Abstract
Background:
Prior research suggests a link between menopausal hormone therapy (MHT) use, memory function, and diabetes risk. The menopausal transition is a modifiable period to enhance long-term health and cognitive outcomes, although studies have been limited by short follow-up periods precluding a solid understanding of the lasting effects of MHT use on cognition.
Objective:
We examined the effects of midlife MHT use on subsequent diabetes incidence and late life memory performance in a large, same-aged, population-based cohort. We hypothesized that the beneficial effects of MHT use on late life cognition would be partially mediated by reduced diabetes risk.
Methods:
1,792 women from the Wisconsin Longitudinal Study (WLS) were included in analysis. We employed hierarchical linear regression, Cox regression, and causal mediation models to test the associations between MHT history, diabetes incidence, and late life cognitive performance.
Results:
1,088/1,792 women (60.7%) reported a history of midlife MHT use and 220/1,792 (12.3%) reported a history of diabetes. MHT use history was associated with better late life immediate recall (but not delayed recall), as well as a reduced risk of diabetes with protracted time to onset. Causal mediation models suggest that the beneficial effect of midlife MHT use on late life immediate recall were at least partially mediated by diabetes risk.
Conclusion:
Our data support a beneficial effect of MHT use on late life immediate recall (learning) that was partially mediated by protection against diabetes risk, supporting MHT use in midlife as protective against late life cognitive decline and adverse health outcomes.
Keywords: Adult onset diabetes mellitus, Alzheimer’s disease, estrogen, immediate memory, mediation, menopause
INTRODUCTION
Aside from a central role in reproductive behaviors and physiology, circulating estrogens are also an important modifier of non-reproductive systems including the brain and metabolic functioning. Endogenous estrogen levels decrease 10-fold at menopause [1], where long-term hormonal deprivation has been shown to have direct functional consequences on the brain [2, 3], often compounded by indirect interactions with other age-related physiological alterations such as changes in glucose uptake and insulin signaling pathways [4–6]. In the brain, prior research supports a direct neuroprotective effect of estrogen through promotion of enhanced synaptic formation and spine density in the hippocampus and prefrontal cortex [7, 8], greater choline acetyltransferase activity in the basal forebrain [9], increased cerebral blood flow and glucose metabolism [10], as well as a reduced deposition of amyloid-β [11, 12] and tau [13]. Circulating estrogen acutely influences metabolic functioning in the brain, particularly within regions subserving learning and memory that show a high concentration of estrogen receptors [14]. For example, women taking menopausal hormone therapy (MHT) containing estrogen show higher glucose uptake in brain regions preferentially implicated in Alzheimer’s disease compared to non MHT users [6]. Estrogen levels in cerebrospinal fluid have also been positively correlated with glucose uptake in the left hippocampus [15]. In community-based cohorts, declining estrogen levels during menopause has been shown to exacerbate age-related declines in learning and memory [16], whereas the use of estrogen replacement therapy during menopause predicts higher late life cognitive functioning (particularly in memory) [16–19] and slower rates of age-related cognitive decline [19, 20].
Considering evidence of the beneficial effects of estrogen on brain structure and function, the use of MHT to prolong estrogen exposure in perimenopausal women has emerged as a target of interest in prevention trials [21]. The Women’s Health Initiative (WHI) was one of the first and largest trials evaluating the effects of MHT use on cognitive and health outcomes although it was prematurely terminated based on the combined form of MHT assignment leading to increased risk of breast cancer, with higher rates of adverse cardiovascular events also reported [22]. However, women recruited into this study were well beyond menopausal age (average age of 63 years) and had a high burden of preexisting cardiovascular conditions, raising concern about the generalizability of these findings to younger and healthier perimenopausal women. Smaller observational and clinical studies among women closer to the menopausal transition have found neutral to positive net effects of MHT use on all-cause mortality [23] and cognitive outcomes particularly within the domain of learning and memory [24, 25]. These discrepant findings of MHT use on cognition dependent on when MHT is initiated in relation to menopause support a ‘critical window’ hypothesis where the perimenopausal phase is seen as the optimal period for estrogen replacement to benefit later life cognitive function and modify disease risk [26, 27]. Although randomized clinical trials typically lack adequate follow-up periods to assess the protracted effects of midlife MHT use on cognitive functioning decades later in late life, one follow-up study reported a 64% reduced risk for late life cognitive impairment among women enrolled in an active arm of earlier MHT trials [28]. Favorable findings on the effects of midlife MHT use on late life cognition are also apparent in population-based studies, where MHT use initiated close to the menopausal transition predicts higher late life global cognition [20], as well as a 26% reduced risk of dementia [29]. Although other longitudinal studies have failed to show an association between MHT use and late life cognition [30, 31], methodological differences in MHT timing, length of treatment, duration of follow-up period, and variability in participant characteristics may account for this observed heterogeneity across cohorts [32].
In addition to studies elucidating the direct neuroprotective effects of estrogen in the brain and lasting impacts on cognition, estrogen may also indirectly influence cognition through metabolic pathways [5]. Estrogen levels are positively associated with insulin sensitivity and glucose utilization in women [4, 6], and when natural estrogen levels decline during menopause, diabetes risk significantly increases [33]. Diabetes is a well-established risk factor for dementia [34], with a recent pooled analysis of over 2.3 million individuals suggesting a 60% increased risk for all-cause dementia among those with a history of type 2 diabetes [35]. Diabetes is also associated late life cognitive deficits most notably in the domains of processing speed, attention, learning and memory [34, 36–38]. The mechanisms relating diabetes to cognitive decline are multifactorial, with diabetes shown to incur endothelial changes, inflammation, mitochondrial dysfunction, and contribute to neurodegenerative processes [39]. Similar to dementia prevention trials discussed above, a beneficial effect of MHT use on diabetes risk is most readily apparent in studies of younger perimenopausal women, whereas MHT initiation in late life seems to interact with pre-existing diabetes to double the risk for dementia [40]. Following similar logic as the critical window hypothesis, optimal MHT initiation in relation to diabetes status lends support for a “healthy cell bias” of estrogen action, where the beneficial effects of MHT on glucose metabolism depend on the base-line metabolic health of the estrogen targeted cells [2]. Although no known clinical trial has examined the effects of MHT on diabetes as a primary endpoint, several large studies have shown that early MHT use is protective against diabetes risk by improving β-cell insulin secretion, glucose effectiveness, and insulin sensitivity [33]. Notably, the incidence of new onset type 2 diabetes increases significantly at midlife, with a prevalence of 14% among those aged 45–65, and 25% for those over age 65 [41]. Prolonging the beneficial metabolic actions of estrogen through MHT use during the menopausal transition has been shown to protect against incident diabetes risk [42, 43] as well as optimize neuronal function and metabolism [2, 4–6, 10].
Taken together, converging veins of research have demonstrated a link between optimally timed midlife MHT use and protection against incident diabetes in women, as well as support that perimenopausal MHT use may offer late life cognitive benefit and reduce dementia risk [21]. Although several direct mechanisms relate estrogen exposure to both acute and prolonged optimization of brain function and cognition, evidence also exists for indirect estrogen-mediated effects on cognition through metabolic pathways that impact diabetes risk. Despite this plausible temporal cascade linking early hormonal changes during menopause to increased risk for diabetes to subsequent late life cognitive decline [40], no known study has examined whether the beneficial effects of midlife MHT use on cognition may be partially explained by protection against incident diabetes-associated cognitive decline in a cohesive statistical model. Existing longitudinal cohort studies have either neglected diabetes status, or treated it as a confounding nuisance variable of no interest, precluding understanding of the magnitude of potential direct and indirect mechanisms of prolonged estrogen exposure on late life cognition. To address this gap in knowledge, we leverage the Wisconsin Longitudinal Study (WLS), a large randomly selected population-based cohort with over 60 years of prospective follow-up data spanning the full life course to examine the prolonged effects of MHT use history reported at midlife on 1) emergent diabetes incidence and 2) objective late life memory performance. We hypothesize that a history of midlife MHT use will be associated with better late life memory performance in this large randomly selected population-based cohort, and that this effect would be at least partially mediated by an indirect and protective effects of MHT in reducing diabetes risk and associated cognitive decline.
