Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Dr S Suissa, Division of Clinical Epidemiology, Royal Victoria Hospital, 687 Pine avenue west, Ross 4.29, Montreal, Québec, Canada H3A 1A1. Email firstname.lastname@example.org
Objective To determine the effect of different types and formulations of hormone replacement therapy (HRT) on the risk of breast cancer in postmenopausal women.
Design Population-based case–control study.
Setting UK, 1988–2004.
Participants Women 50–75 years between 1998 and 2004.
Main outcome measures Breast cancer incidence to estimate the rate ratio (RR) associated with use of various HRTs over a 30-year period.
Results We identified 6347 incident cases of breast cancer that were matched with 31 516 controls. Cases were on average 61 years at diagnosis and 22% had undergone a hysterectomy. The rate of breast cancer was increased with the use of opposed estrogens in oral form (adjusted RR 1.38; 95% CI 1.27–1.49) in contrast to patch form (RR 1.08; 95% CI 0.81–1.43). This rate was similarly elevated with both continuous (RR 1.29; 95% CI 1.07–1.56) and sequential (RR 1.33; 95% CI 1.21–1.46) forms of opposed estrogen. The rate of breast cancer was not increased among exclusive users of unopposed estrogens (RR 0.97; 95% CI 0.86–1.09) or of tibolone (RR 0.86; 95% CI 0.65–1.13). Users of tibolone who had switched from opposed estrogens, however, had an elevated rate (RR 1.29; 95% CI 1.09–1.52). The rate of breast cancer increased by 25% (95% CI 20–30%) with every ten prescriptions of orally administered opposed estrogen.
Conclusions The risk of breast cancer varies with the formulation and preparation of HRT. Opposed estrogens (progesterone–estrogen) in oral form are associated with an increased risk of breast cancer, which increases with use. Transdermal opposed estrogens, unopposed estrogens and tibolone do not increase this risk. However, this study is an observational study that carries risks of various biases, and thus the findings need to be interpreted with caution.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Hormone replacement therapy (HRT) has been used extensively in postmenopausal women as a proven and effective therapy for climacteric symptoms and in osteoporosis treatment. Its widespread long-term use for indications such as primary prevention of cardiovascular disease or cognitive decline was called into question with the publication of several epidemiological studies, most notably the Million Women Study (MWS) and the Women’s Health Initiative (WHI), which reported several adverse effects among users of HRT, including breast cancer.1–5
The magnitude of this risk, however, has been found to vary across studies. In particular, the breast cancer risk appears to vary with different HRT formulations (e.g. opposed versus unopposed estrogen), different modes of delivery (transdermal versus oral), the agent and the length of use.6–8 The addition of a progestin appears to increase the risk of breast cancer. However, it is unclear whether this risk varies with different classes of progestins. With respect to the mode of delivery, the transdermal form of delivery means that the user is exposed to continuous low doses of hormones with first pass clearance by the liver. This contrasts with oral absorption with an initial peak dose and subsequent clearance from the blood long before the next dose. This difference in hormone exposure may have effects on breast tissue and consequently breast cancer risk over a long duration of time. A new agent on the market, tibolone is a synthetic compound, classified as a selective tissue estrogenic activity regulator (STEAR) for which it has been suggested that its use may be associated with an increase in breast cancer.1
Using the UKs General Practice Research Database (GPRD), we examined the risk of breast cancer in postmenopausal women associated with different formulations of HRT, different agents and different modes of delivery of HRT.
The UKs GPRD was used as the source of data for this study, and its characteristics have been published previously.9 Briefly, the GPRD is a prospectively collected and regularly updated database containing information on approximately 8% of the English population from 1994 to the present and contains several million patient-years of data. Since health care delivery in the UK is GP-based, all women encounters with the GP, diagnoses, consultations with specialists and prescribed outpatient medications are recorded in the database. The women within the database reflect the demographics of the entire general English population, and quality control of data entry is maintained and enforced on a regular basis. Information available from this database includes numerous demographic characteristics, clinical diagnoses, drugs prescribed, referrals to specialists and hospitalisations. Studies using chart review or comparison with other national registries have validated numerous data points in the GPRD.9,10
Study design: cohort definition
We conducted a case–control study within a defined cohort of the GPRD population.11,12 We identified the source population as all women 50–75 years between 1 January 1998 and 31 December 2004, who were members of an ’up to standard’ practice.
