Lifetime Exposure to Endogenous Estradiol and Markers of Dementia Risk: Associations with Later Life Cognitive, Behavioural, and Functional Complaints

1. Introduction

Alzheimer disease (AD) is the most common cause of dementia, accounting for approximately 60-70% of all cases worldwide [1]. An estimated 55 million people currently live with dementia, a number projected to nearly triple by 2050 as populations age [2]. Notably, females represent nearly two-thirds of individuals with AD3, which cannot be fully explained by greater longevity alone [4]. This sex disparity underscores the importance of understanding biological and life course factors that may confer susceptibility or resilience to AD in females.

Although recent advances in disease-modifying therapies have shown promise for some patients, their impact on clinical outcomes remains modest [5,6]. This has increased the importance of identifying early risk markers and modifiable factors that precede the onset of AD. One such avenue involves understanding how hormonal exposures across the female lifespan correlate with later-life brain health. From menarche (first menstruation) through the reproductive years (including pregnancy) to menopause (permanent cessation of menses), females experience dynamic fluctuations in reproductive hormones, particularly estradiol (E2), the most abundant and potent form of circulating estrogen. E2 exerts widespread neuroprotective effects [7] including supporting vascular integrity [8]; promoting synaptic plasticity [9]; modulating neurotransmitter [10] and inflammatory systems [11]; and facilitating the clearance of amyloid-beta and phosphorylated tau [7], hallmark proteinopathies of AD. Consequently, variations in E2 exposure have been proposed as one mechanism linking female reproductive aging to later-life brain health and AD risk.

Beyond post-menopausal decline in E2, lifetime exposure to endogenous estradiol (L_EE2) may also play a role in shaping brain aging trajectories. L_EE2 duration can be approximated by the interval of time between menarche and menopause [12], with additional years of pregnancy contributing to exposure given the markedly elevated E2 levels during gestation [13,14,15]. Shorter L_EE2 has been associated with poorer performance on cognitive tasks (including poorer delayed memory recall [12] and global cognition [15]), and greater risk of white matter hyperintensity burden [8,14] and dementia16 in later life. Conversely, longer LEE2 may confer neuroprotective benefits through prolonged exposure to the supportive effects of E2 on neural [15], vascular [14], and inflammatory [17] processes.

While emerging evidence links L_EE2 to cognitive outcomes, few studies have examined the association of L_EE2 with other early markers of neurodegenerative disease, such as neuropsychiatric symptoms (NPS) and functional decline. Both NPS and functional decline are predictive of dementia risk and progression and may precede overt cognitive impairment [18]. When NPS are later-life emergent and persistent, they are classified as mild behavioural impairment (MBI) [19], which has been associated with neurodegeneration [20] and AD cerebrospinal21 and plasma [22,23] fluid biomarkers, even after accounting for cognitive status. Similarly, subtle functional difficulties (e.g., managing medications, preparing meals, travelling, and managing finances) may also signal early-stage neurodegenerative changes [24,25]; recent work has linked mild functional impairment in cognitively unimpaired older adults to both incident dementia and AD biomarkers [26,27]. Although no studies to date have directly examined LEE2 in relation to MBI or functional decline, premature menopause has been linked to greater depressive symptoms [28] and poorer physical function [29] in later life, suggesting that duration of E2 exposure may influence behavioural and functional outcomes. Thus, extending investigations of L_EE2 beyond cognition may strengthen early identification of females at elevated risk of neurodegeneration.

In addition to endogenous E2 exposure, menopause hormone therapy (MHT) may also influence brain aging and dementia risk, though findings remain inconsistent [30]. Variability in MHT timing of initiation, duration of use, and formulation of E2, as well as individual health characteristics may contribute to these inconsistencies [31]. These findings underscore the need to account for MHT use in L_EE2 models when evaluating their effects on brain aging.

The present study examined the associations between L_EE2, cognition, behaviour, and function in a sample of postmenopausal females. We hypothesized that longer L_EE2 would associate with lower severity of cognitive, behavioural, and functional symptoms in later life, reflecting a potential protective effect of prolonged E2 exposure on the aging female brain.

