Sleep and Breathing

, Volume 13, Issue 3, pp 213–219

Evaluation of a portable recording device (ApneaLink™) for case selection of obstructive sleep apnea

  • Hui Chen
  • Alan A. Lowe
  • Yuxing Bai
  • Peter Hamilton
  • John A. Fleetham
  • Fernanda R. Almeida
Original Article

DOI: 10.1007/s11325-008-0232-4

Cite this article as:
Chen, H., Lowe, A.A., Bai, Y. et al. Sleep Breath (2009) 13: 213. doi:10.1007/s11325-008-0232-4

Abstract

Objective

This study was designed to assess the sensitivity and specificity of a portable sleep apnea recording device (ApneaLink™) using standard polysomnography (PSG) as a reference and to evaluate the possibility of using the ApneaLink™ as a case selection technique for patients with suspected obstructive sleep apnea (OSA).

Materials and methods

Fifty patients (mean age 48.7 ± 12.6 years, 32 males) were recruited during a 4-week period. A simultaneous recording of both the standard in-laboratory PSG and an ambulatory level 4 sleep monitor (ApneaLink™) was performed during an overnight study for each patient. PSG sleep and respiratory events were scored manually according to standard criteria. ApneaLink™ data were analyzed either with the automated computerized algorithm provided by the manufacturer following the American Academy of Sleep Medicine standards (default setting DFAL) or The University of British Columbia Hospital sleep laboratory standards (alternative setting, ATAL). The ApneaLink respiratory disturbance indices (RDI), PSG apnea–hypopnea indices (AHI), and PSG oxygen desaturation index (ODI) were compared.

Results

The mean PSG-AHI was 30.0 ± 25.8 events per hour. The means of DFAL-RDI and ATAL-RDI were 23.8 ± 21.9 events per hour and 29.5 ± 22.2 events per hour, respectively. Intraclass correlation coefficients were 0.958 between PSG-AHI and DFAL-RDI and 0.966 between PSG-AHI and ATAL-RDI. Receiver operator characteristic curves were constructed using a variety of PSG-AHI cutoff values (5, 10, 15, 20, and 30 events per hour). Optimal combinations of sensitivity and specificity for the various cutoffs were 97.7/66.7, 95.0/90.0, 87.5/88.9, 88.0/88.0, and 88.2/93.9, respectively for the default setting. The ApneaLink™ demonstrated the best agreement with laboratory PSG data at cutoffs of AHI ≥ 10. There were no significant differences among PSG-AHI, DFAL-RDI, and ATAL-RDI when all subjects were considered as one group. ODI at 2%, 3%, and 4% desaturation levels showed significant differences (p < 0.05) compared with PSG-AHI, DFAL-RDI, and ATAL-RDI for the entire group.

Conclusion

The ApneaLink™ is an ambulatory sleep monitor that can detect OSA and/or hypopnea with acceptable reliability. The screening and diagnostic capability needs to be verified by further evaluation and manual scoring of the ApneaLink™. It could be a better choice than traditional oximetry in terms of recording respiratory events, although severity may be under- or overestimated.

Keywords

Home monitoring Obstructive sleep apnea Polysomnography Oral appliance ApneaLink™ 

Introduction

Obstructive sleep apnea (OSA), a syndrome characterized by repeated episodes of upper airway obstruction during sleep, requires an accurate diagnosis [1, 2]. OSA is common and often associated with disabling symptoms and long-term health consequences. In the employed population, the Wisconsin Sleep Cohort Study revealed that as high as 93% of women and 82% of men with moderate to severe OSA were underdiagnosed [3], and increasing awareness among physicians and the public has led to an increase in the demand for diagnostic sleep monitoring [4].

Overnight polysomnography (PSG) remains the gold standard to diagnose OSA [5, 6] and the reference to which the other types of sleep monitors are compared. The primary advantage of PSG is that it provides detailed neurophysiological and respiratory data. However, PSG studies are labor intensive [7], time consuming, and expensive to perform [8]. Waiting time is another issue that affects the actual effectiveness of PSG diagnosis. Waiting times from a non-urgent referral until a hospital-based PSG has been completed vary widely around the world from 2 weeks to more than 2 years [2]. Moreover, the in-laboratory environment is far different from the patient’s own bedroom, which may affect the validity of PSG data.