Cases and controls
Within the above-defined cohort of postmenopausal women, cases of invasive breast cancer were identified using a computerised algorithm. This algorithm used codes for breast cancer as well as specific combinations of clinical information, such as operations (mastectomies, lumpectomies and auxiliary node dissections), consultations with oncologists, chemotherapy, radiotherapy and use of postoperative antihormone therapy to identify breast cancer cases. The computerised medical record of a random stratified sample of 265 subjects, including 228 potential cases and 37 negative controls was assessed blindly by two independent physicians. As a result, the algorithm was modified to achieve a validation of more than 95% between physician and computer diagnosis of breast cancer, which occurred after two iterations.
For each case of breast cancer, we selected five population controls from the source population, after matching on year of birth (±1 year), GP practice and year of entry into the GPRD. Cases and controls with a history of breast cancer before the age of 50 years were excluded to ensure incident cases of breast cancer. For around 1095 cases, the matching criteria for year of birth and year of cohort entry had to be expanded. Finally, 171 cases ended up with less than 5 controls.
For each case and control, we obtained information on all prescriptions for any HRT given prior to the index date. This included particularly information on the type of estrogen used, as well as the modality (oral versus transdermal), the type of progestin class used in combination with estrogen; the use of more than one class of HRT and the duration of use of HRT. Thus, use of HRT was grouped into several categories. Among users with a history of only one type of HRT, the following groups were analysed: oral estrogen, transdermal estrogen, estrogen + progesterone and tibolone. Progesterones were grouped into carbon-21 (C-21) based progestins (including natural progesterone, hydroprogesterone and the synthetic hydroprogesterone and medroxyprogesterone acetate) and carbon-19 (C-19) based progestins (including norethisterone and levonorgestrel). Users with a history that involved more than one type of HRT were considered separately.
We also identified information about recognised risk factors for breast cancer. Odds ratios were adjusted for endometrial cancer, hysterectomy, oophorectomy, family history of breast cancer, documented history of oral contraceptive use, obesity, smoking history (ever versus never) and alcohol consumption (heavy versus other).
Conditional multivariate logistic regression was used to estimate the odds ratio of breast cancer associated with various hormone formulations, including unopposed estrogens, opposed estrogens (both oral and transdermal forms) and tibolone, compared with never-users of HRT. Because we used incidence density sampling, the odds ratio is an estimate of the RR of events among exposed women. Separate odds ratio estimates of breast cancer were calculated for users of only one type of HRT formulation versus those who had also been exposed to multiple formulations of HRT during the time period prior to the index date. Exposure to HRT in the 2-year period preceding the index date was not included to allow for adequate latency. The number of prescriptions of each HRT category was also analysed as a continuous measure. SAS version 9.1.3 was used.
The study protocol was approved by the Scientific and Ethical Advisory Group of the GPRD.
We identified a population of 514 852 women in the GPRD who were 50–75 years between 1 January 1998 and 31 December 2004 and belonged to an ‘up to standard’ practice. Of these, 12 032 women were excluded due to a history of breast cancer before the age of 50 years, leaving a final cohort of 502 820 subjects. Within this cohort, we identified 6347 cases of incident breast cancer using the physician-validated computerised algorithm, after excluding 125 with less than 2 years of history, who were matched with 31 516 controls (Figure 1).
Subjects were on average of 61 years old and had a mean duration of follow up of approximately 7 years and around 30 years of information history within the GP record in the database (Table 1). Cases and controls were similar in their smoking history (ever smokers 48.9% among cases versus 47.2% controls) and weight (38.6% cases with a normal body mass index versus 38.3% of controls). A majority of women had never been exposed to any formulation of HRT (60.6% of cases versus 64.1% of controls).