MATERIALS AND METHODS
Participants
This study used data from the WLS, a same-aged cohort of 1/3 of all Wisconsin high school graduates in 1957 (age 18), with over 60 years of prospective follow-up data obtained at ages 25 (1964), 36 (1975), 54 (1993), 64 (2003–4), and 71 (2011–12) [44]. Data collection waves referenced in the current study include “midlife” (2003–4; age 64) and “late life” (2011–12; age 71). A new data collection wave is currently underway to assess dementia prevalence in WLS, with the study protocol and cohort profile described elsewhere [45]. Retention for this sample is remarkably high compared to other longitudinal studies, with a 75% response rate among living participants (2,060 individuals in the original sample were deceased and 2,015 declined to continue participation at last follow-up). Participants were excluded from analyses based on a reported history of Type 1 diabetes (n = 7), an onset of diabetes prior to reported age of MHT start (n = 14), a history of stroke (n = 264), or missing primary endpoints (n = 596). A total of 1,792 women were included based on complete data for late life memory measures, MHT use history, and diabetes status. Figure 1 reports overall sample retention across study waves and participant selection criteria for the current study.
Fig. 1.
Flow chart of overall participant retention data across all Wisconsin Longitudinal Study (WLS) data collection waves (1957, 1975, 1993, 2004, and 2011), as well as sample selection for the current study. The current study excluded males, individuals who were not in the 80% randomly selected for cognitive testing, those with a history of stroke, those with Type I Diabetes, those with diabetes incidence prior to starting MHT, and those missing primary endpoints. The final sample size included 1,792 women.
Standard protocol approvals, registrations, and patient consents
This study was performed in compliance with the guidelines of human experimentation and the protocol was approved by the Institutional Review Board at the University of Wisconsin School of Medicine and Public Health (2020–1277). All participants provided informed consent and procedures were done in accord with the Helsinki Declaration of 1975.
Primary study variables
Basic demographic data include highest level of educational attainment (years of education equivalent), age at each data collection wave, and self-reported race. Apolipoprotein E4 allele (APOE4) positive status (≥1 E4 allele) was determined through saliva-based genotyping. Depression was assessed using the modified Center for Epidemiological Studies – Depression inventory (CES-D) [46], and scored continuously as the sum of the number of days (1–7) each of the 20 items were endorsed per week (range = 0–140). Late life memory was quantified by total correct words on immediate and delayed recall trials of a 10-item word list memory task, with two versions of the task randomly assigned across participants.
Health variables were collected during the inperson WLS Health Module fielded at late life (2011). The presence of hypertension, hyperlipidemia, heart disease, stroke, cancer, and diabetes were obtained from self-reported history of a physician-based diagnosis. History of heart disease was derived from the question: “Has a doctor ever told you that you had a heart attack, coronary heart disease, angina, congestive heart failure, or other heart problems?” Duration of diabetes was calculated as the reported age of diabetes diagnosis subtracted from age at the 2011 survey. Diabetes management questions were asked of all participants who endorsed a history of diabetes and included treatment by use of oral medication and/or insulin, and subjective assessment of habitual glucose control. Self-reported smoking data was obtained at late life and quantified as the number of years of regular smoking behavior across the lifespan. Midlife obesity was defined as a body mass index (BMI) at or over 35 calculated by self-reported heights and weights.
Reproductive variables and MHT use history were obtained from surveys on women’s health collected in 2004 (age ~64 years), when most participants had already entered the menopausal transition. Reproductive duration was calculated as the difference between reported ages of menarche and last menstruation. Suspected surgical menopause was defined as a reported surgical history of bilateral oophorectomy and/or hysterectomy occurring before the age of 40. MHT formulation type was ascertained by response to survey questions asking whether MHT was comprised of Estrogen only (unopposed Estrogen) or Estrogen plus Progesterone (opposed Estrogen), along with the corresponding age that each formulation was initiated. Because MHT formulation was based on self-report, and because 105 women reported a history of taking both formulations, primary analyses were based on the history of ever taking MHT containing Estrogen (regardless of formulation), with the effects of MHT formulation type examined as sensitivity analyses. Optimal MHT use was defined as initiating MHT before the age of 60, and less than 10 years after the reported age of last menstruation [47–49].
Statistical analyses
Analyses were carried out in SPSS version 26 for Mac and R version 1.2.5033 (R Foundation for Statistical Computing). A critical α of 0.05 defined statistical significance and all statistical tests were 2-sided. Regression diagnostics and collinearity were assessed for all models. To test our hypothesis, we first conducted a set of hierarchical ordinary least squares (OLS) linear regression analyses to assess whether MHT use was associated with late life memory (immediate and delayed recall for 10-item word list), and whether adding diabetes status to the model could account for additional variance. In the first block we modeled the effect of midlife MHT use on late life memory measures while accounting for the following covariates: word list version, age at late life cognitive assessment, education, APOE4 status, suspected surgical menopause, midlife obesity, and late life reported depression symptoms (mCES-D). The second block retained all of the initial predictors while also adding diabetes status and additional health-related covariates (hypertension, hyperlipidemia, heart disease, and lifetime smoking years) to the model. A follow-up sensitivity analyses was conducted among MHT users only to assess whether adding reproduction or MHT-related variables (MHT formulation type, MHT use duration, total reproductive duration, and optimal MHT use) to the first block of the hierarchical regression model could account for any additional significant variance in late life memory performance. Regression coefficients are reported as both unstandardized coefficients (to interpret the effect of each independent variable on the outcome variable in relative units) and as standardized coefficients (to aid in interpreting the relative importance of each predictor within the full model).
We next modeled diabetes risk as a function of MHT use history using a log-rank test, followed by a Cox regression model, to determine whether MHT use remained a significant predictor of time to diabetes onset after accounting for additional covariates: high school IQ, years of education, midlife obesity, hyperlipidemia, hypertension, heart disease, lifetime smoking years, and suspected surgical menopause.
Finally, a mediation model was fit to evaluate whether any associations between MHT use and late life memory were mediated by diabetes risk. Using the “Mediate” package in R [50], the mediator model was specified as a probit regression with MHT use predicting late life diabetes status and the outcome model was specified as a generalized linear regression, where MHT use and diabetes predicted late life memory performance. The significance of the indirect effect was tested using bootstrapping procedures. Unstandardized indirect effects were computed for each of the 1,000 bootstrapped samples, and the 95% confidence interval was computed by determining the indirect effects at the 2.5th and 97.5th percentiles.
RESULTS
Participant demographics by MHT use history and diabetes status
Relevant participant demographic, health, and outcome variables classified by history of MHT use is presented in Table 1. Overall, the final sample included 1,792 female WLS graduate respondents who were a mean age of 64 at the midlife data collection wave, and 71 years of age at the late life data collection wave. Although the WLS graduate sample is a same-aged cohort, age was specified as a covariate in models to account for age differences at the time of assessment (data collection at each timepoint spanned several years due to the large size of the cohort). Consistent with the overall WLS cohort, this sample was predominately white and had at least a high school education [44]. Of the 1,792 included participants, 1,088 women (60.71%) reported a history of midlife MHT use, and 220 (12.3%) reported a history of diabetes. MHT users had a lower incidence of diabetes compared to non-users (Odds Ratio (OR) = 0.49 (95% CI: 0.37, 0.66), p < 0.001), along with a protracted age of diabetes onset (Mean Difference (MD) = 5.48 years (95% CI: 3.05, 7.91), p < 0.001). Among those with diabetes, the frequency of diabetes control and diabetes medication management (insulin or oral meds) did not differ as a function of MHT use history. The mean age of reported MHT initiation was 48.77 ± 6.81 years and occurred an average of 1.49 ± 7.23 years from the age of last reported menstruation. Of MHT users, 88.7% (865/975) were classified as having an optimal MHT initiation, and the average duration of MHT use was 10.45 ± 6.89 years. About 40% of the sample (n = 729) reported a history of bilateral oophorectomy or complete hysterectomy, and of those, about 32% (n = 234) underwent hysterectomy or bilateral oophorectomy prior to the age of 40 and were classified as having experienced surgical menopause. In terms of general health, those with MHT use history were less likely to be obese in midlife (p = 0.015) and showed a trend towards increased incidence of hyperlipidemia by late life (p = 0.067), although the two groups did not statistically differ with regards to smoking, hypertension, heart disease, or cancer histories (all >0.05).
