Summary of findings
Despite a recent drop in both the incidence and prevalence of cardiovascular disease (CVD) it still remains the leading cause of death in the developed world (Deedwania 1990). The main forms of CVD are coronary heart disease (CHD) and stroke. In 2007, CVD caused 34% of deaths in the UK, and killed just over 193,000 people, with approximately 25% of these deaths from CHD and 9% from stroke (British Heart Foundation Statistics database). CVD is therefore the most common cause of death in the UK, accounting for one in five deaths in men and one in seven in women.
In terms of premature mortality (death before the age of 75) in 2008, approximately 30% of premature deaths in men and 22% in women were attributable to CVD (Townsend 2012). The burden of CHD is costly both in terms of reduced patient health-related quality of life (HRQoL) and high health care costs in the management of the conditions. Morbidity statistics indicate that CVD is the leading single cause of disability in Europe, with a prevalence of 6.0% to 6.5% in men and 4.0% to 4.5% in women within the UK. CVD is therefore costly in terms of both direct and indirect health care costs, accounting for 9.8% of total disability-adjusted years (Townsend 2012). In 2006 it was estimated that CVD cost the UK health care system approximately £14.4 billion, equating to approximately just under £250 per capita. The costs for the treatment of stroke are similar to those for other forms of CHD (Hsia 2006).
Description of the condition
The risk of CVD is higher in men than in women in younger age groups, with women’s CVD incidence rates found to lag approximately ten years behind those of men. Between 45 and 64 years of age, the prevalence of CVD in men is several times that of age-matched women (Isles 1992; Tracy 1996). Most women experience the menopause (the last menstrual period) in their early fifties, after a phase of changing ovarian function (the peri-menopause) that may last several years and which is characterised by irregular menstrual cycles (Greendale 1999). Following menopause and loss of endogenous estradiol (major ovarian oestrogen), these gender-based differences narrow (Barrett-Connor 1997; Maxwell 1998). Most women who enter menopause are asymptomatic for CVD, and 95% of women who develop CVD do so after menopause. Evidence suggests that younger age at natural menopause is associated with CVD (Hu 1999) and CVD mortality (Jacobsen 1997; van der Schouw 1996) when comparing age-matched post-menopausal women. Menopause has an adverse effect on lipid profile. Low-density lipoprotein (LDL) rises for approximately 10 to 15 years after the menopause, and high-density lipoprotein (HDL) drops (Matthews 1989). Weight gain and a change in body fat distribution, increases in blood pressure and a host of other metabolic factors are amongst the other changes that may affect the risk for the development of CVD.
Description of the intervention
The term “hormone replacement therapy” has been replaced by “hormone therapy” as the older term infers that hormone therapy is replacing the function of a defective organ. Hormone therapy (HT) is now the preferred term for this intervention. Hormone therapy includes either oestrogen alone (estrogen-only HT) or oestrogen in combined with a progestogen (combined HT). It is used in a variety of formulations and doses which can be taken orally, vaginally, intra-nasally or as an implant, skin patch, cream or gel. The clinical effects vary according to the type of HT and the duration of its use. Formulations of oral oestrogen generally comprise 1 - 2 milligrams of estradiol daily and may include oestradiol (an oestrogen derived from wild Mexican wild yam), oestradiol valerate (a pro-drug for oestradiol), or conjugated equine oestrogen (CEE) a blend of equine estrogens extracted from horse urine. The progestogens used for HT include synthetic derivatives of progesterone, synthetic derivatives of testosterone, and natural progesterones derived from plants. These differ in their metabolic action and potential for adverse effects, and the risk-benefit profile of each type of progestogen for use in HT is currently unclear. In combined HT, progestogen can be taken either every day (continuous combined therapy), cyclically with estrogens taken daily and progestogens taken for part of the month (sequentially combined HT), or less frequently.
The addition of a progestogen to oestrogen reduces the risk of endometrial hyperplasia associated with the use of oestrogen alone in women with a uterus (Furness 2009). However, the addition of progestogens can be problematic as they have adverse effects on blood lipid profiles and may cause symptoms such as headaches, bloating and breast tenderness (McKinney 1998).
How the intervention might work
The finding that CVD rates in women rise sharply after the menopause has led to the suggestion that endogenous estradiol may attenuate age-related vascular remodelling in pre-menopausal women. Age-associated vascular remodelling involves endothelial dysfunction, enhanced growth of intimal smooth muscle cells (SMCs), and increased prevalence of vascular plaques. The same cellular processes participate in atherosclerosis (Lakatta 2003). Additionally, changes in the androgen-to-estradiol ratio may contribute to the negative effects observed on the cardio-vascular system post menopause. The decline in estradiol levels during menopause leads to a higher androgen-to-estradiol ratio. Androgens induce vasoconstriction and SMC growth and exacerbate diet-induced atherosclerosis, plaque formation, and pro-atherosclerotic arterial remodelling. This suggests that the increase in the androgen-to-estradiol ratio in post-menopausal women may be another mechanism which contributes to the acceleration of atherosclerosis observed.
The exact mechanism by which CVD risk may be reduced by oestrogen is not completely understood, but it is widely known that estradiol inhibits many processes involved in age-associated vascular remodelling, including SMC proliferation and endothelial dysfunction, and lowers cholesterol and improves vascular tone (Dubey 2001; Mendelsohn 1999). Other factors that may play a role are changes in coagulation factors, blood pressure, insulin, and body fat distribution (Lieberman 1994; PEPI Trial Writing Group 1995). Exogenous ovarian steroid hormones have multiple target tissues in addition to the vascular system, including the bones, endometrium, and breast, and HT has the potential to effect the risk of several additional conditions including osteoporosis and dementia (Grady 1992).
Why it is important to do this review
Hormone therapy to treat menopausal oestrogen deficiency has been in widespread use for more than 60 years (Wallach 1959). Long-term treatment was assumed to prevent atherosclerosis, and the increased CVD and mortality risk observed following the menopausal transition (Robinson 1959; Wallach 1959; Wilson 1963). Since the early 1980’s several observational studies consistently showed that HT users, many of whom started treatment shortly after menopause, had a significant reduction in total mortality and risk of CVD events of approximately 30% to 50% relative to women who chose not to use HT (Grady 1992; Grodstein 1999; Grodstein 2000; Mann 1994; Psaty 1994; Rosenberg 1993; Stampfer 1991). This reduction in risk was apparent whether the HT regimen used was oestrogen alone or oestrogen in combination with progestogen. However, most observational data sets suggest that the risk reduction in mortality and CHD events, is coupled with a higher impact of the risk of venous thromboembolic events and an apparent increased incidence of stroke but lower stroke mortality (Paganini-Hill 2001). Overall, the accumulated available epidemiological evidence supported the use of HT to increase longevity in post-menopausal women (Mishell 1989).
It was recognised that there was a need for randomised controlled trials (RCTs) in the area (Barrett-Connor 2001; Hemminki 2000a), and that the wide prescribing of HT in the 1990’s, despite the lack of RCT evidence of its effects, might reflect a conflict between commercial and professional interest groups and good public policy (Hemminki 2000b).
The publication of the results from the Heart and Estrogen/progestin Replacement Study (HERS I 1998) and the Women’s Health Initiative (WHI I 2002) trials appeared to strongly contradict conventional clinical practice based on the evidence from observational studies. HERS I 1998 was a secondary prevention trial studying the effects of combination therapy (oestrogen and progestogen) on the risk of CHD events (non-fatal myocardial infarction plus CHD-related death) in 2763 post-menopausal women with established CHD. Whilst, there was an excess risk of CHD events in the HT group in the first year on treatment; for the overall 4.2 years of follow-up, there were no differences in CHD events between the HT and placebo groups, coupled with an increased risk of both venous thromboembolism and pulmonary embolism. A further 2.7 year of follow-up still showed no CHD benefit (Grady 2002). The WHI I 2002 was a primary prevention trial conducted in 16,609 post-menopausal women without hysterectomy. Participants were randomly assigned to combination therapy (the same regimen as used in HERS I 1998) or placebo. There was an excess risk of CHD in the first year and a nearly 30% increased risk of coronary events after 5.6 years. A sub-group analysis of the 400 women included in the trial who had a history of myocardial infarction or coronary revascularization showed a similar risk.
