Dietary and Reproductive Risk Factors for Breast Cancer in the Nurses' Health Studies

Transcription

1 Dietary and Reproductive Risk Factors for Breast Cancer in the Nurses' Health Studies The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters Citation Sisti, Julia Shafto Dietary and Reproductive Risk Factors for Breast Cancer in the Nurses' Health Studies. Doctoral dissertation, Harvard T.H. Chan School of Public Health. Citable link Terms of Use This article was downloaded from Harvard University s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at nrs.harvard.edu/urn-3:hul.instrepos:dash.current.terms-ofuse#laa

2 DIETARY AND REPRODUCTIVE RISK FACTORS FOR BREAST CANCER IN THE NURSES HEALTH STUDIES JULIA SHAFTO SISTI A Dissertation Submitted to the Faculty of The Harvard T.H Chan School of Public Health In Partial Fulfillment of the Requirements for the Degree of Doctor of Science in the Department of Epidemiology Harvard University Boston, Massachusetts May 2015

3 Advisor: A. Heather Eliassen, ScD Julia Shafto Sisti Dietary and Reproductive Risk Factors for Breast Cancer in the Nurses Health Studies ABSTRACT Breast cancer is the most frequently diagnosed cancer and the second leading cause of cancer death among US women. Several hormonal, anthropometric, lifestyle, and genetic factors are known to be associated with breast cancer, though these associations may differ by menopausal status and molecular subtype. Studying risk factors in relation to these subtypes can help enhance our understanding of breast cancer etiology. Here, we aim to further explore the mechanisms through which several established and suspected risk factors may influence risk of breast cancer, with an emphasis on modifiable exposures, which may have direct implications for prevention strategies, particularly for premenopausal and non-luminal breast cancers. In Chapter 1, we evaluated the cross-sectional relationship between intakes of caffeine, coffee, decaffeinated coffee, and tea, and comprehensive profiles of urinary estrogens and estrogen metabolites in premenopausal women. We found significant associations between coffee intake and metabolites in the 2-hydroxylation pathway, suggesting a possible mechanism through which coffee may affect breast cancer risk. In Chapter 2, we examined associations between premenopausal plasma carotenoid levels and markers of oxidative stress and subsequent breast cancer risk. In contrast to previously published analyses, which largely focused on postmenopausal carotenoid exposure, we did not find significant inverse associations between circulating carotenoids and risk; additionally we did not observe positive associations between fluorescent oxidation products and risk. However, we did find some evidence that the effects of carotenoids on risk may be modified by single nucleotide polymorphisms in genes related to ii

4 carotenoid availability and oxidative/antioxidative processes. In Chapter 3, we explored whether the associations of reproductive risk factors with breast cancer vary by intrinsic molecular subtype. We observed evidence that many risk factors are most strongly associated with the hormone receptor-positive luminal A subtype, which comprises the majority of breast cancers, though tests of heterogeneity did not reach significance in many cases. Consistent with previous studies, we observed that breastfeeding may reduce risk of basal-like tumors, and may represent a potential preventive strategy for this aggressive subtype. In conclusion, these analyses, while varied in scope, help elucidate mechanisms by which risk factors may influence breast cancer risk. iii

5 TABLE OF CONTENTS INTRODUCTION 1 References 4 CHAPTER 1: Caffeine, coffee and tea intakes and urinary estrogens and estrogen metabolites in premenopausal women Abstract 8 Introduction 9 Methods 11 Results 15 Discussion 18 References 24 Tables 29 CHAPTER 2: Premenopausal plasma carotenoids, fluorescent oxidation products and subsequent breast cancer risk in the Nurses Health Studies Abstract 35 Introduction 37 Methods 38 Results 44 Discussion 46 References 52 Tables 58 CHAPTER 3: Reproductive risk factors in relation to molecular subtype of breast cancer in the Nurses Health Studies Abstract 72 Introduction 74 Methods 76 Results 81 Discussion 84 References 90 Tables 97 iv

6 LIST OF TABLES CHAPTER 1 Table 1.1 Characteristics of the premenopausal study population in the Nurses Health Study II, by quartiles of caffeine intake Table 1.2. Multivariate-adjusted geometric means of estrogen metabolism measures by quartiles of caffeine intake in the Nurses Health Study II Table 1.3. Multivariate-adjusted geometric means of estrogen metabolism measures by category of coffee intake in the Nurses Health Study II Table 1.4. Multivariate-adjusted geometric means of estrogen metabolism measures by category of decaffeinated coffee intake in the Nurses Health Study II Table 5. Multivariate-adjusted geometric means of estrogen metabolism measures by category of tea intake in the Nurses Health Study II CHAPTER 2 Table 2.1. Characteristics of the premenopausal study population at blood draw, Nurses Health Study and Nurses Health Study II Table 2.2 Relative risks (RR) of breast cancer and 95% confidence intervals (CI) according to quartile of premenopausal carotenoid levels (ug/dl) in the Nurses Health Studies Table 2.3. Relative risks (RR) of breast cancer and 95% confidence intervals (CI) according to quartile of premenopausal FlOP levels (FI/mL) in the Nurses Health Studies Table 2.4. Relative risks (RR) of breast cancer and 95% confidence intervals (CI) according to quartiles of premenopausal carotenoid levels, by menopausal status at diagnosis Table 2.5. Summary of significant interactions between carotenoid levels (>/< median) and SNPs in CAT, SOD2, GPX1, MPO and BCMO1 on breast cancer risk in the Nurses Health Study II Supplemental Table 2.1. Relative risks (RR) of breast cancer and 95% confidence intervals (CI) according to quartiles of premenopausal carotenoid and FlOP levels in the Nurses Health Studies, by tumor estrogen receptor (ER) status Supplemental Table 2.2. Minor allele frequencies for SNPs in CAT, SOD2, GPX1, MPO and BCMO1 genes and age-adjusted per-allele odds ratios for breast cancer in the Nurses Health Study II Supplemental Table 2.3. Age-adjusted geometric means of plasma carotenoid levels (ug/dl) by number of minor alleles in CAT, SOD2, GPX1, MPO and BCMO1 genes among controls in the Nurses Health Study II CHAPTER 3 Table 3.1. Tumor characteristics at diagnosis according to breast cancer phenotypes in the Nurses Health Studies Table 3.2. Multivariate-adjusted HRs (95% CI) for reproductive risk factors in relation to breast cancer subtypes in the Nurses Health Studies Table 3.3 Multivariate-adjusted HRs (95% CI) for birth index and time between menarche and age at first birth in relation to breast cancer subtypes in the Nurses v

