) (Figure 2.37) [406]. PVE refers to the tissues of varying densities (e.g., bone versus soft-tissue) within the same voxel, which could underestimate BMD. Smaller bones (e.g., radius, Ct.Th in particular) are more likely to be biased by PVE. Thus, acquisition protocols should consider how to minimize PVE. Third, long-term precision of pQCT can be limited by the technologist’s ability to position the scan start line at the same location [407]. This is especially challenging in pediatric studies as the ROI is a moving target in growing long bones. Thus, it is not always possible to compare across studies, as some studies use a fixed ROI where others use a relative (%) ROI. In addition, pQCT is unable to assess clinically relevant sites, such as the proximal femur or lumbar spine [408]. However, pQCT measures at the tibia demonstrated relatively good agreement with DXA measurements at central sites [409]. Additionally, pQCT can assess the clinically relevant distal radius [406].   93   Figure 2.37 Illustration of the partial volume effect (PVE), whereby voxels at bone edges (dark grey) contain both soft and bone tissue densities, resulting in a lower density for the dark grey voxels. Smaller bones (e.g., radius) have more voxels close to the bone edge and are at greater risk for PVE. Reprinted from Zemel et al. [406], with permission from Elsevier.     With recent advances in pQCT technology we are now able to attain high-resolution (HR-pQCT) images, quantify bone microstructure (e.g., trabecular thickness, Tb.Th (µm) and number, Tb.N (mm-1), cortical porosity, Co.Po (%), amongst others), and estimate bone strength with micro-finite element analysis (µFEA). The system acquires 110 parallel CT slices at the distal radius or distal tibia. These slices are stacked to form a 3D image. In imaging resolution ranges from 82 µm (first generation scanner) to 61 µm (second generation scanner). Although I do not report HR-pQCT result in my dissertation, I agree that assessing bone’s microstructure enhances our understanding of the microstructural elements that impact bone strength and fracture risk [410, 411].  2.2.4 Factors that Influence Bone in Childhood and Adolescence   In this section, I describe factors that influence bone accrual. In Section 2.2.4.7, I describe the relationship between maturity and bone accrual, which is a central focus of Chapter 7 in this dissertation.  94  2.2.4.1 Genetics  With mapping of the human genome, more than 70 loci associated with bone metabolism (e.g., those which encode bone-specific hormone receptors), bone density, or fracture were identified [412, 413], some of which contribute to development of PBM [414, 415]. Heritability explains most of the variation (i.e., up to 90%) in bone mass, structure, and strength [416-424]. Heritability of aBMD is greatest in twins (80%, MZ and DZ), and slightly less in families (60%, e.g., siblings, parents); though, most of the variation in aBMD is due to differences in body size (e.g., height), as I discussed previously (Section 2.1.6.1). Heritability of aBMD in the axial skeleton (e.g., spine and hip) is often higher than in the appendicular skeleton (e.g., forearm, lower leg) [415]. The mechanisms for site-specific genetic influence on aBMD are unknown, though the extremities are often subject to mechanical loads and are influenced to a greater extent by lifestyle factors [425].   Fewer studies assessed heritability in bone geometry and strength [426-429]; variance accounted for by genetic factors ranges from 19 to 91% [424, 430]. Heritability was lowest in trabecular bone (e.g., radius trabecular number, 19%) and highest in cortical bone (e.g., cortical thickness, 91%) [430]. A recent study used µFEA to assess the radius and tibia of men and women in the Framingham Heart Study (mean age 72 years, 57% women). Heritability estimates for bone strength were 52% at the distal radius and 49% at the distal tibia, as measured by HR-pQCT [429].   2.2.4.2 Ethnicity  Ethnic-related differences in bone mass, structure, and strength are irrefutable, though much of the variation can be explained by genetics, sex, and body size. Ethnicity is also highly interrelated with culture, lifestyle, and SES, making associations with bone mass, structure, and strength difficult to interpret [431].   There is evidence to suggest that fracture rates are lower in Asian children compared with white children [432]; however, it is unclear if this difference is due to bone-related factors, or if Asian children have reduced risk for injury for other reasons (e.g., less risky play, less contact sports participation, etc.). 95  In our HBSIII cohort, we found ethnic-differences in bone density and structure in childhood, where Asian boys and girls had smaller Ct.Ar, but greater Ct.Th, greater Ct.BMD, and less Co.Po compared with their white peers [433]. These data suggest that despite smaller bone structure in Asian boys and girls, bone adapts to maintain its strength. In another study, black children had greater bone strength at the tibia and midshaft tibia compared with their white peers [434]. Additionally, black adolescent girls demonstrated greater Tt.Ar, Ct.BMD, and SSIp at the midshaft tibia compared with white adolescent girls [435].   Ethnic differences in bone were shown to continue into adulthood. Asian men and women (living in Asia) were shorter and had smaller bone area compared with their white counterparts (living in North American and Europe) [436]. Chinese-American women had lower total body, total hip, femoral neck, and lumbar spine aBMD compared with white women [437, 438], where (like linear height) older age at immigration negatively impacted aBMD [437]. When Asians were taller and heavier, aBMD was also higher [431]. Given that much of the ethnic-related variation in bone is related to body (and bone) size in Asians, other parameters besides aBMD should be considered [439]. Using HSA-derived strength indices, Ishii et al. [440] found that compressive and bending strength were greater at the FN for black and Asian women, compared with white women. In another study, Asians had less trabecular spacing, greater Tb.Th,  greater Ct.Th and Ct.BMD compared with their white counterparts [441]. Asian women had greater Ct.Th, less Co.Po and greater Tb.Th (but less Tb.N) compared with white women [442, 443]. Asians also had greater estimates of bone strength by µFEA [441]. Asian women also had the highest unadjusted risk for osteoporosis, though risk was similar when weight was accounted for, and interestingly, fracture risk was considerably lower [444]. Further, a large multi-ethnic study, the National Osteoporosis Risk Assessment (NORA), demonstrated that black men and women had the lowest risk of osteoporosis and related fracture, whereas Hispanic and Aboriginal men and women had risks that were not significantly different from white men and women [445]. Similarly, data from NHANES demonstrated that black women and men from the US had greater femoral neck and lumbar spine aBMD compared with whites and Hispanic (Mexican) Americans at all ages [446-448].  96  2.2.4.3 Sex  In adults, lower fracture prevalence in men compared with women has been explained by men achieving higher PBM during growth, and women producing lower amounts of sex hormones at mid-life, resulting in a negative BMU balance (bone resorption > bone formation) [449].   A multi-ethnic study used DXA to assess total body BMC and aBMD in 64 children at birth. Boys had greater BMC, whereas aBMD did not differ between boys and girls [450]. BMC was no longer significant between the sexes after adjusting for body size. In a study of 428 five years old boys and girls (primarily white), boys had greater total body and proximal femur BMC and aBMD than did girls, even after adjusting for body size (i.e., height and weight). There were no sex-differences at the lumbar spine [451]. Bone mass continued to increase during growth in both sexes during mid-childhood [452]; sex differences in BMC and aBMD became more pronounced at puberty due to rapid linear growth [453, 454]. During pubertal growth, boys have a 2-year longer accelerated growth period compared with girls, accounting for greater differences in BMC and aBMD. Post-pubertal boys have longer bones of greater diameter than post-pubertal girls, on average. Thus, aBMD is greater in post pubertal boys compared with girls, on average [452]  There are also sex differences in bone structure and strength, but these are more complex as the results are more varied by bone site and compartment. Despite larger bone size in post-pubertal boys and men, Ct.Th by pQCT (midshaft tibia) did not differ substantially between sexes [455, 456]. In a cross-sectional study of 665 children and adults (ages 5 to 35 years), Ct.BMD, periosteal circumference, and Z were higher at the midshaft tibia in boys compared with girls at all ages and all stages as per the methods of Tanner, though Ct.BMD was lower and endosteal circumference was higher in peri- and post-pubertal boys compared with girls [434]. After puberty and into young adulthood, boys had greater Tb.Th and less Co.Po (at the distal radius) compared with girls into young adulthood [434, 457].  