2. Materials and Methods

2.1. Study Design

Data were drawn from the Canadian Platform for Research Online to Investigate Health, Quality of Life, Cognition, Behaviour, Function, and Caregiving in Aging (CAN-PROTECT) [32], a digital epidemiology platform longitudinally investigating risk and resilience in brain aging [33,34,35]. To be eligible for CAN-PROTECT, participants must be aged 18 years or older, reside in Canada, be dementia-free at enrollment, and have access to an internet-connected computer or tablet. Participants complete annual mandatory neuropsychological assessments and demographic questions, as well as optional assessments of cognition, behaviour, function, quality of life, medical and psychiatric history, and lifestyle. Participants who reported female sex at birth were also invited to complete a fertility and menopause questionnaire. All participants provide informed consent electronically at registration. The CAN-PROTECT study was approved by the Conjoint Health Research Ethics Board at the University of Calgary, with recruitment ongoing since March 8, 2023.

2.2. Participants

Baseline demographic, cognitive, behavioural, functional, and reproductive health data were available for 1999 participants. Inclusion in the analysis required complete data on neuropsychological battery tests, subjective cognitive, behavioural, and functional measures, reported female sex at birth, self-reported post-menopausal status, and age at menarche and menopause. Of these, 1168 met inclusion criteria; however, two participants were excluded due to biologically implausible values for age at menarche (i.e., menarche after 40 years of age). The final analytic sample comprised 1166 participants (Figure 1).

2.3. Measures

2.3.1. Lifetime Exposure to Endogenous E2

Lifetime exposure to endogenous E2 (L_EE2) was estimated by using self-reported reproductive history from the fertility and menopause questionnaire. Specifically, reproductive span was calculated as the difference between reported age at menopause and age at menarche [12]. To account for elevated E2 during pregnancy [13], an additional 0.75 years (equivalent to nine months) was added for each reported biological child, an approach previously used to examine LEE2 and white matter hyperintensities [14]. For regression analyses, LEE2 was scaled in five-year increments to improve interpretability of effect estimates. Data on miscarriages and preterm births are not collected in CAN-PROTECT and therefore were not included in L_EE2 estimates.

2.3.2. MHT Use

Ever use of MHT was self-reported and included unopposed estrogen (e.g., E2, conjugated estrogens) and opposed estrogen (e.g., combination estrogen and progestin) forms. Data on MHT use indicated for reasons other than menopause symptom treatment are not collected in CAN-PROTECT and thus, were not identified as MHT users. MHT use was categorized as either never or ever use.

2.3.3. E2-Related Reproductive Variables

Menopause type was categorized as spontaneous, surgical, or due to other reasons. CAN-PROTECT captures menopause resulting from oophorectomy or hysterectomy as a single category, precluding distinction between these two surgical procedures.

2.3.4. Objective Cognition

Objective measures of cognition were assessed using the validated CAN-PROTECT neuropsychological battery [36], which is comprised of six assessments including: Trail Making B, Switching Stroop, Self-Ordered Search, Paired Associate Learning, Verbal Reasoning, and Digit Span.

Raw test scores were extracted across the six domains and winsorized at the first and 99^th percentiles to minimize outlier effects. Each task was then standardized into a z-score based on the sample distribution, with time-based measures reverse-coded so that higher z-scores consistently reflected better performance. Domain composite scores were calculated as the mean of all non-missing standardized task scores within each domain, and a global composite neuropsychological score was derived as the mean of all available task z-scores per participant.

2.3.5. Subjective Cognition

Subjective cognition was assessed using the Revised Everyday Cognition (ECog-II) scale [37], a 41-item measure developed to detect subtle cognitive impairment in populations at risk for dementia. Items cover memory, language, visual-spatial and perceptual, planning, organizational, and executive function, with participants rating perceived change over the past ten years on a scale from 0-3 (0=no change, 1=occasionally worse, 2=consistently a little worse, 3=much worse). Total ECog-II scores were calculated by summing all item severity scores, with higher scores indicating greater perceived impairment.

