Abstract
As we age, most of us experience a certain degree of cognitive decline. In most cases, this decline is gradual. However, in some cases, cognitive impairment is so severe that it can be classified as dementia and this impacts greatly on activities of daily living. Alzheimer's disease, the most common form of dementia, has been linked to the reduction in estrogen levels that comes with aging. More specifically, many researchers have hypothesized that estrogen, and hence estrogen replacement via hormone therapy, could protect against cognitive decline in women. However, recent randomized, controlled trials did not reflect this. In fact, some reports showed that hormone therapy could have detrimental effects on cognitive function in older postmenopausal women. The most publicized of these has been the Women's Health Initiative Memory Study. Studies have thus yielded conflicting results and conclusions. The reasons for this may be due to a number of factors, such as the age of participants, the time of hormone therapy onset (‘window of opportunity’ theory), type of treatment, type of menopause (surgical or natural) and, possibly, genetic risk factors. We performed quantitative and qualitative meta-analyses and reviewed each of these factors in detail. The future may lie in combining these factors in order to fully understand the potential mechanisms behind estrogen and its effect on cognition.
We live in a world where people have increasingly longer, but not necessarily healthier, lives. As we age there is a discernible decline in some cognitive functions, such as verbal memory, speed of information processing and complex spatial and verbal skills. The hippocampus, which is particularly important to memory function, loses on average 5% of its volume every decade after the age of 65 years [1]. In pathological cognitive aging, such as dementia, people experience a more pronounced decline of cognitive function, which impacts on activities of daily living. The most common form of dementia is Alzheimer's disease (AD). AD is a progressive dementia that usually manifests initially with memory deficiencies, one of its earlier and more pronounced symptoms, which are later followed by other cognitive deficits [2]. The speed and degree of cognitive decline shows great variability between individuals, ranging from ‘successful’ aging to more severe cognitive problems, such as dementia [3]. What determines this variability is not certain, but women may have a greater risk of developing AD than men [4]. This could be attributed to the fact that women reach an older age than men and AD shows a greater incidence with age. However, the age-specific incidence of AD was also reported to be higher in women than in men [5]. This distinction could be linked to the more pronounced deficiency in sex hormones, such as estrogen, in women after the menopause. Owing to its biological plausibility to protect the brain, deficiency in estrogen after the menopause has been suggested as a potentially important factor in the development of dementia. Therefore, it was believed that hormone therapy (HT) would be a beneficial intervention for treating or preventing dementia in older women [6,7].
Historical background
The relationship between sex hormones (both endogenous and via HT) and cognitive function has been under extensive investigation over the last three decades. Results of many animal and cell-culture studies have repeatedly shown potential protective effects of estrogen on the brain. For instance, synaptic plasticity was shown to be enhanced through binding of estrogen to its receptors in the hippocampus and nucleus basalis of Meynert, areas which are both implicated in AD [8]. In animal models, estrogen has been found to increase dendritic density in the hippocampus, which has the highest concentration of estrogen receptors (ERs) [9]. In affected AD brains, the hippocampal region is severely afflicted, suggesting that the biological effects of estrogen in this area of the brain could perhaps preserve cognitive functioning by having an impact on cognitive decline prior to the onset of AD. It has been found that estrogen use preserves regional cerebral metabolism, protecting against metabolic decline in postmenopausal women, again especially in the areas of the brain found to be adversely affected in AD [10]. Detailed reviews of the possible effects of estrogen in protecting the brain are already available [11,12]. In fact, it was stated that there are almost innumerable biological reasons why estrogen could be protective against both benign memory loss and AD [13].
The nineties showed great optimism regarding the benefits of estrogen therapy in the treatment of AD, encouraged by an increasing number of positive observational studies and a number of small treatment trials, which provided evidence supporting the potentially beneficial biological effects of estrogen. The basic biological mechanisms, as well as the supportive evidence from human observational studies, have been extensively reviewed [14–17]. The theoretical basis behind the idea that estrogen could protect against AD was summed up as follows [18]:
There is a positive relationship between AD and decreasing estrogen levels after menopause;
A sudden earlier loss of gonadal function seen in postmenopausal women compared with men is associated with an increased prevalence of AD in women;
There is generally decreased incidence and delay in the onset of AD in observational studies of women using HT after the menopause [18].
While these arguments follow a logical system, they are, unfortunately, not always supported by research. First, in contrast to initial findings, several well-controlled observational studies actually reported higher levels of estrogen in women with AD compared with controls [19]. Second, in contrast to European and Asian studies, observational studies in the USA did not find increased incidence of AD in elderly women compared with age-matched men [20], and men have often been found to show a faster age-related cognitive decline than women [19]. Last, more recent, but also several earlier, observational studies did not show protective associations of HT use against AD [19]. Importantly, several sources of bias (e.g., healthy user bias and recall bias) exist in these observational studies, which also limits their predictive value [19].