7 Health Studies Table 3.4. Multivariate-adjusted HRs (95% CI) for birth index and time between menarche and age at first birth in relation to breast cancer subtypes in the Nurses Health Studies 101 vi

8 ACKNOWLEDGEMENTS I extend my deepest thanks to a number of individuals for their invaluable contributions to this dissertation. First and foremost, I thank my advisor Dr. A. Heather Eliassen, who provided me with an incredible amount of support and mentorship along every step of this process. Her help and guidance were instrumental not only in the completion of this dissertation, but also to my career development and growth as a scientist. Her encouragement and flexibility allowed me to complete my studies under a variety of changing circumstances, and I could not be more appreciative. I also acknowledge my committee members, Drs. Rulla Tamimi and Bernard Rosner, whose input and advice strengthened my work and often led me to a deeper understanding of my own research. Their insights and help are truly appreciated. I also thank Dr. Shelley Tworoger for her participation in my orals committee, and Dr. Susan Hankinson for taking me on as her student at a time when I had barely considered the possibility of pursuing a doctoral degree. I also appreciate the support I received from National Institutes of Health training grants, as well as Dr. Meir Stampfer for his role in securing and coordinating that funding. Many thanks are also due to the many friends and colleagues I have met over my five years at Harvard, who provided help in various ways, academic and otherwise. I thank my family, particularly my parents, for their unwavering love and help throughout this process, as well as in all aspects of my life. Lastly, but perhaps most importantly, I thank my husband, Aaron Katzman, whose support and ability to make me laugh have been so crucial. I vii

9 could not ask for a greater partner in life or in parenting our son, Milo Katzman, whose arrival in our lives has truly been the greatest gift. viii

10 INTRODUCTION Breast cancer is the most frequently diagnosed cancer and the second leading cause of cancer death among US women. 1 To date, a number of risk factors for breast cancer have been identified, and include: 1) age; 2 2) family history 3 and highly penetrant inherited genetic mutations such as BRCA1 and BRCA2; 4 3) benign breast disease and high breast density; 5 4) reproductive factors including late age at first birth and increased years of menstrual cycling; 6,7 5) dietary factors (i.e. alcohol intake); 8 6) current use of estrogen plus progestin postmenopausal hormones; 9,10 7) exposure to ionizing radiation; 11,12 and 8) anthropometric factors including height 13 and adult weight gain. 14 Though identification of non-modifiable risk factors can help identify populations that may benefit from increased surveillance, preventive efforts targeted at modifiable factors may help reduce the overall burden of disease. Breast cancer is a heterogeneous disease, and examining how risk factors may vary by menopausal status and molecular subtype may enable greater understanding of the underlying mechanisms of pathogenesis. Chapters 1 and 2 of this dissertation focus on the role of modifiable dietary factors in breast cancer risk, with an emphasis on premenopausal exposures and the use of biomarkers to elucidate potential biologic pathways. Chapter 3 focuses on how the associations of reproductive risk factors may differ across molecular subtypes of breast cancer, which may provide insight into the etiology of these phenotypically distinct tumor types. Many breast cancer factors are hypothesized to act by altering lifetime exposure to circulating sex hormones, particularly estrogens and androgens. 15 Circulating levels of these hormones after menopause are associated with breast cancer risk, 16 though evidence of an association with premenopausal levels has been less consistent. 15 Most epidemiologic studies have examined the 1

11 associations of the parent estrogens (estrone and estradiol) with risk; however, metabolites formed in the catabolism of the parent estrogens may also have relevance to risk. Attention has largely focused on the role of two metabolites, 2-hydroxyestrone (2-OHE1) and 16αhydroxyestrone (16α-OHE1), and some evidence of an inverse association between their ratio (2:16α-OHE1) and risk has been observed among premenopausal women. 17 However, premenopausal urinary levels of another metabolite, 17-epiestriol, have also been associated with risk, 18 suggesting that more comprehensive metabolic profiles should be incorporated in future studies. In Chapter 1, we used a recently developed liquid chromatography/tandem mass spectrometry (LC-MS/MS) assay with high sensitivity, accuracy, and reproducibility to measure urinary levels of 15 estrogens and estrogen metabolites (EM) in a sample of premenopausal women in the Nurses Health Study II (NHSII). In a cross-sectional analysis, we investigated whether profiles of estrogen metabolism were associated with intakes of caffeine, coffee and tea, in order to gain insight into whether the observed modest protective effect of coffee on breast risk may be mediated through changes in estrogen metabolism. The role of fruit and vegetable intake in breast cancer development has attracted substantial interest as a possible avenue for prevention, though results of many large-scale epidemiologic studies have not demonstrated a clear protective effect of these foods on risk However, carotenoids, a class of nutrients abundant in many red, orange, and dark green fruits and vegetables, may reduce breast cancer risk, potentially due to their antioxidant properties. Studies examining circulating carotenoids have tended to find greater effect sizes than those that have relied upon food frequency questionnaire (FFQ) data to estimate exposure. 24 Use of these biomarkers avoids measurement error inherent in recalled dietary intakes, as well as the effect of 2

