Original Contribution

Transcription

1 American Journal of Epidemiology ª The Author Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please Vol. 172, No. 12 DOI: /aje/kwq321 Advance Access publication: November 4, 2010 Original Contribution Menstrual and Reproductive Factors, Exogenous Hormone Use, and Gastric Cancer Risk in a Cohort of Women From the European Prospective Investigation Into Cancer and Nutrition Eric J. Duell*, Noémie Travier, Leila Lujan-Barroso, M. C. Boutron-Ruault, F. Clavel-Chapelon, Domenico Palli, Vittorio Krogh, Amalia Mattiello, Rosario Tumino, Carlotta Sacerdote, Laudina Rodriguez, Emilio Sanchez-Cantalejo, Carmen Navarro, Aurelio Barricarte, Miren Dorronsoro, Kay-Tee Khaw, Nicholas Wareham, Naomi E. Allen, Konstantinos K. Tsilidis, H. Bas Bueno-de-Mesquita, Suzanne M. Jeurnink, M. E. Numans, Petra H. M. Peeters, Pagona Lagiou, Elisabeth Valanou, Antonia Trichopoulou, Rudolf Kaaks, Annekatrin Lukanova- McGregor, Manuela M. Bergman, Heiner Boeing, Jonas Manjer, Björn Lindkvist, Roger Stenling, Göran Hallmans, Christina C. Dahm, Kim Overvad, Anja Olsen, Anne Tjonneland, Kjersti Bakken, Eiliv Lund, Mazda Jenab, Valerie McCormack, Sabina Rinaldi, Dominique Michaud, Traci Mouw, Gabriella Nesi, Fatima Carneiro, Elio Riboli, and Carlos A. González * Correspondence to Dr. Eric J. Duell, Unit of Nutrition, Environment and Cancer, Cancer Epidemiology Research Program, Catalan Institute of Oncology, Avda Gran Via , L Hospitalet de Llobregat, Barcelona, Spain ( eduell@iconcologia.net). Initially submitted May 10, 2010; accepted for publication August 24, The worldwide incidence of gastric adenocarcinoma (GC) is lower in women than in men. Furthermore, cancer patients treated with estrogens have been reported to have a lower subsequent risk of GC. The authors conducted a prospective analysis of menstrual and reproductive factors, exogenous hormone use, and GC in 335,216 women from the European Prospective Investigation Into Cancer and Nutrition, a cohort study of individuals aged years from 10 European countries. After a mean follow-up of 8.7 years (through 2004), 181 women for whom complete exposure data were available developed GC. Adjusted hazard ratios and 95% confidence intervals were estimated using Cox proportional hazards models. All statistical tests were 2-sided. Women who had ovariectomy had a 79% increased risk of GC (based on 25 cases) compared with women who did not (hazard ratio ¼ 1.79, 95% confidence interval: 1.15, 2.78). Total cumulative years of menstrual cycling was inversely associated with GC risk (fifth vs. first quintile: hazard ratio ¼ 0.55, 95% confidence interval: 0.31, 0.98; P trend ¼ 0.06). No other reproductive factors analyzed were associated with risk of GC. The results of this analysis provide some support for the hypothesis that endogenous ovarian sex hormones lower GC incidence in women. estrogens; hormones; menstruation; reproduction; stomach neoplasms; women Abbreviations: BMI, body mass index; CI, confidence interval; EPIC, European Prospective Investigation Into Cancer and Nutrition; GC, gastric adenocarcinoma; HR, hazard ratio; HRT, hormone replacement therapy; OC, oral contraceptive; TFF, trefoil factor. Although the incidence of gastric adenocarcinoma (GC) has been steadily declining, it is still the second leading cause of cancer death worldwide. In most regions, men have a higher age-standardized incidence than do women, with a ratio of about 2:1 (1). This sex difference is consistent across international populations with different prevalences of environmental risk factors such as Helicobacter pylori infection and tobacco smoking and different dietary patterns (1, 2). A possible explanation involves biologic differences related to sex hormones such as estrogen (2). 1384

2 Menstrual Factors and Gastric Cancer Risk 1385 The most compelling evidence for an association between hormonal factors and GC comes from follow-up studies of cancer patients treated with estrogen or antiestrogen compounds (3 5). In one large follow-up study of prostate cancer patients treated with estrogen (in Sweden before the 1980s), men receiving these treatments had a significantly lower subsequent risk of GC than did similar men who did not receive estrogen treatment (4). In 2 cohorts of breast cancer patients treated with the antiestrogen tamoxifen, women receiving treatment had a significantly increased risk of subsequent GC (3, 5). The above studies suggest that estrogen compounds may prevent or slow the growth of gastric tumors. To date, there have been 13 epidemiologic studies of the relation between reproductive and menstrual factors and GC, comprising 9 retrospective studies, including 1 mortality study (6 14), and 4 prospective cohort studies (15 20). These studies have generally indicated that menstrual factors associated with longer total time of menstrual cycling and prolonged use of exogenous hormones are associated with a lower risk of GC; however, many of the studies lacked power, and some may have been influenced by recall bias. We investigated the role of reproductive and menstrual factors and exogenous hormone use in the cause of GC in women from the European Prospective Investigation Into Cancer and Nutrition (EPIC), a large prospective cohort study of over 300,000 women recruited through 23 research centers in 10 European countries. MATERIALS AND METHODS Study populations EPIC is a multicenter prospective cohort study of 521,457 participants (368,010 women and 153,447 men) recruited through 23 centers in 10 European countries, including Denmark (Aarhus and Copenhagen), France, Germany (Heidelberg and Potsdam), Greece, Italy (Florence, Turin, Varese, Naples, and Ragusa), the Netherlands (Bilthoven and Utrecht), Norway, Spain (Asturias, Granada, Murcia, Navarra, and San Sebastian), Sweden (Malmö and Umeå), and the United Kingdom (Oxford and Cambridge) (21). Most of the EPIC participants were enrolled from 1992 to 1998, when they were between the ages of 35 and 70 years. Participants were recruited from the general population residing in a geographic area, including towns and provinces. Exceptions were the French cohort (teacher s organization health insurance); cohorts consisting of women attending breast cancer screening programs (Utrecht and Florence); parts of the Italian and Spanish cohorts, in which participants were recruited among blood donors; most of the Oxford cohort, in which participants were recruited among vegetarian volunteers; and the German cohort in which participants were recruited from a health care insurance organization. Eligible participants gave written informed consent and completed questionnaires on diet, lifestyle, and medical history. Participants were excluded from analyses if they had widespread cancer at recruitment (n ¼ 23,633) or if no follow-up information was available (n ¼ 3,448).We also excluded men (n ¼ 153,447), women with missing information on all menstrual and reproductive variables (n ¼ 8,166), women with missing dietary information (n ¼ 2,603), and women with GC for whom it was unknown whether their cancer was a primary or secondary tumor (n ¼ 5). The final number of EPIC cohort women available for these analyses was 335,216. Baseline diet and lifestyle questionnaires Usual diet over the previous 12 months was measured using mostly country-specific validated food questionnaires (21). A separate lifestyle questionnaire included items on lifetime smoking and alcohol consumption, education, occupation, menstrual and reproductive history, use of exogenous hormones (oral contraceptives (OCs) and hormone replacement therapy (HRT)), physical activity, and history of illness, including surgical procedures. Identification of GC cases Follow-up of cohort members for these analyses was performed through December Case identification was based on information in population cancer registries except in France, Germany, Greece, and Naples, where a combination of different methods, including health insurance records, hospital-based cancer and pathology registries, and active follow-up (participant tracking), was used. Followup began on the date of EPIC recruitment and ended on the date of GC diagnosis, the date of death, or the end of followup (in 2004), whichever came first. Because these analyses were based primarily upon exposure information obtained at baseline, loss to follow-up was negligible. Cancer of the stomach included diagnoses coded as C16 according to the International Classification of Diseases, Tenth Revision. Among women in EPIC, a total of 237 cases of GC were observed (200 adenocarcinomas and 37 nonadenocarcinomas), of which a total of 193 were incident, primary GCs. All cases were validated for histologic (diffuse or intestinal, using Lauren s classification) (22) and anatomical subtypes (cardia (C16.0) or noncardia (C16.1-9)) by a panel of pathologists who reviewed original pathology reports, tumor slides, and paraffin blocks from GC cases obtained from EPIC centers (23). Nested case-control study of H. pylori infection A nested case-control study within the EPIC cohort (EurGast II, which included 150 incident primary cases of GC in women and 571 matched controls) was conducted to perform analyses of baseline H. pylori seropositivity status (24), as well as other studies of biomarkers and genetic factors. Each incident GC case with an available blood sample was matched by sex, age group (62.5 years), center, and date of blood collection (645 days) to 4 control participants who were randomly selected from the cohort at risk at the time of diagnosis of the index case. H. pylori status was determined using a combination of ELISA (Pyloriset EIA- GIII kit, Orion Diagnostica, Espoo, Finland) and Western blot (HELICO Blot 2.1 kit, Genelab Diagnostics, Singapore) assays. A participant was considered H. pylori

3 1386 Duell et al. seropositive at baseline if she was positive for enzymelinked immunosorbent assay antibody or CagA antigen by Western blot. Statistical methods Hazard ratios as estimates of relative risk for menstrual factors and GC were calculated using Cox proportional hazards regression models with age as the time scale and stratification by EPIC study center and age at recruitment. The proportional hazards assumption was evaluated in all models using graphical methods and likelihood ratio tests. When the data were analyzed according to baseline H. pylori seropositivity status, unconditional logistic regression modeling was used to estimate the odds ratio. Tests for interaction were based on a likelihood ratio test. Chi-square and Wilcoxon rank sum tests were used to compare categorical and continuous baseline cohort characteristics, respectively, stratified by reproductive variables of interest. All statistical tests were 2-sided, and all analyses were performed using STATA (version 10.0, StataCorp LP, College Station, Texas). We estimated hazard ratios for the following factors: age at menarche and menopause (in years), duration of OC use (in years), ever use of HRT, parity, age