These counter arguments are thus concerned with observational studies. However, there remains an abundance of evidence from in vitro and in vivo studies suggesting that estrogen could act favorably upon almost all mechanisms known to be affected in cognitive decline and AD [21]. Prior to recent findings, the idea that HT could potentially reduce the risk for dementia in women was widely credited and accepted. Well-controlled treatment studies ultimately provide the most compelling evidence. Most of the early treatment studies showed favorable effects of HT, in line with biological plausibility studies [15]. However, several of the initial treatment trials had methodological problems (e.g., size and statistical analyses), which had already earlier led to skepticism about their results [22]. In addition, many better-controlled treatment studies performed in recent years have yielded opposing results. For example, several recent large, well-controlled, randomized, controlled trials (RCTs) showed that HT did not prevent cognitive decline or dementia and did not improve cognitive abilities in postmenopausal women with or without dementia [23–25].
Women's Health Initiative Memory Study
The largest and best known of these RCTs to date has been the Women's Health Initiative-Memory Study (WHIMS) [26]. The WHIMS was a large, multicenter, randomized, double-blind, placebo-controlled trial in which a subgroup of several thousand women from the Women's Health Initiative (WHI) study were assessed for the effects of HT on incidence rate of dementia and mild cognitive impairment. Participants were aged 65 or older (and were thus all postmenopausal), and received conjugated equine estrogens (CEE) plus medroxyprogesterone (MPA) versus placebo [27] or continuous unopposed CEE versus placebo [28]. The combined trial arm was discontinued in 2002 after concerns arose regarding its safety (mainly an increased risk of breast cancer). The CEE and MPA trial found, contrary to the earlier wealth of scientific evidence (which hailed HT as having a positive influence on cognitive function), that the widely used Prempro™ increased the risk for dementia [34]. Several scientists concurred with the WHIMS authors that HT should not be recommended as an effective preventive treatment against dementia [29]. A smaller subset of WHIMS participants receiving combined treatment (n = 1417) showed some positive effects on visual memory, but only after 3 years of treatment [30]. This demonstrates, once again, the variable results of HT and cognition studies.
In February 2004, the estrogen-only (CEE) arm of the WHIMS (WHI) was also discontinued due to an unacceptable increased risk of stroke in the treatment group. The results were published in April of the same year [28,31]. Similar to the findings of the WHIMS CEE and MPA trial, it reported an increased risk of dementia onset, although it was of a smaller magnitude [31]. Our meta-analyses of cognitive function (measured with the Modified Mini-Mental State Examination [3MSE]) show an overall effect in favor of placebo (z = 2.04; p < 0.05; mean difference: −0.15; 95% confidence interval [CI]: −0.29 to −0.01 for the combined trial and z = 2.49; p < 0.0005; mean difference: −0.38; 95% CI: −0.60 to −0.17 for CEE alone). However, we performed individual analyses per year, which showed inconsistencies in risk over time in both studies. While the effect of the combined trial only showed a significant difference in favor of the placebo group at year 4 (z = 2.26; p < 0.05), this was seen after CEE alone only at year 1 (z = 2.25; p < 0.05), when taking into account the number of women who dropped out each year.
It is not clear why negative effects of HT on neurocognitive health were found in the WHIMS. Lacunar infarcts and high-grade deep white matter lesions are risk factors for dementia [32,33]. It was suggested that multi-infarcts may have played a role in the results reported, as MPA has been associated with an increased risk of vascular disease [19]. However, in the WHIMS, the presence of cardiovascular disease or hypertension at baseline did not confound study analyses. In addition, controlling for infarcts in post hoc analyses by excluding women with stroke did not alter the association between HT and dementia, and the effect remained when only AD cases were selected for analyses [34]. However, it remains to be investigated whether microvascular events could have played a mediating role in the increased dementia risk [34].
The results of the WHIMS are a far cry from earlier positive reports of the potential effects of HT. It has been argued that there may be some important reasons for the negative results found in the WHIMS, as well as some rationale behind the general discrepancy seen in the field as a whole. It has been suggested that differences in the age of participants (resulting in the ‘window of opportunity’ theory [35]), type, form and route of treatment, and type of menopause that participants had undergone (surgical or natural) are mostly responsible for the lack of uniformity in results. These factors shall be addressed in more detail below.
Factors explaining the variability in outcomes between studies
Age
Some researchers have argued that the negative results found in the WHIMS were partially due to the age of participants [36]. The advanced age (65+ years of age) and obesity of participants in this study was deemed not representative of the population that HT is aimed at. The majority of participants may have been beyond the scope of help in averting most of the negative outcomes (e.g., stroke and dementia). However, that would not explain the ‘negative’ effects seen in the WHIMS. Another review reported little support for beneficial effects of estrogen (both alone and in combination with progesterone) in women older than 65 years of age and it was noted that potentially beneficial effects on specific cognitive functions were mainly seen in younger and more recently menopausal women [37]. In a recent study, the relationship between HT use and a lowered risk of developing AD in 971 postmenopausal women showed a protective association, but only in women of the youngest age tertile (50–63 years) [38]. This finding is in line with animal studies suggesting protective brain effects in young, but not older, mice [39].
‘Window of opportunity’ theory
Several authors have thus suggested that there is a critical period for HT treatment in order to obtain positive effects on the brain. It is hypothesized that initiating or continuing HT beyond this critical period would have little effect on the brain and cognitive function [35,38,39]. The ‘window of opportunity’ theory suggests that there is a critical time for the initiation of HT and could explain why no protective effects were seen in the WHIMS. Results from animal studies suggest that the longer the delay between ovariectomy and onset of treatment, the less chance there is of detecting the favorable effects of HT on the brain [40].
