Alzheimer’s Disease and (Phyto) Estrogen Treatment: Modification of Effects by Age,Type of Treatment,and Duration of Use

Abstract

Background:

There is a continued debate on whether menopausal hormone therapy (MHT) protects women against Alzheimer’s disease (AD). It is also unclear whether phytoestrogen could be an alternative treatment for AD.

Objective:

To investigate whether mixed study findings may be due to differences in age at initiation of MHT and duration of prescription of different types of MHT using meta-analyses.

Methods:

After a systematic literature search, meta-analyses were carried out using Cochrane Revman 5.4.1.software including data from large nationwide studies of registered medically diagnosed AD and prescribed MHT. These analyses were stratified for duration and type of treatment, by age at start of prescription of therapy. Insufficient quality data were available for phytoestrogen treatment and AD meta-analyses.

Results:

A total of 912,157 women were included from five registries, of whom 278,495 had developed AD during follow-up. Meta-analyses suggested a small increased AD risk after 5–10 years prescription of combination MHT regardless of age, and over 10 years only in women younger than 60 years of age. No association was seen for estrogen alone for women younger than 60 years of age, but AD risk did increase for women over 60 years of age for up to 5 years of MHT prescriptions.

Conclusions:

Combination MHT should probably be prescribed for less than 5 years after menopause to reduce risk for AD, while estrogen alone should not be prescribed to women over 60. For phytoestrogen, small treatment trials suggested some benefit of tempeh (fermented soy), which should be investigated further.

Keywords

Alzheimer’s disease risk estrogen hormone therapy memory menopause meta-analyses

INTRODUCTION

Alzheimer’s disease (AD), the most common type of dementia, is characterized by progressively and gradually developing worsening of memory, planning, and other cognitive functions, which impact activities of daily life and occur in the presence of biomarkers indicating AD type brain pathology [1]. At postmortem, when a definite AD diagnosis can be made, often multiple types of brain pathologies are present, including vascular lesions and deep white matter disease [2]. The risk for AD is higher in older women than in men in many, but not all, cohorts [3]. This increased AD risk in women has sometimes been attributed to loss of estrogens after the menopause.

Estrogen loss after the menopause and replacing loss of sex hormones using MHT

Menopause is usually defined as women’s endocrinological status a year after the last menstrual period (LMP). The LMP results in a significant reduction in estrogen levels, unless women use menopausal hormone therapy (MHT). We used the term MHT although we also included studies in this paper which had investigated women with dementia, and/or those over 62 years of age who would have been a decade older than the average age at menopause and for whom MHT would not be indicated. In addition, several papers still use the term hormone replacement therapy. MHT usually consists of an estrogen and, if women have a womb, a progestogen is added (this is named ‘combination MHT’ in this paper), to protect against endometrial hyperplasia and cancer. According to most medical guidelines, e.g., [4], estrogen alone should only be given to women without a womb, who also have often undergone surgical menopause with oophorectomy (removal of the ovaries).

Unexpected negative effects of MHT on dementia risk in the WHIMS

A large seminal randomized controlled trial (RCT), the Women’s Health Initiative Memory Study (WHIMS) showed (in sharp contrast to the initial expectations) that already after a year of using combination MHT [5], a significant increased risk of dementia was seen in women over 65 years of age. This was not found for treatment with estrogen alone [6]. For this age-group (older than 65), normally MHT would not be indicated according to most medical guidelines (e.g., [4]) Guidelines emphasize the alleviation of neuroendocrine effects of the menopause and most women over 60 would be well past that stage. In WHIMS, there was also an increased risk after combination MHT for breast and some other cancers, stroke, venous thrombosis, and cardiovascular disease (CVD) [7]. The trial was discontinued early because of these perceived risks and as a result many women stopped using MHT. The WHIMS [6] and several shorter and smaller studies treating older women with AD with MHT had been undertaken due to increasing enthusiasm for estrogens’ perceived potential to protect against CVD and AD [8]. Recently this research area again received much media attention, after similar earlier cycles of enthusiasm in time periods such as the early fifties, mid-seventies, late eighties and early nineties [9]. It was again suggested that MHT should be taken for longer periods of time to prevent AD [10], despite WHIMS’s negative findings as described above.

Possible mechanisms for MHT’s benefits for general health and AD prevention

The reason for this continued belief in MHT’s potential protective effects on brain function is partly based on observational studies, but also on substantial biological evidence from animal, cell culture, and brain scan studies. Estrogens can act to prevent AD by regulating brain glucose metabolism, increasing cerebral blood flow and dendritic outgrowth, by acting on nerve growth factors through co-localization of receptors, counteracting oxidative stress; via effects on neurotransmitter synthesis and turn-over and many more potentially beneficial mechanisms [11, 12]. In addition, estrogens have been found to confer reduced risk for CVD, which also shares other risk factors with AD, such as midlife obesity, hypertension, total cholesterol levels, insulin resistance, and diabetes mellitus [13 –15]. Reduced estrogen receptor sensitivity in the brain after menopause was also hypothesized to be responsible for increased AD pathology and more selective acting phytoestrogens (such as genistein, derived from soy beans) were hypothesized to be a more promising treatment for AD [16]. A flurry of recent reviews and meta-analyses of observational studies suggested both reduced and increased risk of AD with MHT (e.g., [17]). Similarly, observational findings and small treatment trials also reported both reduced and increased dementia risk with high consumption of phytoestrogen containing foods, particularly of tofu in older (over 65 years of age) East Asian populations, adding to further confusion [18, 19].

Duration limited benefits of MHT on some cognitive functions sensitive to AD regardless of age

Observational data and biological plausibility do not necessarily translate to experimentally established clinical benefits on brain function in women, such as improved memory in midlife, or prevention of AD progression in later life. Earlier reviews of randomized controlled trials including Cochrane meta-analyses reported only time-limited beneficial effects of estrogen treatment of 2–3 months in highly symptomatic middle-aged surgically menopausal women treated with estradiol [20], but also for 1–2 months treatment in studies on some cognitive test outcomes in women with AD [21]. By 4–6 months of treatment with conjugated equine estrogens (also used in WHIMS), these effects in women with AD were no longer seen and, in fact, reversed on some outcomes [21]. Other studies also suggested that longer treatment with MHT seemed to show reversal of benefits in both middle-aged women and women with AD [22].

The ‘window of time hypothesis’ is not upheld by randomized controlled trials

The short-term benefits on cognition seen in AD, as well as in middle-aged women, do not support the ‘timing’ or ‘window of time’ hypothesis [23], as women with AD were at least 15 years from the average age at menopause, which is around 51 years of age [18]. The timing hypothesis poses that estrogen should be given close to the age at menopause to show beneficial effects on brain and vascular function. Two trials [24, 25] investigated this timing hypothesis and whether type of estrogen (natural estradiol or conjugated equine estrogen, as used in WHIMS) would make a difference to outcomes. Both treatments showed no effects on cognitive functions sensitive to AD (or on most CVD risk outcomes), regardless of type of treatment, or whether this was given closer to age at menopause (within 6 years from LMP or after 10 years). However, these studies assessed cognition after MHT at 4–5 years, which may have been too late to show any initial potential short-term benefits, as shown in the Cochrane meta-analyses to occur at best only at 2–3 months, but not thereafter [20].

