Rice MM, Graves AB, McCurry SM, Gibbons LE, Bowen JD, McCormick WC, Larson EB. Postmenopausal Estrogen and Estrogen-Progestin Use and 2-Year Rate of Cognitive Change in a Cohort of Older Japanese American WomenThe Kame Project. Arch Intern Med. 2000;160(11):1641-1649. doi:10.1001/archinte.160.11.1641
Copyright 2000 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2000
The relation between estrogen and cognition among postmenopausal women remains controversial. Also uncertain is whether the proposed association varies between women taking unopposed estrogen and those taking estrogen combined with progestin.
To determine whether unopposed estrogen and combined estrogen-progestin use were associated with the rate of cognitive change in a cohort of older, Japanese American, postmenopausal women.
A prospective observational study in a population-based cohort of older Japanese Americans (aged ≥65 years) living in King County, Washington. Cognitive performance was measured in 837 women at baseline (1992-1994) and 2-year follow-up (1994-1997) examinations using the 100-point Cognitive Abilities Screening Instrument (CASI). Least squares means general linear models were used to estimate the 2-year rate of cognitive change according to categories of postmenopausal estrogen use.
Approximately half of this cohort (n=455) had never used estrogen at any time since menopause, 186 were past users, 132 were current unopposed estrogen users, and 64 were current estrogen-progestin users. The majority of current estrogen users were taking conjugated estrogens, and all women receiving combined therapy were taking medroxyprogesterone acetate. After adjusting for age, education, language spoken at the interview, surgical menopause, and baseline CASI score, women who had never used postmenopausal estrogen improved slightly on the CASI scale (mean adjusted change, 0.79; SEM, 0.19). This change was significantly greater for current unopposed estrogen users (mean adjusted change, 1.68; SEM, 0.36; P=.04) and significantly worse for current estrogen-progestin users (mean adjusted change, −0.41; SEM, 0.50; P=.02) compared with never users. The improvement observed in past users (mean adjusted change, 1.12; SEM, 0.29) was intermediate between the changes for never users and current unopposed estrogen users and not significantly greater than that for never users (P=.35).
Our findings support a modest beneficial association between current unopposed estrogen use and the rate of cognitive change. We also observed a modest detrimental association between current estrogen-progestin use and the rate of cognitive change. The clinical significance of these modest differences, however, is uncertain. Data from large, long-term randomized trials are required before applying this information to the clinical setting.
ESTROGEN THERAPY has been examined as a possible preventive or treatment strategy for postmenopausal women with Alzheimer disease (AD) and normal age-related memory loss. Recent prospective studies1,2 have shown that women receiving postmenopausal estrogen therapy are about half as likely to develop AD as women not receiving estrogen therapy. In a cross-sectional population-based study,3 the prevalence of AD was significantly less in women who had ever used postmenopausal estrogen compared with never users. One case-control study4 found a significant relation between current estrogen use and AD, and another5 found estrogen dose and duration to be associated with AD and related dementia, whereas other case-control studies6- 11 did not find a statistically significant relation between postmenopausal estrogen use and AD. The potential use of estrogen as a treatment for AD has been studied only in very small, nonrandomized studies12- 14 and one placebo-controlled randomized study.15 Generally, these trials have found that women taking estrogen improved in some, but not all, of the global cognitive instruments used.
Observational and experimental studies16- 32 conducted in postmenopausal women without dementia have generally been inconclusive. These studies have been reasonably consistent in observing a positive association between estrogen therapy and recent verbal memory16,20,21,23,25,26,31 and reasoning,19,20,28,31 whereas other cognitive domains have either shown no association or mixed results.16- 24,27- 30,32
The strongest support for a relation between estrogen and brain function comes from studies in basic neurobiology. Estrogen may influence brain structure and function by stimulating neuronal growth,33- 39 interacting with neurotrophins,40,41 stimulating choline acetyltransferase (the enzyme that synthesizes the neurotransmitter acetylcholine),42,43 protecting neurons from glutamate and β-amyloid toxic effects,44- 46 and enhancing cerebral blood flow.47,48 In addition, estrogen may alleviate depressive symptoms by decreasing the activity of type A monoamine oxidase,49- 51 the enzyme that catabolizes serotonin.
In contrast to the proposed beneficial influence of estrogen, progesterone may have a negative influence on brain structure and function. Progesterone initiates involution of neuronal structures36,37 and increases type A monoamine oxidase activity.49 It is not known whether the proposed association between estrogen and cognition is modified by the addition of a progestin.
