Estrogen Therapy in Postmenopausal Women:  Effects on Cognitive Function and Dementia FREE

DEMENTIA or marked loss of intellectual function is a common and important medical problem. In 1 prospective study of ambulatory, independent persons aged 75 to 85 years, new cases of dementia occurred as frequently as myocardial infarction and more frequently than stroke.¹ Approximately 10% of persons older than 65 years and almost 50% of those older than 85 years have dementia.² Alzheimer disease (AD), the most common cause of dementia, is estimated to affect 3.75 million people in the United States and to cost $67 billion annually.³ Despite the severity and prevalence of dementia, there are few effective treatments or prevention strategies.

Some studies have suggested that postmenopausal estrogen therapy might improve cognition in nondemented perimenopausal and postmenopausal women, prevent the development of dementia, or improve the severity of dementia. We reviewed the basic biochemical and neurophysiologic evidence to support the role of estrogen in cognition and reviewed the current literature on the effects of estrogen therapy on cognitive function and dementia.

We conducted a computerized MEDLINE search from January 1966 through June 1997, using the key words hormone replacement therapy, estrogen, estrogen replacement therapy, menopause, cognition, dementia, Alzheimer's disease, and memory, reviewed the bibliographies of identified articles, and consulted experts. We limited our review to peer-reviewed articles.

We first reviewed studies that provided evidence of possible biological mechanisms of the effect of estrogen on the central nervous system. We then located 40 articles containing primary data that addressed the effect of estrogen on cognitive function or dementia in women. After eliminating case reports and studies in premenopausal women, 27 studies were reviewed. Of these, 13 addressed the effect of estrogen use on cognitive function,^4- 16 10 addressed the effect of estrogen on risk for developing AD or other dementia,^17- 26 and 4 addressed the effect of estrogen treatment in women with AD.^27- 30 Our goal was to quantitatively summarize the results of these studies using formal meta-analytic techniques. However, studies of the effect of estrogen therapy on cognitive function and as treatment for AD used various designs and multiple different outcome measures, precluding quantitative summary of the data. In contrast, the predictor variable in each of the 10 studies of the effect of estrogen use on risk for dementia was previous or current postmenopausal hormone use (with or without concurrent progestin therapy) and the outcome was dementia, allowing us to perform a formal meta-analysis.

We performed separate meta-analyses of risk estimates for developing any dementia (AD and other dementia types)^17- 26 and for developing AD^17- 26 in estrogen users compared with nonusers. Studies that reported risk for AD based on criteria described by the National Institute of Neurologic and Communicative Disorders and Stroke–Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA)³¹ were summarized separately.^{17,19- 22,24- 26} For each study, we used the most adjusted risk estimate and 95% confidence intervals (CIs) to calculate the summary odds ratio (OR). When 95% CIs were not reported, we calculated them from the adjusted exact P value or χ² provided in the publication^17- 18 or obtained them from the author.²² In studies where the control group included men, the ORs and CIs were recalculated for women subjects.¹⁹ For case-control studies with more than 1 control group, we included only the comparison with community controls.¹⁸ When authors published 2 studies with overlapping subjects, we reviewed the original study.²³ The summary OR estimate and 95% CIs were calculated using a random-effects model and the general variance–based method.^32- 33 A test for heterogeneity was performed; P<.10 was considered statistically significant for this test.³⁴

There are estrogen receptors in the hypothalamus, the preoptic area, anterior pituitary, CA1 region of the hippocampus, and several other brain regions.³⁵ The role of these receptors in the function of the central nervous system is under investigation. How sex hormones may affect neuropsychologic function remains unknown, but several mechanisms have been suggested, including modulation of neurotransmitters, neurite and synapse reorganization, prevention of cerebral ischemia, and alterations in lipoproteins.