However, the strongest argument against this theory is that in RCTs of women with AD, both Premarin® and estradiol (E2) were also seen to have positive effects, but only up to 2–3 months [41]. These women with AD (of whom the majority were aged over 65 years) were at least 15–20 years older than the recently menopausal women (with a mean age of 48 years) in the successful E2 trials described in another meta-analysis [19].
This meta-analysis suggested that positive effects of E2 in women without dementia were also time-limited, resulting in an increase in memory, accuracy and abstract reasoning functions, but again only up to 2–3 months. For both older and younger more recently menopausal women, a possible reversal of positive effects was seen on some cognitive functions after 1 year of treatment [19].
Therefore, it seems more likely that positive effects, if they were found, are only short-lived, regardless of age. These findings also tie in with the WHIMS data.
Differential effects of HT on cognitive function: effects of treatment & age
It has often been suggested that the effects of HT are limited to particular aspects of cognitive function, which could be an alternative reason for some studies not finding any effects. It is likely that not all cognitive domains are affected equally by sex steroids. However, results from studies show inconsistencies in findings. One meta-analysis found that the most common cognitive domain found to be affected by HT was memory, specifically verbal memory [42]. Data from many cross-sectional studies suggest this indeed to be the case [43–47], although RCTs and cohort studies report conflicting results [15].
For this review, we performed quantitative and qualitative meta-analyses using RevMan software provided by the Cochrane library standardized review system, employing both their inclusion criteria and their statistical methods (update) (
Hormone replacement therapy for cognitive function in postmenopausal women: characteristics of included studies.
Randomization key: A: Adequate randomization; B: Unclear randomization.
3MSE: Modified Mini-Mental State Examination; BETAM: BErlin Test for Associative Memory; BNT: Boston Naming Test; BVRT: Benton Visual Retention Test; B–W: Bourdon–Wiersma test; CAMCOG: CAMbridge dementia examination COGnitive tests; CANTAB: CAmbridge Neuropsychological Test Automated Battery; CEE: Conjugated equine estrogen; CRT: Complex reaction time; CVLT: Californian Verbal Learning Test; DR: Delayed recall; DSM: Diagnostic and Statistical Manual of Mental Disorders; DSST: Digit Symbol modalities or Substitution Test; E2: Estradiol; E1: Estrone; E3: Estriol; ED: Extradimentional; HRT: Hormone replacement therapy; ID: Intradimensional; i.m.: Intramuscular; IR: Immediate recall; KVT: Konzentrations Verlaufs Test; MPA: Medroxyprogesterone acetate; MMSE: Mini-Mental Status Examination; MWT: Multiple-choice word test; PAL: Paired associate learning; PASAT: Paced Auditory Serial Addition Test; PMA: Primary mental abilities; RT: Reaction time; SD: Standard deviation; SRT: Simple reaction time; STM: Space-time modulation; T: Testosterone; TMT: Trial Making Test; USTM: Unitary space–time modulation; VRT: Visual reaction time; WAIS-R: Wechsler Adult Intelligent Scale – Revised; WCST: Wisconsin Card-Sorting Test; WHIMS: Women's Health Initiative-Memory Study; WMS-R: Wechsler Memory Test – Revised.
Qualitative meta-analysis of effect of hormone therapy on different aspects of cognitive function in postmenopausal women.
+; Positive result; -: Negative result; X: No effect. For example: (2+, 1X)/3 means there were two positive results and one no effect found out of three tests; (1X)/1 means that one test found no effect out of one test.
This author's report found one positive effect while our meta-analysis found none.
This author's report found no positive effects while our meta-analysis found one.
MMSE: Mini-Mental State Examination.
Qualitative meta-analysis of effect of hormone therapy on different aspects of cognitive function in postmenopausal women.
This author's report found one positive effect while out meta-analysis found none.
This author's report found no positive effects while our meta-analysis found one.
CEE: Conjugated equine estrogens; CO: Crossover; E1: Estrone; E2: Estradiol; E3: Estriol.
Summarizing these results, our meta-analysis shows that E2 given orally or as a bolus injection (but not transdermally) was mainly effective in improving verbal memory, but only in very recently menopausal younger symptomatic or surgically menopausal women. This could be explained because these highly symptomatic women had fewer hot flashes, and better sleep and mood after treatment, which could have affected their performance. However, if this were the case, perhaps overall better cognitive performance would be expected on a broad range of cognitive domains [42]. By contrast, it seems only some domains are affected by HT. In addition, this positive effect on verbal memory was only tested up to 3 months. Several earlier studies suggested that the effects of E2 on memory actually reversed after 6–12 months of treatment [19]. The positive effect was also small, as it only held-up in meta-analyses for one aspect of one test of verbal memory.
For visual memory, the effects were not clear and might again be modified by age. We saw some positive effects on visual memory after E2 and CEE and MPA, but only in older women. This may argue against the ‘window of opportunity’ theory.
Similarly, the results for tests of speed of information processing indicated that the successful type of estrogen treatment could depend on age: the younger symptomatic group seemed to profit most from an oral or bolus injection of E2 on these tests, while the older women could possibly profit from transdermal E2 or CEE and MPA, which both render lower levels of estrogen [19].
