A quick roundup of research and news that caught my attention this week. The through-line: noninvasive signals are getting more credible as early-warning tools, large-scale real-world datasets are reshaping how we study dementia, and biomarkers are moving closer to symptoms rather than just pathology.
1. Brain age from sleep EEG may signal dementia risk
Researchers applied machine learning to sleep EEG recordings to estimate “brain age.” When estimated brain age exceeded chronological age, future dementia risk went up, roughly 40% higher risk per 10-year gap. The pooled dataset covered ~7,000 participants across five studies (ages 40 to 94, follow-up ranging from 3.5 to 17 years), with ~1,000 incident dementia cases.
The model used 13 microstructural EEG features rather than broad sleep metrics like sleep stages or efficiency. As Yue Leng put it: “Broad sleep metrics don’t fully capture the complex multidimensional nature of sleep physiology.”
The implication is that fine-grained passive signals, potentially from wearables, could flag risk in nonclinical settings long before symptoms appear.
2. Turning living rooms into dementia research labs
A UK initiative (Cera, GlobalMinds/AkriviaHealth, Cardiff University) is trying to fix a persistent research blind spot: older adults account for two-thirds of illness in the UK but only one-third of clinical trial participants. Their approach is to bring the study into people’s homes, capturing data during the “silent hours” outside clinics, and build a multimodal longitudinal dataset from NHS records, questionnaires, and biological samples including DNA.
Target: ~1,000 people living with dementia. Dr Ben Maruthappu’s framing is memorable: “We aren’t just monitoring health; we are turning every living room into a laboratory for a cure.”
The partnership model here (home care provider, data platform, academic center) is worth watching as a template.
3. A metaplatform for Alzheimer’s at 10-million-patient scale
Columbia and collaborators describe the M3AD Study: an effort to harmonize EHRs from nearly 10 million patients across health systems in NYC, Chicago, and Miami, including ~60,000 with AD/ADRD. The platform treats Alzheimer’s not as an isolated condition but as one node in a web of interacting chronic diseases, behaviors, and social circumstances.
A few things stand out: the scale (roughly 35% of Americans over 85 have Alzheimer’s; more than 7.2 million older Americans are living with it today), the multimorbidity framing (~90% of adults over 60 have 2+ chronic diseases), and tools like eRADAR that flag potentially undiagnosed dementia from routine EHR data. The platform is designed to be extensible: imaging, genetics, and additional biomarkers can be layered in.
4. Siemens launches blood-based biomarker assays for brain health
Siemens Healthineers announced automated immunoassays for phosphorylated tau 217 (pTau217) and brain-derived tau (BDTau) running on their Atellica analyzer platform. The pitch is simpler scaling: blood tests instead of lumbar punctures for CSF.
Henrik Zetterberg’s quote captures the appeal neatly: “Blood tests are much easier for both patients and doctors: you can scale testing, follow patients, or perhaps prepare a biomarker portfolio.”
The entry of a major diagnostics incumbent into “brain health portfolios” as a product category signals that this market is maturing fast.
5. Synaptic CSF markers predict Alzheimer’s progression with AUC ~0.8
A Nature Communications study evaluated CSF synaptic proteins NPTX1 and NPTXR as progression markers across two cohorts (635 participants total, from China and Norway). Both proteins declined as AD worsened in amyloid-positive individuals, correlated with cortical thinning, and were associated with faster clinical decline.
Among MCI participants followed ~2.5 years, lower baseline levels predicted conversion to dementia with AUC ≈ 0.8, outperforming several classical CSF markers in that cohort.
Why this matters conceptually: synaptic damage is “closely tied to symptoms,” so synaptic markers may be more proximal to functional decline than amyloid or tau alone. For staging and progression forecasting, that’s a meaningful distinction.
6. Type 1 diabetes and dementia risk: nearly 3x
A large AAN study (~283,772 people, average age 64) found that people with type 1 diabetes were nearly three times as likely to develop dementia as those without diabetes, and the association appears stronger than for type 2 (roughly 2x). Over ~2.4 years average follow-up, incidence was 2.6% (T1D), 1.8% (T2D), 0.6% (no diabetes).
The usual caveats apply: this is association, not causation, and diagnoses came from EHR and survey data. But as a signal for risk stratification in older adults with metabolic comorbidities, it’s notable.
More dispatches from the aging and brain health research frontier as they come.