Table 1.
Basic Demographics and outcome measures by MHT use
Missing Data | Full Sample | MHT Use | No MHT Use | MHT versus No MHT Comparison | ||
---|---|---|---|---|---|---|
N (%) | N = 1,792 | (n = 1,088) | (n = 704) | MD/OR (95% CI) | p | |
Demographics | ||||||
Mid Life Age, M(SD) [range], y | 53 (2.96) | 64.23 (0.62) [62–67] | 64.21 (0.61) | 64.26 (0.62) | MD = −0.04 (−0.10, 0.02) | 0.172 |
Late Life Age, M(SD) [range], y | 0 | 71.16 (0.88) [69–74] | 71.11 (0.87) | 71.23 (0.89) | MD = −0.11 (−0.19,−0.03) | 0.009 |
Education, M (SD) [range], highest year equivalent | 0 | 13.52 (2.14) [12–21] | 13.66 (2.20) | 13.36 (2.06) | MD = 0.33 (0.13, 0.54) | 0.001 |
Race, No./total No. (%), white | 12 (0.01) | 1,771/1,792 (99.5%) | 1,080/1,088 (99.3%) | 691/704 (98.2%) | OR = 0.32 (0.08, 1.28) | 0.090 |
mCES-D, M (SD) [range], sum of all items | 162 (8.00) | 15.67 (14.71) [0–132] | 15.45 (14.21) | 15.79 (15.24) | MD = −0.53 (−2.00, 0.95) | 0.484 |
APOE4 Status, No./total No. (%), positive | 0 | 457/1,792 (25.5%) | 286/1,088 (26.3%) | 171/704 (24.3%) | OR = 1.11 (0.89, 1.38) | 0.344 |
Late Life Cognition | ||||||
Memory (10-item word list) | ||||||
Immediate Recall, M (SD) [range], raw score | 0 | 5.82 (1.42) [0,10] | 5.91 (1.39) | 5.70 (1.44) | MD = 0.23 (0.10, 0.37) | <0.001 |
Delayed Recall, M (SD) [range], raw score | 2 (0.00) | 3.87 (1.80) [0, 10] | 3.92 (1.79) | 3.85 (1.80) | MD = 0.11 (−0.06, 0.28) | 0.233 |
Reproductive Variables | ||||||
Age at Menarche, M {SD} [range], y | 86 (4.80) | 12.63 (1.49) [8, 19] | 12.64 (1.71) | 12.74 (2.34) | MD = 0.04 (−0.10, 0.19) | 0.563 |
Age at Last Menstruation, M {SD} [range], y | 81 (4.52) | 47.50 (7.41) [14, 65] | 47.27 (7.59) | 48.06 (6.83) | MD = 0.60 (−0.12, 1.32) | 0.100 |
Reproductive Duration, M {SD} [range], y | 157 (8.76) | 34.81 (7.49) [1, 54] | 34.44 (8.19) | 35.27 (7.11) | MD = 0.42 (−0.32, 1.17) | 0.266 |
Bilateral oophorectomy/hysterectomy, No./total No. (%), yes | 9 (.00) | 729/1,783 (40.9%) | 524/,1083 (48.4%) | 205/700 (29.3%) | OR = 2.26 (1.85, 2.77) | <0.001 |
Suspected Surgical Menopause, No./total No. (%), yes | 9 (.00) | 234/1,783 (13.1%) | 159/1,083 (14.7%) | 75/700 (10.7%) | OR = 1.43 (1.07, 1.92) | 0.015 |
MHT Type | 0 | |||||
No MHT Use, No./total No. (%), yes | N/A | 704/1,792 (39.3%) | N/A | N/A | N/A | |
Unopposed Estrogen, No./total No. (%), yes | N/A | 463/1,792 (25.8%) | N/A | N/A | N/A | |
Opposed Estrogen, No./total No. (%), yes | N/A | 625/1,792 (34.9%) | N/A | N/A | N/A | |
Time between Menopause and MHT Start, M (SD) [range], y | 114 (10.48) | N/A | 1.49 (7.23) [0, 39] | N/A | N/A | N/A |
Age Started MHT, M (SD) [range], y | 74 (6.80) | N/A | 48.77 (6.81) [21, 65] | N/A | N/A | N/A |
Duration of MHT use, M (SD) [range], y | 328 (30.15) | N/A | 10.45 (6.89) [1,37] | N/A | N/A | N/A |
Optimal MHT Onset, No./total No. (%), yes | 109 (10.0) | N/A | 867/979 (88.6%) | N/A | N/A | N/A |
General Health | ||||||
Lifetime Smoking Duration, Median [interquartile range], y | 0 | 0 [0, 22] | 0 [0, 22] | 0 [0, 21] | MD = −0.15 (−1.86, 1.56) | 0.867 |
Midlife Obesity (BMI >35 kg/m2), No./total No. (%), yes | 0 | 164/1,792 (9.2%) | 85/1,088 (7.8%) | 79/704 (11.2%) | OR = 0.67 (0.49, 0.93) | 0.015 |
Hypertension, No./total No. (%), yes | 0 | 1,057/1,792 (59.0%) | 633/1,088 (58.2%) | 424/704 (60.2%) | OR = 0.92 (0.78, 1.11) | 0.390 |
Hyperlipidemia, No./total No. (%), yes | 0 | 1,058/1,792 (59.0%) | 661/1,088 (60.8%) | 397/704 (56.4%) | OR = 1.20 (0.98, 1.45) | 0.067 |
Heart Disease, No./total No. (%), yes | 5 (0.28) | 308/1,787 (17.2%) | 188/1,083 (17.4%) | 120/704 (17.0%) | OR = 1.02 (0.80, 1.32) | 0.795 |
Cancer, No./total No. (%), yes | 2 (.11) | 320/1,790 (17.9%) | 201/1,088 (18.5%) | 119/702 (17.0% | OR = 1.11 (0.87, 1.43) | 0.412 |
Diabetes, No./total No. (%), yes | 0 | 220/1,792 (12.3%) | 100/1,088 (9.2%) | 120/704 (17.0%) | OR = 0.49 (0.37, 0.66) | <0.001 |
Diabetes | ||||||
Diabetes Onset Age | 7 (3.18) | 60.37 (9.69) [16, 72] | 63.22 (7.20) | 58.68 (10.55) | MD = 5.48 (3.05, 7.91) | <0.001 |
Insulin Use, No./total No. (%), yes | 0 | 55/220 (25.0%) | 19/100 (19.0%) | 36/120 (30.0%) | OR = 0.55 (0.29, 1.03) | 0.061 |
Diabetes Oral Medication, No./total No. (%), yes | 0 | 158/220 (71.8%) | 71/100 (71.0%) | 87/120 (72.5%) | OR = 0.93 (0.52, 1.67) | 0.805 |
Diabetes Duration (y) | 7 (3.18) | 10.54 (9.40) [0, 55] | 7.93 (7.11) | 12.70 (10.66) | MD = −5.67 (−8.11,−3.23) | <0.001 |
Diabetes Controlled | 7 (3.18) | ChiSq. (2) = 4.18 | 0.124 | |||
All | 90/213 (42.3%) | 47/96 (49.0%) | 43/117 (36.8%) | |||
Some | 92/213 (43.2%) | 39/96 (40.6%) | 53/117 (45.3%) | |||
None | 31/213 (14.6%) | 10/96 (10.4%) | 21/117 (17.9%) |
MD, mean difference; OR, odds ratio; BMI, body mass index; MHT, menopausal hormone therapy; mCES-D, modified Center for Epidemiological Studies Depression Inventory.
MHT use and diabetes status on late life memory
We first conducted a set of hierarchical ordinary least squares linear regression analyses to assess whether MHT use was associated with late life memory (immediate and delayed recall), and whether adding diabetes status to the model could account for additional variance. Overall model fit and regression coefficients for the hierarchical linear regression models are reported in Table 2, with results suggesting that both MHT use history and diabetes status predicted late life immediate recall, although only diabetes status predicted delayed recall.
Table 2.
