The Odds Ratio Product (An Objective Sleep Depth Measure): Normal Values, Repeatability, and Change With CPAP in Patients With OSA

Abstract

Study Objectives:

The Odds Ratio Product (ORP) is an objective measure of sleep depth using the relationships of the powers of different electroencephalogram (EEG) frequencies in a single index. The range of the ORP is 0 (deeply asleep) to 2.5 (fully awake). This investigation seeks to elucidate normal values of non-rapid eye movement ORP (ORP_NR) in healthy individuals, repeatability of the measure, and the change in ORP_NR following continuous positive airway pressure (CPAP) treatment.

Methods:

Healthy individuals underwent a home sleep apnea test (HSAT) with EEG followed 1 week later by EEG alone. Another cohort with OSA underwent baseline HSAT with EEG followed by a second EEG study approximately 4 weeks into treatment with CPAP.

Results:

Thirty-eight healthy individuals completed the protocol (mean age of 34.9 ± 7.4 years, Epworth Sleepiness Scale score 3.6 ± 2.4, Insomnia Severity Index score 2.0 ± 1.6 and Functional Outcomes of Sleep Questionnaire - shorter version score 19 ± 1.2). The mean ORP_NR for all nights was 0.52 ± 0.13. The difference between the first night and the second night was 0.024 ± 0.17 (not significant). The intraclass correlation coefficient was 0.525, suggesting only moderate agreement between the first and second nights. The normal value for ORP_NR in healthy individuals is ≤ 0.78 units using two standard deviations as the cutoff. Forty participants completed the OSA protocol (mean age 49 ± 11 years, body mass index 35 ± 6 kg/m², apnea-hypopnea index 33.5 ± 28.4 events/h). The mean pre-CPAP ORP_NR was 0.69 ± 0.24 and the mean post-CPAP ORP_NR was 0.57 ± 0.22 (P = .02).

Conclusions:

The ORP_NR proves to have significant variability from night to night in healthy individuals. ORP_NR objectively improves following CPAP treatment, providing further evidence that it measures sleep depth.

Citation:

Penner CG, Gerardy B, Ryan R, Williams M. The odds ratio product (an objective sleep depth measure): normal values, repeatability, and change with CPAP in patients with OSA. J Clin Sleep Med. 2019;15(8):1155–1163.

BRIEF SUMMARY

Current Knowledge/Study Rationale: The current study is the first to investigate normal values for the newly described Odds Ratio Product in non-rapid eye movement sleep (ORP_NR; a sleep depth scale based on the electroencephalogram signal) in healthy individuals in their home environment and finds that the normal ORP_NR is ≤ 0.78. A second objective was to investigate whether ORP_NR changes following treatment with CPAP in patients with OSA and finds that it decreases by a mean of 0.12.

Study Impact: The sleep depth scale is attractive in that it measures sleep depth in a single number rather than the complex way sleep is currently described in polysomnography.

INTRODUCTION

Good quality sleep has long been recognized as being important in overall health and wellbeing.¹ An easily applied and available objective measure for measuring sleep depth and quality has not been available. Polysomnography (PSG) has been used to measure sleep time, sleep latency, sleep efficiency, wake after sleep onset, arousal index (AI), and time in each of the stages of sleep to get an idea of sleep depth and quality. The number of variables to describe sleep depth and quality makes comparisons between different nights in an individual or between individuals problematic. PSG is also a time- and cost-intensive process, making the feasibility of repeated sleep studies difficult. The artificiality of the laboratory and the intrusiveness of the technology are other barriers to the use of PSG and may affect the usefulness of the information obtained.

A better way of evaluating sleep depth and quality is therefore needed. Recently, the Odds Ratio Product (ORP) has been introduced as a means to describe depth of sleep.² The ORP is a single, continuous variable based on spectral analysis of the EEG pattern ranging from 0 (deeply asleep) to 2.5 (fully awake). The ORP is an attractive measure not only in that it distills sleep depth to a single index but also because it can be obtained in an automated manner. The Rechtschaffen and Kales rules used for other objective measures of sleep require time-intensive manual review. The ORP, with the addition of sleep time and sleep latency, may allow comparisons of interventions in sleep disorders to be done more quickly in a more understandable way as well as making it less costly given the automation of the measure. A simple to use, portable device to measure electroencephalograms is also needed to make the ORP readily available. A number of these devices have recently come to market, making this feasible.

