Associations between pituitary-ovarian hormones and cognition in recently menopausal women independent of type of hormone therapy.

The Kronos Early Estrogen Prevention Study (KEEPS; NCT00154180 ) was a randomized, placebo-controlled, double-blind multicenter clinical trial that evaluated cardiovascular effects of o-CEE, t-E2, and placebo in healthy, i.e. low Framingham scores, women who were within 6–36 months of their last menstrual period and experienced natural menopause between 42 and 58 years of age (no history of hysterectomy or salpingo-oophorectomy). The enrollment was at nine sites in the United States between August 2005 and July 2008, with final visits completed in 2012. Women eligible for KEEPS refrained from estrogen- or progestogen- containing medications for a minimum of 6 months prior to randomization and had a plasma follicle stimulating hormone (FSH) level ≥ 35 ng/mL and plasma 17β-estradiol level < 40 pg/mL. The additional inclusion and exclusion criteria were published previously [ 30 ]. The KEEPS-Cognitive and Affective Study (KEEPs-COG) ( NCT00623311 ) examined the effects of o-CEE, t-E2, and placebo on cognition. Women were randomized to receive 0.45 mg/day o-CEE daily, 50 μg/day t-E2 weekly, or to receive placebo pills and patches for 4 years. Additionally, women in the active HT treatment groups received oral 200 mg/day micronized progesterone the first 12 days of each month to protect against endometrial activation. Institutional review boards (IRBs) at the participating sites approved the study procedures and all participants provided written informed consent. Of the 727 women participating in KEEPS, snap-frozen serum was available from 362 women who were adherent to treatment from all enrollment centers. Serum samples collected prior to and after 48 months of randomization were used to measure hormones at the clinical core laboratory at Mayo Clinic, Rochester, MN. The hormones measured included androstenedione (A4), testosterone (T), estrone (E1), 17β-estradiol (E2), FSH and LH. Cognitive testing was performed at four time points (baseline, month 18, month 36, and month 48). T was measured by liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Agilent Technologies, Santa Clara, CA 95051). E1 and E2 were extracted with methylene chloride and after derivatization with dansyl chloride, high-pressure liquid chromatography (HPLC) was used prior to introduction of the derivatized sample extract into the tandem mass spectrometry (LC-MS/MS) (Agilent Technologies, Santa Clara, CA 95051). FSH and LH were measured by two-site immunoenzymatic assays performed on the DxI 800 automated immunoassay system (Beckman Instruments, Chaska, MN 55318). The intra-assay and inter-assay coefficients of variability (CV’s) were within acceptable ranges for high sensitivity T, E2, E1, FSH, and LH measurements. A deuterated stable isotope (d7-androstenedione) was added to a 0.1 mL plasma sample for the internal standard for A4. Samples were then extracted from specimens using a solid-phase cartridge and methanol was used to elute them from the cartridge. Nitrogen was used to dry down the extracts, and they were reconstituted with 75 μL of 70/30 methanol/H2O containing 1 μg/mL of estriol which prevented adsorption to equipment components. The reconstituted sample was then analyzed by LC-MS/MS using multiple-reaction monitoring. The sensitivity and specificity of all assays were published previously [ 26 ]. Participants completed a comprehensive battery of eleven neuropsychological tests within the KEEPs-COG at four time points: baseline (prior to randomization) and months 18, 36, and 48 after randomization. Using factor analysis, test scores were summarized as five orthogonal cognitive domains: verbal learning and memory (New York University Paragraph Recall, California Verbal Learning Test Trail 1–3 and Long Delay), auditory attention and working memory (Wechsler Adult Intelligence Scale-3 (WAIS-3) Letter Number Sequencing, Wechsler Memory