at first full-term pregnancy, breastfeeding, miscarriage, induced abortion, ovariectomy, hysterectomy, and cumulative duration of menstrual cycling (in years). Cumulative menstrual cycling was calculated for each woman in the EPIC cohort. For postmenopausal women, it was the difference between the age at menopause and the age at menarche minus the total time pregnant (number of full-term pregnancies 3 9 months). For premenopausal and perimenopausal women, cumulative duration of menstrual cycling was the difference between age at recruitment and age at menarche minus the total time pregnant. We also estimated the cumulative duration of menstrual cycling as above, subtracting the total time spent taking OCs. Cutpoints for categorical definitions of cumulative menstrual cycling and other continuous variables were based on distributions (quintiles, quartiles, or tertiles) in the entire EPIC cohort. Self-reported baseline menopausal status was defined as menopausal (natural cessation of menses in the last 12 months or surgical menopause due to bilateral ovariectomy), perimenopausal (no longer naturally menstruating at the time of recruitment or <9 menstrual cycles in the past 12 months), and premenopausal (regular menses or 9 cycles in the past 12 months). The menopausal statuses of women with missing or incomplete questionnaire data were determined based on cutpoints for age at recruitment (postmenopausal if 55 years of age, perimenopausal if years of age, and premenopausal if 46 years of age). Age at baseline menopausal status was based on self-reporting and was missing for 214,063 cohort members (15 GC cases). We adjusted the analyses for average daily intakes of fruits and nuts, vegetables, red meat, and processed meat. Dietary intakes were energy-adjusted using the residual method (25). Total caloric intake was not associated with GC incidence and was not included in the final energyadjusted food-intake models. We included the following variables as potential confounders or known risk factors: baseline and lifetime alcohol consumption, physical activity using the Cambridge index (26), education (none, primary, technical/professional, secondary, or university), cigarette smoking (never, former: time since quitting 10 years, former: time since quitting <10 years, current: <20 cigarettes per day, or current: 20 cigarettes per day), and body mass index (BMI, in weight (kg)/height (m) squared). Alcohol and physical activity were not included in final models because they did not change effect estimates more than 10%. When the data were analyzed according to baseline H. pylori serostatus, adjustment variables included age at recruitment, education level (less than secondary or secondary or higher), cigarette smoking, BMI, and region (north or south) and matching variables. Categories for education level and region were collapsed because of the smaller sample size in the nested case-control study. RESULTS During a mean follow-up of 8.7 years (2,927,994 personyears) starting in 1992, GC was diagnosed in 193 women. Of these women, data on all of the reproductive and menstrual factors analyzed in this study were missing for 10 and dietary information was missing for 2, so they were removed from further analysis, resulting in 181 cases. Details on the EPIC participants who did not have missing data on 1 menstrual and reproductive factor for each of the 10 European countries show that the majority of patients with GC (n ¼ 181) for these analyses were from Italy (n ¼ 36), whereas the fewest were from Greece (n ¼ 9) (Web Table, available at The majority of GC cases in women were located in the non-cardia region (n ¼ 101, 55.8%); in addition, there were 31 cases (17.1%) in the cardia region and 49 cases (27.1%) in unknown or mixed locations. Eighty-two cases (45.3%) were classified as the diffuse histologic subtype, 48 cases (26.5%) were classified as the intestinal subtype, and 53 cases (29.3%) had an unknown, unclassified, or mixed histologic subtype. Twenty-three women with GC were premenopausal at baseline, 19 were perimenopausal, and 139 were postmenopausal (114 naturally and 25 surgically). Baseline characteristics of the women included in these analyses according to HRT use, self-reported ovariectomy, and total years of menstrual cycling are presented in Table 1. On average, women who reported using HRT were older, smoked more, had slightly lower BMIs, attained less secondary or higher education, had lower fruit and nut and vegetable intake, and had higher total red and processed meat intake (Table 1). Women who reported having an ovariectomy on average were older, smoked more, had higher BMIs, attained less secondary or higher education, and had a slightly higher intake of fruits and nuts and of total red and processed meats (Table 1). Women with the highest (vs. lowest) quintile of cumulative years of menstrual cycling on average were older, smoked less, had higher BMIs, attained less education, and had a lower intake of vegetables but a higher intake of fruits and nuts and total red and processed meat (Table 1). The prevalence of baseline