The purpose of this analysis was to determine whether unopposed estrogen and combined estrogen-progestin use were associated with the rate of cognitive change in a cohort of older postmenopausal women.
The Kame Project is a longitudinal, population-based study of memory and aging designed to establish prevalence and incidence rates of dementia in the older Japanese American community in the area of Seattle, Wash. Detailed information regarding the methods of this study have been described previously.52 Briefly, all men and women who were at least 65 years old and of at least 50% Japanese ancestry, and who were living in King County, Washington, on November 1, 1991, were eligible to participate. To identify these individuals, we conducted a census using King County telephone directories, Health Care Financing Administration (Medicare recipients) lists, Japanese American organizational lists, and word-of-mouth. A cohort of 1318 men and 1727 women was assembled from this census, of whom 858 men (65%) and 1127 women (65%) participated in the baseline examination between May 5, 1992, and September 7, 1994. This study was approved by the Human Subjects Review Committee at the University of Washington, Seattle, and written informed consent was obtained from all participants. Examinations were administered in English or Japanese by a trained interviewer, depending on the participant's preference.
At baseline, the 1127 female participants were asked about their medical history, medications and supplements, reproductive history, lifestyle habits, demographics, acculturation, and diet. Participants were asked about past and current postmenopausal estrogen use, and the current medications were inspected. The baseline examination also included measurements of blood pressure, anthropometry, neurologic function, the 11-item Center for Epidemiological Studies Depression Scale (CES-D),53,54 and the Cognitive Abilities Screening Instrument (CASI).55 The CASI is a 100-point measure of global cognitive function designed for trans-Pacific studies in Seattle; Honolulu, Hawaii; and Japan. It combines the Mini-Mental State Examination,56 the Hasegawa Dementia Screening Scale,57 and the Modified Mini-Mental State Examination,58 and can be separated into 9 cognitive domains. These domains include abstract reasoning (12 points), category fluency (10 points), attention (8 points), recent verbal memory (12 points), mental tracking (10 points), long-term memory (10 points), language (10 points), orientation (18 points), and drawing (10 points). All women scoring less than 81 points on the CASI and a weighted sample of those scoring 81 points or greater were selected for further neuropsychological testing and examination by a physician to determine their dementia status according to the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition (DSM-III-R).59 After the sampling weights were applied to the estimated prevalence rates, 96% of all prevalent cases scored less than 81 points, 4% scored between 81 and 86.9 points, and 0% scored 87 points or greater on the CASI. All women who were diagnosed as having dementia at baseline (n=100) or whose dementia status was unknown at baseline (n=32) were excluded from this study. Among the 995 eligible women, 895 (89.9%) completed the 2-year follow-up between May 26, 1994, and October 13, 1997. The 2-year follow-up visit included much of the same data collected at baseline, including a repeated measure of the CASI. All women with a 2-year follow-up CASI score of less than 87 points were selected for further neuropsychological testing and examination by a physician to determine their dementia status according to DSM-III-R. This cutoff score was used because no prevalent cases were found among persons scoring 87 points or greater on the CASI,52 and because previous studies conducted in Seattle found that a cutoff score of 86 points optimized the sensitivity and specificity of the CASI in distinguishing cases from control subjects.55,60
Least squares means general linear models were used to estimate the mean adjusted 2-year rate of cognitive change, according to categories of postmenopausal estrogen use. Multiple regression models were used to examine the association between postmenopausal estrogen use and 2-year rate of cognitive change and to compute P values. The rate of cognitive change was defined as the 2-year follow-up CASI score minus the baseline CASI score. Because the time between the baseline and 2-year follow-up examinations varied (mean ± SD, 738 ± 101.5 days), standardized scores were calculated by dividing the change in the score by the number of days between examinations and then multiplying by 730 days. The 2-year rate of change was estimated for the total CASI score (global cognitive performance) and for the cognitive domains of abstract reasoning, category fluency, attention, recent verbal memory, and mental tracking. We did not investigate rates of change in the cognitive domains of long-term memory, language, orientation, or drawing because of the limited variability of the data for these domains.