In our HBSIII cohort, we observed lower Ct.BMD and higher Co.Po at the distal tibia and radius by HR-pQCT in boys compared with girls at the radius during and after puberty [368]. Further, we recently reassessed sex differences in endocortical and periosteal apposition in children during puberty [458]. We 97  challenged Garn’s cross-sectional radiographic studies of the second metacarpal where he concluded that boys experience more periosteal apposition while girls experience more endocortical apposition during adolescence [459-461]. HBSIII participants were aligned on APHV. Early puberty, boys had higher annual accrual rates in Tt.Ar and Me.Ar, and lower Ct.Ar/Tt.Ar and Ct.BMD compared with girls. At APHV, boys had greater Tt.Ar, Me.Ar, and SSIp, and lower Ct.Ar/Tt.Ar and Ct.BMD compared with girls. Post-APHV, boys had greater annual accrual rates for all bone parameters, except Ct.Ar/Tt.Ar, which was similar between sexes (Figure 2.38). Thus, there was accelerated periosteal apposition during adolescence, moreso in boys than girls, and there was diminished endocortical resorption, moreso in girls than boys.    Figure 2.38 Representation of the differences in total area (Tt.Ar), medullary area (Me.Ar), cortical area (Ct.Ar), cortical BMD (Ct.BMD), and strength-strain index (SSIp) between boys and girls by maturity offset (MO) in children from the Healthy Bones Study III. Reproduced from Gabel et al. [458] , with permission from the Journal of Bone and Mineral Research, John Wiley & Sons, Inc.     98  2.2.4.4 Hormones  I illustrated the relationship between hormone secretion (i.e., IGF-1, GH, estrogen, testosterone) and BMC accrual in Figure 2.26. GH peaks at APHV [462] and IGF-1 slightly thereafter [463]. The interaction between the GH-IGF-1 axis and sex steroids plays a critical role in regulating bone metabolism. Together, GH and IGF-1 stimulate osteoblasts during modeling and remodeling [462, 463]. During puberty pulsatile secretion of GH and IGF-1 increase 1.5 to 3-fold. Sexual dimorphism in GH secretion during puberty parallels the change in growth velocity [464]. Deficiency in GH during growth results in lower bone mass and strength and short stature [465, 466]. Though GH therapy during growth is controversial, it may benefit bone mineralization [467, 468] and linear growth, specifically in those with clinical GH deficiency [469, 470].  Sex steroids (e.g., testosterone and estrogen) regulate sex differences in bone mass accrual throughout growth [452, 471]. When sex hormones are low early puberty, the GH-IGF-1 axis actively stimulates linear growth and bone development. In late puberty, GH and IGF-1 concentrations are lower and sex steroid concentrations are higher, stimulating skeletal maturation and growth plate closure [472]. In both sexes, estrogen induces pubertal growth, promotes osteoblastogenesis, inhibits osteoblast apoptosis, decreases and osteoclast production [119, 473]. Testosterone also stimulates pubertal growth in both sexes through chondrocyte proliferation in the epiphyses. However, in boys, there is a 10- to 20-fold increase in testosterone concentrations during puberty; this stimulates skeletal muscle development (i.e., increased muscle loading on skeleton) and enhances periosteal geometry [474, 475], though testosterone alone is not sufficient to drive periosteal expansion [476].     Further, bone remodeling is influenced by calcitropic hormones (e.g., calcitonin, parathyroid hormone)  [477]. Calcitonin and parathyroid hormone are peptide hormones that modulates of calcium homeostasis through their actions on osteoclasts and osteoblasts, respectively [478]. Calcitonin reduces serum calcium by inhibiting bone resorption and enhances calcium excretion by the kidney [479], whereas parathyroid hormone increases serum calcium by stimulating bone resorption and increasing calcium absorption from 99  the kidneys [480]. Thus, calcitonin and parathyroid hormone work synergistically to regulate bone remodeling [478].  2.2.4.5 Nutrition   Adequate calcium and vitamin D are necessary for skeletal health [481]. Canadian dieticians currently recommend 1000 to 1300 mg (more during adolescent growth) of calcium and 600 IU vitamin D daily during adolescent growth and in young adulthood  [482].   Calcium is the most abundant mineral in the body (1000 to 1500 g in the adult skeleton); 99% of calcium is found in bone as hydroxyapatite calcium [483]. Bone stores and releases calcium for metabolic requirements through the process of bone remodeling [481]. Dietary calcium requirements vary throughout the lifespan (200 mg at birth to 1300 mg during growth pubertal growth) [482]. Though skeletal saturation (i.e., threshold) was determined to be 1500 mg/day in adolescents, with excess calcium excreted [484].   Ingested calcium comes from food sources (e.g., milk, cheese) and dietary supplements (e.g., calcium carbonate, calcium citrate). Calcium is absorbed across the intestinal mucosa and is dependent on the actions of calcitriol and the intestinal vitamin D receptor [481]. During growth, adequate food sources containing calcium [485] and calcium supplementation [486] tend to enhance bone mineral in children and adolescents, though only slightly [387]. Rizzoli et al. [487] summarized the results of 16 calcium supplementation RCTs in children ages 7 to 17 years. Calcium supplements were given from 10 to 48 months and the dose ranged from 800 mg to 1200 mg daily. The difference in BMC or aBMD between calcium supplement and placebo groups ranged from 0.7 to 5.7% depending on skeletal site. Gains in BMC and aBMD were greater in the appendicular skeleton compared with the LS [487]. Further, two meta-analyses of 21 RCTs [488] and 19 RCTs [489] in children ages 3 to 18 years demonstrated that increased dietary calcium (with or without vitamin D) and calcium supplementation (ranging from 300 to 1200 mg daily) significantly increased TB and LS BMC, particularly in children with low baseline intake (e.g., less than 400 mg daily) [488, 489]. It is unclear how long benefits persist after stopping supplementation [488]. 100   No studies have examined the association between dietary calcium or calcium supplementation and bone structure or strength independent of physical activity. However, it appears that supplementation does not have the same benefits in children who participate in high-impact physical activity [490]. For example, gymnasts and controls (ages 9 and 10 years at baseline) received 1250 mg calcium daily. After 12 months, Tb.BMD was increased (5%) in controls taking the supplement, but minimally changed in gymnasts taking the supplements (1%); that is, there were no significant effects of calcium in gymnasts who were already intensely loading their skeleton [490]. Finally, excessive sodium, alcohol, carbonated beverages, and caffeine intake increased excretion and reduced absorption of calcium [491-494]. Prolonged calcium deficiency results in rickets, a disease in which there is defective mineralization and osteoclasts absorb bone to prevent hypocalcemia, causing progressively weaker bones [481].  Vitamin D facilitates bone remodeling by stimulating bone matrix formation and mobilizing calcium from bones. It enables calcium absorption in the intestine and stimulates calcium binding protein, thus maintaining calcium-phosphate homeostasis [482]. Dietary vitamin D requirements increase from 400 IU at birth, to 600 IU during pubertal growth, and 800 IU in older adulthood [482]. Dietary sources of vitamin D (vitamin D2 and D3) are limited to oily fish (e.g., salmon, sockeye, and mackerel). Though vitamin D is also available through sun exposure; UVB radiation from the sun converts 7-dehydrocholesterol to pre-vitamin D3, which is later converted to vitamin D3 [387]. In climates with daily sun (UVB) exposure, upwards of 90% of vitamin D can be synthesized via photo-conversion [495, 496]; serum levels of 25(OH)D (i.e., vitamin D status) varies seasonally.  Though the role of vitamin D in bone formation and calcium homeostasis is well accepted, it is less clear if vitamin D intake or supplementation positively effects bone accrual during growth [387, 487]. Serum levels of 25(OH)D did not predict BMC accrual in 4 to 8 year old girls. In fact 25(OH)D was negatively associated with BMC accrual (IGF-1 was a better predictor) [497]. Conversely, Cashman et al. [498] found that low 25(OH)D levels were associated with low aBMD in 10 to 16 year old girls. The authors suggested that maintaining vitamin D concentrations of ≥50nmol/L may improve bone health in post-101  pubertal girls [498]. Conceivably vitamin D is less necessary when modeling bone during childhood, when the GH-IGF-1 axis is most active. Whereas after puberty, maintaining vitamin D levels may influence bone accrual into young adulthood.   Winzenberg et al. [499] conducted a systematic review to determine the effectiveness of vitamin D supplementation on aBMD in boys and girls ages 8 to 17 years [499]. In six RCTs, there was a trend towards a small positive effect on lumbar spine aBMD (in children with 25(OH)D<35nmol/L), but no significant effects on total body aBMD or BMC, hip aBMD or BMC, or forearm aBMD or BMC [499]. When Du et al. [500] adjusted for menarcheal status, they observed higher total body BMC in post-menarcheal girls with vitamin D supplementation compared with controls. Again, effects of supplementation may vary by pubertal stage and baseline 25(OH)D, where benefits may be greater in post-pubertal adolescents with low baseline serum vitamin D status [500]. There is evidence to suggest that calcium and vitamin D intake are important in bone accrual during the growing years. However, only 10 to 30% of children and adolescents are getting adequate amounts of calcium and vitamin D daily [501].   