Given this context, we were interested in exploring three aspects of the ORP. The first is to estimate normal values of non-rapid eye movement ORP (ORP_NR) in healthy individuals in their home environment. There is some information in the literature regarding the normal range but this was obtained in a laboratory environment.³ The second is to evaluate the variability of the ORP_NR measurement between nights in healthy individuals. There is no published information regarding variability of ORP_NR in healthy individuals. As part of this analysis, we will seek to evaluate sleep parameters that might help explain variations from night to night. The third is to examine the magnitude of change of the ORP_NR after participants have become acclimated to continuous positive airway pressure (CPAP) therapy for 3 to 4 weeks. Qanash et al examined the change in ORP_NR in participants with OSA during a split-night study.³ This investigation seeks to extend these findings.

METHODS

ORP Technology

The ORP is a sleep depth scale based on power spectral analysis of the EEG signal. The ORP can range from 0 (deeply asleep) to 2.5 (fully awake); thus, the lower the ORP, the greater the depth of sleep. Briefly, the ORP is derived using four EEG frequency bands (beta, alpha, theta, and delta) and the power of each of these bands. The EEG is divided into 3-second segments and a fast Fourier transform is performed for each 3-second segment. The relative power for each frequency is divided into 10 deciles. There are 10,000 possible combinations using the 4 frequency bands and 10 power deciles. Each possible combination is designated as a unique bin number. In the original description, each 3-second EEG segment was sorted according to bin number and tagged as being awake or asleep depending on whether it occurred in a 30-second epoch scored as being asleep or awake using Rechtschaffen and Kales rules. For each bin number the probability of being “awake” from (total number designated “awake”/ total number of epochs with the same bin number in the entire set) × 100 is calculated. If a particular bin number is associated with the wake state 60% of the time and the sleep state 40% of the time, this EEG fragment has a corresponding ORP of 1.5. If another bin number is associated with the wake state 20% of the time and 80% with the sleep state, this would be designated with an ORP of 0.5. A mean ORP is calculated for each 30-second epoch. ORP_NR is the mean ORP of all 30-second epochs scored as non-rapid eye movement (NREM) sleep during a study and is considered to represent sleep depth.²

There are at least two lines of evidence suggesting that ORP reflects depth of sleep. The first is that the ORP has been found to continually decline from wakefulness to stage N3 sleep both in the development set and in the validation set during the original description. The second line of evidence is that the likelihood of arousal in the next 30 seconds when the ORP is low is quite low and high when the ORP is high. A detailed discussion of this is found in the original description of the ORP_NR.²

The ORP-9 is an EEG biomarker that has been described as an index of postarousal sleep dynamics.⁴ The ORP-9_NR is the mean ORP during the first 9 seconds after the end of an arousal/awakening (A/AW) in NREM sleep. The ORP-9_NR conceptualizes the depth of sleep achieved immediately following an A/AW. People who fall asleep more slowly after an arousal will have a high ORP-9_NR. A person with a higher ORP-9_NR is more susceptible to another A/AW because of being in lighter sleep. People who fall asleep quickly after an A/AW will have a lower ORP-9_NR and are less likely to have an A/AW in the next 30-second period of sleep. The ORP-9_NR is highly correlated with ORP_NR with a lower ORP-9 being associated with a lower ORP_NR.⁴ People with a lower ORP-9_NR are less likely to have an arousal leading to deeper sleep overall.

The original description of ORP-9 described the mean ORP-9 following all A/AW in NREM sleep. We have designated it as ORP-9_NR for clarity in this article. We also analyzed ORP-9 in rapid eye movement (REM) sleep and have designated this as ORP-9_REM.

