Abstract
‘Fair examination of the evidence concerning HRT use does not support the negative perceptions that have resulted for its mainstream use in women in the early postmenopausal years.’
The publication of the first large-scale randomized controlled clinical trial (RCT) of postmenopausal hormone replacement therapy (HRT) has had a major impact on perceptions of its role in women's health. These perceptions have been predominantly negative and levels of use of HRT have declined across the world. It is important that our understanding of the effects of medicines is based on the best quality evidence. Fair examination of the evidence concerning HRT use does not support the negative perceptions that have resulted for its mainstream use in women in the early postmenopausal years.
The question, ‘Does HRT have a future?’ would have been unthinkable 5 years ago, but much can happen in 5 years. In that time, and particularly since the first publication from the estrogen/progestogen (E/P) component of the Women's Health Initiative (WHI) Study, only 3 years ago in 2002 [1], there has been a flow of papers which have contained both positive and negative data on the effects of HRT. However, the negative information has, understandably, received the greatest attention from the popular and professional media. This negative focus has resulted in a substantial proportion of HRT users around the world becoming concerned and stopping their HRT.
The positive information contained in those same papers received minimal media attention but confirmed much of the previous understanding of the effects of HRT, which had been based on observational evidence. The negative information also provided confirmation of the existing knowledge on risk of breast cancer and thrombosis. What appears to have been so damaging to confidence was the confirmation of these risks, especially breast cancer risk, as well as the overturning of the claims that HRT could reduce the risk of heart attack and stroke. Indeed the early reports suggested that heart attack rates were actually increased. The sequence of papers that has followed through 2002–2005 has provided the necessary detail to permit reasoned debate on the WHI results, but reasoned debate does not often attract headlines. In 2004 the publication of the estrogen-only (E) WHI study [2] largely supported the general picture provided by the WHI E/P study [1] but with the important difference that there was a non-significant trend for a reduction in breast cancer risk in these hysterectomized E HRT users. This lack of evidence of breast cancer risk did have some media coverage but the overall perception of the public and medical professionals has remained negative compared to a few years ago. This negative tone has been accentuated by the publication of data from the observational Million Women Study in the UK, which have attracted considerable, and largely negative, media attention by suggesting increased breast cancer risk in both E and E/P users of HRT, with greater risk in E/P users [3].
I would like to consider how such a turnaround in a well-established field of medicine could take place in such a short time and ask, ‘What is the sensible position of HRT in the care of women?’
Following the revolution in thinking that was termed ‘evidence-based medicine’, we now accept that clinical medicine should base its evaluation of treatments on evidence that is structured, with particular weight given to the highest levels of evidence. In that structure the RCT reporting real clinical events sits at the top of the tree, particularly when clinical effectiveness is the focus. These are regarded as the least affected by bias. RCTs reporting only on effects on surrogate markers are regarded as less important because changes in surrogate markers might not actually reflect changes in the occurrence of real clinical events.
Evidence does not just look at effectiveness but also examines risk. However, risk events tend to be infrequent so an RCT needs to be very large indeed to have sufficient statistical power to report meaningfully on risk events and most RCTs are very underpowered for this purpose. As a result, the best information on risk events tends to be from observational studies since these can involve much larger numbers of patients in case–control or cohort studies that can collect sufficient risk events. These study designs are vulnerable to many potential biases because they look at the real world and, as a result, observational studies may report differences in outcome between users and non-users that are not actually due to the therapy being studied but due to inherent differences in the people between the user and non-user groups.
Against this background the evidence around HRT is especially complicated because of the wide range of benefits and risks that have been associated with its use. The longest established and most straightforward of these is the use of HRT for relief of menopausal symptoms. These symptoms are straightforward to study in placebo-controlled RCTs of relatively short duration over a timescale of months, in reasonably small patient groups. As a result we have robust evidence for the efficacy of HRT for the relief of menopausal symptoms. This evidence has not been challenged and this was not the focus of the recent papers.
The situation is much more difficult when the effects of HRT under investigation are reductions in major disease events, such as osteoporotic fracture and heart attacks, since these occur over extended timescales of years and only in a relative minority of individuals. Since RCTs in this scenario need to be very large and last for years, the field had based its knowledge on less reliable study methods: RCTs reporting surrogate clinical markers (e.g., bone mineral density and blood lipid profiles) and/or observational studies reporting clinical events (e.g., fractures and coronary events). On that basis, the list of long-term health benefits of HRT that were widely discussed included reductions in coronary heart disease, stroke and Alzheimer's disease, as well as osteoporosis. Of these, only osteoporosis prevention achieved the recognition of becoming a licensed indication. The evidence gap over these long-term benefits was a difficulty because the potential health gains for society were huge if they could be confirmed. Even without RCT confirmation, I understand that increasing numbers of American women were using HRT for the prevention of coronary heart disease. This was much less the case in Europe. The motivation for the WHI trials was the need to clarify the coronary heart disease benefit and at the same time provide evidence on the other health gains that might result from HRT used over an extended period of years. With this aim the age range of women was wide (50–79 years) and very different from the generally under 60 years, early-postmenopausal age group relevant to menopausal symptom HRT use.
‘Where women suffer estrogen deficiency through premature menopause or bilateral oophorectomy, there has been no suggestion that the role of HRT is being questioned.’