These issues have led to the development of unattended ambulatory sleep monitors to assess OSA. A portable recording device for OSA was first systematically examined by the Standards of Practice Committee of the American Sleep Disorders Association (ASDA) in 1993, leading to a series of updated recommendations [9, 10, 11]. According to the practice standards, there are three types of portable devices categorized by their channels and complicacy. Patients can use the device in their homes, and the data could be later downloaded by physicians or technicians. Given the costs of in-laboratory PSG, ambulatory methods for diagnosis are desirable [12]. Various portable systems have been tested for ambulatory or home monitoring without supervision, including respirometer [13], peripheral arterial tonometry [14, 15, 16], blood oximetry [17, 18], or blood pressure recordings [19]. The different devices revealed different clinical usage and reliabilities. Ideally, these devices should be easily operated with acceptable accuracy, enough memory storage capability, and not disturb the sleep of the users.

The purpose of this study was to evaluate the sensitivity and specificity of ApneaLink™ (ResMed Corporation, Poway, CA, USA) compared to a standard PSG and further speculate its usage as a case selection tool for patients with suspected OSA. ApneaLink™ records airflow via a nasal pressure signal, and it is considered an ambulatory level 4 sleep monitor.

Materials and methods

Fifty-four patients with suspected sleep disordered breathing seen at The University of British Columbia (UBC) Hospital sleep disorder program were recruited consecutively during a 4-week time period. Patients scheduled for a follow-up study undergoing any therapy were excluded. The split-night studies were also excluded since they are only prescribed for out-of-province patients or for patients with previous continuous positive airway pressure experience at UBC Hospital. The study protocol was approved by the UBC Ethics Committee. Each participant was asked to sign a written informed consent form before the initial recording.

One female declined to take part in the project. Three other patients revealed extremely short records due to technical problems and were excluded from the final analysis. There were a total of fifty patients (aged 19–75 years, mean age 48.7 ± 12.6 years) with valid data, including 18 women (mean age 49.7 ± 14.7 years) and 32 men (mean age 48.2 ± 11.4 years).

The portable home monitoring device we used for the designed study was called ApneaLink™, which is powered by two 1.5-V AA batteries and fixed on to the user’s chest with the supplied belt. The two nasal tubes of the nasal pressure cannula are inserted into the user’s nostrils and looped around both ears. The ApneaLink™ records the patient’s nasal airflow through the cannula and generates data through its built-in pressure transducer. Once the user presses the start button, ApneaLink™ is capable of monitoring breathing patterns and measures apneas or hypopneas as well as flow limitation, snoring sounds, and inspiratory flow. The downloading and analysis system requires a Pentium III 500 MHz or equivalent computer with 128 MB RAM and 500 MB hard disk space. The installation program (version 5.26) downloads the signals from the recorder and automatically analyzes the data based on pre-set analysis parameters. The settings can be changed, and the data can be reanalyzed. ApneaLink™ also provides manual checking or editing so that automated scoring could be optimized with a manual review.

The overnight data are saved in the main memory of the ApneaLink™. Users can record information, including their height, weight, and BMI and log the time they began using ApneaLink™ and the time the study ended. They can also record and delete the time period they are awake. However, in this study, we did not add such information in the analysis. The ApneaLink™ generates a report with a color-coded Risk Indicator for clinical review.

The ApneaLink™ was connected to one end of a Y-shaped nasal cannula. The other peripheral end of the Y-shaped cannula was directly connected to a pressure transducer (the Ultima Airflow Pressure Sensor™, BRAEBON Medical Corporation, Ontario, Canada) to enable nasal pressure to be recorded by both devices simultaneously.

The PSG data were analyzed manually according to the American Academy of Sleep Medicine (AASM) criteria by polysomnographic technologists blinded to the Apnea-Link™ data. Oxygen desaturation index (ODI), the number of desaturations yielding a certain percentage fall in oxygen saturation per hour of recording, was also obtained from the overnight PSG at 2%, 3%, and 4% cutoffs. Apnea and hypopnea were scored by the ApneaLink™ system with an automatic respiratory disturbance index (RDI). The default setting of analysis parameters for ApneaLink™ is in accordance with the AASM criteria [5], with apnea defined as a reduction of inspiratory airflow by 80% to 100% over 10 s or maximum duration of 80 s, and hypopnea was defined as a reduction of tidal breathing of 50% from baseline tidal breathing lasting 10 s or a maximum duration of 100 s. This represents the default ApneaLink™ RDI (DFAL-RDI). An alternative setting, which is in accordance with the Sandman® Sleep Diagnostic System setting (Nellcor Puritan Bennett, Pleasanton, CA, USA) at the UBC Hospital sleep laboratory was also used. In this alternative setting, apneas were defined as 85% or more reduction of normal flow that lasted 10 s to a maximum duration of 100 s. Hypopneas are defined as 40% reduction of normal flow lasting 10 s to a maximum duration of 120 s. This is the alternative ApneaLink™ RDI (ATAL-RDI).

Measurement agreement

Statistical analysis was performed using the Statistical Package for Social Sciences program (SPSS, version 13.0, Chicago, IL, USA). Intraclass correlation coefficient (ICC) was used to assess the reliability of ApneaLink™. The difference between PSG (PSG-AHI) and ApneaLink™ (ApneaLink-RDI) was plotted on a Bland–Altman graph for both default (DFAL-RDI) and alternative (ATAL-RDI) settings. The limits of agreement have been defined as ±2 SD. Receiver operating characteristics (ROC) curves were constructed for assessing the sensitivity and specificity of ApneaLink™. Positive predictive values (PPV) and negative predictive values (NPV) were calculated. One-way ANOVA was performed to compare the means between PSG-AHI and ApneaLink-RDI or between ODI and respiratory indexes. The post hoc test (LSD test) was used for multiple comparisons between any two OSA severity groups.

Results

Six patients (12%) did not have OSA (PSG-AHI < 5). Mild OSA (5 ≤ PSG-AHI < 15) was diagnosed in 12 patients (24%), moderate OSA (15 ≤ PSG-AHI < 30) in 15 patients (30%), and severe OSA (PSG-AHI ≥ 30) in 17 patients (34%). Table 1 demonstrates the mean ± SD of the selected demographic variables for the different category groups. The PSG-AHI ranged from 0.6 to 86.3 events/hour, with a mean of 30.0 ± 25.8 events/hour. The ApneaLink-RDI for the default setting ranged from 1.0 to 81.0 events/hour, with a mean of 23.8 ± 21.9 events/hour and a mean of 29.5 ± 22.2 events/hour using the alternative setting.
Table 1

Demographic and sleep characteristics of the 50 subjects studied

 

Non-OSA

Mild OSA

Moderate OSA

Severe OSA

Total

Gender (number), (F/M)

1/5

3/9

10/5

4/13

18/32

Age (years)

37.7 ± 17.2

49.3 ± 13.0

50.7 ± 13.4

50.4 ± 8.1

48.7 ± 12.6

BMI (kg/m2)

28.1 ± 4.2

30.1 ± 7.4

34.8 ± 7.6

32.8 ± 6.8

32.2 ± 7.2

Sleep time (min)

378.6 ± 90.0

345.4 ± 63.9

337.9 ± 72.4

338.1 ± 65.0

344.6 ± 69.2

Sleep efficiency (%)

