To compare health outcomes during 14-year observational follow-up in women initially randomised to unopposed estrogen or placebo.
To compare health outcomes during 14-year observational follow-up in women initially randomised to unopposed estrogen or placebo.
At recruitment to the Estrogen for the Prevention of Re-Infarction Trial (ESPRIT) women were assigned to estradiol valerate: 2 mg or placebo treatment for 2 years.
Women were recruited from 35 hospitals in the northwest of England and Wales in July 1996–February 2000.
Women aged 50–69 surviving their first myocardial infarction.
All women were followed by data linkage to UK mortality and cancer records; mean follow-up 14.1 and 12.6 years, respectively. In an intention-to-treat analysis, hazard ratios (HRs) were computed, overall and stratified by age at recruitment.
Death (all-cause, cardiac disease, stroke or cancer) and cancer incidence (any, breast or endometrium).
There were 418 deaths in 1017 women randomised. The all-cause mortality HR of 1.07 (95% CI 0.88–1.29) indicated no significant difference between treatment groups. Women aged 50–59 years at recruitment had lower HRs than women aged 60–69 years for all outcomes except ischaemic heart disease. Among 149 incident cancers there were seven cases of breast cancer in the intervention arm and 15 in the placebo; HR 0.47 (95% CI 0.19–1.15). There were no deaths from endometrial cancer but three incident cases, one in the active arm and two in placebo.
These results suggest that unopposed estrogen may be used safely by women with an intact uterus surviving a first myocardial infarction.
The ESPRIT trial was designed to investigate whether estrogen (estradiol valerate) reduced the risk of re-infarction, cardiac death or all-cause mortality in postmenopausal women recruited at the time of hospital admission for a first myocardial infarction (MI). Few other trials of the health effects of hormone replacement therapy (HRT) have used estrogen without progesterone (unopposed estrogen) in women with an intact uterus. The recent Cochrane review lists only two (EPAT and PEPI), both of which used surrogate markers of disease rather than clinical outcomes. The largest trial of unopposed estrogen (Women's Health Initiative Estrogen-Alone Trial [WHI E-A]), which recruited women in good health, initially included women with an intact uterus in the unopposed estrogen arm but later changed the protocol because of concerns about endometrial hyperplasia. The ESPRIT trial used estrogen without progesterone for all women in the treatment arm, including those with an intact uterus, because of concern about the safety of using progesterone in women with recent MI. The follow-up described in this report was planned primarily to determine whether this use of unopposed estrogen was detrimental, specifically whether there was any evidence of excess endometrial cancer that might be attributed to the intervention.
Recently published data from the WHI E-A trial has demonstrated a protective effect of unopposed conjugated equine estrogens (CEE) on invasive breast cancer in women of all ages in that trial and, in younger women (aged 50–59 years at enrolment), on the incidence of coronary heart disease (CHD), MI and death (all causes), with a trend towards more adverse outcomes in those over 60 years. Although smaller than WHI E-A, ESPRIT, with 1017 subjects randomised, is larger than any other trial using unopposed estrogen. Continued follow-up of the women recruited provided the opportunity to assess whether the WHI-E-A results on incident breast cancer and death from all causes, stratified by age, were found also in women recruited in a different country and exposed to a different type of estrogen (estradiol valerate rather than CEE used in the WHI E-A).
As described in the earlier report, 1017 women age 50–69 years who had survived a first MI were recruited at the time of their hospitalisation for MI from 35 collaborating hospitals in the northwest of England and Wales between July 1996 and February 2000. Women who reported a history of cancer or use of hormone replacement therapy in the previous 12 months were excluded. Subjects were randomised to receive either one tablet of estradiol valerate (2 mg; n = 513) or placebo (n = 504), daily for 2 years. Each subject was followed up through their family physician at 3, 6, 12, 18 and 24 months from date of recruitment, with follow-up of the last recruit ending on 2 February 2002. The outcome of this initial ‘active’ follow-up has been reported previously.
