Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Menopausal hormone use and ovarian cancer risk: individual participant meta-analysis of 52 epidemiological studies

Menopausal hormone use and ovarian cancer risk: individual participant meta-analysis of 52... Articles Menopausal hormone use and ovarian cancer risk: individual participant meta-analysis of 52 epidemiological studies Collaborative Group on Epidemiological Studies of Ovarian Cancer* Summary Background Half the epidemiological studies with information about menopausal hormone therapy and ovarian Lancet 2015; 385: 1835–42 cancer risk remain unpublished, and some retrospective studies could have been biased by selective participation or Published Online February 13, 2015 recall. We aimed to assess with minimal bias the eff ects of hormone therapy on ovarian cancer risk. http://dx.doi.org/10.1016/ S0140-6736(14)61687-1 Methods Individual participant datasets from 52 epidemiological studies were analysed centrally. The principal See Editorial page 1804 analyses involved the prospective studies (with last hormone therapy use extrapolated forwards for up to 4 years). See Comment page 1806 Sensitivity analyses included the retrospective studies. Adjusted Poisson regressions yielded relative risks (RRs) *Analysis and writing committee versus never-use. listed at end of paper Correspondence to: Findings During prospective follow-up, 12 110 postmenopausal women, 55% (6601) of whom had used hormone Secretariat, Cancer Epidemiology therapy, developed ovarian cancer. Among women last recorded as current users, risk was increased even with <5 years Unit, NDPH, Richard Doll Building, Oxford OX3 7LF, UK of use (RR 1·43, 95% CI 1·31–1·56; p<0·0001). Combining current-or-recent use (any duration, but stopped [email protected] <5 years before diagnosis) resulted in an RR of 1·37 (95% CI 1·29–1·46; p<0·0001); this risk was similar in European and American prospective studies and for oestrogen-only and oestrogen-progestagen preparations, but diff ered across the four main tumour types (heterogeneity p<0·0001), being defi nitely increased only for the two most common types, serous (RR 1·53, 95% CI 1·40–1·66; p<0·0001) and endometrioid (1·42, 1·20–1·67; p<0·0001). Risk declined the longer ago use had ceased, although about 10 years after stopping long-duration hormone therapy use there was still an excess of serous or endometrioid tumours (RR 1·25, 95% CI 1·07–1·46, p=0·005). Interpretation The increased risk may well be largely or wholly causal; if it is, women who use hormone therapy for 5 years from around age 50 years have about one extra ovarian cancer per 1000 users and, if its prognosis is typical, about one extra ovarian cancer death per 1700 users. Funding Medical Research Council, Cancer Research UK. Copyright © Collaborative Group on Epidemiological Studies of Ovarian Cancer. Open Access article distributed under the terms of CC BY. Introduction should cause no material harm. Most individual studies Use of hormone therapy for the menopause increased have, however, been too small to assess reliably any risks rapidly during the 1990s, then halved abruptly in the associated with use for only a few years (which is early 2000s after publication of the Women’s Health nowadays the usual pattern), so a systematic review of Initiative, a large randomised trial, but has stabilised the worldwide epidemiological evidence is needed. during the 2010s with about 6 million users in the USA Reliable epidemiological assessment of any and UK alone (fi gure 1, appendix p 4). association of hormone therapy use with ovarian cancer See Online for appendix Current hormone therapy guidelines vary in what is requires large numbers and careful control of all said about ovarian cancer. The European drug regulatory potential sources of appreciable bias, and reviews just of guidelines do not mention the disease, nor does the the published evidence cannot provide this. For, US Food and Drug Administration statement (based although many studies of ovarian cancer collected some just on the Women’s Health Initiative, which recorded information about hormone therapy use, some were few ovarian cancers). UK drug regulatory guidelines focused chiefl y on other issues. Hence, published data state that ovarian cancer might be increased by long- about hormone therapy use are available for only about term use, but were dominated by fi ndings from half the studies of ovarian cancer that have relevant data one large study; new UK guidelines are being developed. (appendix pp 5–9). Moreover, in some of the studies The most recent WHO review was completed before with retrospective designs, hormone therapy users results from most large studies were published, so might have been more willing than non-users to merely concluded that there was insuffi cient evidence participate as controls, or there might have been about any ovarian cancer risk. Recently, some non- diff erential recall of hormone therapy use between governmental reviews have argued that a few years of women already diagnosed with ovarian cancer and hormone therapy use starting before the age of 60 years unaff ected women. The Collaborative Group on www.thelancet.com Vol 385 May 9, 2015 1835 Articles and principal investigators from each eligible study had A USA been invited to collaborate. Datasets from 52 studies are included in these analyses and publications from three others are reviewed (appendix pp 5–10). Cases are postmenopausal women with malignant, or borderline-malignant, epithelial or non-epithelial ovarian cancer; controls are postmenopausal women without ovarian cancer or previous oophorectomy. In prospective studies, up to four randomly selected matched controls per case were selected. Individual data from 51 of the 52 studies were analysed centrally as 2 case-control comparisons; because of data protection laws, individual data from the Danish Sex Hormone Register Study could not be exported, so its investigators 1970 1975 1980 1985 1990 1995 2000 2005 2010 provided detailed tabular results to combine with those B UK of other studies (appendix p 6). 2·5 Information was sought for each woman on socio- demographic, reproductive, and other factors, including 2·0 hormone therapy use before cancer diagnosis for cases and to an equivalent time for controls. Postmenopausal was defi ned as having reached natural menopause or 1·5 age 55 years (because >90% of women have a natural menopause before that age ). Hysterectomy can mask 1·0 natural menopause, so women younger than 55 years with a hysterectomy were excluded. Information sought 0·5 about hormone therapy included ever-use, current use, age at fi rst and last use, total duration of use, and constituents of each preparation. The hormone therapy 1970 1975 1980 1985 1990 1995 2000 2005 2010 preparation last used was classifi ed as oestrogen-only or Year oestrogen-progestagen (or other/unknown formulation; appendix p 6). Figure 1: Trends in hormone therapy use in the USA and the UK since 1970 Tumour histology was classifi ed as fully malignant or For source of data, see appendix p 4. borderline-malignant, and as epithelial or not. Epithelial Epidemiological Studies of Ovarian Cancer was tumours were further subdivided into the four most established in 1998 to bring together and analyse common types: serous, endometrioid, mucinous, or centrally individual participant data from all epidemio- clear-cell (or mixed/other; appendix p 6). When logical studies of ovarian cancer, assessing the risks appropriate, the International Classifi cation of Diseases associated with hormonal and other factors. To evaluate for Oncology was used. with minimal bias the association of ovarian cancer Full details about information sources, search with just a few years of hormone therapy use, or with strategy, data collection, and defi nitions are provided in past use, the principal analyses review detailed data the appendix. from those prospective studies with information about both duration and recency of hormone therapy use. Statistical analyses and presentation of results Sensitivity analyses review the evidence from all studies, A protocol was circulated to collaborators and preliminary prospective or retrospective. results were discussed at a meeting of investigators in July, 2011. Poisson logistic regression comparing particular Methods groups of hormone therapy users with never-users yielded Identifi cation of studies and collection of data odds ratios, described here as relative risks (RRs). Since 1998, epidemiological studies, published and When more than two groups were compared (eg, current- unpublished, have been sought regularly by computer- or-recent users, long-term ex-users, and never-users), the aided literature searches, manual searches of review variance of the log risk was estimated for each group articles, written communications, and discussions at (appendix p 6). These group-specifi c variances were scientifi c meetings (see appendix p 5 for search strategy). used to calculate group-specifi c CIs, facilitating valid Eligible studies are those with information on hormone comparisons between any two or more groups, whether or therapy use, parity, oophorectomy and hysterectomy, not one of them was designated as the baseline group. and, if completed after 2006, at least 200 cases of ovarian The principal analyses include the prospective studies cancer. By January, 2013, 58 such studies were identifi ed only, to avoid any possible biases associated with 1836 www.thelancet.com Vol 385 May 9, 2015 Number of hormone therapy users (millions) Number of hormone therapy users (millions) Articles Hormone Median years Cases Prospective studies All studies therapy use of hormone (prospective/ therapy use all studies) (prospective/ all studies) Relative risk (95% group-specific CI)* Relative risk (95% group-specific CI)* Never-users of hormone therapy 5429/10 571 1·00 (0·96–1·04) 1·00 (0·97–1·03) Current users of hormone therapy Duration <5 years 3/2 571/897 1·43 (1·31–1·56) p<0·00001 1·27 (1·18–1·37) p<0·00001 Duration ≥5 years 10/10 1798/2311 1·41 (1·34–1·49) p<0·00001 1·34 (1·28–1·41) p<0·00001 Past users, <5 years since last use Duration <5 years 1/1 158/458 1·17 (0·98–1·38) p=0·08 1·11 (0·99–1·24) p=0·07 Duration ≥5 years 10/10 224/623 1·29 (1·11–1·49) p=0·0008 1·25 (1·13–1·39) p=0·00002 Past users, ≥5 years since last use Duration <5 years 1/1 940/1453 0·94 (0·88–1·02) p=0·1 0·97 (0·91–1·03) p=0·3 Duration ≥5 years 9/9 728/889 1·10 (1·01–1·20) p=0·02 1·11 (1·03–1·20) p=0·008 0·75 1·0 1·25 1·5 0·75 1·0 1·25 1·5 Figure 2: Relative risk of ovarian cancer by duration of use in current and past