Hormone Replacement Therapy for Prevention: Title and subTitle BreakMore Evidence, More Pessimism

Most physicians and researchers were surprised by the results of the Heart and Estrogen/progestin Replacement Study (HERS).¹ This randomized trial assessed the efficacy of hormone replacement for secondary prevention of coronary heart disease (CHD)—that is, the ability of hormone replacement therapy (HRT) to prevent CHD events in women with established coronary disease. After an average of 4.1 years of follow-up, there was no overall difference in CHD end points between women randomized to receive placebo or HRT given as a continuous combined regimen of 0.625 mg/d of estrogen plus 2.5 mg of medroxyprogesterone acetate.

However, when these results were reported, based on a trend of decreasing risk of CHD events with increasing duration of HRT use, the HERS authors speculated that the HERS results were a consequence of domination of the antiatherogenic effects of HRT by its early thrombotic effects.¹ The implication was that a net benefit for CHD would have been observed with longer durations of HRT use. This speculation was translated into a "don't start, don't stop" message for physicians: don't start HRT for women with coronary disease because of the increased risk of CHD events in the early years after initiation of use, but don't discontinue HRT in women who have used it for several years because they have already weathered the period of risk and have entered the late period of putative benefit.

In this issue of THE JOURNAL, Grady et al² report on CHD outcomes in a follow-up study of women recruited to HERS (HERS II). After an additional 2.7 years of follow-up, the anticipated overall net benefit of HRT use for coronary events did not materialize. The lower rates of coronary disease in long-term users of HRT were not sustained.

In HERS II, treatment assignment was unblinded in 1998. Subsequent use of HRT was based on decisions made by women and their physicians. HERS II is not as robust as a design in which randomized assignment to placebo or HRT would have continued for all women recruited to the study. However, a remarkably low percentage of women who had been assigned to placebo (only 8%) began taking HRT after treatment assignment was unblinded, and 45% of women assigned to HRT remained adherent to HRT at the end of the follow-up period. Furthermore, the results of analyses of CHD end points based only on women who continued to use HRT according to their assigned treatment, as well as analyses that adjusted for differences between women who continued and did not continue treatment assignment, yielded results that were the same as the analysis based on original treatment assignment. Although women assigned to placebo were more likely to use statins than women assigned to HRT, differences in statin use did not explain the lack of benefit of HRT for prevention of CHD events in women with coronary disease.

The findings of HERS and HERS II with regard to a lack of benefit of HRT in high-risk women are supported by findings of 2 randomized trials that assessed the effect of HRT on progression of atherosclerosis.^{3 - 5} The Estrogen Replacement Atherosclerosis trial³ randomized 309 postmenopausal women with angiographically verified coronary disease to receive placebo, estrogen replacement therapy (ERT) as 0.625 mg/d of conjugated equine estrogen, or HRT as daily estrogen plus progestin, as in HERS. After 3.2 years of follow-up, the mean minimal coronary artery diameter on repeat coronary angiograms was not significantly different among women assigned to placebo, ERT, and HRT. The findings were unchanged in an analysis limited to patients who complied with treatment. In the Postmenopausal Hormone Replacement Against Atherosclerosis trial,^{4 - 5} 321 postmenopausal women with increased carotid intima-media thickness based on B-mode ultrasound were randomly assigned to receive placebo or 1 of 2 HRT regimens (1 mg/d of 17 β-estradiol, with administration from day 17 through day 28 of 0.025 mg of gestodene either every 4 weeks or every 12 weeks). After 48 weeks, there was no difference between the placebo and active treatment groups in progression of atherosclerosis measured as a change in carotid⁴ or femoral⁵ intima-media thickness.

In HERS II, HRT provided no benefit for prevention of ischemic stroke, confirming a prior report from the randomized portion of HERS.⁶ The HERS and HERS II findings for ischemic stroke among high-risk women are consistent with the results of the Women's Estrogen for Stroke Trial.⁷ In this trial, 664 postmenopausal women who had recently had an ischemic stroke or transient ischemic attack were randomized to receive placebo or ERT as 1 mg/d of 17 β-estradiol. After 2.8 years of follow-up, the relative risk of nonfatal stroke in women assigned to ERT was 1.0 (95% confidence interval, 0.7-1.4).