In light of these trials not confirming a cardioprotective effect of estrogens, the age of the women enrolled in both HERS I 1998 and WHI I 2002 (mean age: 65 years), and subsequent WHI I 2002 analyses, in which non-significant data trends suggested HT did not lead to excess coronary risk when started close to the menopause, interest alighted upon the timing of initiation of HT in relation to the time of menopause. This ‘timing hypothesis’ suggests that there is a window of opportunity where HT may be beneficial for prevention of CVD in women 10-years post menopause, but that in older women, it does not appear to have the same benefits and may be associated with excess CVD risk. Biological plausibility exists to support the timing hypothesis, in which it is posited that oestrogen therapy has a negative impact on atherosclerotic arteries (i.e. causes events when vulnerable plaques are present) but prevents atherosclerosis if begun early enough. This hypothesis fits with results in the Clarkson non-human primate model, where conjugated equine oestrogen (CEE) prevented atherosclerosis only in animals treated early after castration (within the calculated equivalent of six human post-menopausal years) before the onset of diet-induced atherosclerosis (Mikkola 2002). It is plausible therefore that oestrogen effects differ with the stage in the natural history of the disease and the severity of subclinical disease.
In support of the ‘timing hypothesis’ a stratified meta-analyses by Salpeter 2004 indicated differential treatment effects with HT relative to placebo according to the participants' baseline age. The Salpeter 2004 analyses assessed 30 RCTs which compared HT with placebo that included 26,798 participants and reported at least one death, to assess the effect of HT on total mortality, mortality due to CV disease, cancer, or other causes. Results indicated a significantly reduced risk of death in women with a mean age of under 60 years taking HT compared to a placebo group, though no difference was found when older women were compared. However, this meta-analysis pooled trials which differed widely with respect to the type of HT used, the clinical status of the participants, and in many of the trials death was not a pre-specified outcome. Furthermore, 60% of the events in the meta-analysis of trials in younger women were observed in women with poor prognosis ovarian cancer. It is therefore unclear, as to how applicable the results of this meta-analysis are to either healthy post-menopausal women or those with an existing CVD taking either oestrogen alone or oestrogen in combination with progestogen.
The original Cochrane Review on HT for the prevention of CVD in post-menopausal women (Gabriel-Sanchez 2005) identified a total of ten RCTs which included 24,283 post-menopausal women (12,353 randomised to HT and 11,930 to placebo). The review reported no protective cardiovascular effects for HT observed in either healthy women or women with one or more pre-existing CVD risk factors, but a higher risk of stroke, venous thromboembolic events and pulmonary embolism was observed.
Since the publication of the original Cochrane review a number of further trials, for example, the Estonian Postmenopausal Hormone Trial (EPHT 2006), and the Women’s Health Initiative trial on the effects of oestrogen alone in women with hysterectomy (WHI II 2004) have been published. These, along with other trials that have been published in the interim time, will up-date the evidence base regarding the risks and benefits observed with HT use compared with placebo, and provide up to date evidence on these to help aid both clinicians and patients in their decision making regarding the potential use of HT. Additionally, the previous review (Gabriel-Sanchez 2005) did not include Health Related Quality of Life (HRQoL) as an outcome measure, which may be of importance to patients.
Moreover, the original review, statistically pooled trial data for each outcome measure of interest from trials with different lengths of follow-up. Whilst it is acknowledged that this is standard practice, the addition of further trial evidence may allow stratified analyses to be conducted to assess the impact of age (as a proxy for time since menopause) on CVD outcomes, and the effect of time on treatment, or provide the additional power (from more trials assessing the same outcome) for this to be explored using either univariate or multivariate meta-regression models (Higgins 2010).
To assess the effects of HT for the prevention of CVD in post-menopausal women, and whether there are differential effects between use of single therapy alone compared to combination HT and use in primary or secondary prevention.
Secondary aims were (i) to undertake exploratory analyses to assess the impact of mean age of trial participants at baseline as a proxy for time since menopause (> 60 versus < 60 years of age) and (ii) effects of length of time on treatment.
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) comparing oral HT with either placebo or a no treatment control for a follow-up duration of six months or longer were included. RCTs which compared two or more different types of oral HT were included provided that they were additionally compared with a placebo or a no treatment control arm.
Types of participants
Post-menopausal women (with either spontaneous or induced cessation of menstrual bleeding for a continuous period of six months or more), either with or without evidence of existing CVD.
Types of interventions
Oral Hormone Therapy (HT), consisting of either oestrogen alone or in combination with a progestogen compared with either a placebo or a no treatment control. Combined HT (oestrogen plus progestogen) could be delivered continuously daily (continuous combined HT) or sequentially (oestrogen taken daily with progestogens taken for part of the month).
RCTs in which HT was delivered to the body via either patches, tablets, creams, troches, an intrauterine device (IUD), vaginal ring, gels or injections compared with placebo or no treatment were excluded in accordance with the inclusion criteria from the previous review (Gabriel-Sanchez 2005). Likewise RCTs assessing the effects of selective oestrogen receptor modulators (SERMs) (e.g. raloxifene) compared to placebo or a no treatment control were not included.
Types of outcome measures
- Death from any cause.
- Cardiovascular death.
- Non-fatal acute myocardial infarction.
- Pulmonary emboli.
- Venous thromboemboli (pulmonary emboli plus deep vein thromboses).
- Coronary artery by-pass graft (CABG).
- Angioplasty (with or without a stent).
Any included trials were then searched for additional assessment and reporting of health-related quality of life (HRQoL) obtained using a validated outcome measure.
Trials reporting only intermediate CVD outcomes, such as blood pressure, cholesterol levels, or coagulation factors were not included.
Search methods for identification of studies
Randomised controlled trials (RCTs) that assessed the effects of HT compared to placebo with a minimum of 6-months duration were identified through searching electronic databases. Electronic searches of the Cochrane Central Register of Controlled Trials (CENTRAL) on The Cochrane Library (April 2010), MEDLINE (OVID;1950 to April 2010), EMBASE (OVID; 1966 to Week 15 2010) and LILACS (OVID; 1982 to April 2010) were conducted. Additionally the National Research Register (NRR) and www.clinicaltrials.gov were searched for any ongoing trials on CV diseases (2002 to September 2010).
No language restrictions were applied and appropriate consideration was given to variations in terms and the spelling of terms in different countries so that potentially relevant studies would not be missed by the search strategy due to these variations. A full list of the search strategies applied are detailed in Appendix 1.
Searching other resources
Reference lists of all eligible RCTs and systematic reviews were searched for additional relevant trials. All references were managed using Reference Manager.
Data collection and analysis
Selection of studies
Relevant studies were identified in two stages. Two authors independently screened the titles and abstracts returned by the database searches for relevance. The full texts of any references that were considered as potentially relevant by either author were obtained. The relevance of each paper was then assessed according to the criteria set out above for the review question by two authors independently. This assessment was performed unblinded. Any discrepancies between the authors were resolved by recourse to the papers, and if necessary a third author was consulted.
Data extraction and management
Data were extracted from the included studies using a standardised data extraction form in Microsoft Access by two authors independently. This was checked for agreement and any discrepancies were resolved through recourse to the papers. The following study details were assessed:
- Method of randomisation.
- Method of allocation concealment.
- Use of stratification.
- Adequacy of double blinding.
- Means of recruitment.
- Number of participants screened for eligibility, randomised, analysed, excluded, lost to follow-up or dropped-out (i.e. withdrew from the trial but were followed-up).
- Baseline equality of treatment groups.
- Level of adherence to therapy.
- Whether analyses were conducted on an intention-to-treat (ITT) basis.
- Study design (parallel versus multi-arm, single centre or multi-centre).
- Funding source.
Characteristics of the trial participants
- Inclusion and exclusion criteria.
- Age and other recorded prognostic baseline variables.
- Menopausal status (definition of menopause and how this was defined, surgical or natural menopause) of participants.
- Type of HT (estrogen-only or combination oestrogen and progestogen).
- Duration of therapy (minimum six-months).
- Which relevant primary and secondary outcomes were measured.
- How relevant outcomes were defined and measured.
Assessment of risk of bias in included studies
Risk of bias was assessed according to the risk of bias assessment criteria detailed in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2010). These criteria focus upon the quality of random sequence generation and allocation concealment, blinding (participants, trial personnel and outcome assessors), incomplete outcome data, selective outcome reporting and other sources of bias. Assessment of risk of bias was undertaken by two review authors independently, with any disagreements resolved by discussion.
Statistical analyses were undertaken following the guidelines of the Handbook of the Cochrane Collaboration (Higgins 2010).
For dichotomous data, two by two tables were generated for each study and expressed as a risk ratio (RR) with 95% confidence intervals (CI). The data were grouped firstly according to intervention (single versus combination therapy) and secondly whether the intervention was primary or secondary prevention. Further analyses were undertaken to assess the effect of both single and combination therapy in the overall patient population (both primary and secondary prevention). Data were combined for meta-analysis in RevMan software using the Peto-modified Mantel-Haenszel method using a fixed effect model to provide an overall estimate of treatment effect. For comparisons showing statistically significant differences between treatment groups, the number needed to treat harm (NNH) was calculated.