users of hormone therapy *Risk relative to never-users of hormone therapy, stratifi ed by age at diagnosis, study, and body-mass index, and adjusted for age at menopause, hysterectomy, oral contraceptive use, and parity. p values are two-sided and include the eff ects of the group-specifi c variance in never-users. diff erential participation or recall in retrospective studies, (natural menopause before age 50 years, natural but throughout the main report sensitivity analyses are menopause after age 50 years, or previous hysterectomy). given that include both the prospective and the Unknowns for each variable were assigned to separate retrospective studies. Results for the retrospective studies strata. Sensitivity analyses investigated additional only are given in the appendix (pp 17, 19, 20), and adjustment for eight other potential confounding factors. heterogeneity tests were done to compare results from Standard χ² tests for heterogeneity were used. prospective and retrospective studies. Results were weighted by the amount of statistical Because women can change their use of hormone information in each stratum (inverse of the variance of therapy over time, follow-up in prospective studies was log RR) and are presented as squares and lines, censored 4 years after hormone therapy use was last representing RRs and CIs (or, where appropriate, recorded (sensitivity analyses explored other cutoff s); group-specifi c CIs). Study-specifi c results give 99% CIs duration and recency of use were estimated as if the (to allow for multiple testing), but most other results in last recorded use had continued (ie, duration of use in the fi gures and all results in the text have 95% CIs. To those who were current users when last asked increased illustrate the correspondence between relative and by 1 year for each year of follow-up, as did time since absolute risks in hormone therapy users, absolute risks last use of hormone therapy in ex-users). Hence, if in were estimated from ovarian cancer incidence rates in a prospective study of hormone therapy use the England (appendix p 11). Analyses were done with information last recorded before diagnosis is correct, STATA 13. then in analyses that combine current users with recent ex-users (ie, women who stopped <5 years before Role of the funding sources diagnosis), the current-or-recent users would include The study funders had no role in study design, data no misclassifi ed women. Likewise, the never-users are collection, analysis or interpretation, report preparation, contaminated with few hormone therapy users: only or the decision to publish. The analysis and writing those who started in the interval of less than 4 years committee had full access to all the data and had fi nal before diagnosis, which would be so uncommon responsibility for the decision to submit for publication. (appendix p 6) as to dilute only slightly any real eff ects of hormone therapy use on risk. Results To ensure that women in one study were compared Overall information was provided for 21 488 post- directly only with similar women in that same study, all menopausal women with ovarian cancer (cases) from analyses were stratifi ed by study, centre within study, age 52 studies (17 prospective and 35 retrospective; (5-year age groups up to 85–89 years), and body-mass appendix pp 7–10). The prospective studies contributed index (<25, 25–29, or ≥30 kg/m²), and were adjusted for more than half of the cases (12 110), with mean parity (0, 1–2, or ≥3), past use of oral contraceptives (never, diagnosis year 2001 (SD 6), 55% (6601) of whom had <5 years of use, or ≥5 years of use), and age at menopause used hormone therapy, with median duration 6 years www.thelancet.com Vol 385 May 9, 2015 1837 Articles Study (country) Number of cases in current- Relative risk for current- or-recent users/never-users or-recent users vs never-users of hormone therapy of hormone therapy (99% CI)* Prospective studies Europe EPIC (eight European countries) 57/140 1·71 (0·96–3·06) DaHoRS (Denmark) 600/1384 1·49 (1·31–1·69) Million Women Study (UK) 1500/2500 1·28 (1·14–1·44) Other (Netherlands, Norway, Sweden) 39/243 1·08 (0·54–2·18) All Europe 2196/4267 1·37 (1·26–1·50) North America BCDDP (USA) 85/120 1·40 (0·87–2·24) CPS–II Mortality (USA) 23/485 1·21 (0·61–2·40) CPS–II Nutrition (USA) 91/116 1·57 (0·99–2·51) Nurses' Health Study (USA) 173/177 1·17 (0·84–1·63) NIH—AARP (USA) 135/157 1·69 (1·04–2·74) Other (Canada, USA) 48/107 1·14 (0·47–2·79) All North America 555/1162 1·35 (1·11–1·65) All prospective studies 2751/5429 1·37 (1·27–1·48) Retrospective studies Europe Negri/Franceschi/La Vecchia (Italy) 32/1107 1·97 (1·04–3·75) Riman (Sweden) 145/468 1·29 (0·96–1·74) Other (Denmark, Germany, Greece, Poland) 57/272 0·86 (0·51–1·44) All Europe 234/1847 1·25 (0·99–1·59) North America Rosenberg (USA) 51/377 1·21 (0·75–1·96) Pike/Wu (USA) 149/131 1·02 (0·64–1·63) Goodman/Wu (USA) 173/185 0·60 (0·38–0·94) SHARE (USA) 97/270 0·98 (0·63–1·52) Newcomb (USA) 113/174 1·18 (0·81–1·72) HOPE (USA) 121/207 0·91 (0·60–1·40) Other (Canada, USA) 260/1050 1·11 (0·84–1·48) All North America 964/2394 1·01 (0·87–1·17) Elsewhere Green/Purdie (Australia) 49/385 0·84 (0·46–1·52) AOCS (Australia) 289/487 0·93 (0·69–1·25) Other (China, Israel) 2/29 Insufficient data All elsewhere 340/901 0·90 (0·69–1·18) All retrospective studies 1538/5142 1·04 (0·93–1·16) 0 0·5 1·0 1·5 2·0 2·5 Figure 3: Study-specifi c results for the relative risk of ovarian cancer for current-or-recent users versus never-users of hormone therapy For study-specifi c details and references, see appendix pp 7–10. Dotted lines represent totals for all prospective studies and, separately, for all retrospective studies. Study-specifi c results are arranged by study design and region; results are given for individual studies with the most statistical information (ie, with variance of log relative risk <0·03). Results for the remaining studies are grouped together here (and given separately for every study in appendix p 14). In comparisons of relative risks in prospective versus retrospective studies, overall heterogeneity p<0·0001; for European studies, heterogeneity p=0·4; and for North American studies, heterogeneity p=0·002. In a comparison of relative risks in prospective studies, Europe versus North American heterogeneity p=0·9; for retrospective studies, Europe versus North American heterogeneity p=0·04. References provided in the appendix. *Risk relative to never-users of hormone therapy, stratifi ed by age at diagnosis, study, and body-mass index, and adjusted for age at menopause, hysterectomy, oral contraceptive use, and parity. (IQR 2–10). By contrast, in the retrospective studies Ovarian cancer risk was signifi cantly greater in only 29% (2702) of the women had used hormone ever-users than in never-users of hormone therapy therapy, with median duration 4 years (IQR 1–10), and (RR 1·20, 95% CI 1·15–1·26, p<0·0001 for prospective the mean diagnosis year was 1992 (SD 8), well before studies; 1·14, 1·10–1·19, p<0·0001 for all studies peak hormone therapy use (fi gure 1). combined; every study-specifi c result is provided in 1838 www.thelancet.com Vol 385 May 9, 2015 Articles Cases Prospective studies All studies (prospective/ all studies) Relative risk for current-or- Relative risk for current-or- recent users vs never-users recent users vs never-users of hormone therapy (95% CI)* of hormone therapy (95% CI)* Serous tumours All types of hormone therapy 1286/2208 1·53 (1·40–1·66) 1·40 (1·31–1·49) Oestrogen-only hormone therapy 487/914 1·58 (1·39–1·80) 1·59 (1·45–1·75) Oestrogen-progestagen hormone therapy 543/838 1·55 (1·38–1·74) 1·39 (1·26–1·53) Endometrioid tumours All types of hormone therapy 298/508 1·42 (1·20–1·67) 1·28 (1·13–1·45) Oestrogen-only hormone therapy 114/222 1·34 (1·05–1·72) 1·42 (1·19–1·69) Oestrogen-progestagen hormone therapy 127/192 1·58 (1·26–1·98) 1·29 (1·08–1·55) Mucinous tumours All types of hormone therapy 203/303 0·93 (0·77–1·12) 0·80 (0·69–0·93) Oestrogen-only hormone therapy 79/122 1·00 (0·75–1·33) 0·88 (0·71–1·10) Oestrogen-progestagen hormone therapy 88/122 0·95 (0·73–1·24) 0·83 (0·67–1·03) Clear-cell tumours All types of hormone therapy 92/172 0·75 (0·57–0·98) 0·80 (0·65–0·98) Oestrogen-only hormone therapy 28/57 0·81 (0·53–1·25) 0·91 (0·66–1·24) Oestrogen-progestagen hormone therapy 43/75 0·70 (0·47–1·04) 0·76 (0·57–1·02) 0·5 1·0 1·5 2·0 0·5 1·0 1·5 2·0 Figure 4: Relative risk of the four most common subtypes of ovarian cancer in current-or-recent users versus never-users of hormone therapy Numbers do not add to totals, because some hormone therapy users were classifi ed as using other or unknown types of hormone therapy and some epithelial tumours are classifi ed as mixed types, other type, or type not specifi ed. *Risks relative to never-users of hormone therapy, stratifi ed by age at diagnosis, study, and body-mass index, and adjusted for age at menopause, hysterectomy, oral contraceptive use, and parity. appendix p 12). Subsequent analyses were restricted to current users seemed to be somewhat smaller (fi gure 2, women with information both on duration of use and appendix p 17). on time since last use of hormone therapy; this In prospective studies the risk for current-or-recent exclusion of studies without information on duration of hormone therapy use (ie, use within the past 5 years) use or time since last use slightly increased these RRs was 1·37 (95% CI 1·27–1·48; fi gure 3). This RR did (appendix p 13). not diff er signifi cantly between European and North Risk was strongly related to recency of use (fi gure 2). American studies (1·37 vs 1·35; heterogeneity p=0·9). In prospective studies, risk was greatest in women who However, the RR for current-or-recent hormone when last asked had been current users (RR 1·41, therapy use did diff er signifi cantly between prospective 95% CI 1·32–1·50; p<0·0001). Among them, risk was and retrospective studies (1·37 vs 1·04; heterogeneity substantial even in those who, at diagnosis, had less p<0·0001; fi gure 3, appendix p 14). This diff erence was than 5 years (median duration 3 years) of hormone due to the lack of apparent eff ect in the aggregated therapy use (RR 1·43, 95% CI 1·31–1·56; p<0·0001). North American retrospective studies (fi gure 1, appendix Risk was, however, also signifi cantly increased in pp 12–14). In these retrospective studies, however, the women who had been recent ex-users and would at the design might have made unbiased recruitment of time of diagnosis have still have been within 5 years of controls diffi cult, and the average year of diagnosis for last use (RR 1·23, 95% CI 1·09–1·37; p=0·0006 