Western University Scholarship@Western

Electronic Thesis and Dissertation Repository

8-1-2012 12:00 AM

Weight Status Underestimation among Canadian Adolescents: An Important and Frequently Overlooked Aspect of the Childhood Obesity Epidemic

Mary Ellen Kuenzig The University of Western Ontario

Supervisor Piotr Wilk The University of Western Ontario

Graduate Program in Epidemiology and Biostatistics A thesis submitted in partial fulfillment of the equirr ements for the degree in Master of Science © Mary Ellen Kuenzig 2012

Follow this and additional works at: https://ir.lib.uwo.ca/etd

Part of the Epidemiology Commons

Recommended Citation Kuenzig, Mary Ellen, "Weight Status Underestimation among Canadian Adolescents: An Important and Frequently Overlooked Aspect of the Childhood Obesity Epidemic" (2012). Electronic Thesis and Dissertation Repository. 680. https://ir.lib.uwo.ca/etd/680

This Dissertation/Thesis is brought to you for free and open access by Scholarship@Western. It has been accepted for inclusion in Electronic Thesis and Dissertation Repository by an authorized administrator of Scholarship@Western. For more information, please contact [email protected].

WEIGHT STATUS UNDERESTIMATION AMONG CANADIAN ADOLESCENTS: AN IMPORTANT AND FREQUENTLY OVERLOOKED ASPECT OF THE CHILDHOOD OBESITY EPIDEMIC

(Spine title: Weight Status Underestimation among Canadian Adolescents)

(Thesis format: Monograph)

by

Mary Ellen Kuenzig

Graduate Program in Epidemiology & Biostatistics

A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science

The School of Graduate and Postdoctoral Studies The University of Western Ontario London, Ontario,

© Mary Ellen Kuenzig 2012

i

THE UNIVERSITY OF WESTERN ONTARIO School of Graduate and Postdoctoral Studies

CERTIFICATE OF EXAMINATION

Supervisor Examiners

______Dr. Piotr Wilk Dr. Shauna Burke

Supervisory Committee ______Dr. Anita Kothari

______Dr. Greta R. Bauer ______Dr. Kathy N. Speechley

The thesis by

Mary Ellen Kuenzig

entitled:

Weight Status Underestimation among Canadian Adolescents: An Important and Frequently Overlooked Aspect of the Childhood Obesity Epidemic

is accepted in partial fulfillment of the requirements for the degree of Master of Science

______Date Chair of the Thesis Examination Board

ii

Abstract

Objectives: Overweight adolescents frequently fail to recognize that they are overweight. This project examines the magnitude of weight status underestimation among overweight adolescents and identifies predictors of this underestimation.

Methods: Data from the Canadian Community Health Survey (2001-2010) were used. Overweight adolescents (N=11,452) reporting they were underweight or about right were classified as underestimating their weight. The time trend in underestimation and effects of individual-level characteristics on underestimation were examined using logistic regression. Multilevel analysis examined the effect of weight status of community-based reference groups.

Results: For every 5 overweight male adolescents, 3 underestimated their weight; 2 of 5 overweight females underestimated. Exposure to overweight explained some of the variation in underestimation across communities among females.

Conclusions: Weight status underestimation is a significant problem among overweight adolescents. Understanding how adolescents perceive their weight is an important and novel concept in maximizing the effectiveness of current approaches to adolescent obesity.

Keywords adolescent, overweight, obesity, body mass index (BMI), perception, underestimation, multilevel modeling, Canadian Community Health Survey (CCHS), weight status, interval odds ratio (IOR), median odds ratio (MOR)

iii

Acknowledgments

I would like to begin by thanking my supervisor, Dr Piotr Wilk, for his continued guidance over the past two years. His mentorship has proven invaluable throughout this project. Dr Wilk has encouraged work beyond the thesis writing itself, supporting poster and oral presentations in Lyon, France and , —advising on the abstracts and applications for these international and national conferences. Without him, none of this would have been possible. I would also like to thank Dr Greta Bauer of my supervisory committee for her insight and encouragement along the way.

This Master’s thesis could not have been written without the professors in the Department of Epidemiology & Biostatistics. Throughout the many courses I have taken, I have gained a wealth of knowledge in all aspects of epidemiological research which allowed for the final document. I thank them all for sharing their knowledge and skill.

I extend sincere thanks to my colleagues in the Department of Epidemiology & Biostatistics. You have enhanced and broadened my skills and experiences as an individual. In addition, I will cherish the many fond and memorable moments we have shared as I go forward in life.

Financial support for this project has been provided by the Children’s Health Research Institute and the Graduate Thesis Research Fund. This funding has provided me with opportunities far beyond what I ever could have expected.

Lastly, I would like to thank my parents and extended family for the continued love and support.

iv

Table of Contents

CERTIFICATE OF EXAMINATION ...... ii

Abstract ...... iii

Acknowledgments...... iv

Table of Contents ...... v

List of Tables ...... ix

List of Figures ...... xii

List of Appendices ...... xiii

List of Abbreviations ...... xiv

Chapter 1 ...... 1

1 Introduction & Literature Review ...... 1

1.1 Introduction ...... 1

1.2 The Childhood Obesity Epidemic in Canada...... 2

1.2.1 Prevalence of Overweight and Obesity ...... 2

1.2.2 Risk Factors for Adolescent Overweight & Obesity ...... 4

1.2.3 Health Risks & Adolescent Obesity ...... 6

1.3 Weight Status Underestimation ...... 10

1.3.1 Implications of Weight Status Underestimation ...... 10

1.3.2 Prevalence of Weight Status Underestimation ...... 15

1.3.3 Causal Pathways ...... 20

1.3.4 Challenges in Comparing Studies of Weight Status Underestimation ..... 22

1.3.5 Predictors of Weight Status Underestimation among Adolescents ...... 26

1.4 Summary ...... 31

Chapter 2 ...... 33

v

2 Objectives & Hypotheses ...... 33

2.1 Objectives ...... 33

2.2 Hypotheses ...... 34

Chapter 3 ...... 36

3 Methods ...... 36

3.1 Data Source ...... 36

3.1.1 Content of the CCHS ...... 36

3.1.2 Sampling Design ...... 37

3.1.3 Study Population ...... 38

3.2 Measurement Instruments ...... 42

3.2.1 Perceived Weight Status ...... 42

3.2.2 Severity of Overweight ...... 43

3.2.3 Age ...... 43

3.2.4 Ethnicity ...... 43

3.2.5 Time ...... 44

3.2.6 Weight Status of Community-Based Reference Groups: Exposure to Overweight ...... 44

3.3 Overview of Multilevel Logistic Regression ...... 47

3.3.1 Interpretation of Multilevel Logistic Regression Analyses ...... 49

3.4 Statistical Analyses ...... 52

3.4.1 Preliminary Analyses ...... 52

3.4.2 Objective 1 ...... 52

3.4.3 Objective 2 ...... 53

3.4.4 Objective 3 ...... 55

3.5 Additional Statistical Considerations ...... 55

3.5.1 Software & Algorithms ...... 55

vi

3.5.2 Missing Data ...... 57

3.5.3 Survey Weights ...... 57

3.5.4 Interview Mode ...... 58

Chapter 4 ...... 60

4 Results ...... 60

4.1 Sample Characteristics ...... 60

4.2 Magnitude of Weight Status Underestimation ...... 62

4.3 Effect of Individual-Level Characteristics ...... 65

4.4 Multilevel Analyses ...... 75

4.4.1 Variation across Clusters ...... 75

4.4.2 Exploring the Definition of Community ...... 77

Chapter 5 ...... 82

5 Discussion ...... 82

5.1 Overview of Findings ...... 82

5.1.1 Individual-Level Characteristics of Weight Status Underestimation ...... 83

5.1.2 Weight Status of Community-Based Reference Groups on Weight Status Underestimation ...... 85

5.2 Weight Status Underestimation & Public Health ...... 86

5.3 Challenges in Studying Weight Status Underestimation ...... 88

5.4 Reasons for Weight Status Underestimation among Canadian Adolescents ...... 89

5.5 Strengths ...... 94

5.5.1 Representativeness of the CCHS ...... 94

5.5.2 Measure of Perceived Weight Status ...... 95

5.5.3 Expanded Definition of Exposure to Overweight ...... 97

5.6 Limitations ...... 97

5.6.1 Measures ...... 97

vii

5.6.2 Interview Mode ...... 101

5.6.3 Response Rates ...... 101

5.6.4 Temporality ...... 102

5.7 Conclusions and Implications ...... 102

References ...... 103

Appendices ...... 132

Curriculum Vitae ...... 165

viii

List of Tables

Table 1. Sample size and response rates by year for the Canadian Community Health Survey (CCHS) ...... 40

Table 2. Prevalence of overweight and obesity among in the Canadian Community Health Survey (CCHS) ...... 61

Table 3. Characteristics of Adolescent Respondents (aged 12 to 18) to the Canadian Community Health Survey from 2001 through 2010 (CCHS) ...... 62

Table 4. Perceived weight status of overweight adolescents from 2001 to 2010 among adolescent (aged 12-18) participants of the Canadian Community Health Survey (CCHS) ...... 63

Table 5. Logistic regression models exploring the trend in weight status underestimation among Canadian overweight adolescents between 2001 and 2010 in the Canadian Community Health Survey (CCHS) ...... 64

Table 6. Logistic regression models exploring the effect of the severity of overweight on weight status underestimation among overweight adolescents ...... 67

Table 7. Adjusted logistic regression models combining all individual-level effects (age, severity of overweight, and ethnicity) and the time trend from 2001 through 2010 ...... 68

Table 8. Logistic regression models exploring the effect of age on weight status underestimation among overweight adolescents ...... 70

Table 9. Crude odds ratios comparing weight status underestimation across ethnic groups among all overweight adolescents ...... 72

Table 10. Crude odds ratios comparing weight status underestimation across ethnic groups among male overweight adolescents ...... 73

Table 11. Crude odds ratios comparing weight status underestimation across ethnic groups among overweight female adolescents ...... 74

ix

Table 12. Variance across reference health regions and census subdivisions in weight status underestimation among all overweight adolescents ...... 75

Table 13. Variance in weight status underestimation among overweight male adolescents across health regions and census subdivisions...... 76

Table 14. Variance in weight status underestimation among overweight female adolescents across health regions and census subdivisions...... 76

Table 15. Multilevel regression models examining the effect exposure to overweight within an adolescent’s health region on weight status underestimation among male overweight adolescents ...... 78

Table 16. Multilevel regression models examining the effect exposure to overweight within an adolescent’s census subdivision (CSD) on weight status underestimation among male overweight adolescents ...... 79

Table 17. Multilevel regression models examining the effect exposure to overweight within an adolescent’s health region on weight status underestimation among female overweight adolescents ...... 80

Table 18. Multilevel regression models examining the effect exposure to overweight within an adolescent’s census subdivision (CSD) on weight status underestimation among female overweight adolescents ...... 81

Table 19. Overview of changes made to health regions in the Canadian Community Health Survey (CCHS) to ensure their consistency across time. Numbers provided represent the code assigned to each health region by Statistics Canada. The final code provided reflects all combined health regions...... 153

Table 20. Comparison of variance in weight status underestimation across census subdivisions (CSDs) with different minimum sample sizes ...... 161

Table 21. Adjusted odds ratios comparing weight status underestimation across ethnic groups among male overweight adolescents ...... 163

x

Table 22. Adjusted odds ratios comparing weight status underestimation across ethnic groups among female overweight adolescents ...... 164

xi

List of Figures

Figure 1. Predicted probability of weight status underestimation among overweight adolescents across time (2001 to 2010) in the Canadian Community Health Survey (CCHS) ...... 65

Figure 2. Predicted probability of weight status underestimation across a range of BMI z-scores. Note: Diamonds correspond to the 85th percentile; squares, 90th percentile; triangles, 95th percentile...... 66

Figure 3. Predicted probability of weight status underestimation for overweight adolescents between the ages of 12 and 18 ...... 69

xii

List of Appendices

Appendix A. Overview of prior studies examining weight status underestimation among overweight and obese adolescents using a Likert-type question to measure weight status underestimation. Note that estimates for the magnitude of weight status underestimation are for overweight adolescents unless otherwise stated...... 133

Appendix B. Detailed list of variables from the Canadian Community Health Survey used in the analysis of this project ...... 150

Appendix C. Overview of changes made to health regions to ensure their comparability across cycles of the Canadian Community Health Survey ...... 152

Appendix D. Comparison of minimum sample size per cluster for multilevel analysis ...... 160

Appendix E. Comparison of weight status underestimation across ethnic groups adjusted for age, severity of overweight, and the effect of time, as well as interview mode ...... 162

xiii

List of Abbreviations BMI Body mass index CAI Computer-assisted interviewing CAPI Computer-assisted personal interviewing CATI Computer-assisted telephone interviewing CI Confidence interval CCHS Canadian Community Health Survey CDC US Centers for Disease Control and Prevention CSD Census subdivision ICC Intraclass correlation coefficient IOR Interval odds ratio IOTF International Obesity Task Force MLR Robust maximum-likelihood estimation MOR Median odds ratio OR Odds ratio RR Relative risk SD Standard deviation SE Standard error PCV Proportional change in variance WHO World Health Organization

xiv 1

Chapter 1 1 Introduction & Literature Review 1.1 Introduction

Child and adolescent overweight and obesity are increasingly important public health problems, with substantial increases in the prevalence of both in Canada1,2 and around the world.3 Overweight children and adolescents face an increased risk of developing chronic diseases, including cardiovascular disease and type II diabetes. These diseases are being diagnosed at younger ages than ever before.

Current public health strategies aim to raise awareness of the obesity epidemic and help mitigate the effects of increasing weight on health. Despite this increased attention, a large proportion of overweight adolescents fail to recognize that they are overweight.4-60 Instead, they think of themselves as being normal weight, and in some cases, underweight. Overweight adolescents who fail to recognize that they are overweight tend to be less motivated to lose weight61-63 and subsequently do not engage in weight management behaviours.8,12,14,15,17-19,31,32,44,61,64,65 As a result, these overweight adolescents face an increased risk of developing weight-related comorbidities later in life. Further, individuals who underestimate their weight status may also have poorer physical health than those who accurately recognize that they are overweight. A failure to recognize oneself as overweight, however, may prove beneficial for other aspects of adolescents’ well-being. In particular, adolescents who accurately identify themselves as being overweight typically have poorer mental health and psychosocial well-being than those who do not recognize that they are overweight.

Addressing weight status underestimation has been identified as an important next step in addressing the obesity epidemic.66 However, because of both the positive and negative implications of weight status underestimation, this is a challenging task. With this in mind, it is important that we identify characteristics of overweight Canadian adolescents who are most likely to underestimate their weight status. A better understanding of the factors associated with weight status underestimation will enable us to better target

2 healthy weight promotion strategies towards adolescents most likely to underestimate their weight status. It will also provide key information to be used in the design of strategies that promote the concept of healthy weight, including which adolescents these programs are targeted towards. These approaches will not only promote the adoption of accurate weight status perception, but also assist these adolescents in understanding the health risks associated with their body weight and steps they can take to help achieve a healthy weight. If these individuals continue to underestimate their weight status, they will continue to be at an increased risk of developing other health-related complications. At the same time, any programs aimed at addressing accurate perception of weight status must be designed in such a way as to protect the mental health and psychosocial well- being of these adolescents.

The remainder of this chapter provides an overview of the obesity epidemic among Canadian children and adolescents (Section 1.2), including risk factors for the development of overweight during childhood and adolescence (Section 1.2.2) and the health risks associated with being overweight during youth (Section 1.2.3). A discussion of weight status underestimation is provided in Section 1.3, including the implications of weight status underestimation (Section 1.3.1). An overview of current estimates for the degree of weight status underestimation is provided in Section 1.3.2. Potential pathways that lead to weight status underestimation are then outlined in Section 1.3.3. The chapter concludes with a discussion of the challenges in comparing studies of weight status underestimation (Section 1.3.4) and some of the factors that may be associated with weight status underestimation among Canadian adolescents (Section 1.3.5), including sex, actual weight status, age, ethnicity, and the weight status of an adolescent’s community.

1.2 The Childhood Obesity Epidemic in Canada

1.2.1 Prevalence of Overweight and Obesity

Recent evidence has pointed to a change in the distribution of body mass index (BMI) among Canadian adolescents. In 2004, the average BMI of Canadians between the ages of 12 and 17 was 22.1 kg/m2—1.3 kg/m2 higher than the mean BMI for adolescents in

3

1978/1979.1,2 An increase in the number of obese adolescents is largely to blame for this increase in mean BMI—a phenomenon that shifts the distribution of adolescent BMI towards the heavy end of the spectrum.2

As is suggested by the increasing average weight of Canadian adolescents, the prevalence of overweight and obesity among Canadian children and adolescents is rising rapidly. In 2004, it was estimated that 26% of Canadians between the ages of 2 and 17 were either overweight or obese—an increase of 70% since 1978/1979.1,2 Specifically, the percentage of Canadian children and adolescents who were overweight increased from 12% to 18%; the prevalence of obesity almost tripled, increasing from 3% to 8%. This increase in prevalence has been particularly pronounced among adolescents.1,2 Among those aged 12 to 17, the prevalence of overweight doubled, increasing from 14% to 29%. The prevalence of obesity in this age group tripled, increasing from 3% to 9%. An estimated 1.1 million Canadians between the ages of 2 and 17 were overweight in 2004; an additional half million were obese.1

Another national study using data collected in 2001 reported that the prevalence of overweight among Canadians aged 11 to 16 was 15.0% (95% CI 13.9-16.1); 4.6% (95% CI 4.0-5.2) were obese.67 Among boys, 18.3% (95% CI 16.5-20.1) were overweight, while 5.8% (95% CI 4.8-6.8) were obese. The prevalence of overweight among females was 13.3% (95% CI 11.8-14.8); 3.5% (95% CI 2.7-4.3) were obese.

Comparisons of the prevalence of overweight and obesity reported in different studies are complicated by both the lack of a standard definition for overweight and obesity and the use of different anthropometric measures.68 The differences in the estimates of the prevalence of overweight and obesity above likely result from how overweight and obesity were measured. The data collected in 2004 used measured height and weight to calculate an adolescents’ BMI,1,2 while the study conducted in 2001 based estimates on BMI calculated from self-reported height and weight.67 Despite the differences in the methodology used, it is clear that the prevalence of overweight and obesity among Canadian adolescents has reached alarming levels.68

4

Geographic Differences

There are substantial disparities in the prevalence of child and adolescent overweight and obesity across regions of Canada.2,69 The highest rates are observed in Atlantic Canada, with combined rates of overweight and obesity in 2004 being 36% in Newfoundland, 34% in New Brunswick, and 32% in Nova Scotia.2,70 The prevalence of overweight and obesity among adolescents in Prince Edward Island (30%) was not significantly different than the national average (26%). Obesity alone was also significantly more prevalent in these provinces than the national average (8%). The prevalence of overweight and obesity in Manitoba (31%) is also significantly higher than the national average (26%). While Quebec (23%) and Alberta (22%) both have combined overweight and obesity rates significantly below the national average, the rate of obesity in both provinces (7% in Quebec and 8% in Alberta) remains similar to what is observed nationally (8%).70

Similar differences across regions of Canada were observed in 1996.71 A child living in an Atlantic province was more likely to be overweight than a child living elsewhere in Canada (OR 1.45, 95% CI 1.28-1.65). Children living in a Prairie province (Manitoba, Saskatchewan, and Alberta) were less likely to be overweight or obese than children living elsewhere in Canada (OR 0.69, 95% CI 0.60-0.78). These regional differences persisted after controlling for child and family characteristics (i.e. income, parental education, and number of siblings).

Increases in the prevalence of overweight and obesity have been observed in all Canadian provinces; however, the magnitude of these increases is not distributed uniformly across the country.71 Between 1981 and 1996, the Atlantic provinces (Newfoundland, Nova Scotia, Prince Edward Island, and New Brunswick) saw a significantly greater increase in the prevalence of overweight and obesity than the rest of Canada, while the prairie provinces saw a slower increase.71

1.2.2 Risk Factors for Adolescent Overweight & Obesity

Obesity is defined as the presence of excess fat and results from a broad spectrum of risk factors that work together to influence energy balance, ultimately leading to the

5 accumulation of excess body fat.72-74 These factors include, but are not limited to, genetic susceptibility72; pre-existing medical conditions such as Prader-Willi, Bardet- Biedl, Alstrom, and Cohen syndromes75; socioeconomic status71,76-80; early life characteristics such as high birthweight and rapid weight gain during the first year of life,81-84 as well as being bottle-fed72,82,85-87; energy consumption and expenditure habits3,35,67,72,74,75,78,81,88-92; and socio-environmental factors3,75 including family characteristics,71,76,82,85,86,93,94 neighbourhood safety78,95,96 and socioeconomic status.97 In addition, the prevalence of overweight is higher among males than females78,80,97-99 and varies across ethnic groups. Although little research has examined the association between ethnicity and adolescent obesity in the Canadian context, Aboriginal adolescents in Canada are more likely to be overweight than their non-Aboriginal peers.100-102 In the United States, adolescents of Asian background are the least likely to be overweight, while those who are Black have the highest rates of overweight and obesity.78,103,104 Further, the influence of immigration status on overweight and obesity among Canadian adolescents is not well understood among Canadian adolescents. Outside Canada, the association between immigrant status and overweight also remains unclear—some studies suggest immigrant children are at increased risk of being overweight while others suggest the opposite.103,105

The remainder of this discussion focuses on risk factors that can be modified at the individual level—these are the behaviours that adolescents can change when they are aware that their weight is problematic. These individual-level risk factors are focused around the idea of energy balance and the behaviours that influence energy consumption and expenditure. When caloric consumption exceeds energy expenditure, there is a positive energy balance leading to the development of obesity.

The changes in energy balance result from both an increase in energy consumption and a decrease in energy expenditure. Since 1987, there has been a substantial increase in soft drink consumption.72 The effect of this increasing consumption on the increase in prevalence of overweight is made evident through significant associations between soft drink consumption and childhood obesity.67,72,74,75 However, not all studies have shown that increased soft drink consumption is associated with overweight.76 Increased

6 consumption of diet soft drinks has also been linked to higher levels of overweight among adolescents.74 Evidence has also suggested that increased fast food consumption is also related to the increase in overweight and obesity but not to the same extent as soft drink consumption.72 Increased portion sizes may also play a role.72,75 The relationship between adolescent overweight and obesity and energy consumption is not necessarily a straightforward one. Eating breakfast regularly and having high fruit and vegetable consumption decreases adolescents’ risk of being overweight.35,67,74 Dieting, binge eating, and engaging in other unhealthy weight control behaviours are also associated with a higher probability of being overweight.74,88

This excess of energy consumption is coupled with a decrease in energy expenditure, particularly through increased engagement in sedentary behaviours. Together, these have led to an overall energy imbalance in favour of overconsumption of energy. Decreases in energy expenditure resulting from increases in sedentary behaviours may be to blame for the majority of the increase in overweight and obesity.89 Children engage in physical activity less often, instead spending time engaging in sedentary activities including watching TV and playing video games.72 Consequently, those who engage in sedentary behaviours are at a much higher risk of being overweight than their active peers.67,72,74,75,78,81,90-92 The relationship between sedentary activity and obesity is independent of the effects of physical activity on health.106 The relationship between increased participation in physical activity and decreased risk of overweight is not as clear.67,72,74,76,78,90,98

1.2.3 Health Risks & Adolescent Obesity

This section provides a general overview of the relationship between overweight and obesity and adolescents’ health and well-being. Overweight children and adolescents face several challenges with regard to their physical, emotional, and social well- being.107,108 The impact of childhood obesity begins at a young age and persists into adulthood.67,109-115 Childhood obesity is also an important predictor of adult obesity86,116- 119 and this persistence of excess weight may explain the continued effects of excess risk into adulthood.113 The health risks faced by adolescents with severe obesity may be particularly pronounced.120-122 If current obesity trends continue unchanged, this

7 increasing weight could potentially outweigh the increases in life expectancy that have been achieved through smoking reduction.123

Physical Health

Overweight adolescents report that, in general, their health is poorer than their normal weight peers.98,124 Risk factors for cardiovascular disease (including blood pressure, altered lipid metabolism), along with insulin resistance, are commonly observed among overweight and obese adolescents.75,117,121,122,125-131 Clustering of these risk factors is common among overweight and obese adolescents75,117,121,125-128,131; together, these risk factors comprise the metabolic syndrome. Being overweight during childhood and adolescence increases the risk of developing cardiovascular disease as an adult,67,111-114 independent of adult overweight status.112-114 In addition, obesity is associated with insulin resistance131 and glucose intolerance75 among youth and is an important risk factor for the early onset of type II diabetes.117,121,132 Adolescent obesity and weight gain from adolescence to adulthood are both independent predictors of type II diabetes later in life.111,115 Overweight and obese children are also at increased risk of musculoskeletal complications,117,126,133 liver disease,117,126,134 early onset of puberty,117 sleep apnea and other forms of sleep-disordered breathing,75,117,135 chronic inflammation,75 and asthma.75 Minor injuries, particularly sprains of the lower extremities, are more common among overweight136 and obese adolescents.137,138 Overweight and obese adolescents are also more likely to report functional limitations than their normal weight peers.124

Psychosocial Well-Being

In addition to the impact of overweight on adolescents’ physical health, it is also important to consider the relationship between adolescent overweight and mental health. However, the causal relationship between obesity and mental health is not straightforward. Child and adolescent obesity can lead to poor psychological well-being; at the same time, poor psychological well-being can contribute to the development and persistence of obesity.117,139-141 This relationship is further complicated by evidence suggesting that weight itself is not associated with poorer mental health, but how an individual perceives their weight and their subsequent weight concerns.39,49,139,142-150 The

8 psychosocial impact of obesity is more pronounced among females30,142,151-154 and overweight adolescents seeking treatment for their weight.139,141,144,152,155 It has been hypothesized that these effects largely arise from weight-based teasing and stigma towards obesity—something that begins at an early age.117,141

Overweight adolescents describe themselves as being ‘unhappy’ more often than their normal weight peers.98 Adolescents with an obesity-related diagnoses (i.e. type II diabetes) are more likely to have a psychiatric diagnoses than are adolescents with other chronic conditions.156 Having a high BMI leads to the development of low self- esteem,39,43,51,109,124,139,151-153,157-162 depression,7,26,39,49,80,109,124,128,139,143,148,159,162-164 and negative self-image and body dissatisfaction.98,109,110,144,152 Suicidal behaviours are also more common among overweight and obese adolescents than among their normal weight peers.28,146,147

Overweight and obese individuals experience higher rates of ADHD.126,139,165,166 Binge eating and other eating disorder symptoms are also relatively common in the overweight population117,151,167 and are frequently associated with higher levels of anxiety and depressive symptoms.151 Further, because children who are overweight or obese tend to be taller than their normal-weight peers, adults often think of them as being older than they actually are and expect a higher level of maturity.117 This may lead to these children feeling socially isolated.117 Overweight and obese adolescents are further marginalized by their peers168,169 and experience difficulties making friends.64 This stigmatization may be a more important predictor of obesity-related morbidity and mortality than obesity itself.170 This stigmatization is more common among overweight females and this may help explain the disproportionate degree of obesity-related morbidity and mortality faced by females.170-172 These psychosocial effects cluster in overweight and obese children.152

Overweight and obese adolescents also have a lower perceived quality of life than their normal weight peers.30,107,108,124,141,152,173-175 This is especially true with respect to coping,173 physical health and social functioning,108,124,139,141,174,175 and school-related domains.124,139,141,175 Obese adolescents being treated in a clinic scored similar to cancer patients with regard to their quality of life.107 In addition, overweight adolescents

9 experience poorer academic achievement than their normal weight peers23,64,117,124,139,159,167,176,177 and frequently require extra academic assistance.141

Studies have suggested that overweight adolescents are more likely to engage in risky behaviours, including smoking, using dieting pills, gambling, binge drinking, and are more likely to have eating disorders, when compared to normal weight adolescents.149,167 This association may be mediated by the presence of other psychosocial risk factors, such as low self-esteem160,167 and perception of overweight.149 However, substance abuse and dependence diagnoses are not more prevalent among obese adolescents than their normal weight peers.153

Health Care Utilization & Costs

The combination of these physical and psychosocial impacts of being overweight or obese during adolescence culminates in the use of the health care system by overweight and obese adolescents. An extra $14.1 billion are spent annually in the United States on the health care for children and adolescents between the ages of 6 and 19 who are overweight and obese.178 This includes the extra costs of outpatient visits, prescription drugs, and emergency room visits. These children not only visit their family physicians or paediatricians more frequently,99,179 but also have more visits to specialists,99 mental health services,99 and emergency departments.178 They are also admitted to hospital more often, and when hospitalized, have longer hospital stays than children who are not overweight.179 Overweight children, as well as those with both diagnosed and undiagnosed obesity, use more laboratory services than do normal weight adolescents.180 Overweight and obese adolescents also have increased medication costs, both prescription and non-prescription, than normal weight adolescents.179

Overweight and obese children in the fifth grade in Nova Scotia (N=4,380) had higher total health care costs than did normal weight children in the province.181 This included lifetime physician cost, number of visits to primary health care providers, and specialist referrals. Each overweight child cost an extra $156 per year, while an obese child costs an extra $349.181 In another study conducted among adolescents in Ontario, those who were overweight were not found to use excess physician costs compared to normal

10 weight adolescents182 This study calculated physician costs from physician billing data obtained from the Ontario Health Insurance Plan (OHIP) and linked with the Canadian Community Health Survey (CCHS). It was suggested that, although there were no differences for Ontario adolescents, the impact of their overweight on the health care system would be observed in the future since obesity persists from adolescence to adulthood.182

1.3 Weight Status Underestimation

The section on weight status underestimation first addresses the implications of weight status underestimtion to health and well-being (Section 1.3.1), followed by an overview of previous studies examining the magnitude of weight status underestimation among adolescents (Section 1.3.2). Section 1.3.3 provides a discussion of possible pathways that lead to weight status underestimation. A discussion of the challenges in making comparisons across studies of weight status underestimation are provided in Section 1.3.4. To close out this section on weight status underestimation, a discussion of risk factors for weight status underestimation is provided (Section 1.3.5).

1.3.1 Implications of Weight Status Underestimation

The focus of the current literature on weight status underestimation is on the relationship between underestimation and engagement in weight management behaviours. Engagement in these behaviours subsequently improves physical health, decreasing the risk of developing obesity-related comorbidities. For example, a recent systematic review found that physical activity alone can signficantly decrease the likelihood of an overweight adolescent developing diabetes.183 Despite the important role of lifestyle modifcations in minimizing the downstream risk of poor physical health, these interventions tend to have limited effectiveness in minimizing the risks of overweight outside clinical settings.184 Lack of motivation to lose weight is a frequently cited reason for the failure of these weight management programs.184 Overweight adolescents who do not recognize that they are overweight may not be aware of the increased health risks they face (i.e. developing obesity-related pathology, including cardiovascular disease, diabetes, and adult obesity). As a result, they are subsequently less likely to engage in

11 weight management behaviours than are those who accurately recognize that they are overweight. Engaging in weight management behaviours may mediate the relationship between underestimation of weight status and poor physical health. There may also be a direct relationship between weight status underestimation and risk of negative health outcomes. For example, obese adults participating in the Dallas Heart Study (N=2,056) who underestimated their weight status were more likely to have hypertension than those accurately recognizing that they are overweight.66 Among diabetic individuals, those who underestimated their weight status were also less likely to report being aware of their disease. In addition, those who underestimated their weight status thought of themselves as healthier and at a lower risk of having a heart attack, diabetes, or high blood pressure than those who did not underestimate their weight status.

Although weight status underestimation may lead to poorer physical health outcomes, adolescents who do not recognize that they are overweight exhibit better psychosocial well-being compared to those who accurately recognize that they are overweight. Underestimating one’s weight status has protective effects for the mental health of overweight individuals, and consequently must be taken into consideration when developing strategies targeting weight status underestimation among overweight adolescents. The role of engagement in weight management behaviours in understanding the relationship between weight status underestimation and improved psychosocial well- being is not yet understood and is an important area of future research.

Engagement in Weight Management Behaviours

Several models of behaviour change can be used to explain the relationship between perceived weight and motivation to engage in weight management behaviours. As examples, the Health Belief Model and the Precaution Adoption Process Model are discussed. These models share a common basis in that they emphasize the importance of perception and awareness of risk on engaging in weight management behaviours. The Health Belief Model in particular points to the importance of “perceived susceptibility” to health problems as a crucial mediating factor for engagement in preventive health behaviours.185,186 This model suggests that overweight and obese adolescents who do not

12 recognize that they are overweight consequently may not be aware of their increased susceptibility to the health risks associated with being overweight. As a result, they may not have the motivation required to engage in weight management behaviours.

The Precaution Adoption Process Model is a stage theory of behaviour change and posits that individuals are first unaware of the risks associated with their behaviour (i.e. being overweight); they then move into being aware of their weight but having no plans to engage in any weight management behaviours.187 Adolescents in the final stages of this model are actually engaging in weight management behaviours. The applicability of this model to the relationship between weight status underestimation and engagement in weight management behaviours is similar to the Health Belief Model. That is, adolescents who do not recognize that their current weight-related behaviours are increasing their risk of developing weight-related diseases are unlikely to engage in behaviours that will help to mitigate this risk. For these overweight adolescents to appropriately engage in weight management behaviours, they need to first perceive that they are overweight and that their current weight increases their risk of developing obesity-related comoribidities, such as diabetes and cardiovascular disease.

Several studies have tested the applicability of these models to the relationship between weight status underestimation and adolescent engagement in weight management behaviours. Adolescents accurately perceiving that they are overweight are not only more motivated to lose weight61-63 but are also significantly more likely to engage in weight management behaviours, than those who inaccurately perceive their weight as normal or underweight.8,12,14,15,17-19,29,31,32,44,50,52,58,60,61,64,65 This association is likely mediated by perceived pressure to lose weight.61 However, overweight adolescents, especially males, who underestimate their weight status are more likely to engage in physical activity than those who accurately perceive themselves as overweight.18,58 Similarly, studies have shown that females who accurately perceive that they are overweight are more likely to be inactive than those who underestimate their weight status.32 Not all studies have found an association between weight underestimation and weight management behaviours among overweight and obese adolescents.20,188 It is also important to note that adolescents facing increased pressure to lose weight, both in the

13 form of parental concerns about weight and weight-related teasing, are at an increased risk of overweight.74,88 This is likely the result of increased body dissatisfaction and weight concern, both of which can lead to the development of overweight.74,88 Peer dieting and parent overweight (as reported by the adolescent) were not found to be associated with overweight among females, but peer dieting was found to be associated with overweight among males.74

Some overweight adolescents recognize the need to lose weight without necessarily recognizing that they are overweight.14,16,21,35,52 Although levels of weight status underestimation appear to be relatively stable across time, data from Finland suggest that overweight adolescents are increasingly engaging in weight management behaviours.38 The prevalence of weight control behaviours among overweight adolescent males increased from 3% in 1994 to 18% in 2006; the number of overweight female adolescents engaging in weight management behaviours doubled, increasing from 19% in 1994 to 39% in 2010. Although the proportion of adolescents engaging in weight management behaviours increased between 1994 and 2010, the differences across years were not significant for females and only the difference between 1994 and 2006 was significant for males. There are, however, overweight and obese adolescents who report wanting to weigh more than they currently do.35 For example, a study conducted in Australia found that among those who are obese, 6.8% reported that they wanted to be heavier than their current weight; this includes 2.4% of obese respondents who reported they wanted to be a lot heavier than their current weight.35 Overweight adolescents reported that they wanted to be heavier than they currently are 5.0% of the time. Making effort to gain weight was reported by 2.6% of overweight and 2.6% of obese children and adolescents.35

When being asked to select one of seven body figures that best described their current weight, Portuguese adolescents engaging in dieting behaviours were more likely to select a larger body figure than did individuals not engaging in dieting behaviours, controlling for BMI, sex, age, and perceived weight status.64 A large gap in weight management behaviours even exists among overweight adolescents diagnosed with type II diabetes,45 an effect possibly mediated by high levels of weight status underestimation in the sample

14 of type II diabetes patients included in the study. Adolescents who accurately reported their weight status reported fewer barriers to engaging in these behaviours.45

Accurately recognizing one’s weight as overweight also increases the probability that adolescents engage in unhealthy weight management behaviours.15,18,19,29,32,44,50 As such, a cautionary note is required when discussing weight management behaviours among those who accurately perceive that they are overweight. These individuals are more likely to engage in unhealthy weight management behaviours,32,50 such as skipping breakfast,44,50 fasting (males only),19,29 purging,15,18,29,50,189 use of diet pills15,19,50 and laxatives18,50,189 than those who fail to recognize that they are overweight.