Longer duration of hormone treatment shows no benefit on cognition

WHIMS-Y (a WHI ancillary study of younger women around 50–55 years of age) also showed no difference in the MHT or placebo treated women on several cognitive functions at a follow-up (7 years after treatments had stopped), when the 1,326 women tested were on average 67 years old. Only verbal fluency was worse in women who had used conjugated equine estrogen alone, who had undergone a hysterectomy, and/or had used MHT prior to the trial for an average of 2 years longer [26]. Another study (Cognitive Complaints in Early Menopause) stopped when the WHIMS results came out and had investigated the same combination MHT regime in 180 women aged 45–55 years for 4 months. They also found no overall negative cognitive effects, save for trends for worse verbal memory after MHT [27]. Verbal memory is sensitive to early AD and to undergoing menopausal change and hormone fluctuations [8]. These data could suggest that with longer use, MHT has no beneficial effects, but could instead have negative effects on some cognitive functions sensitive to AD.

The ‘healthy cell bias theory’ possibly explains negative effects of MHT in older women

The increased risk of dementia in the older WHIMS women (>65 years) and with longer MHT use on some cognitive functions sensitive to AD in middle-aged women versus the protective effects seen of MHT in observational studies [28] could be explained by the ‘healthy cell bias’ theory [29]. Cell studies showed that estrogens can help sustain healthy neurons. However, when given to neurons undergoing pathological change (of mitochondria and calcium channels) associated with AD risk (which is more likely as women age), estrogens were seen to accelerate this damage [29]. Similar timing/healthy bias hypotheses were posed for CVD risk associated with [30] clinical atherosclerosis [7], but not all treatment studies substantiated this[15].

Unclear if cognitive decline in menopause with/out HT relates to later life AD risk

The remaining question is whether, despite only showing (at best) short time-limited benefits on cognition (if at all) after menopause, MHT in midlife for a short duration could still protect against subclinical AD/CVD risk parameters and ultimately reduce later life AD risk. If estrogens would benefit the brain in midlife, the significant changes in estrogen levels during and after the menopausal transition would be expected to affect cognitive functions sensitive to AD in some way in that period. Decline in memory and other cognitive functions in midlife [31, 32] can predict AD onset in later life, which is usually after 65 years of age (unless it is the rarer early onset form with a strong genetic/familial component). However, objective changes in cognitive function during the menopausal transition were found to be subtle, related to reduced (implicit) memory (i.e., learning from repeated exposure to a task to improve performance) on that task and were mainly seen when following the same women over the menopausal transition, but not when comparing women in different stages of the menopause cross-sectionally [18]. In one relatively small longitudinal study [33], objectively reduced cognitive performance was limited to a subgroup of 23% of women, whereas the other women in this study showed no change, or even improvements, over the menopausal transition. Several longitudinal studies also suggested that the implicit memory issues (not improving in performance after repeated exposure to a task) reversed back to premenopausal functioning in later postmenopausal follow-up assessments[18].

The ‘domino theory’ explaining cognitive decline during the menopausal transition

In contrast, significant declines in memory and other functions, as well as increased levels of biomarkers associated with AD have been seen in several studies following women after surgical menopause, with removal of the ovaries resulting in a sharp acute drop in estrogens and testosterone levels. MHT reversed this cognitive decline when tested for 2–4 months in our Cochrane analyses [34]. It was hypothesized that disrupted sleep, common during menopausal transition and/or vasomotor symptoms, such as hot flushes and (night)sweats, at least partly explained these cognitive issues, which is known as the ‘domino theory’[35].

Two studies reported that poor sleep possibly mediated worse memory and other cognitive functions after following women through time while they were undergoing natural menopause [33, 36]. Another study did not find that poor sleep, flushes or anxiety explained worse cognition in the early stages of the menopausal transition using a cross sectional study [37]. Others reported that improved memory after MHT was not explained in analyses by improved sleep, mood or vasomotor flushes in highly symptomatic surgical menopausal women [38], and some suggested that there was a separate ‘tonic’ effect of estrogens on cognition [35]. An earlier age at surgical menopause without estrogen add-back treatment until the average age at natural menopause could induce longer term increased cognitive impairment risk, as Professor Walter Rocca’s group suggested [18].

Sleep and depression in menopausal transition and later life AD risk

Poor sleep is associated with an increased build-up of AD and CVD pathology and depression is also a risk factor for both AD and CVD [13]. However, while mood can fluctuate in the perimenopausal stage (and many women complain of mental health issues, including poor memory), a clinical depression is not more common after the menopause [18]. With a sharp drop in estrogen levels, as well as worse symptoms after surgical menopause, longer term protective effects in preventing AD would perhaps be expected in women who could be prescribed estrogens alone, whether that is mediated by improved sleep or is a direct effect of undergoing surgical menopause. However, small UK observational follow up data also showed negative associations with worse cognition, when estrogens alone was used for a longer period of time (over 10 years of use) in these women [39]. Duration of use, type of regimen (with or without a progestogen) and/or age of women at onset of MHT could play an important modifying role.

The healthy user bias theory

A significant source of bias in observational studies is that women who took MHT after the menopause in several studies were at reduced risk for AD before they started MHT, such as by having obtained higher levels of education and having fewer CVD risk factors already 8–10 years before deciding to use MHT [40]. Conversely, for women undergoing surgical menopause and/or menopause at an earlier age, in some studies the reverse was found, where these women were characterized by having had obtained on average less education and also already had more CVD/AD risk factors (e.g., obesity, smoking) before their early menopause onset [18]. While the majority of observational studies controlled for these confounds, systematic inherent bias, where characteristics of (educated, wealthy) people choosing MHT are closely bound to engaging in (other) health promoting behaviors, is difficult to control for statistically.

The duration of treatment hypothesis and AD risk versus meta-analyses finding protective effects

These data combined suggest that perhaps temporary beneficial effects on memory and other cognitive symptoms are most clearly seen when menopausal symptoms are reduced in highly symptomatic women, but that these benefits (if they occur) do not persist after a couple of months, then become neutral, and may even reverse to incur AD risk after longer (more than 10 years) durations of MHT, regardless of the type of treatment. However, a recent meta-analysis in 2020 suggested that overall MHT reduced AD risk significantly by 33%. Inspection of the 16 included observational studies in this review showed that, of the seven studies using medical records or prescription data, five studies (the last one published in 2006) showed no significant association between MHT use and AD. In the two studies that did find a significant association, the 95% confidence interval (CI) was very close to 1 [41].

Recall bias in self-report investigating AD risk

This meta-analysis [41] which also included studies which had assessed dementia or AD diagnosis via self-report could be sensitive to report bias, due to denial or shame because of the stigma related to the AD diagnosis, leading to false negatives. Retrospective self-report data collected with questionnaires and/or interviews, when women already have AD, can also be confounded by recall bias, caused by the very characteristics of the disease, i.e., memory loss. Indeed, women with AD were found to be twice more likely to forget having used MHT compared to controls, when self-reported use was compared against by medical prescription data [42]. This aligns with our earlier meta-analysis of observational studies done before the year 2000, with many studies that used self-reported AD and MHT use finding a halved AD risk with MHT use [28].

A recent meta-analysis finding increased and decreased dementia risk with MHT

An even more recent, very comprehensive review [17] and meta-analyses including the WHIMS with several follow-ups at 12, and 13 years after treatment suggest an increased risk for dementia with combination, but not estrogen alone treatment, but this was no longer significant after the 18 years of follow-up (possibly due to attrition). The meta-analyses also included 45 observational studies [17] which suggested reduced risk of dementia, with protective effects for estrogen alone (but not for combination MHT) in midlife. Non-significant increased risks were seen in later life (>65) for both types of treatment in combined analyses. Most of the included observational studies were carried out in the USA (12 versus 6 studies from elsewhere which had used medical registries).