in the cases was 1990, well before hormone therapy use prospective studies). Risk decreased the longer ago had become common (fi gure 1). hormone therapy had last been used, although women Sensitivity analyses left the main fi ndings in prospective who had used hormone therapy for at least 5 years studies largely unchanged (appendix p 18). For example, (median duration 9 years) and then stopped were still at adjustment for eight additional factors (year of birth, signifi cantly increased risk more than 5 years ethnic origin, education, age at menarche, height, alcohol (median time since last use 10 years) later (RR 1·10, consumption, smoking, and family history of ovarian or 95% CI 1·01–1·20; p=0·02). For prospective and breast cancer) altered the RRs in current-or-recent users retrospective studies combined, the risks were similar to by 0·02 or less (the main fi ndings had already been those in prospective studies alone, except that the risks in stratifi ed by study, age, and body-mass index, and www.thelancet.com Vol 385 May 9, 2015 1839 Articles In prospective studies, risks in current-or-recent users 5 year incidence of Absolute 5 year excess Absolute 5 year excess were defi nitely increased only for the two most common ovarian cancer per incidence per 1000 users incidence per 1000 users 1000 never-users of with 5 years of hormone with 10 years of hormone tumour types, serous (RR 1·53, 95% CI 1·40–1·66, hormone therapy therapy use therapy use p<0·0001) and endometrioid (1·42, 1·20–1·67; p<0·0001). Age 50–54 years 1·2 0·52 0·52 In the aggregate of these two types, the risk more than Age 55–59 years 1·6 0·37 0·67 5 years since last use for past users who had used hormone therapy for more than 5 years (RR 1·25, Age 60–64 years 2·1 0·10 0·61 95% CI 1·07–1·46; p=0·005) was more defi nite than in Excess incidence ·· 0·99 per 1000; 1·80 per 1000; 1 in 1000 users 1 in 600 users the aggregate of all types (1·10, 1·01–1·20, p=0·02). Risk Excess deaths ·· 0·6 per 1000; 1·2 per 1000; might have been somewhat decreased for the least 1 in 1700 users 1 in 800 users common type, clear-cell tumours (RR 0·75, 95% CI 0·57–0·98; p=0·04 before any allowance for multiple Methods and sources of data are provided in appendix p 11. hypothesis testing), but this protective eff ect is not Table: Estimated excess incidence of ovarian cancer in England associated with 5 years and with 10 years statistically defi nite, and in the aggregate of both of the of hormone therapy use, starting at age 50 years less common types the risk reduction was not signifi cant (RR 0·86, 0·74–1·01; p=0·07). Within each tumour type adjusted for parity, use of hormonal contraceptives, age there was little diff erence between the RRs for oestrogen- at menopause, and hysterectomy). only and oestrogen-progestagen prepara tions (fi gure 4), Furthermore, the main fi ndings were robust against or for borderline and fully malignant tumours (appendix variation in follow-up duration. Censoring either earlier p 19). or later than year 4 (appendix p 18) made little diff erence Age at initiation of hormone therapy had little eff ect; the to the results. Two prospective studies (in the UK and RRs in current-or-recent users were similarly elevated with Denmark ) contributed the most statistical information, hormone therapy use beginning before age 50 years but the RR in all other prospective studies was much the (RR 1·35, 95% CI 1·24–1·47) and at age 50–59 years same when these two were excluded. One prospective (1·31, 1·22–1·40), with little information about older ages study used fatal ovarian cancer as the outcome, but (1·15, 0·93–1·43; appendix p 20). Likewise, the available evidence suggested no major heterogeneity across again its fi ndings were typical. subgroups defi ned by smoking, body size, parity, past use In current-or-recent users, ovarian cancer risk was of oral contraceptives, hysterectomy, or other characteristics signifi cantly increased with use of both oestrogen-only (appendix pp 15–16). and oestrogen-progestagen preparations, with little Application of the RRs in the prospective studies to age- heterogeneity between the risks: RR 1·37 (95% CI specifi c ovarian cancer incidence and death rates in 1·26–1·50) and 1·37 (1·26–1·48), respectively, in the England suggested that 5 years of hormone therapy use, prospective studies (heterogeneity p=0·9); and starting at around 50 years of age, would result in 1·32 (1·23–1·41) and 1·25 (1·16–1·34) in all studies about one additional ovarian cancer per 1000 users and (heterogeneity p=0·3; appendix p 17). Few women had one additional ovarian cancer death per 1700 users used both these classes of hormone therapy, and within (appendix p 11); 10 years of hormone therapy use from the two classes there was insuffi cient information to around 50 years of age would result in about one additional assess whether risk varied by hormone therapy ovarian cancer per 600 users and one additional ovarian formulation or mode of delivery (appendix p 6). cancer death per 800 users (table). All but three studies provided tumour histology (appendix p 6). Of tumours with known histology, 98% (14 862 of 15 090) were epithelial. The RR in Discussion current-or-recent users versus never-users did not seem This collaboration brought together and analysed to diff er between epithelial (RR 1·28, 95% CI 1·22–1·34) centrally individual data from 52 epidemiological studies, and non-epithelial (1·35, 0·90–2·02) tumours, although in which about half the postmenopausal women with the confi dence interval for non-epithelial tumours was ovarian cancer had used hormone therapy. Ovarian wide. There were only 228 non-epithelial tumours, too cancer risk was signifi cantly increased in current users, few for further analysis. even in those with less than 5 years hormone therapy Almost all epithelial tumours of known, unmixed use. In ex-users, risks decreased the longer ago hormone histology were adenocarcinomas of four main tumour therapy use had ceased, but risks during the fi rst few types which were, in decreasing order of frequency, years after stopping remained appreciable. Furthermore, serous (7406 cases), endometrioid (1749), mucinous about a decade after ceasing long-duration hormone (1434), and clear cell (766). These four epithelial tumour therapy use, there still seemed to be a small excess risk. types had qualitatively diff erent relationships with In current-or-recent users (all of whom had used hormone therapy use, both in prospective studies alone hormone therapy within the past 5 years), the RRs did and in all studies combined (fi gure 4; heterogeneity not diff er signifi cantly between users of oestrogen-only p<0·0001). and of oestrogen-progestagen preparations, or between 1840 www.thelancet.com Vol 385 May 9, 2015 Articles women who had started hormone therapy before the tumour histology could have been classifi ed in slightly age of 50 years or during their 50s. The RR did, diff erent ways in diff erent studies, any misclassifi cation however, vary substantially by tumour type, being would tend to blur diff erences by tumour type, yet we increased only for the two most common histological noted distinctly heterogeneous RRs. This heterogeneity types, serous and endometrioid tumours. In analyses argues strongly for causality, because it implies that the restricted to these two types, the excess risk about a hormone-therapy-associated risks were not due just to decade after ceasing long duration hormone therapy confounding and that diff erent ovarian cancer types use became more defi nite. have diff ering causes. The reasons for this heterogeneity An important strength of prospective studies is that are unclear, partly because the sites of origin of the recruitment takes place and information about hormone four main tumour types are uncertain. The dependence therapy use is recorded before women know whether of risk on ovarian tumour type is quite diff erent for they will develop ovarian cancer. The robustness of other exposures; oral contraceptives decrease serous, prospective data is demonstrated by the stability of the endometrioid, and clear-cell but not mucinous fi ndings in various sensitivity analyses, and by the tumours, whereas smoking decreases endometrioid similarity of the fi ndings in Europe and North America. and clear-cell but increases mucinous tumours. Prospective studies provided more than half the statistical The fi ndings that ovarian cancer risk is greatest in information, so results for all studies (prospective and current users of hormone therapy, falls after use ceases, retrospective combined) were broadly similar to those for and varies by tumour type, strongly suggest a causal the prospective studies alone. relationship—ie, that among otherwise similar women, When the retrospective studies were assessed in use of hormone therapy increases the probability of isolation, their aggregate fi ndings diff ered from those of developing the two most common types of ovarian the prospective studies, however, perhaps because of cancer, and hence ovarian cancer as a whole. There are biases in some retrospective studies. Many retrospective still some 6 million users of hormone therapy in the study results could have been somewhat biased by USA and the UK, in addition to a comparable number in selective participation of hormone therapy users, and in other high-income countries (fi gure 1, appendix p 4). At all retrospective studies information about hormone present, the WHO, European, and US guidelines about therapy use was recorded after cancer diagnosis, so there hormone therapy do not mention ovarian cancer, and the might have been diff erential recall of hormone therapy UK guidelines (which are due to be revised) state only use. Moreover, some retrospective datasets in this that risk may be increased with long-term use. The collaboration have yielded apparently discrepant defi nite risk of ovarian cancer that is observed even with fi ndings on the association of ovarian cancer risk with less than 5 years of use starting at around age 50 years is 14 15 smoking and with body-mass index. directly relevant to current patterns of hormone therapy Almost all the worldwide evidence from eligible use, and hence directly relevant to medical advice, epidemiological studies was included in this personal choices, and the current eff orts to revise UK meta-analysis. The three eligible studies that had and worldwide guidelines. published results but did not contribute data were all Contributors retrospective and North American, and had all reported Analysis and writing committee: Cancer Epidemiology Unit, Oxford, UK: V Beral, K