Table 1. Characteristics of cases of incident breast cancer and age-matched controls
Breast cancer cases
Age (mean ± SD)
61.2 ± 7.3
61.2 ± 7.3
Duration of follow up, years (mean ± SD)
6.7 ± 4.3
7.1 ± 4.3
Duration of observation period, years (mean ± SD)
29.6 ± 14.6
29.7 ± 14.3
Body mass index (%)
Smoking status (%)
Alcohol consumption (%)
Social or abstainer
Medical history (%)
The adjusted odds ratio of breast cancer associated with the exclusive use of opposed estrogens (i.e. combined estrogens with progesterone, henceforth referred to as CEP) was 1.33 (95% CI 1.23–1.44) relative to nonusers of HRT (Table 2). The use of estrogen alone (i.e. without progesterone), present in fewer than 10% of the postmenopausal women studied, was not associated with a higher risk of breast cancer (OR = 0.97; 95% CI 0.86–1.09). The most common estrogens used were Prempak-c and Premarin, representing 19 and 11.6% of prescriptions issued, respectively. Tibolone as the only HRT, used by 1.2% of the controls and representing approximately 30% of all tibolone use, was associated with odds ratio for breast cancer of 0.86 (95% CI 0.65–1.13). Women exposed to progestins alone were uncommon (less than 2% of women) and also did not exhibit an increased risk of breast cancer (OR 1.05; 95% CI 0.85–1.30). The mean time from the first recorded opposed estrogen prescription to the index date among users of these drugs was 2681 days, compared with 2891 days for estrogen only and 2365 days for tibolone.
Table 2. Adjusted* odds ratios of breast cancer incidence associated with lifetime use of single hormone therapy types
Use of HRT, % (n)
Cases (n= 6347)
Controls (n= 31 516)
Adjusted OR (95% CI)
Adjusted for endometrial cancer, hysterectomy, oophorectomy, family history of breast cancer, documented oral contraceptive use, body mass index, smoking and drinking status.
No use ever (reference)
64.1 (20 200)
Opposed estrogens only
The increased risk with the use of opposed estrogens remained elevated for oral medication regardless of whether the progesterone was given in a continuous (OR 1.29; 95% CI 1.07–1.56) or sequential (OR 1.33; 95% CI 1.21–1.46) manner (Table 3). Among users of opposed estrogen, the rate of breast cancer varied significantly depending on the type of progestin used. The use of C-21 based progestins appears to be associated with a slightly higher risk of breast cancer (OR 1.53; 95% CI 1.22–1.92) than that of the C-19 based progestins (OR 1.29; 95% CI 1.18–1.40). However, women using the transdermal formulation of opposed estrogens (patch) did not have a significant association with breast cancer (OR 1.08; 95% CI 0.81–1.43). Finally, Table 3 also shows that tibolone users who had also been previously exposed to traditional HRT had a significant increase in risk when this previous medication was opposed estrogen (OR 1.29; 95% CI 1.09–1.52) but not if it was estrogen alone (OR 1.22; 95% CI 0.74–2.00).
Table 3. Adjusted* odds ratios of breast cancer detailing of more than one type of HRT
Adjusted for endometrial cancer, hysterectomy, oophorectomy, family history of breast cancer, documented oral contraceptive use, body mass index, smoking and drinking status.
There were 7.05% of cases and 6.00% of controls who used more than one type of these agents over time.
No use (ever)
64.1 (20 200)
Both oral and patch
Tibolone preceded by
Subjects who were exposed to oral opposed estrogen therapy have a 25% (95% CI 20–30%) increased risk of breast cancer for every ten prescriptions issued (Table 4). Opposed estrogens were used for a mean of 1320 days ± 1160 (SD). Such a quantity-response effect was not seen with the use of transdermal formulations of opposed estrogen therapy, estrogens alone or tibolone alone. However, women exposed to tibolone who had also previously been exposed to opposed estrogens had a 29% (95% CI 10–51%) increased risk of breast cancer for every ten prescriptions of any of these agents.
Table 4. Adjusted odds ratio of breast cancer per ten prescriptions of HRT
This study confirms that different formulations of HRT prescribed to postmenopausal women confer differing relative risks of breast cancer in a community-based setting. Specifically, we found that the use of oral CEP combinations delivered in either sequential or continuous manner both increased the odds ratio of breast cancer by approximately 30%, compared with nonusers of HRT, which is very similar in magnitude to the results found by the WHI study.2,5 Furthermore, a dose–response effect was seen with the use of these medications, resulting in an increased odds ratio of breast cancer by 25% for every ten prescriptions issued. Also, the risk of breast cancer differed depending on the progestin used in combination with estrogen. In contrast, the use of estrogens alone and transdermal estrogen + progesterone delivery did not increase the RR of breast cancer. The use of tibolone, a pharmacologically distinct form of HRT, also did not increase the odds ratio of breast cancer. Women exposed to tibolone who had also received oral CEP had an increased odds ratio of breast cancer.