ORP Normal Values for Healthy Individuals

Healthy individuals between the ages of 20 and 50 years were recruited by word of mouth and by a general email invitation within a large corporation and enrolled in the study after giving informed consent. Research ethics approval was obtained from the University of Manitoba Health Research Ethics Board. To be enrolled, participants needed to sleep between 7 and 9 hours each night and not habitually nap or fall asleep unintentionally. In addition, they had to affirm that they did not usually have trouble getting through their day because of sleepiness or fatigue. Exclusion criteria included the use of sedatives, antidepressants, antipsychotic medications, recreational drugs, tobacco, narcotics, nicotine, or excessive alcohol intake. Excessive alcohol intake was defined as more than 10 standard alcoholic drinks per week for women and more than 14 standard alcoholic drinks for men. Also excluded were shift workers, women who had reached menopause, and those with chronic disease or a previously diagnosed sleep disorder. Healthy individuals who met the aforementioned inclusion and exclusion criteria were asked to complete the Epworth Sleepiness Score (ESS)⁵, Functional Outcomes of Sleep Questionnaire - shorter version (FOSQ10),⁶ and the Insomnia Severity Index (ISI).⁷ Participants were excluded if they scored higher than 9 on the ESS, higher than 7 on the ISI, and lower than 18 on the FOSQ10.

Healthy individuals proceeded to a home sleep apnea test (HSAT) using the Alice NightOne system (Royal Philips: US headquarters, Andover, Massachusetts, United States) combined with Prodigy 1.0 (Younes Medical Technologies [YMT], Winnipeg, Manitoba, Canada). Healthy individuals were sent the sleep study monitors with written instructions on how to set up. The EEG monitor uses two frontal leads, two eye leads, a chin lead, and mastoid lead for reference.⁸ The two components of the study were combined and scored automatically scored by the Michele Sleep Scoring system (YMT) using Rechtschaffen and Kales rules and the American Academy of Sleep Medicine (AASM) scoring guidelines.⁹ Hypopneas were scored using the AASM 1A definition of a 3% decrease in oxygen saturation or the presence of an arousal following an event. An experienced PSG technologist manually reviewed the studies, spending approximately 10 minutes per study. ORP_NR and ORP-9_NR and ORP-9_REM were calculated using the Michele Scoring System after the brief manual review was complete. Participants with sleep apnea based on an apnea-hypopnea index (AHI) ≥ 6 events/h during the first sleep study were excluded. AHI was calculated as number of respiratory events per hour of sleep. Actigraphy results were collected for 1 week using the Micro Motionlogger sleep watch (Ambulatory Monitoring, Inc, Ardsley, New York, United States). Downtime intervals (the period of time between when actigraphic motion markedly decreased to close to zero and the time when it markedly increased) for each night were adjusted manually and Action-W software (version 2.7.3045; Ambulatory Monitoring) was used to calculate sleep time and wake after sleep onset. Proportional integral mode was used for the analysis. A second sleep study was then performed using Prodigy alone (EEG signals only). The median time to the second study was 18 days (range, 5–136).

ORP in Participants With OSA

Participants were recruited through RANA Respiratory Care Group, a home sleep care company operating in Brandon, Manitoba, and were enrolled in the study after providing informed consent. Research ethics approval was obtained from the University of Manitoba Health Research Ethics Board. At the time of the study, the home sleep care company was trialing the combination of the Prodigy and ApneaLink (ResMed, San Diego, California, United States) home sleep apnea monitor and combining the signals with the Michele Sleep Scoring system. Participants were educated on how to apply the sleep monitor either in person at a visit to the home sleep care company or were given written instructions depending on their preference. The diagnostic study was conducted in each participant’s home. Inclusion criteria were an adequate sleep study, age older than 18 years, AHI ≥ 5 events/h or significant flow limited breathing (> 30%) with symptoms, and a CPAP therapy trial start with the home sleep care company. Exclusion criteria were the use of a CPAP mask that did not allow the application of the EEG device, poor adherence during the trial (average daily use less than 4 hours), and unresponsiveness to a request for a second study or failure to use CPAP on the night of the second study. Adherence was measured using Airview adherence software Version 4.2‐4.5.

Participants filled out the FOSQ10 in addition to the ESS, which is collected normally by the sleep care company. An EEG sleep monitor was couriered to each participant’s home about 3 to 4 weeks after starting CPAP therapy, and the participant was asked to wear CPAP and the sleep monitor simultaneously. Adherence was tracked by proprietary adherence software for the CPAP device during the month trial including the night of the sleep study.

The studies were scored using the Michele Sleep Scoring system. The automatically scored sleep studies were reviewed by an experienced PSG technologist spending about 10 minutes per study. Common factors that were reviewed manually included respiratory events, sleep onset, and REM sleep staging. ORP_NR, ORP-9_NR, and ORP-9_REM were determined for the nights with and without CPAP after manual review was complete. Only participants completing the entire study were included in the final analysis.