The only significant effect on accuracy and abstract reasoning was found in one study of surgically menopausal women after a bolus injection. Importantly, none of the meta-analyses showed significant heterogeneity and the more conservative standardized mean differences with random effects (rather than weighted mean differences with fixed effects) were used throughout.
These analyses thus showed that possibly not all cognitive tests are equally sensitive to HT and that different treatments may show different cognitive effects, which may also be modified by the age of participants tested [19]. Of course, the WHIMS, as the largest study that included older women, reported that women who used CEE and MPA, and CEE alone, had less improvement on the 3MSE and a greater risk of dementia, after 2 and 3 years respectively, although the effects were small [31,34]. However, as also mentioned, while a decrease in verbal memory was seen, the WHIMS substudy actually reported improvements in visual memory after combined HT at the same time point [30].
Hence, a current theory is that estrogens could exert a combination of beneficial and harmful effects and this may be modified by age [30].
Treatment form, route & duration
Whether HT given for a longer period of time to older women could actually increase dementia risk has been reviewed by others and several possible mechanisms for this relationship were suggested [49]. These mechanisms were based on the theory that a longer life span, a lower prevalence of vascular dementia and lower levels of testosterone in women may contribute to the higher prevalence of AD. The mechanisms that may have been active in WHIMS and potentially responsible for an increase in dementia could be:
Qualitative and quantitative meta-analyses of studies investigating the effects of hormone therapy on different aspects of cognitive function.
Qualitative analyses using
Quantitative analyses showed an overall positive effect on only one test, the verbal paired associates immediate recall (z = 2.40; p < 0.05; effect size: 1.02; 95% confidence interval: 0.19–1.85) [81,82], but not on the delayed recall or recall of word lists or stories or on verbal fluency (which also tests language and executive functions). These analyses may be limited as standard deviations (SDs) of the mean difference often had to be recalculated, which, in smaller studies, could lead to an overestimate of the SD, making these analyses vulnerable to type II error. However, it does indicate that effects, if present, are small.
In visual memory tests (where verbalization may occur), positive effects of E2 were seen on the memory of faces in women aged 74 years (after 5 months of oral E2) [86]; on picture recall and visual paired associates in women aged 65 years (3 weeks E2 transdermal) [90]; and also on figure recall after 3 years of CEE and MPA therapy in women aged 74 years [30]. A total of 15 studies reported no effects on visual memory regardless of age or type of treatment [25,81–83,87,88,91–93]. Differences in tests could explain the differences between studies, but quantitative meta-analyses showed no significance for tests of heterogeneity. These analyses also revealed no overall significant effect of HT on visual memory. Traditionally, effects of estrogen have been thought to be most apparent on verbal memory tests [94].
However, several studies reported positive effects of HT on speed of information processing (found on a third of all tests used in studies). This was the case in five studies using transdermal E2 in women aged 65 years [90]; transdermal E2 with women aged 57 years [84]; oral E2 with women aged 57 years [95]; transdermal E2 with women aged 71 years [85]; and bolus E2 with women aged 45 years [57], although several other tests used in studies with transdermal E2 could not reflect this [84,85,87,90,95]. Three studies using CEE reported no effects on these tests [89,91,96].
Quantitative meta-analyses found overall trends on simple tests (Trail Making Test [TMT] Part A and Speech Reception Threshold test; p = 0.06) and more complex tests (TMT Part B and Stroop test; p = 0.10). Surprisingly, for simple speed, these analyses involved studies with CEE and MPA for 9 months [89] and transdermal E2 for 24 weeks [85], both in older women. For complex speed, positive effects were seen in quantitative meta-analyses, again with CEE and MPA [86], after transdermal E2 in older women for 2 weeks [87] and 3 weeks [90], and also after oral E2 in younger women for a duration of 10 weeks [84]. Other tests, such as the cognitive reflection test and the Digit Vigilance and Digit Symbol Substitution tests, were not affected by HT.
Several other tests showed positive effects, but only after bolus injections of E2 in surgical menopausal women [57] on executive functions (abstract reasoning) and accuracy or after oral E2 in symptomatic young women [95]. Meta-analyses showed that there was no overall significant effect of HT on these types of tests (p = 0.14; p = 0.16) Three studies showed no overall effect of HT on the Mini-Mental State Examination (MMSE) or the cognitive and self-contained part of the CAMbridge dementia examination COGnitve tests [25,86,97] and negative effects were found twice after HT on the modified MMSE in the WHIMS [27,28]. However, in subanalyses [25], elderly women without stroke who had a normal MMSE score (28–30) were found to decline significantly less after E2 treatment compared with placebo after 3 years (z = 2.37; p < 0.005).
Use of MPA and estrone (E1), which is a major metabolite of CEE, may have different effects on neuronal and cerebrovascular function than progesterone and E2 (the most potent estrogen);
A greater risk of stroke (caused by the use of MPA [50] or through accidental inclusion of those at risk for stroke) leading to dementia;
Last, a decrease of bioavailable testosterone (through the increase in sex hormone binding globulin [SHBG], which CEE induces to a greater extend than E2 does [51]). Testosterone can, directly or through conversion into E2, protect against AD [52].
