UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Insulin and related factors in breast-cancer mortality Borugian, Marilyn Jean 2003

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

Item Metadata


831-ubc_2003-854264.pdf [ 9.35MB ]
JSON: 831-1.0091205.json
JSON-LD: 831-1.0091205-ld.json
RDF/XML (Pretty): 831-1.0091205-rdf.xml
RDF/JSON: 831-1.0091205-rdf.json
Turtle: 831-1.0091205-turtle.txt
N-Triples: 831-1.0091205-rdf-ntriples.txt
Original Record: 831-1.0091205-source.json
Full Text

Full Text

I N S U L I N A N D R E L A T E D F A C T O R S I N B R E A S T -C A N C f i l l M O R T A L I T Y by M A R I L Y N 11.AN B O R U G 1 A N B . A . , University o f Michigan, 1970 M . S c , University o f British (Columbia, 2000 t\ T H E S I S S U B M I T l ' E D I N P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F D O C T O R O F P H I L O S O P H Y in T H E F A C U L T Y OF' G R A D U A T E S T U D I E S Department o f Health ('are and Epidemiology W e accept this tl^iuo un a* t t£^ jau in^o j"hc requyetl standard T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A A p r i l 2003 © Mari lyn Jean Borugian, 2003 UBC Rare Books and Special Collections - Thesis Authorisation Form Page 1 of 1 I n p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t o f t h e r e q u i r e m e n t s f o r a n a d v a n c e d d e g r e e a t t h e U n i v e r s i t y o f B r i t i s h C o l u m b i a , I a g r e e t h a t t h e L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e a n d s t u d y . I f u r t h e r a g r e e t h a t p e r m i s s i o n f o r e x t e n s i v e c o p y i n g o f t h i s t h e s i s f o r s c h o l a r l y p u r p o s e s may be g r a n t e d b y t h e h e a d o f my d e p a r t m e n t o r b y h i s o r h e r r e p r e s e n t a t i v e s . I t i s u n d e r s t o o d t h a t c o p y i n g o r p u b l i c a t i o n o f t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l n o t be a l l o w e d w i t h o u t my w r i t t e n p e r m i s s i o n . D e p a r t m e n t o f U e A l 71-1 j (zip: £ > i : i m 6 (/:•(•••-j The U n i v e r s i t y o f B r i t i s h C o l u m b i a V a n c o u v e r , C a n a d a http://www.library.ubc.ca/spcoll/thesauth.html 2/5/03 University of British Columbia Department of Health Care and Epidemiology Abstract INSULIN A N D R E L A T E D FACTORS IN B R E A S T - C A N C E R M O R T A L I T Y by Marilyn Jean Borugian Chairperson of the Supervisory Committee: Professor Samuel B. Sheps background. High levels of insulin have been associated with increased risk of breast cancer, and poorer survival after a diagnosis of breast cancer. To study possible modifiable factors affecting breast cancer survival, in 1991 data and sera were collected from 603 breast cancer patients and archived. The current study made use of these prospective 10-year cohort data to test the prognostic value of insulin and C-peptide levels and related lifestyle factors in breast cancer survival. Objectives. The primary aim of this study was to test the hypothesis that elevated insulin at diagnosis is directly related to breast cancer mortality, as well as insulin-related factors including body size and shape, dietary intake and physical inactivity. A secondary aim was to describe the effect on mortality of changes made in the hyperinsulinernia-related lifestyle factors of dietary intake and body size during the first 2 years post-diagnosis. Methods. Using a prospective design, outcomes for cohort members were ascertained by linkage to the BC Cancer Registry and Cancer Agency Information System (CAIS). The primary ii outcome of interest was breast-cancer mortality, but all-cause mortality was also examined. Archived sera were tested for insulin, C-peptide (an index of insulin secretion), fmctosamine (a marker of glycemia over the previous 1 to 3 weeks) and SHBG (sex hormone binding globulin) by personnel blinded to participant outcomes. Lifestyle data, including dietary, physical activity and body size factors were analysed using Cox proportional hazards models to relate the prognostic markers to outcomes, controlling for potential confounders such as age, tumour stage, and menopausal status. Data for the biological variables were analyzed as a nested case-control study due to limited serum volumes, with at least one survivor as a control for each breast cancer death, matched on stage at diagnosis and length of follow-up. Results. Serum insulin levels at diagnosis were directly related to breast-cancer mortality in post-menopausal women, with evidence of a non-linear relationship (OR, 5.2; 95% CI, 1.0-27.8 for the 2 n d quartile; OR, 5.0; 95% CI, 1.0-24.4 for the 3 rd quartile; OR, 2.6; 95% CI, 0.4, 14.6 for 4 th quartile, compared with the 1st quartile), but no relationship was observed in pre-menopausal women. No significant association with breast-cancer mortality was observed for C-peptide, fmctosamine or SHBG levels at diagnosis. Increased waist-to-hip ratio (WHR) at diagnosis was directly related to both breast-cancer mortality (RR, 3.2; 95% CI, 1.3-8.1 comparing highest quartile to lowest quartile) and all-cause mortality (RR, 2.6; 95% CI, 1.3-5.1) in post-menopausal women, but not in pre-menopausal women (RR, 1.0; 95% CI, 0.4-2.7), independent of age and body-mass index (BMI). The association was restricted to women with estrogen receptor (ER)-positive tumours. Dietary intake of total fat and saturated fat was direcdy related to breast-cancer mortality in pre-menopausal women (RR, 4.8; 95% CI, 1.3-18.1, comparing highest to lowest quartile of total fat intake), but not in post-menopausal women. Dietary protein intake was inversely related to breast-cancer mortality (RR, 0.4; 95% iii CI, 0.2-0.8) for all women. No significant association with breast-cancer mortality was found for BMI, physical activity, dietary intake of carbohydrate, fiber, alcohol, total energy or glycemic load (glycemic index times number of grams) at diagnosis. Analysis of changes made during the first 2 years post-diagnosis showed that only changes in BMI were significantly related to mortality, where a decrease of 5% or more was associated with more risk (RR, 4.1; 95% CI, 1.3-13.0 for pre-menopausal women; RR, 2.5; 95% CI, 1.0-5.8 for post-menopausal women) than an increase of 5% or more (RR, 2.0; 95% CI, 0.8-5.0 for pre-menopausal women; RR, 1.6; 95% CI, 0.8-3.4 for post-menopausal women), compared with no change (plus or minus less than 5%). Conclusions. The results are consistent with the study hypothesis, supporting an association between hyperinsulinemia and breast-cancer mortality, modified by menopausal status and ER tumour subtype. IV T A B L E O F C O N T E N T S Abstract : ii List of Tables ix List of Figures xi Acknowledgments xii Glossary xiii Quotation xiv INTRODUCTION 1 Purpose . 1 Impact of the Project on the Burden of Cancer 1 Specific Aims 2 Hypotheses 2 Rationale 3 The Following Sections 3 BACKGROUND 5 OVERVIEW 5 I. EPIDEMIOLOGY OF BREAST CANCER 7 Breast Cancer Incidence and Mortality Trends 7 Canadian and U.S. Trends 7 International Trends 8 Factors Associated with Risk of Breast Cancer 10 Genetics 10 Demographic Factors 11 Hormone Receptor Status 11 Hormone Use 12 Mammographic Density 12 Reproductive Factors 13 Energy Balance 14 Obesity 15 Dietary Factors 16 Cigarette Smoking 21 Hyperinsulinemia 21 Factors Associated with Mortality from Breast Cancer 21 Clinical Factors 21 Modifiable Lifestyle Factors 22 H. INSULIN AS A RISK FACTOR 25 The Biological Role of Insulin and Insulin Resistance 25 Insulin as a Growth Factor 25 Insulin and Reproductive Hormones 26 The Normal Role of Insulin Resistance 26 An Evolutionary Perspective 27 v Problems of Excess Insulin _ _ _ _ _ 27 Insulin Resistance and Cancer .28 Experimental Data 29 Human Observational Data 29 Insulin Resistance and Breast Cancer Risk , 30 Evidence for an Insulin Role in Breast Cancer Mortality 32 Possible Mechanisms 32 Sex Steroid Hormones • 33 Direct Action_ 34 III. LOWERING THE RISK _35 Biological Factors Affecting Insulin Resistance 35 Genetic Factors . 35 Age . 36 Menopause 36 Pregnancy ____ 36 Puberty _ _ 37 Nervous System Input_ 37 Lifestyle Factors Affecting Insulin Resistance 37 Dietary Fat . 39 Dietary Carbohydrate 40 Stress 41 Cigarette Smoking . 41 Obesity . 41 Physical Activity 42 SUMMARY 43 METHODS AND MA TERIALS 45 Design 45 Study Design 45 Participants 45 Recruitment 46 Selection for nested case-control 47 Questionnaire 49 Lifestyle Variables 49 Biological Variables 51 Outcomes .52 Sample Size and Power . 52 Procedures . 53 Data Collection and Coding . 53 Collection of Blood Samples 54 Lab Protocols 54 Lab Assay Pilot 55 Chart Abstracts 55 Data Analysis 56 Variable Selection_ 57 Calculated Variables 58 Descriptive Statistics 59 Analysis of Questionnaire Variables 60 Analysis of Lab Assay Variables , 62 vi Ethics Approvals Dissemination of Results 63 63 RESULTS 65 Overview 65 Description of Study Population 66 Distributions and Correlations 67 Body Size and Shape Variables 68 Weight and BMI 68 Waist-to-Hip Ratio 6 9 The Effects of Age and BMI on WHR Risk 70 Multivariate Analyses of WHR 70 Survival by Waist-to-hip Ratio and Stage at Diagnosis 71 Stratification on Estrogen Receptor Status and Family History 72 Physical Activity and Dietary Variables 72 Physical Activity 72 Total Energy Intake 73 Percent of Energy from Macronutrients 74 Fat Consumption 74 Protein Consumption _ 7 5 Carbohydrate Consumption 75 Alcohol Consumption 75 Glycemic Load 76 Biological Variables 76 Serum Insulin Levels 77 C-Peptide and Fractosamine 78 Sex Hormone Binding Globulin Levels 78 The Effect of Tumour and Treatment Variables on Insulin Prognostic Association 78 Changes Made Post-Diagnosis 79 Body Size and Shape Changes 79 Dietary Changes 80 DISCUSSION 143 Overview , 143 Lifestyle Factors at Diagnosis 144 Body Size 144 Body Shape . 145 Physical Activity 149 Macronutrient and Total Energy Intake 151 Glycemic Load 151 Alcohol 152 Biological Factors at Diagnosis 153 Insulin and C-Peptide . 153 Sex Hormone Binding Globulin 154 Changes Made Post-Diagnosis 155 Body Size and Shape Changes 155 Dietary Changes 156 Biological Pathways 156 Strengths and Limitations 157 Validity 159 vii The Role of Chance 160 Sources of Bias 160 Confounding 162 Measurement Error 162 Generalizability 164 Comparison with B.C. Population, 1991-2000 164 The Evidence for Causality 165 Strength of Association 165 Biological Plausibility 166 Consistency 166 Dose-Response Relationship 167 Coherance 167 Temporality 167 Conclusion 168 Summary 168 Implications for Breast Cancer Patients Today 170 Future Research 171 Bibliography 172 Appendices A. Study questionnaire 189 B. Power calculation 203 C. Foods by food group 204 D. Questionnaire return: days post-diagnosis 207 viii LIST OF T A B L E S Number Page Table 1. Selected characteristics of the study population at diagnosis. 98 Table 2. Distribution of causes of death. 101 Table 3. Average survival time from enrolment, by stage and vital status. 102 Table 4. Correlations. 103 Table 5. Body weight at diagnosis of breast cancer and age-adjusted mortality.104 Table 6. Body-mass index at diagnosis of breast cancer and age-adjusted mortality. 106 Table 7. Waist-to-hip ratio at diagnosis and breast-cancer mortality. 107 Table 8. Waist-to-hip ratio at diagnosis and breast-cancer mortality by age subgroup. 108 Table 9. Waist-to-hip ratio at diagnosis and breast-cancer mortality by BMI subgroup. 109 Table 10. Waist-to-hip ratio at diagnosis and breast-cancer mortality by tumour size subgroup. 110 Table 11. Waist-to-hip ratio at diagnosis and breast-cancer mortality by estrogen receptor status and family history subgroups. 111 Table 12. Physical activity at diagnosis and breast-cancer mortality. 112 Table 13. Total energy consumption at diagnosis and breast-cancer mortality. 114 Table 14. Total energy consumption at diagnosis and breast-cancer mortality by menopausal status. 115 Table 15. Percent of energy at diagnosis and breast-cancer mortality. 116 Table 16. Fat consumption at diagnosis and breast-cancer mortality. 117 Table 17. Fat consumption at diagnosis and breast-cancer mortality by menopausal status. 118 Table 18. Protein consumption at diagnosis and breast-cancer mortality. 119 Table 19. Protein consumption at diagnosis and breast-cancer mortality by menopausal status. 120 Table 20. Carbohydrate consumption at diagnosis and breast-cancer mortality. 121 Table 21. Alcohol consumption at diagnosis and breast-cancer mortality. 122 Table 22. Average daily glycemic load and breast-cancer mortality. 124 Table 23. Insulin, C-peptide, fructosamine and SHBG. 125 Table 24. Serum insulin levels at diagnosis and breast-cancer mortality. 126 Table 25. Serum C-peptide levels at diagnosis and breast-cancer mortality. 127 Table 26. Serum fructosamine levels at diagnosis and breast-cancer mortality. 128 ix Table 27. C-peptide-to-fructosamine ratio at diagnosis and breast-cancer mortality. 129 Table 28. Serum SHBG levels at diagnosis and breast-cancer mortality. 130 Table 29. Insulin and post-menopausal breast-cancer mortality, adjusted for family history, estrogen receptor status, and both. 131 Table 30. Multivariate analysis of insulin and post-menopausal breast-cancer mortality. ; 132 Table 31. Changes in body srze and food intake during the first 2 years post-diagnosis. 133 Table 32. Change in BMI in the first 2 years post-diagnosis and breast-cancer mortality. ; ; 134 Table 33. Change in BMI in the first 2 years post-diagnosis and breast-cancer mortality by menopausal status. 135 Table 34. Change in WHR in the first 2 years post-diagnosis and breast-cancer mortality. 136 Table 35. Change in carbohydrate intake in the first 2 years post-diagnosis and breast-cancer mortality. 137 Table 36. Change in fat intake in the first 2 years post-diagnosis and breast-cancer mortality. 138 Table 37. Change in protein intake in the first two years post-diagnosis and breast-cancer mortality. ; 139 Table 38. Change in energy intake in the first 2 years post-diagnosis and breast-cancer mortality. ; 140 Table 39. British Columbia Cancer Agency new female breast-cancer patients (BC Residents only). 141 x LIST OF FIGURES Number Page Figure 1. Study Overview. 48 Figure 2. Survival by stage at diagnosis. 82 Figure 3. Distribution of waist-to-hip ratio at diagnosis. 83 Figure 4. Distribution of body-mass index at diagnosis. 84 Figure 5. Distribution of dietary fat intake at diagnosis. 85 Figure 6. Distribution of dietary protein intake at diagnosis. 86 Figure 7. Distribution of dietary e-carbohydrate intake at diagnosis. 87 Figure 8. Distribution of dietary fiber intake at diagnosis. 88 Figure 9. Distribution of dietary energy intake at diagnosis. 89 Figure 10. Distribution of serum insulin level at diagnosis. 90 Figure 11. Distribution of serum C-peptide level at diagnosis. 91 Figure 12. Distribution of serum fmctosamine levels at diagnosis. 92 Figure 13. Distribution of serum C-peptide-to-fructosamine ratio at diagnosis. 93 Figure 14. Distribution of serum SHBG level at diagnosis. 94 Figure 15. Survival curves by WHR and stage at diagnosis. 95 Figure 16. Survival curves by insulin level and stage at diagnosis 96 Figure 17. Biological pathways. 97 XI A C K N O W L E D G M E N T S The author wishes to acknowledge the generous time commitment, caring guidance and insightful comments of Dr. Sam Sheps, without whom this dissertation would not have been possible; the cheerful support of Dr. Greg Hislop; the long, idea-generating talks with Mr. Richard Gallagher; the helpful statistical advice of Dr. Andy Coldman; the considerable expertise and valued feedback of Dr. John D. Potter; the preview of early drafts by Dr. John Spinelli; and the gracious permission to use the data given by the original investigators, as well as their support throughout, Dr. Bruce Dunn, Dr. Ivo Olivotto, Dr. Charmaine Kim-Sing and Mrs. Cheri Van Patten; and most importantly, the participation of the women in the study, whose struggle with breast cancer inspired this work. This study was made possible by the generous support of the Canadian Breast Cancer Foundation / B.C. and Yukon Chapter, and the Lions Gate Healthcare Research Foundation. xu G L O S S A R Y Effective carbohydrates. Total carbohydrates minus fiber, or the portion of carbohydrate consumption that has an effect on blood glucose. Glycemic Index. A ranking of foods based on the postprandial (following a meal) blood glucose response compared with a reference food, typically either glucose or white bread. Calculated as the integrated area under the curve of blood glucose response plotted over time. Hyperinsulinemia. The state of having excess circulating insulin (normal range, 43-194 pmol/L). Insulin resistance. Loss of normal response to msulin by tissue receptors so glucose uptake is decreased. Postprandial. Following a meal. xui The dragon is the spirit of change ... therefore of life itself ... taking new forms according to its surroundings, yet never seen in its final shape. It is the great mystery itself. Hidden in the caverns of inaccessible mountains, or coiled in the unfathomed depth of the sea, he awaits the time when he slowly arouses himself into activity. He unfolds himself in the storm-cloud, he washes his mane in the darkness of the seething whirlpools. His claws are the fork of the Hghtning ... His voice is heard in the hurricane ... The dragon reveals himself only to vanish. - Kakusu Okakura, The Awakening of Japan xiv C h a p t e r 1 INTRODUCTION Purpose The dragon has come to symbolize breast cancer (1). Each year, over 20,000 Canadian women will face the dragon, and more than 5000 will die from it. Many environmental factors have been investigated with respect to breast cancer risk (2-8). Less work, however, has been done on environmental factors affecting breast-cancer mortality. In particular, it is critical to learn more about the effect of lifestyle factors on mortality because these factors are potentially modifiable. A cluster of such modifiable factors related to hypermsulinemia (high insulin levels) may provide new avenues for clinicians and patients to achieve better outcomes. This dissertation describes the research associated with a cohort study of insulin and related biological and lifestyle factors in women diagnosed with breast cancer. The primary aim was to estimate the effects on breast-cancer mortality of post-diagnosis insulin and C-Peptide levels and lifestyle factors related to hyperinsulinernia, mcluding carbohydrate, fat, alcohol and fiber consumption, glycemic index, physical activity, body mass index (BMI) and waist-to-hip ratio (WHR). Impact of the Project on the Burden of Cancer If breast-cancer mortality is positively related to high uisulin levels, which are subject to modifiable lifestyle factors, there are broad implications for breast-cancer patients even for 1 relatively modest increases in relative risk. The goal of this study is to contribute to evidence-based patient counselling, as well as targeted interventions to reduce insulin resistance. This work may thereby help to improve survival rates for those breast cancer patients with elevated insulin levels. Specific Aims 1. The primary aim was to estimate the relative risk of breast-cancer mortality associated with hyperinsulinemia and associated lifestyle factors. 2. A secondary aim was to describe the effect on mortality of changes made in the hyperinsulinemia-related lifestyle factors during the first 2 years post-diagnosis. Hypotheses 1. Breast-cancer mortality is positively related to insulin and C-peptide levels and related factors in hyperinsulinemia, and women in the highest quartile of insulin or C-peptide are expected to have at least 2-fold greater risk of dying (odds ratio 2.0), compared to women in the lowest quartile. 2. Improvements in lifestyle factors related to hyperinsulinemia during the first 2 years post-diagnosis will be associated with better survival rates, estimated to be similar in magnitude to the main effect. These factors include increased fiber consumption, decreased body weight particularly abdominal, and decreased consumption of high energy nutrients such as saturated fat and refined carbohydrate. 3. The association of insulin and hyperinsulinemia-related factors with breast-cancer mortality will be modified by age and menopausal status, with a stronger association post-menopause, when the effects of endogenous estrogen no longer overshadow more moderate risk factors. 2 Rationale The study rationale is based on preliminary work regarding carbohydrates and colorectal cancer risk by the author (9), on recent work at the University of Toronto with a similar breast-cancer cohort (10), and on peer-reviewed literature about the role of insulin in carcinogenesis, both breast and colorectal. Insulin and C-peptide levels have been shown to be associated with increased risk of breast cancer (11, 12) and recently have been associated with increased breast-cancer mortality (10). If this association can be confirmed in other prospective human studies, it will provide more support for evidence-based patient lifestyle counselling. This work may also suggest opportunities for risk reduction intervention in newly diagnosed breast-cancer patients to modify high-risk endocrine profiles (13, 14). This is timely, as the authors of a randomized trial, the Oslo Diet and Exercise Study, reported that a 1 year diet and exercise intervention significandy lowered insulin levels. The authors concluded that such lifestyle changes may reverse development of the msulin resistance syndrome (15). In addition, the The Canadian Breast Cancer Initiative Workshop on the Primary Prevention of Breast Cancer (16) has called for further prospective and intervention studies of the associations between lifestyle factors and breast cancer. The Following Sections Chapter 2 reviews the literature and summarizes the evidence for an insulin role in breast-cancer risk and mortality. First, the epidemiology of breast cancer is briefly reviewed. Next, insulin's role as a risk factor is evaluated, and finally, biological and lifestyle factors that affect uisuHn resistance and might lower risk of breast cancer are discussed. Chapter 3 contains the methodological details of the study design, data collection and analysis. Chapter 4 presents the study results, and Chapter 5 discusses some of the implications and limitations of the study, 3 and links the results with the existing body of literature. Chapter 6 provides a summary, then looks forward to future research to answer remaining questions. 4 Chapter 2 B A C K G R O U N D OVERVIEW For women, breast cancer is the most common incident cancer and the most common cause of death from cancer worldwide (4). The impact of breast cancer is very large, not only because of its incidence but also because of the average years of life lost (20 years), which is higher for breast cancer than for all cancers combined (16 years) (17). There is a compelling need for research on potentially modifiable factors related to breast cancer risk and mortality. This research is needed to provide women at risk and women diagnosed with breast cancer the information they and their clinicians can use to help reduce their risk and maximize their survival. As a breast cancer risk factor, insulin is a relative newcomer, with limited data available. Studies to date have been consistent in direction and magnitude of effect observed, and they represent a balance of experimental and observational designs but no randomized controlled trial of insulin lowering regimens in breast cancer patients has yet been published. The available data suggest that high insulin levels (hyperinsulinemia), usually a result of insulin resistance and positive energy balance, may be stimulating growth of breast tumours. Further studies are important because hyperinsulinemia is potentially amenable to improvement by changes in diet and physical activity that women can implement themselves, as well as a range of pharmaceutical interventions available to clinicians. 5 This literature review will look at the possible role of insulin in breast cancer, with an emphasis on effects relevant to breast-cancer mortality specifically. Medline (1966-2002) and Cancerlit (1975-2002) were searched combining the Medical Subject Headings "breast neoplasms" and "insulin resistance". No limits were used with respect to language, subjects or type of study. Additional references were obtained by examination of the references from the first set of articles, as well as by personal communications and reviews of unpublished research. The first section of this chapter briefly reviews aspects of breast cancer epidemiology; trends, risk factors, and prognostic factors. The second section looks at insulin and its actions relevant to breast carcinogenesis and prognosis, while the third section considers factors that affect insulin levels and how they might be used in breast-cancer risk reduction strategies. 6 I. EPIDEMIOLOGY OF BREAST CANCER Breast cancer has an inttiguing 5-fold variation in international incidence (6), though modest compared with many other cancer sites. This variation suggests environmental causes, supported by evidence from migrant studies. Such studies show that when people move from areas of low breast-cancer incidence such as China or India to areas of high breast-cancer incidence such as North America or Australia, they reach or approach the risk profile of their adopted country (7, 8). Breast Cancer Incidence and Mortality Trends Canadian and U.S. Trends After a small but steady increase in incidence over three decades, breast cancer incidence in Canada began to level off in 1993 (18), while mortality from breast cancer has declined since 1986, due in part to intensive treatment and earlier detection through screening mammography. In Canada, an estimated 20,700 people will be diagnosed with breast cancer this year, and 5,400 will die from it (18). The magnitude of the problem therefore remains significant, justifying continued study of factors that can further reduce breast cancer incidence and mortality rates. In the United States, while breast-cancer mortality increased slightly in the 1980's, there was a sharp decrease after 1989, with the age-adjusted breast-cancer mortality rate for U.S. white females dropping 6.8% from 1989 through 1993 (19). Although decreasing birth cohort risks for younger women may be pardy responsible, early detection and successful treatment are also probable causes of this decline. 7 International Trends The incidence of breast cancer is increasing generally throughout the world, and particularly in areas where rates have historically been low. Trends in breast-cancer incidence and mortality, however, are not uniform world-wide. In developed countries, where over 50% of breast-cancer incidence occurs (4), breast cancer incidence is levelling off and mortality has begun to decline (20). In developing countries, however, where breast-cancer incidence and mortality were historically low, both have been increasing, which supports migrant studies with respect to the role of environmental causes. The developed or industrialized countries include Canada, the U.S., Australia, and Western European countries among others, though this category varies somewhat from study to study. In an international comparison of cancer patterns in 15 industrialized countries, Hoel et al. (21) looked at the time period from 1969 to 1986, and reported that breast-cancer mortality in older women has been increasing in Europe and East Asia, but declining in pre-menopausal women in the U.S. and in Nordic countries (21). The greatest increases in breast-cancer mortality at young ages were in Eastern Europe. East Asian females were found to have significantly lower mortality rates than females in other regions, but their percent annual change (1.0%) is higher than U.S. females (0.3%) for the period 1969-86, expressed as a percentage of the 1986 rate. Within the framework of E U R O C A R E (31 population-based cancer registries covering a population of 100 million Europeans), Sant et al. analyzed data from 119,139 women diagnosed with breast cancer between 1978 and 1985 in 12 countries and followed for at least 6 years (22). They found the highest survival rates at 5 years in Finland and Switzerland (about 74%), intermediate rates for Italy, France, the Netherlands, Denmark and Germany 8 (about 70%), and lowest survival rates for Spain, the United Kingdom, Estonia and Poland (55-64%). Women aged 40-49 at diagnosis had the best prognosis in all countries. There are several possible explanations for these observations. Part of the decreased mortality in younger women could be explained by advances in treatments such as adjuvant chemotherapy. Mamrhographic screening programs would be unlikely to explain all of the mortality decreases in the younger age groups because screening mammography was not generally conducted for younger women during that time period, though early detection through breast self-examination (BSE) may account for some of the decreased mortality. The increases in breast-cancer mortality for women over age 65 could be partly explained by the shifting of breast cancer deaths to the older age group as a result of improved treatments which increase survival time. A large part of the explanation, however, remains unresolved. Rose et al. (23) found the relationship between breast-cancer mortality and animal fat consumption to be particularly strong in post-menopausal women. They also noted increases in both mortality and fat intake for countries at relatively low breast-cancer risk such as Japan, Greece, Spain and Poland, as well as overall correlations with meat, milk and total calories consumed from animal sources. International comparisons, however, do not measure actual consumption and cannot control for the many other differences between countries that can affect risk of and mortality from cancer. Examination of international trends, however, does offer plausible directions for further experimental and clinical research. 9 Factors Associated with Risk of Breast Cancer Despite considerable research, a large proportion of breast cancer remains unexplained. Of the risk factors identified to date, few offer the potential for intervention to reduce risk to any significant degree. This section will focus on factors associated with the risk of developing breast cancer, while the next section will address factors associated with survival after a diagnosis of breast cancer. Environmental factors play a major role in breast cancer risk (2-4). A large twin study suggests that environmental factors may account for over 70% of breast cancer cases (5) compared with less than 30% for hereditary factors, though some would estimate environmental causes as responsible for more than 90% of breast cancer cases (24). Among dietary factors suspected in breast carcinogenesis are total and saturated fat intake, alcohol and total energy intake (4), while dietary fiber, anti-oxidant vitamins, and vegetables have been associated with decreased risk (25). Other risk factors include age, obesity (post-menopausal) (26), lack of physical activity, family history, mammographic density, and inheritance of specific genes (4). Reproductive risk factors include parity (higher is protective), age at first birth (earlier is protective), age at menarche (later is protective) and age at menopause (earlier is protective) (25, 27), each of which defines an aspect of lifetime estrogen exposure. Genetics The contribution of hereditary factors to the etiology of breast cancer is unclear. Lichtenstein et al. (5) compiled data on 44,788 pairs of twins and reported a statistically significant effect of heritable factors in 27% (95% CI, 4-41) of breast cancer cases. Doll (24) estimates the breast cancer risk explainable by inherited susceptibility with a high penetrance and a neutral environment to be much less, not more than a few percent. Several germline 10 mutations have been investigated in genes suspected to be involved in breast cancer including BRCA1, BRCA2, p53 and Ataxia Telangectasia. BRCA1 mutations may account for 5% of breast cancer cases in women < 40 years old, 2% in women 40-49, and 1% in women 50-70 (28) . Germ-line mutations in p53 are present in the majority of families with Li-Fraumeni syndrome, which predisposes to early-onset of a diverse range of carcinomas and sarcomas (29) . The tumour suppressor gene product p53, often functionally inactivated in breast tumours, is a transcription factor that can bind to D N A sequences in various promotors, and can function as a transcription repressor of growth factor-regulated genes, including the insulin receptor gene (30). Interactions between genetic factors and environmental and lifestyle factors may be key to unlocking some of the unexplained portion of breast-cancer risk. Demographic Factors Age is the single most important factor in breast-cancer risk, but the effect is not necessarily linear, and may be modified by menopause. Both ethnicity and place of residence are important factors as well, suggesting a combination of inherited and environmental risk factors. Education level and other measures of socioeconomic status have been studied with respect to breast-cancer risk and unlike many other cancers, higher socioeconomic status is associated with a higher incidence of breast cancer (31), which is consistent with an hypothesis that the high-energy intake and low-energy output of "the good Hfe" may be associated with higher risk of certain cancers. Hormone Receptor Status ER and PR, the receptors for estrogen and progesterone, are often overexpressed in malignant breast tissue, and it has been suggested that together, they define distinct tumour subtypes with potentially different risk factors and etiologies (32). There may also be an 11 interaction with menopausal status, in that the proportion of receptor-positive women differs by menopausal status, and in one study, significant differences in risk of breast cancer by menopausal status were no longer observed when the data were stratified by ER tumour status (33). Hormone Use Unlike colorectal cancer, past use of oral contraceptives (OC) has not been related to increased risk of breast cancer in most studies (34), however the duration of use before the first pregnancy may be a factor (17), so particular sub-groups of users may be at increased risk from OC use. Post-menopausal hormone use was associated with a 30% increase in breast cancer risk for current users (17), in data from the Nurses' Health Study, though no increase in risk was seen for past users or with increasing duration of use in that study. The Oxford Collaborative Group on Hormonal Factors in Breast Cancer, however, reported increased risk with long-term usage (35) after reanalyzing about 90% of worldwide epidemiological evidence on the relation between risk of breast cancer and use of hormone replacement therapy (HRT). The recent stopping of the Women's Health Initiative clinical trial (36) due to adverse effects including breast cancer outweighing beneficial effects, has underlined the potential risk of post-menopausal estrogen use. Mammographic Density A relatively new, important risk factor for breast cancer is mammographic density. Women with extensive breast density (over 75% of the mammographic area) are 4-6 times more likely to develop breast cancer (37). Mammographic density has also been associated with hyperplasia and atypia (38), histological features known to be related to breast-cancer risk. Mammographic density has been shown to be modifiable by dietary fat reduction (39, 40), 12 which may modify risk. There is some evidence that mammographic density is positively associated with plasma IGF-I levels and negatively associated with plasma IGF-BP3 levels in pre-menopausal women (41), suggesting that IGF-I may be involved in the pathway that relates mammographic density to risk of breast cancer. Reproductive Factors Reproductive factors such as number of children or age at menarche have been hypothesized to influence breast-cancer risk mainly because they are associated with hormonal changes. Age at menarche and age at menopause together define the length of a women's exposure to reproductive hormones, particularly