Statistical Analysis

Mean ORP for each sleep stage, ORP_NR, ORP-9_NR, and ORP-9_REM were calculated. Normality of ORP_NR for healthy individuals was assessed by distribution fitting. A paired t test was used to compare the means. Multiple linear regression analysis was performed to determine factors that could account for changes in ORP_NR between nights both in good sleeper and participants with OSA. The XLSTAT (Version 19.6) software was used for statistical analysis.

RESULTS

Healthy Individuals ORP_NR

Seventy-three individuals consented to participate in the study (Figure 1). Thirty-eight participants (10 females younger than 35 years, 8 males younger than 35 years, 11 males age 35 years or older, and 9 females age 35 years or older) completed the protocol with a mean age of 34.9 ± 7.4 years. The mean questionnaire scores were ESS 3.6 ± 2.4, ISI 2.0 ± 1.6, and FOSQ10 19 ± 1.2. Adequate actigraphy data were collected in 31 of the 38 participants. The mean downtime interval was 513 ± 35 minutes (range 457–607) and estimated sleep time was 455 ± 39 minutes (range 375–553). Common sleep variables for the first and second nights are found in Table 1.

Figure 1. Enrollment of healthy individuals.

Table 1.

Sleep variables for healthy individuals.

ORP values for sleep stages are provided in Table 2. The mean ORP_NR was 0.56 ± 0.17 on the first night and 0.53 ± 0.17 on the second night (P = .42). Excluding participants with no actigraphy data and the one outlier with an average downtime interval greater than 9 hours yielded exactly the same means (n = 30). Eliminating one participant’s data with a prominent beta-delta pattern on manual review during both nights resulted in a mean ORP_NR for all nights of 0.52 ± 0.13. No statistically significant differences were found when comparing sex or age cohorts (20–35 years and 36–50 years). Two standard deviations above the mean for ORP_NR is 0.78. Distribution fitting for all the values in both nights suggested a normal distribution. ORP-9_NR was also calculated and was 0.87 ± 0.22 for the first night versus 0.82 ± 0.2 for the second night (P = .38).

Table 2.

ORP by sleep stage.

There was significant variability in the ORP_NR between the first and second nights in individual participants with a mean difference of 0.024 ± 0.17. The intraclass correlation coefficient between nights was 0.525, indicating only moderate agreement (Figure 2). Multiple regression analysis determined which sleep parameters might explain the difference in ORP_NR between nights. Factors included in the analysis were total sleep time, total recording time, stage N1, N2, N3 sleep percentages, ORP-9_NR, wake after sleep onset, and AI. Most of the change for ORP_NR from the first night to the second night could be accounted for by ∆ORP-9_NR alone (R² = 0.75) (Figure 3). The regression equation is ∆ORP_NR = 0.0102 + 0.5986 × ∆ORP-9_NR. No other measured sleep parameters contributed to the model.

Figure 2. Bland-Altman plot for healthy individuals ORP_NR.

LOA = limit of agreement, ORP_NR = Odds Ratio Product in non-rapid eye movement sleep.

Figure 3. Linear regression model for ∆ORP_NR in healthy individuals.

ORP_NR = Odds Ratio Product in non-rapid eye movement sleep, ORP-9_NR = mean Odds Ratio Product in the first 9 seconds after arousal in NREM sleep, ∆ORP_NR = 0.0102 + 0.5986 × ORP-9.

Technical failure of sleep data collection in this section of the study was 6.7%.

Findings for OSA ORP_NR

Sixty-one participants gave consent to participate in the study. Details of enrollment are provided in Figure 4. The final OSA cohort was composed of 17 females and 23 males. The mean age was 49 ± 11 years and the mean body mass index was 35.3 ± 6.1 kg/m². The mean AHI for the population was 33.5 ± 28.4 events/h. The mean pre/post CPAP FOSQ10 was 15 ± 3.1 versus 18.2 ± 2 (higher is better) and the mean pre/post CPAP ESS was 10.2 ± 4.9 versus 4.4 ± 3.1. Common sleep variables for the first night and the second night are provided in Table 3. The mean hours of CPAP use during the first month of therapy was 6.74 ± 1.3 hours with a 95% CPAP pressure of 11.1 ± 2.6 cm of water. Mean hours of CPAP use during the second night for the cohort was 7.3 ± 1.3 hours.

