Abstract
Study Objectives:
Thermistors, nasal cannulas, and respiratory inductance plethysmography (RIP) are the recommended reference sensors of the American Academy of Sleep Medicine (AASM) for the detection and characterization of apneas and hypopneas; however, these sensors are not well tolerated by patients and have poor scorability. We evaluated the performance of an alternative method using a combination of tracheal sounds (TSs) and RIP signals.
Methods:
Consecutive recordings of 70 adult patients from the Pays de la Loire Sleep Cohort were manually scored in random order using the AASM standard signals and the combination TS and RIP signals, without respiratory sensors placed on the patient’s face. The TS-RIP scoring used the TS and RIP-flow signals for detection of apneas and hypopneas, respectively, and the suprasternal pressure and RIP belt signals for the characterization of apneas.
Results:
Sensitivity and specificity of the TS-RIP combination were 96.21% and 91.34% for apnea detection and 89.94% and 93.25% for detecting hypopneas, respectively, with a kappa coefficient of 0.87. For the characterization of apneas, sensitivity and specificity were 98.67% and 96.17% for obstructive apneas, 92.66% and 99.36% for mixed apneas, and 96.14% and 98.89% for central apneas, respectively, with a kappa coefficient of 0.94. The TS-RIP scoring revealed a high agreement for classifying obstructive sleep apnea into severity classes (none, mild, moderate, and severe obstructive sleep apnea) with a Cohen’s kappa coefficient of 0.96.
Conclusions:
Compared with the AASM reference sensors, the TS-RIP combination allows reliable noninvasive detection and characterization of respiratory events with a high degree of sensitivity and specificity. TS-RIP combination could be used for diagnosis of obstructive sleep apnea in adults, either as an alternative to the AASM sensors or in combination with the recommended AASM sensors.
Citation:
Sabil AK, Marien C, LeVaillant M, Baffet G, Meslier N, Gagnadpix F. Diagnosis of sleep apnea without sensors on the patient’s face. J Clin Sleep Med. 2020;16(7):1161–1169.
Keywords: sleep apnea diagnosis, respiratory event scoring, tracheal sounds, suprasternal pressure
BRIEF SUMMARY
Current Knowledge/Study Rationale: The American Academy of Sleep Medicine recommends the use of an oronasal thermistor and a nasal cannula for the detection of respiratory events during sleep studies; however, these sensors are uncomfortable for certain patients and can even affect their sleep. They are sometimes unreliable because they could be displaced or even removed by the patients during sleep studies.
Study Impact: The combination of tracheal sounds and respiratory inductance plethysmography could be used as a reliable method for detection and characterization of respiratory events without the need to place sensors on the face of patients and could be used either as an alternative or in combination with the recommended American Academy of Sleep Medicine sensors.
INTRODUCTION
Accurate and reliable detection and classification of respiratory events during sleep as obstructive or central are crucial issues for treatment decision in patients with sleep-disordered breathing. Full-night polysomnography (PSG), the gold standard procedure for sleep apnea diagnosis, requires sensors to measure airflow and respiratory efforts.
In the absence of pneumotachograph, the American Academy of Sleep Medicine (AASM)1 recommends using oronasal thermal airflow sensors (thermistors or thermocouples), nasal cannulas, and respiratory inductance plethysmography (RIP) as alternative techniques to measure airflow. It is also recommended to use, as a first choice, thermal airflow sensors for the detection of apneas and a nasal pressure (NP) transducer for the detection of hypopneas; however, these two sensors are poorly tolerated because they are uncomfortable for certain patients and can even affect their sleep.2,3 They are therefore often displaced or even removed by patients during sleep studies. The validity of the thermistor and NP signals for more than 6 hours of unattended PSG recording is satisfactory for <60% for both children and adults.2,4 Goodwin et al reported that for the thermistor sensor, only 59% of the recordings were without artifact and scorable for more than 6 hours and 52% for the nasal cannula.2 Redline et al have reported that for the thermocouple sensor, only 58% of the recordings were without artifact and scorable for more than 6 hours.4
In the absence of NP and thermistor signals or if the signals are of poor quality, the AASM recommends using one of the RIP-derived signals, either the RIPsum or the RIPflow, as surrogate respiratory signals.1 Using two belts placed around the thorax and the abdomen, the RIP method allows semiquantitative assessment of volume changes (RIPsum) and/or air-flow changes (RIPflow).5–7
Several studies have evaluated the performance of tracheal sounds (TSs) compared with the reference sensors for air-flow measurement and respiratory event detection.8–16 TSs reflect the superficial vibrations of the body set in motion by pressure fluctuations.17,18 Placed just above the sternal notch, the TS sensors can detect these vibrations and thus measure tracheal flow sound. In a recent study, Sabil et al have compared the detection of apneas using thermistor, NP, RIPsum, and TS. Using the thermistor as the reference method, this study showed that TSs score apneas better than both the NP and the RIPsum. The study showed that TS transducers could also be used as reliable sensors for detecting mouth breathing and therefore could be used as an alternative sensor for apnea detection in the absence of oronasal thermistor.19
For characterization of apneas into obstructive and central events, evaluation of respiratory effort during sleep is required. The AASM recommends the use of RIP belts as an indirect method for the evaluation of respiratory efforts for routine sleep studies1; however, thoracoabdominal movements may underestimate obstructive events compared with esophageal pressure, especially in young children or obese adults.20–22 Recent studies have shown that the suprasternal pressure (SSP), a low-frequency signal derived from TS recordings, is a reliable method for respiratory effort assessment in both children23,24 and adults.25–27 The SSP signal corresponds to the pressure variations induced by respiratory efforts. The patient’s respiratory efforts cause variations in pharyngeal pressure, which cause pressure changes at the respiratory rate in the TS sensor chamber.17,18
The aim of this study was to investigate whether the TS-RIP combination compared with the AASM-recommended thermistor-NP-RIP combination could allow reliable detection and characterization of respiratory events and without the need to place sensors on the faces of patients.