estrogen. Pregnancies, lactation, anovulatory cycles and age at first birth further modify exposure levels to a variety of hormones with powerful and wide-ranging effects. Early work in this area showed increased breast cancer rates among nulliparous women who would likely have a greater lifetime hormone exposure, compared with women who have carried a child to term (42). In addition to nulliparity, late age at first birth, early menarche and late menopause, all factors that increase total hormone exposure, have been associated with increased breast-cancer risk (43). Menopausal status may affect breast-cancer risk, and has been associated with differences in risk factors such as age at first birth, and age at menarche (33), supporting etiological differences in pre- and post-menopausal breast cancer. It is unclear how much of such a difference would be attributable to menopausal status, to age, or to related factors such as ER and PR status (32). Associations of reproductive factors with breast-cancer risk are generally of a modest magnitude; for example, age at menarche (OR, 0.92 per year after age 11), reported by Wu et 13 al. (44) who also found that odds ratios for breast-cancer risk among Asian women living in Los Angeles changed only slightly after adjustment for menstrual and reproductive factors. There are large international differences in some reproductive factors such as average age at menarche which is 12 or 13 in the US and other developed countries, compared with 17 in rural China (43). These differences may be indicative of other interacting factors affecting both age at menarche and breast-cancer risk. Two factors that may fit in this category are energy intake and energy output — that is, energy balance. Energy Balance A combination of intake and output, energy balance may be a significant factor in chronic disease causation for people in developed countries, where energy intake often exceeds output and obesity is reaching epidemic proportions (45). Low levels of physical activity have been associated with increased breast-cancer risk and have been rated as "possibly" increasing risk of breast cancer by the 1997 World Cancer Research Fund/American Insitite for Cancer Research policy document on food, nutrition and the prevention of cancer (4). Physical exercise influences many metabolic and endocrine processes and hormones. Regular physical activity is recommended for cardiovascular and bone health, and increasingly is being associated with reduced risk of cancers, particularly colorectal, breast and prostate. Though not entirely consistent, the evidence suggests a protective role for exercise in the prevention of breast cancer. The effect sizes, however, are modest. For example, McTiernan et al. found a slightly decreased risk of breast cancer in women who exercised more than 1.5 hours per week or engaged in at least some high-intensity physical exercise (OR, 0.7; 95% CI, 0.4-1.1) (46). Gilliland et al. (47) found decreased breast-cancer risk with increased physical activity in their case-control study in both pre- and post-14 menopausal Hispanic women, but only in post-menopausal white women. Luoto et al. (48) found only a small association between activity and breast-cancer risk in a cohort of 30,548 women. It is not yet clear what type of physical activity is best, whether it is effective both pre- and post-menopause, or whether there are specific critical periods of early exposure. On the intake side of energy balance, markers of high energy intake, including greater adult height and high body mass or adult weight gain have been rated as having "convincing" or "probable" evidence respectively for their role in breast-cancer etiology (4). International breast-cancer rates correlate positively with adult height, a surrogate for energy balance during growth (49), and Western children are taller and reach menarche earlier now than during the early part of the 20 th century (17). High energy intake in relation to physical activity accelerates growth and the onset of menarche, which may contribute significantly to increased breast-cancer risk (50). Specific high-energy dietary components are covered separately in subsequent sections. Obesity Obesity, usually the result of a chronic positive energy balance, has been related to risk of breast cancer in post-menopausal women. Null or inverse associations have been reported among pre-menopausal women, however, suggesting complex endocrine-metabolic interactions (51, 52). Elevated waist-to-hip ratio (WHR), representing higher abdominal fat distribution, has been associated with incidence of multiple health outcomes, including breast cancer (53, 54). In a nested case-control study using the New York University Women's Health Study cohort, high WHR was a predictor of breast-cancer risk in pre-menopausal women, especially if overweight 15 (52). Abdominal obesity is also considered to be a marker for insulin resistance and hyperinsulinemia (55, 56), which have been independently associated with risk of developing breast cancer in pfe-menopausal women (11, 12). A Canadian population-based case-control study of 1233 incident breast cancer cases (53) found strong evidence that WHR and lifetime weight gain were post-menopausal breast-cancer ask factors. The Carolina Breast Cancer Study of black women arid white women reported that BMI was inversely related to breast-cancer among pre-menopausal women for whites but not for blacks, and there was no association found among post-menopausal women. Higher WHR, adjusted for BMI, increased risk for all women (57). Dietary Factors In dietary studies, apart from work on energy intake, there has been considerable research conducted to identify specific dietary components that may be related to breast-cancer risk. Of these, dietary fat has received the most attention. Fat. Some of the initial ecologic studies looking at dietary factors and breast cancer reported high correlations of per capita fat consumption and breast-cancer incidence, while animal feeding experiments indicated that a high intake of dietary fat may be an important breast-cancer risk factor (reviewed in Willett 2001). The international comparisons were enticing, but they could not control for potential confounders, most notably total energy intake. Data from rodent models must also be interpreted with caution; first because animals are often given high doses of carcinogens and experience feeding regimens to which humans are rarely exposed, and second, because in those animal studies designed to separate the effect of total energy from that of fat intake, the relationship of fat to breast-cancer risk was weak or non-existent in relation to that of total energy intake. Further analysis of the fat hypothesis using case-control 16 and cohort designs has so far failed to support the international and animal data. In his review of diet and breast cancer (50), Willett reports results for eight large prospective studies of fat and breast cancer, each with at least 200 incident cases. No association was observed with total fat intake. This does not preclude the possibility that fat intake during childhood or adolescence may be important. Willett also notes that consumption of fat as a percentage of intake has been falling in the United States while breast-cancer incidence has been rising. He highlights findings that in some cases, a low fat/high carbohydrate diet could actually increase breast-cancer risk by exacerbating hypermsulinemia. When fat intake is broken down into lipid sub-groups, however, some evidence points to a possible beneficial effect of mono-unsaturated fats such as olive oil, and omega-3 polyunsaturated fatty acids, derived mainly from fish oils, on the hormonal profile of post-menopausal women (14). Further research is needed to determine whether such changes can reduce the risk of developing breast cancer. With dietary fat, then, perhaps the only thing we can say with certainly is that fat intake is correlated with breast-cancer incidence, but there is not yet enough evidence to support a causal relationship. Many have argued that the evidence for dietary fat intake reduction is strong enough to merit definitive clinical studies, including the Women's Health Initiative (WHI) (58) and the Breast Cancer Prevention Trial (Boyd & Hislop, unpublished data), looking at the possible preventive effect of a low-fat dietary intevention in women at high risk of breast cancer. Fat intake, of course, is also highly correlated with a number of other possible causative factors including total energy, carbohydrate and protein intake, and obesity among others that may be independent risk factors. Protein. Protein intake as a percentage of total energy has not been related to breast-cancer risk, and studies on red meat specifically have been largely null as well (50), despite a consistent 17 association with colon cancer. Some evidence for an inverse association with fish protein consumption has been reported, though this may be confounded by differences in fatty acid composition (59). Carbohydrate. Carbohydrate, like fat, is a diverse nutrient group, made up of several subgroups with quite different properties. Just as saturated fat is different structurally and functionally from monounsaturated fat, so too are monosaccharides, or simple sugars, different from pectin, a soluble type of fiber, yet both are called carbohydrates. Except for the fiber sub-component, very little work has been done to date on the relationship between carbohydrate intake and breast-cancer risk. With fiber and colorectal cancer, the prospective data have not supported earlier observational and animal studies with respect to possible protective effects (60, 61). Subgroups of fiber, like fat, may have different effects and these remain to be elucidated. The non-fiber carbohydrate types of interest in this study are the starches and sugars. This digestible portion of total carbohydrate will be referred to as "effective" carbohydrate (eCarb) (9), because it is the portion of total carbohydrate that has an effect on blood glucose. Within this group, carbohydrate foods can be further classified by glycemic index '(GI), where foods are assigned values according to how fast they are digested and absorbed relative to a reference food. GI is measured as the incremental area under the blood glucose curve after consumption of a fixed amount (usually 50 or 100 grams) of a carbohydrate test food, divided by the area under the curve after consumption of the same amount of the reference food, 1 Glycemic index is a ranking of foods based on the postprandial (following a meal) blood glucose response compared with a reference food, typically either glucose or white bread. 18 usually glucose or white bread. A limited amount of recent case-control evidence demonstrates that a low-glycemic index diet may protect against the development of colon cancer (62) and breast cancer (63). The glycemic index concept is, in part, an extension of the fiber hypothesis, which was itself partly derived from previous work on sugar (64, 65). One effect of the fiber portion of carbohydrate foods is to slow the rate of glucose absorption, and lower the GI. The glycemic index, as a research variable, still has several technical limitations (66), including the variability of GI for individual foods, the effects of refining and cooking, the higher GI associated with smaller food particle size, and the effect of mixed meals on measurement accuracy. Sufficient positive findings have accumulated to support the potential importance of dietary GI (67). The Food and Agriculture Organization (FAO) of the United Nations and the World Health Organization (WHO) have endorsed the use of glycemic index to classify carbohydrate foods, and have recommended that GI be used together with other information to guide food choices (FAO/WHO Rome 14-18 Apr 1997). Toward this end, Australia has begun consumer labelling of foods with glycemic index (68). Alcohol. Of the dietary risk factors, alcohol intake has the most consistent supportive evidence (50) and is rated as "probably" increasing breast cancer risk by the WCRF/AICR report (4), but the effect is generally modest. Alcohol intake has also been shown to increase levels of estrogens in the bloodstream (49), suggesting a possible mechanism of action. A recent case-control study (69) found the risk related to high levels of alcohol intake was modified significantly by education, with greater intake and greater risk in the low education group. The risk associated with alcohol intake may also be modified by ER (estrogen receptor) and PR (progesterone receptor) status such that risk is increased primarily for those who are receptor-negative (70, 71). 19 Micro-nutrients. Several anti-oxidant micro-nutrients have been the focus of research including Vitamins A, C, E and the mineral selenium. With the exception of a modest inverse association between Vitamin A intake and breast cancer, the available prospective data do not support a benefit of high intakes of these micro-nutrients for reducing breast cancer risk (4). There is some evidence to support the hypothesis that Vitamin D intake and sunlight exposure may be related to reduced risk of breast cancer (72). The active form of Vitamin D, 1-25-clihydroxyvitamin D(3) acts as a modulator of cell growth and differentiation, including breast-cancer cells (73). Polymorphisms in the Vitamin D receptor have been linked to insulin resistance (74), suggesting one possible mechanism. Vegetables and Fruits. Higher consumption of vegetables may be associated with a reduction of breast-cancer risk by about 20% (49), however, which vegetables specifically are responsible for this reduction remains unclear. Brassica or cruciferous vegetables (broccoli, cabbage, cauliflower), allium vegetables (onions, garlic), dark green leafy vegetables (spinach, romaine lettuce), and high carotene vegetables (carrots) have all been the subject of study. In the 14-year follow up of the Nurses' Health Study, a protective effect was found for total vegetable intake, but only for pre-menopausal women (50). Little relation was seen with fruit consumption. A variety of phytoestrogens, potentially anti-carcinogenic compounds found in vegetables and fruits, have attracted research attention, including daidzen and genistein from soy, and glucosinolates from cruciferous vegetables (75). These compounds, as well as the fiber content and energy-dilute nature of vegetables and fruits all may contribute independently to any observed protective effect. High fiber foods like fruits and vegetables are filling, and may also decrease intake of high energy fats and carbohydrates by substitution. 20 Cigarette Smoking In the Nurses' Health Study no relationship was found between cigarette smoking and the risk of either pre-menopausal or post-menopausal breast cancer (17). There are no data at this time to support an earlier proposed protective effect, based on a suggested association between cigarette smoking and reduced age at menopause (76). Cigarette smoking, however, may be associated with a lower BMI in some people, which could also independently affect risk. Cigarette smokers have also been found to be relatively glucose intolerant, hyperinsulinemic and dyslipidemic (77), all potential independent risk factors. Hyperinsulinemia Metabolic and reproductive hormones interact and have profound effects on each other. Many factors, including diet, activity and obesity, contribute to the development of insulin resistance and hypermsulinemia, which may in turn be an important breast-cancer risk factor. Hyperihsulinemia is covered in detail in Section II, Insulin as a Risk Factor. Factors Associated with Mortality from Breast Cancer Clinical Factors Most known factors that affect breast-cancer mortality are clinical factors, such as tumour stage (size and spread), tumour grade (differentiation), histology (ductal, lobular), treatments received (local and systemic), comorbidity and hormone receptor status (ER and PR) (78, 79). Regrettably, these factors are not modifiable. 21 Modifiable Lifestyle Factors Epidemiological studies have identified some modifiable factors associated with increased breast-cancer mortality including high levels of energy and fat intake, and obesity, especially abdominal (80-84). It is interesting to note that these factors all have in common their relationship with insulin resistance. It is not yet clear whether changes in these factors would result in reduced mortality from breast cancer. Obesity In a comprehensive review of current evidence, women with breast cancer who are overweight were found to be at greater risk of recurrence and death compared with lighter women (85). In a combined case-control and follow-up study in Denmark, Ewertz et al. (86) examined survival of breast-cancer patients in relation to risk factors for developing breast cancer, to determine the prognostic significance, if any, of known risk factors. Among the 2445 patients, body weight or BMI was the only consistent relationship with survival. Lower survival was associated with obesity only in women with early-stage disease. The authors also reported that the shape of the BMI hazard function was quadratic rather than linear, suggesting that low as well as high BMI is associated with an unfavourable outcome. Other studies have also suggested a quadratic relationship between BMI and breast-cancer outcome (10). In a U.S. cohort study, Hebert et al. (82) found the relationship between body weight and death among early-stage breast cancer patients to be strongest in pre-menopausal women. They report that relative body weight increased risk of death by 12% per kg/m 2 (RR, 1.12; 95% CI, 1.03-1.22). Similarly, in a Canadian cohort study of 1121 breast cancer patients, there was a significant trend to worsening survival with heavier weight at diagnosis (87). Zhang et al. 22 (88) found that, adjusted for age, women in the highest tertile of BMI had 1.9-fold higher risk of dying (95% CI, 1.0-3.7) after breast cancer than those in the lowest tertile. Android body fat distribution was a statistically significant (p < 0.0001) prognostic indicator for breast-cancer mortality in a prospective study by Kumar et al. (89). Folsom et al reported a positive association of waist-to-hip ratio with breast-cancer mortality in the Iowa Women's Health Study cohort (90), though the authors note that this finding is limited by small numbers. A later report (54) confirmed a dose-response relationship of WHR and age-adjusted all-cancer mortality (RR, 1.3; 95% CI, 1.1- 1.6 comparing fifth to first quintile), but this relationship was attenuated and lost significance after adjustment for other risk factors. In addition, an interaction with family history has been observed in the Iowa Women's Health Study cohort (91), such that a high WHR was associated with increased risk of breast cancer in the presence of a family history of breast'cancer (RR, 2.10; 95% CI, 1.43-3.09), but not in the absence of a family history. A recent report by Sellers et al. (92) notes that the interaction appears to reflect risk primarily for PR-negative tumours, underlining the complex relationship of genetics, tumour subtype and lifestyle. Energy Balance High caloric intake without a corresponding level of physical activity (positive energy balance) is a key factor in insulin resistance (93, 94), and a recent intervention study has suggested that changing energy balance can improve insulin resistance (15). Energy balance is also a probable factor in breast-cancer mortality. Saxe et al. (84) found total energy intake to be associated with increased risk of death from breast cancer (HR, 1.58 per lOOOkCal/day; 95% CI, 1.03-2.43) in a cohort of U.S. breast-cancer patients. Although their research focus was fat intake rather than total energy, 23 Zhang et al's study (88) also supported energy intake as a contributor to poorer breast-cancer prognosis. Fat Intake In a study of breast-cancer mortality, Zhang et al. (88) examined 698 post-menopausal breast-cancer patients from the Iowa Women's Health Study Cohort, and reported a statistically significant risk of death after breast cancer greater than 2.0 for the highest tertiles of total fat, saturated fat and monounsaturated fat intake. Saxe et al. (84) reported a relationship between poly-unsaturated fat intake and breast cancer mortality (HR, 1.84 per 10 gram/day; 95% CI, 1.09-3.13). In a study of breast cancer survival among Caucasians and Japanese women in Hawaii, Nomura et al. (80) reported that Caucasian patients with a high fat intake had an RR of 3.2 (95%CI, 1.2-8.6), compared to those with a low fat intake. In Japanese patients, however, fat intake was not related to survival (RR, 0.66; 95% CI, 0.25-1.76). In a study of dietary habits and breast-cancer outcomes in patients who have undergone treatment, Holm et al (95) reported that women with ER-rich tumours who had treatment failure during follow-up had higher intakes of total and saturated fat than those who did not have treatment failure (OR 1.08 for each 1% increment in total energy from fat. No association between fat intake and treatment failure was observed for women with ER-poor cancers. Not all studies have found fat intake to be related to breast cancer survival. Holmes et al (96) examined dietary factors uicluding fat in the Nurses' Health Study Cohort, and reported no apparent association between fat intake and mortality. Protein Intake Higher protein consumption, mainly from poultry, fish and dairy sources, has been associated with a better prognosis in a large study of diet before and after a breast-cancer diagnosis (96). The relative risk of mortality was 0.65 (95% CI, 0.47-0.88), comparing highest 24 to lowest quintile of protein intake. No association was observed between red meat and mortality. II. INSULIN AS A RISK FACTOR The Biological Role of Insulin and Insulin Resistance Insulin can be considered the master metabolic hormone which, along with its counterpart glucagon, directs metabolic, reproductive and growth activities in response to food availability and energy demand. The human body utilizes a multi-level control system, with controls on the secretion of insulin as well as the sensitivity to insulin in different tissue types. Although perhaps best known for its blood glucose-lowering function, insulin is primarily a growth hormone and its effect on blood glucose is just one of many diverse activities related to growth. Insulin as a Growth Factor The primary role of insulin is to stimulate growth in response to energy intake. A kind of transducer, insulin responds to environmental signals of rich food sources with actions such as cell construction for energy storage, or increased availability of growth and reproductive hormones. Without the ability to store excess glucose when it is available, and to switch to other energy sources when it is not, humans would be forced to eat continuously. Fat stores are controlled by insulin and glucagon secreted by the liver in response to the changing context of energy intake and output demands that an individual is experiencing (97). Insulin has direct effects on growth, particularly with respect to cell construction. Insulin stimulates the production of L D L cholesterol, while decreasing H D L cholesterol concentration (97). Insulin facilitates lipid accumulation by expressing lipoprotein lipase (98). 25 Indirectly, insulin also affects growth by down-regulating IGF binding proteins (99), thereby increasing available IGF-I to act at the tissue level and stimulate growth. Insulin and Reproductive Hormones Besides growth, signals of a rich food source also affect reproductive hormones, both in terms of hormone production and bio-availability. Insulin stimulates the production of estrogens and also androgens such as testosterone. Similar to its action with IGF-I binding proteins, insulin can down-regulate sex hormone binding globulin (SHBG), the binding globulin that controls availability of both estrogen and testosterone (100). Insulin signaling may be enhanced by estradiol at the level of IRS-1 (insulin receptor substrate-1) gene expression (101). It is plausible that insulin's interactions with sex steroid hormones may be involved with the observed correlation of early menarche and high energy childhood diet (102). The Normal Role of Insulin Resistance Differential tissue sensitivity to insulin provides a way for the human body to fine-tune the effects of energy signals by directing tissues to use or not use specific fuels, depending on the situation. Not all tissues can use alternate fuels. Obligate glucose users include brain, mammary and placental tissues. Muscle tissue, however, is able to use glucose, but can also switch to lower-energy, "dirtier" fuels such as amino acids or fatty acids as needed, under insulin's control. This option provides adaptability to changing circumstances which greatly increases survival advantage. Tissue sensitivity to insulin is affected by dietary behaviours (15), as well as by non-dietary factors such as physical activity, age and genetic predisposition (103). In most young active 26 people, very small amounts of insulin can control large amounts of dietary glucose satisfactorily without adverse effects. There are groups of people, however, with particularly high rates of insulin resistance, such as the Pima Indians or the Australian aborigines, perhaps due in part to their late introduction to highly refined, low-fiber foods (104). Insulin resistance is also common in pregnancy, preferentially sending glucose to the baby via the placenta, and during puberty, when significant growth must occur (105). A.n Evolutionary Perspective Judging by its estimated prevalence of 35% (105), insulin resistance was probably an adaptive trait until fairly recently in our evolutionary past. Since the introduction of agriculture, however, carbohydrates and indeed all foods have become available in large quantities year-round, while the last 200 years have brought dramatic increases in consumption of highly-refined, easily-obtained "junk" carbohydrates, foods that deliver quickly-absorbed energy without accompanying fiber, vitamins or minerals (106). During most of human evolution, survival and reproductive success were improved by taking a thrifty approach and diverting scarce high-energy fuel to the high priority systems that are critical to species survival (107, 108). In today's environment, however, the thrifty approach becomes maladaptive because there is an excess, not a scarcity of glucose, and more insulin is required on a regular basis to compensate and keep glucose below toxic levels. This compensation happens automatically and without symptoms, but when chronic, can have damaging consequences (109). Problems of Excess Insulin Although insulin is critical to survival, more is not better. As with most physiological compounds, there are tightly regulated bounds. Levels outside the normal range, high or low, 27 present problems. In the case of insulin, those problems can be very serious and wide-ranging, and are the direct result of msulin's normal functions carried to extremes. For example, to reduce blood glucose concentrations, insulin can act at the kidney to stimulate sodium and water retention (97) and thereby dilute the excess glucose in the bloodstream, but water retention also has the effect of increasing blood pressure. Though an acceptable way to deal with an acute glucose excess, it will be a problem if it becomes a chronic adaptation. The adverse effects of excess insulin have been described by Gerald Reaven (110) and others as Syndrome X or the Metabolic Syndrome, and are now known collectively as the Insulin Resistance Syndrome. The syndrome includes hypertension, atherosclerosis, hypermsulinemia, impaired glucose tolerance (IGT), increased plasma triglyeride concentrations, and decreased H D L cholesterol concentrations. Insulin Resistance Syndrome is suspected to be involved in the etiology of non-insulin-dependent diabetes mellitus (NIDDM) and coronary artery disease, independent of obesity (111). Given insulin's ability to stimulate growth and reproductive hormones, Insulin Resistance Syndrome is also suspected of involvement in the etiology of some hormonally-sensitive cancers including breast, colorectal and pancreatic cancer among others (112, 113). Insulin Resistance and Cancer A recent unifying hypothesis suggests that obesity, physical inactivity, alcohol and the consumption of a typical Western diet are all associated with the development of insulin resistance and hyperinsulinemia, and that hyperinsulinemia may be stimulating growth of tumours (11, 12, 114-117). Insulin is a powerful anabolic (growth) hormone and a risk factor for several major diseases, including cancer, particularly colorectal and breast (11, 12, 110, 118). Hyperinsulinemia may predispose to cancer at a number of, but not all sites. In a review 28 of evidence for a link between hypermsulinernia, diabetes and cancer, Moore et al. (119) present data for an association at several cancer sites — colon, rectum, liver, gall bladder, pancreas, kidney, urinary tract, non-melanoma skin, breast, prostate and endometrium. Several lines of research support a role for insulin in carcinogenesis. Experimental Data Insulin is an important growth factor for human epithelial cells, which have both insulin and IGF-I (insulin-like growth factor I) receptors (118). In contrast to its blood glucose regulation role, insulin may act through IGF-I receptors in its mitogenic effect on carcinoma cells in vitro (120, 121), but the binding is weak. Insulin may also increase available IGF-I by down-regulation of IGF-I binding proteins IGF-BP1 and perhaps IGF-BP2 (99). In animal studies, inhibition of msuHn secretion or action has been shown to decrease growth of mammary tumours (122). A recent review related to biologic interactions among insulin, IGF-I and IGF-BP's points out that insulin may also increase the IGF-I/IGF-BP3 ratio by increasing hepatic growth hormone sensitivity (123). Human Observational Data Support for insulin's role also comes from human prospective studies of subjects with N I D D M (an insulin-resistant state) who have an increased risk of colorectal cancer (124, 125), as well as studies linking N I D D M and breast cancer (119). The link between N I D D M and cancer may be due to the long period of hyperinsurinemia which precedes clinical N I D D M promoting tumour development. N I D D M and cancer also share a number of risk factors, including high-energy diet, especially saturated fats, physical inactivity, and obesity. Shared risk factors, however, do not mean the two diseases would necessarily be found together in a given individual. 29 Insulin Resistance and Breast Cancer Risk This section summarizes the epidemiologic evidence for an association between insulin and breast-cancer risk (breast-cancer mortality studies are discussed separately in the next section). Both insulin levels and C-peptide-to-glucose or C-peptide-to-fructosamine ratio (measures of insulin sensitivity) have been shown to be higher in breast-cancer cases than in controls (11, 12), particularly among pre-menopausal women (126). In 1992, Bruning et al. (12) published results of their case-control study of 223 women with stage I or II breast cancer and 441 age-matched population controls. They also had a second "other cancer" control group of 100 early stage melanoma, lymphoma or cervical cancer patients. The researchers reported an odds ratio, adjusted for age, family history, BMI and WHR, of 2.9 (95% CI, 1.1-5.1) for the highest quintile of serum C-peptide compared with the lowest quintile. Serum C-peptide levels were also found to be positively related to BMI and WHR, markers of msulin resistance. Also significandy higher in breast-cancer cases than controls were C-peptide-to-glucose or C-peptide-to-fructosamine ratios, indicating insulin resistance, independent of general adiposity or body fat distribution. In a 1995 publication, Bruning et al. (127) noted that pre-menopausal breast cancer cases showed elevated IGF-I levels and decreased IGF-BP3 levels. The relative risk was 7.3 (95% CI, 1.6-32.1) for the highest quintile of IGF-I/IGF-BP3 ratio compared with the lowest quintile. These findings are consistent with hyperinsulinemia as a risk factor because insulin has a profound effect on the availability of IGF-I. A Canadian case-control study by DelGuidice et al. (11) provided some confirmation of Bmning's early results. After adjustment for age and weight, elevated insulin levels were significantly associated with breast cancer in 99 pre-menopausal node-negative women. The 30 control group was made up of 99 age-matched women with biopsied non-proliferative breast disease. The OR for women in the highest versus the lowest quintile of insulin was 2.8 (95% CI, 1.2-6.5), independent of diet and other known breast-cancer risk factors. One limitation of case-control studies is the inability to determine which came first, the high insulin or the disease. One prospective study has examined the association of insulin with breast-cancer risk and found no significant association with fasting serum insulin level or WHR, but did find the incidence of breast cancer was 60% higher among diabetic women (128). More prospective studies are needed to determine the possible effects of the disease on the balance of metabolic hormones, including disease effects on weight, diet and activity, as well as effects of disease-related medication and stress. Finally, an Italian cohort study by Muti et al. (126) looked at markers of insulin resistance such as abdominal adiposity and sex steroid hormone activity in relation to breast-cancer risk. They noted the different relationships these risk factors have with breast cancer in pre- and post-menopausal women, and stressed the role of hormones and metabolic alterations in breast cancer etiology. In a nested case-control design, 144 incident breast-cancer cases were identified from the inception cohort of 10,786 women who were cancer-free at the beginning of the study. With an average of 5.5 years of follow-up, each incident breast cancer case was matched with 4 controls from the same cohort, who did not develop breast cancer. WHR was found to be associated with breast cancer in pre-menopausal women, with an age and BMI adjusted RR of 2.2 (95% CI, 1.0-4.7). When stratified by BMI, the effect was confined to women with BMI less than 24 (RR for highest WHR tertile, 3.4; 95% CI, 1.2-9.5). Although, to date, only a small number of studies have addressed hyperinsulinemic insulin resistance as a factor in breast-cancer etiology, and the strength of the association is relatively 31 modest, 2 to 3-fold, the risk is clinically very relevant in view of the numbers of women potentially affected. Thus insulin, a known growth factor, is a plausible agent in breast carcinogenesis, particularly at chronically elevated levels. Evidence for an Insulin Role in Breast Cancer Mortality-High fasting insulin levels have been associated with distant recurrence and death in a Canadian cohort of 512 breast cancer patients analysed by Goodwin and colleagues (10). Adjusted hazard ratios and 95% CI's for those in the highest versus the lowest insulin quartile were 2.1 (95% CI, 1.2-3.6) and 3.3 (95% CI, 1.5-7.0) for distant recurrence and death, respectively. There was some evidence, as well, to suggest that the relationship of insulin with breast-cancer mortality is non-linear. Although these data are statistically significant, plausible, prospective and consistent with expectations, there are few if any other studies to provide confirmation at this time. Insulin levels may be clinically important in identifying those patients expected to have poor outcomes, in order to target more aggressive treatments. There is a compelling need for more research to provide evidence-based care for patients with high insulin levels who may be able to lower their risk by lowering their insulin. Possible Mechanisms A key growth hormone, insulin is a biologically plausible cancer promoter with several possible modes of action. Current research has not suggested a direct role for insulin in cancer initiation. Potential mechanisms for promotion include the influence of insulin levels on circulatory steroid hormone levels and their bioavailability, as well as on the circulating peptide growth hormone IGF-I and its tissue availability. Insulin can also act directly to cause growth, either through its own receptor or through the IGF-I receptor. 