Article Text
PDF
Abstract
Introduction Nocturnal pulse oximetry can be used to screen for obstructive sleep apnoea (OSA) using the McGill Oximetry Score (MOS). The MOS has a time threshold for a technically adequate study of 6 hours. It has been suggested that one night of oximetry is sufficient to screen for OSA using the MOS.
Aims (1) To evaluate night-to-night variation of the MOS. (2) To determine the impact of recording three nights of oximetry on the screening yield for OSA. (3) To explore whether useful MOS data are discarded when a threshold of 6 hours of oximetry recording is used.
Methods A retrospective study of nocturnal pulse oximetry done at home over three consecutive nights in paediatric patients with suspected OSA. Studies were scored (MOS) using thresholds of ≥6 and ≥4 hours of recording.
Results A total of 329 patients were studied. MOS scores over three nights showed only fair to moderate agreement. On the first night 126 patients (38%) screened positive for OSA. When three nights of oximetry were done 195 patients (59%) screened positive on at least one of the nights. There were 48 patients with studies of between 4 and 6 hours duration on one or more nights. If these studies are scored 20 patients (42%) would screen positive for OSA on at least one night based on scoring these studies alone.
Conclusion One night of oximetry screening may not be sufficient to screen for OSA. Lowering the time threshold to ≥4 hours may increase the screening capability of nocturnal oximetry.
- respiratory
- sleep
Statistics from Altmetric.com
What is already known on this topic?
Nocturnal pulse oximetry is an alternative to screen for obstructive sleep apnoea (OSA) when polysomnography is not available.
It has been suggested that one night of nocturnal pulse oximetry is sufficient in order to diagnose OSA.
The McGill Oximetry Score is a recognised score for estimating the severity of OSA but requires a minimum of 6 hours of recording duration.
What this study adds?
There is significant night-to-night variability in screening for OSA using nocturnal pulse oximetry.
Performing three nights of oximetry instead of one increases the screening yield for OSA.
OSA can be detected in studies of between 4 and 6 hours duration.
Introduction
Overnight polysomnography (PSG) in a sleep laboratory is the gold standard diagnostic tool for obstructive sleep apnoea (OSA). Nocturnal pulse oximetry is an alternative to screen for OSA when PSG is not available and is low cost and easy to use.1 2 Brouillette et al demonstrated that abnormal oximetry, based on cyclical clusters of oxygen desaturations, has a high positive predictive value but a low negative predictive value for detecting OSA compared with PSG.3 Nixon et al subsequently used this to develop the McGill Oximetry Score (MOS) as a classification of the severity of OSA.4
A previous study comparing oximetry studies done over two consecutive nights using the MOS showed very good consistency—suggesting that one night of pulse oximetry may be sufficient to screen for OSA.5 PSG has been shown to have significant night-to-night variation when two consecutive nights were compared.6 7 This has been termed the ‘first night effect’.8 This may also apply to nocturnal pulse oximetry. Further research using the MOS needs to be done before accepting that a single night study is adequate to screen for OSA.
The MOS was developed to evaluate patients referred for investigation of possible OSA secondary to adenotonsillar hypertrophy.4 9 It has also been used to evaluate OSA in patients with Down syndrome (DS).10 However, there is limited evidence of it being used in other conditions that have an increased incidence of OSA such as Prader-Willi syndrome and Pierre Robin sequence.
The MOS uses a minimum recording duration of at least 6 hours.3 Other studies reporting the interpretation of oximetry recordings have used a minimum duration of 5 hours.11 12 More recently ≥3 hours of artefact-free (AF) oximetry has been used to assess myofunctional training on OSA in children with DS.13 The absolute minimum duration for a sleep recording to get useful data is currently unknown.14
Local practice in the Royal Belfast Hospital for Sick Children (RBHSC) has been to perform three consecutive nights of home pulse oximetry to screen for OSA. The aims of this study are:
To evaluate the extent of night-to-night variation in the MOS over three consecutive nights in otherwise healthy children and in those with a condition predisposing to OSA.