METHODS
Study population
This study was nested in the Pays de la Loire Sleep Cohort. Since May 15, 2007, consecutive patients aged ≥18 years, investigated by overnight polysomnography (PSG) or overnight respiratory polygraphy for suspected obstructive sleep apnea (OSA) in seven centers from the Pays de la Loire, were eligible for inclusion in this cohort. Patients with learning difficulties, who were unable to fill in the questionnaires or read and/or speak French, and patients with neuromuscular diseases or chronic respiratory failure were excluded from this cohort.28 Approval was obtained from the University of Angers ethics committee and the Comité Consultative sur le Traitement de l’Information en matière de Recherche dans le domaine de la Santé (CCTIRS) (07.207bis). The database is anonymous and complies with the restrictive requirements of the Commission Nationale Informatique et Liberté (CNIL) and the French information technology and personal data protection authority.
The present study included 70 patients enrolled consecutively, between 10/19/2016 and 13/02/2017, from the Pays de Loire Sleep Cohort, investigated by full-night in-laboratory PSG in Angers Sleep Center for clinical suspicion of OSA. Comparison between the selected patients and the overall cohort population (3,318 patients) has shown no statistically significant difference in terms of patients’ characteristics and PSG results apart from age, sex, systemic hypertension, and diabetes (Table S1 in the supplemental material). All patients gave their written consent to participate in the study.
Baseline evaluation and questionnaires
Each patient enrolled in the Pays de la Loire Sleep Cohort completed surveys that included anthropometric data, smoking habits, alcohol consumption, medical history, and medication use. Systemic hypertension and diabetes were defined as a physician diagnosis combined with treatment with appropriate medication. Comorbid cardiovascular disease was defined as a physician diagnosis of one or more of the following cardiovascular disease: ischemic heart disease, cardiac arrhythmia, congestive heart failure, and stroke. Excessive daytime sleepiness was assessed by the Epworth Sleepiness Score.29
Sleep recordings
All patients were investigated by in-laboratory overnight PSG using the CID102L8D system (CIDELEC, Sainte Gemmes sur Loire, France). Recorded data included electrophysiological signals (six electroencephalographic, two electromyelographic, two electro-oculographic, and electrocardiographic studies) for sleep and cardiac evaluation, as well as airflow by thermistor and NP, RIP belts, pulse oximetry, body position, limb movements, actigraphy, light, and synchronized infrared video. As part of the laboratory routine, TSs and SSP signals using the PneaVoX (CIDELEC) technology were recorded. The PneaVoX sensor was correctly placed on the skin at 1 cm above the sternal notch using a double-sided ring tape and then secured in place using an extra adhesive tape on the top of the sensor. A well-sealed contact surface of the sensor is an essential element to obtain good-quality SSP signals. The PneaVoX, whose technical details have been described previously,17,23,26,27 is a threefold sensor that detects snoring, breathing sound, and respiratory effort.
Data analysis
In addition to an automatic validation of the TS signals by the CIDELEC software, overall visual-quality validation for all signals was performed, and the analysis was done only on the validated segments of the recordings. For each PSG recording, periods with major body movements, artifacts, or absence of any of the used signals were invalidated for analysis. Thereafter, each tracing was anonymized and prepared for the two random scorings as follows.
AASM scoring
Respiratory events were scored according to AASM recommendations1 using oronasal thermistor signal to detect apnea and nasal pressure and oxygen saturation (SpO2) signals for hypopnea detection. Apnea was defined as a ≥90% reduction of the flow signal for at least 10 seconds. Hypopnea was defined as a ≥30% drop in the flow signal for at least 10 seconds associated with a ≥3% decrease in SpO2 and/or an arousal. Signals from thoracic and abdominal movements measured with the RIP were used for respiratory event characterization.
TS-RIP scoring
Respiratory events were scored using TS for the detection of apneas and the respiratory inductance plethysmography derived flow (RIPflow) and SpO2 for hypopnea detection. As illustrated in Figure 1, a normal respiratory cycle on TS signal is characterized by a wave composed of two distinct intensity surges corresponding to inspiration and expiration. Apnea was defined as an absence of respiratory cycle on the TS signal for a duration of at least 10 seconds. Hypopnea was defined as a reduction of RIPflow signal of at least 30% for a duration of at least 10 seconds with a reduction of ≥3% decrease in SpO2 and/or an arousal. Both SSP and RIP belt signals were used for the characterization of respiratory events. In case of discordance, priority was given to the signal that showed respiratory effort. This should increase the reliability of apnea characterization and reduce the risk of underestimating obstructive events.