The use of national registries of medical databases

Some studies included in the meta-analyses [17] had used self-report and interviews to obtain data on MHT use and/or AD diagnoses. Self-selection into observational trials often does not include people who have undergone less education, with accompanying poor health and lower socioeconomic status (SES) [43]. National medically registered databases on MHT use and medically diagnosed AD perhaps offer a more generalizable overview of risks and benefits of MHT, as these might be expected to also include women of lower SES, who often have a higher risk of morbidity, including AD. The studies included in the meta-analyses [17] that had used medical prescription data were generally small (n < 500). The only included (US) study in the meta-analyses with n > 5000 had used an insurance database (Kaiser Permanente) from North California of data collected between 1964 and 1994. Analyses showed protective effects of MHT use for AD when prescribed in midlife (average age 49) by 26%, but when using MHT in later life (around age 76), the medical specialist diagnosed AD risk was increased by 48%. For women who used MHT in midlife and late life, AD risk was not significantly affected [44]. In the meta-analysis [17], duration of use was not analyzed for estrogen or combination therapy separately or by age at MHT initiation (younger than 60 years of age or older).

Objectives and hypotheses

We hypothesized, based on the literature, that when using national registries of medical databases, an association of longer combination MHT duration of prescription would predict (worse AD risk) in women younger than 60 years of age. For estrogen alone (as a proxy for surgical menopause) there may be protective effects, but again longer prescriptions (>10 years) were expected to confer increased AD risk in women younger than 60 years of age. For women older than 60, increased AD risk was expected already after a year of MHT. We also discuss phytoestrogen treatment for AD as an alternative approach using our recent review.

MATERIALS AND METHODS

We only included data on medically diagnosed and registered AD as an outcome, as this is the most common type of dementia and other types of dementia are not always recognized in life, when compared to postmortem diagnosis (validity). However, diagnoses in life can also have poor inter-rater reliability [2]. MHT was defined as medically prescribed and registered presumed use of an estrogen (estradiol, estrone, conjugated equine estrogen, etc.) with or without progestogens (named combination MHT versus estrogen alone). National medical databases included national health insurance databases and national GP and other (e.g., hospital) registries. Excluded were studies which had not used a national database to avoid self-selection bias, (as these studies often include more higher educated women who are more likely to use MHT and have less AD risk) [40]. Other sex steroids treatments (e.g., testosterone) or drugs acting on sex steroid levels or selective estrogen receptor modulators (e.g., tamoxifen) wereexcluded.

A literature search was conducted by all authors independently following the PRISMA guidelines in Google Scholar and PubMed/MEDLINE until November 2023 using key search terms ‘Alzheimer’ OR ‘Alzheimer’s disease’ OR ‘dementia’ AND ‘menopausal hormone therapy’ OR ‘hormone replacement therapy’ AND ‘risk’, OR ‘odds ratio’. This rendered 465,008 references on PubMed with many irrelevant papers, versus 1,880 publications via Google Scholar. Adding AND ‘Registry’ resulted in n = 19,812 publications in PubMed and 1,070 in Google Scholar, where adding AND ‘estr*’ OR ‘oestr’ resulted in n = 788 papers. Using MEDLINE search terms (see the Supplementary Material, which allowed exclusion of studies) resulted in n = 381 abstracts. From this database, duplicates, reviews, editorials, and any studies focusing mainly on estrogen receptors or endogenous estrogen levels and/or interactions with genotypes (APOE, estrogen metabolism like COMT genotypes), women with cancer (e.g., breast), other morbidity (e.g., depression, HIV) including neurological disease (e.g., Parkinson’s disease, stroke), as well as other types or forms of dementia (e.g., mild cognitive impairment, a prodromal stage of AD) or only investigating cognition could be excluded.

The ultimate selection of nine studies was checked against the most recent meta-analyses mentioned in the introduction [17], which had included studies via a systematic literature search in PubMed/MEDLINE, Web of Science, and Cochrane databases from 1975 through to July 2023. Only one additional study [45] was identified via these meta-analyses. Screening of selected studies and data extraction from the papers were discussed by all authors and reviewed using consensus for any disagreements. Only one set of analyses was included per national registry [e.g., studies for Denmark [45]; for Finland [46], and the UK [47] were excluded from the search, as these nationwide cohorts had already been included: [48] (the Denmark national registry), [49] (the Finish national registry), and [50] (the UK registry), respectively]. The remaining six national registry studies and the nationwide RCT [5, 6] are described in Table 1. We had included the WHIMS RCT [5, 6] for the meta-analyses, as the other US registry study [10] had not given analyses for different types of treatment by duration of use, but had instead combined the different treatments. The nationwide WHI had included 40 clinical centers and had enrolled originally over 161,000 women, which had a long follow-up and had included medically registered MHT data (combined and estrogen alone) in the over 65 years of age. Only four studies ([48 –50] and [5, 6] for WHIMS) that described duration of prescription for the separate MHT treatments by age (estrogen alone or combination MHT) could be included in the meta-analyses.

