O034 Development and Evaluation of Two Non-Invasive Approaches for Estimating Breath-by-Breath Ventilatory Drive in Obstructive Sleep Apnea Patients

Abstract Introduction Estimation of breath-by-breath ventilatory drive is an important step in quantifying key obstructive sleep apnea (OSA) endotypes. However, the gold-standard for measuring ventilatory drive requires invasive instrumentation. Here, we extend two published methods for estimating breath-by-breath ventilatory drive non-invasively from polysomnography. Methods We retrospectively analysed sleep studies of 33 OSA patients (apnea-hypopnea index>5) who underwent a clinical in-lab polysomnography with the addition of intraoesophageal diaphragmatic electromyography (Edi). The ventilatory drive was estimated on a breath-by-breath basis using two independent methods: (a) a flow-shape-based machine learning approach (Flow-Drive method), Mann et al 2019 ERJ; and (b) a chemoreflex model-based approach (Loop-gain method), Terrill et al 2014 ERJ. The Spearman’s rank correlation (ρ) between the two estimated breath-by-breath ventilatory drive signals and Edi (gold-standard) was calculated: (i) for 7-min overlapping windows throughout the night for each individual; and (ii) an overnight ensemble average of obstructive respiratory events at onset and termination periods (three breaths before and after event onset and termination respectively). Results The Spearman’s ρ in the 7-minute window was similar between the Flow-Drive and Loop-gain methods, and Edi [0.40(0.10)vs.0.27(0.13), mean(SD)]. Flow-Drive and Loop-gain methods both showed similar Spearman’s ρ with Edi at event onset [0.69(0.34)vs.0.69(0.25) respectively]. At event termination, the Loop-gain method [0.85(0.13)] was more strongly correlated with Edi than the Flow-Drive method [0.78(0.30)]. Conclusions These promising results set a foundation for integrating the advantages of a machine learning approach and a physiological model-based approach in a unified method to improve the non-invasive estimation of breath-by-breath ventilatory drive.