Hierarchical linear regression models predicting late life immediate and delayed recall memory by MHT use (Model 1) and diabetes status (Model 2)
Immediate Recall | Delayed Recall | |||||||
---|---|---|---|---|---|---|---|---|
Model 1 | Model 2 | Model 1 | Model 2 | |||||
Unstandardized | Standardized | Unstandardized | Standardized | Unstandardized | Standardized | Unstandardized | Standardized | |
B (95% CI) | B | B (95% CI) | B | B (95% CI) | B | B (95% CI) | B | |
Constant | 16.05 (10.74, 21.36) | 15.87 (10.56, 21.17) | 12.59 (5.77, 19.41) | 12.17 (5.36, 18.98) | ||||
Word List Version | −0.250 (−0.377,−0.122) | −0.088*** | −0.246 (−0.373,−0.119) | −0.087*** | −0.212 (−0.375, −0.048) | −0.059* | −0.216 (−0.379, −0.053) | −0.060** |
Age At Assessment (y) | −0.155 (−0.229,−0.082) | −0.097*** | −0.153 (−0.226,−0.079) | −0.095*** | −0.139 (−0.233.−0.044) | −0.068** | −0.133 (−0.228,−0.039) | −0.066** |
Education (y) | 0.080 (0.049, 0.110) | 0.120 *** | 0.077 (0.046, 0.108) | 0.116 *** | 0.106 (0.066, 0.145) | 0.126 *** | 0.105 (0.066, 0.145) | 0.125 *** |
APOE4 Carrier (positive) | −0.241 (−0.386, −0.095) | −0.074** | −0.253,−0.399,−0.107) | −0.078*** | −0.352 (−0.539,−0.165) | −0.086*** | −0.532 (−0.539,−0.165) | −0.086*** |
mCES-D | −0.012 (−0.017, −0.007) | −0.118*** | −0.012 (−0.017, −0.007) | −0.119 *** | −0.008 (−0.014, −0.002) | −0.060* | −0.008 (−0.014, −0.002) | −0.061* |
Midlife Obesity (BMI <35) | 0.050 (−0.172, 0.273) | 0.010 | 0.119 (−0.108,0.346) | 0.024 | −0.094 (−0.380, 0.192) | −0.015 | −0.018 (−0.309, 0.274) | −0.003 |
Surgical Menopause | 0.034 (−0.155,0.223) | 0.008 | 0.044 (−0.145, 0.234) | 0.011 | 0.134 (−0.108, 0.377) | 0.025 | 0.085 (−0.193, 0.363) | −0.003 |
MHT Use History | 0.183 (0.052, 0.314) | 0.063 ** | 0.154 (0.023, 0.286) | 0.053 * | 0.045 (−0.123,0.213) | 0.012 | 0.014 (−0.155, 0.183) | 0.004 |
Diabetes | −0.320 (−0.524,−0.115) | −0.074** | −0.443 (−0.706,−0.180) | −0.081*** | ||||
Hyperlipidemia | 0.154 (0.020, 0.289) | 0.054 * | 0.010 (−0.163, 0.183) | 0.003 | ||||
Hypertension | −0.030 (−0.166, 0.106) | −0.010 | 0.098 (−0.077, 0.273) | 0.027 | ||||
Heart Disease | 0.007 (−0.165, 0.179) | 0.002 | 0.048 (−0.173, 0.269) | 0.010 | ||||
Lifetime Smoking Years | 0.001 (−0.003, 0.004) | 0.010 | 0.004 (−0.001, 0.008) | 0.037 |
p < 0.001;
p < 0.01;
p < 0.05;
mCES-D, modified Center for Epidemiological Studies Depression Inventory Score; MHT, menopausal hormone therapy; BMI, body mass index.
When predicting immediate recall performance, the first block of the hierarchical model reached statistical significance (F(8,1783) = 16.49, p < 0.001) and accounted for 6.9% of the variance in immediate recall scores (Adj. R2 = 0.065). After accounting for covariates, a reported history of MHT use was associated with higher immediate recall scores in late life compared to no MHT use (B = 0.183 [95% CI: 0.052, 0.314, p = 0.006). After the addition of the second block containing diabetes status and other health variables/comorbidities, the model fit significantly improved (F-Change = 2.75, p = 0.018; R2-Change = 0.007). A reported history of diabetes predicted worse immediate recall performance (B = −0.320 [95% CI: −0.524, −0.115], p = 0.002) while a history of MHT use remained a significant predictor of higher immediate word recall (B = 0.154 [95% CI: 0.023, 0.286], p = 0.022).
When predicting delayed recall performance, the first block of the hierarchical regression model was statistically significant (F(8,1783) = 10.17, p < 0.001) and accounted for 4.4% of the variance in delayed recall scores (Adj. R2 = 0.039). In the first block of the model, MHT use history was not a significant predictor of delayed recall performance (B = 0.045 [95% CI: −0.123, 0.213], p = 0.528), as only word list version, age at time of assessment, education, APOE4 status, and depression scores reached a level of statistical significance. After the addition of the second block containing additional health variables/comorbidities, the model fit significantly improved (F-Change = 2.78, p = 0.016; R2-Change = 0.007), and diabetes status was a significant predictor of lower delayed recall in late life (B = −0.443 [95% CI: −0.706, −0.180], p < 0.001).
Follow-up sensitivity analysis were conducted to assess whether reproduction or MHT-related factors (MHT formulation type, MHT use duration, total reproductive duration, and optimal MHT start) could account for additional variance in late life memory performance when added to the first block of the hierarchical regression model above. Results are presented in Table 3. Among women with MHT use history, having experienced a longer total reproductive duration (B = 0.016 [95% CI: 0.000, 0.031], p < 0.05), and primarily taking unopposed Estrogen compared to opposed Estrogen (B = 0.239 [95% CI: 0.060, 0.418], p < 0.01) was associated with higher performance on immediate recall trials, whereas MHT use duration, and optimal MHT start did not emerge as significant predictors in the model. For delayed recall, a similar pattern was observed, with history of unopposed Estrogen formulation (B = 0.299 [95% CI: 0.068, 0.530], p < 0.05), and longer total reproductive duration (B = 0.021 [95% CI0.001, 0.041], p < 0.05), also predicting higher delayed recall performance.
Table 3.
Sensitivity analysis among MHT users only (n = 1,088) to assess whether additional MHT or reproduction related factors account for additional variance in late life memory performance
Immediate Recall | Delayed Recall | |||||||
---|---|---|---|---|---|---|---|---|
Model 1 | Model 2 | Model 1 | Model 2 | |||||
Unstandardized | Standardized | Unstandardized | Standardized | Unstandardized | Standardized | Unstandardized | Standardized | |
B (95% CI) | B | B (95% CI) | B | B (95% CI) | B | B (95% CI) | B | |
Constant | 14.55 (8.93–20.17) | 10.92 (4.99, 16.85)] | 3.98 (−4.32, 11.28) | 3.62 (−4.18, 11.41) | ||||
Word List Version | −0.263 (−0.403,−0.123) | −0.094*** | −0.259 (−0.399,−0.120) | −0.093*** | −0.239 (−0.422, −0.055) | −0.067* | −0.236 (−0.419,−0.053) | −0.066* |
Age at Assessment (y) | −0.088 (−0.169,−0.006) | −0.056* | −0.091 (−0.173,−0.010) | −0.058* | −0.013 (−0.119, 0.094) | −0.006 | −0.016 (−0.122, 0.091) | −0.008 |
Education (y) | 0.013 (−0.023, 0.048) | 0.020 | 0.008 (−0.028, 0.044) | 0.013 | 0.026 (−0.021 (0.073) | 0.032 | 0.027 (−0.020, 0.074) | 0.033 |
APOE4 Carrier (positive) | −0.167 (−0.328,−0.005) | −0.052* | −0.172 (−0.334,−0.011) | −0.053* | −0.304 (−0.516,−0.092) | −0.073** | −0.312 (−0.525,−0.100) | −0.075** |
mCES-D | −0.007 (−0.012, −0.002) | −0.073* | −0.007 (−0.012, −0.002) | −0.073* | −0.006 (−0.013, 0.001) | −0.047 | −0.006 (−0.013, 0.001) | −0.049 |
Midlife Obesity (BMI <35) | 0.015 (0.003, 0.027) | 0.064 * | 0.018 (0.005, 0.030) | 0.078 ** | 0.010 (−0.006, 0.025) | 0.033 | 0.014 (−0.003, 0.030) | 0.046 |
Surgical Menopause | 0.008 (−0.204, 0.220) | 0.002 | 0.025 (−0.187,0.237) | 0.006 | 0.075 (−0.203, 0.354) | 0.014 | 0.085 (−0.193, 0.363) | 0.016 |
MHT Type | ||||||||
Unopposed Estrogen | 0.195 (0.010, 0.380) | 0.062 * | 0.176 (−0.009, 0.361) | 0.056 | 0.137 (−0.105, 0.380) | 0.034 | 0.120 (−0.123, 0.363) | 0.030 |
Opposed Estrogen | −0.015 (−0.184, 0.155) | −0.005 | −0.045 (−0.214, 0.125) | −0.015 | −0.101 (−0.323, 0.122) | −0.027 | −0.132 (−0.355, 0.091) | −0.035 |
Diabetes | −0.401 (−0.626, −0.175) | −0.095** | −0.519 (−0.815,−0.223) | −0.096** | ||||
Hyperlipidemia | 0.074 (−0.072, 0.220) | 0.026 | 0.074 (−0.117, 0.266) | 0.021 | ||||
Hypertension | −0.068 (−0.218, 0.083) | −0.024 | 0.070 (−0.128, 0.268) | 0.019 | ||||
Heart Disease | 0.091 (−0.099, 0.281) | 0.025 | 0.157 (−0.093, 0.407) | 0.033 | ||||
Lifetime Smoking Years | 0.001 (−0.004, 0.005) | 0.008 | 0.000 (−0.007, 0.006) | −0.004 |
p < 0.001;
p < 0.01;
p < 0.05;
mCES-D, modified Center for Epidemiological Studies Depression Inventory Score; MHT, menopausal hormone therapy; BMI, body mass index.
