Oxygen Desaturation Index Differs Significantly Between Types of Sleep Software

Transcription

1 pii: jc SCIENTIFIC INVESTIGATIONS Oxygen Desaturation Index Differs Significantly Between Types of Sleep Software Yvonne Ng, MBBS(Hons), BMedSc, FRACP 1 ; Simon A. Joosten, MBBS, BMedSc, FRACP, PhD 1,2,3 ; Bradley A. Edwards, PhD 4,5 ; Anthony Turton, BSc(Hons) 1 ; Helen Romios, BSc (MLS) 1 ; Thilini Samarasinghe, BBiomedSc(Hons), PhD 1,6 ; Shane Landry, PhD 4,5 ; Darren R. Mansfield, MBBS, FRACP, PhD 1,3,5 ; Garun S. Hamilton, MBBS, FRACP, PhD 1,2,3 1 Department of Lung and Sleep Medicine, Monash Health, Victoria, Australia; 2 School of Clinical Sciences, Monash University, Victoria, Australia; 3 Monash Partners, Epworth, Victoria, Australia; 4 Sleep and Circadian Medicine Laboratory, Department of Physiology, Monash University, Victoria, Australia; 5 School of Psychological Sciences and Monash Institute of Cognitive and Clinical Neurosciences, Monash University, Victoria, Australia; 6 Hudson Institute of Medical Research, Victoria, Australia Study Objectives: The aim of this study was to compare the oxygen desaturation index (ODI) generated by two different sleep software systems. Methods: Participants undergoing diagnostic polysomnography for suspected obstructive sleep apnea underwent simultaneous oximetry recording using the ResMed ApneaLink Plus device (AL) and Compumedics Profusion PSG3 system (Comp). The ODI was calculated by the algorithms in the respective software of each system. To determine if differences were due to algorithm or recording devices, the Comp software was also used to generate ODI values using oximetry data from the AL. Results: In 106 participants, there was good correlation but poor agreement in the ODI generated by the two systems. AL ODI values tended to be higher than Comp ODI values, but with significant variability. For ODI4%, bias was 4.4 events/h (95% limits of agreement 5.8 to 14.6 events/h). There was excellent correlation and agreement when the same oximetry raw data was analyzed by both systems. For ODI4%, bias was 0.03 events/h (95% limits of agreement 2.7 to 2.8 events/h). Similar results were evident when the ODI3% was used. Conclusions: There is a clinically significant difference in ODI values generated by the two systems, likely due to device signal processing, rather than difference in ODI calculation algorithms. Keywords: cardiovascular diseases, correlation, obstructive sleep apnea, oximetry, polysomnography Citation: Ng Y, Joosten SA, Edwards BA, Turton A, Romios H, Samarasinghe T, Landry S, Mansfield DR, Hamilton GS. Oxygen desaturation index differs significantly between types of sleep software. J Clin Sleep Med. 2017;13(4): INTRODUCTION Obstructive sleep apnea (OSA) is characterized by repetitive collapse of the upper airway during sleep. OSA is highly prevalent, with moderate-severe OSA reported to affect between 17% and 49% of middle-aged men. 1,2 The gold standard for diagnosis of OSA is in-laboratory overnight polysomnography. Polysomnography is complex and labor intensive and incorporates the measurement of multiple physiological signals with manual interpretation of data. OSA is most commonly diagnosed based on the calculation of an apneahypopnea index (AHI). The rules for scoring hypopneas have changed over time 3 5 and a key component of the definition of a hypopnea is the extent of oxygen desaturation accompanying the respiratory event. Over recent years there has been increasing use of homebased, limited-channel recording devices to diagnose OSA. These devices are classified either as type 3 or type 4 monitors, depending on the number of signals recorded. 6,7 The cornerstone signal of type 3, and particularly, type 4 devices is the oxygen saturation (SpO 2 ), measured via pulse oximetry. This enables automatic calculation of an oxygen desaturation index (ODI), which is the average number of desaturation episodes per hour of recording. Typically, ODI is reported as BRIEF SUMMARY Current Knowledge/Study Rationale: The oxygen desaturation index (ODI) is a commonly used metric in the assessment of obstructive sleep apnea. However, it has been unclear whether the ODI differs according to the type of system used to measure it. Study Impact: This study highlights the significant difference in the ODI measured by two common sleep diagnostic systems, which has implications for the usage of the ODI as a diagnostic tool or marker of cardiovascular risk in obstructive sleep apnea. Further research is required to confirm whether other sleep diagnostic systems differ in their measurement of the ODI. the number of 3% desaturations (ODI3%) and/or the number of 4% desaturations (ODI4%). ODI has validity in the diagnosis of OSA. 8 When there is a high clinical probability of OSA, an elevated ODI3% or ODI4% performs well at ruling in the diagnosis, 9 13 particularly in obese subjects, and has higher diagnostic utility in severe compared to moderatesevere OSA. 14 However, a negative result does not exclude OSA in those at high risk and still requires further evaluation. Furthermore, there has been increasing interest in the role of ODI4% in determining the risk for hypertension and cardiovascular disease in patients with OSA. In a report from 599 Journal of Clinical Sleep Medicine, Vol. 13, No. 4, 2017

2 the European Sleep Apnea Database, although both ODI4% and AHI were associated with prevalent hypertension, only ODI4% remained independently associated with hypertension when both were entered into a multiple regression model. 15 Moreover, the majority of studies showing an independent association between OSA and cardiovascular disease have embedded a 4% desaturation in the hypopnea definition. 16,17 This hypopnea definition correlates very strongly to ODI4%. 18 A large randomized controlled trial assessing whether CPAP lowers cardiovascular risk the SAVE (Sleep Apnea cardio- Vascular Endpoints) study utilized ODI4% 12 events/h as the main inclusion criterion for OSA diagnosis and treatment. 19 Importantly, this study used the ResMed ApneaLink device to measure the ODI4%. A potential advantage of ODI over AHI is that it is based on information from a single channel (oximetry) rather than multiple respiratory channels and is calculated by an algorithm rather than manual scoring, and thus is highly reproducible if recorded and analyzed using the same device and software. The variability of AHI between and within observers is well characterized and appreciated by clinicians. 20 However, there has not been any systematic comparison between ODI values measured using different oximeters and software. Such a comparison is important given that oximeter data processing and ODI algorithms differ between manufacturers, 7 and it is unclear whether different systems would produce the same ODI values in a given patient, and whether this affects either OSA diagnosis or the assessment of cardiovascular risk. We hypothesize that there will be systematic differences between ODIs measured using different systems due to these algorithm differences. As a result, the current study aims to compare the ODI measured using two common home- and laboratory-based OSA testing systems, the ResMed ApneaLink Plus device (ResMed, Sydney, Australia) hereafter referred to as AL and Compumedics Grael Profusion PSG3 system (Compumedics Limited, Abbotsford, Victoria, Australia) hereafter referred to as Comp. METHODS Institutional ethics approval was obtained from the Monash Health Human Research Ethics Committee (approval number 14358L). Consecutive adult patients with suspected OSA who were undergoing an overnight in-laboratory diagnostic sleep study at Monash Medical Centre from Sunday to Thursday nights between October 9, 2014 and February 8, 2015 were eligible for the study. Up to three participants were enrolled each night. When the number of eligible patients exceeded the number of AL available, they were randomly selected for inclusion. Written informed consent was obtained from each participant. Enrolled participants underwent simultaneous oximetry recording using both the AL and Comp. The Comp was part of the patients diagnostic sleep study recording, which proceeded according to standard criteria. 21 The AL was used in addition to the full polysomnography montage. Each system s in-built oximeter was connected to individual Nonin pulse oximetry sensors (Nonin Medical, Inc., Minnesota, United States), which were attached to separate fingers on opposite hands. Consequently, the Comp ran from a finger sensor on one hand and the AL ran from a finger sensor on the opposite hand. The ODI4% and ODI3% for the simultaneous, matched total recording time (irrespective of whether the participant was awake or asleep) were automatically calculated by the software of each system using its respective oximetry data. Recording time rather than sleep time ODI was used for analysis so that the