2.3.6. Subjective Behaviour

Subjective behavioural symptoms were assessed using the Mild Behavioral Impairment Checklist (MBI-C) [38], a 34-item measure design to capture late-life onset, persistent (≥6 months), and impactful behavioural changes not explained by established psychiatric conditions or other diagnoses. The MBI-C covers five domains: decreased motivation, affective dysregulation, impulse dyscontrol, social inappropriateness, and abnormal perception and thought content. Items are rated from 0-3 (0=no symptom, 1=mild symptom, 2=moderate symptom, 3=severe symptom). Total MBI-C scores were calculated by summing all item severity scores, with higher total scores indicating greater behavioural disturbance.

2.3.7. Subjective Function

Subjective functional ability was measured using the Standard Assessment of Global Everyday Activities (SAGEA) scale [39], a 15-item, multidomain measure of functional capacity over the past month. The SAGEA assesses instrumental activities of daily living, basic activities of daily living, cognition, social participation, and mobility. Items are rated on a 0-3 scale (0=no impairment, 1=mild impairment, 2=moderation impairment, 3=severe impairment), with higher scores indicating greater functional impairment. For four activities, participants are additionally asked whether they require assistance from another person; endorsement of assistance contributes additional points, up to a maximum of 3 points per item. Total SAGEA scores were calculated as the sum of all domain scores.

2.4. Statistical Analysis

Participant demographics and outcome variables were summarized using descriptive statistics (count, percentages, mean, and standard deviations). Distributions of outcome variables were visually inspected using histograms and Q-Q plots and further assessed for skewness, kurtosis, and dispersion to guide model selection. L_EE2 was modeled as a continuous exposure variable, scaled in five-year increments to enhance interpretability of effect estimates.

The relationship between L_EE2 (exposure) and neuropsychological performance (global and domain) was modeled using linear regressions. Comparatively, the associations of L_EE2 (exposure) and the outcome variables, ECog-II, MBI-C, and SAGEA total scores, were modeled as overdispersed count outcomes using separate negative binomial regressions. Exponentiated coefficients from these models (using a log link) are presented as count ratios (CRs), representing the proportional change in the expected count of the outcome per one-unit increase in the predictor. This terminology was used instead of incidence rate ratios because no offset term –used to account for differing exposure times or observation periods– was included; thus, the models estimated counts rather than rates.

All models were adjusted for MHT use (never versus ever), menopause type (spontaneous, surgical, or other), age, years of education, and ethnocultural (European versus no European) background.

3. Results

A total of 1166 females were included in the analyses. Participants had a mean age of 64.0±7.4 years and reported a mean number of 15.8±4.5 years of education. Most participants (85.2%) reported at least some European ethnocultural background (Table 1).

The mean age at menarche was 12.7±1.5 years, while menopause onset occurred at a mean age of 49.6±5.8 years. Most participants experienced spontaneous menopause (76.1%). While 23.2% of the sample reported having no biological children, the majority reported at least one, with an average of 1.3±0.9 years pregnant. Combining reproductive period with years pregnant, the mean years of L_EE2 was 38.2±5.9. Approximately 35.2% reported MHT use at some point.

On average, participants reported a global neuropsychological performance z-score of 0.0±0.5, ECog-II severity score of 12.5±11.7, MBI-C severity score of 5.9±7.7, and SAGEA severity score of 2.8±3.5.

3.1. L_EE2 and Cognition

L_EE2 was neither associated with global or domain-specific neuropsychological performance (Table 2), nor with ECog-II score (CR= 0.97, 95% CI [0.92, 1.02], p=0.205) (Figure 2).

3.2. L_EE2 and Behaviour

L_EE2 was associated with severity of MBI symptoms (Figure 2). Each additional five years of L_EE2 corresponded to an estimated 9.0% lower expected MBI-C total score (CR= 0.91, 95% CI [0.84 0.98], p=0.009).