Heterogeneity between studies was explored qualitatively (by comparing the characteristics of included studies) and quantitatively using the chi-squared test of heterogeneity and the I
To assess the potential impact of time since menopause trials were stratified according to the mean age of participants at baseline (> 60 versus < 60 years of age). Where data were reported for more than two different time points for each outcome (either between different trials or longitudinally within the same trial) meta-analysis were conducted stratified by length of time on treatment. These analyses were a priori classified as exploratory given the heterogeneity between the different HT regimens assessed and the patient populations in the different trials. To conduct the analyses time points for the reporting of outcomes in the trials had to be rounded up or down.
Data were rounded as follows:
WISDOM 2007 reported results after a median follow-up of 11.9 months (range 7.1 – 19.6). Results are therefore reported as though there was one year of follow-up. EVTET 2000 was conducted for a 1.3 year period; results from this trial are therefore also reported for one year follow-up. EPAT 2001, ESPRIT 2002 and HALL 1998 are all reported for two-years of follow-up. EPHT 2006 [median length of follow-up: 3.4 years (range: 2 – 4.9)], EAGAR 2006 [mean follow-up: 3.5 years (range: 25 – 41)], ERA 2000 [mean follow-up: 3.2 years (range: 2.8 – 3.8)], WAVE 2002 [mean follow-up: 2.8 years (range: 2.1 – 3.9)] and WEST 2001 [mean follow-up: 2.8 years (range: 1.6 – 4.1)] were all classified as having a three-year follow-up period.
HERS I 1998 was classified as having a four-year follow-up period. Results within the blinded part of the trial were reported at a mean of 4.1 years, with selected clinical outcomes reported for each year of follow-up. Outcome data for the 4 – 6.8 (unblinded open label) follow-up period were not included in the standard pair-wise meta-analyses but were included in the relevant stratified analyses. WHI I 2002 reported results after a mean follow-up of 5.2 or 5.6 years (range: 3.5 – 8.5). Selected clinical outcomes were also reported for each year of follow-up. Since all women had been enrolled on active treatment for at least 3.5 years at study termination, data for each of the first three year time points, and final follow-up were included in the analyses. WHI II 2004 reported results for a mean follow-up duration of 6.8 and 7.1 years (range: 5.7 - 10.7) depending on the outcome. Data were therefore classified as having 6.8 or 7.1 years of follow-up as appropriate.
Given the lack of standardisation in the HRQoL measures used in the trials, and the variation in methods and reporting, we were unable to undertake meta-analyses for these measures. Instead differences between outcomes in the HT versus placebo group for outcomes within each domain were compared and the p-value presented.
Univariate meta-regression analyses were undertaken to assess whether any particular features of the trials were potential predictors of the CVD outcomes for all cause mortality, CVD mortality, non-fatal MI, stroke, angina, venous thromboembolism and pulmonary embolism. Due to the small number of trials included in the review, we limited our exploration of outcome predictors to three variables: length of trial follow-up; whether treatment was single or combination HT; and whether treatment was for primary or secondary prevention. All analyses were conducted using Stata version 11.
Description of studies
Results of the search
The literature searches retrieved 4508 references before de-duplication, and 3728 unique references after de-duplication. Fifty-seven were ordered as full paper copies and considered for inclusion. Thirteen trials (reported in 38 papers) were included and 19 were excluded. The process of study selection for the review is displayed in Figure 1.
|Figure 1. Figure 1: Process of study selection for the review|
In total we identified 13 randomised controlled trials with at least six-months follow-up that compared HT to placebo or a no treatment control published between 1998 and 2007 (EAGAR 2006; EPAT 2001; EPHT 2006; ERA 2000; ESPRIT 2002; EVTET 2000; HALL 1998; HERS I 1998; WAVE 2002; WEST 2001; WHI I 2002; WHI II 2004 and WISDOM 2007). Nine of the identified trials had been included in the previous review, and one trial also originally included in the previous review (HERS II) was excluded from this up-date. This trial was the long-term open label follow-up phase of HERS I 1998, and therefore not included as a separate trial, as done in the original review. Additonally, data from the single therapy oestrogen alone trial arm from the three-armed trial ERA 2000 was also included in the analyses. This had been excluded from the previous review, with only data from the combination arm being included in the analyses. Four new trials were therefore identified for the up-date review (EAGAR 2006; EPHT 2006; WHI II 2004; WISDOM 2007).
The 13 trials included 38,171 post-menopausal women; 19,302 randomised to receive some form of HT and 18,869 to receive either placebo or a no treatment control. WISDOM 2007 also included a further 1307 women who were randomised to a comparison of two active HT therapies, and EPHT 2006, also included 1001 women who were randomised to either open label HT, or a no treatment control, to examine the effect of blinding upon recruitment and retention rates within the trial. The data from these further 2306 women randomised into either of these trials (EPHT 2006; WISDOM 2007) are not included in this review.
The trials varied dramatically in size, ranging from 40 (HALL 1998) to 16,608 (WHI I 2002). Likewise, there was large variation in the length of follow-up within the trials ranging from 11.9 months (WISDOM 2007) to 7.1 years (WHI II 2004). Overall, three large trials (HERS I 1998; WHI I 2002; WHI II 2004) with a mean follow-up duration of 5.6 years (range: 4.1 – 7.1) randomised 30,110 women to either HT treatment or placebo, and therefore contributed approximately 79% of the data available from the 13 trials.
The majority of the trials (n = 7) had been conducted in the USA, two were international (one in USA and Canada, and one in England, New Zealand and Australia), with one trial conducted in each of the following countries: England and Wales, Norway, Sweden, and Estonia.
Six trials were stopped early (EAGAR 2006; EPHT 2006; EVTET 2000; WHI I 2002; WHI II 2004; WISDOM 2007) either as other trial results were published showing no beneficial effect, or a detrimental effect of HT on CVD outcomes, (EAGAR 2006; EPHT 2006; EVTET 2000; WISDOM 2007) or due to it being established that the overall risks (adverse events) associated with HT use were unlikely to be outweighed by any potential benefits of HT use on CVD outcomes within the time frame of the trial (WHI I 2002; WHI II 2004).
A summary of the main characteristics of the included trials in displayed in Table 1.
All the trials included post-menopausal women, whether menses was natural or an artefact of hysterectomy or oophorectomy, with a mean age of 63.2 years (range: 42 - 91 years). In 11 out of the 13 trials the mean participant age was over 60 years at baseline. The hysterectomy status of the women in three of the trials was related to the inclusion criteria and therefore in both HERS I 1998 and WHI I 2002 was 0%, and in WHI II 2004 100%. In the other five trials reporting baseline hysterectomy status this ranged from 10% - 61% (EPAT 2001; EPHT 2006; ERA 2000; ESPRIT 2002; WEST 2001)
The trial inclusion criteria varied according to the primary study objectives. Five of the trials were designed to assess the effects of HT in the primary prevention of CVD, and therefore enrolled predominantly healthy patient populations (EPAT 2001; EPHT 2006; WHI I 2002; WHI II 2004; WISDOM 2007). Whilst eight of the trials aimed to assess the impact of HT in secondary prevention, and therefore enrolled women with established CVD (ERA 2000; HERS I 1998; WAVE 2002) or after a designated specific CVD event of interest, namely coronary artery by-pass graft (CABG) (EAGAR 2006), angina (HALL 1998) myocardial infarction (MI) or transient ischaemic attack (TIA) (ESPRIT 2002; WEST 2001), or pulmonary embolism (PE) or deep vein thrombosis (DVT) (EVTET 2000).
Primary prevention trials
Five studies enrolled relatively healthy women (EPAT 2001; EPHT 2006; WHI I 2002; WHI II 2004; WISDOM 2007). Although one of the studies enrolled women with one CVD risk factor, namely hypercholesterolaemia (EPAT 2001) and a small minority (approximately ≤ 5%) of women within all trials had a history of CVD, the trial participants were representative of population samples of fit women in this age group without overt disease. Four of these trials (EPHT 2006; WHI I 2002; WHI II 2004; WISDOM 2007) assessed the impact of HT on both CVD, as well as a wide range of other endpoints, including cancer, osteoporosis and gallbladder disease, and therefore reported detailed lists of participant inclusion and exclusion criteria.