Much has been published with respect to HRT and the subsequent risk of breast cancer, and some studies have come to conflicting conclusions.1–5,13–15 Our finding that oral CEP is associated with an increased risk of breast cancer is consistent with that of many large observational studies and randomised control trial data including the WHI and MWS.1,5
In this article, we confirm the findings of the WHI estrogen-only arm that women exposed to estrogen alone do not have an increased risk of breast cancer.2 Results of studies have varied in their findings on the risk of breast cancer and exposure to estrogen. Both our study and the WHI results conflict with the MWS, which reported an increased risk of breast cancer in women exposed to estrogen alone. However, almost 30% of women in the MWS had been exposed to more than one formulation of HRT, which was not corrected in the calculations of risk. In contrast, the WHI controlled exposure to other HRT formulations through its randomisation process, while our study was able to analyse risk separately in women who were exposed to estrogen as their only HRT. Therefore, it may be that the increased RR seen in the MWS among users of estrogen alone was the result of previous CEP exposure rather than from the use of estrogen itself. The possible biological explanation of lack of estrogen association with breast cancer may be breast cancer cell apoptosis on exposure to estrogen in certain conditions including after an estrogen-deprived state as is the suggestion by some in vitro models.16,17 The increase in breast cancer risk seen with the addition of progestin may reflect a biological potentiation of estrogen on breast cancer cells.
We also compared the odds ratio of breast cancer among users of oral versus transdermal HRT. We found that users of the transdermal form did not have an increased risk of breast cancer. The result is possibly due to differences in the pharmacokinetics and pharmacodynamics between these two formulations. The transdermal route provides a low constant level of hormone in the blood, which furthermore avoids hepatic protein synthesis stimulation in contrast to the oral form, which has a peak level following gastric absorption after which there is a trough prior to the next dose.18 Some observational studies have shown the risk of breast cancer with the use of transdermal HRT to be of similar magnitude to oral formulations.1,19 This may perhaps be explained by the fact that in these studies, a subset of women had been exposed to more than one type of HRT formulation, which was not taken into account in the analysis. Reanalysis of available data from other large observational studies in the same method as our study could clarify these conflicting results.
We further analysed CEP according to the progestin type used. Interestingly, we found that the use of C-19 progestins had a lower risk of breast cancer than the C-21 progestins when used in combination with estrogens. The two progestins differ as the C-19 possesses some androgenic qualities, and thus, the difference in breast cancer risk may be explained by differential effects on breast tissue.20 Of note, progestin use alone was not associated with an increased risk of breast cancer. This would suggest that the progesterone may potentiate the proliferative effect of estrogen on the breast and warrants further investigation in future studies.
Tibolone is used as a treatment of menopausal symptoms outside of North America. It is classified as a STEAR, given its estrogenic activity on tissues including the vagina, bone and brain without exerting estrogenic activity in the breast or endometrium.21,22 Recent in vivo studies have also demonstrated evidence that tibolone stimulates apoptosis and inhibits cell proliferation on breast cancer cells.20 Our study confirms what was postulated by experimental data that there is no increased risk of breast cancer among postmenopausal women who had used tibolone alone (Table 2). However, women who had also been exposed to oral estrogen or oral CEP did have an increased risk of breast cancer (Table 2), although to a lesser extent than exposure to CEP alone. Therefore, this increased breast cancer risk may be due to the effect of oral CEP rather than from the use of tibolone. Previous studies, including the MWS, have classified women according to the last HRT preparation they used, not taking into account previous HRT exposures. This may well explain the conflicting results previously seen.
Due to the nature of the database, we were unable to look at specific histological subtypes of breast cancer and their relation to HRT. Also, the database provides information on medications prescribed, not dispensed. This should not affect our results since we looked at effect of HRT over years of exposure, where medications prescribed are very likely to accurately reflect medications consumed.
