Abstract
Objective: APOE-ε4 increases the risk of cognitive decline, while elderly women who take estrogen may have less risk of cognitive decline. The authors sought to determine whether estrogen use modifies the association between APOE-ε4 and cognitive decline. Method:— As part of the Cardiovascular Health Study, 3,393 Medicare-eligible women (≥65 years) were randomly selected and recruited from Sacramento County, CA; Washington County, MD; Forsyth County, NC; and Pittsburgh, PA. Cognitive testing was administered annually; the authors studied the 2,716 women with cognitive testing on ≥2 visits. They analyzed change in score on the Modified Mini-Mental State Examination (3MS) as a function of estrogen use, APOE genotype, and baseline common and internal carotid artery wall thickening.
Results: A total of 297 (11%) women were current estrogen users and 336 (12%) were past estrogen users. Over the 6-year average follow-up, baseline current users declined 1.5 points on the 3MS whereas never users declined 2.7 points (p = 0.023). Compared with ε4-negative women, ε4-positive women had a greater adjusted hazard ratio of cognitive impairment (3MS < 80), hazard risk [HR] = 1.47; 95% CI, 1.13 to 1.90. There was an interaction between estrogen use and ε4 presence (p = 0.037). Among ε4-negative women, current estrogen use reduced the risk of adjusted cognitive impairment compared with never users by almost half (HR = 0.59; 95% CI, 0.36 to 0.99), whereas, it did not reduce the risk among ε4-positive women (current use, HR = 1.33; 95% CI, 0.74 to 2.42). Compared with never use, current estrogen use was associated with less internal and common carotid wall thickening in ε4-negative women but not in ε4-positive women (p for interaction < 0.05 for both). Differences remained after adjusting for age, education, race, and stroke.
Conclusions: Estrogen use was associated with less cognitive decline among ε4-negative women but not ε4-positive women. Potential mechanisms, including carotid atherosclerosis, by which ε4 may interact with estrogen and cognition warrant further investigation.
Several studies have suggested that estrogen use in elderly women improves cognition,1 prevents development of dementia,2-4 and lessens severity of dementia.5 However, other studies have not found an association between estrogen replacement and improved cognition or dementia prevention.6,7 APOE-ε4 is a known genetic risk factor for Alzheimer’s disease (AD) and preclinical cognitive decline.8,9 Some studies have reported an interaction between gender and APOE with women having a higher ε4-associated risk of AD than men.10,11 Several lines of evidence from in vitro and animal studies, including estradiol-induced synaptic sprouting and expression of APOE messenger RNA in rodents, support an estrogen–APOE interaction.12,13 We investigated whether estrogen, in particular, carefully documented use, modifies the association between APOE and cognition. We also sought to explore a possible mechanism for an APOE–estrogen interaction by examining carotid atherosclerosis.
Methods.
The design of and subject recruitment in the Cardiovascular Health Study (CHS) has been published in detail.14 The original sample included 5,201 men and women over age 65 years who were randomly selected and recruited from Medicare eligibility files during 1989 and 1990 from four communities in the United States: Sacramento County, CA; Washington County, MD; Forsyth County, NC; and Pittsburgh, PA. The subjects were community-dwelling, able to give informed consent, and expected to remain in the geographic area for at least 3 years. In 1992 and 1993, 687 additional African-Americans were recruited by similar methods. Of the total CHS subjects, 3,393 (58%) were women and 84% were white. Because we were interested in the effect of unopposed estrogen on cognitive function, we excluded the 154 (5%) women who, either at baseline or during the study, were receiving progestins (alone or in combination with estrogen). Of the remaining 3,239 women, 2,716 (84%) had cognitive testing at the first available visit and at least one other visit and were the subjects of this study. The institutional review board of each site approved the study; all participants signed informed consent at study onset.
Cognitive assessment.
Beginning with the first follow-up visit, a trained interviewer administered the Modified Mini-Mental State Examination (3MS) annually for 5 to 7 years. The 3MS is a global cognitive scale (scored from 0 to 100), with higher scores indicating better performance. The 3MS is more sensitive than the Mini-Mental State Examination for mild cognitive impairments and tests a broader range of cognitive domains.15 A score of <80 is associated with cognitive impairment.16
Estrogen use.