Figure 4. Enrollment of patients with obstructive sleep apnea.

Table 3.

Sleep variables for patients with OSA.

Changes in ORP for each sleep stage are provided for participants with OSA in Table 2. The mean pre-CPAP ORP_NR was 0.69 ± 0.24 and the mean post-CPAP ORP_NR was 0.57 ± 0.22 (P = .02). The mean pre-CPAP ORP_NR for participants with an AHI less than 20 events/h (n = 16) was 0.66 ± 0.19 and the mean post-CPAP ORP_NR was 0.59 ± 0.21 (P = .31). The mean pre-CPAP ORP_NR for participants with an AHI greater than 20 events/h (n = 24) was 0.71 ± 0.27 and the mean post-CPAP ORP_NR was 0.56 ± 0.23 (P = .04). Twelve of 40 participants did not objectively improve with CPAP therapy. There was no correlation between self-reported measures (ESS or FOSQ10) and baseline ORP_NR. There was no correlation following CPAP between ∆ESS or ∆FOSQ10 and the ∆ORP_NR. The ORP-9_NR pre-CPAP was 0.96 ± 0.25 versus 0.87 ± 0.29 on CPAP (P = .11). The ORP-9_REM pre-CPAP was 1.01 ± 0.42 versus 0.89 ± 0.39 on CPAP (P = .19). A multilinear regression model was constructed to account for ∆ORP_NR from the first night to the second night. Factors that were included in the analysis were total sleep time, total recording time, stage N1, N2, N3 sleep percentages, ORP-9_NR, ORP-9_REM, wake after sleep onset, and AI. Changes in ORP_NR were best modeled by changes in AI, ORP-9_NR and ORP-9_REM in decreasing statistical importance, with an R² of 0.59 (Figure 5). The regression equation is ∆ORP_NR = 0.001 + (0.00583 × ∆AI) + (0.397 × ∆ORP-9_NR) + (0.244 × ∆ORP-9_REM). One participant did not experience REM sleep and was excluded from the multilinear regression analysis.

Figure 5. Linear regression model for ∆ORP_NR in OSA posttreatment.

AI = arousal index, ORP_NR = Odds Ratio Product in non-rapid eye movement sleep, ORP-9_NR = mean Odds Ratio Product in the first 9 seconds after arousal in NREM sleep, ORP-9_REM = mean Odds Ratio Product in the first 9 seconds after arousal in REM sleep, ∆ORP_NR = 0.001 + (0.00583 × ∆AI) + (0.397 × ∆ORP-9_NR) + (0.244 × ∆ORP-9_REM), OSA = obstructive sleep apnea.

Technical failure of sleep data collection in this section of the study was 8.3%.

DISCUSSION

Normal Values for Healthy Individuals

The current investigation establishes approximate normal values for ORP_NR in healthy individuals using the Prodigy EEG recorder. Normal values for ORP_NR obtained from this study are the first reported in the literature in the home environment. Qanash et al reported ORP_NR values for the first and second half of the night of 34 participants with “no pathology” studied in a laboratory and found values of 0.64 ± 0.28 and 0.72 ± 0.33, respectively.³ Total values for the night were not reported but a simple calculation, if the halves of the night were exactly equal, yields a value of about 0.68 compared to 0.52 found in the current study. There are several factors that could have accounted for the difference between the two studies, including different environments for the recordings. First, participants may have slept more deeply in their home environment. Second, the age range between the two cohorts is different with an average age of 34.9 years in the current investigation compared with 43 years in the Qanash study. Third, the “no pathology” group in the Qanash paper was not well characterized except that they had undergone PSG that did not show any pathology compared to the well-defined healthy individuals in the current study. The Qanash cohort was not enrolled specifically for their study but consisted of a collection of PSG results that had been obtained in the provision of clinical care, making it possible the “no pathology group” did not contain healthy individuals.

Recording equipment may have made a difference in values obtained. The performance of the Prodigy has been previously investigated and compared to a standard PSG collection device.⁸ Two problems were identified during that investigation. The first problem was lower impedance in the portable EEG recorder when compared to a large PSG recorder. The lower impedance resulted in higher beta frequencies that raised the ORP in the portable EEG device. The portable EEG device was subsequently adjusted to mimic signals coming from larger recording devices by raising the apparent impedance. The second issue identified was that cardiac artifact in the EEG signal obtained by the portable EEG device also elevated the ORP. An algorithm was developed to remove the cardiac artifact in the portable EEG recordings. After these two adjustments were made, the ORP agreement between signals obtained with a large PSG recorder and the portable EEG resulted in good agreement.

Variability of ORP_NR in Healthy Individuals

The current study found a large variability in ORP between nights in single participants. The standard deviation of the change in ORP_NR between nights was 0.17 and the intraclass correlation coefficient of 0.525 demonstrated only moderate agreement. The change in mean ORP_NR between nights for the group was not significant. Several study design features may have played a role in the variability of sleep depth found between nights. First, it is possible that there was a first-night effect where sleep quality was poorer because of the annoyance of the recording device. However, if there is a first-night effect it is thought to be small for the whole population studied based on the small difference between the first and second night mean ORP_NR. The first- night effect could have had a noticeable effect on individual variation. Second, participants did not wear the cardiorespiratory monitoring equipment for the second night of the study and this may also have contributed to the variability found between nights. The finding of ORP_NR variability suggests that more than a single night will be needed to obtain a good estimate of a single person’s ORP_NR.

A multiple linear regression analysis looked at whether changes in sleep architecture could explain the differences between nights. The only important factor revealed by the analysis was the ∆ORP-9_NR, which was highly correlated with the ∆ORP_NR. This suggests that the ORP-9_NR is variable in individuals between nights. The ORP-9_NR has been described as a measure of postarousal sleep dynamics or, to put it more simply, a measure of sleep pressure. Younes and Hanly⁴ previously speculated that the ORP-9_NR may be stable between nights in individuals based on the finding that ORP-9 did not change with the addition of CPAP. Most studies examined during their investigation into postarousal sleep dynamics were split-night studies rather than two separate nights, with and without CPAP. Qanash et al³ demonstrated that ORP_NR did not change in split-night studies after treatment with CPAP and speculated that this was because ORP_NR tends to rise as the night progresses, as demonstrated in participants with “no pathology”, and that CPAP prevented this rise. It is possible that Younes and Hanly⁴ did not see a change in the ORP-9_NR in split-night studies during treatment with CPAP because the ORP_NR did not change.

A Decrease in ORP_NR During Treatment With CPAP

The current study demonstrates that ORP_NR improves in a population of patients with OSA after treatment with CPAP for 1 month. This is statistically significant in patients with OSA and AHI > 20 events/h, but not in those with AHI < 20 events/h, though there was a trend toward improvement as well in the less severely affected group. Not all patients objectively improved their sleep depth with CPAP. Data from the healthy individuals reported in this study demonstrated that there can be wide variation in the ORP_NR between nights. Based on this finding, lack of improvement in ORP_NR in some of the patients with OSA might be explained by night-to-night variability. However, it cannot be excluded that some people with OSA have less deep sleep with CPAP therapy where the stimulus of CPAP outweighs the removal of the stimulus of sleep- disordered breathing. It is also possible that the first-night effect of the monitoring equipment may have worsened sleep during the diagnostic study. A better study design would have included two diagnostic studies to look at the first-night effect.

There was no correlation between sleep depth as measured by the ORP_NR and self-reported measures such as ESS or FOSQ10 at baseline or between ∆ESS or ∆FOSQ10 and the ∆ORP_NR following treatment with CPAP. This finding is provocative in that evaluating sleep quality using an objective measure versus using self-reported measures seems not to be measuring the same thing. Another possibility is that the measurement of ORP_NR for a single night during CPAP therapy may not provide a complete picture of sleep quality over the course of 1 month of therapy. The ∆ESS or ∆FOSQ10 provide a more global measure of improvement in daytime functioning and sleepiness, taking into account all nights of sleep after starting CPAP. There was significant variability in the ORP_NR of healthy individuals and it would seem likely that this may be present in those with OSA. It may be necessary to measure more than a single night to find a correlation between self-reported measures and an objective measure such as ORP. It is also possible that the measurement device itself could have introduced variability in objective sleep quality that may not be present on other nights when the EEG is not collected. Further research is necessary to determine whether 2 or 3 nights may provide a better correlation to self-reported measures.