32 IGF-I Insulin can control IGF-I availability at the tissue level by down-regulating its binding proteins (99). Insulin can also cross-react with IGF-I receptors. As the name implies, IGF-I receptors have some affinity for insulin which is further enhanced by increased insulin concentration. Insulin's mitogenic effect is suspected to occur via the IGF-I receptor (121). Insulin may amplify its effect by down-regulating IGF-BP concentrations and therefore increasing availability of biologically active IGF-I in the target organs such as the breast and ovary (99, 129). Tissue IGF-I bioactivity has been correlated with turnover rate of epithelial cells (130, 131). Even a small increase in turnover rate over time would produce many more cells at risk of somatic cell mutations. IGFs are also thought to influence cancer risk through anti-apoptotic properties, whereby partially-transformed cells can survive longer in hosts with relatively high IGF bioactivity than in hosts with relatively low IGF bioactivity (130). Dietary restriction was found by Dunn et al. (132) to reduce IGF-I levels, and to modulate apoptosis, cell proliferation and tumour progression in mice, an effect that might be mediated in part by reduced insulin levels. Sex Steroid Hormones Similar to its actions with IGF-I, insulin can control the production and the tissue availability of the sex steroid hormones mcluding estrogens and androgens. Insulin is a known factor in steroidogenesis and high levels of insulin potentiate the luteinizing hormone-stimulated production of ovarian androgens, particularly testosterone. Insulin has also been found to inhibit production of sex hormone binding globulin (SHBG) (100) which would result in higher plasma concentrations of available estrogen and testosterone. It has been 33 proposed that insulin resistance can result in an hormonal shift, especially if experienced during adolescence, and may be related to a hyper-androgenic profile, anovulation and PCOS (polycystic ovary syndrome) (13, 114), that could increase risk of hormone-sensitive cancers including breast cancer. Estrogen may have effects on insulin signaling, possibly by increasing IRS-1 expression through direct regulation of the IRS-1 promoter in estrogen receptor-positive cells (101). This regulation results in a greater response to insulin and insulin-like growth factors in the presence of estradiol (133). Conversely, growth factors such as insulin and IGF's can stimulate the estrogen receptor (ER) alpha subunit in breast-cancer cells (134), increasing the cancer cell's growth response to estrogen exposure. Direct Action Normally, insulin receptor levels in epithelial cells are low and they respond only weakly to insulin. Insulin receptor content has been found to be 6-fold higher in breast malignancies compared to normal breast tissue (135), correlated with tumour size and grade. Webster et al. (30) reported that p53 inactivation up-regulates insulin receptor gene expression which can lead to insulin-dependent transformation. The shared use of IRS-proteins by multiple receptors (e.g., IGF-I, interleukin-4) may soon reveal connections with other hormones, cytokines and growth factors not yet recognized (136). 34 III. LOWERING T H E RISK Biological Factors Affecting Insulin Resistance Over 65 years ago, Himsworth (137) differentiated insulin-sensitive and insulin-insensitive types of diabetes, and suggested that some types of human disease could be secondary to a defect in insulin action. Resistance to insulin-stimulated glucose uptake is now recognized as a common phenomenon, present in the majority of those with impaired glucose tolerance (IGT) or NIDDM. Insulin resistance is also present in an estimated 25% of non-obese people with normal glucose tolerance (110). There are a number of biological factors influencing insulin resistance, though most are not modifiable. They may, however, interact with modifiable environmental factors in important ways. Genetic Factors The estimated prevalence of insulin resistance is 35% (105), and this suggests conservation of a trait important to survival, and a potential genetic component. Clausen et al. (1995) reported on polymorphisms in codons for the protein insulin receptor substrate-1 (IRS-1) associated with changes in insulin sensitivity in normal adults. The IRS-1 gene variants also appeared to interact with obesity, resulting in greater reductions in insulin sensitivity, up to 50%, for obese subjects. Some indirect evidence for a genetic component in insulin resistance comes from observed ethnic differences in kisulin sensitivity and the related disorders IGT and NIDDM. Examples include Pima Indians, African-American individuals, and Australian aboriginal peoples (138, 35 139). A relatively late introduction to a Western diet is common to most of these groups, and may indicate insufficient time for metabolic selection to occur. Age Within each individual's range of response, age has the effect of decreasing insulin sensitivity overall. It is not clear if this is a necessary part of aging, or if other aspects of aging such as decreased physical activity may be the cause. Menopause This age-related event has a significant impact on the levels of numerous reproductive hormones, and menopause may affect insulin sensitivity as well. Potischman et al. (140) reported observing a reversal of the relationship between body mass index and endogenous estrogen concentrations with menopause. As BMI increased, total estradiol decreased in pre-menopausal women, but increased in post-menopausal women. A part of this effect may be explained by anovulation in pre-menopausal women and estrogen production in adipose tissue in post-menopausal women. Although there is no evidence of a direct effect of menopause on insiilih sensitivity, the common problem of weight gain after menopause, especially abdominal gain, may indicate increased insulin resistance. Pregnancy During pregnancy, insulin resistance increases (105), and some women develop a temporary condition called "gestational diabetes". This response is biologically plausible because the placenta requires glucose, and considerable growth activity is required. The mother's muscle tissue switches to other fuels and the glucose is used for fetal growth or to increase the mother's energy stores as fat. There is some evidence that intrauterine under-nutrition may be 36 associated with subsequent decreased insulin sensitivity in the offspring, with possible long-term metabolic consequences (Hernandez M - Amer Diabetes Assoc Jun 2000). Puberty Insulin resistance is common around the time of puberty, when substantial growth is required. The high-energy, low-fiber Western diet is linked to earlier menarche and to hyperinsulinemia, especially when associated with inadequate physical activity (27). Puberty is also marked by reduced SHBG and IGF-BP1, and a reduction in msulin sensitivity that correlates with the onset of pulsatile G H secretion (114). It is possible that the development of breast cancer may be determined primarily by endocrine dysregulation during the pre- or peri-pubertal years. Nervous System Input In addition to environmental cues like food intake, the secretion of insulin can be stimulated by the brain, through sympathetic innervation of the pancreas. An example is the secretion of insulin in anticipation of food (141, 142). It is possible that the primary event is hyperinsulinemia rather than insulin resistance, and hyperinsulinemia could result from an underlying central nervous system disorder of chronic hypothalamic arousal (143). Lifestyle Factors Affecting Insulin Resistance Many North Americans are overweight, physically inactive, and eat excessive amounts of both fats and high glycemic index carbohydrates, all of which are determinants of insulin resistance and hyperinsulinemia (118, 144-146). A diet high in refined carbohydrates causes rapid intestinal absorption of glucose and postprandial (following a meal) hyperinsulinemia. 37 Chronic high glycemic load (glycemic index multiplied by carbohydrate quantity consumed), in the context of pre-existing msulin resistance can produce particularly high insulin levels because the muscle tissue is not taking up the extra glucose and more insulin must be produced to compensate. The body does this initially without symptoms or signs, but there is a price associated with maintaining high insulin levels over the long term. A World Health Organization (WHO) study reports that obesity and lack of exercise contribute to between one-fourth and one-third of all cancers of the colon, breast, kidney and digestive tract (45). The WHO panel concluded that adiposity and inactivity appear to be the most important avoidable causes of post-menopausal breast cancer. Weight gain and physical inactivity are also two of the main environmental determinants of insulin resistance. Although there may be a significant genetic contribution to an individual's range of response to insulin, several modifiable lifestyle factors can have profound effects on insulin sensitivity. A few key factors such as diet, physical activity and weight have been the subject of many studies, while others such as stress-related increases in Cortisol levels are just beginning to attract interest. It makes biological sense for a function like insulin sensitivity to be programmable by environmental cues. Insulin signals a rich food source, and the physiological response to that signal must be conditioned by the individual's context, or setting in the real world. For example, in the context of a very low energy demand as with a sedentary modern lifestyle, msulin resistance results in storage of excess energy as fat. In the context of high energy demand, however, increased insulin sensitivity ensures the tissues take up and use the glucose rather than store it. If elevated insulin levels are associated with increased risk of breast cancer, their responsiveness to environmental changes will be key to possible strategies 38 for risk reduction. Such strategies may involve dietary changes, increases in physical activity, weight reduction, insulin sensitizers, or some combination. Dietary Fat The evidence for an effect of fat consumption on insulin resistance differs for mono-unsaturated, poly-unsaturated and saturated fat subgroups. Intake of omega-6 polyunsaturated fatty acids (PUFA) (138), especially a high omega-6/omega-3 ratio of intake (147), has been associated with msulin resistance and hypermsulinemia (148). Yam et al. found this in Israeli Jews (149), perhaps helping to explain the paradox of high disease rates and relatively low total fat but high PUFA intake in this group. Dietary omega-6 PUFA's are derived mainly from certain vegetable oils such as soybean oil. Overall, the data suggest that a high-fat diet reduces insulin sensitivity, but the pattern of consumption of fatty acids may be a more significant factor than total quantity of fat consumed (150) and increasingly, focus is being placed on the type of fat and the type of carbohydrate consumed rather than only the total amount (50). Derived mainly from seafood, omega-3 polyunsaturated fatty acids improve msuHn sensitivity (Lenhard J - Amer Diabetes Assoc Jun 2000, Storken 1996), and can inhibit occurrence of mammary tumours in animals, but epidemiologic studies to date have not found reduced risk of breast cancer with higher intake of omega-3 fatty acids. Fiigh mono-unsaturated fat/low carbohydrate diets have been found to improve insulin sensitivity in N I D D M subjects (145, 151). In most animal studies, high intake of omega-6 fatty acids increases both insulin resistance (147) and occurrence of mammary tumours (50). Incorporation of more saturated fatty acids into muscle membrane phospholipid is associated with impairment of insulin action (147). The literature suggests that the type of fat consumed 39 is critical, and the effect on insulin resistance may come about via changes in membrane lipid composition (150), affecting membrane fluidity and activity of the insulin receptors. Dietary Carbohydrate Some evidence suggests that a high carbohydrate diet increases glucose intolerance and insulin levels in normal as well as N I D D M subjects (144). The effect appears strongest for high levels of sucrose and fructose consumption (146, 152). This contradicts the currently-advocated high carbohydrate / low fat diet for N I D D M patients. Seval et al. (146) studied the relationship of hyperinsulinemia to dietary intake in South Asian and European men residing in London, in an attempt to explain the increased mortality and morbidity from coronary heart disease among migrants of South Asian origin (Indian, Pakistani and Bangladeshi). The results suggested that a high intake of carbohydrate, especially sucrose, may worsen metabolic disturbances associated with insulin resistance. Carbohydrate foods can be broken down into subgroups like fat, and these subgroups have very different effects on glucose and insulin levels. Fiber, a subgroup of dietary carbohydrate, is thought to improve insulin levels by slowing the release of glucose during digestion, thereby reducing the need for insulin (153), while the subgroup of non-fiber carbohydrate or eCarb increases the need for insulin. A high-fiber diet also improves satiety and may displace other energy-dense foods from the diet. Fiber can be divided further, into a number of subgroups, each of which has different properties and physiological effects. Pectins, for example, are water-soluble gel-like substances that can coat the small intestine and slow glucose absorption (153), while an insoluble fiber like cellulose might simply displace excess glucose from the diet as a result of its bulky, filling properties. 40 Stress Insulin resistance may be affected by stress, both physical and psychological. Stress-triggered hormones such as Cortisol and epinephrine increase arterial concentrations of glucose and also increase output of glucose by the liver (154, 155). Acute pain has been found to decrease insulin sensitivity (156), suggesting that pain relief and stress reduction may be important to the maintenance of normal glucose metabolism. The hypothesis that insulin resistance can be induced through psychoneuroendocrine pathways was addressed by a population-based Swedish study (157) using salivary Cortisol measurements. They found a relationship between insulin resistance and abnormalities in stress-related Cortisol secretion, pointing to a polymorphism of the glucocorticoid receptor gene as a possible mechanism. This could result in an exaggerated somatic response to both food intake and stress in genetically susceptible individuals (158). Cigarette Smoking Smoking has been found to impair insulin sensitivity in some (109, 159), but not all (160) studies. Possible explanations for inconsistent results include small sample sizes, and different methods of quantifying insulin resistance. A larger study (77) in 152 healthy subjects (76 smokers and 76 nonsmokers) found smokers are relatively glucose intolerant, hyperinsulinemic and dyslipidemic. Plasma glucose and insulin after glucose loading were higher in smokers, who also had higher triglyceride and lower H D L cholesterol concentrations than nonsmokers. Obesity As discussed previously, obesity, WHR and weight gain have all been investigated for their connection with breast cancer. Obesity, especially abdominal obesity is also widely considered 41 a hallmark of insulin resistance, because hyperinsulinemia is known to favour abdominal fat storage (161). In a cross-sectional study, Rock et al. (162) looked at factors influencing weight gain after a diagnosis of breast cancer using data from the Women's Healthy Eating and Living (WHEL) Study. Several of the significant factors are also factors related to hyperinsulinemia. Factors positively associated with weight gain included current energy intake, post-menopausal status, African-American ethnicity and adjuvant chemotherapy. Factors inversely associated with weight gain were pre-diagnosis BMI, age at diagnosis, education level and exercise index score. Physical Activity Increased physical activity is thought to improve insulin sensitivity, but the evidence is not consistently supportive. The Oslo Diet and Exercise Study (15), a randomized trial, looked at the single and joint effects of diet and exercise on insulin resistance. A 1-year 2 X 2 factorial intervention in 219 men and women, it showed that while both diet-alone and diet-and-exercise arms gave significant decreases in insulin resistance, the exercise-alone intervention did not significantly change insulin resistance. Their results also showed that BMI was a strong predictor of insulin resistance. One possible contributing factor was noted by Trichopoulou et al. (163) in their study of physical activity and WHR, a marker of insulin resistance. They reported that physical activity has a strong association with WHR, independent of BMI, only in men which may help explain why physical activity is more effective in reducing disease risk in men than in women. 42 SUMMARY Despite a large proportion of unexplained breast cancer risk, only a few emerging factors show promise while some "known" factors such as dietary fat intake may be weakening as prospective evidence comes in. New factors, especially modifiable ones are urgently needed. This chapter has reviewed the evidence for msulin levels and msulin sensitivity as intermediate factors related to both breast-cancer risk and breast-cancer mortality. Obesity, inactivity and diet each have complex antecedents and effects. A possible relationship to breast cancer by means of effects on insulin sensitivity and glucose metabolism is one way these factors may separately or jointly affect cancer risk and prognosis. Perhaps by virtue of its relationship with other breast-cancer risk factors (obesity, energy balance, fat intake), insulin resistance may offer a unifying causal explanation. A connection with insulin resistance provides a convenient and modifiable intermediate endpoint that can be targeted, changed, and measured. Changes in insulin resistance may also be used to monitor progress in lifestyle modification programs. Measures to improve insulin resistance, then, may offer a general approach to both prevention and better survival. The literature reviewed to date supports diet and exercise interventions to normalize insulin sensitivity, but further research is needed before pharmaceutical interventions, which may have adverse side effects, would be indicated. Insulin level, as determined by a combination of lifestyle and individual sensitivity, is a risk factor that affects all people to some degree. Insulin's function as a master growth hormone suggests several possible modes of action. Most important, insulin can be measured 43 and it can be modified. Although a relatively new area of research, the potential of a modifiable factor responsible for some of the unexplained breast-cancer incidence and mortality is exciting indeed. 44 Chapter 3 M E T H O D S A N D MATERIALS Design Study Design This study tests the hypothesis that elevated insulin, C-peptide and lifestyle correlates of hyperinsulinemia are directly related to breast-cancer mortality. The prospective design provides the strongest observational evidence in support of hypotheses that, to date, have been primarily supported by retrospective evidence. Participants The inception cohort included 603 female breast cancer patients, age range 19 to 75, seen at the Vancouver Cancer Centre of the British Columbia Cancer Agency for consultation between July 1991 and December 1992, an average of 2 months after surgery but before any adjuvant treatment. The Dunn and Hislop study (164) collected data and sera post-diagnosis (Phase I), and again 2 years later (Phase II, n=465). Of 700 patients approached, the original study protocol excluded patients for whom the referral was not for a newly diagnosed tumour, who had had any hormone, ablation or chemotherapy for breast cancer prior to referral (n=7), who were Stage IV (n=9), who did not speak English (n=4), or were over age 75 (n=2). The current study will exclude subjects known to be diabetic at the time of enrolment from analyses of insulin levels, because exogenous insulin and insulin resistonce-modifying therapies, as well as insulin antibodies will affect measured serum levels of insulin. Reasons for non-45 participation included; not feeling well (n=12), not interested (n=24), unable to contact (n=4), and contacted but questionnaire not returned (n=35). With the exception of menopausal status, subjects with missing values were excluded from analysis. For example, there were 17 missing values for waist-to-hip ratio, so the total number available for waist-to-hip ratio analysis was 586. Data on menopausal status, an important stratification variable, were taken from the Cancer Agency Information System (CAIS), but values were missing for 50 (8%) cohort members. Menopausal status was imputed for these patients using as the cut-off the median (and modal) age at menopause (age 50) for the cohort. Recruitment Figure 1 describes the relationship of Phase I, Phase II and the current study. Consecutive eligible patients presenting for initial assessment at the Vancouver Cancer Centre of the BCCA during the study period of July 1991 to December 1992 were invited to participate in a study of dietary and psychosocial influences thought to be related to breast-cancer outcomes. Patients were contacted by telephone to answer any questions and confirm their willingness to participate. A single non-fasting blood sample was obtained at the time of their second Agency visit. Time of last meal was recorded at that time. Potential participants were sent a self-administered Phase I questionnaire (Appendix A) by mail, with a followup phone call approximately two weeks later to encourage response and answer questions. A response rate of 87% was achieved. For Phase II, patients attending the VCC were identified using a computerized scheduling system for booking appointments. Patients were contacted by letter, explaining the current status of the project, and inviting their participation in Phase II. Of the 603 participants from Phase I, 48 had died of breast cancer, 8 had died of other causes and 138 refused or were lost 46 to follow up, leaving 465 who returned questionnaires. Participating subjects were then sent a Phase II self-administered questionnaire by mail, with a follow up phone call 2 weeks later. A request for a blood sample was attached to the patient's chart, or if the patient was not scheduled for an appointment, an invitation to come to the clinic and give a blood sample was sent by mail. For the current study, no further patient contact was required. Selection for nested case-control For analysis of biologic variables only, a nested case-control design was used to accommodate limited resources. Breast-cancer deaths (n=105) were frequency-matched to survivors from the original cohort, on stage at diagnosis and length of follow-up such that controls had at least as long a period of follow-up as cases. Known diabetics were excluded. 47 PHASE I 1991-1992 Invite patients from Vancouver Cancer Centre 700 Sera - Store —•/ Liq Nitrogen 603 Questipnnaires Store—•( Database T Update Database Chart Review PHASE II 1993-1994 Invite cohort to participate in Phase II Current Study 2002 277 Sera Store Liq Nitrogen 465 Questionnaires—Store—w Database II Lab Analysis for Insulin and related markers Outcome update via linkage, validate by chart review Data Analysis Figure 1. Study Overview. Questionnaire The study questionnaire (Appendix A) gathered information on lifestyle factors including diet, cigarette smoking, alcohol consumption, exercise and social support, and recognized breast cancer risk factors. It included a validated semi-quantitative food frequency questionnaire designed by Dr. Gladys Block (165). This food frequency questionnaire (the Health Habits and History Questionnaire, short interactive form) was selected because it is widely-used, relatively brief, and can be analyzed by standard software (166). The time interval between diagnosis and completion of the questionnaire was an average of 2 months after surgery. The distribution of time intervals between diagnosis and questionnaire return, for both Phase I and Phase II, is shown in Appendix D. The questionnaires used in Phase I and Phase II were similar, but the Phase II questionnaire had additional sections. The first part of the Phase II questionnaire was a repeat of that used in Phase I, with minor modifications to clarify the time interval for which information is being requested. The second part of the questionnaire was new, and asked about changes in diet and lifestyle in the 2 years since diagnosis, as well as sources of health information the women encountered and used during that time interval. Lifestyle Variables Hyperinsulinemia-related lifestyle factors were selected from the questionnaire data, including dietary intake (in grams per day and percent of total caloric intake) of carbohydrate, fat, fiber and alcohol. Standard reference values of glycemic index for individual foods (68) were used to calculate daily average glycemic load, a weighted average of glycemic index multiplied by grams consumed. Categorical physical activity variables for stairs climbed per 49 day, city blocks walked per day, and frequency of different types of exercise were also analyzed, as well as body mass index and waist-to-hip ratio. A calculated variable was created to represent total number of times per year the patient participated in any exercise activity, using the individual data elements for each exercise type (swimming, sports, jogging, exercise and gardening). Body mass index (BMI) was calculated as the ratio of weight in kilograms to height in meters squared. BMI was categorized using tertiles, because the study population had too few members in the lowest and highest categories of the Statistics Canada standards (<18.6 as underweight, 18.6-24.9 as average, 25-29.9 as overweight, and 30+ as obese) (167). Waist-to-hip ratio was calculated as the ratio of waist measurement to hip measurement. The questionnaire did not provide a definition for waist or hip, but the waist was assumed to be measured 2.5 cm above the umbilicus, and the hip was assumed to be measured at the widest part. WHR was categorized into quartiles for analysis, chosen to provide the most detail possible while maintaining acceptable subgroup sizes. Additional covariates were defined as follows: age at diagnosis (continuous); menopausal status (pre or post); daily intake of energy (kj/day), fiber (grams/day), carbohydrate (grams/day), protein (grams/day), total fat (grams/day) and saturated fat (grams/day); employment status (yes or no); marital status (single, married, widowed, divorced); family history of breast cancer in first degree relative (yes if mother, sister or daughter; otherwise no); physical activity (blocks walked per day, flights of stairs climbed per day, times participated in exercise, sports, swimming, jogging and gardening per year); tumour size (0-1 cm, 1.1-2.0 cm, 2.1-5.0 cm, 5.1-9.9 cm); tumour grade (well, moderately or poorly differentiated); nodal status (positive or negative); number of positive nodes; estrogen receptor status (positive or negative); local treatment (lumpectomy, lumpectomy plus radiation therapy, complete mastectomy, complete mastectomy plus radiation therapy, other); systemic treatment 50 (chemotherapy alone, tamoxifen alone, both, none); cigarette smoking (ever/never); age at menarche (continuous); number of live births (continuous); age at first birth (continuous); age at menopause (if applicable); and ethnicity (Caucasian, Asian, East Indian, Black, Other). Data on use of hormone replacement therapy before diagnosis were not available for analysis. Age at menarche was defined as the age at which the patient began menstruation. Menopausal status was classified as "pre" if menstruation had occurred witiiin the previous year, or if the patient had had a hysterectomy (ovaries not removed) and no menopausal symptoms were present; "post" if the last menstruation was more than one year previous to assessment. Age at menopause was defined as the age at which the patient stopped menstruation or the date of hysterectomy if both ovaries were removed. Body size (BMI, WHR) and dietary intake variables were analyzed as continuous variables as well as in categories such as quartiles. biological Variables Four biological variables were examined: insulin, C-peptide, fmctosamine and SHBG. A measure of insulin secretion over approximately the previous three months, C-peptide can serve as a valuable index to insulin secretion (168), though it still shows significant variability in relation to food intake. Fmctosamine was assayed in place of glucose because plasma was not available to do a glucose assay. The concentration of fmctosamine reflects the average of the continuously varying blood glucose concentrations during the previous 1-3 week period, serving as a blood glucose "memory" (169). Fmctosamine is also required to calculate the C-peptide-to-fmctosamine ratio, a measure of msulin sensitivity. SHBG was assayed to address one possible mechanism of msulin action. Additional assays were not possible due to limited blood volumes. With respect to laboratory quality control, the selected laboratory specializes in research and has documented professional quality control standards (170). 51 Outcomes Mortality was chosen as the outcome for this study rather than recurrence because mortality provides a simple, well-documented endpoint that does not require further subgroup analysis as with recurrence (e.g., local, regional and systemic), which the combined sample size and effect size does not allow. Record linkage, used to obtain updated outcome data, was performed in 2001 by data analysts with the Data and Evaluation Department of the BC Cancer Agency. The data were matched on Agency ID (patient identifier), and transferred by secured data format. The outcome data obtained via linkage included vital status, date and cause of death if applicable. BCCA data are updated monthly from Vital Statistics and the Canadian Cancer Registry with national death certificate information, so follow-up included national data A patient was considered to have died of breast cancer if breast cancer was either the primary or secondary cause of death in either Vital Statistics or BCCA records. Sample Si%e and Power Of 700 eligible patients who provided blood samples, 612 (87%) returned valid questionnaires, of which 9 were Stage IV, leaving 603 available for analysis in the current study. For the lab assay data, a smaller sample size (N=289) was necessary, due to limited volumes and resources. Power was calculated assuming a level of significance, or alpha, of 0.05, two-sided. For the questionnaire data, with a sample size of 301 (half of total; comparing highest to lowest quartiles ), and a 17% average breast-cancer mortality rate (105/603), the study has approximately 80% power to detect at least a 2-fold difference in relative risk from the highest to the lowest quartile. Subgroup analysis by menopausal status (approximately 40% pre-menopausal) reduced the power to 58% for a 2-fold difference and requires at least a 2.5-fold 52 difference to achieve better than 80% power. The details of the calculation (171) are presented in Appendix B. For the lab assay data (insulin, C-peptide, fructosamine and SHBG), with a sample size of 145 (comparing highest to lowest quarnles), the study has 67% power to detect at least a 2-fold difference in relative risk from the highest to the lowest quartile. Power is reduced when stratification by stage and menopausal status is carried out, such that there is only 28% power to detect a 2-fold difference. To achieve 80% power or better it would be necessary to have differences greater than 3-fold. Procedures Data Collection and Coding Questionnaires for both Phase I and II were coded by the study coordinator. Missing or unclear information was followed up by phone call at the time. Al l data were collected under conditions which respect patient confidentiality. Subjects were assigned study numbers at the time of enrolment, and the information was stored in a secure computer system within the Cancer Control Research Department of the BC Cancer Research Centre. No names or other personal identifiers appear on the questionnaires or blood samples. A common protocol was used for questionnaire coding and data entry. Responses to questions 1 to 30 were coded and entered into a computer data base. Responses to question 31 (the diet questionnaire developed by Dr. G. Block) were coded and entered into a computer program associated with the questionnaire (Dietsys) (166) to estimate average daily consumption. 53 Collection of Blood Samples For Phase I, in a procedure worked out in consultation with the Clinical Investigations Committee of BCCA, a single non-fasting blood sample was obtained from eligible patients at the time of their second Agency visit (a pre-therapy assessment), using the same venipuncture normally required for the standard clinical workup. Blood was allowed to clot, and the serum was separated, transferred to a cryotube and immediately stored in liquid nitrogen. The sera were subdivided into 4 aliquots of approximately 1 ml each. For Phase II, a request for a blood sample was attached to the patient's chart. This system of requesting a blood sample is routinely used for other research projects. Greater Vancouver area patients who were not scheduled for an appointment were requested to come to the clinic at their convenience to give a blood sample. Patients living outside Greater Vancouver were invited to give a blood sample should they be in Vancouver witiiin the appropriate time period. Lab Protocols All samples were handled by individuals blinded to patient outcomes. Frozen aliquots were removed from the liquid nitrogen, packaged with dry ice in approved 1A transport boxes and sent to Hospitals In-Common Laboratory (HICL) in Toronto for assay of insulin, C-peptide, fmctosamine and SHBG. The HICL laboratory was chosen because it was able to do all the assays, and division of the aliquots for more than one laboratory location was not desirable. Lab assays were done in the sequence shown below to ensure the highest priority assays were completed despite limited volumes. 54 1. C-Peptide was assayed using Irhmulite solid-phase competitive chemHuminescent enzyme immunoassay from Diagnostic Products Corporation (Los Angeles, CA) (168). There is no detectable cross-reactivity with msulin and 13% cross-reactivity with proinsulin. 2. Fmctosamine was assayed using Spectrophotometry (Cobas Integra) from Roche (USA) (169), using a colorimetric test by reaction with nitroblue tetrazolium. 3. Insulin was measured using the Immunolite immunometric assay from Diagnostic Products Corporation (Los Angeles, CA) (172). There is no detectable cross-reactivity with C-peptide and 11.9% cross-reactivity with proinsulin. 