Principal component analysis identified a high respiratory-burden phenotype in older White males (d=0.83) and a protective sleep-continuity phenotype in younger non-White women (d=0.34; p<0.001).
Cohort (n=4,329)
No
Principal component analysis of polysomnography data identified distinct sleep-burden phenotypes that align with specific demographic and cardiometabolic profiles.
Effect estimate: Cohen's d 0.83 (PC1); 0.34 (PC2)
p-value: p=<0.001
Abstract Introduction Obstructive sleep apnea (OSA) is one of the most common sleep disorder and can be characterized with multiple physiological burdens: hypoxic burden (HB), ventilatory burden (VB), and arousal burden (AB). To better capture patterns across these dimensions, we applied principal component analysis (PCA) to derive axes integrating multiple burdens and examine their relation to demographic and cardiometabolic phenotypes in sleep clinic patients. Methods We analyzed 4,329 adults (mean age 49.7 ± 15.4 y, 50.8% female, 20.4% non-White, 41.6% overweight) who underwent diagnostic in-laboratory polysomnography at Massachusetts General Hospital (2010–2020) through the Human Sleep Project. PCA was applied to 31 features, including HB (cumulative oxygen desaturation), VB (low-amplitude breathing), AB (total arousal duration), Apnea-Hypopnea Index (AHI), and Respiratory Disturbance Index (RDI), measured across the whole night (WN), NREM, and REM, along with sleep-architecture features. Associations with demographic or cardiometabolic variables were assessed using t-tests with Cohen’s d. For each principal component, we calculated the associations of the component score with variables including age (50y vs ≤50y), sex, BMI, race, hypertension, and diabetes. Results PCA revealed two primary components (PC1: 29.7%; PC2: 12.9% variance explained), subsequent components contributed 10% each. PC1 represented a global respiratory-burden phenotype, primarily driven by WN and NREM HB (apnea-related HB loadings: 0.28, 0.27; overall HB loadings: 0.25, 0.25), AHI and RDI (loadings: 0.30, 0.30), with additional contributions from AB and VB, and minimal influence from sleep-architecture features. Higher PC1 scores occurred in older, male, White, overweight individuals (p 0.001, d = 0.83). PC2 defined a sleep-continuity phenotype, with higher sleep efficiency (loading: 0.42), total sleep time (loading: 0.36), lower wake after sleep onset (loading: –0.39) and sleep fragmentation (loading: –0.31), minimally influenced by respiratory burden. Higher PC2 scores occurred in younger, non-White women without hypertension or diabetes (p 0.001, d = 0.34). Conclusion In 4,329 adults, PCA identified two interpretable sleep-burden phenotypes: a high respiratory-burden phenotype in older White males with elevated BMI, and a protective sleep-continuity phenotype in younger, non-White women without hypertension or diabetes. These findings demonstrated that demographic and cardiometabolic factors aligned with distinct physiologic burdens. Support (if any) NIA (R21AG085495) NIA (R01AG083836)
Albrecht et al. (2026) conducted a cohort in Obstructive sleep apnea (n=4,329). Principal component analysis of sleep-burden features was evaluated on Sleep-burden phenotypes (principal components) and their association with demographic and cardiometabolic variables (Cohen's d 0.83 (PC1); 0.34 (PC2), p=<0.001). Principal component analysis identified a high respiratory-burden phenotype in older White males (d=0.83) and a protective sleep-continuity phenotype in younger non-White women (d=0.34; p<0.001).