Cells in the basal forebrain, specifically the nucleus basalis of Meynert, send cholinergic projections to cortical, hippocampal, and amygdala regions. Degeneration of these neurons and their tracts has been shown to be one of the earliest and most pronounced neuropathologic changes in brains of patients with AD and age-related cognitive deficits.^36- 37 Estradiol administration to oophorectimized rats is associated with an increase in choline acetyltransferase and potassium-stimulated acetylcholine release in certain brain regions.^38- 40 Estradiol also prolongs survival of cholinergic neurons.⁴¹ Estrogen receptors colocalize with nerve growth factor receptors in cholinergic neurons in the rat basal forebrain.⁴² Estrogen-treated oophorectomized adult rats have superior performance on behavioral memory tasks compared with estrogen-deprived animals, and performance is associated with increased choline uptake and higher levels of choline acetyltransferase in the hippocampus and frontal cortex.⁴³ Furthermore, estrogen counteracts cognitive impairments induced by scopolamine (a cholinergic muscarinic receptor blocker) in oophorectomized rats.⁴⁴ These studies support the hypothesis that estrogen may, either independently or in synergy with other neurotrophic factors, improve cognitive function by promoting cholinergic activity in the brain.

Monoamine oxidase is a major enzyme in the central nervous system that metabolizes catecholamines, including serotonin, norepinephrine, and dopamine. Modulation of monoamine oxidase activity affects neurotransmitter activity in many neural pathways, including those involved in learning and memory. Abnormalities of the adrenergic, serotonergic, and dopaminergic systems have been described in AD.⁴⁵ Estradiol administered to oophorectomized adult rats results in decreased monoamine oxidase activity in the amygdala and basomedial hypothalamus.⁴⁶ Human studies have shown that estrogen treatment reduces serum and platelet monoamine oxidase levels^47- 48 and displaces tryptophan, the precursor of serotonin, from plasma albumin binding sites, possibly allowing more free tryptophan for central nervous system conversion to serotonin.^49- 50 In female rats, estradiol administration induces serotonin receptors in forebrain regions involved in cognition and behavior, including the frontal lobe, cingulate, and nucleus accumbens.⁵¹ The possible involvement of monoamine oxidase B inhibition with attendant increased adrenergic activity as a treatment for AD has been suggested by a recently published placebo-controlled, double-blind study in which selegiline (a monoamine oxidase B inhibitor) showed a modest beneficial effect on certain dementia-related outcome measures.⁵² Though the true value of such therapy is under debate,⁵³ increased central adrenergic tone may be a mechanism whereby estrogens exert a portion of their therapeutic benefit.

Early neuron loss in the CA1 region of the hippocampus, a region associated with memory and learning, is found in patients with AD and age-related cognitive dysfunction. Decreased hippocampal pyramidal cell density, a change seen in AD and other dementias, has been correlated with verbal memory impairment.⁵⁴ Estrogen regulates synaptic plasticity by stimulating axonal sprouting and dendritic spine formation in the adult rat hypothalamus and CA1 hippocampal pyramidal neurons.^35,55 Removal of circulating gonadal steroids by oophorectomy results in a profound loss of dendritic spine density in CA1 pyramidal cells in rats.⁵⁶ Thus, at least in rats, estrogen plays a role in maintaining or promoting neuronal circuitry in several brain regions important in cognition.

Endogenous estrogens may protect against cerebral ischemia⁵⁷ by inducing central nervous system vasodilatation, reducing central arterial smooth muscle injury response, or reducing platelet aggregation.^58- 60 Estrogen therapy protects rodent neurons against oxidative stress, excitotoxins, and β-amyloid–induced toxicity in cell culture.^61- 62 Postmenopausal estrogen therapy reduces serum low-density lipoprotein cholesterol and increases high-density lipoprotein cholesterol.⁵⁸ These favorable lipoprotein changes may slow progression of cerebral atherosclerosis and prevent vascular dementia and other cognitive decline.

The apolipoprotein E isoform ∊4 has recently been identified as a major biological risk factor for AD and preclinical cognitive decline.^63- 64 Estrogen modulates the expression of the apolipoprotein E gene in rodent tissues and theoretically could reduce risk of AD in humans via apolipoprotein E alterations.^65- 66 Physiologic concentrations of estradiol in tissue culture promote the nonamyloidogenic metabolism of the amyloid precursor protein, which gives rise to the β-amyloid peptide.^67- 68 Thus, estrogens may reduce the risk of AD through a variety of mechanisms.