Gaitskell, C Hermon, K Moser, G Reeves; and Clinical Trial increased risks of ovarian cancer associated with some Service Unit and Epidemiological Studies Unit (CTSU), Oxford, UK: R Peto. aspect of hormone therapy use (appendix p 5). Had Contributors: The CEU secretariat (V Beral, K Gaitskell, C Hermon, they been included, the fi ndings in North American K Moser, G Reeves) identifi ed studies, received and checked data, retrospective studies might not have been as diff erent conducted analyses, and had full access to all materials and results. V Beral, C Hermon, R Peto, and G Reeves drafted the report, and all from those in other groups of studies. writing committee members helped revise it before and after circulation As long as current users and recent ex-users are to collaborators. combined together as current-or-recent users, there are Collaborative Group on Epidemiological Studies of Ovarian Cancer fewer potential sources of bias in prospective than in American Cancer Society, Atlanta, GA, USA: S M Gapstur, A V Patel; retrospective studies, and there is now for the fi rst time Australian National University, Caberra, ACT, Australia: E Banks; suffi cient evidence from prospective studies alone for Aviano Cancer Center, Italy: L Dal Maso, R Talamini; Gertner Institute, Tel Hashomer, Israel: A Chetrit, G Hirsh-Yechezkel, F Lubin, statistically stable meta-analyses. Hence, it is now S Sadetzki; Cancer Epidemiology Unit, Oxford, UK: V Beral, D Bull, possible to base the main conclusions on prospective B Cairns, B Crossley, K Gaitskell, A Goodill, J Green, C Hermon, T Key, study results. K Moser, G Reeves; Cancer Council NSW, NSW, Australia: F Sitas; The overall relative risk for any type of ovarian cancer Clinical Trial Service Unit and Epidemiological Studies Unit (CTSU), Oxford, UK: R Collins, R Peto; Catalan Institute of Oncology, is the key public health outcome. There are, however, Barcelona, Spain: C A Gonzalez; Centers for Disease Control and four main types of ovarian cancer, and in the prospective Prevention, Atlanta, GA, USA: N Lee, P Marchbanks, H W Ory, studies risk was defi nitely increased only for the H B Peterson, P A Wingo; Chiang Mai University, Thailand: N Martin, two most common types, serous and endometrioid. S Silpisornkosol, C Theetranont; Chulalongkorn University, Bangkok, Thailand: B Boosiri, S Chutivongse, P Jimakorn, P Virutamasen, Risk was possibly, although not defi nitely, decreased for C Wongsrichanalai; Cedars-Sinai Medical Center, Los Angeles, CA, the least common type, clear-cell tumours. Although www.thelancet.com Vol 385 May 9, 2015 1841 Articles Acknowledgments USA: M T Goodman; Copenhagen University Hospital, Denmark: O Lidegaard, S K Kjaer, L S Morch; Danish Cancer Society Research The chief acknowledgment is to the participants and to the investigators in these studies. Data checking and analysis in the Oxford Cancer Center, Copenhagen, Denmark: S K Kjaer, A Tjonneland; Colorado Epidemiology Unit was supported by Cancer Research UK and the UK School of Public Health, Denver, CO, USA: T Byers; Albert Einstein Medical Research Council. Adrian Goodill drew the graphs. College of Medicine, Bronx, New York, NY, USA: T Rohan; Herlev University Hospital, Copenhagen, Denmark: B Mosgaard; Department References of Public Health, Oxford, UK: M Vessey, D Yeates; Department of 1 Writing Group for the Women’s Health Initiative Investigators. Social and Preventive Medicine, State University of New York, Buff alo, Risks and benefi ts of estrogen plus progestin in healthy NY, USA: J L Freudenheim; Geisel School of Medicine at Dartmouth, postmenopausal women: principal results from the Women’s Health Hanover, NH, USA: L J Titus; German Cancer Research Center Initiative randomized controlled trial. JAMA 2002; 288: 321–33. (DKFZ), Heidelberg, Germany: J Chang-Claude, R Kaaks; University of 2 European Medicines Agency. Guidelines on clinical investigation of Minnesota School of Public Health, Minneapolis, MN, USA: medicinal products for hormone replacement therapy of oestrogen defi ciency symptoms in postmenopausal women www.ema.europa. K E Anderson, D Lazovich, K Robien; Fred Hutchinson Cancer eu/docs/en_GB/document_library/Scientifi c_guideline/2009/09/ Research Center, University of Washington, Seattle, WA, USA: WC500003348.pdf (accessed Jan 1, 2014). J Hampton, P A Newcomb, M A Rossing, D B Thomas, N S Weiss; 3 US Food and Drug Administration. Estrogen and estrogen with Hillerod Hospital, University of Copenhagen, Denmark: E Lokkegaard; progestin therapies for postmenopausal women www.fda.gov/ Imperial College London, UK: E Riboli; INSERM U1018 and Paris Drugs/DrugSafety/InformationbyDrugClass/ucm135318.htm South University, Institut Gustave-Roussy, Villejuif, France: (accessed Jan 1, 2014). F Clavel-Chapelon; Harvard Medical School, Cambridge, MA, USA: 4 Medicines and Healthcare products Regulatory Agency (MHRA). D Cramer, S E Hankinson, R M Tamimi, S S Tworoger; International Hormone-replacement therapy: safety update. UK Public Agency for Research on Cancer, Lyon, France: S Franceschi; Istituto di Assessment Report www.mhra.gov.uk/home/groups/pl-p/ Ricerche Farmacologiche Mario Negri, University of Milan, Italy: documents/websiteresources/con2032228.pdf (accessed Jan 1, 2014). C La Vecchia, E Negri; Karolinska Institutet, Stockholm, Sweden: 5 Beral V, Reeves G, Green J, Bull D. Ovarian cancer and hormone H O Adami, C Magnusson, T Riman, E Weiderpass, A Wolk; replacement therapy in the Million Women Study. Lancet 2007; Maastricht University, Netherlands: L J Schouten, P A van den Brandt; 369: 1703–10. Mahidol University, Bangkok, Thailand: N Chantarakul, S Koetsawang, 6 International Agency for Research on Cancer. Monograph D Rachawat; Cancer Research and Prevention Institute, Florence, Italy: Volume 91: combined estrogen-progestogen contraceptives and D Palli; National Cancer Institute, Bethesda, MD, USA: A Black, combined estrogen-progestogen menopausal therapy. Lyon, France: L A Brinton, D M Freedman, P Hartge, A W Hsing, J V Lacey Jnr, International Agency for Research on Cancer, 2007. J Lissowska, R N Hoover, C Schairer; Cancer Epidemiology Research 7 de Villiers TJ, Pines A, Panay N, et al. Updated 2013 International Menopause Society recommendations on menopausal hormone Group, National Health Laboratory Services, Medical Research Council, therapy and preventive strategies for midlife health. Climacteric Johannesburg, South Africa: C Babb, M Urban; Norwegian Institute of 2013; 16: 316–37. Public Health, Oslo, Norway: S Graff -Iversen, R Selmer; QIMR 8 Mørch LS, Løkkegaard E, Andreasen AH, et al. Hormone therapy Berghofer Medical Research Institute and University of Queensland, and ovarian cancer. JAMA 2009; 302: 289–305. Brisbane, QLD, Australia: C J Bain, A C Green, D M Purdie, V Siskind, 9 Collaborative Group on Hormonal Factors in Breast Cancer. P M Webb; Roswell Park Cancer Institute, Buff alo, NY, USA: Menarche, menopause, and breast cancer risk: individual participant K Moysich, S E McCann; Royal College of General Practitioners’ Oral meta-analysis, including 118 964 women with breast cancer from Contraception Study, London, UK: P Hannaford, C Kay; School of 117 epidemiological studies. Lancet Oncology 2012; 13: 1141–51. Public Health, Curtin University, Perth, WA, Australia: C W Binns, 10 Fritz AG, Percy C, Jack A, et al, eds. International classifi cation of A H Lee, M Zhang; School of Public Health, University of Texas, diseases for oncology (ICD-O). Geneva: World Health Houston, TX, USA: R B Ness; School of Public Health and Health Organization, 2000. Sciences, University of Massachusetts, Boston, MA, USA: P Nasca; 11 Plummer M. Improved estimates of fl oating absolute risk. Stat Med Slone Epidemiology Center, Boston University, MA, USA: P F Coogan, 2004; 23: 93–104. J R Palmer, L Rosenberg; Stanford University, CA, USA: A Whittemore; 12 Offi ce for National Statistics. National Cancer Statistics for England. University of Athens Medical School, Athens, Greece: K Katsouyanni, http://www.ons.gov.uk/ons/taxonomy/index.html?nscl=Cancer A Trichopoulou, D Trichopoulos, A Tzonou; University of Chile, (accessed Jan 1, 2014). Santiago, Chile: A Dabancens, L Martinez, R Molina, O Salas; 13 Rodriguez C, Patel AV, Calle EE, Jacob EJ, Thun MJ. Estrogen University of Hawaii, Honolulu, HA, USA: G Lurie, M E Carney, replacement therapy and ovarian cancer mortality in a large L R Wilkens; University Hospital, Lund, Sweden: L Hartman, J Manjer, prospective study of US women. JAMA 2001; 285: 1460–65. H Olsson; University Hospital of North Norway, Tromsø, Norway: 14 Collaborative Group on Epidemiological Studies of Ovarian Cancer. M Kumle; University of Pennsylvania, Philadelphia, PA, USA: Ovarian cancer and smoking: individual participant meta-analysis J A Grisso, M Morgan, J E Wheeler; University of Pittsburgh, PA, USA: including 28 114 women with ovarian cancer from 51 epidemiological R P Edwards, J L Kelley, F Modugno; University Medical Center, studies. Lancet Oncol 2012; 13: 946–56. Utrecht, Netherlands: N C Onland-Moret, P H M Peeters; University of 15 Collaborative Group on Epidemiological Studies of Ovarian Cancer. Southern California, Los Angeles, CA, USA: J Casagrande, M C Pike, Ovarian cancer and body size: individual participant meta-analysis including 25 157 women with ovarian cancer from A H Wu; University of New South Wales, NSW, Australia: K Canfell; 47 epidemiological studies. PloS Med 2012; 9: e1001200. University of Toronto, ON, Canada: A B Miller; Arctic University of DOI:10.1371/journal.pmed.1001200. Tromsø, Norway: I T Gram, E Lund; George Washington University, 16 Kurman RJ, Shih IeM. Molecular pathogenesis and extra-ovarian Washington, DC, USA: L McGowan; Vanderbilt University, Nashville, origin of epithelial ovarian cancer—shifting the paradigm. TN, USA: X O Shu, W Zheng; World Health Organization, Geneva, Hum Pathol 2011; 42: 918–31. Switzerland: T M M Farley, S Holck, O Meirik; and Yale School of 17 Collaborative Group on Epidemiological Studies of Ovarian Cancer. Public Health, CT, USA: H A Risch. Ovarian cancer and oral contraceptives: collaborative reanalysis of Declaration of interests data from 45 epidemiological studies including 23 257 women with All analysis and writing committee members declare no competing ovarian cancer and 87 303 controls. Lancet 2008; 371: 303–14. interests. 1842 www.thelancet.com Vol 385 May 9, 2015 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Lancet Unpaywall

Menopausal hormone use and ovarian cancer risk: individual participant meta-analysis of 52 epidemiological studies

Download PDF
8 pages

Loading next page...