4 Table 1. Baseline Characteristics of Women in the European Prospective Investigation Into Cancer and Nutrition Cohort by Hormonal Factors a, EPIC Any HRT Use Ovariectomy Total Menstrual Cycling, years Cohort Women No Yes No Yes Quintile 1 ( ) Quintile 5 ( ) Mean (SD) % Mean (SD) % Mean (SD) % Mean (SD) % Mean (SD) % Mean (SD) % Mean (SD) % Age, years (9.7) (10.32) (5.76) (10.21) (7.71) (11.50) (5.47) Ever tobacco smoker Body mass index, kg/m (4.46) (4.63) (4.03) (4.54) (4.63) (4.49) (4.60) Secondary or higher education Energy intake, kcal/day Total vegetable intake, g/day Total fruit and nut intake, g/day Total red meat intake, g/day Total processed meat intake, g/day Prevalence of Helicobacter pylori positive infection in noncases b 1, (597.52) 1, (606.68) 1, (578.25) 1, (606.98) 1, (592.13) 1, (603.18) 1, (573.04) (146.23) (153.47) (134.65) (150.12) (149.48) (163.54) (138.82) (184.33) (192.26) (174.44) (190.89) (187.71) (200.77) (184.83) (32.14) (32.88) (32.43) (33.66) (32.36) (31.61) (31.54) (25.90) (24.43) (24.55) (25.93) (24.89) (28.71) (23.53) Abbreviations: EPIC, European Prospective Investigation Into Cancer and Nutrition; HRT, hormone replacement therapy; SD, standard deviation. a P values were based on a Wilcoxon rank sum test for continuous variables, and a chi-square test for categorical variables (all P values 0.001, except for total energy by total menstrual cycling, P ¼ 0.68). b Based on a nested case-control (EurGast II) study with 571 controls from EPIC; chi-square P values for Helicobacter pylori positive prevalence by any HRT use: 0.02; ovariectomy: 0.8; and years of menstrual cycling: 0.2. Menstrual Factors and Gastric Cancer Risk 1387

5 1388 Duell et al. Table 2. Reproductive and Menstrual Risk Factors for Gastric Adenocarcinoma in Women From the European Prospective Investigation Into Cancer and Nutrition Cohort, Variable Person-Years Adenocarcinoma Cases (n 5 181) HR a 95% CI Age at menarche, years <12 443, , , , (Referent) 748, Referent , , , , 1.56 Missing 2 Age at menopause b, years <46 238, Referent , , , , , , , , 2.99 Missing 15 Duration of OC use, years 0 1,205, Referent 1 300, , , , , , 2.02 Missing 9 Hormone replacement therapy No 1,927, Referent Yes 715, , 1.25 Missing 15 Postmenopausal hormone replacement therapy b No 732, Referent Yes 570, , 1.24 Missing 12 Parity (no. of full-term pregnancies) 0 402, Referent 1 429, , ,124, , , , , , 2.18 Missing 6 Age at first full-term pregnancy, years <22 536, Referent , , , , , , 2.07 Missing 24 Breastfeeding No 789, Referent Yes 1,889, , 1.17 Missing 11 Table continues

6 Menstrual Factors and Gastric Cancer Risk 1389 Table 2. Continued Variable Person-Years Adenocarcinoma Cases (n 5 181) HR a 95% CI Miscarriage No 1,656, Referent Yes 530, , 1.30 Missing 35 Induced abortion No 1,636, Referent Yes 403, , 1.69 Missing 47 Ovariectomy No 2,267, Referent 1 ovary removed c 100, , ovaries removed 86, , 3.24 Missing 30 Hysterectomy (in women without ovariectomy) No 2,104, Referent Yes 159, , 2.14 Missing 1 Total cumulative menstrual cycles in years, by quintile Quintile 1: , Referent Quintile 2: , , 1.22 Quintile 3: , , 1.34 Quintile 4: , , 1.06 Quintile 5: , , 0.98 Missing 20 P trend Total cumulative menstrual cycles in years, accounting for OC use, by quintile Quintile 1: , Referent Quintile 2: , , 2.55 Quintile 3: , , 2.23 Quintile 4: , , 1.25 Quintile 5: , , 1.65 Missing 25 P trend Abbreviations: CI, confidence interval; HR, hazard ratio; OC, oral contraceptive. a Adjusted for age, center, smoking status, education, body mass index (weight (kg)/height (m) 2 ), and calorie-adjusted vegetable, fruit, red meat, and processed meat intakes. b Among postmenopausal women (includes natural and surgical menopause); 15 cases were missing data on age at menopause, 23 cases were classified as premenopausal, and 19 cases were classified as perimenopausal (and were not considered postmenopausal). c For 1 case, it was unknown whether 1 or 2 ovaries were removed. H. pylori seropositivity in noncases was higher in those who did not use HRT and in women in the lowest quintile of total years of menstrual cycling, but it was similar by history of ovariectomy (Table 1). Neither age at menarche nor age at menopause was statistically significantly associated with GC risk (Table 2). Similarly, duration of OC use, HRT, parity, age at first full-term pregnancy, breastfeeding, miscarriage, and induced abortion were not associated with GC risk (Table 2). Women who reported having had an ovariectomy had a nearly 2-fold increased risk of GC (hazard ratio (HR) ¼ 1.79, 95% confidence interval (CI): 1.15, 2.78; data not

7 1390 Duell et al. Table 3. Ovariectomy and HRT Use and Gastric Adenocarcinoma in Women From the European Prospective Investigation Into Cancer and Nutrition Cohort a, Ovaries Present/ Ever Used HRT Person-Years All Women Adenocarcinoma Cases HR c 95% CI Person-Years Postmenopausal Women b Adenocarcinoma Cases Yes/yes 503, Referent 407, Referent Yes/no 1,610, , , , 2.29 No/yes 88, , , , 4.42 No/no 92, , , , 4.62 HR c 95% CI Abbreviation: CI, confidence interval; HR, hazard ratio; HRT, hormone replacement therapy. a Data on ovariectomy and/or HRT use were missing for 34 women (24 postmenopausal). b Postmenopausal status at baseline. c Adjusted for age, center, smoking status, education level, body mass index (kg/m 2 ), and calorie-adjusted vegetable, fruit, red meat, and processed meat intakes. shown). The magnitude of the association was similar for unilateral and bilateral ovariectomy (Table 2). One woman with GC who reported ovariectomy (unilateral) was classified as premenopausal at baseline. Of the 34 women who reported having had a hysterectomy, 19 also reported having had an ovariectomy. In women who reported never having an ovariectomy, there was no association between hysterectomy and GC (Table 2). The hazard ratios for ovariectomy and GC were similar when the analysis was