SLEEP-STAGE DYNAMICS may help identify both sleep-disordered breathing and long-term cardiovascular risk, according to a machine learning analysis of 2,579 adults followed for more than 10 years.
Researchers found that patterns of sleep architecture contained substantial prognostic information, even without direct measures of respiratory disturbance.
The study analysed data from 2,579 participants without previous cardiovascular events or sleep-altering medication use.
Investigators examined sleep-stage architecture, transition dynamics between sleep stages, and common cardiovascular risk factors including age, body mass index, and smoking status.
Using a random classifier, the team assessed whether sleep-stage information alone could identify moderate-to-severe sleep-disordered breathing (defined as an apnoea-hypopnoea index above 15).
The model achieved an area under the receiver operating characteristic curve (AUROC) of 76.1%, indicating good discriminatory performance without the use of direct respiratory measurements.
The findings suggest that sleep-disordered breathing is reflected in characteristic alterations in sleep organisation, including increased fragmentation, reduced continuity of restorative sleep stages, and abnormal transitions between stages.
Researchers then evaluated whether the same sleep features could predict future cardiovascular events.
A random survival model achieved a concordance index of 73.3% over more than a decade of follow-up.
Importantly, model performance remained essentially unchanged when the apnoea-hypopnoea index was excluded, suggesting that sleep-stage dynamics independently capture cardiovascular risk information.
The model demonstrated comparable discriminative ability to established Framingham risk scores.
However, sleep-based predictions showed improved calibration, with the mean predicted 10-year cardiovascular risk differing from the observed event rate by less than 1% in the primary cohort.
Analysis of individual sleep characteristics revealed that cardiovascular risk was often associated with non-linear, U-shaped patterns.
The lowest risk was observed within specific ranges of total sleep time, wake after sleep onset, and continuity of rapid eye movement and deep sleep.
Notably, several uncommon sleep-stage transitions emerged as sensitive markers of cardiovascular vulnerability.
Transitions such as N3 to N1 and REM to N3, even when occurring only once per night, were associated with increases in cardiovascular risk of up to 10%.
The researchers concluded that sleep-stage dynamics may serve as promising non-invasive biomarkers for both sleep-disordered breathing detection and long-term cardiovascular risk stratification.
The findings suggest that these measures could eventually be incorporated into automated polysomnography analysis and future sleep-monitoring technologies.
Reference
Bechny M et al. Sleep-stage dynamics predict current sleep-disordered breathing and future cardiovascular risk. Sci Rep. 2026;DOI:10.1038/s41598-026-69352-2.
Featured image: kittyfly on Adobe Stock
Each article is made available under the terms of the Creative Commons Attribution-Non Commercial 4.0 License.