Also in this issue of THE JOURNAL, Hulley et al⁸ report on noncardiovascular end points in HERS II. An overall increase in the risk of venous thromboembolism and biliary tract surgery in HRT users, reported in other HERS publications,^{9 - 10} persisted. Of considerable importance, the risk of fracture was not decreased in HRT users in HERS II. Randomized trials have established the effectiveness of ERT and HRT in preventing postmenopausal bone loss.¹¹ Epidemiologic studies are consistent in showing a decreased risk of hip fracture outcomes in users of estrogen.¹² The lack of information about use of bisphosphonates and raloxifene after unblinding makes the HERS II findings on fracture somewhat difficult to interpret. These drugs have been shown in randomized trials to prevent fracture.^{13 - 14} For instance, if more women initially assigned to placebo initiated use of bisphosphonates or raloxifene after unblinding, any benefit of HRT use for fracture prevention would be obscured. However, the difference in use of these drugs between women initially assigned to placebo and those assigned to HRT would have to be very large to change the observed fracture risk in HRT users into a large overall decrease in fracture risk.

Since the HERS results were published, many physicians have clutched at every hint that HRT might have benefit in preventing disease, death, and long-term disability in postmenopausal women. More evidence brings more pessimism about the preventive benefits of HRT and ERT. What then remains?

It is possible that ERT or HRT might prevent CHD in women free of coronary disease. The Estrogen Prevention of Atherosclerosis Trial¹⁵ provides limited evidence. This trial assessed the effect of ERT on carotid intima-media thickness in 222 postmenopausal women without preexisting cardiovascular disease and low-density lipoprotein cholesterol levels of at least 130 mg/dL (≥3.37 mmol/L) who were randomized to receive placebo or 1 mg/d of 17 β-estradiol. After 2 years of follow-up, the rate of atherosclerosis progression measured as change in intima-media thickness was less in the ERT-treated group than in the placebo group, but the benefit of ERT was limited to women who did not also take a lipid-lowering medication.

The Women's Health Initiative (WHI) is a randomized trial that has recruited 27 348 participants to its hormone replacement component.¹⁶ Women without a uterus have been randomized to receive placebo or ERT as 0.625 mg/d of conjugated equine estrogen; women with a uterus have been randomized to receive placebo or HRT as continuous combined estrogen plus progestin, as in HERS. While full results have not been published, following meetings of the WHI study's data and safety monitoring board in April 2000 and February 2001,¹⁷ it was announced that the risks of venous thromboembolism, stroke, and myocardial infarction were increased in women assigned to hormone replacement. A net benefit of ERT, HRT, or both in preventing vascular disease in the WHI would require these early increases in the risk of vascular events—stroke and CHD—to be offset by decreases in vascular risk in women with longer durations of use. This scenario was more plausible before publication of the HERS II results than now.

Prospective observational studies continue to suggest that ERT might protect against cognitive decline¹⁸ and the development of dementia.¹⁹ Data on ERT and HRT to prevent cognitive decline and dementia are not consistent,^{20 - 22} however, and randomized trials of estrogen in the treatment of Alzheimer disease show no evidence of benefit.^{23 - 24} Observational studies of HRT and CHD were consistent, yet randomized trials have failed to confirm benefit. The WHI²⁵ will address these issues, and observational research should continue since some studies show benefit of HRT only in the oldest old.¹⁸

Pessimism about HRT and ERT does not mean pessimism about disease prevention in postmenopausal women. Randomized trials that included women provide strong support for the use β-blockers²⁶ and aspirin to prevent CHD events and ischemic stroke in women with coronary disease or a history of ischemic stroke or transient ischemic attack.²⁷ The Heart Outcomes Prevention Evaluation (HOPE) trial showed that angiotensin-converting enzyme inhibitors prevent CHD events in women at high risk of cardiovascular disease.²⁸ Randomized trials show that blood pressure control prevents stroke and CHD in women with hypertension.²⁹ Aspirin use has a favorable risk-benefit ratio for primary prevention of CHD in high-risk women.³⁰ Randomized trials support screening and treatment of dyslipidemia to prevent first CHD events in women with average serum cholesterol levels,^{31 - 32} and some lipid-lowering drugs may also prevent stroke.³³ Randomized trials show that bisphosphonates and raloxifene prevent fracture in women with low bone mineral density or a history of fracture.^{13 - 14} An appropriately pessimistic view of HRT as an omnibus agent to prevent disease in postmenopausal women should focus more attention on these preventive interventions, which have a strong evidence base.