The two biggest primary prevention trials (WHI I 2002 and WHI II 2004) both set enrolment targets to establish set fractions for baseline age categories and to achieve racial and ethnic group representation within participant groups in the proportions recorded in the USA census for the 50 - 79 year old age group. This was achieved, with it being noted that baseline cardiovascular risk factors in the trial participants in both WHI I 2002 and WHI II 2004 were low and consistent with those observed in a generally healthy population of post-menopausal women (Manson 2003; Stefanick 2003). WISDOM 2007 recruited women with no major health problems from general practice registers in England, Australia and New Zealand, whilst EPHT 2006 included healthy women with no major health problems drawn from population samples in Estonia. In both trials participant baseline cardiovascular risk factors were low and consistent with those observed in the general population of postmenopausal women within this age group.
Secondary prevention trials
Eight studies included women with established CVD (EAGAR 2006; ERA 2000; ESPRIT 2002; EVTET 2000; HALL 1998; HERS I 19988; WAVE 2002; WEST 2001). Both ERA 2000 and WAVE 2002 included women who had coronary artery stenosis evidenced by angiogram. HERS I 1998 and EAGAR 2006 both included women who had undergone a revascularization procedure [CABG or percutaneous coronary intervention (PCI)], whilst ESPRIT 2002 and WEST 2001 included women who had had a previous MI or TIA. HALL 1998 included women previously hospitalised with angina, and EVTET 2000 included women who had suffered a thrombo-embolic event, PE or DVT.
The largest of the eight trials (HERS I 1998) compared the baseline characteristics of the trial participants with a similar group of women presumed to have coronary heart disease who were participants in a survey designed to produce nationally representative data. The HERS I 1998 participants had significantly fewer smokers, women with hypertension and diabetics than the comparison group but were comparable with respect to blood pressure, body mass index, physical activity and cholesterol levels (Grady 1998).
A number of different oestrogen alone or oestrogen and progestogen combinations had been assessed in the different trials. One trial (ERA 2000) was a three armed trial, and therefore assessed both oestrogen alone and in combination with a progestogen versus placebo. Most of the included comparisons used a moderate does of oestrogen, for example, oestradiol 1 mg or conjugated equine oestrogen (CEE) 0.625 mg daily. The following interventions assessed were:
2 mg oestradiol valerate (ESPRIT 2002).
Combined HT regimes
Combined HT regimens included one of the above types of oestrogen in combination with one of the two progestogens:
- medroxyprogesterone acetate (MPA)norethisterone
The continuous combined regimens were composed of the following
CEE 0.625 mg with MPA 2.5 mg daily (EPHT 2006; ERA 2000; HERS I 1998; WAVE 2002; WHI I 2002; WISDOM 2007).
Oestradiol 2 mg with 1 mg norethisterone daily (EVTET 2000).
Whist the combined sequential regimes included:
Oestradiol 1 mg daily with MPA 5 mg for 12 days once a year (WEST 2001).
CEE 0.625 mg for 18 days followed by a combination with oral 5 mg MPA (HALL 1998).
The control arm in each of the trials received placebo tablets.
The duration of HT use varied widely across the trials, with follow-up duration ranging from 11.9 months (WISDOM 2007) to 7.1 years (WHI II 2004). Three trials reported outcomes after HT use for around one-year (EVTET 2000; HALL 1998; WISDOM 2007); two for 2-3 years (EPAT 2001; ESPRIT 2002), and five for approximately 3 years (EAGAR 2006; ERA 2000; EPHT 2006; WAVE 2002; WEST 2001). HERS I 1998 measured outcomes after 4.1 years, and continued the study unblinded for a further 2.7 years follow-up (HERS II) (Grady 2002). Both the WHI I 2002 and WHI II 2004 trials were planned to continue for 8.5 years, but both trials were terminated early. Outcomes in WHI I 2002 were reported at 5.2 years and subsequently for a further 4 months of follow-up (total follow-up 5.6 years) for primary and selected secondary outcome measures. WHI II 2004 reported outcomes at 6.8 years and for a subsequent further three-months of follow-up (7.1-years) for primary and selected secondary outcomes, with a median time of 5.9 and 5.8 years on treatment for the HT and placebo groups respectively.
The outcomes assessed in the individual trials varied according to the trial objectives. One primary prevention trial (EPAT 2001) and three secondary prevention trials (ERA 2000; ESPRIT 2002; WAVE 2002) aimed to assess the effects of HT upon intermediate outcomes, namely carotid artery intima-media thickness, and the impact on coronary atherosclerosis as measured by the minimal lumen diameter of the arteries respectively. However, all four trials also reported one or more of the clinical outcomes of interest as secondary outcomes and therefore were included in the analyses. The primary aim in the largest two trials, WHI I 2002 and WHI II 2004 was to assess the potential cardio-protective effect of HT in relatively healthy post-menopausal women, and therefore both trials reported cardiovascular clinical endpoints as the primary outcome. Invasive breast cancer was the designated primary adverse outcome in both trials, with the incidence of other cancers, fractures, gallbladder disease and death reported as secondary outcomes. Two further primary prevention trials, EPHT 2006 and WISDOM 2007 also measured similar outcomes, with CVD outcomes designated as the primary ones of interest. The remaining five secondary prevention trials aimed to examine the effects of HT in women with already established clinical disease, with the primary outcome designated according to the underlying patient pathology. Their primary outcomes were myocardial infarction or death (ESPRIT 2002; HERS I 1998), thrombo-embolism (EVTET 2000), stroke (WEST 2001), and angina (HALL 1998).
Five out of the 13 trials (EPHT 2006; HERS I 1998; WHI I 2002; WHI II 2004; WISDOM 2007) additionally reported HRQoL outcomes obtained using one or more validated measures. These outcomes focused on overall health or functional status, and the specific domains of energy/fatigue, depressive symptoms, sleep disturbance, sexual satisfaction, and psychological well being. Outcomes were reported at baseline and one- (WHI I 2002; WHI II 2004; WISDOM 2007) and three-year follow-up (HERS I 1998) in four of the trials. Whilst in EPHT 2006 no baseline scores were reported, with only follow-up scores at two- and 3.6 years presented
All 13 trials reported the funding source. Only one of the trials, HERS I 1998 was exclusively funded by the Pharmaceutical Industry (Wyeth-Ayerst), whilst EVTET 2000 was part funded by a grant from Novo-Nordisk Pharmaceutical. The study medication for ERA 2000 and WHI I 2002 and WHI II 2004 was provided by Wyeth-Ayerst Research, and for ESPRIT 2002 and WEST 2001, Schering AG and Mead Johnson laboratories respectively.
Nineteen (as well as the HERS II study) studies were excluded. The primary reason for the exclusion were:
Fourteen studies reported no relevant outcomes of interest to this review
Two assessed a different intervention
One was not a randomised controlled trial
One was not the relevant population
One assessed a non-comparison
Risk of bias in included studies
|Figure 2. Figure 2: Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.|
|Figure 3. Figure 3: Risk of bias summary: review authors' judgements about each risk of bias item for each included study.|
The generation of randomised sequence was adequate in 11 out of the 13 trials; in all of these cases it was computer-generated. Neither EAGAR 2006 or HALL 1998 reported the methods used to generate random allocation, and therefore it is unclear as to whether the method used was satisfactory. Ten trials described a satisfactory method of allocation concealment: in these trials allocation to treatment was either generated by the computer once information about an eligible participant had been entered, or was completed by remote contact between the recruiting centre and the study co-ordinating centre or pharmacy. One of these ten trials, EPHT 2006 randomised women who expressed an interest in participating, but did not open the randomisation envelope until their eligibility had been checked and they had consented. Three of the trials (EAGAR 2006; EVTET 2000; HALL 1998) did not report methods of allocation concealment.
All the trials except HALL 1998 described themselves as double blind. Ten of the trials explicitly stated that all participants, clinical staff and outcome assessors were blinded to treatment allocation, and all 13 trials reported ‘hard’ outcomes; the verification of which is unlikely to be effected by blinding. Unblinding of participants occurred in 331 women initially randomised into the active single HT treatment arm in WHI II 2004, whom after a protocol change were unblinded, and changed arms into the WHI I 2002 combined therapy arm. Eight of the trials additionally described an unblinding mechanism to be used in the management of adverse events (ERA 2000; ESPRIT 2002; WAVE 2002; WEST 2001; WHI I 2002; WHI II 2004; WISDOM 2007).
Incomplete outcome data
Twelve of the trials analysed all participants on an intention-to-treat basis at least for the outcomes of interest in the present review, whilst data in WAVE 2002 were analysed on an ITT basis for over 97% of participants. Drop-out rates (medication non-compliance) were generally high, particularly in the active treatment groups, and tended to increase over time. In the 11 trials that reported data on adherence, these ranged from greater than 90% compliance rates in EPAT 2001 at two-year follow-up, to less than 40% compliance in EPHT 2006 at four-year follow-up. In the two WHI trials with the longest follow-up, 42% of the active treatment group and 38% of the placebo group were no longer taking their allocated treatment at 5.2 years in WHI I 2002, and 10.7% of the placebo group had initiated active HT treatment outside of the trial. Whilst in WHI II 2004 53% of participants overall were no longer taking their allocated treatment at 6.8 years and a further 5.7% had initiated hormone use outside the study. A summary of medication compliance within the trials is given in Table 2.