Figure 1. Normal respiratory cycle concordance on the thermistor, nasal pressure, RIPflow, and TS intensity signals.
On TS signal, respiratory cycles are characterized by a wave composed of two distinct intensity surges corresponding to inspiration and expiration. RIP = respiratory inductance plethysmography, TS = tracheal sound
Scoring for the 70 recordings lasted several weeks. Both scorings were done in random order by the same investigator (C.P.), and a lapse in time between the two scoring was respected so that the scorer could not be influenced from one scoring to another for the same patient. For each scoring, the relevant signals from the alternate scoring criteria were removed to “blind” the scorer. For the AASM scoring, only the oronasal thermistor, the NP, and the SpO2 signals were available for the detection of respiratory events and the RIP belts for the characterization of apneas. For the proposed method, TS-RIP, only the TS, RIPflow, and SpO2 signals were available for the detection of respiratory events and the SSP and RIP belts for the characterization of apneas. An example of the two different scoring methods is illustrated in Figure 2. The results were then compared for each patient for (1) apnea and hypopnea identification; (2) apnea and hypopnea differentiation as the same respiratory event may be scored as apnea or hypopnea by the different scorings; and (3) characterization of the type of apneas as obstructive, mixed, or central. The apnea-hypopnea index (AHI) was calculated for each patient with the two scorings. For each scoring, patients were classified according to their OSA severity (no OSA for AHI <5 events/h, mild for 5 events/h ≤ AHI <15 events/h, moderate for 15 events/h ≤ AHI <30 events/h, and severe for AHI ≥30 events/h).
Figure 2. Concordance of detection of an apnea and a hypopnea.
The events could be detected and characterized on the AASM recommended signals (nasal pressure-thermistor-RIP) and using the proposed method with only tracheal sounds-RIP combination. Respiratory effort is seen on the RIP belts and the suprasternal pressure as well. AASM = American Academy of Sleep Medicine, RIP = respiratory inductance plethysmography
Statistical analysis
Statistical analysis was performed using Prism 6.0C (GraphPad Software, San Diego, California) and MATLAB R2016a software (MathWorks, Natick, Massachusetts). Values are presented as mean ± SD or median and range. The Cohen’s kappa, sensitivity, and specificity were calculated. Correlation analyses and Bland-Altman plots were also used to evaluate agreement visually between the AASM scoring as a reference and the TS-RIP scoring.
RESULTS
Patient characteristics
As shown in Table 1, the study population consisted of typical mild‐to‐severe OSA patients (AHI = 23.0 ± 21.0 events/h), predominantly male (n = 53 [75%]), obese or overweight (body mass index = 28.9 ± 6.4 kg/m2), and frequently presenting with systemic hypertension (n = 36 [51%]), diabetes (n = 15 [21%]), and cardiovascular diseases (n = 12 [17%]). The average total sleep time was 7.15 ± 1.35 hours.
Table 1.
Patient characteristics.
| No. of patients | 70 |
| Age, y | 55.1 ± 13.6 |
| Body mass index (kg/m2) | 28.9 ± 6.4 |
| Females, n (%) | 17 (25) |
| Epworth Sleepiness Scale | 9.8 ± 5.1 |
| Apnea-hypopnea index (events/h) | 23 ± 21 |
| Apnea index (/h) | 7.4 ± 15.6 |
| 3%, Oxygen desaturation index (/h) | 20.7 ± 19.6 |
| Arousal index (/h) | 30.3 ± 16.8 |
| Total sleep time (min) | 428.9 ± 81.1 |
| N1-N2 sleep (%) | 59.9 ± 12.2 |
| N3 sleep (%) | 21.8 ± 8.4 |
| REM sleep (%) | 19.0 ± 6.0 |
| Systemic hypertension, n (%) | 36 (51.3) |
| Diabetes, n (%) | 15 (21.6) |
| Cardiovascular diseases, n (%) | 12 (17.4) |
Data are mean ± SD otherwise specified.
Signal validation
In addition to CIDELEC software’s automatic signal-quality checking, visual checking was performed to validate the quality of all signals used in the different scorings. The validation was 89.56% of the total recording time for the AASM scoring signals and 96.51% for the TS-RIP scoring signals. The mean validation time was 8.18 ± 1.61 hours for the AASM scoring signals and 8.81 ± 0.63 hours for the TS-RIP scoring signals.
Respiratory events detection
For AASM scoring, the total number of respiratory events was 10,883, including 3,803 apneas and 7,080 hypopneas. The total number was 11,071 respiratory events (3,764 apneas and 7,306 hypopneas) for the TS-RIP scoring. On close examination, 3,659 apneas and 6,368 hypopneas were correctly identified by the TS-RIP scoring; however, 113 events were classified as hypopneas by the TS-RIP scoring but as apneas according to the AASM scoring. Similarly, 71 events were described as apnea by the TS-RIP scoring but as hypopneas according to the AASM scoring (Table 2). The sensitivity and specificity of the TS-RIP scoring with respect to the AASM scoring were 96.21% and 91.34% for detecting apneas, 89.94% and 93.25% for detecting hypopneas, and 92.24% and 92.13% for not detecting an event when there is none. For overall detection using TS-RIP scoring, the Cohen’s kappa coefficient was .87.
Table 2.