3.4. L_EE2 and Function

L_EE2 was associated with severity of subjective functional impairment (Figure 2). Each additional five years of L_EE2 corresponded to an estimated 9.0% lower SAGEA total score (CR=0.91, 95% CI [0.86, 0.98], p=0.009).

3.5. MHT Use

MHT ever use was not associated with subjective cognitive, behavioural, or functional outcomes (Table 3). Similarly, no associations were observed between MHT use and most neuropsychological measures, though MHT use was associated with better verbal reasoning performance (b=0.127, 95% CI [0.02, 0.23], p=0.016).

4. Discussion

Among 1166 postmenopausal females without a diagnosis of dementia, longer L_EE2 was associated with lower severity of behavioural and functional symptoms. No statistically significant associations were observed between L_EE2 and objective or subjective cognitive outcomes.

In the present study, L_EE2 associations with all measures of cognition were not statistically significant. Although longer L_EE2 has been proposed to confer neuroprotective effects on cognition, findings across studies remain mixed [40,41] and are further compounded by limited investigation. Some studies report that among cognitively unimpaired females, longer reproductive spans are associated with reduced subjective cognitive complaints [15,42], lower white matter hyperintensity burden [14], and better memory [12] and verbal fluency [43]. Differences between our findings and prior work may reflect variation in how L_EE2 is operationalized. In this study, L_EE2 was defined as reproductive span and adjusted for parity, whereas other studies additionally adjusted for breastfeeding [15,42], a factor not captured in CAN-PROTECT. Notably, studies incorporating breastfeeding [15,42] into L_EE2 calculations have reported positive associations with cognition, while studies limited to reproductive span alone have not consistently observed such relationships [40]. These findings suggest that L_EE2 characterization may influence observed relationships. Additionally, our participants were highly educated (mean of 15.8 years of schooling), and greater years of education are associated with higher cognitive reserve [44,45], an established resilience factor against cognitive decline and dementia. Higher educational attainment may contribute to attenuating detectable associations between L_EE2 and cognitive performance.

Extending beyond cognition, the present findings link longer L_EE2 to lower severity of subjective behavioural and functional complaints, markers that have been largely overlooked in understanding dementia risk early in the disease course. Although limited, preliminary work has shown that earlier menopause [28] and shorter LEE2 [46,47] is associated with more depressive symptoms in later life. In parallel, another study found that among postmenopausal females, spontaneous premature menopause was associated with later-life physical functional decline [29], including lower gait speed and grip strength, when adjusted for age and study site. However, the association between premature menopause and grip strength was attenuated when adjusted for education, body-mass index, MHT use, and history of smoking. Our findings further align with prior work showing that greater menopause symptom burden [48], which may reflect the intensity of hormonal fluctuations during the menopause transition [49], is associated with more severe MBI symptoms. These results collectively suggest that both the timing and duration of estrogenic exposure may shape later-life brain health. Although functional symptoms were not examined in the prior work, evidence links menopause to greater functional difficulties in later life [50,51], potentially due to long-term impact of E2 loss on neural [52], vascular [53], and skeletal [54,55] systems, independent of age. Thus, lower E2 exposure -whether from earlier menopause or shorter L_EE2- may contribute to greater behavioural and functional complaints. In comparison, prolonged exposure may preserve brain integrity through its effects on vascular health [8], synaptic plasticity [9], neurotransmitter regulation [10], inflammation [11], and the clearance of AD pathological proteins [7,52]. These mechanisms may underpin behavioural and functional resilience, analogous to the protective effects of greater years of education or factors that promote higher cognitive reserve [44,45]. Collectively, these findings expand current evidence by emphasizing that associations between L_EE2 and AD risk may manifest not only through cognitive pathways, but also through behavioural and functional changes that are equally relevant to dementia risk in females.

