Prediction of severity of obstructive sleep apnea by awake impulse oscillometry

Obstructive sleep apnea (OSA) is a common disease, which poses a significant health threat. Initial diagnostics with polygraphy or polysomnography are time consuming and expensive. Therefore, there is an unmet medical need for simplification, especially to exclude healthy patients from elaborate and unnecessary diagnostics. Impulse oscillometry (IOS) is a simple, cheap and noninvasive tool to asses upper airway resistance, which is increased in patients with OSA. The objective was to examine the relationship between IOS parameters and polysomnography in order to evaluate the applicability of IOS as a supplementing tool in OSA diagnostics. We performed a prospective, cross-sectional, observational study across 107 participants. Pulmonary function tests with IOS, bodyplethysmography and overnight polysomnography were performed. We computed direct and partial correlations between IOS- and PSG-results. ROC analysis was performed to evaluate the most impactful predictive IOS parameter for diagnosing OSA. In ROC analysis the predicted probability of resistance at 5Hz (R5%) combined with age showed the highest AUC of 0.919, while R5 at 0.4325kPa/(l/s) provided the optimal cut-off. Correlations between IOS parameters and OSA severity as well as the duration and severity of oxygen desaturation were observed. However, they could not be reproduced as partial correlations after eliminating the BMI as confounding variable. Our results cannot indicate the usefulness of IOS in OSA diagnostics. The lack of BMI-independent partial correlations between IOS- and PSG-results suggest a correlation without causality fallacy between IOS- and PSG-results. Therefore, the initial impression of good test quality for IOS might be invalid. •Impulse oscillometry showed good test quality for diagnosing obstructive sleep apnea.•No BMI-independent correlation between impulse oscillometry and polysomnography.•Use of impulse oscillometry in obstructive apnea diagnostics questionable.