Respiratory event detection.
| AASM | TS-RIP | Total | ||
|---|---|---|---|---|
| Apneas | Hypopneas | None | ||
| Apneas | 3,659 | 113 | 31 | 3,803 |
| Hypopneas | 71 | 6,368 | 641 | 7080 |
| None | 34 | 825 | 10,212 | 11,071 |
| Total | 3,764 | 7,306 | 10,884 | |
| Sensitivity | 96.21% | 89.94% | 92.24% | |
| Specificity | 91.34% | 93.25% | 92.13% | |
Comparison of apnea and hypopnea detection between the TS-RIP method and the AASM method. AASM = American Academy of Sleep Medicine, RIP = respiratory inductance plethysmography.
Apnea characterization
A total of 3,659 apneas common to AASM and TS-RIP scorings were detected in all patients and then characterized using each method separately. Using AASM scoring, 2,774 were considered obstructive, 500 mixed, and 385 central. Using TS-RIP scoring, 2,778 apneas were classified as obstructive, 518 as mixed, and 363 as central. Table 3 shows apnea classification comparison of AASM scoring and TS-RIP scoring. The sensitivity for the characterization of obstructive, mixed, and central apneas was 98.67%, 92.66%, and 96.14%, respectively. Specificity for the characterization of apneas as obstructive, mixed, and central was 96.17%, 99.36%, and 98.89%, respectively. Cohen’s kappa coefficient for the characterization of apneas using the TS-RIP scoring compared with the recommended AASM scoring was .94.
Table 3.
Apnea characterization.
| AASM | TS-RIP | Total | ||
|---|---|---|---|---|
| OA | MA | CA | ||
| OA | 2741 | 23 | 10 | 2774 |
| MA | 16 | 480 | 4 | 500 |
| CA | 21 | 15 | 349 | 385 |
| Total | 2,778 | 518 | 363 | 3,659 |
| Sensitivity | 98.67% | 92.66% | 96.14% | |
| Specificity | 96.17% | 99.36% | 98.89% | |
Comparison of apnea characterization TS-RIP method and the AASM method. AASM = American Academy of Sleep Medicine, CA = central apnea, MA = mixed apnea, OA = obstructive apnea, RIP = respiratory inductance plethysmography, TS = trachea sound.
AHI and OSA severity
The mean AHI was 23.0 ± 21.0 events/h using the AASM scoring and 23.4 ± 23.7 events/h using TS-RIP scoring.
Agreement between the two methods for calculating the AHI was confirmed by correlation analysis (Figure 3) and Bland-Altman analysis (Figure 4) for the apnea index, hypopnea index, and AHI. There was a strong positive correlation between AASM and TS-RIP for the apnea index (r =.998, n = 70, P < .0001), the hypopnea index (r = .984, n = 70, P < .0001), and the AHI (r = .994, n = 70, P < .0001). The Bland-Altman analysis was performed for the apnea index, hypopnea index, and AHI evaluations using the AASM scoring as the reference. The TS-RIP scorings underestimated the apnea index, hypopnea index, and AHI in average by .14, .47, and .97 events/h, respectively.
Figure 3. Correlation graphs illustrating the strength of the linear association between TS-RIP scoring and AASM scoring.
Strength of the linear association between TS-RIP scoring and AASM scoring for apnea index (A), hypopnea index (B), and apnea-hypopnea index (C). AASM = American Academy of Sleep Medicine, RIP = respiratory inductance plethysmography, TS = tracheal sound.
Figure 4. Bland-Altman plots comparing TS-RIP scoring and AASM scoring.
Comparing apnea index (A), hypopnea index (B), and apnea-hypopnea index (C) scoring. AASM = American Academy of Sleep Medicine, AHI = apnea hypopnea index, AI = apnea index, HI = hypopnea index, RIP = respiratory inductance plethysmography, SD = standard deviation, TS = tracheal sound.
The TS-RIP scoring revealed a high level of agreement for classifying OSA into severity classes with a Cohen’s kappa coefficient of 0.96. Of 70 patients, only four were misclassified by TS-RIP vs AASM with observed scoring misclassification by only one severity class (Figure 5). Two patients classified as severe by TS-RIP (AHI values: 30.7 events/h and 32.5 events/h, respectively) and moderate by AASM (AHI values: 26.6 events/h and 29.2 events/h), one classified as moderate (AHI value: 28.5 events/h) rather than severe (AHI value: 30.3 events/h), and one scored as no OSA (AHI value: 4.6 events/h) rather than mild OSA (AHI value: 5.0 events/h).
Figure 5. Distribution of AHI severity categories for the AASM method and the TS-RIP method.
Of 70 patients, four changed categories with the TS-RIP method with observed scoring misclassification by only one severity class. Two patients classified as severe by TS-RIP were moderate by AASM, one classified as moderate was severe by AASM, and one classified as no OSA was mild by AASM. AASM = American Academy of Sleep Medicine, AHI = apnea-hypopnea index, OSA = obstructive sleep apnea, RIP = respiratory inductance plethysmography, TS = tracheal sound.
DISCUSSION
This study is the first to investigate the use of TSs in combination with RIP belt signals for the diagnosis of OSA in adult patients without placing any respiratory sensors on the patient’s face. Results were compared with those obtained with the use of the combination thermistor-nasal cannula-RIP sensors, as recommended by the AASM. Under the assumption that the TS sensor is well placed just above the sternal notch and that the scorer is familiar with the TS signal analysis, the proposed method provides a correct detection and characterization of respiratory events and thus a highly accurate evaluation of OSA.