Diet is a complex and variable behavior to study and measure (e.g., food frequency and 24-hour recall; described in Section 4.2.3.2). Though I have only described the functions of calcium, and vitamin D above, it is important to note that a number of macro and micronutrients are also required for normal bone metabolism (e.g., protein, magnesium, zinc, copper, iron, fluoride, vitamins A, C, K) [482]. Protein makes up the organic matrix of bone where mineralization occurs, and other minerals and vitamins carry out reactions and metabolic processes in bone tissue [482]. When RCTs combine dietary supplementation with weight-bearing activity, there tends to be an even greater effect on bone accrual [502].    2.2.4.6 Physical Activity   During growth, bones continually adapt their mass, density, structure, and strength to withstand loads from increases in length, muscle mass (i.e., muscle-bone unit [503]), or external forces (e.g., physical activity) [504]. Briefly, physical activity strains whole bone and activates mechanosensing cells which 102  signal the bone to adapt to loads through bone remodeling [375]. Further, participation in physical activity increases bone accrual into young adulthood [505]. This is in part due to the close relationship between bone and muscle, where muscle development precedes bone development [503]. The deformation in bone caused by physical activity and increasing muscle mass must be greater than habitual loads (i.e., a strain threshold) to positively adapt bone [506]. Strain threshold may vary between children of different ages, sex, ethnicity, maturity level, habitual physical activity levels, and different skeletal sites, amongst others [455]. Canadian physical activity guidelines currently recommend children and adolescents (ages 5 to 17 years) participate in 60 minutes of moderate to vigorous physical activity (MVPA) daily, including bone-strengthening exercises at least 3 days per week [507, 508]. Physical activity is frequently measured using self-reported questionnaires ([509] as described in Section 4.2.3.2). Objective measures of physical activity (such as accelerometry) are increasingly common and preferred for research [510]. Physical activity is unquestionably critical to optimize the growing skeleton (particularly in pre- and early-puberty [289]); however, the type, quantity, and timing of physical activity to optimize bone accrual in adolescence and young adulthood are still unclear [387]. In this section, I highlight studies that describe the relationship between physical activity and enhanced bone mass, structure, and strength. Given the scope of literature in this area, I defer to published systematic reviews and higher levels of evidence where they exist.  In a systematic review and narrative synthesis from our laboratory [511], we found that 26 of 37 trials (observational, recreational physical activity, organized sports) reported positive relationships between physical activity and bone strength in children. Further, Weaver et al. [387] found that of 20 prospective observational trials, 18 reported statistically significant differences in bone mass or density by DXA between physically active children and their non-active peers [387]. I will highlight some of the studies within these reviews.  In a 15-year prospective follow-up study on bone health in boys and girls (age 8 years at baseline) from PMBAS, men and women who were physically active (measured using self-report) during puberty had 8 to 10% greater BMC at TB (boys only), LS and FN compared with less active peers [512]. In a 12-year 103  prospective follow-up study on bone accrual in boys and girls (age 5 years at baseline) from the Iowa Bone Development Study, men and women who were physically active (measured using accelerometry) during childhood had 10 to 16% greater FN BMC and 8% greater proximal femur aBMD than their non-active counterparts [513]. These studies demonstrate the potential long-term benefits of childhood activity on skeletal health. Further, Weaver et al. [387] found that 30 of 36 RCTs reported statistically significant differences in bone mass and density between intervention (physical activity: sports, games, jumping, high-impact exercises, from 10 to 60 minutes, 2 to 5 days per week, from 7 to 24 months) and control groups. Greater differences were found in pre- and early-puberty compared post-puberty in both sexes. Some RCTs highlighted in the Tan et al. [511] and Weaver et al. [387] reviews were from HBSIII physical activity and jumping interventions. For example, in HBSII (boys and girls; ages 8 to 11 years at baseline) we found early pubertal girls gained 1.5 to 3.1% more BMC at FN and LS compared with controls [514] and early-pubertal boys gained 1 to 1.6% more total body BMC and PF aBMD compared with controls [515]. In B@B, we observed a 2% greater gain in PF BMC in exercisers compared with controls. Finally, in AS!BC, boys in the exercise group had greater gains (1.7 to 2.7%) in TB and LS BMC compared with controls, whereas girls had greater gains (3.5%) in FN BMC compared with controls [169]. The results show that well-timed and well-designed interventions can increase bone mass and density in children; whether enhanced bone continues after the intervention ceases is less clear.  Similarly, physical activity has been shown to positively impact bone structure and strength, though with less consistency. Observational trials reported consistent positive effects on bone structure and strength. For example, at 7-years follow-up, PBMAS showed MVPA was positively associated with FN section modulus and CSA by HSA [516]. After 15-years, PBMAS reported active youth (at APHV) had 8 to 12% greater section modulus and CSA compared with less active youth by HSA [517]. Finally, in an 18-year follow-up of the same group, men who were physically active teens (at APHV) had 13% greater SSIp, and 10% greater Tt.Ar at the tibia, women who were active pre-teens (at APHV) had 10% greater CoA by pQCT compared with their less active counterparts [518]. In HBSIII, we observed differences in bone 104  structure and strength between intervention participants and control participants. For example, exercising boys and girls had a 7.5% and 4% greater change, respectively, in NN Z by HSA compared with non-exercisers [519, 520]. Further, in AS!BC, exercising boys had a 3% greater grain in Imax by pQCT compared with controls, though no other structural parameters differed between groups [168, 521].   In summary, the evidence shows a clear positive effect of physical activity on bone mass and density, and slightly less consistent but positive effect on bone structure and strength. The process of bone accrual is complex; the timing of PBM and strength varies by skeletal site and bone compartment. It seems that physical activity in pre-puberty may be more beneficial to enhance bone mass and density, though it is unclear if this is true for bone structure and strength.    2.2.4.7 The Role of Maturity in Bone Accrual  Given that such a large proportion (40%) of bone is accrued during four to five years around the adolescent growth spurt (Figure 2.29), maturational timing could play a role in determining PBM [342]. A dearth of studies explore this topic. Some studies suggest that late-maturing children are at risk for reduced PBM, structural and microstructural deficits, and increased fracture risk in later life [32, 522-526], while others propose that late-maturing children ‘catch-up’ by late-adolescents or young adulthood and are left with no bone health deficits [33, 79, 80]. The maturity-bone relationship is a central tenet of my dissertation and is assessed in Chapter 7. Thus, to better understand the current position on the relationship between maturational timing and post-pubertal bone, I summarize nine studies (Table 2.9; 2006 to present) that specifically aimed to describe the relationship between maturational timing (defined by menarche, onset of PH2, and APHV) and late-adolescence or young adult bone mass, density, structure, and/or strength.   105  Table 2.9 Overview of studies that described the relationship between maturational timing and bone outcomes in young adulthood.   Author, Study Location Sample Description Methods Results and Conclusions Kindblom et al. [522]  GOOD Study;  Cross-sectional, Retrospective Chart Review  Gothenburg, Sweden Sample:  642 boys  Age:  18 to 20 years at baseline  Ethnicity:  not stated  Maturity indicator:  APHV (ICP model); categorized as early- average- and late-maturers by tertiles according to sample APHV  Adulthood defined as:  Young adulthood, ages 18 to 20 years  Bone measures: TB, LS, FN, and Radius BA, aBMD, BMC by DXA  Ct.BMD, Ct.Th, and CSA at the 25% site of the radius and tibia, and Tb.BMD at the 4% site of the radius and tibia by pQCT Late-maturing boys had lower TB and Radial aBMD, Radial Ct.BMD, Tb.BMD, Tibial Ct.BMD and Tb.BMD compared with early-maturing boys in young adulthood (age 18 to 20 years).  