Psychosocial Well-Being

Perceived weight status also plays an important role in understanding an individual’s psychosocial well-being. As discussed in Section 1.2.3 above, the effect of overweight on psychosocial well-being appears to be mediated by how an overweight individual perceives their weight. The effect of perceiving one’s self as overweight on psychosocial well-being may be mediated by body dissatisfaction9 and is stronger for females than for males.7,162 Adolescents who recognize that they are obese report being the victim of bullying more frequently than those who fail to recognize their obesity status; reports of bullying are higher among males than females.32,150 Adolescents who report weight- based teasing subsequently perceive that they are under increased pressure to lose weight61 and consequently experience poorer psychosocial well-being. Perceived overweight is associated with feelings of stress,54,190 lower self-esteem,9,39,42,162 decreased body appreciation,191 poorer quality of life,30,61 poorer social adaptation,49,192 social isolation,56 poorer academic performance,23,54 behavioural problems,49,192 and increased risk of depression and other emotional problems.7,9,16,24,26,32,48,49,54-56,143,148,162,163,189,192 Underestimation of weight status is also protective against engagement in suicide behaviours,28,32,146-148,189 especially among females.193 The association between weight status underestimation and suicide ideation does not persist when controlling for all weight-related attitudes, including body dissatisfaction; this suggests that this relationship is mediated by weight dissatisfaction.28

15

Risk-taking behaviours are more common among those who are overweight, with female adolescents who underestimate their weight status being more likely to engage in risky sexual behaviours than those who do not underestimate their weight status.5 This includes being less likely to use oral contraceptives and being more likely to have multiple sexual partners.5 However, other studies have not found this same relationship between sexual behaviours and weight status underestimation.189 Adolescent smoking, marijuana use, and alcohol consumption are not associated with weight status underestimation.189 Other studies have found that a desire to control weight is associated with increased smoking among adolescents.149,194

The majority of studies examining the relationship between adolescent mental health and perceived weight compare those who perceive their weight as overweight to those who do not. Whether or not studies control for actual weight in the analysis, there is a clear negative effect of perceiving oneself as overweight on psychosocial well-being. Those adolescents who perceive their weight status as normal tend to have the best mental health, comparing to those who perceive their weight as overweight.

1.3.2 Prevalence of Weight Status Underestimation

Overweight adolescents frequently fail to recognize that they are overweight. For example, previous studies conducted in Canada have found that more than 35% of overweight adolescents underestimate their weight status.13,195 Adolescent perception of weight status has a high specificity but a low sensitivity for predicting actual weight status; the sensitivity is especially low for males.52,196 This means that the majority of adolescents reporting that they are overweight are in fact overweight, while a large percentage of those who are actually overweight report being normal weight. The agreement between actual and perceived weight status is poor to moderate for both males and females.15,52,57

The degree of weight status underestimation across studies varies greatly (Appendix A). The highest rates of weight status underestimation are typically found in the United States and Australia.4,5,9-11,13,16-20,22-24,26,32-37,40-43,45,51,52,57-60 Studies conducted in European and Asian countries tend to have lower rates of weight status underestimation

16 among overweight and obese adolescents.6,12,14,15,21,25,27-31,38,39,44,46,48-50,53-56,59,60 In the United States, estimates of weight status underestimation typically range from around 40% to 50% for males and from 15% to 30% for females among overweight and obese adolescents. The highest level of weight status underestimation previously reported was 85.7% for overweight and 54.7% for obese males; 69.1% for overweight and 38.2% for obese females.11 The data for this study came from a nation-wide methodological study conducted as part of the Youth Risk Behavior Survey in the United States. The sample used in this study had a higher prevalence of overweight and obesity than that of the general American adolescent population. This may have contributed to the high level of underestimation observed in this study.11 It is also important to note that this study relied on a convenience sample and was consequently not representative of the population of American adolescents. In contrast, a study of Dutch adolescents found that only 1.5% of those who were overweight underestimated their weight status; the prevalence of overweight among these adolescents was very low by comparison (6%).12

Despite the plethora of research studies examining the prevalence of weight status underestimation globally, there has been very little research conducted on this issue in the Canadian population. Further, what causes adolescents to underestimate their weight status is not well understood, particularly among Canadian adolescents. The results of the few studies examining weight status underestimation among Canadian adolescents are highlighted below. Other notable findings from the previous studies examining weight status underestimation presented in Appendix A are also discussed.

Two studies have examined weight status underestimation in Canada. One utilized data available from the Quebec Child and Adolescent Health and Social Survey.195 The other was conducted among attendees of a gastroenterology clinic in Hamilton, Ontario whose health concerns were not related to their weight.13

The study conducted in Hamilton, Ontario occurred in 2005 and included only 53 patients between the ages of 12 and 18.13 When asked to describe their weight status as underweight, slightly underweight, average, slightly overweight, or overweight, 44.0% (11/25) of all males and 35.0% (7/20) of females underestimated their weight status,

17 relative to measured height and weight. Normal weight adolescents perceiving themselves as underweight were included with overweight adolescents who underestimated their weight status. Participants in this study were also asked to identify their weight status using a visual scale developed by the study’s authors; similar results were observed for both measures of perceived weight status.

The second Canadian study used data collected in 1999 as part of the Quebec Child and Adolescent Health and Social Survey.195 It involved a provincially representative sample of Quebec children and adolescents, aged 9, 13, and 16. Over 1000 students in each age group participated. This study used a visual tool to measure weight perception. Participants were presented with a series of seven body figures ranging from underweight to overweight. Each figure was assigned a BMI z-score (i.e. -3 through +3). A BMI z- score was also assigned to each individual based on their measured height and weight (using growth curves provided by the US Centers for Disease Control and Prevention; CDC). The two z-scores were compared. An individual with a negative misperception z- score (i.e. perception z-score < BMI z-score) was considered to underestimate their weight status. Among overweight adolescents, 71.4% underestimated their body size (i.e. had a misperception z-score between -1 and -3); 59.4% of obese adolescents underestimated their weight status.

In addition to the two studies of Canadian adolescents, a similar study examined weight status underestimation among adults living in the Canadian province of Alberta in 2004.197 Perceived weight status was measured by asking participants if they would describe themselves as underweight, about the right weight, or overweight. Their response was compared to both their BMI and waist circumference. Overweight and obese males and females were accurate in describing their weight status 83% of the time. Among females deemed to be overweight or obese based on their BMI, 93.0% (95% CI 92.0-94.0) accurately perceived themselves as overweight; 71.4% (95% CI 69.6-73.3) of males were overweight or obese. When comparing weight status underestimation to actual weight status by waist circumference, 93.1% (95% CI 91.8-94.4) of high risk females and 86.7% (95% CI 84.6-88.8) of high risk males accurately perceived themselves as overweight. Those at high risk by both waist circumference and BMI were

18 the most accurate in their perceptions: 95.7% (95% CI 94.6-96.8) of females and 87.4% (95% CI 85.4-89.4) of males accurately perceived their weight status. The results of this study may have been biased since participants were asked to measure their height, weight, and waist circumference as part of the questionnaire. Because of these instructions, participants may have been more aware of their weight status than if they had not measured and weighed themselves. Further, the high degree of accuracy observed in this study compared to the studies involving adolescents may have been the result of a cohort effect, with adults in this study growing up in an era prior to the rapid increases in overweight and obesity. This contrasts with the studies of adolescents who have grown up in an era dominated by the increasing rates of overweight and obesity.

It is important not only to consider adolescents’ perceptions of their own weight, but also their perceptions of how their parents and peers view the adolescents. Among males, 64.3% of overweight and 17.6% of obese adolescents believed that their parents thought of them as being normal weight; 39.4% of overweight and 9.5% of obese females believed the same.6 Among overweight males, 70.4% reported that their peers perceived them to be normal weight; 3.3% of obese males reported that their peers considered them to be underweight.6 An additional 30.0% of obese males thought their friends considered them normal weight.6 For females, 59.5% of overweight and 21.1% of obese adolescents believed their peers considered themselves to be normal weight; none believed their peers thought of them as underweight.6 In this same study, 60.0% of overweight and 13.9% of obese males perceived their current weight status to be either underweight or normal weight, while 36.7% of overweight and 4.3% of obese females underestimated their weight status.6 In general, levels of weight status underestimation are higher when the comparison is made to how others perceive one’s own weight.

Trends in Weight Status Underestimation

Although there are no Canadian studies that examine time trends in weight status underestimation, some studies have examined these trends in other populations. Overall, the results of these studies are inconclusive. Despite increasing media and public health attention focused on the obesity epidemic, there has been no change, from 1999 to 2007,

19 in the proportion of American overweight and obese adolescents accurately recognizing that they are overweight.22 Although not significant, it is important to note that the prevalence of weight status underestimation among overweight females has increased slightly; in contrast, males have become slightly better at recognizing that they are overweight.22 Another study compared levels of weight status underestimation among American adolescents between 1999 and 2010.34 This study found that the proportion of overweight males accurately perceiving that they were overweight remained unchanged. There was an increase in the proportion of overweight females who accurately perceived that they were overweight.

Exploration of trends in weight status underestimation among ethnic subgroups reveals that overweight Black male adolescents are becoming more accurate in recognizing that they are overweight.22 While the proportion of overweight adolescents recognizing that they are overweight has remained relatively stable, the overall prevalence of overweight is increasing.22 This suggests that there is an increase in the overall number of overweight adolescents who fail to recognize that they are overweight.22

A study of overweight Finnish adolescents similarly found that levels of weight status underestimation remained unchanged between 1994 and 2010.38 However, in a different study of Finnish adolescents, increasing levels of weight status underestimation were observed between 1979 and 1999.27 Overall, the number of overweight adolescents who underestimated their weight status increased.27

The proportion of overweight Spanish males (aged 20 and older) who failed to recognize that they were overweight remained relatively stable between 1987 and 2006/2007; an increasing number of females failed to recognize that they were overweight.198 Between 1995 and 2006/2007, there was an 8% increase in the proportion of parents failing to recognize their child’s overweight status in both males and females (aged 5 to 15). Weight status in children was determined using IOTF reference values and parent- reported height and weight.

20

1.3.3 Causal Pathways

A Shift in Societal Norms

Evidence has suggested that overweight is becoming the new ‘normal.’199 Although the weight status underestimation appears to be relatively stable across time, given that there is an increasing number of overweight individuals, the overall number of overweight adolescents who underestimate their weight status may actually be increasing.22 As the prevalence of overweight continues to increase, these trends emphasize the importance of societal characteristics (i.e. the average weight of individuals in one’s community) in influencing our own weight perceptions and consequently our weight. This concept can be explained using the Theory of Endogenous Weight Norms.200 This theory posits that people want to weigh less than the average individual and is based on two general concepts: (1) an individual’s preference to be thinner than the average person; and (2) that individuals tend to compare themselves to others. This theory can thus be summed up by recognizing that, as people become heavier, ideal weight also becomes heavier. That is, individuals prefer to be thinner than the average (relative weight), but the actual weight this preference corresponds to is increasing (absolute weight). As a result of these changing preferences, Burke and Heiland200 hypothesize that a normalization of a heavier ideal weight may be an important contributing factor in the current obesity epidemic. Perceptions of one’s weight are becoming increasingly based on subjective rather than objective criteria for overweight.199,200 This includes adolescents increasingly comparing themselves to their peers and less to external sources, such as the media.27 This theory is supported by evidence from a study conducted by Maximova et al195 finding that adolescents exposed to overweight at school and at home (i.e. parental overweight) were at an increased risk of underestimating their weight status.

As an alternative to the Theory of Endogenous Weight Norms, Neighbors et al201 has suggested two competing hypotheses that may explain the adoption of overweight as a new normal: individuals are comparing themselves to those in their environment and those individuals are becoming increasingly overweight. Alternatively, women may be becoming more accepting of larger body sizes. These hypotheses are based on the Social Comparison Theory.

21

All of this has led to an increasing threshold for overweight.199 That is, the cut point or reference value individuals use to decide whether or not they are overweight is increasing. It has been hypothesized that the current rates of childhood obesity may increase this threshold even further, resulting in further misclassifications of overweight status among those who are overweight.199 While the Theory of Endogenous Weight Norms and the hypotheses proposed by Neighbors et al201 are attractive explanations for this change, other possible explanations cannot be ruled out. These include the influence of the media and the effects of public health campaigns.199

Social Contagion of Obesity

It is evident that there is a relationship between exposure to overweight and risk of weight status underestimation, and that this relationship may play a role in the further propagation of overweight. This is supported by evidence suggesting that there is a contagious component to the obesity epidemic. Blanchflower et al202 found that there is evidence of a spread of obesity across European adults. Further, who someone compares his or her weight to depends on his or her sociodemographic characteristics. For example, those most highly educated compare themselves to others who are also highly educated, a group of the population that tends to be thinner. Thus, individuals demonstrate social comparison of their weight status to their peers. These comparisons are based on relative measures of weight status rather than absolute measures.

Trogdon et al169 found similar results among American adolescents. There was a high correlation between an individual’s weight and the mean weight of his or her peers, after controlling for demographic characteristics, smoking, birth weight, and parental and household characteristics. Adolescents whose parents have a high BMI are also more likely to be overweight. Trogdon and colleagues169 hypothesize three mechanisms that may lead to this contagious aspect of obesity. These are: (1) the direct effects of peers’ weight, ‘endogenous or causal effects;’ (2) characteristics of their peers other than their weight, ‘exogenous or contextual effects;’ and (3) factors common to both the adolescent and their peers, ‘correlated effects.’ Exogenous or contextual effects include characteristics of a peer that influences that peer’s weight, which in turn affects an

22 adolescent’s weight. Correlated effects are underlying factors that make an adolescent and his or her peers similar, such as a school exercise policy. Females and those with the highest BMI tend to be the most influenced by their peers, compared to males and those with a lower BMI, respectively. The authors hypothesize that increasing comparisons may similarly play an important role in the increasing prevalence of overweight among adolescents.169

Younger high school students are more likely to be overweight if there is a high prevalence of overweight among the senior students at their school.203 This effect persisted when controlling for individual-level risk factors for overweight. However, this study did not consider school-level predictors of overweight. It has also been shown that the BMI of individuals in an adolescent’s peer group, particularly those of the same sex, are important predictors of increased risk of overweight.204 Using techniques to adjust for the bi-directionality of the relationship between peer and individual weight status, the effect of overweight status on individual weight remained significant for females only.

This problem is exacerbated by the fact that overweight and obese adolescents are more likely to perceive that the average weight of their peers is higher than it actually is.205 Adolescents overestimating the weight of their peers are more likely to underestimate their own weight.205 As a result, these adolescents perceive the norm as being higher than it actually is, and may in fact see themselves as having a weight that is similar to that of their peers. These adolescents are consequently more likely to underestimate their weight status.

1.3.4 Challenges in Comparing Studies of Weight Status Underestimation

In addition to differences observed in estimates of weight status underestimation possibly due to the context of the study (i.e. country of data collection), two main challenges exist in comparing the results of these studies. Both relate to how actual weight status was determined. The first challenge is the use of measured vs. self-reported height and weight to calculate an individual’s body mass index (BMI). The second challenge is the

23 lack of a consistent means of identifying whether or not a specific BMI is considered overweight. These two issues are explained below in Sections 1.3.4.1 and 1.3.4.2 below.

1.3.4.1 Measured vs. Self-Reported Height and Weight

Estimates of the degree of weight status underestimation among adolescents vary greatly. These differences may partly be explained by differences in the methodology used. In particular, it is important to consider how actual weight status was determined (i.e. using self-reported or measured height and weight). Brener et al11 provides a comparison of estimates based on self-reported and measured height and weight: 57.5% underestimated their weight status when compared to BMI based on self-reported height and weight; 76.4% of overweight adolescents underestimated their weight status when height and weight were measured objectively. There was a similar degree of difference between the two measures of BMI (self-report vs. measured height and weight) among obese adolescents: 39.87% vs. 46.2%, respectively. The differences between self-reported and measured height and weight were significant (p<0.0001), with the concordance between self-reported height and weight with perception being higher (κ=0.17±0.02) than the concordance between measured height and weight with perception (κ=0.09±0.02).

1.3.4.2 Definition of Overweight & Obesity in Adolescents

Body mass index (BMI) is commonly used to identify individuals who are overweight or obese and is calculated by dividing weight by the square of height in metres (units: kilograms per metre squared; kg/m2). It has been recommended as a screening tool for overweight and is a good measure of body fatness in children and adolescents.206 Increasing BMI is predictive of increasing metabolic risk in youth125-128 and of disease later in life.67,111 BMI is suggested for use when measuring the weight status of children and adolescents in epidemiological studies207-211 and it is commonly used in surveys and population-based studies when direct measures of body fatness are not feasible.3,212,213

The BMI at which a child or adolescent is determined to be overweight or obese depends on their age and sex. BMI follows a J-shaped curve, with young children declining in BMI after birth, followed by adiposity rebound around six years of age. The curve levels out during the late teenage years and differs slightly for males and females.214-216 These

24 growth curves are then used to determine which children and adolescents are overweight or obese. However, several growth curves have been created and there is no widely adopted growth chart for school-aged children and adolescents. Instead, several sets of age- and sex-specific growth charts and reference values have recently been developed; these include international charts (i.e. those published by the World Health Organization216 and the International Obesity Task Force215) and country-specific charts such as the one published by the US Centers for Disease Control and Prevention.214 There are no growth charts created specifically for Canadian children and adolescents.

World Health Organization

The World Health Organization (WHO) growth curves (for children aged 5-19) are a re- creation of the 1977 NCHS/WHO growth curves using the same measured height and weight data as the original growth curves.216 Only American data were used in growth curve creation because international datasets varied in the study methodology used and the quality of the data. Improved statistical techniques were used with the data that had been collected for the previous growth curves. The purpose in creating these growth curves was to create a single growth curve that could be used from birth to adulthood, building on a previously created growth standard for children from birth to 5 years of age. The goal of this growth reference was to establish a conservative definition of overweight and obesity, since the association between elevated BMI and health risks is not well established in the adolescent population.217

The WHO growth reference defines overweight as having a BMI one standard deviation above the mean for age and sex and obesity as more than two standard deviations above the mean. In 19 year olds, one standard deviation above the mean corresponds to a BMI of 25.4 kg/m2 in males and 25.0 kg/m2 in females. A BMI of 29.7 kg/m2 is two standard deviations above the mean in 19-year-old males and females. These values are similar to the cut points recommended for adult overweight and obesity of 25 kg/m2 and 30 kg/m2, respectively.

25

International Obesity Task Force

The International Obesity Task Force (IOTF) provides reference values for overweight and obesity for children and adolescents aged 2 to 18. Nationally representative cross- sectional surveys from six countries (Brazil, 1989; Great Britain, 1978-1993; Hong Kong, 1993; the Netherlands, 1980; Singapore, 1993; and the United States, 1963-1980) were used in the calculation of these reference values.215 The percentile corresponding to the adult cut-points of 25 and 30 kg/m2 for overweight and obesity at age 18, respectively, were determined for each data set. These percentiles were extrapolated across the span of ages included in the dataset. The percentiles defining overweight and obesity in each dataset were then averaged to obtain the percentiles for the international population. This was possible because of the similarities in the shape of the percentile curves representing 25 kg/m2 and 30 kg/m2 in each of the data sets. The techniques used in the construction of this growth curve minimized national variations in the prevalence of overweight. The IOTF definition is better at predicting overweight than obesity.215 The IOTF growth reference is limited in that it only provides a categorical measure of adolescent overweight and obesity. That is, after applying these criteria, an individual is either overweight, obese, or neither overweight nor obese.

US Centers for Disease Control and Prevention

The US Center for Disease Control and Prevention (CDC) revised their growth charts for children and adolescents aged 2 to 20 years of age in 2000.214 These growth charts are based on data from five American cross-sectional health surveys that took place from 1963 through 1994. The samples in each survey were representative of the American population. Those over the age of 72 months were excluded from the most recent survey to ensure that the growth references were not influenced by the increasing prevalence of overweight and obesity among American children and adolescents. Children and adolescents between the 85th and 94th percentiles for their age and sex are overweight; those above the 95th percentile are obese. In addition, the CDC growth curves allow for the calculation of a continuous measure of weight status, specifically BMI z-scores.

26

Comparison of Definitions

The prevalence of overweight and obesity in Canadian children and adolescents varies substantially depending on which of the three above growth references is used.218 The differences are largely a consequence of the different methodologies used in the construction of each of these growth curves. This includes (1) the different samples used; (2) the statistical techniques used in the growth curve construction; and (3) the selection of the reference value used to define overweight and obesity.

The prevalence of overweight and obesity among Canadians aged 12 to 17 in 2004 ranged from 28.0% (CDC) to 33.2% (WHO), depending on which growth reference was used.218 The differences among the definitions is most pronounced for males aged 2 to 5, with a difference of 18 percentage points being observed in the prevalence of overweight between the WHO and IOTF definitions. The prevalence of obesity is similar across all definitions. Among adolescents, the prevalence of obesity ranged from 9.4% (IOTF) to 12.4% (WHO). These prevalence estimates were nearly identical when comparing WHO and CDC definitions (12.4% vs. 12.1%, respectively).

The recommendations for which growth reference to use in the Canadian context are constantly evolving. In 2004, Canadian guidelines recommended that the growth of individual children be monitored using the CDC references, while the IOTF references be used for epidemiological purposes.207,219 These recommendations have been mirrored in the United States.212 These guidelines were reinforced in 2010 in a study that compared the prevalence of overweight and obesity among Canadian children and youth across all three growth curves.218 However, also in 2010, an updated version of the 2004 guidelines was published, now recommending that the WHO guidelines be used to monitor child and adolescent growth.220

1.3.5 Predictors of Weight Status Underestimation among Adolescents

Since weight status underestimation is such a common problem among overweight adolescents, there has been a recent push for increased research in this area. In particular, identifying influential factors for weight status perception is of increasing importance.4

27

Included in this review are the effects of sex, severity of overweight, age, ethnicity, and exposure to overweight. Despite socioeconomic status playing an important role in increasing the risk of being overweight, there appears to be no association between socioeconomic status and weight status underestimation.12,44,51,57,221

1.3.5.1 Individual Characteristics

Sex

Studies examining the accuracy of weight status perception among overweight adolescents consistently find significant sex differences, with overweight male adolescents being significantly less likely to recognize that they are overweight status than their female counterparts.4,8,10-12,14-16,18,19,21,22,24,26,28,30,33,38,40,43,45-47,49- 52,56,57,64,195,196,221,222 This likely results from greater cultural desires for thinness among females, as well as differing muscle to fat ratios for females and males.223 Although not statistically significant, the last decade has seen slight increases in the proportion of overweight males who accurately recognize that they are overweight.22,34 While Foti and Lowry22 observed a slight decline in overweight perception among overweight females, Neumark-Sztainer et al34 found a significant increase in the accurate perceptions of overweight among overweight female adolescents.

Severity of Overweight

Obese adolescents are more likely than overweight adolescents to accurately perceive that they are overweight; however, there is still a large disconnect between actual and perceived weight status among those who are overweight.4,6,9,11,16,20,33,35,39,40,50-52,195 Further, as an adolescent’s BMI or BMI z-score increases, they become less likely to underestimate their weight status.57,64,65,221 However, Viner et al51 found that increasing BMI z-score was associated with decreasing accuracy in males but increasing accuracy in females; it is important to note that this study included individuals ranging from underweight to obese. The same relationship between weight status underestimation and BMI z-score may not be observed when only those who are overweight and obese are included.

28

Age

The relationship between weight status underestimation and age (or alternatively pubertal status) is not clear. Some studies have found that there is no relationship between these two variables,12,222 while others have found a significant association.27,40,57,64,195 Those who have found an association suggest that increasing age is associated with increased risk of underestimating one’s weight status.40,57,64 Skinner et al45 found that adolescents between the ages of 13 and 16 are the most accurate in their perceptions of their weight status, when compared to those less than 13 years of age and those more than 16 years of age. The effect of age on weight status underestimation may be different for males and females: younger males are less likely to underestimate their weight status than younger females, while older males are more likely to underestimate their weight status than older females.27,195

Ethnicity

There has been no previous research examining the relationship between ethnicity and weight status underestimation among Canadian adolescents. Within the Canadian context, only one study has assessed the relationship between ethnicity and body image. This study found that Aboriginal adults selected a larger body size to represent their ideal weight than did non-Aboriginal adults.224

In the United States, significant differences in weight status underestimation among overweight and obese adolescents exist across ethnic groups. Black adolescents are consistently more likely to underestimate their weight status than White adolescents.5,11,16,18,22,33,58,221 The differences in weight status underestimation across other ethnic groups are not as clear. Some studies have suggested that Hispanic adolescents are less likely to underestimate their weight status than Black adolescents, but more likely than White adolescents.5,11 Other studies have either found that there are no differences between Hispanic adolescents and those of other ethnic groups or that these differences exist only in certain subpopulations of adolescents. For example, several studies demonstrate that levels of weight status underestimation are similar among Black and Hispanic adolescents.16,18,221 In contrast, Foti and Lowry22 found that

29 levels of weight status underestimation among Hispanic female adolescents was higher than among White female adolescents; the same was not observed among males. In addition, levels of weight status underestimation are higher among Asian American adolescents than among White adolescents.33 This contrasts with a study comparing adolescents in the United States to adolescents in China, finding that levels of weight status underestimation were substantially lower among Chinese adolescents than American adolescents.60 Female American Indian/Native American adolescents are more likely to underestimate their weight status than female Hispanic adolescents.58

Among adolescents in the Netherlands, those who are of non-Dutch descent were more likely to underestimate their weight status than those who were of Dutch descent.12 In contrast, there were no differences in weight status underestimation comparing those who were born in the United Kingdom to those who immigrated there.51 American-born male adolescents were more likely to underestimate their weight status than were immigrants to the United states.221 Among New Zealand adolescents, those of East Asian descent were the least likely to underestimate their weight status; Europeans and South Asians were the next least likely; Pacific Islanders were the most likely to underestimate their weight status.17 Similar ethnic differences are found in the United States and the United Kingdom with regard to weight status underestimation: Black British adolescents were less likely to accurately perceive that they were overweight, when compared to White and Asian adolescents.46,51 No differences were found comparing weight perceptions in Black Jamaican adolescents with Jamaican adolescents of other ethnicities.222

However, since these studies have been conducted outside Canada, their results cannot be extrapolated to Canadian adolescents. These ethnic differences may result from different cultures placing different meanings on body size.11

1.3.5.2 Community Characteristics

Weight Status of Community-Based Reference Groups: Exposure to Overweight

Adolescents and their behaviours are influenced by an array of sources and the exposures that these sources provide. . This includes exposures found within the home (i.e. the

30 weight status of parents or siblings), as well as from schoolmates and peers, the community in which an adolescent lives, and the media. Each of these spheres may have a unique effect on adolescents. Previous literature has focused both on the role of overweight among parents and schoolmates on weight status underestimation among overweight adolescents.

Parental weight is an important predictor of weight status underestimation among overweight and obese adolescents. Martin et al221 found that having two obese parents was significantly associated with increased weight status underestimation. This was significant when males and females were combined, and in males alone, but not in the model that included only females. Similarly, having only an obese mother was a significant predictor of underestimation in the full sample and in males only. It is important to note that the effect of maternal obesity was not a significant predictor of weight status underestimation when an interaction between sex and BMI was included in the model. Paternal obesity was not significantly associated with weight status underestimation in either males or females. Similarly, Strauss65 found that paternal weight did not influence perception of overweight, controlling for actual weight. In contrast, adolescents with thin mothers were more likely to consider themselves as overweight controlling for actual weight status—an effect likely mediated by pressure to lose weight.65 It is not only parents’ actual weight that influences weight status perception among adolescents. When parents underestimate their child’s weight status, that child is also more likely to underestimate their weight status when compared to an adolescent whose parents accurately perceive that their child is overweight.45

Peers are considered to be a very important comparison group for adolescents in their perceptions of their own attractiveness—peers are more important than models and celebrities.225 In addition, peers have been identified as being an important source of information about one’s weight; adolescents frequently base their opinions of what defines healthy weight by comparing themselves to their peers.226 Peers are also influential in determining whether or not an adolescent engages in weight-related behaviours, including exercise, sports, and fast food consumption, after adjustment for school-level effects.227

31

In a study of Quebec adolescents, Maximova et al195 found that increased parental and schoolmate BMI were both independent predictors of decreased accuracy of weight status estimation. The magnitude of the effect of schoolmate BMI on accuracy of perceived weight status was stronger than the effect of parental BMI. Parental BMI was not a significant predictor of underestimation among 13-year-olds, suggesting that adolescents of this age rely heavily on their peers for comparisons of their weight.195 However, these results were not stratified by sex, so the different effects of social comparisons on underestimation in males and females could not be ascertained. As the prevalence of overweight continues to increase, adolescents appear to becoming more desensitized to overweight.226

It is interesting to note that studies based on samples with a high prevalence of overweight tend have higher rates of weight status underestimation among overweight adolescents than do studies with a lower prevalence of overweight. This supports evidence suggesting that adolescents exposed to higher amounts of overweight and obesity in their homes and at school tend to underestimate their weight status more often.195,221 Brener et al11 examined weight status underestimation in a sample of American adolescents. Their sample had a higher prevalence of overweight and obesity than the general American population and very high levels of weight status underestimation were observed. In contrast, studies that have a lower prevalence of overweight tend to have lower estimates of weight status underestimation among overweight adolescents. For example, Brug et al12 examined weight status underestimation among Dutch adolescents between the ages of 13 and 19. The prevalence of overweight in this sample was 6%, with 1.5% of these overweight adolescents underestimating their weight status.

1.4 Summary

With almost 1 in 3 Canadian adolescents being either overweight or obese, it is important that efforts made to address this epidemic be effective. This is especially true because, left untreated, these adolescents face significant health risks down the road, including being at increased risk of developing cardiovascular disease and type II diabetes. However, these adolescents also tend to have poorer psychosocial well-being than those

32 who do not recognize that they are overweight. Weight status underestimation represents an important area of focus for future strategies aimed at decreasing overweight and obesity among Canadian adolescents but the approach taken must be on that balances the positive effects of underestimation on mental health with the negative effects on physical health. Adolescents who accurately recognize that they are overweight are more likely to engage in weight management behaviours than those who underestimate their weight status. Since these adolescents do not engage in healthy weight-related behaviours, they may be more likely to develop obesity-related comorbidities than those who accurately recognize that they are overweight. Future approaches to the obesity epidemic need to focus on weight status underestimation.

This thesis begins to address weight status underestimation in a Canadian context. Specifically, it focuses on understanding how common weight status underestimation is among adolescents and if the proportion of overweight adolescents that underestimate their weight status has changed over the past decade. It also examines characteristics of adolescents who underestimate their weight status and the influence of exposure to overweight as a predictor of weight status underestimation. The objectives are further elaborated on in Chapter 2.

33

Chapter 2 2 Objectives & Hypotheses 2.1 Objectives

There are three main objectives for this thesis. Firstly, this thesis aims to examine the magnitude of weight status underestimation among overweight and obese Canadian adolescents and whether or not the levels of weight status underestimation have changed across time. The second objective is to examine the role of individual-level characteristics (i.e. severity of overweight, age, and ethnicity) on weight status underestimation and whether the effects of these characteristics are different for males and females. Thirdly, this thesis explores differences in weight status underestimation across communities and the role of the prevalence of overweight in these communities plays in accounting for any variation across communities. These objectives are explicitly stated below:

Objective 1 (a) Assess the magnitude of weight status underestimation among overweight adolescents in Canada between 2001 and 2010, separately for males and females; (b) Examine the time trend in weight status underestimation for overweight adolescents from 2001 to 2010 among all adolescents and separately for males and females

Objective 2 (a) Assess if individual-level characteristics, including age, severity of overweight, and ethnicity, play a role in predicting the weight status underestimation among overweight adolescents; (b) Assess if the effects of the above characteristics on weight status underestimation are different for males and females;

34

Objective 3 (a) Assess the variation in weight status underestimation among overweight and obese adolescents across communities; (b) Examine if the prevalence of overweight in an adolescent’s community (i.e. exposure to overweight) explains the variation in weight status underestimation identified in sub-objective (a).

2.2 Hypotheses

Objective 1

Based on similar studies of adolescents outside Canada, it is expected that more than 20% of overweight and obese adolescents will underestimate their weight status. Weight status underestimation is expected to be more prevalent among males than females. In terms of the trend across time, it is expected that the levels of weight status underestimation will remain relatively stable from 2001 to 2010 for both males and females.

Objective 2

Severity of Overweight: It is hypothesized that As the severity of overweight increases, adolescents are more likely to underestimate their weight status. It is expected that the effect of severity of overweight on weight status underestimation will be similar in males and females.

Age: It is expected that the relationship between age and weight status underestimation will be different for males and females. Specifically, based on previous literature, it is hypothesized that, among young adolescents, females will be more likely to underestimate their weight status than males. The reverse is expected for older adolescents; specifically, in this age group, males are expected to underestimate their weight status more often than females.

Ethnicity: No previous studies have examined the relationship between ethnicity and weight status underestimation in a Canadian context. This objective explores the

35 differences across ethnic groups in Canada. It is expected that the overall effect of ethnicity on weight status underestimation will be significant. The effects of ethnicity are expected to be similar for males and females.

Objective 3

Significant variation in weight status underestimation among overweight and obese adolescents across communities is expected and that exposure to overweight will be an important predictor of this variation across communities.

36

Chapter 3 3 Methods

This chapter begins with an overview of the data used in this thesis (Section 3.1), followed by the measures used in the analysis (Section 3.2), and an introduction to multilevel logistic regression (Section 3.3). The specific details of the analysis completed are provided in Section 3.4 and additional statistical considerations, including the software used, missing data, survey weights, and the role of interview mode in the analysis, is provided in Section 3.5.

