23 February, 2015 

Many of us were exposed to the news in mid-February that a meta-analysis by Beral and colleagues [1] published inThe Lancet concluded that even a short-term use (< 5 years) of postmenopausal hormone therapy (HT) is associated with increased risk for ovarian cancer. Analyses used individual participant datasets from 52 epidemiological studies. The principal analyses involved the prospective studies (with last HT use extrapolated forwards for up to 4 years). Sensitivity analyses included the retrospective studies. Adjusted Poisson regressions yielded relative risks (RRs) versus never-use.

During prospective follow-up, 12,110 postmenopausal women, 55% (6601) of whom had used HT, developed ovarian cancer. Among women last recorded as current users, risk was increased even with < 5 years of use (RR 1.43, 95% CI 1.31–1.56;p< 0.0001). Risk was significantly increased in women who had been recent HT users (any duration, but stopped < 5 years before diagnosis) (RR 1.23, 95% CI 1.09–1.37;p= 0.0006). Risk decreased the longer ago HT had last been used, although women who had used HT for at least 5 years (median duration 9 years) and then stopped were still at significantly increased risk more than 5 years (median time since last use 10 years) later (RR 1.10, 95% CI 1.01–1.20;p= 0.02). Combining current or recent use 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 estrogen-only and estrogen–progestin preparations, but differed across the four main tumor types (heterogeneity,p< 0•0001), being definitely increased only for the two most common types, serous (RR 1.53, 95% CI 1.40–1.66;p< 0.0001) and endometrioid (RR 1.42, 95% CI 1.20–1.67;p< 0.0001). Age at initiation of HT had little effect. The increased risk may well be largely or wholly causal; if it is, women who use HT 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.

Comment

Comment by Anne Gompel
A full-size commentary is being prepared for submission to Climacteric

It appears that a large proportion of the database was not included in the main article by Beral and colleagues, but is to be found only in the Appendix [1]. The investigators explained that the retrospective studies were conducted many years ago, when HT use was less prevalent than later. Here are few examples for the important data that do not appear in the body of the article but rather in the Appendix: in the retrospective studies (n= 32), only 29% (2702) of the women had used HT, with median duration 4 years, and the mean diagnosis year was 1992 (SD 8). The retrospective studies, either from Europe, USA or pooled, did not show any increase in the ovarian cancer risk with a RR of 1.04 (95% CI 0.93–1.16). Furthermore, analysis of the global risk for ever-users compared to non-users yielded a RR of 1.20 (95% CI 1.13–1.28) in the prospectives tudies, 1.02 (95% CI 0.93–1.11) in the retrospective studies and 1.14 (95% CI 1.09–1.20) in all studies.

Relative risks were claimed to be adjusted for body mass index, age at menopause, oral contraceptive use and parity. This is not true for some of the studies which presented incomplete data. Another major point to be considered is that, in meta-analyses, the results reflect mostly the more powerful studies included. In the current case, two prospective studies weighted heavily: the Million Women Study (MWS) [2] and the Danish Sex Hormones Register Study (DaHoRS) [3], respectively accounting for 1500 cases and 2500 controls and 600 cases and 1384 controls, i.e. 2100 of the total of 2751 (76%) cases and 3884 of the total number of 5429 (71.5%) controls for the prospective studies. However, in the DaHoRS, the information on oral contraceptive use, body mass index and age at menopause was lacking as this was a study based on data from registers. It should be emphasized that the results published in the Lancet paper actually represent, to a large extent, the MWS data. Women who have no pelvic examinations, especially postmenopausal women, have been shown to be at increased risk of ovarian cancer. Recruitment in the MWS was through breast cancer screening; no data were provided for pelvic examinations. The association between use of HT and risk of ovarian cancer seems higher in women who have had hysterectomies than in those who have not. The hypothesis that such women have fewer pelvic examinations might explain the higher risk. Another puzzling finding is that estrogen-only therapy (ET) and combined estrogen–progestin therapy (EPT) were associated with similar risks. This is surprising, since the MWS showed more risk with ET. Similarly, a previous meta-analysis suggested that ET was associated with an increased risk partially reversed by progestins [4].

