Reimagining the diagnosis of obstructive sleep apnea beyond the apnea-hypopnea index

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT Obstructive sleep apnea (OSA) is a common chronic condition variably associated with neurocognitive impairment, hypertension, and incident cerebrovascular and cardiovascular disease. OSA is currently defined by measuring the frequency of obstructive events (apnea-hypopnea index [AHI]) defined by the number/hour of obstructions to breathing lasting at least 10 seconds and associated with cortical arousals and/or oxygen desaturation during sleep. The AHI has been only modestly useful for diagnosis of disease severity and for predicting short- and long-term consequences in large epidemiologic studies and mounting evidence of AHI's limitations on both physiological as well as methodological grounds has led to an outcry for rethinking the way we currently diagnose OSA. This project proposes to extract, from standard sleep studies, metrics to characterize the burden of OSA in three distinct domains: ventilatory, hypoxic and arousal. The ventilatory burden will be quantitated automatically from the distribution of breath sizes in a whole night combined with the probability that each breath is obstructive. Hypoxic burden is quantitated in an automated fashion from overnight oxygen desaturation (area between baseline saturation and desaturation events). Arousal burden is quantitated using EEG metrics that indirectly relate to sleep depth and/or arousability overnight. The two overarching aims of this project are: 1) use ventilatory burden to identify presence of OSA better than AHI, and 2) use combination of ventilatory/hypoxic/arousal burdens, where the equation for combination is machine-learned, to predict risk of short- (daytime sleepiness) and long-term (cardiovascular mortality, stroke, and incident congestive heart failure/myocardial infarction) consequences of OSA. By using existing data (in-lab and at-home polysomnography) from >9000 subjects, the proposed project will harness the power of big data to refine and improve the way we diagnose OSA and assess risk of adverse outcomes.