Improving the quantification of endotypic traits in obstructive sleep apnea from standard poly(somno)graphy

Obstructive sleep apnea (OSA) is characterized by repeated airway collapse during sleep, resulting in oxygen desaturation or arousal. There are four underlying causes (endotypes): increased upper airway collapsibility, low muscle responsiveness, low arousal threshold, and elevated loop gain. The Phenotyping Using Polysomnography (PUP) model may be applied to polysomnography (PSG) recordings to perform non-invasive endotype quantification. This study aims to improve the current PUP-model, and make it applicable to polygraphy (PG) measurements, a more widely used scaled-back version of PSG. Improving the model optimization technique resulted in 19.7% less model error and 27.2% more physiologically plausible model fits. This also significantly impacted endotype quantification. Furthermore, an arousal detector (AD) that estimates arousals from PG-data was designed. The PGAD-method (PG with AD-detected arousals) increased loop gain correlation with PSG (r=0.797 vs. r=0.700 for PG alone) and enabled quantification of the other endotypes. This study highlights the impact of the model optimization technique and proved added value of the AD in PG-based analysis. Future work should find the best optimization technique, enhance arousal detection accuracy, and compare PUP-model outputs to gold standard methods. Our work increased the accessibility of endotyping in clinical practice, making an important step towards personalized treatment of OSA.