The WHI and the brain: what have we learned?☆

The Women's Health Initiative (WHI) study 1, 2 has dramatically changed our approach to the management of the postmenopausal woman with headlines reading: “Hormone therapy found to kill women” and now more recently, “Hormone therapy causes Alzheimer's disease.” Professional societies and the US Preventive Health Services Task Force admonish physicians to agree with this sort of headlines because the risk of hormone therapy (HT) exceeds the benefits. Hormone therapy should not be used for the treatment or prevention of chronic disease. In other words, HT should not be used for the prevention not only of cardiovascular disease but also osteoporosis (estrogen is still approved by the FDA for the prevention of osteoporosis but not treatment), osteoporotic fractures, and Alzheimer's disease (AD). Its use should be restricted to the treatment of menopausal symptoms for the shortest duration possible. So how did we get it so wrong? Were we misled by 50 years of observational data because of the “healthy user bias”? A bias based on the fact that women who elect to initiate HT are healthier and more highly educated and therefore less likely to develop heart disease and AD. Let's look at that research, which appears to be refuted by the WHI–Memory Study (WHIMS), and then examine what we have really learned from WHIMS.

First, what is the scientific rationale for believing that HT may affect the expression of AD and brain aging? Neurobiologists have demonstrated in experimental animal models during the past two decades that estrogen stimulates the growth of neurons and the connections between neurons (synapses). Estrogen also stimulates the production of the neurotransmitters acetylcholine and serotonin involved in memory and the cytokines involved in the repair of damaged neurons. Finally, estrogen has been shown to protect neurons from ischemic injury and the neurotoxicity of the β-amyloid protein implicated in the pathogenesis of AD. Not surprisingly then, estrogen replacement in ovariectomized animals restored memory performance. However, this effect of estrogen replacement on memory, as seen in young and middle-aged animals, was not seen in older animals (3). Thus, there is a compelling scientific rationale for estrogen having a positive effect on the brain and memory, at least in middle-aged women.

View Large Image Download to PowerPoint

Fig 2.

What do we know about the effects of estrogen on the brain in women? One of the manifestations of the menopause is memory, concentration, and mood changes. Approximately two-thirds of women report adverse effects on memory during perimenopause. Dr. Sherwin (4) has demonstrated that surgical menopause or the pharmacological induction of the estrogen deficiency state with Lupron results in a decrement in memory performance that is restored with estrogen replacement. Numerous observational studies indicate that postmenopausal women on HT perform better on selected measures of cognitive function than women not on HT. These women also demonstrate a slowing of the age-related decline in cognitive function (5). However, clinical trials in older postmenopausal women, including the WHI, have failed to demonstrate any effect of HT on the slowing of the age-related decline in cognitive function. The study of Dr. Mathews and her colleagues (6) suggests that estrogen's effect on brain aging is dependent on the initiation of HT at the time of the menopause and that the window of opportunity to effect brain aging may be limited to this critical physiologic transition from the reproductive state to the estrogen deficiency state. The vast majority of women included in observation studies are women who initiated HT at the time of menopause.

View Large Image Download to PowerPoint

Fig 1. HT Therapy: A Window of Opportunity.

Can this discrepancy between observational studies and clinical trials be explained by the “healthy user bias”? The level of education is a significant confounder in studies of cognitive function. However, when differences in education exist, statistical adjustment for education does not change the association between HT and better cognitive performance. Another approach to this question is to examine the association between endogenous estrogen or biological markers of endogenous estrogen and cognitive function. Again, we find that women with higher levels of estrogen or lower rates of bone loss perform better on measures of cognitive function and are less likely to develop AD (7). These studies are significant in that they obviate the healthy user bias and suggest that the doses of estrogen required to slow brain aging and to delay the expression of AD may be relative low.

What about AD? Does HT cause AD as claimed by the press? As noted, osteoporosis is associated with AD. Women who have lost more than 2 inches in height have a ninefold increase in their risk of developing AD (8). Multiple observational studies demonstrate that women who have used HT have a lower risk of developing AD. Of 15 studies, four failed to demonstrate a reduction in risk and in fact suggested that the risk was actually increased with HT. Of interest, these studies excluded past users of HT. This difference is best illustrated by the Cache County Study (9), which examined past and current users of HT separately. The incidence of AD was ascertained by a three-tiered screening process similar to the one used in the WHI. The investigators first demonstrated that the incidence of AD was about two-and-half times greater in women than in men, consistent with other studies. They also demonstrated that women who initiated HT after age 60 years had a twofold greater incidence of AD than nonhormone users. In contrast, women who initiated HT in the past, presumably at the time of the menopause, had between a 40%–80% reduction in their incidence of AD depending on the duration of HT use.

