A million women's worth of data on hormone therapy and dementia returns a flat line

Dementia is more common in women than in men, and one long-running hypothesis for why involves the menopause transition — specifically, what happens to cognition when circulating sex steroids fall [s1]. If that hypothesis is right, hormone therapy ought to alter dementia risk in one direction or the other.

A systematic review and meta-analysis published in The Lancet Healthy Longevity on 22 December looked for that signal across the available literature and did not find one [s1].

What was done

The researchers first searched MEDLINE, Embase, Cochrane, and PsycINFO for systematic reviews published between 1 January 2000 and 19 December 2024 [s1]. Finding that no existing review met their quality or scope criteria, they conducted their own review and meta-analysis of primary studies published from 1 January 2000 to 20 October 2025 [s1].

Eligible studies were randomised controlled trials, non-randomised intervention studies, and prospective observational studies examining the association between menopause hormone therapy — oestrogen-only, combined, testosterone, or tibolone — and incident mild cognitive impairment or dementia [s1]. Two reviewers independently screened and extracted, with risk of bias assessed using RoB 2 and ROBINS-E, and certainty of evidence rated with GRADE [s1]. Estimates were pooled in a random-effects model. The protocol was preregistered on PROSPERO (CRD42025639384) and the work was funded by the Public Health Agency of Canada [s1].

What was found

Of 5,914 records screened, ten studies met criteria — one randomised controlled trial and nine observational studies — with 1,016,055 participants in total [s1].

No significant association was found between hormone therapy use and risk of mild cognitive impairment or dementia [s1]. Subgroup analyses by timing of initiation, duration of use, and type of hormone therapy likewise showed no significant effects [s1].

Certainty of evidence ranged from moderate to very low [s1].

The ratio that should shape how you read this

Ten studies. One of them randomised [s1].

That ratio is the most important number in the paper. Observational studies of hormone therapy carry a well-known structural problem: women who take it have historically differed from women who do not — in health, in income, in engagement with healthcare — in ways that also predict dementia. Whatever the pooled estimate is, it is being driven overwhelmingly by designs that cannot fully separate the drug from the woman taking it.

A null result from such a body of evidence is therefore weaker than a null result from a body of trials. It is compatible with no effect. It is also compatible with a modest effect in either direction that this evidence base is not built to detect.

The GRADE ratings say as much. "Moderate to very low" certainty is not a description of a settled question; it is a description of a literature that has not yet been able to answer one [s1].

What the review explicitly did not cover

Two gaps stand out and the authors name them [s1]:

• No included study examined testosterone, despite testosterone being within the review's scope.
• No included study examined use in premature ovarian insufficiency — women whose ovaries stop functioning well before the typical age of menopause, for whom hormone therapy is a replacement of what should be there rather than a treatment of symptoms. This is arguably the group for whom the sex-steroid hypothesis of dementia risk is most relevant, and it has essentially no evidence.

The authors call for high-quality long-term studies addressing formulation, dose, route, timing, and duration, with attention to women with premature ovarian insufficiency, early menopause, or existing mild cognitive impairment [s1].

The interpretation the authors offer

Their conclusion is narrow and worth quoting in structure rather than in slogan: the review found no evidence that hormone therapy either increases or decreases dementia risk in post-menopausal women, which they say reinforces existing clinical guidance that hormone therapy should be prescribed on the basis of its other benefits and risks, and not for dementia prevention [s1].

That framing cuts against two different popular claims at once. It does not support taking hormone therapy to protect the brain. It also does not support avoiding it out of fear for the brain.

Why this arrives into a loud argument

Cognition has been one of the more emotionally charged threads in the long public dispute over menopausal hormone therapy — invoked both as a reason to prescribe and as a reason to refuse. What this review supplies is not a verdict but a current, preregistered, GRADE-rated account of how much the existing evidence can carry on this one endpoint. The answer is: not much, in either direction [s1].

That is an unsatisfying result to argue with, which is part of why it is useful. A null estimate with low certainty is not ammunition for anyone.

What to watch

Whether any randomised trial with dementia or MCI as a prespecified outcome is funded — the single thing that would move certainty above "moderate." Whether registry studies in premature ovarian insufficiency populations begin to accumulate. And whether guideline bodies revise their language on cognition in light of a null pooled estimate carrying low certainty, which is a genuinely awkward thing to write guidance around.

This article describes a research synthesis. It is not medical advice, and decisions about hormone therapy belong with a clinician who knows the individual case.

Sources

1. Menopause hormone therapy and risk of mild cognitive impairment or dementia: a systematic review and meta-analysis — The Lancet Healthy Longevity, 22 December 2025