 
/lp/unpaywall/menopausal-hormone-use-and-ovarian-cancer-risk-individual-participant-CgVGUg01oy

References (82)

Publisher
Unpaywall
ISSN
0140-6736
DOI
10.1016/s0140-6736(14)61687-1
Publisher site
See Article on Publisher Site

Abstract

Articles Menopausal hormone use and ovarian cancer risk: individual participant meta-analysis of 52 epidemiological studies Collaborative Group on Epidemiological Studies of Ovarian Cancer* Summary Background Half the epidemiological studies with information about menopausal hormone therapy and ovarian Lancet 2015; 385: 1835–42 cancer risk remain unpublished, and some retrospective studies could have been biased by selective participation or Published Online February 13, 2015 recall. We aimed to assess with minimal bias the eff ects of hormone therapy on ovarian cancer risk. http://dx.doi.org/10.1016/ S0140-6736(14)61687-1 Methods Individual participant datasets from 52 epidemiological studies were analysed centrally. The principal See Editorial page 1804 analyses involved the prospective studies (with last hormone therapy use extrapolated forwards for up to 4 years). See Comment page 1806 Sensitivity analyses included the retrospective studies. Adjusted Poisson regressions yielded relative risks (RRs) *Analysis and writing committee versus never-use. listed at end of paper Correspondence to: Findings During prospective follow-up, 12 110 postmenopausal women, 55% (6601) of whom had used hormone Secretariat, Cancer Epidemiology therapy, developed ovarian cancer. Among women last recorded as current users, risk was increased even with <5 years Unit, NDPH, Richard Doll Building, Oxford OX3 7LF, UK of use (RR 1·43, 95% CI 1·31–1·56; p<0·0001). Combining current-or-recent use (any duration, but stopped [email protected] <5 years before diagnosis) resulted in an RR of 1·37 (95% CI 1·29–1·46; p<0·0001); this risk was similar in European and American prospective studies and for oestrogen-only and oestrogen-progestagen preparations, but diff ered across the four main tumour types (heterogeneity p<0·0001), being defi nitely increased only for the two most common types, serous (RR 1·53, 95% CI 1·40–1·66; p<0·0001) and endometrioid (1·42, 1·20–1·67; p<0·0001). Risk declined the longer ago use had ceased, although about 10 years after stopping long-duration hormone therapy use there was still an excess of serous or endometrioid tumours (RR 1·25, 95% CI 1·07–1·46, p=0·005). Interpretation The increased risk may well be largely or wholly causal; if it is, women who use hormone therapy for 5 years from around age 50 years have about one extra ovarian cancer per 1000 users and, if its prognosis is typical, about one extra ovarian cancer death per 1700 users. Funding Medical Research Council, Cancer Research UK. Copyright © Collaborative Group on Epidemiological Studies of Ovarian Cancer. Open Access article distributed under the terms of CC BY. Introduction should cause no material harm. Most individual studies Use of hormone therapy for the menopause increased have, however, been too small to assess reliably any risks rapidly during the 1990s, then halved abruptly in the associated with use for only a few years (which is early 2000s after publication of the Women’s Health nowadays the usual pattern), so a systematic review of Initiative, a large randomised trial, but has stabilised the worldwide epidemiological evidence is needed. during the 2010s with about 6 million users in the USA Reliable epidemiological assessment of any and UK alone (fi gure 1, appendix p 4). association of hormone therapy use with ovarian cancer See Online for appendix Current hormone therapy guidelines vary in what is requires large numbers and careful control of all said about ovarian cancer. The European drug regulatory potential sources of appreciable bias, and reviews just of guidelines do not mention the disease, nor does the the published evidence cannot provide this. For, US Food and Drug Administration statement (based although many studies of ovarian cancer collected some just on the Women’s Health Initiative, which recorded information about hormone therapy use, some were few ovarian cancers). UK drug regulatory guidelines focused chiefl y on other issues. Hence, published data state that ovarian cancer might be increased by long- about hormone therapy use are available for only about term use, but were dominated by fi ndings from half the studies of ovarian cancer that have relevant data one large study; new UK guidelines are being developed. (appendix pp 5–9). Moreover, in some of the studies The most recent WHO review was completed before with retrospective designs, hormone therapy users results from most large studies were published, so might have been more willing than non-users to merely concluded that there was insuffi cient evidence participate as controls, or there might have been about any ovarian cancer risk. Recently, some non- diff erential recall of hormone therapy use between governmental reviews have argued that a few years of women already diagnosed with ovarian cancer and hormone therapy use starting before the age of 60 years unaff ected women. The Collaborative Group on www.thelancet.com Vol 385 May 9, 2015 1835 Articles and principal investigators from each eligible study had A USA been invited to collaborate. Datasets from 52 studies are included in these analyses and publications from three others are reviewed (appendix pp 5–10). Cases are postmenopausal women with malignant, or borderline-malignant, epithelial or non-epithelial ovarian cancer; controls are postmenopausal women without ovarian cancer or previous oophorectomy. In prospective studies, up to four randomly selected matched controls per case were selected. Individual data from 51 of the 52 studies were analysed centrally as 2 case-control comparisons; because of data protection laws, individual data from the Danish Sex Hormone Register Study could not be exported, so its investigators 1970 1975 1980 1985 1990 1995 2000 2005 2010 provided detailed tabular results to combine with those B UK of other studies (appendix p 6). 2·5 Information was sought for each woman on socio- demographic, reproductive, and other factors, including 2·0 hormone therapy use before cancer diagnosis for cases and to an equivalent time for controls. Postmenopausal was defi ned as having reached natural menopause or 1·5 age 55 years (because >90% of women have a natural menopause before that age ). Hysterectomy can mask 1·0 natural menopause, so women younger than 55 years with a hysterectomy were excluded. Information sought 0·5 about hormone therapy included ever-use, current use, age at fi rst and last use, total duration of use, and constituents of each preparation. The hormone therapy 1970 1975 1980 1985 1990 1995 2000 2005 2010 preparation last used was classifi ed as oestrogen-only or Year oestrogen-progestagen (or other/unknown formulation; appendix p 6). Figure 1: Trends in hormone therapy use in the USA and the UK since 1970 Tumour histology was classifi ed as fully malignant or For source of data, see appendix p 4. borderline-malignant, and as epithelial or not. Epithelial Epidemiological Studies of Ovarian Cancer was tumours were further subdivided into the four most established in 1998 to bring together and analyse common types: serous, endometrioid, mucinous, or centrally individual participant data from all epidemio- clear-cell (or mixed/other; appendix p 6). When logical studies of ovarian cancer, assessing the risks appropriate, the International Classifi cation of Diseases associated with hormonal and other factors. To evaluate for Oncology was used. with minimal bias the association of ovarian cancer Full details about information sources, search with just a few years of hormone therapy use, or with strategy, data collection, and defi nitions are provided in past use, the principal analyses review detailed data the appendix. from those prospective studies with information about both duration and recency of hormone therapy use. Statistical analyses and presentation of results Sensitivity analyses review the evidence from all studies, A protocol was circulated to collaborators and preliminary prospective or retrospective. results were discussed at a meeting of investigators in July, 2011. Poisson logistic regression comparing particular Methods groups of hormone therapy users with never-users yielded Identifi cation of studies and collection of data odds ratios, described here as relative risks (RRs). Since 1998, epidemiological studies, published and When more than two groups were compared (eg, current- unpublished, have been sought regularly by computer- or-recent users, long-term ex-users, and never-users), the aided literature searches, manual searches of review variance of the log risk was estimated for each group articles, written communications, and discussions at (appendix p 6). These group-specifi c variances were scientifi c meetings (see appendix p 5 for search strategy). used to calculate group-specifi c CIs, facilitating valid Eligible studies are those with information on hormone comparisons between any two or more groups, whether or therapy use, parity, oophorectomy and hysterectomy, not one of them was designated as the baseline group. and, if completed after 2006, at least 200 cases of ovarian The principal analyses include the prospective studies cancer. By January, 2013, 58 such studies were identifi ed only, to avoid any possible biases associated with 1836 www.thelancet.com Vol 385 May 9, 2015 Number of hormone therapy users (millions) Number of hormone therapy users (millions) Articles Hormone Median years Cases Prospective studies All studies therapy use of hormone (prospective/ therapy use all studies) (prospective/ all studies) Relative risk (95% group-specific CI)* Relative risk (95% group-specific CI)* Never-users of hormone therapy 5429/10 571 1·00 (0·96–1·04) 1·00 (0·97–1·03) Current users of hormone therapy Duration <5 years 3/2 571/897 1·43 (1·31–1·56) p<0·00001 1·27 (1·18–1·37) p<0·00001 Duration ≥5 years 10/10 1798/2311 1·41 (1·34–1·49) p<0·00001 1·34 (1·28–1·41) p<0·00001 Past users, <5 years since last use Duration <5 years 1/1 158/458 1·17 (0·98–1·38) p=0·08 1·11 (0·99–1·24) p=0·07 Duration ≥5 years 10/10 224/623 1·29 (1·11–1·49) p=0·0008 1·25 (1·13–1·39) p=0·00002 Past users, ≥5 years since last use Duration <5 years 1/1 940/1453 0·94 (0·88–1·02) p=0·1 0·97 (0·91–1·03) p=0·3 Duration ≥5 years 9/9 728/889 1·10 (1·01–1·20) p=0·02 1·11 (1·03–1·20) p=0·008 0·75 1·0 1·25 1·5 0·75 1·0 1·25 1·5 Figure 2: Relative risk of ovarian cancer by duration of use in