Earlier menopause, less gray matter: a Cambridge cohort finds both, and can't say why
747 postmenopausal women were tested; 182 were scanned. Gray matter volume partially mediated the link between menopause age and cognition — in a cross-sectional design that cannot establish direction.
Menopause is filed under reproductive medicine, which is a filing decision rather than a biological one. Estrogen receptors are distributed through the brain, and the question of what the transition does to neural tissue has been asked mostly in small samples, usually with either cognitive testing or imaging but not both.
A study published in Menopause on March 10 has both, in the same women.
What was measured
Researchers analysed data from the Cambridge Centre of Neuroscience and Aging, including 747 postmenopausal women who underwent cognitive testing [s1]. A subset was additionally given a fluid intelligence test, and a subset of 182 had structural brain scans [s1].
Multiple linear regression models tested the association between age at menopause, cognitive performance and gray matter volume, controlling for chronological age, education, depressive symptoms and physical activity [s1]. Controlling for chronological age is the crucial adjustment here: without it, any association between earlier menopause and worse later-life cognition could simply reflect how old someone was when tested.
What was found
Earlier menopause was associated with lower cognitive performance (P = 0.028) and with lower fluid intelligence (P = 0.026) [s1].
In the imaging subset of 182 women, earlier menopause was associated with decreased total gray matter volume (P = 0.024) [s1].
An exploratory mediation analysis found that total gray matter volume partially mediated the relationship between age at menopause and cognitive performance (P = 0.006) [s1]. The authors describe this as the first test of that mediation within a single sample [s1].
The design constraint
The study is cross-sectional, and the authors state the limitation directly: the design prevents causal conclusions, and longitudinal research is needed to establish causal links [s1].
That is not boilerplate in this case. At least three explanations fit these data equally well.
Earlier loss of ovarian hormones could accelerate brain aging — the interpretation the framing invites. Alternatively, whatever processes determine when ovaries stop could be the same processes that determine how brains age, with neither causing the other. Or the cohort could be affected by factors upstream of both: early menopause is associated with smoking, socioeconomic position and chronic illness, and while the models adjust for education, depressive symptoms and physical activity [s1], adjustment is not the same as control.
Reading the numbers honestly
The P values cluster just under 0.05. The imaging finding rests on 182 women, and the fluid intelligence result on a smaller subset than the main cognitive analysis [s1]. The mediation analysis is labelled exploratory by the authors [s1].
None of that makes the finding wrong. It makes it preliminary, which is the correct word for a single cross-sectional cohort producing effects at the edge of conventional significance.
What the study does not report is effect magnitude in terms a reader can attach meaning to — how much gray matter, how many years of menopause timing, how much difference on a test. Statistical significance and practical significance are different questions, and this analysis answers only the first.
What "partially mediated" means
Mediation analysis asks whether the relationship between two variables runs through a third. Here the claim is that part of the association between menopause age and cognitive performance is accounted for by gray matter volume [s1] — that is, women with earlier menopause had less gray matter, and less gray matter went with lower scores.
"Partially" is doing real work in that sentence. It means gray matter volume did not account for the whole association, so something else is carrying the remainder. It does not identify what.
Mediation also inherits every assumption of the underlying design. In cross-sectional data, the statistical model imposes an ordering — menopause age, then brain volume, then cognition — that the data themselves cannot verify. The arrow is a modelling choice, and the authors label the analysis exploratory [s1].
Why it is worth reporting anyway
Because the alternative framing — menopause as a set of symptoms to be managed until they stop — has shaped both research funding and clinical attention for decades, and evidence that measurable structural change accompanies the transition argues for treating it as a whole-body event with a neurological component.
That argument does not require this study to be definitive. It requires the question to be asked with larger samples, longitudinal designs, and imaging before and after rather than decades later. This paper is a reason to run those studies, not a substitute for them.
This article is informational and is not medical advice.
Sources
Sources
More on
Social support and memory in postmenopausal women: a careful null
A secondary analysis of the ELITE hormone-therapy trial found hints that social ties tracked with cognition, but none survived correction for multiple testing.
Eighteen years of follow-up tie menopause age to brain change. The effects are tiny.
A cohort with serial MRI and autopsy data found earlier menopause tracked faster cognitive decline and white matter damage. The per-year coefficients are very small.
A million women's worth of data on hormone therapy and dementia returns a flat line
Pooling ten studies, researchers found menopause hormone therapy neither raised nor lowered dementia risk. The certainty of that evidence ranged from moderate to very low.
Vaginal oestrogen around prolapse surgery: better symptoms, same anatomy
A 293-woman Dutch trial found more women reported improvement at 12 months with perioperative oestriol cream than placebo. Anatomical success, surgical success and reintervention were unchanged.