Sample EPID 168 commentary (Fall 1997)

Exploring Estrogen and Alzheimer's Disease

Background. Alzheimer's disease (AD) afflicts more women than men [3], and the risk of developing AD increases as an individual grows older [6]. Menopause, also a function of age, decreases the level of circulating estrogens in a woman's body. These dual occurrences have lead researchers to hypothesize that estrogen loss associated with menopause may contribute to the development of AD. Conversely, the introduction of exogenous estrogen may prevent or delay the onset of AD. Biological mechanisms support this association: estrogen increases cerebral blood flow, prevents neuronal atrophy, and reverses nerve damage, particularly in the area of the forebrain damaged by AD [2]. Furthermore, prior studies have demonstrated a temporal relationship between cognitive function and estrogen [2].

Encouraged by these findings, investigators have attempted to demonstrate this relationship through epidemiologic study (case-control and, more recently, cohort). These studies have suggested a protective relationship between ever and never use of estrogen [5, 7, 8, 10], but they are not universally confirmed [1]. Relative risks range from 0.40 to 1.10; however, studies vary both in their statistical power and in their ability to control for confounders. Brenner et al [1] conducted a case-control study based on membership in an HMO, and found a relatively weak association between estrogen and AD (1.1), determined to be statistically insignificant. Paganini-Hill & Henderson [7] conducted a case-control study nested within a cohort of predominantly middle-class White women, and found an association of 0.65 (p=0.005). Cohort studies conducted by Tang et al [10] and Kawas et al [5] have also demonstrated statistically significant relationships: Tang et al [10] found an association of 0.40 but apparently failed to control for age, while studying a cohort with a significant minority representation. Kawas et al [5] studied a cohort of predominantly White, well-educated women in Baltimore and found an association of 0.46.

Where a dose-response relationship has been explored, no conclusive relationship has been found. Several studies support this relationship [8, 10], while others demonstrate no obvious relationship [5, 7]. However, it must be noted that categorization of estrogen exposure varied widely in these studies.

Methodological Issues. Although these studies appear to begin to demonstrate a protective relationship, their results are overshadowed by the measurement difficulties and confounding. Measurement is the most problematic issue, both in determining disease status and exposure. Determining onset of AD is difficult since it has no "gold standard" for diagnosis. One case-control study [7] determined AD through a review death certificates, which undoubtedly underestimates the true number of AD cases; while cohort studies [10, 5] have used a battery of extensive physical and psychological examinations, which introduce less bias so long as those who make diagnoses are blinded to exposure status.

Ascertaining estrogen exposure varies widely. The literature has recently attempted to discern a dose-response relationship, but generally limits itself to ever-never use. Dosage, duration, regimen, and route of administration (oral, vaginal, transdermal patch, creams) all further complicate amount of exposure. For instance, is using cream daily equal to taking a pill daily? Paganini-Hill & Henderson [7] obtained estrogen use for a nested case-control study via questionnaire, where the cohort was not known to be disease-free when the questionnaire was administered. Thus, exposure status has some inherent limitations due to memory and misclassification. Misclassification with respect to estrogen exposure, though, was most likely nondifferential [3], which would bias the resulting risk estimate towards the null. Another case-control study [1] defined exposure as filling a prescription for estrogen, which may be more accurate than subject recall given the nature of AD, although it introduces selection bias. A cohort study [10] determined estrogen exposure by re-testing exposure questions to ensure a consistent response; another study [5] requested exposure be identified by one of several ranges in an attempt to diminish recall bias in the exact durations.

Controlling for confounders also important in evaluating the estrogen-AD relationship. Age is both a risk factor for AD and (estrogen loss from) menopause. If a study fails to control for age, then it will overestimate the true association. Although not explicitly stated in the text, this is most likely the case for the strong protective relationship (0.40) found by Tang et al [10]. Education has also been suggested as a possible confounder as some previous studies have named higher education as a protective factor against developing AD. Body weight has also been suggested as a possible confounding factor, as higher body weight increases the level of circulating estrogens and lower body weight is common among AD cases. The implications of education and body weight continue to be controversial, but most studies have nonetheless tried to control for education and, to some degree, body weight, in their analyses.

