Hypnogram-derived structured data reveal differences in sleep stage transition characteristics between high and low arousal threshold patients.

BACKGROUND

Arousal threshold (ArTH) is associated with disease severity and sleep stability, but conventional PSG metrics cannot capture dynamic stage transitions. Hypnogram-based analysis provides information on the temporal structure of sleep. How variations in ArTH phenotype affect sleep architecture and transition dynamics remains unclear.

METHODS

This retrospective study included 1792 patients with OSA stratified by arousal threshold. Hypnograms were converted into structured sleep-stage sequences using Python. After propensity score matching for age, gender, and BMI, group differences were evaluated using survival analysis for stage persistence, Poisson regression for transition frequency, and multistate Cox and semi-Markov models for transition dynamics.

RESULTS

After propensity score matching, 580 OSA patients with a high arousal threshold and 580 with a low arousal threshold were analyzed. The High ArTH group showed longer N1 (80.6 ± 58.8 vs. 74.4 ± 54.5 min, p < 0.05), N2 (230.7 ± 76.3 vs. 216.8 ± 81.2 min, p < 0.01) and REM (68.9 ± 32.6 vs. 57.7 ± 35.9 min, p < 0.001), but lower N3 percentage (7.0 ± 7.4% vs. 8.2 ± 7.7%, p < 0.01). Transition analyses revealed unstable N2 with frequent NREM oscillations (N1→N2 RR = 1.076, p < 0.001; HR = 1.099, p < 0.001; N2→N1 RR = 1.219, p < 0.001; HR = 1.316, p < 0.001), reduced N3 progression (N2→N3 RR = 0.843, p < 0.01; HR = 0.837, p < 0.001), and lower NREM-to-wake probability (N1→Wake RR = 0.922, p < 0.05; HR = 0.911, p < 0.001; N2→Wake RR = 0.944, p < 0.05), indicating impaired sleep stability.

CONCLUSIONS

High ArTH patients exhibit unstable N2 sleep, frequent NREM oscillations and reduced deep sleep progression, particularly in moderate and severe disease.