Qanash et al previously investigated the change in ORP_NR in split-night studies for OSA.³ The change in ORP_NR was variable and bidirectional in that study after treatment with CPAP. The mean ORP_NR declined by 0.04 units in the second half of the night and was not statistically different than the first half of the night. The authors speculated that the removal of the respiratory stimulus for arousal was counterbalanced in some participants by the added stimulus of CPAP therapy, which may have accounted for the lack of change in the ORP_NR in the group. Another finding that was reported in the split-night study was that ORP_NR tends to increase as the night progresses. A “no pathology” group in that study demonstrated an increase in the ORP_NR by 0.12 units from the first half to the second half of the night, indicating lighter sleep (P < .0002). This raises the possibility that CPAP did have a significant effect on sleep depth, given that a rise in the ORP_NR under normal conditions would have been expected rather than no change. The study by Qanash et al does not address the effect of CPAP on sleep depth over an entire night or in the individual’s home environment.

Studies of Sleep Architecture Following Treatment With CPAP

Previously demonstrated objective changes in sleep architecture in participants with OSA after treatment with CPAP have included increased percentages of slow wave sleep and REM sleep, and reduced stage N1 sleep.^10-16 In the current report, similar changes were identified (Table 3). Changes in stage N2 sleep have been more variable during treatment with CPAP with reports of increased stage N2 sleep,¹⁷ unchanged stage N2 sleep,^12,13 and reduced stage N2 sleep.¹⁵ Most of the studies looking at changes in sleep architecture after CPAP therapy included a control night and a CPAP titration night.^10,13-15,17

Three studies included a PSG done at least 1 month after CPAP was started. Parrino et al¹¹ studied 10 participants with OSA and found that changes in REM sleep and deep sleep were similar to controls without OSA 1 month after starting CPAP, but were still improved compared to the baseline study without treatment. McArdle et al studied 22 participants on 5 nights including 1 night 30 days after CPAP initiation and found that stage N1 sleep was less and deep and REM sleep were greater than at baseline on all nights after treatment with CPAP, but that the absolute change decreased from the first night of CPAP to the 30th day with CPAP.¹² Statistical analysis on comparisons between nights in participants with OSA was not reported. Antczak et al¹⁶ studied three groups of participants with OSA (nonobese, obese, and very obese) with a baseline study, 2 nights during treatment with CPAP and 3 months after initiation of CPAP. Stage N1 sleep decreased in all groups after treatment with CPAP and slow wave sleep and REM sleep increased in all groups. Three months after treatment with CPAP similar changes were found compared to the second night of CPAP in all groups with the exception of the very obese group, in which slow wave sleep and REM sleep decreased statistically compared with the second night of CPAP therapy.

Changes in AI have also been studied in participants with OSA after treatment with CPAP when compared to a control group. A meta-analysis was reported by Liu et al¹⁸ which found that AI decreased by 10.76 arousals/h (95% confidence interval: 7.84–13.69) after treatment with CPAP, which is similar to what is reported here (Table 3).

Determinants of Changes in ORP_NR in Patients With OSA After CPAP Initiation

The changes in ORP_NR between nights are mostly accounted for by the change in AI, ORP-9_NR, and ORP-9_REM. It seems intuitive that if the AI decreases that sleep depth will improve. If there are fewer arousals because of treatment with CPAP, it would be expected that the ORP_NR would decrease because there are fewer times the ORP rises. Younes et al⁴ demonstrated that the ORP decreases quickly initially after an arousal, followed by a slower decline if another arousal does not occur.

The ORP-9_NR is also part of the explanation for change in ORP_NR. The low correlation coefficient of 0.107 between AI and ORP-9_NR in the correlation matrix of the linear regression analysis suggests minimal correlation exists between AI and ORP-9_NR. The correlation between ORP_NR and ORP-9_NR is very strong in the healthy individuals where the mean AI did not change significantly (R² = 0.75). Younes et al⁴ found a similar correlation but did not find a significant change in ORP-9_NR before and after CPAP as mentioned previously. In the current investigation, the change in ORP-9_NR before and after CPAP did not reach significance (P = .11) but was trending in the direction of improvement. If a larger group of participants had been studied, a statistically significant change may have been found. Another possible reason why the ∆ORP-9_NR is important in the determination of overall ∆ORP_NR is that respiratory arousals may elicit higher arousal intensity and a higher ORP-9 compared with other nonrespiratory arousals. The elimination of respiratory arousals may account for the role the ORP-9_NR plays in the model of ∆ORP_NR.