MHT use and diabetes incidence
We next evaluated whether MHT use history was associated with diabetes risk, with results indicating a 45.8% lower risk of diabetes among midlife MHT users compared to non-users, along with a 1.71-year protracted time to diabetes onset among MHT users who developed diabetes compared to non MHT users. A log-rank test was first implemented to determine if there were differences in the survival distribution of age to diabetes onset by MHT use history. The overall model reached statistical significance (χ2(1) = 26.31, p < 0.001), indicating differing time to diabetes onset by MHT use history. Mean survival time among those with no MHT use history was 71.38 years [95% CI: 70.84, 71.93], whereas mean survival time among MHT users was 73.09 years [95% CI: 72.89, 73.30]. We next fit a Cox regression model to determine whether MHT use remained a significant predictor of time to diabetes onset after accounting for additional covariates. The overall model reached statistical significance, χ2(9) = 216.58, p < 0.001, with the resultant hazard ratios (HR) predicting diabetes risk presented in Table 4. Women with a reported history of MHT use were at reduced risk of diabetes compared to women without MHT use history (HR = 0.542 [95% CI: 0.414, 0.710], p < 0.001), with cumulative hazard plots by group presented in Fig. 2. Other disease comorbidities also influenced diabetes risk, with increased incidence among those with history of hypertension (HR = 2.787 [95% CI: 1.909, 4.070], p < 0.001), hyperlipidemia (HR = 1.863 [95% CI: 1.347, 2.577]; p < 0.001), heart disease (HR = 1.911 [95% CI: 1.433, 2.548]; p < 0.001), and midlife obesity (HR = 2.678 [95% CI: 1.957, 3.663], p < 0.001).
Table 4.
Hazard ratios from Cox Regression model predicting diabetes risk by MHT use and additional covariates
Hazard Ratio (95% CI) | p | |
---|---|---|
Adolescent IQ | 0.998 (0.992, 1.003) | 0.459 |
Education (y) | 0.910 (0.839, 0.988) | 0.024 |
Midlife Obesity (BMI ≤35) | 2.678 (1.957, 3.663) | <0.001 |
Lifetime Smoking Years | 0.999 (0.992, 1.006) | 0.753 |
Hypertension | 2.787 (1.909, 4.070) | <0.001 |
Hyperlipidemia | 1.863 (1.347, 2.577) | <0.001 |
Heart Disease | 1.911 (1.433, 2.548) | <0.001 |
MHT Use | 0.542 (0.414, 0.710) | <0.001 |
Suspected Surgical Menopause | 1.242 (0.868, 1.779) | 0.236 |
Fig. 2.
Cox cumulative hazard ratio plotting time to diabetes onset age in years by history of Menopausal Hormone Therapy (MHT) use. Risk of incident diabetes was lower among women with reported history of MHT use compared to no MHT use.
Finally, we ran a causal mediation analysis to formally assess the mediating role of diabetes on the association between MHT use and late life memory performance. Because the direct effect of MHT use on delayed recall performance did not reach a level of statistical significance in the hierarchical linear regression described above, the mediation model was fit for immediate recall performance only. As shown in Fig. 3, the results of mediation analysis supported our initial hypothesis that the effect of MHT use history on late life memory performance was partially mediated by diabetes incidence.
Fig. 3.
Causal mediation analyses evaluating the mediating effects of diabetes on the association between MHT use and late life immediate recall performance. Diabetes was observed to partially mediate the association between MHT use and Late Life Immediate Recall. Average Causal Mediating Effect (ACME); Average Direct Effect (ADE).
DISCUSSION
Based on a lifespan of prospectively collected data from women graduate respondents in the Wisconsin Longitudinal Study, MHT use in midlife was associated with better late life performance on immediate word list recall but not delayed recall, as well as predicted a reduced incidence and delayed onset of diabetes. Consistent with established literature [36], diabetes was independently associated with worse performance on both immediate and delayed recall performances in late life. When these variables were combined into a causal mediation model, a history of diabetes was found to partially mediate the association between MHT use and late life immediate recall performance.
Our finding of a preferential association between MHT use history and immediate recall performance (as opposed to delayed recall) is supported by prior investigations elucidating mechanisms by which estrogen selectively impacts prefrontal and hippocampal brain function and metabolism—brain regions critical for the initial learning and encoding of information. One study evaluating the short-term effects of estradiol versus progesterone treatment in a small sample of perimenopausal women found that compared to placebo, women assigned to the estradiol group showed greater fMRI activation in bilateral superior frontal cortices during a verbal processing task, possibly reflecting more efficient verbal encoding—a finding not observed in the progesterone only group [51]. Other work has also supported beneficial effects of postmenopausal estrogen use on prefrontal metabolism [52] and selective enhancement of associated prefrontal cognitive processes including verbal learning and retrieval [53, 54]. Further, the effects of estrogen on brain structure appear lasting as demonstrated by a recent population-based longitudinal study (the Cardiovascular Health Study (CHS)) of over 500 older women (71–94 years of age) where a history of estrogen use in midlife was associated with larger gray and white matter volumes within frontal, temporal, and parietal lobes in late life [55].
In addition to late life cognitive benefits, we also evidenced a reduced risk and delayed time to onset of diabetes as a function of MHT use. This finding is consistent with findings that estrogen replacement improves β-cell insulin secretion, glucose utilization, and insulin sensitivity [33]. Several prior randomized controlled trials have demonstrated an antidiabetic effect of MHT use, including findings from the Postmenopausal Estrogen/Progestin Interventions study [56], the Heart and Estrogen/Progestin Replacement Study (HERS) [57], and the previously discussed WHI study [58, 59]. These results were synthesized in a 2006 meta-analysis reporting that MHT use was associated with a 13% drop in insulin resistance and an overall 30% reduction in new-onset diabetes [60].