To determine the impact of recording three nights of oximetry instead of one on the screening yield for OSA.
To explore whether useful MOS data are discarded when a threshold of 6 hours is used.
Methods
Design
A retrospective observational study evaluating night-to-night variation of at-home nocturnal pulse oximetry over three consecutive nights was conducted.
Patients and setting
Patients under 18 years old referred to the paediatric respiratory clinic in RBHSC with clinically suspected OSA between December 2010 and November 2016 were included. Patients were divided into three groups: group 1 included otherwise healthy children with tonsillar±adenoidal hypertrophy; group 2 comprised children with DS; group 3 included patients with conditions other than DS that have an increased risk of OSA (eg, Prader-Willi syndrome). The study was part of a quality improvement project and ethics committee approval was not required.
Nocturnal pulse oximetry
Parents attended the RBHSC respiratory clinic and were given a motion-resistant oximeter—either the Nonin 9600 Avant Digital Pulse Oximeter (Nonin Medical, Plymouth, MN, USA) or the Masimo Radical 5 Pulse Oximeter (Masimo, Irvine, CA). They were provided with verbal and written instruction on how to use it for three consecutive nights at home. The sensor probe was to be attached to the child’s finger or big toe and they were asked to remove it if the child awakens. Parents were also asked to keep a sleep record documenting the time of sleep onset, morning wake-up time and any night-time awakenings. Continuous AF trace that aligned with the sleep record was assumed to be recording sleep. Recordings were analysed using nVision software (Nonin Medical) for the Nonin 9600 Avant and Visi-Download (Stowood Scientific, Oxford, UK) for the Masimo Radical 5. Poor perfusion, low signal and sensor displacement artefact were extracted. Oxygen saturations (SpO2) were averaged over a maximum of 3 s increments. Length of time of AF trace (hours) and MOS were recorded. Pulse oximetry studies were independently reviewed by three researchers (NCG, BM, DOD) and data were entered into an anonymised study database.
The MOS scores the severity of OSA. A positive oximetry recording for OSA has ≥3 desaturation clusters and at least three SpO2 drops to <90% (MOS 2).3 Increasingly severe OSA is scored as MOS 3 or 4 if there are ≥3 clusters of desaturation and at least three SpO2 drops to <85% and 80%, respectively. If the recording is normal or inconclusive (<3 desaturation clusters and <3 SpO2 drops to <90%) it is scored MOS 1. A technically adequate pulse oximetry study has been defined as one with ≥6 hours of continuous AF data.4
An additional subanalysis of MOS scoring was performed on studies with 4–6 hours of continuous AF data. These studies would otherwise have been rejected as technically inadequate.
Statistical analysis
Statistical analysis was performed using Stata V.15 (Stata Statistical Software: Release 15, StataCorp, College Station, Texas) and StatsDirect V.3 (StatsDirect Statistical Software, http://www.statsdirect.com). Cohen’s kappa (κ) statistic with 95% CIs was used to express the level of agreement between nights with κ >0.6 indicating good agreement. Quantitative agreement statistics were not calculated given the categorical nature of the MOS. Normally distributed data were expressed as the mean±SD. Data which were not normally distributed were expressed as the median with IQRs.
Results
Nocturnal home oximetry was performed on 329 children (60% male) with median age 4 years (IQR: 2–7 years) in the study time frame. Group 1 consisted of 204 patients (62%), group 2 had 35 (11%) patients and group 3 had 82 (25%) patients. Clinical information was missing from eight patients and these patients were not assigned a group. There were 126 patients (38%) with ≥6 hours of continuous AF oximetry trace on night 1 who had MOS ≥2 suggesting possible OSA. There were 195 patients (59%) with ≥6 hours of continuous AF oximetry trace (on one or more nights) who had MOS ≥2 on at least one of the nights.
Trace duration
The mean AF trace duration on night 1 was 8.7 hours, on night 2 was 8.1 hours and on night 3 was 7.8 hours. On nights 1, 2 and 3 the numbers of patients with <6 hours of AF traces (out of the total 329 patients) were 53 (16%), 69 (21%) and 76 (23%), respectively. There were 19 (6%) patients with <6 hours of AF tracings on all three nights.