The observation that the ORP-9_REM is also significant in the linear regression model for ∆ORP_NR is an interesting finding. Nothing has been previously reported on ORP-9_REM. What happens to depth of sleep in REM sleep may be important for subsequent NREM sleep depth. If the depth of sleep coming out of REM sleep is deeper after CPAP therapy, then it might be expected that the subsequent NREM sleep may also be deeper. A possible criticism of including the ORP-9_REM in the multiple regression analysis is that the number of arousals in REM sleep is smaller compared to NREM sleep making this measurement less reliable.

Limitations

This study has a number of limitations. First, a much larger group of healthy individuals would have been desirable to estimate normal values. A wider range of ages also would have been helpful. The investigators were surprised at how difficult it was to find healthy individuals who met the criteria for inclusion. Second, it would have been preferable to have an adjustment night for healthy individuals to take into account a first-night effect. Third, study design would have been improved if healthy individuals had worn the cardiorespiratory monitor the second night to eliminate this as a variable. Fourth, each patient with OSA was only studied once before and after treatment. Given the wide variation in ORP_NR from night to night in the healthy individuals, a better estimate in treatment effect may have been obtained if 2 nights before and after treatment had been done. Fifth, because of the number of participants enrolled, this study was not able to confirm significant improvement in sleep depth in more mildly affected participants with OSA. Sixth, our study does not include anyone who did not adapt well to CPAP. These participants may be quite different from those who successfully completed the protocol.

CONCLUSIONS

In conclusion, this study has provided information regarding ORP_NR for healthy individuals obtained in the home environment that can be used as normal values in persons age 20 to 50 years. The study has also demonstrated that the treatment of sleep apnea, in persons who can adhere to CPAP therapy, decreases ORP_NR but only in more severely affected individuals, providing objective evidence of improvement in sleep depth with a single variable. The lack of correlation between self-reported measures such as the ESS and FOSQ10 and the ORP_NR may limit its usefulness in clinical evaluation of sleep disorders. The study has also demonstrated that postarousal sleep dynamics (ORP-9_NR), or the change in sleep pressure immediately following an arousal, is not fixed as has been previously suggested. The study also confirms that EEG recordings in the home environment are logistically possible and can be accomplished without a high technical failure rate.

DISCLOSURE STATEMENT

Charles Penner has received remuneration from Cerebra Health Inc. in a consulting capacity and received remuneration for conducting the study. He has done clinical work in a RANA Respiratory Care Group clinic (Brandon) but received no compensation from RANA Respiratory Care Group for this work. He also has served as an unpaid medical director for RANA Respiratory Care Group pulmonary function laboratories in Brandon, Manitoba and Calgary, Alberta. He does not have a financial stake in either company. Bethany Gerardy is an employee of Cerebra Health Inc. Rob Ryan was an employee of RANA Respiratory Care Group at the time the study was conducted. Mark Williams was an employee of Cerebra Health at the time the study was conducted. Work for this study was performed at RANA Respiratory Care Group offices and Cerebra Health offices and research participants’ homes. All authors have reviewed and approved the manuscript.

ABBREVIATIONS

A/AW

arousal/awakening

AASM

American Academy of Sleep Medicine

AHI

apnea-hypopnea index

AI

arousal index

CPAP

continuous positive airway pressure

EEG

electroencephalogram

ESS

Epworth Sleepiness Scale

FOSQ10

Functional Outcomes of Sleep Questionnaire - shorter version

HSAT

home sleep apnea test

ISI

Insomnia Severity Scale

NREM

non-rapid eye movement

ORP

Odds Ratio Product

ORP_NR

Odds Ratio Product NREM sleep

ORP-9_NR

Odds Ratio Product mean 9 seconds after the end of an arousal or awakening in NREM sleep

ORP-9_REM

Odds Ratio Product mean 9 seconds after the end of an arousal or awakening in REM sleep

OSA

obstructive sleep apnea

PSG

polysomnography

REM

rapid eye movement

YMT

Younes Medical Technologies