Table 1

National Medical Registry Data based observational studies of AD and HT use risk

							RCT
Author	Pourhadi	Savolainen	Vinogradova	Kim	Yuk	Sung	Shumaker
Year	2023*	2019	2021	2021	2023	2002	2003/2004
Country	Denmark	Finland	UK	USA	S Korea	Taiwan	USA
Number of AD/total cohort	1,099/11,851	190,284/380,568	37,461/230,813	7,775/379,352	73,664/1,399,256	182/105,072	108/7,479
3	4– to 8	3 to 5	3 to 5	3 to 6	<5		4 to 5
4	8 to 12	5 to 10	5 to 10	6+	5 to 10	<13.5 y
5	12+	10+	10+		10+	>13.5 y
	nationwide prescription reg/hospitals	nationwide drug/neurologist registry	nationwide GP/hospital registry data	health insurance (claims submitted)	health insurance (universal coverage)	health insurance (universal coverage)	randomized trial
Age-group at start MHT use	Median age 53, range 50–60	Average age 58 (E only)56.1 (E + P)	Majority 50–59 (10–30%), 2% over 70	Average age 68	Median age 54, range51–59	Majority 40–49 (58.6%)	Majority 65–74(82%)
MHT type investigated	no E, E + P	E, E + P	E, E + P	MHT	E, E + P	E, E + P	E, E + P
Covariates	Age, education, annual household income, cohabitation, medical history (hypertension, diabetes, thyroid disease)	Age, districts, hormone use, age at hormone initiation, exposure time no data on education	Age, ethnicity, social deprivation, BMI, smoking, alcohol, chronic conditions, drug use, mental health, early menopause, hysterectomy/ oophorectomy, menopausal symptoms, family history of dementia	Age, ethnicity, comorbidities, CCI no data on education/SES, age at initiation or type of menopause	Age, BMI, SES, age at menarche, age at menopause, smoking, alcohol, exercise, Charlson comorbidity index (CCI)	Age, index date, baseline comorbidities, hypertension, diabetes mellitus, hyperlipidemia, coronary artery disease, stroke, COPD, CKD, cancer, CCI, age at menopause, ‘other demographics?’	Age, education, smoking, history of stroke and diabetes, adherence, other prior medication use (statins and aspirin), prior HT use
Time frame of MHT prescribing	2000–2018	1994–2013	1998–2020	2007–2016	2002–2019	2000–2005	1995–2002/4 (combined/Ealone)
Duration of MHT use (average or median)	3.8 y	11.8 (9.3) E only 9.8(7.7) E + P	16 y	60% used 1 y or less, 5% >6 y	75% used less than 5 y	<13.5 y (20,169)>13.5 y (14,885)	7 y with decreasing adherence rate
Average duration of follow-up to AD (y)	18	19	22	5	15	12	5
Outcome	Increased AD risk with longer duration of E + P use, no effect of vaginal estrogen or progestogen alone	Increased AD risk w/ both E alone and E + P and longer duration of use (>10 y in those <60 y of age) AND > 3–5 y in those > 60 y of age)	Increased AD risk w/ E + P (5–9 y, and 10 + y and < 1 y only in 60>) for E alone:<1 y HT OR = 1.11 (95% CI=1.02–1.22). Reduced risk for E alone > 10 y of use <60 all dementia	Decreased risk with greater duration of MHT, oral HT AD: 0.42 (0.40–0.44), for transdermal HT no effect AD. No difference between treatment types (data not shown for AD)	Increased AD risk with oral estrogen alone. No difference in AD risk with combination or transdermal estrogen (but no data given for AD alone)	Increased AD risk with HT use, no significant trend of dementia risk associated with duration of HT.	increased risk for ‘dementia’ w CEE + MPA but not CEE alone in women over 65
OR overall, E alone, combination for AD only	combination HR: 1.22 (95% CI 1.07–1.39)	E alone OR = 1.09, (95% CI 1.05–1.14) Combination OR = 1.17, (95 CI1.13–1.21)	E alone > 10 y OR = 0.85, (95% CI 0.76–0.94) Combination 5–9 y OR = 1.11, (95% CI 1.04–1.20) Combination>10 y OR = 1.19, (95% CI 1.06–1.33)	all MHT RR = 0.43, (95% CI 0.41–0.46)	oral E alone OR = 1.08, 95% CI 1.03–1.11	overall HR: 1.20 (95% CI 1.07–3.86)	Overall ns HR below cut-off dementia for CEE alone HR 1.77 (95% CI, 0.74–4.23)

AD, Alzheimer’s disease; BMI, body mass index; CC, Charlson comorbidity index; OR, odds ratio; RR, relative risk; HR, hazard ratio; COPD, chronic obstructive pulmonary disease; CKD, chronic kidney disease; CEE, conjugated equine estrogen; E, estrogen; P, progestogen; MPA, medroxyprogesterone acetate; y, years; SES, socioeconomic status; CI, confidence interval; MHT, menopausal hormone therapy; HRT, hormone replacement therapy. Authors/years of studies in bold indicate inclusion in the meta-analyses (see Fig. 1aA–D).

Fig. 1

A–D. MHT use (HRT) by treatment and prescription duration in women stratified by age. (A) Combined MHT in women under 60.

Fig. 1B

Combined MHT (HRT) in women over 60.

Fig. 1C

Estrogen alone (HRT) in women under 60.

Fig. 1D

Estrogen alone (HRT) in women over 60.

Our specific hypothesis included the investigation of additional AD risk with MHT for the older age group. On the basis of the literature, we investigated risk by duration of prescription for categories of <1, 1–3, 3–5, 5–10, and 10+ years. For Pourhadi et al.’s study [48], the duration of prescription category was converted to fit: 1–4 into 1–3; 4–8 into 3–5; 8–12 into 5–10; and >12 into 10+. Data were extracted and inputted to Excel tables and then checked and inputted into Revman 5.4.1., the software for the Cochrane medical review system to calculate the odds ratio (OR). The OR was used as outcome measure without adjustments for confounds, so these may slightly deviate from the reported OR, hazard ratio (HR) or relative risk (RR) as published in the papers and shown inTable 1.

We recalculated the OR for all studies, per group (<60 years of age, >60 years of age) by treatment (estrogen alone or combination MHT) and by duration (see above), using the formula:

$\frac{\begin{matrix} # AD cases having been prescribed MHT \\ / total # of people prescribed MHT \end{matrix}}{\begin{matrix} # AD cases not having been prescribed MHT \\ / total # of people not having been prescribed MHT \end{matrix}}$

The software calculated the significance of the z-score for overall effects for the ORs with a 95% confidence interval (CI) using random effects. OR (including CI) greater than 1.0 indicated an increased risk with MHT prescription, OR (including CI) less than 1.0 indicated a reduced risk with MHT prescription, and OR of 1.0 or CI including 1.0 indicated no difference in risk between the groups having been prescribed MHT or not. Heterogeneity was established by examining the significance of the Chi-square test and I², interpreted as <40% (not important), 30–60% (moderate), and 50–90% (substantial). Sensitivity analyses were also conducted by removing each study in turn (see Supplementary Tables) to investigate the effect this had on the overall OR and 95% CI.

RESULTS

In total, four large medical registry studies were included in the meta-analysis. The three observational studies were recently published from 2019 to 2023, and data were extracted from the national registries of Denmark [48], Finland [49], the USA [5, 6], and the United Kingdom [50], respectively. The medically registered nationwide RCT conducted in the US included analyses of women over 65 and was published in 2003/2004. However, the time frame of MHT prescribing across the studies was similar, e.g., from 1994 for Savolainen’s study [49] and from 1996 for Shumaker’s [5, 6] studies (see Table 1).

A total of 912,157 women were included in the meta-analyses, of whom 278,495 (30.5%) had developed medically diagnosed AD during the follow-up after they had been prescribed medically database registered HRT 1 or MHT. Women were between the ages of 50–80+ when they first started MHT, with a range of duration of use between 1 to 12 years. Please see Table 1 for the key characteristics of thesestudies.

Descriptive analyses of the studies in Table 1 suggested that only one US study [10] showed reduced AD risk with longer duration MHT prescriptions, whereas the six others all showed increased risk with larger MHT use. Whereas the UK study [50] showed a reduced risk for all dementia with estrogen-alone when prescribed for over 10 years in women under the age of 60, this was not reported in their separate AD analyses. However, estradiol alone prescriptions (but not conjugated equine estrogens) also rendered a small, reduced risk for AD for 1–5 years of prescription in this study. Transdermal and vaginal treatment overall conferred no risk.

Meta-analyses (see Fig. 1A–D) revealed that in women younger than 60 years of age (Fig. 1A) combination MHT did not reduce or increase risk for AD when this had been prescribed for less than 5 years, with a trend for a small increased risk between 5–10 years of prescription (OR = 1.13, 95% CI = 1.00–1.29, p = 0.06). But after 10 years of combination MHT prescriptions (OR = 1.18, 95% CI = 1.12–1.24), a significant increase in AD risk was observed. Heterogeneity was significant by subgroups only for MHT prescriptions between 5–10 years (p = 0.03 for Chi-square). Using sensitivity analyses, after removal of Pourhadi’s study [48] (but not after removal of either other study [49,50, 49,50]), heterogeneity was no longer significant (p = 0.99) and the OR now indicated a significant, but smaller, increased AD risk (Chi-square p = 0.99, OR = 1.06, 95% CI=1.00–1.13, p = 0.05) for 5–10 years of combination MHT prescription as well. Overall OR and 95% (OR = 1.98, 95% CI 1.02–1.13) did not change substantially after removal of the studies systematically (sensitivity analyses, see Supplementary Tables).