Conclusion: Late-maturers had lower bone mass, density, and structural parameters compared with early-maturers in young adulthood. Chevalley et al. [523]  Prospective Longitudinal  Geneva, Switzerland Sample:  124 girls  Age:  mean age 7.9 years at baseline  Ethnicity:  not stated Maturity indicator:  Age at menarche, categorized as early-maturers or late-maturers, compared with mean age at menarche  Adulthood defined as:  Young adulthood, mean age 20.4 years  Bone measures: Radial metaphysis, diaphysis, FN, LS aBMD by DXA Late-maturing girls had lower FN aBMD (0.4 T-score lower) at adulthood compared with early-maturing girls.  Conclusion: Late-maturers had lower aBMD compared with early-maturers in young adulthood. Chevalley et al. [527]  Prospective Longitudinal  Geneva, Switzerland Sample: 124 girls  Age: mean age 7.9 years at baseline  Ethnicity:  not stated   Maturity indicator:  Age at menarche, categorized as early-maturers or late-maturers, compared with mean age at menarche  Adulthood defined as:  Young adulthood, mean age 20.4 years  Bone measures: TB and FN aBMD by DXA Distal tibia (% not stated) BMD, Tb.BMD, Ct.BMD, Ct.Th, Tb.N, Tb.Th, BV/TV by HR-pQCT  Age at menarche was inversely correlated with early adult FN aBMD, and distal tibia BMD and Ct.Th, where late maturation was deleterious. Other variables showed no difference between early- and late-maturers.  Conclusion: Late-maturers had lower bone density and thinner cortices compared with early-maturers in young adulthood.  106  Table 2.9 Overview of studies that described the relationship between maturational timing and bone outcomes in young adulthood (cont’d).  Author, Study Location Sample Description Methods Results and Conclusions Gilsanz et al. [526]  Multi-Centre Study; Prospective Longitudinal  Los Angeles, Cincinnati, Omaha, Philadelphia, New York, United States  Sample: 85 boys; 78 girls  Age:  mean age 11.7 and 10.7 for boys and girls, respectively, at baseline  Ethnicity:  55% white, 20% black, 18% Hispanic, 6% Asian Maturity indicator:  Age of pubertal onset (PH2), categorized as  early-maturers or late-maturers, compared with mean age pubertal onset (PH2)  Adulthood defined as:  Late adolescence (skeletally mature by epiphyseal closure of phalanges and metacarpals, age 17 and 16 for boys and girls, respectively)  Bone measures: TB, LS, PF and ND forearm aBMD, BMC by DXA  Age of onset of puberty positively predicted aBMD and BMC at all sites in both sexes at adulthood (as defined by skeletal maturity).  Conclusion: Late-maturers had lower aBMD and BMC compared with early-maturers in young adulthood. Jackowski et al. [79]   PBMAS; Prospective Longitudinal  Saskatoon, Canada Sample:  108 boys; 118 girls  Age:  8 to 15 years at baseline  Ethnicity:  primarily white (98%)  Maturity indicator:  APHV (cubic splines), categorized as early-maturers if APHV preceded average APHV by one year or more, late-maturers if APHV was one year or more after average APHV  Adulthood defined as:  Young adulthood (ages 23 to 30 years)  Bone measures: NN, IT, FS CSA and Z by HSA Early-maturing boys and girls had greater NN, IT, and FS CSA and Z compared with average- and late-maturers during childhood and adolescence. After age 16 years in boys and 13 years in girls there were no differences between early-, average- and late-maturers.  Conclusion: Late-maturers started with lower NN CSA and Z compared with early-maturers but ‘caught-up’ by young adulthood. Jackowski et al. [33]  PBMAS; Prospective Longitudinal  Saskatoon, Canada Sample:  108 boys; 118 girls  Age:  8 to 15 years at baseline  Ethnicity:  primarily white (98%)  Maturity indicator:  APHV (cubic splines), categorized as early-maturers if APHV preceded average APHV by one year or more, late-maturers if APHV was one year or more after average APHV  Adulthood defined as:  Young adulthood (ages 23 to 30 years)  Bone measures: TB, FN, LS BMC by DXA Early-maturing boys and girls had greater BMC compared with late-maturing boys during childhood and adolescence. After age 16 years in boys and 13 years in girls, there were no differences between early-, average- and late-maturers.  Conclusion: Late-maturers started with lower bone content compared with early-maturers but ‘caught-up’ by young adulthood. 107  Table 2.9 Overview of studies that described the relationship between maturational timing and bone outcomes in young adulthood (cont’d).  Author, Study Location Sample Description Methods Results and Conclusions Darelid et al. [80]  GOOD Study; Retrospective Chart Review   Gothenburg, Sweden Sample:  501 boys  Age:  18 to 20 years at baseline  Ethnicity:  not stated   Maturity indicator:  APHV (ICP model), categorized as early- average- and late-maturers by tertiles according to sample APHV  Adulthood defined as:  Young adulthood (ages 19 to 24 years)  Bone measures: TB, LS, PF, FN ND Radius aBMD and BMC by DXA Radial Ct.BMD, Ct.Th, PeriCirc, EndoCirc, SSIp, at the 25% site, and Tb.BMD at the 4% site by pQCT APHV was a positive predictor of adult (19 years) TB, LS, and radius aBMD and explained 11.7%, 11.9%, and 23.5% of the variation of change in these variables, respectively. APHV also positively predicted TB, LS, and radius BMC and explained 4.3%, 10.5%, and 22.3%, respectively.   APHV was a positive predictor of adult (19 years) Ct.BMD, Tb.BMD and explained 14.3% and 6.2% of the variation of change in these variables, respectively. APHV was also a positive predictor of adult (19 years) Ct.Th, PeriCirc, SSIp.   By age 24 years, there were no differences between early-, average-and late-maturing boys, except radial aBMD and BMD.  Conclusion: Late-maturers started with lower bone content compared with early-maturers but ‘caught-up’ by young adulthood. Chevalley et al. [525]  Prospective Longitudinal  Geneva, Switzerland Sample: 124 girls  Age: mean age 7.9 years at baseline  Ethnicity:  not stated  Maturity indicator:  Age at menarche, categorized as early-maturers or late-maturers, compared with mean age at menarche   Adulthood defined as:  Young adulthood (mean age 20.4 years)  Bone measures: Radial metaphysis, diaphysis aBMD by DXA (at all study visits)  Distal radius (% not stated) BMD, Ct.BMD, Ct.Th, Tb.BMD by HR-pQCT; Stiffness, Failure Load by µFEA (acquired once at adulthood) Late-maturing girls had lower aBMD at radial diaphysis and metaphysis, lower radial BMD, reduced stiffness, and failure load, compared with early-maturing girls.  Conclusion: Late-maturers had lower bone density and strength compared with early-maturers in young adulthood.  108    Table 2.9 Overview of studies that described the relationship between maturational timing and bone outcomes in young adulthood (cont’d).  Author, Study Location Sample Description Methods Results and Conclusions Kuh et al. [528]  NSHD Study; Retrospective Cohort   England, Scotland, Wales, UK Sample: 704 men and 655 women  Age:  60 to 64 years at baseline  Ethnicity:  not stated Maturity indicator:  Age at menarche in girls, categorized as  early-maturers or late-maturers, compared with mean age at menarche; voice and genital change in boys; obtained from medical records; categorized as  early-maturers (advanced development of genitalia, voice broken) and late-maturers (delayed development of genitalia, voice not broken)  Adulthood defined as:  Older adulthood (ages 60 to 64 years)  Bone measures: LS, PF aBMD by DXA Radial SSIp, BMD, Ct.BMD at the 50% site and Tb.BMD at the 4% site by pQCT  Late maturing girls had lower total BMD and Tb.BMD, and LS and PF aBMD at older adulthood (age 60 to 64 years) compared with early-maturing girls. No differences were observed in other bone parameters between maturity groups.  Late maturing boys had lower Tb.BMD, and LS and PF aBMD, but higher Ct.BMD, at older adulthood (age 50 to 64 years) compared with early-maturing boys. No differences were observed in other bone parameters between maturity groups.  Conclusion: The deleterious effects of late-maturation MAY continue into older adulthood. *GOOD=Gothenburg Osteoporosis and Obesity Study; APHV=age at peak height velocity; ICP=infant-childhood-puberty; TB=total body; LS=lumbar spine; FN=femoral neck; BA=bone area; aBMD=areal bone mineral density; BMC=bone mineral content; DXA=dual energy X-ray absorptiometry; Ct.BMD=cortical (volumetric) bone mineral density; Ct.Th=cortical thickness; CSA=cross sectional area; BMD=(volumetric) bone mineral density; Tb.BMD=trabecular bone mineral density; pQCT=peripheral quantitative computed tomography; Tb.N=trabecular number; Tb.Th=trabecular thickness; BV/TV=(trabecular) bone volume fraction; HR-pQCT=high resolution pQCT; PH2=pubic hair stage 2; PF=proximal femur; ND=non-dominant; PBMAS=Pediatric Bone Mineral Accrual Study; NN=narrow neck; IT= intertrochanteric; FS=femoral shaft; Z=section modulus; HAS=hip structural analysis; PeriCirc=periosteal circumference; EndoCirc=endosteal circumference; SSIp=polar strength strain index; µFEA=micro finite elemental analysis; NSHD=National Survey of Health and Development109   These studies illustrate the equivocal relationship between maturational timing and late-adolescence or young adult bone health. In most (67%) of the studies, post-pubertal or young adult bone health was compromised in late-maturers [522, 523, 525-528], whereas only three of nine studies suggested that late-maturers catch up and achieve similar bone values to their early- and average-maturing peers by late-adolescence or young adulthood (i.e., BMC and aBMD at all sites by DXA; NN, IT, FS, CSA, Z by HSA; Radial Ct.BMD, Ct.Th., periosteal and endosteal circumference at 25% site, and Tb.BMD at 4% site by pQCT) [33, 79, 80]. Ambiguity in the literature may be due to variability in methods used to define maturity (i.e., menarche in girls, voice changes in boys, or pubic hair development or APHV in both sexes), or the age at which the relationship is assessed (i.e., skeletally mature (ages 17 and 16 years for boys and girls, respectively) to older adulthood (ages 60 to 64 years)). Historically, the maturity-bone relationship was examined in girls using age at menarche as an indicator of maturation [32, 523, 524, 529], but more recent studies include boys or both sexes and use other maturity indicators such as APHV [522, 526]. Notably, APHV is the most commonly employed indicator of somatic maturity and is related to skeletal development in both sexes. My dissertation addresses the need for studies that use APHV (as an accurate indicator of maturity) to better understand the relationship between timing of maturation and bone mass, density, structure, and strength in late adolescence.      