For Twenty Years, Women Were Warned That Estrogen Raised Their Dementia Risk. Stanford's Autopsies Point the Other Way.

In the summer of 2002, the National Institutes of Health did something it almost never does: it stopped one of its own trials early and told the country the treatment was hurting people. The Women’s Health Initiative, the largest study of menopausal hormone therapy ever run, had found that the pills millions of women were taking raised their risk of breast cancer, stroke, and clots. The next year, a follow-up analysis reached the brain. In women who started the hormones at 65 or older, combined estrogen-plus-progestin roughly doubled the rate of dementia. Prescriptions collapsed. A generation of women learned to treat estrogen as something that ate the aging brain, and their doctors stopped offering it.

This week a team at Stanford Medicine published a study in the journal Neurology that reads like the mirror image of that warning. Drawing on records from 21,462 women, the researchers reported that women who had used estrogen-only hormone therapy had 39 percent lower odds of a dementia diagnosis during their lives. Then they did the thing almost no dementia study can: they looked inside the brains. Among nearly 3,000 women who came to autopsy, estrogen users had 35 percent lower odds of the physical hallmarks of Alzheimer’s: the amyloid plaques, the tau tangles, the neuritic deposits that define the disease under a microscope.

“For a long time, the going recommendation was ‘Don’t use MHT for memory decline,’” said senior author Hadi Hosseini. “This study flies in the face of that recommendation.” It is a striking dataset, and the look inside the brain is what makes it more than another database-mining exercise. Read against the trial it appears to overturn, though, it cannot carry the weight its headline invites.

Start with the shape of the evidence. The Women’s Health Initiative was a randomized controlled trial: women were assigned to hormones or a placebo by chance, the only design that lets you say the drug did the thing. The Stanford analysis was not. It is observational. It watches which women happened to take estrogen and asks how their brains turned out, and women who take estrogen-only therapy are not a random slice of the population. They are, typically, women who have had a hysterectomy, and they skew healthier and more plugged into medical care than women who never fill the prescription. Hosseini says so himself: it could be, he told Scientific American, that the estrogen users were “healthier overall” or “had better access to health care”, and something other than the drug did the protecting. The datasets did not even record why any given woman was on estrogen, or whether she’d had the hysterectomy that usually comes with the prescription.

So there are two studies pointing in opposite directions on the same molecule. One was randomized and found harm. One is observational and finds benefit. The randomized one is the stronger instrument. It is also not a clean rematch, because the harm that frightened everyone came mostly from older women on estrogen-plus-progestin, while the Stanford benefit is estrogen alone.

But the WHI has a problem of its own, and it is why this argument has run for twenty years without resolving. The women in that trial were old. The dementia arm enrolled women 65 and up, and the harm landed hardest in those who started hormones a decade or more past menopause. The estrogen-only arm, the apples-to-apples comparison for the new data, pointed toward roughly 50 percent higher dementia risk too, though that arm never reached statistical significance on its own. Out of that mess came the timing hypothesis: the idea that estrogen may protect the brain when it is started near menopause, in a woman’s late forties or early fifties, and may do nothing or do harm when it is started years later into a brain that has already begun to change.

The new study carries the same flaw, pointed the other way. Its participants used estrogen-only therapy later in life, well past the menopause window the timing hypothesis says matters, which is not how the therapy is prescribed today. Hosseini’s read is that late initiation would only have blunted the benefit, so the true effect might be larger, and the team notes the finding may understate the protection for that reason. Maybe. But it means the flashy result and the trial it seems to contradict share a defect: both studied women who took estrogen at the wrong age to answer the question everyone actually cares about, which is whether a fifty-year-old starting hormones at menopause is protecting her brain or endangering it. The most recent systematic review, published in Lancet Healthy Longevity, lands where the honest reading has to: the data are mixed, formulation and timing matter, and the blanket verdicts of the past two decades were built on thinner ground than they were sold with.

The institutional failure survives either result. The WHI’s blunt public message folded a synthetic progestin combination together with estrogen alone, said little about the age at which women started, and frightened a generation off a therapy that, for the right woman at the right time, the evidence never clearly condemned. Women were handed a warning with the confidence of settled science and the nuance of a fire alarm. The correction is not a press release built on an association, however large the number attached to it. The Stanford team adjusted for age, the APOE4 risk gene, blood pressure, education, and race. It could not adjust for the one thing that most separates estrogen users from everyone else, which is the kind of woman who ends up on estrogen in the first place. The work was paid for by the National Institutes of Health.

Jennifer Bruno, the lead author, calls the effect “modest but meaningful” and says the work “sets the stage for the future randomized controlled trials.” She is right about what it would take. Whether estrogen guards the aging brain will not be settled by another look back through autopsy records; it will take a trial that enrolls women at menopause, assigns the drug by chance, and follows them for decades. Which is to say the finding that looks like it refutes the Women’s Health Initiative can only be confirmed by building a better one. Twenty-three years after the NIH pulled the plug on the first, that trial still has not been run.

Sources

1. Neurology – Bruno et al., estrogen-only menopausal hormone therapy and Alzheimer’s pathology (2026), DOI 10.1212/WNL.0000000000218413 (peer-reviewed report; no open-access URL)
2. Stanford Medicine – Study ties estrogen-based menopausal hormone therapy to lower Alzheimer’s risk (2026)
3. ScienceDaily / American Academy of Neurology – summary of the Neurology study (2026)
4. Scientific American – Estrogen-only hormone therapy could lower risk of Alzheimer’s (independent coverage with caveats, 2026)
5. PubMed – The Women’s Health Initiative Memory Study: findings and implications for treatment (2005)
6. Lancet Healthy Longevity – Menopause hormone therapy and risk of dementia: systematic review and meta-analysis (2025)