Given the unconventional use of unopposed estrogen in women with an intact uterus (373/513 in the active treatment arm), strict trial procedures sought to maximise the notification of all episodes of vaginal bleeding, and provide effective management of such episodes. Women with vaginal bleeding during the trial were clinically assessed and, if appropriate, had a pipelle endometrial biopsy followed by a 14-day course of medroxyprogesterone acetate 20 mg, increased to three courses each lasting 14 days if the biopsy showed complex hyperplasia. When women with an intact uterus stopped active or placebo treatment, at 2 years after study entry or earlier, those who had not been investigated previously for vaginal bleeding were offered an endometrial biopsy. Women in the active arm who had not had a hysterectomy were also sent a letter each year for 5 years after stopping treatment to remind them to seek medical attention if they experienced vaginal bleeding. There was no attempt after the end of the 2-year intervention period to obtain information from women or their physicians about medication (including use of estrogen) or health events. However, all 1017 women were flagged at recruitment on the UK National Health Service central register so that deaths and cancers could be notified through record linkage. The analyses reported here are based on these notifications.
Follow-up for this paper was to 31 December 2010 for cancer incidence (mean follow-up 12.6 years, range 10.9–14.5) and to 30 June 2012 for mortality (mean follow-up 14.1 years, range 12.4–16.0). Cancer incidence, vital status and cause of death were determined from data routinely collected by the Office of National Statistics for England and Wales, and by parallel systems in Scotland for women who had moved there after recruitment into the study. Cancer incidence data are routinely collected by regional cancer boards and forwarded to the national register with some delay (hence the shorter follow-up for this endpoint). Mortality is reported centrally with cause of death (both underlying and ‘mentioned’, i.e. a condition that contributed to but did not cause the death) on the death certificate coded routinely using the World Health Organization International Classification of Diseases (ICD) codes, version 9 until 2000 and version 10 in subsequent years. For this follow-up the outcomes of interest were death (all-cause) and deaths from ischaemic heart disease (ICD-9 410-414; ICD-10 I20-I25), any cardiac diagnosis (ICD-9 codes 393-398, 410-414, 415-417, 420-429; ICD-10 I05-I09, I20-I25, I26-I28, I30-I52), stroke (ICD 9: 430-438, ICD-10: I160-168), any cancer (ICD-9 140-209, ICD-10 C00-C97), breast cancer (ICD-9 174; ICD-10 C50) and endometrial cancer (ICD-9 182; ICD-10 C54.1). Incidence of cancer in this analysis, for any cancer, breast cancer and endometrial cancer, was taken from cancer notifications with the same ICD codes as those in the mortality analysis.
Hazard ratios (HRs) comparing treatment arms were estimated using Cox regression. In the all-cause mortality analysis, data for those still alive at the end of follow-up period were treated as censored observations; for analysis of specific causes or cancer incidence, censoring was at the end of the follow-up period or time of death if earlier. All HRs were adjusted for age at risk, using six 5-year age bands (50–55 to 75–80). A binary variable indicated the randomised treatment arm; thus each analysis was by the intention to treat during the 2-year intervention phase. Each outcome was considered separately. For mortality, the analysis was first by underlying cause and then by ‘any mention’ on the death certificate. For cancer incidence, the time to event for ‘any cancer’ was taken as time to the first cancer if more than one had occurred: in situ and benign neoplasms were excluded. Confidence intervals (95% CI) were calculated. For each outcome the effect of treatment group was estimated for all women and then stratified by age at recruitment (age 50–59 years [n = 301] or 60–69 years [n = 716]): again, adjustment was made in 5-year bands for age at risk. Tests of whether the HRs differed significantly between the two age groups were based on a test of interaction between age and treatment in the regression model. The HR for endometrial cancer was calculated only for those with an intact uterus at the time of recruitment.