4. SHBG was assayed using Immunolite immunometric assay from Diagnostic Products Corporation (Los Angeles, CA) (173). There is no detectable cross-reactivity with estradiol or testosterone. We recognized that the average minimum volume stored per patient, 2 mL, might not be sufficient to complete all assays, but only two patients had incomplete data for that reason. Lxib Assay Pilot Using BCCA internal funds, a small pilot group of blood samples (N=12) randomly selected from the same cohort and liquid nitrogen container was prepared and sent to the HICL laboratory, to work out the logistics of the assays, ensure the sera were still usable, and ensure the assays could reasonably be completed with available volumes. Chart Abstracts Abstracts of clinical charts were undertaken at enrolment to obtain patient and tumour information, including the following: size and stage of tumour at diagnosis, histologic type of 55 tumour, pathological nodal status, estrogen receptor status of tumour, age and menopausal status of the patient, height and weight, and primary treatment information. Stage at diagnosis was coded using t-stage (T), n-stage (N) and m-stage (M) data as follows, where X=unknown: Stage I = TX,1 N0,X, MX,0; Stage II = TX,0,1 N l MX,0 or T l NX,0 MX,0 or T2 NX,0,1 MX,0 or T3 NX,0 MX,0; Stage III = TX,0,1,2 N2 MX,0 or T3 Nl ,2 MX,0 or T4 AnyN MX,0 or AnyT N3 MX,0. Both tumour size and nodal status represent pathological assessment if available (over 90% of cases had pathological assessment data). Additional chart abstraction was done on a 10% random sample to validate survival and recurrence details obtained via linkage with the Cancer Agency Information System for the cohort. Data Analysis Phase I data were analysed for the relationship between 10-year survival and biological, dietary, lifestyle and patient variables. Data from Phase II were analyzed for the relationship between survival and any changes in diet and lifestyle variables over the 2 years post-diagnosis. Breast-cancer and all-cause mortality were examined in relation to the co-variates of interest by the multi-variate regression models, Cox proportional hazards for questionnaire variables which were analyzed for the whole cohort, and conditional logistic regression for the biological variables that were analyzed as a nested case-control. In the latter case, sampling was done by first stratifying breast-cancer deaths by stage at diagnosis, then randomly selecting up to 2 survivors for each breast-cancer death within stage. Survivors were also matched to breast-cancer deaths by length of follow-up. For both proportional hazards and conditional logistic regression analyses, univariate regression models were done first, followed by age-adjusted models for variables of interest. Then, exploratory multivariate models were built. Clinical prognostic factors such as tumour 56 stage at diagnosis were included first in the multivariate models. Other potential confounders were added and their significance assessed by their effect on the model's likelihood ratio and on the primary variable's relative risk, including family history, parity, menopausal status and tobacco smoking among others (for full list, see pp 52-53). Analyses of insulin, C-peptide and fmctosarnine were controlled for the time since last meal. Al l p values were two-tailed. Bonferroni adjustment for multiple comparisons was not carried out because joint confidence regions were not required by the research objectives of this study (174). Independent T tests (for continuous variables) or Chi-Square (for categorical variables) were used to examine whether statistically significant differences exist between subgroups, in any of the factors being considered for the multivariate model. Additional comparisons of outcome rates and prognostic variables (age at diagnosis, tumour stage) from the larger population on the BC Cancer Registry were carried out on population data from BCCA patients diagnosed during 1991-2000. We compared our study population to the population of cancer cases who received treatment at the BCCA, and also to those who did not receive treatment at BCCA but were recorded in the BC Cancer Registry for those years. Variable Selection Initially, variables of interest with respect to hyperinsulinemia and potential confounders/known risk factors were selected from the research dataset, based on the literature, and on the investigator's previous work and hypothesis about the role of insulin in breast-cancer mortality. Covariates were included in an initial multivariate model through iterative stepwise modelling. Al l known risk factors for which data were available were considered for inclusion in the model, such as tumour stage, histology and treatment. Each 57 variable was checked for its effect on the primary variable before being included in a final model, and only those which caused greater than 20% change in the relative risk of the primary variable (e.g., insulin level) were included. A calculated variable representing the time since last meal was required as a co-variate to control partially for the non-fasting state of the subjects when analyzing insulin, C-peptide and fmctosamine levels. Calculated Variables Interval variables for breast-cancer and all-cause mortality were calculated using the following algorithms. The date of last contact was obtained by linkage, then validated by manual review of patient records to ensure the most recent and complete information was used. 1. If deceased, overall survival duration = date of death minus date of diagnosis. 2. If alive, overall survival duration = last contact date minus date of diagnosis. 3. For Phase II, the interval represents time from the year 2 questionnaire to date of last contact or death. Estimates of dietary intake were calculated using the nutrient analysis function of the DIETSYS software, based on the participants' responses to the Block-NCI Health Habits and History (HHHQ) questionnaire which includes frequency of consumption and portion size data. The nutrient per week contributed by each food was calculated using the following formula (175): (Portion Size X Nutrient Content X Weekly Food Frequency X Seasonality) 100 58 Portion size is the grams per portion. Nutrient content is the amount of nutrient in 100 grams of the food. The participant's reported food frequency is converted to a weekly frequency by multiplying the number of times by a conversion factor for each unit of time (day, month or year). Seasonality factor refers to foods specifically shown on the questionnaire as "in season". For these foods, the program assumes the frequency pertains to consumption during a short season, and adjusts the reported frequency downward to represent a year-long average. The calculated weekly estimate is divided by 7 to yield the average gram amount of food consumed per day. Percent of energy from fat, protein, carbohydrates, alcohol and sweets was calculated as the weekly estimate of consumption, divided by weekly total calories. For fat, this value is multiplied by 9, and for protein and carbohydrates it is multiplied by 4 to correct for the greater caloric content of fat. For fat, protein and carbohydrate calculations, any food which is a member of the Alcohol food group is ignored. For the calculation of glycemic load, the DIETSYS program database was updated with glycemic index values for each food using the most recent international tables (68). This value was then multiplied by the daily estimated intake of each food and these were summed to estimate the daily glycemic load, a factor which has been associated with risk of developing breast cancer (63) but not with breast-cancer mortality. When divided by daily carbohydrate intake in 100's of grams, this becomes daily average glycemic index. Descriptive Statistics Initially, frequency distributions were generated for each variable, and examined to determine if the distribution was normal, and where cut-points for categories should be set if 59 these were not already indicated by the literature. A correlation matrix was constructed to examine interrelationships between variables which might indicate the necessity to adjust the final model. Summary statistics to describe the population were calculated using SPSS Version 11.0, including means, standard deviations and quartiles (if continuous) of independent variables in the model, such as age, carbohydrate consumption, and saturated fat consumption . Independent T tests (for continuous variables) or Chi-Square (for categorical variables) were used to examine whether statistically significant differences existed between various subgroups of interest. Analysis of Questionnaire Variables Questionnaire data were collected for the entire cohort (N=603), and were analyzed using Cox proportional hazards survival analysis regression models. The lab assay data, in contrast, were collected only for approximately half the cohort (N=289) and were analyzed as a nested case-control, using logistic regression, discussed in a later section. Strategies to reduce the number of comparisons included doing a limited number of overall comparisons, then proceeding to subgroup analysis only when there was significant interaction, or theoretical reasons for doing so. The Multivariate Cox Regression Model. Crude hazard ratios were calculated to determine the univariate relationship between each variable of interest and breast-cancer mortality. A multivariate Cox proportional hazards regression model was built for each independent variable of interest, with mortality as the outcome variable and co-variates selected as described. A. priori, adjustment for age and stage at diagnosis was done since these are among the most powerful influences on survival. The independent variables of interest were in three 60 categories; 1) body size and shape, including weight, BMI and WHR, 2) food and exercise, and 3) biological factors including serum levels of insulin, C-peptide, fmctosamine and SHBG. Each independent variable of interest was analyzed as a continuous variable, or categorized into quartiles where possible, unless standard categories existed, to provide the maximum amount of information with acceptable power, and to provide comparability with othe published studies. The remaining explanatory variables were analyzed as continuous variables where the data were available, to conserve degrees of freedom in the model. The Cox Regression model can analyze data that contain censored observations, and can handle co-vahates without stratification, avoiding problems of too many groups with too few cases, and allowing an assessment of the impact of multiple covariates in the same model. In survival analysis, unless each cohort member is followed until the event of interest occurs, there can be losses to follow up or competing risks, which are treated as forms of censoring. The assumption is that censoring is unrelated to risk. Participants, however, may be censored due to competing risks such as coronary disease, caused by common risk factors, not necessarily independent of the study disease. If censoring by competing risk did not occur, the altered risk factors would lower the risk of breast cancer as well, so censored observations may overestimate risk for those censored by competing risks. For the Phase II analysis, values were calculated for the changes in dietary and body size variables between the Phase I and Phase II questionnaires, and entered into the model to assess the elevated or reduced risk associated with changes made in the first 2 years post-diagnosis. For body size changes, categories were; 1) 5% or more increase, 2) 5% or more decrease, and 3) no change (+/- less than 5%). For dietary changes, categories were; 1) 5%-25% increase, 2) more than 25% increase, 3) 5%-25% decrease, 4) more than 25% decrease, 61 and 5) no change (+/- less than 5%). Survival was measured from the date of the second questionnaire for these analyses. Model Checking. The Cox model assumes that observations are independent, and that the proportionality of hazards from one case to another does not vary over time. The assumption of proportional hazards was checked by examining survival function plots mcluding the log-minus-log plot, using the variables selected for the model, and found to not be violated. Diagnostic variables (e.g. DFBeta and partial residuals) were saved during the regression runs, to allow checking for outliers and influential points. Specific outlier data cases were selected for examination, and no errors were found. Models were run excluding the most influential points, but the results were not significantly changed so outliers were included in subsequent analyses. First-level interaction terms were tested for coefficients significantly different from zero by being entered individually after the main effects. Analysis of Lab Assay Variables A different method was used for the analysis of the lab data than the one used for the questionnaire data. The lab data were analyzed as a nested case-control study, including all breast-cancer deaths, and 2 survivors for each, matched on stage at diagnosis and length of follow-up. This was done because of limited serum volumes, competing requests for use of the sera, and limited funding, since this method reduces the sample size and therefore the cost and serum usage. In order to avoid selection bias introduced by matching in the design, the matching factor, stage at diagnosis, was controlled in the analysis. Matched Analysis. The analysis was done using a conditional logistic regression model. The matching factor, stage at diagnosis, was treated as a potential confounder, and each matching category defined its own stratum. A variable for "time since last meal" was calculated as the 62 difference between "time of blood draw" and "time of last meal", and was entered into all models of lab assay data to partly control for the non-fasting nature of the blood samples, a method used in some other studies where fasting blood was not available (176). Variables for insulin and C-peptide were log-transformed before analysis to compensate for their skewed distributions. Ethics Approvals Consent to participate was indicated by completing the self-administered questionnaire, according to UBC human ethics policy. A paragraph at the begmning of the questionnaire explained the study and how the data would be used, and this consent applied to both the questionnaire and the blood sample, which was standard procedure at the time of cohort assembly. The study was approved by both the Clinical Screening Committee of Research involving Human Subjects of UBC, and the Clinical Investigations Committee of BCCA. Additional ethical reviews for the current analysis were completed and approvals received from both the UBC Clinical Reseach Ethic Board, and the BCCA Cancer Registry. Permission was received to review clinical patient charts and access the BCCA Cancer Registry for mortality and recurrence outcomes. Dissemination of Results • Three manuscripts will be written and submitted for publication, including body size and shape results, dietary and physical activity results, and biological variable results. • The study results will be presented at scientific conferences and professional research gatherings. 63 Funds permitting, the investigator will prepare a written and an oral presentation of the results for the benefit of the study participants. After the study is completed, we will send a written lay summary to those who requested the results of the study (approximately 92%) and are still alive (approximately 425 — updated just before the mailing). We will invite the cohort members to an "Education Evening" on diet and breast cancer, to highlight the results of the study. The participants will be encouraged to bring a friend. 64 Chapter 4 RESULTS Overview Of the body size and shape variables at diagnosis, only waist-to-hip ratio was significantly associated with mortality in this cohort. Increased WHR was observed to be directly related to both breast-cancer and all-cause mortality in post-menopausal women but not in pre-menopausal women, and the association was restricted to those with estrogen receptor (ER)-positive tumours. For peri-menopausal status, not captured in this study, no observation can be made, as these women were probably classified pre-menopausal. Of the diet and activity variables at diagnosis, total fat and saturated fat intake were directly related to risk of breast-cancer and all-cause mortality, though only for pre-menopausal women. Increased protein intake, either in grams per day or percentage of total energy was inversely associated with risk, most strongly for pre-menopausal women. Percentage of total energy from foods in the sweets group (see Appendix C for full list) was directly related to mortality in both pre- and post-menopausal women. No significant association of alcohol consumption and mortality was observed after adjustment for stage at diagnosis. No significant association with physical activity at diagnosis and mortality was observed for pre- or post-menopausal women. 65 Of the biological variables, only insulin was significantly related to breast-cancer mortality, and then only marginally so. The increased risk associated with increased insulin was restricted to the post-menopausal women. No association was observed between C-peptide, fmctosamine or SHBG levels and breastcancer mortality. Changes in dietary intake during the 2 years post-diagnosis were not significantly associated with mortality, but both gain and loss of more than 5% of BMI were independently associated with increased risk of mortality, with BMI loss the greater risk. Although a reduction in BMI may be secondary to the cancer, this is less likely because the median time to death was 4.9 years, so changes in the first 2 years post-diagnosis still occurred almost 3 years before the outcome, on average. It is possible that the reduction in BMI was related to recurrences in the first 2 years post-diagnosis, but this outcome was not examined. Description of Study Population The average participant was 54 years of age, overweight by Statistics Canada standards (BMI 26), and married with 2 or 3 children (Table 1) 2 Almost half of the study participants were employed at diagnosis (46.7%). The vast majority were Caucasian (88.4%), though 7.1% were of Asian ethnicity. On average, physical activity levels were fairly low (43.7% did physical exercise rarely or not at all). Over three-quarters of participants tested for estrogen receptor status were positive, and over half of participants had a tumour greater than 2 centimeters in size. Most tumours were moderately or poorly differentiated, and 27.9% of participants had positive nodes. Tamoxifen 2 Figures and tables are located at the end of the chapter, figures first followed by tables, in numerical sequence. 66 treatment was primarily given to post-menopausal women while more aggressive chemotherapy was primarily given to pre-menopausal women. Breast cancer was the primary cause of death for approximately 72% of deceased cases, and cancer generally accounted for approximately 86% of mortality in the cohort (other cancers that contributed being lung, ovarian, intestinal cancer and leukemia), as shown in Table 2. Some misclassification is possible, given the number with unknown cause of death (n=13), but the high proportion of breast-cancer deaths minimizes this issue. In Table 3, the mean and median survival times are given for those deceased versus censored at the study cut-off date of June 30, 2001, by stage at diagnosis. In Figure 2, survival time by stage is shown. These data demonstrate the strong influence of stage at diagnosis on survival. Distributions and Correlations The frequency distribution for each independent variable of interest was examined using histograms. Figure 3 shows the distribution of waist-to-hip ratio at diagnosis and Figure 4 shows the same for body mass index at diagnosis. The WHR histogram reveals a near-normal distribution. Outliers were examined using the original source documents (in this case the questionnaires) to ensure the values were entered correctly and were consistent with other related data. No participants were excluded based on having an outlier value because the values were plausible and there was no reason to suspect they were incorrect. The histogram for BMI shows a near-normal distribution. The outliers were investigated and as with WHR, the values were within reason. The outliers for WHR were also outliers for BMI which is consistent with the relationship between WHR and BMI generally. 67 In Figures 5-9, the frequency distributions of dietary intake at diagnosis in grams per day are presented; fat (Figure 5), protein (Figure 6), eCarb (Figure 7), fiber (Figure 8), and total energy (Figure 9), showing near-normal distributions. Figures 10-14 show the frequency distributions of serum variables; insulin (Figure 10), C-peptide (Figure 11), fmctosamine (Figure 12), C-peptide-to-fructosarnine ratio (Figure 13), and SHBG (Figure 14). As with the dietary variables, outlier values were investigated and all but one were found to be within reason and were retained for analysis. One insulin value (878 pmol/L), almost twice the next largest value, was dropped because it was suspected that this person may be a misclassified diabetic on exogenous insulin. A matrix was constructed for the independent variables of interest showing Pearson correlation coefficients (Table 4). These data indicate that, as expected, WHR and BMI are correlated, and total energy intake is correlated with intake of individual macronutrients. In addition, age was significantly positively correlated with WHR (r = 0.205, p< 0.001) and with BMI (r = 0.121, p< 0.01). Age was also weakly negatively correlated with intake of energy (r = -0.081, p< 0.05), with intake of protein (r = -0.087, p< 0.05) and with intake of fat (r = -0.107, p< 0.01). A negative correlation with age was observed for physical activity, but no statistically significant correlation was observed with stage at diagnosis and level of physical activity. Body Size and Shape Variables Weight and BMI Weight at diagnosis was not related to all-cause or breast-cancer mortality for pre- or post-menopausal women (Table 5). Heaviest weight relative to current weight, weight as a teen 68 relative to current weight, and weight 5 years before diagnosis relative to current weight were also not significantly related to mortality. The BMI data in Table 6 indicate that BMI at diagnosis is not related to risk of breast-cancer mortality for pre-menopausal women (RR, 1.0; 95% CI, 0.5-2.0 for highest BMI category compared to lowest) or for post-menopausal women (RR, 1.2; 95% CI, 0.7-2.1 for highest BMI category compared to lowest). BMI was also analyzed as a continuous variable, and a quadratic term was tested, but the results were not significant. Waist-to-Hip Ratio By current standards (167), the average BMI for the cohort, 26.0, is classified as overweight. The cohort mean WHR, 0.8, is also the value generally considered in the literature to be the cut-point above which the risk of obesity-related health problems increases (177). Baseline body shape, as measured by WHR, was found to predict both all-cause and breast-cancer mortality for post-menopausal women (Table 7). Waist or hip measurements alone were not associated with mortality (data not shown), but for waist-to-hip ratio as a continuous variable, we observed a statistically significant age-adjusted RR of 1.5 (95% CI, 1.1-2.1) for breast-cancer mortality, meaning a 50% increase in risk of dying of breast cancer for each 0.1 increase in WHR (e.g. WHR increase from 0.8 to 0.9). For all-cause mortality, the RR was 1.6 (95% CI, 1.2-2.1), or a 60% increase in risk of dying from any cause per 0.1 increase in WHR. No statistically significant increased risk associated with elevated WHR was observed in the pre-menopausal women (RR, 1.0; 95% CI, 0.6-1.6) for either breast-cancer or all-cause mortality. Analyses were performed with and without women with ductal carcinoma in-situ 69 (DCIS), to verify that the association applies to in-situ as well as invasive cases, and the observed relative risks for WHR were not altered.. The Effects of Age and BMI on WHR Risk Analysis by 10-year age categories showed that the difference between pre- and post-menopausal women was not solely due to the effects of age differences (Table 8). Although the number was small (n=25), reducing the power, in pre-menopausal women age 50-59 years at diagnosis, increasing WHR was not associated with increased breast cancer mortality (RR, 0.6; 95% CI, 0.2-2.3). In contrast, post-menopausal women age 50-59 years at diagnosis showed a relative risk of breast-cancer mortality of 1.6 (95% CI, 0.9-2.8) for a 0.1 increase in WHR and a statistically significandy increased risk for all-cause mortality (RR, 1.9; 95% CI, 1.2-2.9). An analysis adjusted for ER status (69% positive in pre-menopausal women; 81% positive in post-menopausal women) was carried out but the results were not changed. The relative risk of breast-cancer mortality in post-menopausal women age 60-69 years, 1.6, was also statistically significant, but wider confidence intervals were observed as expected for the smaller 70-75 year age group. Stratification by BMI category showed that post-menopausal women were at increased risk of breast-cancer mortality from increasing WHR, irrespective of BMI (Table 9). In contrast, pre-menopausal women showed no relationship between elevated WHR and mortality at any BMI. The lowest BMI category, less than 18.6, contained too few women for analysis. Multivariate Analyses of WHR Multivariate models for post-menopausal subjects did not greatly alter the relative risk associated with WHR. The only breast-cancer mortality factors that modified the relative risk associated with WHR were age at diagnosis, ethnicity, family history, estrogen receptor status 70 and severity of disease at diagnosis, including tumour size, nodal status, tumour grade and systemic treatment. Sub-group analysis was performed for tumour size and nodal status, but not for systemic treatment and ethnicity, due to the small size of some sub-groups. The relationship of estrogen receptor status and family history with WHR is shown in the next section, using stratification. Post-menopausal women with tumours larger than 2 cm were at somewhat greater risk from a 0.1 increase in WHR (RR, 1.9; 95% CI, 1.2-3.0) than women with tumours 2 cm or less in size (RR, 1.4; 95% CI, 0.9-2.3) (Table 10). This analysis was adjusted for nodal status, age, ethnicity and systemic treatment. Adjustment for BMI did not alter the results, perhaps because the mean BMI for the two groups is very similar (26.4 for those with tumours 2 cm or less, 26.5 for those with tumours more than 2 cm). The fully-adjusted RR for all post-menopausal women was essentially unchanged from the univariate analyses. When node-negative and node-positive cases were analyzed separately (data not shown), controlling for age, ethnicity and tumour size, the breast cancer mortality risk associated with an increase in WHR was similar for node-positive and node-negative women (RR, 1.6; 95% CI, 0.9-2.9 for node-negative and RR, 1.7; 95% CI, 1.1-2.9 for node-positive). Survival by Waist-to-hip Ratio and Stage at Diagnosis Figure 15 compares the survival curves for women at or above the WHR median of 0.8 with women below the median, by stage at diagnosis. For Stage I women, those at or above the WHR median exhibit poorer survival than those below the WHR median over the entire follow up period. In contrast, the risk for Stage II women is similar, regardless of WHR category. In Stage III women, after approximately 5 years post-diagnosis, being below the WHR median is associated with poorer survival. A test of interaction between stage at 71 diagnosis and WHR as a continuous variable was performed, but the interaction term was not significant (p=0.61), perhaps due to low power. Stratification on Estrogen Receptor Status and Family History In Table 11, WHR is represented as a continuous variable and data are given for pre- and post-menopausal women separately, stratified by estrogen receptor (ER) status and by family history, with unstratified values given for comparison. In this way, it is clear that the association of elevated WHR with post-menopausal breast-cancer mortality is confined to those women who are positive for estrogen receptors (RR, 1.6; 95% CI, 1.1, 2.5 for ER-positive, compared with RR, 0.8; 95% CI, 0.3, 2.2 for ER-negative). No significant association was seen for pre-menopausal women, regardless of ER status. The stratification on family history indicates a stronger association of WHR with breast-cancer mortality in post-menopausal women without a family history of breast cancer in a first-degree relative (RR, 1.6; 95% CI, 1.2-2.3 for family history-negative, compared with RR, 1.2; 95% CI, 0.6-2.6 for family history-positive post-menopausal women). The family history-positive group is much smaller, so the wider confidence intervals may be the only real difference between the groups. Physical Activity and Dietary Variables Physical Activity The risk ratios for various frequencies of different leisure-time physical activities are presented in Table 12. The reference category in each case is the category of individuals doing the activity rarely or not at all. Age-adjusted relative risks are given for breast-cancer mortality and for all-cause mortality. Adjustment for stage at diagnosis did not change the relative risk. This is consistent with the correlation data which indicated that there is no significant 72 correlation between stage at diagnosis and self-reported physical activity at diagnosis in this population. Age, however, was significantly and negatively correlated with all physical activities except walking, so age-adjustment was retained in the models. No significant association was seen for total exercise (number of times per year for all exercise activities combined), either as a continuous variable or categorized into quintiles. Overall, leisure-time physical activity at diagnosis was not related to mortality. The one activity with a marginally significant inverse association with all-cause mortality was gardening more than once a week (RR, 0.6; 95% CI, 0.4-1.0, age-adjusted), compared with doing no gardening. Gardening, of course, may be a marker for more than just physical activity, and this will be discussed in a later section. Also, it may be a chance finding. Data on occupational physical activity were not available for analysis. Total Energy Intake Higher total energy intake may be inversely associated with breast-cancer mortality (Table 13), at least to a point. Compared with women in the lowest quartile of energy intake, women in the second quartile had an RR of 0.6 (95% CI, 0.3-1.0). Stratification by menopausal status (Table 14) shows that the inverse association is restricted to post-menopausal women (RR, 0.5; 95% CI, 0.2-0.9). The analyses were adjusted for age and stage at diagnosis. Because energy intake may simply be a marker for physical activity, analyses were carried out controlling for exercise variables, including number of blocks walked per day, flights of stairs climbed per day, frequency of physical exercise, sports, swimming and jogging, but the results were unchanged. An analysis excluding those participants who died in the first year post-diagnosis was performed, but the results were not changed. 73 Percent of Energy from Macronutrients Dietary intake of each macronutrient was analyzed in two ways, as a percentage of total energy intake, and as an average number of grams per day (shown in subsequent sections). A higher proportion of energy derived from protein consumption showed a statistically significant inverse association with risk of both breast-cancer and all-cause mortality, strongest in pre-menopausal women (RR, 0.1, 95% CI, 0.04-0.3 for each 10% increase in percent energy from protein) but still statistically significant in the post-menopausal women (RR, 0.4, 95% CI, 0.2-0.9) (Table 15). The only food group with a statistically significant result was sweets, which was associated with a 2.1-fold increase in mortality for each 10% increase in total energy derived from foods in the sweets group for pre-menopausal women (95% CI, 1.3-3.4), and a 1.5-fold increase in mortality for post-menopausal women (95% CI, 1.1-2.1). Fat Consumption Saturated fat consumption at diagnosis was directly related to both breast-cancer and all-cause mortality (Table 16). Women in the fourth (highest) quartile of saturated fat consumption were 2 Vz times more likely to die of their disease during the 10-year follow-up period than women in the first (lowest) quartile (RR, 2.5; 95% CI, 1.2-5.3), and this association was stronger for all-cause mortality (RR, 2.7; 95% CI, 1.4-5.4). Analyses were adjusted for total energy intake, age, and stage at diagnosis. The association of total fat consumption with risk of dying was similarly elevated but the 95% CI included 1.0 (RR, 2.1; 95% CI, 0.9-4.8). Mono-unsaturated and poly-unsaturated fat consumption did not demonstrate any statistically significant association with mortality in this cohort (data not shown). When stratified on menopausal status (Table 17), the risk associated with increased fat consumption, both total and saturated fat, is confined to the pre-menopausal women, with an increased risk of mortality of nearly 5-fold associated with being in the highest quartiles of both, compared with 74 the lowest. No statistically significant association with fat intake was observed for the post-menopausal women. Protein Consumption Breast-cancer mortality was inversely associated with intake of protein, the result being stronger for breast-cancer mortality (RR, 0.4; 95% CI, 0.2-0.8) than for all-cause mortality (RR, 0.6; 95% CI, 0.3-1.1), adjusted for total energy intake, age, and stage at diagnosis (Table 18). An inverse dose-response relationship was suggested for breast-cancer mortality. When stratified on menopausal status (Table 19), the inverse association is seen among both pre-and post-menopausal women, though the relationship is strongest in the fourth (highest) quartile among pre-menopausal women (RR, 0.2; 95% CI, 0.1-0.9), and in the third quartile for post-menopausal women (RR, 0.3; 95% CI, 0.1-0.8). The inverse association is weaker for all-cause mortality. Carbohydrate Consumption No relationship between carbohydrate consumption and mortality was observed in this cohort (Table 20). Carbohydrate intake (excluding alcohol, shown separately in a subsequent section) is shown in total, then divided into fiber and non-fiber (effective carbohydrate or eCarb) components. The point estimates suggest an inverse association with high (fourth quartile) fiber consumption and an elevated risk associated with high eCarb consumption, but the confidence intervals include one. Additional sub-group analysis by menopausal status did not reveal any differences (data not shown). Alcohol Consumption Self-reported data on alcohol consumption at diagnosis is presented in Table 21, suggesting an inverse association (RR, 0.7; 95% CI, 0.4-1.1), comparing drinkers with rion-drinkers, 75 though the confidence limits include 1.0 (p=0.12). There was a significandy increased risk of mortality associated with stopping drinking, compared with not stopping (RE., 2.1; 95% CI, 1.3-3.2). Further stratification on ER status showed this association to be absent in women with ER-negative tumours (data not shown). The questions related to consumption of specific alcoholic beverages reveal that if there is an inverse association, it is most probably associated with beer consumption rather than wine or liquor consumption. These data would indicate that drinking beer at least once a month but less than once a week may be associated with approximately a 50% reduction in risk of mortality from breast cancer (RR, 0.5, 95% CI, 0.2-1.1), or mortality from any cause (RR, 0.4, 95% CI, 0.2-0.9), but given there is not a plausible explanation for this isolated finding, it could be a chance observation. Further adjustment for stage at diagnosis weakened these relationships, suggesting that earlier-stage participants may simply feel well enough to drink alcohol while later-stage participants do not. Glycemic ljoad No statistically significant association between average daily glycemic load and mortality was observed for pre- or post-menopausal women in this cohort (Table 22). Unadjusted analyses and also those adjusted for age, stage and total energy intake produced risk estimates close to unity, and confidence intervals that included 1.0. This is interesting in light of the observed risk associated with the sweets food group, but sweets often have high fat and energy content and low fiber content, in addition to high glycemic index. Biological Variables Descriptive data on insulin, C-peptide, fructosamine and SHBG variables are presented in Table 23. Insulin values below the minimum detectable limit, designated as less than 30 pmol/L in the lab reports, were set to V2 the minimum, or 15 