Abstract

Objective To investigate the long term effect of hormone replacement therapy on cardiovascular outcomes in recently postmenopausal women.

Design Open label, randomised controlled trial.

Setting Denmark, 1990-93.

Participants 1006 healthy women aged 45-58 who were recently postmenopausal or had perimenopausal symptoms in combination with recorded postmenopausal serum follicle stimulating hormone values. 502 women were randomly allocated to receive hormone replacement therapy and 504 to receive no treatment (control). Women who had undergone hysterectomy were included if they were aged 45-52 and had recorded values for postmenopausal serum follicle stimulating hormone.

Interventions In the treatment group, women with an intact uterus were treated with triphasic estradiol and norethisterone acetate and women who had undergone hysterectomy received 2 mg estradiol a day. Intervention was stopped after about 11 years owing to adverse reports from other trials, but participants were followed for death, cardiovascular disease, and cancer for up to 16 years. Sensitivity analyses were carried out on women who took more than 80% of the prescribed treatment for five years.

Main outcome measure The primary endpoint was a composite of death, admission to hospital for heart failure, and myocardial infarction.

Results At inclusion the women on average were aged 50 and had been postmenopausal for seven months. After 10 years of intervention, 16 women in the treatment group experienced the primary composite endpoint compared with 33 in the control group (hazard ratio 0.48, 95% confidence interval 0.26 to 0.87; P=0.015) and 15 died compared with 26 (0.57, 0.30 to 1.08; P=0.084). The reduction in cardiovascular events was not associated with an increase in any cancer (36 in treated group v 39 in control group, 0.92, 0.58 to 1.45; P=0.71) or in breast cancer (10 in treated group v 17 in control group, 0.58, 0.27 to 1.27; P=0.17). The hazard ratio for deep vein thrombosis (2 in treated group v 1 in control group) was 2.01 (0.18 to 22.16) and for stroke (11 in treated group v 14 in control group) was 0.77 (0.35 to 1.70). After 16 years the reduction in the primary composite outcome was still present and not associated with an increase in any cancer.

Conclusions After 10 years of randomised treatment, women receiving hormone replacement therapy early after menopause had a significantly reduced risk of mortality, heart failure, or myocardial infarction, without any apparent increase in risk of cancer, venous thromboembolism, or stroke.

Trial registration ClinicalTrials.gov NCT00252408.

Introduction

Hormone replacement therapy for postmenopausal women has been subject to much discussion and speculation since the 1960s. Before 2002 the effects of hormone replacement therapy were believed to be beneficial, owing to a reduction in risk of cardiovascular disease, osteoporosis, and colon cancer.1 The negative side effects—an increased risk of breast cancer and thromboembolic disease—were thought to be outweighed by the advantages, principally on the basis of results from observational studies.2 3 In 2002 the primary results from the Women’s Health Initiative showed no cardiovascular benefit from hormone replacement therapy. These conflicting results have led to the “timing hypothesis”; the idea that the differences in cardiovascular outcome can be accounted for by time since menopause until the start of hormone therapy.4 5 The observational studies mainly have shown positive cardiovascular effects, probably as a result of hormone therapy starting shortly after menopause, and the randomised studies have shown no or negative cardiovascular effects, often in women who start hormone therapy many years (5 to 20) after menopause. In meta-analyses taking age into special consideration, use of hormone therapy in younger women has been associated with a lower risk of coronary heart disease6 and reduced overall mortality.7

We used data from the Danish Osteoporosis Prevention Study (DOPS) to test whether hormone replacement therapy can reduce cardiovascular endpoints in women if started early after menopause.

Methods

The Danish Osteoporosis Prevention Study is a prospective investigator initiated multicentre trial evaluating the effect of hormone replacement therapy as primary prevention of osteoporotic fractures. The criteria for inclusion in the study were healthy, recently postmenopausal white women aged 45-58, with last menstrual bleeding 3-24 months before study entry or perimenopausal symptoms (including irregular menstruations) in combination with recorded postmenopausal serum follicle stimulating hormone values (>2 standard deviations over the premenopausal mean). We included women who had had hysterectomy if they were aged 45-52 and had records showing an increase in serum follicle stimulating hormone levels. Exclusion criteria were a history of bone disease (including non-traumatic vertebral fractures on radiography), uncontrolled chronic disease, previous or current cancer or thromboembolic disease, current or past treatment with glucocorticoids for more than six months, current or previous use of hormone replacement therapy within the past three months, and alcohol or drug dependency.

Between 1990 and 1993, 2016 women were enrolled in a prospectively followed cohort. Of these, 1006 were randomly allocated (open label) to receive hormone replacement therapy (n=502) or no treatment (n=504); the remaining 1010 women had a personal choice (of these, 221 opted for hormone replacement therapy). The results presented here are based solely on the randomised groups. Recruitment has been described in detail elsewhere.8Participants were recruited by direct mailing to a random sample (n=47 720) of women from the general background population.9 In the invitation letter women were asked to participate in a study on prevention of postmenopausal osteoporosis. Participants were stratified according to centre and randomly allocated to either hormone replacement therapy or no treatment in blocks of 10, using sealed envelopes. Participants gave informed consent before the study.

