Estrogen and Dementia: Title and subTitle BreakInsights From the Women's Health Initiative Memory Study

Alzheimer disease (AD) increases exponentially with age, with an annual incidence of approximately 0.08% at age 60 to 64 years, and more than doubling every 5 years to an incidence of 0.7% at age 75 to 79 years and 1.4% at age 80 to 85 years.¹ Approximately 43% of people with AD are between 75 and 85 years.¹ A substantial number of observational studies have suggested that hormone therapy decreases incidence or delays onset of dementia, primarily AD.^{2 - 7} Laboratory studies have shown that estrogen receptors are present on cholinergic neurons and have neurotropic, antioxidant, and anti-inflammatory effects.^{8 - 9} Estrogen deficiency in middle age has been postulated to account, at least in part, for the somewhat higher incidence of AD in women than men and may constitute a risk factor for AD.

The Women's Health Initiative Memory Study (WHIMS), the estrogen-alone portion of which is reported in this issue of JAMA,^{10 - 11} was designed with this observational and basic science background in mind. WHIMS is an ancillary trial nested within the Women's Health Initiative (WHI) placebo-controlled trial of combined hormone therapy with 0.625 mg of daily conjugated equine estrogens (CEE) and 2.5 mg of medroxyprogesterone acetate (MPA) for women with a uterus,^{12 - 13} or CEE alone for those who had hysterectomy.^{14 - 15} The WHI presented the opportunity to piggyback longitudinal cognitive assessment onto a clinical trial large enough that dementia outcomes and a treatment effect might be detected.

However, the WHI, originally planned to have 9 years of follow-up for the 27 347 women aged 50 to 79 years who enrolled, ended early. First, the CEE plus MPA group of the trial was discontinued early after approximately 5 years of mean follow-up because of increased risks for coronary heart disease outcomes, stroke, venous thromboembolism, and breast cancer, although risks for hip fracture and colon cancer were decreased.¹⁴ These results were followed up with the report of the WHIMS trial of CEE plus MPA vs placebo, which revealed an increased risk of dementia with CEE plus MPA and no benefit for mild cognitive impairment or global cognition (assessed by longitudinal assessments of a brief cognitive screening test, the Modified Mini-Mental State Examination [3MSE]).^{12 - 13}

In February 2004, the CEE-only group was discontinued after nearly 7 years of mean follow-up because of increased risk for stroke and an absence of benefit for the primary cardiovascular outcomes or for cancer, although a decreased risk for hip fracture was observed.¹⁵ This study did not include the WHIMS results, and so left unanswered the issue of whether estrogen could protect women from dementia or mild cognitive impairment. If estrogen were found to have a significant protective effect on dementia and cognition, it possibly could have outweighed the adverse effects observed in the main trial.

The overall WHIMS study of CEE either alone or combined with MPA vs placebo enrolled WHI participants who were free of dementia and aged 65 to 79 years. Of 8094 eligible women, 7479 (92.4%) agreed to be followed up and assessed at yearly intervals during a planned follow-up period of 4 to 6 years (instead of the 9-year follow up for WHI as a whole).¹⁶ Women who screened positive on the 3MSE were further assessed by study physicians and with laboratory tests as indicated to diagnose dementia according to standard protocols.¹⁶

The main clinical questions were fairly specific. Does starting or restarting (31% of the sample used hormone therapy previously) either CEE alone or CEE plus MPA delay the diagnosis of dementia or mild cognitive impairment? Additional analyses evaluating global cognition, as assessed by serial 3MSEs, were performed in parallel.^{11 ,13} WHIMS was designed to have at least 80% statistical power to detect a 40% risk reduction in the incidence of dementia based on an enrollment of 8300 women and expectation of about 170 incident cases. This large risk reduction was the level considered clinically important by an outside advisory committee,¹⁶ although a smaller reduction would still have had substantial public health implications. WHIMS actually enrolled 821 fewer women and observed only 108 incident cases of dementia during a shorter than expected follow-up period of 4 to 5 years, a result of the lower than expected incidence of dementia; therefore, statistical power to find this level of risk reduction was further diminished.

After WHIMS was under way, randomized trials of CEE alone in women who already were diagnosed with mild to moderate AD showed no significant cognitive or clinically advantageous effects during the course of 4 months to a year and some evidence for deleterious cognitive effects.^{17 - 20} These results were sufficient to definitively inform clinical practice that the use of CEE for women with AD, at least at doses of 0.625 per day or higher, was ineffective, possibly harmful, and certainly not indicated.

WHIMS found similar results: CEE alone did not protect against dementia, as 47 participants were diagnosed with probable dementia, 28 assigned to CEE, and 19 to placebo (hazard ratio [HR], 1.49; 95% confidence interval [CI], 0.83-2.66). When the estrogen-alone trial was combined with the estrogen plus progestin trial, per the original WHIMS protocol, the overall HR for probable dementia was 1.76 (95% CI, 1.19-2.60; P = .005). The risk remained when women who probably had cognitive impairment at baseline were removed. The increase in dementia incidence appeared to become more pronounced with longer follow-up.

Results for mild cognitive impairment did not show clear evidence of harm or benefit. In the CEE-alone trial, mild cognitive impairment was diagnosed in 76 participants in the CEE group compared with 58 in the placebo group (HR, 1.34; 95% CI, 0.95-1.89), and when this was combined with the estrogen plus progestin trial, the HR was 1.25 (95% CI, 0.97-1.60).

The trial was underpowered to adequately assess efficacy separately in women receiving combined hormone therapy or CEE alone, or to reliably assess subgroups. However, when the increased risk for stroke is factored in, clinical uncertainty is virtually eliminated.

The incidence of types of dementia is consistent with observations from population-based studies, in which AD and AD with cerebrovascular disease, or mixed dementia, comprise about two thirds of all dementia, and vascular dementia alone about 10%. Alzheimer disease and vascular pathology often coexist with vascular pathology present in about one third of autopsied dementia cases in population-based cohorts, and pure vascular pathology accounts for about 10% of dementia.^{21 - 23} The idea that vascular effects of estrogens—increases in thrombin, fibrinolysis, triglycerides, and C-reactive protein—may exacerbate the course of dementia or may differentially contribute to an increase in vascular or mixed dementia is relevant but cannot be tested in this trial because of too few cases for subgroup analysis and no direct measures of vascular effects of estrogen.

It would have been possible to design a trial better suited to answering more definitively whether estrogen or estogen plus progesterone reduce the risk of dementia, in different age ranges and possibly with different interventions. However, conducting such a study without also assessing coronary heart disease and stroke outcomes would not have been justified. Considering the necessary compromises, the WHI trial succeeded in resolving the important issue that hormone therapy should not be given to women older than 65 years to prevent or delay onset of dementia, or with any expectation for meaningfully improving cognitive function.

The appropriateness of combining the 2 analyses, CEE alone with CEE plus MPA, as was specified in the original protocol,¹⁶ is open to debate. According to the authors, the women who had undergone hysterectomy and were enrolled in WHIMS had lower mean education and lower baseline 3MSE scores, were more likely to have had a history of stroke or coronary heart disease, and more likely to have used hormone therapy previously. The groups may differ in additional ways as a result of hysterectomy and subsequent hormone therapy. Although numbers are small, the incidence of dementia or mild cognitive impairment is approximately 33% higher in the women with hysterectomy than in those women with a uterus regardless of treatment with hormone therapy or placebo (the incidence of dementia alone appears the same). Although it is speculative to consider that women with hysterectomy may have had a longer period of estrogen deprivation than women who progressed through natural menopause because of the earlier, sudden, and complete loss of ovarian function and a failure to effectively replace estrogen over a long period, while those who underwent natural menopause had a more gradual loss over several years, such a consideration is relevant to comparing the 2 WHIMS interventions. About 74% of the sample had hysterectomy before age 50 years, and at least 40% of the women knew they had bilateral oophorectomy as well. A proportion of women received no hormone therapy after hysterectomy. Sensitivity analyses estimating each woman's postmenopausal exposure to endogenous and exogenous estrogens, in terms of years since menopause or hysterectomy, would be worthwhile, although the numbers of women may be too small to obtain reliable estimates.

Some important questions regarding estrogen therapy remain. Most important is whether short-term use of estrogen over several years in early postmenopause is effective in reducing dementia 2 or 3 decades later. This is the crux of the observational data, suggesting that previous hormone therapy during a critical period is protective while recent or current use is not.^{7 ,24} By initiating hormone therapy at an approximate mean age of 71 years and following up patients to 4 to 5 years, WHIMS is intervening fairly late in life while seeking to identify relatively infrequent earlier-onset AD cases around age 75 years. The one third of women with past history of hormone therapy did not benefit from estrogen alone either, suggesting that earlier treatment may not be effective, but this possibility remains an open issue.

A second major issue is whether attempting to prevent a diagnosis of AD within 5 years of starting hormone therapy constitutes primary or secondary prevention. For women who developed dementia during the trial, the pathophysiological processes of substantial and widespread neurodegeneration may have been well under way at the trial's outset. This can be observed clinically in the lower and progressively worsening annual cognitive scores during 3 years before dementia was diagnosed. Global cognitive function was not improved, particularly among women with lower baseline scores, leaving open the possibility that the effects of estrogen in women without early neurodegeneration has not been sufficiently tested.

The WHI can still address some of these remaining issues. First, WHIMS participants can be followed up indefinitely to evaluate potential continuing effects from the 4 to 5 years of treatment. If risk differences were due to cerebrovascular effects of CEE, those differences may attenuate after estrogen is stopped. Although the adverse effects of estrogen would likely overshadow any cognitive benefit, any protective effect would be provocative for understanding cognitive decline and dementia.

Second, the approximately 10 000 WHI participants who were aged between 50 and 60 years can be screened cognitively biyearly during the next decade or longer to assess whether their early exposure to CEE affects their rate of developing dementia. This unique population, randomly and systematically exposed to hormone therapy or placebo for an average of about 5 to 7 years beginning when they were aged approximately 50 years, could provide the opportunity to conduct a true prevention trial to assess whether hormone therapy in early postmenopause may have neuroprotective effects against an illness that will not become apparent for 2 decades.

The WHIMS results do not prove that estrogen therapy has no effect on AD or dementia, but they do clearly indicate that women older than 65 years should not be treated with CEE with or without MPA to attempt to prevent dementia or enhance cognition. Whether with different populations, lower doses of CEE, other forms of estrogen or receptor modulators, or delivered in lower more physiological doses, estrogen therapy could eventually be proven beneficial remains to be seen. However, the harmful to neutral effects of estrogen in the WHI and WHIMS trials will make further development of and research with estrogen therapy a daunting task.