Women over the age of 60 (Fig. 1B) only showed significant risk for AD with combination MHT prescriptions between 5–10 years (OR = 1.22, 95% CI = 1.05–1.43), but not when this was prescribed for over 10 years or less than 5 years (where analyses showed significant heterogeneity, and so indicated low consistency and reliability). Heterogeneity only became non significant (p = 0.15, I² = 52%) when Vinogradova’s [50] study (but not Shumaker’s [5] or Savolainen’s [49] studies) was removed and AD risk was now also significant between 3–5 years of prescription (OR = 1.49, 95% CI = 1.02–2.18). Between 1–3/4 years of prescription, heterogeneity was significant and only one study [49] showed a significant risk, while the other did not [50]. This tentatively suggests increased AD risk up to 10 years of combination MHT prescription in women over 60. When Savolainen et al. [49] was removed, the overall OR was no longer significant (OR = 1.06, 95% CI = 0.97–1.16, compared to the overall OR with all studies included of 1.13, 95% CI1.03–1.23).

For estrogen alone in women younger than 60 years of age (Fig. 1C), no increased AD risk was seen using meta-analyses and heterogeneity was not significant (overall OR = 0.99, 95% CI = 0.93–1.05, Chi-square p = 0.16, I² = 32%, z = 0.24, p = 0.81). Removal of either study did not change the OR or 95% CI (see Supplementary Tables).

Prescription of estrogen alone in women over 60 years of age (Fig. 1D) showed a significant increased risk for AD with prescriptions from less than 1 year (OR = 1.17, 95% CI = 1.04–1.31), and 1–3/4 years (OR = 1.13, 95% CI = 1.04–1.22), but after this period, AD risk was no longer significant. Heterogeneity was significant for analyses of prescription over years 3–5 (Chi-square=11.43, p = 0.003, I² = 83%). After removal of [50] (but not after removal of [6]), heterogeneity was no longer significant (p = 0.74) and AD risk now was significantly increased (OR = 1.37, 95% CI = 1.10–1.71) for estrogen alone for 3–5 years in women over 60 years of age. Overall OR was 1.11 (95% CI = 1.03–1.19), removal of Savolainen et al. [49] changed this significantly (OR = 1.07, 95% CI = 0.96–1.19) but not after removal of the other individual studies (see Supplementary Tables).

DISCUSSION

Only one of our four included studies of large national registry based databases showed a reduced risk with MHT use [10]. The other large registry-based studies in Table 1 all showed increased risk for AD. Meta-analyses suggested that this increased risk was only significant after combination MHT prescriptions over 5 years, regardless of age. For women over 60, combination MHT prescribed for less than 5 years showed considerable heterogeneity, suggesting low consistency and low reliability. For women younger than 60, prescription of estrogen alone was not associated with increased AD risk regardless of duration, which was unexpected. Estrogen alone in women over 60 s years of age conferred significant AD risk when prescriptions had been given for up to 5 years. Importantly, however, the increased AD risk was overall small. Sensitivity analyses showed no differences after removal of individual studies for the overall OR for women younger than 60 for whom MHT might be considered. The fact that one study [49] was instrumental in driving significant effects in overall analyses in women older than 60 (for both estrogen alone and combination MHT) suggests that these data are perhaps less reliable, so this should be taken into account when reviewing thedata.

Limitations

There are several limitations to this meta-analysis. It was not possible to register the protocol for meta-analyses via PROSPERO (as this had already been done by others, but had not been carried out), but our search strategy complied with that of other meta-analyses and reviews. Another limitation is that the one study that showed benefits of MHT [10] could not be included in the meta-analyses, as they had not provided AD risk by duration of use for the separate treatments (estrogen alone or combination MHT) by age. We also pooled the other studies and recalculated OR. The other larger meta-analysis [17] had used adjusted HR to better approximate OR in pooled analyses, but inspection of our data suggested that their reported HR and 95% CI in the papers were not substantially different from our calculated OR. We could not use adjusted analyses, as all studies adjusted for different covariates and/or used these in different formats, so adjustments were not comparable. Instead, we recalculated the unadjusted OR for a better comparison. Inspection of the data again did not suggest substantial differences in the studies between our calculated OR and their adjusted OR for covariates, the 95% CIs and levels of significance. However, in one set of analyses [48], MHT use between 1–4 years conferred significantly higher risk for AD in their adjusted, but not in our unadjusted, analyses.

In addition, medical registries are not necessarily always updated by GP or other medical specialists. Reverse causality could also play a role, with GP or medical specialists deciding to stop MHT once women developed AD, as adherence to treatment could be expected to be problematic due to memory issues seen in AD. Lastly, the clinical diagnosis of AD in life is not always reliable, with many people with dementia suspected not to be diagnosed in the community and inter-rater reliability and validity of dementia diagnoses being moderate at best, even between well trained clinicians [2].

The MHT health belief bias

The Danish study [48] which had different outcomes in their analyses as compared to our (unadjusted) analyses, had alarmingly suggested that AD risk with combination MHT already increased after 1 year of prescriptions in women who were 45–55 years of age, with longer MHT use increasing the risk significantly, to almost doubling after 12 years of use [48]. We could not replicate their result of AD risk already found after 1, or even up to 5, years of prescription of combination MHT. Women who had undergone hysterectomy (who would have been prescribed estrogen alone) were not included in the Danish study. In this study, women with dementia were also more likely to have had less education, low SES, to live alone, and have hypertension, diabetes, and thyroid disease, but these variables were all controlled for in their statistical analyses [48]. This use of covariates could have explained differences between their OR and the unadjusted ORs in our study. Alternatively, in this study, MHT prescriptions were measured between 1995–1998, the height of enthusiasm for protective effects of MHT. This was before the publication of WHIMS in 2003 suggesting health risks with MHT [5], after which MHT prescriptions dropped significantly [50]. The Danish women could have been prescribed MHT at that time because they were thought to be at risk for AD and CVD (e.g., they had hypertension, diabetes, etc.) and that MHT would help prevent this. The belief was at that time that these risks could be reduced by giving MHT during the menopausal transition and beyond, which ideas had led to WHIMS. The authors controlled for some of these risk factors, but not for all. Another analysis of this Danish registry controlling for education suggested increased dementia risk with systemic combination MHT of 14% (95% CI = 1.01–1.28), but only in women who had been prescribed MHT between 1995–2003, and not in those who were prescribed MHT later, between 2005–2011 [45]. These analyses could substantiate the increased AD/CVD risk selection bias described above, as after 2003 (the year of publication of WHIMS data [5]) most women would no longer be prescribed MHT for prevention of AD or CVD. Alternatively, the shorter follow-up in the latter group could have prevented showing increased AD risk.

MHT health belief bias and age of women using MHT

Other studies that included prescription data from a different time frame (including the period after 2003) and controlling for various risk factors also found an increased overall risk with combination therapy (see Table 1). In these studies, the duration of MHT prescription, but also the age of women was important. For instance, the large Finnish study showed increased AD risk, but only after10 years of prescriptions in women younger than 60 of age, whereas duration of MHT prescription was not important for women >60 years to increase AD risk [49]. There was no difference in this study between using estradiol alone or combination therapy including progestogens. However, the analyses only controlled for age at initiation of MHT. In contrast, earlier analyses of this cohort (over a shorter time frame, with fewer AD cases) showed an increased risk of AD with shorter MHT prescriptions. Conversely, a reduced AD risk with prescriptions over 10 years was also reported in that paper [46]. This study had analyzed MHT prescriptions between 1995–2011. A co-morbidity score (which included morbidity associated with dementia) was included as a covariate in analyses, as well as low SES [43], gynecological cancers and surgery, which all were more common in women with AD. Women with AD in this study were also more likely to have used MHT, but also to have undergone surgical menopause. These Finnish women had used MHT for an average of 10 years, but were on average aged between 62–64 (range 57–69), when they initiated treatment. This would not be done for menopausal complaints. Similar to the Danish data, the shorter MHT duration AD risk increase in this older group may thus again be related to the enthusiasm at the time for MHT to be believed to protect against AD and CVD (i.e., before 2003). The WHIMS study reported increased dementia risk in women over 65 years of age already after 1 year of treatment, but only for combination MHT. However, WHIMS had very small numbers of women who developed dementia, which also included other types of dementia than the more common AD or vascular dementia types and where potentially biologically plausible benefits of MHT were less clear (e.g., for alcoholic dementia). Analyses of WHIMS could not be done for AD alone. These data combined suggest that perhaps AD (and CVD) showed a small increased risk after 5 years of use because women who were believed to be at risk for these morbidities selectively were prescribed combined MHT.

Possible reduced AD risk with use of estrogen alone

In contrast to some of the other studies, in the Finnish data, prescription of estrogen alone longer than 10 years reduced AD risk, but the CI was close to 1 (OR = 0.91, 95% 0.84–0.99). In a comprehensive UK study of GP databases linked to hospital and other databases [50], longer (5–10 or more years) of combination MHT prescription was associated with an increased AD risk, but estradiol (the most potent estrogen) alone, when prescribed for 1–5 years was associated with reduced AD risk. Overall, in our meta-analyses estrogen alone did not affect risk in women younger than 60 but conferred increased AD risk in women over 60 with prescriptions less than 5 years. This data confirms the healthy cell bias theory, in that (older) cells undergoing potential pathological change could show detrimental effects of estrogens. The short or long-term protective effect of estrogens was not seen overall, though.

In the UK study [50], reflecting the other observational studies, AD cases also had more risk of cardiovascular disease, hypertension, diabetes, and used anticholinergics, which were all less common in women who used MHT, but which were all controlled for in analyses. The UK data also confirmed the increase in MHT use since 1988 which peaked in 2000 (especially in women aged 60–69 years,conform the MHT health belief bias) and dropped after 2003 (the WHIMS publication date).

Health insurance registries and wealthy healthy user bias

The US study not included in the meta-analyses, but shown in Table 1 to reduce AD risk regardless of age with longer oral MHT used health insurance data from 2007–2016 [10]. Importantly, the majority of women in this registry had been prescribed MHT for less than a year. In contrast to the other studies using nationwide population and hospital registries, this data came from health insurance registries, which included perhaps a less representative group of the US population [45] and could perhaps be confounded by the wealthy-healthy user bias, while SES and education of these women were also not controlled for in analyses. The South Korean retrospective cohort study [51] also not included in our meta-analyses had also used a nationwide health insurance database but was perhaps more comprehensive (with > 90% of people in South Korea using this system) and SES was controlled for, as well as other factors (see Table 1), including age at, but not type of, menopause. In contrast, this South Korean study showed an increased risk of AD with oral estrogen alone, but not when prescribed transdermal, or when prescribed in combination with a progestogen [51]. In this study, the majority of women had been prescribed MHT for less than 2 years, but average follow-up was 15 years. A long follow-up of 12 years was also seen in a Taiwanese health insurance register also not included in the meta-analyses but also showing an increased risk of dementia with MHT, which was independent of duration of prescription [52]. Women in both East Asian cohorts were younger than 60 years old when MHT was first prescribed. However, with a lack of stratification of type of prescriptions by duration and age, it is unclear whether these East Asian data are comparable to the other Western studies, which also were longitudinal case controlstudies.

Risk of dementia with duration and type of progestogen

While even within the same registries mixed results were seen (e.g., [46, 49]), the combined data from all studies may suggest that negative effects on brain function may be seen in Western/North European women at risk for CVD/AD and/or with longerduration (5–10 years, >10 years) of combination MHT prescriptions. These risks may further depend on the type of progestogen used for combination therapy, where medroxyprogesterone acetate (MPA) was associated with CVD risk, but micronized progesterone was not [18]. Importantly, risk for breast cancer also starts increasing exponentially after 5–10 years of use, especially after combination systemic MHT. This was shown in the UK registered medical database, but this risk reduced after cessation of MHT [53]. In a worldwide meta-analyses, similar outcomes were found, but some residual breast cancer risk remained after stopping MHT for up to 5 years. This risk also further depended on larger duration of use [54].

Medical guidelines and prevention

In line with our meta-analyses, most medical guidelines (e.g., [4]) do not recommend MHT for prevention of CVD, dementia, or mood disorders [18]. Possibly MHT could be considered when women have surgical or early onset menopause (before age 45, although data for estrogen alone are mixed and some studies also suggested increased AD risk over 10 years of prescription or when initiated in women over the age of 60). MHT should also be considered if women have severe menopausal symptoms and there are no counterindications, such as deep vein thrombosis or stroke [55]. The role of sleep deprivation (and perhaps also stress resilience) brought about by hormonal changes, particularly in the perimenopausal stages, on later life dementia risk needs to be more closely investigated, with the potentially successful non-hormonal treatments including cognitive behavioral therapy [18].

Phytoestrogens treatment and AD risk

For phytoestrogens, in Indonesian observational studies [56], we found that tempeh, a fermented soy bean product, reduced AD risk, also when consumed by older people (over 68 years of age). However, in similar aged Indonesian, Chinese, and Japanese American older people, high tofu (soybean curd) consumption was associated with increased dementia and cognitive impairment risk [19, 56]. We also found that in older Japanese American men, high tofu consumption and high estrogen levels had a further conferred risk for possible dementia [19]. Similar negative effects were found for Japanese American women using MHT combined with high tofu consumption [56]. This may also explain the data from the East Asian National Health Insurance registries [51, 52] where estrogen treatment together with soy-based products may have rendered increased AD risk. Such risk for tofu consumption by itself was not seen in Western populations who consume on average much less soy and often have different gut microbiome, producing less estrogenic metabolites, such as equol [19]. Tempeh is a fermented whole soy bean probiotic with vitamin B12 and folate, which may offset its high estrogenic compounds [57] and which can also stabilize blood glucose levels to reduce risk for AD. Animal studies showed that tempeh reduced amyloid- β levels and improved memory after ovariectomy [56]. A RCT study found benefits on memory in 47 menopausal Indonesian and UK women already 24 hours after tempeh treatment [56]. Effects were most pronounced in the Indonesian women, suggesting differences in gut microbiota with higher exposure to phytoestrogens in their daily diet being a possible mediating factor [56]. Our recent review suggested some benefits of tempeh in older people with AD (in 2/3 studies, but not in the study of people with severe AD); with mild cognitive impairment (in 2/4 studies) and some beneficial effects on cognition in 3 studies of healthy older volunteers [58]. None of the included RCT had assessments for longer than 12–16 weeks. Tempeh would perhaps be particularly beneficial for women who have undergone chemical or surgical menopause for hormone sensitive cancers, as it is not known to increase cancer risk. Larger and longer better controlled trials remain to be done for both women undergoing early (surgical) menopause and people living with AD.