Arguments against these mechanisms include the following. First, as we saw in the meta-analyses, while only E2 had positive effects on verbal memory in relatively younger postmenopausal symptomatic women, small positive effects of CEE with or without MPA were seen on some cognitive tests, but only in older women.
In addition, exclusion of women with stroke, in the CEE alone or CEE with MPA trial, did not substantially alter the negative effect of HT on the overall improvement on the 3MSE in all arms [28].
Last, while SHBG may increase with CEE, so would the amount of estrogen when treating with CEE. The role of bioavailable testosterone and E2 (and the ratio of testosterone:E2) on cognitive function in elderly women is not well understood and requires further investigation. It is unclear whether CEE metabolites would have worse effects than E2, but brain ERs are more sensitive to E2 than E1 [53]. In addition, in a recent RCT focusing on event-related potentials in women with an average age of 60 years, a positive effect of CEE was found [54]. The results showed a shortening of P300 latency, which is consistent with normalization of cognitive function. This result casts further uncertainty into the understanding of the role of CEE in cognitive treatment. One observational study reported that higher E1 levels were associated with detrimental effects on cognition. In the Study of Osteoporotic Fractures, women in the highest E1 quartile showed worse scores than women in the lower E1 quartiles on two cognitive tests [55].
By contrast, the majority of observational studies showed a protective association (although tainted by several potential confounds, such as the healthy user bias [19]). Most women in the USA-based studies would have used CEE (in the form of Prempro and Premarin) and would have thus been exposed to high levels of E1 metabolites.
As mentioned above, CEE (similar to E2) has also shown positive effects on cognition in women with dementia [48] and, as our most recent meta-analyses suggest, also on some tests in older women without dementia.
Taken together, these findings suggest that, as well as the form of HT, the route of administration may also be of importance and may depend on the age of the woman. Where oral and bolus injections of E2 with or without progesterone may be more suitable for relatively recently menopausal women, transdermal E2 and possibly CEE could be more suitable for older women. However, it is not clear whether and which positive effects, if they exist at all, reverse after longer periods of usage.
Cyclical versus continuous administration
Following on from the influence of duration of treatment, it has also been suggested that continuously elevated levels of estrogen could potentially have long-term negative effects on cognition. A similar pattern has been seen in estrogen studies, suggesting that effects were positive up to 3 months of hormone treatment, stabilized at 6 months and then showed a decline after 1 year [19]. Adding progesterone does not seem to substantially alter results (see also [19]). A possible reason for the negative effects of continuous high-level estrogen treatment in elderly women [27,28] could be downregulation of receptors [56]. A potential way to prevent this could be by using an altered treatment regimen [19]. Animal studies have shown that intermittent (2 days/week) or low-dose estrogen treatment is more effective than continuous normal-to-high dose treatment [39]. It is also a possibility that these regimens may prevent reversal of hippocampal sprouting, which is a by-product of continuous (>2 days) elevated estrogen levels [39]. We hypothesized that this may mimic the natural estrogen fluctuations seen in the female reproductive cycle and hence may prevent the reversal of positive effects on the brain [19].
Surgical menopause
Our quantitative meta-analyses of women without dementia and the current updated meta-analyses indicated that significant effects of HT were strongest in women who had undergone surgical menopause [48]. A significant drop in cognitive function after ovariectomy has been shown [57–59]. By contrast, researchers could not find any evidence suggesting that natural menopause causes a drop in cognitive function [60]. Therefore, one conclusion that could be drawn from these results is that surgical menopause in itself is a risk factor for accelerated cognitive impairment and that estrogen therapy would be particularly indicated for this group. However, in another observational study [61], surgically menopausal women who had an E2 implant for approximately 10 years had worse long-term episodic memory and mental flexibility, and more psychological and somatic menopausal symptoms, than untreated surgically menopausal women. These negative results are in line with other findings of an association between better verbal fluency and HT use in younger (<58 years) surgically menopausal women, but not in older, surgically or younger, naturally menopausal women [43]. These findings also tie in with an observational study that indicated that former –but not current – users of HT were protected against AD [62]. However, interactions of duration of use and protective effects were not always found. For example, in another observational study, surgically menopausal women using HT compared with surgically menopausal women not using HT had better verbal memory and constructional abilities at the age of 65 years [63]. As duration of treatment was not reported in this study, it is possible that the majority of women in this study had only used HT for a limited period of time.
Thus, it must be kept in mind that several clinically important questions remain unanswered, such as the effect of duration of use and the generalizability of the WHIMS to women for whom HT is an indication, that is, perimenopausal women who have menopausal symptoms [64].
While our quantitative meta-analyses identified surgically menopausal women as those most effectively treated for cognitive deficiency by HT [48], other reviewers suggested that women who were highly symptomatic (which would include most surgically menopausal women) showed largest treatment effects [42]. Although when statistically tested, menopausal symptoms did not explain improvement on cognitive function, these issues need to be further elucidated [19].
There is no doubt that the findings of these trials have had and will continue to have a huge influence on the industry and the use of HT. Furthermore, more controlled studies need to be done in order to fully understand and support or refute the findings so far. The key to future understanding may lie in focusing on interactions of the factors mentioned above and genetic risk factors.