Our study has a number of strengths. The large numbers of postmenopausal women allow for a well-powered study. Also, exposure data from the GPRD is recorded prospectively with each prescription. This makes it ideal for pharmacoepidemiological studies, removing the real potential for subject recall bias of HRT exposure that is present in some other large observational studies including the MWS and E3N-EPIC.1,19 Also, the cases of breast cancer were identified through a validated protocol described in our methods. A 2-year latency period following HRT exposure was used which also helps to avoid detection bias. This is evidenced in the MWS, where the risk of breast cancer among women was seen from the very first month of HRT treatment, highly suggestive of detection bias rather than a cause-effect relationship.1 Our analysis provides information on breast cancer risk for different users of specific HRT formulations, thereby giving clean data results. We also considered only incident cases of breast cancer to remove women who are likely to be prescribed HRT in a different manner, if at all, which would have resulted in a bias by indication.
Our study has several potential limitations. The exposure to HRT is based on analysis of prescriptions written rather than prescriptions filled. However, for long-term exposure to medications that must be picked up and filled on a monthly basis, as is the case with HRT, this is not likely to be a source of bias. Also, due to the population-based nature of the GPRD, the exposure to medications reflects real-world context where people are exposed to multiple medications rather than one single medication. However, we were able to take this into account in the analysis.
Our study confirms that the type and formulation of HRT results in differing risks of breast cancer. Oral estrogen–progestin combinations appear to be the only form associated with an increased breast cancer risk, while transdermal HRT, estrogens alone and tibolone are not.
Conflict of interest
S.S. has received funding for this research project from Organon, the manufacturer of tibolone, and research funding from Wyeth, a manufacturer of hormone replacement therapy, for a project unrelated to these therapies.
Commentary on ‘Hormone replacement therapy use and variations in the risk of breast cancer’
Opatrny et al. have applied the case–control design rigorously to the question of breast cancer risk from various hormone replacement therapy (HRT) preparations. Dependence on case–control design for evaluating rare and long-term outcomes such as breast cancer is often necessary in epidemiological enquiries, but the potential for biases and erroneous inferences are immense: they can suffer from selection bias, recall bias and ascertainment bias to name but a few, and no amount of meticulous adjustments for confounders will deal with all the biases. The literature on HRT and cardiovascular risk serves as an example of the need to maintain a healthy dose of scepticism when evaluating evidence from observational studies.
Let us take the finding that ‘tibolone does not increase the risk of breast cancer’. An examination of Table 2 shows that this finding comes from less than 1% of 6347 cases. Two important consequences of such a small number are a lack of power to detect associations between the exposure and breast cancer and the instability of estimates. Therefore, it should come as no surprise that the 95% CI around the breast cancer risk for tibolone includes an increased risk of up to 13% (adjusted RR = 0.86, 95% CI 0.65–1.13). Let us now remind ourselves of the findings from the larger Million Women Study (Banks, Lancet 2003;362:419–27), which is of a stronger design as it was a cohort study: ‘(in).women, who were likely to have used tibolone exclusively, the relative risk of breast cancer in current users was 1.48 (1.20–1.83, P < 0.0001). Current users of tibolone at baseline did not differ substantially from other current users of HRT for known risk factors for breast cancer. Additional adjustment for factors which were not already included in the analysis model did not alter the estimates of relative risk of breast cancer associated with current use of tibolone’. It would, therefore, be very prudent not to rush out with the message that tibolone is safe from the breast cancer point of view.
The finding that oral combined HRT is associated with an increased risk of breast cancer is consistent with the findings of the Women’s Health Institue (WHI) (JAMA 2002;288:321–33) and Million Women studies. Furthermore, Opatrny et al. confirm the finding of the WHI estrogen-only arm that women exposed to estrogen alone do not have an increased risk of breast cancer (JAMA 2004;291:1701–12). These are useful confirmations. However, for the other findings, they are a source for further research questions and should not serve as the basis for recommendations in clinical practice.
The database access was funded by grants from the Canadian Institutes of Health Research (CIHR) and the Canadian Foundation for Innovation. The study was funded by an unconditional grant from Organon. L.O. is the recipient of a research salary support grant from the Fondation de recherché en sante du Quebec. S.S. is the recipient of a Distinguished Investigator award from CIHR.