AUTHOR CONTRIBUTIONS

Eef Hogervorst (Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Resources; Software; Supervision; Validation; Visualization; Writing – original draft; Writing – review & editing); MJ Kuck (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Validation; Visualization; Writing – original draft; Writing – review & editing); Ahmet Begde (Data curation; Formal analysis; Investigation; Methodology; Software; Validation; Visualization; Writing – review & editing); Katie Hawkins (Data curation; Formal analysis; Writing – original draft).

Footnotes

ACKNOWLEDGMENTS

We would to acknowledge Dr. Yana Vinogradova and Professor Tom Dening for the timely delivery of their data for our calculations and Loughborough University for their IT support.

FUNDING

The authors have no funding to report.

CONFLICT OF INTEREST

The authors have nothing to disclose and were not funded for this review from pharmaceutical industry or other potential beneficiaries. Eef Hogervorst acted as dementia expert for the NICE committee to investigate risk and benefit of menopausal hormone therapy to update their guidelines until end January 2024. The current review was commissioned separately and was carried out independently of the work of the NICE committee, which used different search terms and analyses. This paper does not reflect the work or opinions from the NICE committee members or staff. Loughborough University funds EH and the Turkish Government funded AB for his PhD. KH and MJ funded themselves.

DATA AVAILABILITY

Data used for the meta-analyses will be made available for third parties upon request to the authors and are held in a repository at Loughborough University.

The supplementary material is available in the electronic version of this article: .

HRT was used in the figures as the studies referred to HRT in the results.

References

Tahami Monfared

, Byrnes

, White

, Zhang

(2022) Alzheimer’s disease: Epidemiology and clinical progression. Neurol Therapy11, 553–569.

Hogervorst

, Bandelow

, Combrinck

, Irani

, Smith

(2003) The validity and reliability of 6 sets of clinical criteria to classify Alzheimer’s disease and vascular dementia in post-mortem confirmed cases: Added value of a decision-tree approach. Dementia16, 170–180.

Gustavsson

, Norton

, Fast

, Frölich

, Georges

, Holzapfel

, Kirabali

, Krolak-Salmon

, Rossini

, Ferretti

, Lanman

, Chadha

, van der Flier

(2023) Global estimates on the number of persons across the Alzheimer’s disease continuum. Alzheimers Dement19, 658–670.

NICE-National Institute for Health and Care Excellence (2015) Menopause: Diagnosis and management report available via. https://www.nice.org.uk/guidance/ng23/chapter/recommendations. Accessed November 30, 2023.

Shumaker

, Legault

, Rapp

, Thal

, Wallace

, Ockene

, Hendrix

, Jones III

, Assaf

, Jackson

, Kotchen

(2003) Estrogen plus progestin and the incidence of dementia and mild cognitive impairment in postmenopausal women. JAMA289, 2651–2662.

Shumaker

, Legault

, Kuller

, Rapp

, Thal

, Lane

, Fillit

, Stefanick

, Hendrix

, Lewis

, Masaki

(2004) Conjugated equine estrogens and incidence of probable dementia and mild cognitive impairment in postmenopausal women: Women’s Health Initiative Memory Study. JAMA291, 2947–2958.

Hodis

, Mack

(2022) Menopausal hormone replacement therapy and reduction of all-cause mortality and cardiovascular disease: It is about time and timing. Cancer J28, 208–223.

Hogervorst

(2021) Hormone treatment and Alzheimer’s disease: Déjà vu or confused all over again?Drugs Aging38, 793–795.

Hogervorst

, O’Donnel

, Hardy

(2022) Menopause: HT’s protective effect on the brain may be overstated. The Conversation 182449. Online https://theconversation.com/menopause-HTs-brain-protecting-effect-may-be-over-stated-182449

10.

Kim

, Soto

, Branigan

, Rodgers

, Brinton

(2021) Association between menopausal hormone therapy and risk of neurodegenerative diseases: Implications for precision hormone therapy. Alzheimers Dement (N Y)7, e12174.

11.

Bustamante-Barrientos

, Méndez-Ruette

, Ortloff

, Luz-Crawford

, Rivera

, Figueroa

, Molina

, Bátiz

(2021) The impact of estrogen and estrogen-like molecules in neurogenesis and neurodegeneration: Beneficial or harmful?Frontiers Cell Neurosci15, 636176.

12.

Sahab-Negah

, Hajali

, Moradi

, Gorji

(2020) The impact of estradiol on neurogenesis and cognitive functions in Alzheimer’s disease. Cell Mol Neurobiol40, 283–299.

13.

Irving

, Hogervorst

, Oliviera

, Kivipelto

(2018), New developments in Dementia Prevention Research, Routledge, New York.

14.

Livingston

, Huntley

, Sommerlad

, Ames

, Ballard

, Banerjee

, Brayne

, Burns

, Cohen-Mansfield

, Cooper

, Costafreda

, Dias

, Fox

, Gitlin

, Howard

, Kales

, Kivimäki

, Larson

, Ogunniyi

, Orgeta

, Ritchie

, Rockwood

, Sampson

, Samus

, Schneider

, Selbæk

, Teri

, Mukadam

(2020) Dementia prevention, intervention and care. Lancet396, 413–446.

15.

Hogervorst

, Temple

, O’Donnell

(2023) Sex differences in dementia. Curr Top Behav Neurosci62, 309–331.

16.

Duan

, Li

, Xu

, Ding

(2021) Study on the neuroprotective effects of Genistein on Alzheimer’s disease. Brain Behav11, e02100.

17.

Nerattini

, Jett

, Andy

, Carlton

, Zarate

, Boneu

, Battista

, Pahlajani

, Loeb-Zeitlin

, Havryulik

, Williams

, Christos

, Fink

, Brinton

, Mosconi

(2023) Systematic review and meta-analysis of the effects of menopause hormone therapy on risk of Alzheimer’s disease and dementia. Front Aging Neurosci15, 1260427.

18.

Hogervorst

, Craig

, O’Donnell

(2022) Cognition and mental health in menopause: A review. Best Pract Res Clin Obstet Gynaecol81, 69–84.

19.

Soni

, Rahardjo

TBW

, Soekardi

, Sulistyowati

, Lestariningsih , Yesufu-Udechuku

, Irsan

, Hogervorst

(2014) Phytoestrogens and cognitive function: A review. Maturitas77, 209–220.

20.

Hogervorst

, Yaffe

, Richards

, Huppert

(2002) Hormone replacement therapy to maintain cognitive function in women with dementia. Cochrane Database Syst Rev3, CD003799.

21.

Hogervorst

, Yaffe

, Richards

, Huppert

(2009) HT to maintain cognitive function in women with dementia. Cochrane Database Syst Rev21, CD003799.

22.

Hogervorst

(2006) The short-lived effect of HT on cognition function. In The Effects of Estrogen on Brain Function., Rasgon

, ed. Johns Hopkins Press, Baltimore, pp. 46–78.

23.

Resnick

, Henderson

(2002) Hormone therapy and risk of Alzheimer disease: A critical time. JAMA288, 2170–2172.

24.

Gleason

, Dowling

, Wharton

, Manson

, Miller

, Atwood

, Brinton

, Cedars

, Lobo

, Merriam

, Neal-Perry

, Santoro

, Taylor

, Black

, Budoff

, Hodis

, Naftolin

, Harman

, Asthana

(2015) Effects of hormone therapy on cognition and mood in recently postmenopausal women: Findings from the randomized, controlled KEEPS-Cognitive and Affective Study. PLoS Med12, e1001833.