3.1 Data Source

The objectives of this thesis were accomplished by conducting a secondary analysis of the Canadian Community Health Survey (CCHS). The CCHS is a population-based cross-sectional survey conducted by Statistics Canada and designed to gather information from Canadians on their health and its determinants, as well as their use of the health care system. The CCHS was conducted biennially from 2001 through 2005, with more than 130,000 Canadians surveyed during each wave. Starting in 2007, there were substantial changes to the methodology of the CCHS. Instead of collecting data biennially, data collection occurred on an ongoing basis. The total sample size was kept constant across a span of two years (i.e. 2007-2008), with half being collected in each year. Despite these changes, a sample size of at least 130,000 was maintained every two years. The data collected since 2007 are available either as an annual component (file consists of all data collected over one year) or as a combination of two years (i.e. 2007-2008). More detailed information about the CCHS is provided in publicly available documentation from Statistics Canada.228 All CCHS data from 2001 through 2010 were used in the present study. An indicator variable for the year in which data were collected was computed (see Section 3.2.5 below). Files were then combined to create one dataset for analysis.

3.1.1 Content of the CCHS

The content of the CCHS specifically reflects the role of this survey as a tool to gain information about the health and health care utilization patterns of Canadians. To

37 accomplish this task, the CCHS included three components: (1) common content; (2) optional content; and (3) theme content. The common content included demographic characteristics, height and weight, general health, health care utilization, and other basic health information. This portion of the questionnaire was asked to all respondents in each survey cycle. Optional content was selected for each survey wave by health regions. Possible topics for the optional content included drug use, several mental health scales, and changes made to improve health. Lastly, each wave of the CCHS had specified theme content. Questions pertaining to the selected theme were asked to participants across the country. These modules were typically selected from among the optional content. There was a rotation of CCHS themes across cycles to allow for comparisons from previous cycles. This project focuses specifically on survey questions related to the weight of adolescents and their demographic characteristics.

3.1.2 Sampling Design

The target population of the CCHS was non-institutionalized Canadians, 12 years of age and older, residing in each of the ten provinces and the three territories. The sample population excluded those living on Indian Reserves and Crown Lands, institutionalized individuals, full-time members of the Canadian Forces, and individuals living in remote areas. Approximately 2% of the Canadian population was missed as a result of these exclusions.

For the purposes of sample allocation, each province was broken down into several health regions; each territory represents a single health region. The health regions typically corresponded to local public health units or health authorities. A multistage approach was taken to sample allocation. The goal of this sample allocation was to ensure that reliable estimates could be obtained for each health region. The first stage of the sampling strategy ensured a minimum of 500 respondents in each provincial health region. The second stage of the sampling design distributed the remainder of the 130,000 respondents across health regions proportional to their population. The technique used for sample allocation in the three territories was slightly different. It was pre-determined that the Yukon and the Northwest Territories would each have a sample of 600, while Nunavut would have a sample of 350.

38

Statistics Canada used three complementary sampling frames to obtain a representative sample of the Canadian population. Half of the sample was selected using an area frame—or list of dwellings—based on the sampling frame developed for the Canadian Labour Force Survey. This area frame was obtained using a multistage stratified cluster design to ensure the sample was representative of all geographic regions and socioeconomic strata. Another half was selected from a list frame of telephone numbers, obtained from the Canada Phone directory. Telephone numbers were matched with postal codes and then assigned to their corresponding health region. Random sampling techniques were then used to obtain the required number of telephone numbers in each health region. A small percentage (1%) of the sample was selected using random digit dialling to account for unlisted telephone numbers.

Data were collected using computer-assisted interviewing techniques (CAI). Those sampled by an area frame were interviewed either in person (computer assisted personal interviewing, CAPI) or by telephone (computer assisted telephone interviewing, CATI). Individuals sampled from either the list of telephone numbers or random digit dialling were interviewed by telephone (CATI). The proportion of the population interviewed using either mode varied across time and geography. The effect of these differences in interview mode across time was taken into account during the data analysis; details are provided in Section 3.5.4 below.

3.1.3 Study Population

The analysis for this project was based on CCHS data collected in 2001 (cycle 1.1), 2003 (cycle 2.1), 2005 (cycle 3.1), 2007, 2008, 2009, and 2010. The total sample size and response rates for each survey cycle are provided in Table 1. All analyses were conducted on adolescents between the ages of 12 and 18. Adolescents who were pregnant at the time of the survey or those who had assistance in completing the survey (i.e. responded via proxy) were excluded from the analysis. Substantial physiological changes occur during pregnancy. As a result, pregnancy is of important consideration when assessing data for an individual’s health, including their weight, and consequently their body mass index (BMI). Since BMI does not maintain its usual interpretation in pregnant women, no accurate measure of weight status in pregnant adolescent females

39 was available, resulting in their exclusion from the analysis. Individuals reporting by proxy were also excluded since an accurate estimate of an individual’s perceived weight status cannot be obtained when someone else is answering on their behalf. The final sample consisted of all adolescents who were either overweight or obese, as defined below.

Table 1. Sample size and response rates by year for the Canadian Community Health Survey (CCHS) Survey Year 2001 2003 2005 2007 2008 2009 2010 Sample Size (N) Total 131,535 135,573 132,947 65,946 66,013 61,679 63,191 Adolescents only (aged 12-18) 15,419 16,610 14,424 6,220 6,737 6,474 6,422 Overweight & obese adolescents (aged 12-18) 2,189 2,653 2,375 973 1,068 1,126 1,068 Response Rates (%) Overall 84.7 80.7 79.0 77.6 75.2 73.2 71.5 By health region 76.2-92.3 71.6-89.1 68.3-87.1 66.3-87.6 66.1-86.3 62.5-84.7 61.7-84.8 Note: Sample sizes are calculated using rescaled sampling weights provided by Statistics Canada (See Section 3.5.3).

40

41

Defining Overweight in Adolescents

Several tools can be used to measure adolescent overweight, including methods that are based on the level and distribution of fat in an individual, such as hydrodensitometry, dual-energy x-ray absorptiometry, bioelectrical impedance analysis, and anthropometric indices such as skinfold measurements and waist circumference. However, use of these measures is not always feasible, particularly in large national surveys such as the CCHS. Instead, the CCHS used body mass index (BMI) to identify individuals who were overweight. BMI is defined as the ratio of weight to squared height (units: kilograms per metre squared; kg/m2). Values for height and weight were obtained by asking adolescents to report their height (without shoes on) and weight.

The exact BMI at which an adolescent is considered to be overweight or obese depends on their age and sex. Growth curves created by the United States Centers for Disease Control and Prevention (CDC) were used to (1) identify which adolescents were overweight and obese and (2) compute BMI z-scores.214 The CDC growth curves were chosen over other growth references (i.e. the IOTF or WHO growth references) because they best reflected Canadian recommendations at the time of data collection.219 Consequently, if comparing their weight to objective standards, adolescents may have been more likely to compare their weight to the CDC growth curves than other objective criteria. Individuals above the 85th percentile for their age and sex were considered overweight, while those above the 95th percentile were considered obese. Only those defined as either overweight or obese were included in the analysis. BMI z-score was used to measure severity of overweight and is described in Section 3.2.2 below.

BMI percentiles and z-scores were calculated using a SAS program available from the CDC website.229 This program uses adolescents’ self-reported height and weight, as well as their age in months, to calculate their BMI z-score. Since the age provided as part of the CCHS is in years, a more exact measure of age was calculated by taking the difference between date of interview and date of birth. Missing month of birth was assigned a value of 7 (July); 4.0% of all adolescents had missing data for their month of

42 birth. For respondents with missing day of birth, a value of 1 was assigned; this was the case for 4.2% of all adolescents.

In addition, the SAS program provided by the CDC flagged extreme (i.e. biologically implausible) values for height, weight, and BMI.229 Individuals whose self-reported height, weight, and BMI exceeded these extreme values were excluded from the analysis. This included individuals with: (1) a height-for-age z-score of less than -5 or greater than 3; (2) a weight-for-age z-score of less than -5 or greater than 5; and (3) a weight-for- height z-score of less than -4 or greater than 5.229 These extreme values are typically the result of measurement or other administrative errors and were based on criteria proposed by the World Health Organization and used in the growth charts published by NCHS and WHO in 1977.229

Adolescents with missing data for height and weight were also excluded from the analysis since the weight status of these individuals could not be identified. Consequently, it could not be determined if these adolescents met the inclusion criteria of being either overweight or obese. A total of 7.5% of adolescents were excluded from the analysis either because of missing data for height and/or weight, or because their self- reported values were considered extreme.

3.2 Measurement Instruments

This project aimed to estimate the effects of sex, age, severity of overweight, ethnicity, time, and exposure to overweight in the community on weight status underestimation among overweight and obese Canadian adolescents. How each of these constructs is measured and how they were used in the statistical model is explained below. A complete list of CCHS variables used in the analysis is provided in Appendix B.

3.2.1 Perceived Weight Status

Perceived weight status was measured with the following survey question, “Do you consider yourself: overweight, underweight, or just about right?” This question was included in all cycles of the CCHS. A similar measure, including five possible responses, for perceived weight status had a test-retest reliability score of r=0.69.230

43

3.2.2 Severity of Overweight

Severity of overweight is measured using BMI z-score (based on age and sex). A description on how BMI z-score is calculated is provided in Section 3.1.3 above. A quadratic variable for BMI z-score was computed by squaring the above variable. All regression models centred both the linear and quadratic indicators for severity of overweight to improve the interpretability of the intercept in the model.

3.2.3 Age

The CCHS provided a measure of the age of respondents in years. This variable was a derived variable calculated by taking the difference between the date of interview and date of birth. This value for age was confirmed with the respondent during the interview. A quadratic variable for age was computed by squaring the above variable. All regression models centred the indicator for age to improve the interpretability of the intercept in the model.

3.2.4 Ethnicity

The measure of ethnicity used was based on the question in the CCHS asking respondents about their cultural/racial background. Specifically, they were asked:

“People living in Canada come from many different cultural and racial backgrounds. Are you: White, Chinese, South Asian, Black, Filipino, Latin American, South East Asian, Arab, West Asian, Japanese, Korean?”

Prior to June 2005, Aboriginal Peoples of North America was a potential response option in the question stated above. After June 2005, a separate variable was used to identify participants of Aboriginal descent, based on the question: “Are you an Aboriginal person that is North American, Indian, Métis, or Inuit?” After these changes were made, respondents identifying as Aboriginal were not asked if they belonged to any other racial groups. Those who reported belonging to multiple racial groups were grouped as belonging to those of mixed cultural background. The above responses were grouped as follows: White, Asian (combination of the Chinese, South Asian, Filipino, South East Asian, Arab, West Asian, Japanese, and Korean categories), Black, Latin American,

44

Aboriginal, other, and mixed cultural background. Dummy variables were created for each ethnic group. An additional dummy variable was created for those with missing data on ethnicity. This included individuals who refuse to answer and those who did not know their cultural/racial background.

3.2.5 Time

A discrete variable was created to reflect the timing of data collection. This variable took on seven values, and was coded as follows: 0, 2001; 2, 2003; 4, 2005; 6, 2007; 7, 2008; 8, 2009; 9, 2010. A quadratic variable for time was computed by squaring the above variable.

3.2.6 Weight Status of Community-Based Reference Groups: Exposure to Overweight

There are several spheres of influence of adolescents. Previous literature has suggested that both peer and parental overweight are important sources of comparison for overweight.65,195,221,225,226 This present study expands on what has already been established with regards to adolescents’ weight-based comparisons. In particular, this study examines weight status underestimation among all individuals living in an adolescent’s community. It is expected that individuals compare their weight to others in their community when forming perceptions of their own weight. However, it is not known who comprises the reference population within this community for this comparison. For example, adolescents’ perceptions of their weight may be differently influenced by the weight of individuals within different reference populations, including both age- and sex-specific reference populations. It is, consequently, important to identify what this reference population is in order to better understand what adolescents base their perceptions of their weight on. For the purpose of this project, several reference populations were explored. These included: (1) all individuals over the age of 12 within a community; (2) all individuals of the same sex as the respondent and of all ages in a community; (3) all individuals between the ages of 12 and 18 in a community; and (4) individuals between the ages of 12 and 18 and of the same sex as the respondent

45 in a community. The effect of sex-specific reference populations were examined only for adolescents of that sex.

The effect of exposure to overweight was examined using the prevalence of overweight within each of the reference groups identified above. Among those 19 years of age and older, those with a BMI greater than 25 kg/m2 were classified as overweight. Adolescents (aged 12 to 18) were overweight if their BMI exceeded the 85th percentile for their age and sex. Individuals with missing data for height and weight did not contribute to the prevalence of overweight in their respective community.

There is no clear or well-established geographic reference community for defining an adolescent’s exposure to overweight. Previous studies have focused only on classmate and/or parental BMI as a source of exposure to overweight. This study expanded on these conceptualizations by exploring two different operationalizations of community: health regions and census subdivisions (CSDs). While neither is perfect given the CCHS survey design, each has its own unique set of strengths and weaknesses. Since each operationalization has strengths that outweigh the limitations of the other, the role of identifying the heterogeneity across communities was examined.

Health Regions

The majority of health regions correspond to local public health units. Consequently, the use of health regions enhances the applicability of the findings to public health care providers. Public health strategies are often based at the level of individual health units. Incorporating the findings of this study to population-based strategies aimed at adolescent obesity is an important outcome for this project.

A second strength of using health regions to define community involves the ability to make stable population estimates at the health region level. Since the CCHS was designed to provide estimates of population-level parameters at the level of individual health regions, this ensures that accurate, stable estimates of aggregate variables could be obtained.

46

The use of health regions, however, was not without significant limitations. The number of health regions included in each cycle of the CCHS ranged from 121 to 136. Over the course of the survey waves included in the analysis, the boundaries of some health regions changed for administrative purposes. To ensure comparability of health regions across time, modifications were made to those health regions with substantial boundary changes. This included considerable changes to health regions in the provinces of Alberta, British Columbia, and Prince Edward Island. As an example of such a modification, all health regions in Prince Edward Island were grouped into one provincial health region, since the number of health regions in the province changed from 2 to 4, then to 3. Each of these changes resulted in substantial changes to the boundaries of the province’s health regions. The modifications made to all health regions are outlined in Appendix C. The final number of health regions included in the analysis was 109. These combined health regions may be heterogeneous and mask potential differences across communities. In addition, some individual health regions covered a large geographic area. As an example, each territory corresponded to its own health region.

Census Subdivisions

CSDs were the second operationalization of community used in this thesis. CSDs corresponded to municipalities or other geographic region considered equivalent by Statistics Canada. The use of these geographic areas to define one’s reference community addresses the key limitations associated with the use of health regions to define community, in particular, the heterogeneity of these regions. Specifically, CSDs are smaller in nature, resulting in more homogeneous communities.

However, the use of CSDs is also not without important limitation. Because CSDs are smaller geographic areas than health regions, the number of respondents in a CSD is much smaller than the sample size of a health region. As a result, which CSDs are represented in each wave of the survey is not consistent across time. Consequently, even when cycles are pooled across years, the sample sizes in some CSDs remain small. Only CSDs with at least two male and two female overweight adolescents were included to ensure sufficient power for the multilevel analyses. This ensured that there were at least

47 two respondents in each cluster since multilevel analyses were stratified by sex. This resulted in the exclusion of 1856 of 2402 (77.3%) CSDs and 3603 of 12683 (28.4%) overweight adolescents. It has been suggested that the number of clusters is more important than the number of individuals per cluster in obtaining accurate estimates with multilevel analysis.231 The use of singleton clusters (i.e. clusters with only one individual) introduces bias. A sensitivity analysis of cluster size compared results obtained using the sample described above with analysis excluding CSDs with fewer than 5 male and 5 female overweight adolescents (Appendix D). Including only those CSDs with at least five adolescents of each sex resulted in the exclusion of 2221 of 2402 (92.5%) CSDs and 6,636 of 12,683 (50.0%) of overweight adolescents. Similar results were observed for both cluster sizes. The use of CSDs with at least two male and two female overweight adolescents ensured both a greater number of clusters and respondents, as well as a greater representation of all CSDs in the CCHS.

3.3 Overview of Multilevel Logistic Regression

Multilevel regression analysis is a tool that is commonly used to assess the effect of community-level covariates on an individual-level characteristic. It is also a statistical technique that allows for the assessment of variation across communities. These attributes of multilevel regression analysis made it an ideal tool to address the questions raised in Objective 3. This section provides readers with an overview of multilevel regression analysis with a particular focus on the use of multilevel techniques to binary outcomes (i.e. multilevel logistic regression).

Multilevel regression accounts for clustering within the data since individuals within clusters are correlated.232 That is, individuals within a cluster are more alike than individuals from different clusters. Because individuals are not independent, the independence of observations assumption of usual regression analyses is violated. Multilevel analysis relaxes the assumption of fixed effects for covariates in the model by allowing these effects to vary across clusters. This is done by allowing the intercept and slopes of individual-level characteristics (i.e. age, BMI z-score, and ethnicity), as well as the slope for time, to vary. When these effects are not allowed to vary (i.e. are fixed effects), the model is similar to a usual regression model. The main difference is that

48 multilevel analysis assumes non-independence of individuals within a given cluster and adjusts the standard errors accordingly.

When these regression parameters are allowed to vary across clusters (i.e. are random effects), unique regression equations are estimated for each cluster. The output of the analysis provides an average value of these parameters, as well as the variance in these parameters across clusters. Significant variation for a random intercept indicates that different clusters have different baseline levels of the outcome of interest, controlling for all individual-level covariates in the model. When the variation for a random slope is significant, the rate of change across values of that dependent variable is different across clusters. Random effects are obtained by adding additional regression equations to the model. The outcome for each of these regression equations is the parameter from the initial regression equation. When no cluster-level covariates are included in the model, the added regression equations include an intercept and a term for the variation. Cluster- level covariates are added to this equation to examine the effect of cluster-level covariates on the variance across clusters. These cluster-level characteristics predict the parameters in the individual-level model. That is, they are used to predict the intercept or slope for a particular cluster. The general equation for a multilevel logistic regression model is as follows:

Individual-level equation:

Cluster-level equation:

where is the probability of the outcome is the number of individual-level covariates is the number of clusters is the number of cluster-level covariates

is an individual-level covariate

is a cluster-level covariate

49

2 is the random variation associated with person i (fixed to π /3 for a logistic model)

is the random variation associated with cluster j

3.3.1 Interpretation of Multilevel Logistic Regression Analyses

3.3.1.1 Measures of Association

Individual-Level Effects

It is important to note that the regression coefficients in a multilevel model do not maintain their usual interpretations.233,234 After exponentiation, the individual-level coefficients are odds ratios for within-cluster comparisons relative to the residual variation.233 That is, an odds ratio maintains its usual interpretation provided the comparison is made for two individuals in the same cluster.

Cluster-Level Effects: The Interval Odds Ratio

The cluster-level coefficients are, after exponentiation, also odds ratios; however, like their individual-level counterparts, they cannot be interpreted as a typical odds ratio.234 Instead, they are considered to be odds ratios comparing individuals from clusters differing on the value of the cluster-level variable, but having the same random effect 234 (unj). In other words, the estimate obtained is an average odds ratio. The interval odds ratio (IOR) has been recommended as an alternative measure to quantify the effect of a cluster-level covariate.233,235 The IOR provides an interval of odds ratios used to compare individuals with different cluster-level covariates. The middle 80% of odds ratios are represented in this interval. Although the initial choice of an 80% interval is somewhat arbitrary (see Larsen et al235), the use of an 80% interval has been recommended233,234 and is being used with increasing frequency in the epidemiological literature.236-238

The IOR incorporates both the average regression coefficient and the variance of that coefficient across clusters.233-235 This provides a statistic that allows for assessment of the importance of that covariate in explaining the cluster-level variance. An interval containing 1 indicates large cluster variability in relation to the magnitude of the effect of

50 that cluster-level variable.233,234 An interval that does not include 1 reflects a large effect of that cluster-level covariate in relation to residual variation across clusters.233,234 The formula to calculate the IOR is as follows:

where and are the 10th and 90th percentile of the standard normal distribution, respectively

is the average parameter (random-effect) from the regression equation

is the residual cluster-level variance of the parameter of interest ( )

Cluster-Level Effects: Proportional Change in Variance

The proportional change in variance (PCV) allows for an assessment of how much variance is explained by introducing a cluster-level covariate into the multilevel regression equation. It is important to note that, in the case of a binary outcome, only models with identical individual-level regression equations can be compared. The formula for the PCV is:

where is the variance for crude model (i.e. the model without cluster-level covariates)

is the variance of the model adjusted for cluster-level covariates

51

3.3.1.2 Measures of Variance

Intraclass Correlation Coefficient

In a multilevel linear regression analysis, the cluster-level variation can be easily quantified by computing the intraclass correlation coefficient (ICC).232 The ICC is calculated by dividing the cluster-level variance by the total variance (sum of the individual-level variance and the cluster-level variance). However, for multilevel logistic regression analysis, the two variances are on different scales: the individual-level variance is on the probability scale, while the cluster-level variance is on the logistic scale.232-234,239 To calculate the ICC for the logistic case, the individual-level variance is fixed to the variance of the standard logistic distribution (2/3).232-234,239 The formula for the ICC then becomes:

where is the cluster-level variance

Median Odds Ratio

To ease interpretability of cluster-level variation, the median odds ratio (MOR) has been recommended.233-235 This measure translates the cluster-level variance to an odds ratio scale.233-235 The MOR is an odds ratio that compares two individuals with the same individual-level characteristics but randomly chosen from different clusters. The MOR then reflects the median odds ratio of all possible comparisons of identical individuals from different clusters.233-235 The possible values for a MOR are greater than or equal to 1. A large MOR is indicative of large variation across clusters while a value of 1 indicates no variation across clusters. The MOR can be compared directly to odds ratios for fixed effects. The formula to calculate the MOR is:

52 where is the 75th percentile of the cumulative distribution function of the standard normal distribution.

is the residual cluster-level variance of the parameter of interest ( )

3.4 Statistical Analyses

3.4.1 Preliminary Analyses

The prevalence of overweight and obesity at each survey cycle was determined for all adolescents (aged 12 to 18) and all Canadians over the age of 12. Descriptive statistics were computed for all other key variables in the analysis, both for all adolescents and all overweight adolescents. This includes the calculation of frequencies for categorical variables (sex and ethnicity), and means and standard deviations for continuous variables (age and BMI z-score). Sampling weights (see Section 3.5.1 below) were used to ensure these estimates reflected the unequal probability of selection inherent in the design of the CCHS.

3.4.2 Objective 1

The first objective examined the magnitude of weight status underestimation among overweight adolescents and how this has changed between 2001 and 2010. The first half of this objective was assessed by determining the frequency of weight status underestimation among overweight adolescents for both the full sample (i.e. combining males and females) and separately for males and females. This analysis was completed separately for each survey year. The complex sampling design of the CCHS was taken into consideration in the analysis by using sampling weights provided by Statistics Canada (see Section 3.5.1 below).

The time trend (both linear and curvilinear) in weight status underestimation was determined using logistic regression. The overall time trend, as well as the separate time trend for males and females, was estimated. The derivation of the time variable is explained in Section 3.2.5 above. Interview mode was controlled for to ensure that observed results were not the result of the changes made regarding the mode of data collection across the survey cycles, as described in Section 3.5.4 below. All regression

53 models were computed using sampling weights, as described in Section 3.5.3 below, and robust maximum likelihood estimation, as described in Section 3.5.1 below.

3.4.3 Objective 2

The second objective examined the effect of three individual-level factors on weight status underestimation: (1) severity of overweight; (2) age; and, (3) ethnicity. This objective also identified the differences in these effects for males and females. This was achieved using logistic regression. To examine the differences in each of these effects for males and females, regression models including a term for the interaction between the main effect of interest and sex were computed. Following evidence of a significant interaction between these two variables, models were stratified by sex. All models controlled for the effect of interview mode and used sampling weights (see Section 3.5.1 below). Robust maximum likelihood estimation (see Section 3.5.1 below) was employed.

The first set of regression models measured the effect of severity of overweight on weight status underestimation by including BMI z-score (see Section 3.2.2 above) as the covariate of interest. Both the linear and the quadratic effects of BMI z-score were evaluated. The interaction between sex and BMI z-score was assessed to determine if the relationship between weight status underestimation and severity of overweight was different for males and females. The effect of severity of overweight, adjusted for the effects of age and ethnicity, was also estimated.

The second set of regression models measured the effect of age on weight status underestimation by including age (see Section 3.2.3 above) as the covariate of interest. Both the linear and quadratic effects of age on the risk of weight status underestimation were assessed. The interaction between sex and age was also assessed to determine if the relationship between underestimation and age was different for male and female adolescents. The effect of age, adjusted for the effects of severity of overweight and ethnicity, was also assessed.

54

The predicted probabilities of weight status underestimation were computed across a range of BMI z-score values and ages. Predicted probabilities were computed from logistic regression models using both the intercept and corresponding regression coefficients by first calculating the logarithm of the odds, then converting this to odds by exponentiating the result of the previous step. Odds were then transformed into probabilities. Given the following equation for a logistic regression model, the formula to compute the predicted probability of an event occurring for a given value of the covariate of interest (i.e. the probability that an adolescent underestimates their weight status at a given age) is:

These predicted probabilities were plotted separately for males and females.

The third set of regression models examined the effect of ethnicity on weight status underestimation. These models included the dummy variables defined above (see Section 3.2.4) to represent each of the ethnic groups under study. The interaction between sex and ethnicity was assessed to determine if the relationship between weight status underestimation and ethnicity was different for male and female adolescents. This was modeled by including an interaction term between each of the dummy variables for ethnicity and the dummy variable for sex. The overall effect of ethnicity on weight status underestimation was determined by comparing nested models (i.e. models with and without dummy variables for ethnicity); Section 3.5.1 below outlines the likelihood ratio test used to compare these models. In addition, comparisons across different ethnic groups were conducted by repeating regression models with different ethnic groups serving as the reference for each model. This was done for the crude effect of ethnicity (i.e. the model controlling only for the effect of interview mode) and the model additionally adjusted for the effects of severity of overweight and age.

55

3.4.4 Objective 3

The third objective examined the variation in weight status underestimation across communities and explored the effect of the weight status of community-based reference groups on variation in weight status underestimation across communities. Multilevel logistic regression, described in Section 3.3 above, was used to address this objective.

The variation across communities was assessed by allowing the intercept in the multilevel regression model to vary. The analysis was repeated using both operationalizations of community (health regions and CSDs). Separate models were run for males and females. To ensure variation across communities was not the consequence of differences in how data were collected in different regions, the model testing for community-level variation controlled for the effect of interview mode. The amount of variation in weight status underestimation across clusters was assessed by both the significance of the cluster-level variation and the MOR.

This objective also aimed to determine whether exposure to overweight explained any of the identified variation in weight status underestimation across communities. The effect of exposure to overweight in each of the different reference populations described above on weight underestimation was examined. This was done by introducing the prevalence of overweight in a reference population (i.e. CSD) to cluster-level equations for the intercept of the individual-level model. Analyses were completed separately for males and females and controlled for the effect of interview mode. The importance of exposure to overweight on weight status underestimation in each reference population was assessed using the IOR. The PCV was used to compare models with and without exposure to overweight as a cluster-level predictor of weight status underestimation.

3.5 Additional Statistical Considerations

3.5.1 Software & Algorithms

Descriptive statistics were computed using SAS software.240 All regression analyses, including the multilevel models, were computed using Mplus.241 Maximum likelihood estimation with robust standard errors (MLR) was used for all regression models. The

56 use of this estimation technique accounted for the complex survey design through the use of sampling weights. This estimation technique computes standard errors using a sandwich estimator. The chi-squared tests produced in these models are asymptotically equivalent to the Yuan-Bentler T2* test statistic.241 Satorra and Bentler242 developed a simple formula for a likelihood ratio test that can be computed using the both the loglikelihood and the value of a scaling correction factor provided by Mplus. Loglikelihood ratio tests were used to compare the effects of multiple regression coefficients (i.e. for the effect of ethnicity) simultaneously. The first required step for this test was to compute a new scaling factor, based on the scaling factors from the models under the null and alternative hypotheses using the following formula:

where is the number of parameters in the model specified under the null hypothesis

is the number of parameters in the model specified under the alternative hypothesis

is the scale correction for the model specified under the null hypothesis

is the scale correction for the model specified under the alternative hypothesis

The new scale correction factor was then combined with the loglikelihood values for each of the models being compare, using the following formula to get a chi-squared test statistic:

where is the loglikelihood for the model specified under the null hypothesis

is the loglikelihood for the model specified under the alternative hypothesis

57

Mplus uses numerical integration techniques when estimating multilevel models.241 When compared to other estimation techniques for multilevel logistic regression (i.e. quasi-likelihood), the precision of estimates is increased.232

3.5.2 Missing Data

Due to the nature of sample selection and the derivation of variables for the final model (i.e. the use of a dummy variable for information on missing ethnicity), no eligible respondents had missing data for any of the independent variables included in the regression models. Less than 1% (n=45) of those in the final sample had missing data for the outcome of interest. These individuals were excluded from the analysis.

3.5.3 Survey Weights

All analyses were computed using the sampling weights provided by Statistics Canada. These weights reflected the complex nature of the sampling design used in the CCHS and thus the unequal probabilities of selection, as well as non-response rates. Use of these weights ensured that all estimates reflect the true estimates for the Canadian population.

Weights provided by Statistics Canada were designed to scale estimates to the population level rather than the sample level. As a result, some software, including SAS, requires that these weights be rescaled prior to analyses so the total weighted sample size is equal to the actual sample size. Weights were rescaled by dividing an individual’s sampling weight by the mean sampling weight for each cycle.

Descriptive statistics and variables aggregated only for adolescents were computed using weights that had been rescaled for the adolescent sample only. That is, the weight provided was divided by the mean weight for all adolescents. Descriptive statistics and variables aggregated for all respondents were computed using weights that had been rescaled for all respondents. That is, the sample weight provided was divided by the mean weight for all respondents at each cycle. All survey weights were rescaled prior to excluding ineligible respondents from the sample. All regression analyses (both logistic and multilevel logistic) used the weights as provided by Statistics Canada, as Mplus does not require that weights be rescaled.241

58

3.5.4 Interview Mode

A mode study conducted as part of the CCHS (2003, Cycle 2.1) found that the accuracy of reported height and weight varied significantly with interview mode among adults.243 The difference between self-reported and measured height and weight was greater for those who were interviewed by telephone than those interviewed in person. The average BMI of telephone respondents was significantly lower than the average BMI reported by those interviewed in person. The authors of the mode study cited social desirability and interviewer variability as reasons for the observed differences across interview modes.243 When being interviewed, respondents have a tendency to report in a way that is seen to be more socially acceptable.244 The effects of social desirability bias are particularly pronounced for interviews conducted in person. For example, reports of smoking tend to be lower for in person interview than for either in a telephone interview or a web- or paper-based questionnaire. Smoking is seen by many as a negative behavior and responding in person limits the anonymity of the respondent, leading the respondent to feel like their behavior is being judged by the interviewer. Overall, this leads to decreased accuracy of reporting for in person interviews. The opposite is true with regard to reporting of height and weight. When being interviewed in person, respondents were more likely to accurately report their height and weight since the interviewer would be better able to determine if their response was accurate than for respondents answering by telephone.

Similarly, interviewer variability may also have been responsible for some of the variation across interview modes.243 Canada is a very large and diverse country. To complete interviews of 130,000 Canadians, a large number of interviewers would have been required. These interviewers would also have been spread out over great geographic distance, making continued interviewer training difficult. Differences also may have existed between modes in the amount of deviation from the script across modes.

The above study compared the effect of interview modes on responses for several health- related questions included in the CCHS. However, this comparison was made only for those over the age of 18.243 Among overweight adolescent CCHS respondents, those

59 interviewed by telephone were significantly more likely to underestimate their weight status than those interviewed in person (RR 1.13, 95% CI 1.09-1.17). This was true for both males (RR 1.09, 95% CI 1.05-1.13) and females (RR 1.17, 95% CI 1.09-1.25). The average BMI z-score for overweight adolescents interviewed by telephone was 0.031 (95% CI 0.018, 0.045) standard deviation units lower than those who were interviewed in person. Since the proportion of individuals interviewed by each mode varied across survey cycles and geography, and it is associated with outcome, interview mode was controlled for in all analyses. Individuals selected from the telephone frame were all interviewed over the telephone, whereas those selected from the area frame could have participated in either in-person or telephone interviews.

The coding for the interview mode variable in the CCHS identifies individuals who either were (1) interviewed in person, (2) interviewed by telephone, and (3) mixed-mode interviews. Very few individuals participated in mixed-mode interviews. Since these individuals would likely first have been interviewed in person, these individuals are grouped with those interviewed in person. A dummy variable for interview mode was created so that interview by telephone was the reference group. This variable was centred in all regression models, allowing all estimates to reflect the ‘average’ value for interview mode. That is, parameter estimates (i.e. the intercept) reflect the overall population instead of being specific to only those interviewed by telephone.

60

Chapter 4 4 Results

This chapter begins with a discussion of the characteristics of the sample, including the prevalence of overweight among adolescents (Section 4.1). This is followed by a discussion of the magnitude of weight status underestimation among Canadian adolescents (Section 4.2) and individual-level characteristics that are associated with the likelihood of underestimating one’s weight status (Section 4.3). This chapter concludes with a presentation of the results from the multilevel analysis, including assessment of variation in weight status underestimation across health regions and CSDs and the role of exposure to overweight in explaining variation in weight status underestimation across communities (Section 4.4).

4.1 Sample Characteristics

Among the Canadian population aged 12 and older, 30.7% were overweight; 14.6%, obese (Table 2). While the prevalence of overweight remained relatively stable between 2001 and 2010, there was a slight increase in the prevalence of obesity (from 14% to 16%). The chi-square test for differences across survey years was statistically significant (p<0.001). Among those between the ages of 12 and 18, the overall prevalence of overweight was 11.5%, with yearly prevalence estimates ranging from 10.8% (2007) to 12.6% (2009); see Table 2. The prevalence of obesity was 5.1% overall, with prevalence estimates ranging from 4.8% (2009 and 2010) to 5.5% (2001). The chi-square test for differences across survey years was statistically significant (p=0.0015). Overweight and obese adolescents are together referred to as overweight for the remainder of this chapter. The characteristics of all overweight adolescents are provided in Table 3.