So what have we learned from the WHIMS (2)? In this study, women were recruited from the WHI who were participants of the HT component of the study and who were more than 65 years. The results of the CEE/MPA arm revealed that the incidence of dementia was increased twofold by HT. The point estimate of this increased incidence in AD was essentially identical to the increase reported earlier in the Cache County Study (9). The WHIMS also demonstrated that prior use of HT resulted in a 62% reduction in a woman's risk of developing AD. It is therefore disingenuous to suggest that WHIMS repudiates decades of prior research on estrogen and the brain. In truth, WHIMS has confirmed what we have learned from preceding clinical trials and observational studies: Initiating HT after age 60 years results in an increased risk of AD, whereas earlier initiation may dramatically reduce a woman's risk of this devastating disease.

The WHIMS also provides some clues as to potential mechanisms of this adverse effect on the expression of this neurodegenerative disorder, a disorder in which the neurodegenerative changes are evident in the brain some 10–15 years before the clinical manifestations of the disease. Mild cognitive impairment (MCI), the precursor to dementia, was also increased by HT. This increase occurred in the second and third years of exposure to HT with a trend toward protection in the fourth and fifth years (10), whereas the increase in dementia peaked in the third and fourth years. These data suggest that there is a vulnerable population of patients representing approximately 1% of the population in whom the initiation of HT after age 65 years accelerates the expression of AD. We also know that in the largest two clinical trials in patients with AD, HT did not slow the progression of the disease (11).

Can these changes reflect a similar mechanism as is seen with the transient increase in incidence of cardiovascular disease? As reported recently in the New England Journal of Medicine (12), this increase occurred in the first year followed by a statistically significant trend toward no effect or benefit. This report also revealed that initiation of HT within 10 years of menopause was not associated with an increased risk. In fact, the point estimated suggested a slight benefit, although this was not significant. In contrast, the increase in the incidence of stroke was independent of the age because menopause, and like MCI, occurred in the second and third year of exposure. Another argument against this effect being an acceleration of or destabilization of pre-existing atheromatous disease is that women with hypertension and prior history of cardiovascular disease actually demonstrated dramatic improvement in cognitive function. At this time we just do not know what the genotypic characteristics are of the small population of older women who are vulnerable to this adverse effect of HT on the brain.

So how do we apply the information learned from WHI and WHIMS to the management of our menopausal patients? These are patients transitioning from the reproductive state to the estrogen deficiency state. Should they initiate HT and for how long? This question can best be answered by examining a woman's lifetime risks and benefits of initiating HT for 10 years at the time of the menopause. Unfortunately, in WHI and WHIMS the average age of initiating HT was after age 63 years. Thus, we are forced to rely on observational data. With respect to cardiovascular disease, let us assume that there is a null effect that would be consistent with the reanalysis of the estrogen plus progestin and the recently published estrogen-alone WHI data 12, 13. For breast cancer, we can use the data from the Collaborative Group on Hormonal Factors in Breast Cancer (14). This figure may overestimate the risk for women on unopposed estrogen. In the unopposed estrogen arm of the WHI (13), the risk of breast cancer actually decreased by 7 cases per 10,000 person-years after an average of 6.8 years of observation. The WHI and observational studies appear to be consistent with respect to the reduction in the risk of colon cancer with HT. It should be noted that mortality from colon cancer is about three times that of breast cancer. Prevention of osteoporotic fractures is not included because that data is not available. For AD, let us assume a prevalence of 30% at age 84 years, the life expectancy of a woman at age 50 years. Thus, we can anticipate 300 cases of AD in every 1,000 women who elect not to initiate HT at the time of the menopause. Prospective observational trials suggest that HT for 10 years will reduce a woman's risk by 80% (9) or for every 1,000 women who initiate HT at the time of the menopause, there will be 240 fewer women who will develop AD in their lifetime. This analysis of a woman's lifetime risk of disease suggests that the benefits of initiating HT and continuing for 10 years clearly outweighs the risks.

Where do we go from here in this era of evidence-based medicine? There is compelling observational data and experimental animal model data that suggest that the clinical trial data of WHIMS cannot be extrapolated to women initiating HT at the time of the menopause. Should we wait until that clinical trial data are available? Those clinical trials would require randomizing symptomatic perimenopausal women to HT or placebo and following them for 30 years to observe a sufficient number of cases AD. Clearly, those trials are not possible. When clinical trials are not feasible, we must rely on observational data. The risk of lung cancer from smoking was not based on any clinical trial nor is the recommendation that exercise prolongs life. Finally, if we are to rely on observational data, we must demonstrate that the benefits far exceed the known risks of treatment. The effects of HT on AD, when initiated at the time of the menopause, satisfy these criteria for recommending HT as a public health policy.