Vitamin D and Breast Cancer Risk

by

Laura Nicole Anderson

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy

Dalla Lana School of Public Health University of Toronto

© Copyright by Laura Nicole Anderson 2010

ii

Vitamin D and Breast Cancer Risk

Laura Nicole Anderson

Doctor of Philosophy

Dalla Lana School of Public Health University of Toronto

2010

Abstract

It has long been known that vitamin D is important for calcium absorption and bone health. More

recently, vitamin D has been found to modulate breast cancer cell growth and increasingly

epidemiologic studies suggest vitamin D may be associated with reduced breast cancer risk. The

primary objective of this thesis was to evaluate the associations between vitamin D from all

sources (food, supplements and sunlight exposure) and breast cancer. Secondary objectives were

focused on methodological issues including the development of a solar vitamin D score and

adapting the measurement of vitamin D from foods for use among Canadians. The data source

for this study was the “Ontario Women’s Diet and Health Study”, a population-based case-

control study of women in Ontario. Cases (n = 3,101) diagnosed between 2002 and 2003 were

identified through the Ontario Cancer Registry and controls (n = 3,471) were identified through

random digit dialing of Ontario households. Study participants completed mailed risk factor and

food frequency questionnaires. Vitamin D intake from supplements (>400 IU/day compared to

none) was found to be associated with reduced breast cancer risk (OR = 0.76; 95% CI: 0.59,

0.98). However, total vitamin D intake (from food and supplements) and intake from food alone

were not associated with breast cancer risk. Time spent outdoors during 4 periods of life

(including adolescence) was associated with reduced breast cancer (e.g., highest versus lowest

categories of exposure at age 40 to 59: OR = 0.74; 95% CI: 0.61, 0.88). The novel solar vitamin

iii

D score, derived from time spent outdoors, skin color, sun protection practices, and ultraviolet

radiation of residence, was also associated with reduced breast cancer risk. In summary, there is

some evidence to suggest that vitamin D intake from supplements and determinants of cutaneous

vitamin D production are associated with reduced breast cancer risk.

iv

Acknowledgments

I am grateful to numerous people who contributed to this thesis. I extend my sincerest

appreciation to my supervisor, Dr. Michelle Cotterchio, for her guidance and support and for

making this process an enriching and valuable experience. I would like to acknowledge my

thesis committee members Drs. Vicki Kirsh, Julia Knight, and Reinhold Vieth for contributing

substantial time and effort to my training and this resultant thesis. I would also like to thank Dr.

Julia Knight for providing additional mentorship and direction.

It is a pleasure to thank many people at Cancer Care Ontario including Nancy Deming, Noori

Chowdhury, Todd Norwood, Patrick Brown, Lucia Mirea, and Beatrice Boucher. I am

particularly grateful to Beatrice Boucher for her constant encouragement and nutritional

expertise. I would also like to recognize Torin Block at NutritionQuest and Dr. Vitali Fioletov at

Environment Canada for sharing their expertise. Many thanks also to the vitamin D journal club

members for provoking interesting discussion and the valuable exchange of relevant information.

I would also like to express many thanks to my friends, including my fellow PhD students, for

their encouragement over the past 4 years. Last of all, I would like to extend my heartfelt

gratitude to my family, and to Sean McIntyre and his family for their enduring support.

This research would not have been possible without the generosity of all women who

participated in this study. The study was supported by grants from the Canadian Breast Cancer

Research Alliance and Canadian Breast Cancer Foundation - Ontario Chapter. I received

financial support through a Doctoral Award in Public Health Research from the Canadian

Institutes for Health Research.

v

Table of Contents

Acknowledgments........................................................................................................................iv

Table of Contents .........................................................................................................................v

List of Tables ...............................................................................................................................viii

List of Figures ..............................................................................................................................xi

List of Appendices .......................................................................................................................xii

List of Abbreviations ...................................................................................................................xiii

Chapter 1 Introduction and Objectives ........................................................................................1

1.1 Introduction ......................................................................................................................1

1.2 Study Objectives ..............................................................................................................2

Chapter 2 Background and Literature Review.............................................................................3

2.1 Breast Cancer ...................................................................................................................3

2.1.1 Breast Cancer Biology, Screening and Treatment ...............................................3

2.1.2 Burden of Breast Cancer in Canada .....................................................................4

2.1.3 Breast Cancer Risk Factors ..................................................................................5

2.2 Vitamin D Background ....................................................................................................11

2.2.1 Sources of Vitamin D...........................................................................................11

2.2.2 Vitamin D Biologic Action ..................................................................................14

2.2.3 Recommended Vitamin D Intake from Diet ........................................................16

2.2.4 Determinants and Optimal Level of 25(OH)D ....................................................17

2.3 Epidemiologic Studies of Vitamin D and Breast Cancer.................................................18

2.3.1 Reviews and Meta-analyses .................................................................................18

2.3.2 Trials ....................................................................................................................19

2.3.3 Biomarker Studies ................................................................................................20

2.3.4 Cohort Studies ......................................................................................................23

vi

2.3.5 Case-Control Studies ...........................................................................................27

2.3.6 Ecologic and Sun Exposure Proxy Studies ..........................................................30

2.4 Factors that May Influence or Modify the Vitamin D and Breast Cancer Association ...31

2.4.1 Calcium ................................................................................................................31

2.4.2 Timing of Exposure to Vitamin D .......................................................................32

2.4.3 Hormone Receptor Status (ER/PR) .....................................................................33

2.4.4 Genetic Variants...................................................................................................33

2.5 Vitamin D and Breast Cancer Mortality, Prognosis or Precursors ..................................34

2.6 Appraisal of the Vitamin D and Breast Cancer Literature ...............................................35

2.7 Summary and Rationale for the Current Study ................................................................38

Chapter 3 Study Methods.............................................................................................................40

3.1 Data Source and Study Design ........................................................................................40

3.2 Identification of Cases and Controls ................................................................................40

3.2.1 Cases ....................................................................................................................40

3.2.2 Controls ................................................................................................................41

3.3 Data Collection ................................................................................................................41

3.4 Variable Definitions .........................................................................................................43

3.4.1 Vitamin D and Calcium from Food and Supplements .........................................43

3.4.2 Individual Variables Related to Cutaneous Vitamin D Production .....................44

3.4.3 Derivation of a Solar Vitamin D Score ................................................................48

3.4.4 Potential Confounders ..........................................................................................49

3.5 Statistical Analysis ...........................................................................................................54

3.6 Ethics................................................................................................................................57

Chapter 4 Study Results ...............................................................................................................58

4.1 Overview of Results .........................................................................................................58

vii

4.2 Paper 1: Vitamin D Intake from Food and Supplements among Ontario Women Based

on the US Block Food Frequency Questionnaire with and without Modification for

Canadian Food Values ....................................................................................................59

4.3 Paper 2: Vitamin D and Calcium Intakes and Breast Cancer Risk in Pre- and

Postmenopausal Women ..................................................................................................72

4.4 Paper 3: Ultraviolet Sunlight Exposure and Breast Cancer Risk: A Population Based

Case-Control Study in Ontario ........................................................................................97

Chapter 5 Discussion and Conclusions ........................................................................................123

5.1 Summary of Findings and Comparison to the Literature.................................................123

5.1.1 Objectives 1 and 5 ................................................................................................123

5.1.2 Objectives 2 and 4 ................................................................................................125

5.1.3 Objective 3 ...........................................................................................................126

5.2 Limitations and Methodological Issues ...........................................................................127

5.2.1 Selection Bias.......................................................................................................127

5.2.2 Information Bias ..................................................................................................129

5.2.3 Confounding and Effect Modification .................................................................132

5.2.4 Analytical Issues ..................................................................................................133

5.2.5 External Validity ..................................................................................................134

5.2.6 General Limitations .............................................................................................135

5.3 Study Strengths ................................................................................................................135

5.4 Causation and Future Studies...........................................................................................136

5.5 Conclusions and Public Health Importance .....................................................................137

References ....................................................................................................................................140

Appendices ...................................................................................................................................162

viii

List of Tables

Chapters 1-3

Table 1. Amount of vitamin D in selected foods and supplements (values obtained from Health

Canada, Canadian Nutrient File, 2007)......................................................................................... 14

Table 2. Summary of previous studies of serum 25(OH)D and breast cancer risk ..................... .22

Table 3. Summary of previous cohort studies of vitamin D (from diet or sunlight) and breast

cancer risk ..................................................................................................................................... 25

Table 4. Summary of previous case-control studies of vitamin D (from diet or sunlight) and

breast cancer risk........................................................................................................................... 29

Table 5. Distribution of breast cancer cases and controls and age-group adjusted odds ratio

(AOR) estimates for selected known and suspected breast cancer risk factors ............................ 51

Table 6. Spearman rank correlations (rs) between physical activity and time outdoors per week at

4 age periods of exposure ............................................................................................................. 55

Table 7. Spearman rank correlations (rs) between vitamin D and calcium from food, supplements

and total combined (food and supplements) intake ...................................................................... 56

Chapter 4

Paper 1

Table 1. Vitamin D food values assigned to the standard (US) nutrient analysis and the modified

Canadian analysis.......................................................................................................................... 69

Table 2. Distribution of subject characteristics among all participating Ontario women ............ 70

Table 3. Distribution of vitamin D intake among Ontario women (total and stratified by age

group) ............................................................................................................................................ 71

Paper 2

Table 1. Distribution of selected characteristics and age-group adjusted odds ratio (OR) estimates

among 3,101 breast cancer cases and 3,471 controls in the Ontario Women’s Diet and Health

study .............................................................................................................................................. 87

ix

Table 2. Distribution of breast cancer cases (n = 3101) and controls (n = 3471) and odds ratio

(OR) estimates for intake of selected foods and supplements (frequency and duration) known to

contain vitamin D or calcium among Ontario women ...................................................................89

Table 3. Distribution of breast cancer cases (n = 3101) and controls (n = 3471) and odds ratio

(OR) estimates for derived vitamin D and calcium nutrient intake from food, supplements and

total combined among Ontario women ......................................................................................... 92

Table 4. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for

vitamin D intake variables stratified by total calcium intake .......................................................94

Table 5. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for

vitamin D and calcium variables stratified by menopausal status among Ontario women ...........95

Paper 3

Table 1. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for sun

exposure related variables during 4 age periods ..........................................................................114

Table 2. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for

derived proxy measures of vitamin D from sunlight during 4 age periods, recent exposure only,

and cumulative life exposure ......................................................................................................117

Table 3. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for

derived proxy measures of vitamin D from sunlight during 4 age periods stratified by vitamin D

supplement use .............................................................................................................................119

Table 4. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for

combined solar vitamin D score and vitamin D from supplements created by cross-classification

......................................................................................................................................................120

Table 5. Application of the predicted 25(OH)D model from the Health Professionals Follow-Up

Study to our Ontario Women’s Diet and Health study data ........................................................121

Table 6. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for

predicted 25(OH)D using the Health Professionals Study algorithm ..........................................122

Chapter 5

Table 1. Corrected ‘true’ risk estimates for selected measures of validity (γ) and observed risk

estimates of 0.76 and 0.85............................................................................................................130

x

Appendix 3

Table 1. Missing data for each main variable examined .............................................................166

Appendix 4

Table 1. The association between ethnicity and sun protection practices at each age group of

exposure .......................................................................................................................................170

Table 2. Spearman rank correlations (rs) between time spent outdoors and sun protection

practices, erythemal UV and latitude during each age period of exposure .................................171

Table 3. Spearman rank correlations (rs) between time spent outdoors and parity, income and

education during each age period of exposure …………………………………………………172

Table 4. Distribution of breast cancer cases and controls and odds ratio (OR) estimates of

variables associated with cutaneous vitamin D production created by cross-classification during

4 age periods ................................................................................................................................173

Table 5. Sensitivity analyses for solar vitamin D score ...............................................................174

Table 6. Odds ratio (OR) estimates for derived solar vitamin D score during 4 age periods and

breast cancer risk overall and among lifelong residents of Canada only……………………… 175

xi

List of Figures

Chapter 3

Figure 1. Map of erythemal UV radiation (mW/m2) worldwide and locations where cases and

controls resided in their teenage years ......................................................................................... 47

Figure 2. Hypothesized model of vitamin D and breast cancer with details of the proposed

algorithm for the measurement of UV radiation conditional on factors that affect vitamin D

production……………………………………………………………………………………… 49

Appendix 4

Figure 1. Distributions of vitamin D intake from a) foods, b) supplements (multivitamins and

vitamin D or cod liver oil), and c) total vitamin D intake (food and supplements)…………… 169

xii

List of Appendices

Appendix 1. Questionnaires ......................................................................................................163

Appendix 2. Ethics approval .....................................................................................................165

Appendix 3. Missing data. ........................................................................................................166

Appendix 4. Supplementary analyses .......................................................................................168

Appendix 5. Power calculations ...............................................................................................176

xiii

List of Abbreviations

25(OH)D 25-hydroxyvitamin D

1,25(OH)2D 1,25-dihydroxyvitamin D

AI adequate intake

BBD benign breast disease

BDDS Block dietary data system

BMI body mass index

CI confidence interval

CNF Canadian Nutrient File

DRI dietary reference intake

IARC International Agency for Research on Cancer

ER estrogen receptor

FFQ food frequency questionnaire

HRT hormone replacement therapy

NHANES National Health and Nutrition Examination Survey

OCR Ontario Cancer Registry

OR odds ratio

PR progesterone receptor

RDA recommended dietary allowance

RR relative risk

SZA solar zenith angle

VDR vitamin D receptor

WHI Women’s Health Initiative

UV ultraviolet

USDA United States Department of Agriculture

1

Chapter 1 Introduction and Objectives

1.1 Introduction

A growing body of literature emerging from epidemiologic and laboratory studies has led to the

hypothesis that vitamin D may reduce breast cancer risk (as reviewed in Bertone-Johnson, 2007;

Bertone-Johnson, 2009; Colston, 2008; Cui & Rohan, 2006; Lipkin & Newmark, 1999; Perez-

Lopez, Chedraui, & Haya, 2009; Rohan, 2007); however, the epidemiologic study findings are

inconclusive and many gaps exist in the literature. The International Agency for Research in

Cancer (IARC) recently concluded there is evidence of an inverse association between vitamin D

and breast cancer risk but not sufficient evidence to conclude a causal effect exists (IARC,

2008). Sources of vitamin D include supplements and a few foods and, unlike any other vitamin,

vitamin D is produced in the skin following sunlight exposure. Thus, in addition to the usual

challenges of measurement in nutritional epidemiology, there are unique challenges in the

measurement of vitamin D for observational epidemiologic studies (Millen & Bodnar, 2008).

Furthermore, the association between vitamin D and breast cancer risk may be modified by other

factors (e.g., menopausal status, body fatness) or depend upon timing of exposure (Bertone-

Johnson, 2007; Bertone-Johnson, 2009; Rohan, 2007), but few previous studies have explored

these differences. Vitamin D, from both sunlight and diet, is important for calcium absorption

(Heaney, 2008) and both vitamin D and calcium have some common dietary sources (e.g.,

vitamin D fortified milk) (McCullough et al., 2005). Calcium may also have anti-cancer

properties but has less consistently been associated with reduced breast cancer risk (as reviewed

in Al Sarakbi, Salhab, & Mokbel, 2005; Cui & Rohan, 2006). Despite the biologic relationship

between calcium and vitamin D, few studies have included calcium in their investigations of

vitamin D and breast cancer.

The overall aim of this thesis was to investigate the associations between vitamin D from all

sources (food, supplements, and sunlight exposure) and breast cancer risk in a population-based

case-control study of over 6,000 Ontario women aged 25-74 years. This study improves upon the

aforementioned limitations and addresses some of the current gaps in knowledge. An emphasis

was placed on the measurement of vitamin D, including adaptation of a measure of dietary

vitamin D specific to our Canadian population, and the development of a unique algorithm

2

(incorporating time outdoors, ultraviolet radiation of residence, sun protection practices and skin

color) to measure vitamin D from sunlight. Vitamin D intake is modifiable and research

identifying modifiable breast cancer risk factors is important for prevention, as most well-

established risk factors are not modifiable (Rockhill, Weinberg, & Newman, 1998). The ultimate

aim of this study was to develop a better understanding of the association between vitamin D and

breast cancer among Ontario women.

A detailed literature review and rationale for the study is included in chapter 2, followed by

methods and related methodological results in chapter 3. This thesis was written in manuscript

format with 3 manuscripts in chapter 4. Chapter 5 contains a detailed discussion of study results,

methodological issues and overall conclusions.

1.2 Study Objectives

Primary Objectives

Objective 1. To evaluate the associations between breast cancer risk and vitamin D intakes from

food and supplements (and potential effect modification by calcium, menopausal

status or body mass index).

Objective 2. To evaluate the association between breast cancer risk and sunlight exposure

variables from adolescence through adulthood – as reported and using the solar

vitamin D score from objective 4 (and potential effect modification by calcium,

menopausal status or body mass index).

Objective 3. To evaluate the association between breast cancer risk and total vitamin D from all

sources (food, supplements and sunlight exposure).

Secondary Objectives

Objective 4. To develop and apply an algorithm to derive a solar vitamin D score (for objective

2).

Objective 5. To modify and compare vitamin D intake from a standard US nutrient analysis of a

food frequency questionnaire (FFQ) versus a modified nutrient analysis that reflects

additional vitamin D sources and Canadian food fortification (for objective 1).

3

Chapter 2 Background and Literature Review

2.1 Breast Cancer

2.1.1 Breast Cancer Biology, Screening and Treatment

Cancer of the breast results from unregulated cell growth and the new growth of undifferentiated

tissue (Medline Plus Merriam-Webster, 2009). Breast tissue consists of mammary glands and

ducts, made up of epithelial tissue, in addition to adipose, connective tissue and vessels of the

lymphatic and blood system. Development begins during puberty, is generally quite advanced by

menarche and final differentiation only occurs during pregnancy and lactation (Colditz, Baer, &

Tamimi, 2006). After menopause, hormone levels decline and breast cells do not continue to

divide. Breast cells are potentially more susceptible to exposures during the period from

menarche to first birth when breast tissue is undifferentiated (Colditz & Frazier, 1995; Okasha,

McCarron, Gunnell, & Smith, 2003; Russo, Moral, Balogh, Mailo, & Russo, 2005) or during

pregnancy when breast tissue is growing (Kelsey & Berkowitz, 1988). There is also some

evidence that earlier life exposures may be important (Kelsey & Berkowitz, 1988; Okasha et al.,

2003). Thus, exposures during critical periods of breast development may be more likely to

influence breast cancer risk.

Breast cancers are classified as in situ (non-invasive) or invasive (invading the breast stroma). In

situ breast cancers occur in the duct or lobule whereas the majority (>95%) of invasive breast

cancers are infiltrating ductal adenocarcinomas, cancers of the glandular epithelium (Colditz et

al., 2006; Kelsey & Bernstein, 1996). Breast cancers are further classified by stage and grade,

based on histopathology and differentiation of the tumor cells. Stage is determined based on

tumor size, regional lymph node involvement and distant metastasis (Singletary & Connolly,

2006). Treatment and prognosis are determined based on breast cancer stage (Bland et al., 1998).

Breast cancers are often further classified by time of diagnosis (pre- versus post- menopause),

and by their expression of human epidermal growth factor receptor 2 (HER2) and positive or

negative expression of hormone receptors (estrogen-receptors (ER) and progesterone-receptors

(PR)) (Colditz et al., 2006).

4

Women in Ontario between the ages of 50-69 years are recommended to undergo mammography

and clinical breast exams every 2 years; guidelines vary for women at high risk (e.g., family

history of breast cancer) (Lipskie, 1998). Mammography rates among these women have

increased from 40% in 1990 to 73% in 2008 (Shields & Wilkins, 2009). Breast cancer screening

is important for the early detection of breast cancer and has been associated with a 30%

reduction in mortality among women 50-69 years of age (Fletcher, Black, Harris, Rimer, &

Shapiro, 1993). Breast cancer diagnosis is confirmed by histology from tissue biopsy. Treatment

options depend on stage, hormone receptor status and other characteristics of the primary tumor.

Treatment of breast cancer often involves surgery, radiation therapy, and adjuvant systematic

therapies (e.g., chemotherapy, tamoxifen, aromatase inhibitors) (National Cancer Institute, 2009;

Steering Committee on Clinical Practice Guidelines for the Care and Treatment of Breast

Cancer, 1998; Veronesi, Boyle, Goldhirsch, Orecchia, & Viale, 2005).

2.1.2 Burden of Breast Cancer in Canada

The Ontario Cancer Registry (OCR) collects data on all cancer cases in Ontario (excluding non-

melanoma skin cancers) and data from the OCR are combined with other provincial and

territorial registries or data sources to create the Canadian Cancer Registry. Canadian Cancer

Statistics are published annually using these data and provide detailed information on the burden

of cancer in Canada (Canadian Cancer Society/National Cancer Institute of Canada, 2009).

Breast cancer is the most common cancer among Canadian women and it is estimated that

22,700 Canadian women were diagnosed with breast cancer in 2009. The lifetime probability of

developing breast cancer among females is 11%; this statistic takes into account the risk at each

stage of life using life tables to estimate average life expectancy. Age-standardized annual breast

cancer incidence rates remained relatively stable from 1991 to 2002 with about 100 cases per

100,000. A slight decrease was observed from 2003 to 2005 with approximately 97 cases per

100,000 (actual data not yet available for more recent years). In contrast, age-standardized breast

cancer mortality rates have decreased steadily from about 30 deaths per 100,000 in 1980 to an

estimated 22 deaths per 100,000 in 2009. This decrease is likely due to improved treatment and

early detection. Despite this decrease, breast cancer mortality remains a leading cause of cancer

mortality among females in Canada, second only to lung cancer. The 5-year relative survival

ratio for breast cancer is 87%. Breast cancer incidence rates in the US are highest among White

women, lower among Black women and the lowest among Asian women (Altekruse, Kosary,

5

Krapcho, Neyman et al., 2009); Canadian statistics are not available by race/ethnicity. Breast

cancer in men is much less common (<1% of all breast cancer cases) and not the focus of this

thesis.

The burden of breast cancer has significant public health implications in Canada. The average

lifetime treatment cost per case in Canada was $25,661 in 1995 corresponding to a total cost of

$454 million for all cases diagnosed in 1995 (Will et al., 2000). To the best of our knowledge

more recent data on costs in Canada are not available, however, we would suspect that costs have

continued to rise with increased survival. Canadian women with breast cancer also report

significant personal financial impact. In the year following breast cancer diagnosis working

women reported an average wage loss of 27% (Lauzier et al., 2008). Beyond the economic costs

and physical health effects, breast cancer has a significant impact on the psychological wellbeing

of patients, caregivers and ‘healthy’ women with a family history of breast cancer (e.g., Badger,

Segrin, Dorros, Meek, & Lopez, 2007; Lauzier et al., 2009; Maheu, 2009).

2.1.3 Breast Cancer Risk Factors

Although breast cancer is the most common cancer among women in North America, relatively

few breast cancer risk factors are well-established and most that are established are non-

modifiable. Multiple lines of evidence suggest that genetic and reproductive factors influence

breast cancer risk; however these factors are for the most part not readily modifiable for cancer

prevention. The subsequent section briefly reviews the known and suspected breast cancer risk

factors.

Genetic factors

Studies of breast cancer risk among twins suggest some breast cancer is attributable to hereditary

factors (Lichtenstein et al., 2000). Mutations in BRCA1 and BRCA2 genes lead to high-risk

genotypes associated with up to 80% lifetime risk of developing breast cancer; however, there is

much variation in penetrance which is likely due to non-genetic factors (as reviewed by Narod,

2006). Mutations in BRCA1 and BRCA2 are relatively rare and account for only 5-10% of all

breast cancers (Claus, Schildkraut, Thompson, & Risch, 1996; Martin & Weber, 2000). Other

more common genetic variants have also been associated with breast cancer risk, but these

explain only a small amount of the variation in breast cancer risk (as reviewed in Martin &

6

Weber, 2000). Approximately 10% of breast cancer cases have at least one first degree relative

with a history of breast cancer (Collaborative Group on Hormonal Factors in Breast Cancer,

2001). The risk of breast cancer increases substantially with number of affected relatives; the

relative risks of having 1, 2 or 3 first degree relative versus none are, respectively, 1.80 (99% CI:

1.69–1.91), 2.93 (99% CI: 2.36–3.64), and 3.90 (99% CI: 2.03–7.49) (Collaborative Group on

Hormonal Factors in Breast Cancer, 2001).

Breast cancer rates also differ substantially by country: the age-adjusted standardized incidence

rates in North American and Northern European countries are two to four times greater than in

Asian and Latin American countries (Althuis, Dozier, Anderson, Devesa, & Brinton, 2005).

Some of this variation may be attributable to differences in screening/diagnosis or reporting, but

lifestyle and genetic factors are also possible contributors (Althuis et al., 2005). Migrant studies

have shown that women moving from countries of low to high cancer risk develop the higher

rates of breast cancer of the new population (Andreeva, Unger, & Pentz, 2007; Kliewer & Smith,

1995). These findings suggest that although genetic factors may be important, environmental

factors also influence breast cancer risk.

Reproductive factors

As early as the 1700’s there was evidence suggesting that reproductive factors may be important

for breast cancer risk (Mustacchi, 1961). It is now well-established that nulliparity or lower

parity, older age at first full-term pregnancy, younger age at menarche and older age at

menopause are all associated with increased breast cancer risk (as reviewed in Colditz et al.,

2000; Kelsey, Gammon, & John, 1993; Veronesi et al., 2005). Nulliparous women have a 20-

70% increased risk of breast cancer compared to parous women and a comparable increase in

risk is observed among women greater than 30 years of age at first full-term pregnancy (Kelsey

et al., 1993). Among parous women, breast cancer risk decreases 7% on average for each

additional birth (Veronesi et al., 2005). However, breast cancer risk temporarily increases for a

few years following a birth, particularly the first one (Kelsey et al., 1993; Pathak, 2002). Breast

tissue becomes further differentiated during pregnancy and it is through these changes to breast

tissue, that parity and age at first birth are suspected to influence breast cancer risk (Kelsey et al.,

1993).

7

Breastfeeding has also been associated with breast tissue differentiation and several reviews have

concluded that lactation, or ever breastfeeding, is associated with reduced overall breast cancer

risk (Colditz et al., 2000; World Cancer Research Fund, 2007), premenopausal breast cancer

only (Lipworth et al., 2000) or only among premenopausal women with a family history of

breast cancer (Stuebe, Willett, Xue, & Michels, 2009). The association between being breastfed

as an infant and later breast cancer risk is less well-established, findings from individual studies

are inconclusive (e.g., Freudenheim et al., 1994; Nichols et al., 2008; Wise et al., 2009).

Whereas, results from a meta-analysis suggest having been breastfed may be protective against

pre- but not postmenopausal breast cancer (Martin, Middleton, Gunnell, Owen, & Smith, 2005).

Younger age at menarche and older age at menopause are two additional well-established breast

cancer risk factors (as reviewed in Colditz et al., 2000; Kelsey, Gammon, & John, 1993;

Veronesi et al., 2005), these reproductive-related variables are associated with increased estrogen

exposure. Consistent with the hypothesis that lifetime estrogen exposure increases breast cancer

risk (Colditz et al., 2006; Martin & Weber, 2000), circulating levels of estrogen (Bernstein &

Ross, 1993; Hankinson & Eliassen, 2007) and testosterone (Hankinson & Eliassen, 2007) have

also been associated with increased breast cancer risk, particularly among postmenopausal

women. Use of exogenous hormones including hormone replacement therapy, oral

contraceptives, and diethylstilbestrol (DES) (during pregnancy) also increases breast cancer risk

(Colditz et al., 2000; Veronesi et al., 2005). The Women’s Health Initiative (WHI) trial, where

postmenopausal women were randomized to hormone replacement therapy (estrogen plus

progestin) (n = 8506) or placebo (n = 8102), found HRT use increased breast cancer risk

(HR=1.26; 95% CI: 1.00, 1.59) (Rossouw et al., 2002). Immediately following publication of the

WHI trial (between 2002 and 2003), breast cancer incidence decreased by nearly 6% in Canada

and HRT prescription rates decreased by 27% (Kliewer, Demers, & Nugent, 2007), similar

decreases were observed in the US over this one year period (Ravdin et al., 2007). Between 2000

and 2005, HRT use has decreased by more than 50% in the US and it has been suggested that

this may explain the corresponding 9% decrease in breast cancer incidence observed over this

same time period (Coombs, Cronin, Taylor, Freedman, & Boyages, 2010).

Well-established non-reproductive risk factors

8

Most well-established non-reproductive breast cancer risk factors not directly related to genetics

are not highly modifiable and thus not very influential for cancer prevention. These include age,

height, higher socioeconomic status, benign breast disease (BBD), mammographic density,

ionizing radiation exposure (Colditz et al., 2000; Kelsey & Bernstein, 1996; Veronesi et al.,

2005). The risk of breast cancer increases markedly with increasing age; with the rate of increase

more pronounced under 50 years of age (Pike, Krailo, Henderson, Casagrande, & Hoel, 1983).

Risk of breast cancer is also positively associated with height (Colditz et al., 2000; Kelsey &

Bernstein, 1996); this may be due to increased energy intake during childhood or confounding by

body size or mammographic density (Hunter & Willett, 1993). Unlike many chronic diseases,

higher socioeconomic status is associated with increased breast cancer risk (Kelsey & Bernstein,

1996; Veronesi et al., 2005). This association may be due to confounding by other factors such

as delayed childbirth, increased screening, differences in body weight or exogenous hormone use

(Kelsey & Bernstein, 1996; Mettlin, 1992).

Benign breast disease (BBD) and mammographic density are considered well-established breast

cancer risk factors. BBD is a non-cancerous condition characterized by atypical hyperplasia or

cysts. Breast cancer risk among women with a history of BBD is 4 to 5 times greater (dependent

upon histological type) than women with no history of BBD (Hartmann et al., 2005; Kabat et al.,

2010). Mammographic density is determined based on the appearance of fat versus

epithelial/stromal breast tissue from mammogram images and is also strongly associated with

breast cancer risk (Boyd et al., 2005; Veronesi et al., 2005). High density (> 75% fibro-glandular

tissue) is associated with 5 times increased risk of breast cancer (Boyd et al., 2005; Veronesi et

al., 2005). Exposure to ionizing radiation, from x-rays or other sources of radiation, is also

associated with increased breast cancer risk (Colditz et al., 2000; Kelsey & Bernstein, 1996;

Veronesi et al., 2005), particularly high exposure during puberty (Ronckers et al., 2005; Veronesi

et al., 2005).

There are few well-established lifestyle breast cancer risk factors that are potentially modifiable:

alcohol intake, body fat and low physical activity. Alcohol consumption has been consistently

associated with increased breast cancer risk (e.g., Key, Hodgson, Omar, Jensen, Thompson, et

al., 2006; Colditz et al., 2000; Giovannucci et al., 1993; Hamajima et al., 2002; Singletary &

Gapstur, 2001; Terry et al., 2006; Tjonneland et al., 2007). A positive linear relationship has

been observed between daily alcohol intake and breast cancer risk and suggesting a 9 -10%

9

increased risk with each drink consumed (Smith-Warner et al., 1998) or per 10g ethanol (Key et

al., 2006). The association between alcohol intake and breast cancer risk is modified by common

genetic polymorphisms (e.g., Boffetta & Hashibe, 2006; Platek, Shields, Marian, McCann et al.,

2009) and limited evidence suggests the association may be attenuated by folic acid intake (as

reviewed in Linos, Holmes, & Willett, 2007).

The association between BMI or body fat and breast cancer risk varies substantially by

menopausal status. Increased BMI is inversely associated with premenopausal breast cancer risk

and positively associated with postmenopausal breast cancer risk (Colditz et al., 2000; Linos et

al., 2007; Stephenson & Rose, 2003; Veronesi et al., 2005). Increased body fatness is associated

with higher estrogen levels in postmenopausal women which may explain the increased breast

cancer risk observed among postmenopausal women (Cleary & Grossmann, 2009). In contrast,

body fatness may decrease risk in premenopausal women through anovulation (Linos et al.,

2007). Body fatness is also commonly associated with reduced physical activity, and higher

levels of physical activity have been associated with a 20 to 30% reduction in breast cancer risk

(as reviewed in Friedenreich & Cust, 2008; Warburton, Katzmarzyk, Rhodes, & Shephard,

2007). This association is strongest among postmenopausal women and, limited evidence

suggests, among hormone receptor negative breast cancers, women with normal BMI, and no

family history of breast cancer (Friedenreich & Cust, 2008). Physical activity may exert a

protective effect on breast cancer risk through decreased levels of BMI, hormones (e.g.,

estrogens), insulin resistance, and/or inflammation (e.g., c-reactive protein) (Neilson,

Friedenreich, Brockton, & Millikan, 2009). Changes in hormone levels, insulin resistance and

inflammation may also be influenced by BMI and perhaps contribute to the observed

associations between BMI and breast cancer risk.

Possible dietary risk factors

Only a few of the aforementioned established risk factors are readily modifiable: exogenous

hormone use, alcohol intake, physical activity, and obesity. There is conflicting evidence

regarding the association between other more common modifiable factors and breast cancer risk.

For example, a recent report by the World Cancer Research Fund (WCRF) and IARC (World

Cancer Research Fund, 2007) reviewed more than 40 specific dietary factors or groups of foods

(including milk, dairy products, vitamin D, and calcium) and concluded there was limited or no

10

evidence for nearly all dietary factors. Only the evidence for total dietary fat intake was judged

to be suggestive with respect to increasing breast cancer risk among postmenopausal women

only. However, a low-fat dietary intervention was not found to reduce breast cancer risk among

postmenopausal women in the WHI (Prentice et al., 2006). Other comprehensive reviews of

breast cancer risk factors have also concluded there are no well-established dietary breast cancer

risk factors (with the exception of alcohol and weight gain) (Colditz et al., 2000; Colditz, 2005;

Kelsey & Bernstein, 1996; Linos et al., 2007; Lof & Weiderpass, 2009; Michels, Mohllajee,

Roset-Bahmanyar, Beehler, & Moysich, 2007; Veronesi et al., 2005).

Several reviews have identified dietary factors for which there is limited evidence suggesting a

possible association with reduced breast cancer risk. These include phytoestrogens or soy foods

(Bissonauth, Shatenstein, & Ghadirian, 2008; Kelsey & Bernstein, 1996; Linos et al., 2007),

fruits (Bissonauth et al., 2008), vegetables (Bissonauth et al., 2008; Colditz et al., 2000), fish

(Bissonauth et al., 2008), monounsaturated fat (Bissonauth et al., 2008; Colditz et al., 2000),

dairy products (Bissonauth et al., 2008; Linos et al., 2007), calcium (Bissonauth et al., 2008; Cui

& Rohan, 2006; Linos et al., 2007) and vitamin D (Bissonauth et al., 2008; Cui & Rohan, 2006;

Linos et al., 2007). Possible dietary risk factors associated with increased risk include saturated

or total dietary fat (Bissonauth et al., 2008; Colditz et al., 2000; Linos et al., 2007), total energy

(Bissonauth et al., 2008), high glycemic index foods (Linos et al., 2007) and meat or specifically

red meat intake (Bissonauth et al., 2008; Linos et al., 2007; Taylor, Misra, & Mukherjee, 2009).

However, the literature on all of these factors is inconclusive. Emerging interest in dietary

patterns provides some evidence that a high fat or ‘Western’ pattern may be associated with

increased breast cancer risk (Lof & Weiderpass, 2009). Future studies are needed on all of these

dietary factors that should take into consideration timing of exposure, gene-environment

interactions and breast cancer characteristics (Bissonauth et al., 2008; Linos et al., 2007; Lof &

Weiderpass, 2009). Relatively few studies have evaluated childhood or adolescent diet and risk

of adult breast cancer and the results are generally inconsistent with few statistically significant

results (e.g., Frazier et al., 2004; Frazier et al., 2003; Linos et al., 2010)

Smoking and other possible environmental risk factors

Tobacco smoking, a well-established modifiable risk factor for other cancers (e.g., lung,

pancreas) (Colditz et al., 2000), has not been consistently associated with breast cancer risk

11

(Morabia, 2002; Palmer & Rosenberg, 1993; Terry & Rohan, 2002). Smoking and alcohol intake

are highly correlated and after adjustment for smoking, alcohol remains associated with breast

cancer risk but no effect of smoking is observed (Hamajima et al., 2002). There is somewhat

more consistent evidence, although not complete agreement, that environmental tobacco smoke,

also known as passive or secondhand smoke, is associated with increased breast cancer

especially in premenopausal women (Lee & Hamling, 2006; Miller et al., 2007, Johnson &

Glantz, 2008). Furthermore, the associations between smoking (active or passive) and breast

cancer risk may be modified by genotype (Terry & Goodman, 2006). Overall, the associations

between smoking and breast cancer are not well understood. There is great interest and public

perception that environmental contaminants and consumer products are associated with breast

cancer risk but there is currently insufficient evidence to support these associations (Brody et al.,

2007).

Risk factors by hormone receptor status

Increasingly studies are suggesting that risk factors may vary by breast cancer subtype. Triple

negative breast cancers (ER-, PR- , and HER2 -) are associated with a poorer prognosis than

hormone receptor positive or luminal type cancers and occur more frequency in African

Americans women than white women, are often diagnosed at younger ages and frequently

associated with BRCA1 mutations (as reviewed in Ray & Polite, 2010). Reproductive factors

traditionally associated with reduced overall breast cancer risk (e.g., later age at menarche, parity

and younger age at first birth) are associated with hormone receptor positive but not receptor

negative tumors (as reviewed in Althuis et al., 2004).

2.2 Vitamin D Background

2.2.1 Sources of Vitamin D

A unique property of vitamin D is that it can be produced endogenously in the skin following

sufficient sunlight exposure; specifically exposure to ultraviolet (UV) B radiation is required.

UVB wavelengths range from 280 nm to 315 nm; although shorter than the UVA range, UVB

wavelengths are more damaging. UVB radiation drives the conversion of provitamin D3 (7-

dehydrocholesterol) in the skin to previtamin D3 through isomerization. Previtamin D3 is then

12

spontaneously converted to vitamin D31 (Holick et al., 1980; MacLaughlin, Anderson, & Holick,

1982). The body strictly regulates the amount of vitamin D that can be produced in the skin at

any given time and additional sun exposure is not beneficial once maximal vitamin D production

is reached (Wolpowitz & Gilchrest, 2006). Within 24 to 48 hours of sufficient exposure to UVB

radiation, levels of the circulating form of vitamin D, 25-hydroxyvitamin D increase 5 to 10 fold

(Holick, 1987). The amount of time required for maximal vitamin D production depends on

several factors (both ecologic and person-specific); this may be as short as 5 minutes for

individuals with highly exposed lightly pigmented skin and is usually reached at suberythemal

(before skin reddening) doses (Wolpowitz & Gilchrest, 2006).

The strength of UVB that reaches the skin depends upon a number of factors. Sun exposure

varies depending upon the solar zenith angle (SZA) – the distance between the sun and zenith (a

point directly overhead) – which changes with time of year, time of day and geographic location.

Sun exposure is correlated with latitude; it is greatest near the equator and decreases with

increasing or decreasing latitude. At higher latitudes there is insufficient sunlight during the

winter months for the skin to produce vitamin D (Webb, Kline, & Holick, 1988). For example, in

Boston (42.2°N) no vitamin D is produced by skin from November to February, and in

Edmonton (52°N) this period extends from October to March. Other ambient factors that affect

UVB exposure, and hence vitamin D synthesis, include altitude, ozone, and other aerosols.

In addition to geographic factors and ambient UVB radiation there are person-specific variables

that affect endogenous vitamin D production: time spent outdoors, sun protection practices and

skin colour (Webb, 2006). Time spent outdoors is generally positively associated with increased

vitamin D, although maximal vitamin D production can be reached within short time periods

(Wolpowitz & Gilchrest, 2006) and frequent short periods of exposure may maximize vitamin D

intake. Clothing and sunscreen block the ability of UVB radiation to penetrate the skin, hence

decreasing vitamin D production (Matsuoka, Ide, Wortsman, MacLaughlin, & Holick, 1987).

Skin pigmentation (i.e., more melanin) acts as a natural sunscreen and people with darker skin

require more UVB exposure (e.g., longer time outdoors) to produce vitamin D. Adults with

whole body exposure to either one minimal erythemal dose (amount of time to produce minimal

1 Vitamin D3 is also known as cholecalciferol

13

skin reddening) or daily sun exposure (time unknown) have serum 25(OH)D increases equivalent

to oral doses of >10,000 IU vitamin D supplements (Stamp, Haddad, & Twigg, 1977; Holick,

1995). This value is regularly reported in the literature but is supported by limited data from

small studies.

Exogenous sources of vitamin D include foods and supplements. Very few foods naturally

contain vitamin D. Fatty fish, such as salmon or mackerel, contain relatively high amounts,

whereas other foods, such as meats, eggs, and shellfish, contain low quantities (Calvo, Whiting,

& Barton, 2004; Health Canada, 2007). Additionally, there has been mandatory fortification of

all fluid cows’ milk and margarine with vitamin D in Canada since the 1970’s and voluntary

fortification occurred much earlier (Health Canada, 2005). Some milk products (e.g., cheese,

yogurt) are made with fortified milk and discretionary fortification of some other items, such as

orange juice and milk beverage substitutes (e.g., soy milk), has been permitted since 2005 (as

reviewed in Sacco & Tarasuk, 2009) but these items are not universally enriched. Food

fortification practices differ between countries; using US nutrient values has been found to

underestimate Canadian vitamin D intakes (Csizmadi et al., 2007). Multivitamins, single product

vitamin D supplements and cod liver oil are all supplemental sources of vitamin D available to

Canadians. Table 1 summarizes the current vitamin D content of some foods in Canada.

There are two types of vitamin D from diet and supplements: D2 and D3. Vitamin D22 is derived

from plant sources, yeast and fungi whereas D3 is found in animal sources such as fish (and is

identical to vitamin D3 produced in skin). Current nutrient databases do not distinguish between

the two forms (Health Canada, 2007; United States Department of Agriculture, 2009). Either

form of vitamin D can be used in supplements and food fortification but most supplements now

contain vitamin D3. Historically vitamin D2 was thought to be equivalent to D3, however, some

studies (Armas, Hollis, & Heaney, 2004; Trang et al., 1998), but not all (Holick et al., 2008),

suggest that vitamin D2 may be less biologically active than D3 and may not be as effective as

vitamin D3 at maintaining serum 25(OH)D levels.

2 Vitamin D2 is also known as ergocalciferol

14

Table 1. Amount of vitamin D in selected foods and supplements (values obtained from Health

Canada, Canadian Nutrient File, 2007)

Source Amount of vitamin D

IU (µg) per serving 3

Serving size

Cod liver oil 1280 (32) 15 ml (14g)

Salmon 200 - 640 (5 - 16 ) 75g

Milk 100 (2.6 ) 260g (250mL)

Enriched orange juice 100 (2.6 ) 263g (250mL)

White fish (e.g., sole or halibut) 60 - 144 (1.5 – 3.6) 75g

Mushroom, shiitake 76 (1.9 ) 77g (125mL)

Margarine 76 (1.9 ) 14g (15ml)

Canned Tuna 60 (1.5) 165g (1 can)

Egg 28 (0.7) 50g (1 large egg)

2.2.2 Vitamin D Biologic Action

Vitamin D4 from all sources is not biologically active and must undergo hydroxylation in the

liver by 25-hydroxylase (CYP27A1) to produce 25-hydroxyvitamin D (25(OH)D) – the

circulating form of vitamin D (Holick, 2003; Schwartz & Blot, 2006). 25(OH)D circulates bound

to vitamin D binding protein (DBP). Limited evidence suggests that 25(OH)D3 has a higher

binding affinity for DBP than vitamin 25(OH)D2 (Houghton & Vieth, 2006). A second

hydroxylation is necessary to produce the biologically active form of vitamin D: 1,25-

dihydroxyvitamin D (1,25(OH)2D)5. It has been long known that 1,25(OH)2D is produced in the

kidney by 1-α-hydroxylase and it is through this well-established pathway that vitamin D

regulates calcium metabolism, important for the maintenance of healthy bones and the

prevention of rickets in children. More recently it has been identified that many other cells in the

body can also express the 1-α-hydroxylase enzyme and locally produce 1,25(OH)2D from

3 40 IU = 1 µg

4 The term vitamin D without a subscript is used to refer to either D2 or D3

5 Also known as Calcitriol

15

25(OH)D (Hewison et al., 2000; Hewison, Zehnder, Bland, & Stewart, 2000; Hewison et al.,

2007).

1,25(OH)2D is a fat soluble hormone that can be stored in adipose tissue (Holick, 2002) and has

both endocrine (regulation of calcium metabolism) and non-endocrine functions. The vitamin D

receptor, through which 1,25(OH)2D interacts, has been found to be present in most cells in the

body including intestine, bone, kidney, brain, breast, prostate, colon and some immune cells

(Buras et al., 1994; Holick, 2002; Norman, 2008; Stumpf, Sar, Reid, Tanaka, & DeLuca, 1979).

The biologically active form of vitamin D (1,25(OH)2D) has a short half life (~5 hours) and is

very tightly regulated, with circulating 1,25(OH)2D levels up to 1000 times less than 25(OH)D

(Holick, 2009). The enzyme 24-hydroxylase (CYP24A1) breaks down 1,25(OH) 2D to

1,24,25(OH)D. Low vitamin D levels, and thus decreased intestinal calcium absorption, cause an

increase in parathyroid hormone (PTH) levels promoting the re-absorption of calcium from

bones and the conversion of 25(OH)D to 1,25(OH)D (as reviewed in Holick, 2009).

Paradoxically, 1,25(OH)2D serum levels that are in the high or normal range can be observed

when 25(OH)D levels are in the low range. Thus, 25(OH)D is the preferred biomarker for

determining vitamin D status (Holick, 2009).

The finding that vitamin D receptors (VDR) are present in cells throughout the body, and not

restricted to bone, intestine and kidney, has led to the rapidly expanding body of research

suggesting that vitamin D may be associated with reduced risk of all-cause mortality (Autier &

Gandini, 2007), some autoimmune diseases (in particular multiple sclerosis), certain cancers, and

other chronic diseases (Giovannucci, 2008; Holick, 2007; Holick, 2008). The vitamin D receptor

is present in both normal and cancerous breast cells, enabling them to respond to 1,25(OH)2D

(Buras et al., 1994; Colston & Hansen, 2002; Holick, 2003; Welsh, 2004). In vitro and animal

studies have shown 1,25(OH)2D inhibits cell proliferation and angiogenesis, and promotes cell

differentiation and apoptosis (Buras et al., 1994; Colston & Hansen, 2002; Deeb, Trump, &

Johnson, 2007; Giovannucci, 2005; Holick, 2003; Holick, 2006; Welsh, 2004). These anti-cancer

properties of vitamin D occur through both genomic (mediated by VDR) and non-genomic

(direct) pathways (Norman, 2008). A recent review (Krishnan, Swami, & Feldman, 2010)

identified 3 mechanisms by which 1,25(OH)2D inhibits the growth of breast cancer cells: 1) cell

cycle arrest and differentiation, 2) apoptosis and 3) inhibition of invasion and metastasis.

Additionally, these researchers propose 6 newly discovered mechanisms which may further

16

explain the beneficial effects of 1,25(OH)2D on breast cancer cell growth involving anti-

inflammatory effects, inhibition of estrogen synthesis and signaling or down-regulation of the

estrogen receptor and aromatase inhibition (Krishnan et al., 2010).

2.2.3 Recommended Vitamin D Intake from Diet

Dietary Reference Intakes (DRI), used by Canada and the US, are established by the Institute of

Medicine in partnership with Health Canada. Currently, there is no Recommended Dietary

Allowance (RDA) for vitamin D as there was not considered sufficient evidence to calculate an

Estimated Average Requirement of vitamin D for based on adequacy for health or disease

reduction when the DRIs were released in 1995. In its absence, the Adequate Intake (AI) is used

as the reference level. An AI is the average daily nutrient intake level recommended based on

healthy people who are assumed to be in an adequate nutritional state. The AIs for vitamin D are

200 IU/day, 400 IU/day and 600 IU/day (5, 10 and 15 µg/day)6 for adults ≤ 50, 51-70 and >70

years of age, respectively (Health Canada, 2006; Office of Dietary Supplements National

Institutes of Health, 2006). The tolerable upper limit (the highest level at which no adverse

effects are expected if consumed daily) is 2,000 IU/day (Institute of Medicine National Academy

of Sciences, 2009) .