denominator (ie, time) for ODI calculation was the same for both the AL and Comp. This is because the AL does not have EEG input unlike the Comp in our current setup. This also mimics how the AL and other limited-channel devices perform in the home setting. The ResMed ApneaLink software system version 9.20 (ResMed, Sydney, Australia) and the Compumedics Profusion PSG3 software system version 3.4, build 401 (Compumedics Limited, Abbotsford, Victoria, Australia) were used for analysis. To determine if discrepancies were due to software algorithm or recording device data processing differences, the Comp software was used to generate ODI values utilizing the AL oximetry data exported in European Data Format (EDF). Data were entered into an Excel spreadsheet (2010, Version , Microsoft Corporation, Redmond, Washington, United States) and subsequently exported into Prism version 6.0, 2014 (GraphPad, San Diego, California, United States) for analysis. Correlation between the AL ODI and Comp ODI was assessed using the Pearson correlation coefficient for ODI3% and ODI4%. Bland-Altman plots were used to determine agreement between the ODIs calculated by the two systems. 22 We aimed to enroll at least 100 participants to achieve a confidence interval (CI) for the 95% limits of agreement of ± 0.34s, where s is the standard deviation of the differences between measurements by the two methods. 22,23 To determine the effect of differences in ODI between systems on potential clinical decision-making, categories of OSA diagnosis for the different systems were made utilizing ODI3% 5 events/h for any OSA and 15 events/h for moderate-severe OSA. These were compared with the AHI calculated through scoring of the respiratory signals recorded by the Comp using the American Academy of Sleep Medicine 2012 rules. 21 The chi-square test and odds ratio were used to compare proportions of OSA diagnosis. RESULTS A total of 116 patients consented to the study. Data from 10 participants had to be excluded due to technical issues with the oximetry recording, as demonstrated in Figure 1. Table 1 summarizes the baseline characteristics of the remaining 106 participants. Device ODI Comparison Correlation and agreement between the ODI measured by each device, and calculated by their respective software, are summarized in Figure 2 and Figure 3. For AL ODI4% versus Comp ODI4%, r = 0.96 (P <.0001; 95% CI ) and r 2 = For AL ODI3% versus Comp ODI3%, r = 0.94 (P <.0001; 95% Journal of Clinical Sleep Medicine, Vol. 13, No. 4,

3 CI ) and r 2 = AL ODI values tended to be higher than Comp ODI values, but with significant variability. For ODI4% bias = 4.4 events/h (95% limits of agreement 5.8 to 14.6 events/h). For ODI3%, bias = 7.1 events/h (95% limits of agreement 6.4 to 20.6 events/h). Figure 1 Study population flow diagram. Software ODI Comparison The performance of the two software systems was assessed by using a single recording source of data. The data recorded using the AL were analyzed by both the AL software and the Comp software (having been exported in EDF from the AL), thus allowing for a specific comparison of the effect of software algorithm on the ODI via correlation and agreement. For ODI4%, r = (P <.0001, 95% CI ) and bias = 0.03 events/h (95% limits of agreement 2.7 to 2.8 events/h). For ODI3%, r = (P <.0001; 95% CI Table 1 Characteristics of study population (n = 106). Age, y, mean (SD) 47 (15.5) Sex, n (%) Male 62 (58.5) Female 44 (41.5) Body mass index, kg/m 2, mean (SD) 31.9 (7.1) Epworth Sleepiness Scale score, mean (SD) 7.8 (4.8) Apnea-hypopnea index, events/h, mean (SD) 23.2 (25.0) ODI4%, events/h, mean (SD) ApneaLink 15.0 (17.8) Compumedics 10.6 (16.3) ODI3%, events/h, mean (SD) ApneaLink 22.0 (19.4) Compumedics 14.9 (18.8) ODI4% = oxygen desaturation index for 4% desaturation, ODI3% = oxygen desaturation index for 3% desaturation, SD = standard deviation. Figure 2 Pearson correlation comparing ODI generated by the ApneaLink Plus and Compumedics software systems. (A) Comp versus AL ODI4%. (B) Comp versus AL ODI3%. (C) AL-Comp versus AL ODI4%. (D) AL-Comp versus AL ODI3%. Correlation between systems improved when using the ApneaLink Plus oximetry data instead of the respective oximetry data. ODI4% = oxygen desaturation index for 4% desaturation, ODI3% = oxygen desaturation index for 3% desaturation, Comp ODI = Compumedics ODI using Compumedics oximetry data, AL ODI = ApneaLink Plus ODI using ApneaLink Plus oximetry data, AL-Comp ODI = Compumedics ODI using ApneaLink Plus oximetry data. 601 Journal of Clinical Sleep Medicine, Vol. 13, No. 4, 2017

4 Figure 3 Bland-Altman plots comparing ODI generated by the ApneaLink Plus and Compumedics software systems. (A) Difference versus average for AL and Comp ODI4%. (B) Difference versus average for AL and Comp ODI3%. (C) Difference versus average for AL and AL-Comp ODI4%. (D) Difference versus average for AL and AL-Comp ODI3%. Agreement between systems improved when using the ApneaLink Plus oximetry data instead of the respective oximetry data. ODI4% = oxygen desaturation index for 4% desaturation, ODI3% = oxygen desaturation index for 3% desaturation, Comp ODI = Compumedics ODI using Compumedics oximetry data, AL ODI = ApneaLink Plus ODI using ApneaLink Plus oximetry data, AL-Comp ODI = Compumedics ODI using ApneaLink Plus oximetry data ) and bias = 0.08 events/h (95% limits of agreement 3.1 to 3.3 events/h). Device Bias on Diagnostic Categories When considering OSA as ODI3% 5 events/h, Comp ODI3% resulted in a diagnosis of OSA in 66 subjects, compared to 90 subjects for AL ODI3%, which is 36% more subjects with OSA when the ApneaLink was used; P = Odds ratio for OSA diagnosis with AL ODI3% compared to Comp ODI3% was 3.4 (95% CI ). For moderate-severe OSA (ODI3% 15 events/h), AL ODI3% resulted in a diagnosis in 59 subjects (56%), compared to 32 subjects (30%) for Comp ODI3%, P = Odds ratio for diagnosing moderate-severe OSA with AL ODI3% compared to Comp ODI3% was 2.9 (95% CI ). When looking at the diagnosis of moderate-severe OSA (based on AHI 15 events/h), the AL ODI3% and Comp ODI3% performed differently. AL ODI3% 15 events/h diagnosed moderate-severe OSA with a sensitivity of 96.4% (95% CI 87.5% to 99.6%) and a specificity of 88.2% (95% CI 76.1% to 95.6%). Comp ODI3% 15 events/h diagnosed moderatesevere OSA with a sensitivity of 58.2% (95% CI 44.1% to 71.3%) and a specificity of 100% (95% CI 93% to 100%). DISCUSSION Our study demonstrates a clinically significant difference between the ODI3% and ODI4% values generated by the AL software and Comp. Despite good correlation between devices, our results show that there is a bias for AL ODI values to be higher than Comp ODI. The bias is larger for ODI3% (7.1 events/h) than for ODI4% (4.4 events/h). More importantly, the agreement between devices was poor, with wide 95% limits of agreement, particularly for ODI3%. These differences have clinical relevance, with significantly more patients crossing potential diagnostic thresholds for OSA and moderate-severe OSA with the AL compared to Comp. Oximetry is a key signal measured during polysomnography and is usually considered the most robust and reproducible signal. As a result, oximetry is the cornerstone of type 4, and many type 3, limited-channel home sleep test devices. The ODI is commonly used in some countries as a metric for diagnosing OSA 8 13 and/or as a marker for cardiovascular risk Implicit in the clinical utility of ODI is that the result is both accurate and reproducible. However, previous literature assessing the performance characteristics of individual oximeters has demonstrated that due to differences in internal signal processing, there is significant variability in the SpO 2 values obtained during simulated OSA. 29 Although an ODI from individual oximeters has been validated as a rule in test for the diagnosis of moderate-severe OSA, 10,13 there have been negligible data comparing the ODIs between oximeters. To date, the only previous study that directly compared ODI between devices is that by Zou et al. 30 In this study, the investigators compared the Watch_PAT 100 (Itamar Medical Ltd., Caesarea, Israel) to unattended in-home polysomnography using the Embla A10 system (Medcare, Reykjavik, Iceland). As part of their broader analysis, the ODI4% automatically calculated by these two devices was compared. Similar to the current study, they demonstrated that despite good correlation there was poor agreement Journal of Clinical Sleep Medicine, Vol. 13, No. 4,

5 between devices. The Watch_PAT 100 tended to score higher ODI4% (particularly in mild-moderate OSA), with a bias of 4.4 (± 6.5) events/h, a bias which is identical to that observed in the current study. Importantly, the wide limits of agreement ( 5.8 to 14.6 events/h) in our study highlights that clinicians cannot be confident that an ODI4% recorded in the AL is the same as that recorded in the Comp. The variability was even greater for ODI3%. In this case the bias was for the AL to record a higher ODI3% than the Comp by an average of 7.1 events/h. The limits of agreement were even wider ( 6.4 to 20.6 events/h). These findings are of importance to clinicians. Mean difference and limits of agreement between ODI were shown to cross diagnostic and severity thresholds for any OSA, and particularly for moderate-severe OSA. This is highlighted when one considers the proportion of patients who receive a diagnosis of either OSA (based on ODI3% 5 events/h) or moderatesevere OSA (based on ODI3% 15 events/h) with the ODI from either device. The current work demonstrates that significantly more patients would receive a diagnosis of OSA, or more particularly, moderate-severe OSA with the AL ODI compared to the Comp ODI. Furthermore, when compared to the gold standard of AHI for diagnosing moderate-severe OSA, at a threshold of ODI3% 15 events/h the AL ODI3% had a significantly higher sensitivity (96.4%) compared to Comp ODI3% (58.2%). This finding appears to partly contradict the findings of a previous study by Ward and colleagues. 31 Contrary to our results, they found that the AL was less sensitive than the Comp at diagnosing OSA. However, there are several important differences between their study and ours. Most importantly, instead of ODI, they compared a different metric, the AHI, generated by both the AL and Comp, and found that the AL tended to underestimate the Comp AHI (mean difference 13.5 events/h during simultaneous recording). This is likely due to the different denominator used to calculate the AHI (monitoring time for the AL and sleep time for the Comp), leading to a systematically lower AHI for the AL, particularly for patients with poor sleep efficiency. In contrast, we used recording time instead of sleep time ODI in our analysis to ensure the denominator for ODI calculation was the same for both the AL and Comp. In our study, based on the calculated odds ratio, a patient has 2.9 times greater odds of a diagnosis of moderate-severe OSA if an AL oximeter is used compared to a Comp oximeter. Differences as large as this are likely to affect clinical decisionmaking about whether to pursue further evaluation (eg, in the patient with a high pretest probability of OSA but normal ODI) and when to implement treatment. There are three potential reasons why oximeters could give different ODI values: (1) oximeter acquisition and internal processing, (2) patient factors related to the use of different fingers, or (3) algorithm differences in the way desaturation events are defined. Pulse oximeters rely on light-emitting diodes, light sensors, and the differing light absorption characteristics of oxygenated and deoxygenated hemoglobin. 32 Multiple stages of signal processing are required to produce a SpO 2. These include: light and wavelength measurement characteristics, artefact rejection, sampling frequency, signal averaging times, accuracy and reproducibility. Manufacturers typically validate oximeters to an accuracy of ± 2%. All of these may influence both the baseline SpO 2 and the temporal characteristics of any change in SpO Signal averaging time is likely to have the greatest potential effect in the setting of OSA, as apneas and hypopneas lead to short-lived oxygen desaturations, 33,34 and longer signal averaging may therefore underestimate the extent of oxygen desaturation during a respiratory event. AHI is known to vary significantly with different signal averaging times. 33,34 With respect to patient factors, the variable most likely to affect SpO 2 is poor peripheral tissue perfusion, which could potentially differ between fingers or hands. Poor perfusion reduces the pulsatile flow in the finger and hence the accuracy of measured SpO Finally, the ODI calculation algorithm is important. Just as AHI will vary according to the scoring criteria used, 4,5 the rules used to score an oxygen desaturation event will also influence the ODI. Key factors include the SpO 2 level used as baseline prior to the dip and rise in SpO 2, but also the duration of SpO 2 required for a desaturation to be recognized by the algorithm. Other temporal characteristics such as the rate of SpO 2 decline could affect the ODI. Each oximeter and/or sleep system manufacturer uses its own algorithm to calculate ODI and there is no standardized set of rules in place to guide manufacturers or clinicians, such as there is with