In the WHI study, relative risk of clinical events was similar across the different age groups but the absolute risks were very different because the incidence of events in the placebo group rose markedly with increasing age. Thus, the change in relative risk between those given HRT and those given placebo in the trial must be considered in the context of the incidence to indicate the relevance of the identified effects of HRT. Overall, the absolute risks of HRT use were not found to be substantial. However, in the oldest group of women in the study the excess of events, whilst not substantial, was sufficient that many women in that age group (70–79 years) will prefer to avoid HRT. This will be a value judgement but rates for different events, as follows, will deter many. For the 70–79 year age group the excess absolute risk associated with E/P use was as follows: coronary heart disease (23/10,000/year); stroke (13/10,000/year); thrombosis (35/10,000/year) and breast cancer (13/10,000/year) [4–7]. For the E group the excess risk was noticeably lower at 4/10,000/year; 14/10,000/year; 12/10,000/year; and −2/10,000/year, respectively [2]. This level of risk of unwanted events must be understood in the context that HRT doses designed for women generally below 60 years of age may be inappropriate in women over 70 years of age. Indeed, if there is interest in reducing the risk of major adverse health events, such as coronary heart disease and dementia, in women over 70 years old, there is a prominent view that this may be best delivered by the use of HRT in the decade before 60 years of age, based on the evidence provided by mechanistic and animal studies. Such questions are the subject of current debate but they are not addressed by the WHI trial data.
The converse of this is that the absolute risk of events in women under age 60 years using HRT is low, especially in those who do not use P because they have had a hysterectomy. This group in the WHI report experienced 2/10,000/year additional thrombosis events and no excess of any other events. The excess absolute risk associated with E use for the 50–59 year age group were as follows: coronary heart disease (−10/10,000/year); stroke (0/10,000/year); thrombosis (2/10,000/year) and breast cancer (−8/10,000/year). For the E/P group the excess risks were higher at 5/10,000/year; 4/10,000/year; 9/10,000/year; and 5/10,000/year, respectively. Thus, in younger HRT users, the risk is negligible in hysterectomized women and low in E/P users. From that same study it is important to point out that these levels of excess risk are similar to the range of variation seen between the risk in the two placebo groups from the E/P and E studies. The difference in risks between placebo groups in women aged 50–59 years were: coronary heart disease (7/10,000/year); stroke (6/10,000/year); thrombosis (5/10,000/year) and breast cancer (3/10,000/year).
My comment on these comparisons has not included any mention of statistical significance because the WHI study was powered to address events in the whole study population and the 10-year age subgroups are relatively underpowered for valid statistical questions. On the other hand, these age-based subgroups provide the largest RCT information we have on the effects of HRT.
‘Where the future of HRT is most in doubt is in the uses that have not been licensed, and which relate to other long-term health benefits.’
I believe that the data on risk in the youngest women in the WHI study should not be a source of concern for clinicians or women, and this is supported by the position taken by the regulatory agencies. Sadly, there is a common perception that HRT use is now regarded as generally dangerous. There is an understandable tendency for busy people to get their information in digested form, but in the case of HRT broad-brush messages hide very important detail, as I have described above. Those who take time to read the detailed WHI papers and the European Agency for the Evaluation of Mecicinal Products (EMEA) or the UK Medicines and Healthcare products Regulatory Agency (MHRA) regulatory statements on HRT, issued at the end of 2003, will find reassurance for the most common uses of HRT [8]. The MHRA statement highlighted its view of the place of HRT in the management of osteoporosis and emphasized that it does not recommend HRT as a first-line option in that context [9]. The statement emphasized that when HRT is used in the management of menopausal symptoms, the benefit outweighs the risk. Menopausal symptom relief is by far the commonest reason that women use HRT, yet, in the wake of the media and regulatory attention, large numbers of menopausal women have abandoned use of HRT, despite being mostly below 55 years of age. There is no good scientific basis for these fears.
The regulatory statement also made it clear that it was referring to the use of HRT in women over 50 years of age. Where women suffer estrogen deficiency through premature menopause or bilateral oophorectomy, there has been no suggestion that the role of HRT is being questioned. However, the generally negative atmosphere around HRT is leading these younger women to be concerned.
The role of HRT in the management of osteoporosis is the most controversial of the licensed indications for a number of reasons. There are alternative bone-sparing drugs available, about which there is good evidence of effectiveness against fracture in older women with osteoporosis, and the priority for drugs in osteoporosis management is in the secondary prevention of fracture. Those at highest risk tend to be women over 65 years where the HRT burden of risk is more significant, hence the argument that HRT should not be the first-line option. A different question concerning HRT and osteoporosis is whether HRT can have a role in the prevention of the first fracture in women at risk, who may well be below 65 years. We have the RCT evidence that HRT use is associated with fewer fractures than placebo use [2,10], even in women below 60 years of age, which is a group in which the major osteoporosis drugs such as bisphosphonates have been little tested. There is room for a legitimate argument on whether the use of drug therapy for the prevention of osteoporosis in the under 60s can be cost effective, but if intervention is to be applied in this situation it can be argued that HRT is the least expensive drug option and that its risk burden in this age group is low. Thus, in primary prevention of osteoporotic fracture, the placing of HRT in a second-line category may be inappropriate.
The question posed in this article is whether HRT has a future. I have put forward my view that the future of HRT has been seriously undermined by the presentation of the findings of the field's only major RCT in the popular and professional media. Whilst you can argue about aspects of that trial, its data cannot be ignored. Where the future of HRT is most in doubt is in the uses that have not been licensed, and which relate to other long-term health benefits. The major postulated benefits were not observed in the WHI study and, whilst the debate around that continues, we may never have the necessary RCT data to clarify the matter. However, I believe that the WHI study does not undermine the solid core of HRT practice and I believe that the researchers involved agree with this. That solid core is estrogen replacement for premature deficiency, HRT use for menopausal symptoms, and possibly primary prevention of osteoporotic fracture in women under 65 years. In these principal areas of HRT practice the evidence has not changed. In my opinion, HRT has a well-established past in these areas of clinical practice and has a solid future in the care of women, if common sense and evidence can prevail.