There were 418 deaths among the 1017 women recruited to ESPRIT, 214 among the 513 in the treatment arm and 204 among the 504 in the placebo arm (Table 1). The overall HR for all-cause mortality was 1.07 (95% CI: 0.88–1.29) (Table 2). Heart disease was recorded as the underlying cause for 177 (42.3%) of the 418 deaths and as either the underlying cause or as a condition mentioned elsewhere on the death certificate (‘any mention’ in the tables) in 269 (64.4%) deaths. Estimates of HRs (Table 2) for all women showed no significant detrimental or protective effect of treatment, although the ratios were above unity for heart disease and stroke. When stratified by age, women recruited at age 50–59 years had lower HRs (except for IHD) than those aged 60–69 years at recruitment, although none of the differences between age groups was significant (P > 0.05).
|n = 167||n = 134||n = 346||n = 370||n = 513||n = 504|
|Ischaemic heart disease (IHD)||23||13.8||14||10.5||66||19.1||58||15.7||89||17.3||72||14.3|
|Any heart disease||24||14.4||16||11.9||73||21.1||64||17.3||97||18.9||80||15.9|
|Cancer – all sites||10||6.0||11||8.2||28||8.1||28||7.6||38||7.4||39||7.7|
|Ischaemic heart disease (IHD)||31||18.6||20||14.9||95||27.5||89||24.1||126||24.6||109||21.6|
|Any heart disease||32||19.2||27||20.2||109||31.5||101||27.3||141||27.5||128||25.4|
|Cancer – all sites||13||7.8||11||8.2||33||9.5||34||9.2||46||9.0||45||8.9|
|Cancer incidence b|
|HR||95% CI||HR||95% CI||HR||95% CI|
|Ischaemic heart disease (IHD)||1.23||0.63–2.41||1.23||0.87–1.76||1.24||0.91–1.70|
|Any heart disease||1.14||0.60–2.16||1.24||0.89–1.73||1.22||0.91–1.64|
|Cancer – all sites||0.71||0.30–1.67||1.09||0.65–1.84||0.91||0.62–1.52|
|Ischaemic heart disease (IHD)||1.19||0.67–2.08||1.16||0.87–1.54||1.18||0.91–1.52|
|Any heart disease||0.91||0.54–1.51||1.18||0.90–1.54||1.12||0.88–1.42|
|Cancer – all sites||0.91||0.41–2.04||1.06||0.65–1.71||1.03||0.68–1.55|
Five deaths were attributed to breast cancer (one in the active treatment group, four in placebo) as the underlying cause. There were seven women with ‘any mention’ of breast cancer on the death certificate, three in the active arm and four in placebo. Endometrial cancer was not reported on any death certificate, either as an underlying or as a ‘mentioned’ cause. Although not an outcome specified a priori, three death certificates with codes for ovarian cancer (ICD-9 183.0; ICD-10 C56) were noted, all in the active arm (P = 0.25, Fisher's exact test).
There were 149 incident malignant cancers among 142 women, plus 13 in situ or benign tumours (none breast or genital; excluded from the analysis). The HR was close to unity for any cancer (all types) among the 1017 women, with lower HRs for those under 60 years of age at recruitment (Table 2). There were 22 incident cases of breast cancer, seven in the treatment group and 15 in placebo. HRs for breast cancer were below unity for the treatment arm overall and for both younger and older women when examined separately, with women aged 50–59 years having the smallest risk estimate (0.33). For the group as a whole, the (two-sided) probability of an HR of 0.47 or smaller, under the null hypothesis, was 0.097, having adjusted for age at risk.
There were three incident cases of endometrial cancer (one in the active arm and two in placebo), giving an HR of 0.52 (95% CI 0.05–5.80) and one cervical cancer (placebo). Five ovarian cancers were listed, four in the active arm (P = 0.37, Fisher's exact test).
This follow-up of ESPRIT was primarily to investigate whether any harm had been done by using unopposed estrogen in women with an intact uterus. No harm has been demonstrated. Although during the trial, more than half of the women in the active arm with a uterus reported bleeding, no woman was diagnosed with endometrial cancer during active follow-up. The unexpected observation of more ovarian cancer in the treatment arm is consistent with chance and has not been reported in other trials.
As in the initial report, there was no significant difference between women in the active and treatment arms on any of the outcomes considered. In the current analysis, the lowest HR (0.47, suggesting a protective effect of active treatment) was for incident breast cancer, with the likelihood of such an effect arising by chance being <0.10. For all the outcomes (except IHD as underlying cause) considered in this extended ESPRIT follow-up, the HR for women aged 50–59 years at recruitment was lower than for older women.