12 food preparation and meal composition on carotenoid availability. Currently, few analyses have examined the effects of carotenoid exposure earlier in life, as most have measured postmenopausal carotenoid levels. 25 However, studying levels during premenopause may help us better understand the relevant etiologic period for carotenoid exposure, and allow us to examine whether the effect of carotenoids varies by hormonal milieu. In Chapter 2, we examined the associations between circulating carotenoid levels and breast cancer risk in a nested case-control study in the NHS and NHSII. In order to further elucidate the roles of antioxidative and oxidative factors in breast carcinogenesis, we additionally investigated whether circulating markers of oxidative stress and single nucleotide polymorphisms in genes hypothesized to affect the burden of oxidative stress were associated with risk. There is an evolving understanding that breast cancer is a highly complex and heterogeneous disease. While breast tumors were once primarily defined by histologic characteristics or hormone receptor expression, patterns of gene expression have more recently been used to classify breast tumors into at least four intrinsic subtypes (luminal A, luminal B, HER2-type, and basal-like) with distinct phenotypic and clinical characteristics. 26,27 In Chapter 3, we prospectively examined the associations between reproductive risk factors for breast cancer in relation to each of the currently recognized intrinsic subtypes. Studying whether the associations of breast cancer risk factors vary across subtypes provides insight into whether distinct etiologic pathways underlie these subtypes. Many reproductive risk factors are non-modifiable (i.e. age at menarche, age at menopause); however, breastfeeding, which has previously been linked to a reduced risk of basal-like tumors, could represent a potential avenue of prevention for this aggressive subtype. 3

13 REFERENCES 1. Siegel, R., Naishadham, D. & Jemal, A. Cancer statistics, CA. Cancer J. Clin. 63, (2013). 2. Howlader, N. et al. SEER Cancer Statistics Review, , National Cancer Institute. Bethesda, MD, based on November 2013 SEER data submission, posted to the SEER web site, April Collaborative Group on Hormonal Factors in Breast Cancer. 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, (2001). 4. Chen, S. & Parmigiani, G. Meta-analysis of BRCA1 and BRCA2 penetrance. J. Clin. Oncol. Off. J. Am. Soc. Clin. Oncol. 25, (2007). 5. Tice, J. A. et al. Benign breast disease, mammographic breast density, and the risk of breast cancer. J. Natl. Cancer Inst. 105, (2013). 6. Reeves, G. K. et al. Reproductive factors and specific histological types of breast cancer: prospective study and meta-analysis. Br. J. Cancer 100, (2009). 7. Collaborative Group on Hormonal Factors in Breast Cancer. Menarche, menopause, and breast cancer risk: individual participant meta-analysis, including women with breast cancer from 117 epidemiological studies. Lancet Oncol. 13, (2012). 8. Hamajima, N. et al. 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. Br. J. Cancer 87, (2002). 4

14 9. Chlebowski, R. T. et al. Estrogen plus progestin and breast cancer incidence and mortality in postmenopausal women. JAMA 304, (2010). 10. Beral, V. & Million Women Study Collaborators. Breast cancer and hormone-replacement therapy in the Million Women Study. Lancet 362, (2003). 11. Land, C. E. et al. Incidence of female breast cancer among atomic bomb survivors, Hiroshima and Nagasaki, Radiat. Res. 160, (2003). 12. Boice, J. D., Harvey, E. B., Blettner, M., Stovall, M. & Flannery, J. T. Cancer in the contralateral breast after radiotherapy for breast cancer. N. Engl. J. Med. 326, (1992). 13. Green, J. et al. Height and cancer incidence in the Million Women Study: prospective cohort, and meta-analysis of prospective studies of height and total cancer risk. Lancet Oncol. 12, (2011). 14. Eliassen, A. H., Colditz, G. A., Rosner, B., Willett, W. C. & Hankinson, S. E. Adult weight change and risk of postmenopausal breast cancer. JAMA 296, (2006). 15. Endogenous Hormones and Breast Cancer Collaborative Group et al. Sex hormones and risk of breast cancer in premenopausal women: a collaborative reanalysis of individual participant data from seven prospective studies. Lancet Oncol. 14, (2013). 16. Key, T., Appleby, P., Barnes, I. & Reeves, G. Endogenous sex hormones and breast cancer in postmenopausal women: reanalysis of nine prospective studies. J. Natl. Cancer Inst. 94, (2002). 17. Dallal, C. M. et al. Urinary estrogen metabolites and breast cancer: a combined analysis of individual level data. Int. J. Biol. Markers 0 (2012). doi: /jbm

15 18. Eliassen, A. H. et al. Urinary estrogens and estrogen metabolites and subsequent risk of breast cancer among premenopausal women. Cancer Res. 72, (2012). 19. Van Gils, C. H. et al. Consumption of vegetables and fruits and risk of breast cancer. JAMA J. Am. Med. Assoc. 293, (2005). 20. Lissowska, J. et al. Intake of fruits, and vegetables in relation to breast cancer risk by hormone receptor status. Breast Cancer Res. Treat. 107, (2008). 21. Jung, S. et al. Fruit and vegetable intake and risk of breast cancer by hormone receptor status. J. Natl. Cancer Inst. 105, (2013). 22. Smith-Warner, S. A. et al. Intake of fruits and vegetables and risk of breast cancer: a pooled analysis of cohort studies. JAMA J. Am. Med. Assoc. 285, (2001). 23. Aune, D. et al. Fruits, vegetables and breast cancer risk: a systematic review and metaanalysis of prospective studies. Breast Cancer Res. Treat. 134, (2012). 24. Aune, D. et al. Dietary compared with blood concentrations of carotenoids and breast cancer risk: a systematic review and meta-analysis of prospective studies. Am. J. Clin. Nutr. 96, (2012). 25. Eliassen, A. H. et al. Circulating carotenoids and risk of breast cancer: pooled analysis of eight prospective studies. J. Natl. Cancer Inst. 104, (2012). 26. Perou, C. M. et al. Molecular portraits of human breast tumours. Nature 406, (2000). 27. Sørlie, T. et al. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc. Natl. Acad. Sci. U. S. A. 98, (2001). 6