Detection of respiratory events
Ranges of the number of apnea and hypopnea events detected were closely comparable, with values ranging from 3 to 708 events for the TS-RIP scoring and from 3 to 711 events for AASM scoring; however, the number of events detected with the TS-RIP scoring was 1.7% more than that detected with the AASM reference scoring. This difference concerns mainly hypopneas, with 226 (+3.2%) more events detected with the TS-RIP scoring, which could be explained by the fact that in some cases, patients could have upper airway resistance, which can cause the RIP abdominal and thoracic signals to be slightly out of phase and could result in a RIPflow drop of >30%, whereas the nasal flow drop, if there is any, does not exceed the 30% mark. Nonetheless, the TS-RIP scoring had high sensitivity and specificity in detecting respiratory events. In addition, among the commonly detected events, the sensitivity and specificity of differentiating between apneas and hypopneas were high, with a kappa coefficient of 0.94. These results confirm the reliability of the TS-RIP scoring.
Apneas are easily detected on TS signals and/or the absence of definite respiratory cycles during monitoring. 8,12,30 Apneas are defined as events where the respiratory flow is absent or reduced by >90% of the reference value for at least 10 seconds.1,31,32 A first-generation PneaVoX has been already validated against a pneumotachograph for apnea detection, and there were no differences in apnea number and duration recorded by the TS method and the PNT.8 A recent study with a new-generation PneaVoX has shown that the TS signal detects fewer apneas than the NP signal and more apneas than the RIPsum, placing the TS signal as the closest signal to the thermistor in term of apnea detection compared with the NP and the RIPsum signals.19 We relied on the TS for the detection of apneas because the RIPsum and the RIPflow could sometimes be unreliable in detection of apneas. During obstructive apnea, the RIPflow signal is not always reduced >90% as it should during apneas because the thoracic and abdominal signals are not always necessary in paradoxical movements. Thus, the thorax and abdominal belt signals do not exactly sum up to zero. This problem is minimized by calibration of the RIP signals; however, even the calibrated RIPflow may not remain accurate overnight because of changes in patient position and/or belt movements. Detecting apneas using the RIPflow signal is not reliable and may underestimate the apnea index. Thus, using the RIPflow signal may have an impact on the clinical diagnosis of the OSA and minimize the severity of the disease. We have shown in a recent study that obstructive apnea could be well detected by the TS, NP, and the thermistor sensors, whereas the RIPsum (the integral of RIPflow) signal could have mistaken it for a hypopnea.19 In our study, the TS detected 1% less apneas, and 2.7% of events detected as apneas with the AASM method were considered hypopneas with the TS-RIP scoring; however, these differences were not significant and had no impact on the diagnosis.
Finally, it is important to note that some apneic events are somewhat equivocal on the TS signal. This is the case where chocking noise is recorded during obstructive apnea; however, choking sounds are characterized by an amplified burst of the TS with an absence of respiratory cycles during the event, and an experienced scorer should be able to discern the difference between chocking noise and breathing sounds with definite respiratory cycle.18 Thus, the scorer’s familiarity with the TS signal characteristics is quite important for successful analysis.
AHI and OSA severity
From a clinical point of view, the TS-RIP scoring categorized patients correctly into severity groups as no OSA, mild, moderate, or severe OSA. Only four of the recorded 70 patients (5.7%) changed the severity class and by only one class. The AHI index difference was small, however, with a minimum change of 0.4 event/h for the misclassification from “no OSA” to “mild OSA” and a maximum of 4.1 for the change in class from “severe” to “moderate” OSA and not clinically significant. Given the small changes in AHI, treatment of these OSA patients would presumably not have changed. In all four cases, the difference was in hypopnea index, whereas the apnea index (AI) was relatively similar. In fact, the apnea index was 8.0 ± 17.2 /h and 8.0 ± 17.0 /h, and the hypopnea index was 15.0 ± 12.7 /h and 15.5 ± 12.9 /h for AASM and TS-RIP scoring, respectively; however, there was a high correlation between the scorings for AHI, AI, and HI. Given that apneas were scored on the TS signal, this confirms the reliability of the TS for apnea detection. Our results confirm those of previous studies showing that apneas could be reliably identified by the cessation of TS and/or the absence of definite respiratory cycles during monitoring and quantification of sleep apnea by TS recordings.12,19,25,30 The ability of the TS-RIP method to detect apneas and hypopneas is also supported by the Bland-Altman plot (Figure 4), illustrating that the patient-to-patient variability of differences was small. The AHI average difference seen in patients was not significant, and it was mostly present among four patients with severe OSA and an average AHI of 62.8 events/h. Thus, this variability did not make a difference in the diagnosis of these four patients.