Table 2. Prevalence of overweight and obesity among Canadians in the Canadian Community Health Survey (CCHS) Survey Year N (%) 2001 2003 2005 2007 2008 2009 2010 Total All Canadians aged 12 and over (p<0.001) Normal weight 73,861 74,404 71,892 36,277 36,130 33,318 33,655 359,537 (56.2%) (54.9%) (54.1%) (55.0%) (54.7%) (54.0%) (53.3%) (54.7%) Overweight 39,793 42,220 41,671 19,874 19,858 18,616 19,348 201,379 (30.3%) (31.1%) (31.3%) (30.1%) (30.1%) (30.2%) (30.6%) (30.7%) Obese 17,881 18,949 19384 9795 10,025 9744 10,189 95,968 (13.6%) (14.0%) (14.6%) (14.9%) (15.2%) (15.8%) (16.1%) (14.6%) Adolescents ( aged 12-18) only (p=0.0015) Normal weight 12,816 13,846 11.983 5247 5669 5348 5354 60,262 (83.1%) (83.4%) (83.1%) (84.4%) (84.2%) (82.6%) (83.4%) (83.3%) Overweight 1763 1884 1661 669 777 816 763 8334 (11.4%) (11.3%) (11.5%) (10.8%) (11.5%) (12.6%) (11.9%) (11.5%) Obese 840 880 780 304 291 310 305 3710 (5.5%) (5.3%) (5.4%) (4.9%) (4.3%) (4.8%) (4.8%) (5.1%)

61

62

Table 3. Characteristics of Adolescent Respondents (aged 12 to 18) to the Canadian Community Health Survey from 2001 through 2010 (CCHS) All Adolescents Overweight Adolescents Frequency (%) Frequency (%) Perceived Weight Status Perceived overweight 9057 (13.3%) 5261 (46.1%) Perceived normal or underweight 59,065 (86.7%) 6145 (53.9%) Sex Male 36,963 (51.1%) 7233 (63.2%) Female 35,343 (48.9%) 4219 (36.8%) Ethnicity White 55,560 (76.8%) 8835 (77.1%) Black 1933 (2.7%) 422 (3.7%) Asian 7952 (11.0%) 934 (8.2%) Aboriginal 2520 (3.5%) 536 (4.7%) Latin American 718 (1.0%) 121 (1.1%) Other 994 (1.4%) 178 (1.6%) Mixed 1211 (1.7%) 200 (1.7%) Missing 1428 (2.0%) 226 (2.0%) Mean (SD) Mean (SD) Age 15.0 (2.0) 15.0 (1.9) BMI z-score 0.13 (1.07) 1.55 (0.37) Abbreviations: SD (standard deviation); BMI (body mass index) 4.2 Magnitude of Weight Status Underestimation

Overall, 53.9% of overweight adolescents underestimated their weight status (Table 4). The prevalence of weight status underestimation among males ranged from 56.1% (2001) to 65.8% (2008). The prevalence of weight status underestimation in females ranged

62

from 37.6% in 2001 to 47.3% in 2010 (Table 4). Females were significantly less likely to underestimate their weight status than males—females had about half the odds of underestimation of males (OR 0.46, 95% CI 0.41-0.52; Table 5). Overall, there was a significant trend in weight status underestimation for both males and females: curvilinear in males and linear in females (Figure 1). A significant interaction term between the quadratic term for time and sex (p=0.037) supported the difference in trends for males and females.

Table 4. Perceived weight status of overweight adolescents from 2001 to 2010 among adolescent (aged 12-18) participants of the Canadian Community Health Survey (CCHS) Survey Year 2001 2003 2005 2007 2008 2009 2010 Total N (%) N (%) N (%) N (%) N (%) N (%) N (%) N (%) Perceived Weight Status All Overweight Adolescents (p<0.001) Overweight 1,113 1,173 1,089 433 464 521 468 5261 (51.0%) (44.4%) (46.0%) (46.0%) (43.8%) (46.7%) (43.9%) (46.1%) Normal or 1,072 1,470 1,277 539 595 595 598 6145 Underweight (49.0%) (55.6%) (54.0%) (55.5%) (56.2%) (53.3%) (56.1%) (53.9%) Total 2,185 2,644 2,366 972 1,059 1,116 1,066 11,407 Male Overweight Adolescents (p<0.001) Overweight 593 625 579 232 228 295 253 2806 (34.9%) (36.9%) (38.5%) (36.2%) (34.2%) (42.7%) (38.5%) (38.9%) Normal or 759 1067 926 409 439 397 405 4402 Underweight (56.1%) (63.1%) (61.5%) (63.8%) (65.8%) (57.4%) (61.5%) (61.1%) Total 1,352 1,692 1,505 640 668 693 658 7,208 Female Overweight Adolescents (p=0.0082) Overweight 520 548 510 201 236 226 215 2455 (62.4%) (57.6%) (59.2%) (60.7%) (60.3%) (53.3%) (52.7%) 58.5%) Normal or 313 403 351 130 155 198 193 1743 Underweight (37.6%) (42.4%) (40.8%) (39.3%) (39.7%) (46.7%) (47.3%) (41.5%) Total 832 951 861 331 391 423 408 4,199

63

Table 5. Logistic regression models exploring the trend in weight status underestimation among Canadian overweight adolescents between 2001 and 2010 in the Canadian Community Health Survey (CCHS) Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) All Overweight Adolescents Time 1.02 (1.00, 1.04) 1.07 (1.00, 1.14) 1.02 (1.00, 1.04) 1.06 (0.99, 1.14) 1.01 (0.99, 1.04) 1.11 (1.02, 1.22) Time2 1.00 (0.99, 1.00) 1.00 (0.99, 1.00) 0.99 (0.98, 1.00) Sex (male=0) 0.46 (0.41, 0.52) 0.46 (0.41, 0.53) 0.41 (0.34, 0.50) 0.50 (0.40, 0.64) Time*sex 1.02 (0.98, 1.06) 0.88 (0.76, 1.01) Time2*sex 1.02 (1.00, 1.03) Male Overweight Adolescents Time 1.01 (0.9, 1.04) 1.11 (1.02, 1.22) Time2 0.99 (0.98, 1.00) Female Overweight Adolescents Time 1.04 (1.00, 1.07) 0.97 (0.87, 1.09) Time2 1.01 (0.99, 1.02) Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All models controlled for the effect of interview mode. Models: (1) linear time trend; (2) curvilinear time trend; (3) linear time trend controlling for the effect of sex; (4) curvilinear time trend controlling for the effect of sex; (5) interaction between linear time trend and sex; (6) interaction between curvilinear time trend and sex. Abbreviations: OR (odds ratio); CI (confidence interval)

64

65

0.7

0.6

0.5

0.4

0.3 Males Females 0.2

Probability of Underestimation of Probability 0.1

0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Year

Figure 1. Predicted probability of weight status underestimation among overweight adolescents across time (2001 to 2010) in the Canadian Community Health Survey (CCHS)

4.3 Effect of Individual-Level Characteristics

Severity of Overweight

As the severity of an adolescent’s overweight increased, the probability that an adolescent underestimated their weight status decreased significantly (Figure 2). There was a curvilinear association between weight status underestimation and BMI z-score

(Table 6). Both the linear and quadratic components of this relationship were 65

significantly different for males and females (i.e. both interaction terms were significant at the 5% level). The rate of decline in underestimation occurred sooner and was much steeper for females than for males. The effect of severity of overweight remained significant in both males and females, after controlling for age, ethnicity, and the effect of time (Table 7).

66

1

0.9

0.8 0.7 0.6 0.5 Males 0.4 Females 0.3

0.2 Probability of Underestimation of Probability 0.1 0 1 1.5 2 2.5 3 BMI z-Score

Figure 2. Predicted probability of weight status underestimation across a range of BMI z-scores. Note: Diamonds correspond to the 85th percentile; squares, 90th percentile; triangles, 95th percentile.

66

Table 6. Logistic regression models exploring the effect of the severity of overweight on weight status underestimation among overweight adolescents Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) All Overweight Adolescents BMI z-score 0.22 (0.18, 0.26) 0.02 (0.01, 0.08) 0.17 (0.14, 0.20) 0.01 (0.00, 0.05) 0.14 (0.11, 0.18) 0.02 (0.00, 0.13) BMI z-score2 2.01 (1.36,2.97) 2.13 (1.24, 3.22) 1.75 (1.03, 2.98) Sex (male=0) 0.35 (0.31, 0.41) 0.35 (0.30, 0.40) 0.16 (0.08, 0.30) 3.46 (0.31, 39.0) BMI z-score*sex 1.73 (1.13, 2.66) 0.03 (0.00, 0.54) BMI z-score2*sex 3.71 (1.55, 9.86) Overweight Male Adolescents BMI z-score 0.14 (0.11, 0.18) 0.02 (0.00, 0.14) BMI z-score2 1.74 (1.02, 2.95) Overweight Female Adolescents BMI z-score 0.25 (0.17, 0.35) 0.00 (0.00, 0.01) BMI z-score2 6.54 (3.27, 13.1) Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All models controlled for the effect of interview mode. Models: (1) linear effect of BMI z-score; (2) curvilinear effect of BMI z-score; (3) linear effect of BMI z-score controlling for the effect of sex; (4) curvilinear effect of BMI z-score controlling for the effect of sex; (5) interaction between linear effect of BMI z-score and sex; (6) interaction between curvilinear effect of BMI z-score and sex Abbreviations: OR (odds ratio); CI (confidence interval); BMI (body mass index)

67

Table 7. Adjusted logistic regression models combining all individual-level effects (age, severity of overweight, and ethnicity) and the time trend from 2001 through 2010 Males Females OR (95% CI) OR (95% CI) Age (units: years) 0.94 (0.89, 0.98) 0.78 (0.74, 0.83) Severity of Overweight BMI z-score 0.02 (0.00, 0.13) 0.001 (0.000, 0.006) BMI z-score2 1.66 (1.02, 2.69) 6.06 (2.98, 12.34) Time Time 1.15 (1.04, 1.27) 1.05 (1.01, 1.08) Time2 0.99 (0.98, 1.00) - Ethnicity† White Reference Reference Black 3.39 (1.51, 7.61) 1.50 (0.76, 2.95) Asian 0.58 (0.40, 0.83) 0.56 (0.31, 1.02) Aboriginal 0.94 (0.67, 1.32) 0.76 (0.48, 1.21) Latin American 1.33 (0.48, 3.65) 0.67 (0.24, 1.89) Other Cultural Origin 1.68 (0.68, 4.13) 1.11 (0.35, 3.46) Multiple Cultural Origins 0.86 (0.46, 1.62) 0.67 (0.32, 1.39) Ethnicity Missing 1.24 (0.75, 2.06) 0.90 (0.45, 1.79) Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All models controlled for the effect of interview mode; (3) Only terms that were significant when exploring the crude effects of each of these covariates were included. Abbreviations: OR (odds ratio); CI (confidence interval); BMI (body mass index) †White is the reference group. Odds ratios for all adjusted comparisons across ethnic groups are provided in Appendix E.

68

69

Age

Younger adolescents were more likely to underestimate their weight status than older adolescents (Figure 3). While males had higher rates of weight status underestimation at all ages, the gap between males and females widened as adolescents increased in age. Logistic regression models exploring the relationship between age and weight status underestimation, and the differences in this relationship for males and females, are provided in Table 8. In females, the odds ratio for weight status underestimation associated with a one-year increase in age was 0.81 (95% CI 0.77-0.85). In males, this same odds ratio was 0.92 (95% CI 0.88-0.96; Table 8). The effect of age remained significant in both males and females, after controlling for BMI z-score, ethnicity, and the effect of time (Table 7).

1

0.9 0.8 0.7 0.6 0.5 Males 0.4 Females 0.3

0.2

Probability of Underestimation of Probability

0.1 0 12 13 14 15 16 17 18

69

Age (years)

Figure 3. Predicted probability of weight status underestimation for overweight adolescents between the ages of 12 and 18

Table 8. Logistic regression models exploring the effect of age on weight status underestimation among overweight adolescents Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) All Overweight Adolescents Age 0.88 (0.86, 0.91) 1.45 (0.85, 2.48) 0.88 (0.85, 0.91) 1.27 (0.73, 2.20) 0.92 (0.88, 0.96) 1.74 (0.86, 3.49) Age2 0.98 (0.97, 1.00) 0.99 (0.97, 1.01) 0.98 (0.96, 1.00) Sex (male=0) 0.45 (0.40, 0.51) 0.46 (0.40, 0.52) 3.12 (1.14, 8.55) 181.0 (0.04, 844081.812) Age*Sex 0.88 (0.82, 0.94) 0.51 (0.16, 1.59) Age2*Sex 1.02 (0.96, 1.06) Overweight Males Adolescents Age 0.92 (0.88, 0.96) 1.74 (0.87, 3.50) Age2 0.98 (0.96, 1.00) Overweight Female Adolescents Age 0.81 (0.77, 0.85) 0.88 (0.36, 2.16) Age2 1.00 (0.97, 1.03) Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All models controlled for the effect of interview mode. Models: (1) linear effect of Age; (2) curvilinear effect of Age; (3) linear effect of Age controlling for the effect of sex; (4) curvilinear effect of Age controlling for the effect of sex; (5) interaction between linear effect of Age and sex; (6) interaction between curvilinear effect of Age and sex Abbreviations: OR (odds ratio); CI (confidence interval)

70

71

Ethnicity

Ethnicity was an important predictor of weight status underestimation, both among all overweight adolescents (i.e. males and females together), as well as in males only (p<0.001 for both overall tests). The overall effect of ethnicity was not significant for females (p=0.5570). The overall effect of the interaction between sex and ethnicity could not be determined due to a negative value for the adjusted likelihood ratio test.

Black adolescents were consistently the most likely to underestimate their weight status, while Asian adolescents were the least likely to underestimate their weight status. There were no significant differences between individual ethnic groups for females. Although not significant, Aboriginal adolescents were less likely to underestimate their weight status than White adolescents. Table 9 provides odds ratios (adjusted for the effects of interview mode) for the comparison of weight status underestimation across all ethnic groups in the full sample of adolescents; comparisons of individual ethnic groups for males and females are provided in Tables 10 and 11, respectively

The overall effect of ethnicity on weight status underestimation remained significant for males after controlling for the effects of age, severity of overweight, and time. Individual comparisons across ethnic groups remained unchanged from the unadjusted model, with the addition of Aboriginal male adolescents being significantly more likely to underestimate their weight status than Asian male adolescents (Appendix E).

71

After controlling for these characteristics, the overall effect of ethnicity remained non- significant in females (p=0.1427). However, differences between individual ethnic groups were observed after controlling for the effect of age and the quadratic effect of BMI z-score (Appendix E). Specifically, Asian females were significantly less likely to underestimate their weight status than Black females (OR 0.38, 95% CI 0.15-0.93). Additionally, a marginally significant (p=0.060) difference was observed when comparing levels of weight status underestimation among Asian females to White females, with Asian females being less likely to underestimate their weight status than White females (OR 0.56, 95% CI 0.31-1.02).

Table 9. Crude odds ratios comparing weight status underestimation across ethnic groups among all overweight adolescents Reference Group White Black Asian Aboriginal Latin American Other Mixed Ethnic Group OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Black 1.8 (1.2, 2.7) - Asian 0.7 (0.6, 1.0) 0.4 (0.3, 0.7) - Aboriginal 0.8 (0.7, 1.1) 0.5 (0.3, 0.7) 1.1 (0.8, 1.6) - Latin American 1.0 (0.5, 1.9) 0.5 (0.2, 1.2) 1.3 (0.6, 2.7) 1.1 (0.6, 2.3) - Other 1.7 (0.9, 3.1) 0.9 (0.5, 1.9) 2.3 (1.2, 4.5) 2.0 (1.1, 3.9) 1.8 (0.7, 4.4) - Mixed 0.8 (0.5, 1.2) 0.4 (0.3, 0.8) 1.1 (0.7, 1.8) 1.0 (0.6, 1.6) 0.8 (0.4, 1.9) 0.5 (0.2, 1.0) - Missing 1.1 (0.7, 1.7) 0.6 (0.4, 1.1) 1.5 (0.9, 2.4) 1.3 (0.8, 2.1) 1.2 (0.5, 2.5) 0.7 (0.3, 1.4) 1.4 (0.8, 2.4) Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All OR are controlled for the effect of interview mode. Abbreviations: OR (odds ratio); CI (confidence interval).

72

Table 10. Crude odds ratios comparing weight status underestimation across ethnic groups among male overweight adolescents Reference Group White Black Asian Aboriginal Latin American Other Mixed Ethnic Group OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Black 2.9 (1.6, 5.3) - Asian 0.7 (0.4, 0.9) 0.2 (0.1, 0.4) - Aboriginal 0.9 (0.7, 1.3) 0.3 (0.2, 0.6) 1.4 (0.9, 2.1) - Latin American 1.2 (0.5, 3.2) 0.4 (0.1, 1.3) 1.9 (0.7, 5.0) 1.3 (0.5, 3.6) - Other 2.0 (0.9, 4.2) 0.7 (0.3, 1.8) 3.0 (1.3, 6.6) 2.2 (1.0, 4.8) 1.6 (0.5, 5.3) - Mixed 0.8 (0.5, 1.4) 0.3 (0.1, 0.6) 1.2 (0.7, 2.2) 0.9 (0.5, 1.6) 0.7 (0.2, 1.9) 0.4 (0.2, 1.0) - Missing 1.2 (0.8, 2.0) 0.4 (0.2, 0.9) 1.9 (1.1, 3.3) 1.4 (0.8, 2.4) 1.0 (0.4, 2.9) 0.6 (0.3, 1.5) 1.5 (0.8, 3.0) Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All OR are controlled for the effect of interview mode. Abbreviations: OR (odds ratio); CI (confidence interval).

73

Table 11. Crude odds ratios comparing weight status underestimation across ethnic groups among overweight female adolescents Reference Group White Black Asian Aboriginal Latin American Other Mixed Ethnic Group OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Black 1.5 (0.9. 2.7) - Asian 0.7 (0.4, 1.3) 0.5 (0.2, 1.1) - Aboriginal 0.8 (0.6, 1.2) 0.6 (0.3, 1.1) 1.1 (0.6, 2.1) - Latin American 0.7 (0.3, 1.9) 0.5 (0.2, 1.5) 1.0 (0.3, 3.0) 0.9 (0.3, 2.5) - Other 1.3 (0.4, 4.6) 0.9 (0.2, 3.6 1.8 (0.4, 6.9) 1.6 (0.4, 6.0) 1.8 (0.4, 9.1) - Mixed 0.8 (0.4, 1.7) 0.6 (0.2, 1.4) 1.1 (0.5, 2.7) 1.0 (0.5, 2.3) 1.2 (0.4, 3.9) 0.6 (0.2, 2.7) - Missing 0.9 (0.5, 1.8) 0.6 (0.3, 1.6) 0.7 (0.6, 0.9) 1.2 (0.5, 2.5) 1.3 (0.4, 4.4) 0.7 (0.2, 3.0) 1.1 (0.4, 4.1) Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All OR are controlled for the effect of interview mode. Abbreviations: OR (odds ratio); CI (confidence interval).

74

75

4.4 Multilevel Analyses

4.4.1 Variation across Clusters

There was significant variation in weight status underestimation when using health regions and census subdivisions to define community (Table 12). The variance across health regions was 0.107 (p<0.001); this corresponds to a MOR of 1.37 and an ICC of 0.032. The variance across clusters was higher when using CSDs as the definition of community than when health regions were used to define community: the variance across CSDs was 0.298 (p<0.001) with a MOR of 1.68 and ICC of 0.083. These models controlled for the effect of interview mode.

Table 12. Variance across reference health regions and census subdivisions in weight status underestimation among all overweight adolescents Reference Community Health Region Census Subdivision Variance (95% CI) Variance (95% CI) Intercept 0.107 (0.062, 0.152) 0.298 (0.176, 0.391) MOR 1.37 1.68 ICC 0.032 0.083 Notes: (1) Results significant at the 5% level are highlighted in bold; (2) Both models controlled for the effect of interview mode. Abbreviations: CI (confidence interval); MOR (median odds ratio); ICC (intraclass correlation coefficient)

Similar patterns in the variance across clusters were observed when analyses were stratified by sex (Tables 13 and 14 for males and females, respectively). The variation in

75 weight status underestimation for both sexes was lowest when health regions were used to define clusters: after adjusting for the individual-level effects of age, severity of overweight, and ethnicity, as well as the effect of time, the MOR was 1.44 for males and 1.67 for females. The ICCs for these models were 0.042 and 0.080, respectively. The variation in weight status underestimation was similarly higher when using census subdivisions to define community than for health regions. The MORs for the adjusted models were 1.73 for males and 2.17 for females; the ICCs for these same models were 0.091 and 0.167, respectively.

Table 13. Variance in weight status underestimation among overweight male adolescents across health regions and census subdivisions Reference Community Health Regions Census Subdivisions Model 1 Model 2 Model 1 Model 2 Variance (95% CI) Variance (95% CI) Variance (95% CI) Variance (95% CI) Intercept 0.124 (0.076, 0.172) 0.145 (0.068, 0.203) 0.332 (0.202, 0.461) 0.328 (0.177, 0.478) MOR 1.40 1.44 1.73 1.73 ICC 0.036 0.042 0.092 0.091 Notes: (1) Results significant at the 5% level are highlighted in bold; (2) Both models controlled for the effect of interview mode. Models: (1) null model: controlled only for the effect of interview mode; (2) model controlled for individual-level characteristics (age, BMI z- score, and ethnicity), time, and interview mode. Abbreviations: CI (confidence interval); MOR (median odds ratio); ICC: intraclass correlation coefficient

Table 14. Variance in weight status underestimation among overweight female adolescents across health regions and census subdivisions Reference Community Health Region Census Subdivisions Model 1 Model 2 Model 1 Model 2 Variance (95% CI) Variance (95% CI) Variance (95% CI) Variance (95% CI) Intercept 0.206 (0.106, 0.306) 0.286 (0.154, 0.418) 0.570 (0.322, 0.818) 0.661 (0.373, 0.949) MOR 1.54 1.67 2.05 2.17 ICC 0.059 0.080 0.148 0.167 Notes: (1) Results significant at the 5% level are highlighted in bold; (2) Both models controlled for the effect of interview mode. Models: (1) controlled only for the effect of interview mode; (2) model controlled for individual-level characteristics (age, BMI z-score, and ethnicity), time, and interview mode. Abbreviations: CI (confidence interval); MOR (median odds ratio); ICC: intraclass correlation coefficient

76

77

4.4.2 Exploring the Definition of Community

The effect of exposure to overweight on weight status underestimation among overweight male adolescents is presented in Tables 15 and 16 when health regions and census subdivisions, respectively, are used to define an adolescent’s community. In males, exposure to overweight was not a significant predictor of the variance across health regions or CSDs. This was true regardless of the reference group: all individuals in the community, all males in the community, all adolescents in the community, or all male adolescents in the community. The variance across clusters remained relatively stable across all models: the MOR for the residual variance was consistently 1.72 for CSDs and 1.43 for health regions.

The effect of exposure to overweight within an adolescent’s health region on weight status underestimation was also not significant among overweight females when health regions were used to define community (Table 17). However, when CSDs were used to define community, exposure to overweight was a significant predictor of weight status underestimation in females when all adolescents (Model 3) and female adolescents only (Model 4) were used to define the reference community (Table 18). The IORs for the effect of prevalence of overweight in all adolescents was 0.24-4.29 and 0.24-4.27 for females only. The residual variance in each model remained significant (p<0.001 for both) and the amount of variation explained by the prevalence of overweight was small (PCVs of 3.9% and 4.1%, respectively).

77

Table 15. Multilevel regression models examining the effect exposure to overweight within an adolescent’s health region on weight status underestimation among male overweight adolescents Model 1 Model 2 Model 3 Model 4 β (95% CI) β (95% CI) β (95% CI) β (95% CI) Exposure to Overweight Everyone 0.012 (-0.002, 0.026) IOR 0.51-1.99 All Males 0.008 (-0.007, 0.023) IOR 0.51-2.00 All Adolescents 0.013 (-0.008, 0.033) IOR 0.51-2.00 Male Adolescents 0.009 (-0.010, 0.028) IOR 0.51-2.00 Variance (95% CI) Variance (95% CI) Variance (95% CI) Variance (95% CI) Intercept* 0.139 (0.081, 0.197) 0.143 (0.084, 0.202) 0.141 (0.082, 0.199) 0.143 (0.085, 0.201) MOR 1.43 1.43 1.43 1.43 ICC 0.041 0.042 0.041 0.042 Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All models controlled for the age, severity of overweight, ethnicity, time, and interview mode. Models: Each model examines the effect of a different reference group on weight status underestimation among male adolescents. These reference groups are: (1) all respondents in an adolescent’s health region; (2) all male respondents in an adolescent’s health region; (3) all adolescent respondents in an adolescent’s health region; (4) all male adolescents in an adolescent’s health region. Abbreviations: CI (confidence interval); IOR (interval odds ratio); MOR (median odds ratio); ICC: intraclass correlation coefficient *Residual variance in the intercept after controlling for predictors of the variation across health regions.

78

Table 16. Multilevel regression models examining the effect exposure to overweight within an adolescent’s census subdivision (CSD) on weight status underestimation among male overweight adolescents Model 1 Model 2 Model 3 Model 4 β (95% CI) β (95% CI) β (95% CI) β (95% CI) Exposure to Overweight Everyone 0.008 (-0.040, 0.021) IOR 0.36-2.81 All Males 0.002 (-0.010, 0.015) IOR 0.36-2.82 All Adolescents 0.005 (-0.005, 0.016) IOR 0.34-2.94 Male Adolescents 0.006 (-0.002, 0.014) IOR 0.36-2.82 Variance (95% CI) Variance (95% CI) Variance (95% CI) Variance (95% CI) Intercept* 0.320 (0.171, 0.469) 0.326 (0.176, 0.477) 0.320 (0.170, 0.471) 0.323 (0.173, 0.472) MOR 1.72 1.72 1.72 1.72 ICC 0.087 0.090 0.087 0.089 Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All models controlled for the age, severity of overweight, ethnicity, time, and interview mode. Models: Each model examines the effect of a different reference group on weight status underestimation among male adolescents. These reference groups are: (1) all respondents in an adolescent’s CSD; (2) all male respondents in an adolescent’s CSD; (3) all adolescent respondents in an adolescent’s CSD; (4) all male adolescents in an adolescent’s CSD. Abbreviations: CSD (census subdivision); CI (confidence interval); IOR (interval odds ratio); MOR (median odds ratio); ICC: intraclass correlation coefficient *Residual variance in the intercept after controlling for predictors of the variation across CSDs.

79

Table 17. Multilevel regression models examining the effect exposure to overweight within an adolescent’s health region on weight status underestimation among female overweight adolescents Model 1 Model 2 Model 3 Model 4 β (95% CI) β (95% CI) β (95% CI) β (95% CI) Exposure to Overweight Everyone 0.016 (-0.003, 0.034) IOR 0.39-2.67 All Females 0.015 (0.000, 0.028) IOR 0.39-2.65 All Adolescents 0.018 (-0.008, 0.044) IOR 0.39-2.66 Female Adolescents 0.014 (-0.006, 0.035) IOR 0.39-2.64 Variance (95% CI) Variance (95% CI) Variance (95% CI) Variance (95% CI) Intercept* 0.284 (0.150, 0.418) 0.281 (0.148, 0.414) 0.281 (0.149, 0.412) 0.279 (0.148, 0.409) MOR 1.66 1.66 1.66 1.66 ICC 0.079 0.079 0.079 0.078 Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All models controlled for the age, severity of overweight, ethnicity, time, and interview mode. Models: Each model examines the effect of a different reference group on weight status underestimation among male adolescents. These reference groups are: (1) all respondents in an adolescent’s health region; (2) all female respondents in an adolescent’s health region; (3) all adolescent respondents in an adolescent’s health region; (4) all female adolescents in an adolescent’s health region. Abbreviations: CI (confidence interval); IOR (interval odds ratio); MOR (median odds ratio); ICC: intraclass correlation coefficient *Residual variance in the intercept after controlling for predictors of the variation across health regions.

80

Table 18. Multilevel regression models examining the effect exposure to overweight within an adolescent’s census subdivision (CSD) on weight status underestimation among female overweight adolescents Model 1 Model 2 Model 3 Model 4 β (95% CI) β (95% CI) β (95% CI) β (95% CI) Exposure to Overweight Everyone 0.008 (-0.008, 0.023) IOR 0.23-4.39 All Females 0.010 (-0.002, 0.022) IOR 0.23-4.37 All Adolescents 0.011 (0.000, 0.022) IOR 0.24-4.29 Female Adolescents 0.009 (0.001, 0.018) IOR 0.24-4.27 Variance (95% CI) Variance (95% CI) Variance (95% CI) Variance (95% CI) Intercept* 0.659 (0.370, 0.948) 0.653 (0.364, 0.942) 0.635 (0.348, 0.922) 0.634 (0.349, 0.919) MOR 2.17 2.16 2.14 2.14 ICC 0.167 0.166 0.162 0.162 Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All models controlled for the age, severity of overweight, ethnicity, time, and interview mode. Models: Each model examines the effect of a different reference group on weight status underestimation among male adolescents. These reference groups are: (1) all respondents in an adolescent’s health region; (2) all female respondents in an adolescent’s health region; (3) all adolescent respondents in an adolescent’s health region; (4) all female adolescents in an adolescent’s health region. Abbreviations: CSD (census subdivision); CI (confidence interval); IOR (interval odds ratio); MOR (median odds ratio); ICC (intraclass correlation coefficient) *Residual variance in the intercept after controlling for predictors of the variation across CSDs.

81

82

Chapter 5

5 Discussion

This study aimed to estimate the degree of weight status underestimation among all overweight Canadian adolescents, as well as examine the time trend in weight status underestimation between 2001 and 2010. This project also assessed the effect of demographic characteristics (i.e. age and ethnicity) and severity of overweight on weight status underestimation among these adolescents. Lastly, the effect of exposure to overweight within an adolescent’s community on weight status underestimation was examined. Section 5.1 provides an overview of the findings of this study and how they relate to current knowledge of weight status underestimation among overweight adolescents. The public health implications of these findings are discussed (Section 5.2), followed by a discussion of the challenges in making comparisons of weight status underestimation across studies (Section 5.3) and a general commentary on why levels of weight status underestimation may be so high among Canadian adolescents (Section 5.4). The chapter concludes with an overview of the strengths and limitations of this study (Sections 5.5 and 5.6, respectively) and some general conclusions and implications (Section 5.6).

5.1 Overview of Findings

More than half of all overweight Canadian adolescents do not recognize that they are

82

overweight. For every 5 overweight males, 3 do not recognize that they are overweight; 2 of 5 overweight females do not recognize that they are overweight. Not only are male adolescents more likely to be overweight than their female counterparts, but they are also more likely to not recognize that they are overweight. Estimates of weight status underestimation in this study are among the highest of all previous reported studies (Appendix A).

The time trend in weight status underestimation from 2001 to 2010 is of particular interest. The increasing prevalence of weight status underestimation among overweight

83 females is particularly noteworthy, especially when considering that the proportion of Canadian adolescents who are overweight has remained relatively stable between 2001 and 2010. This suggests that the overall number of females adolescents who are overweight yet fail to recognize their overweight status is increasing. The trend in male adolescents is not as clear. Any increase in weight status underestimation among adolescents poses important challenges for future strategies to address the obesity epidemic in adolescents.

5.1.1 Individual-Level Characteristics of Weight Status Underestimation

Severity of Overweight

As the severity of overweight increased the probability that an adolescent underestimated his or her weight status decreased substantially. Adolescents with a BMI close to the 85th percentile for their age and sex rarely recognized that they were overweight, while the majority of those at the 95th percentile and above recognized that they were overweight. These findings were consistent with previous studies examining differences in weight status underestimation for overweight and obese adolescents,4,6,9,11,16,20,33,35,39,40,50-52,195 as well as those that have examined the trend across a range of BMI or BMI z-score values.57,64,65,221

Females were more accurate in recognizing that they were overweight at a lower BMI than males. This may relate to the differences in overall accuracy for males and females, suggesting that the threshold adolescents use to define overweight may be lower among 83

females than among males. Identifying these thresholds has not been previously addressed and is an important area of future research in the area of weight status underestimation. Having a better understanding of what adolescents consider overweight may allow public health providers to better target overweight and obesity strategies at changing what this threshold currently is.

The high probability of weight status underestimation among adolescents at the lower end of the overweight spectrum points to the need for a special emphasis of strategies aimed at helping these adolescents understand what a healthy weight is. Without such an

84 intervention, these adolescents likely will not recognize the health risks they face and continue to engage in overweight-promoting behaviours.

Age

As adolescents increase in age, they become increasingly accurate in recognizing when they are overweight, although the magnitude of this association was much stronger for females than for males. These results contradict previous studies suggesting, instead, that increasing age is actually associated with increased risk of weight status underestimation.40,57,64 The observed differences in the effect of age on weight status underestimation, for both males and females, were consistent with previously reported studies finding that, at older ages, females were less likely to underestimate their weight status than males.27,195 Although female adolescents at the youngest ages in this study (12 years of age) did not underestimate their weight status more frequently than males of the same age—as was expected (i.e. see Kaltiala-Heino et al27 or Maximova et al195)—the predicted probabilities for the two sexes were much closer than for older adolescents. It is possible that the female adolescents outside the age range included in this study may, in fact, have higher rates of underestimation than their male counterparts.

Ethnicity

Consistent with previous studies examining the effect of ethnicity on weight status underestimation, Black adolescents were more likely to underestimate their weight status than White adolescents.5,11,16,18,22,33,58,221 This study also found that Asian adolescents

84 were the least likely to underestimate their weight status (compared to White and Black Canadian adolescents). Previous studies did not demonstrate a clear relationship between being of Asian descent and the risk of weight status underestimation, compared to those of other ethnic backgrounds.33,60

One particularly noteworthy finding regarding weight status underestimation in adolescents of different cultural backgrounds is the risk of underestimation among those of Aboriginal descent. This study found that Aboriginal adolescents were less likely to underestimate their weight status than both Black adolescents and those of other cultural

85 origin (i.e. were none of black, Asian, Aboriginal, Latin American, or multiple origins) in the full sample of adolescents. Aboriginal male adolescents were also less likely to underestimate their weight status than Black male adolescents; no significant differences were observed when comparing levels of weight status underestimation in Aboriginal male adolescents to those of other cultural backgrounds. Although not significant, Aboriginal adolescents were also less likely to underestimate their weight status than White adolescents; this was true for both the full sample, as well as for both males and females. This finding contradicts findings from the United States suggesting that Aboriginal adolescents were more likely to underestimate their weight status when compared to those of other ethnic backgrounds.58 Further, based on the suggestion that adolescents exposed to high levels of overweight are more likely to underestimate their weight status195,199 and the high prevalence of overweight among Aboriginal adolescents,100-102 it would be expected that levels of weight status underestimation would be high in this population. This was not the case in the present study. When considering the differences between Aboriginal adolescents and those of other ethnic backgrounds, it is important to note that the CCHS only included off-reserve Aboriginal people. The comparisons across ethnic groups may have been different if the available data were also representative of Aboriginal adolescents living on reserves.