Five observational studies have evaluated the association of estrogen therapy and cognitive performance (Table 1). Two of the studies reported inconclusive results,^5- 6 2 reported no association between estrogen and cognitive performance,^4,8 and 1 found that estrogen use improved cognitive function.⁷ In a cross-sectional study, Kampen and Sherwin⁵ administered 14 cognitive tests to 71 healthy, recently postmenopausal women and found that scores on only one of the tests (Paragraph Recall) were significantly higher in women taking estrogen compared with nonusers. There were no other differences between the estrogen users and nonusers, and there was no association of test scores with serum estrogen level. The analyses were not adjusted for education or depression. If estrogen users are better educated or less depressed than nonusers, these factors could account for the observed findings. Robinson et al⁶ conducted a cross-sectional study of 72 postmenopausal estrogen users and 72 nonusers matched for age and education to examine the association of estrogen use and performance on 2 recall tests. Women treated with estrogen performed significantly better on proper name recall but not on word recall compared with those not receiving estrogen.⁶ Kimura⁷ assessed performance on 10 cognitive tests and 1 mood scale in 21 postmenopausal women receiving estrogen replacement and 33 postmenopausal women not taking estrogen. Kimura concluded that the estrogen group had better overall performance on the cognitive tests, but no specific test scores or statistical analyses were reported. In the only prospective cohort study to date, Barrett-Connor and Kritz-Silverstein⁴ determined postmenopausal estrogen use among 800 white, upper-middle-class elderly women. Fifteen years later, cognitive function was measured using a battery of 12 cognitive function tests. Age- and education-adjusted comparisons showed no difference between users and nonusers, and no effect of dose or duration of estrogen use. Paganini-Hill and Henderson⁸ conducted a nested case-control study of 214 elderly upper-middle-class women to assess the association of estrogen replacement therapy and performance on a visual-spatial task, the Clock-Drawing Task. There was no statistically significant difference in percentage of normal and abnormal clock drawers in the estrogen users compared with the women not using estrogen.

Observational studies of the association of estrogen use and cognitive performance are susceptible to confounding, especially by such strong predictors of cognitive performance as age, education, and depression.⁶⁹ All of the studies described above either matched or statistically adjusted for age; only the findings of Barrett-Connor and Kritz-Silverstein⁴ and Robinson et al⁶ were matched or adjusted for education, and no study controlled for depression. Three of the studies compared estrogen users with nonusers based on the results of multiple cognitive tests.^4- 5,7 Such multiple comparisons between users and nonusers substantially increase the probability of obtaining at least 1 positive result by chance alone.

Seven randomized controlled trials and 1 nonrandomized controlled trial of the effect of estrogen therapy on cognitive function have been published (Table 2). ^9- 16 Six of these 8 trials concluded that estrogen therapy improved cognitive function.^9,11- 14,16 However, substantial methodologic problems make the results of these trials difficult to evaluate. The trials are small, including only 18 to 64 subjects. Two of the older trials used neuropsychologic tests that are not validated or commonly used.^9,12 Two of the trials concluded that estrogen use improved cognitive function, despite the fact that there was improvement in only 1 or 2 of multiple cognitive tests.^9,16 Participants in 5 of the 6 studies that reported benefit from estrogen therapy included many recently menopausal women who had menopausal symptoms.^11- 14,16 Relief of vasomotor symptoms and insomnia might have resulted in improved cognitive performance. A recent placebo-controlled trial in asymptomatic postmenopausal women found little or no effect of estrogen therapy on cognitive test scores.¹⁵ Finally, several of these trials reported that scores on cognitive tests improved after treatment with estrogen, but did not compare these changes in the estrogen-treated group with results in the placebo-treated group.^11,15- 16 In summary, all these trials have substantial methodologic problems. While there does appear to be evidence that estrogen therapy may improve cognitive performance in postmenopausal women with symptoms, there is no clear evidence of a beneficial effect in asymptomatic women.

