Answers are grounded in this paper and cite their sources.
The prevalence of dementia in women over age 85 is about 50%, and there is no known effective preventive treatment. Thus, even a small effect of estrogen in preventing dementia or delaying its onset could have a substantial public health benefit. In this issue of theJournal, a cross-sectional, observational study of older women shows a modestly higher score on a screening test for dementia in users of hormonal replacement therapy.1 How does this fit into the current evidence for or against an effect of estrogen for the prevention of dementia? At this time, both animal and human studies provide mixed support for a beneficial effect of estrogen replacement for primary prevention in postmenopausal women. The ongoing Women's Health Initiative, a randomized trial of hormone replacement therapy (HRT), includes all of the outcomes of importance in older women: coronary heart disease, osteoporosis, breast cancer, and dementia,2 and this should be completed early in the next century. In the meantime, new observational studies and small trials must be considered in the context of the evidence to date. Observational studies of estrogen use have focused either on its association with dementia or on its associations with cognitive test performance. The strongest designs are prospective studies where HRT use has been ascertained before the onset of dementia. There have been several fairly recent, prospective, observational studies with dementia defined as the outcome. One of these studies had only 1 to 5 years follow-up, and only nine cases of Alzheimer's disease in 156 estrogen users.3 Another study had similar findings but with 34 cases of dementia and with a much longer follow-up of 16 years.4 A case-control design based the diagnosis of dementia on death certificate diagnosis.5 As with most observational studies of estrogen use, it is difficult to exclude “healthy user” effects, recall, or survival bias. These findings might also be mediated by reductions in cerebrovascular disease. Though limited by design issues, these studies all showed about a 30 to 50% lower rate of Alzheimer's disease in users of HRT. A meta-analysis of these recent studies with other earlier studies supports about a 29% reduction in risk of incident dementia in estrogen users.6 Because dementia is fairly infrequent until after age 80 and is labor intensive to document, some studies have examined tests of cognitive function and decline in test scores as surrogate outcomes for dementia. In these studies, a finding of lower cognitive function at follow-up could be attributable to the presence of cases of early dementia in the cohort that brings down the mean scores, or to variation in the overall rate of decline in cognitive function, or to both of these factors. Most studies of cognitive decline have found no evidence for a protective effect of estrogen use on future cognitive function or decline. Strong evidence against an effect comes from a longitudinal study of 800 women, aged 65 to 95, living in Rancho Bernardo, California. This study was designed to measure cognitive function 15 years after baseline assessment of HRT status.7 The cognitive battery broadly assessed cognitive function, including memory, language, and visual spatial skills. Test scores were not associated with current or past use of HRT or duration of use or dose of estrogen. Half were past users and a third were current users of estrogen at baseline; the vast majority reported unopposed estrogen. Another more recent study examined the association of self-reported estrogen use to both baseline cognitive function, as well as to cognitive decline at follow-up 4 to 6 years later, in the 9651 women participating in the Study of Osteoporotic Fractures (SOF).8 Estrogen use was associated cross-sectionally with higher baseline performance on the Folstein Mini-Mental State Exam (MMSE) as well as on two other tests, the Trails B test and Digit Symbol Substitution Task. At baseline, the proportion scoring less than 23 on the MMSE was 20.5% in never users compared with 14.3% and 14.5% in current and past users, respectively. Age and educational attainment related were more strongly to the MMSE score in the never users than the past or current users, but the cross-sectional association remained significant with adjustment. However, all groups showed similar rates of decline over 4 to 6 years of follow-up, and education was the strongest predictor of decline. The study reported here by Steffens et al.1 actually shows cross-sectional results similar to those found in the SOF study, yet it comes across as a much more positive study. This study was designed to screen for dementia in a community (Cache County, UT). An education-adjusted cutoff on the modified (100 point) MMSE was used as initial criteria for further assessment of dementia. It is not possible to evaluate this from the data presented, but it seems that the use of an education-adjusted cut-off to screen for dementia might have biased the study group. Education adjustment essentially gives lower education subjects “extra credit” and thus could decrease the proportion ascertained for dementia in those with borderline scores and less education. Potentially, this could allow more of those with lower MMSE scores into the study, and these may be individuals with lower estrogen use. Another group has reported that the use of age-adjusted cut-points did not effect the sensitivity and specificity of dementia diagnosis in a lower SES group.9 This may or may not present a bias. After evaluation, some of those who initially screened positive for dementia were determined not to have dementia, and they were included in the nondemented group. In these 2338 women without dementia, the mean MMSE scores were 92.1 for never users, 93.5 for past users, and 94.4 (out of 100 possible points) for current users of estrogen. This relationship of HRT use to MMSE score was substantially diminished, but it remained statistically significant after adjustment for age, education, the presence of an apolipoprotein E isoform epsilon 4 (Apo E e4) allele and depression. No significant dose-response relationship of hormone replacement to test score was found. In a series of models, the variance in the MMSE score explained by estrogen was only R2= .035, and this increased to .274 with addition of age, education, and disability to the model. Thus, the effect of estrogen use on the total cognitive score was quite small. Interestingly, current and past users of estrogen in this study were more likely to meet criteria for depression, a finding that is contrary to other studies and might be related to the increased use of combination therapy with progestin.10 This finding might have biased the results away from a larger effect. Further work is needed on the effects of combination therapy on mood. Of note, Apo E e4 was associated independently with lower score, similar to other studies that have shown that higher age, Apo E e4, and lower education have been the most consistent risk factors for dementia. The potential interaction of estrogen use with Apo E e4 was not examined. It is also important to point out that the modified MMSE was the only test performed in these women and that the scores for all user groups were clearly in the normal range. Additional tests without such a ceiling effect may have been more revealing in that the MMSE underestimates cognitive decline in those with higher education and higher premorbid cognitive function. Finally, as mentioned above, this analysis was performed after excluding cases of dementia. Presumably, a description of the HRT use in these cases of dementia will be reported elsewhere. Together, these issues make it very difficult to judge whether the small, but positive, statistical association of HRT use with the MMSE score is attributable to these or other factors in the design. Therefore, this and other studies of cognitive function seem to indicate that die effects of estrogen on decline are very small and, at most, are much less than the effects of age and education. Perhaps the effects of estrogen are only apparent in those already experiencing significant decline, and there could be too few cases of dementia in the longitudinal studies to cause differences in mean test scores, especially when the majority of the women seem to be having minimal change. Perhaps there is an interaction with genetic factors, such that only certain subgroups benefit. It has been shown that estrogen can effect the expression of the Apo E gene in animal models, and in women, Apo E e4 predicts the rate of decline in cognitive function. Or, perhaps, estrogen replacement is not required for the maintenance of cognitive function in late life. The good news is that the majority of the postmenopausal women in these studies, especially those with higher education, show minimal decline in cognitive function with age, regardless of history of HRT use. In the future, ongoing observational studies of older women may have the power to look at dementia as an outcome if care has been taken to collect adequate data to define dementia in an unbiased fashion. Dementia is a difficult diagnosis to make from death certificates or medical records, and more definitive studies should incorporate a clinical evaluation for dementia diagnosis. Although not as powerful as a randomized clinical trial design, observational studies of estrogen use have provided strong and consistent evidence for a beneficial effect of estrogen use for other important health outcomes: mortality,13 coronary artery disease,13, 15 and osteoporotic fracture.16 With further follow-up and more cases of dementia, these same studies may, in the future, provide better evidence of an independent effect of HRT on the reduction of dementia incidence or may delay its onset. Ultimately, a clinical trial would still be needed to provide definitive evidence.17 The evidence for widespread use of estrogen to prevent dementia or cognitive decline is not yet sufficient. Currently available evidence for an observed effect for either prevention of dementia or cognitive decline is quite limited, especially when compared with the consistent risk reductions observed for the outcomes of mortality, coronary heart disease, and osteoporotic fracture. Dementia prevention should not be oversold to women as one of the indications for estrogen use in late life. Longitudinal follow-up of well defined cohorts and, we hope, a definitive outcome from the Women's Health Initiative, may give us more to go on in the future.
No takes yet. Share an insight, caveat, or question.
Anne B. Newman (1999) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: