Menopause Blood Proteins Tracked Alzheimer's Risk, Not All Dementia

It is two in the morning in a house where everyone else is asleep, and a woman of forty-nine has pushed the duvet onto the floor for the third time this week. The night sweats themselves are not what frightens her. What lands the next morning, in a headline read standing up at the kitchen counter, is a different proposition altogether: that menopause leaves a signature in the blood linked to brain aging and Alzheimer's risk. Read that way, a broken night stops being an inconvenience and starts looking like evidence of something.

The paper underneath that headline, published in Nature Medicine on 22 September 2026, is both more interesting and considerably more modest than the framing around it. It is a biomarker-discovery study. It asks whether the hormonal transition of menopause leaves measurable traces in circulating proteins, and whether those traces carry information about brain aging decades later. It does not report a test, a threshold, or a number any woman could look up. And on the specific question of whether hot flashes mark a brain under strain, its midlife measurements point the other way.

The work runs in three stages, across three different kinds of sample. The discovery stage used the UCSB Healthy Aging Study: 80 women aged 43 to 58, staged as pre-, peri- or postmenopausal by STRAW+10 criteria in roughly equal numbers, plus 36 age-matched men. Their serum was run on a NULISAseq CNS panel of 118 proteins chosen for relevance to neurologic aging. Recruitment was deliberately clean of exogenous hormones: participants were not taking menopausal hormone therapy, were required to have an intact uterus and no history of bilateral oophorectomy, and just one was on an oral contraceptive.

The validation stage repeated the analysis in 2,814 age-matched pre-, peri- and postmenopausal women whose plasma had been run on a different platform, Olink. Agreement across two dissimilar assays matters here, because a protein signature that appears on only one platform is as likely to describe the assay as the biology.

The third stage asked whether any of it mattered later. Four independent cohorts of older women, with mean baseline ages from 60.7 to 72.1 years and 11,925 participants in total, were scored: the Alzheimer's Disease Neuroimaging Initiative (673 women, 54.3% with mild cognitive impairment or dementia, 49.0% amyloid-positive) and the UCSF Brain Aging Network for Cognitive Health (100 women) followed longitudinally over roughly five years; the Wisconsin Registry for Alzheimer's Prevention (93 women, 64.5% amyloid-positive) cross-sectionally; and 11,059 women in the UK Biobank for incident dementia. The authors also ran the analysis again under narrower conditions - restricting the discovery sample to ages 47 to 53, adjusting for cardiometabolic conditions, excluding women with bilateral oophorectomy, stratifying by hormone therapy history, removing p-tau231 from the score, and excluding anyone with cognitive impairment at baseline.

Sixteen proteins were elevated in postmenopausal women relative to premenopausal women, spanning inflammatory biology (CCL13, IL-12p70, CXCL1, CCL26, CCL2, CCL11, TEK, CD63) and synaptic and neuronal signalling (CNTN2 and ACHE among them). Combined into a single menopause proteomic score, they tracked hormones more closely than they tracked age: the score moved inversely with estradiol (beta = -0.89, 95% CI -1.37 to -0.41, P = 0.0004) and positively with follicle-stimulating hormone (beta = 1.29, 95% CI 0.78 to 1.80, P = 0.000003).

One result in that same analysis complicates the label on the box. After adjusting for age, men scored higher on the menopause proteomic signature than women did overall (beta = 1.13, 95% CI 0.27 to 1.90, P = 0.009). Whatever these sixteen proteins capture about inflammatory and metabolic aging, it is plainly not unique to menopause, and the authors report it plainly.

In later life the score did carry information. Higher baseline scores tracked steeper cognitive decline over about five years in the Alzheimer's Disease Neuroimaging Initiative (beta = -0.30, 95% CI -0.42 to -0.18, P = 0.000003) and, more faintly, in the San Francisco cohort (beta = -0.03, 95% CI -0.05 to -0.002, P = 0.03), and lower global cognitive performance in the Wisconsin registry (beta = -0.27, 95% CI -0.51 to -0.04, P = 0.02).