restricted to postmenopausal women (HR ¼ 1.84, 95% CI: 1.16, 2.93). Hazard ratios for ovariectomy and GC by age at surgery were 1.71 (95% CI: 1.03, 2.83) for women <50 years of age and 2.17 (95% CI: 0.94, 4.98) for women 50 years of age. When GC was restricted to location (cardia vs. noncardia), patterns of association for ovariectomy were statistically nonsignificant (noncardia: HR ¼ 1.39, 95% CI: 0.73, 2.67; cardia: HR ¼ 0.67, 95% CI: 0.16, 2.88). Hazard ratios for ovariectomy and GC were elevated (but not statistically significant) when the cases were restricted to diffuse (HR ¼ 1.84, 95% CI: 0.92, 3.69) or intestinal (HR ¼ 1.31, 95% CI: 0.54, 3.16) histologic subtypes. The total cumulative number of years of menstrual cycling was inversely associated with risk of GC (Table 2). Inverse associations between total years of cycling and risk of GC also were observed when we accounted for total duration of OC use (Table 2). Associations between total menstrual cycling and GC in postmenopausal women also showed inverse relations, but trend tests and individual hazard ratios were not statistically significant (HR for quartile of total menstrual cycling in years: first quartile: years (referent), second quartile: years (HR ¼ 0.91, 95% CI: 0.51, 1.65); third quartile: years (HR ¼ 0.69, 95% CI: 0.38, 1.24); fourth quartile: years (HR ¼ 0.82, 95% CI: 0.47, 1.43)). Patterns of association for total years of menstrual cycling and GC were similar when stratified by smoking status and by anatomic subsite (data not shown). We analyzed HRT use together with ovariectomy in 1 combined variable, with women who had both ovaries present and who used HRT (highest potential estrogen exposure) as the referent group. Although some estimates lacked precision, the group of women who had ovariectomy and who never had used HRT were at the highest risk of GC (Table 3). A likelihood ratio test for interaction between ovariectomy and HRT use in relation to risk of GC in all women was not statistically significant (P ¼ 0.87). Associations for ovariectomy-hrt use and GC were of similar magnitude when the data were restricted to postmenopausal women (Table 3). We evaluated associations between ovariectomy, total years of menstrual cycling, and risk of GC by baseline H. pylori serostatus (positive or negative) in the nested case-control study of GC in EPIC. There were insufficient numbers of women with GC who were H. pylori negative and who reported having an ovariectomy; thus, odds ratios in this group could not be estimated. In H. pylori positive women, associations between ovariectomy and HRT use and GC were very similar in magnitude and direction to the hazard ratios presented in Table 3 (data not shown). Tertiles of total cumulative years of menstrual cycling were defined as follows: tertile 1: ; tertile 2: ; and tertile 3: years. For associations between total cumulative years of menstrual cycling and GC, we observed the following odds ratios and 95% confidence intervals in H. pylori positive women: tertile 1: referent (33 cases and 94 controls); tertile 2: odds ratio ¼ 1.11, 95% confidence interval: 0.64, 1.94 (43 cases and 112 controls); and tertile 3: odds ratio ¼ 1.05, 95% confidence interval: 0.60, 1.81 (44 cases and 126 controls). In H. pylori negative women, we observed the following odds ratios and 95% confidence intervals: tertile 1: referent (4 cases and 48 controls); tertile 2: odds ratio ¼ 0.75, 95% confidence interval: 0.19, 2.98 (6 cases and 65 controls); and tertile 3: odds ratio ¼ 0.73, 95% confidence interval: 0.18, 3.02 (6 cases and 76 controls). DISCUSSION In the present prospective cohort analysis of menstrual factors, exogenous hormone use, and GC, we observed a hazard ratio of 1.79 in women who at baseline reported having had an ovariectomy (compared with women who did not). Furthermore, women who reported ovariectomy and nonuse of HRT had a hazard ratio of 2.25 compared with women who had intact ovaries and who reported ever use of HRT. We also observed some evidence of lower risk of GC

8 Menstrual Factors and Gastric Cancer Risk 1391 in women who had greater total cumulative years of menstrual cycling. On the basis of our data, it does not appear that baseline H. pylori serostatus modifies the associations that we report here, which corresponds to results from a recent prospective study from Shanghai, China (16). We also did not observe modification of associations by smoking status. To our knowledge, the present study is the first to observe a statistically significant higher risk of GC in women who at baseline reported having had an ovariectomy. Using data from the National Institutes of Health-AARP cohort, Freedman et al. (20) reported elevated risks for esophageal and gastric cardia adenocarcinomas combined (n ¼ 65 cases) in patients who had undergone ovariectomy, but none of the associations were statistically significant. Regarding total years of menstrual cycling, 4 retrospective studies (8, 9, 12, 14) and 2 prospective studies (10, 16) have reported inverse associations with GC risk of about the same magnitude as reported here. Our observations, along with previous evidence, including the studies of hormone treatment in cancer patients (3 5), support the notion that ovarian sex hormones and related compounds alter the risk of GC. The direction of the relative risks described in this and previous reports suggests that greater or prolonged estrogen-related exposure is associated with a lower risk of GC (2). Our results showing that the highest