current and past users of hormone therapy *Risk relative to never-users of hormone therapy, stratifi ed by age at diagnosis, study, and body-mass index, and adjusted for age at menopause, hysterectomy, oral contraceptive use, and parity. p values are two-sided and include the eff ects of the group-specifi c variance in never-users. diff erential participation or recall in retrospective studies, (natural menopause before age 50 years, natural but throughout the main report sensitivity analyses are menopause after age 50 years, or previous hysterectomy). given that include both the prospective and the Unknowns for each variable were assigned to separate retrospective studies. Results for the retrospective studies strata. Sensitivity analyses investigated additional only are given in the appendix (pp 17, 19, 20), and adjustment for eight other potential confounding factors. heterogeneity tests were done to compare results from Standard χ² tests for heterogeneity were used. prospective and retrospective studies. Results were weighted by the amount of statistical Because women can change their use of hormone information in each stratum (inverse of the variance of therapy over time, follow-up in prospective studies was log RR) and are presented as squares and lines, censored 4 years after hormone therapy use was last representing RRs and CIs (or, where appropriate, recorded (sensitivity analyses explored other cutoff s); group-specifi c CIs). Study-specifi c results give 99% CIs duration and recency of use were estimated as if the (to allow for multiple testing), but most other results in last recorded use had continued (ie, duration of use in the fi gures and all results in the text have 95% CIs. To those who were current users when last asked increased illustrate the correspondence between relative and by 1 year for each year of follow-up, as did time since absolute risks in hormone therapy users, absolute risks last use of hormone therapy in ex-users). Hence, if in were estimated from ovarian cancer incidence rates in a prospective study of hormone therapy use the England (appendix p 11). Analyses were done with information last recorded before diagnosis is correct, STATA 13. then in analyses that combine current users with recent ex-users (ie, women who stopped <5 years before Role of the funding sources diagnosis), the current-or-recent users would include The study funders had no role in study design, data no misclassifi ed women. Likewise, the never-users are collection, analysis or interpretation, report preparation, contaminated with few hormone therapy users: only or the decision to publish. The analysis and writing those who started in the interval of less than 4 years committee had full access to all the data and had fi nal before diagnosis, which would be so uncommon responsibility for the decision to submit for publication. (appendix p 6) as to dilute only slightly any real eff ects of hormone therapy use on risk. Results To ensure that women in one study were compared Overall information was provided for 21 488 post- directly only with similar women in that same study, all menopausal women with ovarian cancer (cases) from analyses were stratifi ed by study, centre within study, age 52 studies (17 prospective and 35 retrospective; (5-year age groups up to 85–89 years), and body-mass appendix pp 7–10). The prospective studies contributed index (<25, 25–29, or ≥30 kg/m²), and were adjusted for more than half of the cases (12 110), with mean parity (0, 1–2, or ≥3), past use of oral contraceptives (never, diagnosis year 2001 (SD 6), 55% (6601) of whom had <5 years of use, or ≥5 years of use), and age at menopause used hormone therapy, with median duration 6 years www.thelancet.com Vol 385 May 9, 2015 1837 Articles Study (country) Number of cases in current- Relative risk for current- or-recent users/never-users or-recent users vs never-users of hormone therapy of hormone therapy (99% CI)* Prospective studies Europe EPIC (eight European countries) 57/140 1·71 (0·96–3·06) DaHoRS (Denmark) 600/1384 1·49 (1·31–1·69) Million Women Study (UK) 1500/2500 1·28 (1·14–1·44) Other (Netherlands, Norway, Sweden) 39/243 1·08 (0·54–2·18) All Europe 2196/4267 1·37 (1·26–1·50) North America BCDDP (USA) 85/120 1·40 (0·87–2·24) CPS–II Mortality (USA) 23/485 1·21 (0·61–2·40) CPS–II Nutrition (USA) 91/116 1·57 (0·99–2·51) Nurses' Health Study (USA) 173/177 1·17 (0·84–1·63) NIH—AARP (USA) 135/157 1·69 (1·04–2·74) Other (Canada, USA) 48/107 1·14 (0·47–2·79) All North America 555/1162 1·35 (1·11–1·65) All prospective studies 2751/5429 1·37 (1·27–1·48) Retrospective studies Europe Negri/Franceschi/La Vecchia (Italy) 32/1107 1·97 (1·04–3·75) Riman (Sweden) 145/468 1·29 (0·96–1·74) Other (Denmark, Germany, Greece, Poland) 57/272 0·86 (0·51–1·44) All Europe 234/1847 1·25 (0·99–1·59) North America Rosenberg (USA) 51/377 1·21 (0·75–1·96) Pike/Wu (USA) 149/131 1·02 (0·64–1·63) Goodman/Wu (USA) 173/185 0·60 (0·38–0·94) SHARE (USA) 97/270 0·98 (0·63–1·52) Newcomb (USA) 113/174 1·18 (0·81–1·72) HOPE (USA) 121/207 0·91 (0·60–1·40) Other (Canada, USA) 260/1050 1·11 (0·84–1·48) All North America 964/2394 1·01 (0·87–1·17) Elsewhere Green/Purdie (Australia) 49/385 0·84 (0·46–1·52) AOCS (Australia) 289/487 0·93 (0·69–1·25) Other (China, Israel) 2/29 Insufficient data All elsewhere 340/901 0·90 (0·69–1·18) All retrospective studies 1538/5142 1·04 (0·93–1·16) 0 0·5 1·0 1·5 2·0 2·5 Figure 3: Study-specifi c results for the relative risk of ovarian cancer for current-or-recent users versus never-users of hormone therapy For study-specifi c details and references, see appendix pp 7–10. Dotted lines represent totals for all prospective studies and, separately, for all retrospective studies. Study-specifi c results are arranged by study design and region; results are given for individual studies with the most statistical information (ie, with variance of log relative risk <0·03). Results for the remaining studies are grouped together here (and given separately for every study in appendix p 14). In comparisons of relative risks in prospective versus retrospective studies, overall heterogeneity p<0·0001; for European studies, heterogeneity p=0·4; and for North American studies, heterogeneity p=0·002. In a comparison of relative risks in prospective studies, Europe versus North American heterogeneity p=0·9; for retrospective studies, Europe versus North American heterogeneity p=0·04. References provided in the appendix. *Risk relative to never-users of hormone therapy, stratifi ed by age at diagnosis, study, and body-mass index, and adjusted for age at menopause, hysterectomy, oral contraceptive use, and parity. (IQR 2–10). By contrast, in the retrospective studies Ovarian cancer risk was signifi cantly greater in only 29% (2702) of the women had used hormone ever-users than in never-users of hormone therapy therapy, with median duration 4 years (IQR 1–10), and (RR 1·20, 95% CI 1·15–1·26, p<0·0001 for prospective the mean diagnosis year was 1992 (SD 8), well before studies; 1·14, 1·10–1·19, p<0·0001 for all studies peak hormone therapy use (fi gure 1). combined; every study-specifi c result is provided in 1838 www.thelancet.com Vol 385 May 9, 2015 Articles Cases Prospective studies All studies (prospective/ all studies) Relative risk for current-or- Relative risk for current-or- recent users vs never-users recent users vs never-users of hormone therapy (95% CI)* of hormone therapy (95% CI)* Serous tumours All types of hormone therapy 1286/2208 1·53 (1·40–1·66) 1·40 (1·31–1·49) Oestrogen-only hormone therapy 487/914 1·58 (1·39–1·80) 1·59 (1·45–1·75) Oestrogen-progestagen hormone therapy 543/838 1·55 (1·38–1·74) 1·39 (1·26–1·53) Endometrioid tumours All types of hormone therapy 298/508 1·42 (1·20–1·67) 1·28 (1·13–1·45) Oestrogen-only hormone therapy 114/222 1·34 (1·05–1·72) 1·42 (1·19–1·69) Oestrogen-progestagen hormone therapy 127/192 1·58 (1·26–1·98) 1·29 (1·08–1·55) Mucinous tumours All types of hormone therapy 203/303 0·93 (0·77–1·12) 0·80 (0·69–0·93) Oestrogen-only hormone therapy 79/122 1·00 (0·75–1·33) 0·88 (0·71–1·10) Oestrogen-progestagen hormone therapy 88/122 0·95 (0·73–1·24) 0·83 (0·67–1·03) Clear-cell tumours All types of hormone therapy 92/172 0·75 (0·57–0·98) 0·80 (0·65–0·98) Oestrogen-only hormone therapy 28/57 0·81 (0·53–1·25) 0·91 (0·66–1·24) Oestrogen-progestagen hormone therapy 43/75 0·70 (0·47–1·04) 0·76 (0·57–1·02) 0·5 1·0 1·5 2·0 0·5 1·0 1·5 2·0 Figure 4: Relative risk of the four most common subtypes of ovarian cancer in current-or-recent users versus never-users of hormone therapy Numbers do not add to totals, because some hormone therapy users were classifi ed as using other or unknown types of hormone therapy and some epithelial tumours are classifi ed as mixed types, other type, or type not specifi ed. *Risks relative to never-users of hormone therapy, stratifi ed by age at diagnosis, study, and body-mass index, and adjusted for age at menopause, hysterectomy, oral contraceptive use, and parity. appendix p 12). Subsequent analyses were restricted to current users seemed to be somewhat smaller (fi gure 2, women with information both on duration of use and appendix p 17). on time since last use of hormone therapy; this In prospective studies the risk for current-or-recent exclusion of studies without information on duration of hormone therapy use (ie, use within the past 5 years) use or time since last use slightly increased these RRs was 1·37 (95% CI 1·27–1·48; fi gure 3). This RR did (appendix p 13). not diff er signifi cantly between European and North Risk was strongly related to recency of use (fi gure 2). American studies (1·37 vs 1·35; heterogeneity p=0·9). In prospective studies, risk was greatest in women who However, the RR for current-or-recent hormone when last asked had been current users (RR 1·41, therapy use did diff er signifi cantly between prospective 95% CI 1·32–1·50; p<0·0001). Among them, risk was and retrospective studies (1·37 vs 1·04; heterogeneity substantial even in those who, at diagnosis, had less p<0·0001; fi gure 3, appendix p 14). This diff erence was than 5 years (median duration 3 years) of hormone due to the lack of apparent eff ect in the aggregated therapy use (RR 1·43, 95% CI 1·31–1·56; p<0·0001). North American retrospective studies (fi gure 1, appendix Risk was, however, also signifi cantly increased in pp 12–14). In these retrospective studies, however, the women who had been recent ex-users and would at the design might have made unbiased recruitment of time of diagnosis have still have been within 5 years of controls diffi cult, and the average year of