AHI. 35 Moreover, the algorithm used by each device is not readily available for clinicians to refer to when interpreting ODI data. All of these factors may lead to important differences in ODI between oximeters and future research is needed to address how some of these factors differ between devices. A key strength of our study is that our results demonstrate that the differences in ODI between the AL and the Comp are likely due to oximeter acquisition and processing factors, not patient-related factors or ODI calculation algorithms. We exported the oximeter data from the AL into the Comp software. ODI3% and ODI4% were then generated using the Comp software and compared to the ODIs measured and calculated by the AL software. In this situation, there was not only an extremely strong correlation between ODIs (r 2 = 0.99, see Figure 2C and Figure 2D), but the Bland-Altman analysis demonstrated excellent agreement. There was also no clinically significant bias between software systems, and 95% limits of agreement were extremely narrow (see Figure 3C and Figure 3D). These results mean that there was no significant effect of the ODI calculation algorithm to explain the variability seen in our primary analysis, rather that the data including any prealgorithm processing was the difference. Although it is possible for the differences in ODI to be due to the use of separate fingers and arms, we think this is unlikely given the consistent bias seen when AL ODI was compared with Comp ODI in Bland-Altman analysis. If random patient effects such as finger tissue perfusion were to be the cause, one would expect minimal bias as the variability would extend in either direction, but wide limits of agreement (although note that we did not use a formal randomization process in selecting fingers for pulse oximetry monitoring). Moreover, oximeter processing factors are the likely explanation for our results as the processing parameters between the AL and the Comp are known to be different. The AL samples and records SpO 2 data at 1 Hz and uses a 3 second signal averaging time to create the 603 Journal of Clinical Sleep Medicine, Vol. 13, No. 4, 2017

6 final output value. In contrast, the Comp samples SpO 2 every heartbeat. The most recent seven samples are used to calculate the SpO 2 output value; the highest and lowest values in that block of data are excluded and the remaining five data points are averaged to provide the final SpO 2 data point. In addition, how each device determines and rejects artefact (as compared to a valid signal) is not known, but each is unlikely to deal with artefact in the same way. Our study has a few limitations. Although we demonstrated that the ODI generated by the AL and the Comp are different, we cannot say which system produces a more accurate ODI. There is no universally accepted gold standard for either oximeter processing or ODI algorithm calculation, and there is not even an accepted standard for manual scoring of ODI. The most important limitation, however, is that our results relate only to these two devices and their respective software, and results cannot be generalized to other oximeters and sleep diagnostic systems. Nevertheless, the important take-home message from our study is that ODI values cannot be directly compared between patients, unless one is comparing data acquired with the same oximeter and software. Clinicians need to know how an oximeter processes SpO 2 data and how an ODI is calculated, and whether there is a bias for one oximeter to overestimate or underestimate an ODI compared to the other. It is also important that other comparative studies such as ours are performed with other manufacturers equipment and software, and hopefully this study will stimulate further research into the area. In summary, our study has demonstrated that there is a clinically significant difference in ODI measured by two common sleep diagnostic systems the ResMed ApneaLink Plus device (AL) and the Compumedics Grael Profusion PSG3 system (Comp). There is a bias for the AL to report higher ODI values, both for ODI3% and ODI4%, but with wide limits of agreement. The differences are large enough to significantly affect diagnostic thresholds for OSA and, in particular, moderate-severe OSA. The differences are likely the result of signal processing rather than patient factors or manufacturer algorithms for scoring desaturations. This has implications for how clinicians use ODI as a diagnostic tool, or as a marker of cardiovascular risk in OSA. Caution is advised when comparing ODI between patients or when performing posttreatment reassessment in the same patient, unless the same oximeter and software algorithm have been used. ABBREVIATIONS AHI, apnea-hypopnea index AL, ResMed ApneaLink Plus device AL-Comp ODI3%, Compumedics ODI3% using ApneaLink Plus oximetry data AL-Comp ODI4%, Compumedics ODI4% using ApneaLink Plus oximetry data AL ODI3%, ApneaLink Plus ODI3% using ApneaLink Plus oximetry data. AL ODI4%, ApneaLink Plus ODI4% using ApneaLink Plus oximetry data CI, confidence interval Comp, Compumedics Profusion PSG3 system Comp ODI3%, Compumedics ODI3% using Compumedics oximetry data Comp ODI4%, Compumedics ODI4% using Compumedics oximetry data EDF, European Data Format ODI, oxygen desaturation index ODI3%, oxygen desaturation index for 3% desaturation ODI4%, oxygen desaturation index for 4% desaturation OSA, obstructive sleep apnea s, standard deviation of the differences between measurements by the two methods SD, standard deviation SpO 2, oxygen saturation REFERENCES 1. Peppard PE, Young T, Barnet JH, Palta M, Hagen EW, Hla KM. Increased prevalence of sleep-disordered breathing in adults. Am J Epidemiol. 2013;177(9): Heinzer R, Vat S, Marques-Vidal P, et al. Prevalence of sleep-disordered breathing in the general population: the HypnoLaus study. Lancet Respir Med. 2015;3(4): Berry RB, Budhiraja R, Gottlieb DJ, et al. Rules for scoring respiratory events in sleep: update of the 2007 AASM Manual for the Scoring of Sleep and Associated Events. J Clin Sleep Med. 2012;8(5): Ruehland WR, Rochford PD, O Donoghue FJ, Pierce RJ, Singh P, Thornton AT. The new AASM criteria for scoring hypopneas: impact on the apnea hypopnea index. Sleep. 2009;32(2): Duce B, Milosavljevic J, Hukins C. The 2012 AASM respiratory event criteria increase the incidence of hypopneas in an adult sleep center population. J Clin Sleep Med. 2015;11(12): Collop N, Anderson W, Boehlecke B, et al. Clinical guidelines for the use of unattended portable monitors in the diagnosis of obstructive sleep apnea in adult patients. J Clin Sleep Med. 2007;3(7): Flemons WW, Littner MR, Rowley JA, et al. Home diagnosis of sleep apnea: a systematic review of the literature: an evidence review cosponsored by the American Academy of Sleep Medicine, the American College of Chest Physicians, and the American Thoracic Society. Chest. 2003;124(4): Netzer N, Eliasson AH, Netzer C, Kristo DA. Overnight pulse oximetry for sleep-disordered breathing in adults: a review. Chest. 2001;120(2): Chiner E, Signes-Costa J, Arriero JM, Marco J, Fuentes I, Sergado A. Nocturnal oximetry for the diagnosis of the sleep apnoea hypopnoea syndrome: a method to reduce the number of polysomnographies? Thorax. 1999;54(11): Chung F, Liao P, Elsaid H, Islam S, Shapiro CM, Sun Y. Oxygen desaturation index from nocturnal oximetry: a sensitive and specific tool to detect sleepdisordered breathing in surgical patients. Anesth Analg. 2012;114(5): Dawson A, Loving RT, Gordon RM, et al. Type III home sleep testing versus pulse oximetry: is the respiratory disturbance index better than the oxygen desaturation index to predict the apnoea-hypopnoea index measured during laboratory polysomnography? BMJ Open. 2015;5(6):e Kunisaki KM, Bohn OA, Wetherbee EE, Rector TS. High-resolution wrist-worn overnight oximetry has high positive predictive value for obstructive sleep apnea in a sleep study referral population. Sleep Breath. 2016;20(2): Ling IT, James AL, Hillman DR. Interrelationships between body mass, oxygen desaturation, and apnea-hypopnea indices in a sleep clinic population. Sleep. 2012;35(1): Hang L-W, Wang H-L, Chen J-H, et al. Validation of overnight oximetry to diagnose patients with moderate to severe obstructive sleep apnea. BMC Pulm Med. 2015;15:24. Journal of Clinical Sleep Medicine, Vol. 13, No. 4,

