A landmark drug-target Mendelian randomisation study published in npj Women’s Health has used genetic proxies to cut through decades of conflicting observational data on menopausal hormone therapy. The findings offer some reassurance on dementia risk – but sound a cautionary note about depression that clinicians and patients will need to weigh carefully.
For years, women approaching the menopause and their doctors have faced an uncomfortable paradox: the therapy most commonly used to ease menopausal symptoms sits at the centre of a scientific controversy that shows no sign of resolving itself. Does menopausal hormone therapy (MHT) protect the ageing brain, harm it, or do nothing at all? A new study led by researchers at King’s College London and the University of Oxford has taken an ingenious approach to cutting through the noise – and the results are both reassuring and thought-provoking in equal measure.
The research, published in npj Women’s Health, deployed a technique called drug-target Mendelian randomisation (MR) to sidestep the confounding and reverse causation that have long plagued observational studies of MHT. Rather than following women through their hormonal choices and health outcomes, the team used genetic variants as stand-ins for the drug itself – effectively asking what happens to the brain when the molecular targets of MHT are perturbed, independent of lifestyle, socioeconomic factors, or clinical indication.
How genetics can mimic a drug trial
MHT works primarily by binding to two oestrogen receptors, ERα and ERβ , encoded by the genes ESR1 and ESR2 respectively. The researchers identified naturally occurring single-nucleotide polymorphisms (SNPs) within these genes that influence downstream bio-markers known to reflect oestrogen receptor activity – bone mineral density and sex hormone-binding globulin (SHBG) levels for ERα, and haemoglobin levels for ERβ.
These SNPs were then used as instrumental variables in two-sample MR analyses, drawing on genome-wide association study (GWAS) data from tens of thousands of participants. Because genetic variants are assigned at conception and cannot be influenced by later environmental factors, this approach largely avoids the confounding that makes observational studies so difficult to interpret.
The outcomes of interest were chosen with care: Alzheimer’s disease (AD), cortical grey matter volume, hippocampal volume, white matter hyperintensity (WMH) volume, depression, and anxiety – all conditions with a marked female preponderance and all plausibly linked to oestrogen biology.
No smoking gun for Alzheimer’s
Perhaps the most clinically significant finding is what the study did not find. Genetically proxied perturbation of either ERα or ERβ showed no statistically significant association with Alzheimer’s disease risk or with any of the brain structural measures examined – cortical grey matter volume, hippocampal volume, or white matter hyperintensities.
This stands in notable contrast to the Women’s Health Initiative Memory Study (WHIMS), which reported a roughly doubled risk of dementia in postmenopausal women aged 65 and over who received oral conjugated equine oestrogen combined with progesterone. The authors acknowledge this discrepancy but offer a plausible explanation rooted in the so-called critical window hypothesis, which proposes that the neurological benefits of MHT depend heavily on when it is initiated relative to the onset of menopause.
As the authors write, “MR estimates can provide insights into lifelong effects of genetic variants, they do not equate to the impacts of pharmacological interventions initiated at specific times.” In other words, the genetic approach captures a kind of averaged, lifetime exposure to oestrogen receptor activity – it cannot tell us whether starting MHT at 51 versus 65 makes a meaningful difference to dementia risk. That question, the authors are careful to stress, remains open.
Although ERα perturbation was nominally associated with higher cortical grey matter volume and lower hippocampal volume, neither finding survived correction for multiple testing (false discovery rate adjustment), and the authors treat them accordingly – as hypothesis-generating signals rather than conclusions.
A depression signal that demands attention
Where the study does flag a significant concern is in the domain of depression. Genetically proxied ERβ perturbation – using the haemoglobin SNP rs1256061 in ESR2 as the instrumental variable – was significantly associated with a higher risk of depression, with a Wald β ratio of -0.656 (95% CI -0.992 to -0.319, FDR-corrected p = 0.002).
The directionality here requires a moment’s parsing: because lower haemoglobin is used as a proxy for higher oestrogenic activity at ERβ, the negative beta coefficient indicates that greater ERβ perturbation is associated with increased depression risk.
The authors are measured in their interpretation. They write: “Our study highlights psychiatric considerations when targeting oestrogen receptors with MHT, but provides no evidence for either harmful or protective effects on AD risk.” They also note that the biological plausibility is supported by ERβ ’s particularly high expression in the thalamus and hippocampus – brain regions closely implicated in mood disorders.
This finding sits in tension with a substantial body of clinical trial evidence suggesting that MHT is broadly neutral or even protective for depressive symptoms in menopausal women. The authors do not dismiss that literature but point out that their Mendelian randomisation approach is asking a subtly different question: not whether taking MHT improves mood in women who are already symptomatic, but whether oestrogen receptor perturbation itself has a causal relationship with depression risk across a population.
Caveats the authors are candid about
To their credit, the researchers are unusually forthright about the limitations of their ERβ findings. The use of haemoglobin as a proxy for ERβ activity is less robustly validated than bone mineral density is for ERα. There are also plausible biological pathways from low haemoglobin to fatigue and depressive symptoms that could violate the exclusion restriction – a core assumption of Mendelian randomisation requiring that the instrumental variable affects the outcome only through the exposure of interest.
A follow-up standard MR analysis of haemoglobin on depression showed an association by inverse-variance weighted method, though not across all sensitivity analyses, suggesting partial violation of this assumption is possible. The authors conclude that “further research with alternative ERβ proxies is needed to confirm this finding.”
There are further structural limitations. The analyses were restricted to European-ancestry samples, limiting generalisability. Most GWAS summary statistics were derived from mixed-sex cohorts rather than female-only samples, which may dilute the precision of female-specific estimates. And, by design, the MR framework assumes constant genetic exposure effects over a lifetime – a simplification that may not reflect the reality of hormone biology, which shifts dramatically across reproductive stages.
What this means for clinical practice
The study will not – and should not – immediately change prescribing practice. What it does is add a methodologically distinct layer of evidence to an already complex picture. Taken together with the existing RCT and observational literature, the most defensible reading is that MHT, when used appropriately and initiated at the right time, is unlikely to substantially increase Alzheimer’s risk – a message many clinicians will find broadly consistent with their clinical intuition.
The depression signal, however, warrants closer attention. As the authors put it, “following future research confirming a differential role of ERα and ERβ in depression, oestrogen receptor modulators that selectively act on ERα could be explored to avoid inducing depressive symptoms with hormone therapy use.” That is a genuinely interesting therapeutic hypothesis – one that points towards more targeted, receptor-selective approaches to hormonal treatment in the future.
For now, the study reinforces a familiar but important message: the effects of MHT on the brain are real, biologically plausible, and still incompletely understood. Genetic epidemiology is proving a powerful tool for untangling causality in this field – but it works best as a complement to, not a replacement for, well-designed randomised trials and carefully conducted longitudinal studies.
Reference:
Schindler, L. S., Gill, D., Oppenheimer, H., et al. (2026). Menopausal hormone therapy and risk of neuropsychiatric disease: a drug target Mendelian randomisation study. npj Women’s Health, 4, 10. https://doi.org/10.1038/s44294-026-00130-1