Scale-3 Digit Span Forward and Backward), visual attention and executive function (Stroop Color Word Interference test, Trail Making Test A & B, WAIS-3 Digit Symbol, Benton Visual Retention Test), speeded language and mental flexibility (Phonemic Fluency S and Category Fluency: Animals, Fruits and Vegetables), and global cognition (all previous tests). Further information about the cognitive battery and the development of factor scores have been published previously [ 10 , 31 ]. Additional analyses focused on individual cognitive tests to discern complex attention skill-specific hormone level effects, especially on working memory dependent tasks such as Digit Span and Trail Making test A, and tests of higher difficulty executive function that require planning and inhibitory control, including Trail Making test B and the Stroop Color Word test. Participant characteristics are summarized in Table 1 . Descriptive summaries by treatment group were provided for all hormone levels at baseline and at month 48 ( Table 2 ). The analysis-of-variance F-test was used to compare groups on continuous measures and chi-square tests of independence to examine the association between groups and categorical measures. FSH and LH measures were normally distributed at both data collection time points. All androstenedione, testosterone, 17β-estradiol, and estrone measures were log transformed to normalize their distributions. Latent growth curve models (LGM) were used to examine the association between circulating hormone levels on differences in rates of change in each cognitive domain scores at baseline, month 18, month 36, and month 48 ( Fig. 1 ) [ 32 ]. Unconditional (no covariates) models were examined first to determine whether cognitive growth trajectories from baseline to month 48 had a linear trend. All models controlled for education level, age, waist circumference, APOE allele status, and treatment allocation. Hormone levels at baseline and at month 48 were modeled as time-dependent covariates. For all cognitive outcomes, the time factor included discrete values representing months elapsed since baseline, i.e., 0, 18, 36, and 48 months. Full information maximum likelihood (FIML) was used in the analyses to address missing data and a robust parameter estimator was used to address non-normality [ 33 ]. In all statistical tests an alpha level of 0.01 was used to control for multiple testing. All models were estimated using R 3.6.2 [ 34 ]. A linear multivariate mixed model examined the effect of changes in circulating hormone levels on changes on trajectories in working memory and complex attention, contrasting outcomes based on varying degree of difficulty or cognitive demand – “low vs high difficulty” ( Fig. 2 ). For example, the simple task of quickly connecting numbers in order – Trails A – provides a “low difficulty” measure of speeded visuomotor attention. Trails B and the requirement to alternate between numbers and letters, increases working memory demands, and represents a “high difficulty” outcome. The first set of analyses modeled jointly the performance on Trails A (“low difficulty”) and Trails B (“high difficulty”) outcomes across four time points (from baseline to month 48). In the second set of analyses, the unweighted average of longitudinal measures of Trails A and Digit Span Forward (“low difficulty”) were jointly modeled with an unweighted average of longitudinal measures of Trails B with Stroop Color Word (“high difficulty”) as complex attention outcomes. Change in circulating hormone levels was defined as a difference score or “change from baseline to month 48.” We examined the effect of this difference score on the slope of “low” and “high” difficulty tasks. That is, we examined whether changes of circulating hormones from baseline to month 48 had an effect on working and complex attention tasks of varying degrees of difficulty. All models controlled for age, waist circumference, education, APOE allele status, and treatment allocation.