Postmenopausal estrogen use was grouped according to timing and type (never, past, current unopposed estrogen, and current estrogen-progestin). Past use included both unopposed estrogen and combined estrogen-progestin use because the specific type of therapy previously used was not available. Regression models were constructed to determine the potential confounding effects of numerous factors. Variables considered were age, education, income, marital status, place of birth, language spoken at the interview, place of formal education, nonsteroidal anti-inflammatory drug use, regular health checkups, thyroid disease, history of cancer, history of vascular events, blood pressure, body mass index (calculated as the weight in kilograms divided by the square of height in meters), physical activity, smoking, alcohol, dietary fiber, dietary and supplemental antioxidants, age at menarche, age at menopause, type of menopause, history of a hysterectomy, history of depression, and CES-D score. The appropriate scale of each variable was determined, and the influence of outlying data was evaluated. All analyses were performed using Statistical Analysis System software (SAS Institute, Inc, Cary, NC).
A total of 837 women were included in this analysis. The characteristics of these women are summarized in Table 1. In addition, Table 1 summarizes the characteristics of 158 women who were excluded from this analysis either because they did not participate in the 2-year follow-up examination owing to death (n=31), refusal (n=62), or moving out of the area (n=7) or because they did complete the 2-year follow-up, but had missing baseline data on postmenopausal estrogen use (n=50) or had invalid CASI scores owing to sensory impairment (n=8). In comparison, the women who were not included in the present analysis were older, less educated, and more likely to have spoken Japanese at the interview, and performed worse on the baseline CASI. A large percentage of the women who were not included in the present analysis had missing data regarding type of menopause (43.0%) and postmenopausal estrogen use (39.2%).
Of the 837 women included in this analysis, 455 (54.4%) had never used postmenopausal estrogen, 186 (22.2%) were past users, 132 (15.8%) were current unopposed estrogen users, and 64 (7.6%) were current estrogen-progestin users. Compared with never users, current unopposed estrogen and estrogen-progestin users were younger, more educated, and more likely to have spoken English at the interview, and performed better on the baseline CASI (Table 2). Past estrogen users, compared with never users, were more educated and more likely to have spoken English at the interview. Current unopposed estrogen users and past estrogen users were more likely to have undergone a surgical menopause (removal of uterus and/or bilateral oophorectomy prior to onset of natural menopause) than never users, whereas current estrogen-progestin users were less likely to have undergone a surgical menopause. The mean ± SD age at menopause was younger for current unopposed estrogen users (45 ± 6.6 years) compared with current estrogen-progestin users (51 ± 4.7 years), past users (47 ± 6.7 years), and never users (49 ± 5.7 years). Correspondingly, the mean ± SD duration of postmenopausal estrogen use was longer for current unopposed estrogen users (15 ± 11.7 years) compared with current estrogen-progestin users (7 ± 7.8 years) and past users (4 ± 5.8 years). Almost all current unopposed estrogen and estrogen-progestin users had regular health checkups (97.6% and 95.2%, respectively), higher than the percentages of never and past users (80.5% and 87.9%, respectively). A history of cancer was less frequent among current estrogen-progestin users (1.6%) than never (10.5%), past (14.3%), or current unopposed (9.5%) estrogen users. A history of depression was uncommon in this cohort, and scores on the CES-D scale were very low and did not differ between groups.
The majority of current unopposed estrogen and estrogen-progestin users were taking 0.625 mg of conjugated estrogens (55.3% and 57.8%, respectively). Conjugated estrogens in other doses were taken by 11.3% and 11.0%, respectively. Esterified estrogens were taken by 21.2% and 20.3%, micronized estradiol was taken by 6.1% and 3.1%, and transdermal estradiol was taken by 1.5% and 3.1%, respectively. The remaining 4.6% of current unopposed estrogen users and 4.8% of current estrogen-progestin users were taking less common orally administered estrogens, including estropipate and ethinyl estradiol. All current estrogen-progestin users were taking medroxyprogesterone acetate (MPA), with 65.6% taking 2.5 mg, 17.2% taking 5.0 mg, and 17.2% taking 10.0 mg. Data regarding the type and dose of estrogen and progestin were not available for past users.
Variables included in the multivariate least squares means and regression models were age (continuous), education (continuous), language spoken at the interview (English or Japanese or mixed), type of menopause (natural, surgical, or unknown), and baseline CASI score (continuous). Allowance for other potentially confounding variables (income, regular health checkups, history of cancer, history of depression, CES-D score, or any of the other variables examined for confounding) did not substantially modify any of the estimates.