Losses to follow up were low in most of the trials, with no women lost to follow-up in six trials (EPAT 2001; ERA 2000; ESPRIT 2002; EVTET 2000; HALL 1998; WEST 2001) and between 0.1% to 5.2% lost in five other trials(EPHT 2006; HERS I 1998; WAVE 2002; WHI I 2002; WHI II 2004; WISDOM 2007).
Only one trial HALL 1998 may have been subject to selective reporting. All the other 11 trials reported all expected outcomes.
Effects of interventions
Results are reported below. In most cases details of effect measures are only reported in the text where results are statistically significant. It was not possible to conduct any analyses stratified by the participants mean age at baseline (> 60 versus < 60 years of age), as only two trials (EPHT 2006; EVTET 2000) included participants with a mean age < 60 years at baseline.
Estrogen alone versus placebo in primary prevention
This comparison was reported in two trials (EPAT 2001; WHI II 2004) with a total of 10,961 participants. No protective effects for HT on all cause mortality, or any CVD outcomes (death, non-fatal MI or angina), the number of angioplasties, or PE was observed ( Analysis 1.1; Analysis 1.2; Analysis 1.3; Analysis 1.5; Analysis 1.7; Analysis 1.8). However, at 7.1 year follow-up HT use was associated with a increased risk of stroke, RR 1.35 (95% CI 1.08 to 1.70) ( Analysis 1.4), and a borderline significant increase in the number of venous thromboembolism, RR 1.32 (95% CI 1.00 to 1.74) relative to placebo ( Analysis 1.6). The associated number needed-to-harm (NNH) was 121 for stroke and 197 for venous thromboembolism respectively. Due to the limited evidence for this outcome no analyses stratified by time on treatment were conducted. However, the WHI II 2004 authors noted the excess risk in the intervention arm was due to an increased risk of ischaemic rather than haemorrhagic stroke which become apparent after four years of follow-up (Hendrix 2006). There was no significant statistical heterogeneity between the two studies for any outcome.
Combination HT versus placebo in primary prevention
Combination HT versus placebo in primary prevention was assessed in three trials, EPHT 2006, WHI I 2002 and WISDOM 2007, with a total of 21,770 participants. Results showed no significant impact on all cause mortality, CVD mortality, or angina ( Analysis 2.1; Analysis 2.2; Analysis 2.5; Analysis 2.8). There was an increased risk of non-fatal MI ( Analysis 2.3) and stroke ( Analysis 2.4). The overall RR for non-fatal MI of 1.38 (95% CI: 1.06 to 1.78) had an associated NNH of 295, whilst the RR for stroke of 1.31 (95% CI 1.03 to 1.68) had a NNT of 231. Analysis stratified by time on treatment for stroke, indicated the excess risk was apparent after three-years on treatment and remained significant with further follow-up time until 5.6 years ( Analysis 2.9). There was moderate statistical heterogeneity between the trials (I
The RR for venous thromboembolism and PE also increased compared to placebo: 2.29 (95% CI 1.76 to 2.97) ( Analysis 2.6) and 2.29 (95% CI 1.59 to 3.31) ( Analysis 2.7) respectively. The NNH for the outcome of venous thromboembolism was 100, and the NNH for PE 197. Analysis by time on treatment for venous thromboembolism indicated excess risk at each follow-up time. This was highest early on treatment (one-year follow-up) and diminished with time, but remained significantly higher at each time point ( Analysis 2.10). Significant statistical heterogeneity between trials (WHI I 2002 and WISDOM 2007) with an I
Estrogen alone versus placebo in secondary prevention
Four trials with a total of 1917 participants assessed oestrogen alone versus placebo in women with established CVD disease (EAGAR 2006; ERA 2000; ESPRIT 2002; WEST 2001). No significant differences between treatment arms were observed for mortality (all cause or CVD), any CVD outcomes (non-fatal MI, stroke, angina), or either venous thromboembolism or PE ( Analysis 3.1; Analysis 3.2; Analysis 3.3; Analysis 3.4; Analysis 3.5; Analysis 3.6; Analysis 3.7). Significantly more patients on HT underwent an angioplasty, with a RR of 8.60 (95% CI 1.13 to 65.73), but this result was based on the results of one small trial (EAGAR 2006; n = 83) with a low number of event rates (n = 9). There was no statistical heterogeneity between trials for any of the outcomes.
Combination HT versus placebo in secondary prevention
Combination HT versus placebo in secondary prevention was assessed in five trials with a total of 3523 participants (ERA 2000; HALL 1998; HERS I 1998; EVTET 2000; WAVE 2002). No protective effects for combination therapy were observed for mortality (all cause or CVD), any CVD outcomes (non-fatal MI, stroke, angina), or the number of CABG or angioplasty procedures ( Analysis 4.1; Analysis 4.2; Analysis 4.3; Analysis 4.4; Analysis 4.5; Analysis 4.8; Analysis 4.9). Significantly more venous thromboembolic events (including PE) occurred in the HT trial arms, with a RR of 2.59 (95% CI 1.51 to 4.42) for all venous thromboembolic events and a RR of 3.77 (95% CI 1.41 to 10.06) for PE alone ( Analysis 4.6; Analysis 4.7). The associated NNH for venous thromboembolism and PE were 60 and 104 respectively. Analysis by time on treatment for venous thromboembolism indicated excess risk at each follow-up time. This was highest early on treatment (one-year follow-up) which diminished but remained significantly higher compared to placebo at each follow-up time until 6.8 years ( Analysis 4.12). There was no significant heterogeneity between the trials for any of the outcomes.
Single and combination HT (oestrogen plus progestogen) in both primary and secondary prevention
Consistent with effects observed in both primary and secondary prevention, HT overall had no protective effects for death, or any of the CVD outcomes, or CVD related surgical procedures compared to placebo ( Analysis 5.1; Analysis 5.2; Analysis 5.3; Analysis 5.8; Analysis 5.9). Again, an increased risk of stroke, RR 1.26 (95% CI 1.11 to 1.43), venous thromboembolism RR 1.89 (95% CI 1.58 to 2.26) and PE RR 1.84 (95% CI 1.42 to 2.37) relative to placebo was observed ( Analysis 5.4; Analysis 5.6; Analysis 5.7). The associated NNH were 164, 109 and 243 for stroke, venous thromboembolism and PE respectively.
Analysis for stroke by time on treatment indicated that excess risk was evident from approximately three-year follow-up, and remained higher until the longest follow-up time of 7.1-years ( Analysis 5.13). Whilst the analysis for time on treatment for venous thromboembolism evidenced an excess risk at all follow-up points (1 - 7.1 years), which was highest at one-year follow-up and attenuated with longer follow-up, but remained significantly higher relative to placebo ( Analysis 5.15).
Significant statistical heterogeneity was present between trials for the outcome of venous thromboembolism (I
No evidence of reporting bias was evident on examination of funnel plots (Figure 4)
|Figure 4. Funnel plot of comparison: HT versus placebo in primary and secondary prevention, outcome: 5.1 Death (all causes).|
None of the three predictor variables of length of trial follow-up, whether treatment was single (i.e.) oestrogen alone or in combination with progestogen, and whether the patient population were being treated for primary or secondary prevention of CVD were statistically significant predictors for the outcomes of (i) all cause mortality, (ii) CVD mortality, (iii) non-fatal MI, (iv) stroke, (v) angina, (vi) venous thromboembolism or (vii) PE in the trials. This was related to the fact that there was no overall substantive statistical heterogeneity in terms of the results between the trials, and stratified analyses for both stroke and venous thromboembolism indicated that effects did not vary linearly as a function of treatment time, but were likely to be curvlinear. The results of the meta-regression are presented in Table 3.