The women in the treated group with an intact uterus started treatment with 2 mg synthetic 17-β-estradiol for 12 days, 2 mg 17-β-estradiol plus 1 mg norethisterone acetate for 10 days, and 1 mg 17-β-estradiol for six days (Trisekvens; Novo Nordisk, Denmark). In women who had undergone hysterectomy, first line treatment was 2 mg 17-β-estradiol a day (Estrofem; Novo Nordisk, Denmark). Other treatment modalities were offered to those who experienced side effects or insufficient relief of symptoms.10

All participants underwent a physical examination and biochemical screening at baseline. They were subsequently seen after six months, one year, and two, three, five, and 10 years. The study drugs were posted to the women randomised to hormone replacement therapy, and they were offered an annual visit. We advised the women that if they had health concerns they should contact their own general practitioner or gynaecologist. The planned duration of the study was 20 years. However, as data published from other trials at the time of the 10 year visit indicated that use of hormone replacement therapy might result in more harm than benefit in postmenopausal women we advised our study participants to stop treatment.11 After their 10 year visit we followed the participants in national registers, which provide data on all hospital contacts or deaths.

Assessment of mortality and admission to hospital due to cardiovascular events or cancer

The primary endpoint for this study was a composite of death, admission to hospital for myocardial infarction, or heart failure. We prespecified and adjudicated cardiovascular disease as well as cancer as safety outcome measures. Secondary endpoints were the individual components of the primary endpoint and admission to hospital for stroke. Safety endpoints included death or a diagnosis of breast cancer or other cancer grouped together, and admission to hospital for pulmonary embolism or deep venous thrombosis. Evaluations of endpoints were carried out with a PROBE (Prospectively, Randomised, Open with Blinded Endpoint evaluation) design.

On 16 June 2008 we ended our follow-up period by retrieving data on all participants from the Danish civil registration system and the national hospital discharge register. In the Danish civil registration system we identified all women who had died or emigrated during follow-up, as this register has electronic records on all changes in vital status, including change of address and date of death for the entire Danish population since 1968.

Using the Danish national hospital discharge register, which covers all contacts to Danish hospitals, we identified women who had been admitted to hospital for a cardiovascular event. The register was founded in 1977 and includes information on discharge diagnoses and date of discharge assigned exclusively by the doctor at discharge according to the International Classification of Diseases, eighth revision until the end of 1993 and the 10th revision from 1994. The register has nationwide coverage of hospitals with an almost 100% completeness of recordings and a high precision of diagnoses.12 Using this register, we identified all study participants who had been assigned a diagnosis of cardiovascular disease classified as myocardial infarction (ICD-10 code I21), heart failure (ICD-10 code I42 and I50), and stroke (ICD-10 code I60 to I69, which covers ischaemic as well as non-ischaemic stroke). For the composite endpoint we used the date of the first incident.

Total cancer included all cancer diagnoses (ICD-10 codes C01 to C99) except non-melanoma skin cancer (ICD-10 codes C44). We defined other cancer as total cancer except for breast cancer (ICD-10 code C50), which was independently surveyed and monitored. Breast cancer was analysed both as a composite endpoint (with mortality) and as breast cancer only. For composite endpoints, we used the date of first incident. Pulmonary embolism (ICD-10 code I26.9) and deep vein thrombosis (ICD-10code I80.1 to 80.3) were registered separately.

Statistical analysis

Only the randomised participants are considered in this study (n=1006), and all analyses are done on the intention to treat population, except when mentioned specifically in sensitivity analyses. The analyses were carried out, with 1 august 2002 as the stopping date, about 10 years after randomization (when the randomised treatment was stopped). So as not to miss major long term effect, we carried out secondary analyses with an additional six years of non-randomised follow-up. Because women who had undergone hysterectomy in the treatment group received oestrogen only, post-hoc analyses were carried out for this group; a total of 192 women had undergone hysterectomy, of whom 95 were randomised to treatment. As only three women had undergone complete oophorectomy, no subgroup analyses were done. Unless otherwise stated, baseline data are expressed as means (standard deviations). We tested dichotomous variables with a χ2 test and continuous variables with students t test. All tests were two sided, and we considered P<0.05 to be statistically significant. The survival data and the composite endpoint are presented using the Kaplan-Meier method and analysed by a log rank test. We carried out analyses as time to first event, thus we counted the women only once at their first endpoint of that analysis. Hazard ratios (95% confidence intervals) were determined using Cox proportional hazards regression analyses. We repeated the Cox regression analyses adjusted for age. Age was included owing to a difference of 0.5 years between the two randomised groups. With a hazard ratio for the primary endpoint of less than 0.7 the study had a power of at least 80% to detect a difference between the two groups. We tested model assumptions including linearity of continuous variables, the proportional hazards assumption, and absence of interactions and found them to be valid unless otherwise indicated. We carried out sensitivity analyses on women who took more than 80% of the prescribed treatment for five years. Statistical analyses were done with SAS version 9.2.

Results

Of 1006 recently postmenopausal women or women with perimenopausal symptoms and recorded postmenopausal serum follicle stimulating hormone levels included in the study, 502 were randomly allocated to hormone replacement therapy and 504 to no treatment fig 1⇓). The women had a mean age of 49.7 (2.8 years), mean body mass index of 25.2 (4.4), and mean time since menopause of 0.59 (0.64) years (about seven months). Their mean blood pressure was 130/81 mm Hg and 43% of the women were smokers at the time of inclusion. Women in the control group were 0.47 years (about 5.7 months) older (P=0.006) than those in the treated group, whereas other variables did not differ significantly between the groups (table 1⇓). Only 22 (2%) of the women had used hormone replacement therapy previously, for a median duration of 1 year (interquartile range 0-5 years).