One major criticism of the study designs used to date in establishing this relationship is that women are voluntarily choosing to take exogenous estrogens. This may overestimate the relationship between estrogen and AD since these women are likely to be "healthy users.

To evaluate the true association of estrogen and AD, if any, a randomized, placebo-controlled intervention trial is necessary. However, ethical dilemmas exist in blindly assigning women to estrogen [9] due to its established benefits (decreased risk of osteoporosis and possibly coronary heart disease) and risks (endometrial cancer and possibly breast cancer). Assigning women to one of these conditions may be detrimental, depending on the woman's risk factors for these conditions and diseases.

Conclusions and Recommendations. A protective relationship of estrogen use on AD is suggested by the data, but not conclusively. The relationship is biologically plausible, prior research has demonstrated a temporal relationship, and several recent studies have shown protective associations, yet dose-response is inconclusive. Bias (especially misclassification through measurement error) and confounding have alternatively overestimated and underestimated the estrogen-AD relationship.

Future intervention trials to determine the true association between estrogen and AD would be ideal from an epidemiological point of view, as they can account for the Ahealthy user@ effect as well as differences in duration, dosage, regimen, and route of administration experienced by many studies. In the meantime, ongoing intervention trials which attempt to ascertain the relationship between hormone replacement therapies (estrogen and progestin) and all of their proposed effects, such as the ongoing Women's Health Initiative, must suffice. Future prospective studies may also yield useful and encouraging results, providing that they can adequately minimize or control for both confounding and bias.

References

[1] Brenner, D.E., Kukull, W.A., Stergachis, A., van Belle, G., Bowen, J.D., McCormick, W.C., Teri, L., & Larson, E.B. (1994). Postmenopausal estrogen replacement therapy and the risk of Alzheimer's disease: A population-based case-control study. American Journal of Epidemiology, 140, 262-267.

[2] Burns, A., & Murphy, D. (1996). Protection against Alzheimer's disease? [Commentary]. The Lancet, 348, 420-421.

[3] Henderson, V.W. (1997). The epidemiology of estrogen replacement therapy and Alzheimer's disease. American Academy of Neurology, 48 (Supplement 7), S27-S35.

[4] Henderson, V.W., Paganini-Hill, A., Emanuel, C.K., Dunn, M.E., & Buckwalter, J.G. (1994). Estrogen Replacement Therapy in Older Women. Archives of Neurology, 51, 896-900.

[5] Kawas, C., Resnick, S., Morrison, A., Brookmeyer, R., Corrada, M., Zonderman, A., Bacal, C., Donnell Lingle, D., & Metter, E. (1997). A prospective study of estrogen replacement therapy and the risk of developing Alzheimer's disease. American Academy of Neurology, 48, 1517-1521.

[6] Paganini-Hill, A. (1997). Does estrogen replacement therapy protect against Alzheimer's disease? Osteoporosis International, Supplement 1, S12-S17.

[7] Paganini-Hill, A., & Henderson, V.W. (1996). Estrogen replacement therapy and risk of Alzheimer's disease. Archives of Internal Medicine, 156, 2213-2217.

[8] Paganini-Hill, A., & Henderson, V.W. (1994). Estrogen deficiency and risk of Alzheimer's disease in women. American Journal of Epidemiology, 140, 256-261.

[9] Skolnick, A.A. (1997). Evaluating estrogen for Alzheimer's disease poses ethical and logistical challenges. Journal of the American Medical Association, 277, 1831-1833.

[10] Tang, M., Jacobs, D., Stern, Y., Marder, K., Schofield, P., Gurland, B., Andrews, H., & Mayeux, R. (1996). Effect of estrogen during menopause on risk and age at onset of Alzheimer's disease. The Lancet, 348, 429-432.