Eighteen years of follow-up tie menopause age to brain change. The effects are tiny.

In March, a Cambridge cohort reported that earlier menopause was associated with lower cognitive performance and lower total gray matter volume, in a cross-sectional sample of 747 postmenopausal women with imaging in 182 of them [s2]. The obvious objection was that a snapshot cannot show direction, and the authors said so themselves [s2].

A study published in JAMA Network Open on August 25 addresses that objection with longitudinal data, serial imaging and autopsy [s1]. It also demonstrates why answering the design question does not automatically produce a large answer.

The cohort

The analysis used the Religious Orders Study and the Rush Memory and Aging Project, with up to 18 years of follow-up [s1]. Religious Orders Study enrollment began in January 1994; the Rush Memory and Aging Project enrolled from September 1997 to April 2005; follow-up in both is ongoing [s1]. Data were extracted and analysed in August 2025 [s1].

Three nested samples were used: 2,603 women with cognitive data, 1,287 with neuropathology at autopsy, and 774 with serial 3T magnetic resonance imaging [s1].

Mean age at enrollment was 78.30 years (SD 7.85). The cohort was 92.2% White non-Hispanic (2,400 women) and 6.6% Black or African American (172) [s1] — a composition that limits how far the findings generalise.

Age at menopause was examined overall and stratified by type, spontaneous versus surgical [s1]. Analyses used linear mixed-effects, accelerated failure time, proportional odds and inverse-probability-weighted generalized additive mixed models, with a prespecified three-family Benjamini-Hochberg correction for multiple comparisons [s1].

What was found

Earlier menopause was associated with faster global cognitive decline (β = −0.0009 SD per year of earlier menopause age; SE 0.0004; FDR-corrected P = .04) and faster episodic memory decline (β = −0.0014; SE 0.0005; FDR-corrected P = .03) [s1].

It was also associated with earlier Alzheimer disease diagnosis (time ratio 0.998; 95% CI 0.997–0.999; FDR-corrected P = .02), with associations described as directionally stronger in surgical menopause (time ratio 1.000; 95% CI 0.998–1.000; P = .008) [s1].

The largest effect was in white matter. Earlier menopause was associated with substantially faster accumulation of white matter hyperintensity volume in women with spontaneous menopause (f² = 0.30; FDR-corrected P < .001) — an association that increased with advancing age and was absent in surgical menopause [s1]. The authors read that contrast as indicating specificity to the gradual hormonal trajectory of spontaneous menopause [s1].

The size of the effect

The cognitive coefficients are per single year of menopause age. A decline of 0.0009 standard deviations per year of earlier menopause means the difference between menopause at 45 and menopause at 55 corresponds to roughly 0.009 SD per year of subsequent cognitive decline. That is a real signal in a cohort of 2,603 women followed for up to 18 years, and it is not something visible in any individual.

The Alzheimer diagnosis result is similar. A time ratio of 0.998 with a confidence interval of 0.997 to 0.999 excludes the null while describing a very small shift in time to diagnosis.

The white matter finding is the one with substantial magnitude, and it is also the one confined to a subgroup defined after the fact.

Where it is stronger than what came before

Three things. The follow-up is longitudinal with serial imaging rather than a single scan, which means the outcome is a trajectory rather than a level [s1]. Neuropathology at autopsy in 1,287 women gives a hard endpoint that no imaging study can substitute for [s1]. And the stratification by menopause type separates two exposures — gradual hormonal decline and surgical removal — that most studies pool.

The spontaneous-versus-surgical contrast in the white matter result is the most mechanistically interesting output of the study, precisely because it runs against the simplest hypothesis. If less lifetime estrogen exposure were the whole story, surgical menopause should show the effect most strongly. It did not [s1].

Where it is limited

The cohort is old at enrollment, overwhelmingly White, and drawn from religious orders and a Chicago-area volunteer study — populations selected in ways that differ from the general population. Age at menopause in women enrolled in their late seventies is recalled, not recorded.

And the exposure is not randomisable. Whatever determines when menopause occurs may also, directly, determine how brains age, without menopause timing being on the causal path at all.

What it changes

The authors position menopause age as a midlife-identifiable marker for risk stratification, offering a window for prevention before structural or cognitive change becomes apparent [s1].

As a marker, that is defensible: age at menopause is already in most medical histories and costs nothing to record. As a target, nothing here supports it. This study does not test any intervention, and it does not address whether hormone therapy alters any of these trajectories.

This article is informational and is not medical advice.

Sources

• Age at Menopause and Brain Atrophy Among Older Women — JAMA Network Open, 2026-08-25
• When the clock shifts: menopause timing is associated with reduced cognitive performance and gray matter volume in a population-based cohort — Menopause, 2026-03-10