August 26, 2026 · Science · Women's health · Hormone therapy

What the Study Reports

Postmenopausal women who used hormone replacement therapy (HRT) for at least a year had a lower rate of dementia diagnosis than women who did not. The cohort came from UK Biobank, with baseline assessment between 2006 and 2010 and follow-up through October 2022. It covered 183,450 postmenopausal women with no dementia diagnosis at baseline, mean age 60.3 years at entry.

Over 2,433,320 person-years of follow-up, 3,948 women were diagnosed with dementia. Of those, 1,993 were classified as Alzheimer's disease and 1,955 as non-Alzheimer's dementia. The analysis used Cox proportional hazards regression, a standard method for comparing event rates between groups over time.

HRT use was associated with a lower rate of all-cause dementia, at a hazard ratio of 0.90 with a 95 percent confidence interval of 0.84 to 0.96. Three subgroups showed stronger associations. Women who had surgical menopause came in at 0.76 (0.67 to 0.85). Women with lower cumulative lifetime exposure to their own estrogen came in at 0.78 (0.70 to 0.86). Carriers of at least one APOE e4 allele, the most common genetic risk factor for Alzheimer's disease, came in at 0.87 (0.80 to 0.95). The authors also report that starting HRT between ages 46 and 56 was associated with reduced risk.

The work comes from teams at the University of East Anglia and the University of Exeter. The figures above are taken from the medRxiv preprint, which publishes the full result set with confidence intervals. A peer-reviewed version has since appeared in the journal Alzheimer's & Dementia.

The short version

This is a large, well-powered observational cohort reporting an association, and the authors say plainly that causality is not established. The overall effect is modest. A hazard ratio of 0.90 on a cohort where about 2 percent of women were diagnosed with dementia over roughly thirteen years is a small shift in absolute terms. The more interesting part of the result is that the association was not uniform: it was roughly twice as large in women who had surgical menopause or low lifetime estrogen exposure. That pattern is what a real biological effect would look like, and it is also what confounding by indication would look like. Nobody should start or stop HRT because of this paper. Anyone already weighing HRT for menopausal symptoms now has one more input for a conversation with their clinician.

The Absolute Numbers Behind a 10 Percent Reduction

A relative risk reduction sounds larger than the change it describes. Across the whole cohort, 3,948 of 183,450 women were diagnosed with dementia, which is about 2.2 percent over a mean follow-up of roughly thirteen years. Applying a 10 percent relative reduction to a baseline of that size moves the figure by roughly two tenths of a percentage point. The published hazard ratio comes from an adjusted model rather than these crude counts, so the arithmetic illustrates the scale of the effect instead of estimating any individual's risk.

The subgroup figures follow the same rule. A hazard ratio of 0.76 in women with surgical menopause is the largest association in the paper, and it still describes a relative shift in a low-frequency outcome. Reporting that quotes only the percentage without the denominator gives a reader no way to judge the size of the decision in front of them.

What It Means for You

Hormone therapy is sold as a longevity product, and this is the kind of finding that reaches marketing copy fast. In the guides we publish here, the recurring problem with hormone marketing is that it borrows the strength of a clinical claim while dropping the conditions attached to it. Expect "reduces dementia risk" to appear in hormone marketing within the year. That includes menopause services such as Alloy and Midi Health, and broader hormone services such as Hone Health. The study does not support that sentence. It supports "was associated with a lower rate of dementia in one observational cohort, with the effect concentrated in specific groups."

The practical value here sits in the timing result rather than the headline number. The reduced risk appeared in women who started HRT between 46 and 56, which is the same window covered on our page about when to start HRT. That turns into a concrete question for an appointment. The question is not whether HRT prevents dementia. It is whether your own timing, menopause type, and symptom picture sit inside or outside the groups where the association was largest. Our guide to hormone therapy options for women covers the formulations and routes those conversations turn on, and what HRT costs covers the part telehealth pricing pages make hard to compare.

One caution applies to the APOE e4 result in particular. Consumer genetic tests report APOE status, and a carrier reading this paper may treat the 0.87 hazard ratio as a personal instruction. It is a subgroup association inside a single cohort, and it was not the study's primary comparison. Knowing APOE status changes what a clinician weighs; it does not by itself qualify anyone for a prescription.

