Night-to-night reliability and agreement of obstructive sleep apnea pathophysiologic mechanisms estimated with phenotyping using polysomnography in cognitively normal elderly subjects.

STUDY OBJECTIVES: Phenotyping using polysomnography (PUP) is an algorithmic method to quantify physiologic mechanisms underlying obstructive sleep apnea (OSA): loop gain (LG1), arousal threshold (ArTH), and upper airway collapsibility (Vpassive) and muscular compensation (Vcomp). The consecutive-night test-retest reliability and agreement of PUP-derived estimates is unknown. From a cohort of elderly (age ≥55 years), largely non-sleepy, community-dwelling volunteers who underwent in-lab polysomnography (PSG) on 2 consecutive nights, we determined the test-retest reliability and agreement of PUP-estimated physiologic factors.

METHODS: Subjects who had an apnea-hypopnea index (AHI3A) of at least 15 events per hour on the first night were included. PUP analyses were performed on each of 2 PSGs from each subject. Physiologic factor estimates were derived from NREM sleep and compared across nights using intraclass correlation coefficients (ICC) for reliability and smallest real differences (SRD) for agreement.

RESULTS: Two PSGs from each of 43 subjects (86 total) were analyzed. A first-night effect was evident with increased sleep time and stability and decreased OSA severity on the second night. LG1, ArTH, and Vpassive demonstrated good reliability (ICC >0.80). Vcomp had modest reliability (ICC = 0.67). For all physiologic factors, SRD values were approximately 20% or more of the observed ranges, suggesting limited agreement of longitudinal measurements for a given individual.

CONCLUSION: For NREM sleep in cognitively normal elderly individuals with OSA, PUP-estimated LG1, ArTH, and Vpassive demonstrated consistent relative ranking of individuals (good reliability) on short-term repeat measurement. For all physiologic factors, longitudinal measurements demonstrated substantial intraindividual variability across nights (limited agreement).