Some in the scientific community have proposed that a higher intake of vitamin D should be

recommended (Vieth et al., 2007). Recent reviews have indicated that vitamin D intake above

the current AIs is not associated with adverse effects (Cranney et al., 2007; Hathcock, Shao,

Vieth, & Heaney, 2007) and a safe upper limit for adults may be as high as 10,000 IU/day

(Hathcock et al., 2007). DRIs for vitamin D are currently under review by the Institutes of

Medicine (Office of Dietary Supplements National Institutes of Health, 2006; Yetley et al.,

2009). Previous studies of vitamin D intake have reported a high proportion of the Canadian

population did not meet the current AIs (Gozdzik et al., 2008; Poliquin, Joseph, & Gray-Donald,

2009; Statistics Canada, 2004; Whiting, Green, & Calvo, 2007). The Canadian Cancer Society

has recommended that Canadians take 1,000 IU vitamin D supplements to reduce cancer risk

“based on the growing body of evidence about the link between vitamin D and reducing risk for

colorectal, breast and prostate cancers” (Canadian Cancer Society, 2007); this recommendatation

6 To convert vitamin D from micrograms (µg) to international units (IU) multiply by 40 (25 µg = 1 IU)

17

was made shortly after publication of the vitamin D and calcium trial by Lappe et al. (Lappe,

Travers-Gustafson, Davies, Recker, & Heaney, 2007) (described in detail in section 2.3.2).

2.2.4 Determinants and Optimal Level of 25(OH)D

It is regularly reported that more than 90% of vitamin D intake is from sun exposure (Holick,

2004; Holick, 2003; John et al., 2007). There is, however, little evidence to support this

statement and it is not appropriate to assume sufficient year round sun exposure in all

populations. The true contribution of diet versus sunlight to 25(OH)D is not well understood.

Even in Hawaii, a population with high year-round sun exposure, low 25(OH)D levels have been

observed (Binkley et al., 2007); which may be due to genetics, or other differences that limit

cutaneous production of vitamin D. Studies of 25(OH)D predictors have generally found

variables related to sun exposure and dietary vitamin D are associated with serum levels

(Brustad, Alsaker, Engelsen, Aksnes, & Lund, 2004; Burgaz, Akesson, Oster, Michaelsson, &

Wolk, 2007; Gozdzik et al., 2008; Sahota et al., 2008; van der Meer et al., 2008).

Additional well-established predictors include age and body fat, which are inversely associated

with 25(OH)D levels (Vieth, Ladak, & Walfish, 2003). Age may affect the ability for cutaneous

synthesis of vitamin D through a decline in either skin thickness (Need, Morris, Horowitz, &

Nordin, 1993) or 7-dehydrocholesterol in the skin (MacLaughlin & Holick, 1985). It has been

suggested that since vitamin D is fat soluble, body fat may sequester vitamin D resulting in lower

serum 25(OH)D levels (Harris & Dawson-Hughes, 2007; Need et al., 1993). However, the levels

of vitamin D found in adipose tissue are not especially high and people experiencing weight loss

do not become vitamin D intoxicated (as reviewed in Heaney et al., 2009). An alternative

hypothesis is that low vitamin D levels may be associated with body fat due to increased surface

area or confounding by diet and time spent outdoors, although the latter was not found in one

study (Harris & Dawson-Hughes, 2007). Other less well-established predictors associated with

higher 25(OH)D levels include oral contraceptives (Harris & Dawson-Hughes, 1998) and

hormone replacement therapy (Heikkinen et al., 1998). In contrast, smoking has been found to be

associated with lower 25(OH)D levels (Brot, Jorgensen, & Sorensen, 1999).

18

There are no well-established standard reference values for 25(OH)D. In 1997 the Institute of

Medicine considered vitamin D deficiency as 25(OH)D levels <27.5 nmol/L7 (Institute of

Medicine National Academy of Sciences, 2009). More recently vitamin D deficiency has been

defined as serum 25(OH)D levels <50 nmol/L and insufficiency is considered 50 to 74 nmol/L

(Holick, 2009). Optimal vitamin D is often defined as serum 25(OH)D levels >75 nmol/L

(Bischoff-Ferrari, 2008; Holick, 2008; Holick, 2009; Vieth, 2006). This optimal level has been

proposed to meet the needs of the non-endocrine (non-calcitropic) pathway. In populations with

high sun exposure (e.g., lifeguards, farm workers) 25(OH)D levels are often greater than 130

nmol/L (Hollis, 2005; Vieth, 1999). It has been shown that vitamin D intakes of at least 1600

IU/day are required to maintain optimal wintertime 25(OH)D levels in Northern populations,

such as Canada (Whiting et al., 2007; Barake, Weiler, Payette, & Gray-Donald, 2010; Cashman

et al., 2008; Hall et al., 2010). The Canadian Health Measures Survey found that 25(OH)D levels

were above 75 nmol/L in only 37.8% (95% CI: 27.6, 38.9) of women age 6 to 79 (Langlois,

Greene-Finestone, Little, Hidiroglou, & Whiting, 2010).

2.3 Epidemiologic Studies of Vitamin D and Breast Cancer

There is now a relatively large body of literature from epidemiologic studies of vitamin D and

breast cancer. Most of it has been published within the past 4 years, since the start of this thesis.

The literature is reviewed here by study design: reviews/meta-analyses, randomized controlled

trials, observational studies (cohort and case control), and ecologic studies. Previous reviews and

meta-analyses are described first.

2.3.1 Reviews and Meta-analyses

In the past ten years at least seven review papers specific to vitamin D and breast cancer risk

have generally concluded there is some evidence from epidemiologic studies to support the

hypothesis that intake of vitamin D from diet or supplements or sunlight exposure may reduce

breast cancer risk (Bertone-Johnson, 2007; Bertone-Johnson, 2009; Colston, 2008; Cui & Rohan,

2006; Lipkin & Newmark, 1999; Perez-Lopez et al., 2009; Rohan, 2007). All of these reviews,

however, have highlighted inconsistencies in the literature and identified areas that still require

investigation. A meta-analysis of 6 studies concluded there was no overall association between

7 2.5 nanomoles per litre (nmol/l) = 1 nanograms/millilitre (ng/ml)

19

vitamin D from diet and supplements and breast cancer risk (pooled RR = 0.98, 95% CI: 0.93-

1.03), but a significant association was observed when intakes ≥400 IU/day were compared to

<150 IU/day (pooled RR = 0.92, 95% CI: 0.87-0.97) (Gissel, Rejnmark, Mosekilde, &

Vestergaard, 2008). An IARC working group review concluded there was not sufficient evidence

of a causal association between 25(OH)D and breast cancer risk but results from their meta-

analyses of 5 studies suggested an inverse association although not statistically significant; for

every 25 nmol/L increase in 25(OH)D (as a continuous variable) the pooled RR was 0.85 (95%

CI: 0.71-1.02) and when comparing highest versus lowest categories (cutpoints not provided) the

RR was 0.46 (95% CI: 0.21-1.03) (IARC, 2008). However, significant heterogeneity was

observed (p<0.001) and both prospective and post-diagnosis studies were combined.

2.3.2 Trials

Two trials of vitamin D plus calcium have reported on cancer risk as the outcome (Chlebowski et

al., 2008; Lappe, et al., 2007). The trial by Lappe et al. evaluated cancer at all sites (not specific

to breast cancer) and was a four-year trial designed to investigate bone fracture as a primary

outcome. Postmenopausal women (n = 1179) from rural Nebraska ( 41° N) were randomized to

a placebo, 1500 mg calcium, or 1500 mg calcium plus 1100 IU vitamin D3. A significant

reduction in all cancer risk was observed for the vitamin D and calcium group (RR = 0.40; 95%

CI, 0.20-0.82) and a similar, although non-statistically significant, association was observed for

calcium only (RR = 0.53; 95% CI, 0.27-1.03). Only 19 breast cancer cases developed throughout

the study: 5 (1.1%) in the vitamin D and calcium arm, 6 (1.4%) in the calcium only arm, and 8

(2.8%) among the placebo group. Specific to breast cancer, results have also been published

from the Women’s Health Initiative (WHI) (Chlebowski et al., 2008). The WHI trial evaluated

calcium and vitamin D intervention with the a priori objectives of evaluating fractures and

colorectal cancer; among the WHI there was also a low fat diet and HRT intervention and

observational study (The Women's Health Initiative Study Group, 1998). Postmenopausal

women from across the US were randomized to placebo (n = 18,106) or 400 IU vitamin D3 with

1000 mg calcium (n = 18,176). After a mean follow-up time of 7 years no significant difference

in breast cancer incidence was observed (HR, 0.96; 95% CI, 0.85, 1.09), although, tumor size

was significantly smaller in the group receiving calcium and vitamin D (Chlebowski et al.,

2008).

20

This trial by Lappe et al., has been criticized for limitations in statistical analysis, and insufficient

information on randomization, loss to follow-up and compliance (Ojha, Felini, & Fischbach,

2007). Additional concerns have been raised regarding lack of statistical power or discussion of

adverse effects (Sood & Sood, 2007). Furthermore, high rates of cancer were observed among

the placebo group but cancer rates among the intervention group were similar to the Nebraska

population rates, suggesting a failure of randomization (Bolland & Reid, 2008; Schabas, 2008).

One strength of the Lappe trial is that a relatively high dose of vitamin D (1100 IU/day) was

administered. It is possible that the findings by Lappe et al. reflect a true effect, but in

consideration of the many limitations raised and the very few breast cancer cases, the results

from this study are largely uninformative with respect to vitamin D and breast cancer. In

contrast, a major limitation of the WHI trial was the low dose of vitamin D used (400 IU/day)

and there was a high potential for contamination as controls were allowed to take vitamin D

supplements and likely also received vitamin D from diet or sunlight (Speers, 2008). There were

also issues of compliance in the intervention group and measures of 25(OH)D serum levels

following randomization were not provided, thus, it is unknown if this trial was successful in

increasing vitamin D in the intervention group. Similar to the trial by Lappe et al., the WHI was

restricted to postmenopausal women only with a relatively short follow-up and did not include a

vitamin D-only trial arm. Neither of these trials was designed specifically to assess vitamin D

and breast cancer risk.

2.3.3 Biomarker Studies

Many studies of 25(OH)D and breast cancer risk (summarized in Table 2) have reported

statistically significant inverse associations (Abbas et al., 2008; Abbas et al., 2009; Crew et al.,

2009; Lowe et al., 2005; Rejnmark et al., 2009;), or non-significant inverse associations

(Bertone-Johnson et al., 2005; Chlebowski et al., 2007; McCullough et al., 2009), few have

reported null associations (Freedman et al., 2008; Janowsky et al., 1999). There was considerable

range in the cutpoints used for comparison of serum 25(OH)D levels. Cutpoints for the highest

categories ranged from greater than 60-150 nmol/L versus less than 30-60 nmol/L for the low

categories; some significant results were observed at both the low (Abbas et al., 2009) and high

(Crew et al., 2009) upper cutoffs.

21

One limitation of these biomarker studies is that only 5 of these studies measured 25(OH)D prior

to cancer diagnosis (Bertone-Johnson et al., 2005; Chlebowski et al., 2008; Freedman et al.,

2008; Rejnmark et al., 2009; McCullough et al., 2009) and although inverse associations were

observed, only one was statistically significant (OR = 0.52; 95% CI, 0.32-0.85) (Rejnmark et al.,

2009). In contrast, four of the five post-diagnosis case-control studies reported statistically

significant inverse associations, however, post-diagnosis 25(OH)D levels may not reflect pre-

diagnostic levels. Breast cancer patients are often recommended to take vitamin D supplements

to prevent treatment-associated bone loss (Wang-Gillam, 2008) thus post-diagnosis serum

25(OH)D might over-estimate usual pre-diagnosis levels potentially biasing study results

towards the null. It is also plausible that women with cancer spend less time outdoors or make

dietary changes that limit their intake of vitamin D which may explain why post-diagnosis case-

control studies have more consistently found 25(OH)D is associated with reduced breast cancer

risk.

Serum 25(OH)D is the established biomarker of vitamin D, reflecting vitamin D from all

sources, but it may not be the best measure of long-term or usual vitamin D status (as reviewed

in Millen & Bodnar, 2008). Moderate correlation was found between two measures of 25(OH)D

measured 14 years apart (correlation between 0.42-0.52 depending on seasonal adjustment

method); stronger correlations have been observed for one to 5-year periods (correlation from

0.53 to 0.80 depending on study population and length of time) (Hofmann, Yu, Horst, Hayes, &

Purdue, 2010; Jorde et al., 2010). None of the studies in Table 2 had more than one measure of

25(OH)D, ideally a long-term average 25(OH)D level would be the best measure. Seasonal

variations in 25(OH)D levels are expected and bias can be introduced by not adequately

adjusting for seasonal variation (Wang et al., 2009). Furthermore, it is unknown how well

circulating 25(OH)D-levels reflect breast cell specific levels.

22

Table 2. Summary of previous studies of serum 25(OH)D and breast cancer risk

Study details

1st author, yr

N cases/ N controls 25 (OH)D cutpoints

for comparison in

nmol/L1

OR (95% CI)

Case-control studies

(post-diagnosis

25(OH)D)

Abbas 2008

1394/1365 Postmenopausal only

>75 vs <30 0.31 (0.24-0.42)

Abbas 2009 289/595 Premenopausal only

≥60 vs <30 0.45 (0.29-0.70)

Crew 2009

1026/1075 >100 vs. <50 0.56 (0.41-0.78) Premenopausal: 0.83 (0.36-1.30) Postmenopausal: 0.46 (0.09-0.83)

Janowsky 1999

156/184 Not provided No association (data not provided)

Lowe 2005

179/179

<50 vs >150 5.83 (2.31-14.7) (reciprocal = 0.17)2

Nested case-control

(pre-diagnosis

25(OH)D)

Bertone-Johnson 2005 Nurses’ Health Study

701/ 724 ≥100 vs. ≤50 0.73 (0.49-1.07) Age <60: 0.92 (0.57-1.48) Age >60: 0.57 (0.31-1.04)

Chlebowski 2007 Women’s Health Initiative

1067/1067 Postmenopausal only

<32.4 vs ≥67.6 1.22 (0.89-1.67) (reciprocal = 0.82)2

Freedman 2008 PLCO

1005/1005 Postmenopausal only

≥84.25 vs <45.75 1.04 (0.75-1.45)

McCullough 2009

516/516 Postmenopausal only

≥75 vs <50 0.86 (0.59-1.26)

Rejnmark 2009

142/420 >84 vs <60 0.52 (0.32-0.85) Premenopausal: 0.38 (0.15-0.97) Postmenopausal: 0.71 (0.38-1.30)

1 Converted from ng/ml to nmol/L (multiplied by 2.5) for the studies by Crew, Bertone-Johnson and Freedman

2 Results were reported comparing lowest to highest categories – opposite direction of other studies

23

2.3.4 Cohort Studies

Table 3 summarizes the six large US (John, Schwartz, Dreon, & Koo, 1999; Lin et al., 2007;

McCullough et al., 2005; Millen et al., 2009; Robien, Cutler, & Lazovich, 2007; Shin et al.,

2002), and one Swedish (Kuper et al., 2009), cohort studies that have evaluated the association

between vitamin D and breast cancer risk. Despite variable measures of vitamin D and diverse

populations, all of these studies, except one (Kuper et al., 2009), reported some inverse

associations between vitamin D and breast cancer risk. However, none of the studies reported

significant inverse associations consistently for all sources of vitamin D intake measured or

among all women. Some studies reported overall risk estimates suggestive of an inverse

association (e.g., HR<1.0) but not statistically significant for all measures of vitamin D (John et

al., 1999; Robien et al., 2007). Other studies reported statistically significant associations

consistent with the vitamin D hypothesis for time spent outdoors (Millen et al., 2009) or for

dietary measures among only premenopausal women (Lin et al., 2007; Shin et al., 2002) or

women living in states with low UV (McCullough et al., 2005). Only two studies reported any

effect estimates that were opposite of the expected direction; one suggested dietary vitamin D

may be associated with increased breast cancer risk (not statistically significant) among post- but

not pre-menopausal women (Lin et al., 2007); the other reported low solar radiation was

significantly associated with reduced breast cancer risk in postmenopausal women but reported

effect estimates that were consistent with the vitamin D hypothesis for other sun exposure

measures (Millen et al., 2009).

A major strength of cohort studies is that temporality can be established; the exposure is

measured before the outcome occurs. The average follow-up length ranged from around 10 years

(Lin et al., 2007; McCullough et al., 2005; Millen et al., 2009) to 18 years or greater (John et al.,

1999; Robien et al., 2007). In one study, the association between total vitamin D and breast

cancer risk was stronger and statistically significant when the analysis was restricted to within 5

years of baseline (RR = 0.66; 95% CI: 0.46-0.94) (Robien et al., 2007). Despite the advantage of

cohort studies – in terms of reduced recall bias and establishing temporality – none of these were

designed specifically for vitamin D; hence, comprehensive measures of vitamin D from all

sources were not used, introducing the potential for nondifferential measurement error which

often biases results towards null although not always.

24

Among the studies of dietary vitamin D, only three studies had measures of vitamin D from food,

multivitamins and single product vitamin D supplements (John et al., 1999; Robien et al., 2007;

Shin et al., 2002), the remaining studies included food and multivitamins only (Kuper et al.,

2009; Lin et al., 2007; McCullough et al., 2005). None of the studies specify the inclusion of cod

liver oil. The highest categories of vitamin D intake varied from approximately 200 IU/day from

food only (John et al., 1999; Kuper et al., 2009) to greater than 700 or 800 IU/day for total intake

(Robien et al., 2007; McCullough et al., 2005). The studies with upper cutoffs at only ≥200

IU/day were, unsurprisingly, not significantly associated with breast cancer risk with risk

estimates were only slightly less than 1.0 (John et al., 1999; Kuper et al., 2009). Although the

study with the highest category of intake (≥800 IU/day) was not significant, the risk estimate for

food was 0.55 with a 95% CI from 0.24 to 1.22 (Robien et al., 2007), suggesting a possible

reduced risk; however, despite the large confidence interval the sample size was quite large with

34,321 women and 2,440 cases. Significant inverse associations (or borderline significant where

upper 95% CI limit equals 1.00) were observed at intakes >500 IU/day among premenopausal

women only (Shin et al., 2002; Lin et al., 2007) and >300 IU/day among women living in states

with low UV (McCullough et al., 2005).

Only three studies included any measures of sunlight exposure as proxy measures for vitamin D

exposure and a range of variables were assessed: ecologic-level measures of solar or ultraviolet

radiation (John et al., 1999; Millen et al., 2009), variables associated with sunburn/skin damage

(John et al., 1999; Kuper et al., 2009) and time spent outdoors (John et al., 1999; Millen et al.,

2009). Only one study investigated the combined effect of diet and sunlight by comparing a

measure of high versus low sun and diet (John et al., 1999). There is some evidence of an

interaction between diet and sunlight (McCullough et al., 2005) but this was not found elsewhere

(Millen et al., 2009).

25

Table 3. Summary of previous cohort studies of vitamin D (from diet or sunlight) and breast cancer

risk

Study details

1st author, yr

N cases/ N total

cohort

Dietary vitamin D measures

RR or HR1 (95% CI)

Sun exposure measures

RR or HR (95% CI)

NHANES I

(John et al. 1999)

190/ 5009 Food (≥200 vs <100 IU/day): 0.85 (0.59-1.24) Supplements (daily vs never): 0.89 (0.60-1.32) Food or supplements (≥200 or daily vs <100 IU/day or never): 0.86 (0.61-1.20) High sun & diet (vs low sun & diet): 0.71 (0.44-1.14)

Occupational & recreational sun exposure: 0.67 (0.42-1.06) Sun-induced skin damage: 0.80 (0.48-1.29) Region of residence (south vs northeast): 0.71 (0.47-1.09) Solar radiation of birth place (high vs low): 0.73 (0.49-1.09)2

Swedish

Women’s

Lifestyle and

Health Cohort

(Kuper et al, 2009)

840/41,889 Dietary vitamin D3 (>204 vs 116): 0.9 (0.8-1.1) Multivitamin use yes vs no: 1.0 (0.8-1.2)

Exposure during age period 40-49 years4: Annual number of sunburns (≥2 vs never): 0.9 (0.7-1.3) Sunbathing vacations( ≥4wks/y vs never): 1.2 (0.9-1.6) Solarium use (≥1 per month vs never): 0.9 (0.8-1.2)

Women’s

Health Study

(Lin et al. 2007)

Premenopausal: 276/10,578 Postmenopausal: 743/20,909

Premenopausal: Food (≥319 vs <142 IU/day): 1.02 (0.69, 1.53) Multivitamins (≥400 vs 0 IU/day): 0.76 (0.50-1.17) Total (≥548 vs <162 IU/day): 0.65 (0.42-1.00) Postmenopausal: Food ( ≥319 vs <142 IU/day): 1.22 (0.95, 1.55) Multivitamins (≥400 vs 0 IU/day): 0.87 (0.68-1.12) Total (≥548 vs <162 IU/day): 1.30 (0.97-1.73)

Cancer

Prevention

Study II

(McCullough,2005)

2,855/68.567 postmenopausal only

Food (>300 vs ≤100 IU/day): 0.89(0.76-1.03) Total (>700 vs ≤100 IU/day): 0.95 (0.81-1.13) Low UV state of residence5: Food (>300 IU/day vs ≤100): 0.81 (0.67-0.97) High UV state of residence Food (>300 IU/day vs ≤100): 1.05 (0.82-1.35)

Not reported but food results stratified by UV.

Continued…

26

WHI

observational

study (Millen et al, 2009)

2,535 / 71,662 postmenopausal only

Effect measures not reported6 State of residence 7 (north vs. south): 0.97 (0.89-1.07) Clinic centre Watts per m2 (≤0.5 vs ≥1.5): 0.85 (0.74-0.98)8 Time outdoors summer (<30min vs >2 hrs): 1.18 (1.05-1.34)

Iowa Women’s

Health Study

(Robien et al, 2007)

2440 / 34,321 Postmenopausal only

Total (≥800 vs <400 IU/day): 0.89 (0.77-1.03) Supplement (≥800 vs 0 IU/day): 0.89 (0.74-1.08) Diet (≥800 vs <500 IU/day): 0.55 (0.24-1.22)9 Restricted to within 5 years of baseline: Total (≥800 vs <400 IU/day): 0.66 (0.46-0.94)

Nurses’ Health

Study

(Shin, 2002)

3172 / 88,691

Premenopausal: Total (>500 vs ≤150 IU/day): 0.72 (0.55-0.94) Food (>300 vs ≤75 IU/day): 0.66 (0.43-1.00) Postmenopausal: Total (>500 vs ≤150 IU/day): 0.94 (0.80-1.10) Food (>300 vs ≤75 IU/day): 1.06 (0.85-1.34)10

1 Effect estimates from fully adjusted multivariate models are reported. Each study controlled for different variables. 2 Additional measures of sun exposure not reported here 3 Quartile cutpoints were obtained from personal communication with the authors 4 Three additional age periods of exposure were measured (10-19, 20-29, and 30-39 years) and were not associated with

breast cancer risk. Additional measures of sun sensitivity (e.g., skin color, use of sun block) were also not associated

with breast cancer risk 5 No significant associations between dietary vitamin D and breast cancer risk overall. Significant reduced risks were

also observed among ER+ only(not in ER-). 6 Vitamin D intake from foods and supplements did not confound or modify the associations observed for the measures

of sunlight. 7 Also measured at birth, 15 and 35 years of age and results were similar. Additional sunlight measures were provided. 8 This association is not in the expected direction given the vitamin D hypothesis. 9 Data also provided stratified by years of follow-up. With 0-5 years follow-up total vitamin D: RR=0.66 (0.46-0.94). 10 Cumulative average diet model is reported here. Effect estimates are also provided for earlier diet (collected at baseline

in 1980). Estimates were adjusted for outdoor sun exposure and residential area.

27

2.3.5 Case-Control Studies

Table 4 summarizes the case-control studies of vitamin D and breast cancer risk. Similar to the

cohort studies, a range of dietary and sunlight measures of vitamin D have been assessed and

most case-control study results suggest an inverse association between vitamin D and breast

cancer risk but not all effect estimates were statistically significant (Abbas, Linseisen, & Chang-

Claude, 2007; John, Schwartz, Koo, Wang, & Ingles, 2007; Knight, Lesosky, Barnett, Raboud,

& Vieth, 2007; Rossi et al., 2009). Results from two European studies found strong significant

reductions in breast cancer risk at relatively low intakes of vitamin D from food alone. Vitamin

D intake from food alone ≥400 IU/day was significantly associated with a 50% reduced breast

cancer risk among a German population (Abbas et al., 2007) and intake >190 IU/day was

associated with a 64% reduced risk among women living in Southern Italy (Rossi et al., 2009);

the association was attenuated and not significant among women living in Northern Italy.

The study by Knight et al., (Knight et al., 2007) is the most comprehensive case-control study as

it included a wide range of vitamin D measures (including vitamin D rich foods, multivitamins,

supplements and cod liver oil and measures of sunlight exposure) at three specific periods of life

and is the only previous Canadian study. Although overall vitamin D dose from food or

supplements was not measured, specific foods (e.g., milk, and salmon) known to contain vitamin

D and supplements, including cod liver oil, were significantly associated with reduced breast

cancer risk. A range of sun exposure measures that may be important for vitamin D production

were also measured and most were associated with reduced breast cancer risk consistent with the

vitamin D hypothesis. Overall the associations between supplements, foods and sunlight

exposures were strongest for exposure during the ages 10-19 (Knight et al., 2007). The fourth

case-control study was conducted in California and included a sun exposure index (derived from

facultative (sun exposed) and constitutive (usual) skin pigmentation measured by reflectometry),

measures of lifetime outdoor activity and race/ethnicity (John et al., 2007). These sun measures

may not be specific to vitamin D generating potential, however, a high sun exposure index was

associated with up to 50% reduced risk of advanced (but not localized) breast cancer among light

skinned women (John et al., 2007).

In addition to the studies included in Table 4 there are two other small (<300 cases) hospital

(Levi, Pasche, Lucchini, & La Vecchia, 2001) or screening-based studies (Simard, Vobecky, &

28

Vobecky, 1991) that both reported positive associations between vitamin D from food only and

breast cancer risk, however, these studies do not provide sufficient data for interpretation. The

study by Levi et al., assessed a range of micronutrients, not specific to vitamin D, and the values

of vitamin D intake are improbable – the median values for highest and lowest tertiles were 2.7

and 1.4 milligrams/day1 (OR = 1.39; 95% CI: 1.01, 1.92) (the usual units for vitamin D are

micrograms). Simard et al., provided descriptive data only (no measure of effect or statistical

significance – from the crude data the unadjusted OR could be calculated comparing >200 vs

<50 IU/day: OR = 2.78).

One threat to the validity of case-control studies is the potential for recall bias. For this reason,

case-control studies are often considered inferior to cohort studies. However, with respect to

vitamin D intake there is no obvious reason to suspect that cases would differentially recall their

dietary intake of foods containing vitamin D or sun exposure particularly for these studies in

Table 4 which began before the recent publicity regarding the vitamin D hypothesis. All except

for one of the case-control studies (Rossi et al., 2008) were population-based, improving the

likelihood that controls would be similar to cases with respect to everything except disease

outcome. As with cohort studies there is the potential for measurement error in case-control

studies and only one study evaluated all sources of vitamin D (food, supplements, and sunlight)

(Knight et al., 2006), the others evaluated food (Abbas et al., 2007; Rossi et al 2008) or sun

exposures only (John et al., 2007). Although there are fewer case-control studies than cohort

studies of vitamin D and breast cancer risk, more of the results observed were statistically

significant, potentially due to timing of exposure; both North American studies evaluated

lifetime exposures from adolescence through adulthood (Knight et al., 2006; John et al, 2007).

1 2.7 milligram = >100,000 IU/day. If we assume this was a typo and should have been 2.7 µg this would yield a

highest intake category of only 108 IU/day.

29

Table 4. Summary of previous case-control studies of vitamin D (from diet or sunlight) and breast

cancer risk

Study details

1st author, yr

N cases/ N

controls

Dietary vitamin D measures

OR (95% CI)

Sun exposure measures

OR (95% CI)

Abbas, 2007 Population-based Germany

278 / 666 Premenopausal only

Food (≥400 vs <80 IU/day): 0.50 (0.26-0.96)

John, 2007 Population-based California

1786 / 2127 Among women with light skin

pigmentation only1

Advanced breast cancer: Lifetime outdoor activity (4 vs 1 hrs/week): 0.86 (0.51-1.45) Sun exposure index (high vs low): 0.53 (0.31-0.91) Localized breast cancer Lifetime outdoor activity (4 vs 1 hrs/week): 1.05 (0.72-1.54) Sun exposure index (high vs low): 1.10 (0.74-1.63)

Knight, 2006 Population-based Ontario

972 / 1135 Exposure from age 10-19:2 Milk (≥10 vs 0 glasses/week): 0.62 (0.45-0.86) Salmon or tuna (>1 vs 0 per week): 0.86 (0.64-1.14) Supplement/multivitamin (yes vs no):0.53 (0.39-0.73) Cod liver oil (yes vs no): 0.76 (0.62-0.92)

Exposure from age 10-19:2 Days outside (<3 vs 7): 1.49 (1.00-2.22) Outdoor activity episodes (high vs low): 0.65 (0.50-0.85) Outdoor job (≥1yr vs never): 0.61 (0.46-0.80) Limbs covered (yes vs no): 1.68 (1.14-2.50) Skin burned (no vs yes): 1.55 (1.08-2.24) Sunscreen use (yes vs no): 1.04 (0.72-1.51) Winter sun trip (yes vs no): 1.00 (0.77-1.30) Sunlamp use (yes vs no): 0.81 (0.57-1.14)

Rossi, 2008 Hospital-based Italy

2560 / 2588 Food (>190 vs <60 IU/day):0.76 (0.58-1.00) Stratified by geographical area: Northern: 0.86 (0.63-1.16) Southern: 0.36 (0.20-0.68)

1 Effect measures are also provided for medium and dark skin pigmentation – no significant associations were observed

among these groups. 2 Exposures at ages 20-29 and 45-54 were also assessed. Effect estimates were attenuated at age 20-29 and non-

significant at age 45-54.

30

2.3.6 Ecologic and Sun Exposure Proxy Studies

Maps showing higher mortality from cancer among people living in northern latitudes in the US

first led to the hypothesis that vitamin D may be associated with reduced breast cancer risk

(Garland, Garland, Gorham, & Young, 1990). Since then, ecologic studies have shown higher

latitude (a proxy for sun exposure) and lower UVB irradiance are positively associated with

breast cancer rates (Mohr, Garland, Gorham, Grant, & Garland, 2008) or mortality (Grant,

2002a; Grant, 2002b; Porojnicu et al., 2006), respectively; consistent with the vitamin D

hypothesis. The interpretation of these studies is limited since all observations were made only at

a population level; it is not possible to determine if this association remains at an individual level

and there is a high potential for residual confounding.

Other studies that have evaluated the association between skin cancer, as a proxy for UV

exposure, and breast cancer risk have been inconsistent (Cantwell et al., 2009; Levi,

Randimbison, Te, Conconi, & La Vecchia, 2008; Chen et al., 2008; Grant, 2007; Soerjomataram,

Louwman, Lemmens, Coebergh, & de Vries, 2008; Tuohimaa et al., 2007). Contradicting the

vitamin D hypothesis, women with a previous skin cancer diagnosis (squamous cell carcinoma

(SCC), basal cell carcinoma (BCC) or cutaneous malignant melanoma (CMM)) have been

observed to have a higher rate of breast cancer (standardized incidence ratio (SIR) = 1.18; 95%

CI: 1.08-1.30) (Levi et al., 2008). Soerjomataram et al. also observed a positive association

between breast cancer risk and CMM but an inverse (not statistically significant) association

with SCC (Soerjomataram et al., 2008); Soerjomataram et al., suggest the positive association

may be due to confounding by SES or other variables known to be related to both skin and breast

cancer risk. Elsewhere an inverse association was observed between breast cancer risk and SCC

(SIR = 0.58; 95% CI: 0.30-0.86) but not BCC (Cantwell et al., 2009). A meta-analysis of 10

studies of non-melanoma skin cancers, SCC and BCC combined (SCC, BCC or CMM) and

breast cancer risk reported a pooled OR of 1.13 (95% CI: 1.09-1.17) (Soerjomataram et al.,

2008). Paradoxically, 25(OH)D has been found to be inversely associated with non-melanoma

skin cancer among men and thus Tang et al. propose that BCC and SCC are not good proxy

measures of UV exposure or vitamin D status (Tang et al., 2010). Overall these studies provide

little evidence in support of an association between vitamin D and breast cancer risk.

31

2.4 Factors that May Influence or Modify the Vitamin D and Breast Cancer Association

2.4.1 Calcium

It is well established that the active form of vitamin D regulates calcium absorption and that

calcium also has a role in vitamin D metabolism (Heaney, 2008). Therefore the interaction

between calcium and vitamin D may be important for breast cancer risk. Furthermore, vitamin D

and calcium are found in some of the same foods (e.g., vitamin D fortified milk and dairy

products made with fortified milk) and thus it is difficult to tease out the independent effects in

dietary studies. Evidence from animal and in vitro studies suggests that calcium may also have

anticarcinogenic properties that include regulation of cell differentiation, proliferation and

apoptosis (Carroll, Jacobson, Eckel, & Newmark, 1991; Khan et al., 1994; McGrath & Soule,

1984; Sergeev, 2004; Whitfield, Boynton, MacManus, Sikorska, & Tsang, 1979; Xue, Lipkin,

Newmark, & Wang, 1999). Hence, it is important to elucidate the independent roles of both

calcium and vitamin D (as reviewed by Cui & Rohan, 2006; Heaney, 2008).

Results from epidemiologic studies do not strongly support an inverse association between

calcium, or more generally dairy products, and breast cancer risk (as reviewed by Al Sarakbi et

al., 2005; Bissonauth et al., 2008; Cui & Rohan, 2006; Larsson, Bergkvist, & Wolk, 2009;

Moorman & Terry, 2004; Parodi, 2005). As discussed earlier, the trial by Lappe et al. of all

cancers included a calcium only arm and found inverse associations of similar magnitude, but

not significant, to calcium plus vitamin D (Lappe et al., 2007). The WHI trial included only a

calcium and vitamin D group and no protective effect of the combined nutrients was observed

(Chlebowski et al., 2008). Recent prospective cohort studies have found an inverse association

between breast cancer incidence and serum calcium among pre- and post-menopausal women

with BMI ≥ 25 (Almquist, Manjer, Bondeson, & Bondeson, 2007), and dietary calcium intake

among both pre- and post-menopausal women (stronger among premenopausal women) (Kesse-

Guyot et al., 2007).

Few observational studies of vitamin D and breast cancer risk have investigated the interaction

between calcium and vitamin D (Abbas et al., 2007; Lin et al., 2007; Shin et al., 2002). Results

suggest either no interaction (Abbas et al., 2007; Shin et al., 2002) or an interaction among

postmenopausal women only (Lin et al., 2007), such that an inverse association between calcium

32

and breast cancer risk was observed only among the highest category of vitamin D intake. Three

other studies have included both measures of vitamin D and calcium but did not report on the

interaction (John et al., 1999; McCullough et al., 2005; Robien et al., 2007). The combined and

independent associations between calcium and vitamin D and breast cancer risk require further

investigation.

2.4.2 Timing of Exposure to Vitamin D

Most of the case-control and cohort studies described above have focused on recent exposure

during adulthood; few studies have looked at earlier life vitamin D intakes. The most

comprehensive of these studies (Knight et al., 2007) observed the strongest associations for

measures of vitamin D at ages 10-19 (e.g. sun exposure (OR = 0.65; 95% CI: 0.50-0.85), cod

liver oil use (OR = 0.76; 95% CI: 0.62-0.92), and milk consumption (OR = 0.62; 95% CI, 0.45-

0.86)). Weaker associations were observed for vitamin D during ages 20-29 and no associations

were observed during ages 45-54. In contrast, sunburns, sunbathing vacations and solarium use

exposure during the age period 10-19 years, or any later age period, were not associated with

breast cancer risk (Kuper et al., 2009). As for dietary vitamin D, two previous studies

investigating a range of micronutrients reported no associations between adult breast cancer risk

and vitamin D intake from food during adolescence (Frazier, Ryan, Rockett, Willett, & Colditz,

2003; Frazier et al., 2004). Elsewhere, an inverse, but not statistically significant, association was

reported between milk consumption during high school and premenopausal breast cancer risk

(RR = 0.76; 95% CI: 0.48-1.21) (Shin et al., 2002).

In addition to early life exposures and adult breast cancer risk, the association between dietary

vitamin D and breast cancer risk may (Lin et al., 2007; Shin et al., 2002) or may not (Knight et

al., 2007; Rossi et al., 2009) differ by menopausal status. Many studies have been conducted

among postmenopausal women only and unable to evaluate if menopausal status modifies the

association between vitamin D and breast cancer risk (Robien et al., 2007; Millen et al., 2009;

McCullough et al., Chlebowski et al., 2007; Freedman, et al., 2008; McCullough, et al., 2009).

One of these cohort studies among postmenopausal women only, found significant inverse

associations only when the analyses were restricted to within 5-years of baseline (Robien et al.,

2007) and the authors suggest exposure misclassification may increase with time since baseline;

alternatively, recent vitamin D intake may influence cancer development.

33

2.4.3 Hormone Receptor Status (ER/PR)

Some studies of vitamin D and breast cancer risk (Lin et al., 2007; McCullough et al., 2005;

Robien et al., 2007), but not all (Blackmore et al., 2008), have reported differences by hormone

receptor status (ER/PR). Among the studies that observed differences, the results were

inconsistent. Among premenopausal women significant inverse associations were observed

between total dietary vitamin D and breast cancer risk for both ER and PR positive breast cancer

cases; no differences were found among postmenopausal women (Lin et al., 2007). McCullough

et al. observed significant inverse associations among ER+ cases and no association among ER-

cases in their study of postmenopausal women only (McCullough et al., 2005). In contrast

Robien et al. observed stronger associations between vitamin D intake and ER- or PR- cases

(also among a study of postmenopausal women) (Robien et al., 2007). Blackmore et al. found no

differences by ER or PR status and this was not modified by menopausal status (Blackmore et

al., 2008). The representativeness of these findings is limited since receptor status was only

known for a portion of cases in each study: 53% (McCullough, Bostick, & Mayo, 2009), 68%

(Robien et al., 2007), 68% (Blackmore et al., 2008) and was not reported for the fourth study

(Lin et al., 2007).

2.4.4 Genetic Variants

Variants in genes on the vitamin D pathway may result in functional changes that may affect

endogenous vitamin D levels resulting in modification of breast cancer risk. A comprehensive

review of the literature identified three genes in the vitamin D pathway that have been studied in

relation to BC risk: vitamin D receptor (VDR), vitamin D binding protein (Gc), and CYP24A1

(involved in degradation of 1,25-dihydroxyvitamin D) (McCullough et al., 2009). There is some

support suggesting that VDR polymorphisms may be directly associated with breast cancer risk

or may modify the association between vitamin D exposure and breast cancer risk (as reviewed

in McCullough et al., 2009; Slattery, 2007). A pooled analysis of 6 prospective studies of the two

most commonly studied VDR polymorphisms (FokI (rs2228570) and BsmI (rs1544410)) found

that FokI was associated with increased breast cancer risk (ff versus FF: OR = 1.16; 95% CI:

1.04-1.28) (McKay et al., 2009). Only 3 studies have investigated variants in the Gc or

CYP24A1 genes (as reviewed by McCullough et al., 2009) and the results of these studies are

inconclusive. Few studies have evaluated gene-environment interactions and the current

34

evidence on variants in vitamin D related genes and breast cancer risk is minimal and requires

further investigation.

2.5 Vitamin D and Breast Cancer Mortality, Prognosis or Precursors

In addition to breast cancer risk, vitamin D has also been shown to be inversely associated with

breast cancer stage, recurrence and mortality. Among a cohort of women with early breast

cancer, vitamin D deficiency (defined as 25(OH)D levels <50 nmol/L versus >72 nmol/L) was

associated with increased risk of both distant recurrence (HR = 1.71; 95% CI: 1.02-2.86) and

death (HR = 1.60; 95% CI: 0.96-2.64) (Goodwin, Ennis, Pritchard, Koo, & Hood, 2009). A

significant inverse association between serum 25(OH)D and breast cancer mortality was also

reported in a prospective cohort study (n = 16,818) with only 28 breast cancer cases (comparing

≥62.5 versus < 62.5 nmol/L: RR = 0.28; 95% CI: 0.08-0.93) (Freedman, Looker, Chang, &

Graubard, 2007). Similarly, a death certificate based case-control study found significant inverse

associations between residential and occupational sun exposure and breast cancer mortality

(Freedman, Dosemeci, & McGlynn, 2002). In regards to breast cancer stage, tumor size was

significantly smaller among women in the vitamin D and calcium arm of the WHI than the

intervention group (Chlebowski, et al., 2008). Elsewhere, higher levels of serum 25(OH)D have

been observed among women with early versus advanced stage breast cancer (Palmieri,

Macgregor, Girgis, & Vigushin, 2006). In a large study among women in Norway, season of

diagnosis was found to be associated with breast cancer prognosis; women diagnosed in the

summer or fall – when higher vitamin D levels are expected – had a lower risk of breast cancer

death (Robsahm, Tretli, Dahlback, & Moan, 2004).

Mammographic density and benign breast disease (BBD) are known breast cancer risk factors

and may also be intermediate markers. In addition to having a direct effect on breast cancer risk,

the effects of vitamin D may also be mediated by changes to mammographic density or BBD.

However, studies of vitamin D and mammographic density are conflicting. Only one study has

found an overall inverse association between dietary vitamin D and calcium intake and

mammographic density (Berube, Diorio, Verhoek-Oftedahl, & Brisson, 2004). Other studies of

dietary vitamin D and calcium have variously reported inverse associations among pre- but not

postmenopausal women (Berube et al., 2005), or no associations with vitamin D but inverse

35

associations with calcium among postmenopausal women only (Mishra et al., 2008). One

biomarker study of 25(OH)D and calcium intake found no associations of either with

mammographic density (Knight et al., 2006). Elsewhere, serum 25(OH)D was found to be

inversely associated with mammographic density when a four month lag time, reflecting

seasonal changes, was applied (Brisson et al., 2007). In the WHI trial, the effect of calcium plus

vitamin D on benign proliferative breast disease has also been investigated and no significant

association was observed (Rohan et al., 2009). However, this study of BBD is affected by the

same limitations of the WHI trial described earlier in terms of breast cancer risk, namely a low

dose of vitamin D.

2.6 Appraisal of the Vitamin D and Breast Cancer Literature

The results of all epidemiologic studies of vitamin D and breast cancer risk were described above

by study design. The purpose of this section is to provide an appraisal of the vitamin D and

breast cancer literature in consideration of causation (using the guidelines provided by Elwood,

2007) and to propose a minimal ideal vitamin D dose.

Intervention trials are often considered the most rigorous study design and often demonstrate

causation since randomization and blinding reduce the potential for confounding and observer

bias (Elwood, 2007). The WHI trial is currently the only trial to evaluate vitamin D and breast

cancer (Chlebowski et al., 2008). Unfortunately, as discussed previously, the low vitamin D dose

(400 IU/day) in the WHI and lack of information on compliance and contamination among

controls limits our ability to draw any meaningful overall conclusions regarding the association

between vitamin D and breast cancer from this study.

One of the main threats to the validity of observational studies is confounding. However, the vast

majority of the observational studies described in tables 2, 3 and 4 had information available on

many potential confounders (e.g., known breast cancer risk factors including reproductive factors

and physical activity) and adjusted for confounding through matching or analysis; only a few

studies reported limited data on potential confounders (Rossi et al, 2008; Lowe et al., 2005;

Janowsky et al.,1999). Measurement error and bias are also of concern in observational studies.

The primary objectives of most of these observational studies were not specific to vitamin D and

breast cancer and many began before the vitamin D and cancer hypothesis was well-known, thus

reducing the potential for recall or observer bias. However, this may have introduced additional

36

measurement error since vitamin D from all sources was rarely measured; this measurement

error would likely be non-differential which often biases results towards the null. The

measurement of dietary vitamin D was most often through validated FFQ (e.g., Abbas et al.,

2007; Lin et al., 2007; Robien et al., 2007), 24-hour recall (John et al., 1999) or established

serum 25(OH)D assays (e.g., McCullough et al., 2009; Freedman et al., 2008; Bertone-Johnson

et al., 2005; Lowe et al., 2005; Crew et al., 2009); the validity of the various sun exposure

measures are not well-established. The sample sizes for all studies were quite large and only a

few studies had fewer than 500 cases (Abbas et al., 2009; Janowsky et al., 1999; Lowe et al.,

2005; Rejnmark et al., 2009; John et al., 1999; Abbas et al., 2007). Despite the large sample

sizes, many studies reported risk estimates less than 1.0 (suggestive of an inverse association

between vitamin D and breast cancer) but were not statistically significant at p<0.05 thus the

probability that these findings are due to chance is greater than the conventionally accepted 5%

among many studies. The few studies that may be considered of less rigorous design (smaller

sample size, fewer confounders or less established measures of vitamin D) (Rossi et al, 2008;

Lowe et al., 2005; Janowsky et al.,1999; Rejnmark et al., 2009) did not consistently find

different findings than the more rigorous studies.

Despite the aforementioned potential threats to internal study validity, other features support a

causal association between vitamin D and reduced breast cancer risk. The evidence for a biologic

mechanism is strong (as reviewed in section 2.2.2), accordingly the plausibility is high. Study

results have also been relatively highly consistent; of the 21 studies described in tables 2, 3 and 4

all except for 3 (Janowsky et al., 1999; Freedman et al., 2008; Kuper et al., 2009) provide at least

some evidence of an inverse association between vitamin D and breast cancer risk. Consistency

across subgroups has been less dependable such that some differences have been observed by

menopausal status, hormone receptor status, geographic location or timing of exposure.

Furthermore, the specificity of the vitamin D and breast cancer association is not well established

and other potential explanations still need to be ruled out. A dose-response relationship is also

not well-established; most studies have been conducted within a relatively small range of vitamin

D intake or serum 25(OH)D levels. Although most of the observational studies on vitamin D and

breast cancer risk (as described in tables 2, 3 and 4) have been rigorously conducted, we cannot

rule out all potential threats to study validity or conclude that all characteristics of a causal

association have been met therefore ability to determine causality is limited.

37

In terms of external validity, many of the observational studies were population-based and thus

the findings should be generalizeable to at least the source population (e.g., Knight et al., 2006;

John et al., 2007; Kuper et al., 2009; Crew et al., 2009; Abbas et al., 2007, 2008 & 2009). Other

studies were clinic or screening based (Janowsky et al., 1999; Lowe et al., 2005; Freedman et al.,

2008; Rejnmark et al., 2009; Rossi et al., 2008) or conducted among specific populations (e.g.,

the Nurses’ Health Study papers by Bertone-Johnson et al., 2005, and Shin et al., 2002) and may

not be generalizeable to the general population. The sampling methods and response/follow-up

rates were not described in sufficient detail for many of the cohort studies, but the reader was

often referred elsewhere. All studies except for one Canadian (Knight et al., 2006) and 6

European (Kuper et al., 2009; Rejnmark et al., 2009; Rossi et al., 2008; Abbas et al., 2007, 2008

& 2009) were conducted throughout the US. Although the biologic mechanism may be

independent of geographic differences, it is highly plausible that differences in food consumption

patterns (e.g., fatty fish), country-specific food fortification policies, skin color, genetics and

exposure to UVB radiation may influence population 25(OH)D levels and thus vitamin D dosage

information may not be generalizeable.