As all of the women were flagged for death and cancers at the NHS Central registries, we should have captured all such events occurring in the trial participants; in effect there was no loss to follow-up. Although our study had limited statistical power for some of the endpoints included in the analysis, it had 80% power to detect a 20% decrease in all-cause mortality. No increase in risk of endometrial cancer was seen in this study (HR = 0.52) but with a confidence interval 0.05–5.80 the possibility of a true increase in risk cannot be excluded on the basis of these data alone. During the extended follow-up of the original trial, we did not obtain participant data on non-fatal clinical events other than cancer incidence. This trial cannot, therefore, add to reports of other outcomes (e.g. gall bladder disease, venous thrombo-embolism, non-fatal stroke) that have been associated with unopposed estrogen elsewhere. Neither could we assess whether, over time, unopposed estrogen affects the risk of non-fatal myocardial re-infarction. Data were not available about use of hormone replacement therapy after the formal trial ended. Some women may have used these products subsequently, although the number is probably small due to the widespread publicity that occurred in the summer of 2002 concerning the premature cessation of the estrogen plus progestin trial of the Women's Health Initiative, leading to regulatory advice in the UK to avoid long-term use. The absence of adverse effect on the endometrium in this study is likely to be, in large part, a reflection of the management of bleeding during and at the end of the active intervention. Eight women found to have atypical hyperplasia during the trial were immediately started on a 3-month course of cyclical medroxyprogesterone acetate: all endometrial abnormalities resolved and none of the women required a hysterectomy during the 2 years of the trial. The conclusion of no adverse effect, therefore, should not be generalised to situations without such active management.
In its most recent follow-up, the WHI E-A found a significantly reduced HR for breast cancer in those randomised to unopposed estrogen (0.77: 95% CI 0.62–0.95). The size of the protective effect was much the same at 10.7 years as at 7.1 years of follow-up, but with more events increasing the trial's power, a stronger conclusion of benefit was possible. Younger women in the active treatment arm of the WHI E-A study had a similar reduction in risk of breast cancer as older women. In a recently reported trial from Denmark, 192 women aged 45–52 years who had undergone hysterectomy were randomly assigned to take unopposed estrogen (n = 95) or to no treatment (n = 97) and followed up for death, cardiovascular disease and cancer for 16 years. Those on unopposed estrogen as their active treatment were reported to have an HR (0.63: 95% CI 0.23–1.78) for breast cancer similar to that in the WHI E-A trial. The lower breast cancer incidence in the women in ESPRIT is consistent with these earlier reports.
The WHI E-A also showed greater benefit of active treatment in younger than in older women at 10.7 years of follow-up. Rates were lower for incident CHD, MI and death (all causes). The incident data for CHD and MI is not comparable with ESPRIT, where all the women had been recruited post-MI, but the lower HR for death from all causes was seen in younger women in both studies.
The WHI E-A found an increased risk of stroke at 7.1 years, but by 10.7 years there was no significant increase in the active intervention group, overall or in any age-specific group. In ESPRIT, clinical reports of stroke were somewhat higher in the intervention arm at the end of 2 years (risk ratio 1.64, 95% CI 0.60–4.47). Non-fatal strokes were not identified in the follow-up reported here, but there was only a small excess of deaths in which stroke appeared as an underlying or ‘mentioned’ cause among those assigned to estrogen, again consistent with the findings from the larger US trial.
The ESPRIT trial was set up to examine whether estrogen prevented re-infarction in post-menopausal women. The results of both the initial trial and this extended follow-up did not suggest important deficit in cardiac events/mortality in the intervention arm and do not support the use of estrogen for prevention of cardiac mortality in this group. Importantly, there was no evidence that use of unopposed estrogen was harmful: its use by women with an intact uterus did not result in an excess risk of endometrial cancer. As such, if estrogen is indicated in elderly women with heart disease for reasons such as the control of menopausal symptoms, the use of unopposed estrogen may be considered. A recent Cochrane review concluded that ‘for women in their 50s without a uterus taking estrogen-only HT for 5–6 years appears relatively safe and there may even be some health benefits’. Our results support this conclusion, even for women with a uterus, provided that there are processes in place for the investigation and management of any vaginal bleeding arising among users.
