News · Sleep
Broken sleep in perimenopause tracked worse thinking decades on
A study of 2,097 postmenopausal women in Menopause found the worst perimenopausal sleep went with lower cognitive scores years later, but only in women woken by symptoms.
Based on a peer-reviewed cohort analysis in Menopause
- Researchers analyzed 2,097 postmenopausal women without known dementia in the St Louis Baby Tooth-Later Life Health Study, publishing in Menopause.
- Women recalled their perimenopausal sleep at a single later visit and completed the TestMyBrain cognitive battery on the same day.
- Moderate to severe perimenopausal sleep disturbance went with lower cognitive scores (-0.15 SD; 95% CI, -0.29 to -0.02); mild disturbance did not reach significance (-0.07; 95% CI, -0.17 to 0.02).
- Splitting by whether menopausal symptoms woke them changed everything: no association in women never or rarely woken.
- Among women woken sometimes to always, the gap widened (mild: -0.17; 95% CI, -0.28 to -0.06; moderate to severe: -0.31; 95% CI, -0.47 to -0.16).
- Results held after adjusting for hormone therapy and for sleep problems measured at the time of testing.
- Sleep was recalled decades after the fact, at the same visit as the cognitive test. That is the study's central weakness and the authors call for prospective confirmation.
- Effect sizes are small. A third of a standard deviation is a real group difference, not a diagnosis.
Most women going through perimenopause are told the sleep problems will pass. A study in Menopause asked whether anything is left behind, and found a signal in one specific group.
The researchers set out to evaluate whether self-reported sleep disturbances during perimenopause are associated with cognitive function decades later in postmenopausal women.
The cohort
The sample came from an unusual place. The team analyzed 2,097 postmenopausal women without known dementia in the St Louis Baby Tooth-Later Life Health Study, recruited between 2019 and 2023.
Excluding women with diagnosed dementia matters. It means any difference found is happening in the range of normal thinking, not being driven by a handful of people already ill.
At a single postmenopausal visit, participants retrospectively reported perimenopausal sleep, rated current sleep, and completed a cognitive battery. Everything was captured in one sitting, which is the design’s efficiency and also its problem.
What the overall comparison showed
Women were sorted into three groups by how bad their perimenopausal sleep had been: none to slight, mild, or moderate to severe.
Perimenopausal sleep disturbances were associated with lower cognitive scores compared with none to slight. The worst group sat about 0.15 standard deviations lower. The mild group sat lower too, by roughly half that, but with enough uncertainty that the difference could have been nothing.
Read alone, that is a modest and slightly wobbly finding. The paper’s real content is what happened when the researchers split the sample.
The split that mattered
They divided women by how often menopausal symptoms actually woke them: never or rarely, versus sometimes to always.
No association was observed among women never or rarely woken by menopausal symptoms. Flat. Nothing.
Among the women whose nights were being interrupted by symptoms, the gaps roughly doubled: larger deficits were seen among those woken sometimes to always, reaching about 0.31 standard deviations in the worst-sleep group and about 0.17 even in the mild group.
That pattern is more informative than a single overall number. It says the association is not with feeling tired in midlife generally. It is with nights repeatedly broken by hot flashes and night sweats.
The checks that held
Two obvious objections were tested.
The first is hormone therapy, which affects both symptoms and, in some studies, cognition. The second is current sleep, since a woman sleeping badly today might do worse on a test today regardless of what happened 20 years ago.
Findings were unchanged after adjustment for hormone therapy or for concurrent sleep disturbances measured at the time of cognitive testing. Neither explained the result away.
Why this is not proof
The design puts a real ceiling on how far this goes.
Sleep during perimenopause was recalled at the same appointment as the cognitive test, sometimes decades after the event. Memory for how you slept in 1998 is not a precise instrument, and it may be systematically less precise in women who score lower on a memory test now. That single feature could generate the entire association.
The direction is also unsettled. Brain changes can disturb sleep long before they show up on testing, so poor perimenopausal sleep might be an early symptom rather than a cause. From one visit, those two stories look identical.
And the effects are small. A third of a standard deviation describes two heavily overlapping distributions, not two different kinds of people.
What the authors actually claim
They are careful about it. These results suggest that sleep and menopausal symptoms in midlife may be relevant markers of later-life cognitive health, and warrant confirmation in prospective studies.
Markers, not causes. Warrant confirmation, not established. The useful version of this finding is narrow: being woken repeatedly by menopausal symptoms may flag something worth tracking, and it is worth treating regardless, because nobody should have to put up with it for the four to eight years perimenopause tends to run.
People also ask
What is perimenopause?
The transition leading up to menopause, typically starting in the mid-40s and lasting four to eight years. Hormones fluctuate rather than simply decline, which is why symptoms during this window can be more disruptive than after menopause itself. Sleep is one of the first things to go, through night sweats, more frequent waking, and difficulty getting back to sleep.
How big is a difference of 0.31 standard deviations?
Small at an individual level, meaningful at a population level. It is roughly the difference you would see between two groups whose scores overlap heavily; plenty of women in the poor-sleep group scored above the average of the good-sleep group. It is not a level of difference anyone would notice in a person. It is the kind that shows up when you average thousands.
Why does being woken by symptoms matter so much?
It is the most interesting result in the paper. Among women who reported bad sleep but were not being woken by menopausal symptoms, there was no association with later cognition at all. The signal lived entirely in the group whose sleep was being broken by hot flashes and night sweats. That points at symptom-driven fragmented sleep specifically, rather than at poor sleep in general, though a single study cannot settle why.
Does this mean bad sleep in midlife causes cognitive decline?
No, and the design cannot show that. Sleep was reported from memory decades later, at the same appointment where cognition was tested, so women with lower scores may simply recall their past sleep differently. Early brain changes can also disturb sleep years before they affect test scores, which would produce this exact pattern with the causal arrow reversed.
Is it worth treating menopausal sleep problems anyway?
This is general information rather than medical advice, but the case for treating disruptive night sweats and insomnia stands on its own merits, independent of anything about long-term cognition. Effective options exist, including hormonal and non-hormonal approaches, and a clinician can weigh them against individual risk. This study does not add a new reason so much as remove an excuse to leave it untreated.