Women adherent to treatment and who completed cognitive testing at 48 months of the study included 109 in the oCEE group, 107 in the t-E2 group and 146 in the placebo group (362 of the 727 enrolled in KEEPS, 49.7 %). Table 1 displays the participant demographics. Treatment groups did not differ in age (F = 0.22, p = 0.81), years since menopause at baseline (F = 1.50, p = 0.26), racial/ethnic composition (χ 2 = 12.14, p = 0.60), and educational attainment (χ 2 = 14.70, p = 0.14). The proportion of APOE ε4 carriers differed by treatment group (χ 2 = 7.92, p = 0.02). Hormone levels at baseline and month 48 are displayed in Table 2 . At baseline there were no differences in hormone levels among treatment groups. At 48 months, groups differed in testosterone (p = 0.010), FSH (p < 0.001), estrone (p < 0.001), and 17β-estradiol levels (p < 0.001). Overall, the change in cognition was non-significant over the 4 years of the study. Controlling for hormone levels at baseline and at month 48 did not have a significant effect on the rate of change in cognitive trajectories over time ( Table 3 ). Of all the associations tested among the individual ovarian and pituitary hormones modeled as time varying covariates, only higher levels of E1 at baseline associated with poorer cognitive performance at 48 months in auditory attention and working memory (p = 0.009) ( Table 3 ). For the visual attention and executive function factor score, higher baseline E1 levels were significantly associated with lower cognitive performance at all four time points. A trend for higher levels of baseline E1 levels to be associated with poorer global cognition was seen at all four time points (p = 0.015, 0.017, 0.015, 0.013, respectively). We observed a trend for higher levels of 17β-E2 at baseline to be associated with poorer cognitive performance at baseline in speeded language and mental flexibility (p = 0.013), as well as for global cognition (p = 0.012). There were no associations for testosterone, androstenedione, LH or FSH with any cognitive test at any time point. There were no significant associations between the change in individual hormone levels (difference from baseline to month 48) and change in skill specific working memory and complex attention tasks of varying difficulty levels ( Table 4 ).

In cognitively healthy, recently postmenopausal women at low cardiovascular risk, the rate of change in cognition was independent of HT over four years. This includes two forms of estrogens (CEE and E2) and both oral and transdermal routes of administration. These results suggest that postmenopausal HT may not affect the projection of cognitive performance. These findings, however, support the hypothesis that endogenous E1 levels, E1 metabolism, or secondary factors associated with elevations in E1 may affect various cognitive domains, regardless of HT status. Physiologic factors that alter endogenous hormone levels, such as central adiposity, insulin resistance, genetic determinants, or other phenomena which alter hormone metabolism, might still explain the cognitive changes associated with menopause, however, and this should be further explored. These results raise implications for elucidating the systems-level impacts of the pituitary-ovarian hormone milieu, including estrogens and their metabolism, on the brain and its function in the aging woman.

The findings of the present study extend those of previous cognitive analyses of women in KEEPS identifying that higher serum levels of E1 prior to treatment associated with poor cognitive performance in global cognition and specific functional domains, visual attention and executive function, and auditory attention and working memory. E1 is a relatively weak estrogen with less potent receptor binding kinetics compared to 17β-E2, making E1 a mildly antagonistic selective estrogen receptor modulator [ 35 ]. E1 is derived from A4, the primary post-menopausal ovarian androgen, and may be converted to 17β-E2. Aromatization from A4 can occur in the ovary, brain and adipose tissue, via the enzyme aromatase. In preclinical studies using rats, higher levels of circulating A4 correlated with deleterious cognitive effects, particularly in spatial reference and working memory [ 24 ]. Furthermore, blockade of the aromatase enzyme, preventing the conversion of A4 to E1, obviated the negative effects A4 on cognition indicating that it is the conversion to E1 that may be responsible for its memory impairment [ 25 ]. All of this raises the question of whether E1 acts against cognition by antagonizing the effects of E2. In clinical trials, results have been mixed regarding both E1 and E2 and their associations with cognitive outcomes [ 36 – 40 ]. In a crosssectional analysis of hormone levels and cognition in the ELITE trial, baseline endogenous estrogens (E1 and E2) were unrelated to verbal memory, executive functions, global cognition, and mood for both women early after menopause and those >10 years postmenopause with randomization to HT [ 41 ]. The present study provides examination of changes in cognition over 4 years. However, this study was not designed to address whether the present hormonal associations result from direct effects on brain microanatomy (e.g. synapses), or on brain physiology, or from interactions with other physiological factors such as central adiposity, insulin