Large UK Study Links HRT Use to Lower Dementia Risk in Postmenopausal Women
The largest observational study of hormone replacement therapy (HRT) and dementia risk to date has found that postmenopausal women who used HRT for at least one year were 10% less likely to develop any type of dementia and 16% less likely to be diagnosed with Alzheimer's disease compared with non-users. The findings, published in the journal Alzheimer's & Dementia on 26 August 2026, tracked 183,450 postmenopausal women from the UK Biobank over an average of 13 years, during which nearly 4,000 dementia cases were identified (The Guardian).
The study was led by Prof Anne-Marie Minihane, director of the Norwich Institute for Healthy Ageing at the University of East Anglia. Researchers examined whether HRT use of any kind for at least one year altered women's risk of developing dementia. The University of Exeter issued a press release on 26 August confirming the 16% Alzheimer's risk reduction among HRT users (University of Exeter).
The protective signal was not uniform across all subgroups. Women who had undergone surgical menopause — menopause triggered by the removal of the ovaries rather than occurring naturally — showed the greatest risk reduction, at 26% lower risk of any dementia compared with non-users. Women with naturally lower lifetime oestrogen exposure, defined by late menarche (the onset of menstruation) or early menopause, appeared to benefit more, with a 16% reduction in all-cause dementia risk. The associations were also stronger in carriers of the APOE4 gene variant, the most established genetic risk factor for late-onset Alzheimer's disease (The Guardian).
The Science Media Centre published an expert reaction on 26 August corroborating the headline figures: HRT was linked to approximately a 10% reduction in overall dementia rates (Science Media Centre).
These results sit within a body of evidence that has been inconsistent, at times contradictory. A Reuters report from June 2023 found HRT use was associated with a higher risk of dementia shortly after menopause, though causation remained unclear (Reuters). A 2023 systematic review and meta-analysis by Nerattini and colleagues found that randomised controlled trials (RCTs) — the gold-standard study type where participants are randomly assigned to treatment or placebo — in postmenopausal women aged 65 and older showed an increased risk of dementia with hormone therapy compared with placebo (PMC). More recently, Stanford Medicine reported in August 2026 that women on oestrogen-only menopausal hormone therapy regimens had reduced Alzheimer's risk (Stanford Medicine). The Alzheimer's Society also notes a 2021 study of nearly 400,000 women finding that both newer and older HRT formulations reduced the risk of diseases that cause dementia (Alzheimer's Society).
The preprint version of the Minihane-led study has been available on medRxiv since July 2025 under the identifier 10.1101/2025.07.22.25331871, reporting that HRT use for one year or more was associated with reduced dementia risk among postmenopausal women (medRxiv).
The study's observational design is its central methodological constraint. Because the data are drawn from the UK Biobank cohort rather than a randomised controlled trial, a causal link between HRT and reduced dementia risk cannot be proven. Confounding by indication — a statistical problem where the people who choose a treatment differ in important ways from those who do not — remains a concern: HRT users in the study had a higher average level of education than non-users, a difference the authors flagged as a potential limitation despite adjusting for socioeconomic factors in their analysis.
Dr Susan Kohlhaas, executive director of research and partnerships at Alzheimer's Research UK, stated that the study alone is not enough to tell whether HRT should be used to reduce dementia risk (The Guardian).
The broader context here is a research field that has struggled to reconcile observational signals of neuroprotection with RCT evidence that, at least in older women, tilts toward harm. The UK Biobank findings are notable for their sample size, duration of follow-up, and the granularity of subgroup analyses, particularly the surgical menopause and APOE4 carrier cohorts. The surgical menopause result, at 26% risk reduction, is especially striking: these women experience an abrupt loss of endogenous oestrogen, and the magnitude of the HRT effect in this group lends mechanistic plausibility to the oestrogen-depletion hypothesis of cognitive decline.
Yet the observational architecture limits what clinicians can take to the bedside. The education differential between users and non-users is not a trivial confounder; cognitive reserve — the brain's accumulated resilience to damage, closely linked to educational attainment — is itself a robust protective factor against dementia clinical presentation. Whether HRT users in this cohort were systematically healthier, more health-literate, or more proactive about preventive care cannot be fully resolved through covariate adjustment alone.
The tension with the Nerattini meta-analysis, which found increased dementia risk in RCT populations aged 65 and older, raises the critical timing question that has dominated this field for two decades: whether the neuroprotective window for oestrogen replacement is confined to the early postmenopausal period, and whether initiation after age 65 flips the risk-benefit calculus. The Minihane study did not stratify results by age at HRT initiation in the reported findings, which limits direct comparison with the RCT evidence base.
For clinicians and researchers, the study reinforces a consistent observational signal that has now been replicated across multiple large cohorts. It does not, by itself, resolve the causal question. What it does is sharpen the subgroup hypotheses worth testing in a prospective trial: surgical menopause patients, women with shortened reproductive windows, and APOE4 carriers. Whether any of those groups will be enrolled in the kind of adequately powered RCT needed to establish causation is an open question.
