Establishing a physiological normative value of area-based hypoxic burden in obstructive sleep apnea

Background and Objectives: Area-based hypoxic burden, termed as hypoxic dip area (HDA) has been studied as a novel obstructive sleep apnea (OSA) metric that may better characterize intermittent hypoxemia in OSA and thereby better predict outcomes. However, to be of use in routine clinical care, a physiological normative value is necessary. Here we attempt to use several large cohorts to establish a percentile-based threshold that can classify individuals with normal vs. abnormal physiological profiles. Methods: Data from 10 cohorts (EPISONO, FINS, Dayfun, MrOS, MESA, SHHS, APPLES, WSC, CFS, and AIRS [Mount Sinai Clinical Cohort]; n=14,031 subjects) were included. HDA was defined as the area bounded by SpO2 nadirs ([≥]2% desaturation) flanking left/right peaks. The 97.5th percentile of HDA among asymptomatic subjects (n=136) from EPISONO, FINS, and DAYFUN (aged 20-52 years; 46 [33.8%] male) was used to define normal and elevated HDA levels. Physiological features, clinical comorbidities, and incident cardiovascular disease (CVD) events and mortality were compared between normal/elevated HDA groups. Sensitivity analyses were conducted using age-, sex-, and BMI-adjusted threshold derived from multivariable regression model. Results: The 97.5% percentile cutoff of HDA was 8.6%min/h, classifying 4,500 individuals as normal HDA and 9,531 as elevated HDA. Individuals with elevated HDA ([≥]8.6%min/h) demonstrated greater nocturnal hypoxemia (lower baseline SpO2, higher T90, T85, T80 and oxygen desaturation index), higher AHI (apnea-hypopnea index) and arousal index, and greater daytime sleepiness as well as a higher prevalence of lifetime CVD and hypertension across cohorts. Elevated HDA was related with a higher prevalence of incident CVD events and mortality in unadjusted analyses, although these associations were attenuated after adjustment for confounding factors. Results were similar using alternative thresholds derived from multivariable regression analyses. Conclusions: A physiologically derived cutoff of 8.6% min/h for HDA effectively distinguishes physiological profiles across respiratory, arousal, and oxygenation domains and generalizes across multiple cohorts. Those with elevated HDA had a higher prevalence of incident CVD outcomes, although observed relationships were attenuated after further adjustment. Our findings provide a physiologically interpretable threshold for HDA that can be readily used in clinical use.