pmol/L. Insulin values ranged 76 from 15 to 480 pmol/L, and the distribution was skewed, with 75% of the values below 106 pmol/L. Similarly, the C-peptide distribution was skewed and both msulin and C-peptide values have large standard deviations. These two variables were log-transformed when analyzed as continuous variables to approximate a normal distribution. Serum Insulin Levels Insulin levels were positively correlated with C-peptide, WHR and BMI, negatively correlated with SHBG, and not correlated significandy with stage at diagnosis, histology, tumour grade, estrogen receptor (ER) status, or serum fmctosamine. Figure 16 shows a similar pattern to that observed for WHR, with a clear advantage to a lower level of serum insulin for Stage I women, a small advantage for Stage II women, and a disadvantage for Stage III women, after about 5 years post-diagnosis. In Table 24, odds ratios (OR) associated with quartiles of insulin levels at diagnosis are stratified on stage at diagnosis, because the women were matched on stage in the sampling design. The OR's shown are adjusted for age at diagnosis and time since last meal before the blood was drawn. These data indicate an increased risk of mortality in the 2 n d and 3 r d quartiles of serum insulin levels for post-menopausal women, but not for the 4 th quartile. No association was seen in pre-menopausal women. Due to small subgroup sizes, only the Stage II stratum is statistically significant (OR, 5.0; 95% CI, 1.0-26.4 for third quartile compared to first quartile). For stage I and stage III groups, the quartiles were collapsed due to small numbers, into categories representing those above and below the median. The data on all stages show that for post-menopausal women, the risk associated with increased insulin starts with the second quartile, and follows a more linear pattern, with smaller confidence intervals than the subgroups by stage. A test for interaction between stage at diagnosis and insulin level as a continuous variable was 77 performed, but the interaction term was not significant (p=0.62). Analyses were performed with and without women with DCIS, and the observed results were not altered. C-Peptide and Fructosamine Serum levels of C-peptide (Table 25) and fructosamine (Table 26) were not associated with breast-cancer mortality. The pattern of associations with C-peptide-to-fmctosamine ratio in pre-menopausal women is not consistent with elevated risk for an elevated ratio (Table 27). The odds ratios for the post-menopausal women are consistent with the insulin results, but wide CPs cannot exclude a chance result (OR, 2.4; 95% CI, 0.5-11.6 for the highest compared with lowest quartile). Sex Hormone Binding Globulin Levels SHBG levels were inversely correlated with insulin levels (Table 23), but SHBG levels were not associated with breast-cancer mortality in this cohort (Table 28). The Effect of Tumour and Treatment Variables on Insulin Prognostic Association The association of insulin with breast-cancer mortality was examined after adjustment for each variable individually including; ER status, histology, tumour grade and systemic treatment. This method was used in the exploratory analysis to maximize the number of participants in each analysis, given the small sample size and missing values for some variables, notably ER status. The effects of family history and ER status on the insulin association are presented in Table 29 as an adjusted analysis because the subgroups sizes were too small to permit stratification. The adjustment for family history results in a slightly stronger association between insulin and breast-cancer mortality, across all quartiles, while adjustment for ER status 78 strengthens the association, but mainly for the 2 n d quartile. Adjustment for both ER status and family history results in a significantly stronger association across all quartiles. Insulin remained positively associated with mortality after multivariate adjustment (Table 30), but lost statistical significance in the 3 r d quartile. Due to the combination of small subgroups and missing values, all stage strata were considered together, with stage entered as a co-variate. The test for overall association was not significant (p=0.15). Changes Made Post-Diagnosis Body Si%e and Shape Changes The median change in BMI during the first 2 years post-diagnosis was an increase of 0.5, or about 2.1% (Table 31). For WHR, the median change was an increase of 0.007, or less than 1%. In both cases, the changes were normally distributed, with a few outliers whose values were investigated, then retained in the analyses since no errors or inconsistencies were detected. The range of percentage change in BMI was from a decrease of 21.2% to an increase of 54.5%; the observed range of percentage change in WHR was from a decrease of 32.4% to an increase of 65.9%. The median was used to describe the average change because of the few large increases observed which distorted the mean values. The relationship between changes in BMI and mortality is naturally a complex one. These data suggest that weight loss is a more serious risk than weight gain after a breast cancer diagnosis (Table 32), likely due in part to the disease itself. Adjusted for age and stage at diagnosis, the relative risk associated with increasing BMI more than 5% was 1.8 for breast-cancer mortality (95% CI, 1.0-3.1), and was attenuated for all-cause mortality (RR, 1.3; 95% CI, 0.8-2.1). In contrast, decreasing BMI more than 5% was associated with an increase in risk of 79 dying (RR, 2.8; 95% CI, 1.5-5.5 for breast-cancer mortality; RR, 2.2; 95% CI, 1.3-4.0 for all-cause mortality). When stratified by menopausal status at diagnosis (Table 33), the risk is greater for pre-menopausal women (RR, 4.1, 95% CI, 1.3-13.0 for breast-cancer mortality) than for post-menopausal women (RR, 2.5; 95% CI, 1.0-5.8), suggesting different metabolic effects by menopausal status. No significant association was observed between changes in WHR and mortality from breast cancer or from any cause (Table 34). Dietary Changes On average, women surveyed in Phase II reported reducing their total energy intake by a median value of 110 calories per day or about 8% (Table 31). They also reduced their intake of fat by a median 6.8 grams per day or about 13.3%, their intake of protein by 4.2 grams per day or about 7%, and their intake of total carbohydrate by 2.4 grams per day or about 1.5%. When fiber and eCarb are analyzed separately, it shows that fiber intake increased (1.0 gram or 7.9%) and eCarb intake decreased (4.0 grams or 2.3%). Changes in Carbohydrate Intake. In Table 35, the relative risks associated with increasing or decreasing total carbohydrate, fiber and effective carbohydrate are shown, compared with "no change". The no change category was defined as increasing or decreasing not more than 5%, based on the values at diagnosis. The relative risks shown are adjusted for total energy intake, age and stage at diagnosis. No relationship between changes in e-carbohydrate intake post-diagnosis and mortality is indicated by these data. The distribution shows that most women increased their fiber intake, but this did not have an impact on breast-cancer mortality. Increasing intake of total carbohydrate may increase risk of breast-cancer mortality (RR, 2.3; 80 95% CI, 1.0-5.3 for 5-25% increase compared with no change), but a larger sample would be required to confirm this. Changes in Fat Intake. Percentage changes in saturated fat intake in the first 2 years post-diagnosis (Table 36) show a pattern similar to that of carbohydrates. Increased intake of saturated fat (5-25% increase category compared with no change) is associated with an RR of 2.7 (95% CI, 0.9-8.5). Changes in total fat intake are not associated with breast-cancer or all-cause mortality. Again, the distribution in these categories confirms that most women reduced both total and saturated fat intake over the 2 years post-diagnosis. Changes in Protein Intake. Decreases in protein intake were associated with a reduced risk of dying (Table 37) (RR, 0.4; 95% CI, 0.1-0.9 for breast-cancer mortality and a decrease of more than 25%). This analysis was adjusted for total energy intake, but changes in protein intake are also strongly correlated with changes in fat intake (Pearson correlation coefficient 0.773, p<0.0001), which may partially explain the inverse association observed. Additional adjustment for change in fat intake caused the association to lose statistical significance. Changes in Total Energy Intake. No association was seen between changes in total energy intake and mortality (Table 38). Changes in alcohol intake and in average daily glycemic load were not associated with mortality (data not shown). 81 82 100 Waist to Hip Ratio at Diagnosis Figure 3. Distribution of waist-to-hip ratio at diagnosis. 83 Std. Dev = 4.6 Mean = 26.0 N = 603.00 <? c? •<? <? <? o o o o o o o o o Body Mass Index at Diagnosis Figure 4. Distribution of body-mass index at diagnosis. 84 Std. Dev = 24.48 Mean = 61.2 N = 603.00 J U 5fr <5k <5W A , CrU ^ <7 7_ Z ? Z - 7> Estimated total fat, grams/day Figure 5. Distribution of dietary fat intake at diagnosis. 85 140 120 H Estimated protein, grams/day Figure 6. Distribution of dietary protein intake at diagnosis. 86 100 E-Carb Consumption, grams/day Figure 7. Distribution of dietary e-carbohydrate intake at diagnosis. 87 100 H 6.0 10.0 14.0 18.0 22.0 26.0 30.0 34.0 38.0 Estimated dietary fiber, grams/day Figure 8. Distribution of dietary fiber intake at diagnosis. 88 120 — i i — i — i i — i i i i i i i i i i i i i i %%%%%%%%%%%%%%%%%%% ^ ^ o o o o o o -o o o o o o o o o o o Estimated energy, kilocalories/day Figure 9. Distribution of dietary energy intake at diagnosis. 89 100 Insulin level (pmol/L) Figure 10. Distribution of serum insulin level at diagnosis. 90 40 H C-Peptide level (pmol/L) Figure 11. Distribution of serum C-peptide level at diagnosis. 91 Fructosamine level (umol/L) Figure 12. Distribution of serum fructosamine levels at diagnosis. 92 50 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 .50 1.50 2.50 3.50 4.50 5.50 6.50 7.50 8.50 9.50 C-Peptide to fructosamine ratio Figure 13. Distribution of serum C-peptide-to-fructosamine ratio at diagnosis. 93 94 STAGE I 1 02 -i O 90 1000 2000 3000 Days survived-to June 30/01 STAGE II 1000 2000 3000 Waist-to-hip ratio D At/above WHR median + -censored 0 Below WHR median + -censored Waist-to-hip ratio ° At/above WHR median + -censored D Below WHR median + -censored Days survived-to June 30/01 STAGE Waist-to-hip ratio ° At/above WHR median + -censored ° Below WHR median "*" -censored 1000 2000 3000 4000 Days survived-to June 30/01 ure 15. Survival curves by WHR and stage at diagnosis. STAGE I 1.1 T CO E o co 0 1000 2000 Days survived-to June 30/01 STAGE II 1 0 8 Serum Insulin a At/above median + -censored ° Below median + -censored Serum Insulin n At/at»ve median + -censored ° Below median + -censored STAGE CO E a o 1000 Serum Insulin n At/above median + -censored D Below median + -censored 4000 Days survived-to June 30/01 ure 16. Survival curves by insulin level and stage at diagnosis Factors that Increase Factors that Decrease Figure 17. Biological pathways MORE AVAILABLE ESTROGEN 00 00 00 T-H 00 ^ 00 — O ° ° CO CN T-H ^ LO o _^  © , CN 00 T"H CN o CO "3-CN *0 00 CN L O sd 00 CN VO 0 s o -- o -^ CN O O L O CN O 1-H fO CO T-H CN ^ 00 < CN r- o CO Cs LO CN r-o CN CO O 00 O CO 00 r-- T-H T-H CN cO O 00 T-H SO O :=r\ p . o CN N o 00 ° ^ CO CD . o 0 s ©~* 0 s o~- CO sO 00 <^  oo T-; LO SO T-H T-H TH- 0s- 0s- 0s-00 T-| CN CO 00 t~~ CN O Cs ON 00 O CO CN o <=> ON LO i r i r-SO T"H ^ CN "1 l O , CN, CO, 1^ f i o \ o i T—I LO T—I CN SO r-^  o -tf- T-H r--00 ON O CN T-i CN ^p s? Cs C O 00 r-^  T~H CO i r i L O CN CN CN Cs 00 CO l a C s « CN LO SO T-C SO CO O 00 o LO cs T—1 Ci 00 CN T-H T-H SO CN ^p cO ' CN r-- 00 LO LO CO 00 CO CO Cs T-H T-H CN 00 CO LO CO Cs sd O T-H T-H CN LO 00 o CN •5 O r-- os 00 Cs LO LO CN L O SO T-H 4-1 V *d u CN LO CO . 00 CN i £ S C S 2 ^ ^ 00 CN "^ T 00 CN T-H O CO T-H so CN so •d TO •J3 TO > a o •{3 a, a o u TO hi •d I ? W U 3 •d * "3 -+J TO 3 Q H Ji •d 3 a u o <-l t-- o C O O ^ T-H cs u >> TO TO CO </> u v a a •a -a £ * * W r u ru U <u "TO u LO SO T-H CO T-H CN 5 «J > u ^ a w O o -g e I a 5 ° a u 6 <j S * o so LO T-H SO sd so O CO r--CN CN od •d u « •a a u u "2 Q •a § u •a CO 0 o u OH •d u u -a •d TO •a c IT! « o sc .a ^ 0 g o 6 § o q CN LO u ON OS O H N i l l v ° ^P d^ ^p d^ v? NP C O C \ C O T-H C N o NO C O T-H C O C N NO NO CN co 00 C O C O T-H T-H d^ CN vp NP N? d^ vO C O 00 d^ T-H NO C N NO O T-H L O C N T-H ^p vp xP d^ d^ ^p NO CN d^ O r~-C O C O C N oo C N CN T-H L O CN C O NO C O o xp %P •*P NP N? d^ vP sP d^ d~-00 r-- d^ C N C O NO 00 T-H 00 NO C O T-H C N © CN C O CN CN L O T-H L O T-H T-H d^ C N T 1 C N L O NO o T-H T-H CN C O •3- CN T-H CN d^ NO d^ NO NO o CN C N © rH T-H L O T-H T-H > W a « <u •£3 U 5 u II & £ • 9 . „ u .a S <3 T3 CD -a 3 IT o 0 a O 3 3 a o 1 o +-> cn i 31 r 0 +J u u 4-1 CO a! a OH 8 o o ,0 -a CD I ,o u ?7H I Si? _ !/5 oo CL) T-H OJO a a ^ cn crj cn o U M u OH O 1-1 t^Z SO CN 00 o LO CN 00 o LO CN 0s-O T-H TH CN SO CO CN so CO O © o t-~ o o o o cs r-~ T-H CO r-- T-H LO o o CN 0 s d^- T—i cs T-H 00 CN SO CO o o o CS C s o o o o o so TO o -2 £ Q c TO •-3 C TO u 13 u O CO a u TO "TO I C/3 T3 u x TO U U Q C/2 c TO •-3 u C TO u > u Q CN Cs O T-H CN T-I CN CN O CT; h CO CS 00 CO OD T-H cO CS CN co od K od t^" LO SO t^- 00 o so cd o u-i cs oo so t^- r~-r-~-•3-t^- LO _ CN SO ^ CN T-H ^ t--LO O O CO CO CN CN CN CS CO CO SO -^ J- CO CN CO so r>; LO SO -^f CO v « £ x ^ ^ ^ 5^  LO CN so CN CO T-H CS Cs CO LO CN LO LO T-H CO CO so co i—i U Q u u u ba t>a OJO 3 S U CO CO CO CD u H .3 o 4) •cl , o CU 0 0 00 0 LO CO O 0 00 LO 0 CN LO T-H O O O ^-T-H O 00 CO t--00 0 0 00 C N O C N C N LO T-H O C N 00 r-C N 0 C N O LO 00 CN O CO C N ^-0 0 p 0 O 0 1 0 O 1 O O 0 p' 0 O 1 O O 0 0 1 O O O C N 0 NO 00 C N O NO 00 r--T-H O O O C N O 0 CO C N NO 0 0 0 LO O O O CO CO T-H 0 0 00 NO O O O O O O 00 C N 0 CO C N •<*-©• 0 O 0 O O O O 0 0 0 O O 0 O O ^ 0 1 0 00 0 • 00 CN T-H O C N r-r--0 0 T-H r-00 T-H O O C N r--C N NO CO 00 0 0 0 0 N O N O O O O LO T-H LO 0 0 0 O O O 00 NO O O O C N 0 CN O LO 0 O O 0 0 0 O O 0 0 O O O 0 T-H 0 O 0 1 O C N O NO T-H NO LO O r--CN r--0 T-H 00 0 0 LO O C N NO C N r--NO 00 0 0 0 CO NO O O O O O O LO LO O O O CO CO O O T-H T-H 0 C N 00 r-~ O O O 0 O 0 O O 0 0 O O T-H O O O O 0 0 C N O C N NO CO CN O r-1--LO r-00 0 CO O C N LO O NO C N CO 00 0 0 0 O O O CO r--NO 0 0 0 O NO NO O O O t--LO O O O T-H C N 0 C N C N LO O O O 0 0 O O O 0 0 T-H O 0 O O O O 0 O co co 0 00 C N CO CN O NO NO LO T-H 00 O 0 r-0 O r--NO 0 0 p C N CO 00 O O O r--NO 00 0 0 0 NO CO 00 O O O NO O O O 00 0 0 T-H t^" 00 0 O O O O 0 0 O 0 O 0 0 0 O O O 0 0 CO C N O "3-CN O r-~-0 0 NO 00 CN NO O 0 CO T-H 0 0 p r--0 0 NO C N LO O NO T-H LO 0 CN NO C N T-H O O CN t--C N C N O O CO C N T-H 0 00 CO r--O O 0 0 O 1 O T-H 0 O O O 0 O O t O O O 0 0 LO O CN O O O T-H C N T-H CO 0 0 O O O CN NO O O CO T-H T-H 00 0 O r--00 0 CO O r~-0 T-H 00 O O r--0 0 T-H r-00 r--O O O CN LO 0 0 0 O O O 0 ^ O O 0 1 O p' 0 0 O 0 0 0 O 0 1 0 NO T-H cO O O O NO ON 00 0 0 0 C N O CN O CN CN O CO 00 LO 00 0 C N CO C N T|-CN O 00 LO LO NO O C N T-H T-H t--CO O CO NO CO CO O O co C N CO 0 O O 0 0 O O O 0 0 O O O O O O O O 0 0 1 0 T-H NO CO O O O 0 0 p T-H C N T-H CO O O 1--O O NO 00 CO C N 0 NO NO LO CO CN O LO LO O CN O C N r-00 C N O NO 00 1^- t--0 00 LO 0 0 O T-H O O O 0 0 O O 0 0 0 O 0 O 1 0 0 1 0 0 0 p T-H NO CO 0 0 0 LO O C N O O O CO CN O ^J-C N 0 CO CO O 00 C N C N O CN NO T-H CN O NO T-H NO 00 ^-0 00 ^-CN C N T-H O NO 00 0 LO CO O T-H O 0 O O O 0 O O O O O O 0 O O 0 0 O G O •d *—H CU a o U 0 o cn CD £H O •d 2. i a •B u ° u 0 .CN cn bp 2 bp in PH 0 0 a •s o U G 0 cn a PH a 0 •a 4> r 1 ^ U 2 a C N 0 <_> c« bp g in PH CN^ C O a 0 •a JH a o a G G 0 0 0 •d •d •d <u u 0 ^  0 0 U rd U rd U G 0 cn bp g C O PH d G CN 0 <J> cn bp g C O PH C N . bp 'co G O cn c^  OJ PH X CU G O OJO CU bQ B co G O •d ,—. u ^ o u r 1 d G d G o •d CTj 1—H CU o f G -v cn i d H bp g bp g 0 0 PH C O PH CO I u u CU X CU o PH CO o TO o u cn 3 TO l u CN C O C O o o C O ^ L O r-; C O T-H T-H C N C N T-H T-H T-H T-H T-H T-H T-H 1 so" NO" 1 LO" LO" 1 LO" Lo" 1 NO" I--" 1 NO" r-" LO" O O o o © o O © © © © ' o CN C N O p p O 00 00 O O O p O C N O o T-H © ' © © NO NO C N o N O C N N O o CN T-H T-H CN CN T-H C N T-H T-H T-H C N CN NO" r-" NO" LO" LO" NO" NO" CO" LO" LO" o o o o o O O O o o O p p p O T-H p O p T-H O O C N p C N CN NO C N o o CN o CN o C N r--NO NO NO o T-H C O C N T-H C N C O T-H r-o C N L O T-H C N O T-H C N CN CN T-H oo o C O CN C O C/3 _ Q L O !=H L O CN ^ V cO bs L O bs L O bs L O L O L O CN T-H L O C N T-H L O o L O C O CN C O T-H L O C O C N C O L O T-H T-H T-H T-H T-H T-H A V A V A V jj *H NO NO T-C S A U P-i 3 cs OH O a o 3 OH O G <u ht QjO O $> <U ' £ 5 1-1 X <U • M u 2 a <u <u I 8 o 'So CJO ITJ 3 JJ 8 1 W S ^ W J H CN CN CN 00 CN CN 0.2, NO" o O © CN © o O O CN T-H r--© ' r-~-LO CN CN o CN LO CN cO CN 0.3, O o © T-H O CO O T-H o Cs O O T-H NO o CN CN CO CN 00 CN 00 LO o T-H u t-l L2 <u L O 1) 6 o y IS SH <u W -*-> S 3 « i o g ""§0 "5 § ^ <q co so o o a U u a |u • M CO CM U l« U LT> ON o 1 CN C O o T-H L O T-H T-H CN CN T-H T-H Lo" NO" LO" NO" o o o O o o p do p p CN O p NO ON o T-H T-H T-H o o C O L O o C O T-H T-H T"H CN CN T-H CN Lo" oo" r--" LO" co" o o o O o o O T-H O C O O p NO CN T-H o T-H T-H T-H o , - ' C O C O CN NO L O CN 00 O T-H NO NO NO L O CN T-H L O CN 00 CN V 24-27 LZ< <24 24-27 CN A <24 24-27 LZ< z4 < u PH 3 OH 0 a o PH 3 CS OH O a o <u +J CS o a u 1H .1* o OJO & CNJ u +-> co X 4) a 'o ^ CQ S «8 H «8 O PL, c a o </j 3 «8 O c i tH PH u u a LO ON ON 3 PH O W DC • o ^ pi •X * * * T-H T-H cO cO L O CN] CN" CN" co" CN]" O O , " H C N NO NO T-H CN] T-H * -X--X-* C N C N NO* 00 CN] T-H co" T-H T-H o ^ C N CN] L O cvi C O C N r--o o O r--L O C O <N o NO co CN] C O T-H r-" ^ f " LO" NO" o © © © L O C N CN] O o T-H r-- NO C O CN] CN] T-H r-" ^ " NO" o © o © C N p O T-H o L O NO CO L O T-H C N CN] CN] oo C0 u 3 a © o oo oo © NO I £ v o ^ LO o r-© oo 00 V 00 o <=> A 3 O 3 C 1 o U a o I © II V CL, a CTJ U L O o o II V OH co a OS 1) u CU i £ CU CTJ •S T3 CU '-3 H T H S o s S O u ( J O C K \< » OH - O c S o CA 3 TO c en O C u s o "TO co 3 TO a c 4J O u - fa TO u u 4f ON O i -CN CN co " o" co" o CN ON ^ NO CO LO cO CN CN LO ON" T-H" NO" O T-H O NO NO 00 LO CO _,_ CO NO ^ T-H T-H " ' T-H O CO CN CN CN LO" NO" CN" o o o O T-H SO T-H ^H O CN O CO CN CN CN LO" NO" CN" © o o O T-H so T-H ^H O CO LO LO CN ON ON ON CN CO M 6p W> 0 o o S h fa 1 -s -s ON o o ON ON t C NO •5 ftJO 6J0 ftX 3 3 3 O O O h h fa •B $ $ © © © LO NO r~-o . O £ II > V ^ CH.S OS CO fa o .a o © 1 -g O fa 0 •X- Lfa * « . r\ . fa LO „ O fV fa r/N ^ - ! C ^ > u L f a G 0 Q -O Si0 II G in V aor a cn <U G > <a -a a * G "3 s B , o I* cn 3 a cn l«2 CU a , o IB* 13 CO 3 3 3 « U cn « _ £ 2 o ^ o ft* « U « CO LO CN CN CN CO o f CN" 00" O o NO LO NO T - H T - H t-O CN O CN CO CN" CN" oo" o o o LO 00 T - H T - H T - H CN SO T - H O CN LO r~~ 00 LO NO T - " CN CN LO" Tt-" o o o p p T - H o 00 LO T H CN CN LO" co" o o o p p CN o CN cO T - H NO o CO LO co CN CN CN CN CN CN CN CN v i 6 + I—I hH h-< s s s CO PH PH o o II V OH co 3 cu LO o II V OH G CU l o 3 L2 cu cn > x a! CU u U o '0 T ^ T3 O cu o © • CU 6J0 £> 3 co N fa s o a s « ts o a fa U U d cS U CO cs u fa •a d cs o d toJO cs cs o •a cs fa a cs H d u a , o IB* "cs en 3 d CO cs O s fa <u u d , * •<-> CO CS <u u ON CN CO o C co" ^- o T - H CN CO CO jy u O a y CN a y CN d cs o CN CN CN CN CO CN CN T - H © ^r; LO T—1 T - H T - H LO Cs o O LO CN T - H CO CO ^ tU 13 co <u U S <! Is o o II V OH 0 CD a * LO o II V OH d cj CD fa O o crj ^ 8 •a o u II • i i CD CTJ a in ON <u •5 6J0 CU a OH w ON u •I 0 3 o a w NO CN] T - 1 CN co" co" o o r~~ oo o o co r-~ LO o o CO LO CN) CN) NO" T-T © T - H oo LO rH O CO CN cO I U C o u 0 3 O PH »H O c n .3 a PH u NO CO T - H CN NO" CN" © T " H O NO CN 00 CN 00 T " H CN r--oo -4-NO CN NO" © NO CN © T ^ NO CO CN NO CN oo o I P H o oo ai NO T - H T - H CN NO" CN" © T ^ c u a O O NO CN t~-CN LO CN CO NO 00 LO •3 * rt c n c n a 3 « « PH PH O 3 y a 6 i » c n PH PH «s c n c n 3 3 rt « PH PH O o 3 3 u U 3 a 8 +-» i «H 0 3 O PH PH OH LO CU O , o I S V CO 3 S3 CM ON CO CO C/3 a a a ON CN NO 00 t~- L O © o © T-H r>; q C N T-H^ C N r--" in NO" o © © d d d d T-H L O C O - T j -CN C N C N C N r--" t^-" LO" NO" © © o o co co co co d a a c C N CN CN t T - H C N CN T - H NO" OO" CO" r-" o o © ©' O CN ON O C N CN T - H O CN O T - H ( N cO cO o CU CM a a a a a a a cs d d d d u u c CM i y LO ON C O 00 T - H C N CN CN r--" oo" LO" © ' © ' © ' CN NO 00 NO t^" LO o o o T - H NO C N C N C N C N C N NO" ^J-" -st"" ^ " ©' © ©' o C N NO r-~ C N CN C N T - H NO" CO" CO" NO" © • © ' © ' O X CM l « © C O L O T—I © T - H CO CN T - H T - H © © © T - H © © © © CN -3- -*J- © © O 00 O Lfl - t O T - H C N T - H T - H r- L O C N C N T J - C N m CN 2 $ C N co r~- Tt- © NO C N co co C O © CO C N CN NO NO NO N O t^" CN T - H CD a o Z --f i , -f i , CO co c3 O 0 _ i _ +J -f LO CN U V o H Q o u + ,CD ,CD a *> 1 & cM CD U d o -t-i 3 O n cs CD >•» cn co CD CD CD CD CD O a o d cn •a CD 1H O CD d o IH CS CD CD CD cci cd CD U d o CO CO CD CD ,CD ,CD 3 O <D CD CD U d o d cs •3 CD )H o Z < < < 2 co too .3 u W3 0 O H co C/J C/J C/J C/j a a a a C/J C/J C/J C/J a a a a LO CO CO CO (—, a a a § oo L O C O L O C N 00 © L O L O NO © CN S O C N C O T - H T - H C N T H oo" so" so" so" t~~" NO" NO" NO" NO" T+" © © © © O © © © © © ' d © © o C O C O © p C N C N © C N © © NO T - H CN C N T - H T - H © T - H © T - H T - H © CO CO CO CO CO CO CO CO CO CO CO CO a a a G G G G G G G G G CO C N O ON L O NO T - H O C O C O NO © T - H cO C N T-H C N CN T H LO" ON" so" CN" so" LO" NO" LO" LO" r-~" o o © O o o O o o © © d © CN L O 00 © T - H ON © ON O C N C N oo T - H r-H C N C N o T - H T-H © © d T - H T - H d L O C N O so cO SO C N T - H NO 00 ON © C N CN © L O T - H T - H T - H T - H 00 C N C N T - H ^—t © 00 oo CN L O T - H CN CN T - H T - H OS G O 8 OS OS CO CO CU CU .CU ,cu CU CU OS CU u G 0 +J 3 O . a cu OS CU u G 0 G OS -3 cu u O Z < < < S bp 6J0 bp co CU o u .3 c« o U CO O N CS ^-3 CM o u c u I CO <u u « LO a G G LO Tt-NO NO o o o o r~- o CN T - H © T - J © LO o o CO CO T - H T - H T - H co" LO" LO" o o © ' NO 00 00 o © © T - H LO LO LO cs & CD G CD 0 H CD H CS 3 CD H " C J 3 C P a-^ -2 H3 C N cn Tt Ny .52 % CO ^ O & 4 CS CS CD g CD < ^ 8 -a a CS CD cs0 I H L2 CD O A )H 0 -t-T G cs u MG o z II + o Cs 00 Tr-ey C N CN 00 NO LO LO co a LO LO LO CO NO CN CN a CN? NO CN T - H V cs o 1 3 u cS T3 o <u A u b | ° g c j " d «8 3 a O PH u a o cs CO 3 §^  3 u «8 tJ O £ u 3 cs r j t o o u £ +H cn CD PH *H - > PH 0 5 S H H ns ns ns CN 00 NO T - H NO" LO" d d d p r-- p C N d T - H d o CO G CO G d ON L O NO d T - H T H CN" Tf" d d d O L O 00 00 d d d C O co C N ,_ C N o 00 L O CO CO CO a G G NO NO L O T - H co" d d d O oo T - H d d d CO CO CO G G G 00 NO T - H T - H T - H co" co" d d d O 00 T - H d d d r-- 00 C N 00 NO NO C N o NU •3 1) 1) l •3 quarl . quai quar quar +-» CO 13 G -d -a 1—< C N C O CL) o I -cS o CD NTT^ 9 3 i n 1 •» <D y CS M 1 3 CD + O C N 00 C N C N CO i NO L O L O C O a L O L O L O I C O NO C N T - H ( N J d Q> A ^ *-> o NO y ^ u CO Z O H ' 3 u 13 <D <1 Pi CD 13 CD 'S3 "3 P 3 o CD X cs O N o fa ' o ft 33 C J u O N X CS cB IPQ * 2 & i-i <u C W C v CJ *H fa cn cn cn o o © o NO cn cn cn cn CO CO T - H O o CN a a a o o c^  a a a a G G O o © V 0> o © © o LO C N NO p CN LO co CN NO ON T - H 00 T - H T - H T - H o o T - H T - H T - H T - H T - H T - H T - H T - H CO Tfa C N C N 00 CN o CO 00 r-- 00 LO C N C N T - H o o O o o o o © © o o © ^ ' T - H CN CO CO LO o o p 00 oo 00 LO o ^r; T"H T - H T - H o © ' o T-H T - H T - H © ' o o T - H C N T-H T - H T - H T-H O o CO O O C N ns cn a cn G o o o o cn a cn a cn G cn G cn G cn G O O o o O o o © © o o V V V LO C N LO CO C N LO NO o CN T - H T-H T - H T - H o © ' o T - H T - H T - H T - H T - H C N CN CO C N C N ON r-- T - H o CN C N C N LO CO LO co co T - H O O O o © © O o o o © ' © T - H T - H T - H C N CO O CO TCt" T - H p CN C N p T - H LO T - H T - H © © © T - H ^ © o T - H CN CO LO 00 cO LO 00 CO LO 00 CO LO 00 co LO oo O CO NO o CO NO o CO NO o cO NO o CO ^0 NO CN co NO C N co NO C N co N O C N CO NO CN CO fa fa fa fa fa :=> cu fa cn O CJ fa cn O fa cn O CL) fa cn O :=l CL) fa cn O <5 P H P H < P H P H < P H P H <l P H P H <! P H P H fa o fa P H & 0 CS JH -a o U h o O u < CL) CU C O CU "3 o U cu i n y 1 ^ c u 3 ° |U C N cc CS 0 H CO { H 00 CN C N CO (~~, o o o o c—> o d d d d dCN CN C N NO T H CN" C N co" LO r--" CN" CN" CN" T - H o o O d T - H T - H C N r~; o O NO CN r-C N C N su CN ns ns C N CO G su o ns ns o d d 00 00 LO cO C N C N C N LO NO" NO" CN" oo" r--" C N " d d d d o T - H p cO T - H O cO LO T - H T - H C N T - H T - H T - H C N T - H T - H O T - H T - H LO LO LO LO LO LO LO JH CS 3 CD n CS 3 c tr tr ^ -B -fl C N co T^T cs CD cr Q H i l l H cs y CD CS c r ^ -s -S CN TJ-CS CD cS co o CS CL) co T3 G cs CD" CS + NO LTT 06 LO, CO a L  C N a <U T - H CS c r © 5 A y. a co CD i ° o ^ O co o .23 o 5? CD 8P g> T3 CD G cS cG ^ 3 w .§> 6 co cs o" & 11 S co - a 3 u CD -a CD T3 E — CS CS 3 O H a c u a , o I* cn rt , & 1 C U a • cn a a , o is* cn « , C u a U ns ns ns ns ns ns LO NO CO CO CN cO T - H CN CN CN co t--" co" 00" oo" o O O o o © ' p 00 CN O CN NO 00 o T - H o T - H T - H T - H T - H cn co CO CO CO CO G G G G G G CN NO CN CN CN O T - H T - H CN T - H T - H co" co" CN LO" co" LO" O O o o © © p NO r - f - O CN 00 LO o o o T - H O © T - H LO ^_ o CN C \ 00 ,_ O CN o CN 00 C \ CN CN oo T - H CN CO { H CN NO CO Q T - H O o O O O o O O T - H NO 00 00 00 00 LO LO LO NO LO CN NO co" O " Ttf o o T - H o T - H O T - H CN CN p CN 00 T - H CN CN CN r-- CO ( H CN CO ( H o O O o o o © © ' r-~ T - H 00 00 o LO NO oo T - H NO CN" I-"" co" p " oo" o o T - H T - H o O CO o CO p r-- LO CN CN CN T - H CN CN LO O ,_ CCN T - H CO o T - H LO LO NO NO LO LO NO U CU CU cu cu m 1 u CS . qua quai quai HW cn G u -3 C N co cS MH o H G "2 -5 C N cO TJ" T3 ts cS rCS o o P L , o , 1 cu M cs '1 cs 1-3 cS G cr OS J-J 0 .<3 CT o OJO CU -i-) CS cu CU u ,0 +3 CU CO *H G ,CU , . C HH -a cu \< & u SO IS s 0 3 O N s I U 1 S * Pi CU I'-3 t-o IS S U § °^ | y O N 0 3 S3 LO CN o o d d Tt- eO c o o d d o r NO NO d <d> d N O CNI T - H o o o o d d o C N oo H d d Tt"" co" CN" o d d O NO LO T - " d d d o LO <u u CS -t-l o 1H PH LO LO LO rS ~H 1 1 oS cr CU m u CS M3 CN co cr1 cr o CU CO T 3 + CO OO, a 00 00 NO^ cO a NO CO LO^ CN a OS H « i T-H CU .g u •a O LO A V a CS -4-J O co «" JH 43 O <4H (J "3 y cu c a cu CO JH 3 ,CU • M-H T J CU <; pe. * ST" ml ^ ' I §° 8 CO -s & 11 a co -g cu u |"0 CU O C3 3 s £ o "3 x S «* , &' 3 u O PH 3 £ o 13 c e 3 rt , &1 3 u r--o T - H O T - H T - H f>j T T " cn" ^-" o d d o oo in oo T-! o o o CN o CO CO SO T - H O T - H c-T T - T of o d d r-- cn so o d d oo -s t - C N C N C N C N 0 0 so co co ,—, 3 3 • o CN r- o CN C N T - H o d d o m so cn T - H d d d CN i n 3 3 o d cn t-- C N H H d CN" CN" T - T o d d o m NO C N T H d d d t C N t i n i n NO i l l cS CT .3 i) - M O JH PH CU m 3 cr cr "3 -TJ 3 ft CN cn -a o I 1) PT « SH ^ cr si ^ CJ ,_( •a O -3 fj CO o .23 & I T3 <u <! pi J3 CS 1=3 ^ t! o 1—1 | l U CU o~-to i n 3 O N cs ts , o IS fa U u C , c s W CO c s , a u cs U O N ns ns ns ns ns su ns ns ns C N T f cO o T " H T - H CN T - H T - H T - H T - H CN CO L O " t--" I--" N O " r-~" N O " 00" o o O o O o o O o T - H O T - H C N t-- O LO T - H o T - H T - H T - H O o T - H T - H T - H ns ns ns ns ns ns ns ns ns 00 o o 00 CO 00 00 T - H CN cO CN T - H T - H T - H CN CO N O " N O " N O " N O " r--" o O o o o o o o o T - H T - H LO O CN p r-- o o CO T - H ^ T - H o T - H T - H T - H T - H LO T - H T - H LO T - H T - H LO T - H O LO T - H CN LO T - H o LO T - H T - H LO T - H O LO T - H LO T - H T - H LO T - H o LO T - H i I i ^ IH I n gn cr Sr cr fa "3 T 3 J3 CO C u - H J T H CN C O -t fa crj •a o •f U 3 o H o 0 cj 3 0 CU CU CU m fa CO § H cr cr fa "3 T3 J3 to O fa ^3 TH CN co Oj 3 cr CU fa CS 3 . fa CU i CU m cl 3 cr CU CU CU fa Pi ?r u- cr tr p "5 t) j j co G fa 13 T - H CN co ^t-CU fa ci -G o fa cS u CU cj tfa1 fa CU fa 3 0 •3 o •4 cu CO T3 C cS CU 3 c j .a ^ in pi o ^ A V, fa a cS M 3 ; ? O co u" ?r fa H = , 0 cs <-fa cj T3 c u 11 a co -TH Z 0H - d 3 cu U T3 CU 3 • faV CH 03 o « S rt H cy J3 J CN CM , o IS , « u <u U IT) OS CO T - H O CO CN f—N CO CO CO CO CO CO CO CO CO a O G G G G G G G G G G d d T - H CN NO CN LO CN O CN 00 LO 00 LO T - H CN T—» d T - H T - H CN H CN T H T H T H LO" co" CN" LO" CN" NO" LO" NO" NO" LO" d T - H d d d d d d d d d d d o t-- O o p T - H 00 NO O T * H CN 00 cO O CN p 00 T " H O T - H CN" T - H T - H d d d T - H T—1 d d T - H H d T - H d o IS <H u u a lu TH C O 54 U u LO ON CO G o.oo-CO G 0.09 CO G co G CO G CO G CO G CO G CO G CO G CO G T - H CN LO LO CN 00 CN NO O NO T " H CO T H T " H T H CN CO CN CN T - H CN T H co" NO" CN" ^J-" CN" NO" LO" LO" t-~" LO" LO" co" d T - H d o O o d d d d d d d p t - - O T H O CN LO 00 p CN CN CN p ON o H d T H CN T - H d d d d T - H d T - H d o T - H CN CN r-00 CO O T H T - H CO 00 T - H NO NO LO LO r-CN cO LO T + oo oo o O T H LO CN O T H CN T - H CN 00 NO T H o 8 2 J H o £ 2 >H U o a o a OS <U u G O G OS •s cu cu CU cu CO CO cn cs 5? CNI CNJ CNS •a G O u G 0 G G O a C U C J 5T CS CS CS cu <u C U u u o G G G 0 0 0 - M - M 4-> CO CO CO c s CS 03 CU f—i , ^ OJ 4-> ±J , flj -4-J -4-J -4—1 •a G 0 a OS C U o G O G OS -a cu 1-1 G ^ a * OS OS CU cu u o G 0 CO CO OS OS u CU cu u 0 3 cr CO CN O A u 0 a 2  S 'o * So .1 * -s <u II M co CJ .y +-» CO CO I 3 o <! H d ,J2 CO a CO a CO a T - H T - H CN CN CN N O " N O " L O " O o o p T - H T - H T - H T - H CO CO CO a a a ^- CN o CN C N C O L O " Tt"" O O o p T - H T - H p T - H r-- O r-~-00 C N C N C N CO CO CO G a a L O C N T " H CN CN Th" Th" o o © o 00 O © o T - H T - H CO CO CO G Cl G 00 N O CN T - H CN C O Tt"" "sf" o O O p 00 O , - H o T - H co T - H r-- Tf N O N O L O L O 00 N O CO 60 ' 3 .a C U G O cu 60 c « u ,o T 3 cu CO 3 \< & 0 60 cu +J cS u G cu u L O co CN C U CJ T3 3 3 cS c r CU 03 c r t r « "2 J G co G u ~ T - H CN CO Tj-CU m u CS G . I -8 CA) Q O |£ 03 ^3 ,£) LO CN o X •O C 3 8 tS co O u is MH iT T3 1-1 u ft I u e -3 S 3 rO CN HO CO s o •43 H^ u C o U > 3 a 3 0 3 3 3 Q c/j a •x- * i ^ O ^ ^ C N j O L O g c O o o i - o o o ^ ^ p t~- T-H Tt- CO o <=> o - S3 LO ^ ^ ^ CN ^ o 00 o LO CN ON . 00 ON [-- O ^ CN ^ J O vO cO co co O CN CN CN P ^ CO t ^ N N-" CN r~~ LO LO CN NO 00 6 0 X £ C/3 OJO NO tN CS o a u CJ U 3 cs o • +•> X CS CU «H T3 CS cS x O s ox cs cs x > c 3 x 3 a 2 CU CN 3 CS H s (U a , o I* w«H CS x 3 & a X i«2 0 o a , o " « x 3 * , 3 a t o ©N o U t o ON O a -? 3 o x fl 3 « , CO C N Lo" o C N C N O CN C N T - H CN LO CN o o oo o o 3 3 c« cs u u « £ > •2 u -3 c n P Q < oo -rd- NO NO T h C N C N T - H o " o " -«sr T - H ^ H O O CN p NO T - H LO LO CN r-- C N Td-C N C N CN 1^-C N H M t CU CU CU cu u ^ ^ A (4 W) 3 3 3 3 B a oao C/3 co T - H © LO LO NO CN L O C O CN T - H co" CN" T - H © © © O NO T t - O T - H o © O CN L O C N C N T - H C N C N C N o o o 3 CS CU P bio C/J P Q 3 cs <u a u cS 3 C/J o r-- r -00 CN" oo" r~-" o © © o r— LO co T " H CN C N C N LO N O CO T - H co -Hr 00 o LO co CN T - H Tt"" CN" T - H " © © © p CO T - H T - H o © 00 LO T - H C N CO C N CO C N T-H CN m <U U tU U • f i l ! ftil cs cs cs cs 3 3 3 3 oooo o o © II V OH •X-# o © ' 1 1 G v « + o CL) ^ • NO o LO o o II V OH ON NO CJ o jo 3 O CO CO cu oo 2 ^ ? •3 LO i t § 0 LPT -3 v cs ^ tcO .. CS CO T3 <U o & S S 3 > co (J M . E , <U ' w CS ^ co . cS j ft 3 2 o c / cS _Q cj c Cu) a o CO 3 3 u CO O P H s <u a o & CO 3 3 CJ s i « H PH LO ON o u LO ON O • 3 J ° " 3 u HJ L O co" d O co O C N CN] T - H o o T ^ c\i 2 C N s a cs CJ u a a SJD.O C O PH CJ > O [— C N T - H CN! N O Tt T - T L O " o f d d d O N O O N o T - H O T - H T - H O N T - H T ^ CM cO CN] O GO CN] NO TH CN T-H CD O N C~~ ^ C N CN TH TH T - H cN CO T H -CJ CJ CJ CJ N 1 1 1 1 <U C<S c3 Ctf W) 3 3 3 3 8 O OOO C/3 CN L O o oo T - H d 2 O N N O o o r - oo CO CN 3 «s CJ a SJD % C/3 PH CO CJ 3 CJ a u > o T H T N O I~-Tt ui L O " N O " o d d O CN N O C N LO T H - CN] CO co Teh Teh T J - N O LO C N ^ LO N O co o o 00 TH .8,9. 04,3 .1,2. N O LO co C N C N " .8,9. 04,3 .1,2. TH N O " Tf" C N d <=> d ° d o d d 00 TH N O rH TH" O oo N O TH oo co CN CO CN co CN T H CN CO Tt g u t u v eg 03 ^ 3 3 3 3 = oooo o o d II V O H •* # T - H o d II V O H * -X-L O " o d II V O H + O N a 00 r--r-» •4 oo - 3 co CL) CO U N O w CN cO N O CN V u CU co CN 1H cn 1) -3 CS > . i f '8 a ^ "o cu n L2 CJ O 3 u CU w rt Si cs jd cj oo CN en CI s s© o t o ON cS CO 3 CJ d O OR 3 V •i Pos z NO NO T - H CN T - H CO CN T - H co" CN T - H d d d d o o CN t-~- LO O T - H d T - H T - H d d T - H T - H LO T - H 00 NO CN T - H CO CN CN co CN "-T LO LO CN co" o co" ^ " " d d d d p O CN LO O T - H d T - H T - H CN cO CO CN CN T - H r CN T - H CN T - H r 3 3 <rS eS -a "-a Cu) <U a a c/5 P Q CJ > O 3 rH CN CO r t CJ Cu) CJ CJ CN % % u C4 cS cs ts W) 3 3 3 3 cs a a OO C/3 LO C N " d CN d CN CN CN d o CN 3 CJ -I c/j C Q 3 CS •S CJ a Cu) cS 3 C/N T - H NO CO C N T - H LO" C N " C N " d d d O CO LO T - H T - H d d T - H 00 LO CO CO -3- CO C N NO C N C N cO co" LO" LO" d d d O 00 CO H d 00 LO CO CO C N co C N T-H CN fO CJ CJ CJ CJ nil (S (S (S cS 3 3 3 3 oooo CN NO CO CN CS o CN CJ CJ s cs CJ •i cn cs U . f a .fi "d a CS cn O 0 W) cs cs o cu c a CS cn O +-» U . 2 u T3 • fa • f a a CN fan CS H u a o X 3 cs PH O s c j cn O PH CS X 13 PH CU § a CU o CN -cj-" O C N C N LO oo CN © O O CN 3 3 CS CS -3 3 U CU a a cu Ej ^ CJ CA) PH > O CN ON £ NO" "f LO" © ° o O t~~ T - H T t " T - H CN CN CN oo o r-- co T - H CO CN CN r-: C N - * ~ o" <"! • o o p T - H far ^ T - H THT T - H O C N C O r~- C N CN T - H CN T - H H cN flt CU CU CU CU 4> 03 Ctf W> 3 3 3 3 B aaaa C/3 CN o o NO O r--N O CN O o O «r T - H O 00 3 cs c j CU > bio C/J PH 3 cs u a CJ cS 3 o o © II V O, * * * T - H O O II V OH cn S c« 03 CU * LO o © • II V OH' CU CJ cn O fa CU c*° fa . O cu oS c u > fa Q CS J cl a o 7s c n 3 CM OH O 3 u c n O PH 73 c n 3 pa aen o aen 3 G CU O s> 1 <u u OH u o P Q DC C/3 NO o O NO 00 CN o o CN 3 3 CM cM cu U « % > 8 u -o C O P Q < SrH C N CO T f U U C N '3 "3 '3 '3 CJ CM CM CM CM W ) 3 3 3 3 CM +-» a a OO C/3 3 CM CU a - % two c ^ CU CO CU CN ON NO NO CN 00 LO" LO" oo" co" CD o o O p 00 00 O NO T - H T - H T - H CN T - H T - H CO LO LO CO CN CN CN CN oq cO NO T - H cO CO CO T - H T - H co" CN co" T - H © o O O O p CN O p O T - H © , - ' T 1 T - H o CN T - H O CN CN CN CN T - H NO r--3 CM CU a C O C O C O C O LO" r--" r--" o © o o co r-~ r-00 CO NO co oo oo CN CN CN © ' o © ' O NO NO NO T - H O O O 00 CN T - H CN CN CN co CN r--cu C/i P Q < T-H C N cO -cl-eg CU CJ CU CU ^ • f i - f i - f i l +^ c3 c$ <S> 3 3 3 3 sj OOOO o o © II V P H * -* T - H o © II V P H * LO" o o II V O H co G CM CU a * CO CN co 'co o 2 V •w §> -13 + a ic •W CO CN cO^ CNT a h-5 CM . CU fccO CM n L2 cu 3 u CU • 9 '-a ^ Q c « J O c j 3 c y oiD O u +-» CO <U £ O +-» CO .3 a tH CJ +-» CO S3 C O a <H CU C J C cs y • +-» CO 58 U CS CO 3 cS a o 3 a • co M . 