Vitamin D dosage

Optimal vitamin D is often defined as serum 25(OH)D levels >75 nmol/L (Bischoff-Ferrari,

2008; Holick, 2008; Holick, 2009; Vieth, 2006). Previous studies of serum 25(OH)D and breast

cancer risk have evaluated levels >75 nmol/L and most have reported significant inverse

associations (as shown in Table 2). There is some evidence to suggest that vitamin D intakes of

at least 1600 IU/day are required to maintain 25(OH)D levels in Northern populations, such as

Canada (Whiting et al., 2007; Barake, Weiler, Payette, & Gray-Donald, 2010; Cashman et al.,

2008; Hall et al., 2010). No studies of dietary vitamin D intake have assessed intakes as high as

1600 IU/day; the highest cutpoint assessed was >800 IU/day (compared to <400 IU/day) and was

significantly associated with a 34% reduced risk of breast cancer when analyses were restricted

to within 5 years of baseline (Robien et al., 2007). Albeit, a meta-analysis of dietary vitamin D

intake found a small but significant 8% reduction in breast cancer risk only after restricting the

analyses to the 3 studies (Robien et al., 2007; Shin et al., 2002; McCullough et al., 2005) with the

highest vitamin D intakes, which was only ≥400 IU/day (Gissel, Rejnmark, Mosekilde, &

Vestergaard, 2008).

38

Based on the current literature and biologic mechanism the ideal highest category of exposure for

studies of dietary vitamin D intake is likely much higher than those previously reported. The

maximal ideal dose should likely be at least 1000 IU/day or higher. Given the relatively low

intakes observed in study populations this may not currently be feasible but including all sources

of dietary vitamin D (including all foods and supplements) would help to improve measurement

of vitamin D intake. Furthermore, increasingly vitamin D supplements are available on pharmacy

shelves at doses of 1000 IU and the importance of vitamin D supplement use has garnered

attention in recent years. The Canadian Cancer Society now recommends all adult Canadians

take a 1000 IU vitamin D supplement. The minimal ideal dose of vitamin D that future studies

should measure is likely at least 1000 IU/day.

2.7 Summary and Rationale for the Current Study

Despite the emerging body of literature suggesting that vitamin D may reduce breast cancer risk,

study results have been inconsistent and many gaps still exist in our understanding of these

associations thus limiting our ability to conclude a causal association exists. Most previous

studies of vitamin D have included only measures of dietary intake (from food and/or

supplement sources) , few studies have measured vitamin D from sunlight. Ideally a combination

of measures should be used taking into consideration vitamin D from all sources (Millen et al.,

2008). A major focus of this thesis is on the measurement of vitamin D. To improve upon

previous studies, a food frequency questionnaire was modified for Canadian-specific vitamin D

values and used to measure vitamin D from food and supplements. Furthermore, an algorithm

was developed to derive a composite measure of vitamin D from sunlight using individual level

variables associated with endogenous vitamin D production (e.g., time spent outdoors, location

resided, skin color, sun protection practices) and was applied in this study.

In some studies, but not all, the association between vitamin D and breast cancer has been shown

to be stronger in premenopausal women and there is strong biologic evidence that obese women

are more likely to have lower levels of vitamin D suggesting a possible interaction between

vitamin D and obesity. Despite the direct relationship between calcium and vitamin D, few

studies have included calcium in their investigations of vitamin D and breast cancer. As well,

most studies have evaluated only adulthood vitamin D exposure and not the potentially critical

period during breast development. This thesis explored the association between variables related

39

to the production of vitamin D from sunlight at four age periods of exposure (from adolescence

through adulthood). Furthermore, this thesis contributes to our knowledge of the association

between vitamin D and breast cancer risk by incorporating vitamin D exposure at different stages

of life, and testing for effect modification by calcium, menopausal status, and BMI.

This population-based case-control study of over 6,000 women in Ontario improves upon these

limitations and addresses some of the gaps in knowledge. Vitamin D intake from diet or

exposure to UV light is modifiable and research identifying modifiable breast cancer risk factors

is essential since to-date, most well-established risk factors are not modifiable. The results of this

study provide a better understanding of the association between vitamin D and breast cancer in

Ontario women with the ultimate aim of primary prevention.

40

Chapter 3 Study Methods

3.1 Data Source and Study Design

The data source for this study was the “Ontario Women’s Diet and Health Study”, a population-

based case-control study conducted among Ontario women. Data were collected from 2003 to

2004 as part of a Canadian Breast Cancer Research Alliance (CBCRA) – Canadian Breast

Cancer Foundation (CBCF) funded research project entitled “Total phytoestrogen intake during

selected periods of life and breast cancer risk” (primary investigator: M. Cotterchio; co-

investigators: N. Kreiger, L. Thompson, B. Boucher). The primary objectives of the “Ontario

Women’s Diet and Health Study” were to evaluate phytoestrogen intake during various periods

of life and breast cancer risk (Cotterchio, Boucher, Kreiger, Mills, & Thompson, 2008; Thanos,

Cotterchio, Boucher, Kreiger, & Thompson, 2006) and to develop a novel Canadian database for

phytoestrogens (Thompson, Boucher, Liu, Cotterchio, & Kreiger, 2006; Thompson, Boucher,

Cotterchio, Kreiger, & Liu, 2007).

Data collection occurred prior to my involvement in this study. All decisions regarding study

design and measurement of variables were made by the study investigators. My role included

conducting the literature review, deriving the specific hypotheses, developing the specific study

objectives, evaluating methodological issues related to the measurement of vitamin D from diet

and sunlight and conducting all statistical analyses and variable derivation, including

development of the solar vitamin D score. In addition, I wrote the first draft of each manuscript

and revised these papers based on suggested revisions by my thesis committee and co-authors.

3.2 Identification of Cases and Controls

3.2.1 Cases

Cases were women aged 25 – 74 years with a first pathologically confirmed cancer of the breast

diagnosed between June 2002 and April 2003. Cases were identified through the Ontario Cancer

Registry (OCR). The OCR is a population-based registry that obtains information from nearly all

breast cancer cases in the province in Ontario (Hall, Schulze, Groome, Mackillop, & Holowaty,

41

2006). The OCR identifies new cancer cases and deaths in Ontario by computerized probabilistic

record linkage of data from the following sources: 1) pathology reports, 2) Regional Cancer

Centres, 3) hospital discharge and ambulatory care records, and 4) Ontario death certificates.

Consent and contact information was sought from the physicians identified in the registry

pathology reports, and was obtained for 4,109 eligible cases (96%). Cases were eligible for

inclusion in the study if they were women between 25 and 74 years of age with a first primary,

pathology confirmed breast cancer diagnosed between September, 2002 and August, 2003.

Physician cooperation was required to contact patients, obtain contact information and vital

status. The average time between diagnosis and interview was 11 months with a range (5th to 95th

percentile) from 7 to 18 months. In total, 3,101 of the 4,109 cases with physician consent

completed the study (75% response rate or 72% of all cases regardless of physician consent).

3.2.2 Controls

Controls were identified using a modified random digit dialing procedure of households in

Ontario and frequency matched (1:1) within 5-year age groups to the identified cases.

Recruitment of controls was conducted by the survey research unit at the Institute for Social

Research at York University (Toronto, Ontario). A sampling frame of phone numbers was

derived from Ontario telephone directories and other commercially available lists. In addition to

these listed numbers, other numbers on either side of these were added to the sampling frame to

capture unlisted numbers.

In total, 25,250 households were telephoned; approximately, 17,000 of these households were

ineligible, 2,000 did not answer the phone and 2,000 refused (eligibility of these households is

unknown). Of the 4,352 households where an eligible woman was identified, 250 (6%) women

refused, and 4,102 (94%) women agreed to participate. One woman from each household was

randomly selected for inclusion in the study. Overall, 3,471 controls completed the study out of

the 4,352 households with known eligibility (80% response rate). The true response rate

denominator for all eligible Ontario women is unknown.

3.3 Data Collection

Eligible cases and controls were mailed a risk factor (epidemiologic) questionnaire and food

frequency questionnaire (FFQ). The risk factor questionnaire consisted of 20 pages with 79

42

questions and collected information on lifestyle, reproductive and medical history factors. This

risk factor questionnaire was pre-tested among a convenience sample of women before it was

finalized. Food and supplement intake was measured using a modified 178-item version of the

1998 Block Food Frequency Questionnaire (FFQ). This FFQ has been validated several times for

many nutrients (e.g., (Block, Woods, Potosky, & Clifford, 1990; Boucher et al., 2006)). Cases

and controls were asked about their consumption of food and supplements two years prior to the

time of questionnaire, to reduce any bias due to changes in diet in cases after cancer diagnosis.

Questions related to the measurement of vitamin D from both questionnaires are included in

appendix 1 and are described in section 3.4.

To improve response rates a structured mailing timeline with a series of 5 contacts was followed.

After the initial telephone contact, study participants were mailed the questionnaire packages

with a signed letter of information, and return Canada Post business reply envelope. To improve

response rates, a $5 incentive was included in the control questionnaire packages and a magnet

was included in the case questionnaire packages. Two weeks after the initial mailing all subjects

were mailed a thank-you postcard, which served as a reminder for non-respondents. Non-

respondents were phoned after 4 weeks and new questionnaire packages were mailed after 8

weeks. A final telephone call was made 12 weeks after the initial mailing to encourage response

among non-respondents. Trained interviewers conducted scripted phone calls and 8 attempts

were made to contact subjects. If phone numbers were no longer in service or questionnaires

were returned undelivered every attempt was made to find the subjects’ new address, through

online directories for controls and by contact with physicians for cases.

Questionnaire response data were entered into an Access dataset by trained research clerks.

Study participants were contacted by telephone to clarify responses for any missing or

unexpected responses. All participants were assigned a unique 6 digit study identification

number and confidentiality was maintained by using the study id number only in the study

database. Additional logic checks, data cleaning and derivation of all variables was conducted by

the PhD candidate. For the continuous variables (e.g., height, weight, age at menopause and

menarche) all extreme responses, defined as those in the top or bottom percentile, were double

checked with the original hard copy surveys and any obvious errors were changed to missing.

43

3.4 Variable Definitions

3.4.1 Vitamin D and Calcium from Food and Supplements

A modified Block FFQ measured 178 foods and supplements and requested data on both

frequency of consumption and portion size for all foods and frequency and duration of use for

supplements. The validity and reliability of the Block FFQ have been assessed in a random

sample of Ontario women (Boucher et al., 2006). The reliability for vitamin D and calcium were

high with non-deattenuated Pearson correlation coefficients of 0.76 (95% CI: 0.66-0.83) for

vitamin D and 0.80 (95% CI: 0.71-0.86) for calcium. Validity of the FFQ, compared to two 24-

hour recalls, was also moderately high; the deattenuated Pearson correlation coefficient vitamin

D was 0.54 (95% CI: 0.29-0.79) and for calcium was 0.71 (95% CI: 0.35-1.00). Using the

recently proposed classification for correlation coefficients from FFQ validation studies (Willett,

2009; Serra-Majem, Andersen, Henrique-Sanchez, Doreste-Alonso, et al., 2009) the reliability of

vitamin D is considered very good (≥0.7) and the validity is good (0.5-0.69).

Nutrient analysis was conducted by Block Dietary Data Systems (BDDS) using nutrient values

from the USDA Nutrient Database for Standard Reference and national data on food

consumption (NHANES III and CSFII). In our modified version of the Block 1998 FFQ, two

questions were added to better capture vitamin D intake: type of fish most often consumed (fatty

or white fish) and use of vitamin D supplements or cod liver oil. Objective 5 of this thesis was to

modify the nutrient analysis specific to vitamin D to account for these additional items and

differences in fortification of foods between Canada and the U.S. The results of this objective

and more details regarding the methodology are described in paper 1 (section 4.2 of thesis

results).

Daily intakes of vitamin D (IU/day) and calcium (mg/day) were derived individually for foods

and supplements (alone, in a multivitamin or as cod liver oil for vitamin D) from the Block

nutrient analysis and a combined total intake measure (food plus supplements) was also created.

Histograms showing the distributions of these variables are included in appendix 4 (figures 1-3).

Since the variables were not normally distributed, particularly for supplement intake, and for

consistency with the current literature all vitamin D variables were categorized. Variables were

categorized using cut points that correspond to the established Dietary Reference Intakes (DRI)

when the distribution of the data was sufficient for at least 10% of controls in each category.

44

When DRIs were not available or the distribution was not amenable to such cut points variables

were categorized into quartiles or quintiles based on the distribution among controls. In addition

to the derived nutrient values for vitamin D and calcium (from foods, supplements and total),

individual foods rich in vitamin D and/or calcium and vitamin supplement sources were

individually examined. These foods included milk, fish, margarine, and the frequency and

duration of supplement use for vitamin D or cod liver oil, and calcium (alone or combined with

something else) and regular one-a-day multiple vitamins. Food intakes were calculated as

servings per day, week or month (depending on frequency of consumption) by combining

portion size and frequency of consumption data using guidelines provided by BDDS.

3.4.2 Individual Variables Related to Cutaneous Vitamin D Production

The risk factor questionnaire included a sun exposure questionnaire (appendix 1) which collected

information on the following variables related to vitamin D production: time spent outdoors, sun

protection practices, and location of residence. Self-reported ethnicity was also collected and

used as a proxy for skin color. The overwhelming majority (90%) of study participants was

Caucasian ethnicity (proxy for lighter skin color) and thus skin color was categorized as

Caucasian versus non-Caucasian. Other variables related to sun exposure (weekday time

outdoors, weekend time outdoors, sun protection and location of residence) were measured at

four periods of life: teenage years, 20-30s, 40s-50s and 60s-74. All women were at least 25 years

of age, thus, n = 6,572 for exposure during teenage years and 20s-30s, whereas, n = 6,075 for

exposure during 40s-50s, and n = 2438 for exposure during 60s-74.

Time spent outdoors was measured as the number of hours of sun exposure from April to

October during a typical day for weekends and weekdays separately. Data were collected for

these 6 months only since wintertime sun exposure in Ontario is not sufficient for the production

of vitamin D (Webb et al., 1988). Response options were: less than one hour, 1 to 2 hours, 3 to 4

hours, more than 4 hours. Study participants were instructed to include both exposures at work

and during leisure time. The variable “Hours outdoors per week” was created by weighting and

summing weekday and weekend exposures. During each period women were asked how often

they wore sunscreen or protective clothing when in the sun (never, sometimes, and always) and

where they lived during each of the 4 periods of life.

45

The validity and reliability of other similar sun exposure questionnaires have been measured.

Previous sun exposure questionnaires, focusing on time spent outdoors, have been shown to have

fair to moderate reliability (intraclass correlation coefficients range from 0.25-0.77) for both

recent adult measures (English, Armstrong, & Kricker, 1998; Kricker, Vajdic, & Armstrong,

2005; van der Mei, Blizzard, Ponsonby, & Dwyer, 2006; Yu et al., 2009) and recall of adolescent

exposures (van der Mei et al., 2006). In terms of validity, time spent outdoors measured from sun

exposure questionnaires has been significantly associated with skin measures of solar exposure

(Karagas et al., 2007; van der Mei et al., 2006; Weiler, Knight, Vieth, Barnett, & Wong, 2007)

and 25(OH)D (r = 0.17 to 0.58) (Brot et al., 2001; Kim & Moon, 2000; Need et al., 1993; Sahota

et al., 2008; van der Mei et al., 2006; Hanwell et al., 2010). Moderate agreement has been found

between questionnaire measures of sun exposure and calendar methods (Kappa = 0.54 to 0.71)

(van der Mei et al., 2006), detailed face-to-face measures (ICC = 0.54; 95% CI: 0.21-0.76)

(Kricker et al., 2005) or personal UV dosimetry (r = 0.32 to 0.69) (O’Riordan et al., 2008;

Chodick et al., 2008)

Geographic location was collected by city and province of residence for each of the 4 periods of

exposure. All women resided in Ontario when they participated in the study, but many lived

outside the province earlier in life. Women were asked to report the location where they lived

during the 4 specific periods of life but a full residential history was not collected and no

information was obtained on how long participants resided at each location reported.

Approximately 2% of women reported multiple locations lived during a given age period of

exposure and only the first location was used. There were 1628 (25%) study participants who

reported only country or province/state of residence during at least one period of life; these

participants were assigned the most populated city in their country or region (e.g., Shanghai was

assigned to those who reported China). There were 86 (1%) participants who lived in the

southern hemisphere during at least one life period; for these women the reporting period would

have corresponded to wintertime sun exposure and the analysis was performed with and without

these women. Since the results were essentially unchanged with these women included they were

kept in the analysis. Location of residence was converted from place name to latitude and

longitude through the website www.Geocoder.ca for US and Canadian cities (Geocoder, 2007),

and for international cities using http://worldkit.org/geocoder. Latitude and longitude were

46

obtained for 6,119 (93%), 5858 (89%), 5637 (93%) and 2438 (95%) women during their teens,

20-30s, 40s-50s and 60-74, respectively.

Measures of UV radiation were then obtained for each geographic coordinate (latitude and

longitude). UV radiation from sunlight, specifically UVB radiation, is the necessary precursor to

cutaneous vitamin D production. UV radiation data were obtained from National Aeronautics

and Space Administration’s (NASA) Total Ozone Mapping Spectrometer (TOMS) (NASA,

2007). Ground level UV irradiance data are calculated from TOMS onboard spacecraft

instrument measures of atmospheric UV, total ozone, surface reflectivity and cloud cover.

Monthly average noon-time erythemal UV for June 2003 was selected for use in this study. It is

estimated that the change in summertime UV over the last forty years is minimal, i.e., less than

5% (personal conversation Dr. Fioletov, Environment Canada). Wintertime UV would be

expected to be more affected by the changes to the ozone layer that has occurred over time.

These data are weighted using the McKinlay-Diffey erythemal action spectrum (McKinley,

1987) which weights radiation in the UVA (315-400 nm) and UVB (280-315 nm) wavelengths

based on the time required to induce erythema (skin reddening); shorter rays are more likely to

induce erythema. Cutaneous vitamin D is dependent on only UVB exposure and there is a

vitamin D-specific action spectrum based on human skin’s ability to produce previtamin D3

(MacLaughlin, 1982), but vitamin D weighted UV data are not currently available from TOMS.

Although vitamin D production does not always directly correspond with erythemal UV

estimates (Kimlin, 2003), the erythemal action spectra closely approximates the vitamin D action

spectra in summer north of 42⁰ (Ontario, Canada) (Fioletov, 2009; Pope, Holick, Mackin, &

Godar, 2008).

Figure 1 shows the distribution of erythemal UV radiation worldwide and the locations where all

study participants resided in their teenage years (maps for other age periods of exposure not

shown). Maps were created using ArcGIS version 9.2. Latitude was strongly correlated with

erythemal UV during teens, 20-30s, 40-50s, and 60s-74; the Pearson correlation coefficients

were -0.85, -0.84, -0.76, -0.72, respectively (all p-values <0.0001) after excluding women living

in the Southern hemisphere.

47

Fig

ure

1.

Ma

p o

f er

yth

ema

l U

V r

ad

iati

on

(m

W/m

2)

wo

rld

wid

e a

nd

lo

cati

on

s w

her

e ca

ses

an

d c

on

tro

ls r

esid

ed i

n

thei

r te

ena

ge

yea

rs

48

3.4.3 Derivation of a Solar Vitamin D Score

To derive a measure of vitamin D from sunlight, and create one variable that takes into

consideration known factors that affect vitamin D synthesis, an algorithm was created based on

the available literature on determinants of cutaneous production of vitamin D. Weighted

proportions were used to create this solar vitamin D score which included our measures of time

spent outdoors, sun protection practices, ethnicity (proxy for skin colour) and UV radiation of

location lived. The same algorithm was applied to each of the 4 periods of life (teenage

years/adolescence, 20s-30s, 40s-50s and 60s-74). A cumulative lifetime measure was also

derived by adding the solar vitamin D scores from all relevant age periods and a measure of

recent exposure was also be created by using the solar vitamin D score that pertains to the

current age of all study participants.

Figure 2 presents the derived algorithm. In regards to skin color an accommodation factor of one

third was chosen to weight UV production for non-Caucasian individuals. It has been observed

that people with highly pigmented (darker) skin colors in comparison to lighter require at least 3

times the amount of sunlight to produce equivalent vitamin D (Webb & Engelsen, 2006;

Holick,1987), although estimates range upward to 5 to 10 times (Chen et al., 2007). Thus, using

the most conservative estimate, an accommodation factor of one third was chosen to weight UV

production for non-Caucasian individuals. In regards to sun protection practices, sunscreen and

clothing both have the potential to block all vitamin D production. However, it is unlikely that

women apply a complete application of sunscreen (i.e., a thorough application to all locations of

the body prior to going outdoors with frequent reapplication) (as reviewed by Norval & Wulf,

2009) or fully cover-up with clothing. Sunscreen use does not predict 25(OH)D levels (Sahota et

al., 2008; Thieden, Philipsen, Heydenreich, & Wulf, 2009), but coverage of arms and legs does

significantly predict lower 25(OH)D levels (Sahota et al., 2008). Therefore, within this

population it was assumed that the available vitamin D generating UV light was reduced by two

thirds in participants who report “always” using sun protection (i.e., they would have one third

the potential UV in comparison to participants who reported “never” using sunscreen or

protective clothing). Correspondingly, a decrease of one third was estimated for participants

“sometimes protected”, hence a weighting factor of two thirds was assigned to these respondents.

49

Before applying the algorithm, the associations between each of the sun exposure related

variables were evaluated. Although ethnicity was significantly associated with sun protection

practices at each age of exposure (Chi-Square tests p<0.001), there were no consistent patterns

with respect to the differences in sun protection practices between Caucasians and non-

Caucasians (appendix 4, table 1). Time spent outdoors was not strongly correlated with

erythemal UV, latitude, or sun protection practices (Spearman’s r < 0.10) at any period of

exposure (appendix 4, table 2). Furthermore, self-reported time spent outdoors in the summer

did not differ by season of questionnaire completion.

Figure 2. Hypothesized model of vitamin D and breast cancer with details of the proposed algorithm for the measurement of UV radiation conditional on factors that affect vitamin D production

3.4.4 Potential Confounders

Potential confounders were identified as any variable that may be associated with either breast

cancer risk or vitamin D status based on the literature. Few variables were identified from the

literature that were known be associated with both breast cancer risk and vitamin D

50

exposure/intake, and thus, meet the true definition of a confounder (e.g., physical activity, BMI,

HRT use, and phytoestrogen intake). The following 39 variables were considered potential

confounders: marital status, education, ethnicity, BMI, smoking status, pack years smoked,

breastfed, lactation, age at menarche, oral contraceptive use, oral contraceptive duration, parity,

age at first live birth, age at last menstruation, hormone replacement therapy (HRT) use

(postmenopausal women only), duration of HRT use, history of benign breast disease, family

history of breast cancer, screening mammogram, alcoholic drinks, dietary fat intake, energy

intake, phytoestrogen intake, physical activity (strenuous, moderate and daily activity) at selected

periods of life (teenage years, 20-30s, 40-50s and 60s-74). Calcium variables were evaluated as

potential confounders of the vitamin D and breast cancer associations, and vice versa.

These variables were derived using previously established cutpoints from the literature (e.g.,

BMI) or were categorized into quartiles or quintiles based on the distribution among controls.

The distribution for 24 of these 39 variables among cases and controls and the age-group

adjusted ORs are shown in Table 5 and some have been described previously (Cotterchio et al.,

2008). Age was calculated as age at diagnosis for breast cancer cases and age at midpoint of

recruitment for controls. Variables were lagged where possible to ensure the exposure occurred

at least two years prior to interview (e.g., age at menopause, start and stop dates for HRT use and

smoking status). Daily, moderate and strenuous physical activity was measured for 4 age periods

of exposure and the category ‘age not yet reached’ was assigned for women who were not yet in

their 40s-50s and/or 60s-74.

51

Table 5. Distribution of breast cancer cases and controls and age-group adjusted odds ratio

(AOR) estimates for selected known and suspected breast cancer risk factors

Variable Cases

n = 3101

No. (%)

Controls

n = 3471

No. (%)

AOR

Age1 25 - 39 40 - 44 45 - 49 50 - 54 55 - 59 60 - 64 65 - 69 70 - 74

181 (6) 278 (9) 380 (12) 504 (16) 511 (16) 450 (15) 439 (14) 358 (12)

316 (9) 367 (11) 482 (14) 512 (15) 470 (14) 471 (14) 508 (15) 345 (10)

N/A

Ethnic or racial background Caucasian (White) Black Aboriginal (e.g., Indian, Metis) South East Asian (e.g., Japanese, Chinese) South Asian (e.g., East Indian, Pakistani) Other

2749 (89) 48 (2) 33 (1) 155 (5) 63 (2) 42 (1)

3121 (90) 55 (2) 28 (1) 96 (3) 83 (2) 68 (2)

1.00 1.03 (0.70-1.53) 1.31 (0.79-2.18) 1.96 (1.50-2.54) 0.91 (0.65-1.27) 0.75 (0.51-1.11)

Highest level of Education Elementary High school Postsecondary

275 (9) 1384 (45) 1423 (46)

214 (6) 1558 (45) 1684 (49)

1.00 0.71 (0.58-0.86) 0.71 (0.58-0.86)

BMI2 (kg/m2) in premenopausal women < 24.9 25.0 - 29.9 ≥ 30

566 (59) 241 (25) 146 (15)

663 (54) 338 (27) 235 (19)

1.00 0.81 (0.66-0.99) 0.70 (0.54-0.87)

BMI (kg/m2) in postmenopausal women < 24.9 25.0 - 29.9 ≥ 30

780 (37) 771 (36) 578 (27)

871 (40) 834 (38) 506 (23)

1.00 1.03 (0.90-1.19) 1.28 (1.10-1.50)

Smoking status never smoker ex-smoker current smoker

1572 (51) 1176 (38) 331 (11)

1791 (52) 1122 (32) 535 (15)

1.00 1.18 (1.06-1.31) 0.71 (0.61-0.83)

Pack-years smoked Never smoker Q1 Q2 Q3 Q4

1572 (52) 360 (13) 336 (12) 388 (14) 393 (14)

1791 (52) 409 (12) 365 (11) 424 (13) 432 (14)

1.00 1.04 (0.88-1.21) 1.09 (0.92-1.28) 1.03 (0.88-1.20) 0.97 (0.83-1.13)

Ever Breastfeed your infant No Yes

1470 (48) 1622 (52)

1357 (39) 2093 (61)

1.00 0.73 (0.66-0.81)

Age at menarche (years) ≤ 11 12 13

594 (20) 753 (25) 846 (28)

615 (18) 823 (25) 973 (29)

1.00 0.95 (0.81-1.10) 0.90 (0.80-1.04)

52

≥ 14 782 (26) 939 (28) 0.86 (0.74-0.99) Age at menopause3 ≤ 45 45 - 49 ≥50 premenopausal

572 (19) 511 (17) 987 (33) 957 (32)

727 (21) 550 (16) 888 (26) 1237 (36)

1.00 1.15 (0.98-1.36) 1.37 (1.19-1.59) 1.15 (0.95-1.38)

Parity Nulliparous 1 2 - 3 > 4

543 (18) 421 (14) 1677 (55) 415 (14)

404 (12) 413 (12) 2068 (61) 524 (15)

1.00 0.75 (0.62-0.91) 0.57 (0.49-0.66) 0.53 (0.44-0.64)

Age at first live birth Nulliparous 12 - 19 20 - 24 25 - 29 30 - 34 > 35

543 (18) 408 (13) 951 (31) 743 (24) 318 (10) 93 (3)

404 (12) 507 (15) 1207 (35) 861 (25) 335 (10) 95 (3)

1.00 0.55 (0.45-0.66) 0.54 (0.46-0.63) 0.62 (0.53-0.73) 0.70 (0.58-0.86) 0.71 (0.52-0.97)

Duration of HRT use (yrs) Never use ≤ 3 3.1 – 6.0 6.1 – 10.0 >10 premenopausal

1070 (43) 58 (2) 121 (5) 135 (5) 173 (7) 957 (38)

1183 (42) 84 (3) 114 (4) 89 (3) 115 (4) 1237 (44)

1.00 0.74 (0.52-1.04) 1.13 (0.86-1.49) 1.64 (1.23-2.18) 1.67 (1.29-2.15) 1.05 (0.88-1.27)

Breast cancer in a 1st degree relative No Yes

2389 (77) 635 (21)

2973 (86) 415 (12)

1.00 1.86 (1.63-2.13)

Benign breast disease4 No Yes

2014 (66) 1015 (34)

2631 (77) 785 (23)

1.00 1.65 (1.47-1.84)

Mammogram5 No Yes

623 (20) 2473 (80)

940 (27) 2518 (73)

1.00 1.32 (1.15-1.51)

Strenuous physical activity in teenage years Never 1– 3 times per month 1– 2 times per week 3– 5 times per week >5 times per week

233 (8) 373 (13) 638 (22) 906 (31) 744 (26)

217 (6) 388 (12) 669 (21) 1034 (32) 923 (29)

1.00 0.95 (0.75-1.20) 0.94 (0.75-1.16) 0.86 (0.70-1.06) 0.78 (0.63-0.96)

Strenuous physical activity in 20s-30s Never 1– 3 times per month 1– 2 times per week 3– 5 times per week >5 times per week

231 (8) 409 (14) 828 (28) 907 (30) 616 (21)

188 (6) 443 (13) 930 (28) 1057 (32) 700 (21)

1.00 0.80 (0.63-1.02) 0.76 (0.62-0.95) 0.73 (0.59-0.91) 0.72 (0.58-0.90)

Strenuous physical activity in 40s-50s Never 1– 3 times per month 1– 2 times per week 3– 5 times per week >5 times per week

288 (10) 528 (17) 786 (26) 786 (26) 450 (15)

240 (7) 554 (16) 863 (26) 905 (27) 491 (15)

1.00 0.81 (0.66-1.00) 0.76 (0.63-0.93) 0.73 (0.60-0.89) 0.76 (0.61-0.94)

53

Age not yet reached 181 (6) 316 (9) NA Strenuous physical activity in 60s-74 Never 1– 3 times per month 1– 2 times per week 3– 5 times per week >5 times per week Age not yet reached

209 (7) 262 (9) 304 (10) 254 (8) 156 (5) 1854 (61)

175 (5) 249 (7) 339 (10) 353 (10) 144 (4) 2147 (63)

1.00 0.88 (0.67-1.15) 0.75 (0.58-0.97) 0.60 (0.47-0.78) 0.90 (0.67-1.22) n/a

Daily activity at work in 20s-30s Sitting Light Moderate Strenuous

928 (31) 769 (25) 1210 (40) 116 (4)

893 (27) 880 (26) 1428 (43) 158 (5)

1.00 0.83 (0.72-0.95) 0.79 (0.69-0.89) 0.70 (0.54-0.90)

Daily activity at work in 40s-50s Sitting Light Moderate Strenuous Age not yet reached

933 (31) 757 (25) 1080 (36) 77 (3) 181 (6)

874 (26) 886 (26) 1209 (36) 111 (3) 316 (9)

1.00 0.78 (0.68-0.90) 0.81 (0.71-0.92) 0.63 (0.47-0.86) NA

Daily activity at work in 60s-74 Sitting Light Moderate Strenuous Age not yet reached

314 (10) 478 (16) 335 (11) 12 (0) 1854 (62)

275 (8) 546 (16) 388 (12) 12 (0) 2147 (64)

1.00 0.76 (0.62-0.93) 0.75 (0.60-0.93) 0.88 (0.39-2.00) NA

Alcohol intake (drinks/week)6 Never 1 - 6 7 - 35

1558 (51) 894 (29) 610 (20)

1741 (51) 1052 (31) 634 (19)

1.00 0.97 (0.86-1.08) 1.07 (0.93-1.21)

Dietary fat intake (g/day) 6.7 - 47.9 47.9 – 66.1 66.2 - 88.4 88.4 – 412.8

803 (26) 769 (25) 729 (24) 761 (25)

855 (25) 856 (25) 859 (25) 857 (25)

1.00 0.97 (0.84-1.10) 0.92 (0.80-1.06) 0.97 (0.85-1.12)

Total Phytoestrogens (µg/day) 0 - 438 439 - 978 979 - 3077 3077 - 8594 8595 - 657,042

598 (20) 599 (20) 601 (20) 600 (20) 601 (20)

648 (19) 691 (21) 628 (19) 675 (20) 728 (22)

1.00 0.95 (0.82-1.12) 1.05 (0.90-1.23) 0.96 (0.82-1.12) 0.88 (0.76-1.03)

1 Age at cancer diagnosis for cases and age on 15Nov2002 for controls 2 BMI weight two years ago (kg) divided by height in metres squared 3 If menstrual periods are reported within or beyond one year of diagnosis or referent age, the participant is categorized as premenopausal 4 Benign breast disease is defined as as non-cancerous cysts or lumps in breasts 5 Self-reported history of ever having a screening mammogram (lagged by two years) 6 Amount of beer, wine and liquor usually consumed 2 years ago

54

3.5 Statistical Analysis

An overview of the statistical analyses that were used is provided in this section; each of the 3

papers in the results section of this thesis provides additional methods specific to each study

objective. All statistical analyses were conducted using SAS version 9.1. Statistical significance

was defined as P value less than 0.05 and all tests were two-sided. Descriptive data analyses

were conducted first for data cleaning purposes and to describe the study population. Table 1 in

appendix 3 shows the percent of missing data for each variable. Missing data was minimal for all

potential confounders 0 to 3% and less than 2% for the food and supplement data. The missing

data were greatest for the location of residence variables and correspondingly latitude, erythemal

UV and the solar vitamin D score (calculated using location of residence). Some assumptions

were made as described above for location of residence to preserve as much data as possible, no

other assumptions were made when data were missing. Since the study sample size is large and

the missing data were minimal, imputation methods or inverse probability weighting were not

used. Any cases or controls with missing data were allowed to fall out of the multivariable

models and analyses were conducted on subjects with complete data only.

For objectives 1-3, multivariate logistic regression analysis was used to obtain odds ratios (OR)

and 95% confidence intervals for all vitamin D exposures with breast cancer risk as the

dependent variable. Since controls were frequency age-matched to cases, all logistic regression

models were adjusted for age-group using unconditional models (Rothman & Greenland, 1998).

The likelihood-ratio test was used to test the significance of multiplicative interactions.

Interactions were evaluated between each of the vitamin D variables and calcium intake,

menopausal status, and BMI. Stratified analyses were conducted when significant interactions

were observed. If no significant interaction was observed then each of the potential effect

modifiers was investigated as potential confounders.

Confounders were defined as any variable that changed the OR of the exposure variable by more

than 10% when added to the model (Maldonado & Greenland, 1993). This was assessed by

adding each of the 39 potential confounding variables (listed in section 3.4) one at a time to the

age-adjusted model of interest. This was done for all of the dietary vitamin D and calcium

variables (food and nutrient level and supplements) and for each of the sun exposure related

variables and the derived scores for each age group. If a confounder was identified then it was

55

included in the model. Test for linear trend was calculated by treating the median intake for each

exposure category as a continuous variable in the age-adjusted and fully-adjusted logistic

regression models.

For objective 2, associations between breast cancer risk and each of the following exposure

variables were evaluated: 1) variables associated with cutaneous vitamin D production (as

reported from the epidemiologic questionnaire), 2) latitude and UV radiation of residence, and 3)

derived solar vitamin D scores. All of the exposure variables were evaluated for each of the 4

time periods of exposure (adolescence through adulthood) and the cumulative and recent

measures of the solar vitamin D score. For each age period of exposure, women who had not yet

reached that age period were removed from the models.

It was hypothesized that time spent outdoors might actually be a proxy for physical activity, thus

the correlations between time spent outdoors and physical activity were evaluated. As shown in

Table 6, the correlations between time spent outdoors and each measure of physical activity, at

all 4 periods of exposure, were statistically significant but the strength of the associations were

relatively weak (r < 0.26). Physical activity (daily, moderate or strenuous) at any age period of

exposure was not found to confound any of the sun exposure related variables or the solar

vitamin D score.

Table 6. Spearman rank correlations (rs) between physical activity and time outdoors per week at

4 age periods of exposure

Physical Activity Hrs outdoor per week

Teenage years

20-30s

40-50s

60-75

rs (p-value)

Daily 0.14 (<0.0001) 0.17 (<0.0001) 0.15 (<0.0001) 0.18 (<0.0001) Moderate 0.24 (<0.0001) 0.20 (<0.0001) 0.20 (<0.0001) 0.24 (<0.0001) Strenuous 0.26 (<0.0001) 0.20 (<0.0001) 0.18 (<0.0001) 0.17 (<0.0001)

For objective 1, associations between breast cancer risk and the following dietary measures of

daily vitamin D intake were evaluated: 1) derived intake from foods, supplements, and combined

total (as derived by the Block nutrient analysis), 2) individual foods rich in vitamin D (e.g., milk

and fish), and 3) frequency and duration of supplement use (multivitamin and single product

56

vitamin D or cod liver oil). Vitamin D and calcium intakes were highly correlated (table 7), thus,

various ways of accounting for multicollinearity were evaluated (e.g. creation of a combined

variable, or two new variables, or restriction of vitamin D to only UV exposures).

Table 7. Spearman rank correlations (rs) between vitamin D and calcium from food, supplements

and total combined (food and supplements) intake.

Vitamin D (IU/day)

Total

Supplements

Food

rs (p-value)

Total

0.64 (<0.0001) 0.43 (<0.0001) 0.54 (<0.0001)

Calcium

(mg/day) Food 0.50 (<0.0001) 0.10 (<0.0001) 0.79 (<0.0001)

Supplements 0.56 (<0.0001) 0.70 (<0.0001) 0.05 (0.0003)

To evaluate measurement error associated with the development of the solar vitamin D score for

objective 4, sensitivity analyses were conducted. A range of other plausible values were assumed

and substituted for the parameters assigned in the algorithm and the impact of these changes on

the association between solar vitamin D and breast cancer risk were evaluated. An alternative

solar vitamin D measure was also created through cross-classification. UV exposure, time

outdoors, skin colour and sun protection practices were categorized as high versus low based on

vitamin D production potential and an additive score was then created categorizing women into 4

categories.

In addition to our derived algorithms, we also evaluated the use of a previously published

algorithm for the measurement of predicted serum 25(OH)D values (Giovannucci et al., 2006).

Multiple linear regression was used to develop a predictive model among a subset of men in the

Harvard Health Professionals’ Follow-Up Study with serum 25(OH)D measures. The model was

then applied to predict 25(OH)D among other men in the Health Professionals’ Study and has

been previously applied to women in the Nurses’ Health Study (Forman et al., 2007; Ng et al.,

2009). The predictive model by Giovannucci et al was not specific to the measurement of

vitamin D from sunlight and included 6 variables: dietary vitamin D, supplemental vitamin D,

BMI, race, physical activity (included as a proxy for time spent outdoors), and region of

57

residence. This predictive model was able to explain only 28% of the variation in 25(OH)D.

More details regarding the methodology used are provided in Paper 3.

To evaluate the impact of modifying the vitamin D nutrient analysis (objective 5), descriptive

statistics were calculated to compare vitamin D intake before and after modification. The

weighted kappa statistic and 95% CI were calculated to assess the chance-corrected agreement

between the categorical vitamin D from food intake variables obtained from the Canadian and

US nutrient analyses. The paired t-test was used to determine if the mean difference between the

Canadian and US nutrient analyses using the continuous measures of vitamin D from foods was

different from zero.

3.6 Ethics

Ethics approval for the “Ontario Women’s Diet and Health Study” was initially obtained from

the University of Toronto Research Ethics Board on December 4, 2002 by M. Cotterchio. Ethics

approval for this secondary data analysis project “Vitamin D and Breast Cancer Risk” was

obtained from the University of Toronto Research Ethics Board by L. Anderson on December

20, 2007 and one-year renewals were granted in 2008 and 2009 (appendix 2). This study

involves secondary data analysis only and study participants were not contacted for additional

information, thus, the risk was classified as low and expedited review was granted.

58

Chapter 4 Study Results

4.1 Overview of Results

The results of this thesis are presented in manuscript format with 3 independent papers. Although

all papers have multiple authors, the PhD candidate developed the research questions, completed

all literature reviews, variable derivation, statistical data analysis and wrote each manuscript.

The first paper describes vitamin D intake among the controls only and evaluates the impact of

modifying the nutrient analysis applied to the FFQ for Canadian vitamin D food values and

additional vitamin D sources (thesis objective 5). The second paper evaluates the associations

between vitamin D, calcium intake from food and supplements and breast cancer risk, using the

modified Canadian vitamin-D nutrient analysis (thesis objectives 1). The third paper describes

the development of a solar vitamin D score and evaluates the association between this score and

each variable related to the cutaneous production of vitamin D and breast cancer risk (thesis

objectives 2, 3 and 4).

59

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

4.2 Paper 1: Vitamin D Intake from Food and Supplements among Ontario Women Based on the US Block Food Frequency Questionnaire with and without Modification for Canadian Food Values

Accepted for publication by Canadian Journal of Public Health (to be published July/Aug 2010)

Laura N. Anderson1, 2 *, Michelle Cotterchio1, 2, Beatrice A. Boucher1, 2, 3, Julia A. Knight1, 4,

Torin Block5

1Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada;

2Population Studies & Surveillance, Cancer Care Ontario, Toronto, Ontario, Canada;

3Department of Nutritional Sciences, University of Toronto, Toronto, Ontario, Canada;

4Prosserman Centre for Health Research, Samuel Lunenfeld Research Institute, Mount Sinai

Hospital, Toronto, Ontario, Canada; 5Block Dietary Data Systems, NutritionQuest, Berkeley,

CA, USA

RUNNING TITLE: Dietary vitamin D intake among Ontario women

Conflict of Interest: TB is an owner of NutritionQuest, which holds the copyright on the Block

FFQ. LNA, MC, BAB and JAK have none to declare.

60

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

ABSTRACT

Objectives: To measure and compare dietary vitamin D intake among women in Ontario using a

modified Block 1998 (US) food frequency questionnaire (FFQ) before and after modification for

Canadian-specific vitamin D food fortification.

Methods: An age-stratified random sample of 3,471 women in Ontario (aged 25-74) was

identified using random digit dialing methods. Standard US food values, and a modified

Canadian-specific vitamin D nutrient analysis were applied to the FFQ.

Results: Intake of vitamin D from foods (Canadian nutrient analysis) was 5.3 ± 3.4 µg/day

(mean ± SD) and 45% of women reported vitamin D intake from supplements. Total vitamin D

intakes met the current Adequate Intakes of 5, 10 and 15 µg/day for only 62%, 47%, and 28% of

women aged ≤ 50, 51-70 and ≥71, respectively. Relatively high agreement was found between

the US and Canadian nutrient analysis methods of measuring vitamin D from food (weighted

kappa = 0.74, 95% CI 0.72, 0.76). Intake differences (US minus Canadian) ranged from -5.0

µg/day to +2.0 µg/day (1st – 99th percentile); however, the mean difference was only -0.54

µg/day (95% CI: -0.58, -0.50).

Conclusions: Lower than recommended total vitamin D intakes were observed among our study

participants which may negatively impact the health status of women. Adjustment for Canadian

food fortification and the inclusion of fatty fish had little impact on the measurement of vitamin

D from food.

MeSH subject headings: vitamin D; food, fortified; nutrition surveys; female; Canada; United

States

61

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

Vitamin D is important for maintenance of healthy bones, and low vitamin D intake may be a

risk factor for some cancers and other chronic diseases (1-3). Vitamin D is synthesized by the

skin following sunlight exposure and is present in foods and supplements. Since few foods

contain high amounts of vitamin D (4), many countries have their own food fortification policies

to improve vitamin D levels. In Canada, fortification of fluid milk and margarine with vitamin D

is mandatory (5). Manufacturers are permitted to use fortified milk to make milk products (e.g.,

yogourt) and to fortify milk beverage substitutes, and some other foods such as orange juice, but

these items are not universally enriched. In the United States (US), where vitamin D fortification

is optional, most milk and many breakfast cereals are fortified (6).

In Canada and the US, vitamin D intakes are evaluated against Adequate Intakes (AIs). The AIs

for vitamin D are 5, 10 and 15 µg/day for adults ≤ 50, 51-70 and >70 years of age, respectively

(7). Measuring diet is important for nutritional epidemiology studies and surveillance and a

commonly used tool is the food frequency questionnaire (FFQ). American FFQs are frequently

applied in such studies, without modifying nutrient databases for population-specific food

values. Few studies (8) have investigated the impact of this practice on the measurement of

vitamin intakes.

The objectives of the current study were to 1) describe vitamin D intake among women in

Ontario, from food and supplements, and 2) compare vitamin D intakes using a US nutrient

analysis versus a modified analysis that reflects additional vitamin D sources (fatty fish), and

Canadian food fortification.

METHODS

Study description

Women aged 25-74 years were identified using random digit dialing of households in Ontario

between 2002 and 2003. These women were recruited as controls for a case-control study

evaluating various epidemiologic factors and breast cancer risk (9). This study was approved by

the University of Toronto Research Ethics Board. Of 4,352 households with eligible women,

3,471 (80%) completed a mailed self-administered risk factor questionnaire and an FFQ.

Subjects were asked about foods and supplements they “usually ate about two years ago”.

62

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

Measurement of vitamin D (FFQ and database values for nutrient analysis)

Description of the FFQ

The quantitative Block 1998 FFQ (9) used in this study was modified to improve measurement

of specific dietary components, including vitamin D, and included 178 food items. A sub-

question querying the type of fish eaten most often (white or fatty), and a supplement item for

vitamin D or cod liver oil were added to the FFQ. Validity was assessed against two 24-hour

recalls among Ontario women using US nutrient data (10). Based on usual (current) intake, FFQ

reliability for vitamin D was relatively high (non-deattenuated r = 0.76, 95% CI 0.66, 0.83), and

its validity was moderately high (deattenuated r = 0.54, 95% CI 0.29, 0.79) (10).

Description of the standard (US) and modified (Canadian) nutrient analyses

Vitamin D intake from food was initially measured by applying standard US nutrient values from

Block Dietary Data Systems (BDDS) to the FFQ. Nutrient values were based on the US

Department of Agriculture (USDA) National Nutrient Database (11), and published literature.

BDDS uses national US consumption data to estimate and weight the proportionate use of foods

within each FFQ item (12, 13).

To modify the nutrient analysis for Canada, vitamin D values of all relevant foods in the BDDS

FFQ database were compared to corresponding foods in the Canadian Nutrient file (CNF),

Canada’s standard reference database (4). Table 1 presents vitamin D values assigned to the

primary food sources on the FFQ. The added fish question, only incorporated in the Canadian-

analysis, assigned a higher vitamin D value to women reporting they most often consumed fatty

rather than white fish based on average fish values in the CNF. Items with very low levels of

naturally occurring vitamin D were not modified despite some observed differences between the

BDDS (US) nutrient values and those listed in the CNF. BDDS' standard US-based consumption

weighting values were used in both analyses.

Supplement analysis

The analysis for vitamin D from supplements (multivitamins, and vitamin D supplements or cod

liver oil) was not modified as there are no data suggesting a consistent difference in the vitamin

D content of supplements between Canada and the US. BDDS assigned a vitamin D value of 10

63

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

µg to multivitamins; 10 µg was also assigned to the additional vitamin D supplement or cod liver

oil question (14).

Statistical analysis

The frequency distributions of respondent characteristics and vitamin D intake were tabulated.

Four vitamin D intake variables are reported: 1) Canadian vitamin D from foods (values from the

CNF); 2) US vitamin D from foods (US values provided by BDDS); 3) vitamin D from

supplements; and 4) total vitamin D (combined vitamin D from supplements and Canadian food

values). The weighted kappa statistic and 95% CI were calculated to assess the chance-corrected

agreement between categories of vitamin D intakes obtained from the Canadian and US food

analyses. The paired t-test was used to determine if the mean difference between the Canadian

and US analyses using the continuous measures of vitamin D from foods was different from

zero. Statistical analysis was conducted using SAS version 9.1.

RESULTS

Data analysis was completed on 3,393 of the 3,471 (98%) questionnaires; 44 were considered

incomplete due to a large number of missing responses and 34 were excluded due to unlikely

energy intakes (< 500 or > 4500 kcal per day). The maximum daily vitamin D intakes from food

and supplements were 30 µg and 20 µg, respectively, and seemed plausible. Table 2 describes

the distribution of subject characteristics.

Using Canadian values, the proportion of women meeting the AI for vitamin D from food alone

decreased with age (Table 3); 44% of women age 25-50 met the AI of 5µg/day, only 10% of

women age 51-70 met the AI of 10µg/day, and no women age 71-74 met the AI of 15µg/day.

Most women (55%) did not consume any vitamin D from supplements (single product vitamin

D/cod-liver oil, or multivitamin); 38% were multivitamin users and 14% were vitamin D/cod

liver oil users (7% took both). For total vitamin D intake (supplements and Canadian food

values) only 62% of women age 25-50, 48% of women age 51-70, and 28% of women age 71-74

had total vitamin D intake that met the AI for their age range. No women had total dietary

vitamin D intakes greater than the tolerable upper intake level (50 µg/day).