diagnosis for last use (RR 1·23, 95% CI 1·09–1·37; p=0·0006 in the cases was 1990, well before hormone therapy use prospective studies). Risk decreased the longer ago had become common (fi gure 1). hormone therapy had last been used, although women Sensitivity analyses left the main fi ndings in prospective who had used hormone therapy for at least 5 years studies largely unchanged (appendix p 18). For example, (median duration 9 years) and then stopped were still at adjustment for eight additional factors (year of birth, signifi cantly increased risk more than 5 years ethnic origin, education, age at menarche, height, alcohol (median time since last use 10 years) later (RR 1·10, consumption, smoking, and family history of ovarian or 95% CI 1·01–1·20; p=0·02). For prospective and breast cancer) altered the RRs in current-or-recent users retrospective studies combined, the risks were similar to by 0·02 or less (the main fi ndings had already been those in prospective studies alone, except that the risks in stratifi ed by study, age, and body-mass index, and www.thelancet.com Vol 385 May 9, 2015 1839 Articles In prospective studies, risks in current-or-recent users 5 year incidence of Absolute 5 year excess Absolute 5 year excess were defi nitely increased only for the two most common ovarian cancer per incidence per 1000 users incidence per 1000 users 1000 never-users of with 5 years of hormone with 10 years of hormone tumour types, serous (RR 1·53, 95% CI 1·40–1·66, hormone therapy therapy use therapy use p<0·0001) and endometrioid (1·42, 1·20–1·67; p<0·0001). Age 50–54 years 1·2 0·52 0·52 In the aggregate of these two types, the risk more than Age 55–59 years 1·6 0·37 0·67 5 years since last use for past users who had used hormone therapy for more than 5 years (RR 1·25, Age 60–64 years 2·1 0·10 0·61 95% CI 1·07–1·46; p=0·005) was more defi nite than in Excess incidence ·· 0·99 per 1000; 1·80 per 1000; 1 in 1000 users 1 in 600 users the aggregate of all types (1·10, 1·01–1·20, p=0·02). Risk Excess deaths ·· 0·6 per 1000; 1·2 per 1000; might have been somewhat decreased for the least 1 in 1700 users 1 in 800 users common type, clear-cell tumours (RR 0·75, 95% CI 0·57–0·98; p=0·04 before any allowance for multiple Methods and sources of data are provided in appendix p 11. hypothesis testing), but this protective eff ect is not Table: Estimated excess incidence of ovarian cancer in England associated with 5 years and with 10 years statistically defi nite, and in the aggregate of both of the of hormone therapy use, starting at age 50 years less common types the risk reduction was not signifi cant (RR 0·86, 0·74–1·01; p=0·07). Within each tumour type adjusted for parity, use of hormonal contraceptives, age there was little diff erence between the RRs for oestrogen- at menopause, and hysterectomy). only and oestrogen-progestagen prepara tions (fi gure 4), Furthermore, the main fi ndings were robust against or for borderline and fully malignant tumours (appendix variation in follow-up duration. Censoring either earlier p 19). or later than year 4 (appendix p 18) made little diff erence Age at initiation of hormone therapy had little eff ect; the to the results. Two prospective studies (in the UK and RRs in current-or-recent users were similarly elevated with Denmark ) contributed the most statistical information, hormone therapy use beginning before age 50 years but the RR in all other prospective studies was much the (RR 1·35, 95% CI 1·24–1·47) and at age 50–59 years same when these two were excluded. One prospective (1·31, 1·22–1·40), with little information about older ages study used fatal ovarian cancer as the outcome, but (1·15, 0·93–1·43; appendix p 20). Likewise, the available evidence suggested no major heterogeneity across again its fi ndings were typical. subgroups defi ned by smoking, body size, parity, past use In current-or-recent users, ovarian cancer risk was of oral contraceptives, hysterectomy, or other characteristics signifi cantly increased with use of both oestrogen-only (appendix pp 15–16). and oestrogen-progestagen preparations, with little Application of the RRs in the prospective studies to age- heterogeneity between the risks: RR 1·37 (95% CI specifi c ovarian cancer incidence and death rates in 1·26–1·50) and 1·37 (1·26–1·48), respectively, in the England suggested that 5 years of hormone therapy use, prospective studies (heterogeneity p=0·9); and starting at around 50 years of age, would result in 1·32 (1·23–1·41) and 1·25 (1·16–1·34) in all studies about one additional ovarian cancer per 1000 users and (heterogeneity p=0·3; appendix p 17). Few women had one additional ovarian cancer death per 1700 users used both these classes of hormone therapy, and within (appendix p 11); 10 years of hormone therapy use from the two classes there was insuffi cient information to around 50 years of age would result in about one additional assess whether risk varied by hormone therapy ovarian cancer per 600 users and one additional ovarian formulation or mode of delivery (appendix p 6). cancer death per 800 users (table). All but three studies provided tumour histology (appendix p 6). Of tumours with known histology, 98% (14 862 of 15 090) were epithelial. The RR in Discussion current-or-recent users versus never-users did not seem This collaboration brought together and analysed to diff er between epithelial (RR 1·28, 95% CI 1·22–1·34) centrally individual data from 52 epidemiological studies, and non-epithelial (1·35, 0·90–2·02) tumours, although in which about half the postmenopausal women with the confi dence interval for non-epithelial tumours was ovarian cancer had used hormone therapy. Ovarian wide. There were only 228 non-epithelial tumours, too cancer risk was signifi cantly increased in current users, few for further analysis. even in those with less than 5 years hormone therapy Almost all epithelial tumours of known, unmixed use. In ex-users, risks decreased the longer ago hormone histology were adenocarcinomas of four main tumour therapy use had ceased, but risks during the fi rst few types which were, in decreasing order of frequency, years after stopping remained appreciable. Furthermore, serous (7406 cases), endometrioid (1749), mucinous about a decade after ceasing long-duration hormone (1434), and clear cell (766). These four epithelial tumour therapy use, there still seemed to be a small excess risk. types had qualitatively diff erent relationships with In current-or-recent users (all of whom had used hormone therapy use, both in prospective studies alone hormone therapy within the past 5 years), the RRs did and in all studies combined (fi gure 4; heterogeneity not diff er signifi cantly between users of oestrogen-only p<0·0001). and of oestrogen-progestagen preparations, or between 1840 www.thelancet.com Vol 385 May 9, 2015 Articles women who had started hormone therapy before the tumour histology could have been classifi ed in slightly age of 50 years or during their 50s. The RR did, diff erent ways in diff erent studies, any misclassifi cation however, vary substantially by tumour type, being would tend to blur diff erences by tumour type, yet we increased only for the two most common histological noted distinctly heterogeneous RRs. This heterogeneity types, serous and endometrioid tumours. In analyses argues strongly for causality, because it implies that the restricted to these two types, the excess risk about a hormone-therapy-associated risks were not due just to decade after ceasing long duration hormone therapy confounding and that diff erent ovarian cancer types use became more defi nite. have diff ering causes. The reasons for this heterogeneity An important strength of prospective studies is that are unclear, partly because the sites of origin of the recruitment takes place and information about hormone four main tumour types are uncertain. The dependence therapy use is recorded before women know whether of risk on ovarian tumour type is quite diff erent for they will develop ovarian cancer. The robustness of other exposures; oral contraceptives decrease serous, prospective data is demonstrated by the stability of the endometrioid, and clear-cell but not mucinous fi ndings in various sensitivity analyses, and by the tumours, whereas smoking decreases endometrioid similarity of the fi ndings in Europe and North America. and clear-cell but increases mucinous tumours. Prospective studies provided more than half the statistical The fi ndings that ovarian cancer risk is greatest in information, so results for all studies (prospective and current users of hormone therapy, falls after use ceases, retrospective combined) were broadly similar to those for and varies by tumour type, strongly suggest a causal the prospective studies alone. relationship—ie, that among otherwise similar women, When the retrospective studies were assessed in use of hormone therapy increases the probability of isolation, their aggregate fi ndings diff ered from those of developing the two most common types of ovarian the prospective studies, however, perhaps because of cancer, and hence ovarian cancer as a whole. There are biases in some retrospective studies. Many retrospective still some 6 million users of hormone therapy in the study results could have been somewhat biased by USA and the UK, in addition to a comparable number in selective participation of hormone therapy users, and in other high-income countries (fi gure 1, appendix p 4). At all retrospective studies information about hormone present, the WHO, European, and US guidelines about therapy use was recorded after cancer diagnosis, so there hormone therapy do not mention ovarian cancer, and the might have been diff erential recall of hormone therapy UK guidelines (which are due to be revised) state only use. Moreover, some retrospective datasets in this that risk may be increased with long-term use. The collaboration have yielded apparently discrepant defi nite risk of ovarian cancer that is observed even with fi ndings on the association of ovarian cancer risk with less than 5 years of use starting at around age 50 years is 14 15 smoking and with body-mass index. directly relevant to current patterns of hormone therapy Almost all the worldwide evidence from eligible use, and hence directly relevant to medical advice, epidemiological studies was included in this personal choices, and the current eff orts to revise UK meta-analysis. The three eligible studies that had and worldwide guidelines. published results but did not contribute data were all Contributors retrospective and North American, and had