Who Should Not Read This as a Reason to Start HRT

A personal history of breast cancer, a clotting disorder, or unexplained vaginal bleeding changes the calculation entirely. Anyone in that position should treat this finding as background reading. Those histories drive the prescribing decision whatever a cohort study reports about dementia. The same applies to women past their mid-sixties who have never used HRT, since the association in this paper attached to initiation between 46 and 56 and says nothing about starting later. Our page on HRT side effects and risks covers the specific contraindications, and they are the part of the conversation this study does not touch. Anyone currently taking HRT and considering stopping should not read a dementia association as a reason to continue through side effects they were already planning to raise with a prescriber.

What Would Change This Read

A randomized trial of HRT with dementia or cognitive decline as a pre-specified endpoint would change it, and the researchers name that trial as the next step. A replication in a cohort less affected by healthy-user bias would strengthen it. Movement in the other direction is possible too. The confidence interval reaches 0.96 at its upper end. A replication landing near there would describe an effect small enough to be hard to separate from residual confounding. Until one of those arrives, the finding sits alongside the rest of the observational HRT literature. The prescribing guidance covered on our menopause stages guide has not changed.

Related Coverage

This is the second Alzheimer's-related finding on this site in a week that turns on who a test or treatment was studied in. An earlier briefing covered the FDA clearance of the PrecivityAD2 blood test. That clearance rules out screening. It limits the test to adults who already show signs of cognitive impairment. The pattern is the same one a buyer keeps meeting: a result that is real inside its studied population, and marketing that quietly widens the population. For a reader tracking hormones through bloodwork rather than symptoms alone, our perimenopause biomarker panel guide sets out which markers are worth measuring during the window this study points at.

Sources

  • medRxiv preprint, "Hormone replacement therapy and dementia risk among postmenopausal women: evidence from the UK Biobank," posted July 23, 2025, DOI 10.1101/2025.07.22.25331871. Abstract and full result set retrieved via Europe PMC record PPR1054437, accessed August 26, 2026. Source of every hazard ratio, confidence interval and count on this page.
  • Medical Xpress, "HRT linked to lower dementia risk in large UK study," August 26, 2026, medicalxpress.com (accessed August 26, 2026). Source for the peer-reviewed publication in Alzheimer's & Dementia and the named investigators.

Frequently Asked Questions

Does this study show HRT prevents dementia?

No. It is a prospective observational cohort, so it reports an association between hormone replacement therapy (HRT) use and a lower rate of later dementia diagnosis. The authors state directly that further research is required to establish causality. Women who take HRT for a year or more differ from women who do not in ways a statistical model cannot fully remove. Access to care, general health at the point of prescribing, and how often someone sees a doctor all shape who ends up on a prescription. An association of this size is a reason to run a trial. It does not settle the question.

What is a hazard ratio of 0.90?

A hazard ratio compares the rate at which an event occurs in two groups over time. A hazard ratio of 0.90 for HRT users means their rate of dementia diagnosis was about 10 percent lower than the rate among non-users across the follow-up period. It is a relative measure, so it says nothing on its own about how common the event was. The reported 95 percent confidence interval runs from 0.84 to 0.96. That range covers everything from roughly a 16 percent lower rate down to roughly a 4 percent lower rate.

Which women showed the larger associations?

Three subgroups showed stronger associations than the full cohort. Women who had surgical menopause came in at a hazard ratio of 0.76, with a 95 percent confidence interval of 0.67 to 0.85. Women with lower cumulative lifetime exposure to their own estrogen came in at 0.78 (0.70 to 0.86). Carriers of at least one APOE e4 allele came in at 0.87 (0.80 to 0.95). The authors also report that starting HRT between ages 46 and 56 was associated with reduced dementia risk. These are subgroup analyses within one cohort and were not the primary comparison.

Does this change the standard advice about when to start HRT?

Not on its own. Prescribing decisions rest on symptoms, personal and family history, and the risks covered on our page about HRT side effects and risks, and no single observational cohort moves that. It does add weight to a timing question already present in the menopause literature. The reduced risk appeared in women who began HRT between 46 and 56, not at any age. Whether HRT suits a given person is a determination for a clinician who knows their history.

Where do the numbers on this page come from?

Every hazard ratio and count on this page is copied from the medRxiv preprint of the study, which reports the full results including confidence intervals. The peer-reviewed version has since been published in the journal Alzheimer's & Dementia, and figures in a peer-reviewed version can differ from a preprint. Where the two are compared here, the preprint is the source that was read.