resistance, and aging, which also have associations with cognitive performance that may be independent of circulating sex hormone levels [ 42 ]. Results from KEEPS have identified links between indices of central adiposity and poorer cognitive performance as measured by the Modified Mini Mental Status exam [ 43 ]. It may be that central adiposity by increasing levels of total body aromatase activity, or insulin resistance, contributes to elevated levels of estrone which in turn impacts cognition. In KEEPS participants, variability in the age at onset of and severity of postmenopausal symptoms (independent of HT) were explained in part by a genetic variation in SULT1A1, a gene that encodes an enzyme that metabolizes estrogens (E1 and E2) [ 44 ]. Specifically, an increased number of G alleles from the SULT1A1 gene were associated with earlier age at menopause, decreased night sweat frequency and less severe insomnia prior to HT. The SULT1A1 genotype was not associated with postmenopausal symptoms. Factors influencing estrogen metabolism, such as SULT1A1, could also directly influence clinical outcomes. This concept is supported by the observation that in our study higher endogenous E1 was associated with poorer cognition regardless of treatment. o-CEE consists of numerous biologically active hormones, including about fifty percent estrone sulfate [ 35 ], equine estrogens and androgens. If E1 alone is primarily responsible for cognition in postmenopausal women, we should have seen reduced cognitive performance with randomization to o-CEE, which we did not. Multiple individual factors such as prior hormone exposure, presence of ovaries, stage within the menopausal transition, activity of aromatase (and possibly other converting enzymes), body weight, blood pressure and lipid profile all likely interact to affect cognitive function. Further, women vary in how a given type, dose and route of HT affects circulating levels of the various sex hormones. For example, one woman may have a much larger increase in circulating E1 after a given dose of o-CEE compared to other women. Further, the likely variability in the impact of specific hormone levels on neuronal anatomy and function within the CNS remain to be fully explored. In KEEPS it was found that t-E2 was associated with reduced amyloid-β deposition, a finding seen in people with Alzheimer’s disease, demonstrating that neuroprotection is a broad category and may relate to changes in the brain beyond neural function such as amyloid accumulation [ 45 ]. Finally, additional genetic analysis of estrogen metabolism in premenopausal women may provide insight into how specific estrogens affect multiple functions including cognition. There were no associations for the assessed pituitary hormones, LH and FSH, with any cognitive tests at baseline or 48 months. This latter finding is notable since a recent study in mice showed that FSH acted on hippocampal and cortical neurons to accelerate amyloid-β and Tau deposition leading to impaired cognition, effects which were nullified by blocking FSH [ 46 ]. There were also no associations with androstenedione and testosterone with cognition at any time point. Overall, variability in a woman’s individualized and unique factors including pituitary ovarian hormone metabolism and function, prior HT, gynecologic history, overall health, and sociodemographic circumstances together may influence the net impact of HT on cognition as much or even more than do the details (timing of onset, type, dose and route of administration) of HT. Strengths of our study include the randomized control, blinded multicenter trial design, as well as the longitudinally collected data. Hormone assays were done with gold standard assays assuring their accuracy. The limitations include the lack of availability of a premenopausal reproductive history, such as endometriosis, multiple births, preeclampsia, gestational hypertension, gestational diabetes, etc. We do not know free or bioavailable serum hormone levels since sex hormone binding globulin or albumin were not tested. Furthermore, additional hormones, such as progesterone, thyroid hormones, and corticosteroids, which might affect the interpretability of the observed hormone associations and menopausal state of the gonadotropic-pituitary axis were not included in this analysis. The exact composition of CEE is not fully known. Estrone is estimated to comprise about 50% of the total estrogens in CEE, and estradiol less than 1% [ 47 ]. The assays used in this study, therefore, do not measure the other 49% of the estrogens in CEE, including delta-8-e1 and delta 8 17-beta estradiol which in some studies are shown to have greater neuroprotective effects than estrone and estradiol [ 47 ]. It is possible that those estrogens, or the entire complement of estrogens in CEE, if assayed in their entirety, might relate cognitive performance in women. The exclusion of women with early or surgical menopause, and of women with impaired health, and the low participation of racial minorities limit the generalizability of our findings. Since these results are based on observation of associations, the results are hypothesis generating, although as such, they set the stage for future studies to evaluate hormonal interactions in women as they age, with and without exogenous HT relative to adiposity and insulin resistance.