25.

Henderson

, St John

, Hodis

, McCleary

, Stanczyk

, Shoupe

, Kono

, Dustin

, Allayee

, Mack

(2016) Cognitive effects of estradiol after menopause: A randomized trial of the timing hypothesis. Neurology87, 699–708.

26.

Espeland

, Shumaker

, Leng

, Manson

, Brown

, LeBlanc

, Vaughan

, Robinson

, Rapp

, Goveas

, Wactawski-Wende

, Stefanick

, Li

, Resnick

(2013) Long-term effects on cognitive function of postmenopausal hormone therapy prescribed to women aged 50 to 55 years. JAMA Intern Med173, 1429–1436.

27.

Maki

, Gast

, Vieweg

, Burriss

, Yaffe

(2007) Hormone therapy in menopausal women with cognitive complaints. Neurology69, 1322–1330.

28.

Hogervorst

, Williams

, Budge

, Riedel

, Jolles

(2000) The nature of the effect of female gonadal hormone replacement therapy on cognitive function in post-menopausal women: A meta-analysis. Neurosci101, 485–512.

29.

Brinton

(2008) The healthy cell bias of estrogen action. Trends Neurosci31, 529–537.

30.

Sarkar

, Panja

, Sarkar

(2011) Survey of maintenance policies for the last 50 years. Int J Softw Eng Appl2, 130–148.

31.

Gustavson

, Elman

, Sanderson-Cimino

, Franz

, Panizzon

, Jak

, Reynolds

, Neale

, Lyons

, Kremen

(2020) Extensive memory testing improves prediction of progression to MCI in late middle age. Alzheimers Dement (Amst)12, e12004.

32.

Petersen

, Smith

, Waring

, Ivink

, Tangalos

, Kokmen

(1999) Mild cognitive impairment. Arch Neurol56, 303–308.

33.

Weber

, Rubin

, Schroeder

, Steffenella

, Maki

(2021) Cognitive profiles in perimenopause: Hormonal and menopausal symptom correlates. Climacteric24, 401–407.

34.

Hogervorst

, Yaffe

, Richards

, Huppert

(2002) Hormone replacement therapy for cognitive function in postmenopausal women. Cochrane Database Syst Rev3, CD003122.

35.

Campbell

, Whitehead

(1977) Oestrogen therapy and the menopausal syndrome. Clin Obstet Gynaec4, 31–47.

36.

Shieu

, Braley

, Becker

, Dunietz

(2023) The interplay among natural menopause, insomnia, and cognitive health: A population-based study. Nat Sci Sleep15, 39–48.

37.

Weber MT

, Maki

PM.

(2013) Cognition in perimenopause: The effect of transition stage. Menopause20, 511–517.

38.

Sherwin

(1994) Estrogenic effects on memory in women. Ann NY Acad Sci743, 213–230.

39.

File

, Heard

, Rymer

(2002) Trough oestradiol levels associated with cognitive impairment in post-menopausal women after 10 years of oestradiol implants. Psychopharmacology (Berl)161, 107–112.

40.

Matthews

, Kuller

, Wing

, Meilahn

, Plantinga

(1996) Prior to use of estrogen replacement therapy: Are users healthier than nonusers?Am J Epidemiol143, 971–978.

41.

Song

, Li

, Chen

, Wei

, Liu

, Cheng

(2020) The effect of estrogen replacement therapy on Alzheimer’s disease and Parkinson’s disease in postmenopausal women: A meta-analysis. Front Neurosci14, 157.

42.

Petitti

, Buckwalter

, Crooks

, Chiu

(2002) Prevalence of dementia in users of hormone replacement therapy as defined by prescription data.M. J Gerontol A Biol Sci Med Sci57, 532–538.

43.

Ruiz-Cantero

, Vives-Cases

, Artazcoz

, Delgado

, Calvente

MdMG

, Miqueo

, Montero

, Ortiz

, Ronda

, Ruiz

(2007) A framework to analyse gender bias in epidemiological research. J Epidem Commun Health61, ii46–ii53.

44.

Whitmer

, Quesenberry

, Zhou

, Yaffe

(2011) Timing of hormone therapy and dementia: The critical window theory revisited. Ann Neurol69, 163–169.

45.

Løkkegaard

, Thinggaard

, Nygaard

, Hallas

, Osler

, Christensen

(2022) Systemic hormone therapy and dementia: A nested case-control and co-twin control study. Maturitas165, 113–119.

46.

Imtiaz

, Tuppurainen

, Rikkonen

, Kivipelto

, Soininen

, Kröger

, Tolppanen

(2017) Postmenopausal hormone therapy and Alzheimer disease: A prospective cohort study. Neurology88, 1062–1068.

47.

Seshadri

, Zornberg

, Derby

, Myers

, Jick

, Drachman

(2001) Postmenopausal estrogen replacement therapy and the risk of Alzheimer disease. Arch Neurol58, 435–440.

48.

Pourhadi

, Mørch

, Holm

, Torp-Pedersen

, Meaidi

(2023) Menopausal hormone therapy and dementia: Nationwide, nested case-control study. BMJ381, e072770.

49.

Savolainen-Peltonen

, Rahkola-Soisalo

, Hoti

, Vattulainen

, Gissler

, Ylikorkala

, Mikkola

(2019) Use of postmenopausal hormone therapy and risk of Alzheimer’s disease in Finland: Nationwide case-control study. BMJ364, l665.

50.

Vinogradova

, Dening

, Hippisley-Cox

, Taylor

, Moore

, Coupland

(2021) Use of menopausal hormone therapy and risk of dementia: Nested case-control studies using QResearch and CPRD databases. BMJ374, n2182.

51.

Yuk

J-S

, Lee

, Park

(2023) Menopausal hormone therapy and risk of dementia: Health insurance database in South Korea-based retrospective cohort study. Front Aging Neurosci15, 1213481.

52.

Sung

, Tsai

, Kuo

, Lee

, Chou

, Chen

, Chou

, Sun

(2022) Use of hormone replacement therapy and risk of dementia. A nationwide cohort study. Neurology99, e1835–e1842.

53.

Vinogradova

, Coupland

, Hippisley-Cox

(2020) Use of hormone replacement therapy and risk of breast cancer: Nested case-control studies using the QResearch and CPRD databases. BMJ371, m3873.

54.

Collaborative Group on Hormonal Factors in Breast Cancer (2019) Type and timing of menopausal hormone therapy and breast cancer risk: Individual participant meta-analysis of the worldwide epidemiological evidence. Lancet394, 1159–1168.

55.

Hamoda

, Panay

, Pedder

, Arya

, Savvas

(2020) The British Menopause Society & Women’s Health Concern 2020 recommendations on hormone replacement therapy in menopausal women. Post Reproductive Health26, 181–209.

56.

Hogervorst

, Kassam

, Kridawati

, Soni

, Xin

, Shifu

, Rahardjo

(2017) Nutrition research in cognitive impairment/dementia, with a focus on soya and folate. Proc Nutr Soc76, 437–442.

57.

Hogervorst

, Smith

(2002) The interaction of serum folate and estradiol levels in Alzheimer’s disease. Neuro Endocrinol Lett23, 155–160.

58.

Handajani

, Hengky

, Schröder-Butterfill

, Hogervorst

, Turana

(2023) Probiotic supplementation improved cognitive function in cognitively impaired and healthy older adults: A systematic review of recent trials. Neurol Sci44, 1163–1169.