16 CAFFEINE, COFFEE AND TEA INTAKES AND URINARY ESTROGENS AND ESTROGEN METABOLITES IN PREMENOPAUSAL WOMEN JULIA S. SISTI 1,2, SUSAN E. HANKINSON 1, 2, 3, NEIL E. CAPORASO 4, FANGYI GU 4, RULLA M. TAMIMI 1, 2, BERNARD A. ROSNER 2, XIA XU 5, REGINA G. ZIEGLER 4 *, A. HEATHER ELIASSEN 1, 2 * 1 Department of Epidemiology, Harvard School of Public Health, Boston, MA; 2 Channing Division of Network Medicine, Department of Medicine, Brigham and Women s Hospital and Harvard Medical School, Boston, MA; 3 Division of Biostatistics and Epidemiology, School of Public Health and Health Sciences, University of Massachusetts, Amherst, MA; 4 Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD; 5 Cancer Research Technology Program, Leidos Biomedical Research, Inc., Frederick National Laboratory for Cancer Research, Frederick, MD *Co-last authors 7

17 ABSTRACT: Background: Prior studies have found weak inverse associations between breast cancer and caffeine and coffee intake, possibly mediated through their effects on sex hormones. Methods: High-performance liquid chromatography/tandem mass spectrometry was used to quantify levels of 15 individual estrogens and estrogen metabolites (EM) among 587 premenopausal women in the Nurses Health Study II with mid-luteal phase urine samples and caffeine, coffee and/or tea intakes from self-reported food frequency questionnaires. Multivariate linear mixed models were used to estimate geometric means of individual EM, pathways and ratios by intake categories, and P-values for tests of linear trend. Results: Compared to women in the lowest quartile of caffeine consumption, those in the top quartile had higher urinary concentrations of 16α-hydroxyestrone (28% difference; P- trend=0.01) and 16-epiestriol (13% difference; P-trend=0.04), and a decreased parent estrogens/2-, 4-, 16-pathway ratio (P-trend=0.03). Coffee intake was associated with higher 2- catechols, including 2-hydroxyestradiol (57% difference, 4 cups/day vs. 6 cups/week; P- trend=0.001) and 2-hydroxyestrone (52% difference; P-trend=0.001), and several ratio measures. Decaffeinated coffee was not associated with 2-pathway metabolism, but women in the highest (vs. lowest) category of intake ( 2 cups/day vs. 1-3 cups/month) had significantly lower levels of two 16-pathway metabolites, estriol (25% difference; P-trend=0.01) and 17-epiestriol (48% difference; Ptrend=0.0004). Tea intake was positively associated with 17-epiestriol (52% difference; Ptrend=0.01). Conclusion: Caffeine and coffee intake were both associated with profiles of estrogen metabolism in premenopausal women. 8

18 INTRODUCTION Despite investigation in many large-scale epidemiologic studies, the association between coffee and breast cancer risk remains unclear. While some evidence suggests a small inverse association between coffee and breast cancer risk, 1,2 some large prospective studies have reported null associations. 3 6 Two meta-analyses reported a 2-cup/day increase in coffee intake was associated with a nonsigificant 2% lower breast cancer risk. 7,8 Isolating the mechanism through which coffee might affect cancer risk is complicated by the chemical complexity of coffee, which comprises many potentially bioactive compounds. Coffee contains polyphenol antioxidants, which have been hypothesized to reduce breast cancer risk, 9,10 and is a major dietary source of caffeine, which has been inversely associated with breast cancer risk in some, 1,2,7 though not all, 3 6 studies. Caffeine represents a particularly interesting biologic component of coffee with respect to breast cancer, given that enzymes involved in its metabolism also play a role in estrogen metabolism. 11,12 Parent estrogens, estrone and estradiol, are irreversibly metabolized along three different pathways, depending on the initial hydroxylation at the 2-, 4-, or 16-position of the steroid ring. Estrogen metabolites formed in each pathway are believed to have varying degrees of carcinogenic potential. 13 Catechol estrogens, which have two adjacent hydroxyl groups, are formed after the initial hydroxylation of the parent estrogens at the 2- or 4-positions and may be oxidized to produce reactive quinones or inactivated by methylation of one of the adjacent their hydroxyl groups. Quinones may damage DNA directly, and also may undergo redox cycling to produce mutagenic reactive oxygen species. Laboratory studies suggest that 4-pathway catechols have more potential to induce DNA damage than 2-pathway catechols because they form 9

19 covalent, depurinating adducts, and 2-methoxyestradiol may have antiestrogenic properties, inhibiting proliferation of breast cancer cells. Animal and laboratory evidence have demonstrated that metabolites in the 16-pathway are not only genotoxic, causing the formation of depurinating DNA adducts, but also bind tightly to the estrogen receptor, up-regulate it, and increase cell proliferation While there is limited epidemiologic data for premenopausal women, among premenopausal women in the Nurses Health Study II (NHSII), urinary mid-luteal levels of one 16-pathway EM and a higher ratio of 16-pathway metabolites to parent estrogens were associated with increased risk of breast cancer, while metabolites in the 2- and 4-pathways appeared inversely associated with risk. 17 The initial hydroxylation of the parent estrogens is catalyzed primarily by cytochrome P450 enzymes, which play key roles in the metabolism of caffeine. Of particular interest is the CYP1A2 isoform, which catalyzes the hydroxylation of the parent estrogens, 11 and the initial demethylation of caffeine in humans. 12 In prior studies, polymorphisms in the CYP1A2 gene modified the association between coffee consumption and age at breast cancer diagnosis and estrogen receptor status in the general population, 18 and modified the association between coffee intake and risk of breast cancer among BRCA1 mutation carriers. 19 Previously, in premenopausal women in the NHSII, we observed an inverse association between coffee and caffeine intake and mid-luteal plasma levels of estradiol. 20 In another study of premenopausal women, coffee intake was positively associated with plasma levels of 2-hydroxyestrone, and nonsignificantly inversely associated with 16α-hydroxyestrone. 21 However, it is unclear if these associations are attributable to caffeine, or other known or unknown bioactive elements of coffee. 10