5.1.2 Weight Status of Community-Based Reference Groups on Weight Status Underestimation

The variance in weight status underestimation was highly significant across both health regions and CSDs. The magnitude of this variation was higher when CSDs were used as 85

the operationalization of community than when health regions were used. Additionally, more of the variation across communities was explained when exploring the effect of an adolescent’s exposure to overweight in their CSD than in their health region. However, the effect of exposure to overweight on variance across communities was only significant for females when comparisons were made to all adolescents in their CSD and to only female adolescents in their CSD. Although this study did not find a strong association between community-level exposure to overweight and weight status underestimation, a high degree of residual variation in weight status underestimation did exist across these

86 communities. This variation supports the importance of neighbourhoods as targets of both preventive and intervention strategies in combating the obesity epidemic.89,245

Prior studies have identified both classmate and parental BMI as important predictors of weight status underestimation65,195,221 and those in our environment influence perceptions of body weight.199-201 However, these previous studies have not examined all potential spheres that may influence an adolescent’s perception of their weight status. In particular, adolescents may be additionally influenced by those in their extracurricular activities or neighbourhood. This study examined the effect of exposure to overweight on weight status underestimation in a broader context. Although the source of exposure to overweight from previous studies was not a direct match to the current study, similar results were expected. . The results of this study suggest that, although weight status underestimation among females may be influenced by the weight status of community- based reference groups, the evidence from this study is not strong. Future research should focus on understanding on defining “community” from an adolescent’s perspective. These reference groups may not necessarily be defined by geography and may not be reflective of an adolescent’s family or school community. Further, these comparison groups may be different for adolescents of different ages—expanding on findings that suggest that adolescents of different ages are impacted differently by schoolmate and parental weight status.195 By identifying the precise sources of comparison, this may enable improved targeting of weight management interventions addressing weight status underestimation towards these individuals. The limitations

associated with the definitions of community used are explained further in Section 86

5.6.1.3.

5.2 Weight Status Underestimation & Public Health

The majority of overweight and obese adolescents recognize that there are significant health risks associated with being overweight and obese: 95% are aware that being overweight and obese impacts health; 94%, hypertension; 95%, heart disease; and 78%, diabetes.10 Although these adolescents understand the health risks of being overweight, they may not necessarily recognize that they, themselves, face these health risks. Those that do not recognize these risks subsequently may not have the motivation required to

87 engage in weight management behaviours.61-63 In fact, a discrepancy does exist between the perceived health risks associated with being overweight and those associated with being too fat. Adolescents associate the term overweight with unhealthiness; however, this term is rarely used by adolescents to describe their weight. Overweight adolescents instead refer to themselves as being fat or big—terms that are not associated with the same perceptions of unhealthiness.246 This recognition that being overweight poses significant health risks is important in understanding the relationship between weight status underestimation and motivation to engage in weight management behaviours. This study was conducted in Scotland and may reflect culture-specific interpretations of weight-related terminology. Little is known about the role of overweight terminology in adolescents’ recognition of health risks in the Canadian context and the role this may play in the understanding of weight status underestimation among Canadian adolescents and an important area of future research.

Consequently, addressing weight status underestimation among adolescents is an important component of any weight management strategy, particularly as a tool to help overweight adolescents understand their future health risks. The lack of effectiveness of current weight management strategies in the adolescent population may partially stem from their inability to distinguish between those who do and those who do not recognize that they are overweight and address these adolescents accordingly. Although no studies specifically report the role of underestimation in the effectiveness of these programs, motivation to lose weight is imperative to their success184 and accurately recognizing 61-63 oneself as overweight is important in establishing this motivation. 87

Tackling the issue of weight status underestimation among overweight adolescents is not a straightforward task. Adolescents who consider themselves overweight have poorer psychosocial well-being than those who do not, regardless of their actual weight status.7,9,16,24,26,28,30,32,39,42,48,49,54-56,61,143,146-148,162,163,189-192 As a result, any strategies aimed at helping adolescents accurately recognize that they are overweight must do so in a way is protective of their psychosocial well-being.

88

Better understanding weight status underestimation may also play an important role in decreasing the prevalence of overweight—not only for those who are currently overweight, but also to minimize the effect of the socially contagious nature of obesity. Previous literature suggests that exposure to overweight in one’s family and school environments can have a substantial impact on the persistence and development of obesity. Children and adolescents whose parents are overweight are at an increased risk of becoming overweight than those whose parents are not overweight.81,82,85,86,94,128,247,248 Further, high classmate BMI increases the risk of being overweight among males, but not females.205 Junior high school students are influenced by the weight of older students in their high school; schools that have a high prevalence of overweight among senior students are also likely to have a high prevalence of overweight among junior students controlling for individual-level risk factors for overweight.203

5.3 Challenges in Studying Weight Status Underestimation

Comparison of the results found in this study to previous studies of weight status underestimation is made difficult by differences both in how perceived weight status is measured and in how underestimation is defined across studies. For example, the levels of weight status underestimation observed in this study (61.1% for males and 41.5% for females) are higher than those found among patients at a paediatric gastroenterology clinic in Hamilton, Ontario13 but lower than those from a provincially-representative sample of Quebec adolescents.195 The study conducted in Hamilton defined weight status underestimation to reflect all adolescents who selected a weight status that was smaller 88

than their actual weight status, including normal weight adolescents identifying as underweight. The results of this study may have been different if the study focused only on those adolescents who were overweight.

In comparison, the study conducted in Quebec used a figure-rating scale to measure perceived weight status. Each of the seven figures in this scale was assigned a BMI z- score (i.e. -3 through +3). The z-score corresponding to the figure selected by adolescents was then subtracted from the adolescents’ actual BMI z-scores. Adolescents who were considered to underestimate their weight status may have actually accurately

89 recognized that they were overweight had they been asked to describe their weight status. Instead, they may have underestimated the degree to which they were overweight. These individuals are included in the reported levels of weight status underestimation195 but that same individual would not have been identified as underestimating their weight status in the present study. This may, consequently, explain the differences in weight status underestimation between these two studies.

5.4 Reasons for Weight Status Underestimation among Canadian Adolescents

Defining Overweight

A discussion of the magnitude of weight status underestimation among Canadian adolescents would be remiss without considering the standards adolescents use to assess their own weight status. Of particular note is the lack of consistency in the objective standards used to define overweight in the adolescent population.

Although it has been recommended that a single definition be used for identifying overweight and obesity among children and adolescents,249 there remains a lack of consistency in the definitions of overweight and obesity used in the literature on weight status underestimation (Appendix A). This makes a comparison of the levels of weight status underestimation among overweight and obese adolescents across studies challenging. Further, this lack of a consistent definition may actually be contributing to levels of weight status underestimation in the adolescent population. Specifically, some

89

adolescents may be identified as overweight or obese using one growth reference, but not by another. For example, an adolescent may be considered to be overweight using the CDC and WHO growth references, but not by the IOTF reference. If that adolescent bases their weight on the IOTF reference but they participated in a study using the WHO growth reference, that adolescent would be considered to underestimate their weight status. As a consequence of there being no consistent means of determining whether or not an adolescent is overweight based on his or her BMI, one cannot expect that they will be able to accurately describe their weight status. This is especially true for adolescents at the lower end of the overweight spectrum, where the growth references may not

90 consistently describe an adolescent’s weight status. Despite the differences that exist between growth references, the prevalence of overweight and obesity are similar across all three.218

The impact of multiple growth references on identification of an individual’s weight status may be further complicated by differences in the terminology used in each of these growth references. While both the IOTF and WHO references use the terms overweight and obese, this was not the terminology used in the original publication of the CDC growth references. Initially, the CDC growth reference deemed children and adolescents at or above the 95th percentile for their age and sex to be overweight; those between the 85th and 94th percentile for their age and sex were considered at risk of overweight.214 These terminologies and cut points were recommended by Barlow and Dietz62 on the basis that they should be based on both percentiles and the association between BMI and negative health outcomes. Identifying children and adolescents with a BMI at or above the 85th percentile for their age and sex was originally designed to identify those at higher risk of developing obesity—a disease based on risk due to extra fat—and that these individuals should undergo further screening for other health concerns such as cardiovascular risk factors and insulin resistance.212 A BMI in the 95th percentile during adolescence is a strong predictor of adult obesity, while a BMI between the 85th and 94th percentiles increases one’s risk of developing obesity-related disease.

Since the publication of the CDC growth curves, the Institute of Medicine has instead recommended that the term overweight be replaced with obese.250 The rationale behind

90 this change was two-fold: (1) to better reflect the serious nature of such a high BMI and (2) to increase the comparability of the CDC growth reference with other growth references and with adult definitions of overweight and obesity. This recommendation has been adopted by experts in the field of childhood obesity.209,210 It has subsequently been recommended that children and adolescents between the 85th and 94th percentile be considered overweight instead of at risk for overweight.209 Although adolescents with a BMI at or above the 85th percentile for their age and sex are considered overweight for this thesis, those between the 85th and 94th percentiles may not have accurately

91 recognized that they overweight simply because they believed themselves to be at risk for overweight rather than actually being overweight.

The multiplicity of growth references is especially problematic in Canada where no national growth reference exists. Instead, we must rely either on references developed to monitor the growth of children and adolescents in one country (i.e. the CDC growth charts designed for American children and adolescents) or on international references (i.e. the WHO or IOTF growth references). In Canada, recommendations regarding which growth references should be used to monitor child and adolescent growth are continuously evolving. In less than ten years, two different recommendations have been made regarding the monitoring of child and adolescent growth. In 2004, Dietitians of Canada, along with the Canadian Paediatric Society, the College of Family Physicians of Canada, and the Community Health Nurses Association of Canada, recommended that the CDC growth references (using the terminology of at risk for overweight and overweight) be used to monitor growth.219 However, this recommendation was modified only six years later, with the WHO growth reference now being recommended.220 Both recommendations advise the use of the IOTF growth reference for studies examining the prevalence of overweight and obesity in the Canadian population.219,220 Also in 2010, Shields and Tremblay218 advised that the IOTF growth references be used for epidemiologic purposes, while the CDC growth curves be used for monitoring individual growth. Given the inconsistent nature of these recommendations, it is not surprising that there rates of weight status underestimation are so high among Canadian adolescents.

91

This lack of a standard definition to identify overweight adolescents is further complicated by the multiplicity of tools used to measure obesity status.212 In particular, other measures, such as body fat percentage, measure body fatness while BMI is simply a measure of body weight (relative to height).212 Since overweight and overfatness do not directly correspond (i.e. overweight does not differentiate between fat mass and fat-free mass) there is a discrepancy between individuals who would be identified as being either overweight or obese based on anthropometric measures, such as BMI, but would not be overweight or obese based on measures of body composition, such as body fat percentage. As such, these adolescents may be inaccurate in perceiving themselves as

92 not overweight when comparing to their BMI but these individuals accurately recognize that they do not face any additional health risks because of their weight. However, one study examining weight status underestimation among overweight and obese adolescents found that the levels of underestimation were similar when using weight status based on BMI and body fat percentage: 43.4% underestimated relative to their BMI and 44.1% relative to their percent body fat.17

Physicians play an important role in communicating these objective standards for overweight to their adolescent patients. In particular, it is imperative that physicians discuss potential downstream weight-related health complications. However, the discussion of weight between physicians and their child and adolescent patients may be limited at present. A recent study of American parents of children with a BMI greater than or equal to the 85th percentile for their child’s age and sex found that less than one- quarter recalled their child’s weight status being a topic of discussion with the child’s physician.251 Even among children and adolescents who are severely overweight, less than 60% of parents recalled their child’s weight being a topic of discussion with their child’s physician. The proportion of parents who report that their child’s physician discussed their child’s overweight status did increase in 2007-2008, when compared to 1999 through 2006.251 This study, however, did not differentiate between those who saw a physician and those who did not.

In addition, physicians’ perceptions do not accurately reflect the true weight status of their child and adolescent patients. When paediatric gastroenterologists were asked to

92 describe their patients weight as underweight, slightly underweight, average, slightly overweight, or overweight, 28.9% underestimated the weight status of their male patients between the ages of 5 and 18; 38.7% of female patients had their weight status underestimated by their physician.13 Similar results were found when comparing physicians’ perceptions of their patients’ weight status using a figure-rating scale. Although these numbers are lower than those for parental perceptions of their child or adolescent’s weight status (46.5% for males and 40.6% for males) and adolescent self- perceptions of weight status (44.0% for males and 35.0% for females), physicians underestimating their patients’ weight statuses is an important problem. When physicians

93 fail to accurately perceive the weight status of their patients, this presents a further barrier to discussing weight status and the need to engage in appropriate weight management behaviours with that patient.

Further, there is a failure by physicians to diagnose children as being overweight— between 20% and 53% of overweight children do not have their weight status documented on their medical chart.252-255 A lack of confidence in the ability of these children to lose weight following identification as being either overweight or obese is commonly cited as a reason for not diagnosing these children as being overweight.256 In addition, physicians are dissuaded from using these tools to diagnose obesity since there are no clear associations between any of these growth references and the development of obesity-related pathology.256 These guidelines, however, are useful for screening purposes and when a child screens positively for overweight or obesity should undergo further diagnostic testing.212

Parents present further challenges to health care providers in the identification of overweight among children and adolescents. A recent study of American parents found that terms such as fat and obese were associated with negative connotations including being stigmatizing and not motivational for engaging in weight management behaviours.257 The term overweight is, however, considered to be motivational for children and adolescents to lose weight, while terms such as weight, unhealthy weight, and weight problem are preferred by parents. Although these terms may not be considered to be stigmatizing by parents, the degree to which these terms emphasize the

93 importance of the health risks associated with being overweight may be limited. For example, a qualitative study of adolescents suggested that the term overweight is associated with negative health risks but other terminology used to describe one’s weight status was not.246 Adolescents rarely use the term overweight when describing their own weight or the weight of their peers.

The reluctance of parents to have physicians use appropriate terminology to describe the weight status of their children and the failure of overweight adolescents to use this terminology to describe their own weight provides further explanation as to why so many

94 adolescents underestimate their weight status. This raises the question about the ability of physicians to effectively communicate weight-related issues with their overweight and obese adolescent patients and the role parents play in this communication. However, when considering the protective effects of weight status underestimation on the psychosocial well-being of overweight adolescents, these terminologies may be useful in ensuring adolescents recognize the health risks they face without experiencing any potential negative implications to their mental health. It is important to further understand the role of specific terminologies in both inspiring engagement in weight management behaviours and shaping the mental health of overweight adolescents.

Weight-Based Comparisons

The discussion thus far only takes into consideration the possibility that overweight adolescents who underestimate their weight status are using objective criteria as a comparison of their weight. Instead, it is likely that adolescents base their perceptions of their weight on more subjective standards. The Theory of Endogenous Weight Norms plays an important role in understanding why adolescents underestimate their weight status. This theory states that individuals prefer to be thinner than the average person; however, as average BMI increases, preferred weight also increases.200 As a result, it has been suggested that what once was considered overweight is becoming the new normal.199 Based on this theory, it was expected that adolescents living in an area with a high prevalence of overweight would be more likely to underestimate their weight status than adolescents living in an area with a lower prevalence of overweight. However, the

94 present study provided little support for the notion that the exposure to overweight in an adolescent’s community is an important predictor of weight status underestimation.

5.5 Strengths

5.5.1 Representativeness of the CCHS

The primary strength of this project is its use of a national population-based survey and thus its ability to provide estimates of weight status underestimation for all overweight Canadian adolescents. Statistics Canada uses complex sampling designs to ensure the sample obtained is representative of the entire adolescent population. The sampling

95 strategy used by Statistics Canada ensured that 98% of the Canadian population was included in the sample population. Survey weights were used to ensure that estimates obtained accurately reflect the Canadian population and included adjustments for non- response.

5.5.2 Measure of Perceived Weight Status

This project uses a single item to assess perceived weight status. The majority of studies examining weight status perception rely on similar measures. While a figure-rating scale may be beneficial in that it allows respondents to select figures that they consider to be ideal and which figure they perceive as overweight11 most studies use figure rating scales as a means of looking at weight status dissatisfaction (i.e. compare respondents’ ideal weight with that of their current weight). Figure-rating scales may also allow for a continuous measure of perceived weight status provided a sufficient number figures are included in the scale used.195 However, the use of a figure-rating scale does not necessarily require respondents to assign a specific label, such as overweight or too fat, to a figure. As overweight becomes increasingly normal, adolescents may select a larger figure but not accurately recognize that this figure represents overweight. Instead, word- based measure of perceived weight status allows for adolescents to identify their particular weight status as being overweight. This definitive description of weight status would not have been possible if a figure-rating scale had been used. Since people’s preferences for ideal body weight are increasing199 and overweight is becoming the norm,200 an adolescent selecting a figure deemed to be overweight by researchers may not be, in fact, considered to represent overweight by the respondent. An ideal study would 95

rely on a combination of the two measures, allowing for adolescents to describe whether or not they thought of a particular figure (i.e. the one that they selected to represent their own body weight) as overweight.

Levels of weight status underestimation are similar when comparing a figure-rating scale to a single word-based question of perceived weight status. For example, Alwan et al8 found that there was a high degree of similarity between a descriptive and a pictorial measure of body size estimation, and that the relationship between perception and appropriate weight management strategies were independent of the measure used.

96

Chaimovitz et al13 similarly used both a word-based question and a visual scale to measure perceived weight status because they did not want participants’ responses to be influenced by the choice of words used in the measure. Levels of weight status underestimation were, likewise, similar for both the figure-rating scale and the word- based measure of perceived weight status.13

The question used to measure perceived weight status in the CCHS had three possible responses: overweight, just about right, and underweight. This scale, however, may have failed to take into account subtle differences in the degree of overweight or obesity. Although most studies using a scale that can measure these subtle differences by using a 5-point Likert-type scale (i.e. differentiate between slightly overweight and overweight), most group responses into three categories. As a result, minor subtleties in adolescents’ responses may have been missed. For example, Al-Sendi et al6 observed that 66.7% of obese male adolescents perceived themselves as overweight rather than obese. Although for the purposes of the discussion in this thesis, these individuals were not considered to underestimate their weight status, these individuals may in fact be systematically different from the obese adolescents who accurately perceived themselves as being obese (19.4% of obese male adolescents) with regards to recognition of consequent health risks and their motivation to engage in weight management behaviours. As another example, Cheung et al15 found that 29.0% of overweight adolescents considered themselves to be mildly overweight instead of severely overweight.

Overweight is a term that adolescents recognize as being associated with increased health

96 246 risks, but not a term that is commonly used by this population to describe their weight. As Chaimovitz et al13 suggest, there may be differences in adolescents’ responses to questions about their weight status depending on the terminology used. Examples of different terminology used in measures of perceived weight status include fat, overweight, and heavy. These different terminologies have different meanings for adolescents246 and the interchanging of these definitions may result in the measurement of slightly different constructs. Potential variation in estimates of weight status underestimation using these different terminologies has not been explored. This is an important area for future research, particularly with regard to understanding the role of

97 different terminologies in adolescent understanding of future health risks associated with being overweight.

5.5.3 Expanded Definition of Exposure to Overweight

Previous studies have focused only on exposure to overweight among schoolmates and parents.195,221 For example, Maximova et al195 found that both schoolmate and parental weight had independent influence on levels of weight status underestimation among adolescents. However, these are not the only potential sources of weight-based comparisons for the adolescent population—adolescents are likely influenced by other individuals in their community, including both those in their neighbourhood and in those they encounter as part of their extracurricular activities. This study expanded on what was already known by exploring the effect of exposure to overweight on the levels of weight status underestimation among overweight and obese adolescents. This includes a comparison not only of two different reference communities (i.e. health regions and CSDs), but also different people within those communities.

5.6 Limitations

5.6.1 Measures

5.6.1.1 Perceived Weight Status & Recency Effects

Adolescents’ responses to the question used to measure perceived weight status in the CCHS may have biased due to a recency effect. A recency effect typically occurs during interviews being conducted by telephone or in person and involves respondents being 97

likely to select a category that was provided at the end of the survey question rather than one at the beginning.244 The opposite occurs when a survey is self-administered (i.e. web or paper based). The question used in the CCHS listed overweight as the first possible option; just about right as the final option potentially making it more likely to be selected by respondents. Consequently, the survey design may have actually overinflated the estimates of weight status underestimation. In comparison, other studies have asked respondents to describe their weight status by asking a question that had ordered

98 responses (i.e. ranged from very underweight to very overweight); Appendix A provides a comparison of the wording used in previous studies of weight status underestimation.

5.6.1.2 Actual Weight Status

This study used BMI calculated from self-reported height and weight to determine adolescents’ actual weight status. It is important to note that BMI is a screening mechanism for adolescent overweight and obesity—not a diagnostic tool.211 A primary limitation of using BMI to measure overweight and obesity is its inability to differentiate between fat mass and fat-free mass.213 Obesity is technically defined as excess fat mass, not excess fat-free mass.211 This distinction is an important one to make in understanding the relationship between BMI and health risks—BMI has a high specificity but poor sensitivity for predicting obesity-related morbidity later in life.111 Because BMI is not a perfect measure of body fatness, its use may result in a high degree of misclassification of weight status.258 It is possible that individuals overweight based on their BMI were in fact correct in not perceiving that they were overweight based on the results of further diagnostic testing. That is, although these individuals had a high weight relative to their height, their high weight was the consequence of high fat-free mass rather than high fat mass (i.e. they had a low body fat percentage). This may have, consequently, resulted in higher levels of weight status underestimation than if a diagnostic tool was used. This discrepancy may be particularly pronounced for males and for individuals of certain ethnic backgrounds: the ability of BMI to accurately identify an individual as overweight is decreased for males.213,258 These significant ethnic variations in BMI are not reflected in current growth curves213 and may help explain the differences in weight status 98

underestimation across ethnic groups.

In addition to the inability of BMI to differentiate between fat mass and fat-free mass, the tools used to classify adolescents’ BMI as overweight or obese also raise concerns. In exploring the validity of BMI-based growth curves to generate a definition of obesity, it must be asked how accurate it is to base the definitions of overweight and obesity on a sample of one national population given the differences in fat and fat-free mass across ethnic groups, as is the case with both the CDC214 and WHO216 growth curves. The IOTF215 curve tries to counteract this and uses data from six national surveys, but the

99 authors acknowledge that the data used are not necessarily representative of the global population of school-aged children and adolescents (the Americas were well represented, with data from both the United States and Brazil being included, but other areas received little or no representation). The authors adjust for the differing levels of overweight and obese in these six national samples. Parts of the globe that may experience a higher prevalence of underweight or thinness are among those not represented and if included may have had an impact on the construction of these growth curves.

Concern should also be raised with regard to the use of a percentile-based definition of obesity. If the prevalence of childhood obesity continues to climb, the BMI for children and adolescents at the 85th and 95th percentiles will climb since, based on the CDC definition of obesity, these are what define overweight and obesity. In this case, the prevalence of overweight would be 15% and the prevalence of obesity would be 5%. If these definitions are applied to the data on the Canadian prevalence of overweight (26%) and obese (8%),1,2 the prevalence, by definition, would decrease and the BMI corresponding to the 85th percentile for overweight and 95th percentile for obese would increase. In fact, de Onis259 states that the use of a percentile-based definition for obesity will underestimate the prevalence of overweight and obesity, while overestimating the number of underweight individuals. BMI z-scores for overweight individuals are lower with the 2000 CDC growth curves then with its predecessor, the 1977 NCHS growth curves. Should growth curves continue to be based on historical data, these definitions would not be responsive to changes in the prevalence of overweight and obesity over 216 time. On this note, WHO used the same data that was used to construct a set of growth 99

curves in 1977 and both the CDC214 and IOTF215 excluded the most recent American data (1988-1994) in their analysis.

Despite concerns associated with the use of BMI to identify overweight adolescents, similar levels of weight status underestimation were observed when actual weight status was determined using BMI and body fat percentage. Specifically, a study conducted in New Zealand found that 43.4% of overweight adolescents underestimated their weight status when using BMI to identify actual weight status; 44.1% underestimated their weight status relative to actual weight status calculated from percent body fat.17

100

Self-Reported vs. Measured Height & Weight

Additional concerns are raised when using self-reported height and weight to calculate an adolescent’s BMI. Overall, when BMI is calculated from self-reported height and weight, the prevalence of obesity tends to be underestimated.11,260-263 Males are more accurate in reporting their height and weight than females.263 As adolescents age, they become less accurate in reporting their height, but increasingly accurate in reporting their weight and BMI.261-263 Accuracy of self-reported BMI also decreased with increasing weight status.11,262,263 There is no clear relationship between ethnicity and accuracy in reporting.263 A study conducted in the United States found that levels of weight status underestimation were more conservative when using self-reported height and weight to measure actual weight status than when measured height and weight were used.11 As such, it is likely that the results reported in this present study are also an underestimate of the levels of weight status underestimation among overweight Canadian adolescents.

5.6.1.3 Weight Status of Community-Based Reference Groups

The ability to investigate the effect of exposure to overweight within an adolescent’s community on the likelihood that they underestimate their weight status is limited given the survey design and sampling technique used by Statistics Canada for the CCHS. That is, neither health regions nor CSDs may have provided an ideal reference community for adolescents’ perceptions of their weight. A better comparison may be made by exploring the degree of overweight and obesity within an adolescent’s peer group, both inside and

outside of school. Further, without including a measure of the standards adolescents use 100 as a comparison for their weight no definitive answer can be made regarding the influence of exposure to overweight on levels of weight status underestimation.

It is possible that the non-significant results observed are a consequence of within-cluster heterogeneity. This is especially likely in health regions that were combined to maintain consistency across time. It is possible that combined health regions (and other large health regions with low population density) may be variable in the prevalence of overweight and obesity, and adolescents in these areas may only be influenced by a geographic subpopulation within a health region. Non-significant results may also result

101 from no clear means of defining exposure to overweight. For example, this study examined the role of prevalence of overweight in a community-based reference group on weight status underestimation. Different results may have been obtained had median BMI or BMI z-score been used to describe the weight status of community-based reference groups.

In addition, data sparseness may have posed threats to the validity of estimates based on CSDs. Since some CSDs would have only had two overweight adolescents after being stratified by sex, this may have impacted the results. However, it has been suggested that the number of clusters (i.e. CSDs) is more important than the number of individuals per cluster in obtaining accurate estimates for parameters in multilevel regression analysis.231 Further, a sensitivity analysis comparing different minimum sample sizes per CSD was conducted, finding similar results for CSDs with a minimum size of 2 and 5 (Appendix D).

5.6.2 Interview Mode

The CCHS is an important source of data on the health of Canadians. However, challenges are introduced by the sheer complexity of the study. In particular, the vast geographic coverage of the survey requires that more than one mode of data collection be used (i.e. telephone and in-person interviews). The use of multiple interview modes is complicated by the variation in the proportion of interviews conducted using each mode across geography and time, both factors of interest in this study. Further, reporting of

height and weight, as well as perceived weight status, are subject to reporting bias. A 101 mode study conducted as part of the CCHS reports differences in the reporting of these variables across survey modes among adults243 and a comparison of these characteristics for adolescents in this study suggest the same is true for adolescents. Although interview mode was controlled for in the analysis, bias in reporting may not have been eliminated entirely.

5.6.3 Response Rates

Response rates for the CCHS declined consistently across the course of the CCHS (i.e. from 2001 to 2010). Consequently, any findings across time may have been a

102 consequence of the decreasing response rates. To compensate, Statistics Canada incorporates non-response when determining sample weights.

5.6.4 Temporality

Because this study is cross-sectional, temporality cannot be assured. As a result, this study can only identify factors that are associated with underassessment of weight status and not factors that lead to the development of weight status underestimation. Nonetheless, this study identifies important characteristics of adolescents who underestimate their weight status, and may be beneficial in the design and targeting of future weight management interventions.

5.7 Conclusions and Implications

The results of this study point to a large discrepancy between actual and perceived overweight among overweight Canadian adolescents. The importance of understanding weight status underestimation in this population to tackle the obesity epidemic has been highlighted previously.66 Targeting weight status underestimation in this population should be a target of future weight management programs and public health strategies taken towards the overweight and obesity epidemic.146 However, the actual relationship between weight status underestimation and health outcomes (including physical health and psychosocial well-being) is not yet understood and remains an area of future research. Through increased engagement in weight management behaviours, these adolescents decrease their risk of developing obesity-related pathology such as

102 cardiovascular disease and type II diabetes. However, they are also more likely to engage

in unhealthy weight management behaviours. Further, these adolescents also tend to have poorer psychosocial well-being, experiencing more depressive symptoms and having lower self-esteem. The best approach to addressing weight status underestimation among overweight Canadian adolescents is not yet known—but the solution will need to be one that incorporates the complex nature of weight status underestimation for both physical and psychosocial well-being.

103

References

1. Shields M. Overweight and obesity among children and youth. Health Rep. 2006;17(3):27-42. http://www.statcan.gc.ca/pub/82-003-x/2005003/article/9277- eng.pdf. Accessed November 14, 2011.

2. Shields M. Overweight Canadian children and adolescents: Statistics Canada; 2008. http://www.statcan.gc.ca/pub/82-620-m/2005001/article/child-enfant/8061- eng.htm. Accessed September 14, 2011.

3. Lobstein T, Baur L, Uauy R. Obesity in children and young people: a crisis in public health. Obes Res. 2004;5(suppl 1):4-85. http://onlinelibrary.wiley.com/doi/10.1111/j.1467- 789X.2004.00133.x/abstract;jsessionid=9145E38A8868AADB52E88AACE99EB 464.d03t01. Accessed September 14, 2011.

4. Abbott RA, Lee AJ, Stubbs CO, Davies PSW. Accuracy of weight status perception in contemporary Australian children and adolescents. J Pediatr Child Health. 2010;46(6):343-348. http://onlinelibrary.wiley.com/doi/10.1111/j.1440- 1754.2010.01719.x/abstract. Accessed November 21, 2011.

5. Akers AY, Lynch CP, Gold MA, et al. Exploring the Relationship Among Weight, Race, and Sexual Behaviors Among Girls. Pediatrics. 2009 2009;124(5):e913-e920. http://pediatrics.aappublications.org/content/124/5/e913.long. Accessed December 30, 2011.

6. Al-Sendi AM, Shetty P, Musaiger AO. Body weight perception among Bahraini adolescents. Child Care Health Dev. 2004;30(4):369-376. http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2214.2004.00425.x/abstract. Accessed November 21, 2011.

103 7. Al Mamun A, Cramb S, McDermott BM, O'Callaghan M, Majman JM, Williams GM. Adolescents' perceived weight associated with depression in young adulthood: a longitudinal study. Obesity (Silver Spring). 2007;15(12):3097-3105. http://www.nature.com/oby/journal/v15/n12/full/oby2007369a.html. Accessed September 14, 2011.

8. Alwan H, Viswanathan B, Paccaud F, Bovet P. Is accurate perception of body image associated with appropriate weight-control behavior among adolescents of the Seychelles. J Obes. 2011;2011:81742. http://www.hindawi.com/journals/jobes/2011/817242/. Accessed September 14, 2011.

9. Blond A, Whitaker AH, Lorenz JM, et al. Weight concerns in male low birth weight adolescents: relation to body mass index, self-esteem, and depression. J

104

Dev Behav Pediatr. 2008;29(3):166-172. http://journals.lww.com/jrnldbp/pages/articleviewer.aspx?year=2008&issue=0600 0&article=00004&type=abstract. Accessed September 14, 2011.

10. Bodenlos JS, Rosal MC, Blake D, Lemay C, Elfenbein D. Obesity prevalence, weight-related beliefs and behaviors among low-income ethnically diverse national job corps students. J Health Dispar Res Pract. 2010;3(3):106-114. http://escholarship.umassmed.edu/peds_adolescent/22/. Accessed September 29, 2011.

11. Brener ND, Eaton DK, Lowry R, McManus T. The association between weight perception and BMI among high school students. Obes Res. 2004;12(11):1866- 1874.

12. Brug J, Wammes B, Kremers S, Giskes K, Oenema A. Underestimation and overestimation of personal weight status: associations with socio-demographic characteristics and weight maintenance intentions. J Hum Nutr Dietet. 2006;19:253-262. http://onlinelibrary.wiley.com/doi/10.1111/j.1365- 277X.2006.00707.x/abstract. Accessed September 14, 2011.

13. Chaimovitz R, Issenman R, Moffat T, Persad R. Body perception: do parents, their children, and their children's physicians perceive body image differently? J Pediatr Gastr Nutr. 2008;47(1):76-80. http://journals.lww.com/jpgn/pages/articleviewer.aspx?year=2008&issue=07000 &article=00012&type=abstract. Accessed September 14, 2011.

14. Chen M-Y, Fan J-Y, Jane S-W, Wu J-Y. Do overweight adolescents perceive the need to reduce weight and take healthy actions? J Nurs Res. 2009;17(4):270-277.

15. Cheung PCH, Ip PLS, Lam ST, Bibby H. A study on body weight perception and weight control behaviours among adolescents in Hong Kong. Hong Kong Med J. February 2007;13(1):16-21. http://www.hkmj.org/abstracts/v13n1/16.htm. Accessed September 14, 2011.

104 16. Daniels J. Weight and weight concerns: Are they associated with reported depressive symptoms in adolescents? J Pediatr Health Care. 2005;19(1):33-41. http://www.sciencedirect.com/science/article/pii/S0891524504001889. Accessed September 14, 2011.

17. Duncan JS, Duncan EK, Schofield G. Associations between weight perceptions, weight control and body fatness in a multiethnic sample of adolescent girls. Public Health Nutr. 2011;14(1):93-100. http://journals.cambridge.org/action/displayAbstract?fromPage=online&aid=7947 915. Accessed September 29, 2011.

18. Edwards NM, Pettingell S, Borowsky IW. Where perception meets reality: self- perception of weight in overweight adolescents. Pediatrics. 2010;125(3):e452- e458.

105

http://pediatrics.aappublications.org/cgi/pmidlookup?view=long&pmid=2014228 1. Accessed April 13, 2012.

19. Eichen D, Conner B, Daly B, Fauber R. Weight perception, substance use, and disordered eating behaviors: comparing normal weight and overweight high- school students. J Youth Adolesc. 2012;41(1):1-13. http://www.springerlink.com/content/yj15j34w6422106x/. Accessed September 14, 2011.

20. Fagan HB, Diamond J, Myers R, Gill JM. Perception, intention, and action in adolescent obesity. J Am Board Fam Med. 2008;21(6):555-561. http://www.jabfm.org/content/21/6/555.long. Accessed September 14, 2011.

21. Farré Rovira R, Frasquet Pons I, Martínez Martínez MI, Romá Sánchez R. Self- reported versus measured height, weight and body mass index in Spanish Mediterranean teenagers: effects of gender, age and weight on perceptual measures of body image. Ann Nutr Metab. 2002;46(2):68-72.

22. Foti K, Lowry R. Trends in perceived overweight status among overweight and nonoverweight adolescents. Arch Pediatr Adolesc Med. 2010;164(7):636-642. http://archpedi.jamanetwork.com/article.aspx?articleid=383410. Accessed September 29, 2011.

23. Florin TA, Shults J, Stettler N. Perception of overweight is associated with poor academic performance in US adolescents. J Sch Health. 2011;81(11):663-670. http://onlinelibrary.wiley.com/doi/10.1111/j.1746-1561.2011.00642.x/abstract. Accessed July 4, 2012.

24. Frisco ML, Houle JN, Martin MA. The image in the mirror and the number on the scale: weight, weight perceptions, and adolescent depressive symtpoms. J Health Soc Behav. 2010;51(2):215-228.

25. Isomaa R, Isomaa A-L, Marttunen M, Kaltiala-Heino R, Björkqvist K.

Longitudinal concomitants of incorrect weight perception in female and male 105 adolescents. Body Image. 2011;8(1):58-63. http://www.sciencedirect.com/science/article/pii/S1740144510001130. Accessed September 26, 2011.