2.3 Summary of the Growth, Maturation and Bone Accrual Literature   Bone mass and strength achieved by late-adolescence and young adulthood may play a role in preventing osteoporosis and related fracture later in life. Consequently, it is important to assess factors that may influence bone accrual during adolescence. For example, some suggest that late-maturers are disadvantaged with respect to bone accrual and PBM, others suggest late-maturers catch up in post puberty or by young adulthood. Therefore, in my dissertation I will address the need for studies that accurately assess maturity (e.g., APHV) as a predictor of bone accrual in late-adolescence to better understand the relationship between timing of maturation and post-pubertal bone mass, density, structure, and strength.    110  Chapter 3: Rationale, Objectives, Hypotheses, and Contributions I identified three gaps in the literature that I address in my dissertation. First, current prediction models used to estimate APHV require further validation. Second, given known ethnic differences in tempo and timing of maturation [30], validation and calibration of the prediction equations in non-white populations is warranted. Third, the influence of maturational timing on post-pubertal bone is unclear; some reported that young adults who were late-maturing adolescents were disadvantaged with respect to bone outcomes whereas others did not believe this to be the case [80, 526]. Thus, I recognized the need to evaluate the relationship between maturational timing (using APHV) and bone-specific parameters in late adolescence. In this chapter, I provide the rationale, specific objectives, and hypotheses for the three studies that address these gaps.    111  3.1 STUDY 1: Enhancing a Somatic Maturity Prediction Model   3.1.1 Rationale As there are well-known differences in tempo and timing of maturity for boys and girls at the same chronological age, it is vital to assess maturation in all investigations of children and youth. To that end, several methods are commonly used to estimate a child’s maturation status and rate [2, 4, 11, 109]. However, concerns regarding invasiveness, intrusiveness, and/or logistics limit widespread use of such methods [3, 291]. APHV is the most commonly used indicator of somatic maturity; timing of APHV affects many biological parameters (moreso than growth for chronological age), including skeletal health. Serial height measures are needed to accurately determine APHV; however, APHV can also be estimated using prediction equations and one-time measures of body size [26]. Mirwald generated prediction equation using data acquired from white children in Saskatchewan. These equations have been well utilized, but recently their accuracy was questioned [27-29]. Critics did not examine the possibility of over-fitting and within-subject correlation that likely influenced the Mirwald equations, nor did they consider calibrating the equations using external samples. Therefore, I will examine methodologies used to develop Mirwald equations and estimate prediction error in external samples. Subsequently, I will redevelop the Mirwald equations to better represent a greater proportion of children and youth.   3.1.2 Objectives  The specific objectives of this study are to: 1) examine the modeling approach used to generate the Mirwald equations (e.g., check for over-fitting or within-subject correlation); 2) redevelop regression equations employing methods that consider within-subject correlation (cluster-robust variance techniques and k-fold cross-validation); and 112  3) assess fit of newly developed equations in external samples (i.e., validate new equations in white boys and girls from the HBSIII [368, 514, 530, 531] and the HGS [2, 532]).  3.1.3 Hypotheses 1) new equations with fewer predictors will have similar prediction accuracy as the Mirwald equations; and  2) new equations will have high prediction accuracy in external samples.   3.1.4 Contribution  This study will be the first to validate and calibrate commonly used maturity prediction equations. Mirwald equations are widely used in many pediatric subfields (e.g., clinical populations, talent identification in sport). As accuracy of the prediction equations in external samples (including early-maturing children and children of non-white ethnicities) was recently questioned I aim to reexamine these equations. To do so, I will collaborate with lead researchers from growth studies that generated original data (PBMAS, SGDS, LLTS) and related publications and will examine the accuracy of the Mirwald equations in external samples. Using data from our HBSIII [368, 514, 530, 531] and the HGS [2, 532], I will develop new equations to remedy over-fitting and reduce prediction error. Finally, I will provide important recommendations for current users of maturity prediction equations.     113  3.2 STUDY 2: Growth and Maturation of Asian-Canadian Children   3.2.1 Rationale Ethnicity and environment strongly influence timing and tempo of adolescent growth. For example, APHV occurs earlier in Asian compared with white children. It is therefore important to control for maturational timing in growth studies, particularly when assessing children of different ethnic origins. However, prediction equations used to estimate APHV were developed in white children and may overestimate maturity in Asian children. Importantly, ethnic differences in growth and maturation are less apparent, or more conservative, when children of different ethnicities reside in the same environments, as opposed to living in different geographical areas. The HBSIII study acquired prospective data across 14-years from 412 white and 533 first- and second-generation Asian-Canadian boys and girls who reside in the same communities in Metro Vancouver [368, 514, 530, 531]. Therefore, I will utilize these valuable (and rare) longitudinal data and assess the influence of ethnicity on maturational timing of white and Asian children who resided in geographic proximity. I will also examine the utility of maturity prediction equations for use with data acquired from Asian children, and develop new prediction equations that are relevant for Asian boys and girls.   3.2.2 Objectives  The specific objectives of this study are to: 1) describe differences in growth and maturity between Asian and white boys and girls who live in the same geographic area; 2) assess fit of maturity prediction equations for Asian boys and girls; and 3) create ethnic-specific regression equations that more accurately predict MO in Asian boys and girls.  114  3.2.3 Hypotheses 1) Asian children living in Canada will be shorter and mature earlier than their white peers;  2) prediction equations developed using data acquired from white children will perform less well (i.e., have less prediction accuracy) compared with the development cohorts (white children); and 3) ethnic-specific equations will have better prediction accuracy in Asian children compared with equations developed using data acquired from white children.  3.2.4 Contribution  This study will uniquely assess growth and maturation of Asian children compared with white children who resided in the same geographic area. Studies most commonly compare children who live in different countries (if at all); few compared children of different ethnicities who lived in proximal communities (i.e., Metro Vancouver). Thus, I will capitalize on the multi-ethnic nature of the HBSIII sample to assess growth and maturation between Asian and white children exposed to similar living environments (e.g., walkability, seasonal variations, food choice, accessibility, etc.). Importantly, specific maturity prediction equations do not currently exist for Asian children. Therefore, I will modify maturity prediction equations to accurately predict maturity in Asian children.     115  3.3 STUDY 3: Does Maturational Timing Predict Bone Mass, Density, Structure, and Strength in Late-Adolescence?   