20 We sought to further elucidate the relationship between coffee and caffeine intakes and patterns of estrogen metabolism. Using a liquid chromatography/tandem mass spectrometry (LC-MS/MS) method with high sensitivity, accuracy, and reproducibility, we examined the cross-sectional relationship between intake of caffeine, coffee, tea, and decaffeinated coffee and urinary levels of 15 individual estrogens and estrogen metabolites (all 15 referred to as EM), total EM, and estrogen metabolism pathway measures among premenopausal women in the NHSII. METHODS: Study Population: The NHSII is an ongoing prospective cohort study, established in 1989, when 116,430 registered nurses ages years were enrolled. At baseline and biennially since, participants have returned mailed questionnaires with updated information about disease and exposure status. In , participants who were cancer-free and ages years were asked to provide blood and urine samples. Of 29,611 women who provided samples, 18,521 who were premenopausal and had not been pregnant, breastfed, or used oral contraceptives in the 6 months preceding collection provided samples timed within their menstrual cycle. Women collected follicular phase blood samples during days 3-5 of their menstrual cycle, and blood and urine samples during the mid-luteal phase, 7-9 days before the anticipated start of their next cycle. Urine samples were shipped with an ice pack to our laboratory via overnight courier; 93% of samples were received within 26 hours of collection. Upon arrival at the laboratory, urine samples were aliquoted into cryotubes without preservatives and stored in liquid nitrogen freezers. 11

21 The current study population includes 110 women selected for a biomarker reproducibility study, 22 and 493 controls from a nested case-control study of breast cancer. 17 Of these 603 women, 587 had dietary data. Assessment of Exposure: Semi-quantitative food frequency questionnaires (FFQ) were used to assess intake of specific foods every four years, beginning in Possible choices for intake of coffee, caffeinated tea and decaffeinated coffee ranged from never or less than once per month to 6 or more times per day. Caffeine intake was derived from self-reported intakes of coffee, soda, tea, and chocolate using their caffeine content per serving as estimated by the US Department of Agriculture (USDA) food composition sources. The average amount of caffeine was estimated to be 137mg per 8oz serving of coffee, 47mg per 8oz serving of tea, 46mg per 12oz serving of soda, and 7mg per 1oz serving of chocolate. Caffeine was adjusted for energy intake using the residual method, 23 and modeled as quartiles of daily intake (mg), while intake of each specific beverage was modeled as servings per month, week or day. We present analyses using the 1999 self-reported intakes for exposure; results using the mean of the 1995 and 1999 responses were not appreciably different. Assessment of Covariates: The questionnaire completed at urine collection included information about collection time, whether urine was first morning void, and the participant s present weight and tobacco use. To confirm menstrual cycle phase and calculate luteal day at specimen collection, 97% of participants recorded the date of their next menstrual period on a postcard returned by mail. Progesterone levels were measured in the blood sample drawn on the same day 12

22 as urine collection. 24,25 We considered women with luteal progesterone levels 400 ng/dl to have donated samples in an ovulatory cycle, while participants with levels below this cut-point were defined as having an anovulatory cycle. Anthropometric, reproductive, and other lifestyle factors were assessed on biennial questionnaires. Information on menstrual cycle regularity and usual length was obtained in BMI was calculated from weight at time of urine collection and height in Physical activity was queried every four years; we used the average of the 1997 and 2001 surveys to estimate physical activity in metabolic equivalent (MET) hours/week. Average alcohol intake was calculated from the 1995 and 1999 FFQs. Laboratory Methods: Details of the assay have previously been described 26,27. Briefly, frozen 500 ul aliquots of urine were shipped to the Laboratory of Proteomics and Analytical Technologies (Leidos Biomedical Research, Inc., Frederick National Laboratory for Cancer Research, Frederick, MD), where levels of 15 individual EM were quantified using stable isotope dilution LC-MS/MS. Because urinary EM are generally present as glucoronide and sulfate conjugates, an initial hydrolysis step was performed using B-glucuronidase/sulfatase from Helix pomatia. To correct for loss of EM throughout the assay procedure and allow accurate quantification, an internal standard solution of 5 deuterated EM (17β-estradiol-d4, estriol-d3, 2- hydroxy-17β-estradiol-d5, 2-methoxy-17β-estradiol-d5, 16-epiestriol-d3) was added to each thawed aliquot. Each batch of urine samples included masked replicate quality control samples, which were used to assess intra- and inter-batch assay variability. For most metabolites, coefficients of variation were below 7%, though two low concentration EM, 4-methoxyestradiol 13

23 and 4-methoxyestrone, had CVs of 15% and 17%, respectively. The lower limit of quantitation for all EM was approximately 150 fmol/ml urine. Two batches of urinary creatinine were measured at the Endocrine Core Laboratory at Emory University (Atlanta, GA); a third was measured at the laboratory of Dr. Vincent Ricchiuti at Brigham and Women s Hospital. CVs were 9.2%. Plasma progesterone was assayed by chemiluminescent immunoassay using the Immulite Auto-Analyzer (Diagnostic Products Corp, Los Angeles, CA). Overall CVs were 17%, while within-batch CVs were 4%. Statistical Analysis: EM levels were standardized by creatinine level to account for differences in urine volume and concentration. EM were assessed individually, grouped by metabolic pathway and chemical characteristics (i.e. catechols), summed as total EM, and as ratios of individual EM and pathways. To improve normality, all outcome measures were log-transformed. Statistical outliers for each measure were identified using the generalized extreme Studentized deviate many-outlier approach, 28 and removed from analyses (N=0-16 (2-methoxyestradiol)). Multivariate linear mixed models were used to estimate geometric means of the log-transformed metabolites, pathways and ratios by exposure category. Tests of linear trend were conducted by modeling the median of each beverage intake category and the quartile medians of caffeine intake as continuous variables. All models were adjusted for age at urine collection, BMI at collection, height, ovulatory cycle, first-morning urine, alcohol intake, total physical activity, current tobacco use, luteal day, usual menstrual cycle length, menstrual cycle regularity, and age at first birth and parity. Additional adjustment for creatinine did not significantly change our results, and this variable was not retained in final models. 14

24 In secondary analyses, we restricted to women with ovulatory cycles, non-smokers, or women who provided samples between luteal days 4 and 10. We also examined potential effect modification by BMI ( 25 vs. >25). Wald tests were used to assess the significance of interaction terms between dichotomous BMI and the median of the exposure category that were included in models that included all women. All P values were two-sided and tests of significance were performed at the α=0.05 level. All analyses were conducted using SAS v. 9.3 (SAS Institute, Cary, NC). RESULTS: The mean age of the study population was 42.8 years at urine collection, and the mean BMI was 25.1 (Table 1.1). Based on plasma progesterone level, 90% of cycles were ovulatory; 86% of samples were collected 4-10 days prior to next menstrual period. Compared to women with the lowest caffeine intake, those with highest intake were more likely to report being current smokers, had higher alcohol intake, and were less likely to have provided a first-morning urine sample. Women with highest caffeine intake were also slightly older than those with the lowest intake and had a higher BMI. About half of women (49.9%) reported drinking 6 cups of coffee/week, and 40 (6.9%) reported drinking 4 cups/day. Median caffeine intake was mg/day. Consumption of tea and decaffeinated coffee was notably lower than that of coffee; 415 women (71.4%) reported drinking <1 cup of tea/week, while 36 (6.2%) reported drinking 2 cups/day. A total of 369 (63.5%) participants drank decaffeinated coffee <3 times/month and 44 (7.6%) drank 2 cups/day. In multivariate analyses, caffeine intake was associated with several individual EM and 15

25 metabolic ratios (Table 1.2). Higher caffeine intake was associated with higher urinary levels of two 16-pathway metabolites, 16α-hydroxyestrone (28% higher among women in the highest quartile of caffeine intake compared to the lowest quartile, P-trend= 0.01) and 16-epiestriol (13% higher, P-trend=0.04). Furthermore, suggestively, though not significantly, higher 16- ketoestradiol levels were observed at higher levels of caffeine intake (12% higher, P- trend=0.07). However, the combined 16-pathway was only nonsignificantly 5% higher at higher caffeine intake. There were suggestive positive association between caffeine consumption and the 2-catechols, though these associations did not appear linear and were not statistically significant. Compared to women in the lowest quartile of caffeine intake, those with the highest intake had 16% higher levels of 2-hydroxyestrone (P-trend=0.10) and 13% higher levels of 2-hydroxyestradiol (Ptrend=0.14). These suggestive associations with 2-catechols appeared to drive the positive associations observed with the 2-catechols/methylated 2-catechols ratio (P-trend=0.0002), and the catechols/methylated catechols ratio (P-trend=0.001), and the inverse association with the 4- pathway/2-pathway ratio (P-trend=0.01). Lastly, an inverse association between caffeine and the parent estrogens/2-, 4-, 16-pathway ratio was observed (P-trend=0.03), largely because total EM nonsignificantly increased with caffeine intake (P-trend=0.10). Coffee consumption was positively associated with urinary levels of 2-catechols, but not with methylated 2- or 4-catechols (Table 1.3). Compared to women who reported drinking 6 cups/week, those who drank 4 cups/day had 52% higher 2-hydroxyestrone levels (Ptrend=0.001) and 57% higher 2-hydroxyestradiol levels (P-trend<0.001). Overall, 2-catechols were 52% higher (P-trend=0.001), catechols were 45% higher (P-trend=0.002), and the 2-16

26 hydroxylation pathway was 41% higher (P-trend=0.01) among participants in the top category of intake compared to the lowest. A number of ratios comparing 2-catechol levels to other pathways were also associated with coffee consumption. Positive associations were observed for the following ratios where 2- catechols were included in the numerator: 2-pathway/16-pathway (P-trend=0.01), 2- catechols/methylated 2-catechols (P-trend<0.0001), catechols/methylated catechols (Ptrend<0.0001), and 2-pathway/parent estrogens ratio (P-trend=0.003). Conversely, an inverse association with coffee intake was observed for the 4-pathway/2-pathway ratio (P-trend=0.03). As with caffeine, total EM nonsignificantly increased with coffee intake (P-trend=0.08). Markedly fewer associations were observed between individual EM and decaffeinated coffee and tea intakes. Two individual 16-pathway metabolites were inversely associated with decaffeinated coffee intake (Table 1.4). Compared to those who reported drinking 3 cups/month, participants who drank 2 cups/day had 25% lower estriol levels (P-trend=0.01) and 48% lower 17-epiestriol levels (P-trend=0.0004). Decaffeinated coffee also was suggestively but nonsignificantly inversely associated with other 16-pathway metabolites, specifically 16-ketoestradiol (top vs. bottom intake: 14% lower) and 16-epiestriol (16% lower). Overall, levels of all 16-pathway metabolites combined were 16% lower among participants in the highest category, though the test for trend was not significant (P-trend=0.14). Caffeinated tea intake was positively associated with 17-epiestriol (P-trend=0.01), which was 52% higher among those consuming 2 cups/day compared to those who drank 3 cups/month 17

