30-Second Takeaway
- Foundation-model analysis of a single PSG night now predicts broad multimorbidity and mortality with strong discrimination.
- Sleep fragmentation and nocturnal hypoxemia, not just AHI, emerge as key targets for neuroprotection and cardioprotection.
- Systemic inflammation, perioperative cognition, and circadian disruption further link sleep disorders to chronic disease trajectories.
- Environmental and digital measures (ALAN exposure, EMA, ventilator downloads) are becoming clinically actionable levers and tools.
- Sleep clinicians will increasingly triage, treat, and monitor using multimodal data streams rather than AHI alone.
Week ending January 10, 2026
Sleep medicine at the crossroads of AI, cardiometabolic risk, brain health, and perioperative care
Foundation PSG model predicts broad disease risk from a single night
A multimodal foundation model (SleepFM) was trained on over 585,000 hours of PSG from about 65,000 participants across multiple cohorts. From a single-night PSG, SleepFM predicted 130 conditions with C-index ≥0.75, including all-cause mortality, dementia, myocardial infarction, heart failure, CKD, stroke, and atrial fibrillation. The model also transferred well to an external cohort and performed competitively with dedicated sleep-staging tools, including for apnea classification. These findings suggest routine PSG could be repurposed for scalable, label-efficient disease risk stratification and decision support in sleep clinics.
Sleep fragmentation, not AHI, predicts brain atrophy over 7 years
In a population-based cohort of 387 adults with baseline PSG and MRI and 7-year follow-up MRI, traditional OSA indices showed no longitudinal structural brain effects. Lower mean nocturnal SpO₂ was associated with reduced total brain volume over time. Higher arousal index was linked to increased brain age and reduced global and regional gray matter thickness. Subjective sleep quality also correlated with brain volume changes, underscoring the importance of sleep continuity and oxygenation over AHI alone for neuroprotection.
Early ASV after MI sharply reduces nocturnal hypoxemic burden
This ancillary analysis of the multicenter randomized TEAM-ASV I trial enrolled 35 patients with first acute MI and AHI ≥15 events/h. Minute-ventilation–triggered ASV initiated early after MI markedly reduced total T90 and both desaturation-related and nonspecific hypoxemic components on the first treatment night. After 12 weeks, ASV maintained lower desaturation-related T90 versus standard care, though nonspecific drifts were similar between groups. Higher overall T90, particularly nonspecific drifts, correlated with larger infarct size, suggesting nocturnal hypoxemia as a potential therapeutic target post-MI.
Artificial light at night modestly but consistently worsens sleep disturbance
This systematic review and meta-analysis included 15 observational studies with 765,838 participants examining artificial light at night (ALAN) and sleep disturbance. Ten studies (691,851 participants) contributed to quantitative synthesis. Individuals in the highest ALAN exposure group had a 27% higher risk of sleep disturbance compared with those with low or no exposure (pooled PR 1.27, 95% CI 1.12–1.45). Despite high risk of bias in many studies, certainty of evidence was rated high, supporting counseling on light reduction and informing urban-lighting policies.
References
Numbered in order of appearance. Click any reference to view details.
Additional Reads
Optional additional studies from this edition.