83.2 ± 16.5

78.1 ± 12.5

78.9 ± 15.2

78.5 ± 10.9

79.1 ± 13.0

No significant differences were found among PSG-AHI, DFAL-RDI, and ATAL-RDI if all the subjects were considered as one group. There were differences (p < 0.05) among ODI (2%, 3%, and 4%), PSG-AHI, DFAL-RDI, and ATAL-RDI for the entire group. When we divided the group into no, mild, moderate, or severe OSA subgroups based on the PSG-AHI, there were no significant differences between PSG-AHI/DFAL-RDI and PSG-AHI/ATAL-RDI. Only the difference between PSG-AHI and ATAL-RDI showed significance in the no OSA group. ODI with 2% (ODI2), 3% (ODI3), or 4% (ODI4) desaturations showed significant differences from AHI or RDI with the exception that in the mild OSA group, there were no significant differences between ODI2 and any other variable. Table 2 demonstrates the detailed results and the significant variables.
Table 2

Comparison of PSG-AHI, ApneaLink-RDI, and ODI for the entire group and OSA severity subgroups

Events/h

PSG-AHI

Default AL-RDI

Alternative AL-RDI

ODI2

ODI3

ODI4

Non-OSA (n = 6)

2.6 ± 1.2

3.2 ± 1.6

7.2 ± 2.3a

10.0 ± 3.9b,c

3.1 ± 2.1d

1.7 ± 1.2d

Mild OSA (n = 12)

10.5 ± 2.7

8.2 ± 3.7

13.3 ± 6.2

11.5 ± 3.4

3.9 ± 1.6b,c,d

4.1 ± 3.4b,c,d

Moderate OSA (n = 15)

21.5 ± 4.5

17.7 ± 10.4

24.3 ± 10.5

15.4 ± 5.5b,d

6.1 ± 2.6b,c,d

9.3 ± 6.9b,c,d

Severe OSA (n = 17)

61.0 ± 19.0

47.4 ± 20.0

53.5 ± 19.0

15.0 ± 9.6b,c,d

11.1 ± 6.0b,c,d

37.8 ± 25.2b,d

Total (n = 50)

30.0 ± 25.8

23.8 ± 21.9

29.5 ± 22.2

13.6 ± 6.8b,c,d

6.8 ± 4.9b,c,d

16.4 ± 21.2b,d

No significant differences between PSG and default AL-RDI

aSignificant difference between PSG and alternative AL-RDI (p < 0.05)

bSignificant difference between PSG-AHI and ODI (p < 0.05)

cSignificant difference between default AL-RDI and ODI (p < 0.05)

dSignificant difference between alternative AL-RDI and ODI (p < 0.05)

The ICCs based on one factor repeated measures ANOVA were 0.958 and 0.966 for the default and alternative settings (Fig. 1). Bland–Altman plots of DFAL-RDI, ATAL-RDI, and PSG-AHI are demonstrated in Fig. 2. The Bland–Altman plot represents the difference between the ApneaLink™ and the PSG against the average value (ApneaLink-RDI + PSG-AHI/2) for each individual patient. A mean difference of −6.3 events/hour, indicating a systematic bias of underscored ApneaLink™ data, was observed from the default setting, and the limits of agreement were 9.6 (1 SD) and 19.2 (2 SD) events/hour. The alternative setting showed a smaller deviation in the mean ± 1 SD of −0.5 ± 8.7 events/hour. The plots also indicated that the ApneaLink™ tended to estimate disease severity properly at lower cutoffs of AHI (<30 events per hour), while it underestimated severity at higher cutoffs of AHI (>60 events per hour).
Fig. 1

Intraclass correlation coefficient. Left ICC between DFAL-RDI (default) and PSG-AHI. Right ICC between ATAL-RDI (alternative) and PSG-AHI

Fig. 2

Bland–Altman Plot with limits of agreement. Left The differences between DFAL-RDI (default) and PSG-AHI plotted against their mean for each subject. Right The differences between ATAL-RDI (alternative) and PSG-AHI plotted against their mean for each subject