Women who had never used postmenopausal estrogen improved slightly on the CASI scale (mean adjusted change, 0.79; SEM, 0.19) (Table 3). This change was significantly greater for current unopposed estrogen users (mean adjusted change, 1.68; SEM, 0.36; P=.04) and significantly worse for current estrogen-progestin users (mean adjusted change, −0.41; SEM, 0.50; P=.02) compared with never users. The improvement observed in past users (mean adjusted change, 1.12; SEM, 0.29) was intermediate between that for never users and current unopposed estrogen users and not significantly greater than that for never users (P=.35). Current unopposed estrogen use was associated with improved cognitive performance in the domains of abstract reasoning (P=.002) and category fluency (P=.02), but not attention (P=.15), recent verbal memory (P=.75), or mental tracking (P=.87). Current estrogen-progestin use was associated with declined cognitive performance in the domain of mental tracking (P=.005), but not abstract reasoning, (P=.65), category fluency (P=.35), attention (P=.45), or recent verbal memory (P=.11). Past postmenopausal estrogen use was associated with improved cognitive performance in the domain of abstract reasoning (P=.009).
All 837 women included in the present analysis did not have dementia at baseline. However, 30 of these women (21 never, 4 past, 3 current unopposed estrogen, and 2 current estrogen-progestin users) screened positive for mild cognitive impairment (MCI) at baseline (baseline CASI score <81 points). At the 2-year follow-up examination, 12 women (9 never and 3 past users) were diagnosed as having incident dementia. In addition, 13 women (10 never, 2 past, and 1 current unopposed estrogen users) were considered to have unknown incident dementia status because they were selected for but refused neuropsychological testing and examination by a physician. Eight of these women with incident and uncertain incident dementia also screened positive for MCI at baseline. Because incipient dementia and/or cognitive impairment at baseline may have influenced the validity of the data collected, selection of estrogen use, or compliance among those who were prescribed estrogen, we conducted an analysis that excluded all women with incident dementia, all women with unknown incident dementia status, and all women scoring less than 81 points on the baseline CASI (n=790; data not shown). The mean adjusted rates of cognitive change and the SEMs were virtually unchanged from the analysis that included all 837 women.
Because menopause type differed substantially between current unopposed estrogen users and current estrogen-progestin users, we conducted an analysis stratified by the type of menopause. This allowed the use of reference groups specific to women with a natural menopause (n=540) and women with a surgical menopause (n=240) (Table 4). Among women with a natural menopause, the same trends were observed; however, the reduced sample size for women who were currently taking unopposed estrogen (n=38) limited our ability to observe the same significance levels as in the analysis that included all 837 women. Women who had never used postmenopausal estrogen improved slightly on the CASI scale (mean adjusted change, 0.93; SEM, 0.21). Compared with never users, this change was greater for current unopposed estrogen users (mean adjusted change, 1.94; SEM, 0.63; P=.13) although no longer statistically significant, and significantly worse for current estrogen-progestin users (mean adjusted change, −0.20; SEM, 0.51; P=.04). Current unopposed estrogen use was associated with improved cognitive performance in the domains of category fluency (P=.048) and, not significantly, abstract reasoning (P=.07). Current estrogen-progestin use was associated with declined cognitive performance in the domain of mental tracking (P=.004). Women who had used postmenopausal estrogen in the past generally had changes intermediate between those of never users and current unopposed estrogen users. Past postmenopausal estrogen use was associated with improved cognitive performance in the domain of abstract reasoning (P=.04). Women with a surgical menopause showed less of an improvement in their CASI scores than women with a natural menopause. Among women with a surgical menopause, women who had never used postmenopausal estrogen improved slightly on the CASI scale (mean adjusted change, 0.47; SEM, 0.46). Compared with never users, this change was greater for current unopposed estrogen users (mean adjusted change, 1.11; SEM, 0.42; P=.33) although not statistically significant. Current unopposed estrogen use was associated with improved cognitive performance in the domains of abstract reasoning (P=.02) and, not significantly, category fluency (P=.17). Current estrogen-progestin users with a surgical menopause were not included in this analysis because of the small sample size (n=2).
In this cohort of 837 older Japanese American women, current unopposed estrogen use appeared to have a modest beneficial association with the 2-year rate of change in global cognitive performance (total CASI score), abstract reasoning, and category fluency. Estrogen-progestin use was not beneficial to any of the cognitive domains examined and appeared to have a modest detrimental association with the 2-year rate of change in global cognitive performance and mental tracking. The 2-year rate of cognitive change in women who had used postmenopausal estrogen in the past was intermediate between rates in women who had never used postmenopausal estrogen and current unopposed estrogen users.
