Comparative analysis of sleep physiology using qualitative and quantitative criteria for insomnia symptoms

Abstract

Despite decades of research, defining insomnia remains challenging due to its complex and variable nature. Various diagnostic systems emphasize the chronic nature of insomnia and its impact on daily functioning, relying heavily on patient self-reporting due to limitations in objective measures such as polysomnography (PSG). Discrepancies between subjective experiences and objective PSG results highlight the need for more nuanced approaches, such as electroencephalogram (EEG) spectral analysis, which reveals distinct patterns of high-frequency activity in individuals with insomnia. This study explores EEG markers of insomnia by integrating subjective reports with objective physiological markers, specifically ORP (Odds-Ratio-Product) and spectral features, to address inconsistencies found in previous research and clinical settings. Qualitative and quantitative definitions of insomnia are contrasted to highlight differences in sleep architecture and EEG characteristics. The research aims to determine whether groups defined by weekly frequency and daily duration of symptoms have different distribution patterns and which physiological characteristics best distinguish insomnia patients from controls. Our findings suggest that ORP, as a dependent variable, captures the most significant differences in the independent variables across the model. Elevated beta power in insomnia patients indicates increased cortical arousal, supporting the perspective of insomnia as a hyperarousal disorder. Future research should focus on using ORP to enhance the understanding of sleep disturbances in insomnia. Comprehensive evaluation of insomnia requires integrating qualitative, quantitative, and neurophysiological data to fully understand its impact on sleep architecture and quality.

Graphical Abstract

Open in new tabDownload slide

Despite decades of research, defining insomnia remains challenging. Insomnia can encompass symptoms of difficulty initiating or maintaining sleep (DIMS) and/or early morning awakening (EMA), historically classified under the term DIMS [1]. Initially, it was thought that these three distinct symptoms may represent different subtypes of insomnia [2], but this was challenged by data showing that the majority of individuals with insomnia have a combination of these problems [3] and that symptoms can change over time [4]. Various guidelines [5–7] offer diagnostic criteria for insomnia disorder, which include DIMS plus an indicator of chronicity and subjective difficulties with daytime functioning or overall well-being [5–8]. Research suggests that three months is the most appropriate duration to define chronic insomnia [8].

There have been many attempts to improve upon these diagnostic criteria by creating a more concrete operationalization of insomnia symptoms beyond general dissatisfaction with sleep continuity. In terms of subjective criteria, Linchestein and colleagues [9] sought to establish quantitative criteria by examining whether the duration of sleep onset latency (SOL) or wakefulness after sleep onset (WASO) provided optimal differentiation between individuals with insomnia and good sleepers. They found that a threshold of 31 minutes or greater on SOL or WASO was optimal for identifying those with insomnia. The search for objective criteria has been challenging over the past 40 years, with studies comparing individuals with insomnia to good sleepers often failing to show statistically significant differences in PSG measures of sleep continuity and sleep architecture despite large group differences in subjective sleep estimates [10, 11]. It has been argued that this subjective/objective discrepancy is due to methodologic limitations with standard PSG scoring and interpretation rather than a true lack of differences in objective sleep [12]. Indeed, studies of sleep microarchitecture have found elevations of high-frequency EEG activity in the beta range (13–30 Hz) during PSG-scored sleep [13]. Beta EEG activity is more typically seen in awake subjects and is thought to reflect cortical processing of sensory information [14]. This state is perceived as wakefulness, and, indeed, the amount of beta activity correlates with the magnitude of the subjective/objective discrepancy [15]. Others have proposed additional EEG metrics that could differentiate individuals with insomnia from good sleepers [13, 16–18]. Ultimately, subjective and objective sleep metrics seem to capture fundamentally different and mostly uncorrelated sources of information.

To summarize EEG microarchitecture research in insomnia, Zhao et al. [13] conducted a meta-analysis of 24 studies, finding small and inconsistent differences in beta, delta, alpha, and sigma power between insomnia patients and controls. The observed effect sizes (≤0.5) suggested that spectral power alone is not sufficient to diagnose insomnia due to significant overlap between groups. Given these inconsistent and often small effects, there is a need for multiple and novel EEG measures to accurately reflect the physiological state of insomnia. One such metric is the ORP [19], a novel EEG-based continuous measure of sleep depth that provides an advancement over traditional methods of quantifying sleep depth by offering a continuous index of arousal and sleep stage.

Building off this literature, the present study evaluates a range of EEG metrics of macro- and microarchitecture, as well as ORP, between insomnia cases and controls, defined based on the main symptoms of insomnia, daily dysfunction due to insomnia, and symptom duration according to DSM-5, using both qualitative and quantitative definitions. The objectives of this study are (1) to determine which sleep EEG features differentiate individuals with insomnia from good sleepers and (2) to assess differences using both qualitative and quantitative criteria for defining insomnia.