Beginning with the baseline visit, study participants were asked about current and past use of oral estrogen and progestins. Current medication information was verified by inventory of prescription bottles that subjects brought to the clinic at each visit.
APOE genotyping.
Of the 2,716 women, DNA was available for APOE genotyping in 2,586 (95%). Following phlebotomy, the buffy coat was removed from centrifuged blood and stored at −70 °C. DNA was extracted using a modification of the salt precipitation method, following guidelines of the National Heart, Lung, and Blood Institute Working Group on Genetic Studies.17 The three major APOE alleles were determined at the University of Vermont by the method of Hixson and Vernier.18 We defined presence of ε4 to include women with the 4-4, 3-4, and 2-4 genotypes.
Other variables.
At the baseline visit, participants completed a comprehensive questionnaire that included information on age, age at menopause, race, and education. Subjects were asked about alcohol consumption (drinks per week for the past 30 days) and cigarette smoking. Baseline stroke history was determined during baseline examinations; cases were adjudicated by an expert panel. Weight and height were measured at clinic examinations; body mass index was defined as weight in kilograms divided by the square of height in meters. Depression was assessed at baseline using the Center for Epidemiologic Studies Depression Scale (CES-D)19 with scores ranging from 0 to 30 (higher scores indicate greater depression symptoms). Baseline measurement of common and internal carotid wall thickness using Toshiba (Tustin, CA) SSA-280A sonographic units has been reported in detail elsewhere.20 Two-dimensional brightness mode imaging was used to detect maximal media-intima wall thickness in a standardized manner.
Statistical analysis.
Baseline subject characteristics were analyzed according to baseline estrogen use (never, past, current) by χ2 for dichotomous variables and analysis of variance (ANOVA) for continuous variables. Random effects models (REMs)21 were used to examine the association of baseline estrogen use or APOE with repeated cognitive scores. We also examined the association between carotid atherosclerosis (wall thickening), estrogen use, and cognitive change using REMs. The REM adjusts for within-subject variability and repeated tests on the same subjects, taking into account the lack of independence in a participant’s observations across time. Autoregressive and unstructured correlation structures were also evaluated, but compound symmetry showed the best fit to these data. Only subjects who had cognitive data for at least two visits were included in the analyses. The association between estrogen (a fixed effect) or APOE-ε4 and cognition was analyzed by testing whether the average annual 3MS change over the 5 to 7 year period varied significantly by estrogen use or ε4 presence. Modification of the association between APOE-ε4 and cognitive decline by estrogen was examined by including an interaction term to the model. We included covariates of age, education, race (white versus other), and stroke history in the models; both as intercept terms and interacted with time. To examine the association of estrogen, APOE, and a clinically meaningful cognitive outcome, we used the method of Kaplan and Meier to produce survival curves indicating time to cognitive impairment (3MS < 80). Subjects without cognitive impairment are censored at their last nonmissing 3MS score. We then performed adjusted proportional hazards survival analysis with estrogen use (stratified by presence of ε4) or ε4 presence as the predictor variable and time to Modified Mini-Mental State Examination (3MS) < 80 as the outcome variable. All analyses were performed using SAS software (SAS Institute, Cary, NC), with p values <0.05 considered significant.
Results.
At study baseline, 297 (11%) of women were taking oral estrogen. These women tended to be younger, more educated, drank slightly more alcohol, and had lower body mass index than women who had never used estrogen (table 1). Estrogen users had less internal and common carotid wall thickness, younger age at menopause, and were more likely to be white. Of the subjects, 336 (12%) were past estrogen users and did not differ as greatly on baseline characteristics.
Baseline mean ± SD characteristics of the 2,716 women according to baseline estrogen status
At the baseline cognitive testing, the 3MS scores were higher for women taking estrogen (never: 89.7 ± 0.4 (95% CI); past: 90.9 ± 1.0 (95% CI); current: 92.5 ± 1.0 (95% CI); p < 0.0001). During the study, 251 (9%) women died; a similar proportion of women in each estrogen group died (never: 9.8%, past: 7.4%, current: 7.4%; p = 0.20). Of the survivors, 1910 (77%) completed the 5 to 7 years of follow-up cognitive testing. Those who did not complete cognitive follow-up were older, less educated, more depressed, and had lower initial 3MS scores (data not shown; p < 0.05 for all comparisons) but were not less likely to use estrogen or be ε4-positive (p > 0.05). On average, cognitive scores declined over the 5 to 7 year follow-up period. Never users had an average 3MS decline of 2.7 ± 0.4 (95% CI) whereas current users declined 1.5 ± 0.9 (95% CI) points (p = 0.023). Past users were similar to never users with an average decline of 2.4 ± 0.9 (95% CI). When these models for decline were adjusted for age, education, race, and stroke, the differences became nonsignificant (p = 0.59).