3.3.1 Rationale Attainment of maximal amounts of bone during pubertal growth may be a critical factor in deterring or preventing the detrimental effects of osteoporosis and osteoporosis-related fractures in later life. Pubertal timing influences peak bone mineral accrual; up to 40% of adult bone mass is accrued during the four to five years surrounding rapid linear growth [342]. Late-maturing children may be at-risk for reduced bone in adulthood, bone mass and structural deficits compared with average- and early-maturing children [32, 523-526]. However, if tracked into late-adolescence or young adulthood, those whom were previously considered late-maturing children ‘caught up’ to their average- and early-maturity peers, and deficits in bone mass, structure, and strength did not persist [33, 79, 80]. Most studies that considered late-maturation deleterious to post-pubertal bone assessed only girls or women and used retrospective age at menarche to define maturity [32, 523-525, 528]. Far fewer studies assessed boys/men or both sexes [33, 79, 80, 522], or used APHV to define maturational timing [33, 79, 80, 522], despite APHV being a more accurate reflection of somatic and skeletal maturity. Studies that included boys were more likely to find no effects of maturational timing on the adult skeleton [33, 79]. Further, our current understanding of the relationship between maturational timing and bone health is based on results from DXA-based studies that reported BMC or aBMD as primary outcomes [33, 523, 526]. Although aBMD is used clinically to diagnosis osteoporosis and predict fracture risk, it does not differentiate between the components of bone structure that contribute to bone strength. Only a handful of studies assessed bone density, structure, and strength using pQCT or HR-pQCT [80, 522, 528]. Effects of maturational timing were equivocal, and site and/or component specific [528]. Finally, we know little about the influence of ethnicity on the relationship between timing of maturity and post-pubertal bone, as only one previous investigation of this relationship included a multi-ethnic cohort (only 6% Asian) [526]. Thus, there remains a need to confirm the relationship 116  between maturational timing and post-pubertal bone mass, structure, and strength, and to confirm the nature of the relationship in participants of different ethnic origin.  3.3.2 Objectives  The specific objectives of this study are to: 1) describe the relationship between maturational timing and adult BMC (g) and aBMD (g/cm2) by DXA at the whole body, lumbar spine, and femoral neck, adjusting for ethnicity, body size, muscle, physical activity, and calcium intake. 2) describe the relationship between maturational timing and adult Tt.Ar (mm2), Ct.Ar (mm2), Ct.Th (mm), Me.Ar (mm), Ct.BMD (mg/cm3), and SSIp (mm3) by pQCT at the mid-shaft tibia, adjusting for ethnicity, body size, muscle, physical activity, and calcium intake.  3.3.3 Hypotheses 1) APHV will predict BMC and aBMD by DXA in crude, but not adjusted models (adjusted for ethnicity, body size, muscle mass, physical activity, and calcium intake);  2) APHV will predict Tt.Ar, Ct.Ar, Ct.Th, Me.Ar, Ct.BMD, and SSIp by pQCT in crude, but not adjusted models (adjusted for ethnicity, body size, MCSA, physical activity, and calcium intake); and 3) muscle mass/MCSA and body size (height) will be the key predictor variables in each model.  3.3.4 Contribution  It is still unclear whether late maturation is deleterious to post-pubertal and young adult bone. I will clarify the relationship between maturational timing and bone mass, density, structure, and strength accrual in late-adolescence. I will utilize data from our HBSIII. In HBSIII we assessed APHV and MO in all 117  children. I will assess both DXA and pQCT-derived bone outcomes to better understand the maturity-bone relationship.  118  Chapter 4: Methods In this chapter I provide an overview of the history and design of HBSIII, methods used to recruit and retain the HBSIII cohort, data collection methods, and statistical analyses relevant to my dissertation. Where methods and statistical analyses are specific to research studies, I include more detailed information in the relevant chapter.  4.1 Healthy Bones Study III In this section I describe the HBSIII design and study cohort, including participant recruitment and retention across the 14-year study.   4.1.1 Study Design  HBSIII was a mixed-longitudinal observational study designed to assess factors associated with bone accrual in children. Across the 14-year study (1999 to 2012), the HBSIII assessed 1071 participants (515 boys and 556 girls) recruited from 29 elementary schools in Vancouver and Richmond school districts in British Columbia, Canada. Participants were 8.8 to 12.4 years of age at baseline; ages ranged from 8.8 to 23.2 years across the 14-year study.  HBSIII included four waves of data collection: Healthy Bones Study II (HBSII; 1999 entry), Bounce at the Bell (B@B; 2000 entry), Action Schools! BC (AS!BC; 2003 entry), and the final Healthy Bones group (HBS2009; 2009 entry). The first three studies (HBS-II, B@B, and AS!BC) were designed as school-based interventions [515, 521, 530, 533, 534]. At the conclusion of the intervention trials, participants were invited to return for annual follow-up measurements. Given the nearly identical designs across studies (detailed in Appendix A) and that no study or secular effect was apparent (data not shown), datasets were merged in the spring of 2007 to form the HBSIII cohort. HBSIII participants were followed annually through 2011. In 2008, our laboratory acquired a 3D imaging system (HR-pQCT) to assess bone microstructure in the HBSIII cohort. As most of the cohort was post-pubertal in 2008 (93%), we recruited 119  a fourth cohort in 2009 to assess pre- and early-pubertal changes in bone microstructure using HR-pQCT. We followed this group annually until 2012. Data collection protocols for this cohort were identical to those employed with the HBSIII cohort; thus, we merged all data to create the Pediatric Bone and Physical Activity Database [368, 535]. I provide an overview of the mixed longitudinal study design, including dates of measurement and overall sample size by cohort, sex, and ethnic group in Figure 4.1. 120   Figure 4.1 This figure provides an overview of the mixed-longitudinal design of the Healthy Bones Study III (HBSIII). It illustrates sample size by individual study cohorts and total sample size for HBSIII by sex and ethnicity. Light shaded boxes represent when two or more cohorts were measured simultaneously. Dark shaded boxes represent when in school measurements were performed (all other measurements were taken in our bone health laboratory). n(participants) =1071; n(observations; obs) =7943.  121  4.1.2 My Role2 in the Healthy Bones Study III HBSIII was underway when I began my graduate studies in the fall of 2007. Thus, I was not involved in recruitment of the first three study cohorts, the intervention trials, nor data collection from 1999 to 2007. Beginning in 2008 and continuing through 2012, I assisted with scheduling participants for follow-up data collection. I also designed and facilitated training programs for the measurement team. I recruited the new HBS cohort (HBS2009; n=120) as I describe in Section 4.1.3. Further, I was the primary DXA technician for HBSIII from 2008 to 2012. In this role, I acquired and analyzed more than 3800 DXA full body, hip, and lumbar spine scans. I also assisted with quality assurance protocols for all anthropometry and developed checking/cleaning protocols for anthropometry data where appropriate (e.g., tibial length measurement protocol). At the end of each measurement period, I assisted with data cleaning and statistical analysis. Upon conclusion of the study in 2012, I plotted all longitudinal data (n=1071), checked for missing data and data errors not apparent in yearly checks (e.g., decreasing heights over time), and created standard protocols for data cleaning. Lastly, I created standard protocols for estimating APHV and assigning adult height.    4.1.3 Participant Recruitment Recruitment procedures for HBS and AS!BC are described in detail elsewhere and were similar across projects [515, 536]. First, the study team made presentations to school principals at district meetings; interested principals then volunteered to have their school participate. Second, the study team presented to students and teachers in grades 4, 5, and 6 at schools that demonstrated interest. Third, the study team provided letters and consent forms to classroom teachers who distributed the forms to their students. Fourth,                                                       2I clarify my role in section 4.1.2; though in subsequent methods and measurement sections I use the pronoun ‘we’ as over the 14-year duration of the study there were multiple measurers. Using ‘I’ would not reflect the nature of the study design or acknowledge the work of a large research team. Thus, I only use ‘I’ for roles that were exclusively my responsibility for the entire study duration. 122  members of the study team obtained consent for participation in the study in 1999, 2001, 2003, 2006, and 2009 for the HBSII and B@B study participants, and in 2003, 2004, 2006, 2007, and 2009 for the AS!BC participants.  