Estrogen and Alzheimer's Disease: What Is the Connection?

Beverly N. Jones III, MD, John R. Absher, MD

Medscape Psychiatry & Mental Health eJournal. 1997;2(4) 

In This Article

Observational Studies of Estrogen's Relationship to Alzheimer's Disease

The case-control method--or observational study design in which individuals with a disease of interest are matched to individuals without the disease--has been the basis of many studies attempting to identify any relationship between estrogen and dementia. Typically, the cases of women with dementia are compared with controls without dementia in an attempt to identify the impact of suspected factors, such as whether or not women took estrogen after their ovaries stopped functioning during menopause or after an oophorectomy. Use of the case-control method for studying the impact of estrogen has been made easier by the many large groups (cohorts) of women that have been assembled for other research projects or from whom medical information has been collected and is available for analysis, such as HMO members.

The first case-control reports addressing estrogen use and AD were studies that examined relatively few women and found relative risks for developing AD that ranged from 0.78 to 2.38, with no statistically significant results regarding estrogen therapy.[21,22,23] These studies are limited in their usefulness and generalizability by the small sample sizes and by the methodologic aspect that the estrogen-use history was obtained by proxy from informants.

Retirement communities offer the opportunity to gather information about health issues from large populations that can be more easily followed over time. Leisure World Laguna Hills is a large retirement community in southern California. Researchers periodically mailed a detailed health questionnaire to new residents beginning in 1981, gathering information about medical problems and prescription medication use. Using a case-control design nested in a larger cohort study, the researchers examined death certificates and identified 138 women who had died of AD.[24] Each of these cases were matched to 4 controls by year of birth and year of death. Estrogen use was identified based on the women's questionnaire responses at the time of entry into the retirement community. The women without AD were more likely to have taken estrogen (odds ratio 0.69; 95% CI=0.46-1.03), with the risk of Alzheimer's decreasing as dose and duration of estrogen use increased. The risk of AD also decreased with increasing weight (estrogen levels in postmenopausal women tend to increase with increased body fat). Current use of blood pressure medication to control hypertension also was associated with a lower risk of dementia; this observation may reflect the prevention of hypertension-related infarct dementia.

Two years after the 1994 report, an extension of the Leisure World Study was published.[25] The most recent report documented that 3760 women of the original cohort had died; of these, 248 women had AD or a diagnosis likely to represent AD on their death certificate.[25] Five women who were in the control group were matched for year of birth and death and were selected for each case of AD. Regression analyses examined the relative risk of variables possibly associated with AD. The risk of AD for women who had ever used estrogen was reduced about one third below that of nonusers (odds ratio 0.65; 95% CI=0.49-0.88). Increased duration of estrogen use was associated with reduced risk.

In a study that utilized data from the Alzheimer's Disease Research Center of the University of Southern California, Henderson and colleagues[26] compared 143 women with AD to 92 nondemented control subjects. They assessed current use of estrogen as reported by the women who volunteered to be controls or as reported by the caregiver informants of the women with AD. The AD cases were significantly less likely than the control subjects to have taken estrogen (7% vs 18%). Women who had AD who were taking estrogen did not differ from women with AD not taking estrogen in terms of age, education, and symptom duration, but scored higher on the Mini-Mental Status Exam (MMSE; mean score 14.9 vs 6.5, P<.005).

Mortel and Meyer[27] compared estrogen use among 3 groups: 148 nondemented controls, 93 women with probable AD, and 65 with probable vascular dementia. The proportion of nondemented controls who took estrogen was twice that of the women in the 2 dementia groups (19% vs 11% for AD and 10% for VD). The unadjusted odds ratio for the risk of AD in the absence of estrogen use was 1.82 (95% CI=0.859-3.842).

A study recently reported by Kawas and associates[28] limited the problem of potential inaccuracy of relying on informants to retrospectively assess estrogen use by using information gathered prospectively as part of the Baltimore Longitudinal Study of Aging (BLSA). Of 472 women followed for up to 16 years, 38 cases of AD were identified. Based on ERT use documented prospectively at each BLSA visit, women who had ever used oral or transdermal estrogen were compared with nonusers. The relative risk for AD in ERT users versus nonusers was 0.44 (95% CI=0.20-0.97); these analyses included adjustments to account for differences in educational level. Noting that AD was 2.3 times more likely in women who had never used ERT, the investigators concluded that their findings offered additional support for a protective influence of ERT against AD. Although estrogen use was not randomized, the prospective design permitted accurate assessment of estrogen use and diagnosis of AD.

Another analysis of estrogen use among women diagnosed with AD used data from a longitudinal study of aging in New York City.[29] Of 1124 women followed, 156 (12.5%) reported taking postmenopausal estrogen. After adjusting for education, ethnic origin, and apolipoprotein-E genotype, women who had taken estrogen had significantly later ages of onset of AD than those who did not and had a relative risk of 0.40 (95% CI=0.22-0.85). Estrogen use lasting longer than 1 year reduced risk.

Some researchers have failed to find lower risk of AD for women who took estrogen. Brenner and colleagues[30] performed a population-based case control study at the Group Health Cooperative HMO of Seattle using computerized pharmacy data to examine the relationship between ERT and the development of AD. Women who were identified as having AD through the University of Washington's Alzheimer's Disease Patient Registry were matched to nondemented women who were enrolled in the same HMO. Roughly half of both AD cases and controls had received estrogen, showing no association (adjusted odds ratio=1.1; 95% CI=0.6-1.8 ). The investigators examined the number of prescriptions for estrogen filled by the women studied and the method of estrogen administration (oral tablets vs vaginal creams) but found no significant odds ratios.

Graves and associates[31] examined 130 matched pairs of patients with AD and controls and found an odds ratio for previous estrogen use of 1.15 (95% CI=0.50-2.64). Exposure was ascertained for the AD cases by taking a medical history from an informant. Clear evidence was obtained for an increased risk of AD if there was a first-degree relative with AD, but there was no reduction in risk based on estrogen use.

There are a variety of issues limiting extrapolation of results from these studies ( Table III ). Comprehensive assessments of other illnesses or past or current medication use were not uniformly undertaken. Potential bias is present in studies that did not randomize women to ERT and that assessed estrogen exposure years after their estrogen use. Studies that analyzed data from women who participated in other studies unrelated to estrogen may face ascertainment bias; also, as noted by Graves and associates,[31] the likelihood of individuals negotiating a series of medical channels to obtain specialized medical evaluation is related to the education and socioeconomic level of the spouse or caregiver who accompanies the patient. Whether the findings from these women can be generalized to all women is not clear. It is possible that estrogen use (or nonuse) is a marker for some other medical condition, for exposure to some substance, or for the presence of certain health-related attitudes and behaviors. Observing a reduced relative risk for AD in estrogen users does not demonstrate a cause-and-effect relationship. Understanding the exact relationship of estrogen to the findings reported requires clinical trials of estrogen involving sufficient numbers of women.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: