By Jeffrey T. Jensen, MD, MPH

Leon Speroff Professor and Vice Chair for Research Department of OB/GYN, Oregon Health & Science University, Portland

Dr. Jensen reports that he is a consultant for and receives grant/research support from ObstetRx, Bayer, Merck, and Sebela; he receives grant/research support from Abbvie, Mithra, and Daré Bioscience; and he is a consultant for CooperSurgical and the Population Council.

SYNOPSIS: The results of a large Finnish case-control study suggest that postmenopausal hormone therapy results in a 9-17% increase in the risk of Alzheimer’s disease. However, the small effect size, and likely confounding of use effect, does not provide strong evidence for a causal relationship.

SOURCE: Savolainen-Peltonen H, et al. Use of postmenopausal hormone therapy and risk of Alzheimer’s disease in Finland: Nationwide case-control study. BMJ 2019;364:l665.

A number of case control and other epidemiologic observational studies support a decrease in the risk of Alzheimer’s disease (AD) associated with use of postmenopausal hormone replacement therapy (HT). In contrast, the authors of the prospective, randomized Women’s Health Initiative Memory Study found that both oral conjugated equine estrogens alone, or in combination with medroxyprogesterone acetate, increased the risk of dementia and cognitive decline.1

Savolainen-Peltonen et al conducted a large national case-control study that evaluated the relationship between HT and AD among postmenopausal women in Finland. They identified as cases all postmenopausal women in Finland who received an AD diagnosis from a neurologist or geriatrician between 1999 and 2013 to qualify for reimbursement for treatment from national health insurance, a total of 84,739 women. They used the Finnish national population register to identify an equal number of controls, matching by age and hospital district. They used data on HT use available from the Finnish national drug reimbursement register. They calculated odds ratios (OR) and 95% confidence intervals (CI) for AD using conditional logistic regression analysis. Most (99%) cases received the diagnosis of AD at or after 60 years of age, with 56% diagnosed after age 80 years. Compared to never-users of HT, use of estradiol only resulted in a statistically significant overall 9% (OR, 1.09; 95% CI, 1.05-1.14) increase in the risk of AD, and use of combined HT increased the risk by 17% (OR, 1.17; 95% CI, 1.13-1.21). The differences between estrogen-only and combined HT were not significant. Use of vaginal estradiol was not related to risk of AD (OR, 0.99; 95% CI, 0.96-1.01). The use of tibolone, a selective estrogen receptor modulator with estrogenic, androgenic, and progestogenic activities, also was not associated with any increase in risk.

The authors found similar associations in women who started HT at < 60 years of age and 60 years of age. The ORs increased minimally (and inconsistently) with duration of use in both women who started HT at < 60 years of age and 60 years of age.

Based on these findings, the authors concluded that the long-term use of systemic HT “might be” associated with an overall increased risk of AD. They further concluded that the type of progestogen or the age at initiation of systemic HT does not modify the risk.


Aging is not easy. Although a good diet, exercise, and a healthy lifestyle can influence the progressive decline in physical capacity and stamina that comes with aging, all of us have seen cancer and dementia reduce both the quality and length of life. Therefore, it is not surprising that journals and the news media show great interest in reporting findings of studies that suggest a potential risk-modifying factor. Unfortunately, most published studies present more noise than news. As clinicians, we must sort through the mess to help our patients draw reasonable conclusions. How should we evaluate this new study suggesting an association of HT with development of Alzheimer’s dementia?

The short answer is that the limitation of the case-control study design does not permit any reasonable conclusion. First, a case-control study can demonstrate association, but cannot show causality. When evaluating a case-control study, the presence of uncontrolled confounding always should be suspected as the explanation for any association. The glaring confounder in this study is diagnosis bias. Women required a diagnosis of AD to qualify for benefits for AD treatment under the Finnish national insurance system; some benefits included medications. One can reasonably assume that those women who presented for medical care for AD also may have been more likely to present for (and have received) medical care for menopause. This would create a prescription bias favoring more HT use among those women diagnosed with AD. Thus, the selection of controls in this study does not appear appropriate for the comparison.

I am always impressed by how large sample size is used to demonstrate the importance of a finding. Journal editors love these large numbers; this study included almost 85,000 cases and 85,000 controls; it must be important, correct? Wrong. Large numbers allow for very precise and narrow CIs but do not influence or increase the significance of the effect size. This provides the frequent problem of a statistically significant and clinically unimportant result. Here, there was a small (9-34%) increase in the odds ratio, (e.g., 1.09-1.34). These are extremely weak associations. In general, we should ignore ORs less than 2.0. The most likely explanation is bias.2

In 2017, other Finnish investigators published results from the Kuopio Osteoporosis Risk Factor and Prevention cohort, a population-based cohort followed for 20 years. They reported a history of HT use did not change the risk of AD, but that a trend toward protection emerged with longer duration of self-reported use with an approximately 50% reduction in risk seen in women reporting > 10 years of HT (adjusted hazard ratio [HR], 0.53; 95% CI, 0.31-0.91).3 This prospective study provides much better evidence, with a more robust estimate of benefit that approaches a moderate effect (HR, 0.5). The creators of the Cache County Cohort, a prospective study of incident dementia in men and women residing in a single county in Utah, enrolled an older group of subjects (men and postmenopausal women in their 70s) than the Kuopio cohort (perimenopause to early menopause).4 In the Cache County study, men and women were equally likely to develop AD up to about age 80 years, when the relative risk for women more than doubled. Utah women who reported any use of HT exhibited a reduced risk of AD compared with non-HT users (adjusted HR, 0.59; 95% CI, 0.36-0.96). The reduction in risk showed a strong linear relationship, with duration of use becoming significant only among women using HT for more than 10 years.

Again, the prospective nature of these studies provides more confidence in the risk estimates, with the magnitude of the protective effect of long-term (> 10 years) HT on AD risk remarkably consistent. However, a healthy user effect could confound these results, so more research is needed.

As I have stated before, cognitive benefits are not an indication for HT. Clinicians should discuss the limitations of the data and the potential risks and benefits of treatment. Bad studies, like this Finnish case control, can reduce the confidence of women to continue treatment, particularly if they are asymptomatic. I feel our best data still support the notion that HT continued for at least 10 years may maximize protection against the development of AD.


  1. Craig MC, et al. The Women’s Health Initiative Memory Study: Findings and implications for treatment. Lancet Neurol 2005;4:190-194.
  2. Spitzer WO. Bias versus causality: Interpreting recent evidence of oral contraceptive studies. Am J Obstet Gynecol 1998;179(3 Pt 2):S43-S50.
  3. Imtiaz B, et al. Postmenopausal hormone therapy and Alzheimer disease: A prospective cohort study. Neurology 2017;88:1062-1068.
  4. Zandi PP, et al. Hormone replacement therapy and incidence of Alzheimer disease in older women: The Cache County Study. JAMA 2002;288:2123-2129.