A comparison of 2 visual methods for classifying obstructive vs central hypopneas

Study Objectives: Rules for classifying apneas as obstructive, central, or mixed are well established. Although hypopneas are given equal weight when calculating the apnea-hypopnea index, classification is not standardized. Visual methods for classifying hypopneas have been proposed by the American Academy of Sleep Medicine and by Randerath et al (Sleep. 2013;36[3]:363-368) but never compared. We evaluated the clinical suitability of the 2 visual methods for classifying hypopneas as central or obstructive. Methods: Fifty hypopnea-containing polysomnographic segments were selected from patients with clear obstructive or clear central physiology to serve as standard obstructive or central hypopneas. These 100 hypopnea-containing polysomnographic segments were deidentified, randomized, and scored by 2 groups. We assigned 1 group to use the American Academy of Sleep Medicine criteria and the other the Randerath algorithm. After a washout period, re-randomized hypopnea-containing polysomnographic segments were scored using the alternative method. We determined the accuracy (agreement with standard), interrater (Fleiss's kappa), and intrarater agreement (Cohen's kappa) for obtained scores. Results: Accuracy of the 2 methods was similar: 67% vs 69.3% for Randerath et al and the American Academy of Sleep Medicine, respectively. Cohen's kappa was 0.01-0.75, showing that some raters scored similarly using the 2 methods, while others scored them markedly differently. Fleiss's kappa for the American Academy of Sleep Medicine algorithm was 0.32 (95% confidence interval, 0.29-0.36) and for the Randerath algorithm was 0.27 (95% confidence interval, 0.23-0.30). Conclusions: More work is needed to discover a noninvasive way to accurately characterize hypopneas. Studies like ours may lay the foundation for discovering the full spectrum of physiologic consequences of obstructive sleep apnea and central sleep apnea.