Findings from other studies examining the association between estrogen and cognition in women without dementia have been inconclusive. However, estrogen may only affect specific cognitive domains.61,62 Studies have found estrogen to be associated with certain types of recent verbal memory, including paragraph recall,20,21,23,25 proper name recall,26 and associate learning,16,21,23 but not usually word recall.24- 26,29 We did not observe an association between unopposed estrogen use and recent verbal memory, which included word and object recall. Consistent with other studies, we found unopposed estrogen use to be associated with reasoning19,20,28,31 and category fluency.24,29 Also consistent was our failure to observe an association between unopposed estrogen use and attention17,18,22,23,28,32 or mental tracking.16,18,24,32 We observed a beneficial association between unopposed estrogen use and global cognitive performance; previously mixed results have been reported.24,27
Epidemiologic data regarding the independent associations between unopposed estrogen and combined estrogen-progestin use and cognition are limited. Kampen and Sherwin25 compared the cognitive function of women taking unopposed estrogen with that of women receiving estrogen-progestin therapy and found no difference; however, the sample sizes were small (n=12 for unopposed estrogen use; n=16 for estrogen-progestin use, with all but 1 individual taking MPA). In a large cohort study, Szklo et al29 observed an association between word fluency and postmenopausal replacement therapy in young women aged 48 to 57 years who had surgical menopause but not in the young women aged 48 to 57 years who had natural menopause. One explanation may be that the women who had surgical menopause were likely taking unopposed estrogen, whereas the women who had natural menopause were likely receiving estrogen-progestin therapy. Alternately, since these women were in the early stages of menopause, the observed benefit among the women who had surgical menopause may have been because of the relief of menopausal symptoms. In an observational study of clock-drawing ability, women who drew abnormal or blank clocks had higher serum progesterone concentrations than women with normal clock drawings.63
Studies have compared the associations between unopposed estrogen use and combined estrogen-progestin therapy and vascular risk factors in women. Findings from the Postmenopausal Estrogen/Progestin Interventions trial64,65 showed that MPA, but not micronized progesterone, blunted the beneficial association between estrogen and high-density lipoprotein cholesterol. Furthermore, 2-hour glucose concentrations increased twice as much in the estrogen plus MPA group as in the estrogen alone or estrogen plus micronized progesterone groups.64,65 Similarly, in a double-blind randomized trial, Gelfand et al66 found that estrogen in combination with dydrogesterone resulted in better vascular risk profiles (lipoprotein concentrations and glucose tolerance) than estrogen plus MPA. Another study67 similarly found that women taking unopposed conjugated estrogen had a greater decrease in lipoprotein(a) concentrations and a greater increase in high-density lipoprotein cholesterol concentrations than women taking conjugated estrogen plus MPA. These data provide support for the differential effects of unopposed estrogen and estrogen combined with MPA on vascular risk factors, and perhaps on cognition. Vascular-related changes, including less favorable lipid concentrations and reduced glucose tolerance, have been observed in patients with AD.68
Other explanations for our findings should be considered. It is well established that women taking postmenopausal estrogen tend to be healthier. Although we considered the potential confounding effect of numerous demographic, health, and lifestyle variables, unknown characteristics may explain the observed beneficial association between unopposed estrogen use and the rate of cognitive change. In our cohort, women taking postmenopausal estrogen were more likely to have had regular health checkups. Although further adjustment for this variable did not change our results, this variable may have been too crude of a measure of medical care. Women who regularly visit their physicians will have more opportunities to be diagnosed as having hypertension, hyperlipidemia, hypothyroidism, and other conditions. These conditions may go undetected and untreated in women who do not regularly see their physicians. This bias, however, cannot explain the observed differences between unopposed estrogen and estrogen-progestin users in this study. Estrogen-progestin users appeared to have the same access to health care, educational attainment, and healthful habits as unopposed estrogen users. The differences that were observed either were adjusted for (age and surgical menopause) or were found not to influence the results (age at menopause, duration of postmenopausal estrogen use, and history of cancer). In the analysis restricted to women with a natural menopause, the differences in cognitive change between unopposed estrogen and estrogen-progestin users persisted. In addition, with regard to duration of use, we found no evidence that the longer mean duration of use in unopposed estrogen users, compared with estrogen-progestin users, explained the differences observed in 2-year cognitive change. Among a subset of women who used postmenopausal estrogen for less than 5 years, CASI scores of current unopposed estrogen users improved during the 2-year period (mean adjusted change, 2.07; SEM, 0.72), whereas scores of current estrogen-progestin users declined (mean adjusted change, −0.48; SEM, 0.69).