APOE, estrogen use, and cognitive decline
The APOE allelic frequencies for the subjects were as follows: 1% with 2-2, 14% with 2-3, 3% with 2-4, 59% with 3-3, 21% with 3-4, and 2% with 4-4; 677 (26%) of the subjects had one or more APOE-ε4 allele. APOE frequencies were similar in the white and nonwhite groups. There was no difference in ε4 distribution across baseline estrogen use group (never users: 530/1975 (27%), past users: 80/325 (25%), current users: 67/286 (23%); p = 0.39). Women who were ε4-negative declined an average of 1.7 ± 0.4 (95% CI) points over the average 6 years of follow-up, whereas those who were ε4 heterozygous declined 4.6 ± 0.8 (95% CI) points and those who were ε4 homozygous declined 11.9 ± 2.6 (95% CI) points. After adjusting for age, education, race, and stroke history, women heterozygous for ε4 had a 3.6 ± 0.7 (95% CI) greater point decline on the 3MS compared with those without ε4 (p = 0.0001); those who were ε4 homozygous declined 7.8 ± 2.4 (95% CI) more points (p = 0.0001). Having one or more ε4 allele was associated with greater cognitive decline for both whites and nonwhites.
Among ε4-negative women, those currently taking estrogen had a 1.5 ± 1.0 (95% CI) smaller 3MS point decline over 6 years compared with the never users (p = 0.003) (table 2). Past estrogen users’ change in scores did not differ from never users (p = 0.83). Among ε4-positive women, current or past estrogen use was not associated with the amount of cognitive decline (compared with never use: p = 0.37 for current use; p = 0.79 for past use). There was an interaction between estrogen use, APOE-ε4, and cognitive decline (p = 0.037). After adjusting for age, education, race, and stroke history, the interaction between APOE and estrogen use remained but lessened somewhat in statistical significance (p = 0.06).
Baseline estrogen use, APOE-ε4 status and change in 3MS (± 95% CI) score over the 6-year follow-up
Estrogen use and risk of cognitive impairment.
At the initial cognitive testing, 267 (10%) had 3MS scores consistent with cognitive impairment (<80) and 369 (13%) women had an incident score <80 on at least one follow-up visit. After excluding women with baseline impairment, time to cognitive impairment differed according to estrogen use (current or never) and ε4 presence (figure 1). Among ε4-negative women, 8% of current estrogen users became cognitively impaired whereas 15% of never users became cognitively impaired. Among ε4-positive women, 18% of current and 18% of never users became impaired. After multivariate adjustments, ε4-positive women had a greater adjusted hazard risk (HR = 1.47; 95% CI, 1.13 to 1.90) of cognitive impairment compared with ε4-negative women. Among women negative for ε4, current estrogen use reduced the risk of cognitive impairment (adjusted for age, education, race, and stroke history) by almost half compared with never users (current use HR = 0.59; 95% CI, 0.36 to 0.99; past use OR = 0.72; 95% CI, 0.48 to 1.08). Among women positive for ε4, current estrogen use was not associated with risk of developing cognitive impairment (current use HR = 1.33; 95% CI, 0.74 to 2.42; past use HR = 0.93; 95% CI, 0.51 to 1.72). The proportional hazards assumption held for these adjusted models.
Figure 1. Time to cognitive impairment (3MS score <80) according to APOE-ε4 status and baseline estrogen use after adjusting for age, education, race, and stroke history. ERT = estrogen replacement therapy.
Estrogen use, APOE, and carotid atherosclerosis.
Women positive for ε4 had greater internal carotid wall thickness than women negative for ε4 (p = 0.002) (figure 2). Current estrogen use was associated with less common carotid wall thickness among ε4-negative but not ε4-positive women (see figure 2). Internal carotid artery wall thickness was associated with greater 3MS decline (p = 0.0001). Common carotid wall thickness was also lower among the ε4-negative women currently taking estrogen but not among the ε4-positive women (data not shown; p = 0.019).