Despite emerging benefits of midlife MHT use on health and cognitive outcomes in a growing body of clinical and observational studies, MHT use has decreased 80% since the seminal WHI trial data was first published in 2002 [61, 62]. Although the WHI study reported deleterious effects of MHT when administered in late life, subsequent studies have challenged the generalizability of these findings, as well as proposed a critical window hypothesis where MHT is considered most advantageous when administered closer to the menopausal transition. Randomized controlled clinical trials that have accounted for this critical window hypothesis yield more promising effects of MHT use among younger perimenopausal women. The Research into Memory, Brain function, and Estrogen Replacement (REMEMBER) study enrolled 428 women to investigate the timing of initiation and duration of MHT on cognitive function. Results indicated that early MHT use (before age 56) was associated with better performances on the MMSE, Trail Making Test Part A and phonemic fluency compared to no history of MHT use, whereas late MHT onset was found to be detrimental to late life cognition [63]. A smaller study of 34 perimenopausal women found that early MHT users exhibited better performance on a fMRI-based verbal memory task, the degree to which associated with increased activation of the left hippocampus and decreased activation in the bilateral hippocampus gyrus [64]. However, MHT use was found to not offer cognitive benefits in the ancillary Cognitive and Affective study within the Kronos Early Estrogen Prevention Study (KEEPS-Cog) after a 4-year follow-up period [65]. It is possible that such discrepant findings among existing MHT studies are subsequent to methodological differences in MHT formulation type, MHT timing, length of treatment, duration of follow-up period, and variability in participant characteristics [32]. Yet despite overall accumulating evidence that MHT is safe and offers benefit in younger postmenopausal women, prescription rates have remained extremely low in some settings, with only 2–4% of women initiating MHT following menopause in the year 2020 in data coming out of Australia and Canada, compared to 25–30% prior to the WHI study [66].
Although we did not observe a main effect of optimally timed MHT initiation on cognitive outcomes in support of the critical window hypothesis in the present study, our sample was not well suited to test this hypothesis given that the majority of WLS participants (~89%) reported an optimal MHT initiation (prior to the age of 60 and within 10 years of reported last menstruation). However, other large-scale observational studies have yielded support for optimally timed MHT use by demonstrating that midlife MHT users had a 26% decreased risk for dementia, whereas late-life MHT users showed a 48% increased risk [29]. A more recent population-based study of over 2,000 women in Cache County, Utah found that MHT use was positively associated with late life global cognition on the modified Mini-Mental Status Examination, with the strongest association for women who initiation MHT within 5 years of menopause [20]. Our findings add to this literature by extending the long-term cognitive benefit of midlife MHT use beyond a reduction in dementia risk to also include better initial learning and encoding of verbal information (i.e., immediate recall) in a large community-based aging population in late life. Although dementia status was not defined for the WLS participants during the data range in which this study is based, a new NIH-funded study is currently underway to assess dementia prevalence in this cohort [45], thereby allowing promising new opportunities to explore the effects of MHT use history on late life dementia outcomes in a well characterized population-based sample with full life course data. Taken together, results of the present study support an indirect effect of midlife MHT use on aspects of late life memory that are found to be partially mediated by lowered diabetes risk.
Strengths and limitations of the present study
A major strength of the present study is the large population-based cohort with detailed prospectively collected life course data from which our findings are drawn. The broad scope of available WLS data allows consideration of a range of individual-level factors that may impact late life cognition that are not typically available in other cohorts including prospectively collected high school IQ as well as a wide range of well-characterized sociodemographic indicators. The same-aged nature of this cohort also reduces confounds between biological and reproductive aging effects, allowing greater precision into the nature of reproductive aging and related factors on cognitive outcomes. On the other hand, this sample is racially homogenous thereby limiting generalizability of current findings, and only represents individuals with at least a high school education. Given the restricted range of the word list task (0–10), this measure may have limited sensitivity to detect subclinical memory changes within a general aging population. This data was also limited by self-reported data collection of medically diagnosed health conditions and reproductive history. Report of MHT use history, MHT formulation type, as well as age of menarche and menopause were based on retrospective recall, although they were assessed at midlife in close proximity to the menopausal transition. Finally, from a clinical perspective, despite the observed benefits of MHT use on health and cognitive outcomes reported in this study, global consensus does not currently recommend the use of MHT for protection against cognitive decline or as a means of diabetes prevention. Any potential benefit of MHT use in a clinical setting must be balanced against established risks, with MHT generally considered safe and most advantageous if initiated before the age of 60 years and within 10 years of menopause to treat menopausal symptoms [47, 48]. Future directions should aim to replicate these findings in a prospective and controlled study with observed data to support medical conditions and detailed MHT related variables.
ACKNOWLEDGMENTS
The authors would like to thank the National Institute on Aging for funding the newest collection wave currently underway in the WLS cohort (R01 AG060737-01), the University of Wisconsin Survey Center for their integral role in developing and implementing all data collection survey instruments, as well as the National Alzheimer’s Coordinating Center (NACC) and the Wisconsin Alzheimer’s Disease Research Center (ADRC) for their support and resources in study design and instrumentation.
FUNDING
Since 1991, the Wisconsin Longitudinal Study (WLS) has been supported principally by the National Institutes for Health, National Institute on Aging (R01 AG009775; R01 AG033258; R01 AG041686), with additional support from the Vilas Estate Trust, the National Science Foundation, the Spencer Foundation, and the Graduate School of the University of Wisconsin-Madison. A newly supported data collection wave is currently underway with the aim of detecting dementia prevalence in this cohort, granted by the National Institute on Aging (R01 AG060737).
Footnotes
CONFLICT OF INTEREST
Victoria J. Williams is an Associate Editor of this journal but was not involved in the peer-review process nor had access to any information regarding its peer-review. Victoria J. Williams owns stock equity in Niji Corporation and receives scientific consulting fees from Amylyx Pharmaceuticals and Cognito Therapeutics. The remaining authors have no conflicts of interest to report.
DATA AVAILABILITY
Data from the Wisconsin Longitudinal Study, along with full documentation, are free and publicly available and can be accessed at http://www.ssc.wisc.edu/wlsresearch/. Requests for protected data or assistance should be sent to E-mail: wls@ssc.wisc.edu.
REFERENCES
- [1].Rannevik G, Jeppsson S, Johnell O, Bjerre B, Laurell-Borulf Y, Svanberg L (2008) A longitudinal study of the perimenopausal transition: Altered profiles of steroid and pituitary hormones, SHBG and bone mineral density. Maturitas 61, 67–77. [DOI] [PubMed] [Google Scholar]