26. Kaplan SL, Busner J, Pollack S. Perceived weight, actual weight, and depressive symptoms in a general adolescent sample. Int J Eat Disord. 1988;7(1):107-113. http://onlinelibrary.wiley.com/doi/10.1002/1098- 108X(198801)7:1%3C107::AID-EAT2260070111%3E3.0.CO;2-%23/abstract. Accessed September 27, 2011.

27. Kaltiala‐Heino R, Kautiainen S, Virtanen SM, Rimpelä A, Rimpelä M. Has the adolescents' weight concern increased over 20 years? Eur J Public Health. 2003;13(1):4-10. http://eurpub.oxfordjournals.org/content/13/1/4.long. Accessed September 26, 2011.

106

28. Kim J-S, Lee K. The relationship of weight-related attitudes with suicidal behaviors in Korean adolescents. Obesity (Silver Spring). 2010;18(11):2145-2151. http://www.nature.com/oby/journal/v18/n11/full/oby201062a.html. Accessed September 14, 2011.

29. Kurdak H, Bozdemir N, Saatci E, Ozturk P, Ozcan S, Alkpinar E. Self-perceived body weight status and weight-control behaviors of high school students in a southern city of Turkey. Coll Antropol. 2010;4(34):1295.

30. Kurth B-M, Ellert U. Perceived or true obesity: which causes more suffering in adolescents? Findings of the Gernman Health Interview and Examination Survey for Children and Adolescents (KIGGS). Dtsch Arztebl Int. 2008;105(23):406-412. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2696846/. Accessed September 14, 2011.

31. Lee G, Ha Y, Vann JJ, Choi E. Weight perception and dieting behavior among Korean adolescents. J Sch Nurs. 2009;25(6):427-435.

32. Lenhart CM, Daly BP, Eichen DM. Is accuracy of weight perception associated with health risk behaviors in a diverse sample of obese adolescents? J Sch Nurs. 2011;27(6):416-423.

33. Martin MA, Frisco ML, May AL. Gender and race/ethnic differences in inaccurate weight perceptions among U.S. adolescents. Womens Health Issues. 2009;19(5):292-299. http://www.sciencedirect.com/science/article/pii/S1049386709000553. Accessed September 14, 2011.

34. Neumark-Sztainer D, Wall MM, Larson N, et al. Secular trends in weight status and weight-related attitudes and behaviors in adolescents from 1999 to 2010. Prev Med. 2012;54(1):77-81. http://www.sciencedirect.com/science/article/pii/S0091743511003963. Accessed July 4, 2012.

106 35. O'Dea JA, Amy NK. Perceived and desired weight, weight related eating and exercising behaviors, and advice received from parents among thin, overweight, obese or normal weight Australian children and adolescents. Int J Behav Nutr Phys Activity. 2011;8(1):68. http://www.ijbnpa.org/content/8/1/68. Accessed April 17, 2012.

36. O'Dea JA, Caputi P. Association between socioeconomic status, weight, age and gender, and the body image and weight control practices of 6- to 19-year-old children and adolescents. Health Educ Res. 2001;16(5):521-532. http://her.oxfordjournals.org/cgi/pmidlookup?view=long&pmid=11675800. Accessed September 14, 2011.

37. O'Haver J, Szalacha LA, Kelly S, Jacobson D, Melnyk BM. The relationships among body size, biological sex, ethnicity, and healthy lifestyles in adolescents. J

107

Spec Pediatr Nurs. 2011;16(3):199-206. http://onlinelibrary.wiley.com/doi/10.1111/j.1744- 6155.2011.00290.x/abstract;jsessionid=D6639659AC35A47C94D799D364DF86 B7.d03t02. Accessed September 14, 2011.

38. Ojala K, Tynjala J, Valimaa R, Villberg J, Kannas L. Overweight adolescents' self-perceived weight and weight control behaviour: HSBC study in Finland 1994-2010. J Obes. 2012;2012:180176. http://www.hindawi.com/journals/jobes/2012/180176/ref/. Accessed June 22, 2012.

39. Ozmen D, Ozmen E, Ergin D, et al. The association of self-esteem, depression and body satisfaction with obesity among Turkish adolescents. BMC Public Health. 2007;7:80. http://www.biomedcentral.com/1471-2458/7/80. Accessed September 14, 2011.

40. Park E. Overestimation and underestimation: adolescents' weight perception in comparison to BMI-based weight status and how it varies across socio- demographic factors. J Sch Health. 2011;81(2):57-64. http://onlinelibrary.wiley.com/doi/10.1111/j.1746-1561.2010.00561.x/abstract. Accessed September 14, 2011.

41. Pasch KE, Klein EG, Laska MN, Velazquez CE, Moe SG, Lytle LA. Weight misperception and health risk behaviors among early adolescents. Am J Health Behav. 2011;35(6):797-806. http://www.ingentaconnect.com/content/png/ajhb/2011/00000035/00000006/art00 015?token=00531e2311f88fec4e18e9437a63736a6f3b47742148662a774523446f 644a467b4d616d3f4e4b34a6c. Accessed June 25, 2012.

42. Perrin EM, Boone-Heinonen J, Field AE, Coyne-Beasley T, Gordon-Larsen P. Perception of overweight and self-esteem during adolescence. Int J Eat Disord. 2010;43(5):447-454. http://onlinelibrary.wiley.com/doi/10.1002/eat.20710/abstract. Accessed

September 14, 2011. 107

43. Pritchard ME, King SL, Czajka-Narins DM. Adolescent body mass indices and self-perception. Adolescence. 1997;32(128):863-880.

44. Shi Z, Lien N, Nirmal Kumar B, Holmboe-Ottesen G. Perceptions of weight and associated factors of adolescents in Jiangsu Province, China. Public Health Nutr. 2007;10(3):298-305. http://journals.cambridge.org/action/displayAbstract?fromPage=online&aid=6943 40. Accessed September 14, 2011.

45. Skinner AC, Weinberger M, Mulvaney S, Schlundt D, Rothman RL. Accuracy of perceptions of overweight and relation to self-care behaviors among adolescents with type 2 diabetes and their parents. Diabetes Care. 2008;31(2):227-229.

108

http://care.diabetesjournals.org/content/31/2/227.long. Accessed September 14, 2011.

46. Standley R, Sullivan V, Wardle J. Self-perceived weight in adolescents: over- estimation or under-estimation? Body Image. 2009;6(1):56-59. http://www.sciencedirect.com/science/article/pii/S1740144508001010. Accessed September 14, 2011.

47. Stigler M, Arora M, Dhavan P, Shrivastav R, Reddy K, Perry C. Weight-related concerns and weight-control behaviors among overweight adolescents in Delhi, India: A cross-sectional study. Int J Behav Nutr and Phys Act. 2011;8(1):9. http://www.ijbnpa.org/content/8/1/9. Accessed June 28, 2012.

48. Tang J, Yu Y, Du Y, Ma Y, Zhu H, Liu Z. Association between actual weight status, perceived weight and depressive, anxious symptoms in Chinese adolescents: a cross-sectional study. BMC Public Health. 2010;10(1):594. http://www.biomedcentral.com/1471-2458/10/594. Accessed September 14, 2011.

49. ter Bogt TFM, van Dorsselaer SAFM, Monshouwer K, Verdurmen JEE, Engels RCME, Vollebergh WAM. Body mass index and body weight perception as risk factors for internalizing and externalizing problem behavior among adolescents. J Adolesc Health. 2006;39(1):27-34. http://www.sciencedirect.com/science/article/pii/S1054139X05004301. Accessed September 14, 2011.

50. Ursoniu S, Putnoky S, Vlaicu B. Body weight perception among high school students and its influence on weight management behaviors in normal weight students: a cross-sectional study. Wien Klin Wochenschr. 2011;123(11):327-333. http://www.springerlink.com/content/n21q3p58272v3880/. Accessed September 14, 2011.

51. Viner RM, Haines MM, Taylor SJC, Head J, Booy R, Stansfeld S. Body mass, weight control behaviours, weight perception and emotional well being in a

multiethnic sample of early adolescents. Int J Obes (Lond). 2006;30:1514-1521. 108 http://www.nature.com/ijo/journal/v30/n10/full/0803352a.html. Accessed September 14, 2011.

52. Wang Y, Liang H, Chen X. Measured body mass index, body weight perception, dissatisfaction and control practices in urban, low-income African American adolescents. BMC Public Health. 2009;9:183. http://www.biomedcentral.com/1471-2458/9/183. Accessed September 26, 2011.

53. Xie B, Chou C-P, Spruijt-Metz D, et al. Weight perception and weight-related sociocultural and behavioral factors in Chinese adolescents. Prev Med. 2006;42(3):229-234. http://www.sciencedirect.com/science/article/pii/S0091743505002689. Accessed September 14, 2011.

109

54. Xie B, Chou C-P, Spruijt-Metz D, et al. Weight perception, academic performance, and psychological factors in Chinese adolescents. Am J Health Behav. 2006;30(2):115-124.

55. Xie B, Chou C-P, Spruijt-Metz D, et al. Longitudinal analysis of weight perception and psychological factors in Chinese adolescents. Am J Health Behav. 2011;35(1):92-104.

56. Xie B, Liu C, Chou C-P, et al. Weight perception and psychological factors in Chinese adolescents. J Adolesc Health. 2003;33(3):202-210. http://www.sciencedirect.com/science/article/pii/S1054139X03000995. Accessed September 14, 2011.

57. Yan AF, Zhang G, Wang MQ, Stoesen CA, Harris BM. Weight perception and weight control practice in a multiethnic sample of US adolescents. South Med J. 2009;102(4):354-360. http://journals.lww.com/smajournalonline/pages/articleviewer.aspx?year=2009&i ssue=04000&article=00007&type=abstract. Accessed September 14, 2011.

58. Yost J, Krainovich-Miller B, Budin W, Norman R. Assessing weight perception accuracy to promote weight loss among U.S. female adolescents: a secondary analysis. BMC Public Health. 2010;10(1):465. http://www.biomedcentral.com/1471-2458/10/465. Accessed September 14, 2011.

59. Zaborskis A, Petronyte G, Sumskas L, Kuzman M, Iannotti RJ. Body image and weight control among adolescents in Lithuania, Croatia, and the United States in the context of global obesity. Croat Med J. 2008;49(2):233-242. http://www.cmj.hr/2008/49/2/18461679.htm. Accessed September 14, 2011.

60. Zhang J, Seo D-C, Kolbe L, et al. Comparison of overweight, weight perception, and weight-related practices among high school students in three large Chinese cities and two large U.S. cities. J Adolesc Health. 2011;48(4):366-372. http://www.sciencedirect.com/science/article/pii/S1054139X10003538. Accessed

September 29, 2011. 109

61. Gillison F, Standage M, Skevington S. Relationships among adolescents' weight perceptions, exercise goals, exercise motivation, quality of life and leisure-time exercise behaviour: a self-determination theory approach. Health Educ Res. 2006;21(6):836-847. http://her.oxfordjournals.org/content/21/6/836.long. Accessed September 14, 2011.

62. Barlow SE, Dietz WH. Obesity evaluation and treatment: expert committee recommendations. Pediatrics. 1998;102(3):e29. http://pediatrics.aappublications.org/content/102/3/e29.long. Accessed September 16, 2011.

63. Story MT, Neumark-Stzainer DR, Sherwood NE, et al. Management of child and adolescent obesity: attitudes, barriers, skills, and training needs among health care

110

professionals. Pediatrics. 2002;110(1, pt 2):210-214. http://pediatrics.aappublications.org/content/110/Supplement_1/210.long. Accessed September 14, 2011.

64. Fonseca H, de Matos MG. Perception of overweight and obesity among Portuguese adolescents: an overview of associated factors. Eur J Public Health. 2005;15(3):323-328. http://eurpub.oxfordjournals.org/content/15/3/323.long. Accessed September 14, 2011.

65. Strauss RS. Self-reported weight status and dieting in a cross-sectional sample of young adolescents: National Health and Nutrition Examination Survey III. Arch Pediatr Adolesc Med. 1999;153(7):741-747. http://archpedi.jamanetwork.com/article.aspx?articleid=347363. Accessed September 14, 2011.

66. Powell TM, de Lamos JA, Banks K, et al. Body size misperception: a novel determinant in the obesity epidemic. Arch Intern Med. 2010;170(18):1695-1697. http://archinte.jamanetwork.com/article.aspx?articleid=226035. Accessed September 14, 2011.

67. Janssen I, Katzmarzyk PT, Boyce WF, King MA, Pickett W. Overweight and obesity in Canadian adolescents and their associations with dietary habits and physical activity patterns. J Adolesc Health. 2004;35(5):360-367. http://www.sciencedirect.com/science/article/pii/S1054139X04000588. Accessed April 25, 2012.

68. Ball GDC, McCargar LJ. Childhood obesity in Canada: a review of prevalence estimates and risk factors for cardiovascular diseases and type 2 diabetes. Can J Appl Physiol. 2003;28(1):117-140.

69. Bélanger-Ducharme F, Tremblay A. Prevalence of obesity in Canada. Obes Rev. 2005;6(3):183-186. http://onlinelibrary.wiley.com/doi/10.1111/j.1467- 789X.2005.00179.x/abstract. Accessed September 14, 2011.

110 70. Shields M, Tjepkema M. Regional differences in obesity. Health Rep. 2006;17(3):61-67. http://www.statcan.gc.ca/pub/82-003-x/2005003/article/9280- eng.pdf. Accessed February 20, 2012.

71. Willms JD, Tremblay MS, Katzmarzyk PT. Geographic and demographic variation in the prevalence of overweight Canadian children. Obes Res. 2003;11(5):668-673.

72. Anderson PM, Butcher KF. Childhood obesity: trends and potential causes. Future Child. 2006;16(1):19-45.

73. World Health Organization. Obesity: preventing and managing the global epidemic: report of a WHO consultation on obesity. Geneva: World Health Organization;1998.

111

http://whqlibdoc.who.int/hq/1998/WHO_NUT_NCD_98.1_(p1-158).pdf. Accessed April 2, 2012.

74. Haines J, Neumark-Sztainer D, Wall M, Story M. Personal, behavioural, and environmental risk and protective factors for adolescent overweight. Obesity (Sliver Spring). 2007;15(11):2748 - 2760. http://www.nature.com/oby/journal/v15/n11/full/oby2007327a.html. Accessed July 7, 2012.

75. Philippas NG, Lo CW. Childhood obesity: etiology, prevention, and treatment. Nutr Clin Care. 2005;8(2):77-88.

76. Veugelers PJ, Fitzgerald AL. Prevalence of and risk factors for childhood overweight and obesity. CMAJ. 2005;173(6):607-613. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1197160/?tool=pubmed. Accessed February 8, 2012.

77. McLaren L. Socioeconomic status and obesity. Epidemiol Rev. 2007;29:29 - 48.

78. Singh GK, Kogan MD, Van Dyck PC, Siahpush M. Racial/ethnic, socioeconomic, and behavioral determinants of childhood and adolescent obesity in the United States: analyzing independent and joint associations. Ann Epidemiol. 2008;18(9):682-695. http://www.sciencedirect.com/science/article/pii/S1047279708001129. Accessed December 30, 2011.

79. Morgen C, Mortensen L, Rasmussen M, Andersen A-M, Sorensen T, Due P. Parental socioeconomic position and development of overweight in adolescence: longitudinal study of Danish adolescents. BMC Public Health. 2010;10:520. http://www.biomedcentral.com/1471-2458/10/520. Accessed February 15, 2012.

80. Revah-Levy A, Speranza M, Barry C, et al. Association between body mass index and depression: the "fat and jolly" hypothesis for adolescent girls. BMC Public

Health. 2011;11:649. http://www.biomedcentral.com/1471-2458/11/649. 111 Accessed February 28, 2012.

81. Shankaran S, Bann C, Das A, et al. Risk for obesity in adolescence starts in early childhood. J Perinatol. 2011;31(11):711-716. http://www.nature.com/jp/journal/v31/n11/full/jp201114a.html. Accessed February 11, 2012.

82. Reilly JJ, Armstrong J, Dorosty AR, et al. Early life risk factors for obesity in childhood: cohort study. BMJ. 2005;330(7504):1357. http://www.bmj.com/content/330/7504/1357?view=long&pmid=15908441. Accessed February 7, 2012.

83. Baird J, Fisher D, Lucas P, Kleijnen J, Roberts H, Law C. Being big or growing fast: systematic review of size and growth in infancy and later obesity. BMJ.

112

2005;331(7522):929. http://www.bmj.com/content/331/7522/929?view=long&pmid=16227306. Accessed January 29, 2012.

84. Mehta SH, Kruger M, Sokol RJ. Being too large for gestational age precedes childhood obesity in African Americans. Am J Obstet Gynecol. 2011;204(3):265.e1-e5. http://www.sciencedirect.com/science/article/pii/S0002937810024257. Accessed December 30, 2011.

85. Bergmann KE, Bergmann RL, von Kries R, et al. Early determinants of childhood overweight and adiposity in a birth cohort study: role of breast-feeding. Int J Obes Relat Metab Disord. 2003;27(2):162-172.

86. Salsberry PJ, Reagan PB. Dynamics of early childhood overweight. Pediatrics. 2005;116(6):1329-1338. http://pediatrics.aappublications.org/content/116/6/1329.long. Accessed October 19, 2011.

87. Arenz S, Ruckerl R, Koletzko B, von Kries R. Breast-feeding and childhood obesity: a systematic review. Int J Obes Relat Metab Disord. 2004;28(10):1247- 1256.

88. Haines J, Neumark-Sztainer D. Prevention of obesity and eating disorders: a consideration of shared risk factors. Health Educ Res. 2006;21(6):770 - 782. http://her.oxfordjournals.org/cgi/pmidlookup?view=long&pmid=16963727. Accessed January 6, 2012.

89. Pagani LS, Huot C. Why are children living in poverty getting fatter? Paediatr Child Health. 2007;12(8):698-700.

90. Parsons TJ, Power C, Manor O. Physical activity, television viewing and body mass index: a cross-sectional analysis from childhood to adulthood in the 1958

British cohort. Int J Obes Relat Metab Disord. 2005;29(10):1212-1221. 112

91. Landhuis CE, Poulton R, Welch D, Hancox RJ. Programming obesity and poor fitness: the long-term impact of childhood television. Obesity (Silver Spring). 2008;16(6):1457-1459. http://www.nature.com/oby/journal/v16/n6/full/oby2008205a.html. Accessed December 30, 2011.

92. Wells JCK, Hallal PC, Reichert FF, Menezes AMB, Araujo CLP, Victora CG. Sleep patterns and television viewing in relation to obesity and blood pressure: evidence from an adolescent Brazilian birth cohort. Int J Obes (Lond). 2008;32(7):1042-1049. http://www.nature.com/ijo/journal/v32/n7/full/ijo200837a.html. Accessed February 16, 2012.

113

93. Adams AK, Harvey HE, Prince RJ. Association of maternal smoking with overweight at age 3 y in American Indian children. Am J Clin Nutr. 2005;82(2):393-398. http://www.ajcn.org/content/82/2/393.long. Accessed October 10, 2011.

94. Fasting MH, Nilsen TIL, Holmen TL, Vik T. Changes in parental weight and smoking habits and offspring adiposity: Data from the HUNT-study. Int J Pediatr Obes. 2011;6(2, pt. 2):e399-e407. http://informahealthcare.com/doi/abs/10.3109/17477166.2010.518238. Accessed February 14, 2012.

95. Cecil-Karb R, Grogan-Kaylor A. Childhood Body Mass Index in Community Context: Neighborhood Safety, Television Viewing, and Growth Trajectories of BMI. Health Soc Work. 2009;34(3):169-177. http://web.ebscohost.com/ehost/detail?sid=f493cfb5-2cec-4974-8eea- 12bc0a687834%40sessionmgr114&vid=1&hid=108&bdata=JnNpdGU9ZWhvc3 QtbGl2ZQ%3d%3d#db=a9h&AN=43781332. Accessed December 30, 2011

96. Nicholson L, Browning C. Racial and ethnic disparities in obesity during the transition to adulthood: the contingent and nonlinear impact of neighborhood disadvantage. J Youth Adolesc. 2011;41(1):1-14. http://www.springerlink.com/content/f927524150m54737/?MUD=MP. Accessed February 12, 2012.

97. Oliver LN, Hayes MV. Neighbourhood socio-economic status and the prevalence of overweight Canadian children and youth. Can J Public Health. 2005;96(6):415-420.

98. Fonseca H, Matos Mg, Guerra A, Gomes-Pedros J. How much does overweight impact the adolescent development process? Child Care Health Dev. 2010;37(1):135-142. http://onlinelibrary.wiley.com/doi/10.1111/j.1365- 2214.2010.01136.x/abstract. Accessed September 27, 2011.

99. Estabrooks PA, Shetterly S. The prevalence and health care use of overweight 113 children in an integrated health care system. Arch Pediatr Adolesc Med. 2007;161(3):222-227. http://archpedi.jamanetwork.com/article.aspx?doi=10.1001/archpedi.161.3.222. Accessed September 26, 2011.

100. Hanley AJ, Harris SB, Gittelsohn J, Wolever TM, Saksvig B, Zinman B. Overweight among children and adolescents in a Native Canadian community: prevalence and associated factors. Am J Clin Nutr. 2000;71(3):693-700. http://www.ajcn.org/content/71/3/693.long. Accessed September 14, 2011.

101. Willows ND. Overweight in First Nations children: prevalence, implications, and solutions. J Aboriginal Health. 2005;2(1):76-86. http://www.naho.ca/jah/english/jah02_01/JournalVol2No1ENG9overweightchildr en.pdf. Accessed December 15, 2011.

114

102. Willows ND, Johnson MS, Ball GDC. Prevalence estimates of overweight and obesity in Cree preschool children in northern Quebec according to international and US reference criteria. Am J Public Health. 2007;97(2):311-316. http://ajph.aphapublications.org/doi/abs/10.2105/AJPH.2005.073940. Accessed December 15, 2011.

103. Singh G, Kogan M, Yu S. Disparities in obesity and overweight prevalence among US immigrant children and adolescents by generational status. J Community Health. 2009;34(4):271-281. http://www.springerlink.com/content/v734047648873288/. Accessed January 11, 2012.

104. Taylor SJC, Viner R, Booy R, et al. Ethnicity, socio-economic status, overweight and underweight in East London adolescents. Ethn Health. 2005;10(2):113-128. http://www.tandfonline.com/doi/abs/10.1080/13557850500071095?url_ver=Z39. 88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_pub%3dpubmed. Accessed February 15, 2012.

105. Singh A, Chinapaw M, Brug J, Kremers S, Visscher T, van Mechelen W. Ethnic differences in BMI among Dutch adolescents: what is the role of screen-viewing, active commuting to school, and consumption of soft drinks and high-caloric snacks? Int J Behav Nutr Phys Act. 2009;6(1):23. http://www.ijbnpa.org/content/6/1/23. Accessed December 30, 2011.

106. Tremblay M, LeBlanc A, Kho M, et al. Systematic review of sedentary behaviour and health indicators in school-aged children and youth. Int J Behav Nutr Phys Act. 2011;8(1):98. http://www.ijbnpa.org/content/8/1/98. Accessed June 22, 2012.

107. Schwimmer JB, Burwinkle TM, Varni JW. Health-related quality of life of severely obese children and adolescents. JAMA. 2003;289(14):1813-1819. http://jama.jamanetwork.com/article.aspx?articleid=196343. Accessed September 14, 2011.

108. Williams J, Wake M, Hesketh K, Maher E, Waters E. Health-related quality of 114 life of overweight and obese children. JAMA. 2005;293(1):70-76. http://jama.jamanetwork.com/article.aspx?doi=10.1001/jama.293.1.70. Accessed September 14, 2011.

109. Loth KA, Mond J, Wall M, Neumark-Sztainer. Weight status and emotional well- being: longitudinal findings from project EAT. J Pediatr Psychol. 2011;36(2):216-225.

110. Stunkard A, Burt V. Obesity and the body image: II. age at onset of disturbances in the body image. Am J Psychiatry. 1967;123(11):1443-1447.

111. Li L, de Moira AP, Power C. Predicting cardiovascular disease risk factors in midadulthood from childhood body mass index: utility of different cutoffs for childhood body mass index. Am J Clin Nutr. 2011;93:1204-1211.

115

http://www.ajcn.org/cgi/pmidlookup?view=long&pmid=21430113. Accessed September 15, 2011.

112. Vanhala M, Vanhala P, Kampusalo E, Halonen P, Takala J. Relation between obesity from childhood to adulthood and the metabolic syndrome: population based study. BMJ. 1998;317(7154):319-320. http://www.bmj.com/content/317/7154/319?view=long&pmid=9685277. Accessed November 30, 2011.

113. Raitakari OT, Juonala M, Viikari JSA. Obesity in childhood and vascular changes in adulthood: insights into the Cardiovascular Risk in Young Finns Study. Int J Obes (Lond). 2005;29(suppl 2):S101-S104. http://www.nature.com/ijo/journal/v29/n2s/full/0803085a.html. Accessed September 14, 2011.

114. Must A, Jacques PF, Dallal GE, Bajema CJ, Dietz WH. Long-term morbidity and mortality of overweight adolescents. N Engl J Med. 1992;327(19):1350-1355. http://www.nejm.org/doi/full/10.1056/NEJM199211053271904. Accessed September 14, 2011.

115. Tirosh A, Shai I, Afek A, et al. Adolescent BMI trajectory and risk of diabetes versus coronary disease. N Eng J Med. 2011;364(14):1315-1325. http://www.nejm.org/doi/full/10.1056/NEJMoa1006992. Accessed February 20, 2012.

116. Plourde G. Preventing and managing pediatric obesity. Recommendations for family physicians. Can Fam Physician. 2006;52(3):322-328. http://www.cfp.ca/content/52/3/322.long. Accessed October 10, 2011.

117. Dietz WH. Health consequences of obesity in youth: childhood predictors of adult disease. Pediatrics. 1998;101(suppl 1):518-525. http://pediatrics.aappublications.org/cgi/pmidlookup?view=long&pmid=1222465 8. Accessed September 14, 2011.

115 118. Guo SS, Wu W, Chumlea WC, Roche AF. Predicting overweight and obesity in adulthood from body mass index values in childhood and adolescence. Am J Clin Nutr. 2002;76:653-658. http://www.ajcn.org/content/76/3/653.long. Accessed September 16, 2011.

119. Patton GC, Coffey C, Carlin JB, et al. Overweight and obesity between adolescence and young adulthood: a 10-year prospective cohort study. J Adolesc Health. 2011;48:275-280. http://www.sciencedirect.com/science/article/pii/S1054139X10003113. Accessed September 27, 2011.

120. Woo JG. Using body mass index z-score among severely obese adolescents: a cautionary note. Int J Pediatr Obes. 2009;4:405-410.

116

http://informahealthcare.com/doi/abs/10.3109/17477160902957133. Accessed September 26, 2011.

121. I'Allemand D, Wiegand S, Reinehr T, et al. Cardiovascular risk in 26,008 European overweight children as established by a multicenter database. Obesity (Silver Spring). 2008;16(7):1672-1679. http://www.nature.com/oby/journal/v16/n7/full/oby2008259a.html. Accessed September 14, 2011.

122. Weiss R, Dziura J, Burgert TS, et al. Obesity and the metabolic syndrome in children and adolescents. N Engl J Med. 2004;350(23):2362-2374. http://www.nejm.org/doi/full/10.1056/NEJMoa031049. Accessed February 29, 2012.

123. Stewart ST, Cutler DM, Rosen AB. Forecasting the effects of obesity and smoking on U.S. life expectancy. N Engl J Med. 2009;361(23):2252-2260. http://www.nejm.org/doi/full/10.1056/NEJMsa0900459. Accessed February 14, 2012.

124. Swallen KC, Reither EN, Haas SA, Meier AM. Overweight, obesity, and health- related quality of life among adolescents: the National Longitudinal Study of Adolescent Health. Pediatrics. 2005;115(2):340-347. http://pediatrics.aappublications.org/cgi/pmidlookup?view=long&pmid=1568744 2. Accessed September 16, 2011.

125. Katzmarzyk PT, Tremblay A, Pérusse L, Després J-P, Bouchard C. The utility of the international child and adolescent overweight guidelines for predicting coronary heart disease risk factors. J Clin Epidemiol. 2003;56(5):456-462. http://www.sciencedirect.com/science/article/pii/S0895435602005954. Accessed September 14, 2011.

126. l'Allemand-Jander D. Clinical diagnosis of metabolic and cardiovascular risks in overweight children: early development of chronic diseases in the obese child. Int

J Obes (Lond). 2010;34(suppl 2):S32-S36. 116 http://www.nature.com/ijo/journal/v34/n2s/full/ijo2010237a.html. Accessed September 16, 2011.

127. Freedman DS, Dietz WH, Srinivasan SR, Berenson GS. The relation of overweight to cardiovascular risk factors among children and adolescents: the Bogalusa Heart Study. Pediatrics. 1999;103(6):1175-1182. http://pediatrics.aappublications.org/cgi/pmidlookup?view=long&pmid=1035392 5. Accessed February 18, 2012.

128. Bell LM, Byrne S, Thompson A, et al. Increasing body mass index z-score is continuously associated with complications of overweight in children, even in the healthy weight range. J Clin Endocrinol Metab. 2007;92(2):517-522. http://jcem.endojournals.org/cgi/pmidlookup?view=long&pmid=17105842. Accessed September 14, 2011.

117

129. Polderman J, Gurgel RQ, Barreto-Fillho JA, et al. Blood pressure and BMI in adolescents in Aracaju, Brazil. Public Health Nutr. 2011;14(6):1064-1070. http://journals.cambridge.org/action/displayAbstract?fromPage=online&aid=8269 931. Accessed September 27, 2011.

130. Sant'Anna TC, Vieira Braga CM, Moreira GM, Sousa L. Overweight, physical activity and atherosclerotic disease risk in Brazilian adolescents. Int J Cardiol. 2011;146(2):236-237. http://www.sciencedirect.com/science/article/pii/S0167527310008223. Accessed September 27, 2011.

131. Savino A, Pelliccia P, Giannini C, et al. Implications for kidney disease in obese children and adolescents. Pediatr Nephrol. 2011;26:749-758. http://www.springerlink.com/content/m1764811j6616465/. Accessed February 21, 2012.

132. Amed S, Dean HJ, Panagiotopoulos C, et al. Type 2 diabetes, medication-induced diabetes, and monogenic diabetes in Canadian children. Diabetes Care. 2010;33(4):786-791. http://care.diabetesjournals.org/content/33/4/786.long. Accessed February 20, 2012.

133. Pollock NK, Bernard PJ, Gutin B, Davis CL, Zhu H, Dong Y. Adolescent obesity, bone mass, and cardiometabolic risk factors. J Pediatr. 2011;158:727-734. http://www.sciencedirect.com/science/article/pii/S0022347610010462. Accessed September 16, 2011.

134. Imhof A, Kratzer W, Boehm B, et al. Prevalence of non-alcoholic fatty liver and characteristics in overweight adolescents in the general population. Eur J Epidemiol. 2007;22(12):889-897. http://www.springerlink.com/content/67413n4x231tq273/. Accessed March 27, 2012.

135. Redline S, Tishler PV, Schluchter M, Aylor J, Clark K, Graham G. Risk factors

for sleep-disordered breathing in children: associations with obesity, race, and 117 respiratory problems. Am J Respir Crit Care Med. 1999;159(5, pt 1):1527-1532.

136. Doan Q, Koehoorn M, Kissoon N. Body mass index and the risk of acute injuries in adolescents. Paediatr Child Health. 2010;15(6):351-356.

137. Bazelmans C, Coppieters Y, Godin I, et al. Is obesity associated with injuries among young people? Eur J Epidemiol. 2004;19(11):1037-1042. http://www.springerlink.com/content/l5416426g626344r/. Accessed Ocrtober 11, 2011.

138. Pomerantz WJ, Timm NL, Gittelman MA. Injury patterns in obese versus nonobese children presenting to a pediatric emergency department. Pediatrics. 2010;125(4):681-685.

118

http://pediatrics.aappublications.org/content/125/4/681.long. Accessed February 21, 2012.

139. Puder JJ, Munsch S. Psychological correlates of childhood obesity. Int J Obes (Lond). 2010;34 (suppl 2):S37-S43. http://www.nature.com/ijo/journal/v34/n2s/full/ijo2010238a.html. Accessed October 11, 2011.

140. Goodman E, Whitaker Rc. A prospective study of the role of depression in the development and persistence of adolescent obesity. Pediatrics. 2002;110(3):497. http://pediatrics.aappublications.org/content/110/3/497.long. Accessed September 16, 2011.

141. Warschburger P. The unhappy obese child. Int J Obes (Lond). 2005;29(suppl 2):127-129.

142. Needham BL, Crosnoe R. Overweight status and depressive symptoms during adolescence. J Adolesc Health. 2005;36(1):48-55. http://www.sciencedirect.com/science/article/pii/S1054139X04002630. Accessed December 14, 2011.

143. Erickson SJ, Robinson TN, Haydel KF, Killen JD. Are overweight children unhappy? Body mass index, depressive symptoms, and overweight concerns in elementary school children. Arch Pediatr Adolesc Med. 2000;154(9):931-935. http://archpedi.jamanetwork.com/article.aspx?articleid=351173. Accessed September 16, 2011.

144. Wardle J, Cooke L. The impact of obesity on psychological well-being. Best Pract Res Clin Endocrinol Metab. 2005;19(3):421-440. http://www.sciencedirect.com/science/article/pii/S1521690X05000382. Accessed September 16, 2011.

145. Newman D, Sontag L, Salvato R. Psychosocial aspects of body mass and body

image among rural American Indian adolescents. J Youth Adolesc. 118 2006;35(2):281-291. http://www.springerlink.com/content/54675715335rp410/. Accessed February 28, 2012.

146. Dave D, Rashad I. Overweight status, self-perception, and suicidal behaviors among adolescents. Soc Sci Med. 2009;68(9):1685-1691. http://www.sciencedirect.com/science/article/pii/S0277953609001038. Accessed September 27, 2011.

147. Eaton DK, Lowry R, Brener ND, Galuska DA, Crosby AE. Associations of body mass index and perceived weight with suicide ideation and suicide attempts among US high school students. Arch Pediatr Adolesc Med. 2005;159(6):513- 519. http://archpedi.jamanetwork.com/article.aspx?articleid=486040. Accessed September 27, 2011.

119

148. Jansen W, van de Looij-Jansen PM, de Wilde EJ, Brug J. Feeling fat rather than being fat may be associated with psychological well-being in young Dutch adolescents. J Adolesc Health. 2008;42(2):128-136. http://www.sciencedirect.com/science/article/pii/S1054139X07003333. Accessed March 28, 2012.

149. Caria MP, Bellocco R, Zambon A, Horton NJ, Galanti MR. Overweight and perception of overweight as predictors of smokeless tobacco use and of cigarette smoking in a cohort of Swedish adolescents. Addiction. 2009;104(4):661-668. http://onlinelibrary.wiley.com/doi/10.1111/j.1360-0443.2009.02506.x/abstract. Accessed September 27, 2011.