Mean (± SD) vitamin D intake from supplements alone was 4.4 ± 5.7 µg/day and total combined

intake was 9.7 ± 6.9 µg/day. Mean (± SD) intake of vitamin D from food was 5.3 ± 3.4 and 4.8 ±

64

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

3.2 µg/day for the Canadian and US nutrient analyses, respectively. The mean US minus

Canadian food difference was -0.54 µg/day (95% CI: -0.50, -0.58) (p<0.0001) and the

distribution of differences (1st - 99th percentile) ranged from -5.0 µg/day to 2.0 µg/day. There

was relatively high agreement between the categories of vitamin D intake from food alone, using

US and Canadian values (weighted kappa = 0.74, 95% CI: 0.72, 0.76). However, an additional

4%, 3% and 0% of women age 25-50, 51-70 and 70-74, respectively, would be misclassified as

‘inadequate’ from food if US values were relied upon (Table 3).

DISCUSSION

Even after modification for Canadian specific values, low intake of total combined vitamin D

(foods and supplements) was observed in our study and was most pronounced among women age

71-74, despite their higher use of supplements. High agreement and limited misclassification

were observed between the two food measures, suggesting the standard US Block FFQ and

nutrient analysis may be adequate for the measurement of vitamin D foods among Canadians.

Using a standard FFQ with US food values can both over- and under-estimate Canadian vitamin

D intakes, although the magnitude of the mean difference was relatively small.

One previous study examined differences in US versus Canadian vitamin intakes using an FFQ

and also found mean Canadian vitamin D intake was slightly underestimated using US values

(8). This study, conducted in Alberta among 7,659 women age 35-69, also reported few women

meeting the AI for vitamin D from food only (30% of women age 31-50, and only 3% of women

age 51-70); supplement intake was not described (8). The Canadian Community Health Survey

(CCHS), a population-based survey of food only using 24-hour recalls (Canadian nutrient

analysis), suggests the AI are met by only 36%, 42% and 9.3% of Canadian women ages 19-30,

31-50, and 51-70, respectively (supplement data are not currently available) (15). The CCHS

(15) reports higher intakes of vitamin D from food than measured by FFQ in our study or

Csizmadi et al (8) but still suggests a large proportion of women are not meeting the current AIs.

The inclusion of supplements in our study increased the proportion of women meeting the AIs

but still indicates that many women may have inadequate intakes of vitamin D. Since our data

were collected in 2002-2003 reflecting 2000-2001 intakes, it is possible that recent mass media

describing the potential benefits of vitamins D may have led to increased vitamin D intake from

food and/or supplements.

65

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

A potential limitation of any dietary study is measurement error. There are concerns regarding

the accuracy of the analysis of vitamin D content of foods (16, 17) and the amount (as reported

in nutrient databases) has been found to vary greatly in both fish (18) and fortified milk (19, 20).

FFQs are also a source of measurement error in nutritional epidemiology and are best used to

capture relative rather than absolute individual intake (21, 22). However, the measurement of

usual vitamin D using this FFQ was shown to have moderately high validity and mean intake

was not significantly different when compared to two 24-hour recalls (10). Although, our study

measured usual diet ‘two years ago’, diet is expected to be stable over time (23). An additional

limitation is that nutrient databases change over time and we applied the most recent version of

the CNF to earlier intake data. The lack of modification of nutrient databases for country specific

vitamin D fortification regulations likely introduces error that may bias disease association study

findings, although, we found high agreement and little misclassification between the two

measures of vitamin D intake. Measurement error due to respondent memory is always a concern

in epidemiologic studies and we were unable to evaluate this.

Sun exposure is another important source of vitamin D, yet, many studies of vitamin D and

disease risk have focused only on vitamin D from diet/supplements. Optimal vitamin D status

has been proposed at 25 hydroxyvitamin D [25(OH)D] serum levels >30 ng/mL (1, 24) and it has

been shown that intakes >12.5 µg/day (25, 26) are required to maintain optimal wintertime

25(OH)D levels in Northern populations, such as Canada. The mean total intake among all

women in our study was only 9.7 µg/day suggesting the potential for sub-optimal serum

25(OH)D levels. There are two forms of vitamin D: vitamin D3 (from fatty fish, most

supplements and fortified foods) and vitamin D2 (from plants). These forms of vitamin D may

differ in biologic activity (27) and were not measured in this study, but we would suspect that

most would be vitamin D3.

Many researchers have concluded that the current AIs are not sufficient to maintain optimal

serum levels of 25(OH)D (25, 28) and the dietary reference intakes for vitamin D are under

review (29, 30). Considering the low proportion of women who met current AIs in this study,

more work is needed to ensure that women are consuming sufficient vitamin D. Despite

fortification differences between Canada and the US there is relatively high agreement in vitamin

D intake from food using a US FFQ before and after modification for Canadian values (and

66

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

including fatty fish). Lower than recommended total vitamin D intakes were observed among our

study participants which may negatively impact health (1-3).

Literature Cited

1. Holick MF. Vitamin D: A D-lightful health perspective. Nutr Rev. 2008;66(10 Suppl 2):S182-94.

2. Holick MF. Vitamin D deficiency. N Engl J Med. 2007;357(3):266-81.

3. Giovannucci E. Vitamin D status and cancer incidence and mortality. Adv Exp Med Biol. 2008;624:31-42.

4. Health Canada. Canadian Nutrient File, 2007b version. 2007 [cited 2009 Mar 30]. Available from: www.healthcanada.ca/cnf

5. A Health Canada. Addition of Vitamins and Minerals to Foods. 2005 [cited 2009 Mar 30]. Available from: http://www.hc-sc.gc.ca/fn-an/nutrition/vitamin/fortification_final_doc_1-eng.php#c6

6. Calvo MS, Whiting SJ, Barton CN. Vitamin D fortification in the united states and canada: Current status and data needs. Am J Clin Nutr. 2004;80(6 Suppl):1710S-6S.

7. Standing Committee on the Scientific Evaluation of Dietary Reference Intakes, Food and Nutrition Board, Institute of Medicine. Dietary Reference Intakes for calcium, phosphorus, magnesium, vitamin D and flouride. Washington, DC: National Academy Press; 1997.

8. Csizmadi I, Kahle L, Ullman R, Dawe U, Zimmerman TP, Friedenreich CM, Bryant H, Subar AF. Adaptation and evaluation of the national cancer institute's diet history questionnaire and nutrient database for Canadian populations. Public Health Nutr. 2007;10(1):88-96.

9. Cotterchio M, Boucher BA, Kreiger N, Mills CA, Thompson LU. Dietary phytoestrogen intake--lignans and isoflavones--and breast cancer risk (Canada). Cancer Causes Control. 2008;19(3):259-72.

10. Boucher B, Cotterchio M, Kreiger N, Nadalin V, Block T, Block G. Validity and reliability of the Block98 food-frequency questionnaire in a sample of Canadian women. Public Health Nutr. 2006;9(1):84-93.

11. United States Department of Agriculture, Agricultural Research Service. USDA National Nutrient Database for Standard Reference. 2009 [cited 2009 Mar 30]. Available from: http://www.ars.usda.gov/ba/bhnrc/ndl

12. Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, Gardner L. A data-based approach to diet questionnaire design and testing. Am J Epidemiol. 1986;124(3):453-69.

13. Block G. Invited commentary: Another perspective on food frequency questionnaires. Am J Epidemiol. 2001;154(12):1103,4; discussion 1105-6.

14. Moore C, Murphy MM, Keast DR, Holick MF. Vitamin D intake in the United States. J Am Diet Assoc. 2004;104(6):980-3.

67

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

15. Health Canada, Statistics Canada. Canadian Community Health Survey (CCHS) Cycle 2.2, Nutrition (2004). Nutrient intakes from food. Provincial, regional and national summary data tables. 2nd vol. Ottawa: Health Canada; 2008. Report No.: H164–45/2–2008E-PDF.

16. Holden JM, Lemar LE, Exler J. Vitamin D in foods: Development of the US department of agriculture database. Am J Clin Nutr. 2008;87(4):1092S-6S.

17. Byrdwell WC, Devries J, Exler J, Harnly JM, Holden JM, Holick MF, Hollis BW, Horst RL, Lada M, et al. Analyzing vitamin D in foods and supplements: Methodologic challenges. Am J Clin Nutr. 2008;88(2):554S-7S.

18. Lu Z, Chen TC, Zhang A, Persons KS, Kohn N, Berkowitz R, Martinello S, Holick MF. An evaluation of the vitamin D3 content in fish: Is the vitamin D content adequate to satisfy the dietary requirement for vitamin D? J Steroid Biochem Mol Biol. 2007;103(3-5):642-4.

19. Faulkner H, Hussein A, Foran M, Szijarto L. A survey of vitamin A and D contents of fortified fluid milk in Ontario. J Dairy Sci. 2000;83(6):1210-6.

20. Chen TC, Shao A, Heath H,3rd, Holick MF. An update on the vitamin D content of fortified milk from the United States and Canada. N Engl J Med. 1993 Nov 11;329(20):1507.

21. Kristal AR, Peters U, Potter JD. Is it time to abandon the food frequency questionnaire? Cancer Epidemiol Biomarkers Prev. 2005;14(12):2826-8.

22. Schatzkin A, Subar AF, Moore S, Park Y, Potischman N, Thompson FE, Leitzmann M, Hollenbeck A, Morrissey KG, Kipnis V. Observational epidemiologic studies of nutrition and cancer: The next generation (with better observation). Cancer Epidemiol Biomarkers Prev. 2009;18(4):1026-32.

23. Goldbohm RA, van 't Veer P, van den Brandt PA, van 't Hof MA, Brants HA, Sturmans F, Hermus RJ. Reproducibility of a food frequency questionnaire and stability of dietary habits determined from five annually repeated measurements. Eur J Clin Nutr. 1995 Jun;49(6):420-9.

24.Vieth R. What is the optimal vitamin D status for health? Prog Biophys Mol Biol. 2006;92(1):26-32.

25. Whiting SJ, Green TJ, Calvo MS. Vitamin D intakes in North America and Asia-Pacific countries are not sufficient to prevent vitamin D insufficiency. J Steroid Biochem Mol Biol. 2007;103(3-5):626-30.

26. Cashman KD, Hill TR, Lucey AJ, Taylor N, Seamans KM, Muldowney S, Fitzgerald AP, Flynn A, Barnes MS, et al. Estimation of the dietary requirement for vitamin D in healthy adults. Am J Clin Nutr. 2008;88(6):1535-42.

27. Houghton LA, Vieth R. The case against ergocalciferol (vitamin D2) as a vitamin supplement. Am J Clin Nutr. 2006;84(4):694-7.

28. Vieth R, Bischoff-Ferrari H, Boucher BJ, Dawson-Hughes B, Garland CF, Heaney RP, Holick MF, Hollis BW, Lamberg-Allardt C, et al. The urgent need to recommend an intake of vitamin D that is effective. Am J Clin Nutr. 2007;85(3):649-50.

29. US Department of Agriculture, Human Nutrition Information Service. Provisional table on the vitamin D content of foods. Washington, DC: US Government Printing Office, 1980. Revised 1999.

68

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

30. Yetley EA, Brule D, Cheney MC, Davis CD, Esslinger KA, Fischer PW, Friedl KE, Greene-Finestone LS, Guenther PM, et al. Dietary reference intakes for vitamin D: Justification for a review of the 1997 values. Am J Clin Nutr. 2009;89(3):719-27.

69

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

Table 1. Vitamin D food values assigned to the standard (US) nutrient analysis and the modified

Canadian analysis

Vitamin D values

Foods US* Canada

†

µg per 100g

Breakfast cereals 3.5 0

Margarine 1.5 13.3

Fish (not fried)‡ 1.5 -

Fatty fish - 10.0 White fish - 1.5

Milk 1.0 1.0

* Values from Block Dietary Data Systems

† Values from Canadian Nutrient File

‡ Fish type was added to the modified Canadian analysis and a higher value was assigned to fatty fish. The type of

fish question was not applied to the US analysis.

70

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

Table 2. Distribution of subject characteristics among all participating Ontario women (n =

3,393)

Variable n (%)*

Age (years) 25 - 50 51 - 70 71 - 74

1251 (37) 1902 (56) 240 (7)

Highest level of Education Elementary or High school Postsecondary

1716 (51) 1664 (49)

BMI† (kg/m2)

< 24.9 25.0 - 29.9 ≥ 30

1506 (45) 1145 (34) 719 (21)

Smoking status Never smoker Ex-smoker Current smoker

1741 (51) 1107 (33) 523 (15)

Ethnicity Caucasian

South East Asian‡

Black Other

3062 (91) 95 (3) 50 (2) 169 (5)

* Numbers may not add to totals because of missing values and/or rounding † BMI calculated as weight two years ago in kilograms divided by height in meters squared ‡ e.g., Japanese, Chinese, Korean, Vietnamese

71

Paper 1: Dietary Vitamin D Intake among Ontario Women (Anderson et al, in press)

Table 3. Distribution of vitamin D intake among Ontario women (total and stratified by age

group)

Age group (years)

Vitamin D intake (µg/day)

Total

(n = 3,393)

n (%)

25-50

(n = 1,251)

n (%)

51-70

(n = 1,902)

n (%)

71-74

(n = 240)

n (%)

Foods - Cdn values <5 5 – 9.9 10 – 14.9 ≥15

1877 (55) 1193 (35) 280 (8) 43 (1)

699 (56) 435 (35) 102 (8) 15 (1)

1043 (55) 670 (35) 162 (8) 27 (2)

135 (56) 88 (37) 16 (7) 1 (0)

Foods - US values <5 5 – 9.9 10 – 14.9 ≥15

2096 (62) 1049 (31) 223 (7) 25 (1)

751 (60) 394 (31) 94 (8) 12 (1)

1195 (63) 573 (30) 121 (6) 13 (1)

150 (63) 82 (34) 8 (3) 0 (0)

Supplements* 0 1 - 4.9 5 – 9.9 10 – 14.9 ≥15

1875 (55) 230 (7) 178 (5) 940 (28) 170 (5)

787 (63) 113 (9) 74 (6) 246 (20) 31 (2)

973 (51) 104 (5) 99 (5) 606 (32) 120 (6)

115 (48) 13 (6) 5 (2) 88 (37) 19 (8)

Total† <5 5 – 9.9 10 – 14.9 ≥15

1132 (33) 825 (24) 729 (21) 707 (21)

475 (38)‡ 363 (29) 229 (18) 184 (15)

583 (31)‡ 413 (22)‡ 449 (24) 457 (24)

74 (31)‡ 49 (20)‡ 51 (21)‡ 66 (28)

* Supplemental vitamin D includes multivitamins and vitamin D supplements or cod liver oil. † Total vitamin D from food (Canadian nutrient values) and supplements ‡ Intakes below the AI for that age group; daily AI for ages 25-50 = 5 µg; 51-70 = 10 µg; and ≥71 = 15 µg.

72

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

4.3 Paper 2: Vitamin D and Calcium Intakes and Breast Cancer Risk in Pre- and Postmenopausal Women

American Journal of Clinical Nutrition (AJCN). Vol. 91, No. 6, 1699-1707, June 2010. First

published April 14, 2010; doi: 10.3945/ajcn.2009.28869.

Laura N. Anderson1,2, Michelle Cotterchio1,2, Reinhold Vieth3, Julia A. Knight 2,4

1Population Studies & Surveillance, Cancer Care Ontario, 620 University Ave., Toronto, ON;

2Dalla Lana School of Public Health, University of Toronto, Toronto, ON; 3Department of

Pathology and Laboratory Medicine, Mount Sinai Hospital, Toronto, ON; 4Prosserman Centre

for Health Research, Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, ON

Short title: Vitamin D, calcium and breast cancer risk

Abstract

Background: Some evidence suggests that vitamin D may reduce breast cancer risk. Despite the

biologic interaction between vitamin D and calcium, few studies have evaluated their joint

effects on breast cancer risk.

Objective: To evaluate the associations and potential interaction between vitamin D and calcium

(from food and supplements) and breast cancer risk in a population-based case-control study.

Design: Breast cancer cases aged 25 - 74 years (diagnosed 2002-2003) were identified through

the Ontario Cancer Registry. Controls were identified using random digit dialing. 3101 cases and

3471 controls completed epidemiologic and food frequency questionnaires. Adjusted odds ratios

(OR) and 95% confidence intervals (CI) were estimated using multivariate logistic regression.

Results: Vitamin D and calcium intakes from food only and total combined intake (food and

supplements) were not associated with breast cancer risk, although mean intake of vitamin D was

low. Vitamin D supplement intake > 10µg (400 IU) per day versus no intake was associated with

a reduced risk of breast cancer (adjusted OR: 0.76; 95% CI: 0.59, 0.98). No categories of

calcium supplement intake were significantly associated with reduced breast cancer risk but a

significant inverse trend was observed (p = 0.04). There were no significant interactions

involving vitamin D, calcium or menopausal status.

73

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

Conclusions: No associations were found between overall vitamin D or calcium intake and breast

cancer risk. Vitamin D from supplements was independently associated with reduced breast

cancer risk. Further research is needed to investigate the effects of higher doses of vitamin D and

calcium supplements.

INTRODUCTION

Breast cancer is the most common cancer among Canadian women (1) and few modifiable risk

factors have been identified (e.g., alcohol consumption, hormone replacement therapy, obesity

among postmenopausal women) (2-4). Several review papers concluded that, despite

inconsistencies in the literature and identified areas that still require investigation, low vitamin D

intake may also increase breast cancer risk (5-11). One meta-analysis found no overall

association between vitamin D, from diet and supplements, and breast cancer risk, but did

suggest an inverse association may exist at higher intakes (12). Vitamin D is synthesized in the

skin following sufficient ultraviolet B exposure from sunlight and is found in some foods (e.g.,

fortified milk and fatty fish), and vitamin supplements (13). Vitamin D (from diet and sunlight)

is hydroxylated by the liver to the circulating form 25-hydroxyvitamin D (25(OH)D) (the

preferred biomarker for vitamin D). A second hydroxylation in the kidney or in other cells,

including breast cells, produces the active hormone 1,25-dihydroxyvitamin D (1,25(OH)2D). The

vitamin D receptor is present in many cells, including normal and cancerous breast cells,

enabling these cells to respond to 1,25(OH)2D (14-16). Laboratory studies have shown

1,25(OH)2D promotes cell differentiation and inhibits cell growth (14-17). Some studies of

25(OH)D and breast cancer risk have found an inverse association (18-21), though not all (22-

24).

It is well established that 1,25(OH)2D regulates calcium metabolism and that vitamin D and

calcium are found in some of the same foods (e.g., vitamin D fortified milk). Calcium may also

have anticarcinogenic properties that include regulation of cell differentiation, proliferation and

apoptosis (25-27). However, results from epidemiologic studies do not strongly support an

inverse association between calcium, or more generally dairy products, and breast cancer risk (8,

28-32). The Women’s Health Initiative trial –in which postmenopausal women were randomized

to 10 µg (400 IU) vitamin D and 1000 mg calcium daily or a placebo - found no reduction in

breast cancer risk after a mean follow-up time of 7 years (22). Few observational studies of

74

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

dietary vitamin D and breast cancer risk have investigated the interaction between calcium and

vitamin D (33-35).

Currently, it is unclear whether the possible association between dietary vitamin D and reduced

breast cancer risk is confounded or modified by calcium and vice versa. Furthermore, there is

limited evidence suggesting the association between dietary vitamin D and breast cancer risk

may (34, 35) or may not (36, 37) differ by menopausal status. The objectives of this study were

to evaluate the associations and potential interaction between vitamin D and calcium (from food

and supplements) and breast cancer risk in a population-based case-control study of pre- and

post-menopausal women in Ontario.

SUBJECTS AND METHODS

Data were collected as part of the Ontario Women’s Diet and Health Study, a large population

based case-control study evaluating various epidemiologic factors and breast cancer risk (38).

The study protocols for this study were approved by the University of Toronto Research Ethics

Board.

Cases

Cases were women aged 25 - 74 with a first, pathologically confirmed cancer of the breast

identified from the Ontario Cancer Registry and diagnosed between June 2002 and April 2003.

The Ontario Cancer Registry is a population-based registry that obtains information from nearly

all breast cancer cases in the province in Ontario (39, 40). Physician cooperation was required to

contact patients, obtain contact information and vital status. Consent was obtained for 4,109

eligible cases (96%). The average time between diagnosis and interview was 11 months, with a

range (5th to 95th percentile) from 7 to 18 months.

Controls

Random digit dialing methods were used to identify eligible controls among households in

Ontario and frequency-matched (1:1) within 5-year age groups to the identified cases, described

in detail elsewhere (38). Only one woman from each household was randomly selected for

inclusion in the study. Approximately 25,250 households were telephoned; ≈17,000 of these

households were ineligible (e.g., no woman between the ages 25 to 74 with no history of breast

75

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

cancer), 2,000 did not answer the phone and 2,000 refused (eligibility of these households is

unknown). Of the 4,352 households where an eligible woman was identified, 250 women

refused, and 4,102 women (94%) agreed to participate.

Data Collection & Response rate

Cases and controls were mailed an epidemiologic questionnaire and food frequency

questionnaire (FFQ). The epidemiologic questionnaire consisted of 79 questions and collected

information of lifestyle, reproductive and medical history factors. These questionnaires were

completed and returned by 3,101 cases (75% response rate) and 3,471 controls (80% response

rate).

Measurement of vitamin D and calcium

A modified Block FFQ measured 178 foods and supplements and requested data on both

frequency of consumption and portion size for all foods and frequency and duration of use for

supplements. The validity and reliability of the Block FFQ have been assessed in a random

sample of Ontario women (41). The reliability was high for vitamin D, the non-deattenuated

Pearson correlation coefficients was 0.76 (95% CI: 0.66, 0.83), and for calcium (r = 0.80; 95%

CI: 0.71, 0.86). Validity of the FFQ, compared to a 24-hour recall, was also moderately high for

vitamin D (deattenuated Pearson correlation coefficient was 0.54; 95% CI: 0.29, 0.79) and for

calcium (r = 0.71; 95% CI: 0.35, 1.00) (41). Cases and controls were asked about their

consumption of food and supplements two years prior to the time of questionnaire, to reduce any

bias due to changes in diet following cancer diagnosis. Nutrient analysis was conducted by Block

Dietary Data Systems using nutrient values from the USDA Nutrient Database for Standard

Reference and national data on food consumption (NHANES III and CSFII) (42, 43). In the

modified version of the FFQ used for this study two questions were added to better capture

vitamin D intake: type of fish most often consumed (fatty or white fish) and use of vitamin D

supplements or cod liver oil. The nutrient analysis specific to vitamin D was modified to account

for these additional items and food fortification differences between Canada and the US.

Calcium supplement use, alone or combined with something else, was measured but data on

combined calcium plus vitamin D supplements were not available.

76

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

Derived variables from the Block nutrient analysis were obtained for daily intake of vitamin D

and calcium (mg/day), from all foods, and supplements (single product supplements,

multivitamins, or cod liver oil for vitamin D). A combined total intake measure (foods plus

supplements) was also created. In addition to the derived total nutrient values for vitamin D and

calcium (from foods, supplements and total), the following individual foods rich in vitamin D

and/or calcium and vitamin supplement sources were individually examined: milk, fish,

margarine, single product supplement measures of vitamin D (or cod liver oil) and calcium

(alone or combined with something else), and regular one-a-day multiple vitamins. Individual

food intakes were calculated as servings per day, week or month (depending on frequency of

consumption).

Measurement of other variables

The following variables were tested as potential confounders: marital status, education, ethnicity,

body mass index (BMI), smoking status, pack years smoked, breastfeeding history, breastfed as

an infant, age at menarche, oral contraceptive use, oral contraceptive duration, parity, age at first

live birth, age at menopause, hormone replacement therapy (HRT) use (postmenopausal women

only), duration of HRT use, history of benign breast disease, family history of breast cancer,

screening mammogram, alcoholic drinks, dietary fat intake, calorie intake, phytoestrogen intake,

physical activity (strenuous, moderate and daily activity) at selected periods of life (teenage

years, 20-30s, 40-50s and 60s-74), and sun exposure variables (time spent outdoors, location of

residence, skin color, and sun protection practices) at selected periods of life (teenage years, 20-

39, 40-59 and 60-74 years).

Statistical data analysis

Unconditional logistic regression analysis was used to obtain age-adjusted odds ratios (OR) and

95% CIs for all vitamin D and calcium variables (at both the food and nutrient level). Age was

calculated as age at diagnosis for breast cancer cases and age at midpoint of recruitment for

controls. Confounders were defined as any variable that changed the OR of the exposure variable

by >10% when added to the model (44). None of the variables met our definition of a

confounder; however, to be conservative we also constructed multivariate models that adjusted

for age, education, age at menarche, age at first live birth, parity, menopausal status, breast

cancer in first degree relative, total energy intake, BMI, pack years smoked, moderate physical

77

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

activity during ages 20-30, moderate physical activity during ages 40-59, time spent outdoors per

week during ages 20-39, time spent outdoors per week during ages 40-59, total calcium intake

(included in the models of vitamin D variables only) and total vitamin D intake (included in the

models of calcium variables only). Test for linear trend was calculated by treating the median

intake for each exposure category as a continuous variable in the age-adjusted and fully-adjusted

logistic regression models. Tests for multiplicative interactions were calculated using the

Likelihood Ratio test. To assess the interaction between vitamin D and calcium, calcium was

categorized into two categories: high versus low intake (since calcium and vitamin D intake were

highly correlated resulting in small numbers in the extreme categories, e.g., lowest vitamin D and

highest calcium). Stratified results are presented by calcium intake and menopausal status.

Interactions between vitamin D or calcium intake and BMI and hormone replacement were also

tested. Statistical significance was defined as P value less than 0.05 and all tests were two-sided.

All statistical analyses were conducted using SAS version 9.1.

RESULTS

Overall 6,572 (3,101 cases and 3,471 controls) women completed the questionnaires. The mean

(± SD) age of study participants was 56 years (11). The majority of women in this study (90%)

were Caucasian and many had postsecondary education (46% of cases and 49% of controls).

Characteristics of the cases and controls and age-group adjusted ORs for selected factors that

may be associated with breast cancer risk or vitamin D status are shown in Table 1 and have

been described previously (38).

No significant ORs were observed between milk, margarine, dairy, or fish intake and breast

cancer risk (Table 2). However, OR point estimates increased with milk intake (p for trend =

0.04). Single product vitamin D supplements or cod liver oil were used by only 13% of cases and

14% of controls and although no categories of frequency or duration of use were significantly

associated with breast cancer risk, but a significant inverse dose-response relation was observed

between frequency of supplement use and breast cancer risk (p for trend = 0.04). Calcium

supplement use was more common (33% of cases and 35% of controls) and, similar to vitamin D

supplement use, none of the categories of intake for either frequency or duration were

significant, but OR point estimates decreased with frequency of calcium supplement use (p for

78

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

trend = 0.04). Neither frequency nor duration of multivitamin use was associated with breast

cancer risk.

Results for the vitamin D and calcium nutrient-level variables (intake from foods only,

supplements only and total combined) are presented in Table 3. No confounders were identified.

Although the risk of breast cancer was reduced among women with vitamin D from supplement

intakes >10 µg/day (400 IU/day) compared to none (fully adjusted OR: 0.76; 95% CI: 0.59,

0.98), no dose-response relation was observed between vitamin D supplements and breast cancer

risk. No associations were observed between total combined vitamin D intake or vitamin D

intake from foods alone and breast cancer risk. No statistically significant associations were

observed between calcium and breast cancer risk; however, the OR point estimates decreased

with increased calcium supplement dose (p for trend = 0.04).

Vitamin D and calcium were highly correlated with a strong positive correlation observed

between the continuous measures from food (Pearson’s r = 0.79, P < 0.0001), supplements (r =

0.50, p < 0.0001) and total intake (r = 0.63, p < 0.0001). The interactions between total calcium

(high versus low intake) and all categorical vitamin D variables were not significant (Table 4).

The odds ratios in the stratified analysis do not appear substantially different suggesting no effect

modification. When the interactions were assessed using supplemental calcium intake (yes

versus no), which was less highly correlated to all measures of vitamin D, there were still no

significant interactions. The relation between calcium and vitamin D was not different among

pre- and post-menopausal women (data not shown).

Similarly, no significant interactions between the vitamin D or calcium nutrient-level variables

and menopausal status were observed (Table 5). There were also no statistically significant

interactions and the ORs did not differ substantially by menopausal status for intake of milk,

margarine, and other fish or for duration/frequency of calcium, vitamin D or multivitamin use

(data not shown). However, a significant interaction was observed between tuna intake and

menopausal status (p = 0.02); no association was observed among premenopausal women,

although a significant inverse association existed among postmenopausal women (comparing

highest to lowest category OR: 0.78; 95% CI: 0.65, 0.93). Further tests for interactions revealed

no significant interactions between any of the three vitamin D nutrient-level variables and BMI

or hormone replacement therapy use (among postmenopausal women only).

79

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

DISCUSSION

This study provides some evidence that vitamin D supplement use, but not intake from food

alone or combined intake, is independently associated with reduced breast cancer risk. Our data

suggest there may be a threshold at >10 µg/day (400 IU/day) for vitamin D supplements. No

significant ORs were observed for calcium intake from foods, supplements or total combined

intake and breast cancer risk; however, a significant inverse trend was observed across categories

of calcium supplement use. Measurement of vitamin D or calcium from foods compared to

supplements may be more susceptible to misclassification (potentially biasing results towards the

null). It is also possible that foods containing vitamin D and calcium contain other detrimental

components that counteract the potential benefits from vitamin D (e.g., dietary fat in milk) (29,

32). We cannot rule out the possibility that our observed associations were due to chance or

residual confounding; however, multivitamin use was not associated with breast cancer risk

suggesting the associations are not due to residual confounding by other unmeasured healthy

lifestyle traits among supplement users.

A recent meta-analysis suggested a trend towards reduced breast cancer risk at vitamin D

minimum intakes of 10 µg/day or greater but no association overall (RR: 0.98; 95% CI: 0.93,

1.03) (12). Several large cohort studies (34, 35, 45-47) have all reported some inverse

associations between vitamin D intake (from diet and/or supplements) and breast cancer risk, but

none have reported significant inverse associations consistently for all sources of vitamin D

intake measured or among all women (e.g., pre- and post-menopausal). The Women’s Health

Initiative trial of breast cancer risk among postmenopausal women randomized to 10 µg vitamin

D plus 1000 mg calcium daily or a placebo (HR: 0.96; 95% CI: 0.85, 1.09) (22) was consistent

with our results that vitamin D ≤10 µg/day (400 IU/day) was not associated with breast cancer

risk. A smaller trial of all cancer sites combined (few breast cancer cases) among

postmenopausal women randomized to 27.5 µg (1100 IU) vitamin D plus 1500 mg calcium per

day versus placebo observed a significant reduction in cancers (RR: 0.40; 95% CI: 0.20, 0.82); a

reduction in risk of borderline statistical significance was also observed among calcium only

(RR: 0.53; 95% CI: 0.27, 1.03) (48).

Only one (36) of the previous large population-based case-control studies of vitamin D intake

and breast cancer risk (33, 36, 37) included vitamin D from supplements and found reduced

80

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

breast cancer risk associated with vitamin D supplement use earlier in life only (36). In contrast

to our null results for dietary vitamin D, other case-control studies of diet only (during

adulthood) have found significant inverse associations above the threshold of only 5 µg (200 IU)

daily among European populations (33, 37).

Our results do not suggest an interaction between calcium and vitamin D intake and these two

variables did not confound one another. Elsewhere, the association between dietary calcium and

breast cancer risk was attenuated and no longer statistically significant after adjustment for

vitamin D (33). Although we observed an independent dose-response trend for calcium

supplements and reduced breast cancer risk, we cannot rule out the possibility of residual

confounding from unmeasured vitamin D intake; many calcium supplement users also consume

vitamin D (usually for bone health) and 19% of calcium supplement users did not report taking

vitamin D supplements. Previous studies (33-35) found no interaction between vitamin D and

calcium (33, 34) or an interaction among postmenopausal women only (35), such that an inverse

association between calcium and breast cancer risk was observed only among the highest

category of vitamin D intake. A recent prospective study of serum calcium (with no measures of

vitamin D) and breast cancer incidence found an inverse association among premenopausal

women only (49). The independent association between calcium and breast cancer risk requires

further investigation.

Studies evaluating the vitamin D-breast cancer relationship by menopausal status have reported

inconsistent results (34-37). Consistent with previous case-control studies (36, 37), we found no

significant interactions. Unfortunately, a small proportion of premenopausal women in our study

were taking vitamin D supplements >10 µg/day (400 IU/day). In contrast, cohort studies have

reported inverse associations between dietary vitamin D intake and breast cancer risk only

among premenopausal women (34, 35) and the potential for increased risk among

postmenopausal women (35). Some studies of vitamin D and breast cancer risk (35, 46, 47), but

not all (50), have reported differences by hormone receptor status. Data on hormone receptor

status is not currently available for our study.

While it has been hypothesized that vitamin D exposure during adolescence may be most

important, we were unable to examine early life dietary vitamin D intake, and the current

evidence is inconsistent (34, 36, 51, 52). Vitamin D may also have some short term effects in

81

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

reducing breast cancer risk (18). Among two large cohort studies the inverse associations with

vitamin D and breast cancer were stronger when recent measures of vitamin D (34) or cases only

within 5-years of baseline measurement of vitamin D intake were considered (47), possibly

indicating a role for vitamin D in slowing disease progression.

As with all observational studies, the potential for measurement error and other biases may limit

internal validity. Sun exposure is the primary source of vitamin D in sunny populations;

however, this study population resides north of 43◦ latitude (Ontario, Canada) where sun

exposure is insufficient for vitamin D production at least 4 months of the year (53) and skin is

mostly covered up for at least half of the year. Most studies among Canadian and other northern

populations have found that dietary vitamin D intake is a significant predictor of 25(OH)D (54-

58), but not all (59). There is the potential for misclassification, likely nondifferential, resulting

from the absence of a complete measure of vitamin D (i.e., a composite measure including

diet/supplements and sun exposure). Our study did not measure the different forms of vitamin D;

however, the majority of intake would likely be vitamin D3, which is found in fatty fish and

commonly used in supplements and food fortification in Canada, versus vitamin D2 from plant

sources. Recall bias is likely minimal in this study as there is no obvious reason why cases or

controls would differentially recall their recent intake of vitamin D or calcium. Similarly,

survival bias is expected to be minimal since there is a high rate of survival among breast cancer

cases in Ontario and women were recruited, on average, within 1 year of diagnosis.

This study has several strengths, including a large sample size, population-based recruitment of

cases and controls and high response rates. Overall the study results do not support an

association between vitamin D or calcium from food or total intake and breast cancer risk.

However, vitamin D intake levels are relatively low in this study and supplemental vitamin D

intake greater than 10 µg/day (400 IU/day) was associated with reduced breast cancer risk.

Future studies are needed among populations with higher intakes, possibly carried out as a

chemoprevention/intervention trial. Additional research is also needed to determine if the

association between vitamin D supplements and reduced breast cancer risk varies by timing of

exposure, menopausal status, and tumor characteristics (including hormone receptor status, stage

of diagnosis).

Acknowledgements

82

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

The authors thank the study coordinator, Noori Chowdhury, for her dedication to this study.

The contributions of each author to the manuscript were as follows: LNA drafted the manuscript

and conducted the statistical analysis. MC, RV, and JAK, provided significant advice on the

study design. MC is the PI and obtained funding for the initial Ontario Women’s Diet and Health

Study. All authors reviewed, revised and approved the manuscript.

LNA, MC and JAK have no conflict of interest. RV has been a consultant or speaker for Carlson

Laboratories, DiaSorin, and Yoplait, and is related to a person employed in the dietary

supplement industry.

83

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

Reference List

1. Canadian Cancer Society/National Cancer Institute of Canada. Canadian Cancer Statistics 2008. Toronto, Canada: 2008.

2. Colditz GA. Epidemiology and prevention of breast cancer. Cancer Epidemiol Biomarkers Prev 2005;14:768-72.

3. Veronesi U, Boyle P, Goldhirsch A, Orecchia R, Viale G. Breast cancer. Lancet 2005;365:1727-41.

4. Kelsey JL, Bernstein L. Epidemiology and prevention of breast cancer. Annu Rev Public Health 1996;17:47-67.

5. Bertone-Johnson ER. Prospective studies of dietary vitamin D and breast cancer: more questions raised than answered. Nutr Rev 2007;65:459-66.

6. Bertone-Johnson ER. Vitamin D and Breast Cancer. Ann Epidemiol 2009;

7. Colston KW. Vitamin D and breast cancer risk. Best Pract Res Clin Endocrinol Metab 2008;22:587-99.

8. Cui Y, Rohan TE. Vitamin D, calcium, and breast cancer risk: a review. Cancer Epidemiol Biomarkers Prev 2006;15:1427-37.

9. Perez-Lopez FR, Chedraui P, Haya J. Review article: vitamin D acquisition and breast cancer risk. Reprod Sci 2009;16:7-19.

10. Lipkin M, Newmark HL. Vitamin D, calcium and prevention of breast cancer: a review. J Am Coll Nutr 1999;18:392S-7S.

11. Rohan T. Epidemiological studies of vitamin D and breast cancer. Nutr Rev 2007;65:S80-3.

12. Gissel T, Rejnmark L, Mosekilde L, Vestergaard P. Intake of vitamin D and risk of breast cancer--a meta-analysis. J Steroid Biochem Mol Biol 2008;111:195-9.

13. Calvo MS, Whiting SJ, Barton CN. Vitamin D fortification in the United States and Canada: current status and data needs. Am J Clin Nutr 2004;80:1710S-6S.

14. Holick MF. Vitamin D: A millenium perspective. J Cell Biochem 2003;88:296-307.

15. Colston KW, Hansen CM. Mechanisms implicated in the growth regulatory effects of vitamin D in breast cancer. Endocr Relat Cancer 2002;9:45-59.

16. Welsh J. Vitamin D and breast cancer: insights from animal models. Am J Clin Nutr 2004;80:1721S-4S.

17. Holick MF. Vitamin D: its role in cancer prevention and treatment. Prog Biophys Mol Biol 2006;92:49-59.

18. Crew KD, Gammon MD, Steck SE, et al. Association between plasma 25-hydroxyvitamin D and breast cancer risk. Cancer Prev Res 2009;2:598-604.

19. Abbas S, Chang-Claude J, Linseisen J. Plasma 25-hydroxyvitamin D and premenopausal breast cancer risk in a German case-control study. Int J Cancer 2009;124:250-5.

20. Bertone-Johnson ER, Chen WY, Holick MF, et al. Plasma 25-hydroxyvitamin D and 1,25-dihydroxyvitamin D and risk of breast cancer. Cancer Epidemiol Biomarkers Prev 2005;14:1991-7.

84

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

21. Lowe LC, Guy M, Mansi JL, et al. Plasma 25-hydroxy vitamin D concentrations, vitamin D receptor genotype and breast cancer risk in a UK Caucasian population. Eur J Cancer 2005;41:1164-9.

22. Chlebowski RT, Johnson KC, Kooperberg C, et al. Calcium plus vitamin D supplementation and the risk of breast cancer. J Natl Cancer Inst 2008;100:1581-91.

23. Freedman DM, Chang SC, Falk RT, et al. Serum levels of vitamin D metabolites and breast cancer risk in the prostate, lung, colorectal, and ovarian cancer screening trial. Cancer Epidemiol Biomarkers Prev 2008;17:889-94.

24. Janowsky EC, Lester GE, Weinberg CR, et al. Association between low levels of 1,25-dihydroxyvitamin D and breast cancer risk. Public Health Nutr 1999;2:283-91.

25. Whitfield JF, Boynton AL, MacManus JP, Sikorska M, Tsang BK. The regulation of cell proliferation by calcium and cyclic AMP. Mol Cell Biochem 1979;27:155-79.

26. McGrath CM, Soule HD. Calcium regulation of normal human mammary epithelial cell growth in culture. In Vitro 1984;20:652-62.

27. Sergeev IN. Calcium as a mediator of 1,25-dihydroxyvitamin D3-induced apoptosis. J Steroid Biochem Mol Biol 2004;89-90:419-25.

28. Larsson SC, Bergkvist L, Wolk A. Long-term dietary calcium intake and breast cancer risk in a prospective cohort of women. Am J Clin Nutr 2009;89:277-82.

29. Moorman PG, Terry PD. Consumption of dairy products and the risk of breast cancer: a review of the literature. Am J Clin Nutr 2004;80:5-14.

30. Bissonauth V, Shatenstein B, Ghadirian P. Nutrition and breast cancer among sporadic cases and gene mutation carriers: an overview. Cancer Detect Prev 2008;32:52-64.

31. Parodi PW. Dairy product consumption and the risk of breast cancer. J Am Coll Nutr 2005;24:556S-68S.

32. Al Sarakbi W, Salhab M, Mokbel K. Dairy products and breast cancer risk: a review of the literature. Int J Fertil Womens Med 2005;50:244-9.

33. Abbas S, Linseisen J, Chang-Claude J. Dietary vitamin d and calcium intake and premenopausal breast cancer risk in a german case-control study. Nutr Cancer 2007;59:54-61.

34. Shin MH, Holmes MD, Hankinson SE, Wu K, Colditz GA, Willett WC. Intake of dairy products, calcium, and vitamin d and risk of breast cancer. J Natl Cancer Inst 2002;94:1301-11.

35. Lin J, Manson JE, Lee IM, Cook NR, Buring JE, Zhang SM. Intakes of calcium and vitamin d and breast cancer risk in women. Arch Intern Med 2007;167:1050-9.

36. Knight JA, Lesosky M, Barnett H, Raboud JM, Vieth R. Vitamin D and reduced risk of breast cancer: a population-based case-control study. Cancer Epidemiol Biomarkers Prev 2007;16:422-9.

37. Rossi M, McLaughlin JK, Lagiou P, et al. Vitamin D intake and breast cancer risk: a case-control study in Italy. Ann Oncol 2009;20:374-8.

85

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

38. Cotterchio M, Boucher BA, Kreiger N, Mills CA, Thompson LU. Dietary phytoestrogen intake--lignans and isoflavones--and breast cancer risk (Canada). Cancer Causes Control 2008;19:259-72.

39. Hall S, Schulze K, Groome P, Mackillop W, Holowaty E. Using cancer registry data for survival studies: the example of the Ontario Cancer Registry. J Clin Epidemiol 2006;59:67-76.

40. Marrett LD, Clarke EA, Hatcher J, Weir HK. Epidemiologic research using the Ontario Cancer Registry. Can J Public Health 1986;77 Suppl 1:79-85.

41. Boucher B, Cotterchio M, Kreiger N, Nadalin V, Block T, Block G. Validity and reliability of the Block98 food-frequency questionnaire in a sample of Canadian women. Public Health Nutr 2006;9:84-93.

42. Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, Gardner L. A data-based approach to diet questionnaire design and testing. Am J Epidemiol 1986;124:453-69.

43. Block G. Invited commentary: Another perspective on food frequency questionnaires. Am J Epidemiol 2001;154:1103,4; discussion 1105-6.

44. Maldonado G, Greenland S. Simulation study of confounder-selection strategies. Am J Epidemiol 1993;138:923–936.

45. John EM, Schwartz GG, Dreon DM, 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 1999;8:399-406.

46. McCullough ML, Rodriguez C, Diver WR, et al. Dairy, calcium, and vitamin D intake and postmenopausal breast cancer risk in the Cancer Prevention Study II Nutrition Cohort. Cancer Epidemiol Biomarkers Prev 2005;14:2898-904.

47. Robien K, Cutler GJ, Lazovich D. Vitamin D intake and breast cancer risk in postmenopausal women: the Iowa Women's Health Study. Cancer Causes Control 2007;18:775-82.

48. Lappe JM, Travers-Gustafson D, Davies KM, Recker RR, Heaney RP. Vitamin D and calcium supplementation reduces cancer risk: results of a randomized trial. Am J Clin Nutr 2007;85:1586-91.

49. Almquist M, Manjer J, Bondeson L, Bondeson AG. Serum calcium and breast cancer risk: results from a prospective cohort study of 7,847 women. Cancer Causes Control 2007;18:595-602.

50. Blackmore KM, Lesosky M, Barnett H, Raboud JM, Vieth R, Knight JA. Vitamin D from dietary intake and sunlight exposure and the risk of hormone-receptor-defined breast cancer. Am J Epidemiol 2008;168:915-24.

51. Frazier AL, Li L, Cho E, Willett WC, Colditz GA. Adolescent diet and risk of breast cancer. Cancer Causes Control 2004;15:73-82.

52. Frazier AL, Ryan CT, Rockett H, Willett WC, Colditz GA. Adolescent diet and risk of breast cancer. Breast Cancer Res 2003;5:R59-64.

86

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

53. Webb AR, Kline L, Holick MF. Influence of season and latitude on the cutaneous synthesis of vitamin D3: exposure to winter sunlight in Boston and Edmonton will not promote vitamin D3 synthesis in human skin. J Clin Endocrinol Metab 1988;67:373-8.

54. Gozdzik A, Barta JL, Wu H, et al. Low wintertime vitamin D levels in a sample of healthy young adults of diverse ancestry living in the Toronto area: associations with vitamin D intake and skin pigmentation. BMC Public Health 2008;8:336.

55. Sahota H, Barnett H, Lesosky M, Raboud JM, Vieth R, Knight JA. Association of vitamin d related information from a telephone interview with 25-hydroxyvitamin D. Cancer Epidemiol Biomarkers Prev 2008;17:232-8.

56. van der Meer IM, Boeke AJ, Lips P, et al. Fatty fish and supplements are the greatest modifiable contributors to the serum 25-hydroxyvitamin D concentration in a multiethnic population. Clin Endocrinol 2008;68:466-72.

57. Brustad M, Alsaker E, Engelsen O, Aksnes L, Lund E. Vitamin D status of middle-aged women at 65-71 degrees N in relation to dietary intake and exposure to ultraviolet radiation. Public Health Nutr 2004;7:327-35.

58. Burgaz A, Akesson A, Oster A, Michaelsson K, Wolk A. Associations of diet, supplement use, and ultraviolet B radiation exposure with vitamin D status in Swedish women during winter. Am J Clin Nutr 2007;86:1399-404.

59. Vieth R, Cole DE, Hawker GA, Trang HM, Rubin LA. Wintertime vitamin D insufficiency is common in young Canadian women, and their vitamin D intake does not prevent it. Eur J Clin Nutr 2001;55:1091-7.