Health-related quality of life
Five out of the 13 trials (EPHT 2006; HERS I 1998; WHI I 2002; WHI II 2004; WISDOM 2007) reported validated HRQoL measures. The results are displayed in Table 4 and Table 5 . A number of different measures had been used across the trials, both generic and condition specific. HERS I 1998 (Hlatky 2002) measured physical functioning at baseline, years one, two and three-year follow-up using the 12-item Duke Activity Status Index (Hlatky 1989), energy/fatigue using a four-item RAND scale (Ware 1992), mental health using the Rand Mental Health Inventory (Stewart 1988), and depressive symptoms using an eight-item scale by Burnam and colleagues developed to screen for depression in the National Study of Medical Outcomes (Burnam 1988). Only very limited data from each scale were reported graphically, and therefore only the composite results from baseline and three-year follow-up are presented. HRQoL / functional status was assessed using the RAND 36-Item Health Survey (RAND 36) (Ware 1992) at baseline and one-year follow-up in both WHI I 2002 and WHI II 2004; depressive symptoms were assessed using the same 8-item scale by Burnam 1988 used in HERS I 1998; sleep quality was assessed using a five-item Women’s Health Initiative Insomnia Rating Scale (developed and validated for use in the WHI; Levine 2003) and sexual satisfaction (either with your current partner or alone) was assessed by a single item with a four-point response scale ranging from one (worst) to four (best). In WISDOM 2007 general HRQoL and psychological well being were assessed at baseline and 11.9 month follow-up using a modified version of the Women’s Health questionnaire (Hunter 1992), emotional and physical menopausal symptoms using a trial specific 28 item symptoms questionnaire, depression using the Centre for Epidemiological Studies Questionnaire (CES-D) (Radloff 1977), and generic HRQoL using the EQ-5D (EuroQol 1990) (Kind 2003). EPHT 2006 assessed generic HRQoL also using the EQ-5D at both two- and 3.6-year follow-up. However, no results were reported at baseline, so interpreting any changes in scores longitudinally is problematic.
Given the disparity and wide variation in the HRQoL outcomes used, pooling outcomes across the studies was deemed inappropriate. The HRQoL results at baseline and three-year follow-up for HERS I 1998, baseline and one-year follow-up for WHI I 2002 and WHI II 2004, WISDOM 2007 and EPHT 2006 with the between group differences are therefore presented in Table 4 and Table 5.
Overall, from HERS I 1998 (Hlatky 2002) there was no evidence that HT had any statistically significant impact on any of the four domains of HRQoL assessed (energy/fatigue, physical functioning, mental health or depressive symptoms) compared to placebo from baseline over the three-year follow-up period. Energy/fatigue, physical functioning and mental health remained relatively stable from baseline until year three in both groups. However, a minor statistically significant decline in depressive symptoms was observed in both groups across the trial period. This did not differ between the two treatment groups. Results from WHI I 2002 (Hays 2003) showed that women randomised to HT had a statistically significantly better level of functioning on four out of the eleven domains of functioning assessed, namely sleep disturbance, bodily pain, physical functioning, and related to this, role limitation due to physical problems compared to those on placebo at one-year follow-up. However, when the increment of change between the two groups was compared, the differences in effect size between the two groups was small, and therefore whilst statistically significant was not likely to be clinically meaningful. Likewise, the results from the WHI II 2004 (Brunner 2005) indicated that HT (oestrogen alone) had a statistically significant positive impact on some limited areas of functioning, namely physical functioning and sleep disturbance, but appeared to have a detrimental effect on vitality compared to placebo. Again, a statistically significant positive impact for HT use on overall HRQoL was observed for only two of the 11 domains of functioning assessed, and effect sizes were small. Results from WISDOM 2007 (Welton 2008) indicate that combination HT had a statistically significant positive impact on some areas of functioning across the 11.9 month follow-up period. Most notably, as expected, these were related to menopausal symptoms such as hot flushes, night sweats, vaginal problems (dryness and discharge), breast tenderness and bloating. However, this statistically significant effect was only observed in seven of the 28 areas of symptom related problems assessed, and may be of somewhat limited clinical impact, as no differences between the two treatment groups was observed on either the EQ-5D VAS or questionnaire which assess health impact more broadly on a wider set of functioning domains. Likewise, no significant differences in the distribution of EQ-5D scores were observed between women on HT and placebo in EPHT 2006 at either two- or 3.6-year follow-up. All scores were highly positively skewed, indicating high levels of functioning and little if any impairment, with half of the women in both trials arms having a EQ-5D score of 0.90 at the end of the second year of the trial, and 0.80 at the end of the trial.
Summary of main results
In the overall trial populations there is no evidence that HT has a role in either the prevention or the treatment of CVD. Treatment with HT had no significant impact on either overall death rates, CVD related death, non-fatal MI, angina, or the number of patients undergoing revascularization procedures. On the contrary it is associated with an increased risk of stroke, venous thromboembolism and pulmonary embolism, and combination HT in primary prevention also increased the risk of non-fatal MI. This increased risk of non-fatal MI was not observed in secondary prevention combination HT trials, and it is unclear why differential effects are observed for this outcome between these patient populations particularly given that both HERS I 1998 and WHI I 2002 used the same combination HT preparation (continuous combination CEE with MPA). In contrast to combination HT, oestrogen only HT does not appear to have any statistically significant negative effect on coronary disease.
The excess risk of coronary events in women in the HT group was observed in the first year of treatment in women taking combination HT in both HERS I 1998 and WHI I 2002. Although there was a significant trend in both WHI I 2002 and the blinded phase of HERS I 1998 for CVD risk in the HT group to diminish with time on treatment, subsequent analysis of HERS I 1998 data (including both the blinded and non-blinded follow-up phase) indicated no statistically significant variation in risk over time. The WHI I 2002 investigators suggest the apparent decline in CVD risk in later years may be due to an acceleration of events in earlier years among susceptible women in the HT group, and highlight that the trend towards a decreasing CVD risk over time with combination therapy should be interpreted with caution (Manson 2003). Results from our analyses by time on treatment for both CVD death and non-fatal MI broadly agree with the WHI I 2002 and HERS I 1998 findings. Excess risk (although not significant) was highest for both outcomes in the first year on treatment and then gradually declined. Both WHI I 2002 and WHI II 2004 undertook pre-specified sub-group analyses to evaluate whether any clinical characteristics of the trial populations may potentially moderate the effects of HT. The potential predictor variables examined included: age, time since menopause, presence or absence of vasomotor symptoms, prior hormone use, CHD risk factor status and presence or absence of pre-existing CVD (Hsia 2006; Manson 2003). None of these variables significantly effected results, although a non-significant trend for a reduction in CHD risk for women who initiated HT use within ten-years of menopause was observed.
The significant excess risk of stroke in our analyses was observed in both primary prevention analyses (i.e.) those randomised to either oestrogen alone or oestrogen in combination with progestogen compared to placebo. These findings are based on the two largest trials, WHI I 2002 and WHI II 2004 with follow-up of 5.6 and 7.1 years respectively. Whilst, no significant excess risk was observed in any of the secondary prevention trials, including the largest trial HERS I 1998, it is probable that the results from the primary prevention trials are applicable to secondary prevention populations, and that sub-group analyses of these trials were underpowered due to small trial sizes, low event rates and shorter length of follow-up to detect any statistically significant differences in stroke rates between HT and placebo treatment arms. In both WHI I 2002 and WHI II 2004 the excess risk of stroke observed with HT use was driven by an excess of ischaemic rather than haemorrhagic stroke, with 79.8% and 80.3% of strokes respectively observed within the trials being ischaemic (Hendrix 2006; Wassertheil-Smoller 2003). In our analyses increased risk of stroke was apparent after three-years on treatment in women taking combination HT, and after four-years for women randomised to oestrogen alone (Hendrix 2006). In both trials the hazard ratios for ischaemic stroke did not differ significantly in sub-groups based on age, years since menopause, prior CVD, hypertension status or diabetes mellitus, body mass index, or statin or aspirin use at baseline (Hendrix 2006; Wassertheil-Smoller 2003).
The finding of a significant increase in risk for both venous thromboembolism and PE within the overall trial populations appears in our analyses to be driven largely by the excess risk observed in both primary and secondary prevention combination HT trials. Estrogen alone use in primary prevention was associated with a marginally significantly increased risk of venous thromboembolism , and there was no significant excess risk associated with oestrogen alone use in secondary prevention; although it should be noted that in both analyses the point estimates favoured treatment with placebo, and therefore these results should be interpreted with some degree of caution. The risk of venous thromboembolism and PE in both primary and secondary prevention trials with combination HT indicated a more than two-fold risk increase of venous thromboembolism and PE on HT relative to placebo. Analyses by time on treatment showed that for both primary and secondary prevention, the risk was highest close to the initiation of treatment, and attenuated with time, but remained significantly higher on HT compared to placebo. Both WHI I 2002 and WHI II 2004 undertook further pre-specified subgroup analyses to evaluate the association between participant baseline characteristics and venous thromboembolism and PE risk. Not surprisingly, given the fact that no excess risk was observed within the trial WHI II 2004 investigators found no significant interactions between oestrogen alone use and age, body mass index, or most other venous thromboembolism risk factors. The authors did note however, the hazard ratios for combination therapy in WHI II 2004 were significantly higher than those for oestrogen alone even after adjusting for venous thromboembolism risk factors (Curb 2006). In WHI I 2002, increasing age, overweight and obesity, and having a factor V Leiden mutation (a blood coagulation disorder) were associated with a higher risk of venous thromboembolism compared to placebo (Cushman 2004).