2 rt 3 .a « 3 co co 3 3 -P l - H C3 ON CN cS H CU y u o +•> co • i H y ^ t! S co 3 <8 ^ •o « 3 cS P H W •3 y +-» CO 3 eS fk W 2 -o •3 3 •3 CO a « CS P H •3 y CO 3 I y 9 ON P i o ON P i o LO ON Pi O U LO ON P H o 3 ^ co d CO co O N 00 d L O N O CN T H H CN O N " 00" H d d p C N 00 oo C O C O CN L O C O O N (N C O co CN * C N oo C N CN C N 00 © " r-" N O " T - H d d O N O co T " H C O CN CN C O L O C O CN C O co C O * •X-L O N O O N 00 00 o" p" oo" T - H T - H d O C N O N L O T - H CN CN CN CN L O CN N O C O Tt- Tf C O O r--00 O N " oo" r-" d d d p r-- L O co T - H CN CN CN L O N O C O T H C O Tt- T + T H CN CO Ji y 1-H Ji € cS s (S € c« 6 cs 3 3 3 3a o OO O N C O TH CJ <=> .9 o "> II I O H „ * .23 * co •x- 0 N O o C N N O C O a oo NO L O CN a L O V o d ; II V OH •X *~L2 L O ^ 0 -d d B 1 i co II 3 V ^ OH OS a ^ ^ H CJ 0 a OH pi o i=l .a o 3 u CJ 3 + OS - Q U a c s ts o a u o 3 c s u CO cS c s X 3 CS a o a u a CO O a •a a CS s a 3 co s •P4 o co c s 3 CS V c s CS .65 3 s o CO CS H o •a <u 00 o C N TCJ- L O T-H CN C N oi 1^ 00 oi C O ^ o f xr" L O " CN" NO" co" oo" C N " TH-" d d d d T-H d d d C O r-- p 00 C O r~- co L O oi oi T-H T-H oi cO T-H T-H T-H CD 1 C3 3 CN CD CD 0 if o/o/o/o/ CD 3 CO n 0 OH CD O CD *H G CD too O a CD t>J0 CD _ ba 60 S <J w W H G CD 0J CD o •3 CTj »H S CO 3 0O ! CTj PH cO co a 3 a cS 3 > L O C N L O L O NO CO CO o © NO CN CN r-; T - H r-^ CN co co co -rt1 00 CN co CO CN O CN t~-co CN O NO CN T-H oo o LO CN CO CN 00 CO C N O c O t~~ xt" C N CN NO CO C N _ J CN CO tN T t Tt o h; CO N c\ T ^ C N c d c N o d c O ' T f O L O C N C O Nor~-Nor~-oocNNO a 3 c s > 00 CO r-^  © T}- LO LO CO T)- CN T - H LO CO CO CN CO S£5 t NO CN LO _ T T - H 1 T — l 1 , oo u 3 * c s U '•a -i T - H C7N CN © LO r--CD .3 o CD .a •I* *T B B 4 ^ J t L O C N C O C N C O NO O T - H l - C N NO C O C O 00 •<* O O CN 00 T f O tN H T t T f vd tN I ] I I I I I T - H CD fa CS •a o -a CS u I i r 0 o CD o CD ,CD 3 : cs R O rS « 0 c q ^ Q ^ ^ B (H H w H cS 2 >•> CD ~ 0 1 3 VH CD ^ O H co U fa CS § CJ cS •1 s a CD cu 5b > & in , CD CS l z > cs _ Q Th cS U u C cs u •M CO CS CJ £> d c s X O c O X c s O X c s CJ CN X CJ J3 CJ •a a X X c s o C If c s U CN cn u c s H 4 ^ CS o CJ X 3 cs U u cS u ON CJ u C x a c s e « S CJ Is LO CJ ON cq .3 C J a oS .a U o cO o" r--o o d o CN CO d CO O CN T-H TH Ol CN o o d LO LO 00 O 00 T-H T-H CN CO Th CO CN 00 Th T-H CN LO a cs •5 C J u O cj C co cs Si - 3 cj u 0* ° 1-1 0 6 O ^ CQ cu C J 1-1 u C J £ Z Q o I 3^ C J OJO T3 a OS C J T3 C J CO 3 *H O O 8P A U 0 C J u a C J IH L w C J 1-1 o Z bD II £ to u ^ top C cl 3 -3 u O OS _ D CS ts o S O a c s a 3 ^ c s & o u •i *H u • f a CA cS U in ON u CS ft. ON LO o co oo" o o o r--o o oo CO LO oo" o " O T - H NO O LO T - H ^ H CN r- £ * \o 2 M PH +il PH .3 CU G° CTJ -G U o o o NO ON xh 00" o O O ON CN T - H T H / o LO oo" © o © p cO cO O O T - H CN T - H T H / NO ON ON LO 00 T - H LO cl CTJ CU ~B >H O cn 6^ tn Ji fa I O \ u fa cn CTJ S LO cu cu a u £ 0 cn cu fa u cu Z Q NO C O -3 CS o s CJ x CS U o cs L O O N L O d Tt-CN] co" d CO O TJ-O T-H T-H cs o u CJ 3 cs U • X cs V 4-1 u cs ft, L O O N o CNJ LO" d ON d O O CN T - H T-H O O S3 I 3 CD CTJ CD SP L2 CD CO 3 I Ha x V If cs O Tt- 00 C N C N C O L O CN T3 CD 0 a t-i 0 CD L O Pi CD — i TO CD CJ y o CD 3 Z Q S O 8P A CD 0 CD U 3 CD »H M H CD H : i cr £ o Z M II B CO s z ID ^ CD T J £P « 'I 4 cO cS o u co c s U cs ft, C N NO r--o o a a 00 O CN LO CN cO CN CN NO LO o o xh LO O T - H CN NO TJ-•Tvt CO r-" m" o o CN x f co" NO" o o 00 CO O CN NO G G NO CO CN CN NO" NO" O © G G 00 CO T - ! CN LO" NO" © © © CN T - H © So u a cs U CU L O C N NO o o NO CO xh LO* oo" o " © T - H LO NO CO CN NO" xt-" © O G G 00 T - H Tt CO NO" CO" © o G G LO 00 CN ^ CN" LO" o © T - H CN "^4" co oo" [--" o o G G t~~ NO CN CN LO" T*-" © ci CJ bD 8 CT! l.s CJ • « 13 u ce cS CJ fa P Q CN CO O LO O T—I CNj T—I T - H T - H r~; p p 00 LO T - H T - H T - H O T - H 00 t~- O CN o CU OJO CS co oo NO r— co CO CO t cO F~-CN LO 00 CN cO T4- O co CN r~~ T*- O O LO CO oo CN r— co r-~ CU .3 CJ • f a cs fa o •e cs CJ If cs U LO CN G CT! -a CU cn C3 CU fa o G G < ^ LO CN LO T3 CU cn CS CU fa CJ G LO CN G cn cn fa Ji -.o -S \ LO + N 1 LO CU _ - i fa cn £3 cr) J3 cu <-> H cu cn OS CU fa CJ CU LO CN G cn cu cn OS cu fa CJ G G cn -5 . o * LO <N LO T3 CU cn OS CU fa CJ G \ LO LO CN G cs •B <u fa o 6 LO ^ 1) CJ fa cn cn cn cn os cu cj J3 CJ CJ cu fa cu fan LO CN G cn ^ Z Q Q £ £ Z Q Q « U 3 o H LO G cS •3 * cu cn OS CU fa CJ G LO CN LO -a cu cn cn cu fa CJ G 03 U CU LO CN LO CN G cn •3 cu fa 0 a LO ba -a CJ CJ fa cn cn OS oS 03 G cu cu CJ fa u CU fa CJ CU Z Q Q G 1^ \< o o A fa o bJO cu fa OS CJ CU CJ G CU fa . C J G cn CJ LG O 3 Z tuO II B cn * Z CJ <^ > CU T-j bp cj ,03 - f a '•8 o 12 o 03 faO 00 cn a c s tJ o u CO S c s U "3 i n ON T - H C N uri Th d d G G o Th oo CN d d C N 00 O N O C N G G N O C N Th in r--" oo" d d CO CO G G T - H C N cn C N irf u-T d d 00 O O C N C N T—< C N T - H T - H T - H CS o «H CU o 3 CS O -*-» CO C8 V tH u c s ON C N T - H Th Th N O " in" d d oo Th O CO CO G G r~~ C N C N T-H Th" <n" d d TH 00 T-H O 00 o C N m oo oo" C N " d d G G C N CN cn cn Th" Th" d d m r~- o C N C N T - H C N Th C N C N m Th i N O Th cn Th C N oo o Th cn oo in Th TH N O CJ %P 8 co "1 0! l.s u •3 o 13 u 3 o u \& r 3 CJ CJ CS If c s -3 U m C N G OS m G OS * . o ^ CJ IH O m C N B in CJ CJ OS CJ u u G + m CN in 3 CO S3 os rG cj m CN G os •B CJ IH 0 -3 cj CO OS CJ n o CJ OS PH m C N G o! in G os CJ IH 0 cj cj £ H ^ Q Q £ OS CJ U O G OS PH -tH OS C O \ m + ^ m m C N G OS •5 CJ u O a 2^  CJ CJ os cj IH u CJ Z Q Q 3 ITS A u 0 CJ u G CJ IH Irt CO « £? ° o Z ?P II s Z U ^ H CJ T J '•3 i z o os _D ON CO a CS +•> u O a u U O 3 c s u i +-» x cs u u X> T 3 3 c s x O 3 6J0 cs x O CM X tH c s U X m CJ a CJ c s 3 CJ o ft 3 CU 3 ° cs a u r-rO 3 cs H a cs tt o u X 3 cs U CJ c s ft, in C N a CS tt o u V u 3 c s U X cs CJ u n CJ " c s ft, m C N CS +"> 3 •fH 8 CJ +•> o u ft 3 V £? cs a u t--CO 3 CO G CO G o d cn Tt- C N C N m" cn" cn" C N " d d d d NO o NO m d r , d d G G C N O d 00 o CN ON CN CN T-H d m" cn" CN" T-H d d d d C N 00 p m Th d d d 00 00 oo NO o m o T-H m C N G CS -s CJ n O a T J CJ co cn cj IH u G m C N "0 cj CO CS CJ u u G in + m C N m C N G CTJ •Q CJ 1-1 0 m G CO cS CJ u cj CJ o CJ 60 •I cs , u '-S l.a u •o o ?——H cs <j o Cj" - 3 CJ -M CO 3 | T 3 l< cr H o o 60 CJ cs u CJ u G CJ t-i M-H CJ H A. t-i o G cs U 13 0 Z Q Q cr | o Z 60 II « 'co * z CJ CJ -r) M CJ G -a CS • -H U O . ° * IZ H cs -Q CS o o CO s cs U cu "ST in ON cS tt o s u CU u a u X CS cu i-i PH CJ CS > m ON a C/3 0 cn a cn a ON °) 00 CN T - H T - H T - H NO" IT)" Xf" CD o o CD CO CN CD CD CN T - H o T - H T - H CD cn cn cn cn a a a a CN o NO LO CO CN T - H T - H NO" xf co" co" O CD CD CD CN CD r- ^0 T - H o T - H CD CD NO o r-- CN CN NO oo NO LO CN T - H CU CS CU a CJ .3 CU cs J3 U LO C N a CS •B CU u 0 B LO CU CJ CS CU G CS •3 LO C N a cS -3 LO C N LO C N cu H cn cS CU u CJ CU CS u 0 3 T3 CU cn cS cu u CJ cu o Z Q Q Non-Referred* & ReFerred* Alive 1 • 71.6 i • I 72.4 • i • 77.5 i • • 80.1 i i 1 84.0 Non-Referred* & ReFerred* Mean Age At Diagnosis 1 i • 58.1 i i I 57.6 i i I 57.5 • i i 56.6 i i I 57.8 Non-Referred* & ReFerred* Actual Cases 1 i I 1390 l i I 1446 i • I 1490 l i l 1339 i i • 1631 lUBt * 2 L . CD a: c o z * 2 k. £ Alive 1 • I 72.9 i i I 76.1 • • I 78.7 i • • 82.4 i i • 86.7 Mean Age At Diagnosis ' • • 60.5 i i I 61.9 • i ' 61.3 • • • 60.0 • • • 61.5 Actual Cases 1 i • CM CO CO i i I •<* CO co i i I o> CM CO i i • co o> i • I o % Post menopaus al** 72.0 61.8 67.7 66.9 68.1 56.5 60.0 62.2 70.1 55.2 59.7 62.5 68.6 54.6 55.6 61.4 67.7 63.6 54.5 65.0 % Pre-menopausal** 28.0 38.2 32.3 33.1 31.9 43.5 40.0 37.8 29.9 44.8 40.3 37.5 31.4 45.4 44.4 38.6 32.3 36.4 45.5 35.0 % Alive*** 82.2 66.2 26.2 71.2 85.2 63.4 31.3 71.2 87.0 71.4 43.3 77.2 86.9 76.1 46.0 79.7 91.8 78.0 49.4 83.0 Mean Age At Diagnosis 58.6 56.2 57.2 57.4 57.9 54.4 56.5 56.2 58.1 54.7 55.1 56.4 57.7 54.4 55.3 56.1 57.4 55.2 56.3 56.4 Pathologic al TNM * at re (0 CN 00 Total* CM CO Total* CM co Total* CM CO Total* CM co Total* "35 0 re Q Year 1991 1992 1993 1994 1995 T t T f «* 1 1 • CO 1 1 1 o> 1 1 1 CN 1 1 i LO 1 1 • CO o 00 00 CD CD CD 00 c o x" E '8 CO 'E Z •D 1-< o 00 T f co Tr- T — < | 1 1 • 1^  1 1 1 CO 1 1 1 1 1 i CO 1 1 i CO 1^  o o m LO LO IT) LO LO sto Bi >o'xw >. ro Q CO CO CD 00 o CM —H> o , 1 i T t 1 1 1 CO 1 1 1 CO 1 1 • o , , , CD T T 00 If) CM CO 00 CD CO i Date > T~ T - T _ i Date > o o X" CO E 8 LO CO LO CD CO o o CN o. 1 1 1 CD 1 1 1 d 1 1 1 CO 1 1 1 CN 1 1 1 CN T f CO ro b Z o oo CD CO CD CD 00 P 0 t o cu o CJ) cu o A § X" X CN CN CN co CO t o cu CO S ro 1 1 • d 1 1 1 CD 1 1 1 d 1 1 1 d 1 1 • CD d CM Z CO LO CO CO If) CO erred CE 2" XO.1,21 /here agi cu 0 B n T3 1 1 • CO CO T* 1 1 1 TT If) CN 1 1 1 00 co 1 1 i © LO CO 1 1 • If) CD 3151 xcludes ssion Ds Procedui MX,0; 3 D 'pre' an L U E o ro • D 0 cu x" 0 0 < z t o LO CN CO Ti- LO oo CD T f O) CD CN CN o 1^  CO co co CD CD a) CO > CCA n, Sp CO 1-o x<50 CD LO oo CN ai CJ) ai T* CN T— T— co co CN d co d CN CO m o T3 CO LO LO CO CD LO CD CO r-- CO CO CO CD LO LO CO CD LO CO CO CO 8 o |c X" ro X cu cu i Prior "3 Z E 0 D ) CO CO i Prior C L o " ro cu 10 8 ro Q o X X" z cases l l j l o c Q pi cases ary (50 cu >> CO LO d co T f CD d LO d CM d T ~ LO CD CO CO CO CN CO CN co o CN CO Ti-ed CN CN d Tf— CO ary (50 eatm c o .92 X,0o CN T > P CO TJ- T T co co T t co co CN co co co co T t T f co co T t co co co iple prim or RT Tr d cases Vi O'XN dx entei "6 H ro t o 3 E ra o co CO CD LO co CO CM CO q o T t T. 00 O o CO CO CN CO r- ro o CO cu CO 8 o X" B t o CN CN CO LO T t CD CO CD CN T— CO LO CD CO CD CO CN CO i f ) cu 8 ro CD oo CD 00 cx> oo CD 00 CD CD oo CD CD CD CD CD CD CD CD 00 c o Q c cu t o in c o CO z ro * • o cu T— o A —^- t o o c M cu t o a> x" 0 E CO H Tf' CO LO o> LO co r-Tt' LO T t CO CD CO N-T t CO CD CO 00 CD o LO LO CO d o LO o d 00 If) co CO o c c o ro Q lavaila eludes II CN O o c LO LO LO LO LO LO LO LO LO LO LO LO LO LO LO If) LO LO LO If) If) <C 3 X L U X" >; cu E c cu CO t o o CO B cu </> ro x" t o < t o o" case o O z case 0 Tf case E * 4t « * 4t l i i II 0 CO p I ra CO ra CO ra ra ra CO < II B T— CN CO *-> T— CN *-• T— CN T— CN CO X — CN co ra CD S 0 o o o O o es Cas w L L tatus (R 1- 1- 1- 1- O K- es Cas o '81 II (0 0 0 CO B i z tatus (R "O T3 O t o co >< tatus (R —I r— ro c pausal S CD co CD o Grar 1 Excli v Pati O logical <rl c pausal S CD CD CD CD o Grar ro CZ 0 o < pausal S CD CD T— CD CD o CN * - Ovei c ro £ ro 0. II o c 0 { Tj" Chapter 5 DISCUSSION Overview What emerges from the analysis of insulin and insulin-related risk factors is that some, but not all such factors are related to breast-cancer mortality in this cohort. The study hypothesis was that all hyperinsulinemia-related factors would be associated with higher breast-cancer mortality. These factors are WHR, BMI, dietary fat, carbohydrate and energy intake, glycemic load, physical inactivity, serum insulin, C-peptide and C-peptide-to-fmctosamine ratio. Each of these, however, has a different strength of association with hyperinsulinemia, and each may be subject to different patterns of confounding. Also, each factor was measured in a different way, with potentially different amounts of error. Finally, each factor may have a different magnitude of risk and some may be too modest to detect, given the sample size. With many factors operating at once, modest associations may be overshadowed by stronger risk factors. Overall, the results indicate that elevated serum insulin levels were weakly associated with post-menopausal breast-cancer mortality, but C-peptide, fmctosamine, C-peptide-to-fmctosamine ratio and SHBG showed no association with mortality. WHR at diagnosis was a powerful indicator of prognosis for post-menopausal ER-positive women, though the 2-year change data suggest that WHR is difficult to reduce, as might be expected in light of data on 143 heritability of WHR (91, 92). Increased protein intake at diagnosis was found to be inversely associated with mortality for all women, while increased fat intake at diagnosis was associated with higher risk of mortality only for pre-menopausal women. Total energy intake at diagnosis did not show a consistent significant association with mortality. Physical activity at diagnosis was not found to be a prognostic factor in this cohort, though it may have a relatively modest association which was not detected in this sample. The following sections will discuss the significant findings in the larger context of other research, as well as discuss possible explanations for null findings. External and internal validity will be examined, and some proposed explanatory mechanisms will be offered. Finally, the evidence to date supporting hyperinsulinemia as a causal factor in breast-cancer mortality will be presented. Lifestyle Factors at Diagnosis Body Si%e Body size at diagnosis, as measured by weight and BMI, showed no association with mortality in this study, although the point estimates suggest that both high and low BMI can increase risk, while maintaining an average BMI seems best. This result does not support the study hypothesis that increased weight or BMI would be associated with increased bresat-cancer mortality. The U-shaped or J-shaped pattern for BMI has been observed in other studies (90). In some cases, this result could indicate reverse causation, where more advanced disease may have already begun to cause wasting in the underweight group. This is unlikely 144 here, however, because models which excluded those who died in the first year after diagnosis did not produce different results, and the women who died in the first two years did not have a different mean BMI at diagnosis from the rest of the cohort (data not shown). In addition, reverse causation is not likely to be a significant factor because of the relatively long time between measurement of BMI and the outcome, on average 4.9 years. The inverted J pattern observed in the pre-menopausal women is interesting, and although overlapping CPs mean this may be a chance observation, the idea of a reversal at menopause of the relationship between BMI and endogenous estrogen concentrations suggested by Potishman et al. (140) would also fit the data. Low and high BMI may be associated with reduced ovulation, and the resulting protection could mask the risks of BMI extremes until menopause. ER-positive status, more common in post-menopausal breast-cancer cases, defines a distinct subtype of breast cancer with different relations to risk factors (32), and might also account for observed differences in risk by menopausal status. Overall, body size variables were not related to either breast-cancer or all-cause mortality and therefore the results do not support the study hypothesis. Though not statistically significant, the findings on weight 5 years before diagnosis are consistent with data that adult weight gain increases both risk of breast cancer and risk of breast cancer mortality (26, 178, 179). Body Shape In this cohort, we showed that WHR was strongly and positively related to risk of dying from breast cancer, and from any cause, for post-menopausal but not for pre-menopausal breast cancer patients, and the association was restricted to women with ER-positive tumours. 145 The association may appear stronger in post-menopausal women because the effects of estrogen have diminished. Furthermore, the results indicated that the difference in risk between pre- and post-menopausal women was not solely due to age. Finally, the results suggest that greater risk may be associated with elevated WHR for women with larger tumours. These data support the study hypothesis, that elevated WHR is associated with increased breast-cancer mortality. The results suggest that WHR can pose a risk for slim and normal women as well as for obese women. The development of the android or abdominal fat distribution pattern ("apple" shape vs. "pear" shape) is independent of BMI. Studies have shown insulin resistance to also be independent of BMI (12). Post-menopausal women, then, may be at risk from an increasing waistline even if their BMI stays in the normal range. The post-menopausal breast-cancer mortality risk associated with the 2 n d WHR quartile (0.756-0.800) compared with the 1st quartile (RR, 2.7; 95% CI, 1.1-6.9) suggests that the accepted cut-point of 0.8 for increased health risk in women may be too high, if risk is already elevated at a WHR value of 0.756. In a prospective study of pre-menopausal women, Hollman et al. (56) found insulin resistance and hyperinsulinemia to be greater with increasing WHR. Hyperinsulinemia, in turn, has been associated with a 3.3-fold increased risk of mortality (95% CI, 1.5-7.0) for breast-cancer patients in the highest quintile of insulin levels compared with the lowest quintile (10). Insulin, a known growth factor, is a biologically plausible agent in breast carcinogenesis and mortality, and WHR is a marker for excess insulin. 146 Some researchers have observed a direct association between WHR and breast-cancer risk while others have observed no association. In a large population-based case-control study by Friedenreich et al, the odds ratio for the highest quartile of WHR compared with the lowest was 1.43 (95% CI, 1.07-1.93) for post-menopausal women (53). In the New York University Women's Health Study, among pre-menopausal women, the relative risk of breast cancer increased to 1.72 (95% CI, 1.0-3.1) in the upper quartile of WHR (52), but an analysis of the Iowa Women's Health Study showed no additional risk associated with high WHR in post-menopausal women, after controlling for BMI (179). Few studies have examined the relationship between WHR and breast-cancer mortality. Folsom et al. (54) reported a significant trend for increasing mortality from all cancers with increasing WHR, and a positive association of WHR with breast-cancer mortality (90), but they did not report a stratified analysis on menopausal status. The difference observed in WHR relative risk between pre- and post-menopausal women may indicate important metabolic consequences of sex hormone changes during menopause, and it is consistent with the different distributions of receptor-defined subsets of breast cancer, as shown in studies of breast-cancer incidence (32, 33). Stratification on ER status showed that the association in post-menopausal women was restricted to those with ER-positive tumours, and the lack of association in pre-menopausal women was not related to ER subtype. Progesterone receptor (PR) status cannot be ruled out as a potential confounder, as these data were not collected during the time the cohort was assembled. Differences in relative risk post-menopause might also be due to the reduction of estrogen's effects, allowing more moderate associations to be observed. 147 Pre-menopausal women are often treated with chemotherapy that induces menopause, reducing estrogen exposure, so most of them were iatrogenically post-menopausal during the period of follow-up. The observed relationship of WHR with breast cancer mortality for those post-menopausal at diagnosis may reflect the consequences of long-term exposure to excess insulin, whether through genetic predisposition to insulin resistance, or environmental influences like physical inactivity combined with a high-energy diet, or to multiple factors. At menopause, decreasing sex steroid levels may result in adverse metabolic changes that, over time, affect survival. Insulin is one of the major lipid accumulating hormones (cortisol is the other) which act through stimulation of lipoprotein lipase (LPL) expression (98). These effects are countered by sex steroids and growth hormone (180). A significant drop in estrogen, as happens at menopause, including chemotherapy-induced menopause, would change the balance and could have adverse effects on the individual's lipid profile, especially if that individual was also msulin resistant, with associated dyslipidemia. The effects may be more pronounced in visceral tissue because it has a higher density of steroid hormone receptors than other fat depots (98). A high WHR might become a threat to survival only when combined with the post-menopausal metabolic environment over some period of time. The observed association of WHR relative risk with tumour size, but not with nodal status suggests that the underlying metabolic defect may be related to tumour growth but not spread to the nodes. This is consistent with insulin's known capabilities as a growth factor. ER-positivity had a significant impact on the association of WHR with post-menopausal breast-cancer mortality, providing evidence that epidemiological risk factors for breast-cancer mortality differ by ER status. This is consistent with other reports of breast-cancer incidence and ER tumour subtype (32). 148 The stronger association for family history-negative women than for family history-positive women is not consistent with other reports, where an interaction of family history and WHR was shown to increase risk only in family history-positive women and for progesterone-negative tumours (92). Since the current study did not have PR status available for analysis, the apparent contradiction may be explained by this potentially important uncontrolled factor. Physical Activity The data on leisure-time physical activity do not support the study hypothesis in that physical activity was not inversely related to mortality from breast cancer. There are a number of possible explanations for these null findings. The possibly beneficial association observed for frequent gardening also may have several explanations. Null findings may be explained by inadequate sample size, by narrow range of variation in the study population, by lack of accuracy in measurement, by bias in the design, or a combination of factors. Finally, null findings may represent a true lack of association. In this study, sample size is adequate for effects larger than 2-fold, but the relative risks associated with physical activity may be smaller. Luoto et al. reported no statistically significant association of leisure-time physical activity with breast-cancer incidence in a Finnish cohort of 30,548 women (48), while McTiernan et al., in a case-control study, found a slightiy decreased risk of breast cancer in women who exercised more than 1.5 hours per week or engaged in at least some high-intensity physical exercise (OR, 0.7; 95% CI, 0.4-1.1) (46). Moore et al. reported a relative risk of 0.92 (95% CI, 0.80-1.05) for women in the highest level of physical activity at baseline, in the Iowa Women's Health Study (181). As is apparent from an inspection of the frequencies in Table 12, the majority of the study population was sedentary, 149 so the limited range in activity levels may have contributed to the null findings. Recruiting from the Vancouver Cancer Centre, a largely urban population, may have contributed to limited variation. The lack of occupational physical activity data could also weaken any observed relationship by not including this important component of daily activity. A true null finding in a female cohort would not be entirely surprising however, if WHR is related to mortality, because Trichopoulou et al. (163) recently reported that physical activity was not an independent predictor of WHR in women, although it is in men. Also, if the risk of mortality is mediated by insulin resistance, the results of the Oslo Trial (15) are relevant, in that exercise alone did not significantly improve insulin resistance while diet alone, or diet and exercise combined did result in improvement. Gardening activity was inversely associated with all-cause mortality when performed more than once a week. This result may be explained by the physical activity of gardening, but could also be related to other aspects of gardening, such as being outdoors, working with growing tilings, expressing creativity, and planting sometliing with the expectation of seeing it grow over time. A Canadian qualitative pilot project explored the personal meanings attributed to gardens and gardening by people living with a diagnosis of breast cancer (182), and the participants described gardening as restorative, absorbing and as providing a vehicle to reflect about meaning and purpose in life. In addition, having a garden to work in (that is, living in a house or townhouse with a garden) indicates a relatively high socioeconomic status, which has been related to better breast cancer survival (183). 150 In summary, estimated level of leisure-time physical activity at diagnosis was not sufficient on its own to confer protection from breast-cancer mortality in this cohort. Macronutrient and Total Energy Intake Dietary intake is an area where risk factors for mortality from breast cancer may differ from those for developing breast cancer. The importance of adequate nutrition during treatment and recovery suggests a possible U-shaped or even inverse relationship between energy intake and risk of mortality. Although only marginally significant, this was observed, with the strongest association seen for dietary protein. This association might be partly explained by greater physical activity, but adjustment for activity variables did not appreciably alter the result. The activity variables in this study, however, captured relatively broad categories and may not provide sufficient detail to adequately control confounding in this case. Fat, and particularly saturated fat intake was associated with a 2-fold increased risk of dying, in agreement with the study hypothesis, while carbohydrate intake was unrelated to risk of dying. The lack of a significant result for carbohydrate intake does not support the study hypothesis, but it is not Surprising, because the study lacks the power to detect differences below an RR of 2.0, and the carbohydrate point estimates are probably lower, in the 1.5 range. Future studies would require approximately 1400 participants, or more than twice as many as the current study to achieve 80% power for detecting effects sizes of 1.5 or greater, assuming no further subgroup analysis. Glycemic Ijoad The study hypothesis predicts that increased glycemic load would be associated with increased breast-cancer mortality, but no association was observed between daily average 151 glycemic load and mortality in this cohort. Although relative risks have not been reported previously for an association of glycemic load with breast-cancer mortality, it is likely that the magnitude of risk, if any, would be similar to the modest RR's observed for glycemic load and risk of developing breast or colorectal cancer (62, 63), from 1.3 to 1.8, and therefore may have been missed in this study. Alcohol Though limited by relatively small numbers, in this cohort moderate beer consumption (at least once a month) was inversely associated with breast-cancer and all-cause mortality. This result is not consistent with other studies, such as Hebert et al. (82), who reported that beer drinking increased the risk of mortality in early stage breast cancer. That study, however, suffered from even smaller numbers of breast cancer deaths than our study (73 compared with 105), so their result may have been a spurious finding. With respect to breast-cancer incidence, the association with alcohol differs by ER and PR status, and increased risk may be limited primarily to women with ER-negative and PR-negative tumours (70, 71). The finding that women who had stopped drinking before diagnosis were at significandy higher risk of mortality from their breast cancer is difficult to interpret without information on how long before diagnosis drinking stopped. It is most plausible that women with more advanced disease might have stopped drinking due to feelings of ill health caused by the cancer. This was supported by a reduction in hazard ratio when the analysis was adjusted for stage at diagnosis. It is also possible that the group who stopped drinking contained more 152 alcoholics than average, and alcoholism may be related to metabolic imbalance, either direcdy or through dietary modifications. Biological Factors at Diagnosis Insulin and C-Veptide Higher insulin levels were associated with an increased risk of breast-cancer mortality in post-menopausal women but not in pre-menopausal women. This is consistent with the study hypothesis that higher insulin would be associated with higher mortality, and that menopausal status would modify the association. The non-linear pattern, with rates significantly elevated in the 2 n d and 3 r d quartiles but not in the 4 t h quartile, may be the result of chance variation in this small sample, or alternatively, may suggest the presence of a confounder related to high insulin levels that independently affects survival. When all stages are considered together, the 4 t h quartile point estimate (2.3) and 95% CI (0.7-7.7) are similar to those for the 2 n d quartile (RR, 2.7; 95% CI, 0.9-8.0) and the 3 r d quartile (RR, 2.5; 95% CI, 0.8-7.7), so the apparent non-linear relationship for the Stage II women is most probably a result of small sample size. The insulin results are consistent with the observed association of WHR and breast-cancer mortality, and together these findings support the study hypothesis that hyperinsulinemia and related factors may predict higher breast-cancer mortality. Insulin resistance, however, was not directly supported as a breast-cancer mortality factor by the C-peptide-to-fmctosamine ratio results in this study. The C-peptide-to-fmctosarnine ratio, an indicator of insulin resistance, Was not associated with breast-cancer mortality except for a single marginally significant result, and this may be explained by the nature of a ratio, which can increase from either high C-153 peptide, or low fmctosamine, or both, representing a potential mix of insulin resistant individuals and hypoglycemic individuals. A low ratio might represent a mix of people with low C-peptide (normal) or high fmctosamine (glycemia). Additional stratification to address this issue further was not possible due to the sample size. The association between insulin and breast-cancer mortality was modified by menopausal status, ER status and to a lesser extent, family history. Due to missing values and small subgroup sizes, it was not possible to stratify on ER status or family history in the insulin models as was done with WHR models, but adjustment of the insulin model for ER status and family history resulted in a strengthened association of insulin with breast-cancer mortality. A mechanism involving estrogen is consistent with insulin's role in estrogen production and bioavailability. Similar associations of ER-positivity and endogenous hormone variables like WHR have been reported in the Iowa Women's Health Study cohort (32), but in that study, this association was further modified by progesterone receptor (PR) status, which was not available for analysis in the current study. Despite the lack of PR status, the effect of ER status on the WHR association supports the hypothesis of distinct risk factors for tumour subtypes. Sex Hormone Binding Globulin SHBG levels at diagnosis were not significantly associated with breast-cancer mortality, but the point estimates suggest that if there was a protective effect of higher SHBG levels, by binding more estrogen, then it would be more likely for pre-menopausal women, biologically plausible because they have higher estrogen levels. SHBG levels were inversely correlated with insulin levels, which is consistent with msulin's down-regulation of SHBG. 154 Overall, the insulin results are consistent with those reported by Goodwin et al. (10) and, although limited by sample size, these results contribute information on the importance of menopausal status and ER status to the relationship between insulin and breast-cancer mortality. Changes Made Post-Diagnosis Body Si%e and Shape Changes On average, the women in the Phase II cohort reduced their BMI by 2.1%, but reduced their WHR less than 1%. This observation indicates that the weight loss did not reduce abdominal fat stores proportionately, attesting to the difficulty of modifying WHR, and is consistent with data on the heritability of WHR (92). Reduction of BMI by more than 5% was associated with a significant increase in risk of mortality, most strongly in women who were pre-menopausal at diagnosis. This is the reverse of the study hypothesis, that weight loss would improve survival. There are several possible explanations for these observations. The null finding for change in WHR might be due to the limited amount of change, as well as the reduced Phase II sample size. The BMI finding of greater risk for loss than for gain might be due to reverse causation, but this is unlikely, as discussed previously. If reduction in BMI post-diagnosis is causally related to mortality, perhaps deliberate efforts at excessive weight loss (as opposed to weight maintenance) should be discouraged, especially in women pre-menopausal at diagnosis. 