Comment by Samuel Shapiro

What is most astonishing is that there is not a single word about the only plausible explanation of the findings, namely, that early symptoms (dyspareunia, urinary symptoms) of as yet undiagnosed ovarian cancer ‘caused’ short-duration HT use, not the reverse, and that this would have occurred across multiple studies.

The authors argue that the higher risk of endometrioid and serous cancers strongly supports causality. That argument ignores the fact that among pathologists there is major discordance in making histological diagnoses, that different diagnoses commonly exist in different sections of ovarian tumors, and that there is a strong likelihood that pathologists selectively labelled the tumors as endometrioid if they were aware that the women were using HT, or failing that, as the most common histological type, i.e. serous. I have previously made the point that, without blinded central audit, the histological classification cannot be accepted [5]. This applies with added force to a meta-analysis in which the quality of the pathological evidence, and even the classification, would have varied considerably among the studies. Borderline tumors could selectively have been diagnosed in HT users, because they would have undergone vaginal examinations when they renewed their prescriptions.

There is not one word about completeness of follow-up (although this may be mentioned in the Appendix). However, in the MWS, it was only 64%, and cases may selectively have been identified on follow-up if they were using HT. Moreover, the evidence was mainly driven by the MWS: in the prospective studies almost half of the cases (4000/8180) were identified in that single study.

Nor is there any acknowledgement that a large and unknown proportion of hysterectomized women may have lost their ovaries and have been unaware of this. In the MWS data the association was almost entirely confined to hysterectomized women. The authors have previously argued that unidentified oophorectomy would have resulted in underestimation of the relative risk, which is nonsense. No mention is made of the risks among women who were not hysterectomized. Possibly this information is provided in the Appendix, but all they say in the paper is that adjustment for hysterectomy made no difference.

Assuming that the estimated RRs were correct, the representation that, for 5 years of use, there would be one additional case per 1000 users and one additional death per 1700 users is deliberately, provocatively, and needlessly alarmist. For 5 years of use at age 50, the respective absolute risks per 1000 women per year would be 0.2 (1/5), and 0.1 (1000/1700/5), i.e. of no material significance in public health terms. Also, as a general rule, for a relative risk of 1.37, i.e. a small RR, it is impossible to discriminate among bias, confounding, and causation as alternative explanations.

Anne Gompel
Unité de Gynécologie Endocrinienne, Université Paris Descartes, APHP, Cochin Port Royal, Paris, France

and

Samuel Shapiro
Department of Public Health and Family Medicine, University of Cape Town, South Africa

References

1. Collaborative Group on Epidemiological Studies of Ovarian Cancer. Menopausal hormone use and ovarian cancer risk: individual participant meta-analysis of 52 epidemiological studies. Lancet 2015; Feb 12. Epub ahead of print
http://www.ncbi.nlm.nih.gov/pubmed/25684585
2. Beral V, Reeves G, Green J, Bull D, for the Million Women Study Collaborators. Ovarian cancer and hormone replacement therapy in the Million Women Study. Lancet 2007;369:1703-10
http://www.ncbi.nlm.nih.gov/pubmed/17512855
3. Mørch LS, Løkkegaard E, Andreasen AH, Krüger-Kjaer S, Lidegaard O. Hormone therapy and ovarian cancer. JAMA 2009;302:298-305
http://www.ncbi.nlm.nih.gov/pubmed/19602689
4. Pearce CL, Chung K, Pike MC, Wu AH. Increased ovarian cancer risk associated with menopausal estrogen therapy is reduced by adding a progestin. Cancer 2009;115:531–9
http://www.ncbi.nlm.nih.gov/pubmed/19127543
5. Shapiro S. False alarm: postmenopausal hormone therapy and ovarian cancer. Climacteric 2007;10:466-9
http://www.ncbi.nlm.nih.gov/pubmed/18049939