risk of GC is in women who had ovariectomy and who never used HRT support this; however, a number of studies of endogenous hormone levels in ovariectomized women suggest that postmenopausal women who undergo bilateral ovariectomy have reduced serum concentrations of androgens (rather than estrogens) compared with postmenopausal women with intact ovaries (27 32). When we investigated hazard ratios for ovariectomy and GC by age at ovariectomy (<50 vs. 50 years of age), we observed hazard ratios of similar magnitude, but only the hazard ratio in women <50 years of age was statistically significant. Furthermore, the majority of women with GC in this study (n ¼ 139, or 76.8%) were postmenopausal at baseline. In general, women who undergo these surgeries are older (>45 years of age), and reasons include ovarian cysts and cancer prevention (31). Steroid hormone metabolism is extremely complex, and multiple external and internal factors are believed to influence circulating levels of sex steroids and metabolites. Furthermore, postmenopausal hormone production and metabolism are still poorly understood. Potential biologic mechanisms for the association between hormonal factors and GC may include pathways related to the estrogen receptor or other related pathways, for example, the trefoil factor (TFF) genes (2). TFF gene products protect the gastric mucosa from endogenous and exogenous insults (33), and TFF gene expression is known to be reduced in precancerous gastric lesions and in GC (34). Estrogen is known to increase the expression of TFF genes and could be one mechanism by which estrogens exert a protective effect in gastric carcinogenesis (2, 35, 36). Estrogens also interact with estrogen receptors that are known to be present in gastric mucosal cells and in gastric cancer cells and may repress the growth and clonal expansion of precancerous gastric lesions and cancerous gastric cells (2, 37). It is important to note that we were unable to evaluate HRT formulation in our study (e.g., whether it comprised unopposed estrogen or estrogen plus progesterone). Furthermore, estrogens are derived from androstenedione, and estrogen and other steroid levels were not directly evaluated in our study. Thus, our observations could be consistent with those seen with 1 or more other sex steroids, such as androgens or progesterone, and even with other factors that are associated with the menstrual factors we evaluated. For example, diet, smoking, and socioeconomic status may influence some menstrual factors (and are GC risk factors); thus, they may have been potential confounders in these analyses. Nevertheless, we were able to adjust for energy-adjusted fruit, vegetable, and meat intakes at baseline, as well as smoking status, BMI, and educational level. It is possible that we were unable to completely adjust for these imprecisely measured factors, which may have led to some residual confounding; however, the strength of the associations between these potential confounders and GC and menstrual factors is unlikely to be sufficient to appreciably confound the associations we report here. There is a potential for case underascertainment in some EPIC study centers with active follow-up, but we do not expect this to be associated with the reproductive factors we analyzed or with baseline GC risk; thus, the main consequence would be decreased statistical precision in our estimates of relative risk (38). Additional weaknesses of our study were the fact that the menstrual and reproductive variables, including history of ovariectomy, were based on self reporting, and the relatively small number of women with GC who reported having had an ovariectomy. The strengths of this study include case validation by a panel of pathologists (23) and the prospective cohort design, which allowed us to prospectively collect and analyze information on menstrual factors, exogenous hormone use, and covariates and to avoid the biases associated with retrospective study designs. In conclusion, our prospective analysis provides additional evidence that menstrual and reproductive factors such as cumulative years of menstrual cycling and ovariectomy alter the risk of GC in women. Given the results of this and past studies, it is now time to investigate the potential mechanisms of these protective effects with the hope of translating this knowledge to improve the prevention and treatment of this deadly cancer. ACKNOWLEDGMENTS Author affiliations: Unit of Nutrition, Environment, and Cancer, Cancer Epidemiology Research Program, Catalan Institute of Oncology (ICO-IDIBELL), Barcelona, Spain (Eric J. Duell, Noémie Travier, Leila Lujan-Barroso, Carlos A. González); Centre for Research in Epidemiology and Population Health, INSERM, Institut Gustave Roussy, Villejuif, France (M. C. Boutron-Ruault, F. Clavel-Chapelon); Paris South University, Villejuif, France (M. C. Boutron- Ruault, F. Clavel-Chapelon); Molecular and Nutritional Epidemiology Unit, Cancer Research and Prevention Institute (ISPO), Florence, Italy (Domenico Palli); Department of