Characterization of apneas
For characterization of apneas, the present study confirms previous findings in pediatric12 and adults22,25 patients. Sensitivity and specificity for the characterization of obstructive, mixed, and central apneas were excellent with the TS-RIP scoring; however, hypopneas were not characterized in our study as it is optional according to the AASM.1
This study also confirmed the reliability of the SSP measured using the PneaVox sensor for the characterization of apneas. Furthermore, the combination of the TS and the RIP signals in the proposed scoring method increases the accuracy of respiratory effort evaluation. The sensors are complementary, and when the respiratory effort is not clear on the RIP signal, the characterization could rely on the SSP. Studies comparing the SSP and the RIP to the esophageal pressure (Pes) showed that SSP was slightly more reliable than RIP and could be used as a surrogate signal for respiratory effort evaluation.8,26,27 In addition to the studies that have shown that thoracoabdominal movements may underestimate obstructive events compared with esophageal pressure,20–22 recent studies have shown that apnea could be misclassified by the RIP signals as central while clearly identified as obstructive on both the esophageal pressure and the SSP signals.26,27
Although visual detection and characterization of apneas are of high-quality using TS flow and SSP signals measured using the PneaVox, visual detection and characterization of hypopneas using the PneaVox are not straightforward, particularly for scorers who are not familiar with these two signals. Training and understanding of the TS and SSP signals are crucial to ensure accurate scoring of hypopneas. Thus, relying on the RIPflow in our study to score hypopnea was an important combination that increases the reliability of the scoring. This study shows that a well-trained scorer can use TSs as a complement or as an alternative signal to the standard sensors recommended by the AASM. Our results confirm those of Amaddeo et al, who recently showed that the combination TS/SSP/RIPSum has a good sensitivity and specificity for the detection and characterization of apnea and hypopnea in children24; however, these results need to be confirmed in a larger study, including more patients from a heterogeneous population, with analysis performed by multiple scorers.
An informative polysomnography depends on the good tolerability and scorability of the sensors. The AASM recommends sensors for respiratory event detection that are placed in a sensitive area, inside the nostrils for the nasal cannula and the nasal thermistor and between the nose and the mouth for the oral thermistor. They are therefore bothersome and subject to rejection by certain patients and/or displacement during sleep. This problem is particularly recurrent in home sleep testing and can result in nonvalidation of some recordings. The major advantage of the TS-RIP scoring method is its excellent tolerability compared with nasal and oronasal sensors. In addition, once properly put in place, the TS sensor used in this study cannot be removed or displaced during recording at night. Furthermore, the TS sensor is put in place in a clinic setting, and patients are not relied on for the placement of sensors in the case of home sleep apnea testing. In fact, for home sleep apnea testing, the nasal canula and the thermistor sensors are often left for the patient to put in place, which increases the probability of misplacement or nonuse of the sensors and, thus, the risk of failure of this type of recording; however, further studies are required to establish greater clinical utility of the TS-RIP scoring method in a unsupervised setting during home sleep apnea testing recording and in specific groups of patients, such as obese patients, pregnant women, and children.
Limitations of the study
This study has several limitations. No statistical method was used to randomly select the recordings used for this retrospective study; however, our 70-patient sample is representative of the cohort for the PSG data, it has a high number of events, and the demographic differences observed are unlikely to have influenced the results.
Second, the PSG recordings were analyzed by a single scorer, thereby excluding interscorer variability; however, interscorer variability in the scoring of respiratory events is mainly caused by how well trained the scorers are and how familiar they are with the analyzed signals. Signal processing, such as digital filtering, could also influence the quality and the reliability of the scoring. For characterization of events, the lack of consensus of how long the central part of a mixed apnea should last is another problem that can lead to misclassification, mixed vs obstructive, of events and affect interscorer variability. Furthermore, given the large number of patients (n = 70) and analyzed events (10,883 with the AASM and 11,071 with the ST-RIP), the blinded relevant signals between the two scorings and the lapse of time between the two scorings for the same patient, the scorer could not be influenced from one scoring to the other. We therefore believe this is not a major limitation.
Another limitation of our study is the absence of quantified measurement of how much the patients are bothered by the placement of the nasal cannula and the thermistor sensors, and no quantified measurement has been performed to compare it with feeling bothered by the tracheal sound sensor. Beyond the 70 patients we recorded for our study, however, we have observed in our sleep laboratory more complaints about the nasal cannula sensor placed in the nostrils and the thermistor sensor placed on the upper lip than the TS sensor placed on the skin just above the tracheal notch. A few patients complained about an allergic reaction to the tape used to hold the TS sensor in place. In addition, Redline et al have shown in a study assessing PSG sensors a discomfort level that, for the thermistor, only 35% of the patients (n = 6,323) reported no discomfort, whereas the remainder of the patients reported little (36%), moderate (23%), or a great deal (6%) of discomfort.4 In another study in children, Goodwin et al reported for the thermistor/nasal cannula that only 35% of the patients (n = 157) reported no discomfort, whereas the remainder of the patients reported little (31%), moderate (17%), or considerable (17%) discomfort.2
Finally, the TS sensor used in this study, the PneaVoX, is specific to the CIDELEC PSG system, and the signals obtained cannot be generated and recorded by other PSG equipment on the market. This technology has been extensively used in clinical practice in France for the last 25 years, and tracheal flow sound is recognized in French clinical practice recommendations as a valid signal, associated with nasal pressure, to detect oronasal respiration.30 To our knowledge, the PneaVoX is the only tracheal sound sensor incorporated in a PSG system that is commercially available today for sleep studies; however, although this study evaluates only one TS device with certain specifications, it opens the door for other devices to be tested.
CONCLUSIONS
The present study demonstrates that for in-laboratory setting, the TS-RIP combination allows reliable noninvasive detection and characterization of respiratory events with a high degree of sensitivity and specificity compared with the AASM NP-Therm-RIP combination. The TS-RIP combination could be used either as an alternative or in combination with the recommended AASM sensors.
DISCLOSURE STATEMENT
All authors have seen and approved the manuscript as submitted. We certify that all the authors are fully responsible for the reported research and have participated in the concept, design, drafting, and revising of the manuscript. Work for this study was done at the University Hospital of Angers. AbdelKebir Sabil was employed by CIDELEC at the time of the study. He contributed to the design of the study, analysis of the data, and preparation of the manuscript. Guillaume Baffet was also employed by CIDELEC but contributed only to the study design. All other authors declare that they have no conflict of interest. This work was supported by grants from the Institut de Recherche en Santé Respiratoire des Pays de le Loire.
SUPPLEMENTARY MATERIAL
ACKNOWLEDGMENTS
The authors thank Christelle Gosselin and Professor Jean-Louis Racineux, from the Institut de Recherche en Santé Respiratoire des Pays de le Loire and Julien Godey, Laetitia Moreno, and Marion Vincent, sleep technicians at the Department of Respiratory and Sleep Medicine of Angers University Hospital.
ABBREVIATIONS
- AASM
American Academy of Sleep Medicine
- AHI
apnea-hypopnea Index
- NP
nasal pressure
- OSA
obstructive sleep apnea/
- PSG
polysomnography
- RIP
respiratory inductance plethysmography
- SpO2
pulse oximetry
- SSP
suprasternal pressure
- TS
tracheal sound
REFERENCES
- 1.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. Deliberations of the Sleep Apnea Definitions Task Force of the American Academy of Sleep Medicine. J Clin Sleep Med. 2012;8(5):597–619. 10.5664/jcsm.2172 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Goodwin JL, Enright PL, Kaemingk KL, et al. Feasibility of using unattended polysomnography in children for research: report of the Tucson Children’s Assessment of Sleep Apnea study (TuCASA). Sleep. 2001;24(8):937–944. 10.1093/sleep/24.8.937 [DOI] [PubMed] [Google Scholar]
- 3.Trang H, Leske V, Gaultier C. Use of nasal cannula for detecting sleep apneas and hypopneas in infants and children. Am. J Respir Crit Care Med. 2002;166(4):464–468. 10.1164/rccm.2110114 [DOI] [PubMed] [Google Scholar]
- 4.Redline S, Sanders MH, Lind BK, et al. Methods for obtaining and analyzing unattended polysomnography data for a multicenter study. Sleep Heart Health Research Group. Sleep. 1998;21(7):759–767. 10.1093/sleep/21.7.759 [DOI] [PubMed] [Google Scholar]
- 5.Sabil A, Eberhard A, Baconnier P, Benchetrit G. A physical model of inspiratory flow limitation in awake healthy subjects. Adv Exp Med Biol. 2004;551:211–216. 10.1007/0-387-27023-X_32 [DOI] [PubMed] [Google Scholar]
- 6.Konno K, Mead J. Static volume-pressure characteristics of the rib cage and abdomen. J Appl Physiol. 1968;24(4):544–548. 10.1152/jappl.1968.24.4.544 [DOI] [PubMed] [Google Scholar]
- 7.Konno K, Mead J. Measurement of the separate volume changes of rib cage and abdomen during breathing. J Appl Physiol. 1967;22(3):407–422. 10.1152/jappl.1967.22.3.407 [DOI] [PubMed] [Google Scholar]
- 8.N. Meslier JLRIPJH . Podszus T, von Wichert P. Use of tracheal sound recordings to monitor airflow during sleep. In: Sleep Related Disorders and Internal Diseases. Springer: Berlin, Heidelberg; 1991 [Google Scholar]
- 9.Soufflet G, Charbonneau G, Polit M, et al. Interaction between tracheal sound and flow rate: a comparison of some different flow evaluations from lung sounds. IEEE Trans Biomed Eng. 1990;37(4):384–391. 10.1109/10.52345 [DOI] [PubMed] [Google Scholar]
- 10.Beckerman RC, Wegmann MJ. A comparison of tracheal breath sounds, airflow, and impedance pneumography in the detection of childhood apnea. Sleep. 1985;8(4):342–346. 10.1093/sleep/8.4.342 [DOI] [PubMed] [Google Scholar]
- 11.Beckerman RC, Wegmann MJ, Waring WW. Tracheal breath sounds for detection of apnea in infants and children. Crit Care Med. 1982;10(6):363–366. 10.1097/00003246-198206000-00005 [DOI] [PubMed] [Google Scholar]
- 12.Cummiskey J, Williams TC, Krumpe PE, Guilleminault C. The detection and quantification of sleep apnea by tracheal sound recordings. Am Rev Respir Dis. 1982;126(2):221–224. [DOI] [PubMed] [Google Scholar]
- 13.Peirick J, Shepard JWJr. Automated apnoea detection by computer: analysis of tracheal breath sounds. Med Biol Eng Comput. 1983;21(5):632–635. 10.1007/BF02442390 [DOI] [PubMed] [Google Scholar]