Figure 2. Estrogen use, APOE-ε4 status, and internal carotid artery wall thickness. ▪ = never estrogen use; 
= past estrogen use; FIGURE = current estrogen use.
No caption available.
Discussion.
We found that ε4-negative women who were current estrogen users had less cognitive decline than women who were never or past users. Among ε4-positive women, estrogen use was not associated with less cognitive decline. We also found that current estrogen use was associated with less common and internal carotid wall thickening and that APOE-ε4 modified this association.
A number of observational studies and small trials have examined the association between postmenopausal estrogen use and cognitive status and have yielded conflicting results.22 To date, there have been no results from large, randomized trials of estrogen for dementia prevention that can confirm these associations. Because women who take estrogen differ from women who do not take estrogen,23 it is important to take these differences into account as they potentially could account for the observed differences in cognitive function. Although we have adjusted for a number of these differences, such as age, education, and race, statistical adjustment cannot completely account for sources of bias. Our results, showing an interaction between APOE, estrogen use, and rate of cognitive decline, support a biologically mediated association.
Our results are not supported by one study in which estrogen was associated with better cognitive scores independent of APOE-ε4.1 However, in that study, genotyping was obtained in only half of the subjects and very few women were currently taking estrogen. We found the strongest association among women who were current estrogen users. Several lines of basic science evidence support the observation that estrogen and APOE may act synergistically. Estrogen modulates the expression of APOE in rodent tissues,13 and estradiol promotes synaptic sprouting in response to injury through an APOE-dependent mechanism.12 In neuron-specific enolase–APOE-ε4 transgenic mice, females demonstrated greater cognitive impairment than males.24 In humans, ε4 heterozygous women have a greater risk of developing AD than heterozygous men, especially those with the 2-4 genotype.10,25 The effectiveness of cholinesterase treatment for AD may vary among women according to ε4 presence but not among men; compared with those negative for ε4, women positive for ε4 had less of a treatment effect.26 The mechanism for this gender-specific APOE interaction is not clear but could be related to estrogen levels.
Carotid atherosclerosis has been shown to be associated with increased risk of both AD and vascular dementia as well as milder cognitive decline.27,28 These studies also reported an interaction between atherosclerosis, APOE, and risk of AD. ApoE is an important carrier protein for lipid distribution and ε4 is associated with atherosclerosis in the CNS.29 Estrogen may protect against CNS atherosclerosis by inducing vasodilatation, reducing central arterial smooth muscle injury response and platelet aggregation, and altering serum lipids.30-32 Our observed interaction between APOE, estrogen, and atherosclerosis may be one explanation for the differential effect of estrogen on cognitive decline depending on ε4 status. In the presence of carotid stenosis, and possibly more widespread CNS atherosclerosis, estrogen use was not protective against cognitive decline.
Several limitations of our study deserve comment. We specifically excluded the small number of women taking progestins and cannot determine if estrogen taken with progestins would have the same effect on cognitive decline. However, a recent CHS study did not find differences in carotid wall thickening among women taking estrogen or estrogen and progestins.33 We did not have a clinical evaluation for cognitive impairment or dementia and therefore cannot specify the type or underlying pathophysiology of the cognitive decline. Finally, there was some attrition in the study subjects between the first annual cognitive test and the 5 to 7 years of follow-up. Those subjects without full follow-up were older and had lower baseline cognitive scores; exclusion of these subjects could lead to bias. Our analytic approach reduced this bias, as we included all subjects who contributed a test score for each year they participated in the study.
This study supports the association of estrogen use and prevention of cognitive decline. The modification of this effect by APOE-ε4 has potential treatment and prevention implications. The potential mechanisms of this interaction, particularly carotid atherosclerosis, necessitate further study.
Acknowledgments
Supported by Public Health Service grants N01HC85083 and K23-AG00888.
Footnotes
Presented in part at the annual meeting of the American Academy of Neurology; Toronto, Canada; 1999.
- Received October 19, 1999.
- Accepted in final form February 3, 2000.
References
Disputes & Debates: Rapid online correspondence
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.