Overall completeness and applicability of evidence
There are a number of limitations to the evidence base reviewed. Firstly, it should be highlighted that the results are based on those obtained in 13 RCTs, with the majority of statistically significant findings derived from the results of the three largest trials, HERS I 1998, WHI I 2002 and WHI II 2004 which dominated the results. These three trials all evaluated oral CEE 0.625 mg, with or without continuous methoxy progesterone (MPA 2.5 mg). Other trials evaluating different types of HT tended to be much smaller with a shorter duration of follow-up, and reported few if any major clinical events. There is some debate regarding the external validity of the findings of WHI I 2002 and WHI II 2004, and the degree to which they apply to any type of HT other than continuous combined oral CEE 0.624 mg with or without MPA 2.5 mg. The effects of HT may vary with different estrogens and progestogens, different doses, and routes of administration. However, in order to statistically pool the results of different studies we had to make assumptions regarding a ‘class effect’ of HT, which may not be warranted.
It was not possible to stratify any analyses according to the mean baseline age of trial participants in order to assess the impact of time since menopause on CVD outcomes. Only two of the trials (EPHT 2006; EVTET 2000) included participants with a mean baseline age of less than 60 years, and only WHI I 2002 and WHI II 2004 reported additional sub-group analyses to assess the relationship between both participant age and time since menopause on outcome. It therefore has not been possible within the review to assess any potential impact of the timing of HT treatment in relation to the time of menopause, and therefore contribute to the current debate regarding the 'timing hypothesis'.
The clinical outcomes of interest in the review were secondary outcomes in four of the trials (EPAT 2001; ERA 2000; ESPRIT 2002 and WAVE 2002). It can therefore be postulated that these trials may not have been sufficiently powered in order to detect differences in clinical treatment effects between the HT and placebo arms, as this was not the primary aim of the trial. Furthermore, as previously highlighted six of the trials were stopped early (EAGAR 2006; EPHT 2006; EVTET 2000; WISDOM 2007; WHI I 2002; WHI II 2004) either as other trial results were published showing no beneficial effects on CVD outcomes for HT relative to placebo, or observation of a detrimental effect either on CVD outcomes or adverse events was shown. The mean length of trial follow-up therefore ranged considerably from 11.9 months to 7.1 years, with a mean duration of follow-up of three years across the trials. The early stopping of the trials has implications both for the power to detect differences in treatment effects between the HT and placebo arms, as the sample size will have been predicated based on the original proposed length of follow-up and assumptions regarding the number of events observed, and also limits the availability of evidence on the longer term treatment effects of HT compared to placebo. A further limitation of the evidence base reviewed relates to the impact of patient medication compliance, which ranged dramatically between the trials. A high proportion of women in the trials did not receive the treatment to which they were randomised. Overall, the number of women who discontinued their medication or took less than 80% was disproportionately high in the HT trial arms, presumably due to medication side effects. The authors of WHI I 2002 noted that if discontinuation of treatment and initiation of non-study treatment occurred independently of risk factors for clinical outcomes their intention-to-treat analysis underestimates both the harms and benefits of HT among women who adhere to treatment.
Quality of the evidence
Overall study quality was high (Figure 3).
Potential biases in the review process
There are a number of potential biases in the review process, although attempts have been made to limit these. The bias of most concern is that of patient-selection bias which limits external validity. Nearly all the included trials had a mean participant age of over 60-years at baseline, and none focused on women who were either peri-menopausal or around the time of the menopause. Whilst these inclusion criteria reflected the aims of the trials, it does not reflect usual clinical practice, in which HT is prescribed for the relief of vasomotor symptoms at the time of menopause. This also limited the analyses that could be undertaken, as it was not possible to stratify trials according to the participants mean age at baseline to assess the potential impact of time since menopause on CVD outcomes.
Despite of extensive searches it is possible that we failed to identify all relevant studies. However, given the dominance of WHI I 2002 and WHI II 2004 on the results of the review, it is unlikely that we missed any trials large enough to impact substantially on the results. Additionally, as already indicated, assumptions had to be made in the analyses regarding the effects of different HT preparations in order to undertake meta-analyses. These assumptions may not be warranted, as it is as yet unclear how different preparations and doses may differ.
Agreements and disagreements with other studies or reviews
Magliano 2006 pooled results from seven of the trials included in the current review (ERA 2000, ESPRIT 2002, HERS I 1998, WAVE 2002, WEST 2001, WHI I 2002; WHI II 2004), and concluded that there was no impact of HT compared to placebo on total mortality or non-fatal MI, but a statistically significant increase of 29% in the number of strokes observed with HT use. Likewise, a meta-analysis by Bath 2005 pooled 28 RCTs that reported stroke events. HT was associated with a statistically significant increase in the risk of stroke, particularly ischaemic stroke. Furthermore for those participants who had a stroke the HT groups appeared to have a worse outcome. However, it is unclear to what degree the results of this review are applicable to post-menopausal women, as the review had very broad inclusion criteria, pooled a wide range of trials which used different types of HT for a range of indications, some of which included male participants.
Salpeter 2006 in a second meta-analyses, aimed to examine the effect of HT on coronory heart disease events in younger and older post-menopausal women (defined as participants with a mean time from menopause of less than or greater than ten years, or mean age less than or greater than 60 years). The analyses of 23 trials (ten trials with younger women and 11 trials with older women), included the relevant WHI age-specific sub-group data in one or the other group as though they had originated from separate RCTs. The results showed that HT significantly reduced coronay heart disease events in younger women, but not in older women. Given the wide variety of different HT preparations in the 23 trials included in the analyses, the differing patient populations, and the methods of analyses it is unclear how applicable the results of this review are to the populations studied in the present review.
Implications for practice
Treatment with HT in post-menopausal women for either primary or secondary prevention of CVD events is not effective, and causes an increase in the risk of stroke, or venous thromboembolic events. Furthermore, combination HT in primary prevention is also associated with an increased risk of non-fatal MI.
Implications for research
No trials were identified that have assessed the efficacy and safety of HT for either perimenopausal women or those seeking relief from menopausal symptoms. Currently there is a lack of evidence regarding factors that may modulate the risks involved in HT treatment, such as different oestrogen and progestogen preparations, different time-frames for the use of HT, and different doses and routes of administration (for example, skin patches and creams). The results of both ELITE 2004 (NCT00114517) and KEEPS 2005 (NCT00154180) should lay the foundation for future research in this area. There is an additional need for research on the efficacy and safety of alternative methods for the relief of menopausal symptoms for women who may wish to avoid its use.
We would like to acknowledge the contributions by Dr Rod Taylor, Pensinsular Medical School for undertaking the meta-regression analysis and Dr Philippa Davies, School of Social and Community Medicine, University of Bristol, and Ms Zulian Liu, Aggressive Research Intelligence Facility (ARIF), Unit of Public Health, Epidemiology and Biostatistics, Univeristy of Birmingham, for screening the titles and abstracts of the up-dated searches completed between 2002 and 2008.
Data and analyses
- Top of page
- Summary of findings [Explanations]
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Index terms
Appendix 1. Search strategies
#20 (HYPERLIPIDEMIA or HYPERLIPIDAEMIA)
#22 ((((((((#1 or #2) or #3) or #4) or #5) or #6) or #7) or #8) or #9)
#23 ((((((((((#10 or #11 or #12) or #13) or #14) or #15) or #16) or #17) or #18) or #19) or #20) or #21)
#24 (#22 or #23)
#27 (HORMONE near REPLAC*)
#28 (OESTROGEN near REPLAC*)
#29 (ESTROGEN near REPLAC*)
#30 ((MENOPAUS* or POSTMENOPAUS*) or POSTMENOPAUS*)
#33 (#31 or #32)
#34 (#30 and #33)
#35 ((((#25 or #26) or #27) or #28) or #29)
#36 (#34 or #35)
#37 (#24 and #36)
Cochrane Controlled Trial Register, Issue 1, April 2010 (search date: 20/04/2010)
#1. MeSH descriptor Cardiovascular Diseases explode all trees
#2. MeSH descriptor Cerebrovascular Disorders explode all trees
#7. HEART NEAR/3 ATTACK
#9. MeSH descriptor Embolism and Thrombosis explode all trees
#13. MeSH descriptor Hypertension explode all trees
#15. MeSH descriptor Arteriosclerosis explode all trees
#16. ARTERIOSCLER* OR ARTHEROSCLER*
#17. ISCHAEMIC OR ISCHEMIC
#18. MeSH descriptor Hyperlipidemias explode all trees
#19. HYPERLIPIDEMIA* OR HYPERLIPIDAEMIA*
#20. (#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #18 OR #19)
#21. MeSH descriptor Hormone Replacement Therapy explode all trees
#22. HRT OR ERT OR ORT
#23. HORMONE NEAR/4 (REPLAC* OR THERAP* OR SUPPLEMENT*)
#24. ESTROGEN NEAR/4 (REPLAC* OR THERAP* OR SUPPLEMENT*)
#25. OESTROGEN NEAR/4 (REPLAC* OR THERAP* OR SUPPLEMENT*)
#26. (#21 OR #22 OR #23 OR #24 OR #25)
#27. (menopaus* OR postmenopaus* OR post-menopaus*)
#28. MeSH descriptor Postmenopause, this term only
#29. (#27 OR #28)
#30. oestrogen OR estrogen
#31. (#29 AND #30)
#32. (#26 OR #31)
#33. (#20 AND #32)
MEDLINE search 20/04/2010
6. (MYOCARDIAL OR HEART NEAR ATTACK).TI,AB.
15. (ARTERIOSCLERO$5 OR ARTHEROSCLERO$5).TI,AB.
16. (ISCHAEMIC OR ISCHEMIC).TI,AB.
18. (HYPERLIPIDEMIA$4 OR HYPERLIPIDAEMIA$4).TI,AB.
19. 1 OR 2 OR 3 OR 4 OR 5 OR 6 OR 7 OR 8 OR 9 OR 10
20. 11 OR 12 OR 13 OR 14 OR 15 OR 16 OR 17 OR 18
21. 19 OR 20
23. (HRT OR ERT OR ORT).TI,AB.
24. HORMONE NEAR (REPLAC$6 OR THERAP$4 OR SUPPLEMENT$6)
25. ESTROGEN NEAR (REPLAC$6 OR THERAP$4 OR SUPPLEMENT$6)
26. OESTROGEN NEAR (REPLAC$6 OR THERAP$4 OR SUPPLEMENT$6)
27. 22 OR 23 OR 24 OR 25 OR 26
28. MENOPAUS$4 OR POSTMENOPAUS$4 OR POST-MENOPAUS$4
30. 28 OR 29
31. (ESTROGEN OR OESTROGEN).TI,AB.
32. 30 AND 31
33. 27 OR 32
36. (SINGL$4 OR DOUBLE$4 OR TRIPLE$4 OR TREBLE$4) AND (BLIND$4 OR MASK$4)
37. RANDOM$5 OR PLACEBO$2
41. (CLINIC$3 NEAR TRIAL$2).TI,AB.
42. RETRACT$5 NEAR PUBLICATION
43. 34 OR 35 OR 36 OR 37 OR 38 OR 39 OR 40 OR 41 OR 42
44. ANIMAL=YES NOT HUMAN=YES
45. 21 AND 33 AND 43
46. 45 NOT 44
EMBASE search 20/04/2010
6. MYOCARDIAL.TI,AB. OR (HEART NEAR ATTACK).TI,AB.
14. (ARTERIOSCLERO$5 OR ARTHEROSCLERO$5).TI,AB.
15. (ISCHAEMIC OR ISCHEMIC).TI,AB.
16. (HYPERLIPIDEMIA$4 OR HYPERLIPIDAEMIA$4).TI,AB.
18. 1 OR 2 OR 3 OR 4 OR 5 OR 6 OR 7 OR 8 OR 9 OR 10 OR 11 OR 12 OR 13 OR 14 OR 15 OR 16 OR 17
20. (HRT OR ERT OR ORT).TI,AB.
21. HORMONE NEAR (REPLAC$6 OR THERAP$4 OR SUPPLEMENT$6).TI,AB.
22. ESTROGEN NEAR (REPLAC$6 OR THERAP$4 OR SUPPLEMENT$6).TI,AB.
23. OESTROGEN NEAR (REPLAC$6 OR THERAP$4 OR SUPPLEMENT$6).TI,AB.
24. (menopaus$4 OR postmenopaus$4 OR post-menopaus$4).TI,AB.
26. 39 OR 40
27. (oestrogen OR estrogen).TI,AB.
28. 41 AND 42
29. 19 OR 20 OR 34 OR 37 OR 38 OR 43
30. 44 AND 18
32. crossover$2 OR cross ADJ over$2
34. (RANDOM$ OR PLACEBO$).DE,TI,AB.
LILACS search conducted 20/04/2010
"HORMONE REPLACEMENT THERAPY" OR
((hormone OR oestrogen OR oestrogen) AND (replac$ or therap$ or supplement$)) or (hrt OR ert OR ort) AND ("clinical trials, RANDOMIZED" or "controlled clinical trials, RANDOMIZED" OR (( trial$ or ensa$ or estud$) AND (clin$)) OR ((singl$ or doubl$ or doble$ or duplo$ or trebl$ or trip$) AND (blind$ or cego$ or ciego$ or mask$ or mascar$)) OR (random$ or randon$ or casual$ or acaso$ or azar or aleator$)) = 318
(("POSTMENOPAUSE" OR menopaus$ or postmenopaus$ or post-menopause) AND (oestrogen or estrogen))
Question from Jim Thornton, 18 October 2013
This review does not include any of the 22 studies identified in the paper "Impact of postmenopausal hormone therapy on cardiovascular events and cancer: pooled data from clinical trials" (Hemminki E, McPherson K BMJ 1997;315:149-153).
We recognise that some of them may not have included relevant endpoints, but we were surprised not to see them in the list of excluded studies.
We also recognise that most were not initiated with the aim of cardioprotection. However since the Hemminki-McPherson paper identified cardiac events, they surely are informative to he question addressed in the present review.
Finally, since most of these studies included relatively young women they are relevant to the timing hypothesis.
Having assessed the review by Hemminki-McPherson the authors note that the methods section describes 22 studies including 4124 women, but table 1 (included studies) provides details of 23 studies on 4164 women. It is therefore unclear which of the 22 studies were included in the review. The response therefore pertains to all 23 studies listed in table 1.
As Thornton et al, may be aware, this review was an up-date of the original Cochrane review on Hormone replacement therapy for preventing cardiovascular disease in post-menopausal women (Gabriel-Sanchez 2005). All 23 of the studies identified in the Hemminki-McPherson review were listed in the excluded studies section of the original Gabriel-Sanchez 2005 review, but were excluded as they did not meet the inclusion criteria. Likewise, all 23 studies were identified in the searches for the up-date review, but were excluded on the basis of title and abstract, as they clearly did not meet the inclusion criteria specified for the up-date review. It is worth noting that the inclusion criteria for the Hemminki-McPherson review differed considerably from those for the up-date Cochrane review, in terms of study design, intervention, and outcome measures. Therefore the Hemminki-McPherson review includes studies of transdermal HT, a cross-over trial, and studies of less than 6-months duration, all of which were criteria that clearly did not meet the inclusion criteria for the present up-date review, and readers would probably not expect to see listed in the excluded studies section. In terms of outcome measures, the Hemminki-McPherson review included studies in which the primary aim was not to assess the effects of HT on CVD, and therefore only seven of the 23 studies reported any CV outcomes, with a total of 17 events identified. Data on CV events in the review “were given incidentally”, mostly as “reasons for dropping out”. As the aim of the up-dated Cochrane review was to assess the effects of HT on specific types of CV events, as well as death from CV causes, CV events needed to be specified a priori in the trial and reported as either primary or secondary outcome measures. Therefore the review did not include trials in which CV events were reported only as adverse events or reasons for withdrawals.
James Hitchin, Klim McPherson, Jim Thornton
Caroline Main, responded to the feedback.
Last assessed as up-to-date: 20 April 2010.
Protocol first published: Issue 3, 2000
Review first published: Issue 2, 2005
Contributions of authors
Tiffany Moxham developed and ran the search strategies. Study selection, quality assessment and data extraction were performed by Caroline Main and Beatrice Knight. Caroline Main wrote the first draft of the review, and all co-authors commented on this and contributed to the final draft of the review. All authors have approved the manuscript.
Declarations of interest
Sources of support
- No sources of support supplied
- New Source of support, UK.NIHR Cochrane Heart Programme Grant
Medical Subject Headings (MeSH)
*Postmenopause; Cardiovascular Diseases [*prevention & control]; Estrogen Replacement Therapy [adverse effects; *methods]; Hormone Replacement Therapy [adverse effects; methods]; Stroke [chemically induced]; Venous Thromboembolism [chemically induced]
MeSH check words
Adult; Aged; Aged, 80 and over; Female; Humans; Middle Aged
* Indicates the major publication for the study