Genetics
It is possible that an interaction exists between the effects of HT and genetic factors. More specifically, the effect of HT may depend on a woman's particular genetic profile. It is has been found that having at least one apolipoprotein (APO)E ε4 allele confers a 2–3-times increased risk of AD [60]. Some studies found that cognitive decline was prevented by HT, but only in women without the APOE ε4 allele [65,66]. However, at least five other studies did not find any significant difference when taking the APOE genotype into account [38,62,67,68].
We and others, using highly sensitive assays, have found that E2 levels were actually slightly elevated in women with AD [19]. It could be hypothesized that women with AD have a genetic predisposition to higher E2 levels. Certain genes are involved in the metabolism and synthesis of E2. For example, the enzyme cytochrome P450(CYP)c17a is involved in estrogen synthesis, which is regulated by the CYP17 gene. This gene is located on chromosome 10 and has two allele variants. A polymorphism (or variation) of this gene is the A2 allele, which causes an increase in the quantity of enzyme produced, resulting in higher estrogen levels. The allele versions of this gene without the variation are known as A1. It has been reported by several studies that people who are A2 homozygotic (have both A2 gene allele versions: A2:A2) or who are heterozygotic (with just one A2 allele: A2:A1) have higher levels of estrogen than people with two A1 alleles (A1:A1) [69]. Therefore, using HT to increase the levels of estrogen in women who are homozygotic (A2:A2) and already have higher levels of estrogen could have a detrimental effect. It is possible that there may be optimal levels of estrogen for the brain to function at the most efficient level. Therefore, increasing E2 above this optimal level could result in negative effects on the brain [19]. If our review data relating type of treatment to age are correct, this could be related to different optimal levels of estrogen with age (e.g., lower plasma levels of E2 are reached with CEE and transdermal E2 compared with oral and bolus E2 [Figure 1]). Similar mechanisms have been postulated for actions of testosterone in elderly men [19]. Lower levels of estrogen may thus be more effective in older women compared with younger menopausal women, particularly for those who already have higher endogenous estrogen levels [19].
Estradiol levels as reported after different treatment studies.
However, high estrogen levels might be appropriate in those who have less sensitive ERs. The last few years have seen a growth in our understanding of the role of estrogen sites in the brain. The ERα gene is located on chromosome 6, and the β (ERβ) gene is located on chromosome 14 [70]. ERα and ERβ are widely distributed throughout the CNS [71]. They are found in brain regions such as the hippocampus, cerebral cortex, midbrain and brain stem. ERα is found in high levels in the pituitary, hypothalamus, hypothalamic preoptic area and amygdala [72]. Although the brain distribution of ERα is well documented, more controversy surrounds the localization of ERβ. In situ mRNA hybridization experiments have detected ERβ in the cerebellum, cerebral cortex, hypothalamus and hippocampus, although immuno-cytochemical studies have indicated a more restricted expression pattern of ERβ protein [72]. The two ER types coexist in the hypothalamic preoptic area, bed nucleus of the stria terminalis and the medial amygdaloid nucleus [18]. The difference in brain localization suggests that these two types of ER may mediate different functions [73]. The primary function of ERα in the brain is to maintain concentrations of follicle stimulating hormone (FSH) and luteinizing hormone (LH) in the blood. FSH and LH are two gonadotropins produced by the pituitary gland that normally regulate sex steroid hormone production. An alternative splicing isoform of ERβ, with an extra 18 amino acid insertion, has been found to act as a negative regulator for estrogen responses by inhibiting the transcriptional activation effects of ERα and ERβ in a dose-dependent manner [74,75].
It has been suggested that genetic polymorphisms affecting ER sensitivity may be involved in AD pathology. For example, one genetic association study of 193 late-onset AD cases and 202 age-matched controls found that the prevalence of AD was increased in the presence of a combination of ERα variants identified by intronic restriction sites for the endonucleases PvuII and XbaI, and that this genotype interacted with ApoE genotype in determining AD onset susceptibility for one specific combination of ERα and ApoE genotypes (odds ratio [OR]: 7.6; 95% CI: 1.1–62.3) [70]. Although a more recent study of 172 probable AD cases and 172 age-matched controls did not confirm this interaction [76], these authors did find that the ERα XbaI polymorphism correlated with a significantly increased OR for developing late-onset AD.
ERβ polymorphisms that affect estrogen sensitivity have also been implicated as a risk factor for dementia. A study of 387 AD cases and 467 matched controls, which investigated two intronic single nucleotide polymorphisms in the ERβ gene, found a significant increase in AD risk for women (OR: 1.87; 95% CI: 1.21–2.90), but not for men, which was associated with both single nucleotide polymorphisms [77].
On the basis of such results, several extensive reviews concluded that it may be possible to identify a common set of gene variants, including ER variants, which contribute to a common genetic background for neurodegenerative disorders such as AD [70,75,78]. One must note that there are variations in the distribution of ER genotypes, especially ERα, between different ethnic groups, especially Asian and European [71]. Consequently, larger studies with multiethnic populations are needed in order to support the role of ER polymorphisms as a risk factor for AD.