- [2].McEwen BS, Akama KT, Spencer-Segal JL, Milner TA, Waters EM (2012) Estrogen effects on the brain: Actions beyond the hypothalamus via novel mechanisms. Behav Neurosci 126, 4–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Brann DW, Dhandapani K, Wakade C, Mahesh VB, Khan MM (2007) Neurotrophic and neuroprotective actions of estrogen: Basic mechanisms and clinical implications. Steroids 72, 381–405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Yan H, Yang W, Zhou F, Li X, Pan Q, Shen Z, Han G, Newell-Fugate A, Tian Y, Majeti R, Liu W, Xu Y, Wu C, Allred K, Allred C, Sun Y, Guo S (2019) Estrogen improves insulin sensitivity and suppresses gluconeogenesis via the transcription factor Foxo1. Diabetes 68, 291–304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Hugenschmidt CE, Duran T, Espeland MA (2021) Interactions between estradiol, diabetes, and brain aging and the risk for cognitive impairment. Climacteric 24, 359–365. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Eberling JL, Reed BR, Coleman JE, Jagust WJ (2000) Effect of estrogen on cerebral glucose metabolism in postmenopausal women. Neurology 55, 875–877. [DOI] [PubMed] [Google Scholar]
- [7].Smith CC, Vedder LC, Nelson AR, Bredemann TM, McMahon LL (2010) Duration of estrogen deprivation, not chronological age, prevents estrogen’s ability to enhance hippocampal synaptic physiology. Proc Natl Acad Sci U S A 107, 19543–19548. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Wang ACJ, Hara Y, Janssen WGM, Rapp PR, Morrison JH (2010) Synaptic estrogen receptor-α levels in prefrontal cortex in female rhesus monkeys and their correlation with cognitive performance. J Neurosci 30, 12770–12776. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Bohacek J, Daniel JM (2010) The beneficial effects of estradiol on attentional processes are dependent on timing of treatment initiation following ovariectomy in middle-aged rats. Psychoneuroendocrinology 35, 694–705. [DOI] [PubMed] [Google Scholar]
- [10].Brinton RD (2009) Estrogen-induced plasticity from cells to circuits: Predictions for cognitive function. Trends Pharmacol Sci 30, 212–222. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Kantarci K, Lowe VJ, Lesnick TG, Tosakulwong N, Bailey KR, Fields JA, Shuster LT, Zuk SM, Senjem ML, Mielke MM, Gleason C, Jack CR Jr, Rocca WA, Miller VM (2016) Early postmenopausal transdermal 17β-estradiol therapy and amyloid-β deposition. J Alzheimers Dis 53, 547–556. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Jaffe AB, Toran-Allerand CD, Greengard P, Gandy SE (1994) Estrogen regulates metabolism of Alzheimer amyloid beta precursor protein. J Biol Chem 269, 13065–13068. [PubMed] [Google Scholar]
- [13].Wisch JK, Meeker KL, Gordon BA, Flores S, Dincer A, Grant EA, Benzinger TL, Morris JC, Ances BM (2021) Sex-related differences in tau positron emission tomography (PET) and the effects of hormone therapy (HT). Alzheimer Dis Assoc Disord 35, 164–168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Almey A, Milner TA, Brake WG (2015) Estrogen receptors in the central nervous system and their implication for dopamine-dependent cognition in females. Horm Behav 74, 125–138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Schönknecht P, Henze M, Hunt A, Klinga K, Haberkorn U, Schröder J (2003) Hippocampal glucose metabolism is associated with cerebrospinal fluid estrogen levels in postmenopausal women with Alzheimer’s disease. Psychiatry Res 124, 125–127. [PubMed] [Google Scholar]
- [16].Epperson CN, Sammel MD, Freeman EW (2013) Menopause effects on verbal memory: Findings from a longitudinal community cohort. J Clin Endocrinol Metab 98, 3829–3838. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Saleh RNM, Hornberger M, Ritchie CW, Minihane AM (2023) Hormone replacement therapy is associated with improved cognition and larger brain volumes in at-risk APOE4 women: Results from the European Prevention of Alzheimer’s Disease (EPAD) cohort. Alzheimers Res Ther 15, 10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Jacobs DM, Tang MX, Stern Y, Sano M, Marder K, Bell KL, Schofield P, Dooneief G, Gurland B, Mayeux R (1998) Cognitive function in nondemented older women who took estrogen after menopause. Neurology 50, 368–373. [DOI] [PubMed] [Google Scholar]
- [19].Carlson MC, Zandi PP, Plassman BL, Tschanz JT, Welsh-Bohmer KA, Steffens DC, Bastian LA, Mehta KM, Breitner JC, Cache County Study Group (2001) Hormone replacement therapy and reduced cognitive decline in older women: The Cache County Study. Neurology 57, 2210–2216. [DOI] [PubMed] [Google Scholar]
- [20].Matyi JM, Rattinger GB, Schwartz S, Buhusi M, Tschanz JT (2019) Lifetime estrogen exposure and cognition in late life: The Cache County Study. Menopause 26, 1366–1374. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Daniel JM, Witty CF, Rodgers SP (2015) Long-term consequences of estrogens administered in midlife on female cognitive aging. Horm Behav 74, 77–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Rossouw J, Anderson G, Prentice R, LaCroix A, Kooperberg C, Stefanick M, Jackson R, Beresford S, Howard B, Johnson K, Kotcchen J (2002) Risks and benefits of estrogen plus progestin in healthy postmenopausal women: Principal results from the Women’s Health Initiative Randomized Controlled Trial. JAMA 288, 321–333. [DOI] [PubMed] [Google Scholar]
- [23].Stampfer MJ, Colditz GA, Willett WC, Manson JE, Rosner B, Speizer FE, Hennekens CH (1991) Postmenopausal estrogen therapy and cardiovascular disease. N Engl J Med 325, 756–762. [DOI] [PubMed] [Google Scholar]
- [24].Karim R, Dang H, Henderson VW, Hodis HN, John JS, Brinton RD, Mack WJ (2016) Effect of reproductive history and exogenous hormone use on cognitive function in midand late life. J Am Geriatr Soc 64, 2448–2456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Yaffe K, Sawaya G, Lieberburg I, Grady D (1998) Estrogen therapy in postmenopausal women: Effects on cognitive function and dementia. JAMA 279, 688–695. [DOI] [PubMed] [Google Scholar]
- [26].Maki PM (2006) Hormone therapy and cognitive function: Is there a critical period for benefit? Neuroscience 138, 1027–1030. [DOI] [PubMed] [Google Scholar]
- [27].Dumas J, Hancur-Bucci C, Naylor M, Sites C, Newhouse P (2008) Estradiol interacts with the cholinergic system to affect verbal memory in postmenopausal women: Evidence for the critical period hypothesis. Horm Behav 53, 159–169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [28].Bagger YZ, Tankó LB, Alexandersen P, Qin G, Christiansen C, PERF Study Group (2005) Early postmenopausal hormone therapy may prevent cognitive impairment later in life. Menopause 12, 12–17. [DOI] [PubMed] [Google Scholar]
- [29].Whitmer RA, Quesenberry CP, Zhou J, Yaffe K (2011) Timing of hormone therapy and dementia: The critical window theory revisited. Ann Neurol 69, 163–169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Mitchell JL, Cruickshanks KJ, Klein BEK, Palta M, Non-dahl DM (2003) Postmenopausal hormone therapy and its association with cognitive impairment. Arch Intern Med 163, 2485–2490. [DOI] [PubMed] [Google Scholar]
- [31].Kang JH, Weuve J, Grodstein F (2004) Postmenopausal hormone therapy and risk of cognitive decline in community-dwelling aging women. Neurology 63, 101–107. [DOI] [PubMed] [Google Scholar]
- [32].Fischer B, Gleason C, Asthana S (2014) Effects of hormone therapy on cognition and mood. Fertility Sterility 101, 898–904. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [33].Mauvais-Jarvis F, Manson JE, Stevenson JC, Fonseca VA (2017) Menopausal hormone therapy and type 2 diabetes prevention: Evidence, mechanisms, and clinical implications. Endocrine Rev 38, 173–188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [34].Livingston G, Huntley J, Sommerlad A, Ames D, Ballard C, Banerjee S, Brayne C, Burns A, Cohen-Mansfield J, Cooper C, Costafreda SG, Dias A, Fox N, Gitlin LN, Howard R, Kales HC, Kivimäki M, Larson EB, Ogunniyi A, Orgeta V, Ritchie K, Rockwood K, Sampson EL, Samus Q, Schneider LS, Selbæk G, Teri L, Mukadam N (2020) Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet 396, 413–446. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [35].Chatterjee S, Peters SAE, Woodward M, Mejia Arango S, Batty GD, Beckett N, Beiser A, Borenstein AR, Crane PK, Haan M, Hassing LB, Hayden KM, Kiyohara Y, Larson EB, Li C-Y, Ninomiya T, Ohara T, Peters R, Russ TC, Seshadri S, Strand BH, Walker R, Xu W, Huxley RR (2016) Type 2 diabetes as a risk factor for dementia in women compared with men: A pooled analysis of 2.3 million people comprising more than 100,000 cases of dementia. Diabetes Care 39, 300–307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [36].Marden JR, Mayeda ER, Tchetgen EJT, Kawachi I, Glymour MM (2017) High hemoglobin A1c and diabetes predict memory decline in the Health and Retirement Study. Alzheimer Dis Assoc Disord 31, 48–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [37].Awad N, Gagnon M, Messier C (2004) The relationship between impaired glucose tolerance, type 2 diabetes, and cognitive function. J Clin Exp Neuropsychol 26, 1044–1080. [DOI] [PubMed] [Google Scholar]