87

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

Table 1. Distribution of selected characteristics and age-group adjusted odds ratio (OR) estimates

among 3,101 breast cancer cases and 3,471 controls in the Ontario Women’s Diet and Health

study

Variable Cases Controls Model 1 1 OR (95% CI)

n (%) Age-group2 25 - 39 40 - 44 45 - 49 50 - 54 55 - 59 60 - 64 65 - 69 70 - 74

181 (6) 278 (9) 380 (12) 504 (16) 511 (16) 450 (15) 439 (14) 358 (12)

316 (9) 367 (11) 482 (14) 512 (15) 470 (14) 471 (14) 508 (15) 345 (10)

N/A

Breast cancer in a 1st degree relative No Yes

2389 (77) 635 (21)

2973 (86) 415 (12)

1.00 1.86 (1.63, 2.13)

Age at menarche ≤ 11 12 13 ≥ 14

594 (20) 753 (25) 846 (28) 782 (26)

615 (18) 823 (25) 973 (29) 939 (28)

1.00 0.95 (0.81, 1.10) 0.90 (0.80, 1.04) 0.86 (0.74, 0.99)

Age at menopause3 ≤ 45 45 - 49 ≥50 premenopausal

572 (19) 511 (17) 987 (33) 957 (32)

727 (21) 550 (16) 888 (26) 1237 (36)

1.00 1.15 (0.98, 1.36) 1.37 (1.19, 1.59) 1.15 (0.95, 1.38)

Parity Nulliparous 1 2 - 3 > 4

543 (18) 421 (14) 1677 (55) 415 (14)

404 (12) 413 (12) 2068 (61) 524 (15)

1.00 0.75 (0.62, 0.91) 0.57 (0.49, 0.66) 0.53 (0.44, 0.64)

Pack-years smoked Never smoker ≤4 5-12 13-25 ≥26

1572 (52) 360 (13) 336 (12) 388 (14) 393 (14)

1791 (52) 409 (12) 365 (11) 424 (13) 432 (14)

1.00 1.04 (0.88, 1.21) 1.09 (0.92, 1.28) 1.03 (0.88, 1.20) 0.97 (0.83, 1.13)

BMI4 kg/m2 < 24.9 25.0 - 29.9 ≥ 30

1344 (44) 1012 (33) 723 (24)

1535 (45) 1171 (34) 740 (21)

1.00 0.94 (0.84, 1.05) 1.06 (0.93, 1.20)

Moderate physical activity age 20-39 0 – 3 times per month

316 (11)

322 (10)

1.00

88

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

1– 2 times per week 3– 5 times per week >5 times per week

698 (23) 1044 (35) 913 (31)

722 (22) 1261 (38) 1021 (31)

1.00 (0.83, 1.21) 0.85 (0.71, 1.01) 0.90 (0.75, 1.08)

Moderate physical activity age 40-59 0 – 3 times per month 1– 2 times per week 3– 5 times per week >5 times per week Age not yet reached

408 (13) 654 (22) 1115 (37) 669 (22) 181 (6)

375 (11) 711 (21) 1179 (35) 800 (24) 316 (9)

1.00 0.84 (0.70, 1.00) 0.85 (0.73, 1.00) 0.75 (0.63, 0.90) NA

1 Age-group adjusted odds ratios (95% CIs) calculated using multivariate logistic regression. 2 Age at cancer diagnosis for cases and age on 15Nov2002 for controls 3 Women were classified as premenopausal if they had a menstrual period within 12 months of their diagnosis/referent date 4 BMI weight two years ago (kg) divided by height in metres squared

89

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

Table 2. Distribution of breast cancer cases (n = 3101) and controls (n = 3471) and odds ratio (OR)

estimates for intake of selected foods and supplements (frequency and duration) known to contain

vitamin D or calcium among Ontario women

Food/supplement intake (serving size)

Cases Controls Model 11 OR (95% CI)

Model 22 OR (95% CI)

n (%) Glasses of milk (244g (8.5 oz.)) < 1 per week 1 per week 2 - 3 per week 4 - 6 per week ≥ 1 per day P trend

780 (26) 447 (15) 549 (18) 657 (22) 596 (20)

879 (26) 573 (17) 605 (18) 697 (21) 640 (19)

1.00 0.91 (0.77, 1.06) 1.06 (0.91, 1.23) 1.10 (0.96, 1.28) 1.10 (0.95, 1.27) 0.17

1.00 0.94 (0.79, 1.12) 1.10 (0.93, 1.30) 1.13 (0.97, 1.34) 1.15 (0.97, 1.37) 0.04

Margarine (5g (one teaspoon))3 Never or few times per year < 0.5 per week 0.5 – 5 per week 6 – 13 per week ≥ 2 per day P trend

996 (33) 218 (7) 632 (21) 867 (29) 274 (9)

1097 (33) 264 (8) 691 (21) 975 (29) 303 (9)

1.00 0.95 (0.77, 1.16) 1.03 (0.89, 1.18) 0.96 (0.84, 1.09) 0.96 (0.79, 1.15) 0.67

1.00 0.88 (0.70, 1.10) 1.04 (0.89, 1.21) 0.96 (0.83, 1.10) 0.99 (0.80, 1.27) 0.79

Tuna (51g (1/4 cup)) Never or few times per year 1 per month 2 per month 1 per week ≥ 2 per week P trend

836 (28) 521 (17) 545 (18) 547 (18) 580 (19)

884 (26) 557 (16) 620 (18) 652 (19) 674 (20)

1.00 0.99 (0.85, 1.15) 0.94 (0.81, 1.09) 0.89 (0.77, 1.03) 0.91 (0.78, 1.05) 0.95

1.00 1.04 (0.88, 1.24) 0.96 (0.81, 1.14) 0.92 (0.77, 1.08) 0.93 (0.78, 1.10) 0.27

Other fish4 (43 g (1/4 cup)) Never or few times/year 1 per month 2 per month 1 per week ≥ 2 per week P trend

758 (25) 405 (13) 608 (20) 839 (28) 405 (13)

890 (26) 516 (15) 657 (19) 877 (26) 440 (13)

1.00 0.92 (0.78, 1.09) 1.07 (0.92, 1.24) 1.10 (0.96, 1.26) 1.05 (0.89, 1.24) 0.98

1.00 0.93 (0.78, 1.12) 1.09 (0.92, 1.29) 1.07 (0.91, 1.25) 0.96 (0.79, 1.16) 0.59

Fish type consumed most often Did not eat fish or missing White (e.g., haddock, cod) Fatty (e.g., salmon) P trend

523 (17) 1713 (55) 865 (28)

624 (18) 1865 (54) 982 (28)

1.00 1.08 (0.94, 1.23) 1.02 (0.88, 1.19) 0.93

1.00 1.13 (0.97, 1.33) 1.05 (0.88, 1.25) 0.88 continued…

90

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

Frequency of vitamin D supplement use5 Did not use or <1 per month ≥1 per month to ≤6 per week Every day P trend

2701 (87) 114 (4) 286 (9)

2984 (86) 137 (4) 350 (10)

1.00 0.89 (0.69, 1.15) 0.84 (0.71, 1.00) 0.03

1.00 0.85 (0.64, 1.14) 0.84 (0.70, 1.01) 0.04

Duration of vitamin D supplement use Did not use ≤ 2 years 3 -9 years ≥ 10 years P trend

2730 (88) 123 (4) 162 (5) 86 (3)

3022 (87) 135 (4) 201 (6) 113 (3)

1.00 0.98 (0.76, 1.26) 0.83 (0.67, 1.03) 0.78 (0.59, 1.04) 0.03

1.00 0.93 (0.70, 1.23) 0.85 (0.67, 1.08) 0.83 (0.61, 1.14) 0.08

Frequency of calcium supplement use6 Did not use or <1 per month ≥1 per month to ≤6 per week Every day P trend

2064 (67) 247 (8) 790 (25)

2269 (65) 298 (9) 904 (26)

1.00 0.89 (0.74, 1.07) 0.88 (0.78, 0.99) 0.02

1.00 0.91 (0.74, 1.11) 0.88 (0.77, 1.00) 0.04

Duration of calcium supplement use Did not use ≤ 2 years 3 - 9 years ≥ 10 years P trend

2182 (70) 248 (8) 457 (15) 214 (7)

2400 (69) 332 (10) 473 (14) 266 (8)

1.00 0.80 (0.67, 0.95) 0.98 (0.85, 1.14) 0.81 (0.67, 0.98) 0.05

1.00 0.75 (0.62, 0.92) 1.00 (0.85, 1.17) 0.84 (0.68, 1.04) 0.14

Frequency of multivitamin use Did not use or <1 per month ≥1 per month to ≤6 per week Every day P trend

1930 (62) 343 (11) 828 (27)

2179 (63) 380 (11) 912 (26)

1.00 1.05 (0.90, 1.23) 1.00 (0.89, 1.11) 0.99

1.00 1.10 (0.92, 1.31) 0.99 (0.87, 1.12) 0.95

Duration of multivitamin use Did not use ≤ 2 years 3 - 9 years ≥ 10 years P trend

2023 (65) 297 (10) 444 (14) 337 (11)

2256 (65) 320 (9) 539 (16) 356 (10)

1.00 1.05 (0.89, 1.25) 0.91 (0.79, 1.05) 1.01 (0.86, 1.19) 0.57

1.00 0.99 (0.82, 1.19) 0.95 (0.81, 1.11) 0.99 (0.83, 1.18) 0.66

1 Age group adjusted odds ratios (95% CIs) calculated using multivariate logistic regression (note: 39 variables were evaluated as potential confounders and none were identified as confounders i.e., their inclusion in the model did not change the OR>10%).

91

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

2 Adjusted for age group, education, age at menarche, age at first live birth, parity, menopausal status, breast cancer in first degree relative, total energy intake (kcal), BMI, smoking (packyears), moderate physical activity during ages 20-39, moderate physical activity during ages 40-59, time spent outdoors per week during age 20-39, and time spent outdoors per week during ages 40-59. Odds ratios (95% CIs) calculated using multivariate logistic regression. 3 Intake of margarine (not butter) on foods such as bread or vegetables 4 Not fried fish 5 Vitamin D as a single product supplement or cod liver oil 6 Calcium as a single vitamin or combined

92

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

Table 3. Distribution of breast cancer cases (n = 3101) and controls (n = 3471) and odds ratio (OR)

estimates for derived vitamin D and calcium nutrient intake from food, supplements and total

combined among Ontario women

Vitamin D or calcium intake Cases Controls Model 11

Model 22

n (%)

Total combined vitamin D3,4 (µg/day) < 2.5 2.5 – 4.9 5.0 – 9.9 10.0 – 14.9 ≥ 15.0 P trend

384 (13) 607 (20) 731 (24) 708 (23) 632 (21)

443 (13) 702 (20) 831 (24) 734 (21) 717 (21)

1.00 1.02 (0.86, 1.21) 1.03 (0.87, 1.22) 1.10 (0.87, 1.22) 0.99 (0.83, 1.18) 0.96

1.00 1.01 (0.82, 1.25) 1.01 (0.81, 1.26) 1.10 (0.88, 1.37) 0.99 (0.78, 1.26) 0.87

Vitamin D from foods (µg/day) < 2.5 2.5 – 4.9 5.0 – 9.9 ≥ 10.0 P trend

638 (21) 1036 (34) 1066 (35) 322 (11)

717 (21) 1182 (34) 1197 (35) 331 (10)

1.00 1.00 (0.87, 1.14) 1.01 (0.88, 1.15) 1.10 (0.91, 1.33) 0.31

1.00 0.97 (0.82, 1.14) 1.01 (0.85, 1.21) 1.13 (0.88, 1.45) 0.23

Vitamin D from all supplements5 (µg/day) 0 < 10.0 10.0 > 10.0 P trend

1679 (55) 378 (12) 847 (28) 158 (5)

1893 (55) 414 (12) 912 (27) 208 (6)

1.00 1.05 (0.90, 1.22) 1.01 (0.90, 1.13) 0.80 (0.64, 0.99) 0.17

1.00 1.08 (0.90, 1.28) 0.98 (0.85, 1.13) 0.76 (0.59, 0.98) 0.11

Total combined calcium3 (mg/day) < 500 500 - 749 750 - 999 1000 - 1499 >1500 P trend

453 (15) 539 (18) 472 (15) 698 (23) 900 (29)

501 (15) 641 (19) 504 (15) 798 (23) 983 (28)

1.00 0.94 (0.80, 1.12) 1.06 (0.88, 1.26) 0.97 (0.82, 1.14) 0.97 (0.83, 1.14) 0.81

1.00 0.99 (0.80, 1.22) 1.13 (0.90, 1.43) 1.06 (0.85, 1.33) 1.03 (0.82, 1.30) 0.95

Calcium from foods (mg/day) < 500 500 - 749 750 - 999 >1000 P trend

771 (25) 803 (26) 627 (20) 861 (28)

833 (24) 972 (28) 695 (20) 927 (27)

1.00 0.90 (0.79, 1.03) 0.99 (0.86, 1.15) 1.03 (0.90, 1.18) 0.32

1.00 0.93(0.78, 1.10) 1.07 (0.88, 1.30) 1.17 (0.95, 1.45) 0.05

93

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

1 Age group adjusted odds ratios (95% CIs) calculated using multivariate logistic regression. (note: 39 variables were evaluated as potential confounders and none were identified as confounders i.e., their inclusion in the model did not change the OR>10%) (N = 6489). 2 Adjusted for age group, education, age at menarche, age at first live birth, parity, menopausal status, breast cancer in first degree relative, total energy intake (kcal), BMI, smoking (packyears), moderate physical activity during ages 20-39, moderate physical activity during ages 40-59, time spent outdoors per week during age 20-39, time spent outdoors per week during ages 40-59, total calcium intake (included for vitamin D models only) and total vitamin D intake (included for calcium models only). Odds ratios (95% CIs) calculated using multivariate logistic regression (N = 5489). 3 From food and supplements 4 Vitamin D: 10µg = 400 IU 5 Multivitamins and vitamin D single product supplements or cod liver oil. 6 Multivitamins and calcium supplements

Calcium from all supplements6 (mg/day) 0 < 1000 1000 > 1000 P trend

1435 (47) 837 (27) 387 (13) 403 (13)

1612 (47) 911 (27) 433 (13) 471 (14)

1.00 1.03 (0.91, 1.16) 0.92 (0.78, 1.07) 0.88 (0.76, 1.03) 0.05

1.00 0.97 (0.82, 1.14) 0.86 (0.72, 1.03) 0.85 (0.68, 1.05) 0.04

94

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

Table 4. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for vitamin D

intake variables stratified by total calcium intake

Total combined calcium <1000 (mg/day)

Total combined calcium ≥1000 (mg/day)

Vitamin D (µg/day)

Cases n = 1464

Controls n = 1646

Model 11 Cases n=1598

Controls n = 1781

Model 11 P 2

n (%) n (%) Total combined3,4 < 2.5 2.5 – 4.9 5.0 – 9.9 10.0 – 14.9 ≥ 15.0

321 (22) 473 (32) 360 (25) 229 (16) 81 (6)

354 (22) 550 (33) 421 (26) 243 (15) 78 (5)

1.00 0.98 (0.81, 1.19) 0.96 (0.78, 1.19) 1.02 (0.81, 1.30) 1.08 (0.76, 1.53)

63 (4) 134 (8) 371 (23) 479 (30) 551 (34)

89 (5) 152 (9) 410 (23) 491 (28) 639 (36)

1.00 1.25 (0.84, 1.87) 1.30 (0.91, 1.85) 1.38 (0.98, 1.96) 1.22 (0.87, 1.72)

0.49

Foods < 2.5 2.5 – 4.9 5.0 – 9.9 ≥ 10.0

461 (31) 639 (44) 348 (24) 16 (1)

481 (29) 743 (45) 407 (25) 15 (1)

1.00 0.92 (0.78, 1.08) 0.89 (0.73, 1.08) 1.11 (0.53, 2.27)

177 (11) 397 (25) 718 (45) 306 (19)

236 (13) 439 (25) 790 (44) 316 (18)

1.00 1.21 (0.95, 1.54) 1.24 (1.00, 1.55) 1.33 (1.04, 1.71)

0.13

Supplements5 0 < 10.0 10.0 > 10.0

1032(70) 178 (12) 233 (16) 21 (1)

1193(72) 190 (12) 238 (15) 25 (2)

1.00 1.14 (0.91, 1.43) 1.08 (0.88, 1.32) 0.87 (0.48, 1.57)

647 (40) 200 (13) 608 (38) 143 (9)

700 (39) 224 (13) 666 (37) 191 (11)

1.00 0.97 (0.78, 1.20) 0.97 (0.83, 1.13) 0.79 (0.62, 1.01)

0.64

1 Age group adjusted odds ratios (95% CI) calculated using multivariate logistic regression. 2 Likelihood ratio test for interactions 3 From food & supplements 4 Vitamin D: 10µg = 400 IU 5 Multivitamins and vitamin D single product supplements or cod liver oil.

95

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

Table 5. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for vitamin D

and calcium variables stratified by menopausal status among Ontario women

Premenopausal Postmenopausal Variable Cases

n = 948 Controls n =1226

Model 11 Cases n =2111

Controls n =2196

Model 11 P2

n (%) n (%)

Vitamin D

(µg/day)

Total combined3,4 < 2.5 2.5 – 4.9 5.0 – 9.9 10.0 – 14.9 ≥ 15.0 P trend

112 (12) 217 (23) 260 (27) 208 (22) 154 (16)

166 (13) 301 (25) 351 (29) 229 (19) 184 (15)

1.00 1.09 (0.81, 1.47) 1.10 (0.83, 1.47) 1.35 (0.99, 1.84) 1.23 (0.89, 1.70) 0.07

272 (13) 390 (18) 472 (22) 502 (24) 478 (23)

277 (13) 403 (18) 482 (22) 505 (23) 534 (24)

1.00 1.00 (0.80, 1.24) 1.00 (0.81, 1.24) 1.01 (0.82, 1.24) 0.90 (0.73, 1.11) 0.26

0.41

Foods 0 < 10.0 10.0 > 10.0 P trend

188 (20) 325 (34) 332 (35) 106 (11)

240 (20) 452 (37) 421 (34) 118 (10)

1.00 0.92 (0.73, 1.17) 1.00 (0.79, 1.27) 1.16 (0.84, 1.61) 0.11

451 (21) 711 (34) 736 (35) 216 (10)

477 (22) 732 (33) 778 (35) 214 (10)

1.00 1.03 (0.88, 1.22) 1.01 (0.86, 1.19) 1.07 (0.85, 1.34) 0.91

0.66

Supplements5 0 < 10.0 10.0 > 10.0 P trend

562 (59) 158 (17) 202 (21) 29 (3)

773 (63) 178 (15) 245 (20) 35 (3)

1.00 1.24 (0.97, 1.58) 1.12 (0.90, 1.39) 1.07 (0.64, 1.78) 0.38

1117 (53) 222 (11) 640 (30) 135 (6)

1124 (51) 237 (11) 659 (30) 181 (8)

1.00 0.94 (0.77, 1.16) 0.96 (0.84, 1.10) 0.74 (0.58, 0.94) 0.04

0.24

Calcium

(mg/day)

Total combined3 < 500 500 - 749 750 - 999 1000 - 1499 >1500 P trend

133 (14) 191 (20) 180 (19) 233 (25) 214 (23)

190 (15) 277 (23) 217 (18) 318 (26) 229 (19)

1.00 1.00 (0.75, 1.34) 1.19 (0.88, 1.60) 1.05 (0.79, 1.38) 1.32 (0.99, 1.77) 0.04

320 (15) 348 (16) 293 (14) 466 (22) 687 (33)

312 (14) 366 (17) 287 (13) 480 (22) 756 (34)

1.00 0.92 (0.74, 1.14) 1.00 (0.79, 1.25) 0.94 (0.77, 1.15) 0.87 (0.72, 1.05) 0.13

0.13

Foods < 500 500 - 749 750 - 999 >1000 P trend

201 (21) 232 (24) 206 (22) 312 (33)

250 (20) 361 (29) 252 (20) 368 (30)

1.00 0.80 (0.62, 1.02) 1.01 (0.78, 1.31) 1.06 (0.84, 1.35) 0.15

570 (27) 572 (27) 422 (20) 550 (26)

585 (27) 612 (28) 443 (20) 561 (25)

1.00 0.96 (0.81, 1.12) 0.98 (0.81, 1.17) 1.01 (0.85, 1.19) 0.84

0.36

96

Paper 2: Vitamin D, Calcium and Breast Cancer Risk (Anderson et al, 2010)

Supplements6 0 < 1000 1000 > 1000 P trend

527 (55) 309 (32) 48 (5) 67 (7)

709 (58) 383 (31) 60 (5) 79 (6)

1.00 1.09 (0.90, 1.32) 1.03 (0.69, 1.54) 1.09 (0.77, 1.55) 0.68

908 (43) 530 (25) 340 (16) 336 (16)

906 (41) 529 (24) 373 (17) 393 (18)

1.00 0.99 (0.85, 1.15) 0.89 (0.75, 1.06) 0.83 (0.70, 0.99) 0.02

0.68

1 Age-group adjusted odds ratios (95% CI) calculated using multivariate logistic regression

2 Likelihood ratio test for interaction (age-adjusted model only)

3 From food & supplements 4 Vitamin D: 10µg = 400 IU 5 Multivitamins and vitamin D single product supplements or cod liver oil 6 Multivitamins and calcium supplements

97

Paper 3: Sunlight and Breast Cancer Risk

4.4 Paper 3: Ultraviolet Sunlight Exposure and Breast Cancer Risk: A Population Based Case-Control Study in Ontario

Not yet submitted for publication.

Laura N. Anderson, Michelle Cotterchio, Victoria A. Kirsh, and Julia A. Knight

Short title: Sunlight and breast cancer risk

ABSTRACT

Recent studies suggest vitamin D intake may be associated with reduced breast cancer risk, but

most studies have evaluated only dietary vitamin D intake. The associations between ultraviolet

(UV) radiation from sunlight, factors related to cutaneous vitamin D production and breast

cancer risk were evaluated in a population-based case-control study among Ontario women.

Exposure was assessed during 4 periods of life, including adolescence, via mailed lifestyle and

food frequency questionnaires for all cases (n=3,101) and controls (n=3,471). Multivariate

logistic regression was used to estimate odds ratios (OR) and 95% confidence intervals (CI).

Time spent outdoors from age 40 to 59 was associated with reduced breast cancer risk (>21

versus ≤6 hours/week: OR = 0.74, 95% CI: 0.61, 0.88), and significant inverse associations were

also observed for exposure during 3 other periods of life. Sun protection practices and UV

radiation of residence were not associated with breast cancer risk. A combined solar vitamin D

score, including all variables related to vitamin D production, was significantly associated with

reduced breast cancer risk. These associations were not confounded or modified by menopausal

status, dietary vitamin D intake or physical activity. This study suggests that factors related to

increased cutaneous production of vitamin D are associated with reduced breast cancer risk.

Keywords: sunlight; vitamin D; breast neoplasms; case-control studies

98

Paper 3: Sunlight and Breast Cancer Risk

It has been hypothesized that vitamin D, a potentially modifiable factor, is associated with

reduced risk of multiple types of cancer including breast cancer (1, 2). Vitamin D is produced in

the skin through the conversion of 7-dehydrocholesterol to previtamin D3 following sufficient

exposure to ultraviolet (UV) B radiation from sunlight. This process is dependent upon many

extrinsic factors that affect UVB radiation (e.g., geographic location, time of day and season) and

intrinsic, person-specific, factors (e.g., time spent outdoors, sun protection practices, skin color)

(3, 4). Vitamin D is also found in supplements and few foods (e.g., fatty fish, fortified milk) (5).

Vitamin D from sun, diet and supplements undergoes hydroxylation in liver to the circulating

form 25 hydroxyvitamin D [25(OH)D]. Breast cells, among other cells in the body, are capable

of locally converting 25(OH)D to the active hormone 1,25dihydroxyvitamin D [1,25(OH)2D]

which has been shown in laboratory studies to have anti-cancer properties (6-8).

Ecologic studies have shown latitude (inversely correlated with sun exposure) is positively

associated and UVB irradiance is negatively associated with breast cancer incidence (9) or

mortality (10-12). Many observational studies of diet or supplement intakes (13-20) have found

inverse associations between vitamin D intake and breast cancer risk although often among

specific subgroups only. Some studies have found serum 25(OH)D levels are associated with

reduced breast cancer risk (21-25), but not all (26-28).

Despite the ability of vitamin D to be produced in the skin following sun exposure, fewer studies

have evaluated the associations between sun exposure related variables and breast cancer risk

(13, 19, 29-31). Time spent outdoors has been inversely associated with breast cancer risk in

most (13, 19, 30) but not all studies (29). Sunscreen, sunburns/skin damage, winter sun trips or

sunlamp/solarium use have generally not been associated with breast cancer risk (13, 19, 31),

although in one study limb coverage was associated with increased risk (19). Studies of UV or

solar radiation and breast cancer risk in the US are inconsistent (13, 30). No previous breast

cancer studies have created one measure of vitamin D from sunlight that combines person-

specific factors (e.g., time outdoors, skin color, or sun protection practices) and environmental

sun exposure. Challenges to the measurement of vitamin D in population-based studies has been

reviewed (32).

The objective of this study was to evaluate the associations between breast cancer risk and

variables related to the production of vitamin D from sunlight (time spent outdoors, ultraviolet

99

Paper 3: Sunlight and Breast Cancer Risk

radiation at residence, skin color and sun protection practices) measured at different 4 periods of

life in a large, population-based case-control study of women in Ontario, Canada. In addition to

evaluating the contributions of each vitamin D related variable, a solar vitamin D score that

combines all variables was derived and evaluated in relation to breast cancer risk. Secondary

objectives of this study are to evaluate the combined effects of vitamin D exposure from sunlight

and intake from food and supplements, and to evaluate a predictive measure of serum 25(OH)D.

MATERIALS AND METHODS

Study design

A population based case-control study was conducted among women living in Ontario, Canada,

as previously described (33). Breast cancer cases were identified from the Ontario Cancer

Registry between 2002 and 2003. Cases were women between the ages of 25-74 with a first

pathologically confirmed cancer of the breast. Of the 4,109 cases with physician consent for

contact 3,101 completed the study (75% response rate). Controls were recruited by random digit

dialing of households in Ontario and frequency age-matched 1:1 to cases. Of 4,352 households

where an eligible control was identified 3,471 women completed the study (80% response rate).

Ethics approval for this study was obtained from the University of Toronto Research Ethics

Board.

Data collection and exposure variables

Study participants completed a 20-page mailed self-administered risk factor questionnaire and a

modified Block 1998 Food Frequency Questionnaire. Ethnicity was used as a proxy for skin

color. The overwhelming majority (90%) of study participants were Caucasian ethnicity (proxy

for lighter skin color) and thus skin color was categorized as Caucasian versus non-Caucasian.

Variables related to sun exposure (weekday time outdoors, weekend time outdoors, sun

protection and location of residence) were measured at four periods of life: teenage years, 20-

30s, 40s-50s and 60s-74.

To capture time spent outdoors, participants were asked “On a typical weekday in the months

April –October about how many hours per day did you spend outside?” response options

included: less than one hour, 1 to 2 hours, 3 to 4 hours, more than 4 hours. The same question

and response options were repeated for weekend (Saturday and Sunday) exposure. Only

100

Paper 3: Sunlight and Breast Cancer Risk

summertime sun exposure data were collected since wintertime sun exposure in Ontario is not

sufficient for the production of vitamin D (34). The variable “Hours outdoors per week” was

created by weighting and summing weekday and weekend exposures. To measure sun protection

practices all participants were asked “When in the sun did you wear sunscreen or protective

clothing, such as long sleeves, etc.?” response options included: never, sometimes, and always.

To obtain location of residence respondents were asked to report where they lived during 4

specific age periods (a full residential history including duration was not collected).

All women resided in Ontario when they participated in the study but many lived outside the

province at earlier periods of life. Latitude and longitude were obtained for all cities and

provinces/states of residence at the 4 time periods from www.geocoder.ca . There were 1628

(25%) study participants who reported only country or province/state of residence during at least

one period of life; these participants were assigned the coordinates of the most populated city in

their country or region. When multiple locations were reported for a given time period only the

first location was used. There were 86 (1%) participants who lived in the southern hemisphere

during at least one life period; for these women the reporting period would have corresponded to

wintertime sun exposure. The analysis was repeated with these women excluded and the results

did not change.

Latitude and longitude were used to obtain UV radiation data from National Aeronautics and

Space Administration’s (NASA) Total Ozone Mapping Spectrometer (TOMS) (35). Ground

level UV irradiance data is calculated from TOMS onboard spacecraft instrument measures of

atmospheric UV, total ozone, surface reflectivity and cloud cover. Monthly average noon-time

erythemal UV for June 2003 was selected for use in this study. These data are weighted using the

McKinlay-Diffey erythemal action spectra (36) which weights radiation in the UVA (315-400

nm) and UVB (280-315 nm) wavelengths based on the time required to induce erythema (skin

reddening); shorter rays are more likely to induce erythema. Cutaneous vitamin D is dependent

on only UVB exposure and there is a vitamin D-specific action spectra based on human skin’s

ability to produce previtamin D3 (37), but vitamin D weighted UV is not currently available from

TOMS. Although vitamin D production does not always directly correspond with erythemal UV

estimates (38), the erythemal action spectra closely approximates the vitamin D action spectra in

summer north of 42⁰ (Ontario, Canada) (39, 40).

101

Paper 3: Sunlight and Breast Cancer Risk

Derivation of solar vitamin D score

To derive a solar vitamin D score (i.e., one variable that takes into consideration known

determinants of cutaneous vitamin D production) an algorithm was created based on the

available literature on determinants of cutaneous production of vitamin D. Weighted proportions

based on the knowledge that darker skin color and use of sun protection practices limits vitamin

D production were applied to “UV hours per week” (the product of erythemal UV and weekly

time spent outdoors). This algorithm (shown below) was applied to each of the four age periods

of exposure. Sensitivity analyses were conducted to evaluate other plausible values.

Solar vitamin D score (mW/m2 • hrs) = UV hours per week (mW/m2 • hrs) x skin color weight x sun protection weight

Where:

UV hours/week = Erythemal UV radiation (mW/m2) x Time spent outdoors (hrs/ week); skin color weight = 1 if ethnicity = Caucasian; skin color weight = 1/3 if ethnicity = non-Caucasian; sun protection weight = 1 if sun protection use = never; sun protection weight = 2/3 if sun protection use = sometimes; sun protection weight = 1/3 if sun protection use= always.

It has been estimated that people with highly pigmented (darker) skin colors in comparison to

lighter require at least 3 times the amount of sunlight to produce equivalent vitamin D (41)(4),

although estimates range from upward to 5 to 10 times (42). Thus, an accommodation factor of

one third was chosen to weight UV production for non-Caucasian individuals. In regards to sun

protection practices, sunscreen and clothing both have the potential to block all vitamin D

production. However, it is unlikely that women apply a complete application of sunscreen (i.e., a

thorough application to all locations of the body prior to going outdoors with frequent

reapplication) (as reviewed by (43)) or fully cover-up with clothing. Sunscreen use does not

predict 25(OH)D levels (44, 45), but coverage of arms and legs does significantly predict lower

25(OH)D levels (44). Therefore, within this population it was assumed that even individuals who

report “always” using sun protection are still accessing one third of the available vitamin D

generating UV light, in comparison to participants who reported “never” using sunscreen or

protective clothing. Correspondingly, a lesser decrease was estimated for participants who

102

Paper 3: Sunlight and Breast Cancer Risk

responded “sometimes protected” hence a weighting factor of two thirds was assigned to these

respondents.

In addition to our proposed algorithm, another combined measure was created using cross-

classification. UV exposure, time outdoors, sun protection practices and skin color were

classified as high versus low (1 vs 0) based on vitamin D production potential and a combined

additive score was created. Also, we evaluated the use of a previously published algorithm for

the measurement of predicted serum 25(OH)D values among men in the Health Professionals’

Follow-Up Study (46). This algorithm was created using multiple linear regression analysis to

develop a predictive model among a subset of the sample with serum 25(OH)D measures and

included: dietary vitamin D, supplemental vitamin D, BMI, race, physical activity (included as a

proxy for time spent outdoors), and region of residence. The model was then applied to predict

25(OH)D levels among men with no serum data in the Health Professionals’ Study and has also

been previously applied to women in the Nurses’ Health Study (47, 48).

Statistical analysis

Age-adjusted odds ratios (OR) and 95% confidence intervals (CI) were calculated using

unconditional logistic regression. Statistical analysis was conducted using SAS 9.1. All tests

were 2-sided and statistical significance was defined as P < 0.05. Test for linear trend was

calculated by treating the median intake for each exposure category as a continuous variable in

the age-adjusted models. Each variable contributing to vitamin D from sun (i.e., skin color, time

spent outdoors, sun protection and UV) was assessed on its own in addition to the derived solar

vitamin D score. ORs were calculated for each of the four age periods of exposure and a

cumulative measure of lifetime UV exposure was developed by combining the solar vitamin D

score from all periods of life.

Potential confounders were evaluated using the change in OR >10% method. The following

variables were tested as potential confounders: marital status, education, ethnicity, body mass

index (BMI), smoking status, pack years smoked, breastfed, lactation, age at menarche, oral

contraceptive use, oral contraceptive duration, parity, age at first live birth, age at last

menstruation, duration of hormone replacement therapy use (postmenopausal women only),

history of benign breast disease, family history of breast cancer, screening mammogram,

alcoholic drinks, dietary fat intake, calorie intake, phytoestrogen intake, physical activity

103

Paper 3: Sunlight and Breast Cancer Risk

(strenuous, moderate and daily activity) at selected periods of life (teenage years, 20-30s, 40-50s

and 60s-74) and both dietary vitamin D and calcium (from food and supplements). It was

hypothesized a priori that the effect of vitamin D from sunlight may be modified by menopausal

status (pre- versus post), vitamin D or calcium intakes from supplements or total intake (from

diet, and all supplements – including multivitamins ), BMI or smoking status, thus the statistical

significance of these multiplicative interactions was tested using the likelihood ratio test.

RESULTS

The mean age of women in this study was 56 years and many had postsecondary education (46%

of cases and 49% of controls). Breast cancer cases were more likely than controls to have a

family history of breast cancer, younger age at menarche, and lower parity. Among

postmenopausal women (68% of cases and 64% of controls), age at menopause was positively

associated with breast cancer risk. Strenuous physical activity during 20s-30s and 40s-50s was

associated with reduced breast cancer risk. Measures of daily and moderate physical activity

were also associated with reduced breast cancer risk (33). As reported elsewhere, supplemental

vitamin D intake at a dose level of at least 400 IU/day was associated with decreased breast

cancer risk; vitamin D from foods was not associated with breast cancer risk (49).

Non-Caucasian ethnicity was associated with a higher risk of breast cancer than Caucasian

(OR=1.23, 95% CI, 1.05, 1.45) (Table 1). Increasing time spent outdoors (highest vs. lowest

category) was associated with a decreased risk of breast cancer during teenage years (OR = 0.71,

95% CI: 0.60, 0.81), 20s-30s (OR = 0.64, 95% CI: 0.53, 0.76), 40s-50s (OR = 0.74, 95% CI:

0.61, 0.88) and 60s-74 (OR = 0.50, 95% CI: 0.37, 0.66), all with statistically significant trends.

Time spent outdoors was not associated with parity or education (appendix 4, table 3) and there

were no significant interactions between time spent outdoors at any period of exposure and parity

or education and breast cancer risk (data not shown). Sun protection practices, latitude and

erythemal UV of location resided were not associated with breast cancer risk during any of the 4

age periods. There were 86 (1%) participants who lived in the Southern hemisphere during at

least one life period for whom the reporting period would have corresponded to wintertime sun

exposure; the analyses were repeated with these women excluded and the results did not change.

Excluding women who only reported their country of province/state of residence did not

substantially change the results.

104

Paper 3: Sunlight and Breast Cancer Risk

No confounders were identified for any of the models, thus, the age-adjusted only models are

presented. The solar vitamin D score was consistently associated with a reduced risk of breast

cancer across all four age periods of exposure (Table 2). The age adjusted ORs comparing the

highest to lowest quartile for exposure during teenage years, 20s-30s. 40s-50s, and 60s-74, were

0.79 (95% CI: 0.68-0.91), 0.76 (95% CI: 0.65-0.89), 0.75 (95% CI: 0.64-0.88), and 0.59 (95%

CI: 0.46-0.76), respectively, and all trend tests were significant (p <0.001) (Table 2). Similar

inverse associations were also observed for the cumulative and recent measures of exposure. The

solar vitamin D scores for each age period of exposure had moderate to high correlations

(Spearman correlation coefficients ranged from 0.37 to 0.61, all p-values <0.0001) suggesting

that it may not be appropriate to include all 4 in one model due to muticollinearity.

Sensitivity analyses were conducted varying the assumptions for the vitamin D score and the

results changed minimally (appendix 4, table 5). Furthermore when the models are restricted to

lifelong residence in Canada the results were essentially unchanged (appendix 4, table 6). The

measure of high versus low vitamin D production potential created using cross-classification also

yielded inverse associations at all 4 age periods of exposure (appendix 4, table 4), although the

ORs for the highest vs. lowest categories were not all statistically significant, statistically

significant trends were observed suggesting a potential inverse dose-response relationship.

Significant interactions were not found between the derived solar vitamin D score at each of the

4 age periods and total dietary vitamin D intake, menopausal status or smoking status (data not

shown). Significant interactions were observed between the solar vitamin D score at ages 60s-74

(but not for earlier age periods) and both calcium intake and BMI, such that the solar vitamin D

score was associated with a significantly reduced breast cancer risk only among women with

total calcium intake ≥ 1000 mg/day (versus < 1000 mg/day) or BMI >25 (versus <25) (data not

shown). Although no significant interactions were observed between the solar vitamin D scores

and total vitamin D intake, the interactions between vitamin D supplement use (any versus none)

and the solar vitamin D score during 20s-30s and 40s-50s were statistically significant (Table 3).

But, stratified analyses did not reveal large differences and all OR estimates remained <1.0 (i.e.,

no qualitative interactions) (Table 3). The results of a combined solar vitamin D and vitamin D

from supplement score created using cross-classification are not substantially different than the

results for the solar vitamin D score only (Table 4).

105

Paper 3: Sunlight and Breast Cancer Risk

Table 6 shows the factors and corresponding beta coefficients obtained from multiple linear

regression in the Health Professionals’ Study and the variables used and assumptions made to

apply this model to our data in the Ontario Women’s Diet and Health Study. Applying the Health

Professionals Study algorithm to predict 25(OH)D levels in our study yielded a range of

predicted values from 38.8 to 86.8 nmol/L which were very similar to the ranges reported in the

Health Professionals Study and Nurses’ Health Study. A significant inverse association was

observed between predicted 25(OH)D and breast cancer risk (comparing the highest to lowest

quintile of predicted 25(OH)D: OR= 0.84; 95% CI:0.72, 0.98) (Table 6). When we used time

spent outdoors instead of physical activity to calculate predicted 25(OH)D, the range was 39 to

87 nmol/L (5th, 95th = 53, 79). The ORs obtained for the association between predicted 25(OH)D

with time spent outdoors and breast cancer risk were also statistically significant (OR=0.78; 95%

CI: 0.67, 0.91 comparing the highest to lowest quintile of predicted 25(OH)D) (data not shown).

DISCUSSION

The results of this large population-based case control study suggest that time spent outdoors and

our derived proxy measure of vitamin D from sun are inversely associated with breast cancer

risk. Exposure during all 4 periods of life, cumulative life exposure and recent exposure were all

associated with reduced breast cancer, with the strongest inverse associations observed for

exposure during the 60s-74, among the women who had reached this age. We did not find that

erythemal UV radiation, latitude, or sun protection practices were independently associated with

breast cancer risk. However, our measures of skin color and sun protection practices were

relatively crude. And despite a large proportion of study participants’ living outside Canada

during their teenage years there was limited variation in latitude of residence and thus also

limited variation in erythemal UV. The majority of study participants resided in the Greater

Toronto Area (~ 43ºN latitude). Our study did not find that these associations were confounded

by any known breast cancer risk factors or other variables that may be associated with time spent

outdoors (e.g., physical activity or smoking status).

Previous studies evaluating vitamin D from sunlight and breast cancer risk have evaluated

various factors that affect vitamin D production independently or using multivariate analysis to

control for each factor (13, 19, 29-31) or stratified by skin pigment (29), skin type and ethnicity

(30), intensity of outdoor physical activity(19), or solar radiation (13). These studies are

106

Paper 3: Sunlight and Breast Cancer Risk

consistent with the findings for our individual components, time spent outdoors was significantly

associated with reduced breast cancer risk (13, 19, 29-31) but no associations have been

observed for variables related to sun burns, skin damage or sun protection (13), although clothing

was consistently associated with increased breast cancer risk (19). Increasing sun exposure

index, from skin reflectometry measures, was found to be inversely associated with advanced

stage breast cancer risk among light skinned women only (29); we were unable to evaluate stage

of breast cancer in this study. Few studies have included measures of environmental sunlight

exposure (UV or solar radiation), and results have been inconsistent suggesting possible inverse

but non-significant associations(13) or an unexpected positive association such that women at

lower solar radiation had lower breast cancer risk (30). This finding may be explained by the

hypothesis that fluctuating serum 25(OH)D levels may be less desirable than stable levels (50).

Studies of the relationship between skin cancer, as a proxy for UV exposure, and breast cancer

risk have been inconsistent (51-54).

To the best of our knowledge no previous study has combined multiple factors to evaluate the

association between a composite measure of vitamin D from sunlight and breast cancer risk. One

previous study of all cancer mortality among men created a predictive model using serum

25(OH)D measures that were available on a sub-sample of study participants to derive a

predicted 25(OH)D measure for all study participants (46). Giovannucci et al. included dietary

and supplemental vitamin D, BMI, race, physical activity, and residence in their predicted model

and were able to explain 28% of the variation in 25(OH)D levels. We applied the previously

published predictive model to our data and observed that it was also associated with reduced

breast cancer risk. Future studies are needed to develop a predictive vitamin D model that

explains more of the variation in 25(OH)D and can be applied to population-based studies.

Timing of sunlight exposure may be important to confer any potential benefits of vitamin D for

breast cancer risk. Vitamin D exposure during adolescence may be most important because

breast tissue is undifferentiated prior to first pregnancy; hence breast cells are potentially more

susceptible to exposures during the period from menarche to first birth (55). Our study results

and those of another study (of sunburns, sun vacations and solarium use) (31) do not support this

hypothesis, however, one other study did find stronger associations with measures of sun

exposure and dietary vitamin D and breast cancer risk during adolescence (19).

107

Paper 3: Sunlight and Breast Cancer Risk

A limitation of our study is that we were unable to validate our algorithm used to derive the

composite measure of vitamin D from sunlight. Future studies need to be conducted to determine

how well such a measure predicts vitamin D intake from sunlight or 25(OH)D. There is the

potential for misclassification error in this study and more detailed measures of skin type (e.g.

Fitzpatrick skin typing or reflectometry), sun protection practices and data on sun bed/lamp and

wintertime sun holidays may be beneficial. The categories for sun exposure (all >1 hour)

measured in this study are beyond that necessary for vitamin D synthesis. However, it is not

realistic to expect study participants to recall their sun exposure with such a high level of

accuracy and this measure provides us with a relative estimate of high versus low sun exposure.

Additionally, time of day outdoors has a great impact on strength of UV and ultimately vitamin

D production. In observational studies there is always the potential for residual confounding

although our results were independent of many potential confounders (including physical

activity).

The validity and reliability of other similar sun exposure questionnaires has been measured.

Previous sun exposure questionnaires, focusing on time spent outdoors, have been shown to have

fair to moderate reliability (intraclass correlation coefficients range from 0.25-0.77) for both

recent adult measures (56-59) and recall of adolescent exposures (59). In terms of validity, time

spent outdoors measured from sun exposure questionnaires has been significantly associated

with skin measures of solar exposure (59-61) and 25(OH)D (44, 59, 62-64). Strong agreement

has been found between questionnaire measures of sun exposure and both calendar (59) or

detailed face-to-face measures (57).

The strengths of this study include its population-based case-control study design with high

response rates and large sample size. This study included a detailed person-specific sun exposure

questionnaire, with information on exposure during multiple periods of life, and used

environmental data sources to estimate ambient UV irradiation for each participant. Survival bias

is likely minimal in this study as cases were recruited within 11 months of diagnosis (on average)

and 5-year relative survival is 87% among breast cancer cases in Ontario (65). Although

measurement error may be of concern in this study there is no reason to suspect this would be

differential or introduce bias; study participants were not aware of the study hypothesis and data

collection occurred prior to any current media attention regarding the vitamin D hypothesis.

108

Paper 3: Sunlight and Breast Cancer Risk

In conclusion, time spent outdoors during summer and our composite measures of vitamin D

from sunlight were consistently associated with reduced breast cancer risk at multiple ages of

exposure. Future studies are needed to determine if this association is due to vitamin D exposure.

The algorithm created in this study to derive a composite measure combining multiple variables

related to cutaneous production of vitamin D may be useful for other researchers. Future studies

are needed to validate such a measure.

109

Paper 3: Sunlight and Breast Cancer Risk

References

1. Giovannucci E. Vitamin D status and cancer incidence and mortality. Adv Exp Med Biol. 2008;624:31-42.

2. Garland CF, Garland FC, Gorham ED, et al. The role of vitamin D in cancer prevention. Am J Public Health. 2006;96(2):252-61. (doi: AJPH.2004.045260 [pii]; 10.2105/AJPH.2004.045260 [doi]).

3. Webb AR. Who, what, where and when-influences on cutaneous vitamin D synthesis. Prog Biophys Mol Biol. 2006;92(1):17-25. (doi: S0079-6107(06)00020-4 [pii]; 10.1016/j.pbiomolbio.2006.02.004 [doi]).

4. Holick MF. Photosynthesis of vitamin D in the skin: effect of environmental and life-style variables. Fed Proc. 1987;46(5):1876-82.

5. Calvo MS, Whiting SJ, Barton CN. Vitamin D fortification in the United States and Canada: current status and data needs. Am J Clin Nutr. 2004;80(6 Suppl):1710S-6S. (doi: 80/6/1710S [pii]).

6. Holick MF. Vitamin D: A millenium perspective. J Cell Biochem. 2003;88(2):296-307. (doi: 10.1002/jcb.10338 [doi]).

7. Colston KW, Hansen CM. Mechanisms implicated in the growth regulatory effects of vitamin D in breast cancer. Endocr Relat Cancer. 2002;9(1):45-59.

8. Welsh J. Vitamin D and breast cancer: insights from animal models. Am J Clin Nutr. 2004;80(6 Suppl):1721S-4S.

9. Mohr SB, Garland CF, Gorham ED, et al. Relationship between low ultraviolet B irradiance and higher breast cancer risk in 107 countries. Breast J. 2008;14(3):255-60. (doi: 10.1111/j.1524-4741.2008.00571.x).

10. Porojnicu AC, Lagunova Z, Robsahm TE, et al. Changes in risk of death from breast cancer with season and latitude : Sun exposure and breast cancer survival in Norway. Breast Cancer Res Treat. 2006. (doi: 10.1007/s10549-006-9331-8 [doi]).

11. Grant WB. An estimate of premature cancer mortality in the U.S. due to inadequate doses of solar ultraviolet-B radiation. Cancer. 2002;94(6):1867-75. (doi: 10.1002/cncr.10427 [pii]).

12. Grant WB. An ecologic study of dietary and solar ultraviolet-B links to breast carcinoma mortality rates. Cancer. 2002;94(1):272-81. (doi: 10.1002/cncr.10196 [pii]).

13. John EM, Schwartz GG, Dreon DM, et al. 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. 1999;8(5):399-406.

14. Shin MH, Holmes MD, Hankinson SE, et al. Intake of dairy products, calcium, and vitamin d and risk of breast cancer. J Natl Cancer Inst. 2002;94(17):1301-11.

15. McCullough ML, Rodriguez C, Diver WR, et al. Dairy, calcium, and vitamin D intake and postmenopausal breast cancer risk in the Cancer Prevention Study II Nutrition Cohort. Cancer Epidemiol Biomarkers Prev. 2005;14(12):2898-904. (doi: 14/12/2898 [pii]; 10.1158/1055-9965.EPI-05-0611 [doi]).

110

Paper 3: Sunlight and Breast Cancer Risk

16. Lin J, Manson JE, Lee IM, et al. Intakes of calcium and vitamin d and breast cancer risk in women. Arch Intern Med. 2007;167(10):1050-9. (doi: 167/10/1050 [pii]; 10.1001/archinte.167.10.1050 [doi]).

17. Robien K, Cutler GJ, Lazovich D. Vitamin D intake and breast cancer risk in postmenopausal women: the Iowa Women's Health Study. Cancer Causes Control. 2007;18(7):775-82. (doi: 10.1007/s10552-007-9020-x [doi]).

18. Abbas S, Linseisen J, Chang-Claude J. Dietary vitamin d and calcium intake and premenopausal breast cancer risk in a german case-control study. Nutr Cancer. 2007;59(1):54-61. (doi: 10.1080/01635580701390223).

19. Knight JA, Lesosky M, Barnett H, et al. Vitamin D and reduced risk of breast cancer: a population-based case-control study. Cancer Epidemiol Biomarkers Prev. 2007;16(3):422-9. (doi: 16/3/422 [pii]; 10.1158/1055-9965.EPI-06-0865 [doi]).

20. Rossi M, McLaughlin JK, Lagiou P, et al. Vitamin D intake and breast cancer risk: a case-control study in Italy. Ann Oncol. 2009;20(2):374-8. (doi: 10.1093/annonc/mdn550).

21. Crew KD, Gammon MD, Steck SE, et al. Association between plasma 25-hydroxyvitamin D and breast cancer risk. Cancer Prev Res (Phila Pa). 2009;2(6):598-604. (doi: 10.1158/1940-6207.CAPR-08-0138).

22. Abbas S, Chang-Claude J, Linseisen J. Plasma 25-hydroxyvitamin D and premenopausal breast cancer risk in a German case-control study. Int J Cancer. 2009;124(1):250-5. (doi: 10.1002/ijc.23904).

