Validation of a new unattended sleep apnea monitor using two methods for the identification of hypopneas

Abstract

Study objectives: The objective of the present study was to evaluate the accuracy of a home sleep apnea test (HSAT), MATRx plus (Zephyr Sleep Technologies, Calgary, Alberta, Canada), in identifying apneas and hypopneas and estimating indices of obstructive sleep apnea (OSA).

Methods: Individuals with suspected OSA underwent a one-night study wearing both HSAT and polysomnogram (PSG) sensors. The results provided by the overnight HSAT were compared with those from the simultaneously recorded PSG. The PSG data were scored manually, and the HSAT data were analyzed automatically using both preceding peak (PP) and moving average window (MW) methods for determining the reference oxyhemoglobin saturation (O₂ Sat). Accuracy of the HSAT in detecting individual apneic and hypopneic events was evaluated on an epoch-by-epoch basis. The apnea-hypopnea indices from the two recording systems were compared.

Results: Agreement analysis for the individual apneic and hypopneic events yielded median values for sensitivity and specificity of 0.89 and 0.98 and positive and negative likelihood ratios of 37.35 and 0.11, respectively. Comparison of OSA indices between the two systems yielded correlation coefficients in the range of 0.95-0.96 and intraclass correlation coefficients ranging from 0.92-0.96. Bland-Altman analyses showed 0-2 cases lying outside the ± 2 standard deviation (SD) band and biases ranging from 2.1 to 5.3 events/h. The biases were larger for MW than PP.

Conclusions: The MATRx plus HSAT identifies apneic and hypopneic events and estimates OSA indices with accuracy suitable for clinical purposes but not in children, patients with underlying lung disease, and habitual mouth-breathers.

Clinical trial registration: Registry: ClinicalTrials.gov; Name: PSG Validation of MATRx plus AHI; Identifier: NCT03627169.

Keywords: Obstructive sleep apnea; apnea-hypopnea index; home sleep apnea test.

Conflict of interest statement

All authors have reviewed and approved the manuscript. Z.L. Topor, J.E. Remmers, J. Grosse, E.V. Mosca, S.A. Zareian Jahromi, Y. Zhu, and S. Bruehlmann are employees of Zephyr Sleep Technologies. Financial Support was provided by Alberta Innovates Technology Futures, Zephyr Sleep Technologies, and the NRC-IRAP Industrial Assistance Fund. The Alice 6 Sleepware G3 polysomnogram software was provided by Philips Respironics, Inc.

© 2020 American Academy of Sleep Medicine.

Figures

Figure 1. Linear regression between polysomnogram (PSG), apnea-hypopnea index (AHI) 4%, and MATRx plus (preceding peak, PP) scored events.

Slope, intercept, 95% confidence intervals, and y = x identity lines are shown.

Figure 2. Linear regression between polysomnogram (PSG), apnea-hypopnea index (AHI) 3%, and MATRx plus AHI 3% (moving average window, MW) scored events.

Slope, intercept, 95% confidence intervals, and y = x identity line are shown.

Figure 3. Bland-Altman plots.

(A–D) Bland-Altman plots of polysomnogram (PSG) AHI versus MATRx plus event rates. (E, F) Bland-Altman plots of MATRx plus event rates for two different methods of calculating the oxygen desaturation reference point. LOA = limit of agreement, MW = moving average window, PP = preceding peak.

Figure 4. Receiving operator characteristic (ROC) analysis for MATRx plus AHI 4% MW and four clinically relevant cutoff points for PSG AHI 4%.

The numbers below the specific points denote cutoff values for the MATRx plus system.