In contrast to our L_EE2 findings, MHT use was neither associated with most objective and subjective cognitive measures, nor with subjective behavioural or functional outcomes. However, MHT use was associated with better verbal reasoning performance. Some prior studies have similarly reported improved processing speed among MHT users compared to non-users [56,57], although findings remain inconsistent. Variability across studies may reflect differences in MHT duration and formulation, as well as the timing of MHT initiation, which was not directly examined in the present study. Moreover, these results align with broader literature demonstrating inconsistent associations between MHT use and later-life brain outcomes [30,44,58]. For example, while longer LEE2 has been linked to lower white matter hyperintensity burden, no comparable association was observed for exogenous E2 use [14], which comprised of both MHT and hormonal contraceptives used in earlier life. Conversely, another study [59] found that birth control pill and MHT use following premature oophorectomy associated with better episodic memory and visuospatial processing in later life. Similar findings have been noted among females who experienced spontaneous menopause at older ages 60 but these studies only explored age at menopause and not LEE2 duration, which may influence outcomes. Importantly, our primary analyses classified MHT use as never versus ever use, an approach that may obscure potential differences based on recency of use. To explore whether associations differed according to MHT status (never, past, or current use), we conducted post-hoc sensitivity analyses. Although MHT status was not directly assessed in CAN-PROTECT, we derived status using reported age at initiation and duration of use to estimate the end age relative to participants’ current age. In these exploratory models, current MHT use was independently associated with better global neuropsychological performance (p=0.015) and verbal reasoning (p<0.001) compared to non-users, whereas past use was not associated with cognitive performance. No associations were observed between MHT status and behavioural and functional outcomes. These findings raise the possibility that associations between MHT and cognition may depend on recency of exposure rather than lifetime history of use alone.

Nevertheless, given the cross-sectional design and post-hoc nature of these analyses, these results should be interpreted cautiously. Furthermore, longitudinal investigations incorporating detailed characterization of MHT duration, formulation, and timing of initiation, in addition to MHT status, should be conducted to clarify relationships.

Several limitations to the current study warrant consideration. First, the cross-sectional design precludes inference about temporal relationships between L_EE2 and later-life dementia risk markers. Longitudinal data are needed to clarify whether the modest effects observed predict future decline, as even small differences may be meaningful in a largely unimpaired sample. Second, our L_EE2 measure represents a crude proxy of cumulative endogenous E2 exposure and does not capture other medical, lifestyle, and reproductive factors that may influence hormonal variability. Third, MHT use was categorical and lacked critical details about timing of initiation, duration, dosage, or formulation -factors that may modify the study outcomes [30,61]. Relatedly, information on hormonal contraceptive use (i.e., birth control pills) during reproductive years was unavailable, despite widespread use⁶² of estrogen-containing birth control pills that may alter L_EE2. Finally, recall bias in self-reported reproductive history (i.e., age at menarche and menopause) may affect L_EE2 accuracy. This limitation extends to much of the CAN-PROTECT dataset, which relies on retrospective self-reporting. Future longitudinal studies should aim to replicate these findings using biomarker and hormonal data.

Despite these limitations, the study offers several notable strengths. Leveraging a large, well-characterized, population-based Canadian cohort enhances the generalizability of findings. Our operationalization of L_EE2 includes both reproductive duration and pregnancies, providing a more comprehensive estimate of cumulative endogenous E2 exposure. Prior studies often consider only reproductive duration, overlooking pregnancies, an experience that can substantially increase E2 levels [13], and may independently influence dementia risk [63,64] and resilience [65]. Finally, by simultaneously examining cognition, behaviour, and function, this study expands the scope of female brain aging research and supports a multidimensional approach to identifying early indicators of dementia risk.

5. Conclusions

We provide evidence from a cross-sectional study of dementia-free postmenopausal females that longer L_EE2 may associate with lower severity of subjective behavioural and functional complaints. These results underscore the potential role of L_EE2 in supporting later-life brain health. Furthermore, these findings highlight the importance of expanding female dementia risk identification beyond cognition to include behavioural and functional changes.