HRT and dementia risk: why the warning changed, and what it actually means

Short answer: For twenty years, hormone therapy carried a warning that it could cause dementia. In November 2025 the FDA removed that warning — and the removal was justified, because the warning rested on one trial, using one formulation, in women who were already in their late sixties. But the message that replaced it has swung too far the other way. The strongest evidence available, a review of more than a million women commissioned by the World Health Organization, found that hormone therapy neither raises nor lowers dementia risk. The one randomised trial that properly tested "start it early and protect your brain" followed women for a decade and found no cognitive benefit — and no harm. So: the old warning was overstated, the new promise is overstated too, and the sensible conclusion is that dementia shouldn't be part of your decision about hormone therapy in either direction. Decide it on symptoms, on bone health, and on your own medical history, with your doctor.

This is educational, not medical advice, and it is not a reason to start or stop any medication. Hormone therapy decisions belong with your doctor, who knows your symptoms, your history and your other risks. If you're currently taking HRT, nothing here is a reason to change that without a conversation.

What actually changed in November 2025

On 10 November 2025, the FDA announced it was removing the "black box" warning — the agency's most prominent safety label — from menopausal hormone therapy products. Specifically, it asked manufacturers to remove references to cardiovascular disease, breast cancer and probable dementia. The boxed warning about endometrial cancer for estrogen-alone products stayed.

The new labelling recommends starting systemic hormone therapy within 10 years of menopause onset, or before age 60.

(FDA news release, 10 November 2025)

The agency's reasoning was about who was studied. The warning came from the Women's Health Initiative, whose participants averaged 63 years old — more than a decade past the typical age of menopause — and who took a formulation no longer in common use. Applying that to a 52-year-old with hot flushes was, the FDA argued, a distortion of risk. That argument is a fair one, and to see why, it helps to know exactly what the original trial found.

Where the dementia warning came from

The Women's Health Initiative Memory Study (WHIMS) enrolled women aged 65 and older and randomised them to either conjugated equine estrogens plus medroxyprogesterone acetate, or placebo, for about four years.

The result, published in JAMA in 2003, was stark: a hazard ratio of 2.05 (95% CI 1.21–3.48) for probable dementia — roughly double.

But look at the absolute numbers, because they tell a different story than "double" does. Dementia occurred at 45 cases per 10,000 women per year on hormone therapy versus 22 per 10,000 on placebo. That is an additional 23 cases per 10,000 women each year — about 1 in 435.

(Shumaker SA et al., JAMA 2003)

A companion trial testing estrogen alone in women who'd had a hysterectomy found a hazard ratio of 1.49 (95% CI 0.83–2.66) — pointing the same direction, but not statistically significant; the confidence interval comfortably includes "no effect."

(Shumaker SA et al., JAMA 2004)

So the warning was built on: one drug combination, given to women who were on average 15 years past menopause, for four years. It was real evidence about that situation. It was never evidence about a 50-year-old starting estradiol for night sweats — and for two decades it was read as though it were.

The claim that replaced it — and why it needs the same scrutiny

Here's where it's worth being careful, because the correction has picked up momentum of its own.

The HHS fact sheet published alongside the FDA announcement states that hormone therapy "has also been associated with 50% reduction in heart attack risk, 64% reduction in cognitive decline, and 35% lower risk of Alzheimer's."

Two things are worth noticing about that sentence. First, on the fact sheet itself, the claims about all-cause mortality and fractures carry links out to supporting sources — and the sentence containing the cognitive and Alzheimer's figures does not.

(HHS fact sheet, 10 November 2025)

Second, the closest published figure we could find is a 32% reduction — but it comes from pooling observational studies of midlife estrogen-only therapy (RR 0.685, 95% CI 0.513–0.915), and observational studies of hormone therapy have a well-known problem, which we'll come to in a moment.

That same 2023 analysis lays the whole conflict out in one place. Pooling six randomised trials in women 65 and older, hormone therapy was associated with increased dementia risk (RR 1.38, 95% CI 1.16–1.64) — driven by estrogen-plus-progestogen (RR 1.64), with estrogen-only not significant (RR 1.19, 0.92–1.54). Pooling 45 observational studies covering 768,866 cases and 5.5 million controls, hormone therapy was associated with reduced Alzheimer's risk (RR 0.78, 0.64–0.95).

Same treatment. Opposite answers. The difference is the study design.

(Nerattini M et al., Frontiers in Aging Neuroscience, 2023)

The reason to distrust the optimistic number is called healthy-user bias. In the decades those observational studies cover, the women who took hormone therapy were systematically different from those who didn't: more likely to be affluent, better educated, more engaged with healthcare, less likely to smoke, more likely to exercise. Every one of those things independently lowers dementia risk. When you compare hormone users to non-users in the real world, you are partly comparing two different kinds of life. Randomised trials exist precisely to break that link — and the trials don't show the benefit.

The best current answer: no effect either way

In December 2025 — six weeks after the FDA announcement — a systematic review commissioned by the World Health Organization pooled ten studies covering 1,016,055 participants and asked the question directly.

The finding: no significant association between menopause hormone therapy and mild cognitive impairment or dementia. Analyses split by timing, duration and type of hormone therapy also showed no significant effects.