23. Bertone-Johnson ER, Chen WY, Holick MF, et al. Plasma 25-hydroxyvitamin D and 1,25-dihydroxyvitamin D and risk of breast cancer. Cancer Epidemiol Biomarkers Prev. 2005;14(8):1991-7. (doi: 14/8/1991 [pii]; 10.1158/1055-9965.EPI-04-0722 [doi]).

24. Lowe LC, Guy M, Mansi JL, et al. Plasma 25-hydroxy vitamin D concentrations, vitamin D receptor genotype and breast cancer risk in a UK Caucasian population. Eur J Cancer. 2005;41(8):1164-9. (doi: 10.1016/j.ejca.2005.01.017).

25. Rejnmark L, Tietze A, Vestergaard P, et al. Reduced prediagnostic 25-hydroxyvitamin D levels in women with breast cancer: a nested case-control study. Cancer Epidemiol Biomarkers Prev. 2009;18(10):2655-60. (doi: 10.1158/1055-9965.EPI-09-0531).

26. Chlebowski RT, Johnson KC, Kooperberg C, et al. Calcium plus vitamin D supplementation and the risk of breast cancer. J Natl Cancer Inst. 2008;100(22):1581-91. (doi: 10.1093/jnci/djn360).

27. Freedman DM, Chang SC, Falk RT, et al. Serum levels of vitamin D metabolites and breast cancer risk in the prostate, lung, colorectal, and ovarian cancer screening trial. Cancer Epidemiol Biomarkers Prev. 2008;17(4):889-94. (doi: 10.1158/1055-9965.EPI-07-2594).

28. Janowsky EC, Lester GE, Weinberg CR, et al. Association between low levels of 1,25-dihydroxyvitamin D and breast cancer risk. Public Health Nutr. 1999;2(3):283-91.

29. John EM, Schwartz GG, Koo J, et al. Sun Exposure, Vitamin D Receptor Gene Polymorphisms, and Breast Cancer Risk in a Multiethnic Population. Am J Epidemiol. 2007. (doi: 10.1093/aje/kwm259).

111

Paper 3: Sunlight and Breast Cancer Risk

30. Millen AE, Pettinger M, Freudenheim JL, et al. Incident invasive breast cancer, geographic location of residence, and reported average time spent outside. Cancer Epidemiol Biomarkers Prev. 2009;18(2):495-507. (doi: 10.1158/1055-9965.EPI-08-0652).

31. Kuper H, Yang L, Sandin S, et al. Prospective Study of Solar Exposure, Dietary Vitamin D Intake, and Risk of Breast Cancer among Middle-aged Women. Cancer Epidemiol Biomarkers Prev. 2009. (doi: 10.1158/1055-9965.EPI-09-0449).

32. Millen AE, Bodnar LM. Vitamin D assessment in population-based studies: a review of the issues. Am J Clin Nutr. 2008;87(4):1102S-5S.

33. Cotterchio M, Boucher BA, Kreiger N, et al. Dietary phytoestrogen intake--lignans and isoflavones--and breast cancer risk (Canada). Cancer Causes Control. 2008;19(3):259-72. (doi: 10.1007/s10552-007-9089-2).

34. Webb AR, Kline L, Holick MF. Influence of season and latitude on the cutaneous synthesis of vitamin D3: exposure to winter sunlight in Boston and Edmonton will not promote vitamin D3 synthesis in human skin. J Clin Endocrinol Metab. 1988;67(2):373-8.

35. Total Ozone Mapping Spectrometer Data Product: Erythemal UV Exposure , 2007.

36. McKinley A, Deffey BL. A reference action spectrum for ultraviolet induced erythema in human skin. In: Passchier WF, Bosnjakovich BFM, eds. Human Exposure to Ultraviolet Radiation: Risks and Regulations. International Congress Series, 1987:83-7.

37. MacLaughlin JA, Anderson RR, Holick MF. Spectral character of sunlight modulates photosynthesis of previtamin D3 and its photoisomers in human skin. Science. 1982;216(4549):1001-3.

38. Kimlin MG, Downs NJ, Parisi AV. Comparison of human facial UV exposure at high and low latitudes and the potential impact on dermal vitamin D production. Photochem Photobiol Sci. 2003;2(4):370-5.

39. Fioletov VE, McArthur LJ, Mathews TW, et al. On the relationship between erythemal and vitamin D action spectrum weighted ultraviolet radiation. J Photochem Photobiol B. 2009;95(1):9-16. (doi: 10.1016/j.jphotobiol.2008.11.014).

40. Pope SJ, Holick MF, Mackin S, et al. Action spectrum conversion factors that change erythemally weighted to previtamin D3-weighted UV doses. Photochem Photobiol. 2008;84(5):1277-83. (doi: 10.1111/j.1751-1097.2008.00373.x).

41. Webb AR, Engelsen O. Calculated ultraviolet exposure levels for a healthy vitamin D status. Photochem Photobiol. 2006;82(6):1697-703. (doi: 10.1562/2005-09-01-RA-670).

42. Chen TC, Chimeh F, Lu Z, et al. Factors that influence the cutaneous synthesis and dietary sources of vitamin D. Arch Biochem Biophys. 2007;460(2):213-7. (doi: S0003-9861(06)00508-X [pii]; 10.1016/j.abb.2006.12.017 [doi]).

43. Norval M, Wulf HC. Does chronic sunscreen use reduce vitamin D production to insufficient levels? Br J Dermatol. 2009;161(4):732-6. (doi: 10.1111/j.1365-2133.2009.09332.x).

44. Sahota H, Barnett H, Lesosky M, et al. Association of vitamin d related information from a telephone interview with 25-hydroxyvitamin d. Cancer Epidemiol Biomarkers Prev. 2008;17(1):232-8. (doi: 10.1158/1055-9965.EPI-07-0632).

112

Paper 3: Sunlight and Breast Cancer Risk

45. Thieden E, Philipsen PA, Heydenreich J, et al. Vitamin D level in summer and winter related to measured UVR exposure and behavior. Photochem Photobiol. 2009;85(6):1480-4. (doi: 10.1111/j.1751-1097.2009.00612.x).

46. Giovannucci E, Liu Y, Rimm EB, et al. Prospective study of predictors of vitamin D status and cancer incidence and mortality in men. J Natl Cancer Inst. 2006;98(7):451-9. (doi: 10.1093/jnci/djj101).

47. Ng K, Wolpin BM, Meyerhardt JA, et al. Prospective study of predictors of vitamin D status and survival in patients with colorectal cancer. Br J Cancer. 2009;101(6):916-23. (doi: 10.1038/sj.bjc.6605262).

48. Forman JP, Giovannucci E, Holmes MD, et al. Plasma 25-hydroxyvitamin D levels and risk of incident hypertension. Hypertension. 2007;49(5):1063-9. (doi: 10.1161/HYPERTENSIONAHA.107.087288).

49. Anderson LN, Cotterchio M, Vieth R, et al. Vitamin D and calcium intakes and breast cancer risk in pre- and postmenopausal women. Am J Clin Nutr. 2010; 91(6): 1699-1707.

50. Vieth R. How to optimize vitamin D supplementation to prevent cancer, based on cellular adaptation and hydroxylase enzymology. Anticancer Res. 2009;29(9):3675-84.

51. Levi F, Randimbison L, Te VC, et al. Risk of prostate, breast and colorectal cancer after skin cancer diagnosis. Int J Cancer. 2008;123(12):2899-901. (doi: 10.1002/ijc.23816).

52. Soerjomataram I, Louwman WJ, Lemmens VE, et al. Are patients with skin cancer at lower risk of developing colorectal or breast cancer? Am J Epidemiol. 2008;167(12):1421-9. (doi: 10.1093/aje/kwn077).

53. Tuohimaa P, Pukkala E, Scelo G, et al. Does solar exposure, as indicated by the non-melanoma skin cancers, protect from solid cancers: vitamin D as a possible explanation. Eur J Cancer. 2007;43(11):1701-12. (doi: 10.1016/j.ejca.2007.04.018).

54. Grant WB. A meta-analysis of second cancers after a diagnosis of nonmelanoma skin cancer: additional evidence that solar ultraviolet-B irradiance reduces the risk of internal cancers. J Steroid Biochem Mol Biol. 2007;103(3-5):668-74. (doi: 10.1016/j.jsbmb.2006.12.030).

55. Okasha M, McCarron P, Gunnell D, et al. Exposures in childhood, adolescence and early adulthood and breast cancer risk: a systematic review of the literature. Breast Cancer Res Treat. 2003;78(2):223-76.

56. Yu CL, Li Y, Freedman DM, et al. Assessment of lifetime cumulative sun exposure using a self-administered questionnaire: reliability of two approaches. Cancer Epidemiol Biomarkers Prev. 2009;18(2):464-71. (doi: 10.1158/1055-9965.EPI-08-0894).

57. Kricker A, Vajdic CM, Armstrong BK. Reliability and validity of a telephone questionnaire for estimating lifetime personal sun exposure in epidemiologic studies. Cancer Epidemiol Biomarkers Prev. 2005;14(10):2427-32. (doi: 14/10/2427 [pii]; 10.1158/1055-9965.EPI-05-0265 [doi]).

58. English DR, Armstrong BK, Kricker A. Reproducibility of reported measurements of sun exposure in a case-control study. Cancer Epidemiol Biomarkers Prev. 1998;7(10):857-63.

59. van der Mei IA, Blizzard L, Ponsonby AL, et al. Validity and reliability of adult recall of past sun exposure in a case-control study of multiple sclerosis. Cancer Epidemiol Biomarkers Prev. 2006;15(8):1538-44. (doi: 15/8/1538 [pii]; 10.1158/1055-9965.EPI-05-0969 [doi]).

113

Paper 3: Sunlight and Breast Cancer Risk

60. Weiler L, Knight JA, Vieth R, et al. Comparison of self-reported lifetime sun exposure with two methods of cutaneous microtopography. Am J Epidemiol. 2007;165(2):222-30. (doi: 10.1093/aje/kwk004).

61. Karagas MR, Zens MS, Nelson HH, et al. Measures of cumulative exposure from a standardized sun exposure history questionnaire: a comparison with histologic assessment of solar skin damage. Am J Epidemiol. 2007;165(6):719-26. (doi: 10.1093/aje/kwk055).

62. Brot C, Vestergaard P, Kolthoff N, et al. Vitamin D status and its adequacy in healthy Danish perimenopausal women: relationships to dietary intake, sun exposure and serum parathyroid hormone. Br J Nutr. 2001;86 Suppl 1:S97-103.

63. Kim JH, Moon SJ. Time spent outdoors and seasonal variation in serum concentrations of 25-hydroxyvitamin D in Korean women. Int J Food Sci Nutr. 2000;51(6):439-51.

64. Need AG, Morris HA, Horowitz M, et al. Effects of skin thickness, age, body fat, and sunlight on serum 25-hydroxyvitamin D. Am J Clin Nutr. 1993;58(6):882-5.

65. Canadian Cancer Society/National Cancer Institute of Canada. Canadian Cancer Statistics 2008. Toronto, Canada, 2008.

114

Paper 3: Sunlight and Breast Cancer Risk

Table 1. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for sun

exposure related variables during 4 age periods

Variable Cases

n = 3101

No. (%)

Controls

n = 3471

No. (%)

OR (95% CI)1

Ethnicity (proxy for skin color) Caucasian Non-Caucasian

2749 (89) 341 (11)

3121(90) 330 (10)

1.00 1.23 (1.05-1.45)

Teenage years Use of sun protection2 Never Sometimes Always

1601 (54) 1234 (41) 149 (5)

1829 (55) 1354 (40) 169 (5)

1.00 1.07 (0.97-1.19) 1.02 (0.81-1.28) P trend = 0.29

Hrs outdoor per week3 ≤ 6 hours 7 - 12 hours 13 -14 hours 15 - 21 hours >21 hours

365 (12) 558 (19) 505 (17) 566 (19) 944 (32)

324 (10) 549 (17) 569 (17) 679 (20) 1198 (36)

1.00 0.90 (0.74-1.09) 0.81 (0.66-0.98) 0.76 (0.63-0.91) 0.71 (0.60-0.85) P trend <0.0001

Latitude of residence4 ≤ 42.5 ºN 42.6 - 43.5 ºN (Toronto Area) 43.6 - 45.0 ºN >45.0 ºN

531 (19) 1276 (46) 272 (10) 686 (25)

547 (18) 1414 (46) 364 (12) 768 (25)

1.11 (0.94-1.30) 1.03 (0.91-1.17) 0.85 (0.71-1.03) 1.00 P trend = 0.07

Erythemal UV of residence (mW/m2)5 10-170 170-179 180-380

316 (11) 1576 (57) 872 (32)

358 (11) 1706 (55) 1029 (33)

1.00 1.08 (0.94-1.25) 1.06 (0.91-1.23) P trend = 0.45

20s-30s Use of sun protection Never Sometimes Always

1004 (34) 1647 (56) 270 (9)

1119 (34) 1889 (57) 300 (9)

1.00 1.04 (0.93-1.16) 1.11 (0.92-1.36) P trend = 0.31

Hrs outdoor per week ≤ 6 hours 7 - 12 hours 13 -14 hours 15 - 21 hours >21 hours

528 (18) 758 (26) 640 (22) 551 (19) 452 (15)

488 (15) 858 (26) 720 (22) 576 (17) 655 (20)

1.00 0.81 (0.70-0.95) 0.84 (0.71-0.99) 0.88 (0.74-1.05) 0.64 (0.53-0.76) P trend <0.0001

115

Paper 3: Sunlight and Breast Cancer Risk

Latitude of residence ≤42.5 ºN 42.6 - 43.5 ºN (Toronto Area) 43.6 - 45.0 ºN >45.0 ºN

644 (22) 1015 (35) 320 (11) 908 (31)

662 (20) 1137 (35) 383 (12) 1050 (32)

1.15 (1.00-1.33) 1.07 (0.94-1.21) 0.98 (0.83-1.17) 1.00 P trend = 0.22

Erythemal UV of residence (mW/m2) 10-170 170-179 180-380

497 (17) 1370 (47) 1019 (35)

575 (18) 1509 (47) 1148 (36)

1.00 1.07 (0.91-1.26) 0.99 (0.83-1.18) P trend = 0.73

40s-50s6 n = 2920 n = 3155

Use of sun protection Never Sometimes Always

501 (18) 1529 (56) 687 (25)

551 (18) 1736 (58) 697 (23)

1.00 0.98 (0.86-1.13) 1.12 (0.95-1.31) P trend = 0.15

Hrs outdoor per week ≤ 6 hours 7 - 12 hours 13 -14 hours 15 - 21 hours >21 hours

736 (27) 807 (30) 482 (18) 422 (15) 286 (10)

766 (26) 835 (28) 566 (20) 417 (14) 400 (13)

1.00 1.00 (0.87-1.15) 0.89 (0.76-1.04) 1.04 (0.88-1.23) 0.74 (0.61-0.88) P trend = 0.007

Latitude of residence ≤ 42.5 ºN 42.6 – 43.5 ºN (Toronto Area) 43.6 – 45.0 ºN >45.0 ºN

365 (13) 1480 (55) 325 (12) 541 (20)

417 (14) 1516 (52) 416 (14) 577 (20)

0.94 (0.78-1.13) 1.05 (0.92-1.21) 0.83 (0.69-1.01) 1.00 P trend = 0.62

Erythemal UV of residence (mW/m2) 10 - 170 170-179 180-380

152 (6) 1742 (64) 817 (30)

168 (6) 1754 (60) 1004 (34)

1.00 1.11 (0.88-1.40) 0.91 (0.71-1.15) P trend = 0.04

60s-747 n = 1247 n = 1224

Use of sun protection Never Sometimes Always

202 (7) 531 (17) 430 (14)

206 (16) 637 (51) 410 (33)

1.00 0.85 (0.68-1.06) 1.07 (0.84-1.36) P trend = 0.24

Hrs outdoor per week ≤ 6 hours 7 - 12 hours 13 -14 hours 15 - 21 hours >21 hours

402 (35) 370 (32) 115 (10) 171 (15) 93 (8)

366 (29) 352 (28) 143 (12) 210 (17) 172 (14)

1.00 0.96 (0.78-1.17) 0.74 (0.56-0.99) 0.74 (0.58-0.95) 0.50 (0.37-0.66) P trend <0.0001

116

Paper 3: Sunlight and Breast Cancer Risk

Latitude of residence ≤ 42.5 ºN 42.6 – 43.5 ºN (Toronto Area) 43.6 – 45.0 ºN >45.0 ºN

158 (13) 637 (54) 162 (14) 230 (20)

159 (13) 622 (50) 226 (18) 244 (20)

1.05 (0.79-1.40) 1.09 (0.88-1.34) 0.76 (0.58-0.99) 1.00 P trend = 0.17

Erythemal UV of residence (mW/m2) 120-179 180-380

787 (66) 400 (34)

805 (64) 445 (36)

1.00 0.92 (0.77-1.08) P trend = 0.29

1 Age-group adjusted (note: 39 variables were evaluated as potential confounders and none were identified as confounders i.e., their inclusion in the model did not change the OR>10%) 2 Protective clothing or sunscreen use 3 Time spent outdoors from May to September only 4 Geocoded based on location lived reported as: city, and province/state. When respondents indicated country only the most populated city was used. 5 Monthly average local noon erythemal UV radiation for June 2003 obtained from NASA’s Total Ozone Mapping Spectrometer (TOMS) 6 Age 40 not reached by 181 (6%) of cases and 316 (10%) of controls 7 Age 60 not reached by 1854 (61%) of cases and 1324 (63%) of controls

117

Paper 3: Sunlight and Breast Cancer Risk

Table 2. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for

derived proxy measures of vitamin D from sunlight during 4 age periods, recent exposure only,

and cumulative life exposure

Solar vitamin D score1 Cases

n = 3101

No. (%)

Controls

n = 3471

No. (%)

OR2 (95% CI)

By specific age period Teenage years (n = 5924)

Q1 (≤ 1425 mw/m2 · hrs)

Q2 (1426-2295) Q3 (2296-3570) Q4 (>3570)

797 (29) 743 (27) 696 (25) 557 (20)

780 (25) 804 (26) 855 (27) 692 (22)

1.00 0.90 (0.78-1.03) 0.80 (0.70-0.93) 0.79 (0.68-0.91) P trend = 0.0007

20s-30s (n = 5579)

Q1 (≤ 957 mw/m2 · hrs)

Q2 (958-1514) Q3 (1515-2430) Q4 (>2430)

718 (27) 497 (19) 862 (33) 546 (21)

739 (25) 528 (18) 979 (33) 710 (24)

1.00 0.95 (0.81-1.12) 0.89 (0.77-1.02) 0.76 (0.65-0.89) P trend = 0.0003

40s-50s (n = 5263)

Q1 ( ≤617 mw/m2 · hrs)

Q2 ( 618-1178) Q3 (1179 -1785) Q4 (>1785)

665 (26) 560 (22) 754 (30) 541 (21)

623 (23) 615 (22) 841 (31) 664 (24)

1.00 0.85 (0.72-0.99) 0.82 (0.71-0.95) 0.75 (0.64-0.88) P trend = 0.0009

60s-74 (n = 2257) Q1 (≤ 589 mw/m2

· hrs) Q2 (590-1178) Q3 (1179-1767) Q4 (>1767)

295 (27) 313 (29) 285 (26) 197 (18)

257 (22) 300 (26) 318 (27) 292 (25)

1.00 0.91 (0.72-1.14) 0.78 (0.62-0.98) 0.59 (0.46-0.76) P trend = <0.0001

Cumulative 3 (n = 6159)

Low at all age periods Low at 3 age periods Low at 2 ages & high at 2 High at 3 age periods High at all age periods

874 (30) 639 (22) 602 (21) 573 (20) 219 (8)

878 (27) 682 (21) 716 (22) 667 (21) 309 (10)

1.00 0.93 (0.81-1.08) 0.84 (0.72-0.96) 0.81 (0.70-0.94) 0.63 (0.51-0.78) P trend = <0.0001

Recent4 (n = 5805)

Q1 (≤ 589 mw/m2 · hrs)

Q2 (590-1178) Q3 (1179-1603) Q4 (>1603)

692 (25) 799 (29) 624 (23) 619 (23)

669 (22) 856 (28) 711 (23) 838 (27)

1.00 0.90 (0.78-1.04) 0.84 (0.72-0.98) 0.72 (0.62-0.84) P trend = <0.0001

118

Paper 3: Sunlight and Breast Cancer Risk

1Average weekday and weekend hours spent outdoors per week multiplied by erythemal UV radiation of residence weighted for skin color and sun protection practices. Refer to methods for more details. 2 Adjusted for age group 3 Exposure at all 4 age periods was classified as high (>50%le) or low (<50%le) and added for all age periods reached 4 Exposure during age period when questionnaire was completed

119

Table 3. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for

derived proxy measures of vitamin D from sunlight during 4 age periods stratified by vitamin D

supplement use

Vitamin D supplement use

None

Any

Solar vitamin D score1 OR

2 (95% CI) OR

2 (95% CI) P interaction

By specific age period Teenage years (n = 5924) Q1 (≤ 1425 mw/m2

· hrs) Q2 (1426-2295) Q3 (2296-3570) Q4 (>3570)

1.00 0.96 (0.79-1.16) 0.76 (0.63-0.92) 0.79 (0.65-0.97)

1.00 0.84 (0.68-1.04) 0.88 (0.71-1.08) 0.80 (0.64-1.01)

0.3

20s-30s (n = 5579) Q1 (≤ 957 mw/m2

· hrs) Q2 (958-1514) Q3 (1515-2430) Q4 (>2430)

1.00 0.87 (0.70-1.08) 0.97 (0.80-1.17) 0.69 (0.56-0.85)

1.00 1.06 (0.83-1.34) 0.80 (0.65-0.99) 0.88 (0.71-1.11)

0.015

40s-50s (n = 5263) Q1 ( ≤617 mw/m2

· hrs) Q2 ( 618-1178) Q3 (1179 -1785) Q4 (>1785)

1.00 0.83 (0.67-1.03) 0.95 (0.78-1.16) 0.69 (0.56-0.85)

1.00 0.86 (0.68-1.09) 0.70 (0.56-0.87) 0.79 (0.71-1.11)

0.03

60s-74 (n = 2257) Q1 (≤ 589 mw/m2

· hrs) Q2 (590-1178) Q3 (1179-1767) Q4 (>1767)

1.00 0.89 (0.64-1.24) 0.94 (0.68-1.31) 0.59 (0.42-0.84)

1.00 0.94 (0.68-1.31) 0.67 (0.48-0.93) 0.60 (0.42-0.85)

0.29

1Average weekday and weekend hours spent outdoors per week multiplied by erythemal UV radiation of residence weighted for skin color and sun protection practices. Refer to methods for more details. 2 Adjusted for age group

120

Table 4. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for

combined solar vitamin D score and vitamin D from supplements created by cross-classification

1 Age-group adjusted

Solar vitamin D score and vitamin

D supplement use

Cases

n = 3101

No. (%)

Controls

n = 3471

No. (%)

OR (95% CI)1

Teenage years Low solar and no supplement Low solar and any supplement High solar and no supplement High solar and any supplement

833 (30) 688 (25) 676 (24) 573 (21)

841 (27) 726 (23) 878 (28) 658 (21)

1.00 0.94 (0.81-1.08) 0.79 (0.69-0.91) 0.86 (0.74-1.00) P trend = 0.005

20-39 yrs of age Low solar and no supplement Low solar and any supplement High solar and no supplement High solar and any supplement

642 (25) 562 (22) 767 (29) 630 (24)

700 (24) 554 (19) 925 (32) 752 (26)

1.00 1.09 (0.93-1.28) 0.90 (0.78-1.03) 0.87 (0.75-1.01) P trend = 0.013

40-59 years of age Low solar and no supplement Low solar and any supplement High solar and no supplement High solar and any supplement

645 (26) 566 (23) 716 (29) 572 (23)

673 (25) 551 (20) 803 (30) 690 (25)

1.00 1.05 (0.90-1.24) 0.91 (0.79-1.06) 0.83 (0.71-0.97) P trend = 0.008

60-74 years of age Low solar and no supplement Low solar and any supplement High solar and no supplement High solar and any supplement

294 (27) 304 (28) 248 (23) 232 (22)

276 (24) 272 (24) 291 (25) 315 (27)

1.00 1.05 (0.84-1.33) 0.81 (0.64-1.02) 0.69 (0.55-0.88) P trend = 0.0003

121

Table 5. Application of the predicted 25(OH)D model from the Health Professionals Follow-Up

Study to our Ontario Women’s Diet and Health study data.

Health Professionals

Follow-Up Study

Ontario Women’s Diet and

Health Study

Intercept 90.8 Same value assigned Race White African American Asian

0 (ref) -12.8 -13.3

Caucasian=0 Black, Aboriginal, other= -12.8 SE Asian and Asian = -13.3

Residence South Midwest/West Northeast/Mid-Atlantic

0 (ref) -2.4 -6.4

All women were assigned the value for Northeast (since they all lived in Ontario when completing the study)

Quintile of leisure-time physical activity 5 4 3 2 1

0 (ref) -4.5 -7.7 -9.0 -13.5

Quintiles of weekly combined moderate and strenuous physical activity (during 20-30s and 40-50s) were assigned same values 1

Body mass index (kg/m2) <22 22-24.9 25-29.9 30-34.9 ≥35

0 (ref) -1.0 -4.5 -6.5 -8.6

Same values assigned

Dietary vitamin D (IU/day) ≥400 300-399 200-299 100-199 <100

0 (ref) -3.5 -2.6 -7.2 -10.4

Same values assigned

Supplementary vitamin D ≥400 200-399 1-199 <100

0 (ref) -1.8 +2.4 -2.1

Same values assigned

1 Quintiles of time spent outdoors (cumulative over all ages) was also substituted for physical activity since physical activity was considered to be a proxy for time spent outdoors.

122

Table 6. Distribution of breast cancer cases and controls and odds ratio (OR) estimates for

predicted 25(OH)D using the Health Professionals Study algorithm

Predicted 25(OH)D Cases

n = 3101

Controls

n = 3471

OR1 (95% CI)

mean (SD) mean (SD) Per 1 unit increase (nmol/L) 66.2 (7.8) 66.7 (7.6) 0.99 (0.98-1.00)

25 unit increase (nmol/L) 66.2 (7.8) 66.7 (7.6) 0.75 (0.64, 0.88)

No. (%) No. (%) Q1 (39-59.9) Q2 (60-65.2) Q3 (65.3-68.8) Q4 (69-73.2) Q5 (73.3-87)

657 (21) 650 (21) 621 (20) 594 (19) 579 (19)

699 (20) 696 (20) 665 (19) 723 (21) 688 (20)

1.00 0.97 (0.83-1.13) 0.96 (0.82-1.12) 0.84 (0.72-0.98) 0.84 (0.72-0.98)

1 Age adjusted

123

Chapter 5 Discussion and Conclusions

This chapter adds to the discussion section of each paper included in chapter 4 by providing a

brief summary and comparison to the literature for the key findings of each objective followed

by a detailed discussion of methodological issues, suggestions for future research and overall

conclusions. The study results have been discussed in detail and compared to the relevant

literature in the discussion section of each of the 3 papers presented in chapter 4. For

detailed discussion of findings for objective 1 (dietary vitamin D and calcium and breast cancer

risk) see paper 2 in chapter 4.3. For detailed discussion of findings for objective 5 (modification

of FFQ) see paper 1 in chapter 4.2. And for detailed discussion of objectives 2, 3 and 4 (vitamin

D from sunlight, algorithm derivation, and combined diet and sunlight) see paper 3 in chapter

4.4.

5.1 Summary of Findings and Comparison to the Literature

When this study began in 2006, there was a paucity of literature on vitamin D and breast cancer

risk. However, over the past 4 years the number of published epidemiologic studies that assessed

vitamin D and breast cancer risk has rapidly increased, with 15 additional studies published. Our

study contributes substantially to what is now a relatively large body of literature on vitamin D

and breast cancer risk; our findings are consistent with previous literature and we contributed

novel methodology.

5.1.1 Objectives 1 and 5

The results of our study suggest that intake of vitamin D from supplements (>400 IU/day

compared to none) is associated with reduced breast cancer risk (Anderson et al., 2010).

However, no significant associations were observed between breast cancer risk and vitamin D

from foods or total combined (food and supplements) dietary vitamin D. There are a few possible

explanations for this seemingly inconsistent finding. First, vitamin D intake from foods (and

correspondingly total combined intake) may be more susceptible than supplements to

misclassification (potentially biasing results towards the null). Second, foods containing vitamin

124

D may also contain other detrimental components that counteract the potential vitamin D

benefits (e.g., dietary fat in milk or contaminants in fish). Third, despite the similar cut-offs used

for food and supplements, the distribution of data within the upper categories varied

(distributions shown in figures 1-3 of appendix 4). As expected, intake of vitamin D from foods

was relatively low since few foods contain vitamin D and those with the highest amounts of

vitamin D (e.g., fatty fish) were consumed infrequently in our population and the range within

the highest category was not much greater than 400 IU/day. In contrast most supplement users in

the category >400 IU/day were assigned a value of 800 IU/day (the equivalent of one supplement

and one multivitamin). These differences in distribution alone do not explain why total combined

intake was not associated with breast cancer risk. Lastly, we cannot rule out the possibility that

our findings may be due to chance.

Adaptation of the FFQ for values from Canada versus the US did not substantially change the

measured levels of vitamin D from foods or the associations with breast cancer risk; the standard

US values only slightly underestimated modified Canadian values (Anderson et al., in press).

Furthermore, total calcium intake was not associated with breast cancer risk, and was neither an

effect modifier nor confounder of the vitamin D breast cancer association. A significant trend

towards decreased breast cancer risk was observed for calcium from supplements.

Our results are relatively consistent with the growing body of literature that suggests vitamin D

is associated with reduced breast cancer risk (see chapter 2). Most previous studies of vitamin D

from food and/or supplements, but not all (Kuper et al., 2009), report some inverse associations

with breast cancer risk (Abbas et al., 2007; John et al., 1999; Knight et al., 2007; Lin et al., 2007;

McCullough et al., 2005; Robien et al., 2007; Rossi et al., 2009; Shin et al., 2002). But, in

contrast to our study and the previous case-control studies (Abbas et al., 2007; Knight et al.,

2007; Rossi et al., 2009), the results from many cohort studies have not been statistically

significant (Lin et al., 2007; McCullough et al., 2005; John et al., 1999; Robien et al., 2007),

although the effect estimates suggest a modest risk reduction. Only one study reported positive

but non-significant associations between vitamin D from food and breast cancer risk in

postmenopausal women but not premenopausal women (Lin et al., 2007).

When differences are observed between study designs, randomized controlled trials and cohort

studies are often considered superior to case-control study results which may be subject to

125

additional measurement error from recall. There are, however, some limitations of the current

RCT and cohort studies of vitamin D and breast cancer risk that may have biased results. The

cohort studies were not designed with the primary objective of evaluating vitamin D and breast

cancer risk and thus included limited measures of vitamin D (e.g., no supplements/cod liver oil

or sun exposure) which may increase measurement error and bias results towards null. The only

trial specific to vitamin D and breast cancer risk (Chlebowski et al., 2008) was conducted with a

low dose that may have been insufficient to create a difference between the intervention and

control group. Furthermore, participants of long term prospective studies and who are not lost to

follow-up may be different (e.g., more health conscious or in need of health care) than those who

are willing to participate in a less time consuming case-control study which may result in

reduced generalizability and/or reduced variation in exposure). Furthermore, the identification of

cases is usually by routine diagnosis in a case-control study; whereas, it may be conducted

through active surveillance in trials and cohort studies. It is unknown how this may affect studies

of breast cancer but if breast cancer cases are more advanced at diagnosis in case-control studies

than cohort studies or trials, and if vitamin D is important for progression or stage of disease,

then this may explain the tendency towards more null findings from cohort/trials. Prospective

studies often have fewer cases and thus may be not sufficiently powered to detect a significant

association; however, this was not true for most of the vitamin D breast cancer studies to-date.

5.1.2 Objectives 2 and 4

Our study also presents the development and application of a novel solar vitamin D score that

takes into consideration a multitude of factors that affect the cutaneous production of vitamin D.

This score, and the component time spent outdoors, were both inversely associated with breast

cancer risk and there did not appear to be a critical period of exposure; inverse associations were

observed for all age periods of exposure and were only slightly stronger for exposure during age

60 to 74. Time spent outdoors was moderately correlated between all age periods of exposure

(spearman correlation coefficients ranged from 0.37 to 0.66 and all p-values <0.0001); similar

correlations were observed for the solar vitamin D score. The solar vitamin D score appeared to

be driven primarily by time spent outdoors; the correlations between time spent outdoors and the

solar vitamin D score were high (spearman correlation coefficients ranged from 0.75 to 0.80

when comparing the same age-periods and all p-values were <0.0001). Sun protection practices

126

and UV radiation – other components of the solar vitamin D score – were not independently

associated with breast cancer risk. Caucasian ethnicity, our proxy for lighter skin color, was

associated with reduced breast cancer risk consistent with the vitamin D hypothesis. However,

the association between ethnicity and breast cancer risk may be explained by many other

variables (e.g., genetics) and our observed association is inconsistent with the general finding

that breast cancer rates are highest among White women (Altekruse, Kosary, Krapcho, Neyman

et al., 2009). Since the vast majority of women in our study are Caucasian (90%) we are unable

to further explore this association.

Few studies have evaluated the association between person-specific measures of sunlight

exposure (e.g., time spent outdoors) and breast cancer risk (John et al., 1999; John et al., 2007;

Knight et al., 2007; Millen et al., 2009) and, consistent with our results, all but one previous

study (Kuper et al., 2009) report inverse associations. No previous studies of breast cancer risk

have created a combined solar vitamin D score. We did not identify any confounders or effect

modifiers of the associations between vitamin D from each source and breast cancer risk. It has

been hypothesized that the association between time spent outdoors and breast cancer risk may

be explained by physical activity (Knight et al., 2007) but this was not found in our study;

however, our measures of physical activity (daily, moderate or strenuous) did not distinguish

indoor versus outdoor activity.

5.1.3 Objective 3

When we combined both vitamin D from supplements and either our solar vitamin D score or

time outdoors into one variable, the effect estimates for breast cancer risk were not substantially

stronger than our models with only the solar vitamin D score or time outdoors; however, few

women used vitamin D supplements and we may not have had sufficient power to evaluate the

combined effect. Adjusting the solar vitamin D score models for supplemental vitamin D, or vice

versa, did not weaken the associations observed associations for either variable, suggesting that

there is an independent effect of both vitamin D from sun and supplements. Lastly, the

associations between breast cancer risk and both vitamin D from supplements and our solar

vitamin D score were not modified by menopausal status, BMI, or calcium intake and no

significant interactions were observed.

127

One previous study investigated the combined effect of diet and sunlight and found the inverse

associations were slightly stronger when for a measure of high sun and diet than either measure

on its own (John et al., 1999). There is some evidence of an interaction between diet and sunlight

(McCullough et al., 2005) but this was not found elsewhere (Millen et al., 2009). The literature

from biomarker studies of breast cancer risk and 25(OH)D, which reflects vitamin D intake from

both diet and sunlight, is less consistent. Of the 10 studies that we identified, 5 reported

significant inverse associations between 25(OH)D and breast cancer risk (Abbas et al., 2008;

Abbas et al., 2009; Crew et al., 2009; Lowe et al., 2005; Rejnmark et al., 2009); however, only 1

of these studies was prospective and measured pre-diagnosis 25(OH)D levels in cases (Rejnmark

et al., 2009). Other prospective studies found no significant associations between 25(OH)D and

breast cancer risk but in most the ORs were less than 1 (Bertone-Johnson et al., 2005;

Chlebowski et al., 2008; McCullough et al., 2009) and sample sizes were not large. Overall, the

results of our study tend to support the ‘vitamin D hypothesis’ and are consistent with a growing

body of epidemiologic studies, particularly observational studies, that have reported an inverse

association between vitamin D and breast cancer risk.

5.2 Limitations and Methodological Issues

Limitations and methodological issues related to this study have been briefly discussed in each of

the 3 papers included in the results section of this thesis. In this section a more detailed

discussion of issues related to bias (both selection and information) and other threats to the

internal study validity (e.g., confounding, effect modification, and statistical analysis) and

external study is presented. Lastly, a discussion of general limitations is included.

5.2.1 Selection Bias

Selection bias occurs when there are systematic differences between the exposure and disease

relationship in study participants versus non-participants (Rothman & Greenland, 1998 p. 119).

A potential limitation of case-control studies is identifying controls from the same source

population from which the cases arose; controls should be comparable to the cases in all respects

except for the disease of interest (Kopec & Esdaile, 1990; Szklo & Nieto, 2000). Specific to our

study, incident breast cancer cases were identified from the population-based Ontario Cancer

Registry; thus, the population cohort from which the cases arose can readily be identified as all

128

women in Ontario. Controls were identified from the same population using a modified random

digit dialing procedure. A sampling frame of both listed (from multiple directories) and unlisted

(numbers on either side of the listed numbers) telephone numbers was used to conduct random

digit dialing. In 2006 an estimated 93% of Ontario households had a residential phone line and

thus random digit dialing captures the base population well (Statistics Canada, 2006). To

minimize response bias, every effort was made to obtain a high response rate and overall 75%

cases and 80% of identified eligible controls participated in the study.

Although high response rates were obtained for both cases and controls (of those who were

contacted and eligible), it is unknown if the association between vitamin D and breast cancer

differs among those electing to not participate. For example, if the selection of cases is not biased

(i.e., assuming breast cancer cases were equally motivated to participate independent of vitamin

D status) but the selection of controls is biased towards the inclusion of more health conscious

women (e.g., more likely to take supplements and/or spend time outdoors) then the association

between vitamin D and breast cancer risk may be overestimated. For example, if cases were

equally likely to be selected regardless of exposure but controls were more likely to be selected if

they had higher vitamin D exposure than the true risk estimate could be 1.0 or greater

(Greenland, 1996). It is reassuring, however, that vitamin D intake in our study is relatively

consistent with previous studies that have reported mean daily vitamin D intake from food only

to be around 5 µg among Canadian women (Berube et al., 2005; Csizmadi et al., 2007; Statistics

Canada, 2004; Vieth, Cole, Hawker, Trang, & Rubin, 2001). Moreover, we found other

established breast cancer risk factors (e.g., family history of breast cancer, parity) were

associated with breast cancer risk in the expected direction.

Despite the population-based approach to recruitment of cases and controls, and the high

response rate, there is the potential for sampling bias in terms of both who was invited to

participate and who did actually participate in the study. Although we expect the coverage of

cancer cases to be nearly complete in the Ontario Cancer Registry, the population-based

approach to random digit dialing used to recruit controls excludes Ontario residents without a

phone. It also exclude potential participants who were not at home at the various times

throughout the day differences when contact was attempted. Furthermore, if the cases who

survived and participated in our study had higher vitamin D intake/exposure then non-survivors

129

our effect estimates may be biased towards the null; in contrast, if non-survivors had higher

vitamin D intake then our estimates may be overestimated. However, survival bias is likely

minimal since cases were recruited on average within 1 year of diagnosis and the 5 year relative

survival rate for breast cancer is 87% .

5.2.2 Information Bias

An inherent limitation of observational studies is the potential for measurement error.

Differential measurement error between cases and controls may bias study results in either

direction yielding inaccurate effect estimates. Non-differential measurement error is usually

considered less of a concern in observational studies because it often, but not always, biases

results towards the null. Recall bias, which is a type of differential measurement error, is always

a concern in case-control studies. It is possible that cases differentially recall exposures that are

linked to disease risk, particularly those that have received considerable media attention. We do

not expect this to be a concern in our study, since at the time of data collection (2002-2004)

limited information was available linking vitamin D to breast cancer risk. Furthermore,

participants were not made aware of the study hypotheses. Conducting a similar study today may

be more problematic with the considerable media attention towards vitamin D, raising public

awareness about the vitamin D-breast cancer hypothesis.

To minimize measurement error, standardized procedures were employed for data collection

with established quality control mechanisms. Validated measures were used whenever possible

for the measurement of variables. The validity of the Block FFQ was assessed prior to its use in

the Ontario Women’s Diet and Health Study and moderate to high reliability and validity were

observed for most nutrients, including vitamin D and calcium (Boucher et al., 2006). To

minimize the effect of any dietary changes following cancer diagnosis, and to capture exposure

status prior to disease, cases and controls were asked about exposures and diet two years prior to

the study start. For this thesis, we further modified the nutrient analysis of the FFQ to be specific

for vitamin D intake among Canadians. Furthermore, both food and nutrient level analysis were

conducted as recommended to obtain maximal information (Rothman & Greenland, 1998 p.630).

There is no obvious reason to suspect that cases and controls differentially recalled vitamin D

supplement use or foods containing vitamin D.

130

The validity (Pearson’s deattenuated coefficients) of vitamin D and calcium from diet and

supplements combined (using the standard US nutrient analysis) was 0.54 and 0.71, respectively

(Boucher et al., 2006). Re-arranging the formula RRobserved = (RRtrue) γ provided by Willett et al.,

(Willett, 1998) we evaluated what the possible true risk estimate would be given a range of

validity levels (γ) that overlapped these values. Assuming observed risk estimates of 0.75 and

0.85 (as were observed in our study for vitamin D and calcium from supplements, respectively

(Anderson et al., 2010)), Table 1 presents the possible corrected or ‘true’ risk estimates which

range from 0.40 to 0.68 and 0.58 to 0.80, respectively.

Table 1. Corrected ‘true’ risk estimates for selected measures of validity (γ) and observed risk

estimates of 0.76 and 0.85

Observed risk estimates

γ 0.761 0.85

2

Corrected ‘true’ risk estimates3

0.3 0.40 0.58

0.5 0.58 0.72

0.7 0.68 0.80

1 Risk estimate observed in our study for extreme categories of vitamin D supplement use and breast cancer risk

2 Risk estimate observed in our study for the extreme categories of calcium supplement use

3 Calculated as

γ√RRobserved

In regards to the measurement of vitamin D from sunlight, we captured person specific measures

of sun exposure (e.g., time spent outdoors, sun protection practices and skin color) in addition to

ambient UV at location of residence. While capturing sun exposure for 4 specific periods of life

was largely a strength of this study, the long period of recall may increase measurement error. It

is, however, unlikely that this error would differ between cases and controls. An additional

131

limitation of our study is the broad exposure categories for time spent outdoors with the shortest

period being <1hr per day; it has been proposed that a better method to capture sun exposure

relevant to vitamin D production may be to ask about the frequency of exposure instead of total

time outdoors (i.e., how many days per week did you spend at least 30 minutes outdoors)

(Knight et al., 2007). Additionally, the measures of sun protection practices and skin color are

relatively crude and future studies should use more detailed questionnaires. Our study did not

include any measures of tanning bed use or wintertime sun exposure (e.g., “snowbirds” or sun

holidays) which may introduce misclassification error when developing the solar vitamin D

score. The purpose of the solar vitamin D score was to predict vitamin D from sun and validation

of our solar vitamin D score is necessary.

Evaluating the validity of our algorithm used to derive the sun exposure score was beyond the

scope of this thesis due to feasibility constraints. To conduct a validation study, blood samples

would need to be collected from 100-200 women (as estimated a reasonable sample size for

validation studies by Serra-Majem et al., 2009) and laboratory analysis for serum 25(OH)D

would need to be conducted. Due to the length of time between the initial study questionnaires

and start of this thesis project, additional questionnaires would also have to be administered

repeating the sun exposure questionnaire. The increased costs, time and uncertainty to apply for

and secure funding, obtain ethics approval, and carry out such a study was not feasible within the

restricted time range for a PhD thesis. One previous study found a score combining time spent

outdoors and amount of skin exposed was significantly correlated with summer 25(OH)D levels

(r = 0.59); however, this was only slightly better than time spent outdoors alone (r = 0.58)

(Hanwell et al., 2010). Elsewhere a sun index (the product of sun exposure hours per week and

fraction of body surface area exposed) was moderately correlated with summer 25(OH)D levels

(r = 0.49) whereas no significant correlation was observed for exposure hours alone (Barger-Lux

& Heaney, 2002).

Although there is a high likelihood of measurement error for both vitamin D from diet and sun

exposure, it is not expected to differ between cases and controls (i.e., non-differential

misclassification) and, thus, results will typically be biased towards the null. Two exceptions to

this rule are when there are multiple exposure categories (as opposed to a binary exposure) and

when a continuous variable that is not normally distributed is categorized, then bias could be in

132

either direction. A limitation of validating such observational measures is that they are usually

done at one point in time (e.g., one biomarker 25(OH)D measure, or two 24-hour recalls) and we

are interested in usual exposure over the life course. Ideally a prospective long term validation

study would be conducted evaluating stability of 25(OH)D over many years/seasons and later

recall of dietary intake and sunlight exposure.

5.2.3 Confounding and Effect Modification

Confounding is often considered a special type of bias that results in “a confusion of effects”

(Rothman & Greenland, 1998 p. 120). Efforts to minimize the confounding effect of age were

made in both the design (i.e., frequency matching on age) and analysis (i.e., statistical

adjustment) stages of the study. Many potential confounders were evaluated in the multivariable

models. Few of these variables were known a priori to meet the definition of a confounder –

most established breast cancer risk factors were not known to be associated with vitamin D

exposure/intake. Inclusion of any of the potential confounders in the overall multivariable

models did not substantially change the observed age adjusted ORs for any of the primary

exposure variables suggesting that the association between vitamin D and breast cancer risk is

not explained by a confounding factor.

The potential for residual confounding, however, cannot be ruled out. Residual confounding may

arise if possible confounders are inadequately measured (e.g., self-reported history of screening

mammogram) or are not included. For example, residual confounding may arise if healthy

lifestyle was inadequately captured in this study and if healthy lifestyle were a common cause of

both vitamin D exposure and reduced breast cancer risk. Although healthy lifestyle itself was not

measured, neither multivitamin use nor physical activity, components of a healthy lifestyle,

confounded our observed vitamin D and breast cancer associations. Furthermore, time spent

outdoors at all age periods of exposure was not associated with parity, education or income. It is

highly plausible that the observed associations between time spent outdoors and breast cancer

risk may be confounded by other variables not included in this study.

In this study we evaluated the interactions between vitamin D exposure and calcium, menopausal

status, and BMI. No significant interactions were observed and stratified analysis did not reveal

any obvious effect modification. However, few premenopausal women were taking vitamin D

133

supplements and vitamin D and calcium were highly correlated, limiting our power to draw any

conclusions.

5.2.4 Analytical Issues

The above sections considered the study validity with respect study design and systematic error

(bias) whereas this section addresses issues related to statistical analysis and random error. A

strength of this study is that there were minimal missing data. Less than 2% of women had

missing responses for either vitamin D intake from food or supplements. With respect to time

spent outdoors, less than 6% of women had missing responses for each age period of exposure

reached. Logic checks were conducted and responses were deleted if unreasonable. In general,

cases with missing data were not included in the analyses – we did not impute values or employ

other techniques for missing data. Some assumptions were made regarding location of residence

(e.g., city of residence was assumed if only province or country was provided); however, this

was among only a small proportion of participants and sensitivity analyses were conducted to

evaluate the impact of these assumptions.

Multivariable logistic regression was used to model the association between vitamin D and

breast cancer risk and to obtain adjusted ORs. Even though cases and controls were frequency

matched on age, we included age in all our adjusted models to reduce the likelihood of residual

confounding. Other variables were defined as confounders if their inclusion in the model

changed main effect estimates by at least 10% (Maldonado & Greenland, 1993). Another

strategy would be to include all suspected confounders based upon an a priori conceptual model;

however, few variables were expected to fit the definition of a confounder (i.e., associated with

both the exposure and outcome). The goal of this model is to evaluate the association between

vitamin D and breast cancer risk not to develop a full predictive model of breast cancer risk,

thus, variables that did not substantially change the vitamin D and breast cancer effect estimate

were not included in the models.

Multiple comparisons may increase the likelihood of some findings being due to chance. With

statistical significance defined as P value less than 0.05 with 20 comparisons we would expect at

least one to be significant, thus, we cannot rule out the possibility that some of our findings are

due to chance. However, our results appear to be relatively consistent and biologically plausible

134

thus we would expect that our results are reliable. Many of our results are highly significant with

p-values much less than 0.05.