155 Dietary Changes The study hypothesis predicts that decreased consumption of high-energy nutrients like fat and carbohydrate would improve survival, but changes in dietary intake during the first 2 years post-diagnosis were not associated with breast-cancer or all-cause mortality. This result is not surprising, given the reduced power of the Phase II sample size,, and the relatively modest relative risks that characterize dietary risk factors in general. In this cohort specifically, the average participant did not make large changes to dietary intake and this limited variation, combined with the errors in FFQ data may have hampered the ability to detect a modest effect size. Associations with changes in dietary intake may also be confounded by other factors related to diet and outcome, such as loss of appetite and depression, that can occur post-diagnosis and therapy. Some of the potential confounding was controlled by entering stage at diagnosis as a co-variate in the analyses, but residual confounding cannot be ruled out. Biological Pathways Environmental factors influencing msulin sensitivity and insulin secretion, and ultimately influencing breast-cancer risk factors, are summarized in Figure 17. An important premise of the discussion is that the onset of cancer is not necessarily due to a "defect". Rather, each individual has a specific range of variation for a given biological parameter, such as insulin sensitivity, and problems are imposed by environmental changes or constraints at odds with the individual's range of response. The search for genetic "defects", then, would only be expected to turn up a small percentage of cases, as evidenced in breast cancer, where less than 5% of cases are known to have genetic defects (24), although family history is associated with a 156 higher proportion of cases and may drive associations with factors like WHR, which may be heritable. One solution may be to bring the individuars environment into line with their range of response, since it is a general engineering principle that one cannot expect to succeed when operating outside basic design contraints. The first challenge is to identify and measure these "design constraints", and the second much greater challenge is to identify the effective changes and to encourage the individual to make necessary adjustments to lifestyle. An example is insulin resistance. It is not mtrinsically "good" or "bad" to be insulin resistant or insulin sensitive that depends on the environmental demands. In one situation, insulin resistance may be beneficial, and in another situation detrimental. Not everyone can thrive under the same environmental conditions. We need to de-emphasize the search for "defects" and work on studying the normal gene-environment interactions which are expressed as individual responses. Research is needed to define and validate measures that can be applied to individuals, especially early in life, to create an individual blueprint for health, an improvement over current across-the-board public health recommendations. Strengths and Limitations The results of this study must be interpreted with caution. First, the data are from an urban, largely Caucasian cohort, and the results may not generalize to a more mixed population. Second, potential misclassification from self-measured data such as waist and hip circumference can mask a true effect. This study is not able to distinguish self-measurement from self-report of self-perception. The accuracy and reliability of self-measurement of body 157 girths was not evaluated in this cohort, but was examined by Kushi et al. (184) in post-menopausal women from the mid-western United States, a population not unlike our study cohort. Self-measurement was found to be both repeatable and accurate. Intra-class correlations comparing 2 self-measurements (reliability) were 0.96 for waist girth and 0.97 for hip girth. Intra-class correlations comparing self-measurement with technician measurement (accuracy) were 0.93 for waist and 0.96 for hips. Accuracy did not vary with age or educational status, but self-measurements may be under-estimated as girth increases, a finding also reported by Weaver et al. (185). Height and weight taken at the clinic may not have been measured in a standard fashion, but the correlation between questionnaire values and clinic-measured values was strong (r=0.995, p<0.001). Third, although the primary variable, insulin, was a measured one, other variables were selected from questionnaire data that are subject to possible recall bias, and at least one variable, ER status, was missing values for a large number of women (29%), introducing possible bias. Even if perfectly measured, the biologic variables only represent a single point in time, but non-fasting measurement would likely increase measurement error. This potential error may be partly responsible for null results in pre-menopausal women, and may have attenuated the insulin effect size seen in post-menopausal women, so the association may be stronger than that reported. As well, the designation of pre-or post-menopausal is not a hard line, but rather a continuum, including a possibly long transition period about which little is known. Potential misclassification of menopausal status might have attenuated observed differences between pre- and post-menopausal women. Confounding by HRT use and PR status cannot be ruled out. Fourth, the low power, particularly in subgroup analyses, makes the signficant results more noteworthy, but casts doubt on the null results. Finally, risk may have been under-estimated or an effect missed due 158 to the small, relatively homogeneous cohort. For example, Petrelli et al. (186) found BMI to be related to breast cancer mortality in a larger cohort with a wider range of BMI values. The strengths of this study are also worth mentioning. The use of prospective data with up to 10 years of follow up provided strong observational evidence. A 87% participant response rate decreased the likelihood of selection bias. A strong and well-balanced research team, combined with a well-characterized cohort and the excellent follow up, record keeping, infrastructure and resources of the BCCA provided the critical success factors a research study must have. With a focus on biologically plausible modifiable factors, this study is clinically important and timely, and addressed a problem of significant magnitude. Validity To determine whether the results observed represent a valid association, alternative explanations such as chance, bias and confounding must be considered. Judging whether the association is causal requires this step, plus a wider consideration of other studies and biological plausibility, which will be covered in a subsequent section. The presence of a valid statistical association between hypermsulinemia-related factors and breast-cancer mortality is only one explanation for the results of this study. One alternative explanation is chance, which can affect any study where a sample of the population is examined. A second alternative explanation is bias, or systematic error in participant selection or information reporting. A third alternative explanation for some or all of the effects observed is the potential presence of important unmeasured or uncontrolled differences between survivors and deceased that are related to the variable being studied and 159 independently affect outcome, or confounding. Each of these alternative explanations will be discussed in more detail with attention to strategies used to minimize them in this study. The Role of Chance The lack of statistical significance for a given result means that chance cannot be ruled out as an explanation, not that a valid association does not exist. The width of the confidence intervals observed in this study indicate a fairly large amount of variability in the estimates, an indicator of the relatively small sample size. Since the power of the study is dependent on both the size of the study population and the magnitude of the effect, the greatest power is associated with the largest relative risk estimates. In this study, there was only adequate power (80% or more) to detect relative risks of 2.1 or greater. For the statistically significant effects such as WHR, the estimated role of chance is small, but sampling variability, even then, cannot be completely ruled out as a factor in the observed results. Sources of Bias In this study, the most serious potential source of recall bias was the retrospective food frequency questionnaire. Data about dietary intake are subject to error, even when collected as a daily food record, as discussed in a later section. When participants are asked to estimate the number of times per week that they ate broccoli during some time period in the past, the researcher must accept a certain amount of error because the person is estimating from memory, which is imperfect. In this instance, recall bias is minimized by asking about the previous year rather than a year in the distant past. Even so, some random misclassification cannot be ruled out. The effect would be to favor a null finding, and recall bias is therefore likely to be responsible for some part of the null findings related to food consumption in this 160 cohort. Because it is a cohort study, not a case-control, all participants were diagnosed with breast cancer and all of them underwent similar experiences up to the point of study enrolment. There is no reason to believe that those who are msulin resistant would have recalled their diet any differently that those with normal insulin sensitivity, so systematic misclassification is not likely to be a problem. An additional source of bias, this one systematic, was the selection of a mainly urban population, the patients of the Vancouver Cancer Centre. This bias was minimized by the broader base of patients served by the VCC at the time of study enrolment, more closely representing British Columbia as a whole, compared with today, subsequent to opening additional cancer centres in the Fraser Valley and in the Interior. Loss of study subjects to follow up may also introduce bias, but active follow up and regular linkage of patient data to Vital Statistics for outcome updates miriimizes this issue. Bias may also result from the social desirability associated with personal topics such as body measurements, diet and exercise. It is known that under-estimation of self-measurement increases with girth (185). Sedentary people might be expected to over-report their exercise and over-eaters might be expected to under-report their dietary intake. Random misclassification would favor a null result. A bigger problem might arise if insulin-resistant sedentary people were more or less likely to over-report exercise than sedentary non-insulin resistant people, for example, and although this was not examined directly, there is no evidence of it in the literature. 161 In the Phase II questionnaire, a healthy volunteer bias would be likely because this was a convenience sample. Volunteer bias was minimized by the relatively high response rate for a 2-year follow-up (465/603 or 77%). Confounding Confounding occurs when an observed association is due, in whole or in part, to the effect of a third factor that is associated with the exposure and independendy affects risk. In this study, stratification and multi-variate analyses were used to partially control for confounding, though some residual confounding cannot be ruled out. Confounding is a considerable problem when studying lifestyle factors such as diet, which may be associated with a complex web of interrelated potential confounders. One factor in particular was not measured for this study, HRT use, and is a potential confounder because estrogen bioavailability is increased by insulin, and estrogen has an independent relationship with risk. PR status is another unmeasured factor that is a potential confounder. Measurement Error Three types of measurement must be considered here; 1) the lab assays, 2) the questionnaire, and 3) the self-measurements. Lab assays. Replicate assays were not possible due to very limited volumes of serum. Assays were processed by a well-respected research lab, Hospitals In-Common Lab (HICL, Toronto) (170). Assays were completed in small batches, to ensure quality. Results were compared by batch number and batch was not found to have a significant effect on the results. 162 Questionnaire. The questionnaire was a well-known and widely-used instrument (187), but was not validated in this population. Comparison of mean nutrient intakes measured in the target population, using other dietary assessment methods such as diet records would be optimal (188). The Health Habits and History Questionnaire (HHHQ), however, has been compared with multiple dietary records (189), and with other FFQ's, including the Harvard Semi-quantitative Food Frequency Questionnaire (190) in American populations, but not specifically in Canadian populations. As with any FFQ, measurement error cannot be ruled out when people are asked to estimate portion sizes. Also, variations in some food items, for example "vegetable soup", mean that nutrient composition values are only estimates. When portion size and nutrient composition are multiplied, so is the error. The food frequency questionnaire, however, is still the best method to estimate past intake, as serum levels of nutrient components have not proven to be related in any simple way to intake, and intake recorded by self-report using 24-hour recall or food record methods is also subject to problems with respect to recall (191). Self-measurement. Self-measurement was a potential source of error for the waist and hip measurements used in this study. Waist and hip self-measurement has been found to be accurate and repeatable in post-menopausal women (184, 185), with some tendency of larger women to under-estimate, as mentioned above. There is no evidence to suggest that insulin resistant women, for example, might systematically over- or under-report measurements, compared with non-insulin resistant women; similarly, it would not be expected that a women with Stage I disease would systematically report measurements differently than a women with Stage III disease. If random, this error would attenuate the difference, which for WHR 163 remained significant and therefore the risk associated with high waist-to-hip ratio may be greater than reported. Generalizability Conducted mainly in Caucasian women, the results of this study may not be generalizable to ethnic minorities. Given the consecutive nature of the sample, the high response rate and the extensive coverage of the BCCA (about 70% referred for this time period), however, the study results are likely to be generalizable to British Columbian and Canadian urban female populations. Comparison with B.C. Population, 1991-2000 Table 39 presents basic information about outcome, age at diagnosis, staging (where available), and percent alive as of November 2001, for the population of patients who were referred to the BC Cancer Agency for treatment during the years 1991-2000 inclusive. Also shown in Table 39 are outcome and age at diagnosis data for cancer cases recorded in the BC Cancer Registry who were not referred for treatment at a BCCA clinic. These latter data do not include staging information. Finally, Table 39 shows the combined totals for all persons diagnosed with cancer in BC regardless of whether or not they were treated at the BCCA. The selection criteria used for these data are the same as the criteria used for the study cohort. Where multiple breast primaries fit the criteria, only one was selected — the one with the initial diagnosis date. The percent alive from the study cohort was 75.8% (Stage I: 86.5%, Stage II: 68.8%, Stage III; 40.4%), which is similar to the population data for 1991 and 1992. The mean age at diagnosis for the study cohort (54.5) was slighdy less than that for the BC population shown in Table 39, and the proportion of pre-menopausal and post-menopausal women in the 164 study cohort (39% pre; 61% post), was similar to the population data. These data suggest that the results of the current study should be generalizable to the population of breast-cancer cases in British Columbia. The Evidence for Causality Evaluation of the presence of a cause-effect relationship requires more than the establishment of a valid result. Rather, a preponderance of evidence from a number of sources must support a judgment of causality. Additional criteria include strength of association, biological plausibility of the hypothesis, consistency with findings of other studies and investigators, the presence of a dose-response relationship, coherance, and information concerning the temporal sequence of putative cause and effect (192). It would also be helpful to show that reversing hyperinsulinernia/insulin resistance would result in improved breast-cancer outcomes, but this study was not designed to address that issue. Strength of Association Although insulin levels and WHR increase risk 2- to 3-fold, the effect sizes for diet and physical activity variables are generally much more modest, which is consistent with other studies in the literature. The complexity of the association between insulin and lifestyle determinants means that the association may not be a strong one. A modest association, however, is not necessarily an unimportant one, especially when virtually the entire population is exposed. The strength and statistical significance of WHR and msulin effects are consistent with a causal relationship, but do not in themselves indicate causality. 165 biological Plausibility Several lines of research support a role for excess insulin in carcinogenesis. Inhibition of insulin secretion or action has been shown to decrease growth of mammary tumours in animal studies (122). Insulin is an important growth factor for human epithelial cells, which have both insulin and IGF-I (msulin-like growth factor I) receptors. Insulin can also act by increasing available IGF-I, an independent risk factor, through down-regulation of IGF binding proteins (193). Similarly, insulin is believed to down-regulate sex hormone binding globulin (SHBG) (100, 126, 194), thereby increasing bioavailability of estrogen and testosterone. Regulation of binding proteins is a biologically plausible mechanism for insulin action, thereby controlling risk factors like IGF-I (130, 131, 195, 196) and estrogen (13, 197, 198). Consistency Insulin levels and C-Peptide-to-glucose or C-Peptide-to-fmctosamine ratios (measures of insulin sensitivity) have been shown to be higher in breast-cancer cases than in controls in some (11, 12), but not all (199) studies. Interpretation of the null results for insulin reported by Kaaks et al. (199) is complicated by the fact that the authors did not report a glucose or fmctosamine value nor did they exclude diabetics. Del Guidice et al. (11) examined the role of circulating insulin levels in pre-menopausal breast cancer risk, and after adjustment for age and weight, found an odds ratio of 2.83 (95% CI, 1.22-6.58) for the highest msulin quintile compared to the lowest. High fasting insulin levels have been associated with breast-cancer mortality (RR, 3.3, 95% CI, 1.5-7.0 for highest to lowest quartile) (10), consistent with the work on breast-cancer incidence. 166 Dose-Response Relationship A dose-response relationship was observed for total fat and saturated fat intake at diagnosis, and for WHR at diagnosis, as well as an inverse dose-response relationship for protein intake at diagnosis and breast-cancer mortality. It is also possible that some insulin-related factors may be subject to a threshold effect such that the relationship is not strictly linear, such as that observed for serum insulin levels. Coherance The observed results are cohesive, in that serum insulin levels, and some of the lifestyle factors associated with insulin resistance and hyperinsulinemia were found to be associated with breast-cancer mortality (e.g., waist-to-hip ratio, saturated fat intake), but the null association for physical activity, if a real finding, does not support the overall hypothesis. Temporality In this prospective design, the temporal sequence is clear, because the exposures were recorded before the outcomes occurred. One issue with respect to temporality is the possibility that exposure at some other time period, perhaps during childhood, may be a contributing factor. This requires additional research. 167 Chapter 6 CONCLUSION Summary The following conclusions arise from this study. 1. Higher insulin levels are associated with increased risk of dying for post-menopausal breast-cancer patients, and should be investigated further to see if changes made to normalize insulin levels could improve breast-cancer outcomes. 2. Elevated WHR is a significant predictor of mortality after a diagnosis of post-menopausal ER-positive breast cancer and may be useful clinically, both as a simple measurement with prognostic value, and as an opportunity for risk reduction. The risk of post-menopausal breast-cancer mortality associated with increased WHR is independent of BMI, thus elevated WHR may pose as much of a risk for slim or average women as it does for obese women. Menopause and ER status are important modifiers of the relationship between WHR and breast-cancer mortality, independent of the effects of age. It would be useful to undertake a larger study to confirm the observed relationships, and more precisely define the WHR value that should signal a need for intervention in specific populations, though as noted previously, changing WHR may be difficult. 3. High saturated fat intake is a risk factor for pre-menopausal breast-cancer mortality, and low protein intake is a risk factor for breast-cancer mortality in all women. 168 4. Hormone receptor status (ER/PR) is an important modifier of breast-cancer risk factors, and receptor status testing should be carried out for all breast carcinomas, including DCIS, which cases are not currendy being tested in British Columbia, although tamoxifen therapy is recommended. 5. Reduction of BMI by more than 5% in the 2 years post-diagnosis is associated with increased risk of mortality. 6. The available evidence does not yet support a causal relationship for insulin and breast-cancer mortality. Research to determine causality is urgendy needed. If the preliminary data are confirmed by other studies, clinical trials to assess the effects of pharmacologically lowering insulin levels and insulin resistance on breast-cancer incidence and breast-cancer survival would be indicated to establish whether and who this type of intervention might help. The evidence to date, though incomplete, provides a rationale to support these more expensive and time-consuming studies. 7. In the interim, the data presented in this dissertation support lifestyle interventions as appropriate actions now, while evidence for more aggressive pharmaceutical interventions is collected. Although diet and exercise regimens cannot be expected to improve insulin sensitivity to the same degree as pharmaceutical agents, they produce few if any adverse effects and may also protect against other chronic diseases. 169 Implications for Breast Cancer Patients Today The late Dr Ernst Wynder, Chief Medical Director of the American Health Foundation, addressed the topic of what we can reasonably recommend about diet and exercise to breast cancer patients (200). He pointed to ways humans have altered natural behaviours in the course of becoming civilized, and remarked that most of our problems in disease today relate to metabolic overload. The American Health Foundation recommends a diet called the "25/25" diet, which is 25% fat and 25 grams per day of fiber. The World Cancer Research Fund/American Institute for Cancer Research 1997 report on nutrition and cancer (4) recommends that the most effective dietary means of reducing breast cancer risk are to increase consumption of vegetables and fruits, avoid alcohol, and maintain normal body weight by appropriate exercise and diet throughout life. Dr. Basil Stoll of London's St. Thomas' Hospital has called for diet and exercise interventions to improve breast cancer prognosis (27, 93, 113). He recommends a higher dietary fiber to fat ratio, combined with regular physical activity. He points out that hyperinsulinemic insulin resistance may interact synergistically with estrogen to stimulate breast cancer growth, and that dietary and exercise changes can reduce both insulin and estrogen levels in obese and non-obese women, both pre and post-menopause. The implication, then, for breast-cancer patients today is that they can take a hand in their own treatment and begin making changes to help improve their prognosis. Those changes, of which the first 3 relate direcdy to the current study findings, are: 1. Maintain normal weight. 170 2. Ensure adequate dietary protein. 3. Reduce total fat intake. 4. Switch fats to increase intake of omega-3 PUFAs and monounsaturated fatty acids, and decrease saturated fats and omega-6 PUFAs. 5. Eat more vegetables. This step also contributes to reducing fat because high fiber foods such as vegetables are filling but typically energy-dilute and contain low or no fat. 6. Increase regular exercise. Future Research The findings by Bruning, Del Guidice, Muti, Goodwin and this study are enticing, but need confirmation and expansion. With respect to insulin and breast-cancer mortality, the study by Goodwin et al. essentially stands alone although supported by in vitro and animal work. Additional prospective studies are urgendy needed to test if decreased insulin is associated with decreased mortality and for which sub-groups. Certainly, diet and exercise trials have already been called for, to improve breast cancer prognosis . With further epidemiological evidence, clinical trials to lower insulin levels and influence prognosis may also be indicated. 171 BIBLIOGRAPHY 1. Tocher, M. How to ride a dragon: Women with breast cancer tell their stories. Toronto: Key Porter Books, 2002. 2. Doll, R., and Peto, R. The causes of cancer: quantitative estimates of avoidable risks of cancer in the United States today. J Natl Cancer Inst, 66:1191-308, 1981. 3. Doll, R. Chronic and degenerative disease: major causes of morbidity and death. Am J Clin Nutr, 62:1301S-1305S, 1995. 4. World Cancer Research Fund Panel (Potter JD Chair). Food, Nutrition and the Prevention of Cancer: A Global Perspective. Washington (DC): WCRF/American Institute of Cancer Research, 1997. 5. Lichtenstein, P., Holm, N . V., Verkasalo, P. K., Iliadou, A., Kaprio, J., Koskenvuo, M. , Pukkala, E., Skytthe, A., and Hemminki, K. Environmental and heritable factors in the causation of cancer—analyses of cohorts of twins from Sweden, Denmark, and Finland [see comments]. N Engl J Med, 343: 78-85, 2000. 6. Maskarinec, G. Breast cancer—interaction between ethnicity and environment. In Vivo, 7 :^ 115-23, 2000. 7. Parkin, D. M. , Muir, C. S., Whelan, S. L., Gao, J. T., Ferlay, J., and PoweU, J. Cancer incidence in five continents. Lyon (France): International Agency for Cancer Research, 1992. 8. Winter, H. , Cheng, K. K., Cummins, C , Marie, R., Silcocks, P., and Varghese, C. Cancer incidence in the south Asian population of England (1990-92). Br J Cancer, 79: 645-54,1999. 9. Borugian, M . J., Sheps, S. B., Whittemore, A. S., Wu, A. H. , Potter, J. D., and Gallagher, R. P. Carbohydrates and colorectal cancer risk among Chinese in North America. Cancer Epidemiol Biomarkers Prev, / / : 187-93, 2002. 172 10. Goodwin, P. J., Ennis, M. , Pritchard, K. I., Trudeau, M . E., Koo, J., Madarnas, Y., Hartwick, W., Hoffman, B., and Hood, N . Fasting insulin and outcome in early-stage breast cancer: results of a prospective cohort study. J Clin Oncol, 20:42-51., 2002. 11. Del Giudice, M. E., Fantus, I. G., Ezzat, S., McKeown-Eyssen, G., Page, D., and Goodwin, P. J. Insulin and related factors in premenopausal breast cancer risk. Breast Cancer Res Treat, 47:111-20, 1998. 12. Bruning, P. F., Bonfrer, J. M. , van Noord, P. A., Hart, A. A., de Jong-Bakker, M., and Nooijen, W.J. Insulin resistance and breast-cancer risk. Int J Cancer, 52: 511-6, 1992. 13. Kaaks, R. Endogenous hormone metabolism as an exposure marker in breast cancer chemoprevention studies. IARC SciPubl, 154:149-62, 2001. 14. Berrino, F., Bellati, C , Secreto, G., Camerini, E., Pala, V., Panico, S., Allegro, G., and Kaaks, R. Reducing bioavailable sex hormones through a comprehensive change in diet: the diet and androgens (DIANA) randomized trial. Cancer Epidemiol Biomarkers Prev, 10:25-33., 2001. 15. Torjesen, P. A., Birkeland, K. I., Anderssen, S. A., Hjermann, I., Holme, I., and Urdal, P. Lifestyle changes may reverse development of the insulin resistance syndrome. The Oslo Diet and Exercise Study: a randomized trial. Diabetes Care, 20:26-31., 1997. 16. Fnedenrieich, C. M. Review of Lifestyle and Environmental Risk Factors for Breast Cancer., The Canadian Breast Cancer Initiative Workshop on the Primary Prevention of Breast Cancer. Quebec City, 2001. 17. Colditz, G. A. Epidemiology of breast cancer. Findings from the nurses' health study. Cancer, 71:1480-9., 1993. 18. National Cancer Institute of Canada. Canadian Cancer Statistics 2002. Toronto: National Cancer Institute of Canada, 2002. 19. Chu, K. C , Tarone, R. E., Kessler, L. G., Ries, L. A., Hankey, B. F., Miller, B. A., and Edwards, B. K. Recent trends in U.S. breast cancer incidence, survival, and mortality rates. J Natl Cancer Inst, 88:1571-9., 1996. 20. Hermon C, and V, B. Breast cancer mortality rates are levelling off or begmning to decline in many western countries: analysis of time trends, age-cohort and age-period models of breast cancer mortality in 20 countries. British Journal of Cancer, 73: 955-960,1996. 21. Hoel, D. G., Davis, D. L , Miller, A. B., Sondik, E. J., and Swerdlow, A. J. Trends in cancer mortality in 15 industrialized countries, 1969-1986. J Nad Cancer Inst, 84: 313-20, 1992. 173 22. Sant, M. , Capocaccia, R., Verdecchia, A., Esteve, J., Gatta, G., Micheli, A., Coleman, M. P., and Berrino, F. Survival of women with breast cancer in Europe: variation with age, year of diagnosis and country. The EUROCARE Working Group. Int J Cancer, 77:679-83., 1998. 23. Rose, D. P., Boyar, A. P., and Wynder, E. L. International comparisons of mortality rates for cancer of the breast, ovary, prostate, and colon, and per capita food consumption. Cancer, 58:2363-71,1986. 24. Doll, R. Nature and nurture: possibilities for cancer control. Carcinogenesis, 17: 177-84,1996. 25. Graham, S., Hellmann, R., Marshall, J., Freudenheim, J., Vena, J., Swanson, M. , Zielezny, M. , Nemoto, T., Stubbe, N . , and Raimondo, T. Nutritional epidemiology of postmenopausal breast cancer in western New York [see comments]. Am J Epidemiol, 134:552-66,1991. 26. Huang, Z., Hankinson, S. E., Colditz, G. A., Stampfer, M. J., Hunter, D. J., Manson, J. E., Hennekens, C. H., Rosner, B., Speizer, F. E., and Willett, W. C. Dual effects of weight and weight gain on breast cancer risk [see comments]. Jama, 278: 1407-11, 1997. 27. Stoll, B. A. Western diet, early puberty, and breast cancer risk. Breast Cancer Res Treat, ^.•187-93., 1998. 28. Rock, C. L., Lampe, J. W., and Patterson, R. E. Nutrition, genetics, and risks of cancer [In Process Citation]. Annu Rev Public Health, 21:47-64, 2000. 29. Nichols, K. E., Malkin, D., Garber, J. E., Fraumeni, J. F., Jr., and Li , F. P. Germ-line p53 mutations predispose to a wide spectrum of early-onset cancers. Cancer Epidemiol Biomarkers Prev, 10:83-7., 2001. 30. Webster, N . J., Resnik, J. L., Reichart, D. B., Strauss, B., Haas, M., and Seely, B. L. Repression of the msulin receptor promoter by the tumor suppressor gene product p53: a possible mechanism for receptor overexpression in breast cancer. Cancer Res, 56:21%!-%., 1996. 31. Anonymous. Harvard report on cancer prevention. Causes of human cancer. Socioeconomic status. Cancer Causes Control, 7 Suppl 1: S33-5., 1996. 32. Potter, J. D., Cerhan, J. R., Sellers, T. A., McGovern, P. G., Drinkard, C , Kushi, L. R., and Folsom, A. R. Progesterone and estrogen receptors and mammary neoplasia in the Iowa Women's Health Study: how many kinds of breast cancer are there? Cancer Epidemiol Biomarkers Prev, 4: 319-26, 1995. 174 33. Hislop, T. G., Coldman, A. J., Elwood, J. M. , Skippen, D. H., and Kan, L. Relationship between risk factors for breast cancer and hormonal status. Int J Epidemiol, 15: 469-76., 1986. 34. Harras, A. Cancer rates and risks. Bethesda MD: National Cancer Institute, 1996. 35. Anonymous. Breast cancer and hormone replacement therapy: collaborative reanalysis of data from 51 epidemiological studies of 52,705 women with breast cancer and 108,411 women without breast cancer. Collaborative Group on Hormonal Factors in Breast Cancer. Lancet, 350:1047-59., 1997. 36. Rossouw, J. E., Anderson, G. L., Prentice, R. L., LaCroix, A. Z., Kooperberg, C , Stefanick, M. L., Jackson, R. D., Beresford, S. A., Howard, B. V., Johnson, K. C , Kotchen, J. M., and Ockene, J. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results From the Women's Health Initiative randomized controlled trial. Jama, 288:321-33., 2002. 37. Yaffe, M. J., Boyd, N . F., Byng, J. W., Jong, R. A., FisheU, E., Lockwood, G. A., Litde, L. E., and Tritchler, D. L. Breast cancer risk and measured mammographic density. Eur J Cancer Prev, 7 Suppl 1: S47-55, 1998. 38. Boyd, N . F., Jensen, H . M., Cooke, G., Han, H . L., Lockwood, G. A., and Miller, A. B. Mammographic densities and the prevalence and incidence of histological types of benign breast disease. Reference Pathologists of the Canadian National Breast Screening Study. Eur J Cancer Prev, 9:15-24, 2000. 39. Knight, J. A., Martin, L. J., Greenberg, C. V., Lockwood, G. A., Byng, J. W., Yaffe, M . J., Tritchler, D. L., and Boyd,. N . F. Macronutrient intake and change in mammographic density at menopause: results from a randomized trial. Cancer Epidemiol Biomarkers Prev, 8:123-8, 1999. 40. Hislop, T. G., and Boyd, N . An explanatory clinical trial of breast cancer prevention, 1994-2004., 2000: BC Cancer Agency; accessed 4 August 2000; available from http://www.bccancer.bc.ca/research/ccr/people/ghislop/projectl .htm, 2000. 41. Byrne, C , Colditz, G. A., Willett, W. C , Speizer, F. E., Pollak, M. , and Hankinson, S. E. Plasma insulin-like growth factor (IGF) I, IGF-binding protein 3, and mammographic density. Cancer Res, 60:3744-8, 2000. 42. Fraumeni, J. F., Jr., Lloyd, J. W., Smith, E. M. , and Wagoner, J. K. Cancer mortality among nuns: role of marital status in etiology of neoplastic disease in women. J Nad Cancer Inst, 42:455-68,1969. 43. Anonymous. Harvard report on cancer prevention. Causes of human cancer. Reproductive factors. Cancer Causes Control, 7 Suppl 1: S29-31., 1996. 175 44. Wu, A. H , Ziegler, R. G , Pike, M. C , Nomura, A. M . Y , West, D. W , Kolonel, L. N , Horn-Ross, P. L , Rosenthal, J. F , and Hoover, R. N . Menstrual and reproductive factors and risk of breast cancer in Asian-Americans. Br J Cancer, 73: 680-6, 1996. 45. WHO. Controlling the global obesity epidemic, 2002: World Health Organization, 2002. 46. McTiernan, A , Stanford, J. L , Weiss, N . S, Daling, J. R , and Voigt, L. F. Occurrence of breast cancer in relation to recreational exercise in women age 50-64 years. Epidemiology, 7:598-604, 1996. 47. Gilliland, F. D , Li , Y. F , Baumgartner, K , Crumley, D , and Samet, J. M. Physical activity and breast cancer risk in hispanic and non-hispanic white women. Am J Epidemiol, 154:442-50, 2001. 48. Luoto, R , Latikka, P , Pukkala, E , Hakulinen, T , and Vihko, V. The effect of physical activity on breast cancer risk: a cohort study of 30,548 women. Eur J Epidemiol, 16: 973-80, 2000. 49. Anonymous. Harvard report on cancer prevention. Causes of human cancer. Dietary factors. Cancer Causes Control, 7 Suppl 1: S7-9, 1996. 50. Willett, W. C. Diet and breast cancer. J Intern Med, 249:395-411, 2001. 51. Anonymous. Harvard report on cancer prevention. Causes of human cancer. Obesity. Cancer Causes Control, 7 Suppl 1: SI 1-3, 1996. 52. Sonnenschein, E , Toniolo, P , Terry, M. B , Bruning, P. F , Kato, I , Koenig, K. L , and Shore, R. E. Body fat distribution and obesity in pre- and postmenopausal breast cancer. Int J Epidemiol, 28:1026-31, 1999. 53. Friedenreich, C. M , Courneya, K. S, and Bryant, H. E. Case-control study of anthropometric measures and breast cancer risk. Int J Cancer, 99:445-52, 2002. 54. Folsom, A. R , Kushi, L. H , Anderson, K. E , Mink, P. J , Olson, J. E , Hong, C. P , Sellers, T. A , Lazovich, D , and Prineas, R. J. Associations of general and abdominal obesity with multiple health outcomes in older women: the Iowa Women's Health Study. Arch Intern Med, 160:2117-28, 2000. 55. Stoll, B. A. Upper abdominal obesity, insulin resistance and breast cancer risk. Int J Obes Relat Metab Disord, 26: 747-53, 2002. 56. Hollmann, M , Runnebaum, B , and Gerhard, I. Impact of waist-hip-ratio and body-mass-index on hormonal and metabolic parameters in young, obese women. Int J Obes Relat Metab Disord, 21:476-83, 1997. 176 57. Hall, I. J., Newman, B., Millikan, R. C , and Moorman, P. G. Body size and breast cancer risk in black women and white women: the Carolina Breast Cancer Study. Am J Epidemiol, 151:754-64., 2000. 58. Anonymous. Design of the Women's Health Initiative clinical trial and observational study. The Women's Health Initiative Study Group. Control Clin Trials, 19: 61-109., 1998. 59. WiUett, W. C. Specific fatty acids and risks of breast and prostate cancer: dietary intake. Am J Clin Nutr, 66:1557S-1563S., 1997. 60. Alberts, D. S., Martinez, M. E., Roe, D. J., Guillen-Rodriguez, J. M. , Marshall, J. R., van Leeuwen, J. B., Reid, M. E., Ritenbaugh, C , Vargas, P. A., Bhattacharyya, A. B., Earnest, D. L., and Sampliner, R. E. Lack of effect of a high-fiber cereal supplement on the recurrence of colorectal adenomas. Phoenix Colon Cancer Prevention Physicians' Network [see comments]. N EnglJ Med, 342:1156-62, 2000. 61. Schatzkin, A., Lanza, E., Code, D., Lance, P., Iber, F., Caan, B., Shike, M. , Weissfeld, J., Burt, R., Cooper, M . R., Kikendall, J. W., and Cahill, J. Lack of effect of a low-fat, high-fiber diet on the recurrence of colorectal adenomas. Polyp Prevention Trial Study Group [see comments]. N EnglJ Med, 342:1149-55, 2000. 62. Franceschi, S., Dal Maso, L., Augustin, L., Negri, E., Parpinel, M. , Boyle, P., Jenkins, D. J., and La Vecchia, C. Dietary glycemic load and colorectal cancer risk. Ann Oncol, 12:173-8, 2001. 63. Augustin, L. S, Dal Maso, L , La Vecchia, C , Parpinel, M , Negri, E , Vaccarella, S, Kendall, C. W , Jerikins, D. J , and Francesch, S. Dietary glycemic index and glycemic load, and breast cancer risk: a case-control study. Ann Oncol, 12:1533-8, 2001. 64. Yellowlees, W. W. Tribute to Cleave-forgotten prophet. Nutr Health, 7:163-8, 1991. 65. Cleave, T. L , and Campbell, G. D. The saccharine disease. Am J Proctol, 18: 202-10, 1967. 66. Pi-Sunyer, F. X . Glycemic index and disease. Am J Clin Nutr, 76:290S-8S, 2002. 67. Jenkins, D. J , Kendall, C. W , Augustin, L. S, Franceschi, S, Hamidi, M , Marchie, A , Jenkins, A. L , and Axelsen, M . Glycemic index: overview of implications in health and disease. Am J Clin Nutr, 76:266S-73S, 2002. 68. Foster-Powell, K , Holt, S. H , and Brand-Miller, J. C. International table of glycemic index and glycemic load values: 2002. Am J Clin Nutr, 76: 5-56, 2002. 177 69. Kropp, S, Becher, H , Nieters, A , and Chang-Claude, J. Low-to-moderate alcohol consumption and breast cancer risk by age 50 years among women in Germany. Am J Epidemiol, 154:624-34, 2001. 70. Gapstur, S. M , Potter, J. D , Drinkard, C , and Folsom, A. R. Synergistic effect between alcohol and estrogen replacement therapy on risk of breast cancer differs by estrogen/progesterone receptor status in the Iowa Women's Health Study. Cancer Epidemiol Biomarkers Prev, 4: 313-8, 1995. 71. Sellers, T. A , Vierkant, R. A , Cerhan, J. R , Gapstur, S. M , Vachon, C. M , Olson, J. E , Pankratz, V. S, Kushi, L. H , and Folsom, A. R. Interaction of dietary folate intake, alcohol, and risk of hormone receptor-defined breast cancer in a prospective study of postmenopausal women. Cancer Epidemiol Biomarkers Prev, / /: 1104-7, 2002. 72. John, E. M , Schwartz, G. G , Dreon, D. M , and Koo, J. Vitamin D and breast cancer risk: the NHANES I Epidemiologic follow-up study, 1971-1975 to 1992. National Health and Nutrition Examination Survey. Cancer Epidemiol Biomarkers Prev, 8: 399-406,1999. 73. Colston, K. W , and Hansen, C. M. Mechanisms implicated in the growth regulatory effects of vitamin D in breast cancer. Endocr Relat Cancer, 9:45-59, 2002. 74. Chiu, K. C , Chuang, L. M , and Yoon, C. The vitamin D receptor polymorphism in the translation initiation codon is a risk factor for insulin resistance in glucose tolerant Caucasians. BMC Med Genet, 2:2, 2001. 75. Steinmetz, K. A , and Potter, J. D. Vegetables, fruit, and cancer. II. Mechanisms. Cancer Causes Control, 2:427-42, 1991. 