As 93% of the initial cohort was post-pubertal when we acquired a new 3D imaging system (HR-pQCT) in 2008, we sought to recruit a younger pre- and peri-pubertal group (ages 10 to 12 years) to fill the age gap. We acquired measures of bone structure, density, and strength using HR pQCT during the 2009 measurement period. To recruit the younger cohort of 120 grade 4 and 5 students (40 boys and 80 girls), I followed a protocol similar to previous HBSII recruitment approaches. Specifically, I contacted five schools in the Vancouver and Richmond School Districts, met with principals and teachers, and then presented to grade 4 and 5 students during organized assemblies or within their classrooms. I answered questions from the principals, teachers, and children, and distributed information letters, consent forms and health history questionnaires (Appendix B). We aimed to recruit an additional 100 children for this study. We received consent forms from 121 children who were interested in participating and 120 attended spring 2009 data collection. We measured this group annually until 2012. We acquired parental consent at recruitment in 2009 and again in 2012. Of note, I do not include HR-pQCT bone outcomes in my dissertation; however, I include anthropometry, DXA, and pQCT outcomes from this newest HBS2009 cohort.   4.1.4 Participant Retention and Attrition As in any longitudinal study, our ultimate goal was to retain as many participants as possible over time. However, retention in any long-term prospective study is challenging and even moreso as participants move from childhood through adolescence (e.g., reaching maturity; moving to secondary schools). We used several incentives (i.e., snacks, stickers, pencils, Frisbees, etc.) over the study period to retain participants. We also mailed each a results sheet and study update to each participant (Appendix C).  We were unable to document reasons for loss to follow-up in all cases; however, we know that participants moved out of province or to other schools (including graduating to post-secondary education 123  or the work force), some participants were not available for study measurement, some participants declined participation in follow-up measures (i.e., after completion of the intervention study to which they initially consented), and others could not be contacted.  We retained 17% of the original HBSII cohort (13 years; 1999 to 2011), 13% of the B@B study cohort (12 years; 2000 to 2011), 20% of the AS!BC cohort (9 years; 2003 to 2011), and 49% of the HBS2009 cohort (4 years; 2009 to 2012). Each year we invited all children who participated in the previous year’s measurement as well as those who may have missed a measurement period, to participate. In Table 4.1, I provide an extensive overview of the sample by study cohort, sex, and ethnicity for each measurement period.       124  Table 4.1 Description of the Healthy Bones Study III (HBSIII) sample by sex (b=boys; g=girls), ethnicity by category (Asian/white/other), and measurement period; where HBSII is the original Healthy Bones Study II cohort, B@B is the Bounce at the Bell study cohort, AS!BC is the Action Schools! BC study cohort, and HBS2009 is the newly recruited HBSIII cohort. Participants were given the opportunity to re-enter the study if they missed a measurement period. Total number of participants measured at each period is listed as well as the total number of observations (obs) for the study.   HBSII B@B  AS!BC HBS2009 Total Sample Oct-Nov 1999  total: 383 (161/176/46) b:191 (81/87/23) g:192 (80/89/23) - - - total: 383 (161/176/46) b:191 (81/87/23) g:192 (80/89/23) Jan-Feb 2000  total: 348 (149/157/42) b: 174 (74/79/21) g: 174 (75/78/21) - - - total: 348 (149/157/42) b: 174 (74/79/21) g: 174 (75/78/21) May-Jun 2000 total: 366 (153/169/44) b: 185 (78/85/22) g: 181 (75/84/22) - - - total: 366 (153/169/44) b: 185 (78/85/22) g: 181 (75/84/22) Oct-Nov 2000 total: 186 (72/92/22) b: 97 (41/45/11) g: 89 (31/47/11) total: 53 (33/16/4) b: 26 (13/12/1) g: 27 (20/4/3) - - total:239 (105/108/26) b: 123 (54/57/12) g: 116 (51/51/14) Jan-Feb 2001 total: 173 (69/83/21) b: 90 (40/40/10) g: 83 (29/43/11) total: 36 (20/12/4) b: 17 (8/8/1) g: 19 (12/4/3) - - total: 209 (89/95/25) b: 107 (48/48/11) g: 83 (41/47/14) May-Jun 2001  total: 182 (69/93/20) b: 93 (40/43/10) g: 89 (29/50/10) total: 48 (29/15/4) b: 22 (10/11/1) g: 26 (19/4/3) - - total: 230 (98/108/24) b: 115 (50/54/11) g: 115 (48/54/13) Oct-Nov 2001 total: 183 (65/102/16) b: 84 (28/49/7) g: 99 (37/53/9) total: 34 (23/9/2) b: 16 (10/6/0) g: 18 (13/3/2) - - total: 217 (88/111/18) b: 100 (28/55/7) g: 117 (50/56/11) May-Jun 2002 total: 178 (65/99/14) b: 82 (28/48/6) g: 96 (37/51/8) total: 34 (23/9/2) b: 16 (10/6/0) g: 18 (13/3/2) - - total: 212 (88/108/16) b: 98 (38/54/6) g: 114 (50/54/10) Feb-Apr 2003  - - total: 515 (283/167/65) b: 258 (142/88/28) g: 257 (141/79/37) - total: 515 (283/167/65) b: 258 (142/88/28) g: 257 (141/79/37) May-Jun 2003  total: 148 (41/92/15) b: 70 (18/44/8) g: 78 (23/48/7) total: 19 (11/7/1) b: 9 (5/4/0) g: 10 (6/3/1) total: 496 (273/161/62) b: 248 (137/84/27) g: 248 (136/77/35) - total: 663 (325/260/78) b: 327 (160/132/35) g: 336 (165/128/43) Sep-Oct 2003  - - total: 450 (248/153/49) b: 228 (125/81/22) g: 222 (123/72/27) - total: 450 (248/153/49) b: 228 (125/81/22) g: 222 (123/72/27) Jan-Feb 2004  - - total: 443 (245/148/50) b: 225 (125/77/23) g: 218 (120/71/27) - total: 443 (245/148/50) b: 225 (125/77/23) g: 218 (120/71/27) 125  Table 4.1 Description of the HBSIII sample (continued).   HBSII B@B  AS!BC HBS2009 Total Sample May-Jun 2004  total: 138 (41/84/13) b: 64 (18/38/8) g: 74 (23/46/5) total: 21 (13/7/1) b: 9 (5/4/0) g: 12 (8/3/1) total: 451 (245/152/54) b: 228 (124/79/25) g: 223 (121/73/29) - total: 610 (299/243/68) b: 301 (147/121/33) g: 309 (152/122/35) Sep-Oct 2004  - - total: 294 (154/107/33) b: 156 (80/60/16) g: 138 (74/47/17) - total: 294 (154/107/33) b: 156 (80/60/16) g: 138 (74/47/17) Jan-Feb 2005  - - total: 347 (190/116/41) b: 186 (101/66/19) g: 161 (89/50/22) - total: 347 (190/116/41) b: 186 (101/66/19) g: 161 (89/50/22) May-Jun 2005  total: 153 (48/90/15) b: 71 (21/41/9) g: 82 (27/49/6) - total: 350 (191/117/42) b: 191 (105/66/20) g: 159 (86/51/22) - total: 503 (239/207/57) b: 262 (126/107/29) g: 241 (113/100/28) May-Jun 2006  total: 137 (39/85/13) b: 65 (17/39/9) g: 72 (22/46/4) total: 14 (7/6/1) b: 6 (3/3/0) g: 8 (4/3/1) total: 177 (87/72/18) b: 102 (47/45/10) g: 75 (40/27/8) - total: 328 (133/163/32) b: 173 (67/87/19) g: 155 (66/76/13) May-Jun 2007 total: 129 (37/78/14) b: 58 (15/34/9) g: 71 (22/44/5) total: 12 (7/5/0) b: 5 (3/2/0) g: 7 (4/3/0) total: 170 (92/62/16) b: 96 (50/39/7) g: 74 (42/23/9) - total: 311 (136/145/30) b: 159 (68/75/16) g: 152 (68/70/14) May-Jun 2008 total: 106 (34/61/11) b: 45 (15/25/5) g: 61 (19/36/6) total: 10 (6/4/0) b: 5 (3/2/0) g: 5 (3/2/0) total: 163 (83/63/17) b: 97 (50/39/8) g: 66 (33/24/9) - total: 279 (123/128/28) b: 147 (68/66/13) g: 132 (55/62/15) May-Jun 2009 total: 75 (25/44/6) b: 32 (13/17/2) g: 43 (12/27/4) total: 8 (5/3/0) b: 5 (3/2/0) g: 3 (2/1/0) total: 137 (67/56/14) b: 82 (39/37/6) g: 55 (28/19/8) total: 120 (56/53/11) b: 40 (14/20/6) g: 80 (42/33/5) total: 340 (153/156/31) b: 159 (69/76/14) g: 181 (84/80/17) May-Jun 2010 total: 74 (26/43/5) b: 30 (13/16/1) g: 44 (13/27/4) total: 7 (5/2/0) b: 3 (2/1/0) g: 4 (3/1/0) total: 126 (59/51/16) b: 78 (36/35/7) g: 48 (23/16/9) total: 112 (52/50/10) b: 39 (14/20/5) g: 73 (38/30/5) total: 319 (142/146/31) b: 150 (65/72/13) g: 169 (77/74/18) May-Jun 2011 total: 65 (23/38/4) b: 28 (11/16/1) g: 37 (12/22/3) total: 7 (5/2/0) b: 3 (2/1/0) g: 4 (3/1/0) total: 102 (50/40/12) b: 67 (34/27/6) g: 35 (16/13/6) total: 104 (47/47/10) b: 34 (12/17/5) g: 70 (35/30/5) total: 278 (125/127/26) b: 132 (59/61/12) g:146 (66/66/14) May-Jun 2012 - - - total: 59 (16/35/8) b: 24 (7/13/4) g: 35 (9/22/4) total: 59 (16/35/8) b: 24 (7/13/4) g: 35 (9/22/4) Total  Sample Observations total: 3024 (1117/1586/321) b: 1459 (551/746/162) g: 1565 (566/840/159) total: 303 (187/97/19) b: 142 (77/62/3) g:161 (110/35/16) total: 4221 (2267/1465/489) b: 2242 (1195/823/224) g: 1979 (1072/642/265) total: 395 (171/185/39) b: 137 (47/70/20) g: 258 (124/115/19) total: 7943 (3742/3333/868) b: 3980 (1870/1701/409) g: 3963 (1872/1632/459)        126  4.1.5 Comparison Cohorts In my dissertation, I acquired four other datasets from well-known longitudinal studies as a means to externally validate the maturity equations (Chapter 5). I connected with these researchers, discussed my dissertation objectives and the importance of their longitudinal data as comparison metrics for my dissertation, requested and transferred data, and checked and cleaned all data. To reassess the modeling technique used in the Mirwald equations, I used data from the PBMAS, HGS, SGDS and LLTS. I describe these cohorts in detail in the appropriate research chapters, as well as the techniques used for data cleaning, measurement, and analysis related to these cohorts.  4.2 Measurements In this section I describe research team training, participant flow, and data collected for HBSIII.   4.2.1 Planning for Measurement and Participant Flow Prior to each data collection period, the measurement team (students and staff; referred to in subsequent sections as research assistants (RAs)) participated in a full-day training session to learn (or refresh) all procedures and safety protocols. This session was offered to ensure standardized protocols for all measures and acquisition of quality data. The team practiced measurement techniques under the guidance of criterion measurers. In addition, our senior measurement team (myself included) closely observed and supervised measurement during the first week of data collection and provided feedback to RAs at the end of each day. The HBSIII measurement team typically consisted of a driver who was responsible for transporting children to and from school (with a chaperone for the younger children), a team leader (often leading anthropometry and floating where necessary), a secondary anthropometry assistant, two questionnaire administrators, a muscle strength and power assessor, a DXA technician, a pQCT technician (2001 to 2012), an HR-pQCT technician (2008 to 2012), and an accelerometry technician (2008 to 2012). I will not describe procedures related to HR-pQCT or accelerometry as I did not use these data in my dissertation.  