- 14.Nakano H, Hayashi M, Ohshima E, Nishikata N, Shinohara T. Validation of a new system of tracheal sound analysis for the diagnosis of sleep apnea-hypopnea syndrome. Sleep. 2004;27(5):951–957. 10.1093/sleep/27.5.951 [DOI] [PubMed] [Google Scholar]
- 15.Kulkas A, Huupponen E, Virkkala J, et al. New tracheal sound feature for apnoea analysis. Med Biol Eng Comput. 2009;47(4):405–412. 10.1007/s11517-009-0446-z [DOI] [PubMed] [Google Scholar]
- 16.Yadollahi A, Giannouli E, Moussavi Z. Sleep apnea monitoring and diagnosis based on pulse oximetry and tracheal sound signals. Med Biol Eng Comput. 2010;48(11):1087–1097. 10.1007/s11517-010-0674-2 [DOI] [PubMed] [Google Scholar]
- 17.Penzel T, Sabil A. The use of tracheal sounds for the diagnosis of sleep apnoea. Breathe (Sheff). 2017;13(2):e37–e45. 10.1183/20734735.008817 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Penzel T, Sabil A. Physics and applications for tracheal sound recordings in sleep disorders. In: Priftis K, Hadjileontiadis L, Everard M (eds). Breath Sounds: From Basic Science to Clinical Practice. Cham, Switzerland: Springer; 2018:83-104.. 10.1007/978-3-319-71824-8_6 [DOI] [Google Scholar]
- 19.Sabil A, Glos M, Gunther A, et al. Comparison of apnea detection using oronasal thermal airflow sensor, nasal pressure transducer, respiratory inductance plethysmography and tracheal sound sensor. J Clin Sleep Med. 2019;15(2):285–292. 10.5664/jcsm.7634 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Luo YM, Tang J, Jolley C, et al. Distinguishing obstructive from central sleep apnea events: diaphragm electromyogram and esophageal pressure compared. Chest. 2009;135(5):1133–1141. 10.1378/chest.08-1695 [DOI] [PubMed] [Google Scholar]
- 21.Boudewyns A, Willemen M, Wagemans M, De Cock W, Van de Heyning P, De Backer W. Assessment of respiratory effort by means of strain gauges and esophageal pressure swings: a comparative study. Sleep. 1997;20(2):168–170. 10.1093/sleep/20.2.168 [DOI] [PubMed] [Google Scholar]
- 22.Staats BA, Bonekat HW, Harris CD, Offord KP. Chest wall motion in sleep apnea. Am Rev Respir Dis. 1984;130(1):59–63. [DOI] [PubMed] [Google Scholar]
- 23.Amaddeo A, Fernandez-Bolanos M, Olmo Arroyo J, Khirani S, Baffet G, Fauroux B. Validation of a suprasternal pressure sensor for sleep apnea classification in children. J Clin Sleep Med. 2016;12(12):1641–1647. 10.5664/jcsm.6350 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Amaddeo A, Sabil A, Arroyo JO, et al. Tracheal sounds for the scoring of sleep respiratory events in children. J Clin Sleep Med. 2020;16(3):361–369. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Meslier N, Simon I, Kouatchet A, Ouksel H, Person C, Racineux JL. Validation of a suprasternal pressure transducer for apnea classification during sleep. Sleep. 2002;25(7):753–757. 10.1093/sleep/25.7.753 [DOI] [PubMed] [Google Scholar]
- 26.Glos M, Sabil A, Jelavic KS, Schobel C, Fietze I, Penzel T. Characterization of respiratory events in obstructive sleep apnea using suprasternal pressure monitoring. J Clin Sleep Med. 2018;14(3):359–369. 10.5664/jcsm.6978 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Sabil A, Schobel C, Glos M, et al. Apnea and hypopnea characterization using esophageal pressure, respiratory inductance plethysmography, and suprasternal pressure: a comparative study. Sleep Breath. 2019;23(4):1169–1176. 10.1007/s11325-019-01793-8 [DOI] [PubMed] [Google Scholar]
- 28.Gagnadoux F, Le Vaillant M, Goupil F, et al. Influence of marital status and employment status on long-term adherence with continuous positive airway pressure in sleep apnea patients. PLoS One. 2011;6(8):e22503. 10.1371/journal.pone.0022503 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Johns MW. A new method for measuring daytime sleepiness: the Epworth sleepiness scale. Sleep. 1991;14(6):540–545. 10.1093/sleep/14.6.540 [DOI] [PubMed] [Google Scholar]
- 30.Krumpe PE, Cummiskey JM. Use of laryngeal sound recordings to monitor apnea. Am. Rev. Respir. Dis. 1980;122(5):797–801. 10.1164/arrd.1980.122.5.797 [DOI] [PubMed] [Google Scholar]
- 31.Escourrou P, Meslier N, Raffestin B, et al. Quelle approche clinique et quelle procédure diagnostique pour le SAHOS? Rev. Mal. Respir. 2010;27(Suppl 3):S115–S123. 10.1016/S0761-8425(10)70017-6 [DOI] [PubMed] [Google Scholar]
- 32.Mayer G, Arzt M, Braumann B, et al. German S3 Guideline Nonrestorative Sleep/Sleep Disorders: Sleep-Related Breathing Disorders in Adults (short version: German Sleep Society (Deutsche Gesellschaft fur Schlafforschung und Schlafmedizin, DGSM). Somnologie (Berl). 2017;21(4):290–301. 10.1007/s11818-017-0136-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.