76. MacMahon, B , Trichopoulos, D , Cole, P , and Brown, J. Cigarette smoking and urinary estrogens. N EnglJ Med, 307:1062-5, 1982. 77. Zavaroni, I , Bonini, L , Gasparini, P , Dall'Aglio, E , Passeri, M , and Reaven, G. M. Cigarette smokers are relatively glucose intolerant, hypermsulinemic and dyslipidemic. Am J Cardiol, 73:904-5, 1994. 78. Harris, J. R , Lippman, M . E , Morrow, M , and Hellman, S. Diseases of the Breast. Philadelphia: Lippincott-Raven, 1996. 79. Rowlings, P. A , Williams, S. F , Antman, K. H , Fields, K. K , Fay, J. W , Reed, E , Pelz, C. J , Klein, J. P , Sobocinski, K. A , Kennedy, M . J , Freytes, C. O , McCarthy, P. L , Jr., Herzig, R. H , Stadtmauer, E. A , Lazarus, H. M , Pecora, A. L , Bitran, J. D , Wolff, S. N , Gale, R. P , Armitage, J. O , Vaughan, W. P , Spitzer, G , and Horowitz, M. M . Factors correlated with progression-free survival after high-dose chemotherapy and hematopoietic stem cell transplantation for metastatic breast cancer. Jama, 282: 1335-43, 1999. 178 80. Nomura, A. M , Marchand, L. L , Kolonel, L. N , and Hankin, J. H . The effect of dietary fat on breast cancer survival among Caucasian and Japanese women in Hawaii. Breast Cancer Res Treat, 18 Suppl 1: S135-41, 1991. 81. Jain, M , and Miller, A. B. Tumor characteristics and survival of breast cancer patients in relation to premorbid diet and body size. Breast Cancer Res Treat, 42:43-55, 1997. 82. Hebert, J. R , Hurley, T. G , and Ma, Y. The effect of dietary exposures on recurrence and mortality in early stage breast cancer. Breast Cancer Res Treat, 51:17-28, 1998. 83. Stoll, B. A. Obesity, social class and Western diet: a link to breast cancer prognosis. Eur J Cancer, 32A: 1293-5, 1996. 84. Saxe, G. A , Rock, C. L , Wicha, M. S, and Schottenfeld, D. Diet and risk for breast cancer recurrence and survival. Breast Cancer Res Treat, 53:241-53, 1999. 85. Chlebowski, R. T , Aiello, E , and McTiernan, A. Weight loss in breast cancer patient management. J Clin Oncol, 20:1128-43, 2002. 86. Ewertz, M , Gillanders, S, Meyer, L , and Zedeler, K. Survival of breast cancer patients in relation to factors which affect the risk of developing breast cancer. Int J Cancer, 49: 526-30, 1991. 87. Lees, A. W , Jenkins, H. J , May, C. L , Cherian, G , Lam, E. W , and Hanson, J. Risk factors and 10-year breast cancer survival in northern Alberta. Breast Cancer Res Treat, 13:143-51, 1989. 88. Zhang, S, Folsom, A. R , Sellers, T. A , Kushi, L. H , and Potter, J. D. Better breast cancer survival for postmenopausal women who are less overweight and eat less fat. The Iowa Women's Health Study. Cancer, 76:275-83,1995. 89. Kumar, N . B , Cantor, A , Allen, K , and Cox, C. E. Android obesity at diagnosis and breast carcinoma survival: Evaluation of the effects of anthropometric variables at diagnosis, including body composition and body fat distribution and weight gain during life span,and survival from breast carcinoma. Cancer, 88:2151-1., 2000. 90. Folsom, A. R , Kaye, S. A , Sellers, T. A , Hong, C. P , Cerhan, J. R , Potter, J. D , and Prineas, R. J. Body fat distribution and 5-year risk of death in older women. Jama, 269: 483-7,1993. 91. Sellers, T. A , Gapstur, S. M , Potter, J. D , Kushi, L. H , Bostick, R. M , and Folsom, A. R. Association of body fat distribution and family histories of breast and ovarian cancer with risk of postmenopausal breast cancer. Am J Epidemiol, 138: 799-803, 1993. 179 92. Sellers, T. A , Davis, J , Cerhan, J. R , Vierkant, R. A , Olson, J. E , Pankratz, V. S, Potter, J. D , and Folsom, A. R. Interaction of waist/hip ratio and family history on the risk of hormone receptor-defined breast cancer in a prospective study of postmenopausal women. Am J Epidemiol, 155:225-33, 2002. 93. Stall, B. A. Diet and exercise regimens to improve breast carcinoma prognosis. Cancer, 7^:2465-70,1996. 94. Stall, B. A. Obesity and breast cancer. Int J Obes Relat Metab Disord, 20: 389-92, 1996. 95. Holm, L. E , Nordevang, E , Hjalmar, M. L , Lidbrink, E , Callmer, E , and Nilsson, B. Treatment failure and dietary habits in women with breast cancer. J Nad Cancer Inst, <?5:32-6, 1993. 96. Holmes, M. D , Stampfer, M . J , Colditz, G. A , Rosner, B , Hunter, D. J , and Willett, W. C. Dietary factors and the survival of women with breast carcinoma, [comment] [erratum appears in Cancer 1999 Dec 15;86(12):2707-8]. Cancer, 86: 826-35,1999. 97. Berne, R. N , and Levy, M. N . Whole body metabolism and the hormones of the pancreatic islets. Physiology (2nd Edition), pp. 838-74. St. Louis: C.V. Mosby, 1988. 98. Bjorntorp, P. Hormonal control of regional fat distribution. Hum Reprod, 12 Suppl 1: 21-5,1997. 99. Suikkari, A. M , Koivisto, V. A , Rutanen, E. M , Yki-Jarvinen, H , Karonen, S. L , and Seppala, M. Insulin regulates the serum levels of low molecular weight insulin-like growth factor-binding protein. J Clin Endocrinol Metab, 66:266-72,1988. 100. Plymate, S. R , Hoop, R. C , Jones, R. E , and Matej, L. A. Regulation of sex hormone-binding globulin production by growth factors. Metabolism, 39:967-70, 1990. 101. Mauro, L , Salerno, M , Panno, M. L , Bellizzi, D , Sisci, D , Miglietta, A , Surmacz, E , and Ando, S. Estradiol increases IRS-1 gene expression and insulin signaling in breast cancer cells. Biochem Biophys Res Commun, 288: 685-9, 2001. 102. de Waard, F , and Trichopoulos, D. A unifying concept of the aetiology of breast cancer. Int J Cancer, 41:666-9, 1988. 103. Brand-Miller, J. C , and Colagiuri, S. Evolutionary aspects of diet and insulin resistance. World Rev Nutr Diet, 84:74-105,1999. 104. Thorburn, A. W , Brand, J. C , O'Dea, K , Spargo, R. M , and TrusweU, A. S. Plasma glucose and insulin responses to starchy foods in Australian aborigines: a population now at high risk of diabetes. Am J Clin Nutr, 46: 282-5, 1987. 180 105. Stoll, B. A. Nutrition and breast cancer risk: can an effect via insulin resistance be demonstrated? Breast Cancer Res Treat, 38:239-46, 1996. 106. Eaton, S. B , and Cordain, L. Evolutionary aspects of diet: old genes, new fuels. Nutritional changes since agriculture. World Rev Nutr Diet, 81:26-37, 1997. 107. Eaton, S. B , and Konner, M . Paleolithic nutrition. A consideration of its nature and current implications. N EnglJ Med, 312:283-9, 1985. 108. Eaton, S. B , Pike, M. C , Short, R. V , Lee, N . C , Trussell, J , Hatcher, R. A , Wood, J. W , Worthman, C. M , Jones, N . G , Konner, M. J , and et al. Women's reproductive cancers in evolutionary context. Q Rev Biol, 69:353-67, 1994. 109. Facchini, F. S, Hua, N , Abbasi, F , and Reaven, G. M . Insulin resistance as a predictor of age-related diseases. J Clin Endocrinol Metab, 86: 3574-8, 2001. 110. Reaven, G. M . Banting lecture 1988. Role of insulin resistance in human disease. Diabetes, 37:1595-607, 1988. 111. Zavaroni, I , Bonini, L , Fantuzzi, M , Dall'Aglio, E , Passeri, M , and Reaven, G. M. Hyperinsulinaemia, obesity, and syndrome X . J Intern Med, 235: 51-6, 1994. 112. Colangelo, L. A , Gapstur, S. M , Gann, P. H , Dyer, A. R , and Liu, K. Colorectal cancer mortality and factors related to the insulin resistance syndrome. Cancer Epidemiol Biomarkers Prev, / / : 385-91, 2002. 113. Stoll, B. A. Western nutrition and the msulin resistance syndrome: a link to breast cancer. Eur J Clin Nutr, 53: 83-7, 1999. 114. Kaaks, R. Nutrition, hormones, and breast cancer: is insulin the missing link? [see comments]. Cancer Causes Control, 7: 605-25,1996. 115. McKeown-Eyssen, G. Epidemiology of colorectal cancer revisited: are serum triglycerides and/or plasma glucose associated with risk? Cancer Epidemiol Biomarkers Prev, 3:687-95,1994. 116. Kim, Y. I. Diet, lifestyle, and colorectal cancer: is hypermsulinemia the missing link? Nutr Rev, 56:215-9,1998. 117. Kaaks, R , and Lukanova, A. Energy balance and cancer: the role of insulin and insulin-like growth factor-I. Proc Nutr Soc, 60: 91-106, 2001. 118. Giovannucci, E. Insulin and colon cancer. Cancer Causes Control, 6:164-79, 1995. 181 119. Moore, M. A , Park, C. B , and Tsuda, H . Implications of die hyperinsulinaemia-diabetes-cancer link for preventive efforts. Eur J Cancer Prev, 7: 89-107, 1998. 120. Macaulay, V. M . Insulin-like growth factors and cancer. Br J Cancer, 65: 311-20, 1992. 121. Watkins, L. F , Lewis, L. R , and Levine, A. E. Characterization of the synergistic effect of insulin and transferrin and the regulation of their receptors on a human colon carcinoma cell line. Int J Cancer, 45:372-5, 1990. 122. Yam, D. Insulin-cancer relationships: possible dietary implication. Med Hypotheses, 38:111-7,1992. 123. Sandhu, M. S, Dunger, D. B , and Giovannucci, E. L. Insulin, msulin-like growth factor-I (IGF-I), IGF binding proteins, their biologic interactions, and colorectal cancer. J Natl Cancer Inst, 94: 972-80, 2002. 124. Weiderpass, E , Gridley, G , Nyren, O , Ekbom, A , Persson, I , and Adami, H. O. Diabetes mellitus and risk of large bowel cancer [letter] [see comments]. J Natl Cancer Inst, 660-1, 1997. 125. Will, J. C , Galuska, D. A , Vinicor, F , and Calle, E. E. Colorectal cancer: another complication of diabetes mellitus? Am J Epidemiol, 147: 816-25, 1998. 126. Muti, P , Stanulla, M , Micheli, A , Krogh, V , Freudenheim, J. L , Yang, J , Schuneman, H . J , Trevisan, M , and Berrino, F. Markers of insulin resistance and sex steroid hormone activity in relation to breast cancer risk: a prospective analysis of abdominal adiposity, sebum production, and hirsutism (Italy). Cancer Causes Control, / / : 721-30, 2000. 127. Bmning, P. F , Van Doom, J , Bonfrer, J. M , Van Noord, P. A , Korse, C. M , Linders, T. C , and Hart, A. A. Insulin-like growth-factor-binding protein 3 is decreased in early-stage operable pre-menopausal breast cancer [published erratum appears in Int J Cancer 1995 Nov 27;63(5):762]. IntJ Cancer, 62:266-70, 1995. 128. Mink, P. J , Shahar, E , Rosamond, W. D , Alberg, A. J , and Folsom, A. R. Serum insulin and glucose levels and breast cancer incidence: the atherosclerosis risk in communities study. Am J Epidemiol, 156:349-52, 2002. 129. Clausen, P. G , Brismar, K , and Hall, K. Insulin-like growth factor-1 and msulin-like growth factor-1 binding protein in a representative population of type 2 diabetic patients in Sweden. J Clin Lab Invest, 58: 353-60, 1998. 130. Pollak, M. Insulin-like growth factor physiology and neoplasia. Growth Horm IGF Res, 10 Suppl A: S6-7, 2000. 182 131. Pollak, M . The question of a link between msulin-like growth factor physiology and neoplasia. Growth Horm IGF Res, 10 Suppl B: S21-4, 2000. 132. Dunn, S. E , Kari, F. W , French, J , Leininger, J. R , Travlos, G , Wilson, R , and Barrett, J. C. Dietary restriction reduces msulin-like growth factor I levels, which modulates apoptosis, cell proliferation, and tumor progression in p53- deficient mice. Cancer Res, 57:4667-72, 1997. 133. Molloy, C. A , May, F. E , and Westley, B. R. Insulin receptor substrate-1 expression is regulated by estrogen in the MCF-7 human breast cancer cell line. J Biol Chem, 275: 12565-71, 2000. 134. Yee, D , and Lee, A. V. Crosstalk between the insulin-like growth factors and estrogens in breast cancer. J Mammary Gland Biol Neoplasia, 5:107-15, 2000. 135. Papa, V , Pezzino, V , Costantino, A , Belfiore, A , Giuffrida, D , Frittitta, L , Vannelli, G. B , Brand, R , Goldfine, I. D , and Vigneri, R. Elevated msulin receptor content in human breast cancer. J Clin Invest, 86:1503-10, 1990. 136. White, M . F. The insulin signalling system and the IRS proteins. Diabetologia, 40 Suppl Z-S2-17, 1997. 137. Ffimsworth, H. P. Diabetes mellitus: Its differentiation into insulin-sensitive and msulin-insensitive types. Lancet, /: 127-30, 1936. 138. Stoll, B. A. Essential fatty acids, insulin resistance, and breast cancer risk. Nutr Cancer, 31:12-1,1998. 139. Weyer, C , Hanson, K , Bogardus, C , and Pradey, R. E. Long-term changes in insulin action and insulin secretion associated with gain, loss, regain and maintenance of body weight. Diabetologia, 43: 36-46, 2000. 140. Potischman, N , Swanson, C. A , Siiteri, P , and Hoover, R. N . Reversal of relation between body mass and endogenous estrogen concentrations with menopausal status [see comments]. J Natl Cancer Inst, 88: 756-8, 1996. 141. Carlson, N . R. Regulation and control of food intake. In: N . R. Carlson (ed.), Physiology of Behaviour, pp. 373. Boston: Allyn and Bacon, 1981. 142. Teff, K. L , Mattes, R. D , Engelman, K , and Mattern, J. Cephalic-phase insulin in obese and normal-weight men: relation to postprandial insulin. Metabolism, 42: 1600-8,1993. 143. Zimmet, P. Z. Hyperinsulinemia—how innocent a bystander? Diabetes Care, 16 Suppl 3:56-70, 1993. 183 144. Borkman, M , Campbell, L. V , Chisholm, D. J , and Storlien, L. H . Comparison of the effects on insulin sensitivity of high carbohydrate and high fat diets in normal subjects. J Clin Endocrinol Metab, 72:432-7,1991. 145. Parillo, M , Rivellese, A. A , Ciardullo, A. V , Capaldo, B , Giacco, A , Genovese, S, and Riccardi, G. A high-monounsaturated-fat/low-carbohydrate diet improves peripheral insulin sensitivity in non-insulin-dependent diabetic patients. Metabolism, 41:1373-8,1992. 146. Sevak, L , McKeigue, P. M , and Marmot, M. G. Relationship of hyperinsulinemia to dietary intake in south Asian and European men. Am J Clin Nutr, 59:1069-74, 1994. 147. Storlien, L. H , Pan, D. A , Kriketos, A. D , O'Connor, J , Caterson, I. D , Cooney, G. J , Jenkins, A. B , and Baur, L. A. Skeletal muscle membrane lipids and insulin resistance. Lipids, 31 Suppl: S261-5, 1996. 148. Stoll, B. A. Breast cancer and the western diet: role of fatty acids and antioxidant vitamins. Eur J Cancer, 34:1852-6, 1998. 149. Yam, D , Eliraz, A , and Berry, E. M . Diet and disease—the Israeli paradox: possible dangers of a high omega- 6 polyunsaturated fatty acid diet. Isr J Med Sci, 32:1134-43, 1996. 150. Storlien, L. H , Baur, L. A , Kriketos, A. D , Pan, D. A , Cooney, G. J , Jenkins, A. B , Calvert, G. D , and Campbell, L. V. Dietary fats and msulin action. Diabetologia, 39: 621-31, 1996. 151. Garg, A , Bande, J. P , Henry, R. R , Coulston, A. M , Griver, K. A , Raatz, S. K , Brinkley, L , Chen, Y. D , Grundy, S. M , Huet, B. A , and Reaven, G. M . Effects of varying carbohydrate content of diet in patients with non-insulin-dependent diabetes mellitus [see comments]. Jama, 271:1421-8, 1994. 152. Daly, M . E , Vale, C , Walker, M , Alberti, K. G , and Mathers, J. C. Dietary carbohydrates and insulin sensitivity: a review of the evidence and clinical implications. Am J Clin Nutr, 66:1072-85,1997. 153. Hunt, S. M , and Groff, J. L. Dietary fiber: No longer considered inert. Advanced Nutrition and Human Metabolism, pp. 349-59. St Paul: West Publishing, 1990. 154. Waldhausl, W. K , Gasic, S, Bratusch-Marrain, P , Komjati, M , and Korn, A. Effect of stress hormones on splanchnic substrate and insulin disposal after glucose ingestion in healthy humans. Diabetes, 36:127-35, 1987. 155. Khani, S, and Tayek, J. A. Cortisol increases gluconeogenesis in humans: its role in the metabolic syndrome. Clin Sci (Lond), 101:739-47, 2001. 184 156. Greisen, J , Juhl, C. B , Grofte, T , Vilstrup, H , Jensen, T. S, and Schmitz, O. Acute pain induces insulin resistance in humans. Anesthesiology, 95: 578-84, 2001. 157. Bjorntorp, P , Holm, G , and Rosmond, R. Hypothalamic arousal, insulin resistance and Type 2 diabetes mellitus. Diabet Med, 16: 373-83, 1999. 158. Rosmond, R , Holm, G , and Bjorntorp, P. Food-induced Cortisol secretion in relation to anthropometric, metabolic and haemodynamic variables in men. Int J Obes Relat Metab Disord, 24:416-22, 2000. 159. Attvall, S, Fowelin, J , Lager, I , Von Schenck, H , and Smith, U. Smoking induces insulin resistance—a potential link with the insulin resistance syndrome. J Intern Med, 233:327-32, 1993. 160. Godsland, I. F , Wynn, V , Walton, C , and Stevenson, J. C. Insulin resistance and cigarette smoking. Lancet, 339:1619-20, 1992. 161. Stoll, B. A. Perimenopausal weight gain and progression of breast cancer precursors. Cancer Detect Prev, 23:31-6, 1999. 162. Rock, C. L , Flatt, S. W , Newman, V , Caan, B. J , Haan, M . N , Stefanick, M. L , Faerber, S, and Pierce, J. P. Factors associated with weight gain in women after diagnosis of breast cancer. Women's Healthy Eating and living Study Group. J Am Diet Assoc, 99:1212-21, 1999. 163. Trichopoulou, A , Gnardellis, C , Lagiou, A , Benetou, V , Naska, A , and Trichopoulos, D. Physical activity and energy intake selectively predict the waist-to-hip ratio in men but not in women. Am J Clin Nutr, 74: 574-8, 2001. 164. Dunn, B. P , and Hislop, T. G. Personal Communication, 2001. 165. Block, G , Woods, M , Potosky, A , and Clifford, C. Validation of a self-aclministered diet history questionnaire using multiple diet records. J Clin Epidemiol, 43: 1327-35, 1990. 166. NCI. Health Habits and History Questionnaire: Diet History and Other Risk Factors: NIH, 1988. 167. Statistics-Canada. Body Mass Index and Health, 2002: Statistics Canada, 2002. 168. Immulite. C-peptide assay kit insert. Los Angeles CA: Diagnostic Products Corporation, 2002. 169. Cobas-Integra. Fmctosamine assay kit insert. Los Angeles CA: Roche Diagnostics, 2002. 185 170. HICL. Reference Manual, 2001: Hospitals In-Common Laboratory Web Site.; accessed 26 Sep 2001; available from http://www.hicl.on.ca/, 2001. 171. Woodward, M . Sample Size Determination, Epidemiology: Study Design and Data Analysis, pp. 329-65. Boca Raton: Chapman & HaU/CRC, 1999. 172. Immulite. Insulin assay kit insert. Los Angeles CA: Diagnostic Products Corporation, 2002. 173. Immulite. SHBG assay kit insert. Los Angeles CA: Diagnostic Products Corporation, 2002. 174. Rothman, K. J , and Greenland, S. Modern Epidemiology. Philadelphia: Lippincott-Raven, 1998. 175. Dietsys Version 3.0 User's Guide Version 3.0. Bethesda: Block Dietary Data Systems, National Cancer Institute. 176. Kaaks, R , Toniolo, P , Akhmedkhanov, A , Lukanova, A , Biessy, C , Dechaud, H , Rinaldi, S, Zeleniuch-Jacquotte, A , Shore, R. E , and Riboli, E. Serum C-Peptide, Insulin-Like Growth Factor (IGF)-I, IGF-Binding Proteins, and Colorectal Cancer Risk in Women. J Nad Cancer Inst, 92:1592-1600, 2000. 177. Willett, W. C , Dietz, W. H , and Colditz, G. A. Guidelines for healthy weight. N Engl JMed, 34/: 427-34, 1999. 178. Jernstrom, H , and Barrett-Connor, E. Obesity, weight change, fasting insulin, proinsulin, C-peptide, and insulin-like growth factor-1 levels in women with and without breast cancer: the Rancho Bernardo Study. J Womens Health Gend Based Med, 8:1265-72, 1999. 179. Barnes-Josiah, D , Potter, J. D , Sellers, T. A , and Himes, J. H . Early body size and subsequent weight gain as predictors of breast cancer incidence (Iowa, United States). Cancer Causes Control, 6:112-8, 1995. 180. Price, T. M , O'Brien, S. N , Welter, B. H , George, R , Anandjiwala, J , and Kilgore, M . Estrogen regulation of adipose tissue lipoprotein lipase—possible mechanism of body fat distribution. Am J Obstet Gynecol, 178:101-7, 1998. 181. Moore, D. B , Folsom, A. R , Mink, P. J , Hong, C. P , Anderson, K. E , and Kushi, L. H. Physical activity and incidence of postmenopausal breast cancer. Epidemiology, /1: 292-6, 2000. 182. Unruh, A. M , Smith, N , and Scammell, C. The occupation of gardening in life-threatening illness: a qualitative pilot project. Can J Occup Ther, 67: 70-7, 2000. 186 183. Rosso, S, Faggiano, F , Zanetti, R , and Costa, G. Social class and cancer survival in Turin, Italy. J Epidemiol Community Health, 51:30-4, 1997. 184. Kushi, L. H , Kaye, S. A , Folsom, A. R , Soler, J. T , and Prineas, R. J. Accuracy and reliability of self-measurement of body girths. Am J Epidemiol, 128: 740-8, 1988. 185. Weaver, T. W , Kushi, L. H , McGovern, P. G , Potter, J. D , Rich, S. S, King, R. A , Whitbeck, J , Greenstein, J , and Sellers, T. A. Validation study of self-reported measures of fat distribution. Int J Obes Relat Metab Disord, 20:644-50, 1996. 186. Petrelli, J. M , Calle, E. E , Rodriguez, C , and Thun, M . J. Body mass index, height, and postmenopausal breast cancer mortality in a prospective cohort of US women. Cancer Causes Control, 13: 325-32, 2002. 187. Block, G , Hartman, A. M , and Naughton, D. A reduced dietary questionnaire: development and validation. Epidemiology, /: 58-64, 1990. 188. Willett, W. C. Reproducibility and validity of food frequency questionnaires. Nutritional Epidemiology, pp. 101-47. New York: Oxford University, 1998. 189. Block, G , Thompson, F. E , Hartman, A. M , Larkin, F. A , and Guire, K. E. Comparison of two dietary questionnaires validated against multiple dietary records collected during a 1-year period. J Am Diet Assoc, 92: 686-93, 1992. 190. McCann, S. E , Trevisan, M , Priore, R. L , Muti, P , Markovic, N , Russell, M , Chan, A. W , and Freudenheim, J. L. Comparability of nutrient estimation by three food frequency questionnaires for use in epidemiological studies. Nutr Cancer, 35: 4-9, 1999. 191. Willett, W. C. Nutritional Epidemiology. New York: Oxford University, 1998. 192. Hennekens, C. H , and Buring, J. E. Epidemiology in Medicine. Boston: Little, Brown and Company, 1987. 193. Kazer, R. R. Insulin resistance, insulin-like growth factor I and breast cancer: a hypothesis. Int J Cancer, 62:403-6, 1995. 194. Nagata, C , Shimizu, H , Takami, R , Hayashi, M , Takeda, N , and Yasuda, K. Relations of insulin resistance and serum concentrations of estradiol and sex hormone-binding globulin to potential breast cancer risk factors. Jpn J Cancer Res, 91: 948-53, 2000. 195. Giovannucci, E. Insulin-like growth factor-I and binding protein-3 and risk of cancer. HormRes,5/:34-41,1999. 187 196. Giovannucci, E , Pollak, M . N , Platz, E. A , Willett, W. C , Stampfer, M. J , Majeed, N , Colditz, G. A , Speizer, F. E , and Hankinson, S. E. A prospective study of plasma insulin-like growth factor-1 and binding protein-3 and risk of colorectal neoplasia in women. Cancer Epidemiol Biomarkers Prev, 9:345-9, 2000. 197. Flilakivi-Clarke, L. Estrogens, BRCA1, and breast cancer. Cancer Res, 60: 4993-5001, 2000. 198. Pike, M. C , Spicer, D. V , Dahmoush, L , and Press, M . F. Estrogens, progestogens, normal breast cell proliferation, and breast cancer risk. Epidemiol Rev, 15:17-35, 1993. 1.99. Kaaks, R , Lundin, E , Rinaldi, S, Manjer, J , Biessy, C , Soderberg, S, Lenner, P , Janzon, L , Riboli, E , Berglund, G , and Hallmans, G. Prospective study of IGF-I, IGF-binding proteins, and breast cancer risk, in northern and southern Sweden. Cancer Causes Control, 13: 307-16, 2002. 200. Wynder, E. Ernst Wynder, M . D , President, American Health Foundation. Interview by James A. Johnson. J Healthc Manag, 43:107-9, 1998. 188 Appendix A - Questionnaire A STUDY OF LIFESTYLE IN WOMEN WITH BREAST CANCER This study i s conducted by the Division of Epidemiology at the British Columbia Cancer Agency, 600 West 10th Avenue, Vancouver, B.C.. The information that you give v i l l be held in strictest confidence and w i l l be used for research purposes only. Your consent to participate i s indicated by completing the questionnaire. Would you like a copy of the results of our study? • Yes • N O 1. What i s your present marital status? CU single (never have been married) D married (including common-lav) widowed divorced or separated • • 2. a) Are you presently employed? • N O Q Yes, employed outside of my home O Yes, employed within my home b) Are you a student? • N O • Yes 3. What i s your ethnic background? D white D Asian Q East Indian D Black D other 1^0 4. a c t i v ^ " * * ? * 0 Y ° U p a r t i c i p a t e i n t n e Allowing groups or Go to church or tempi* Participate i n group acti v i t i e s (such as clubs, PTA, professional, labor or service groups) More than Once A Week • • About Once A Week • • A Few Times A Month • • A Fev Times A Year • • Rarely or Never • • How often do you v i s i t or talk on the phone to one of the following (excluding those who live with you)? At least once a day Relative Friend At least once a week At least once a month • • • • • • Less often • • In the event of domestic or emotional problems or other stressful situations, do you have anyone on whom you eould c a l l for support or help (eg relative, neighbour, friend)? If No Yes How many are there that you feel you could rely on? Relative • • -» • Neighbour • • -¥ • Friend • • • 7. Do you find that your family doctor i s an important source of support or help? D No, not at a l l • Yes, somewhat • Yes, very much 8. Is faith or religion important to you? • No, not important at a l l • Yes, somewhat • Yes, very important :,..JW9t Is your f a i t h an important source of comfort when you are troubled? • No, i t i s of l i t t l e or no comfort • Yes, i t i s of some comfort • Yes, i t i s of much comfort 9. Do you own a pet? • NO • Yes 10. How t a l l are you? 11. How much do you weigh now? 12. What is your waist measurement? 13. What is your hip measurement? 4 What ia tha heaviest that you have been (not including your weight during and just after pregnancy)? 15. How did your weight compare to others of your height and age: much lighter than most As a teenager: 5 years ago:, • • somewhat lighter than many • • about average somewhat heavier than most • • • • much heavier than most • • 16. How many flights of stairs on average do you climb each day? (10 steps = 1 flight) CU None • 1 to 4 flights 5 to 8 flights 9 or more flights 17. How many c i t y blocks on average do you walk each day? (12 blocks = 1 mile) Q None • 1 to 4 blocks 5 to 8 blocks Cl 9 or more blocks Here i s a l i s t of active things that people do in their free time. How of ten do you do any of these things? (If you do not do the activity, then check "rarely or never".) Active sports More than once a week • About once a week • A few times a month • A few times a year • Rarely or never • Doing physical exercise • • • • • Jogging or running • • • • • Swimming or taking long walks • • • • • Gardening, fishing, hunting • • • • • Something else (Please Specify): • • • • • Have you ever been pregnant? Yes a) How many children (liveborn « stillborn) have you given birth to? b) How old were you when your f i r s t child was born? Do you s t i l l have your periods? C U • Yes a) At what age did they stop? /9</ 20. Do you s t i l l have your periods? (cont'd) b) Have you had a hysterectomy? • N O • Yes c) Have you had your ovaries removed? (oophorectomy) • M O • Yes, one I I Yes, both I I not sure 21. Has any of the following blood relatives ever had breast cancer? I f Yes: How old was she Did i t involve when i t was one or both detected? breasts? No Yea one both Mother • • - • • sister • • - • • Daughter • • - • • 22. Have you ever had a f i r s t degree relative (parents, brother, sister) who died of heart disease or stroke before they were 60? • N O • Yes ( /95 7 23. Have you ever been told by a doctor that you had any of the following conditions? Mo Yes Don't Know High Blood Pressure Asthma Hay Fever Diabetes Rheumatoid Arthritis other Arthritis Abnormal Pap smear Heart Attack Heart Disease or Angina Stroke 24. Do you now smoke or have you ever smoked cigarettes for more than 6 months in all? • NO • Yes S i ! ! ' a) Have you stopped smoking? Yes How old were you when you stopped? b) For how many years have you smoked regularly? c) On average, how many cigarettes did you smoke per day? 26. 25. Do you now drink or have you ever drunk alcoholic beverages? • H O • Yes a) Have you stopped drinking alcoholic beverages? • no • Yes How old were you when you stopped? b) How often do (did) you drink alcoholic beverages? At least At least At least Less Never once a once a once a often day week month Beer • • • • • Vine • • • • • Liquor • • • • • During the past year have you taken any vitamins or minerals? • N O • Yes, f a i r l y regularly • Yes, but not wmmt •X-:«-»»:W»KW::3«-regularly What do you take? # Of FILLS per DAY, WEEK, etc. Multiple Vitamins One-a-day type Stress-tabs type Centrum or Parameds Other Vitamins Vitamin A Vitamin C Vitamin E Calcium or dolomite p i l l s per p i l l s per p i l l s per p i l l s per p i l l s per p i l l s per p i l l s per How many milligrams or IUs per p i l l ? IU per p i l l mg per p i l l IU per p i l l mg per p i l l ( 26. During the past year have you taken any vitamins or minerals? (cont'd) Other (what?) l 4 6 Yeast 2 Iron 5 Cod l i v e r o i l 7 Selenium 3 Beta-carotene Other: Zinc Please l i s t the brand of multiple vitamin/mineral you usually take: 27. How often do you eat the following foods from restaurants or fast food places? RESTAURANT FOOD Almost every day 2-4 times a week Once a week 1-3 times a month 5-10 times a year 1-4 times a year Never, or less than once a year Fried chicken Burgers Pizza Chinese food Fried f i s h pOther foods 28. Seldom/ Sometimes Often/ Never Always How often do you eat the skin on chicken? How often do you eat the fat on meat? How often do you add salt to your food? How often do you add pepper to your food? • • • • • • • • • • • • )18 29 . 30. 31. 10 Mot counting juices, bov many servings of fruits do you usually eat per day or per week? per fruits day, week Mot counting salad or potatoes, about how many servings of vegetables do you eat per day or week? per vegetables day, week This section i s about your usual eating habits over the past year. F i r s t : Seconds Mark whether your usual serving size i s small, medium or large. Please DO NOT OMIT serving size. Mark the column to show how often, on the average, you ate the food during the past year. Please BE CAREFUL which column you put your answer in. Additional Comments: Please DO NOT SKIP any foods. If you never eat a food, mark "Never or less than once a month." A small serving i s about one-half the medium serving size shown, or less. A large serving i s about one-and -a-half times the medium serving size shown, or more. Sample: This person ate a medium serving of rice about twice per month during the past year and never ate squash. TYPE OF FOOD QUANTITY AVERAGE USE LAST 1 fEAF YOUR SERVING NEVER OR LESS THAN ONCE PER MONTH 1 PER 2-3 PER 1 PER 2 PER 3-4 PER 5-6 PER 1 PER 2 PER MEDIUM SERVING S M L MONTH WEEK DAY Rice 3/4 cup X X Winter squash, baked squash 1/2 cup X 1 TYPE OF FOOD QUANTITY AVERAGE USE L A S T YEAR FRUITS AND JUICFS YOUR SERVING NEVER OR LESS 1 PER MC 2-3 PER )NTH 1 PER 2 PER W 3-4 PER EEK 5-6 1 PER I PER ( )A MEDIUM SERVING S M L THAN ONCE PER MONTH Apples, applesauce, pears Cantaloupe (In season) 1 medium or 1/2 cup 1/4 medium Oranges Orange Juice or grapefruit juice Grapefruit 1 medium 6 ounce glass 1/2 medium Other fruit juices, fortified fruit drinks VEGETABLES w • • • w I • • I 6 ounce glass Beans such as baked beans, pintos, kidney, llmas or beans In chilli 374 cup • Tomatoes, tomato juice Broccoli 1 medium or 6 ounce glass 1/2 cup Spinach Mustard greens, turnip greens, collards 1/2 cup 1/2 cup Cole slaw, cabbage, sauerkraut 1/2 cup Carrots, or mixed vegetables containing carrots 1/2 cup Green Salad medium bowl Regular salad dressing & mayonnaise, including on sandwiches 2 tablespoons French fries and fried potatoes Sweet potatoes, yams 3/4 cup /2 cup Other potatoes, including boiled, baked, mashed & potato salad Rice % medium or 1/2 cup V4 cup | 1 T Y P E OF FOOD QUANTITY AVERAGE USE LAST! fEAF 1 YOUR SERVING NEVER OR LESS THAN ONCE PER MONTH 1 PER 2-3 PER 1 PER 2 PER 2-4 PER 54 PER 1 PER p MEDIUM SERVING S M L MONTH WEEK DAY MEAT, FISH, POULTRY LUNCH ITEMS — — — Hamburgers, cheeseburgers, meatloaf 1 medium or 4 ounces Beef (steaks roasts, etc. including on sandwiches) 4 ounces Beef stew or pot pie with carrots or other vegetables 1 cup Liver, including chicken liver 4 ounces Pork, including chops, roasts 2 chops or 4 ounces Fried chicken 2 small or 1 large piece Chicken or turkey (roasted, stewed or broiled, including on sandwiches) 2 small or 1 large piece Fried fish or fish sandwich 4 ounces or 1 sandwich Other fish (broiled or baked) 2 pieces or 4 ounces Spaghetti, lasagna, other pasta with tomato sauce 1 cup Hot dogs 2 hot dogs Ham, bologna, salami and other lunch meats 2 slices or 2 ounces Vegetable & tomato soups, including vegetable beef, ministrone 1 medium bowl BREADS, SNACKS, SPREADS White breads, (including sandwiches, bagels, buns French or Italian bread) 2 slices or 1 large bun Dark breads, such as whole wheat, rye, pumpernickel 2 slices Corn bread, corn muffins, corn tortillas, or grits 1 medium piece Salty snacks such as chips and salted popcorn 2 handfuls Peanuts, peanut butter 2 tablespn. Butter on bread or vegetables 2 pats Margarine on bread or {vegetables 2 pats TYPE OF FOOD BREAKFAST FOODS High fiber bran or granola cereals, shredded wheat Highly fortified cereals, such as Product 19, Total QUANTITY AVERAGE USE LAST YEAR YOUR SERVING MEDIUM SERVING 1 medium bowl 1 medium bowl [other cold cereals, such as com flakes, Rice Krisples 1 medium bowl Icooked cereals 1 medium I bowl Eggs. Bacon Sausage SWEETS Ice cream Doughnuts, cookies, cake, >asti |Ples (Chocolate candy 2 « g g * 2 slices 2 patties or links T scoop or 1/2 cm 1 piece, or cookies 1 med. slice 1 small bar or 1 ounce M NEVER OR LESS THAN ONCE PER MONTH 1 PER 2-3 PER MONTH 1 PER 2 PER 3-4 PER 5-6 PER WEEK 1 PER PI DAY BEVERAGES AND DAIRY PRODUCTS Cheeses and cheese spreads Whole milk and beverages with whole milk (not Including on cereal 2% milk and beverages with 2% milk (not Including on cereal Skim milk, 1% milk or buttermilk (not including on cereal) 2 slices or 2 ounces 8 ounce glass i 18 ounce glass 8 ounce glass Regular soft drinks (not diet soft drinks) 1 can or bottle or 1 large glass Beer 1 can or bottle I Wine or wine coolers ILIquo 1 medium glass 11 shot Milk or cream in coffee or 1 tea tablespoon Sugar In coffee or tea or on |2 teaspoons cereal Thank you very much for taking the time to fill out this questionnaire. Please take a moment to fill In any questions you may have skipped. Appendix B - Power Calculation n{X - \)Jn~r~-za (r + \)Jpc(\-pc) 2 = " yj(r +1) (ATT (1 - XK) + r/r(l - n)) • where X = relative risk to be detected • 7i = estimated chance of death = 105/603 = 0.174 • z a = 1.96 for p=0.05, 2-sided • n = sample size • r = number in highest quartile of insulin/number in lowest quartile of insulin • pc = (7r(rX+l))/(r+l)= [0.174((1)(2) + 1))/(1+1) = 0.261. Appendix C - Food items in each Food Group Fruits and Juices * Apples, applesauce, pears * Cantaloupe (in season) * Oranges * Orange juice or grapefruit juice * Grapefruit * Other fruit juices, fortified fruit drinks Vegetables * Beans such as baked beans, pintos, kidney, limas or beans in chili * Tomatoes, tomato juice Broccoli * Spinach * Mustard greens, turnip greens, collards * Cole slaw, cabbage, sauerkraut * Carrots, or mixed vegetables containing carrots * Green salad * French fries and fried potatoes * Sweet potatoes, yams * Other potatoes, including boiled, baked, mashed and potato salad * Rice Meat, Fish, Poultry * Hamburgers, cheeseburgers, meatloaf * Beef (steaks, roasts, etc) * Beef stew or pot pie with carrots or other vegetables * Liver * Pork Fried chicken * Chicken or turkey (roasted, stewed or broiled) * Fried fish * Other fish (broiled or baked) * Spaghetti, lasagne, other pasta with tomato sauce * Hot dogs * Ham, bologna, salami and other lunch meats * Vegetable & tomato soups including vegetable beef, minestrone Breads, Snacks, Spreads * White breads (including sandwiches, bagels, buns, French or Italian * Dark breads (including wheat, rye, pumpernickel) * Corn bread, corn muffins, corn tortillas, or grits * Salty snacks, such as chips and salted popcorn * Peanuts, peanut butter * Butter on bread or vegetables * Margerine on bread or vegetables Regular salad dressing & mayonnaise Breakfast Foods * High fiber bran or granola cereals, shredded wheat * Highly fortified cereals such as Product 19, Total * Other cold cereals, such as corn flakes, Rice Krispies * Cooked cereals * Eggs Bacon * Sausage Sweets * Ice cream * Doughnuts, cookies, cakes, pastry * Pies * Chocolate candy Beverages and Dairy Products * Cheeses and cheese spreads * Whole milk and beverages with whole milk (not including on cereal) * 2% milk and beverages with2%milk (not including on cereal) * Skim milk and beverages with skim milk (not including on cereal) * Regular soft drinks (not diet) * Beer * Wine or wine coolers * Liquor * Milk or cream in coffee or tea * Sugar in coffee or tea or on cereal 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items