The authors' conclusion is worth quoting nearly in full: "This review found no evidence that MHT use either increases or decreases the risk of dementia in post-menopausal women. This reinforces current clinical guidance, that MHT prescription should be based on other perceived benefits and risks and not for dementia prevention."

(Melville M et al., The Lancet Healthy Longevity, December 2025)

The researchers rated their own certainty honestly: moderate to very low, and only one of the ten studies was a randomised trial. That mixture is part of the finding. This is a large literature that is not, on the whole, a strong one.

On the strength of that review and others like it, the WHO's 2026 dementia guidelines advise against using hormone therapy as a way to reduce dementia risk.

The one trial that tested "start early"

If the timing hypothesis is right — that estrogen protects the brain if you start near menopause, and only harms if you start late — then there should be a trial showing it. There is one, and it's the most useful study in this entire article.

The Kronos Early Estrogen Prevention Study randomised women within three years of their final period (average age 52.6) to oral estrogen, transdermal estradiol, or placebo, for four years. Then the KEEPS Continuation study brought 275 of them back roughly a decade later and re-ran the same cognitive tests.

The result: "mHT poses no long-term cognitive harm; conversely, it provides no cognitive benefit or protective effects against cognitive decline."

(Gleason CE et al., PLOS Medicine, November 2024)

That is the cleanest test available of exactly the claim now being made, and it came back neutral in both directions. Reassuring if you take HRT. Not a reason to start it for your brain.

What about the new brain-autopsy study?

In August 2026, Stanford researchers published a study that got a lot of attention, and it deserves a careful read rather than a headline.

Using two large research databases, they compared women who reported using estrogen-only hormone therapy against non-users, and found lower odds of Alzheimer's changes in donated brain tissue at autopsy (OR 0.65, 95% CI 0.48–0.88), lower odds of a clinical dementia diagnosis (OR 0.61, 0.55–0.67), and lower amyloid levels in blood and spinal fluid.

Three things the coverage mostly left out:

• The headline number is not the study's number. You'll see "a study of 21,000 women's brains." The 21,462 figure is the combined size of the two databases. The autopsy comparison — the study's primary outcome — rests on 258 hormone therapy users against 2,701 non-users.
• It applies to estrogen-only therapy. The sample of women on combined estrogen-plus-progestogen was too small to draw conclusions from, and most women with a uterus need the combined form.
• It's observational, and the authors say so. In their own words, the associations are "small but significant", they "do not address causality", and the study has "limited generalizability due to the retrospective nature". Which means healthy-user bias applies here exactly as it does to the other observational research.

(Bruno J et al., Neurology, 12 August 2026)

It's a genuinely interesting study with a hard endpoint, and it may point somewhere. It is not a reason to take a medication.

The gap nobody is talking about

Buried in the WHO-commissioned review is a sentence that matters enormously for a specific group of readers: no included study examined hormone therapy use in women with premature ovarian insufficiency.

That is a real problem. Women who go through menopause very early — before 40, or after surgery to remove the ovaries — are the group most likely to be prescribed hormone therapy, often for years, and professional guidance generally supports it for them until around the average age of natural menopause, for bone and cardiovascular reasons. And they are precisely the group the dementia evidence skips.

If that's you, the honest position is: the standard reasons for taking hormone therapy still apply, the dementia question is unanswered for your situation specifically, and neither of those facts should change what you do. We've written about that group in more detail in hysterectomy, ovary removal and dementia risk.

So how should you actually decide?

Keep three questions separate, because almost everything confusing about this topic comes from mixing them.

1. Hormone therapy for menopausal symptoms. This is what it's for. Hot flushes, night sweats, disrupted sleep, genitourinary symptoms. The FDA's label change makes the risk picture more accurate for women starting near menopause. That's a real conversation to have with your doctor, and dementia isn't part of it.
2. Hormone therapy after premature or early menopause. A separate, standard indication, driven by bone and cardiovascular health during years you'd otherwise have had your own oestrogen. Also a doctor conversation. Also not about dementia.
3. Hormone therapy to protect your brain. Not supported. WHO advises against it for this purpose, the biggest review found no effect either way, and the one trial designed to test early initiation found nothing after ten years. If someone is selling HRT to you on this basis, that's a reason for scepticism — the same scepticism we'd apply to brain supplements.

And if what's actually driving the question is memory changes around menopause, that's usually a different issue with a much more reassuring answer: menopause brain fog vs. dementia.

Where the real levers are

None of this means dementia risk is out of your hands — the opposite. The 2024 Lancet Commission links around 45% of dementia cases worldwide to 14 modifiable risk factors. Hormone therapy isn't one of them. Blood pressure, blood sugar, cholesterol, hearing, physical activity, sleep, smoking and social connection are.

That's not a consolation prize. It's the actual list, and it's the one the evidence supports.

More context: what the WHO now says about lowering dementia risk, can you reduce your dementia risk? and check your dementia risk.

How we research and review our content →

This article is educational and is not medical advice, diagnosis, or treatment. The studies cited describe findings in general populations, not any individual. Solenna is not a medical device and does not diagnose, prevent, treat, or cure Alzheimer's disease or any other condition. Risk reduction means lowering probability, not eliminating it; individual results vary and no outcome is guaranteed. Decisions about starting, continuing or stopping hormone therapy should be made with a qualified clinician. Do not change any prescribed treatment on the basis of this article.