27 (Table 1.5). Tea intake was also inversely associated with the 2-catechols/methylated 2-catechols (P-trend=0.02) and the catechols/methylated catechols ratios (P-trend=0.02). Results from sensitivity analyses among non-smokers (n=560), samples 4-10 days before next menstrual period (n=516), and ovulatory cycles (n=537) did not differ appreciably from the results of main analyses. We observed some evidence of modification by BMI of the association between coffee and estrogen metabolism, with several associations appearing stronger in overweight women (BMI>25) than in normal weight women (BMI 25). Positive associations between coffee and 2-hydroxyestrone (P-interaction=0.02), 2-hydroxyestradiol (Pinteraction=0.06) and 4-hydroxyestrone (P-interaction=0.11) were observed only among overweight women. Positive associations between coffee intake and 2-catechols (Pinteraction=0.02), catechols (P-interaction=0.02), 2-hydroxylation pathway (P-interaction=0.04), 4-hydroxylation pathway (P-interaction=0.13), and total EM (P-interaction=0.05) were limited to, or stronger among, overweight women. Consequently, the 2-catechols/methylated 2 catechols ratio (P-interaction=0.01), and catechols/methylated catechols ratio (Pinteraction=0.01) were also limited to, or stronger among, overweight women. For other exposures, results stratified by BMI generally appeared similar. DISCUSSION: In this population of premenopausal women, we observed several associations between coffee and caffeine intakes and luteal urinary levels of individual EM, metabolic pathways, and ratio measures. Interestingly, there were relatively few individual EM that were associated with more than one exposure. Coffee intake was positively associated with levels of both 2-catechol EM, 2-18

28 hydroxyestrone and 2-hydroxyestradiol, and several ratio and group measures involving these metabolites. Caffeine, caffeinated tea, and decaffeinated coffee were each associated with individual 16-pathway EM. Prior to the development of the high throughput LC-MS/MS method, measuring multiple estrogen metabolites was largely infeasible and inaccurate. 26 Consequently, prior epidemiologic studies of caffeine, coffee, and EM focused largely on 2-hydroxyestrone and 16αhydroxyestrone. Early evidence suggested that among premenopausal women not using oral contraceptives, daily coffee consumption was positively associated with plasma levels of 2- hydroxyestrone and 2-hydroxyestrone/16α-hydroxyestrone ratio, and nonsignificantly inversely associated with 16α-hydroxyestrone level. 21 Bradlow et al. also reported that coffee intake was positively associated with premenopausal 2-hydroxyestrone/16α-hydroxyestrone ratio in plasma, but not urine. 29 However, in a large, multiethnic group of premenopausal women, coffee intake was not related to urinary 2-hydroxyestrone levels, though a positive association was observed between this metabolite and caffeine from non-coffee sources. 30 In addition to supporting previous findings of a positive association between coffee intake and 2- hydroxyestrone, our study provides evidence that coffee may be associated with 2- hydroxyestradiol, another 2-catechol. The strong association between coffee and 2-catechols in our study contributed to several associations with ratio measures, indicating that coffee intake may increase metabolism in the 2-pathway, relative to the 4- and 16-pathways. Coffee has many constituents that may influence estrogen metabolism, including caffeine and polyphenols. Laboratory evidence has suggested that caffeic acid and chlorogenic acid, two polyphenols found 19

29 in coffee, may inhibit 2-catechol and 4-catechol methylation. 31 This mechanism may explain the finding of a positive association between coffee and 2-catechols; however, no associations with 4-catechols, or any methylated catechols, were observed. Lastly, no significant associations were observed between coffee and any 16-pathway EM, though there was some suggestion of a positive association with 16α-hydroxyestrone. Conversely, we observed no significant association between caffeine and urinary levels of either 2-catechol EM, though some evidence of a non-linear positive association was observed with 2- hydroxyestrone and 2-hydroxyestradiol. Like coffee, caffeine appeared to be associated with both the 2-catechols/methylated 2-catechols and the catechols/methylated catechols ratios; these associations may be attributable to higher levels of 2-catechols among women in the highest level of caffeine intake. Additionally, caffeine was positively associated with 16αhydroxyestrone, while a nonsignificant positive association between this metabolite and coffee was also observed. The similarities in these associations across exposures may be explained by the fact that coffee accounted for 72% of caffeine intake among NHSII participants in Metabolism of caffeine is catalyzed primarily by CYP1A2 enzymes, and laboratory and human studies have suggested that caffeine is as an inducer of CYP1A2 activity. 12 This hepatic enzyme also plays a key role in 2-, 4-, and 16-hydroxylation of parent estrogens, though it is believed to be more active in hydroxylation at the C-2 position of the parent estrogens than at the C-4 or C- 16α positions. 11 Although we did not observe an association with caffeine, our finding of an association between coffee and 2-catechols is notable, as coffee, independent of caffeine, may 20

30 also act as an inducer of CYP1A2, 32,33 potentially by polycyclic aromatic hydrocarbons (PAHs) produced by high-temperature brewing. 34 Tea and decaffeinated coffee were associated with fewer individual EM and ratio measures than coffee and caffeine. We observed a positive association between tea intake and one individual EM, 17-epiestriol. Tea was also inversely associated with the 2-catechols/methylated-2 catechols and the catechols/methylated catechols ratios; interestingly, these associations were in the opposite direction of those observed for caffeine and caffeinated coffee. Decaffeinated coffee was associated with lower levels of two 16-pathway metabolites, estriol and 17-epiestriol, but not with any ratios, groups or pathways. Decaffeinated and regular coffee differ in their caffeine content, and potentially in their antioxidant composition. Chemical analyses have detected lower levels of polyphenols in decaffeinated coffee compared to regular coffee, possibly as a result of the decaffeination process. 35,36 Additionally, it is possible that the relatively low range of decaffeinated coffee and tea consumption in our study limited our ability to detect significant associations. The strengths of our study include large sample size, precisely timed luteal urine samples, and precise and accurate quantification of 15 individual EM using a high-throughput approach. However, the cross-sectional nature of our study limits our ability to conclusively establish temporality. Further, we only collected a single, luteal phase urine sample. Though luteal phase urinary EM have relatively high reproducibility over a 3-year period, 22 we cannot extrapolate the associations with coffee, caffeine, and tea intake over other phases of the menstrual cycle. While unmeasured confounding is always a possibility in an observational study, we carefully 21