ROC curves for the default setting (Fig. 3) with cutoff values of AHI of 5, 10, 15, 20, and 30 demonstrated the different area under curve (AUC) measurements of 0.964, 0.978, 0.944, 0.944, and 0.954 (Fig. 4). Optimum combinations of sensitivity and specificity for default settings were 97.7/66.7, 95.0/90.0, 87.5/88.9, 88.0/88.0, and 88.2/93.9, respectively. For the alternative setting with the same cutoffs, the AUC measurements were 0.951, 0.983, 0.944, 0.934, and 0.955; the optimum combinations of sensitivity and specificity were 93.2/83.3, 97.5/90.0, 90.6/77.8, 92.0/84.0, and 94.1/81.8, respectively. The PPV/NPV for the same cutoff values in default settings were 60%/100%, 50%/100%, 68%/100%, 79%/90%, and 89%/87%, respectively.
Fig. 3

ROC curves with the default settings of ApneaLink

Fig. 4

Area under curves with different PSG-AHI cutoffs (default DFAL-RDI vs. alternative ATAL-RDI)

Discussion

ApneaLink™ is a sleep monitor, which records nasal pressure and is designed to be used unattended at home. In our study, it was used simultaneously with PSG to assess its reliability. When compared to PSG, the ApneaLink™ had a sensitivity of 88.2–97.7% and a specificity of 66.7%–93.9%, with different AHI cutoffs from 5 to 30 events/hour. The best agreement with laboratory PSG data was at cutoffs of AHI ≥ 10. The cutoff AHI of 15 was the best PPV/NPV combination with the PPV of 68% and the NPV of 100%. The ICC was 0.958 for the default settings and 0.966 for the alternative settings. There were no significant differences among PSG-AHI, DFAL-RDI and ATAL-RDI if the patients were considered as one group. There were no significant differences between PSG-AHI/DFAL-RDI and PSG-AHI/ATAL-RDI in the mild, moderate, and severe OSA groups. The only difference was between PSG-AHI and ATAL-RDI in the non-OSA group. This suggests that the default and the alternative settings are both acceptable for different OSA severities. With these results, we believe that the setting for flow analysis needs not to be changed except in special cases, which were not evaluated in the present study. The Bland–Altman plot indicated that a mean underscored 6.26 events per hour for ApneaLink™ might have occurred when using the default setting.

The definition for hypopnea is ≥3% desaturation from pre-event baseline or the event is associated with arousal [20], which is not always associated with desaturations. There have been many studies about the diagnostic ability of oximetry for OSA with rather conflicting results [21, 22]. Our study showed that there were no significant differences between PSG-AHI and ApneaLink-RDI but significant differences between PSG-ODI (desaturation ≥3%, 4%) and PSG-AHI or between PSG-ODI and ApneaLink-RDI. This confirms that the ApneaLink™ demonstrated higher correlation with the PSG if the respiratory disturbance rather than oximetry was focused on to identify OSA patients. While the home oximeter is a good screening tool to concentrate on the oxygen desaturation, ApneaLink™ has probably a higher chance to accurately record hypopneas than home oximetry.

Based on our results and the comparisons of nasal flow and blood oxygen, we believe that to assess case selection and apnea severity, a portable monitor should have at least nasal flow and blood oxygen channels. Simple devices such as ApneaLink™ might be useful for short-term treatment evaluation. For example, we speculated that portable monitors could be used for titration and short-term follow-up of the oral appliance (OA) therapy for OSA. Reports from bed partners are very unreliable in evaluating titration success due to a variety of factors. The introduction of a portable measuring device that shows a good reliability permits a more accurate titration of the OA to improve the quality and consistency of post-treatment success [23]. This could potentially assist the effectiveness of OA therapy and eventually improve current titration protocols.

We did not evaluate the visual-correct manual scoring of the ApneaLink™, and this gives less power to the evaluation of the reliability of the device, in particular from the point of view of the screening and diagnostic capability of this portable monitor. Automated scoring has the advantage that it eliminates variability in manual scoring [7]. The purpose of this study was to assess the automatic analysis ability of ApneaLink™ in scoring apnea and hypopnea events. Even though this machine allows a manual scoring of the events, we assessed only two automatic settings, which were the default (DFAL-RDI) and the alternative (ATAL-RDI). In future studies, it will be important to quantify the improvements of manual scoring of the respiratory events. We did not manually score awake or asleep; instead, we have assessed the respiratory events from ApneaLink™ for the total bed time.