Logistic regression does not account for time-varying exposure and recently new techniques

have been proposed that use weighted Cox models with time-dependent exposures that have

been manipulated for use in case-control studies (L. S. Freedman, Oberman, & Sadetzki, 2009;

Leffondre, Abrahamowicz, & Siemiatycki, 2003; Leffondre et al., 2010). However, the data in

our study are limited by the set age periods asked in our questionnaire and evaluating the models

using these novel statistical procedures was beyond the scope of this thesis.

While the sample size for this study was already determined, post hoc power was calculated to

estimate the power to detect the main study objectives (appendix 5). The study had >80% power

to detect an independent main effect assuming at least a 20% risk reduction. The power to detect

an OR of 0.60 for the joint exposures assuming a multiplicative interaction between calcium and

vitamin D is 78%. A major strength of this study is the large sample size and ability to conduct

stratified analysis.

5.2.5 External Validity

In regards to the external validity of our study, the source population for this study was women

in the Ontario population and the findings are expected to be generalizeable beyond this

population. Patterns of vitamin D exposure/intake are likely relatively consistent across Canada

and the Northern US and, it is expected that the biologic mechanism proposed for vitamin D and

breast cancer is likely generalizeable to most women, thus, our findings are likely applicable to a

much broader population. However, our findings may not be generalizeable to women living

near the equator as they likely have higher vitamin D status all year-round and it has been

proposed that the continual fluctuation of high to low 25(OH)D levels may be more detrimental

than constant year-round levels (Vieth, 2009). Determinants of 25(OH)D status and the

proportion of vitamin D from food versus supplements likely varies by country (e.g., some

populations have high fish consumption) and thus our specific findings regarding oral intake

dose may not be generalizeable.

A major concern when considering the external validity of a study is whether the characteristics

of non-respondents (or the populations with whom you wish to generalize to) are similar to the

135

study population with respect to other characteristics that may modify the vitamin D-breast

cancer association (e.g., effect modifiers). When compared to Ontario data from Statistics

Canada our population is more highly educated and a greater proportion is Caucasian. Although

education may be associated with vitamin D intake it likely does not modify the association with

breast cancer risk. In contrast, ethnicity may be associated with genetic differences that could

modify the vitamin D breast cancer association. Considering our study was 90% Caucasian,

more studies should be conducted in other ethnic groups to determine if the findings are

generalizeable.

5.2.6 General Limitations

A limitation of our study is the lack of serum 25(OH)D measures. Blood samples were not

collected in this study and may not have been feasible given the large sample size and associated

costs and burden of collecting blood samples. Serum 25(OH)D is often considered the gold

standard for the measurement of vitamin D although it may not reflect the desired period of

exposure (Millen et al., 2008). If blood samples were available these would reflect post-diagnosis

25(OH)D levels among cases and may not be generalizeable to pre-diagnostic levels (as

discussed in section 2.3.3). However, if serum had been collected on even a subset of controls

then we could have validated the algorithm used to derive the solar vitamin D score.

Alternatively, we could have employed the strategy used in the Harvard Cohorts to develop a

predictive model of 25(OH)D specific to our population that could be applied to all study

participants (Giovannucci et al, 2006).

An additional limitation of this study is that we were unable to explore differences by breast

cancer stage or subtype (e.g., hormone receptor status) or to evaluate genetic interactions. As

discussed in Chapter 2, there is some evidence that vitamin D may influence disease progression

and there is conflicting evidence suggesting the vitamin D-breast cancer association may vary by

hormone receptor status of tumor. Saliva samples (DNA) were collected in this study and future

studies evaluating variants in vitamin D related genes are planned.

5.3 Study Strengths

Some major strengths of our study are the population-based identification of both cases and

controls, the high response-rate and the large sample size. While some believe case control

136

studies offer less convincing evidence compared to cohort or RCT studies, others argue that

well-designed case-control studies are essentially more efficient cohort studies or studies nested

within a cohort (Rothman & Greenland, 1998 p. 114). We made efforts to minimize common

threats to validity that are characteristic of case-control studies such as using population-based

cases and controls and recruiting cases soon after diagnosis (see section 5.2). Other general

advantages of a case-control study are feasibility, reduced costs and shorter time to complete.

An additional strength of our study is the measurement of vitamin D from all sources. Few

previous studies have includes all foods, supplements and such comprehensive sun exposure

measures. Our use of a validated FFQ and the additional modifications that were made to apply

the nutrient analysis to our Canadian population is a major strength of our study. With respect to

sun exposure, we were able to capture many variables that may affect the cutaneous synthesis of

vitamin D and our study collected these variables for multiple ages of exposure allowing us to

evaluate exposure during adolescence and other critical periods of exposure. Furthermore, our

solar vitamin D score is a novel tool and the methodology may be useful to other researchers.

Lastly, the comprehensiveness of the Ontario Women’s Diet and Health study allowed for the

evaluation of many possible confounders (e.g., calcium, physical activity) and effect modifiers.

5.4 Causation and Future Studies

As reviewed earlier, most studies, including ours, tend to suggest that vitamin D intake/exposure

is inversely associated with breast cancer risk, yet presently we do not have sufficient evidence

to conclude vitamin D prevents breast cancer. Establishing temporality and eliminating unknown

confounding are necessary to determine if there is a causal association. Furthermore, there are

many alternate explanations that have not yet been ruled out. For example, it has been proposed

that the observed associations between sun exposure and cancer risk may be due to changes in

melatonin, seasonal affective disorder, immunological effects, or degradation of folic acid by

UVB exposure (as reviewed in IARC, 2008). Evaluating these hypotheses was not possible

within our study, thus, we cannot determine if the observed associations between our solar

vitamin D score and time spent outdoors and breast cancer are due to vitamin D.

An experimental study of vitamin D from supplements and breast cancer risk is feasible and

warranted given the increasing body of inconclusive evidence from observational studies and the

137

noted limitations of previous trials. A well-designed randomized controlled trial specific to

supplemental vitamin D and breast cancer risk could establish temporality and reduce the

possibility of confounding by unmeasured variables. Such a trial should be conducted with a

sufficiently high vitamin D dose and monitoring of vitamin D levels to ensure serum 25(OH)D

levels in the intervention group are actually greater than control group despite contamination by

sun exposure. Previous trials (Chlebowski et al., 2008; Lappe et al., 2007) have all been

conducted with vitamin D and calcium and unable to tease out the independent effect of vitamin

D. Another option would be a prospective cohort with biomarkers collected at multiple time

periods. Additional studies, likely in the form of observational studies, are also needed to

evaluate the association between time spent outdoors and breast cancer risk and to explore the

aforementioned alternative hypotheses. Any future observational studies of dietary vitamin D

should be conducted in populations with a greater range in vitamin D intake and should include

supplements in addition to food sources.

There is only one study that reported a positive association, although non-significant, between

vitamin D intake and breast cancer risk and this was among postmenopausal women only (Lin et

al., 2007). Although these results were not significant, further study into the subgroup findings is

also needed. Future studies should be designed to evaluate menopausal status, breast cancer stage

at diagnosis, hormone receptor status and genetic variants such as polymorphisms in the vitamin

D receptor gene or vitamin D binding protein gene; the true association between vitamin D and

breast cancer risk may be masked if there are stratum specific differences that are not evaluated.

There is also conflicting results regarding critical period of exposure and future studies should

continue to evaluate vitamin D exposure during multiple periods of life. More generally, there is

also an urgent need for more research on the determinants of individual variation in vitamin D

status and a need for consensus on both serum 25(OH)D levels and recommended dietary intake.

5.5 Conclusions and Public Health Importance

Our research findings that vitamin D from supplements >400 IU/day and time spent outdoors are

inversely associated with breast cancer risk contribute to the growing body of evidence that

suggest vitamin D is associated with reduced breast cancer risk. Yet, even though there is an

established biologic mechanism and increasing body of literature supporting an association

between vitamin D and breast cancer risk, a causal association is not yet established. Low

138

vitamin D levels have generally been associated with breast cancer relative risk estimates of 1.3

to 2 (see chapter 2); although these risk estimates tend to suggest a modest association, the high

prevalence of exposure would result in a relatively large population attributable risk (PAR). For

example, if a true causal effect exists, and we assume a RR of 1.67 and 55% prevalence of low

vitamin D exposure then the PAR would be 25% (i.e., a quarter of all breast cancer cases can be

attributed to low vitamin D status). Despite the current uncertainty, this is an important area of

research, particularly important since breast cancer is the leading type of cancer among women

in Canada (Canadian Cancer Society/National Cancer Institute of Canada, 2009). Furthermore,

few established breast cancer risk factors are modifiable and vitamin D may be highly modifiable

and amenable to population prevention strategies.

Public and physician awareness surrounding the potential benefits of vitamin D seems to be

increasing. A recent review of billing data from the Ontario Ministry of Health and Long-Term

Care indicates that 25(OH)D testing has steadily increased over the past 5-years – from 56,900

tests in 2005 to a projected 696,162 tests in 2009 (Medical Advisory Secretariat, 2010). This

review, conducted by the Ontario Health Technology Advisory Committee, recommended that

Canadians should be advised to increase intake and supplements of vitamin D as per Health

Canada guidelines but “routine vitamin D testing is not warranted in the average risk

population”. Although there is not yet conclusive evidence of an association between vitamin D

and breast cancer risk, increasing vitamin D intake may prevent other diseases (e.g.,

osteoporosis, cardiovascular diseases, and type II diabetes) and prevention strategies to increase

vitamin D for other diseases may be warranted.

A population-based prevention strategy to increase vitamin D intake currently already exists in

Canada for the prevention of rickets and to maintain bone health: mandatory food fortification of

milk and margarine with vitamin D, and Canada’s Food Guide recommends a vitamin D

supplement of 400 IU/day for all Canadians over 50 years of age take since food levels are not

sufficient. Given the large margin of safety for vitamin D, such a strategy is feasible. Yet, in

spite of the current fortification practices, high levels of insufficiency continue to be observed

and concerns have been raised regarding Health Canada’s proposed discretionary fortification

policy (Sacco & Tarasuk, 2009). An alternative to food fortification may be a population-based

program to increase supplement use. Although sun exposure is a good source of vitamin D, it is

139

not likely a feasible prevention strategy to encourage increased sun exposure for 2 reasons: 1)

there is insufficient sun exposure most of the year in Canada for vitamin D synthesis, and 2) sun

exposure is a known skin cancer risk factor. Although the relatively short periods of sun

exposure needed for vitamin D production may not increase skin cancer risk, a message of

moderation may be challenging.

Our study contributes to the growing body of literature that suggests vitamin D may be

associated with reduced risk of breast cancer and explores many of the methodological issues

related to the measurement of vitamin D. Future studies are needed to further elucidate the

relationship between vitamin D and breast cancer risk and to inform public health policy. The

ultimate aim of this research is the primary prevention of breast cancer and the future looks

sunny.

140

References

Abbas, S., Chang-Claude, J., & Linseisen, J. (2009). Plasma 25-hydroxyvitamin D and premenopausal breast cancer risk in a German case-control study. International Journal

of Cancer, 124(1), 250-255.

Abbas, S., Linseisen, J., & Chang-Claude, J. (2007). Dietary vitamin D and calcium intake and premenopausal breast cancer risk in a German case-control study. Nutrition and Cancer,

59(1), 54-61.

Abbas, S., Linseisen, J., Slanger, T., Kropp, S., Mutschelknauss, E. J., Flesch-Janys, D., et al. (2008). Serum 25-hydroxyvitamin D and risk of post-menopausal breast cancer--results of a large case-control study. Carcinogenesis, 29(1), 93-99.

Al Sarakbi, W., Salhab, M., & Mokbel, K. (2005). Dairy products and breast cancer risk: A review of the literature. International Journal of Fertility and Women's Medicine, 50(6), 244-249.

Almquist, M., Manjer, J., Bondeson, L., & Bondeson, A. G. (2007). Serum calcium and breast cancer risk: Results from a prospective cohort study of 7,847 women. Cancer Causes &

Control, 18(6), 595-602.

Altekruse, S. F., Kosary, C. L., Krapcho, M., Neyman, N., Aminou, R., Waldron, W. et al. (eds). (2009) SEER Cancer Statistics Review, 1975-2007, National Cancer Institute. Bethesda, MD, http://seer.cancer.gov/csr/1975_2007/, based on November 2009 SEER data submission, posted to the SEER web site, 2010

Althuis, M. D., Fergenbaum, J. H., Garcia-Closas, M., Brinton, L. A., Madigan, M. P., & Sherman, M. E. (2004) Etiology of hormone receptor-defined breast cancer: a systematic review of the literature. Cancer Epidemiology Biomarkers Prevention., 13(10), 1558-68.

Althuis, M. D., Dozier, J. M., Anderson, W. F., Devesa, S. S., & Brinton, L. A. (2005). Global trends in breast cancer incidence and mortality 1973-1997. International Journal of

Epidemiology, 34(2), 405-412.

Anderson, L. N., Cotterchio, M., Vieth, R., & Knight, J. A. (2010). Vitamin D and calcium intakes and breast cancer risk in pre- and postmenopausal women. The American Journal

of Clinical Nutrition, 91(6), 1699-1707.

Andreeva, V. A., Unger, J. B., & Pentz, M. A. (2007). Breast cancer among immigrants: A systematic review and new research directions. Journal of Immigrant and Minority

Health / Center for Minority Public Health, 9(4), 307-322.

Armas, L. A., Hollis, B. W., & Heaney, R. P. (2004). Vitamin D2 is much less effective than vitamin D3 in humans. The Journal of Clinical Endocrinology and Metabolism, 89(11), 5387-5391.

Autier, P., & Gandini, S. (2007). Vitamin D supplementation and total mortality: A meta-analysis of randomized controlled trials. Archives of Internal Medicine, 167(16), 1730-1737.

141

Badger, T., Segrin, C., Dorros, S. M., Meek, P., & Lopez, A. M. (2007). Depression and anxiety in women with breast cancer and their partners. Nursing Research, 56(1), 44-53.

Barake, R., Weiler, H., Payette, H., & Gray-Donald, K. (2010). Vitamin D supplement consumption is required to achieve a minimal target 25-hydroxyvitamin D concentration of > or = 75 nmol/L in older people. The Journal of Nutrition, 140(3), 551-556.

Barger-Lux, M. J., & Heaney, R. P. (2002) Effects of above average summer sun exposure on serum 25-hydroxyvitamin D and calcium absorption. Journal of Clinical Endocrinology

& Metabolism, 87(11), 4952-6.

Bernstein, L., & Ross, R. K. (1993). Endogenous hormones and breast cancer risk. Epidemiologic Reviews, 15(1), 48-65.

Bertone-Johnson, E. R. (2007). Prospective studies of dietary vitamin D and breast cancer: More questions raised than answered. Nutrition Reviews, 65(10), 459-466.

Bertone-Johnson, E. R. (2009). Vitamin D and breast cancer. Annals of Epidemiology,

19(7):462-7

Bertone-Johnson, E. R., Chen, W. Y., Holick, M. F., Hollis, B. W., Colditz, G. A., Willett, W. C., et al. (2005). Plasma 25-hydroxyvitamin D and 1,25-dihydroxyvitamin D and risk of breast cancer. Cancer Epidemiology, Biomarkers & Prevention, 14(8), 1991-1997.

Berube, S., Diorio, C., Masse, B., Hebert-Croteau, N., Byrne, C., Cote, G., et al. (2005). Vitamin D and calcium intakes from food or supplements and mammographic breast density. Cancer Epidemiology, Biomarkers & Prevention, 14(7), 1653-1659.

Berube, S., Diorio, C., Verhoek-Oftedahl, W., & Brisson, J. (2004). Vitamin D, calcium, and mammographic breast densities. Cancer Epidemiology, Biomarkers & Prevention, 13(9), 1466-1472.

Binkley, N., Novotny, R., Krueger, D., Kawahara, T., Daida, Y. G., Lensmeyer, G., et al. (2007). Low vitamin D status despite abundant sun exposure. The Journal of Clinical

Endocrinology and Metabolism, 92(6), 2130-2135.

Bischoff-Ferrari, H. A. (2008). Optimal serum 25-hydroxyvitamin D levels for multiple health outcomes. Advances in Experimental Medicine and Biology, 624, 55-71.

Bissonauth, V., Shatenstein, B., & Ghadirian, P. (2008). Nutrition and breast cancer among sporadic cases and gene mutation carriers: An overview. Cancer Detection and

Prevention, 32(1), 52-64.

Blackmore, K. M., Lesosky, M., Barnett, H., Raboud, J. M., Vieth, R., & Knight, J. A. (2008). Vitamin D from dietary intake and sunlight exposure and the risk of hormone-receptor-defined breast cancer. American Journal of Epidemiology, 168(8), 915-924.

Bland, K. I., Menck, H. R., Scott-Conner, C. E., Morrow, M., Winchester, D. J., & Winchester, D. P. (1998). The national cancer data base 10-year survey of breast carcinoma treatment at hospitals in the united states. Cancer, 83(6), 1262-1273.

Block, G. (2001). Invited commentary: Another perspective on food frequency questionnaires. American Journal of Epidemiology, 154(12), 1103-4; discussion 1105-6.

142

Block, G., Hartman, A. M., Dresser, C. M., Carroll, M. D., Gannon, J., & Gardner, L. (1986). A data-based approach to diet questionnaire design and testing. American Journal of

Epidemiology, 124(3), 453-469.

Block, G., Woods, M., Potosky, A., & Clifford, C. (1990). Validation of a self-administered diet history questionnaire using multiple diet records. Journal of Clinical Epidemiology,

43(12), 1327-1335.

Boffetta, P., & Hashibe, M. (2006). Alcohol and cancer. Lancet Oncology, 7(2), 149-156.

Bolland, M. J., & Reid, I. R. (2008). Calcium supplementation and cancer incidence. The

American Journal of Clinical Nutrition, 87(3), 792-3; author reply 793-4.

Boucher, B., Cotterchio, M., Kreiger, N., Nadalin, V., Block, T., & Block, G. (2006). Validity and reliability of the Block98 food-frequency questionnaire in a sample of canadian women. Public Health Nutrition, 9(1), 84-93.

Boyd, N. F., Rommens, J. M., Vogt, K., Lee, V., Hopper, J. L., Yaffe, M. J., et al. (2005). Mammographic breast density as an intermediate phenotype for breast cancer. The Lancet

Oncology, 6(10), 798-808.

Brisson, J., Berube, S., Diorio, C., Sinotte, M., Pollak, M., & Masse, B. (2007). Synchronized seasonal variations of mammographic breast density and plasma 25-hydroxyvitamin d. Cancer Epidemiology, Biomarkers & Prevention , 16(5), 929-933.

Brody, J. G., Rudel, R. A., Michels, K. B., Moysich, K. B., Bernstein, L., Attfield, K. R., et al. (2007). Environmental pollutants, diet, physical activity, body size, and breast cancer: Where do we stand in research to identify opportunities for prevention? Cancer, 109(12 Suppl), 2627-2634.

Brot, C., Jorgensen, N. R., & Sorensen, O. H. (1999). The influence of smoking on vitamin D status and calcium metabolism. European Journal of Clinical Nutrition, 53(12), 920-926.

Brot, C., Vestergaard, P., Kolthoff, N., Gram, J., Hermann, A. P., & Sorensen, O. H. (2001). Vitamin D status and its adequacy in healthy Danish perimenopausal women: Relationships to dietary intake, sun exposure and serum parathyroid hormone. The British

Journal of Nutrition, 86 Suppl 1, S97-103.

Brustad, M., Alsaker, E., Engelsen, O., Aksnes, L., & Lund, E. (2004). Vitamin D status of middle-aged women at 65-71 degrees N in relation to dietary intake and exposure to ultraviolet radiation. Public Health Nutrition, 7(2), 327-335.

Buras, R. R., Schumaker, L. M., Davoodi, F., Brenner, R. V., Shabahang, M., Nauta, R. J., et al. (1994). Vitamin D receptors in breast cancer cells. Breast Cancer Research and

Treatment, 31(2-3), 191-202.

Burgaz, A., Akesson, A., Oster, A., Michaelsson, K., & Wolk, A. (2007). Associations of diet, supplement use, and ultraviolet B radiation exposure with vitamin D status in Swedish women during winter. The American Journal of Clinical Nutrition, 86(5), 1399-1404.

Byrdwell, W. C., Devries, J., Exler, J., Harnly, J. M., Holden, J. M., Holick, M. F., et al. (2008). Analyzing vitamin D in foods and supplements: Methodologic challenges. The American

Journal of Clinical Nutrition, 88(2), 554S-557S.

143

Calvo, M. S., Whiting, S. J., & Barton, C. N. (2004). Vitamin D fortification in the united states and canada: Current status and data needs. The American Journal of Clinical Nutrition,

80(6 Suppl), 1710S-6S.

Canadian Cancer Society. (2007). Canadian cancer society announces vitamin D

recommendation.http://www.cancer.ca/ccs/internet/mediareleaselist/0,3208,3172_1613121606_1997621989_langId-en,00.html

Canadian Cancer Society/National Cancer Institute of Canada. (2009). Canadian cancer

statistics 2009. Toronto, Canada. Retrieved 12/05, 2009 from http://www.cancer.ca/Canada-wide/About%20cancer/Cancer%20statistics/Canadian%20Cancer%20Statistics.aspx?sc_lang=en

Cantwell, M. M., Murray, L. J., Catney, D., Donnelly, D., Autier, P., Boniol, M., et al. (2009). Second primary cancers in patients with skin cancer: A population-based study in Northern Ireland. British Journal of Cancer, 100(1), 174-177.

Carroll, K. K., Jacobson, E. A., Eckel, L. A., & Newmark, H. L. (1991). Calcium and carcinogenesis of the mammary gland. The American Journal of Clinical Nutrition, 54(1 Suppl), 206S-208S.

Cashman, K. D., Hill, T. R., Lucey, A. J., Taylor, N., Seamans, K. M., Muldowney, S., et al. (2008). Estimation of the dietary requirement for vitamin D in healthy adults. The

American Journal of Clinical Nutrition, 88(6), 1535-1542.

Chen, J., Ruczinski, I., Jorgensen, T. J., Yenokyan, G., Yao, Y., Alani, R., et al. (2008). Nonmelanoma skin cancer and risk for subsequent malignancy. Journal of the National

Cancer Institute, 100(17), 1215-1222.

Chen, T. C., Chimeh, F., Lu, Z., Mathieu, J., Person, K. S., Zhang, A., et al. (2007). Factors that influence the cutaneous synthesis and dietary sources of vitamin D. Archives of

Biochemistry and Biophysics, 460(2), 213-217.

Chen, T. C., Shao, A., Heath, H.,3rd, & Holick, (1993). An update on the vitamin D content of fortified milk from the united states and canada. The New England Journal of Medicine,

329(20), 1507.

Chlebowski, R. T., Johnson, K. C., Kooperberg, C., Pettinger, M., Wactawski-Wende, J., Rohan, T., et al. (2008). Calcium plus vitamin D supplementation and the risk of breast cancer. Journal of the National Cancer Institute, 100(22), 1581-1591.

Chodick, G., Kleinerman, R., A., Linet, M. S., Fears, T., Kwok, R. K. et al. 2008). Agreement between diary records of time spent outdoors and personal ultraviolet radiation dose measurements. Photochemistry and Photobiology, 84(3), 713-8.

Claus, E. B., Schildkraut, J. M., Thompson, W. D., & Risch, N. J. (1996). The genetic attributable risk of breast and ovarian cancer. Cancer, 77(11), 2318-2324.

Cleary, M. P., & Grossmann, M. E. (2009). Minireview: Obesity and breast cancer: The estrogen connection. Endocrinology, 150(6), 2537-2542.

Colditz, G. A. (2005). Epidemiology and prevention of breast cancer. Cancer Epidemiology,

Biomarkers & Prevention, 14(4), 768-772.

144

Colditz, G. A., Atwood, K. A., Emmons, K., Monson, R. R., Willett, W. C., Trichopoulos, D., et al. (2000). Harvard report on cancer prevention volume 4: Harvard cancer risk index. Risk index working group, Harvard center for cancer prevention. Cancer Causes &

Control, 11(6), 477-488.

Colditz, G. A., Baer, H. J., & Tamimi, R. M. (2006). Breast cancer. In D. Schottenfeld, & J. F. Fraumeni (Eds.), Cancer epidemiology and prevention (3rd ed.). USA: Oxford University Press.

Colditz, G. A., & Frazier, A. L. (1995). Models of breast cancer show that risk is set by events of early life: Prevention efforts must shift focus. Cancer Epidemiology, Biomarkers &

Prevention , 4(5), 567-571.

Collaborative Group on Hormonal Factors in Breast Cancer. (2001). Familial breast cancer: Collaborative reanalysis of individual data from 52 epidemiological studies including 58,209 women with breast cancer and 101,986 women without the disease. Lancet,

358(9291), 1389-1399.

Colston, K. W. (2008). Vitamin D and breast cancer risk. Best Practice & Research.Clinical

Endocrinology & Metabolism, 22(4), 587-599.

Colston, K. W., & Hansen, C. M. (2002). Mechanisms implicated in the growth regulatory effects of vitamin D in breast cancer. Endocrine-Related Cancer, 9(1), 45-59.

Coombs, N. J., Cronin, K. A., Taylor, R. J., Freedman, A. N., & Boyages, J. (2010). The impact of changes in hormone therapy on breast cancer incidence in the US population. Cancer

Causes & Control, 21(1), 83-90.

Cotterchio, M., Boucher, B. A., Kreiger, N., Mills, C. A., & Thompson, L. U. (2008). Dietary phytoestrogen intake--lignans and isoflavones--and breast cancer risk (canada). Cancer

Causes & Control, 19(3), 259-272.

Cranney, A., Horsley, T., O'Donnell, S., Weiler, H., Puil, L., Ooi, D., et al. (2007). Effectiveness and safety of vitamin D in relation to bone health. Evidence report/technology

Assessment, (158)(158), 1-235.

Crew, K. D., Gammon, M. D., Steck, , Hershman, D. L., Cremers, S., Dworakowski, E., et al. (2009). Association between plasma 25-hydroxyvitamin D and breast cancer risk. Cancer

Prevention Research (Philadelphia, Pa.), 2(6), 598-604.

Csizmadi, I., Kahle, L., Ullman, R., Dawe, U., Zimmerman, T. P., Friedenreich, C. M., et al. (2007). Adaptation and evaluation of the national cancer institute's diet history questionnaire and nutrient database for Canadian populations. Public Health Nutrition,

10(1), 88-96.

Cui, Y., & Rohan, (2006). Vitamin D, calcium, and breast cancer risk: A review. Cancer

Epidemiology, Biomarkers & Prevention, 15(8), 1427-1437.

Deeb, K. K., Trump, D. L., & Johnson, C. S. (2007). Vitamin D signalling pathways in cancer: Potential for anticancer therapeutics. Nature Reviews.Cancer, 7(9), 684-700.

Elwood, M. (2007) Critical Appraisal of Epidemiological Studies and Clinical Trials. Oxford: Oxford University Press.

145

English, D. R., Armstrong, B. K., & Kricker, A. (1998). Reproducibility of reported measurements of sun exposure in a case-control study. Cancer Epidemiology, Biomarkers

& Prevention , 7(10), 857-863.

Faulkner, H., Hussein, A., Foran, M., & Szijarto, L. (2000). A survey of vitamin A and D contents of fortified fluid milk in ontario. Journal of Dairy Science, 83(6), 1210-1216.

Fioletov, V. E., McArthur, L. J., Mathews, T. W., & Marrett, L. (2009). On the relationship between erythemal and vitamin D action spectrum weighted ultraviolet radiation. Journal

of Photochemistry and Photobiology.B, Biology, 95(1), 9-16.

Fletcher, S. W., Black, W., Harris, R., Rimer, B. K., & Shapiro, S. (1993). Report of the international workshop on screening for breast cancer. Journal of the National Cancer

Institute, 85(20), 1644-1656.

Forman, J. P., Giovannucci, E., Holmes, M. D., Bischoff-Ferrari, H. A., Tworoger, S. S., Willett, W. C., et al. (2007). Plasma 25-hydroxyvitamin D levels and risk of incident hypertension. Hypertension, 49(5), 1063-1069.

Frazier, A. L., Li, L., Cho, E., Willett, W. C., & Colditz, G. A. (2004). Adolescent diet and risk of breast cancer. Cancer Causes & Control, 15(1), 73-82.

Frazier, A. L., Ryan, C. T., Rockett, H., Willett, W. C., & Colditz, G. A. (2003). Adolescent diet and risk of breast cancer. Breast Cancer Research : BCR, 5(3), R59-64.

Freedman, D. M., Chang, S. C., Falk, R. T., Purdue, M. P., Huang, W. Y., McCarty, C. A., et al. (2008). Serum levels of vitamin D metabolites and breast cancer risk in the prostate, lung, colorectal, and ovarian cancer screening trial. Cancer Epidemiology, Biomarkers &

Prevention, 17(4), 889-894.

Freedman, D. M., Dosemeci, M., & McGlynn, K. (2002). Sunlight and mortality from breast, ovarian, colon, prostate, and non-melanoma skin cancer: A composite death certificate based case-control study. Occupational and Environmental Medicine, 59(4), 257-262.

Freedman, D. M., Looker, A. C., Chang, S. C., & Graubard, B. I. (2007). Prospective study of serum vitamin D and cancer mortality in the united states. Journal of the National Cancer

Institute, 99(21), 1594-1602.

Freedman, L. S., Oberman, B., & Sadetzki, S. (2009). Using time-dependent covariate analysis to elucidate the relation of smoking history to warthin's tumor risk. American Journal of

Epidemiology, 170(9), 1178-1185.

Freudenheim, J. L., Marshall, J. R., Graham, S., Laughlin, R., Vena, J. E., Bandera, E., et al. (1994). Exposure to breastmilk in infancy and the risk of breast cancer. Epidemiology

(Cambridge, Mass.), 5(3), 324-331.

Friedenreich, C. M., & Cust, A. E. (2008). Physical activity and breast cancer risk: Impact of timing, type and dose of activity and population subgroup effects. British Journal of

Sports Medicine, 42(8), 636-647.

Garcia-Closas, M., & Lubin, J. H. (1999). Power and sample size calculations in case-control studies of gene-environment interactions: Comments on different approaches. American

Journal of Epidemiology, 149(8), 689-692.

146

Garland, C. F., Garland, F. C., Gorham, E. D., Lipkin, M., Newmark, H., Mohr, S. B., et al. (2006). The role of vitamin D in cancer prevention. American Journal of Public Health,

96(2), 252-261.

Garland, F. C., Garland, C. F., Gorham, E. D., & Young, J. F. (1990). Geographic variation in breast cancer mortality in the United States: A hypothesis involving exposure to solar radiation. Preventive Medicine, 19(6), 614-622.

Geocoder. (2007). Geocoding for north america - USA and Canada.http://geocoder.ca

Giovannucci, E. (2005). The epidemiology of vitamin D and cancer incidence and mortality: A review (United States). Cancer Causes & Control, 16(2), 83-95.

Giovannucci, E. (2008). Vitamin D status and cancer incidence and mortality. Advances in

Experimental Medicine and Biology, 624, 31-42.

Giovannucci, E., Liu, Y., Rimm, E. B., Hollis, B. W., Fuchs, C. S., Stampfer, M. J., et al. (2006). Prospective study of predictors of vitamin D status and cancer incidence and mortality in men. Journal of the National Cancer Institute, 98(7), 451-459.

Giovannucci, E., Stampfer, M. J., Colditz, G. A., Manson, J. E., Rosner, B. A., Longnecker, M. P., et al. (1993). Recall and selection bias in reporting past alcohol consumption among breast cancer cases. Cancer Causes & Control, 4(5), 441-448.

Gissel, T., Rejnmark, L., Mosekilde, L., & Vestergaard, P. (2008). Intake of vitamin D and risk of breast cancer--a meta-analysis. The Journal of Steroid Biochemistry and Molecular

Biology, 111(3-5), 195-199.

Goodwin, P. J. (2009). Vitamin D in cancer patients: Above all, do no harm. Journal of Clinical

Oncology : Official Journal of the American Society of Clinical Oncology, 27(13), 2117-2119.

Goodwin, P. J., Ennis, M., Pritchard, K. I., Koo, J., & Hood, N. (2009). Prognostic effects of 25-hydroxyvitamin D levels in early breast cancer. Journal of Clinical Oncology : Official

Journal of the American Society of Clinical Oncology, 27(23), 3757-3763.

Gozdzik, A., Barta, J. L., Wu, H., Wagner, D., Cole, D. E., Vieth, R., et al. (2008). Low wintertime vitamin D levels in a sample of healthy young adults of diverse ancestry living in the Toronto area: Associations with vitamin D intake and skin pigmentation. BMC Public Health, 8, 336.

Grant, W. B. (2002a). An ecologic study of dietary and solar ultraviolet-B links to breast carcinoma mortality rates. Cancer, 94(1), 272-281.

Grant, W. B. (2002b). An estimate of premature cancer mortality in the U.S. due to inadequate doses of solar ultraviolet-B radiation. Cancer, 94(6), 1867-1875.

Grant, W. B. (2007). A meta-analysis of second cancers after a diagnosis of nonmelanoma skin cancer: Additional evidence that solar ultraviolet-B irradiance reduces the risk of internal cancers. The Journal of Steroid Biochemistry and Molecular Biology, 103(3-5), 668-674.

Greenland, S. (1996). Basic methods for sensitivity analysis of biases. International Journal of

Epidemiology. 25(6):1107-16.

147

Hall, L. M., Kimlin, M. G., Aronov, P. A., Hammock, B. D., Slusser, J. R., Woodhouse, L. R., et al. (2010). Vitamin D intake needed to maintain target serum 25-hydroxyvitamin D concentrations in participants with low sun exposure and dark skin pigmentation is substantially higher than current recommendations. The Journal of Nutrition, 140(3), 542-550.

Hall S, Schulze K, Groome P, Mackillop W, Holowaty E. Using cancer registry data for survival studies: the example of the Ontario Cancer Registry. J Clin Epidemiol 2006;59:67-76.

Hamajima, N., Hirose, K., Tajima, K., Rohan, T., Calle, E. E., Heath, C. W.,Jr, et al. (2002). Alcohol, tobacco and breast cancer--collaborative reanalysis of individual data from 53 epidemiological studies, including 58,515 women with breast cancer and 95,067 women without the disease. British Journal of Cancer, 87(11), 1234-1245.

Hankinson, S., & Eliassen, A. H. (2007). Endogenous estrogen, testosterone and progesterone levels in relation to breast cancer risk. The Journal of Steroid Biochemistry and

Molecular Biology, 106(1-5), 24-30.

Hanwell, H., Vieth, R., Cole, D., Scillitani, A., Modoni, S., Frusciante, V., et al. (2010). Sun exposure questionnaire predicts circulating 25-hydroxyvitamin D concentrations in caucasian hospital workers in southern italy. The Journal of Steroid Biochemistry and

Molecular Biology, Epub ahead of print.

Harris, S. S., & Dawson-Hughes, B. (1998). The association of oral contraceptive use with plasma 25-hydroxyvitamin D levels. Journal of the American College of Nutrition, 17(3), 282-284.

Harris, S. S., & Dawson-Hughes, B. (2007). Reduced sun exposure does not explain the inverse association of 25-hydroxyvitamin d with percent body fat in older adults. The Journal of

Clinical Endocrinology and Metabolism, 92(8), 3155-3157.

Hartmann, L. C., Sellers, T. A., Frost, M. H., Lingle, W. L., Degnim, A. C., Ghosh, K., et al. (2005). Benign breast disease and the risk of breast cancer. The New England Journal of

Medicine, 353(3), 229-237.

Hathcock, J. N., Shao, A., Vieth, R., & Heaney, R. (2007). Risk assessment for vitamin D. The

American Journal of Clinical Nutrition, 85(1), 6-18.

Health Canada. (2005). Addition of vitamins and minerals to foods, 2005. Retrieved 2009-03-28, 2009, from http://www.hc-sc.gc.ca/fn-an/nutrition/vitamin/fortification_final_doc_1-eng.php#c6

Health Canada. (2006). Dietary reference intakes tables. Retrieved 3/23/2009, 2009, from http://www.hc-sc.gc.ca/fn-an/nutrition/reference/table/index-eng.php

Health Canada. (2007). Canadian nutrient file, 2007b version. www.healthcanada.ca/cnf

Heaney, R. P. (2008). Vitamin D and calcium interactions: Functional outcomes. The American

Journal of Clinical Nutrition, 88(2), 541S-544S.

Heaney, R. P., Horst, R. L., Cullen, D. M., & Armas, L. A. G. (2009). Vitamin D3 distribution and status in the body. Journal of the American College of Nutrition, 28(3), 252-256.

Heikkinen, A., Parviainen, M. T., Tuppurainen, M. T., Niskanen, L., Komulainen, M. H., & Saarikoski, S. (1998). Effects of postmenopausal hormone replacement therapy with and

148

without vitamin D3 on circulating levels of 25-hydroxyvitamin D and 1,25-dihydroxyvitamin D. Calcified Tissue International, 62(1), 26-30.

Hewison, M., Burke, F., Evans, K. N., Lammas, D. A., Sansom, , Liu, P., et al. (2007). Extra-renal 25-hydroxyvitamin D3-1alpha-hydroxylase in human health and disease. The

Journal of Steroid Biochemistry and Molecular Biology, 103(3-5), 316-321.

Hewison, M., Zehnder, D., Bland, R., & Stewart, P. M. (2000). 1alpha-hydroxylase and the action of vitamin D. Journal of Molecular Endocrinology, 25(2), 141-148.

Hofmann, J. N., Yu, K., Horst, R. L., Hayes, R. B., & Purdue, M. P. (2010). Long-term variation in serum 25-hydroxyvitamin D concentration among participants in the prostate, lung, colorectal, and ovarian cancer screening trial. Cancer Epidemiology, Biomarkers &

Prevention , 4, 927-31

Holden, J. M., Lemar, L. E., & Exler, J. (2008). Vitamin D in foods: Development of the US department of agriculture database. The American Journal of Clinical Nutrition, 87(4), 1092S-6S.

Holick, M. F. (2002). Vitamin D: The underappreciated D-lightful hormone that is important for skeletal and cellular health. Curr Opin Endocrinol Diabetes, 9, 87-98.

Holick, M. F. (1987). Photosynthesis of vitamin D in the skin: Effect of environmental and life-style variables. Federation Proceedings, 46(5), 1876-1882.

Holick, M. F. (1995). Environmental factors that influence the cutaneous production of vitamin D. The American Journal of Clinical Nutrition, 61(3 Suppl), 638S-645S.

Holick, M. F. (2003). Vitamin D: A millenium perspective. Journal of Cellular Biochemistry,

88(2), 296-307.

Holick, M. F. (2004).Vitamin D: importance in the prevention of cancers, type 1 diabetes, heart disease, and osteoporosis. American Journal of Clinical Nutrition, 79, 362–71.

Holick, M. F. (2006). Vitamin D: Its role in cancer prevention and treatment. Progress in

Biophysics and Molecular Biology, 92(1), 49-59.

Holick, M. F. (2007). Vitamin D deficiency. The New England Journal of Medicine, 357(3), 266-281.

Holick, M. F. (2008). Vitamin D: A D-lightful health perspective. Nutrition Reviews, 66(10 Suppl 2), S182-94.

Holick, M. F. (2009). Vitamin D status: Measurement, interpretation, and clinical application. Annals of Epidemiology, 19(2), 73-78.

Holick, M. F., Biancuzzo, R. M., Chen, , Klein, E. K., Young, A., Bibuld, D., et al. (2008). Vitamin D2 is as effective as vitamin D3 in maintaining circulating concentrations of 25-hydroxyvitamin D. The Journal of Clinical Endocrinology and Metabolism, 93(3), 677-681.

Holick, M. F., MacLaughlin, J. A., Clark, M. B., Holick, S. A., Potts, J. T.,Jr, Anderson, R. R., et al. (1980). Photosynthesis of previtamin D3 in human skin and the physiologic consequences. Science, 210(4466), 203-205.

149

Hollis, B. W. (2005). Circulating 25-hydroxyvitamin D levels indicative of vitamin D sufficiency: Implications for establishing a new effective dietary intake recommendation for vitamin D. The Journal of Nutrition, 135(2), 317-322.

Houghton, L. A., & Vieth, R. (2006). The case against ergocalciferol (vitamin D2) as a vitamin supplement. The American Journal of Clinical Nutrition, 84(4), 694-697.

Hunter, D. J., & Willett, W. C. (1993). Diet, body size, and breast cancer. Epidemiologic

Reviews, 15(1), 110-132.

Institute of Medicine National Academy of Sciences. (2009). Dietary reference intakes for

vitamin D and calcium. Retrieved 3/23, 2009, from http://www.iom.edu/?id=61170

International Agency for Research on Cancer (IARC). (2008). Vitamin D and cancer / a report

of the IARC working group on vitamin D No. Volume 5). Geneva, Switzerland: World Health Organization Press. Retrieved 5/28, 2010, from http://www.iarc.fr/en/publications/pdfs-online/wrk/wrk5/index.php

Janowsky, E. C., Lester, G. E., Weinberg, C. R., Millikan, R. C., Schildkraut, J. M., Garrett, P. A., et al. (1999). Association between low levels of 1,25-dihydroxyvitamin D and breast cancer risk. Public Health Nutrition, 2(3), 283-291.

John, E. M., Schwartz, G. G., Dreon, , & Koo, J. (1999). Vitamin D and breast cancer risk: The NHANES I epidemiologic follow-up study, 1971-1975 to 1992. National health and nutrition examination survey. Cancer Epidemiology, Biomarkers & Prevention , 8(5), 399-406.

John, E. M., Schwartz, G. G., Koo, J., Wang, W., & Ingles, S. A. (2007). Sun exposure, vitamin D receptor gene polymorphisms, and breast cancer risk in a multiethnic population. American Journal of Epidemiology, 166(12), 1409-19.

Johnson, K. C., & Glantz, S. A. (2008) Evidence secondhand smoke causes breast cancer in 2005 stronger than for lung cancer in 1986. Preventive Medicine, 46(6), 492-6.

Jorde, R., Sneve, M., Hutchinson, M., Emaus, N., Figenschau, Y., & Grimnes, G. (2010). Tracking of serum 25-hydroxyvitamin D levels during 14 years in a population-based study and during 12 months in an intervention study. American Journal of Epidemiology,

171(8), 903-8.

Kabat, G. C., Jones, J. G., Olson, N., Negassa, A., Duggan, C., Ginsberg, M., et al. (2010). A multi-center prospective cohort study of benign breast disease and risk of subsequent breast cancer. Cancer Causes & Control, 21(6), 821-8.

Karagas, M. R., Zens, M. S., Nelson, H. H., Mabuchi, K., Perry, A. E., Stukel, T. A., et al. (2007). Measures of cumulative exposure from a standardized sun exposure history questionnaire: A comparison with histologic assessment of solar skin damage. American

Journal of Epidemiology, 165(6), 719-726.

Kelsey, J. L., & Berkowitz, G. S. (1988). Breast cancer epidemiology. Cancer Research, 48(20), 5615-5623.

Kelsey, J. L., & Bernstein, L. (1996). Epidemiology and prevention of breast cancer. Annual

Review of Public Health, 17, 47-67.

150

Kelsey, J. L., Gammon, M. D., & John, E. M. (1993). Reproductive factors and breast cancer. Epidemiologic Reviews, 15(1), 36-47.

Kesse-Guyot, E., Bertrais, S., Duperray, B., Arnault, N., Bar-Hen, A., Galan, P., et al. (2007). Dairy products, calcium and the risk of breast cancer: Results of the French SU.VI.MAX prospective study. Annals of Nutrition & Metabolism, 51(2), 139-145.

Key, J., Hodgson, S., Omar, R. Z., Jensen, T. K., Thompson, S. G., Boobis, A. R., et al. (2006) Meta-analysis of studies of alcohol and breast cancer with consideration of the methodological issues. Cancer Causes Control, 17, 759–770.

Khan, N., Yang, K., Newmark, H., Wong, G., Telang, N., Rivlin, R., et al. (1994). Mammary ductal epithelial cell hyperproliferation and hyperplasia induced by a nutritional stress diet containing four components of a western-style diet. Carcinogenesis, 15(11), 2645-2648.

Kim, J. H., & Moon, S. J. (2000). Time spent outdoors and seasonal variation in serum concentrations of 25-hydroxyvitamin D in Korean women. International Journal of Food

Sciences and Nutrition, 51(6), 439-451.

Kimlin, M. G., Downs, N. J., & Parisi, A. V. (2003). Comparison of human facial UV exposure at high and low latitudes and the potential impact on dermal vitamin D production. Photochem.Photobiol.Sci., 2(4), 370-375.

Kliewer, E. V., Demers, A. A., & Nugent, Z. J. (2007). A decline in breast-cancer incidence. The

New England Journal of Medicine, 357(5), 509-10; author reply 513.

Kliewer, E. V., & Smith, K. R. (1995). Breast cancer mortality among immigrants in Australia and Canada. Journal of the National Cancer Institute, 87(15), 1154-1161.

Knight, J. A., Lesosky, M., Barnett, H., Raboud, J. M., & Vieth, R. (2007). Vitamin D and reduced risk of breast cancer: A population-based case-control study. Cancer

Epidemiology, Biomarkers & Prevention , 16(3), 422-429.

Knight, J. A., Vachon, C. M., Vierkant, R. A., Vieth, R., Cerhan, J. R., & Sellers, T. A. (2006). No association between 25-hydroxyvitamin d and mammographic density. Cancer

Epidemiology, Biomarkers & Prevention , 15(10), 1988-1992.

Kopec, J. A., & Esdaile, J. M. (1990). Bias in case-control studies. A review. Journal of

Epidemiology and Community Health, 44(3), 179-186.

Kricker, A., Vajdic, C. M., & Armstrong, B. K. (2005). Reliability and validity of a telephone questionnaire for estimating lifetime personal sun exposure in epidemiologic studies. Cancer Epidemiology, Biomarkers & Prevention, 14(10), 2427-2432.

Krishnan, A. V., Swami, S., & Feldman, D. (2010). Vitamin D and breast cancer: Inhibition of estrogen synthesis and signaling. The Journal of Steroid Biochemistry and Molecular

Biology, Epub ahead of print.

Kristal, A. R., Peters, U., & Potter, J. D. (2005). Is it time to abandon the food frequency questionnaire? Cancer Epidemiology, Biomarkers & Prevention, 14(12), 2826-2828.

Kuper, H., Yang, L., Sandin, S., Lof, M., Adami, H. O., & Weiderpass, E. (2009). Prospective study of solar exposure, dietary vitamin D intake, and risk of breast cancer among middle-aged women. Cancer Epidemiology, Biomarkers & Prevention ,18(9), 2558-61.

151

Langlois, K., Greene-Finestone, L., Little, J., Hidiroglou, N., & Whiting, S. (March 2010). Vitamin D status of canadians as measured in the 2007 to 2009 Canadian health measures survey, catalogue no. 82-003-XPE. Statistics Canada Health Reports, 21(1).

Lappe, J. M., Travers-Gustafson, D., Davies, K. M., Recker, R. R., & Heaney, R. P. (2007). Vitamin D and calcium supplementation reduces cancer risk: Results of a randomized trial. The American Journal of Clinical Nutrition, 85(6), 1586-1591.

Larsson, S. C., Bergkvist, L., & Wolk, A. (2009). Long-term dietary calcium intake and breast cancer risk in a prospective cohort of women. The American Journal of Clinical

Nutrition, 89(1), 277-282.

Lauzier, S., Maunsell, E., Drolet, M., Coyle, D., Hebert-Croteau, N., Brisson, J., et al. (2008). Wage losses in the year after breast cancer: Extent and determinants among Canadian women. Journal of the National Cancer Institute, 100(5), 321-332.

Lauzier, S., Maunsell, E., Levesque, P., Mondor, M., Robert, J., Robidoux, A., et al. (2009). Psychological distress and physical health in the year after diagnosis of DCIS or invasive breast cancer. Breast Cancer Research and Treatment, 120(3), 685-91.

Lee, P. N., & Hamling, J. (2006). Environmental tobacco smoke exposure and risk of breast cancer in nonsmoking women: A review with meta-analyses. Inhalation Toxicology,

18(14), 1053-1070.

Leffondre, K., Abrahamowicz, M., & Siemiatycki, J. (2003). Evaluation of Cox’s model and logistic regression for matched case-control data with time-dependent covariates: A simulation study. Statistics in Medicine, 22(24), 3781-3794.

