Background: Unintentional leakage from the mouth or around the mask may lead to cessation of CPAP treatment; however, the causes of unintentional leaks are poorly understood. The objectives of this study were (1) to identify determining factors of unintentional leakage and (2) to determine the effect of the type of mask (nasal/oronasal) used on unintentional leakage.
Methods: Seventy-four polysomnograms from patients with OSA syndrome treated with auto-CPAP were analyzed (23 women; 56 ± 13 years; BMI, 32.9 kg/m2 (range, 29.0-38.0 kg/m2). Polysomnographic recordings were obtained under auto-CPAP, and mandibular behavior was measured with a magnetic sensor. After sleep and respiratory scoring, polysomnographic signals were computed as mean values over nonoverlapping 10-s intervals. The presence/absence of unintentional leakage was dichotomized for each 10-s interval (yes/no). Univariate and multivariate conditional regression models estimated the risk of unintentional leaks during an interval "T" based on the explanatory variables from the previous interval "T-1." A sensitivity analysis for the type of mask was then conducted.
Results: The univariate analysis showed that mandibular lowering (mouth opening), a high level of CPAP, body position (other than supine), and rapid eye movement (REM) sleep increased the risk of unintentional leaks and microarousal decreased it. In the multivariate analysis, the same variables remained independently associated with an increased risk of unintentional leakage. The sensitivity analysis showed that oronasal masks reduced the risk of unintentional leaks in cases of mouth opening and REM sleep.
Conclusions: Mouth opening, CPAP level, sleep position, and REM sleep independently contribute to unintentional leakage. These results provide a strong rationale for the definition of phenotypes and the individual management of leaks during CPAP treatment.
Keywords: CPAP; mask; polysomnography; sleep apnea; unintentional leakage.
Copyright © 2017 American College of Chest Physicians. Published by Elsevier Inc. All rights reserved.