B20-01 Advances in Estimating Sleep Apnea Endotypes From Polysomnography Using Airflow Shapes

Abstract Introduction Obstructive sleep apnea (OSA) is a highly prevalent disorder that is challenging to treat. Characterizing the causes of OSA (endotypic traits), such as pharyngeal collapsibility, is a promising avenue for precision sleep. The leading non-invasive method, phenotyping using polysomnography (PUP), provides a measure of collapsibility through a model-based estimate of ventilatory drive. However, due to uncertainty in drive estimation, resulting collapsibility measures show only modest association with gold standards and have limited predictive value. Here, we develop an advanced method to directly estimate ventilatory drive from airflow shapes (PUPFlowDrive) to provide a more accurate measure of collapsibility. Methods In the MADOx (mandibular advancement device MAD trial, age 53.9±12.4 yrs; 41% female; BMI 32.5±6.2 kg/m2; AHI 39.1±27.8 /hr), N = 59 patients completed overnight polysomnography with ventilation (pneumotachograph) and ventilatory drive assessment using esophageal pressure, providing a gold standard estimate of collapsibility (Vpassive, ventilation at normal drive). For the PUPFlowDrive method, a continuous ventilatory drive signal was derived from a flow-shape-based estimate of airflow obstruction. Collapsibility derived from PUPFlowDrive and the original PUP was compared against the gold standard (standardized linear regression). We further examined associations with future MAD treatment response using multivariable logistic regression (median response = 46% reduction in AHI, N = 38). Replication analysis was performed in the PAtO trial (N = 51, age 48.6±12.4 yrs; 33% female; BMI 32.6±6.6 kg/m2; AHI 38.8±22.9 /hr) using diaphragm EMG as gold standard ventilatory drive. Results PUPFlowDrive-derived collapsibility was strongly associated with the esophageal pressure gold standard (standardized β = 0.78), exceeding the original PUP metric (β = 0.64). Collapsibility by PUPFlowDrive was associated with a favorable MAD response (OR = 3.9 1.2-13.0 per SD increase in collapsibility, p = 0.026) similar to the gold standard (OR = 4.1 1.3-13.3, p = 0.019). By contrast, the original PUP method was not predictive (OR = 1.7 0.5-5.8, p = 0.4). Replication in PAtO confirmed the higher association between PUPFlowDrive -derived collapsibility and the gold standard (β = 0.79), compared to the original PUP method (β = 0.68). Conclusion A novel flow-shape-based translational endotyping method is strongly associated with the gold standard measures of collapsibility, outperforming the original PUP method. The new method exhibited predictive value for MAD treatment response that was similar to the invasive gold standard. PUPFlowDrive can be derived entirely from the non-invasive respiratory flow signal, representing a more powerful and practical means for OSA endotyping in the clinical setting. This abstract is funded by: NIH NHLBI: R01HL168067