150. Brixval CS, Rayce SLB, Rasmussen M, Holstein BE, Due P. Overweight, body image and bullying—an epidemiological study of 11- to 15-years olds. Eur J Public Health. 2012;22(1):126-130. http://eurpub.oxfordjournals.org/content/22/1/126.long. Accessed June 5, 2012.

481. Cornette R. The emotional impact of obesity on children. Worldv Evid-Based Nu. 2008;5(3):136-141. http://onlinelibrary.wiley.com/doi/10.1111/j.1741- 6787.2008.00127.x/abstract. Accessed February 27, 2012.

152. Gibson LY, Byrne SM, Blair E, Davis EA, Jacoby P, Zubrick SR. Clustering of psychosocial symptoms in overweight children. Aust N Z J Psychiatry. 2008;42:118-125. http://anp.sagepub.com/content/42/2/118.long. Accessed September 27, 2011.

153. Sweeting H, Wright C, Minnis H. Psychosocial correlates of adolescent obesity, ‘slimming down’ and ‘becoming obese’. J Adolesc Health. 2005;37(5):409. http://www.sciencedirect.com/science/article/pii/S1054139X05000583. Accessed September 29, 2011.

154. Kostanski M, Gullone E. Adolescent body image dissatisfaction: relationships with self-esteem, anxiety, and depression controlling for body mass. J Child

Psychol Psychiatr. 1998;39(2):255-262. 119

155. Flodmark CE. The happy obese child. Int J Obes (Lond). 2005;29(suppl 2):31-33. http://www.nature.com/ijo/journal/v29/n2s/full/0803060a.html. Accessed October 11, 2011.

156. Janicke DM, Harman JS, Kelleher KJ, Zhang J. Psychiatric diagnosis in children and adolescents with obesity-related health conditions. J Dev Behav Pediatr. 2008;29:276-284. http://journals.lww.com/jrnldbp/pages/articleviewer.aspx?year=2008&issue=0800 0&article=00005&type=abstract. Accessed February 27, 2012.

157. Hesketh K, Wake M, Waters E. Body mass index and parent-reported self-esteem in elementary school children: evidence for a causal relationship. Int J Obes Relat Metab Disord. 2004;28(10):1233-1237.

120

158. Franklin J, Denyer G, Steinbeck KS, Caterson ID, Hill AJ. Obesity and risk of low self-esteem: a statewide survey of Australian children. Pediatrics. 2006;118(6):2481-2487. http://pediatrics.aappublications.org/content/118/6/2481.long. Accessed September 16, 2011.

159. Pesa JA, Syre TR, Jones E. Psychosocial differences associated with body weight among female adolescents: the importance of body image. J Adolesc Health. 2000;26(5):330-337. http://www.sciencedirect.com/science/article/pii/S1054139X99001184. Accessed September 29, 2011.

160. Strauss RS. Childhood obesity and self-esteem. Pediatrics. 2000;105(1):e15. http://pediatrics.aappublications.org/content/105/1/e15.long. Accessed September 16, 2011.

161. Wang F, Veugelers PJ. Self-esteem and cognitive development in the era of the childhood obesity epidemic. Obes Res. 2008;9(6):615-623. http://onlinelibrary.wiley.com/doi/10.1111/j.1467-789X.2008.00507.x/abstract. Accessed September 16, 2011.

162. Ali MM, Fang H, Rizzo JA. Body weight, self-perception and mental health outcomes among adolescents. J Ment Health Policy Econ. 2010;13(2):53-63.

163. Frisco ML, Houle JN, Martin MA. Adolescent weight and depressive symptoms: for whom is weight a burden? Soc Sci Q. 2009;90(4):1019-1038.

164. Kubzansky L, Gilthorpe M, Goodman E. A prospective study of psychological distress and weight status in adolescents/young adults. Ann Behav Med. 2012;43(2):219-228. http://www.springerlink.com/content/j3jpw263g633181q/. Accessed March 28, 2012.

165. Griffiths LJ, Dezateux C, Hill A. Is obesity associated with emotional and

behavioural problems in children? Findings from the Millennium Cohort Study. 120 Int J Pediatr Obes. 2011;6(2, pt 2):e423-e432. http://informahealthcare.com/doi/abs/10.3109/17477166.2010.526221. Accessed September 16, 2011.

166. Erhart M, Herpertz-Dahlmann B, Wille N, Sawitzky-Rose B, Hölling H, Ravens- Sieberer U. Examining the relationship between Attention-Deficit/Hyperactivity Disorder and overweight in children and adolescents. Eur Child Adolesc Psychiatry. 2012;21(1):39-49. http://www.springerlink.com/content/16vv7g135qh85211/. Accessed February 17, 2012.

167. Denoth F, Siciliano V, Iozzo P, Fortunato L, Molinaro S. The association between overweight and illegal drug consumption in adolescents: is there an underlying influence of the sociocultural environment? PLoS One. 2011;6(11):e27358.

121

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.002735 8. Accessed December 6, 2011.

168. Strauss RS, Pollack HA. Social marginalization of overweight children. Arch Pediatr Adolesc Med. 2003;157(8):746-752. http://archpedi.jamanetwork.com/article.aspx?articleid=481398. Accessed December 14, 2011.

169. Trogdon JG, Nonnemaker J, Pais J. Peer effects in adolescent overweight. J Health Econ. 2008;27(5):1388-1399. http://www.sciencedirect.com/science/article/pii/S0167629608000477. Accessed November 29, 2011.

170. Muennig P. The body politic: the relationship between stigma and obesity- associated disease. BMC Public Health. 2008;8(8):128. http://www.biomedcentral.com/1471-2458/8/128. Accessed September 16, 2011.

171. Tang-Péronard JL, Heitmann BL. Stigmatization of obese children and adolescents, the importance of gender. Obes Rev. 2008;9(6):522-534. http://onlinelibrary.wiley.com/doi/10.1111/j.1467-789X.2008.00509.x/abstract. Accessed September 29, 2011.

172. Puhl RM, Latner JD. Stigma, obesity, and the health of the nation's children. Psychol Bull. 2007;133(4):557-580.

173. Keating CL, Moodie ML, Richardson J, Swinburn BA. Utility-based quality of life of overweight and obese adolescents. Value Health. 2011;14(5):752-758.

174. Ostbye T, Malhotra R, Wong H-B, Tan S-B, Saw S-M. The effect of body mass on health-related quality of life among Singaporean adolescents: results from the SCORM study. Qual Life Res. 2010;19(2):167-176. http://www.springerlink.com/content/16714250m8j86703/. Accessed September 15, 2011.

121 175. Pinhas-Hamiel O, Singer S, Pilpel N, Fradkin A, Modan D, Reichman B. Health- related quality of life among children and adolescents: associations with obesity. Int J Obes Relat Metab Disord. 2005;30(2):267-272. http://www.nature.com/ijo/journal/v30/n2/full/0803107a.html. Accessed September 14, 2011.

176. Crosnoe R, Muller C. Body mass index, academic achievement, and school context: examining the educational experiences of adolescents at risk of obesity. J Health Soc Behav. 2004;45(4):393-407.

177. Taras H, Potts-Datema W. Obesity and student performance at school. J Sch Health. 2005;75(8):291-295. http://onlinelibrary.wiley.com/doi/10.1111/j.1746- 1561.2005.00040.x/abstract. Accessed March 27, 2012.

122

178. Trasande L, Chatterjee S. The impact of obesity on health service utilization and costs in childhood. Obesity (Silver Spring). 2009;17(9):1749-1754. http://www.nature.com/oby/journal/v17/n9/full/oby200967a.html. Accessed February 20, 2012.

179. Hering E, Pritsker I, Gonchar L, Pillar G. Obesity in children is associated with increased health care use. Clin Pediatr. 2009;48(8):812-818. http://cpj.sagepub.com/content/48/8/812.long. Accessed February 20, 2012.

180. Hampl SE, Carroll CA, Simon SD, Sharma V. Resource utilization and expenditures for overweight and obese children. Arch Pediatr Adolesc Med. 2007;161(1):11-14. http://archpedi.jamanetwork.com/article.aspx?articleid=569363. Accessed February 20, 2012.

181. Kuhle S, Kirk S, Ohinmaa A, Yasui Y, Allen AC, Veugelers PJ. Use and cost of health services among overweight and obese Canadian children. Int J Pediatr Obes. 2011;6(2):142-148. http://informahealthcare.com/doi/abs/10.3109/17477166.2010.486834. Accessed July 4, 2012.

182. Janssen I, Lam M, Katzmarzyk PT. Influence of overweight and obesity on physician costs in adolescents and adults in Ontario, Canada. Obes Rev. 2009;10(1):51-57. http://onlinelibrary.wiley.com/doi/10.1111/j.1467- 789X.2008.00514.x/abstract. Accessed June 19, 2012.

183. Tompkins CL, Moran K, Preedom S, Brock DW. Physical activity-induced improvements in markers of insulin resistance in overweight and obese children and adolescents. Curr Diabetes Rev. 2011;7(3):164-170. http://www.benthamdirect.org/pages/content.php?CDR/2011/00000007/00000003 /D0003D.SGM. Accessed June 24, 2012.

184. Reinehr T. Symposium III: Metabolic health, weight management and obesity

prevention in childhood and adolescence Effectiveness of lifestyle intervention in 122 overweight children. Paper presented at: 70th Anniversary Conference on 'Nutrition and health: from conception to adolescence'; August 1, 2011; Glasgow, Scotland.

185. Janz NK, Becker MH. The Health Belief Model: a decade later. Health Educ Q. 1984;11(1):1-47.

186. Rosenstock IM, Strecher VJ, Becker MH. Social learning theory and the Health Belief Model. Health Educ Q. 1988;15(2):175-183.

187. Weinstein ND, Rothman AJ, Sutton SR. Stage theories of health behavior: Conceptual and methodological issues. Health Psychol. 1998;17(3):290-299.

123

188. Kant AK. Association of self-perceived body weight status with dietary reporting by U.S. teens. Obes Res. 2002;10(12):1259-1269.

189. Cook SJ, MacPherson K, Langille DB. Far from ideal: weight perception, weight control, and associated risky behaviour of adolescent girls in Nova Scotia. Can Fam Physician. 2007;53(4):678-684. http://www.cfp.ca/content/53/4/678.long . Accessed September 27, 2011.

190. Lo W-S, Ho S-Y, Mak K-K, Wong Y-M, Lai Y-K, Lam T-H. Prospective effects of weight perception and weight comments on psychological health among Chinese adolescents. Acta Pædiatr. 2009;98(12):1959-1964. http://onlinelibrary.wiley.com/doi/10.1111/j.1651-2227.2009.01472.x/abstract. Accessed September 27, 2011.

191. Jauregui-Lobera I, Bolanos-Rios P, Santiago-Fernandex MJ, Garrido-Casals O, Sanchez E. Perception of weight and psychological variables in a sample of Spanish adolescents. Diabetes Metab Syndr Obes. 2011;4:245-251. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3139532/?tool=pubmed. Accessed September 29, 2011.

192. Huang L, Tao F-B, Wan Y-H, et al. Self-reported weight status rather than BMI may be closely related to psychopathological symptoms among Mainland Chinese adolescents. J Trop Pediatr. 2011;57(4):307-311.

182. Seo D-C, Lee C. The effect of perceived body weight on suicidal ideation among a representative sample of US adolescents [published online ahead of print July 4, 2012]. J Behav Med. http://www.springerlink.com/content/l1727524111j5127/. Accessed July 9, 2012.

194. Harakeh Z, Engels RCME, Monshouwer K, Hanssen PF. Adolescent's weight concerns and the onset of smoking. Subst Use Misuse. 2010;45(12):1847-1860. http://informahealthcare.com/doi/abs/10.3109/10826081003682149. Accessed September 29, 2011.

123 195. Maximova K, McGrath JJ, Barnett T, O'Loughlin J, Paradis G, Lambert M. Do you see what I see? Weight status misperception and exposure to obesity among children and adolescents. Int J Obes (Lond). 2008;32(6):1008-1015. http://www.nature.com/ijo/journal/v32/n6/full/ijo200815a.html. Accessed September 14, 2011.

196. Goodman E, Hinden BR, Khandelwal S. Accuracy of teen and parental reports of obesity and body mass index. Pediatrics. 2000;106:52-58. http://pediatrics.aappublications.org/content/106/1/52.long. Accessed Setpember 14, 2011.

197. Linder J, McLaren L, Siou GL, Csizmadi I, Robson PJ. The epidemiology of weight perception: perceived versus self-reported actual weight status among Albertan adults. Can J Public Health. 2010;101(1):56-60.

124

198. Salcedo V, Gutierrez-Fisac JL, Guallar-Castillon P, Rodriguez-Artalejo F. Trends in overweight and misperceived overweight in Spain from 1987 to 2007. Int J Obes (Lond). 2010;34(12):1759-1765. http://www.nature.com/ijo/journal/v34/n12/full/ijo201096a.html. Accessed September 14, 2011.

199. Burke MA, Heiland FW, Nadler CM. From "overweight" to "about right": evidence of a generational shift in body weight norms. Obesity (Silver Spring). 2010;18(6):1226-1234. http://www.nature.com/oby/journal/v18/n6/full/oby2009369a.html. Accessed September 14, 2011.

200. Burke MA, Heiland F. Social dynamics of obesity. Econ Inq. 2007;45(3):571- 591. http://onlinelibrary.wiley.com/doi/10.1111/j.1465-7295.2007.00025.x/full. Accessed September 14, 2011.

201. Neighbors L, Sobal J, Liff C, Amiraian D. Weighing weight: trends in body weight evaluation among young adults, 1990 and 2005. Sex Roles. 2008;59:68-80. http://www.springerlink.com/content/453165738t256703/. September 14, 2011.

202. Blanchflower DG, Oswald AJ, Van Landeghem B. Imitative obesity and relative utility. Cambridge, MA. 2008. ftp://ftp.iza.org/RePEc/Discussionpaper/dp4010.pdf. Accessed September 14, 2011.

203. Leatherdale ST, Papadakis S. A multi-level examination of the association between older social models in the school environment and overweight and obesity among younger students. J Youth Adolesc. 2011;40:361-372. http://www.springerlink.com/content/m616162344508503/. Accessed September 23, 2011.

204. Renna F, Grafova IB, Thakur N. The effect of friends on adolescent body weight. Econ Hum Biol. 2008;6(3):377-387.

http://www.sciencedirect.com/science/article/pii/S1570677X08000312. Accessed 124 May 7, 2012.

205. Perkins JM, Perkins HW, Craig DW. Peer weight norm misperception as a risk factor for being over and underweight among UK secondary school students. Eur J Clin Nutr. 2010;64(9):965-971. http://www.nature.com/ejcn/journal/v64/n9/full/ejcn2010106a.html. Accessed September 14, 2011.

206. Pietrobelli A, Faith MS, Allison DB, Gallagher D, Chiumello G, Heymsfield SB. Body mass index as a measure of adiposity among children and adolescents: a validation study. J Pediatr. 1998;132(2):204-210. http://www.sciencedirect.com/science/article/pii/S0022347698704330. Accessed September 14, 2011.

125

207. Lau DC, Douketis JD, Morrison KM , Hramiak IM, Sharma AM, Ur E; and Obesity Canada clinical Practice Guidelines Expert Panel. 2006 Canadian clinical practice guidelines on the management and prevention of obesity in adults and children [summary]. CMAJ. 2007;176(8):S1-S13. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1839777/?tool=pubmed. Accessed March 4, 2012.

208. Dietz WH, Robinson TN. Use of the body mass index (BMI) as a measure of overweight in children and adolescents. J Pediatr. 1998;132(2):191-193. http://journals1.scholarsportal.info/details- sfx.xqy?uri=/00223476/v132i0002/191_uotbmiooicaa.xml. Accessed September 15, 2011.

209. Krebs NF, Himes JH, Jacobson D, Nicklas TA, Guilday P, Styne D. Assessment of child and adolescent overweight and obesity. Pediatrics. 2007;120(suppl 4):S193-S228. http://pediatrics.aappublications.org/content/120/Supplement_4/S193.long. Accessed November 18, 2011.

210. Barlow SE. Expert committee recommendations regarding the prevention, assessment, and treatment of child and adolescent overweight and obesity: summary report. Pediatrics. 2007;120(suppl 4):S164-S192. http://pediatrics.aappublications.org/content/120/Supplement_4/S164.long. Accessed September 16, 2011.

211. Colombo O, Villani S, Pinelli G, et al. To treat or not to treat: comparison of different criteria used to determine whether weight loss is to be recommended. Nutr J. 2008;7:5. http://www.nutritionj.com/content/7/1/5. Accessed September 15, 2011.

212. Flegal KM, Tabak CJ, Ogden CL. Overweight in children: definitions and interpretations. Health Educ Res. 2006;21(6):755-760. http://her.oxfordjournals.org/content/21/6/755.long. Accessed September 14,

2011. 125

213. Sweeting HN. Measurement and definitions of obesity in childhood and adolescence: a field guide for the uninitiated. Nutr J. 2007;6:32. http://www.nutritionj.com/content/6/1/32. Accessed September 15, 2011.

214. Kuczmarski RJ, Ogden CL, Guo SS, et al. 2000 CDC growth charts for the United States: methods and development. Vital Health Stat 11. 2002;(246):1-190.

215. Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ. 2000;320(7244):1240-1245. http://www.bmj.com/content/320/7244/1240?view=long&pmid=10797032. Accessed February 7, 2012.

126

216. de Onis M, Onyango AW, Borghi E, Siyam A, Nishida C, Siekmann J. Development of a WHO growth reference for school-aged children and adolescents. Bull World Health Organ. 2007;85(9):660-667. http://www.scielosp.org/scielo.php?script=sci_arttext&pid=S0042- 96862007000900010&lng=en&nrm=iso&tlng=en. Accessed September 14, 2011.

217. de Onis M, Lobstein T. Defining obesity risk status in the general childhood population: Which cut-offs should we use? Int J Pediatr Obes. 2010;5(6):458- 460. http://informahealthcare.com/doi/abs/10.3109/17477161003615583. Accessed September 15, 2011.

218. Shields M, Tremblay MS. Canadian childhood obesity estimates based on WHO, IOTF and CDC cut-points. Int J Pediatr Obes. 2010;5(3):265-273. http://informahealthcare.com/doi/abs/10.3109/17477160903268282. Accessed May 25, 2012.

219. Dietitians of Canada, Canadian Paediatric Society, College of Family Physicians of Canada, Community Health Nurses Association of Canada. The use of growth charts for assessing and monitoring growth in Canadian infants and children. Can J Diet Pract Res. 2004;65(1):22-32. http://search.proquest.com/docview/220781170?accountid=15115. Accessed September 16, 2011.

220. Dietitians of Canada, Canadian Paediatric Society, College of Family Physicians of Canada, Community Health Nurses of Canada. Promoting optimal monitoring of child growth in Canada: using the new WHO growth charts. Can J Diet Pract Res. Spring 2010;71(1):e1-e3. http://www.cps.ca/english/statements/n/growth- charts-statement-full.pdf. Accessed May 22, 2012.

221. Martin MA, May AL, Frisco ML. Equal weights but different weight perceptions among US adolescents. J Health Psychol. 2010;15(4):493-504. http://hpq.sagepub.com/content/15/4/493.long. Accessed September 14, 2011.

222. Barrett SC, Huffman FG. Comparison of self-perceived weight and desired 126 weight versus actual body mass index among adolescents in Jamaica. Rev Panam Salud Publica. 2011;29(4):267-276. http://www.scielosp.org/scielo.php?script=sci_arttext&pid=S1020- 49892011000400008&lng=en&nrm=iso&tlng=en. Accessed April 20, 2012.

223. Chang VW, Christakis NA. Extent and determinants of discrepancy between self- evaluations of weight status and clinical standards. J Gen Intern Med. 2001;16(8):538-543. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1495251/?tool=pubmed. Accessed September 14, 2011.

224. Gittelsohn J, Harris SB, Thorne-Lyman AL, Hanley AJG, Barnie A, Zinman B. Body image concepts differ by age and sex in an Ojibway-Cree community in

127

Canada. J Nutr. 1996;126(12):2990-3000. http://jn.nutrition.org/content/126/12/2990.long. Accessed September 29, 2011.

225. Jones D. Social comparison and body image: attractiveness comparisons to models and peers among adolescent girls and boys. Sex Roles. 2001;45(9-10):645- 664. http://www.springerlink.com/content/5f15160lnqcrw6hh/. Accessed April 26, 2012.

226. Williams KJ, Taylor CA, Wolf KN, Lawson RF, Crespo R. Cultural perceptions of healthy weight in rural Appalachian youth. Rural Remote Health. 2008;8(2):932. http://www.rrh.org.au/articles/subviewnew.asp?ArticleID=932. Accessed September 29, 2011.

227. Ali MM, Amialchuk A, Heiland FW. Weight-related behavior among adolescents: the role of peer effects. PLoS One. 2011;6(6):e21179. http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.002117 9. Accessed November 30, 2011.

228. Statistics Canada. Canadian Community Health Survey (CCHS) annual component: User guide 2010 and 2009-2010 microdata files. Ottawa, Canada: Statistics Canada; 2011. http://www23.statcan.gc.ca:81/imdb- bmdi/document/3226_D7_T9_V8-eng.pdf. Accessed January 10, 2012.

229. A SAS program for the CDC growth charts [computer program]. Atlanta, GA; 2011. http://www.cdc.gov/nccdphp/dnpao/growthcharts/resources/sas.htm. Accessed January 10, 2012.

230. Neumark-Sztainer D, Croll J, Story M, Hannan P, French S, Perry C. Ethnic/racial differences in weight-related concerns and behaviours among adolescent girls and boys: findings from project EAT. J Psychosom Res. 2002;53(5):963-974. http://www.sciencedirect.com/science/article/pii/S0022399902004865. Accessed January 4, 2012.

231. Clarke P, Wheaton B. Addressing data sparseness in contextual population 127 research. Sociol Method Res. 2007;35(3):311-351. http://smr.sagepub.com/content/35/3/311.abstract. Accessed June 29, 2012.

232. Hox JJ. Multilevel analysis: techniques and applications. 2nd ed. New York: Routledge; 2010.

233. Larsen K, Merlo J. Appropriate assessment of neighborhood effects on individual health: integrating random and fixed effects in multilevel logistic regression. Am J Epidemiol. 2005;161(1):81-88. http://aje.oxfordjournals.org/content/161/1/81.long. Accessed March 4, 2012.

234. Sanagou M, Wolfe R, Forbes A, Reid CM. Hospital-level associations with 30- day patient mortality after cardiac surgery: a tutorial on the application and interpretation of marginal and multilevel logistic regression. BMC Med Res

128

Methodol. 2012;12(1):28. http://www.biomedcentral.com/1471-2288/12/28. Accessed April 9, 2012.

235. Larsen K, Petersen JH, Budtz-Jørgensen E, Endahl L. Interpreting parameters in the logistic regression model with random effects. Biometrics. 2000;56(3):909- 914. http://www.jstor.org/openurl?volume=56&date=2000&spage=909&issn=000634 1X&issue=3. Accessed May 28, 2012.

236. Turrell G, Haynes M, Burton NW, et al. Neighborhood disadvantage and physical activity: baseline results from the HABITAT multilevel longitudinal study. Ann Epidemiol. 2010;20(3):171-181. http://www.sciencedirect.com/science/article/pii/S1047279709003603. Accessed May 28, 2012.

237. Johnell K, Månsson N-O, Sundquist J, Melander A, Blennow G, Merlo J. Neighborhood social participation, use of anxiolytic-hypnotic drugs, and women's propensity for disability pension: a multilevel analysis. Scand J Public Health. 2006;34(1):41-48. http://journals1.scholarsportal.info/details- sfx.xqy?uri=/14034948/v34i0001/41_nspuoafdpama.xml. Accessed May 28, 2012.

238. Ohlsson H, Lindblad U, Lithman T, et al. Understanding adherence to official guidelines on statin prescribing in primary health care—a multi-level methodological approach. Eur J Clin Pharmacol. 2005;61(9):657-665. http://www.springerlink.com/content/j583823628173113/?MUD=MP. Accessed May 28, 2012.

239. Merlo J, Yang M, Chaix B, Lynch J, Rastam L. A brief conceptual tutorial on multilevel analysis in social epidemiology: investigating contextual phenomena in different groups of people. J Epidemiol Community Health. 2005;59(9):729-736. http://jech.bmj.com/content/59/9/729.long. Accessed September 14, 2012.

240. SAS Software [computer program]. Version 9.2. Cary, NC: SAS Institute Inc; 128 2002-2008.

241. Mplus User's Guide [computer program]. Version 6th. Los Angeles, CA: Muthen & Muthen; 1998-2011.

242. Satorra A, Bentler P. A scaled difference chi-square test statistic for moment structure analysis. Psychometrika. 2001;66(4):507-514. http://rd.springer.com/article/10.1007/BF02296192. Accessed May 24, 2012.

243. St-Pierre M, Beland Y. Mode effects in the Canadian Community Health Survey: a comparison of CAPI and CATI. Paper presented at: 2004 American Statisitical Association Meeting, Survey Research Methods; , Canada. http://www23.statcan.gc.ca:81/imdb-bmdi/document/3226_D16_T9_V5-eng.htm. Accessed September 19, 2011.

129

244. Dillman DA, Smyth JD, Christian LM. Internet, mail, and mixed-mode surveys: the Tailored Design Method. 3rd ed. Hoboken, NJ: Wiley; 2009.

245. Blomquist H, Bergström E. Obesity in 4-year-old children more prevalent in girls and in municipalities with a low socioeconomic level. Acta Paediatr. 2007;96(1):113-116. http://onlinelibrary.wiley.com/doi/10.1111/j.1651- 2227.2006.00018.x/abstract. Accessed October 10, 2011.

246. Wills W, Backett-Milburn K, Gregory S, Lawton J. Young teenagers' perceptions of their own and others' bodies: a qualitative study of obese, overweight and 'normal' weight young people in Scotland. Soc Sci Med. 2006;62(2):396-406. http://www.sciencedirect.com/science/article/pii/S0277953605002960. Accessed September 14, 2011.

247. Steffen LM, Dai S, Fulton JE, Labarthe DR. Overweight in children and adolescents associated with TV viewing and parental weight: Project HeartBeat! Am J Prev Med. 2009;37(suppl 1):S50-S55. http://www.sciencedirect.com/science/article/pii/S0749379709002359. Accessed December 30, 2011.

248. Vandewater EA, Huang X. Parental weight status as a moderator of the relationship between television viewing and childhood overweight. Arch Pediatr Adolesc Med. 2006;160(4):425-431. http://archpedi.jamanetwork.com/article.aspx?articleid=204800. Accessed December 30, 2011.

249. Jebb SA, Prentice AM. Single definition of overweight and obesity should be used. BMJ. 2001;323(7319):999. http://www.bmj.com/content/323/7319/999.1?view=long&pmid=11679395. Accessed May 22, 2012.

250. Institute of Medicine Committee on Prevention of Obesity in Children and Youth. Preventing childhood obesity: health in the balance. Washington, D.C.: The

National Academic Press; 2005. http://www.ncbi.nlm.nih.gov/books/NBK83825/. 129 Accessed March 6, 2012.

251. Perrin EM, Skinner AC, Steiner MJ. Parental recall of doctor communication of weight status: national trends from 1999 through 2008. Arch Pediatr Adolesc Med. 2011:166(4):317-322. http://archpedi.jamanetwork.com/article.aspx?articleid=1148398. Accessed December 6, 2011.

252. O’Brien SH, Holubkov R, Reis EC. Identification, evaluation, and management of obesity in an academic primary care center. Pediatrics. 2004;114(2):e154-e159. http://pediatrics.aappublications.org/content/114/2/e154.long. Accessed September 14, 2011.

130

253. Dorsey KB, Wells C, Krumholz HM, Concato JC. Diagnosis, evaluation, and treatment of childhood obesity in pediatric practice. Arch Pediatr Adolesc Med. 2005;159(7):632-638. http://archpedi.jamanetwork.com/article.aspx?articleid=486069. Accessed May 22, 2012.

254. Louthan MV, Lafferty-Oza MJ, Smith ER, Hornung CA, Franco S, Theriot JA. Diagnosis and treatment frequency for overweight children and adolescents at well child visits. Clin Pediatr. 2005;44(1):57-61. http://cpj.sagepub.com/content/44/1/57.long. Accessed May 22, 2012.

255. Hamilton JL, James FW, Bazargan M. Provider practice, overweight and associated risk variables among children from a multi-ethnic underserved community. J Natl Med Assoc. 2003;95(6):441-448. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2594548/?tool=pubmed. Accessed May 22, 2012.

256. Flower KB, Perrin EM, Viadro CI, Ammerman AS. Using body mass index to identify overweight children: barriers and facilitators in primary care. Ambul Pediatr. 2007;7(1):38-44. http://www.sciencedirect.com/science/article/pii/S1530156706002127. Accessed September 19, 2011.

257. Puhl RM, Peterson JL, Luedicke J. Parental perceptions of weight terminology that providers use with youth. Pediatrics. 2011;128(4):e786-e793. http://pediatrics.aappublications.org/content/128/4/e786.long. Accessed March 6, 2012.

258. Burkhauser RV, Cawley J. Beyond BMI: the value of more accurate measures of fatness and obesity in social science research. J Health Econ. 2008;27(2):519- 529. http://www.sciencedirect.com/science/article/pii/S0167629607001130. Accessed September 29, 2011.

259. de Onis M. The use of anthropometry in the prevention of childhood overweight 130 and obesity. Int J Obes Relat Metab Disord. 2004;28(suppl 3):S81-S85. http://www.nature.com/ijo/journal/v28/n3s/full/0802810a.html. Accessed march 4, 2012.

260. Paccaud F, Wietlisbach V, Rickenbach M. Body mass index: comparing mean values and prevalence rates from telephone and examination surveys. Rev Epidemiol Sante Publique. 2000;49(1):33-40. http://www.em- consulte.com/article/106644/alertePM. Accessed September 14, 2011.

261. Stommel M, Schoenborn C. Accuracy and usefulness of BMI measures based on self-reported weight and height: findings from the NHANES & NHIS 2001-2006. BMC Public Health. 2009;9(1):421. http://www.biomedcentral.com/1471- 2458/9/421. Accessed September 14, 2011.

131

262. Shields M, Gorber SC, Tremblay MS. Estimates of obesity based on self-report versus direct measures. Health Rep. 2008;19(2):61-76. http://www.statcan.gc.ca/pub/82-003-x/2008002/article/10569-eng.pdf. Accessed September 20, 2011.

263. Sherry B, Jefferds ME, Grummer-Strawn LM. Accuracy of adolescent self-report of height and weight in assessing overweight status: a literature review. Arch Pediatr Adolesc Med. 2007;161(12):1154-1161. http://archpedi.jamanetwork.com/article.aspx?articleid=571673. Accessed September 14, 2011.