Materials and Methods

The present study utilized data from the Sleep Apnea Global Interdisciplinary Consortium (SAGIC; http://www.med.upenn.edu/sleepctr/sagic.html). The study received the necessary approval from the Institutional Review Board at the University of Pennsylvania, and all participants provided informed consent.

SAGIC participants

SAGIC is an international collaboration focused on advancing research on obstructive sleep apnea (OSA). The primary goal of SAGIC is to establish a large, multinational cohort characterized by comprehensive phenotyping to better understand both overall and region-specific presentations and risk profiles associated with OSA. Participants in SAGIC included in the present study were recruited from 10 sleep disorders centers across 6 different countries: the United States (specifically, the University of Pennsylvania and The Ohio State University), Australia (Royal North Shore Hospital, Sydney and Sir Charles Gairdner Hospital, Perth), Germany (Charité University Hospital, Berlin), Brazil (Médicado Instituto do Sono), Taiwan (Chang Gung Memorial Hospital), and China (Peking University People’s Hospital and Peking University International Hospital, Beijing and Ruijin Hospital, Shanghai). These participants were originally selected because they visited sleep disorders centers with subjective sleep complaints and underwent a one-night clinical diagnostic PSG. For these analyses, we included only those individuals who had an apnea-hypopnea index (AHI) < 5 events per hour, indicating no evidence of OSA.

SAGIC questionnaire

Participants completed an extensive questionnaire covering various sleep symptom-related inquiries; we focused specifically on insomnia questions, shift work questions, demographics, and daily functioning as indicators of insomnia. Participants indicated the degree to which they experienced problems with SOL, WASO, and EMAs with options ranging from 0 (never or less than once per week) to 4 (every night or almost daily); responses of 5 (“do not know”) were excluded. They also provided quantitative estimates of their average SOL and WASO. The survey also contained nine items in which respondents had to indicate how insomnia impacted their lives during the past month, with response options ranging from “not at all” to “extremely.” Items asked about the impact of insomnia on work, social life, irritability, concentration, and the domains of functioning.

Definitions of insomnia cases based on qualitative criteria and quantitative criteria

We used two subjective methods to define insomnia cases that approximate current diagnostic systems, particularly focusing on the DSM-5-TR. A qualitative case definition required an individual to indicate problems with SOL, WASO, and/or EMA at least 3 nights per week to mirror the frequency criterion from DSM-5-TR. For the quantitative case definition, respondents had to report an average SOL and/or WASO exceeding 31 minutes. For both qualitative and quantitative definitions, symptoms had to have persisted for at least 3 months based on diagnostic criteria. Participants were categorized as the control group if they did not meet either the qualitative or quantitative criteria for insomnia. However, it is possible that participants who met only the quantitative or qualitative criteria could be classified differently between the two approaches (i.e. as having insomnia by one definition but as a control by the other).

EEG measures

The total absolute EEG power and power within specific frequency ranges (0.33–2.33 Hz [slow delta], 2.67–6.33 Hz [theta], 7.33–12.0 Hz [alpha], 12.33–14.0 Hz [sigma], 14.33–20.0 Hz [beta1], 20.33–35.0 Hz [beta2], and >35.0 Hz [gamma-omega]) were computed [20]. We also used absolute EEG power in our analysis. The sampling frequency of the EEG recordings differed across centers, ranging from a minimum of 200–512 Hz. The scoring filters used a band-pass filter of 0.3–35.0 Hz.

ORP was computed as a novel continuous marker of sleep depth, ranging from 0 to 2.5, with higher values indicating greater wakefulness and lower values indicating deeper sleep [21]. ORP is calculated using Fourier analysis in consecutive 3-second artifact-free epochs and involves ranking power in each of four frequency ranges (0.33–2.33, 2.67–6.33, 7.33–14.00, and 14.33–35.00 Hz) on a scale from 0 to 9. The four ranks are concatenated to produce 10 000 four-digit numbers (BIN numbers), with each number describing a unique combination of powers in the four frequency groups. Each BIN number is assigned a probability (0%–100%) of occurring during epochs manually scored wake or during arousals, as determined in development files [25] used as reference in new studies. These probability values are then averaged (and divided by 40, the percentage of awake epochs in the developmental files [25]) to get subject-specific estimates of average ORP overall and during each sleep stage. Additionally, the rate of sleep depth progression (ORP-9) was determined by averaging 3-second ORP values in the first 9 seconds following the end of each arousal, with a lower ORP-9 indicating a swifter return to deep sleep [22]. This index essentially controls sleep depth since a slower decline to deep sleep leaves the patient vulnerable to arousal from weak stimuli and failure to reach deep levels of sleep [22].