127  Generally, we scheduled three groups of participants per day, four days per week. A maximum of six participants attended each session if they were in high school or older, and a maximum of five if participants were in elementary school. The first group was picked-up at the school at approximately 9:00 AM and driven to the Centre for Hip Health and Mobility (CHHM) at Vancouver General Hospital (VGH). The van then returned for the second (10:40 AM), and third (12:15 PM) groups while returning the already measured groups back to their schools. Participants returned to school approximately three hours after their initial pick-up. While at the laboratory, participants rotated through stations for bone densitometry, anthropometry, and questionnaires.  Protocols were standardized across all years of measurement. For six data collection periods when bone was not assessed, height, weight, and questionnaire data were acquired in schools rather than in the laboratory to limit student time away from class. This varied by wave (as noted above) and included winter 2000, winter 2001, spring 2003, fall 2003, fall 2004, and winter 2005. Below I describe protocols specific to in-laboratory measurements; details of school measurement can be found elsewhere [531, 537, 538]. In some cases, I excluded school-based measurement time-points for some studies and provide a rationale for doing so in pertinent chapters.  4.2.2 Anthropometry We assessed each participant’s height (standing and sitting) and weight at each measurement period. We used a divider screen in the measurement room to provide participants with privacy during measurement. Two trained research assistants took all measurements in duplicate. If measures differed by more than the accepted measurement error (height: 4 mm; sitting height: 4 mm; tibial length: 4 mm; weight: 0.2 kg), a third measure was taken. I used the mean of two values or the median of three for all analyses. In our laboratory, reproducibility (coefficient of variation; CV%) was <0.3% for anthropometry, except for tibial length, where CV% was <3.5% (unpublished data).  128  4.2.2.1 Standing Height, Sitting Height, Leg Length, and Tibial Length We measured height in standing (cm) and sitting3 (cm) positions using standard stretch stature techniques [88] to the nearest 0.1 cm using a customized, wall-mounted stadiometer (1999 to 2002) and thereafter using a Seca wall-mounted stadiometer (model 242, Hanover, MD, USA; 2003 to 2012). With shoes removed we asked each participant to put their heels together, place arms at their sides and back against the stadiometer post. If necessary, we asked children to remove ponytails, clips, or hats. For sitting height, we asked the participant to sit tall with his or her knees together, hands on lap, and feet flat against the floor. For sitting and standing height we positioned the head in the Frankfort plane (as described in Section 2.1.3.1). Stretch stature was achieved when one research assistant applied gentle traction at the mastoid process. The second research assistant slowly lowered the headboard of the stadiometer onto the apex of the participant’s head. We later calculated leg length (cm) by subtracting sitting height from standing height.  To assess the 50% site of the tibia using pQCT it was necessary to measure tibial length (2001 to 2012). We asked participants to wear shorts or roll their pant leg past their knee. The participant sat with their left ankle crossed atop of their right knee. We measured length of the left tibia (mm) as the distance from the edge of the medial malleolus to the tibial joint line (palpated manually) to the nearest 1.0 mm using a standardized flexible steel anthropometric tape (Rosscraft, BC, Canada; 2001 to 2012).   4.2.2.2 Weight and Body Mass Index We asked participants to dress lightly (e.g., shorts and t-shirt), empty pockets, and remove shoes and/or any heavy articles of clothing (e.g., jackets). We asked participants to stand on the scale with feet together, hands at his or her side, and to look straight ahead. We measured weight to the nearest 0.1 kg on a calibrated                                                       3Height is comprised of trunk length + leg length. To assess truck length, we measure sitting height. 129  Seca electronic scale (model 840, Hanover, MD, USA). I calculated each child’s BMI as weight (kg) divided by height squared (m2).  4.2.3 Questionnaires We obtained contact information for every child. At each visit, RAs guided each child through the questionnaires and checked each questionnaire for completion and correctness. Questionnaires assessed health history, physical activity, dietary calcium, and maturity. Below I describe the questionnaires relevant to my dissertation. All sample of all questionnaires used the HBSIII is in Appendix D.   4.2.3.1 Health History and Ethnicity Prior to (or at) the initial study visit, a parent or guardian of each consented participant completed a health history questionnaire from which we determined any acute or chronic medical conditions that would prevent the child from participating in the study. Parents/guardians answered questions regarding their family’s medical history and their child’s use of medication. This questionnaire also inquired about parent and child ethnicity. We determined each participant’s ethnicity based on their parents’ or grandparents’ place of birth as reported on their health history questionnaire at baseline. We classified participants as white (42%) if both parents, or three of four grandparents, were born in North America or Europe; and Asian (47%) if both parents or three of four grandparents were born in Hong Kong, China, Japan, Taiwan, Philippines, Korea or India. We classified participants as being of ‘other’ or ‘mixed’ ethnicity (11%) when the child was neither white nor Asian, or if the child had parents of distinct ethnic backgrounds (e.g., black, indigenous). We confirmed our classifications by cross-referencing the parent’s self-reported ethnicity with the reported parent’s and grandparent’s birthplace. We used a similar classification system as the Canadian census on ethnic and ancestral origins. 130  At each subsequent visit, participants completed a follow-up health history questionnaire to identify changes to current medical conditions, family medical history, and medication use (including dosage, frequency of use), and to document any injuries (including fracture history) in the previous year.   4.2.3.2 Physical Activity and Dietary Calcium Intake  We assessed self-reported physical activity over the previous seven days using the Physical Activity Questionnaire for children (in elementary school; PAQ-C) and adolescents (in high-school; PAQ-A) [509]. The PAQ-C used a checklist format with 22 common leisure and sport physical activity; we calculated an average score using a 1-to-5 scale (where 1=low activity and 5=high activity) from nine questions (1-8, 11). The PAQ-A is similar to the PAQ-C except that it does not include activity during recess (item #3 in PAQ-C). Both versions were designed to be used during the school year, rather than during holiday periods. For continuity, we continued to use the PAQ-A after high school graduation and into adulthood, recognizing that not all questions were applicable. I used the mean physical activity score from the PAQ as a covariate in the statistical models in Section 7.2. Based on participants’ estimates of time spent in common sports and activities in Item 1, we also estimated time spent in moderate through vigorous physical activity (MVPA; min/day) and impact/loaded physical activity (hours/week). Validity of the PAQ was determined using an aerobic step test as a criterion measure [539, 540]. The survey was shown to have good internal consistency (standardized Cronbach’s α ranging from 0.72-0.79) [541, 542].   We used a validated food frequency questionnaire (FFQ) [543] to assess dietary calcium intake (mg/day). A research assistant used cues, pictures, and food models to approximate food serving sizes. Participants reported how often they consumed 20 calcium-rich foods per week and per month and how much they consumed each time. Validity (r=0.98) of the FFQ was assessed against a 1-day dietary recall method and reliability (r=0.76) was assessed on two occasions separated by 3 months [543].  131  4.2.3.3 Maturational Status At each visit, participants’ self-assessed pubic hair (in boys and girls) and breast (in girls) stage as per the methods of Tanner [1, 11]. Participants were provided with black and white line drawings representing the different pubertal stages and asked to circle the drawing that best resembled their body. For pubic hair (PH) development stage 1 indicates pre-puberty, stage 2 is considered early puberty, stages 3 and 4 a