Evaluation of AHI agreement between PAP algorithms and polygraphy in CSA patients

Abstract

Introduction: During continuous Positive Air Pressure (PAP) treatment, residual Apnea Hypopnea Index (AHI) determination relies only on Flow. Manufacturers use their own determinants to classify respiratory events and use different technologies (forced oscillation technique, pulse of pressure, cardiac oscillations, machine learning). Residual AHI from devices are challenged when compared to AASM compliant analysis (Li, Q. Y. et al. Sleep 2015; 38:597–605). There is still a need to investigate in reproducible conditions potential discrepancy in AHI because treatment efficacy often relies on it with Central Sleep Apnea (CSA) phenotypes.

Aim: We aimed to compare AHI detection in several PAP devices by using bench test simulation of reproducible respiratory profile recorded in CSA patients by using polygraphy.

Methods: We compared AHI, and related indices of CSA patients obtained with NoxT3 polygraph and from Airsense (Resmed), Sleepstyle (Fischer & Paykel), S.Box (Sefam) and Dreamstation (Philips). 26 hours of breathing profiles from three CSA patients containing 3133 respiratory events (1378 Apneas (59% Central), 1755 hypopneas) were characterized following AASM rules. Breathing profiles were simulated on a bench (Zhu, K. et al. Sleep Med Clin 2012; 12:517-527) including an artificial lung (ASL 5000) and a Starling resistor driven by a custom-made algorithm.

Results: Comparisons revealed mild differences for global AHI whereas there was a tremendous underestimation of hypopnea rate (p<0.05, n=26).

Conclusion: This study suggests that AHI from devices are underestimated in CSA patients. Particular attention would be required when considering treatment efficiency using flow based AHI.