7 15. Tkacova R, McNicholas WT, Javorsky M, et al. Nocturnal intermittent hypoxia predicts prevalent hypertension in the European Sleep Apnoea Database cohort study. Eur Respir J. 2014;44(4): Marin JM, Carrizo SJ, Vicente E, Agusti AGN. Long-term cardiovascular outcomes in men with obstructive sleep apnoea-hypopnoea with or without treatment with continuous positive airway pressure: an observational study. Lancet. 2005;365(9464): Gottlieb DJ, Yenokyan G, Newman AB, et al. Prospective study of obstructive sleep apnea and incident coronary heart disease and heart failure: the Sleep Heart Health Study. Circulation. 2010;122(4): Punjabi NM, Newman AB, Young TB, Resnick HE, Sanders MH. Sleepdisordered breathing and cardiovascular disease: an outcome-based definition of hypopneas. Am J Respir Crit Care Med. 2008;177(10): McEvoy RD, Antic NA, Heeley E, et al. CPAP for prevention of cardiovascular events in obstructive sleep apnea. N Engl J Med. 2016;375(10): Bridevaux PO, Fitting JW, Fellrath JM, Aubert JD. Inter-observer agreement on apnoea hypopnoea index using portable monitoring of respiratory parameters. Swiss Med Wkly. 2007;137(43-33): Berry RB, Brooks R, Gamaldo CE, et al.; for the American Academy of Sleep Medicine. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications. Version 2.0. Darien, IL: American Academy of Sleep Medicine; Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476): Bland JM. How can I decide the sample size for a study of agreement between two methods of measurement? John Martin Bland personal website. www-users.york.ac.uk/~mb55/meas/sizemeth.htm. Updated January 12, Accessed January 28, Baguet JP, Hammer L, Levy P, et al. The severity of oxygen desaturation is predictive of carotid wall thickening and plaque occurrence. Chest. 2005;128(5): Hayashi M, Fujimoto K, Urushibata K, Uchikawa S, Imamura H, Kubo K. Nocturnal oxygen desaturation correlates with the severity of coronary atherosclerosis in coronary artery disease. Chest. 2003;124(3): Mooe T, Franklin KA, Holmstrom K, Rabben T, Wiklund U. Sleep-disordered breathing and coronary artery disease: long-term prognosis. Am J Respir Crit Care Med. 2001;164(10 Pt 1): Mooe T, Rabben T, Wiklund U, Franklin KA, Eriksson P. Sleep-disordered breathing in men with coronary artery disease. Chest. 1996;109(3): Yamauchi M, Nakano H, Maekawa J, et al. Oxidative stress in obstructive sleep apnea. Chest. 2005;127(5): Bohning N, Schultheiss B, Eilers S, Penzel T, Bohning W, Schmittendorf E. Comparability of pulse oximeters used in sleep medicine for the screening of OSA. Physiol Meas. 2010;31(7): Zou D, Grote L, Peker Y, Lindblad U, Hedner J. Validation a portable monitoring device for sleep apnea diagnosis in a population based cohort using synchronized home polysomnography. Sleep. 2006;29(9): Ward KL, McArdle N, James A, et al. A comprehensive evaluation of a twochannel portable monitor to rule in obstructive sleep apnea. J Clin Sleep Med. 2015;11(4): Pretto JJ, Roebuck T, Beckert L, Hamilton G. Clinical use of pulse oximetry: official guidelines from the Thoracic Society of Australia and New Zealand. Respirology. 2014;19(1): Davila DG, Richards KC, Marshall BL, et al. Oximeter performance: the influence of acquisition parameters. Chest. 2002;122(5): Farre R, Montserrat JM, Ballester E, Hernandez L, Rotger M, Navajas D. Importance of the pulse oximeter averaging time when measuring oxygen desaturation in sleep apnea. Sleep. 1998;21(4): Berry RB, Brooks R, Gamaldo CE, et al.; for the American Academy of Sleep Medicine. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications. Version 2.2. Darien, IL: American Academy of Sleep Medicine; ACKNOWLEDGMENTS The authors thank all the patients who participated in this study and the sleep scientists at Monash Health. SUBMISSION & CORRESPONDENCE INFORMATION Submitted for publication September 7, 2016 Submitted in final revised form January 17, 2017 Accepted for publication January 19, 2017 Address correspondence to: A/Prof Garun Hamilton, Department of Lung and Sleep Medicine, Monash Health, 246 Clayton Road, Clayton, Victoria 3168, Australia. Phone Fax garun.hamilton@ monashhealth.org DISCLOSURE STATEMENT Work for this study was performed at Monash Health, Academic University Hospital. The ApneaLink Plus devices used in this study were loaned for use by ResMed (Sydney, Australia). Dr. Bradley Edwards is supported by the National Health and Medical Research Council (NHMRC) of Australia s CJ Martin Overseas Biomedical Fellowship ( ). Dr. Shane Landry and Dr. Yvonne Ng are supported by the NHMRC Centre of Research Excellence (CRE), NeuroSleep. A/Prof Garun Hamilton has received equipment to support research from ResMed, Philips Respironics and Air Liquide. A/Prof Darren Mansfield has received research support from Rhinomed Pty Ltd and Fisher & Paykel. Anthony Turton has received consultancy fees from Compumedics Ltd. The other authors have indicated no financial conflicts of interest. 605 Journal of Clinical Sleep Medicine, Vol. 13, No. 4, 2017