Taking three consecutive nights of oximetry reduced the numbers who would have had <6 hours of AF traces if only one night (taken as the first night) had been performed from 53 (16%) to 19 (6%).
Night-to-night variability
MOS agreement between nights 1, 2 and 3 for the whole study group is tabulated in online supplementary appendix 1. Nights 1 and 2, 2 and 3, and 1 and 3 were compared for the whole study group. Patients were excluded from this and subsequent analyses if they did not have technically adequate traces on both nights being compared.
Supplemental material
Comparing nights 1 and 2, two hundred and thirty-six patients showed fair agreement with κ=0.40 (95% CI 0.32 to 0.49).
Comparing nights 2 and 3, two hundred and twenty-three patients showed fair agreement with κ=0.39 (95% CI 0.31 to 0.48).
Comparing nights 1 and 3, two hundred and twenty-seven patients showed moderate agreement with κ=0.42 (95% CI 0.34 to 0.51).
Night-to-night variability of group 1
There were 204 otherwise healthy patients with tonsillar±adenoidal hypertrophy in this group and MOS agreement between nights 1, 2 and 3 is tabulated in online supplementary appendix 1. Nights 1 and 2, 2 and 3, and 1 and 3 were compared.
Comparing nights 1 and 2, one hundred and fifty-two patients showed moderate agreement with κ=0.41 (95% CI 0.3 to 0.52).
Comparing nights 1 and 3, one hundred and forty-six patients showed moderate agreement with κ=0.48 (95% CI 0.37 to 0.6).
Comparing nights 1 and 3, one hundred and forty-two patients showed fair agreement with κ=0.32 (95% CI 0.21 to 0.43).
Night-to-night variability of group 2
There were 35 patients with DS in this group and MOS agreement between nights 1, 2 and 3 is tabulated in online supplementary appendix 1. Nights 1 and 2, 2 and 3, and 1 and 3 were compared.
Comparing nights 1 and 2, twenty-seven patients showed moderate agreement with κ=0.42 (95% CI 0.17 to 0.66).
Comparing nights 1 and 3, twenty-four patients showed fair agreement between night 1 and night 3 with κ=0.36 (95% CI 0.1 to 0.61).
Comparing nights 1 and 3, twenty-six patients showed moderate agreement with κ=0.47 (95% CI 0.24 to 0.69).
Night-to-night variability of group 3
There were 82 patients with other predisposing factors to OSA in this group. A table of the diagnoses and MOS scores is shown in online supplementary appendix 2. MOS agreement between nights 1, 2 and 3 is tabulated in online supplementary appendix 1. Nights 1 and 2, 2 and 3, and 1 and 3 were compared.
Supplemental material
Comparing nights 1 and 2, fifty-three patients showed fair agreement with κ=0.31 (95% CI 0.14 to 0.5).
Comparing nights 1 and 3, fifty-three patients showed fair agreement with κ=0.27 (95% CI 0.1 to 0.44).
Comparing nights 1 and 3, fifty-one patients showed moderate agreement with κ=0.46 (95% CI 0.28 to 0.64).
MOS scoring of oximetry studies with 4–6 hours of AF trace
There were 48 patients who had at least one oximetry study with 4–6 hours of AF recording on at least one of the three nights. If ≥4 hours of AF recording is deemed acceptable then 20 (42%) of these patients would potentially screen positive for OSA on one night based on MOS scoring of these traces alone. Of these 20 patients, 11 would have an MOS of 2, four would have an MOS of 3 and five would have an MOS of 4 (online supplementary appendix 3). These traces would normally be considered technically inadequate and be rejected.