The suggested beneficial effects of estrogen on cognition in normal postmenopausal women led to the hypothesis that estrogen deficiency associated with menopause may contribute to the development of dementia, including AD. In support of this theory, serum estrogens have been found to be lower in women with AD than in age-matched controls,²⁸ and women may be at increased risk of developing AD compared with men.⁷⁰ There have been 8 case-control studies (Table 3) and 2 prospective cohort studies (Table 4) conducted to evaluate the association between AD and other dementias and postmenopausal estrogen therapy. The results of these studies have ranged from suggesting a protective effect of estrogen on developing AD to suggesting an increased risk of developing AD (Figure 1). One of the case-control studies²² and the 2 prospective studies^25- 26 found a statistically lower risk of developing dementia in postmenopausal women who had taken estrogen compared with those who had not. Of the 7 other case-control studies, 2 showed a nonsignificant increased risk of dementia among estrogen users compared with nonusers,^17- 18 2 showed no difference in the risk of dementia,^20- 21 and 3 found a nonsignificant decreased risk of dementia among estrogen users compared with nonusers.^19,23- 24 All these observational studies either matched cases and controls for age or statistically adjusted for age, but most were not adjusted for education, and none was adjusted for depression. The 1 case-control study that found an association between estrogen use and reduced risk for dementia found this association only for current estrogen users compared with nonusers (OR, 0.3; 95% CI, 0.1-0.7).²² Current use of estrogen might appear to be protective if physicians stop hormone therapy in women who develop cognitive dysfunction or dementia. In addition, retrospective studies might find an association between estrogen use and reduced risk for dementia if women with dementia do not report using estrogen in the past because they cannot remember. Two case-control studies that addressed this problem by using pharmacy records or a verified medication history reported conflicting results.^21- 22 Prospective studies can also avoid recall bias and the bias that may result from different prescribing practices in dementia patients. Both prospective cohort studies found a decreased risk of AD in estrogen users. In one of the prospective studies, estrogen use was recorded among 1124 elderly community-dwelling women who were evaluated for dementia 1 to 5 years later. Women who had ever used estrogen had a 50% reduction (adjusted OR, 0.5; 95% CI, 0.25-0.9) in risk of developing AD.²⁵ The other prospective study found a similar reduction in risk of developing AD (adjusted OR, 0.46; 95% CI, 0.21-1.0) among 472 postmenopausal women enrolled in the Baltimore Longitudinal Study of Aging who were followed up for up to 16 years.²⁶ Both prospective studies and 1 of the case-control studies examined whether increasing duration of estrogen use was associated with increasing protection of developing dementia and found conflicting results.^23,25- 26

To provide a more precise estimate of the effect of estrogen therapy on risk of developing dementia, we performed a meta-analysis of the results of the observational studies described in Table 3 and Table 4. When the results of all studies in which the outcome was any type of dementia (AD and other dementia types),^17- 26 the summary OR was 0.71 (95% CI, 0.53-0.96; P for heterogeneity=.10) (Table 5). When the results of the 10 studies of AD (based on any criteria) are included in the meta-analysis,^17- 26 the summary OR was 0.71 (95% CI, 0.52-0.98; P for heterogeneity=.11). There was significant heterogeneity (P=.07) in the findings of the 8 studies that used the NINCDS-ADRDA criteria for diagnosis of AD.^{17,19- 22,24- 26} Because heterogeneity might have been attributable to study design (case-control vs cohort studies), we conducted a separate analysis of the case-control studies and prospective studies. For the 8 case-control studies, the summary OR for any dementia was 0.79 (95% CI, 0.56-1.12; P for heterogeneity=.09); for any AD diagnosis, the OR was 0.8 (95% CI, 0.56-1.16; P for heterogeneity=.11); and for NINCDS-ADRDA criteria for AD diagnosis, the summary OR was 0.8 (95% CI, 0.56-1.12; P for heterogeneity=.09). The summary OR for the 2 prospective studies, both of which used NINCDS-ADRDA criteria for AD diagnosis, was 0.48 (95% CI, 0.29-0.81; P for heterogeneity=.86).