PH has received research grants from Schering AG, and hospitality and speakers' fees from Schering Health Care Limited.
NC, RMcN and PH were responsible for the design of the original study, with AH contributing to decisions on cardiac aspects of the study and HK to the design and implementation of the gynaecological protocol. During the follow-up reported here, NC and then PH received notifications on deaths and incident cancers. RMcN carried out the statistical analyses. All authors contributed to the writing of the report.
The trial, including follow-up via the Central Registry, was approved by the central ethics committee of the Royal College of General Practitioners, the research ethics committee of the University of Manchester and the local research ethics committees of the 35 collaborating hospitals in the northwest of England and north east Wales.
The initial trial was funded by the UK National Health Services Research and Development Programme on Cardiovascular Disease and Stroke through a grant to the University of Manchester. Medication (active and placebo) used as the intervention was supplied without charge by Schering AG. Additional support came from Schering Health Care Limited. Continued follow-up is supported by the University of Aberdeen. None of the funders has been involved with the collection, analysis or interpretation of data reported here.
Department of Obstetrics and Gynecology, University of Florida College of Medicine, Jacksonville, FL, USA
Conventional wisdom in the 1990s was that menopausal estrogen therapy (ET) prevented coronary artery disease. The ESPRIT investigators employed a 2-year estrogen-only regimen, oral estradiol valerate 2 mg (the equivalent of approximately 1.5 mg of oral micronised estradiol) (Timmer CJ, et al. Eur J Drug Metab Pharacokinet 1999;24:47–53), in women with coronary heart disease, regardless of whether a uterus was present. The long-term follow-up of this cohort does not support the use of ET to prevent a second myocardial infarction. However, ESPRIT's reassuring findings regarding the impact of ET on risk of breast and endometrial cancer will cause some to question our current paradigm of mandating combination estrogen-progestin hormone therapy (EPT) when treating bothersome menopausal symptoms in women with an intact uterus.
Similar to the Women's Health Initiative (WHI), the ESPRIT trial found that long-term use of ET does not increase risk for breast cancer, with both of these large trials finding a reduced risk, contrasting with the findings of WHI and other large studies that have observed an increased risk of breast cancer with EPT.
What prevents us from routinely recommending ET in symptomatic menopausal women with an intact uterus are concerns regarding endometrial cancer. With 2 years of unopposed ET and a protocol which employed endometrial biopsy with short-term cyclical progestin treatment in response to any reported bleeding, the long-term follow-up of the ESPRIT cohort found that 2 years of ET administered to women with an intact uterus did not increase the risk of endometrial cancer. Other placebo-controlled trials which employed 2–3 years of ET (either estradiol 1 mg or conjugated equine estrogen 0.625 mg) in women with a uterus observed no cases of endometrial cancer in women randomised to ET (The Writing Group for the PEPI Trial. JAMA 1996;275:370–5; AZ Steiner, et al. Obstet Gynecol 2007; 109:581–7).
The downside to prescribing ET when a uterus is present relates to uterine bleeding, the need for endometrial surveillance with vaginal ultrasound/endometrial biopsy, and the recognition that some women will develop endometrial hyperplasia, with a few likely undergoing hysterectomy should hyperplasia with atypia be found. Given these issues and the fact that EPT currently represents the standard of care for treatment of symptomatic women with a uterus (North American Menopause Society 2012; 19: 257–71), most clinicians and women will be reluctant to choose ET when a uterus is present. However, some symptomatic women seeking hormonal management may be more concerned about the elevated risk of breast cancer with EPT than the potential for bleeding, vaginal ultrasound, endometrial biopsies and hyperplasia with ET. For such women, the findings of the ESPRIT and other trials indicate that several years of ET, with appropriate endometrial surveillance, represents a reasonable option. Future clinical trials should assess the endometrial safety of longer-term ET.
Financial Support to the Department of Obstetrics and Gynecology, University of Florida College of Medicine-Jacksonville for Clinical Trials: Bayer, Endoceutics, Noven, Teva.
Consultant (Advisory Boards): Actavis, Bayer, Merck, Teva.