131

132

Appendices

Appendix A. Overview of prior studies examining weight status underestimation among overweight and obese adolescents using a Likert-type question to measure weight status underestimation. Note that estimates for the magnitude of weight status underestimation are for overweight adolescents unless otherwise stated. 133

Appendix B. Detailed list of variables from the Canadian Community Health Survey used in the analysis of this project 150

Appendix C. Overview of changes made to health regions to ensure their comparability across cycles of the Canadian Community Health Survey 152

Appendix D. Comparison of Minimum Size for Census Sub-divisions 160

Appendix E. Comparison of weight status underestimation across ethnic groups adjusted for age, severity of overweight, and the effect of time, as well as interview mode 162

132

Appendix A. Overview of prior studies examining weight status underestimation among overweight and obese adolescents using a Likert-type question to measure weight status underestimation. Note that estimates for the magnitude of weight status underestimation are for overweight adolescents unless otherwise stated. Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Abbott et al4 N=895 How would you describe your current Males Age: 14 to 17 body weight: too thin, about right, too Overweight 65.1 BMI: measured fat? Obese 21.4 Growth reference: IOTF Females Location: Queensland, Australia Overweight 44.1 Year: 2006 Obese 12.5 Akers et al5 N=7,193 How do you describe your weight: very Females 15 Age: 12 to 18 underweight, slightly underweight, BMI: self-reported about the right weight, slightly Growth reference: CDC overweight, very overweight? Location: United States Year: 2005 Note: The estimate of weight status underestimation included in this study includes those of normal weight reporting that they are underweight and the percentage provided is of the total sample. Al-Sendi et al6 N=447 Do you think you are thin, about right, Males Age: 12 to 17 fat, too fat? Overweight 60.0 BMI: measured Obese 13.9 th th Growth reference: NHANES1 (85 and 95 Females percentiles) Overweight 36.7 Location: Bahrain Obese 4.3 Year: 2000 133

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Al-Mamun7 N=2017 Do you think of yourself as very Males 27.7 Age: 14 underweight, slightly underweight, Females 14.8 BMI: measured about the right weight, slightly Growth reference: IOTF overweight, very overweight? Location: Brisbane, Australia Year: 1999 Alwan et al8 N=873 How do you describe your weight: very Males 46 Age: 11 to 17 underweight, slightly underweight, Females 26 BMI: measured about the right weight, slightly Growth reference: IOTF overweight, or very overweight? Location: Seychelles Year: 2007 Blond et al9 N=5655 Respondents reported whether they Males Age: 16 considered their bodies to be: Overweight 32.0 BMI: self-reported underweight, average/about right, or Obese 43.75 Growth reference: CDC overweight. (Eating Symptoms Location: New Jersey Inventory) Year: 2001 to 2004 Note: The sample for this study included only boys with a low birthweight Bodenlos et al10 N=344 Respondents reported which BMI Males 55 Age: 16 to 25 category they thought they belonged Females 26 BMI: measured to: underweight, normal weight, Growth reference: age-appropriate overweight, or unsure. Location: Massachusetts Year: not provided

134

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Brener et al11 N=2032 How do you describe your weight: very Measured Age: 13 to 18 underweight, slightly underweight, Males BMI: measured and self-reported about the right weight, slightly Overweight 85.7 Growth reference: CDC overweight, or very overweight? Obese 54.7 Location: United States Females Year: 2000 Overweight 69.1 Obese 38.2 Self-reported Males Overweight 70.2 Obese 46.7 Females

Overweight 46.2 Obese 32.9 Brug et al12 N=1694 Respondents reported about their 1.5 Age: 13 to 19 weight status using a 5-point Likert- BMI: self-reported type scale with response options Growth reference: IOTF ranging from (1) much too light to (5) Location: Netherlands much too heavy. Year: not provided Chaimovitz et al13 N=53 Respondents were asked to describe Males 44.0 Age: 12 to 18 their weight as: underweight, slightly Females 35.0 BMI: measured underweight, average, slightly Growth reference: Roberts SP, Dallal GE. overweight, or overweight. Nutr Rev 2001;59:31-36. Location: Hamilton, Ontario Year: 2005

135

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Chen et al14 N=217 (all overweight) How do you perceive your body weight 15.7 Age: 13 to 17 size: underweight, average, overweight, BMI: measured obesity, or unknown? Growth reference: Taiwanese Location: Toayuan County, Taiwan Year: 2005 to 2006 Cheung et al15 N=1066 Respondents were asked to describe Females Age: 12 to 1 8 their weight as: severely underweight, 97th percentile 41.90 BMI: self-reported mildly underweight, normal, mildly Growth reference: overweight defined as a overweight, or severely overweight. BMI  90th percentile Location: Hong Kong Year: 2003 to 2004 Daniels16 N=17,721 How do you describe your weight: very Males Age: 16 to 18 underweight, slightly underweight, Overweight 60 BMI: self-reported about the right weight, slightly Obese 23 Growth reference: CDC overweight, very overweight? Females Location: United States Overweight 26 Year: 1999, 2001 Obese 11 Duncan et al17 N=954 What do you currently think about your 43.4 Age: 11 to 15 weight: underweight, normal weight, BMI: measured overweight? Growth reference: study-based BMI  85th percentile Location: Auckland, New Zealand Year: not provided

136

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Edwards et al18 N=72,122 How do you describe your weight: very Males 22.6 Age: grades 9 to 12 overweight, slightly overweight, about Females 40.2 BMI: self-reported the right weight, slightly underweight, Growth reference: CDC very underweight? Location: United States Year: 1999 to 2007 Eichen et al19 N=11,103 How do you describe your weight: very Overweight 33.4 Age: 12 to 18 underweight, slightly underweight, BMI: self-reported about the right weight, slightly Growth reference: CDC overweight, very overweight? Location: United States Year: 2007 Fagan et al20 N=2,728 How do you describe your weight: very Overweight 45.8 Age: grades 9 to 12 underweight, slightly underweight, Obese 18.2 BMI: self-reported about the right weight, slightly Growth reference: CDC overweight, very overweight? Location: Delaware, USA Year: 2002 Farré-Rovira et al21 N=568 For your age, you consider your weight Males 34.9 Age: 14 to 20 to be: low, normal, high? Females 20.0 BMI: measured Growth reference: BMI  25 kg/m2 Location: Valencia, Spain Year: not provided Foti and Lowry22 N=72,122 How do you describe your weight: very Males 39.4 Age: 12 to 19 underweight, slightly underweight, Females 20.5 BMI: self-reported about the right weight, slightly Growth reference: CDC overweight, very overweight? Location: United States Year: 1999 to 2007 137

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Florin et al23 N=11,012 How do you describe your weight: very 14.4 Age: 14 to 17 underweight, slightly underweight, BMI: self-reported about the right weight, very Growth reference: CDC overweight? Location: United States Year: 2003 Frisco et al24 N=12,683 How do you feel about yourself in Males 42.2 Age: < 20 years terms of weight: very underweight, Females 19.4 BMI: measured slightly underweight, about the right Growth reference: CDC weight, slightly overweight, very Location: United States overweight? Year: 1995 to 1996 Isomaa et al25 N=595 What do you think about your weight? Males 44.6 Age: 15 Do you consider yourself to be: normal Females 6.7 BMI: measured weight; underweight; overweight? Growth reference: IOTF Location: Jakobstad region, Finland Year: not provided Kaplan et al26 N=244 Respondents reported if they were Males 60.5 Age: 11 to 18 underweight, overweight, or the Females 29.1 BMI: self-reported correct weight (Health Behaviors Growth reference: National Research Questionnaire). Council (1964) Location: United States Year: Not provided

138

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Kaltialo-Heino et al27 N=50,046 Respondents reported if they perceived Males Age: 12, 14, 16, 18 themselves as much overweight, Age 12 27 BMI: self-reported somewhat overweight, normal, Age 14 36 Growth reference: BMI 85th percentile somewhat underweight, or much Age 16 42 Location: Finland underweight. Age 18 42 Year: 1979 to 1999 Females Age 12 21 Age 14 14 Age 16 13 Age 18 10 Kim and Lee28 N=74,698 How do you describe your weight Males 31.5 Age: 12 to 19 compared with your friends': Females 17.6 BMI: self-reported underweight, normal weight, Growth reference: Korean overweight, obese? Location: Korea Year: 2007 Kurdak et al29 N=2,353 Respondents were asked to select if Overweight 79.3 Age: 12 to 21 their body weight was underweight, Obese 81.8 BMI: measured normal weight, overweight, or obese. Growth reference: percentile-based Location: Adana, Turkey Year: 1999 to 2000

139

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Kurth and Ellert30 N=6669 Do you think you are far too thin, a bit Obese Age: 11 to 17 too thin, just about the right weight, a Males 7.1 BMI: measured bit too fat, far too fat? Females 0.6 Growth reference: German Location: Germany Year: not provided Note: This study defined obese as having a BMI ≥ 97th percentile. Lee et al31 N=5,443 Do you consider yourself to be too thin Males 17.3 Age: 13 and 16 (underweight), just right, or Females 9.4 BMI: self-reported overweight? Growth standard: Korean Location: South Korea Year: 2006 Lenhart et al32 N=1,180 How do you describe your weight: very Obese 20.8 Age: grades 9 to 12 underweight, slightly underweight, BMI: self-reported about the right weight, slightly Growth reference: CDC overweight, very overweight? Location: Philadelphia, USA Year: 2009 Martin et al33 N=12,789 How do you think of yourself in terms All Age: grades 7 to 12 of weight: very underweight, slightly Overweight 35.1 BMI: measured underweight, about the right weight, Obese 15.1 Growth reference: CDC slightly overweight, very overweight? Males Location: United States Overweight 49.0 Year: 1995 to 1996 Obese 20.4 Females Overweight 20.4 Obese

7.5 140

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Neumark-Sztainer et N=1,499 (1999 weighted sample); N=1307 At this time do you feel that you are: 1999 al34 (2010 sample) very underweight, somewhat Males 41.0 Age: mean age 14 underweight, about the right weight, Females 34.5 BMI: measured somewhat overweight or very 2010 Growth reference: CDC overweight? Males 39.9 Location: Minneapolis/St. Paul, Minnesota Females 24.9 Year: 1999, 2010 O’Dea and Camputi36 N=1,131 Respondents reported if their weight Males 51.6 Age: 6 to 19 was: too fat, about right, or too thin. Females 31.1 BMI: measured Growth reference: Hammer L et al. Am J Dis Child 1991;145:259-263. (≥ 85th percentile) Location: New South Wales, Australia Year: not provided O’Dea and Amy35 N=8,550 Respondents reported if their weight Overweight 64.6 Age: 6 to 18 was: about right, too thin, too fat Obese 39.5 BMI: measured Growth reference: IOTF Location: Australia Year: 2006 O’Haver et al37 N=404 Participants were asked to report how 50.0 Age: 13 to 18 they compared their body weight to BMI: measured that of their peers. Growth reference: CDC Location: Arizona, USA Year: not provided

141

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Ojala et al38 N=8,236 Respondents were asked to report how 1994 Age: 15 they perceived their body size: much Males 34.3 BMI: self-reported too thin, a bit too thin, about the right Females 0 Growth reference: IOTF size, a bit too fat, or much too fat. 1998 Location: Finland Males 30.6 Year: 1999 to 2010 Females 9.4 2002 Males 37.8 Females 1.2 2006 Males 34.5 Females 6.9 2010 Males 34.4 Females 11.5 Ozmen et al39 N=2,101 How do you perceive yourself: Males Age: 15 to 18 underweight, normal weight, fat? Overweight 59.3 BMI: measured Obese 33.3 Growth reference: IOTF Females Location: Manisa, Turkey Overweight 22.2 Year: not provided Obese 16.7 Park40 N=87,418 At the present time, do you think you 40.7 Age: grades 9 and 12 are underweight, about the right BMI: self-reported weight, or overweight? Growth reference: CDC Location: Minnesota, USA Year: 2007

142

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Pasch et al41 N=3010 How do you think of yourself: very Seventh Grade 35.0 Age: grade 7 (mean age 12.7) with follow- underweight, slightly underweight, Eighth Grade 34.4 up in grade 8 underweight, about the right weight, BMI: self-reported slightly overweight, very overweight? Growth reference: CDC Location: Twin Cities, Minnesota Year: 1997 to 2000 Perrin et al42 N=13,001 How do you think of yourself in terms Males 39.14 Age: 11 to 21 of weight: very underweight, slightly Females 17.57 BMI: measured underweight, about the right weight, Growth reference: CDC slightly overweight, very overweight? Location: USA Year: 1995 to 1996 Pritchard et al43 N=33,196 I am overweight: yes or no. Males 47.1 Age: 15 to 19 Females 15.6 BMI: self-reported Growth reference: BMI  25 kg/m2 Location: United States Year: 1980, 1982 Shi et al44 N=824 What do you think about your own Males 20 Age: 12 to 14 body weight: very underweight, slightly Females 0 BMI: measured underweight, about the right weight, Growth reference: WHO slightly overweight, very overweight? Location: Jiangsu province, China Year: 2002

143

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Skinner et al45 N=104 Do you think your weight is very 54 Age: 12 to 18 overweight, slightly overweight, about BMI: measured right, slightly thin, or very thin? Growth reference: CDC Location: USA (Vanderbilt Eskind Pediatric Diabetes Clinic) Year: not provided Note: All participants in this study had a diagnosis of type II diabetes. Standley et al46 N=4035 Respondents were asked to describe All 26 Age: 14 to 15 their weight on a 5-point scale: much Males 30 BMI: measured too thin, too thin, about right, too fat, Females 21 Growth reference: IOTF much too fat. Location: London, UK Year: 2002 Stigler et al47 N=2,339 At this time, do you feel that you are (a) 43.8 Age: grades 8 and 10 very underweight, (b) somewhat BMI: measured underweight, (c) about the right Growth reference: WHO weight, (d) somewhat overweight, or Location: Delhi, India (e) very overweight? Year: 2006 Tang et al48 N=1144 Which body shape do you think you Males 50.0 Age: 10 to 17 have: too thin, relatively thin, all right, Females 38.2 BMI: measured relatively heavy, too heavy? Growth reference: WHO Location: Wuhan, China Year: 2007

144

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) ter Bogt et al49 N=7556 What do you think of your own body? Males 18.5 Age: 11 to 16 Respondents chose from a 5-point Females 11.2 BMI: self-reported Likert-type scale wit responses ranging Growth reference: Dutch Quotelet from 'far too thin' to 'far too heavy.’ standards Location: Netherlands Year: not provided Note: The estimate of weight status underestimation included in this study includes those of normal weight reporting that they are underweight. Ursoniu et al50 N=2908 Respondents were asked to describe Males Age: 14 to 19 their weight using a 5-point Likert-type Overweight 56.99 BMI: self-reported scale. Responses were: very Obese 19.53 Growth reference: WHO underweight, slightly underweight, Females Location: Timis Couty, Romania about the right weight, slightly Overweight 24.34 Year: 2005 overweight, very overweight. Obese 23.14

145

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Viner et al51 N=2,522 Given your age and height, would you Males Age: 11 to 14 say that you are about the right weight, Overweight 59 BMI: measured too heavy, too light, or not sure? Obese 28 Growth reference: UK Females Location: London, UK Overweight 34 Year: 2001 Obese 10 Note: The reported prevalence is the overall percentage of overweight adolescents who underestimated their weight status. An additional 25% of overweight and 23% of obese males reported that they did not know their weight status; 30% of overweight and 27% of obese females were not sure of their weight status. The levels of underestimation would be higher if those who did not know were excluded from the calculation of these percentages. Wang et al52 N=448 How do you describe your body weight: All Age: grades 5 to 8 (mean 11.9) underweight, normal weight, a little Overweight 43.6 BMI: measured overweight, very overweight? Obese 23.7 Growth reference: CDC Males Location: Chicago, USA Overweight 61.4 Year: 2004 Obese 32.3 Females Overweight 30.8 Obese 19.7

146

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Xie et al56 N=2160 Respondents were asked to describe Males 31.5 Age: 11 to 15 their weight status as one of the Females 11.7 BMI: measured following: too thin, relative thin, alright, Growth reference: IOTF relative heavy, too heavy. Location: Wuhan, China Year: 1998 Xie et al53 N=6863 Respondents described their body Males 12.7 Age: middle and high school students shape as too thin, relatively thin, all Females 8.1 BMI: measured right, relatively heavy, or too heavy. Growth reference: IOTF Location: China Year: 2002 Note: This study uses the same data as what is provided in two other studies of weight status underestimation.168,175 Yan et al57 N=2915 Do you consider yourself now to be Males 61.0 Age: 10 to 18 overweight, underweight, or about the Females 41.4 BMI: measured right weight? Growth reference: CDC Location: United States Year: 2005 to 2006 Yost et al58 N=2216 How do you think of yourself in terms Females 20.95 Age: 13 to 18 of weight: very underweight, slightly BMI: measured underweight, about the right weight, Growth reference: NHANES slightly overweight, very overweight? Location: United States Year: 1996

147

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Zaborskis et al59 N=9820 Do you think your body is: much too Lithuania Age: 13 and 15 thin, a bit too thin, about the right size, Males BMI: measured a bit too fat, much too fat? Age 13 54.17 Growth reference: country specific Age 15 79.61 Location: Lithuania, Croatia, United States Females Year: 2001 to 2002 Age 13 26.67 Age 15 18.64 Croatia Males Age 13 34.23 Age 15 37.08 Females Age 13 28.16 Age 15 11.57 United States Males Age 13 26.72 Age 15 17.76 Females Age 13 22.73 Age 15 17.50

148

Level of Weight Status Study Study Characteristics Measure of Perceived Weight Status Underestimation (%) Zhang et al60 N=14,879 How do you describe your weight: very Hong Kong Age: 14 to 18 underweight, slightly underweight, Males 13.64 BMI: self-reported about the right weight, slightly Females 0 Growth reference: IOTF overweight, very overweight? Macau Location: Hong Kong, Taipei, and Macau, Males 21.31 China; New York and Los Angeles, USA Females 0 Year: 2003 Taipei Males 4.14 Females 0 New York City Males 49.35 Females 26.67 Los Angeles Males 38.53 Females 21.74

149

150

Appendix B. Detailed list of variables from the Canadian Community Health Survey used in the analysis of this project Construct Year CCHS Variable Name Survey Question Outcome Perceived 2001 HWTA_4 Do you consider yourself: overweight, Weight 2003 HWTC_4 underweight, or just about right? Status 2005 HWTE_4 2007-2010 HWT_4 Actual Weight Status Height 2001 HWTADHTM How tall are you without shoes on? 2003 HWTCDHTM 2005 HWTEDHTM 2007-2010 HWTDHTM Weight 2001 HWTADWTK How much do you weight? 2003 HWTCDWTK 2005 HWTEDWTK 2007-2010 HWTDWTK Predictors Sex 2001 DHHA_SEX Completed by interviewer (based on 2003 DHHC_SEX interview’s observation) 2005 DHHE_SEX 2007-2010 DHH_SEX Ethnicity 2001 SDCADRAC People living in Canada come from many 2003 SDCCDRAC different cultural and racial backgrounds. 2005 SDCEDCGT Are you: White, Chinese, South Asian, 2007-2010 SDCDCGT Black, Filipino, Latin American, South East Asian, Arab, West Asian, Japanese, Korean? (included Aboriginal Peoples of North America prior to June 2005) Aboriginal 2001 - Are you an Aboriginal person that is North Status 2003 - American Indian, Métis, or Inuit?

2005 SDCE_41 150

2007-2010 SDC_41 Age 2001 DHHA_AGE Derived from date of birth and confirmed 2003 DHHC_AGE with respondent. 2005 DHHE_AGE 2007-2010 DHH_AGE Variables for Sample Selection Pregnant 2001 MAMA_037 It is important to know when analyzing 2003 MAMC_037 health whether or not the person is 2005 MAME_037 pregnant. Are you pregnant? 2007-2010 MAM_037

151

Construct Year CCHS Variable Name Survey Question Proxy 2001 ADMA_PRX 2003 ADMC_PRX 2005 ADME_PRX 2007-2010 ADM_PROX Geographic Variables Health 2001 GEOA_HR4 Region 2003 GEOCDHR4 2005 GEOEDHR4 2007-2010 GEODHR4 Census 2001 GEOADCSD Subdivision 2003 GEOCDCSD 2005 GEOEDCSD 2007-2010 GEODCSD Other Variables Interview 2001 ADMA_N09 Was this interview conducted on the Mode 2003 ADMC_N09 telephone or in person? (only provided for 2005 ADME_N09 those selected as part of the telephone 2007-2010 ADM_N09 frame) Date of 2001 ADMA_DAT Interview 2003 ADMC_YOI ADMC_MOI ADMC_DOI 2005 ADME_YOI ADME_MOI ADME_DOI 2007-2010 ADM_YOI ADM_MOI ADM_DOI Date of 2001 DHHA_DB What is (respondent’s) date of birth? Birth 2003 DHHC_YOB DHHC_MOB

DHHC_DOB 151

2005 DHHE_YOB DHHE_MOB DHHE_DOB 2007-2010 DHH_YOB DHH_MOB DHH_DOB

Appendix C. Overview of changes made to health regions to ensure their comparability across cycles of the Canadian Community Health Survey

Substantial changes were made to the borders of some health regions between cycles of the Canadian Community Health Survey. To ensure these borders remained consistent for the course of the survey, some health regions were combined. These changes are described in Table 19. Modified health regions are presented in bold font, with the final code number reflecting the final combinations of health regions.

152

Table 19. Overview of changes made to health regions in the Canadian Community Health Survey (CCHS) to ensure their consistency across time. Numbers provided represent the code assigned to each health region by Statistics Canada. The final code provided reflects all combined health regions. Survey Year Health Region* 2001 2003 2005 2007 2008 2009 2010 Final Code

Health/Comm Services St John's Region 1001 1001 1011 1011 1011 1011 1011 1011 Health/Comm Services Eastern Region 1002 1002 Health/Comm Services Central Region 1003 1003 1012 1012 1012 1012 1012 1012 Health/Comm Services Western Region 1004 1004 1013 1013 1013 1013 1013 1013 Grenfell Regional Health Services Board 1005 1005 Newfoundland 1014 1014 1014 1014 1014 1014 Health Labrador Corporation 1006 1006 Urban Health Region 1101

Rural Health Region 1102

West Prince 1101 1101

East Prince 1102 1102

Queens 1103 1103 1199

Kings 1104 1104

Kings County 1101 1101 1101 1101

Prince Edward Island Edward Prince Queens County 1102 1102 1102 1102

Prince County 1103 1103 1103 1103

Zone 1 1201 1201 1201 1201 1201 1201 1201 1201 Zone 2 1202 1202 1202 1202 1202 1202 1202 1202 Zone 3 1203 1203 1203 1203 1203 1203 1203 1203 Zone 4 1204 1204 1204 1204 1204 1204 1204 1204

NovaScotia Zone 5 1205 1205 1205 1205 1205 1205 1205 1205

Zone 6 1206 1206 1206 1206 1206 1206 1206 1206 153

Survey Year

Health Region* 2001 2003 2005 2007 2008 2009 2010 Final Code

Region 1 1301 1301 1301 1301 1301 1301 1301 1301

Region 2 1302 1302 1302 1302 1302 1302 1302 1302 Region 3 1303 1303 1303 1303 1303 1303 1303 1303 Region 4 1304 1304 1304 1304 1304 1304 1304 1304 Region 5 1305 1305 1305 1305 1305 1305 1305 1305

New Brunswick New Region 6 1306 1306 1306 1306 1306 1306 1306 1306 Region 7 1307 1307 1307 1307 1307 1307 1307 1307 Region du Bas-Saint-Laurent 2401 2401 2401 2401 2401 2401 2401 2401 Region du Saguenay - Lac-Saint-Jean 2402 2402 2402 2402 2402 2402 2402 2402 Region du Quebec 2403 2403 2403 2403 2403 2403 2403 2403 Region de la Mauricie-centre-du-Quebec 2404 2404 2404 2404 2404 2404 2404 2404 Region de l'Estrie 2405 2405 2405 2405 2405 2405 2405 2405 Region de -Centre 2406 2406 2406 2406 2406 2406 2406 2406 Region de l'Outaouais 2407 2407 2407 2407 2407 2407 2407 2407 Region de l'Abitibi-Temiscaminque 2408 2408 2408 2408 2408 2408 2408 2408 Region de la cote-nord 2409 2409 2409 2409 2409 2409 2409 2409

Quebec Region du nord-du-quebec 2410 2410 2410 2410 2410 2410 2410 2410 Reg. de la Gaspesie-iles-de-la-madeleine 2411 2411 2411 2411 2411 2411 2411 2411 Region de la Chaudiere-Appalaches 2412 2412 2412 2412 2412 2412 2412 2412 Region de Laval 2413 2413 2413 2413 2413 2413 2413 2413 Region de Lanaudiere 2414 2414 2414 2414 2414 2414 2414 2414 Region des Laurentides 2415 2415 2415 2415 2415 2415 2415 2415 Region de la Monteregie 2416 2416 2416 2416 2416 2416 2416 2416 Region des terres-cries-de-la-baie-James 2418 2418

154

Survey Year

Health Region* 2001 2003 2005 2007 2008 2009 2010 Final Code Algoma Public Health Unit 3526 3526 3526 3526 3526 3526 3526 3526 Brant Public Health Unit 3527 3527 3527 3527 3527 3527 3527 3527 Durham Public Health Unit 3530 3530 3530 3530 3530 3530 3530 3530 Elgin-St Thomas PHU 3531 3531 3531 3531 3531 3531 3531 3531 Bruce-Grey-Owen Sound PHU 3533 3533 3533 3533 3533 3533 3533 3533 Haldimand-Norfolk PHU 3534 3534 3534 3534 3534 3534 3534 3534 Haliburton-Kawartha-Pine Ridge PHU 3535 3535 3535 3535 3535 3535 3535 3535 Halton PHU 3536 3536 3536 3536 3536 3536 3536 3536 Hamilton-Wentworth PHU 3537 3537 3537 3537 3537 3537 3537 3537 Hastings and Prince Edward PHU 3538 3538 3538 3538 3538 3538 3538 3538 Huron PHU 3539 3539 3539 3539 3539 3539 3539 3539 Kent-Chatham PHU 3540 3540 3540 3540 3540 3540 3540 3540 Kingston-Frontenac-Lennox-Addington PHU 3541 3541 3541 3541 3541 3541 3541 3541

Ontario Lambton PHU 3542 3542 3542 3542 3542 3542 3542 3542 Leeds-Grenville-Lanark PHU 3543 3543 3543 3543 3543 3543 3543 3543 Middlesex-London PHU 3544 3544 3544 3544 3544 3544 3544 3544 Muskoka-Parry Sound PHU 3545 3545 3599

Niagara PHU 3546 3546 3546 3546 3546 3546 3546 3546 North Bay PHU 3547 3547 3547 3547 3547 3547 3547 3599 Northwestern PHU 3549 3549 3549 3549 3549 3549 3549 3549 Ottawa Carleton PHU 3551 3551 3551 3551 3551 3551 3551 3551 Oxford PHU 3552 3552 3552 3552 3552 3552 3552 3552 Peel PHU 3553 3553 3553 3553 3553 3553 3553 3553 Perth PHU 3554 3554 3554 3554 3554 3554 3554 3554 Peterborough PHU 3555 3555 3555 3555 3555 3555 3555 3555

155

Survey Year

Health Region* 2001 2003 2005 2007 2008 2009 2010 Final Code Porcupine PHU 3556 3556 3556 3556 3556 3556 3556 3556 Renfrew PHU 3557 3557 3557 3557 3557 3557 3557 3557 Eastern Ontario PHU 3558 3558 3558 3558 3558 3558 3558 3558 Simcoe PHU 3560 3560 3560 3560 3560 3560 3560 3599

Sudbury PHU 3561 3561 3561 3561 3561 3561 3561 3561 Thunderbay PHU 3562 3562 3562 3562 3562 3562 3562 3562 Timiskaming PHU 3563 3563 3563 3563 3563 3563 3563 3563 Ontario Waterloo PHU 3565 3565 3565 3565 3565 3565 3565 3565 Wellington-Dufferin-Guelph PHU 3566 3566 3566 3566 3566 3566 3566 3566 Windsor-Essex PHU 3568 3568 3568 3568 3568 3568 3568 3568 York PHU 3570 3570 3570 3570 3570 3570 3570 3570 City of Toronto PHU 3595 3595 3595 3595 3595 3595 3595 3595 4610 4610 4610 4610 4610 4610 4610 4610 Brandon 4615 4615 4615 4615 4615 4615 4615 4615 Morth Eastman 4620 4620 4620 4620 4620 4620 4620 4620

South Eastman 4625 4625 4625 4625 4625 4625 4625 4625

Interlake 4630 4630 4630 4630 4630 4630 4630 4630 Central 4640 4640 4640 4640 4640 4640 4640 4640

Manitoba Marquette 4650 4645 4645 4645 4645 4645 4645 4645 South Westman 4655 parkalnd 4660 4660 4660 4660 4660 4660 4660 4660 Normal 4670 4670 4670 4670 4670 4670 4670 4670 Burntwood+Churchill 4680 4680 4685 4685 4685 4685 4685 4685

156

Survey Year

Health Region* 2001 2003 2005 2007 2008 2009 2010 Final Code Weyburn (A) Service Area 4701 4701 4701 4701 4701 4701 4701 4701 Moose Jaw (B) Service Area 4702 4702 4702 4702 4702 4702 4702 4702 Swift Curent (C) Service Area 4703 4703 4703 4703 4703 4703 4703 4703 Regina (D) SA 4704 4704 4704 4704 4704 4704 4704 4704 Yorkton (E) SA 4705 4705 4705 4705 4705 4705 4705 4705 Saskatoon (F) SA 4706 4706 4706 4706 4706 4706 4706 4706 Rosetown (G) SA 4707 4707 4707 4707 4707 4707 4707 4707

Saskatchewan Melfort (H) SA 4708 4708 4708 4708 4708 4708 4708 4708 Prince Albert (I) SA 4709 4709 4709 4709 4709 4709 4709 4709 North Battleford (J) SA 4710 4710 4710 4710 4710 4710 4710 4710 North. Hlth. Serv. Branch (K) SA 4711 4714 4714 4714 4714 4714 4714 4714 Chinook Regional health Authority 4801 4820 4820 4821 4821 4821

Palliser Regional Health Authority 4802

Headwaters RHA 4803

Calgary RHA 4804

Health Authority #5 4805

David Thompson RHA 4806

East Central HA 4807

Westview RHA 4808 4899

Alberta Crossroads RHA 4809

Capital Health Authority 4810

Aspen RHA 4811

Lakeland RHA 4812

Mistahia RHA 4813

Peace RHA 4814

Keeweetinok Lakes RHA 4815 157

Survey Year

Health Region* 2001 2003 2005 2007 2008 2009 2010 Final Code Palliser Health Region 4821 4821 4822 4822 4822

Calgary Health Region 4822 4822 4823 4823 4823

David Thompson RHA 4823 4823 4824 4824 4824

East Central Health 4824 4824 4825 4825 4825

Capital Health Authority 4825 4825 4826 4826 4826

Aspen RHA 4826 4826 4827 4827 4827

Peace Country Health 4827 4827 4828 4828 4828 4899 Northern Lights RHA 4816 Alberta 4828 4828 4829 4829 4829 Northwestern RHA 4817

South Zone 4831

Calgary Zone 4832

Central Zone 4833

Edmonton Zone 4834

North Zone 4835

158

Survey Year

Health Region* 2001 2003 2005 2007 2008 2009 2010 Final Code East Kootenay 5901 5911 5911 5911 5911 5911 5911 5911 West Kootenay-Boundary 5902 5912 5912 5912 5912 5912 5912 5912 North Okanagan 5903 5913 5913 5913 5913 5913 5913 South Okanagan Similkameen 5904 5999 Thompson 5905 5914 5914 5914 5914 5914 5914 Fraser Valley/Fraser East 5906 5921 5921 5921 5921 5921 5921 5921 South Fraser Valley/Fraser South 5907 5923 5923 5923 5923 5923 5923 5923

Simon Fraser/Fraser North 5908 5922 5922 5922 5922 5922 5922 5922 Coast Garibaldi 5909 5933 5933 5933 5933 5933 5933 5933 North Shore 5918 Central Island 5910 5942 5942 5942 5942 5942 5942 5942 Upper Island/Central Coast 5911 5943 5943 5943 5943 5943 5943

British Columbia British 5999 Cariboo 5912 5914 5914 5914 5914 5914 5914 North West 5913 5951 5951 5951 5951 5951 5951 5951 Peace Liard/Northeast 5914 5953 5953 5953 5953 5953 5953 5953 Northern Interior 5915 5952 5952 5952 5952 5952 5952 5952 Vancouver 5916 5932 5932 5932 5932 5932 5932 5932 Burnaby 5917 5922 5922 5922 5922 5922 5922 5922 Richmond 5919 5931 5931 5931 5931 5931 5931 5931 Capital/South Vancouver Island 5920 5941 5941 5941 5941 5941 5941 5941 6001 6001 6001 6001 6001 6001 6001 6001

Yukon

Northwest 6101 6101 6101 6101 6101 6101 6101 6101 ries

Territo Nunavut 6201 6201 6201 6201 6201 6201 6201 6201 Note: Modified health regions are shown in bold font. *Only the original names are presented for health regions that underwent a name change between 2001 and 2010.

159

Appendix D. Comparison of minimum sample size per cluster for multilevel analysis

The following provides a comparison of census subdivisions (CSDs) with different minimum sample sizes. The variance across CSDs is compared when including only CSDs with at least five overweight males and five overweight females were compared to including only those with at least two overweight adolescents of each sex. The variance across CSDs in weight status underestimation was highly significant for both minimum sample sizes (Table 20). Since the number of overall clusters is considered to be more important in obtaining accurate estimates than the number of individuals per cluster,228 all analysis was conducted using CSDs that included at least two overweight males and two overweight females.

160

Table 20. Comparison of variance in weight status underestimation across census subdivisions (CSDs) with different minimum sample sizes CSDs with at least 5 overweight adolescents of each sex CSDs with at least 2 overweight adolescents of each sex All Adolescents Males Only Females Only All Adolescents Males Only Females Only Variance (95% CI) Variance (95% CI) Variance (95% CI) Variance (95% CI) Variance (95% CI) Variance (95% CI) Intercept 0.19 (0.01, 0.29) 0.18 (0.08, 0.28) 0.33 (0.11, 0.55) 0.30 (0.18, 0.39) 0.33 (0.20, 0.46) 0.57 (0.32, 0.82) MOR 1.52 1.50 1.73 1.68 1.73 2.05 ICC 0.055 0.052 0.091 0.083 0.092 0.148 Note: (1) All models controlled for the effect of interview mode Abbreviations: CSD (census subdivision); CI (confidence interval); MOR (median odds ratio); ICC (intraclass correlation coefficient)

161

Appendix E. Comparison of weight status underestimation across ethnic groups adjusted for age, severity of overweight, and the effect of time, as well as interview mode

This Appendix provides a comparison of weight status underestimation among overweight adolescents of different ethnic backgrounds, adjusted for the effects of age, severity of overweight, time, and interview mode. Tables 21 and 22 provide these same comparisons separately for males and females, respectively. Note that comparisons across ethnic groups are only provided for males and females separately since the relationship between other characteristics (i.e. time) and weight status underestimation is different for males and females.

162

Table 21. Adjusted odds ratios comparing weight status underestimation across ethnic groups among male overweight adolescents Reference Group White Black Asian Aboriginal Latin American Other Mixed Ethnic Group OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Black 3.4 (1.5, 7.6) - Asian 0.6 (0.4, 0.8) 0.2 (0.1, 0.4) - Aboriginal 0.9 (0.7, 1.3) 0.3 (0.1, 0.7) 1.6 (1.0, 2.6) - Latin American 1.3 (0.5, 3.7) 0.4 (0.1, 1.4) 2.3 (0.8, 2.7) 1.4 (0.5, 4.1) - Other 1.7 (0.7, 4.1) 0.5 (0.2, 1.7) 2.9 (1.1, 7.6) 1.8 (0.7, 4.7) 1.3 (0.3, 4.9) - Mixed 0.9 (0.5, 1.6) 0.3 (0.1, 0.7) 1.5 (0.7, 3.0) 0.9 (0.5, 1.9) 0.7 (0.2, 2.1) 0.5 (0.3, 2.1) - Missing 1.2 (0.8, 2.1) 0.4 (0.1, 0.9) 2.2 (1.2, 3.9) 1.3 (0.7, 2.4) 0.9 (0.3, 2.9) 0.7 (0.7, 1.0) 1.5 (0.8, 3.0) Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All OR are adjusted for the effects of age, severity of overweight, and interview mode. Abbreviations: OR (odds ratio); CI (confidence interval).

163

Table 22. Adjusted odds ratios comparing weight status underestimation across ethnic groups among female overweight adolescents Reference Group White Black Asian Aboriginal Latin American Other Mixed Ethnic Group OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Black 1.5 (0.8, 3.0) - Asian 0.6 (0.3, 1.0) 0.4 (0.15, 0.93) - Aboriginal 0.8 (0.5, 1.2) 0.5 1.4 - (0.2, 1.1) 0.6, 2.9) Latin American 0.7 (0.2, 1.9) 0.5 1.2 0.9 (0.3, 2.7) - (0.1, 1.5) (0.4, 3.9) Other 1.1 (0.4, 3.5) 0.7 2.0 1.5 (0.4, 4.9) 1.7 (0.4, 7.6) - (0.2, 2.8) (0.6, 7.1) Mixed 0.7 (0.3, 1.4) 0.5 1.2 0.9 (0.4, 2.1) 1.0 (0.3, 3.5) 0.6 - (0.2, 1.2) (0.5, 3.0) (0.2, 2.3) Missing 0.9 (0.5, 1.8) 0.6 1.6 1.2 (0.5, 2.7) 1.3 (0.4, 1.6) 0.8 (0.2, 3.1) 1.5 (0.8, 3.0) (0.2, 1.6) (0.7, 3.9) Notes: (1) Results significant at the 5% level are highlighted in bold; (2) All OR are adjusted for the effects of age, severity of overweight, and interview mode. Abbreviations: OR (odds ratio); CI (confidence interval).

164

165

Curriculum Vitae

Name: Mary Ellen Kuenzig

Post-secondary The University of Western Ontario Education and London, Ontario, Canada Degrees: 2005-2010 BMSc

The University of Western Ontario London, Ontario, Canada 2010-present MSc

Honours and CHRI Trainee Travel Award Awards: Children’s Health Research Institute 2012

Graduate Thesis Research Award The University of Western Ontario 2012

Related Work Teaching Assistant, Multivariable Methods in Biostatistics Experience The University of Western Ontario 2012

Graduate Research Assistant The University of Western Ontario 2010-present

Invited Presentations: Corbett B, Kuenzig E. Research Data Centre: When and How to Access Data. Presented at 2011-2012 Population and Life Course Studies and Research Data Centre’s Statistics 165 and Data Series, University of Western Ontario. (November 2011)

Publications: Kuenzig E, Wilk P. Do overweight and obese adolescents accurately perceive their weight status? Obes Facts. 2012; 5(S1):257.

Presentations: Oral Presentations Kuenzig E, Wilk P, Bauer G. Inaccurate Weight Status Perception among Overweight Canadian Adolescents. Oral presentation at Public Health in Canada: Creating and Sustaining Healthy Environments. (Edmonton, Alberta; June 2012)

166

Poster Presentations Kuenzig E, Bauer G, Wilk P. Predictors of Weight Status Misperception among Overweight Canadian Adolescents. Poster presentation at London Health Research Day. (University of Western Ontario; March 2012)

166