Leffondre, K., Wynant, W., Cao, Z., Abrahamowicz, M., Heinze, G., & Siemiatycki, J. (2010). A weighted cox model for modelling time-dependent exposures in the analysis of case-control studies. Statistics in Medicine, 29(7-8), 839-850.

Levi, F., Pasche, C., Lucchini, F., & La Vecchia, C. (2001). Dietary intake of selected micronutrients and breast-cancer risk. International Journal of Cancer, 91(2), 260-263.

Levi, F., Randimbison, L., Te, V. C., Conconi, M. M., & La Vecchia, C. (2008). Risk of prostate, breast and colorectal cancer after skin cancer diagnosis. International Journal of Cancer,

123(12), 2899-2901.

Lichtenstein, P., Holm, N. V., Verkasalo, P. K., Iliadou, A., Kaprio, J., Koskenvuo, M., et al. (2000). Environmental and heritable factors in the causation of cancer--analyses of cohorts of twins from Sweden, Denmark, and Finland. The New England Journal of

Medicine, 343(2), 78-85.

Lin, J., Manson, J. E., Lee, I. M., Cook, N. R., Buring, J. E., & Zhang, S. M. (2007). Intakes of calcium and vitamin d and breast cancer risk in women. Archives of Internal Medicine,

167(10), 1050-1059.

Linos, E., Holmes, M. D., & Willett, W. C. (2007). Diet and breast cancer. Current Oncology

Reports, 9(1), 31-41.

Linos, E., Willett, W. C., Cho, E., Frazier, L. (2010) Adolescent diet in relation to breast cancer risk among premenopausal women. Cancer Epidemiology Biomarkers Prevention. 19(3), 689-96.

152

Lipkin, M., & Newmark, H. L. (1999). Vitamin D, calcium and prevention of breast cancer: A review. Journal of the American College of Nutrition, 18(5 Suppl), 392S-397S.

Lipskie, T. L. (1998). A summary of cancer screening guidelines. Chronic Diseases in Canada,

19(3), 112-130.

Lipworth, L., Bailey, L. R., & Trichopoulos, D. (2000). History of breast-feeding in relation to breast cancer risk: A review of the epidemiologic literature. Journal of the National

Cancer Institute, 92(4), 302-312.

Lof, M., & Weiderpass, E. (2009). Impact of diet on breast cancer risk. Current Opinion in

Obstetrics & Gynecology, 21(1), 80-85.

Lowe, L. C., Guy, M., Mansi, J. L., Peckitt, C., Bliss, J., Wilson, R. G., et al. (2005). Plasma 25-hydroxy vitamin D concentrations, vitamin D receptor genotype and breast cancer risk in a UK caucasian population. European Journal of Cancer (Oxford, England : 1990),

41(8), 1164-1169.

Lu, Z., Chen, T. C., Zhang, A., Persons, K. S., Kohn, N., Berkowitz, R., et al. (2007). An evaluation of the vitamin D3 content in fish: Is the vitamin D content adequate to satisfy the dietary requirement for vitamin D? The Journal of Steroid Biochemistry and

Molecular Biology, 103(3-5), 642-644.

MacLaughlin, J., & Holick, (1985). Aging decreases the capacity of human skin to produce vitamin D3. The Journal of Clinical Investigation, 76(4), 1536-1538.

MacLaughlin, J. A., Anderson, R. R., & Holick, (1982). Spectral character of sunlight modulates photosynthesis of previtamin D3 and its photoisomers in human skin. Science (New York,

N.Y.), 216(4549), 1001-1003.

Maheu, C. (2009). Implications of living with a strong family history of breast cancer. The

Canadian Journal of Nursing Research, 41(2), 100-112.

Maldonado, G., & Greenland, S. (1993). Simulation study of confounder-selection strategies. American Journal of Epidemiology, 138(11), 923-936.

Martin, A. M., & Weber, B. L. (2000). Genetic and hormonal risk factors in breast cancer. Journal of the National Cancer Institute, 92(14), 1126-1135.

Martin, R. M., Middleton, N., Gunnell, D., Owen, C. G., & Smith, G. D. (2005). Breast-feeding and cancer: The boyd orr cohort and a systematic review with meta-analysis. Journal of

the National Cancer Institute, 97(19), 1446-1457.

Matsuoka, L. Y., Ide, L., Wortsman, J., MacLaughlin, J. A., & Holick, (1987). Sunscreens suppress cutaneous vitamin D3 synthesis. The Journal of Clinical Endocrinology and

Metabolism, 64(6), 1165-1168.

McCullough, M. L., Bostick, R. M., & Mayo, T. L. (2009). Vitamin D gene pathway polymorphisms and risk of colorectal, breast, and prostate cancer. Annual Review of

Nutrition, 29, 111-132.

McCullough, M. L., Rodriguez, C., Diver, W. R., Feigelson, H. S., Stevens, V. L., Thun, M. J., et al. (2005). Dairy, calcium, and vitamin D intake and postmenopausal breast cancer risk in the cancer prevention study II nutrition cohort. Cancer Epidemiology, Biomarkers &

Prevention, 14(12), 2898-2904.

153

McCullough, M. L., Stevens, V. L., Patel, R., Jacobs, E. J., Bain, E. B., Horst, R. L., et al. (2009). Serum 25-hydroxyvitamin D concentrations and postmenopausal breast cancer risk: A nested case control study in the cancer prevention study-II nutrition cohort. Breast

Cancer Research : BCR, 11(4), R64.

McGrath, C. M., & Soule, H. D. (1984). Calcium regulation of normal human mammary epithelial cell growth in culture. In Vitro, 20(8), 652-662.

McKay, J. D., McCullough, M. L., Ziegler, R. G., Kraft, P., Saltzman, B. S., Riboli, E., et al. (2009). Vitamin D receptor polymorphisms and breast cancer risk: Results from the national cancer institute breast and prostate cancer cohort consortium. Cancer

Epidemiology, Biomarkers & Prevention, 18(1), 297-305.

McKinley, A., & Deffey, B. L. (1987). A reference action spectrum for ultraviolet induced erythema in human skin. In W. F. Passchier, & B. F. M. Bosnjakovich (Eds.), Human

exposure to ultraviolet radiation: Risks and regulations (pp. 83-87) International Congress Series.

Medical Advisory Secretariat. (2010 Feb). Clinical utility of vitamin D testing: An evidence-

based analysis. ontario health technology assessment series 2010; vol. 10, no. 2. Retrieved 04/01, 2010, from http://www.health.gov.on.ca/english/providers/program/ mas/tech/reviews/pdf/rev_vitamin d_201002.pdf

Medline Plus Merriam-Webster. (2009). Medical dictionary. Retrieved Jan/10, 2010, from http://www.nlm.nih.gov.myaccess.library.utoronto.ca/medlineplus/mplusdictionary.html

Mettlin, C. (1992). Breast cancer risk factors. contributions to planning breast cancer control. Cancer, 69(7 Suppl), 1904-1910.

Michels, K. B., Mohllajee, A. P., Roset-Bahmanyar, E., Beehler, G. P., & Moysich, K. B. (2007). Diet and breast cancer: A review of the prospective observational studies. Cancer, 109(12 Suppl), 2712-2749.

Millen, A. E., & Bodnar, L. M. (2008). Vitamin D assessment in population-based studies: A review of the issues. The American Journal of Clinical Nutrition, 87(4), 1102S-5S.

Millen, A. E., Pettinger, M., Freudenheim, J. L., Langer, R. D., Rosenberg, C. A., Mossavar-Rahmani, Y., et al. (2009). Incident invasive breast cancer, geographic location of residence, and reported average time spent outside. Cancer Epidemiology, Biomarkers &

Prevention, 18(2), 495-507.

Miller, M. D., Marty, M. A., Broadwin, R., Johnson, K. C., Salmon, A. G., Winder, B., et al. (2007). The association between exposure to environmental tobacco smoke and breast cancer: A review by the California environmental protection agency. Preventive

Medicine, 44(2), 93-106.

Mishra, G., McCormack, V., Kuh, D., Hardy, R., Stephen, A., & dos Santos Silva, I. (2008). Dietary calcium and vitamin D intakes in childhood and throughout adulthood and mammographic density in a British birth cohort. British Journal of Cancer, 99(9), 1539-1543.

154

Mohr, S. B., Garland, C. F., Gorham, E. D., Grant, W. B., & Garland, (2008). Relationship between low ultraviolet B irradiance and higher breast cancer risk in 107 countries. The

Breast Journal, 14(3), 255-260.

Moore, C., Murphy, M. M., Keast, D. R., & Holick, (2004). Vitamin D intake in the United States. Journal of the American Dietetic Association, 104(6), 980-983.

Moorman, P. G., & Terry, P. D. (2004). Consumption of dairy products and the risk of breast cancer: A review of the literature. The American Journal of Clinical Nutrition, 80(1), 5-14.

Morabia, A. (2002). Smoking (active and passive) and breast cancer: Epidemiologic evidence up to june 2001. Environmental and Molecular Mutagenesis, 39(2-3), 89-95.

Mustacchi, P. (1961). Ramazzini and rigoni-stern on parity and breast cancer. clinical impression and statistical corroboration. Archives of Internal Medicine, 108, 639-642.

Narod, S. A. (2006). Modifiers of risk of hereditary breast cancer. Oncogene, 25(43), 5832-5836.

NASA Goddard Space Flight Center. (2007). Total ozone mapping spectrometer data product:

Erythemal UV exposure.http://toms.gsfc.nasa.gov/ery_uv/euv_v8.html

National Cancer Institute. (2009). Breast cancer treatment physician data query - health

professional version., 2009, from http://www.cancer.gov/cancertopics/pdq/treatment/breast

Need, A. G., Morris, H. A., Horowitz, M., & Nordin, C. (1993). Effects of skin thickness, age, body fat, and sunlight on serum 25-hydroxyvitamin D. The American Journal of Clinical

Nutrition, 58(6), 882-885.

Neilson, H. K., Friedenreich, C. M., Brockton, N. T., & Millikan, R. C. (2009). Physical activity and postmenopausal breast cancer: Proposed biologic mechanisms and areas for future research. Cancer Epidemiology, Biomarkers & Prevention , 18(1), 11-27.

Ng, K., Wolpin, B. M., Meyerhardt, J. A., Wu, K., Chan, A. T., Hollis, B. W., et al. (2009). Prospective study of predictors of vitamin D status and survival in patients with colorectal cancer. British Journal of Cancer, 101(6), 916-923.

Nichols, H. B., Trentham-Dietz, A., Sprague, B. L., Hampton, J. M., Titus-Ernstoff, L., & Newcomb, P. A. (2008). Effects of birth order and maternal age on breast cancer risk: Modification by whether women had been breast-fed. Epidemiology (Cambridge, Mass.),

19(3), 417-423.

Norman, A. W. (2008). From vitamin D to hormone D: Fundamentals of the vitamin D endocrine system essential for good health. The American Journal of Clinical Nutrition,

88(2), 491S-499S.

Norval, M., & Wulf, H. C. (2009). Does chronic sunscreen use reduce vitamin D production to insufficient levels? The British Journal of Dermatology, 161(4), 732-736.

Office of Dietary Supplements National Institutes of Health. Dietary supplement fact sheet:

Vitamin D. Retrieved 11/26, 2006, from http://dietary-supplements.info.nih.gov/factsheets/vitamind.asp

155

Ojha, R. P., Felini, M. J., & Fischbach, L. A. (2007). Vitamin D for cancer prevention: Valid assertion or premature anointment? The American Journal of Clinical Nutrition, 86(6), 1804-5; author reply 1805-6.

Okasha, M., McCarron, P., Gunnell, D., & Smith, G. D. (2003). Exposures in childhood, adolescence and early adulthood and breast cancer risk: A systematic review of the literature. Breast Cancer Research and Treatment, 78(2), 223-276.

O'Riordan, D.L., Glanz, K., Gies, P., Elliott, T. (2008) A pilot study of the validity of self-reported ultraviolet radiation exposure and sun protection practices among lifeguards, parents and children. Photochemistry and Photobiology 84(3):774-8.

Palmer, J. R., & Rosenberg, L. (1993). Cigarette smoking and the risk of breast cancer. Epidemiologic Reviews, 15(1), 145-156.

Palmieri, C., Macgregor, T., Girgis, S., & Vigushin, D. (2006). Serum 25 hydroxyvitamin D levels in early and advanced breast cancer. Journal of Clinical Pathology, 59(12),1334-6.

Parodi, P. W. (2005). Dairy product consumption and the risk of breast cancer. Journal of the

American College of Nutrition, 24(6 Suppl), 556S-68S.

Pathak, D. R. (2002). Dual effect of first full term pregnancy on breast cancer risk: Empirical evidence and postulated underlying biology. Cancer Causes & Control, 13(4), 295-298.

Perez-Lopez, F. R., Chedraui, P., & Haya, J. (2009). Review article: Vitamin D acquisition and breast cancer risk. Reproductive Sciences (Thousand Oaks, Calif.), 16(1), 7-19.

Pike, M. C., Krailo, M. D., Henderson, B. E., Casagrande, J. T., & Hoel, D. G. (1983). 'Hormonal' risk factors, 'breast tissue age' and the age-incidence of breast cancer. Nature,

303(5920), 767-770. Platek, M. E., Shields, P. G., Marian, C., McCann, S. E., & Bonner, M.R., et al. (2009). Alcohol consumption and genetic variation in methylenetetrahydrofolate reductase and 5-methyltetrahydrofolate-homocysteine methyltransferase in relation to breast cancer risk. Cancer

Epidemiol Biomarkers Prev, 18(9), 2453-9.

Poliquin, S., Joseph, L., & Gray-Donald, K. (2009). Calcium and vitamin D intakes in an adult Canadian population. Canadian Journal of Dietetic Practice and Research : A

Publication of Dietitians of Canada, 70(1), 21-27.

Pope, S. J., Holick, M. F., Mackin, S., & Godar, D. E. (2008). Action spectrum conversion factors that change erythemally weighted to previtamin D3-weighted UV doses. Photochemistry and Photobiology, 84(5), 1277-1283.

Porojnicu, A. C., Lagunova, Z., Robsahm, T. E., Berg, J. P., Dahlback, A., & Moan, J. (2006). Changes in risk of death from breast cancer with season and latitude: Sun exposure and breast cancer survival in norway. Breast Cancer Research and Treatment, 102(3):323-8

Prentice, R. L., Caan, B., Chlebowski, R. T., Patterson, R., Kuller, L. H., Ockene, J. K., et al. (2006). Low-fat dietary pattern and risk of invasive breast cancer: The women's health initiative randomized controlled dietary modification trial. JAMA: The Journal of the

American Medical Association, 295(6), 629-642.

156

Ravdin, P. M., Cronin, K. A., Howlader, N., Berg, C. D., Chlebowski, R. T., Feuer, E. J., et al. (2007). The decrease in breast-cancer incidence in 2003 in the United States. The New

England Journal of Medicine, 356(16), 1670-1674.

Ray, M., & Polite, B. N. (2010) Triple-negative breast cancers: a view from 10,000 feet. Cancer

Journal, 16(1), 17-22. Rejnmark, L., Tietze, A., Vestergaard, P., Buhl, L., Lehbrink, M., Heickendorff, L., et al. (2009).

Reduced prediagnostic 25-hydroxyvitamin D levels in women with breast cancer: A nested case-control study. Cancer Epidemiology, Biomarkers & Prevention , 18(10), 2655-2660.

Robien, K., Cutler, G. J., & Lazovich, D. (2007). Vitamin D intake and breast cancer risk in postmenopausal women: The Iowa women's health study. Cancer Causes & Control,

18(7), 775-782.

Robsahm, T. E., Tretli, S., Dahlback, A., & Moan, J. (2004). Vitamin D3 from sunlight may improve the prognosis of breast-, colon- and prostate cancer (Norway). Cancer Causes &

Control, 15(2), 149-158.

Rockhill, B., Weinberg, C. R., & Newman, B. (1998). Population attributable fraction estimation for established breast cancer risk factors: Considering the issues of high prevalence and unmodifiability. American Journal of Epidemiology, 147(9), 826-833.

Rohan, T. (2007). Epidemiological studies of vitamin D and breast cancer. Nutrition Reviews,

65(8 Pt 2), S80-3.

Rohan, T. E., Negassa, A., Chlebowski, R. T., Ceria-Ulep, C. D., Cochrane, B. B., Lane, D. S., et al. (2009). A randomized controlled trial of calcium plus vitamin D supplementation and risk of benign proliferative breast disease. Breast Cancer Research and Treatment,

116(2), 339-350.

Ronckers, C. M., Erdmann, C. A., & Land, C. E. (2005). Radiation and breast cancer: A review of current evidence. Breast Cancer Research : BCR, 7(1), 21-32.

Rossi, M., McLaughlin, J. K., Lagiou, P., Bosetti, C., Talamini, R., Lipworth, L., et al. (2009). Vitamin D intake and breast cancer risk: A case-control study in Italy. Annals of

Oncology : Official Journal of the European Society for Medical Oncology / ESMO,

20(2), 374-378.

Rossouw, J. E., Anderson, G. L., Prentice, R. L., LaCroix, A. Z., Kooperberg, C., Stefanick, M. L., et al. (2002). Risks and benefits of estrogen plus progestin in healthy postmenopausal women: Principal results from the women's health initiative randomized controlled trial. JAMA : The Journal of the American Medical Association, 288(3), 321-333.

Rothman, K. J., & Greenland, S. (1998). Modern epidemiology. (2nd ed., pp. 127). Philadelphia, PA: Lippincott-Raven.

Russo, J., Moral, R., Balogh, G. A., Mailo, D., & Russo, I. H. (2005). The protective role of pregnancy in breast cancer. Breast Cancer Research : BCR, 7(3), 131-142.

Sacco, J. E., & Tarasuk, V. (2009). Health Canada’s proposed discretionary fortification policy is misaligned with the nutritional needs of Canadians. The Journal of Nutrition, 139(10), 1980-1986.

157

Sahota, H., Barnett, H., Lesosky, M., Raboud, J. M., Vieth, R., & Knight, J. A. (2008). Association of vitamin d related information from a telephone interview with 25-hydroxyvitamin d. Cancer Epidemiology, Biomarkers & Prevention , 17(1), 232-238.

Schabas, R. (2008). Artifact in the control group undermines the conclusions of a vitamin D and cancer study. The American Journal of Clinical Nutrition, 87(3), 792; author reply 793-4.

Schatzkin, A., Subar, A. F., Moore, S., Park, Y., Potischman, N., Thompson, F. E., et al. (2009). Observational epidemiologic studies of nutrition and cancer: The next generation (with better observation). Cancer Epidemiology, Biomarkers & Prevention , 18(4), 1026-1032.

Schwartz, G. G., & Blot, W. J. (2006). Vitamin D status and cancer incidence and mortality: Something new under the sun. Journal of the National Cancer Institute, 98(7), 428-430.

Sergeev, I. N. (2004). Calcium as a mediator of 1,25-dihydroxyvitamin D3-induced apoptosis. The Journal of Steroid Biochemistry and Molecular Biology, 89-90(1-5), 419-425.

Serra-Majem, L., Andersen L. F., Henrique-Sanchez, P., Doreste-Alonso, J., Sanchez-Villegas, A., Ortiz-Andrelluchi, A., et al. (2009). Evaluating the quality of dietary intake validation studies. British Journal of Nutrition, 102, S3-S9.

Shields, M., & Wilkins, K. (2009). An update on mammography use in Canada. Health Reports /

Statistics Canada, Canadian Centre for Health Information, 20(3), 7-19.

Shin, M. H., Holmes, M. D., Hankinson, S., Wu, K., Colditz, G. A., & Willett, W. C. (2002). Intake of dairy products, calcium, and vitamin D and risk of breast cancer. Journal of the

National Cancer Institute, 94(17), 1301-1311.

Simard, A., Vobecky, J., & Vobecky, J. S. (1991). Vitamin D deficiency and cancer of the breast: An unprovocative ecological hypothesis. Canadian Journal of Public Health,

82(5), 300-303.

Singletary, K. W., & Gapstur, S. M. (2001). Alcohol and breast cancer: Review of epidemiologic and experimental evidence and potential mechanisms. JAMA: The Journal of the

American Medical Association, 286(17), 2143-2151.

Singletary, M. L., & Connolly, J. L. (2006). Breast cancer staging: Working with the sixth edition of the AJCC cancer staging manual. CA: A Cancer Journal for Clinicians, 56(1), 37-47.

Slattery, M. L. (2007). Vitamin D receptor gene (VDR) associations with cancer. Nutrition

Reviews, 65(8 Pt 2), S102-4.

Smith-Warner, S. A., Spiegelman, D., Yaun, S. S., van den Brandt, P. A., Folsom, A. R., Goldbohm, R. A., et al. (1998). Alcohol and breast cancer in women: A pooled analysis of cohort studies. JAMA: The Journal of the American Medical Association, 279(7), 535-540.

Soerjomataram, I., Louwman, W. J., Lemmens, V. E., Coebergh, J. W., & de Vries, E. (2008). Are patients with skin cancer at lower risk of developing colorectal or breast cancer?

American Journal of Epidemiology, 167(12), 1421-1429.

Sood, M. M., & Sood, A. R. (2007). Dietary vitamin D and decreases in cancer rates: Canada as the national experiment. The American Journal of Clinical Nutrition, 86(5), 1549; author reply 1549-50.

158

Stamp, , Haddad, J. G., & Twigg, C. A. (1977). Comparison of oral 25-hydroxycholecalciferol, vitamin D, and ultraviolet light as determinants of circulating 25-hydroxyvitamin D. Lancet, 1(8026), 1341-1343.

Statistics Canada. Canadian community health survey, cycle 2.2, nutrition (2004)-nutrient

intakes from food: Provincial, regional and national summary data tables, volume 1,2

and 3

Statistics Canada. (2006) Residential Telephone Service Survey. Retrieved May 2010 from: http://www.statcan.gc.ca/daily-quotidien/070504/dq070504a-eng.htm

Steering Committee on Clinical Practice Guidelines for the Care and Treatment of Breast faCancer. (1998). Clinical guidelines for the care and treatment of breast cancer. CMAJ :

Canadian Medical Association Journal, 158 Suppl 3, S1-2.

Stephenson, G. D., & Rose, D. P. (2003). Breast cancer and obesity: An update. Nutrition and

Cancer, 45(1), 1-16.

Stuebe, A. M., Willett, W. C., Xue, F., & Michels, K. B. (2009). Lactation and incidence of premenopausal breast cancer: A longitudinal study. Archives of Internal Medicine,

169(15), 1364-1371.

Stumpf, W. E., Sar, M., Reid, F. A., Tanaka, Y., & DeLuca, H. F. (1979). Target cells for 1,25-dihydroxyvitamin D3 in intestinal tract, stomach, kidney, skin, pituitary, and parathyroid. Science (New York, N.Y.), 206(4423), 1188-1190.

Szklo, M., & Nieto, F. J. (2000). Epidemiology : Beyond the basics. Gaithersburg, Md: Aspen.

Tang, J. Y., Parimi, N., Wu, A., Boscardin, W. J., Shikany, J. M., Chren, M. M., et al. (2010). Inverse association between serum 25(OH) vitamin D levels and non-melanoma skin cancer in elderly men. Cancer Causes & Control, 21(3), 387-391.

Taylor, V. H., Misra, M., & Mukherjee, S. D. (2009). Is red meat intake a risk factor for breast cancer among premenopausal women? Breast Cancer Research and Treatment, 117(1), 1-8.

Terry, M. B., Zhang, F. F., Kabat, G., Britton, J. A., Teitelbaum, S. L., Neugut, A. I., et al. (2006). Lifetime alcohol intake and breast cancer risk. Annals of Epidemiology, 16(3), 230-240.

Terry, P. D., & Goodman, M. (2006). Is the association between cigarette smoking and breast cancer modified by genotype? A review of epidemiologic studies and meta-analysis. Cancer Epidemiology, Biomarkers & Prevention , 15(4), 602-611.

Terry, P. D., & Rohan, (2002). Cigarette smoking and the risk of breast cancer in women: A review of the literature. Cancer Epidemiology, Biomarkers & Prevention , 11(10 Pt 1), 953-971.

Thanos, J., Cotterchio, M., Boucher, B. A., Kreiger, N., & Thompson, L. U. (2006). Adolescent dietary phytoestrogen intake and breast cancer risk (canada). Cancer Causes & Control,

17(10), 1253-1261.

The Women's Health Initiative Study Group. (1998). Design of the women's health initiative clinical trial and observational study. Controlled Clinical Trials, 19(1), 61-109.

159

Thieden, E., Philipsen, P. A., Heydenreich, J., & Wulf, H. C. (2009). Vitamin D level in summer and winter related to measured UVR exposure and behavior. Photochemistry and

Photobiology, 85(6), 1480-1484.

Thompson, L. U., Boucher, B. A., Cotterchio, M., Kreiger, N., & Liu, Z. (2007). Dietary phytoestrogens, including isoflavones, lignans, and coumestrol, in nonvitamin, nonmineral supplements commonly consumed by women in Canada. Nutrition and

Cancer, 59(2), 176-184.

Thompson, L. U., Boucher, B. A., Liu, Z., Cotterchio, M., & Kreiger, N. (2006). Phytoestrogen content of foods consumed in Canada, including isoflavones, lignans, and coumestan. Nutrition and Cancer, 54(2), 184-201.

Tjonneland, A., Christensen, J., Olsen, A., Stripp, C., Thomsen, B. L., Overvad, K., et al. (2007). Alcohol intake and breast cancer risk: The European prospective investigation into cancer and nutrition (EPIC). Cancer Causes & Control, 18(4), 361-373.

Trang, H. M., Cole, D. E., Rubin, L. A., Pierratos, A., Siu, S., & Vieth, R. (1998). Evidence that vitamin D3 increases serum 25-hydroxyvitamin D more efficiently than does vitamin D2. The American Journal of Clinical Nutrition, 68(4), 854-858.

Tuohimaa, P., Pukkala, E., Scelo, G., Olsen, J. H., Brewster, D. H., Hemminki, K., et al. (2007). Does solar exposure, as indicated by the non-melanoma skin cancers, protect from solid cancers: Vitamin D as a possible explanation. European Journal of Cancer (Oxford,

England: 1990), 43(11), 1701-1712.

United States Department of Agriculture, Agricultural Research Service. (April 2009). USDA

national nutrient database for standard reference. Retrieved 05/13, 2009, from http://www.ars.usda.gov/ba/bhnrc/ndl

van der Meer, I. M., Boeke, A. J., Lips, P., Grootjans-Geerts, I., Wuister, J. D., Deville, W. L., et al. (2008). Fatty fish and supplements are the greatest modifiable contributors to the serum 25-hydroxyvitamin D concentration in a multiethnic population. Clinical

Endocrinology, 68(3), 466-472.

van der Mei, I. A., Blizzard, L., Ponsonby, A. L., & Dwyer, T. (2006). Validity and reliability of adult recall of past sun exposure in a case-control study of multiple sclerosis. Cancer

Epidemiology, Biomarkers & Prevention , 15(8), 1538-1544.

Veronesi, U., Boyle, P., Goldhirsch, A., Orecchia, R., & Viale, G. (2005). Breast cancer. Lancet,

365(9472), 1727-1741.

Vieth, R. (1999). Vitamin D supplementation, 25-hydroxyvitamin D concentrations, and safety. The American Journal of Clinical Nutrition, 69(5), 842-856.

Vieth, R. (2006). What is the optimal vitamin D status for health? Progress in Biophysics and

Molecular Biology, 92(1), 26-32.

Vieth, R. (2009). How to optimize vitamin D supplementation to prevent cancer, based on cellular adaptation and hydroxylase enzymology. Anticancer Research, 29(9), 3675-3684.

160

Vieth, R., Bischoff-Ferrari, H., Boucher, B. J., Dawson-Hughes, B., Garland, C. F., Heaney, R. P., et al. (2007). The urgent need to recommend an intake of vitamin D that is effective. The American Journal of Clinical Nutrition, 85(3), 649-650.

Vieth, R., Cole, D. E., Hawker, G. A., Trang, H. M., & Rubin, L. A. (2001). Wintertime vitamin D insufficiency is common in young Canadian women, and their vitamin D intake does not prevent it. European Journal of Clinical Nutrition, 55(12), 1091-1097.

Vieth, R., Ladak, Y., & Walfish, P. G. (2003). Age-related changes in the 25-hydroxyvitamin D versus parathyroid hormone relationship suggest a different reason why older adults require more vitamin D. The Journal of Clinical Endocrinology and Metabolism, 88(1), 185-191.

Wang, Y., Jacobs, E. J., McCullough, M. L., Rodriguez, C., Thun, M. J., Calle, E. E., et al. (2009). Comparing methods for accounting for seasonal variability in a biomarker when only a single sample is available: Insights from simulations based on serum 25-hydroxyvitamin d. American Journal of Epidemiology, 170(1), 88-94.

Warburton, D. E., Katzmarzyk, P. T., Rhodes, R. E., & Shephard, R. J. (2007). Evidence-informed physical activity guidelines for Canadian adults. Canadian Journal of Public

Health, 98 Suppl 2, S16-68.

Webb, A. R. (2006). Who, what, where and when-influences on cutaneous vitamin D synthesis. Progress in Biophysics and Molecular Biology, 92(1), 17-25.

Webb, A. R., & Engelsen, O. (2006). Calculated ultraviolet exposure levels for a healthy vitamin D status. Photochemistry and Photobiology, 82(6), 1697-1703.

Webb, A. R., Kline, L., & Holick, (1988). Influence of season and latitude on the cutaneous synthesis of vitamin D3: Exposure to winter sunlight in Boston and Edmonton will not promote vitamin D3 synthesis in human skin. The Journal of Clinical Endocrinology and

Metabolism, 67(2), 373-378.

Weiler, L., Knight, J. A., Vieth, R., Barnett, H., & Wong, A. (2007). Comparison of self-reported lifetime sun exposure with two methods of cutaneous microtopography. American

Journal of Epidemiology, 165(2), 222-230.

Welsh, J. (2004). Vitamin D and breast cancer: Insights from animal models. The American

Journal of Clinical Nutrition, 80(6 Suppl), 1721S-4S.

Whitfield, J. F., Boynton, A. L., MacManus, J. P., Sikorska, M., & Tsang, B. K. (1979). The regulation of cell proliferation by calcium and cyclic AMP. Molecular and Cellular

Biochemistry, 27(3), 155-179.

Whiting, S. J., Green, T. J., & Calvo, M. S. (2007). Vitamin D intakes in North America and Asia-pacific countries are not sufficient to prevent vitamin D insufficiency. The Journal

of Steroid Biochemistry and Molecular Biology, 103(3-5), 626-630.

Will, B. P., Berthelot, J. M., Le Petit, C., Tomiak, E. M., Verma, S., & Evans, W. K. (2000). Estimates of the lifetime costs of breast cancer treatment in Canada. European Journal of

Cancer (Oxford, England: 1990), 36(6), 724-735.

Willett, W. (1998). Nutritional epidemiology (2nd ed. ed.). New York: Oxford University Press.

161

Willett, W. (2009) Foreward. The validity of dietary assessment methods for use in epidemiologic studies. British Journal of Nutrition, 102, S1-2.

Wise, L. A., Titus-Ernstoff, L., Newcomb, P. A., Trentham-Dietz, A., Trichopoulos, D., Hampton, J. M., et al. (2009). Exposure to breast milk in infancy and risk of breast cancer. Cancer Causes & Control, 20(7), 1083-1090.

Wolpowitz, D., & Gilchrest, B. A. (2006). The vitamin D questions: How much do you need and how should you get it? Journal of the American Academy of Dermatology, 54(2), 301-317.

World Cancer Research Fund/American Institute for Cancer Research. (2007). Food, nutrition,

physical activity, and the prevention of cancer: A global perspective. Washington DC: AICR.

Xue, L., Lipkin, M., Newmark, H., & Wang, J. (1999). Influence of dietary calcium and vitamin D on diet-induced epithelial cell hyperproliferation in mice. Journal of the National

Cancer Institute, 91(2), 176-181.

Yetley, E. A., Brule, D., Cheney, M. C., Davis, C. D., Esslinger, K. A., Fischer, P. W., et al. (2009). Dietary reference intakes for vitamin D: Justification for a review of the 1997 values. The American Journal of Clinical Nutrition, 89(3), 719-727.

Yu, C. L., Li, Y., Freedman, Fears, T. R., Kwok, R., Chodick, G., et al. (2009). Assessment of lifetime cumulative sun exposure using a self-administered questionnaire: Reliability of two approaches. Cancer Epidemiology, Biomarkers & Prevention, 18(2), 464-471.

162

Appendices

163

Appendix 1 - Questionnaires

Sun Exposure Measure from Epidemiologic Questionnaire:

164

Vitamin D Supplements Measure from Food Frequency Questionnaire:

Note. Vitamin D from foods was measured using additional items from the copyrighted 178-

item Block Food Frequency Questionnaire.

165

Appendix 2 - Ethics Approval

166

Appendix 3 – Missing data

Table 1. Amount of missing data for each variable examined (total n = 6572 for cases and controls combined).

Potential confounders Missing

n (%)

Age 0

Marital status 15 (<1%)

Highest level of Education 34 (<1%)

BMI (kg/m2) 47 (<1%)

Pack-years smoked 102 (1.6%)

Ever breastfeed your infant 27 (<1%)

Age at menarche (years) 247 (3.8%)

Age at menopause 143 (2.2%)

Parity 107 (1.7%)

Age at first live birth 107 (1.7%)

Duration of oral contraceptive use 106 (1.6%)

Duration of HRT use (yrs) 32 (0.5%)

Breast cancer in a 1st degree relative 160 (2.4%)

Benign breast disease 127 (1.9%)

Mammogram 18 (<1%)

Strenuous physical activity in teenage years 447 (6.8%)

Strenuous physical activity in 20s-30s 263 (4.0%)

Strenuous physical activity in 40s-50s (among women >40 years of age) 184 (2.8%)

Strenuous physical activity in 60s-74 (among women >60 years of age) 126 (1.9%)

Moderate physical activity at work in teens 518 (7.9%)

Moderate physical at work in 20s-30s 275 (4.2%)

Moderate physical at work in 40s-50s (among women >40 years of age) 164 (2.5%)

Moderate physical at work in 60s-74 (among women >60 years of age) 96 (1.5%)

Daily activity at work in teens 534 (8.1%)

Daily activity at work in 20s-30s 274 (4.2%)

167

Daily activity at work in 40s-50s (among women >40 years of age) 148 (2.3%)

Daily activity at work in 60s-74 (among women >60 years of age) 211 (3.2%)

Alcohol intake (drinks/week) 83 (1.3%)

Dietary fibre intake 83 (1.3%)

Dietary fat intake 83 (1.3%)

Total phytoestrogen intake 203 (3.1%)

Total calcium intake from food and supplements 83 (1.3%)

Main vitamin D exposure variables (dietary and sunlight)

Vitamin D intake from food 83 (1.3%)

Vitamin D intake from supplements 83 (1.3%)

Ethnicity 31 (<1%)

Solar vitamin D score in teens 648 (9.8%)

Solar vitamin D score in 20s-30s 993 (15.1%)

Solar vitamin D score in 40s-50s (among women >40 years of age) 812 (12.3%)

Solar vitamin D score in 60s-74 (among women >60 years of age) 314 (4.8%)

Sun protection practices in teens 238 (3.6%)

Sun protection practices in 20s-30s 343 (5.2%)

Sun protection practices in 40s-50s (among women >40 years of age) 374 (5.6%)

Sun protection practices in 60s-74 (among women >60 years of age) 155 (2.4%)

UV of residence in teens 454 (6.9%)

UV of residence in 20s-30s 715 (10.9%)

UV of residence in 40s-50s (among women >40 years of age) 438 (6.7%)

UV of residence in 60s-74 (among women >60 years of age) 133 (6.1%)

Time spent outdoors in teens 315 (14.5%)

Time spent outdoors in 20s-30s 346 (5.3%)

Time spent outdoors in 40s-50s (among women >40 years of age) 358 (5.4%)

Time spent outdoors in 60s-74 (among women >60 years of age) 177 (2.7%)

Latitude of residence in teens 453 (6.9%)

Latitude of residence in 20s-30s 714 (10.9%)

Latitude of residence in 40s-50s (among women >40 years of age) 438 (6.7%)

Latitude of residence in 60s-74 (among women >60 years of age) 133 (2.0%)

168

Appendix 4 - Supplementary analyses

169

Figure 1. Distributions of vitamin D intake from a) foods, b) supplements (multivitamins and

vitamin D or cod liver oil), and c) total vitamin D intake (food and supplements) (n = 3,427)

0 120 240 360 480 600 720 840 960 1080 1200

0

5

10

15

20

25

P

e

r

c

e

n

t

0 80 160 240 320 400 480 560 640 720 800

0

10

20

30

40

50

60

P

e

r

c

e

n

t

40 200 360 520 680 840 1000 1160 1320 1480

0

2.5

5.0

7.5

10.0

12.5

15.0

17.5

P

e

r

c

e

n

t

Fig 1a: Vitamin D intake from foods (IU/day)

Fig 1b: Vitamin D intake from supplements (IU/day)

Fig 1c: Total vitamin D intake (IU/day)

170

Table 1. The association between ethnicity and use of sun protection at each age group of

exposure

Use of sun protection1

Caucasian N (%)

Non-Caucasian N (%)

p-value (from Chi square test)

Teen Never Sometimes Always

3065 (54) 2362 (42) 259 (5)

350 (56) 219 (35) 51 (8)

<0.0001

20s-30s Never Sometimes Always

1850 (33) 3258 (58) 494 (9)

265 (44) 269 (45) 68 (11)

<0.0001

40s-50s Never Sometimes Always

894 (17) 2970 (58) 1277 (25)

153 (29) 282 (53) 99 (19)

<0.0001

60s-74 Never Sometimes Always

354 (16) 1089 (49) 798 (36)

52 (34) 67 (44) 35 (23)

<0.0001

1 Self-reported protective clothing or sunscreen use

171

Table 2. Spearman rank correlations (rs) between time spent outdoors and sun protection

practices, erythemal UV and latitude during each age period of exposure

Time spent outdoors (hours per week)1

Teens 20s-30s 40s-50s 60s-74

rs (p-value)

Sun protection2

-0.04 (0.003) -0.02 (0.23) -0.002(0.87) 0.11 (<0.0001)

Erythemal UV of residence

(mW/m2)3

0.05 (0.0003) 0.02 (0.11) 0.04 (0.008) 0.02 (0.39)

Latitude of

residence4

-0.001 (0.64) 0.03 (0.009) 0.01 (0.28) 0.01 (0.59)

1 Typical number of hours spent outdoors from April to October during weekdays and weekends 2 Self-reported protective clothing or sunscreen use 3 Monthly average local noon erythemal UV radiation for June 2003 obtained from NASA’s Total Ozone Mapping Spectrometer (TOMS) 4 Geocoded based on location of residence reported as: city and province/state

172

Table 3. Spearman rank correlations (rs) between time spent outdoors and parity, income and

education during each age period of exposure

Time spent outdoors (hours per week)1

Teens 20s-30s 40s-50s 60s-74

rs (p-value)

Parity 0.06 (<0.001) 0.09 (<0.001) 0.07 (<0.001) 0.03 (0.19)

Income -0.04 (0.005) -0.06 (<0.001) -0.01 (0.34) 0.07 (0.02)

Education -0.07 (<0.001) -0.11 (<0.001) -0.07 (<0.001) 0.01 (0.56)

1 Typical number of hours spent outdoors from April to October during weekdays and weekends

173

Table 4. Distribution of breast cancer cases and controls and odds ratio (OR) estimates of

variables associated with cutaneous vitamin D production created by cross-classification during

4 age periods

1 Sum of the following values: if ethnicity was Caucasian then skin color = 1 and non-Caucasian = 0; if sun protection was used sometimes or always then skin color = 0, if sun protection never used then skin color = 1; if erythemal UV of residence was <50th percentile then UV = 0 and if >50th percentile then UV = 1; if time spent outdoors (calculated as typical number of hours per week during weekdays and weekends from April to October) was <50th percentile then time outdoors = 0 and if >50th percentile then time outdoors = 1. Maximum possible score = 4 for participants with highest UV generating potential. 2 Age-group adjusted

Combined skin color, sun protection, UV,

and time outdoors1

Cases n = 3101 No. (%)

Controls n = 3471 No. (%)

OR (95% CI)2

Teenage years 1 Lowest 2 3 4 Highest

363 (12) 979 (32) 1206 (39) 553 (18)

356 (10) 1009 (29) 1472 (42) 632 (18)

1.00 0.94 (0.79-1.12) 0.79 (0.67-0.93) 0.84 (0.70-1.02)

20-39 yrs of age 1 Lowest 2 3 4 Highest

378 (12) 1053 (34) 1276 (41) 393 (13)

386 (11) 1142 (33) 1472 (42) 469 (14)

1.00 0.93 (0.79-1.10) 0.87 (0.74-1.03) 0.80 (0.66-0.98)

40-59 years of age 1 Lowest 2 3 4 Highest

257 (9) 968 (33) 1463 (50) 231 (8)

245 (8) 1006 (32) 1630 (52) 273 (9)

1.00 0.90 (0.74-1.09) 0.83 (0.69-1.00) 0.77 (0.60-0.99)

60-74 years of age 1 Lowest 2 3 4 Highest

98 (8) 365 (30) 645 (52) 138 (11)

90(7) 352 (27) 714 (54) 167 (13)

1.00 0.95 (0.69-1.31) 0.83 (0.61-1.12) 0.76 (0.53-1.09)

174

Table 5. Sensitivity analyses evaluating a range of assumptions to derive the solar vitamin D score

and the corresponding odds ratio estimates for the derived variables and breast cancer risk

Values applied in algorithm1 OR2 (95% CI) comparing highest to

lowest quartile of exposure

Skin color3 Sun protection

practices 4

Erythemal UV

(mW/m2)5

Time outdoors

(hours/week)6

Exposure during ages 20-39

Exposure during ages 40-59

Caucasian=1.0

Other =0.33

Never=1.0

Sometimes=0.66

Always=0.33

As measured

(continuous)

Hours/week

(continuous)

0.76 (0.65-0.89)7 0.75 (0.64-0.88)

2

Caucasian=1.0 Other = 0.33

Never=1.0 Sometimes=0.66 Always=0.33

As measured (continuous)

Not included 0.77 (0.69-0.89) 0.84 (0.70-0.99)

Caucasians only Never=1.0 Sometimes=0.66 Always=0.33

As measured (continuous)

Hours/week (continuous)

0.79 (0.67-0.93) 0.77 (0.65-0.92)

Non-Caucasians only

Never=1.0 Sometimes=0.66 Always=0.33

As measured (continuous)

Hours/week (continuous)

0.53 (0.22-1.31) 0.50 (0.13-1.91)

Caucasian=1.0 Other=0.33

Never=1.0 Sometimes=0.66 Always=0.33

As measured (continuous)

>1 hr/day = 1.0 <1 hr/day = 0.5

0.83 (0.71-0.96) 0.79 (0.66-0.95)

Caucasian=1.0 Other=0.33

Never=1.0 Sometimes=0.66 Always=0.33

0-120 = 1.0 120-240 = 1.2 240-360 = 1.4 >360 =1.6

>1 hr/day = 1.0 <1 hr/day = 0.5

0.82 (0.71-0.96) 0.80 (0.64-0.99)

Caucasian=1.0 Other=0.33

Never=1.0 Sometimes=0.9 Always=0.7

As measured (continuous)

Hours/week (continuous)

0.75 (0.64-0.87) 0.85 (0.73-0.99)

Caucasian=1.0 Other=0.33

Never=1.0 Sometimes=0.9 Always=0.7

0-120 = 1.0 120-240 = 1.2 240-360 = 1.4 >360 =1.6

>1 hr/day= 1.0 <1 hr/day= 0.5

0.81 (0.69-0.94) 0.81(0.65-1.0)

Caucasian=1.0 Other=0.80

Never=1.0 Sometimes=0.66 Always=0.33

As measured (continuous)

Hours/week (continuous)

0.75 (0.64-0.87) 0.78 (0.66-0.90)

1 Score derived by multiplying assigned values for skin color, sun protection practices, erythemal UV, and time spent outdoors 2 Age group adjusted 3 Ethnicity used as a proxy for skin color with Caucasians assumed to have lighter skin color then non-Caucasian. 4 Self-reported protective clothing or sunscreen use 5 Monthly average local noon erythemal UV radiation for June 2003 obtained from NASA’s Total Ozone Mapping Spectrometer (TOMS) 6 Typical number of hours spent outdoors from April to October during weekdays and weekends 7 Original solar vitamin D score as proposed a priori

175

Table 6. Odds ratio (OR) estimates for derived solar vitamin D score during 4 age periods and

breast cancer risk overall and among lifelong residents of Canada only.

1 Sum of the following values: if ethnicity was Caucasian then skin color = 1 and non-Caucasian = 0; if sun protection was used sometimes or always then skin color = 0, if sun protection never used then skin color = 1; if erythemal UV of residence was <50th percentile then UV = 0 and if >50th percentile then UV = 1; if time spent outdoors (calculated as typical number of hours per week during weekdays and weekends from April to October) was <50th percentile then time outdoors = 0 and if >50th percentile then time outdoors = 1. Maximum possible score = 4 for participants with highest UV generating potential. 2 As reported previously in paper #3, page 117.

Solar vitamin D score 1 Overall

2

(n= 6571)

Lifelong Canadians (n=4447)

Not lifelong Canadian (n= 598)

OR (95% CI) OR (95% CI) OR (95% CI)

Teenage years Q1 Q2 Q3 Q4

1.00 0.90 (0.78-1.03) 0.80 (0.70-0.93) 0.79 (0.68-0.91)

1.00 0.85 (0.71-1.02) 0.78 (0.65-0.94) 0.76 (0.63-0.91)

1.00 1.07 (0.68-1.68) 0.93 (0.58-1.48) 1.02 (0.60-1.75)

20-39 yrs of age Q1 Q2 Q3 Q4

1.00 0.95 (0.81-1.12) 0.89 (0.77-1.02) 0.76 (0.65-0.89)

1.00 1.01 (0.82-1.23) 0.92 (0.77-1.10) 0.84 (0.69-1.01)

1.00 1.19 (0.69-2.07) 0.94 (0.60-1.46) 0.43 (0.25-0.74)

40-59 years of age Q1 Q2 Q3 Q4

1.00 0.85 (0.72-0.99) 0.82 (0.71-0.95) 0.75 (0.64-0.88)

1.00 0.80 (0.65-0.98) 0.85 (0.70-1.02) 0.72 (0.59-0.88)

1.00 0.87 (0.51-1.46) 0.75 (0.46-1.24) 0.82 (0.48-1.42)

60-74 years of age Q1 Q2 Q3 Q4

1.00 0.91 (0.72-1.14) 0.78 (0.62-0.98) 0.59 (0.46-0.76)

1.00 0.79 (0.59-1.05) 0.80 (0.60-1.06) 0.53 (0.39-0.72)

1.00 0.83 (0.37-1.85) 0.99 (0.43-2.28) 0.33 (0.13-0.87)

176

Appendix 5 - Power Calculations

Although the sample size for this study was predetermined, power calculations were conducted

to estimate the power to detect the main study objectives. Power calculations were performed

using Power Program v3.0.0. (Garcia-Closas & Lubin, 1999). All calculations are 2-sided with a

specified type I error (alpha level) of 0.05. An unmatched case control design with 1 control per

case was specified for all calculations. It is acknowledged that power is reduced when

confounders are added to the models but this was not taken into consideration in the power

calculations. Formulas used by Power Program are based on a binary response model.

Vitamin D was categorized into four levels of exposure, based on quartiles, with probabilities of

exposure of 0.25 in each group and a priori an expected odds ratio (comparing the highest versus

lowest quartile) of 0.7 was assumed. This estimated risk reduction of 30% is based on previous

studies results (John et al., 1999; Knight et al., 2007). Given these assumptions, the power to

detect an association between vitamin D and breast cancer was 99.9%. The power to detect an

OR of only 0.8 was 89.8%.

Using the assumptions above for vitamin D, the power to detect a multiplicative interaction (e.g.,

between vitamin D and calcium) with an estimated OR for the interaction effect (theta) of 0.6,

was 78.1%. The assumed interaction effect for the combined effect between calcium and vitamin

D are based on the trial by Lappe et al (Lappe et al., 2007). This assumes calcium intake was also

categorized into quartiles and the expected odds ratio of for the independent effect of calcium

was 0.8 (as observed previously Cui & Rohan, 2006)