Supplemental material
Discussion
The gold standard investigation to diagnose OSA is PSG but access is limited and hence there is a need to optimise the screening capabilities of other investigations such as pulse oximetry. This study has shown that there is significant variability in the night-to-night pulse oximetry recordings of children when using the MOS to screen for OSA. Paediatric PSGs have been shown to exhibit night-to-night variability in sleep architecture with patients having more rapid eye movement (REM) sleep on night 2 than night 1.6 OSA predominantly occurs in children during REM sleep and a susceptible patient is more likely to have the REM-related cyclical desaturations associated with OSA on subsequent nights rather than the first night.6 7 A previous study comparing two consecutive nights of oximetry in otherwise healthy children reported very good ‘consistency’.5 However, this study analysed the correlation rather than the agreement between nights.
We have demonstrated that, by using three consecutive nights of pulse oximetry instead of one, there is a reduction from 16% to 6% of patients having no recordings of adequate duration for analysis. In addition, 59% of patients screened positive for OSA on at least one of the three nights compared with 38% if only one night had been performed. Local practice has been to perform three consecutive nights of home oximetry to reduce the chances of having to do repeat studies if there were insufficient data on the first night. However, if oximetry is abnormal on the first night there is no requirement to do further recordings on subsequent nights. To streamline this decision-making, it is anticipated that the development of Bluetooth oximeters, with the ability to upload overnight studies to a cloud-based repository using a smartphone, will allow healthcare professionals to view recordings remotely on completion of the first night.15 A decision could then be made on whether further nights are required.
Children with DS have an increased risk of OSA with a reported prevalence between 31% and 79%.16–18 The MOS has been shown to be applicable to patients with DS.10 Indeed, a systematic review has suggested that nocturnal pulse oximetry may be the preferred method for diagnosing OSA in patients with DS as they are more likely to display limited cooperation with PSG.19 In the present study, fair to moderate agreement in the MOS was shown over three consecutive nights in patients with DS.
The agreement in MOS between nights for patients with comorbidities that may make them more susceptible to OSA (group 3) showed fair to moderate agreement. Interestingly, a recent study suggests that the positive predictive value of the MOS may be lower in this group due to non-obstructive conditions causing desaturations.20
The lower time threshold required to score nocturnal pulse oximetry tracings has been taken as 6 hours.4 It has been recognised that further work needs to be done to determine the minimum recorded sleep time for children.14 In the present study, reducing the threshold duration for technically adequate oximetry traces to a minimum of 4 hours increased the number of patients with an MOS ≥2 who would have been screened positive for OSA. As the specificity of the MOS as a screening test for OSA is high, this may suggest that an abnormal study (MOS ≥2) could be considered as being positive for OSA. However, as the sensitivity of the MOS is limited, a study of 4–6 hours with a MOS 1 would not provide any additional useful information.
There are a number of limitations to this study. It is a retrospective study and it is recognised that further prospective studies are needed to confirm the study findings. The averaging time of the Masimo Radical 5 pulse oximeter is 2 s and the averaging time of the Nonin 9600 Avant Digital pulse oximeter is 3 s. The pulse oximeters were used in approximately equal numbers. Ideally all patients would have been studied using oximeters with the shorter averaging time. It is known that using oximeters with longer averaging times may underestimate the number of desaturations.21 However, it is unlikely that a 1 s difference in the averaging times will affect the MOS score. Indeed, a comparative study of two oximeters with different averaging times suggested that even a 4 s difference did not affect the classification of severity of OSA.22
Some of the patients in this study had been included in a previous study.23 However, no MOS data on any of these patients have previously been reported.
Conclusion
This study on the use of the MOS to screen for OSA in children suggests that, when it is repeated on three consecutive nights, there is only fair to moderate agreement between MOS scores and that, if only one night of oximetry is performed, the screening yield may be limited. The study also found that OSA can be detected using the MOS in studies of between 4 and 6 hours duration. In order to optimise nocturnal oximetry as a screening test for OSA we suggest that more than one night of nocturnal at home pulse oximetry should be done and that studies ≥4 hours duration should be scored.
References
Footnotes
Contributors NCG collected the data, analysed the data and cowrote the paper. BM collected the data and cowrote the paper. MS helped with study design, analysed the data and cowrote the paper. DOD conceived the study, supervised the data collection and data analysis and cowrote the paper.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.