- [38].Rawlings AM, Sharrett AR, Schneider ALC, Coresh J, Albert M, Couper D, Griswold M, Gottesman RF, Wagenknecht LE, Windham BG, Selvin E (2014) Diabetes in midlife and cognitive change over 20 years. Ann Intern Med 161, 785–793. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [39].Biessels GJ, Despa F (2018) Cognitive decline and dementia in diabetes mellitus: Mechanisms and clinical implications. Nat Rev Endocrinol 14, 591–604. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [40].Espeland MA, Brinton RD, Hugenschmidt C, Manson JE, Craft S, Yaffe K, Weitlauf J, Vaughan L, Johnson KC, Padula CB, Jackson RD, Resnick SM (2015) Impact of type 2 diabetes and postmenopausal hormone therapy on incidence of cognitive impairment in older women. Diabetes Care 38, 2316–2324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [41].(2020) National Diabetes Statistics Report 2020. Estimates of diabetes and its burden in the United States
- [42].Paschou SA, Papanas N (2019) Type 2 diabetes mellitus and menopausal hormone therapy: An update. Diabetes Ther 10, 2313–2320. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [43].Paschou SA, Anagnostis P, Pavlou DI, Vryonidou A, Goulis DG, Lambrinoudaki I (2019) Diabetes in menopause: Risks and management. Curr Vasc Pharmacol 17, 556–563. [DOI] [PubMed] [Google Scholar]
- [44].Herd P, Carr D, Roan C (2014) Cohort profile: Wisconsin longitudinal study (WLS). Int J Epidemiol 43, 34–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [45].Williams VJ, Carlsson CM, Fischer A, Johnson SC, Lange K, Partridge E, Roan C, Asthana S, Herd P (2021) Assessing dementia prevalence in the Wisconsin Longitudinal Study: Cohort profile, protocol, and preliminary findings. J Alzheimers Dis 81, 751–768. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [46].Eaton WW, Smith C, Ybarra M, Muntaner C, Tien A (2004) Center for Epidemiologic Studies Depression Scale: Review and Revision (CESD and CESD-R). In The use of psychological testing for treatment planning and outcomes assessment: Instruments for adults, Volume 3, 3rd ed. Lawrence Erlbaum Associates Publishers, Mahwah, NJ, US, pp. 363–377. [Google Scholar]
- [47].de Villiers TJ, Hall JE, Pinkerton JV, Pérez SC, Rees M, Yang C, Pierroz DD (2016) Revised global consensus statement on menopausal hormone therapy. Maturitas 91, 153–155. [DOI] [PubMed] [Google Scholar]
- [48].Lee SR, Cho MK, Cho YJ, Chun S, Hong S-H, Hwang KR, Jeon G-H, Joo JK, Kim SK, Lee DO, Lee D-Y, Lee ES, Song JY, Yi KW, Yun BH, Shin J-H, Chae HD, Kim T (2020) The 2020 Menopausal Hormone Therapy Guidelines. J Menopausal Med 26, 69–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [49].Scott E, Zhang Q, Wang R, Vadlamudi R, Brann D (2012) Estrogen neuroprotection and the critical period hypothesis. Front Neuroendocrinol 33, 85–104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [50].Tingley D, Yamamoto T, Hirose K, Keele L, Imai K (2014) mediation: R package for causal mediation analysis. UCLA Statistics/American Statistical Association. [Google Scholar]
- [51].Berent-Spillson A, Briceno E, Pinsky A, Simmen A, Persad CC, Zubieta J-K, Smith YR (2015) Distinct cognitive effects of estrogen and progesterone in menopausal women. Psychoneuroendocrinology 59, 25–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [52].Silverman DHS, Geist CL, Kenna HA, Williams K, Wroolie T, Powers B, Brooks J, Rasgon NL (2011) Differences in regional brain metabolism associated with specific formulations of hormone therapy in postmenopausal women at risk for AD. Psychoneuroendocrinology 36, 502–513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [53].Joffe H, Hall JE, Gruber S, Sarmiento IA, Cohen LS, Yurgelun-Todd D, Martin KA (2006) Estrogen therapy selectively enhances prefrontal cognitive processes: A randomized, double-blind, placebo-controlled study with functional magnetic resonance imaging in perimenopausal and recently postmenopausal women. Menopause 13, 411–422. [DOI] [PubMed] [Google Scholar]
- [54].Persad CC, Zubieta J-K, Love T, Wang H, Tkaczyk A, Smith YR (2009) Enhanced neuroactivation during verbal memory processing in postmenopausal women receiving short-term hormone therapy. Fertility Sterility 92, 197–204. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [55].Boyle CP, Raji CA, Erickson KI, Lopez OL, Becker JT, Gach HM, Kuller LH, Longstreth W, Carmichael OT, Riedel BC, Thompson PM (2021) Estrogen, brain structure, and cognition in postmenopausal women. Hum Brain Mapp 42, 24–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [56].Espeland MA, Hogan P, Fineberg E, Howard G, Schrott H, Waclawiw M, Bush T, The Pepi Investigators (1998) Effect of postmenopausal hormone therapy on glucose and insulin concentrations. Diabetes Care 21, 1589–1595. [DOI] [PubMed] [Google Scholar]
- [57].Kanaya AM, Herrington D, Vittinghoff E, Lin F, Grady D, Bittner V, Cauley JA, Barrett-Connor E (2003) Glycemic effects of postmenopausal hormone therapy: The Heart and Estrogen/progestin Replacement Study. A randomized, double-blind, placebo-controlled trial. Ann Intern Med 138, 1–9. [DOI] [PubMed] [Google Scholar]
- [58].Manson JE, Chlebowski RT, Stefanick ML, Aragaki AK, Rossouw JE, Prentice RL, Anderson G, Howard BV, Thomson CA, LaCroix AZ, Wactawski-Wende J, Jackson RD, Limacher M, Margolis KL, Wassertheil-Smoller S, Beresford SA, Cauley JA, Eaton CB, Gass M, Hsia J, Johnson KC, Kooperberg C, Kuller LH, Lewis CE, Liu S, Martin LW, Ockene JK, O’Sullivan MJ, Powell LH, Simon MS, Van Horn L, Vitolins MZ, Wallace RB (2013) Menopausal hormone therapy and health outcomes during the intervention and extended poststopping phases of the Women’s Health Initiative Randomized Trials. JAMA 310, 1353–1368. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [59].Margolis KL, Bonds DE, Rodabough RJ, Tinker L, Phillips LS, Allen C, Bassford T, Burke G, Torrens J, Howard BV, for the Women’s Health Initiative Investigators (2004) Effect of oestrogen plus progestin on the incidence of diabetes in postmenopausal women: Results from the Women’s Health Initiative Hormone Trial. Diabetologia 47, 1175–1187. [DOI] [PubMed] [Google Scholar]
- [60].Salpeter SR, Walsh JME, Ormiston TM, Greyber E, Buckley NS, Salpeter EE (2006) Meta-analysis: Effect of hormone-replacement therapy on components of the metabolic syndrome in postmenopausal women. Diabetes Obesity Metab 8, 538–554. [DOI] [PubMed] [Google Scholar]
- [61].Chester RC, Kling JM, Manson JE (2018) What the Women’s Health Initiative has taught us about menopausal hormone therapy. Clin Cardiol 41, 247–252. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [62].Hersh AL, Stefanick ML, Stafford RS (2004) National use of postmenopausal hormone therapy: Annual trends and response to recent evidence. JAMA 291, 47–53. [DOI] [PubMed] [Google Scholar]
- [63].MacLennan AH, Henderson VW, Paine BJ, Mathias J, Ramsay EN, Ryan P, Stocks NP, Taylor AW (2006) Hormone therapy, timing of initiation, and cognition in women aged older than 60 years: The REMEMBER pilot study. Menopause 13, 28–36. [DOI] [PubMed] [Google Scholar]
- [64].Maki PM, Dennerstein L, Clark M, Guthrie J, LaMontagne P, Fornelli D, Little D, Henderson VW, Resnick SM (2011) Perimenopausal use of hormone therapy is associated with enhanced memory and hippocampal function later in life. Brain Res 1379, 232–243. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [65].Gleason CE, Dowling NM, Wharton W, Manson JE, Miller VM, Atwood CS, Brinton EA, Cedars MI, Lobo RA, Merriam GR, Neal-Perry G, Santoro NF, Taylor HS, Black DM, Budoff MJ, Hodis HN, Naftolin F, Harman SM, Asthana S (2015) Effects of hormone therapy on cognition and mood in recently postmenopausal women: Findings from the randomized, controlled KEEPS–Cognitive and Affective Study. PLOS Med 12, e1001833. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [66].Salagame U, Kliewer EV, Demers A, Banks E, Velentzis LS, Goldsbury D, Egger S, Leslie WD, Canfell K (2020) Trends in prescribing menopausal hormone therapy and bisphosphonates in Australia and Manitoba, Canada and adherence to recommendations. J Womens Health (Larchmt) 29, 177–186. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Data from the Wisconsin Longitudinal Study, along with full documentation, are free and publicly available and can be accessed at http://www.ssc.wisc.edu/wlsresearch/. Requests for protected data or assistance should be sent to E-mail: wls@ssc.wisc.edu.