The fundamental premise for this study was that an in-laboratory PSG is the optimal approach for diagnosing OSA. The validation of the ApneaLink™ was mainly dependent on this premise, but several limitations exist. First, the diagnostic patients in our study had no previous experience with sleeping in a laboratory, and most of them did not sleep as well as at home. The current AASM practice parameters for the use of portable monitoring devices also points out that PSG has limitations due to unfamiliar sleep environments [10]. Second, they tended to sleep in a supine position due to the electrodes attached to their head and chest, so the results may vary from those at home. The limitation revealed that results obtained by either a PSG or the ApneaLink™ should always be accompanied by a thorough clinical examination, daytime sleepiness assessment, and a physician’s evidence-based decision.

We studied patients with a wide range of race, gender and OSA severity. We did not arrange for patients to use the ApneaLink™ at home, so its performance and the percentage of data loss without a technician in attendance are unknown. Erman and coworkers [24] investigated the ApneaLink™ in a group of 59 adults with type 2 diabetes mellitus and discovered that home recordings shorter than 4 h led to more frequent false negative results at AHI levels of ≥15. Therefore, they suggested a longer recording period to minimize the false negative results. Wang and coworkers [25] did a similar investigation on the MicroMESAM (previous brand name of ApneaLink™) with a result of 100% sensitivity and 87.5% specificity at an AHI of 10. They also found that the MicroMESAM-generated flow-time curves corresponded well with pneumotachograph-generated curves.

Although all patients included in this study were referred for a PSG because of suspected OSA, the relatively small number of subjects is a limitation of this study. Moreover, the ApneaLink™ is a one-channel type 4 monitor different from multiple level 2 or 3 monitors. It fails to distinguish between central apnea and obstructive apnea because the recordings are based on airflow with no respiratory effort recorded. Therefore, the ApneaLink™ is not indicated for patients with associated overlap of cardiac or respiratory diseases, limiting its value in patients with suspected central sleep apnea. The signal could also produce false positive events if the patient was mouth breathing [7]. Due to such limitations, ApneaLink™ cannot substitute completely for a regular PSG in the diagnosis of OSA. It can only be used for OSA case selection under a physician’s guidance or as an initial assessment for patients to see if a PSG is necessary.

Conclusion

The ultimate purpose of case selection for OSA diagnosis is to treat and improve the quality of life. The baseline and follow-up overnight sleep studies need to be compared to determine whether the patient has benefited from the treatment. This study suggests that the ApneaLink™ could be as a case selection monitor with an acceptable reliability for patients with OSA. Once the risk indicator has reached or exceeded the limit, it is recommended that a diagnostic PSG should be obtained. Further manual scoring of the ApneaLink™ needs to be done to testify the screening and diagnostic capability of this portable monitor.

Acknowledgments

The authors would like to thank Mrs. Ingrid Ellis for her editorial assistance in the final preparation of this manuscript and the polysomnographic technologists in UBC Hospital sleep disorder laboratory. As a postdoctoral fellow, Dr. Hui Chen was supported in part from royalties paid to The University of British Columbia from worldwide Klearway™ sales.

Copyright information

© Springer-Verlag 2008

Authors and Affiliations

  • Hui Chen
    • 1
    • 2
  • Alan A. Lowe
    • 1
  • Yuxing Bai
    • 2
  • Peter Hamilton
    • 3
  • John A. Fleetham
    • 3
  • Fernanda R. Almeida
    • 4
  1. 1.Division of Orthodontics, Department of Oral Health SciencesThe University of British ColumbiaVancouverCanada
  2. 2.Department of Orthodontics, Faculty of StomatologyThe University of Medical SciencesBeijingPeople’s Republic of China
  3. 3.Division of Respiratory Medicine, Department of MedicineThe University of British ColumbiaVancouverCanada
  4. 4.Department of Oral Biological and Medical SciencesThe University of British ColumbiaVancouverCanada

Personalised recommendations