all reported Analysis and writing committee: Cancer Epidemiology Unit, Oxford, UK: V Beral, K Gaitskell, C Hermon, K Moser, G Reeves; and Clinical Trial increased risks of ovarian cancer associated with some Service Unit and Epidemiological Studies Unit (CTSU), Oxford, UK: R Peto. aspect of hormone therapy use (appendix p 5). Had Contributors: The CEU secretariat (V Beral, K Gaitskell, C Hermon, they been included, the fi ndings in North American K Moser, G Reeves) identifi ed studies, received and checked data, retrospective studies might not have been as diff erent conducted analyses, and had full access to all materials and results. V Beral, C Hermon, R Peto, and G Reeves drafted the report, and all from those in other groups of studies. writing committee members helped revise it before and after circulation As long as current users and recent ex-users are to collaborators. combined together as current-or-recent users, there are Collaborative Group on Epidemiological Studies of Ovarian Cancer fewer potential sources of bias in prospective than in American Cancer Society, Atlanta, GA, USA: S M Gapstur, A V Patel; retrospective studies, and there is now for the fi rst time Australian National University, Caberra, ACT, Australia: E Banks; suffi cient evidence from prospective studies alone for Aviano Cancer Center, Italy: L Dal Maso, R Talamini; Gertner Institute, Tel Hashomer, Israel: A Chetrit, G Hirsh-Yechezkel, F Lubin, statistically stable meta-analyses. Hence, it is now S Sadetzki; Cancer Epidemiology Unit, Oxford, UK: V Beral, D Bull, possible to base the main conclusions on prospective B Cairns, B Crossley, K Gaitskell, A Goodill, J Green, C Hermon, T Key, study results. K Moser, G Reeves; Cancer Council NSW, NSW, Australia: F Sitas; The overall relative risk for any type of ovarian cancer Clinical Trial Service Unit and Epidemiological Studies Unit (CTSU), Oxford, UK: R Collins, R Peto; Catalan Institute of Oncology, is the key public health outcome. There are, however, Barcelona, Spain: C A Gonzalez; Centers for Disease Control and four main types of ovarian cancer, and in the prospective Prevention, Atlanta, GA, USA: N Lee, P Marchbanks, H W Ory, studies risk was defi nitely increased only for the H B Peterson, P A Wingo; Chiang Mai University, Thailand: N Martin, two most common types, serous and endometrioid. S Silpisornkosol, C Theetranont; Chulalongkorn University, Bangkok, Thailand: B Boosiri, S Chutivongse, P Jimakorn, P Virutamasen, Risk was possibly, although not defi nitely, decreased for C Wongsrichanalai; Cedars-Sinai Medical Center, Los Angeles, CA, the least common type, clear-cell tumours. Although www.thelancet.com Vol 385 May 9, 2015 1841 Articles Acknowledgments USA: M T Goodman; Copenhagen University Hospital, Denmark: O Lidegaard, S K Kjaer, L S Morch; Danish Cancer Society Research The chief acknowledgment is to the participants and to the investigators in these studies. Data checking and analysis in the Oxford Cancer Center, Copenhagen, Denmark: S K Kjaer, A Tjonneland; Colorado Epidemiology Unit was supported by Cancer Research UK and the UK School of Public Health, Denver, CO, USA: T Byers; Albert Einstein Medical Research Council. Adrian Goodill drew the graphs. College of Medicine, Bronx, New York, NY, USA: T Rohan; Herlev University Hospital, Copenhagen, Denmark: B Mosgaard; Department References of Public Health, Oxford, UK: M Vessey, D Yeates; Department of 1 Writing Group for the Women’s Health Initiative Investigators. Social and Preventive Medicine, State University of New York, Buff alo, Risks and benefi ts of estrogen plus progestin in healthy NY, USA: J L Freudenheim; Geisel School of Medicine at Dartmouth, postmenopausal women: principal results from the Women’s Health Hanover, NH, USA: L J Titus; German Cancer Research Center Initiative randomized controlled trial. JAMA 2002; 288: 321–33. (DKFZ), Heidelberg, Germany: J Chang-Claude, R Kaaks; University of 2 European Medicines Agency. Guidelines on clinical investigation of Minnesota School of Public Health, Minneapolis, MN, USA: medicinal products for hormone replacement therapy of oestrogen defi ciency symptoms in postmenopausal women www.ema.europa. K E Anderson, D Lazovich, K Robien; Fred Hutchinson Cancer eu/docs/en_GB/document_library/Scientifi c_guideline/2009/09/ Research Center, University of Washington, Seattle, WA, USA: WC500003348.pdf (accessed Jan 1, 2014). J Hampton, P A Newcomb, M A Rossing, D B Thomas, N S Weiss; 3 US Food and Drug Administration. Estrogen and estrogen with Hillerod Hospital, University of Copenhagen, Denmark: E Lokkegaard; progestin therapies for postmenopausal women www.fda.gov/ Imperial College London, UK: E Riboli; INSERM U1018 and Paris Drugs/DrugSafety/InformationbyDrugClass/ucm135318.htm South University, Institut Gustave-Roussy, Villejuif, France: (accessed Jan 1, 2014). F Clavel-Chapelon; Harvard Medical School, Cambridge, MA, USA: 4 Medicines and Healthcare products Regulatory Agency (MHRA). D Cramer, S E Hankinson, R M Tamimi, S S Tworoger; International Hormone-replacement therapy: safety update. UK Public Agency for Research on Cancer, Lyon, France: S Franceschi; Istituto di Assessment Report www.mhra.gov.uk/home/groups/pl-p/ Ricerche Farmacologiche Mario Negri, University of Milan, Italy: documents/websiteresources/con2032228.pdf (accessed Jan 1, 2014). C La Vecchia, E Negri; Karolinska Institutet, Stockholm, Sweden: 5 Beral V, Reeves G, Green J, Bull D. Ovarian cancer and hormone H O Adami, C Magnusson, T Riman, E Weiderpass, A Wolk; replacement therapy in the Million Women Study. Lancet 2007; Maastricht University, Netherlands: L J Schouten, P A van den Brandt; 369: 1703–10. Mahidol University, Bangkok, Thailand: N Chantarakul, S Koetsawang, 6 International Agency for Research on Cancer. Monograph D Rachawat; Cancer Research and Prevention Institute, Florence, Italy: Volume 91: combined estrogen-progestogen contraceptives and D Palli; National Cancer Institute, Bethesda, MD, USA: A Black, combined estrogen-progestogen menopausal therapy. Lyon, France: L A Brinton, D M Freedman, P Hartge, A W Hsing, J V Lacey Jnr, International Agency for Research on Cancer, 2007. J Lissowska, R N Hoover, C Schairer; Cancer Epidemiology Research 7 de Villiers TJ, Pines A, Panay N, et al. Updated 2013 International Menopause Society recommendations on menopausal hormone Group, National Health Laboratory Services, Medical Research Council, therapy and preventive strategies for midlife health. Climacteric Johannesburg, South Africa: C Babb, M Urban; Norwegian Institute of 2013; 16: 316–37. Public Health, Oslo, Norway: S Graff -Iversen, R Selmer; QIMR 8 Mørch LS, Løkkegaard E, Andreasen AH, et al. Hormone therapy Berghofer Medical Research Institute and University of Queensland, and ovarian cancer. JAMA 2009; 302: 289–305. Brisbane, QLD, Australia: C J Bain, A C Green, D M Purdie, V Siskind, 9 Collaborative Group on Hormonal Factors in Breast Cancer. P M Webb; Roswell Park Cancer Institute, Buff alo, NY, USA: Menarche, menopause, and breast cancer risk: individual participant K Moysich, S E McCann; Royal College of General Practitioners’ Oral meta-analysis, including 118 964 women with breast cancer from Contraception Study, London, UK: P Hannaford, C Kay; School of 117 epidemiological studies. Lancet Oncology 2012; 13: 1141–51. Public Health, Curtin University, Perth, WA, Australia: C W Binns, 10 Fritz AG, Percy C, Jack A, et al, eds. International classifi cation of A H Lee, M Zhang; School of Public Health, University of Texas, diseases for oncology (ICD-O). Geneva: World Health Houston, TX, USA: R B Ness; School of Public Health and Health Organization, 2000. Sciences, University of Massachusetts, Boston, MA, USA: P Nasca; 11 Plummer M. Improved estimates of fl oating absolute risk. Stat Med Slone Epidemiology Center, Boston University, MA, USA: P F Coogan, 2004; 23: 93–104. J R Palmer, L Rosenberg; Stanford University, CA, USA: A Whittemore; 12 Offi ce for National Statistics. National Cancer Statistics for England. University of Athens Medical School, Athens, Greece: K Katsouyanni, http://www.ons.gov.uk/ons/taxonomy/index.html?nscl=Cancer A Trichopoulou, D Trichopoulos, A Tzonou; University of Chile, (accessed Jan 1, 2014). Santiago, Chile: A Dabancens, L Martinez, R Molina, O Salas; 13 Rodriguez C, Patel AV, Calle EE, Jacob EJ, Thun MJ. Estrogen University of Hawaii, Honolulu, HA, USA: G Lurie, M E Carney, replacement therapy and ovarian cancer mortality in a large L R Wilkens; University Hospital, Lund, Sweden: L Hartman, J Manjer, prospective study of US women. JAMA 2001; 285: 1460–65. H Olsson; University Hospital of North Norway, Tromsø, Norway: 14 Collaborative Group on Epidemiological Studies of Ovarian Cancer. M Kumle; University of Pennsylvania, Philadelphia, PA, USA: Ovarian cancer and smoking: individual participant meta-analysis J A Grisso, M Morgan, J E Wheeler; University of Pittsburgh, PA, USA: including 28 114 women with ovarian cancer from 51 epidemiological R P Edwards, J L Kelley, F Modugno; University Medical Center, studies. Lancet Oncol 2012; 13: 946–56. Utrecht, Netherlands: N C Onland-Moret, P H M Peeters; University of 15 Collaborative Group on Epidemiological Studies of Ovarian Cancer. Southern California, Los Angeles, CA, USA: J Casagrande, M C Pike, Ovarian cancer and body size: individual participant meta-analysis including 25 157 women with ovarian cancer from A H Wu; University of New South Wales, NSW, Australia: K Canfell; 47 epidemiological studies. PloS Med 2012; 9: e1001200. University of Toronto, ON, Canada: A B Miller; Arctic University of DOI:10.1371/journal.pmed.1001200. Tromsø, Norway: I T Gram, E Lund; George Washington University, 16 Kurman RJ, Shih IeM. Molecular pathogenesis and extra-ovarian Washington, DC, USA: L McGowan; Vanderbilt University, Nashville, origin of epithelial ovarian cancer—shifting the paradigm. TN, USA: X O Shu, W Zheng; World Health Organization, Geneva, Hum Pathol 2011; 42: 918–31. Switzerland: T M M Farley, S Holck, O Meirik; and Yale School of 17 Collaborative Group on Epidemiological Studies of Ovarian Cancer. Public Health, CT, USA: H A Risch. Ovarian cancer and oral contraceptives: collaborative reanalysis of Declaration of interests data from 45 epidemiological studies including 23 257 women with All analysis and writing committee members declare no competing ovarian cancer and 87 303 controls. Lancet 2008; 371: 303–14. interests. 1842 www.thelancet.com Vol 385 May 9, 2015

Journal

The Lancet – Unpaywall

Published: Feb 13, 2015

There are no references for this article.