9 1392 Duell et al. Preventive and Predictive Medicine, Nutritional Epidemiology Unit, Fondazione IRCCS Istituto Nazionale dei Tumori Milan, Milan, Italy (Vittorio Krogh); Department of Clinical and Experimental Medicine, Federico II University, Naples, Italy (Amalia Mattiello); Cancer Registry and Histopathology Unit, Civile M.P. Arezzo Hospital, Ragusa, Italy (Rosario Tumino); Center for Cancer Prevention (CPO Piedmont), Turin, Italy (Carlotta Sacerdote); Human Genetic Foundation (HuGeF), Florence, Italy (Carlotta Sacerdote); Public Health and Participation Directorate, Health and Health Care Services Council, Asturias, Spain (Laudina Rodriguez); Andalusian School of Public Health, Granada, Spain (Emilio Sanchez-Cantalejo); CIBER Epidemiología y Salud Pública (CIBERESP), Granada, Spain (Emilio Sanchez-Cantalejo); Department of Epidemiology, Murcia Health Council, Murcia, Spain (Carmen Navarro); CIBER- ESP, Murcia, Spain (Carmen Navarro); Navarra Public Health Institute, Pamplona, Spain (Aurelio Barricarte); CIBERESP, Pamplona, Spain (Aurelio Barricarte); Public Health Division of Gipuzkoa, Basque Regional Health Department, Gipuzkoa, Spain (Miren Dorronsoro); Department of Public Health and Primary Care, University of Cambridge, Cambridge, United Kingdom (Kay-Tee Khaw, Nicholas Wareham); Cancer Epidemiology Unit, Nuffield Department of Clinical Medicine, University of Oxford, Oxford, United Kingdom (Naomi E. Allen, Konstantinos K. Tsilidis); National Institute for Public Health and the Environment (RIVM), Bilthoven, the Netherlands (H. Bas Bueno-de-Mesquita); Department of Gastroenterology and Hepatology, University Medical Centre Utrecht, Utrecht, the Netherlands (Suzanne M. Jeurnink); Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, the Netherlands (M. E. Numans, Petra H. M. Peeters); WHO Collaborating Center for Food and Nutrition Policies, Department of Hygiene, Epidemiology and Medical Statistics, University of Athens Medical School, Athens, Greece (Pagona Lagiou, Elisabeth Valanou, Antonia Trichopoulou); Hellenic Health Foundation, Athens, Greece (Elisabeth Valanou, Antonia Trichopoulou); Department of Cancer Epidemiology, German Cancer Research Center, Heidelberg, Germany (Rudolf Kaaks, Annekatrin Lukanova- McGregor); Department of Epidemiology, German Institute of Human Nutrition, Potsdam-Rehbrücke, Germany (Manuela M. Bergman, Heiner Boeing); Department of Surgery, Skåne University Hospital Malmö, Lund University, Malmö, Sweden (Jonas Manjer); Department of Internal Medicine, Division of Gastroenterology and Hepatology, Sahlgrenska University Hospital, Gothenburg, Sweden (Björn Lindkvist); Department of Medical Biosciences, Pathology, Umeå University, Umeå, Sweden (Roger Stenling); Department of Public Health and Clinical Medicine, Nutritional Research, Umeå University, Umeå, Sweden (Göran Hallmans); Department of Clinical Epidemiology, Aarhus University Hospital, Aalborg, Denmark (Christina C. Dahm); Department of Epidemiology, School of Public Health, Aarhus University, Aarhus, Denmark (Christina C. Dahm, Kim Overvad); Danish Cancer Society, Institute of Cancer Epidemiology, Diet Cancer and Health, Copenhagen, Denmark (Anja Olsen, Anne Tjonneland); Department of Community Medicine, University of Tromsø, Tromsø, Norway (Kjersti Bakken, Eiliv Lund); International Agency for Research on Cancer, Lyon, France (Mazda Jenab, Valerie McCormack, Sabina Rinaldi); Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom (Dominique Michaud, Traci Mouw); Division of Pathological Anatomy, University of Florence, Florence, Italy (Gabriella Nesi); Institute of Molecular Pathology and Immunology of the University of Porto (IPATIMUP), Porto, Portugal (Fatima Carneiro); Medical Faculty/HS João, Porto, Portugal (Fatima Carneiro); and School of Public Health, Imperial College London, St. Mary s Campus, Imperial College, London, United Kingdom (Elio Riboli). The coordination of EPIC is financially supported by the European Commission (DG-SANCO) and the International Agency for Research on Cancer. The national cohorts are supported by the Health Research Fund (FIS) of the Spanish Ministry of Health (grant Exp P ), the regional governments of Andalucía, Asturias, Basque Country, Murcia and Navarra, La Caixa (grants BM , RTICCC- C03/10, and R06/0020) (Spain); the Danish Cancer Society (Denmark); Ligue contre le Cancer, 3M, Mutuelle Générale de l Education Nationale, Institut National de la Santé et de la Recherche Medicale (France); Deutsche Krebshilfe, Deutsches Krebsforschungszentrum and Federal Ministry of Education and Research (Germany); Ministry of Health and Social Solidarity, Stavros Niarchos Foundation, and Hellenic Health Foundation (Greece); Italian Association for Research on Cancer (AIRC) and National Research Council (Italy); Dutch Ministry of Public Health, Welfare, and Sports (VWS), Netherlands Cancer Registry (NKR), LK Research Funds, Dutch Prevention Funds, Dutch ZON (Zorg Onderzoek Nederland), World Cancer Research Fund (WCRF), and Statistics Netherlands (the Netherlands); Norwegian Cancer Society (Norway); Swedish Cancer Society, Swedish Scientific Council and Regional Government of Skåne and Västerbotten (Sweden); and Cancer Research United Kingdom, Medical Research Council, Stroke Association, British Heart Foundation, Department of Health, Food Standards Agency, and Wellcome Trust (United Kingdom). Conflict of interest: none declared. REFERENCES 1. Parkin DM, Bray F, Ferlay J, et al. Global cancer statistics, CA Cancer J Clin. 2005;55(2): Chandanos E, Lagergren J. Oestrogen and the enigmatic male predominance of gastric cancer. Eur J Cancer. 2008;44(16): Chandanos E, Lindblad M, Rubio CA, et al. Tamoxifen exposure in relation to gastric adenocarcinoma development. Eur J Cancer. 2008;44(7): Lindblad M, Ye W, Rubio C, et al. Estrogen and risk of gastric cancer: a protective effect in a nationwide cohort study of patients with prostate cancer in Sweden. Cancer Epidemiol Biomarkers Prev. 2004;13(12): Matsuyama Y, Tominaga T, Nomura Y, et al. Second cancers after adjuvant tamoxifen therapy for breast cancer in Japan. Ann Oncol. 2000;11(12): Chandanos E, Rubio CA, Lindblad M, et al. Endogenous estrogen exposure in relation to distribution of histological type

