A laboratory validation study of the diagnostic mode of the Autoset system for sleep-related respiratory disorders

We performed a validation study of the diagnostic mode of the Autoset system (ResMed, Australia) on a group of 44 snorers (10 women). We compared the result of the Autoset's automatic analysis of nasal airflow (using nasal prongs) to those of an in-laboratory polysomnographic study with a Fleisch facemask pneumotachograph. For the first 29 patients, the Autoset software was set to recognize only apneas; for the remaining 15, the software was modified to recognize both apneas and hypopneas. Relative to polysomnography, the Autoset overestimated the number of apneas. Oral breathing or displacement of the nasal prongs partially explained these differences. A significant correlation was found between the apnea indices (AI) assessed by the two methods (r = 0.98). For an AI of 20/hour the Autoset was 100% sensitive and 88% specific. The Autoset significantly underestimated the number of hypopneas compared to the polysomnograph with pneumotachograph (62.9 +/- 4.7 vs. 85.5 +/- 73.1, P = 0.04), although for an apnea-hypopnea index of 20, Autoset was 100% sensitive and 88% specific. The lack of linearity of Autoset's volume evaluation at low volumes could explain most of the differences. Our results indicate that the Autoset system, in its diagnostic mode, is a useful tool for identifying patients with significant obstructive sleep apnea syndrome. The system is less useful in patients with mild to moderate sleep disordered breathing, where it may give erroneous results.