Of the 995 women who participated in the baseline examination and were eligible for the 2-year follow-up, 158 (15.9%) were not included in the present analysis. Table 1 shows that these women were older, were less educated, were more likely to have spoken Japanese at the interview, and performed worse on the baseline CASI, and many had missing data regarding postmenopausal estrogen use. Based on these characteristics, it is most likely that these women were not current estrogen users and may have been more likely to have declined cognitively. Therefore, exclusion of these women more than likely decreased the magnitude of the positive association between unopposed estrogen use and cognitive change. On the other hand, exclusion of these women may have increased the magnitude of the negative association between estrogen-progestin use and cognitive change.
The apparent beneficial association of estrogen with cognition may be attributed to changes in mood and well-being.69- 72 Although further adjustment for history of depression and CES-D score did not alter our results, it is likely that these measures were not sensitive enough to pick up subtle differences in mood and well-being that may have influenced cognitive performance. Hormonal effects on mood and well-being also may explain the observed differences between current unopposed estrogen and estrogen-progestin users. Several trials have shown that women receiving estrogen-progestin therapy report more negative moods or depressive symptoms than women taking estrogen only.73- 76 Higher estrogen-progestin dose ratios may attenuate this apparent negative association between progestins and mood and well-being.74,76 Because of limited power, we were unable to examine the influence of varying estrogen-progestin dose ratios on the rate of cognitive change.
We observed modest improvements in cognitive performance among the cohort as a whole. These improvements may be attributable to practice effects, regression toward the mean, cohort characteristics, or attrition. Frank et al77 examined changes in neuropsychological tests administered 2.4 years apart in an elderly community-based sample. Similar to our results, they found that scores on the Mini-Mental State Examination (a measure of global cognitive function) improved by about 1 point among the individuals who were not demented. One explanation suggested for this improvement was related to the subjects' characteristics. Their sample was made up of upper-middle class individuals with a low rate of depression who were highly educated. Likewise, our cohort was more educated than the average population, with a low rate of depression, and included only individuals who were not demented at baseline.
The observed beneficial association between unopposed estrogen use and cognitive change was modest. The clinical significance of these modest differences is yet unknown. Our results reflect cognitive changes during a 2-year period in women who were primarily in their 7th and 8th decades of life. The importance of postmenopausal estrogen use may become more apparent among women as they age into their 9th and 10th decades and are at greater risk of cognitive decline. This article does report statistically significant results that are not likely due to chance and thus help us better understand the potential roles of unopposed estrogen and estrogen-progestin use in cognitive function. Strengths of this study include a prospective examination of postmenopausal estrogen use in cognitive performance in a large cohort of older women and the control of confounding factors, including age, education, language spoken at the interview, and surgical menopause. This analysis also adjusted for baseline cognitive score, which in itself controls for many of the problems associated with cross-sectional analyses and enabled us to examine cognitive change in women with the same baseline cognitive status but different postmenopausal estrogen usage. In addition, current estrogen status was based on inspection of medications, rather than relying on self-report.
Our findings support a modest beneficial association between unopposed estrogen use and the rate of change in global cognitive performance, category fluency, and abstract reasoning. This beneficial association between estrogen and cognitive change appeared to be opposed by the addition of MPA. Because of the observational nature of this study, data from large, long-term randomized trials are required before applying this information to the clinical setting. Although women who still have their uterus cannot be randomly assigned to long-term unopposed estrogen use, random assignment to use of estrogen combined with various progestin types and doses may shed light on this topic. In addition, randomized trials examining the effects of nonuterotrophic estrogens on cognitive function are needed.
Accepted for publication November 17, 1999.
This work was supported by grants AG09769 and AG11143 and a graduate research supplement from the National Institute on Aging, Bethesda, Md.
Presented in part at the Third Annual Graylyn Conference on Women's Health, "Women's Cognitive Health: The Role of Estrogen," Wake Forest University, Winston-Salem, NC, October 16, 1997, and at the American Geriatrics Society Annual Meeting, Seattle, Wash, May 7, 1998.
The authors thank Bruce Psaty, MD, PhD, for providing the medications software program. We gratefully acknowledge the Japanese American community of King County, Washington, and we thank the Kame Project staff for their dedicated work in recruitment, data collection, and data management.
Corresponding author: Madeline Murguia Rice, PhD, University of Washington Kame Project, Pacific Medical Center, Box 358261, 1200 12th Ave S, Quarters 10, Seattle, WA 98144 (e-mail: email@example.com).