10 Menstrual Factors and Gastric Cancer Risk 1393 and estrogen receptors in gastric adenocarcinoma. Gastric Cancer. 2008;11(3): Fernandez E, Gallus S, Bosetti C, et al. Hormone replacement therapy and cancer risk: a systematic analysis from a network of case-control studies. Int J Cancer. 2003;105(3): Frise S, Kreiger N, Gallinger S, et al. Menstrual and reproductive risk factors and risk for gastric adenocarcinoma in women: findings from the Canadian National Enhanced Cancer Surveillance System. Ann Epidemiol. 2006;16(12): Inoue M, Ito LS, Tajima K, et al. Height, weight, menstrual and reproductive factors and risk of gastric cancer among Japanese postmenopausal women: analysis by subsite and histologic subtype. Int J Cancer. 2002;97(6): Kaneko S, Tamakoshi A, Ohno Y, et al. Menstrual and reproductive factors and the mortality risk of gastric cancer in Japanese menopausal females. Cancer Causes Control. 2003;14(1): La Vecchia C, Negri E, Franceschi S, et al. Long-term impact of reproductive factors on cancer risk. Int J Cancer. 1993;53(2): La Vecchia C, D Avanzo B, Franceschi S, et al. Menstrual and reproductive factors and gastric-cancer risk in women. Int J Cancer. 1994;59(6): Lindblad M, García Rodríguez LA, Chandanos E, et al. Hormone replacement therapy and risks of oesophageal and gastric adenocarcinomas. Br J Cancer. 2006;94(1): Palli D, Cipriani F, Decarli A, et al. Reproductive history and gastric cancer among post-menopausal women. Int J Cancer. 1994;56(6): Dorjgochoo T, Shu XO, Li HL, et al. Use of oral contraceptives, intrauterine devices and tubal sterilization and cancer risk in a large prospective study, from 1996 to Int J Cancer. 2009;124(10): Freedman ND, Chow WH, Gao YT, et al. Menstrual and reproductive factors and gastric cancer risk in a large prospective study of women. Gut. 2007;56(12): Heuch I, Kvåle G. Menstrual and reproductive factors and risk of gastric cancer: a Norwegian cohort study. Cancer Causes Control. 2000;11(9): Heuch I, Kvåle G. Does breastfeeding affect the risk of gastric cancer? Int J Cancer. 2003;106(6): Persson C, Inoue M, Sasazuki S, et al. Female reproductive factors and the risk of gastric cancer in a large-scale population-based cohort study in Japan (JPHC study). Eur J Cancer Prev. 2008;17(4): Freedman ND, Lacey JV Jr, Hollenbeck AR, et al. The association of menstrual and reproductive factors with upper gastrointestinal tract cancers in the NIH-AARP cohort. Cancer. 2010;116(6): Riboli E, Hunt KJ, Slimani N, et al. European Prospective Investigation into Cancer and Nutrition (EPIC): study populations and data collection. Public Health Nutr. 2002;5(6B): Lauren P. The two histological main types of gastric carcinoma: diffuse and so-called intestinal-type carcinoma. An attempt at a histo-clinical classification. Acta Pathol Microbiol Scand. 1965;64: Carneiro F, Moutinho C, Pera G, et al. Pathology findings and validation of gastric and esophageal cancer cases in a European cohort (EPIC/EUR-GAST). Scand J Gastroenterol. 2007;42(5): Palli D, Masala G, Del Giudice G, et al. CagAþ Helicobacter pylori infection and gastric cancer risk in the EPIC- EURGAST study. Int J Cancer. 2007;120(4): Willett WC. Nutritional Epidemiology. 2nd ed. New York, NY: Oxford University Press; Wareham NJ, Jakes RW, Rennie KL, et al. Validity and repeatability of a simple index derived from the short physical activity questionnaire used in the European Prospective Investigation into Cancer and Nutrition (EPIC) study. Public Health Nutr. 2003;6(4): Cappola AR, Ratcliffe SJ, Bhasin S, et al. Determinants of serum total and free testosterone levels in women over the age of 65 years. J Clin Endocrinol Metab. 2007;92(2): Davison SL, Bell R, Donath S, et al. Androgen levels in adult females: changes with age, menopause, and oophorectomy. J Clin Endocrinol Metab. 2005;90(7): Fogle RH, Stanczyk FZ, Zhang X, et al. Ovarian androgen production in postmenopausal women. J Clin Endocrinol Metab. 2007;92(8): Judd HL, Lucas WE, Yen SS. Effect of oophorectomy on circulating testosterone and androstenedione levels in patients with endometrial cancer. Am J Obstet Gynecol. 1974;118(6): McTiernan A, Wu L, Barnabei VM, et al. Relation of demographic factors, menstrual history, reproduction and medication use to sex hormone levels in postmenopausal women. Breast Cancer Res Treat. 2008;108(2): Vermeulen A. The hormonal activity of the postmenopausal ovary. J Clin Endocrinol Metab. 1976;42(2): Taupin D, Podolsky DK. Trefoil factors: initiators of mucosal healing. Nat Rev Mol Cell Biol. 2003;4(9): Shi SQ, Cai JT, Yang JM. Expression of trefoil factors 1 and 2 in precancerous condition and gastric cancer. World J Gastroenterol. 2006;12(19): Campbell-Thompson ML. Estrogen receptor alpha and beta expression in upper gastrointestinal tract with regulation of trefoil factor family 2 mrna levels in ovariectomized rats. Biochem Biophys Res Commun. 1997;240(2): Katoh M. Trefoil factors and human gastric cancer (review). Int J Mol Med. 2003;12(1): Oshima CT, Wonraht DR, Catarino RM, et al. Estrogen and progesterone receptors in gastric and colorectal cancer. Hepatogastroenterology. 1999;46(30): Rothman KJ, Greenland S. Modern Epidemiology. 2nd ed. Philadelphia, PA: Lippincott-Raven; 1998.