Cognitive changes are commonly reported during the menopausal transition. These changes may include declines in processing speed and memory with an initial cognitive decline sometimes attenuating in the postmenopausal period [ 1 – 3 ]. Questions remain whether menopausal hormone therapy (HT) impacts these declines, especially processing speed and memory. Sex chromosomes and hormones influence important structural developments in the brain leading to sex differences in cognitive function. The hormonal effects are either organizational - those that are non-reversible and impact later responsiveness to hormone exposure, or activational - effects that are transient and fluctuate with the removal and replacement of hormones [ 4 – 7 ]. Estradiol-induced increases in limbic dendritic spines and synapses have been reported [ 8 ]. In cultured human brain slices, estradiol decreased astroglial spreading [ 9 ]. Furthermore, feedback regulation of the complex hypothalamic-pituitary ovarian hormones could be associated with physiological and neurobehavioral consequences [ 10 – 13 ]. Whether hypothalamic-pituitary ovarian hormones have direct effects on brain structure or function, or if changes are mediated through ovarian hormones remains unclear. Evaluating hormone and cognition changes during HT may provide information about how hormones impact cognitive function after menopause. At the menopause, ovarian follicle depletion leads to low levels of circulating estrogens and progesterone, reflected by high circulating levels of gonadotropins follicle-stimulating hormone (FSH) and luteinizing hormone (LH). However, the menopausal ovary continues to secrete androgens including testosterone and androstenedione [ 11 , 14 – 18 ]. Preclinical studies suggest that higher circulating levels of androstenedione, particularly when locally converted to estrone [ 19 ] correlated with impaired cognition in rats, regardless of administration of HT [ 20 – 25 ]. However, the contribution of androstenedione and estrone to memory and cognition in women who experience natural menopause is unknown. Oral and transdermal formulations of HT differentially impact the hormonal environment, which influences feedback and inter-hormonal associations [ 26 ]. It is however unclear if and how these hormonal milieus associate with clinical outcomes. Cognitive outcomes could be influenced by the timing of initiation and type of HT. For example, in the Women’s Health Initiative (WHI), women randomized to oral conjugated equine estrogens (oCEE) in conjunction with continuous medroxyprogesterone acetate initiated >10 years post menopause had lower verbal learning and memory scores on average than women randomized to placebo [ 27 ], while in the WHI Memory Study of Younger Women those aged 50–55 years of age did not demonstrate benefit or risk to cognitive function with CEE [ 28 ]. In The Kronos Early Estrogen Prevention Study (KEEPS) – Cognitive and Affective Study, there was no treatment related change in cognitive outcomes either with oCEE or transdermal 17β-estradiol (t-E2) vs placebo in healthy women who initiated HT within 3 years after menopause [ 10 ]. Similarly, the Early vs. Late Intervention Trial (ELITE) found no impact of oral estradiol plus vaginal progesterone gel on verbal memory, executive functions, or global cognition in women who initiated treatment within 6 years of menopause or women who initiated 10+ years after menopause [ 29 ]. Evaluating relationships of pituitary-ovarian hormones, particularly androstenedione, the primary steroid hormone secreted by the post-menopausal ovary, and its metabolite estrone (E1), with cognition in recently menopausal women on different types and routes of HT vs placebo could help clarify the factors associated with cognitive changes occurring during menopause. Despite the lack of overall change in the average cognitive effect of oCEE or t-E2 in KEEPS, participants’ hormonal levels differed among treatment groups vs placebo and baseline levels of endogenous hormones varied among volunteers in each randomized group. Further, given the complexity of the interrelationship between hormone levels and cognition, we speculated that changes in circulating levels of specific ovarian and/or pituitary hormones from this study might predict changes in cognitive parameters. This study evaluates, for the first time, the relationships between levels of individual gonadotropins and ovarian hormones and changes in cognition with two common HT regimens (o-CEE, t-E2) or placebo independent of treatment group in healthy, recently postmenopausal women over 4 years in the KEEPS prospective trial.