Thus, whether and how ER sensitivity plays a role in AD is unclear. However, high estrogen levels might have detrimental effects on the brain through other mechanisms [19]. DNA damage has been associated with AD pathology [19,77]. CYP1B1 is a key enzyme in the metabolism of E2. 4-hydroxylation of E2 (regulated by expression of CYP) results in a reduction of estrogenic activity. A toxic metabolite, which has been associated with DNA damage, is a product of this [19,79]. The Val432 allele variant is associated with higher 4-hydroxy E2 levels than the Leu432 variant. In women who would have high estrogen levels, this variant could result in dangerous metabolites, potentially leading to DNA damage that could be implicated in AD risk [19] In addition, the inactivation of reactive metabolites, such as catecholestrogens, is regulated by catechol-O-methytransferase (COMT), which is regulated by the gene COMT (Val108/158 Met). The COMT Met/Met polymorphism may be a risk factor when compared with the COMT Val/Val genotype, in that reactive metabolites are not inactivated [19,80]. If women have genotypes CYP17 A2, CYP1B1 Val432 and COMT Met/Met and are then given HT they might produce very high levels of toxic metabolites, which might put them at risk for dementia [19].
These theories have not been fully investigated, but may have important implications in our knowledge of the causes of AD and need to be substantiated in further studies.
In an observational study from the Mayo Clinic, Minnesota (USA), it was found that women who underwent surgical menopause had a 40% increased risk of dementia. Women who had bilateral ovariectomy by the age of 46 years had a 70% increased risk of dementia. The removal of one ovary before 38 years of age resulted in a 260% increase in the risk of dementia [Rocca W, Mayo Clinic, Rochester, MN, USA. ORAL COMMUN.: Ovary removal elevates risk for Parkinson's disease and parkinsonism. American Academy of Neurology, 2006]. The reason for this trend could be due to early exposure to low levels of potentially protective estrogen levels. However, lifelong exposure to estrogen (calculated from age of menopause and onset of menses) has shown controversial associations with the onset of dementia [19]. Alternatively, surgical menopause and increased risk of dementia may be linked because women usually undergo surgical menopause because of endometrial cancer. The type of genetic risk markers mentioned above, leading to high levels of toxic estrogenic metabolites, could be implicated in the mechanism, explaining the association between endometrial cancer and AD [19]. If this is so, survivors of other steroid-sensitive cancers, such as breast cancer, could also be more likely to develop AD.
Conclusion
The question that needs to be asked is whether oral E2 can have a positive effect on cognitive function in recently postmenopausal women for a longer period of time. This question will be answered by the multicenter Kronos Early Estrogen Prevention Study survey in the USA. It is unclear whether this effect pertains to women who have undergone surgical menopause and/or those with menopausal symptoms, and whether this effect is thus mediated by symptom alleviation. As can be seen from previous studies, it seems that CEE is not an effective form of HT when addressing memory problems in this younger, more recently menopausal group.
It has also previously been stated that the primary reason for a lack of effect of HT on menopausal cognitive decline in women is their age. Some believe that there is a critical window of time after which HT is ineffective, as it will not only have no effect on cognitive decline, but may also result in additional detrimental effect and impair cognition further. However, some treatment studies do show positive effects of HT (transdermal E2 and CEE) in older women and some aspects of cognition are possibly more sensitive to these effects than others (tests of information processing speed). This may be related to optimal estrogen levels for older women.
Whether women have undergone natural or surgical menopause may play a crucial role in this debate. A sudden and early loss of estrogen may not be the most important factor in this association. Instead, it is possible that the same genetic predisposition to conditions leading to surgical menopause underlie the increased risk of AD [19]. In particular, if women who are genetically at risk receive HT, increasing their levels of endogenous E2, this could potentially affect the production of toxic metabolites and lead to DNA damage, which is implicated in a variety of age-related morbidity factors, such as some cancers and AD [19].
To conclude, the studies to date suggest that short-term (up to 3 months) oral E2 treatment (with progesterone for women with an intact uterus) for symptomatic recently naturally or surgically menopausal women is the most favorable combination to benefit verbal memory and also to improve symptoms, the main indication for treating women with estrogen. Whether genetic predispositions associated with estrogen metabolism modify a woman's risk for dementia remains to be investigated.
Future perspective
In a world with constant advances in ways to prolong life and improve standards of living, it is evident that the prevalence of age-related disease, such as dementia, is bound to increase. This creates not only a higher demand on society's resources, but also the economy as a whole. In a report published by the London School of Economics via the Alzheimer's Research Society, the UK alone is estimated to be faced with double the costs of care for the elderly in the next 30 years. Therefore, this is a topic that will have even greater significance and importance in the future.
Executive summary
As we age, there is a decline in certain cognitive functions, such as verbal memory and complex spatial and verbal skills.
Cognitive decline with age ranges from ‘successful’ aging to more severe deterioration and dementia, of which Alzheimer's disease (AD) is the most common.
Women may be at greater risk of developing AD than men.
The fact that women reach an older age than men does not explain this difference in dementia prevalence because age-specific AD incidence is also higher in women.
It was theorized that a more severe deficiency in estrogen in elderly postmenopausal women, when compared with men, may increase their risk of AD.