31 controlled for several factors associated with caffeine intake and urinary EM. For example, smoking was associated with caffeine intake in our population, and has been shown to be associated with patterns of estrogen metabolism; 37 however, we adjusted for smoking, and results were unchanged when restricted to non-smokers. Our assessment of exposure relied on selfreported data. However, results of a recent study in premenopausal women suggest that measurements of caffeine from FFQ and 24-hour recall are highly correlated (r=0.73, P<0.01), though FFQ measurements tend to be higher. 38 Though we were not able to distinguish between specific types of coffees and teas, which may vary in caffeine content and chemical composition, associations between coffee and caffeine intakes derived from FFQ data and other outcomes in the Nurses Health Studies suggest that self-reported intake is relatively well measured. 39,40 Lastly, the large number of multiple comparisons performed could lead to the finding of statistically significant associations by chance. Due to correlations between metabolites, a Bonferroni-correction may be overly conservative. Nonetheless, at a corrected P-value of (0.05/15 individual EM), some results remained statistically significant, including associations of 2-hydroxyestrone and 2-hydroxyestradiol with coffee. In conclusion, our results suggest a relation of coffee and caffeine consumption with patterns of estrogen metabolism in premenopausal women. Our results support previous findings of a relationship between coffee and 2-hydroxyestrone, and possibly with the other 2-catechol, 2- hydroxyestradiol. Additionally, we found associations between caffeine and 16-pathway metabolites, and suggestive associations with methylated 2-catechols. Further studies are needed to confirm our results. 22

32 Funding/support: This study was supported by Infrastructure Grant UM1 CA and Research Grants CA67262 and CA50385 from the National Cancer Institute and the Intramural Research Program of the Division of Cancer Epidemiology and Genetics of the National Cancer Institute, and with federal funds of the National Cancer Institute awarded under Contract HHSN E to SAIC-Frederick. Julia Sisti was funded by R25 CA and T32 CA

33 REFERENCES 1. Ganmaa, D. et al. Coffee, tea, caffeine and risk of breast cancer: a 22-year follow-up. Int. J. Cancer 122, (2008). 2. Li, J. et al. Coffee consumption modifies risk of estrogen-receptor negative breast cancer. Breast Cancer Res. 13, R49 (2011). 3. Boggs, D. A. et al. Tea and coffee intake in relation to risk of breast cancer in the Black Women s Health Study. Cancer Causes Control 21, (2010). 4. Gierach, G. L. et al. Coffee intake and breast cancer risk in the NIH-AARP diet and health study cohort. Int. J. Cancer 131, (2012). 5. Ishitani, K., Lin, J., Manson, J. E., Buring, J. E. & Zhang, S. M. Caffeine consumption and the risk of breast cancer in a large prospective cohort of women. Arch. Intern. Med. 168, (2008). 6. Larsson, S. C., Bergkvist, L. & Wolk, A. Coffee and black tea consumption and risk of breast cancer by estrogen and progesterone receptor status in a Swedish cohort. Cancer Causes Control 20, (2009). 7. Tang, N., Zhou, B., Wang, B. & Yu, R. Coffee consumption and risk of breast cancer: a metaanalysis. Am. J. Obstet. Gynecol. 200, 290.e1 9 (2009). 8. Jiang, W., Wu, Y. & Jiang, X. Coffee and caffeine intake and breast cancer risk: An updated dose-response meta-analysis of 37 published studies. Gynecol. Oncol. (2013). doi: /j.ygyno Buck, K., Zaineddin, A. K., Vrieling, A., Linseisen, J. & Chang-Claude, J. Meta-analyses of lignans and enterolignans in relation to breast cancer risk. Am. J. Clin. Nutr. 92, (2010). 24

34 10. Mense, S. M., Hei, T. K., Ganju, R. K. & Bhat, H. K. Phytoestrogens and breast cancer prevention: possible mechanisms of action. Environ. Health Perspect. 116, (2008). 11. Lee, A. J., Cai, M. X., Thomas, P. E., Conney, A. H. & Zhu, B. T. Characterization of the oxidative metabolites of 17beta-estradiol and estrone formed by 15 selectively expressed human cytochrome p450 isoforms. Endocrinology 144, (2003). 12. Butler, M. A., Iwasaki, M., Guengerich, F. P. & Kadlubar, F. F. Human cytochrome P- 450PA (P-450IA2), the phenacetin O-deethylase, is primarily responsible for the hepatic 3- demethylation of caffeine and N-oxidation of carcinogenic arylamines. Proc. Natl. Acad. Sci. U.S.A. 86, (1989). 13. Yue, W., Yager, J. D., Wang, J.-P., Jupe, E. R. & Santen, R. J. Estrogen receptor-dependent and independent mechanisms of breast cancer carcinogenesis. Steroids 78, (2013). 14. Cavalieri, E., Frenkel, K., Liehr, J. G., Rogan, E. & Roy, D. Estrogens as endogenous genotoxic agents--dna adducts and mutations. J. Natl. Cancer Inst. Monographs (2000). 15. Lippert, C., Seeger, H. & Mueck, A. O. The effect of endogenous estradiol metabolites on the proliferation of human breast cancer cells. Life Sci. 72, (2003). 16. Telang, N. T., Suto, A., Wong, G. Y., Osborne, M. P. & Bradlow, H. L. Induction by estrogen metabolite 16 alpha-hydroxyestrone of genotoxic damage and aberrant proliferation in mouse mammary epithelial cells. J. Natl. Cancer Inst. 84, (1992). 17. Eliassen, A. H. et al. Urinary estrogens and estrogen metabolites and subsequent risk of breast cancer among premenopausal women. Cancer Res. 72, (2012). 25