Background: The impact of leaks has mainly been assessed in bench models using continuous leak patterns which did not reflect real-life leakage. We aimed to assess the impact of the pattern and intensity of unintentional leakage (UL) using several respiratory models.
Methods: An active artificial lung (ASL 5000) was connected to three bilevel-ventilators set in pressure mode; the experiments were carried out with three lung mechanics (COPD, OHS and NMD) with and without upper airway obstruction. Triggering delay, work of breathing, pressure rise time, inspiratory pressure, tidal volume, cycling delay and the asynchrony index were measured at 0, 6, 24 and 36 L/min of UL. We generated continuous and inspiratory UL.
Results: Compared to 0 L/min of UL, triggering delays were significantly higher with 36 L/min of UL (+27 ms) and pressure rise times were longer (+71 ms). Cycling delays increased from -4 [-250-169] ms to 150 [-173-207] ms at, respectively 0 L/min and 36 L/min of UL and work of breathing increased from 0.15 [0.12-0.29] J/L to 0.19 [0.16-0.36] J/L. Inspiratory leakage pattern significantly increased triggering delays (+35 ms) and cycling delays (+263 ms) but decreased delivered pressure (-0.94 cmH2O) compared to continuous leakage pattern. Simulated upper airway obstruction significantly increased triggering delay (+199 ms), cycling delays (+371 ms), and decreased tidal volume (-407 mL) and pressure rise times (-56 ms).
Conclusions: The pattern of leakage impacted more the device performances than the magnitude of the leakage per se. Flow limitation negatively reduced all ventilator performances.
Keywords: air leak; bench study; chronic respiratory failure; mask leak; noninvasive ventilation.