The dementia results are where the headline and the data part company. Over a mean of 15.7 years (SD 2.84) in the UK Biobank, higher scores associated with Alzheimer's dementia at a hazard ratio of 1.15 (95% CI 1.01 to 1.31, P = 0.04) across 219 cases - an interval whose lower bound rests at 1.01. There was no clear association with all-cause dementia (HR 1.08, 95% CI 0.98 to 1.19, P = 0.13, 387 cases), vascular dementia (HR 1.15, 95% CI 0.91 to 1.47, 65 cases) or frontotemporal dementia (HR 0.90, 95% CI 0.59 to 1.36, 23 cases). Carrying APOE4 made little difference either, except for cognitive change in one cohort.

Because symptoms are what a woman can actually observe, this is the part of the paper most likely to be read as personal. In the midlife sample, where symptoms and proteins were measured at the same time, hot flashes showed no association with the menopause score whatsoever: beta = 0.43, 95% CI -0.55 to 1.42, P = 0.38. Irritability showed none either (beta = 0.32, 95% CI -0.74 to 1.38). Night sweats sat precisely on the boundary of significance, with a confidence interval starting at zero (beta = 0.94, 95% CI 0 to 1.88, P = 0.05). Vaginal dryness ran the opposite way, toward lower scores (beta = -1.15, 95% CI -2.17 to -0.13, P = 0.03).

A separate analysis told a different story. Among 89 women in the Wisconsin registry, average age 69.2 years (SD 6.84), recalling their menopause symptoms years or decades after the fact, a history of hot flashes did associate with higher scores (beta = 0.58, 95% CI 0.06 to 1.11, P = 0.03), as did trouble sleeping (beta = 0.45, 95% CI 0.03 to 0.86) and total symptom count (beta = 0.16, 95% CI 0.02 to 0.30). Night sweats, the one symptom that had nearly reached significance in midlife, was flat in this older group (beta = 0.04, 95% CI -0.37 to 0.46, P = 0.84). A single protein, CCL2, was linked to vasomotor symptoms in both groups.

So the two analyses disagree, and they disagree in a direction worth naming: the symptom link appears in the data built on recall across decades and is absent from the data recorded at the time. Menopause symptoms in this study were self-reported throughout, which the authors list among their limitations.

The reason this paper is worth attention has little to do with risk prediction. Research connecting menopause to dementia has mostly worked with blunt instruments - age at menopause, surgical menopause, years of exposure to endogenous hormones - which establish an association without identifying a mechanism anyone could act on. This study replaces those with named pathways: specific inflammatory, metabolic and synaptic proteins, measured on two independent platforms, moving with estradiol and follicle-stimulating hormone rather than with birthdays.

That is the kind of finding a trial can be built on, which is precisely how the authors frame it: molecular signatures that may inform the selection of biomarkers or therapeutic targets for brain health in midlife women. It is a statement about what to measure next, not about what any individual should conclude. The replication across cohorts is real, and the signal held when hormone therapy history was stratified and oophorectomy excluded. What the work does not do - and does not claim to do - is hand one woman a number about her own brain.

The most useful number in this paper is one that did not reach significance. Across 11,059 women and 387 dementia cases, all-cause dementia did not clearly move with the menopause proteomic score, and the Alzheimer's-specific signal that did move has a lower confidence bound of 1.01 - a result compatible with a small effect and not much more. Set that beside the finding that men scored higher on the menopause signature than women once age was accounted for, and the shape of the work becomes clear: this is early biomarker research into shared pathways of inflammatory and metabolic aging, not the discovery that menopause sets a woman on a road toward Alzheimer's. That is a genuine contribution and a narrower one than the week's coverage implied. The part that warrants the most caution is the symptom story, because it is the part a reader can see in her own life. In the sample where symptoms and proteins were recorded at the same time, hot flashes showed no association with the score at all; the later-life link rests on women recalling symptoms decades afterward. A night spent awake and sweating is genuinely miserable, and on this evidence it is not a readout of anyone's brain.