Hormone therapy (HT) consisting of estrogen has been suggested as a way of treating/preventing AD in older women and observational studies have found protective associations.
Animal models have shown that estrogen increases dendritic density in the hippocampus, and estrogen receptors are highly concentrated in the hippocampus.
People with AD have severely affected hippocampal regions, and therefore a lack of estrogen in this area may be a precursor for the development of AD.
Estrogen may protect against metabolic decline of certain brain regions susceptible to AD in postmenopausal women.
The early nineties provided a wealth of research which supported the idea that HT is an effective treatment method against AD.
However, methodological problems of earlier studies and the opposing results found by more recent studies have led to a change in perspective.
The Women's Initiative Memory Study (WHIMS) was a randomized, placebo-controlled trial (RCT) that used conjugated equine estrogens (CEE) with medroxyprogesterone (MPA) versus placebo, or continuous unopposed CEE versus placebo. Both treatments in this study found detrimental effects of HT on cognition in postmenopausal women.
Some WHIMS researchers advise that the popularly used CEEs, Premarin® and Prempro™, are detrimental to neurocognitive health and therefore do not recommend HT as an effective therapy for dementia.
There are several possible reasons for the differences reported in results between WHIMS, earlier studies and other more recent HT trials.
Although multi-infarct factors could have played a mediating role in the WHIMS results, post hoc analysis removing women with stroke at baseline did not change the effects found.
Age – it is believed by some that women aged over 65 years cannot benefit from the beneficial effects of HT. However, some studies found positive short-lived effects of HT in women with dementia.
‘Window of opportunity’ – there may be a critical period of time in which HT is effective. This may be in younger, symptomatic recently menopausal women with a minimal delay between ovariectomy and HT onset. However, studies suggest that these effects, when found, are only short-lived (up to 2–3 months).
Treatment form/route – CEEs are the main treatment used in studies that showed no effects of HT on verbal memory, whereas oral and bolus injections of estradiol (E2) (alone or combined with progesterone) have yielded positive results in recently menopausal women, particularly if they were symptomatic.
Cyclical versus continuous administration – continuous elevated levels of estrogen could have long-term negative effects on cognition. It is possible that the effects are positive up to 3 months, stabilize at 6 months and decline after 1 year. This could be due to downregulation of receptors. Therefore, it has been proposed that an altered treatment regimen of estrogen may be more effective at maintaining cognitive function.
Orally administered CEEs and transdermal E2 have also been found to be effective in older women, suggesting that optimal type and treatment route may be modified by age.
Surgical menopause – some treatment trials have shown that HT may be most effective in women who have undergone surgical menopause.
The effect of HT may depend on a woman's specific genetic polymorphisms associated with estrogen synthesis and metabolism, resulting in higher endogenous E2 levels, which may be detrimental.
DNA damage has been associated with AD onset. 4-hydroxylation of E2 may result in a toxic metabolite, which has been associated with DNA damage. Certain genetic polymorphisms are associated with higher production or lower breakdown of these toxic metabolites
The genotypes cytochrome P450 (CYP)17 A2, CYP1B1 Val432 and catechol-O-methytransferase Met/Met, in combination with HT, may increase levels of E2 and toxic metabolites that could result in DNA damage and could put women at risk of AD.
The role of genotypes relating to estrogen receptor sensitivity is not full elucidated in AD.
Executive summary
From previous studies, it seems that CEE is not an effective form of HT for cognitive complaints in recently menopausal women. E2 has yielded more positive results, with or without progesterone.
It is argued that there is a critical window of time in which women are at the optimal age for the beneficial effects seen with HT, after which it is ineffective or even detrimental. However, other positive treatment studies of older women refute this.
Positive effects of HT may be seen, particularly in younger, surgical menopausal women, using oral E2 as the treatment form, and seem to affect some components of cognition and not others. Effects are short-lived and small, whether they are detected in older or younger postmenopausal women. We may need to further investigate the genetic make-up of the individual before treating women with HT.
There is, and will continue to be, a great demand on societies' resources due to the fact that we are living longer lives and therefore prevalence of age-related cognitive decline is also likely to increase.
The confusion seen in the last 10 years regarding the effects of HT in the treatment of cognitive decline is worrying. Future studies must take into consideration all the variables that have come from the multitude of previous investigations and combine knowledge to come to a more concrete foundation.
Researchers should possibly look carefully at the inclusion criteria of their participants and take into consideration previous findings regarding window of opportunity, treatment form/route, surgical versus natural menopause and also genetic factors, which may tie all this together.
The general confusion and conflicts seen in this area are worrying, particularly for post-menopausal women who are plagued by symptoms, such as hot flushes, night sweats and cognitive and affective complaints. However, just as history has shown a shift in perspective over the last 10 years, what the next 10 years bring will hopefully result in the clarification of the effects that HT has on cognition in menopausal women with and without dementia. The future lies in more RCTs that take into consideration all the variables that have been seen to influence the impact of HT on cognition. A change in perspective is needed to focus on which variables result in a positive effect of HT, instead of just those aspects that may result in negative findings. The inclusion and exclusion criteria used in future studies should take into consideration factors such as the influence of age, window of time, treatment form/route, surgical versus natural menopause and, possibly, genetics.