Epidemiologic evidence suggesting that estrogen use may prevent development of AD and improve cognitive function has sparked interest in estrogen as a treatment for AD. There have been 4 small trials of estrogen therapy in women with AD (Table 6). In 2 uncontrolled trials that each included 7 participants, severity of dementia was measured at baseline and after 6 weeks of estrogen therapy.^27- 28 Each of these studies found improvement on some, but not all measures of dementia severity after treatment. Because there was no control group for comparison, this improvement may represent a practice or learning effect. In addition, these 2 trials were not blinded, which may have biased measurement of outcome. In another trial, 15 women with AD were treated with estrogen, and change in dementia severity was compared with 15 untreated controls matched for age and dementia severity at baseline.³⁰ The treatment group improved on 1 cognitive scale and on 1 dementia rating scale compared with scores prior to treatment, while the control group did not improve, but changes in the severity of dementia in the estrogen-treated group were not compared with changes in the untreated group. In the only placebo-controlled trial, Honjo et al ²⁹ randomly assigned 14 women with AD to receive estrogen (1.25 mg/d of conjugated oral estrogen) or placebo and found greater improvement on 1 dementia scale but not on 2 others in the estrogen-treated compared to the placebo-treated group.²⁹ All 4 of these studies are limited by small sample size, short duration of therapy (3 to 6 weeks), and the inclusion of subjects with a wide range of dementia severity.

There are plausible biological and neurophysiologic mechanisms that might account for a beneficial effect of estrogen therapy on cognition, a reduced risk for developing dementia, or improvement in the severity of dementia. Actual studies in women, however, have substantial methodologic problems and have produced conflicting results.

The largest and most methodologically sound observational study of the effect of estrogen use on cognition in nondemented women showed no benefit.⁴ The results of 8 small trials of estrogen therapy in nondemented women are inconclusive.^9- 16 Many of these trials included primarily recently menopausal women who are likely to have substantial menopausal symptoms; others showed improvement on only 1 or 2 multiple cognitive tests, and several did not compare changes in cognition in the estrogen-treated group with the placebo group. In summary, there is evidence that estrogen therapy improves cognitive performance in recently menopausal women, but no evidence of a beneficial effect in asymptomatic women.

Our meta-analysis of the observational studies of the effect of estrogen therapy on risk for developing dementia suggests a 29% decreased risk among estrogen users. If true, this risk reduction would be of major public health importance. Summary findings from meta-analysis, however, are only as reliable as the findings of the individual studies. Unfortunately, the observational studies that we summarized are susceptible to confounding and compliance bias. For example, women who choose to take estrogen have been reported to be better educated and healthier than nonusers,^71- 74 differences that may result in a lower risk for developing AD.^75- 76 In addition, compliance with assigned medication is a marker for reduced risk of various disease conditions^77- 80 and may also account for the reduced risk of AD in women using estrogen. In summary, the evidence from these observational studies is weak, and large, controlled, blinded trials are necessary to determine if estrogen therapy can reduce the risk of developing AD. The Women's Health Initiative Randomized Trial, which is currently enrolling participants, will include the Women's Health Initiative Memory Study. This ancillary trial will determine the effect of hormone replacement therapy on cognitive function and risk for developing AD and other dementias among approximately 8000 postmenopausal women treated for 10 years.

Should estrogen therapy be used as treatment in persons with AD? Only 58 women have been studied in trials of estrogen therapy for treatment of AD, and most of these trials were uncontrolled, unblinded, and nonrandomized. Only 1 randomized, placebo-controlled, blinded trial has been published. It was conducted in 14 women with AD who were randomly assigned to receive 1.25 mg/d of conjugated estrogen or placebo daily for 3 weeks. Estrogen-treated women improved on 1 severity scale, but not on 2 others. Based on the weakness of this evidence and the potential adverse effects of estrogen therapy (endometrial abnormalities,⁸¹ gallbladder disease,⁸² venous thromboembolic events,^83- 86 and breast cancer^87- 89), we do not believe that estrogen should currently be used to treat women with AD. However, given the lack of available treatment options for women with AD and the suggestive preliminary findings of benefit, we believe that a large randomized trial of estrogen therapy should be completed as soon as possible. Whether estrogen therapy might improve other forms of dementia, such as vascular dementia, should also be explored. Finally, since progestins are typically added to the estrogen regimen in women with a uterus, the role of this hormone must also be evaluated in future studies.