A new pooled analysis led by UCSF researchers suggests that how your brain looks during sleep, its so‑called brain age, may help predict who goes on to develop dementia. In roughly 7,100 U.S. adults followed for as long as 17 years, people whose sleep EEGs made their brains appear older than their actual age had a higher chance of later dementia. Each 10‑year increase in that brain‑age gap lined up with about a 39% higher risk of dementia.
Study details and headline numbers
The findings come from a pooled analysis of five long‑running community cohorts, MESA, ARIC, Framingham Offspring, MrOS and SOF, that together included 7,071 participants aged roughly 54–94 and followed for a median of 3.5 to 17 years, according to a medRxiv preprint. After adjusting for age, sex, race, education, body‑mass index, current smoking, sleep medications and physical activity, each 10‑year increase in the sleep‑based brain‑age index (BAI) was associated with a hazard ratio of 1.39 for incident dementia, which translates to roughly a 39% higher risk. The association held across cohorts and in sensitivity analyses, the authors report.
Researchers' take
"We found that gap is predictive of future risk of dementia," Yue Leng, an associate professor of psychiatry at UCSF and the study’s senior author, told the San Francisco Chronicle. Leng said sleep features such as fragmentation and loss of deep slow‑wave sleep seem to contribute to a larger brain‑age gap. She urged people to notice sleep problems, including snoring and daytime sleepiness, and to consult clinicians about them.
How the brain‑age index works
The BAI is computed by a machine‑learning model that analyzes overnight sleep EEGs and extracts a set of age‑dependent features. In this pooled analysis, the top contributing metric was waveform kurtosis during N2 sleep, according to the medRxiv preprint. The association between higher BAI and dementia persisted after additional adjustment for comorbidities, including APOE ε4 carrier status and apnea‑hypopnea index. This work builds on earlier validation of sleep‑EEG brain‑age measures as a biomarker for neurodegeneration, including a 2020 JAMA Network Open paper that first described a sleep‑based BAI in clinical cohorts.
Bay Area research and next steps
Leng’s MoonLAIT lab at UCSF, which studies sleep, AI and neurodegeneration, is already pursuing NIH‑funded projects to test polysomnography‑derived biomarkers in larger, more diverse populations and in home settings, according to the lab’s website. Investigators say the method could be adapted to wearable EEG headbands for repeated nights of recording, but they underscore the need for standardization before BAI could be used in routine care. The next steps include independent replication, calibration across devices and trials to see if treating sleep disorders changes the brain‑age trajectory.
Limitations and what to do now
The pooled analysis is currently available as a preprint and has not completed peer review, so the findings are still considered preliminary. Experts caution that BAI is a population‑level marker and not yet a diagnostic test; a single night’s recording will not provide a definitive assessment of an individual’s dementia risk. For now, clinicians say the practical takeaway is familiar: prioritize good sleep habits and discuss symptoms such as loud snoring, breathing pauses or excessive daytime sleepiness with a clinician, since treatable conditions like obstructive sleep apnea are linked to cognitive decline.