Statistical analyses

All analyses were performed in R version 4.3.1. Categorical variables were summarized using frequencies and percentages and then compared between two definitions of insomnia cases—one based on the frequency of insomnia symptoms per week (qualitative cases) and the other on the duration of symptoms (quantitative cases)—using Kappa tests. Continuous variables were summarized using means with SD and compared between case and control groups using multivariate analysis of covariance (MANCOVA) with covariates such as sex and age to reduce variance attributed to these influences. Separate models were computed for qualitative and quantitative case definitions and each category of EEG variables (macroarchitecture, microarchitecture, and ORP). Statistically significant MANCOVA models (p < .05) were followed with univariate ANCOVA models. We only considered single factors to be significantly different if the corresponding univariate ANCOVA was significant after applying the Hochberg correction to the significance probabilities. Cohen’s d was used to quantify the effect size of the mean difference between groups for each dependent variable and can be interpreted as small (d = 0.2), medium (d = 0.5), or large (d = 0.8) [23]. This assessment provides insights into the practical significance of differences across various sleep-related variables.

To understand the agreement between quantitative and qualitative insomnia definitions, we calculated a simple Kappa coefficient (K) to compare the agreement between qualitative and quantitative definitions. Using guidelines provided by Landis and Koch [24], Kappa values can be interpreted as poor (K < 0.00), slight (0.00–0.20), fair (0.21–0.40), moderate (0.41–0.60), substantial (0.61–0.80), and almost perfect (0.81–1.00).

Results

Sample characteristics

The SAGIC data included 1693 individuals with an AHI < 5 events/hour. Of these, 599 (35.4%) participants were excluded from the analysis because they reported being shift workers or failed to provide work schedules; an additional three observations with missing values of age were also excluded. Finally, 136 individuals with missing values for all 3 major insomnia symptoms based on qualitative criteria or missing values for all 9 variables for daily dysfunctioning were excluded. This resulted in a total sample of 955 (56%) individuals eligible to be included in the proposed analyses (see Figure 1).

For defining insomnia based on qualitative criteria, there were complete data for 955 individuals, 350 of whom were categorized as cases with insomnia and 605 as controls (see Table 1). Overall, the sample was 45.2 ± 14.6 years old, and 49.2% were female. The average age of the insomnia group was 48.5 ± 14.4 years, and 58.0% were female, while the average age of the control group was 43.3 ± 14.4 years and 44.1% were female.

For the quantitative case definition, 527 individuals had complete data, with 194 classified as insomnia and 333 classified as controls (see Table 2). Overall, the sample was 46.7 ± 14.5 years, and 56.4% were females. The insomnia group had an average age of 49.0 ± 13.6 years, and 62.4% were females, whereas the control group had an average age of 45.4 ± 14.8 years and 52.9% were females.

These qualitative and quantitative sample sizes represent the number of observations included in at least one analysis. The actual number of observations varied across analyses, ranging from a maximum of 824 (Macroarchitecture + qualitative case) to 267 (ORP + quantitative case), depending on the missing values related to macroarchitecture, microarchitecture, and ORP. The variation in sample size depending on analyses was due to missing values in each dependent variable.

Comparative analysis of insomnia definitions using kappa coefficient

A kappa coefficient was utilized to evaluate agreement between qualitative and quantitative criteria. Results indicated moderate agreement between the two definitions, with a Kappa (95% CI) of 0.466 (0.413, 0.518). The overall percent agreement between the two definitions was 73%. As shown in Table 2, there were 142 (26.9%) individuals that were categorized as having insomnia based on only one definition, reflecting some discrepancy in the specific traits being captured by these two definitions.

PSG sleep macroarchitecture

The MANCOVA for the qualitative case definition indicated a statistically significant difference in the dependent variables between insomnia and control groups (Pillai’s Trace = 0.02, F(7, 814) = 2.342, p = .02; see Table 3). Subsequent one-way ANCOVAs on the seven dependent variables revealed nominally significant differences in total sleep time, WASO, SOL, and total stage R sleep. Differences in SOL remained statistically significant after the Hochberg correction, with the insomnia group experiencing 12 minutes longer SOL on average compared with the control group, denoting a small effect size (Cohen’s d = 0.21).

Similarly, the MANCOVA for the quantitative group definition indicated a statistically significant difference in dependent variables (Pillai’s Trace = 0.03, F(7, 448) = 2.226, p = .03; see Table 3). Follow-up ANCOVAs for the seven dependent variables revealed nominally significant differences in WASO (14 minutes longer in the insomnia group; Cohen’s d = 0.21). However, this result did not maintain statistical significance after the Hochberg correction.

PSG sleep microarchitecture

There was a significant difference in absolute EEG power characteristics between insomnia cases and controls using qualitative criteria (Pillai’s Trace = 0.06, F(7, 522) = 4.579, p < .0001; see Table 3), with subsequent one-way ANCOVAs revealing nominally significant differences for alpha, sigma, beta1, and beta2 power. The difference in beta1 remained significant after the Hochberg correction, with the insomnia group experiencing higher beta1 activity on average compared with the control group (Cohen’s d = 0.25).

Significant differences in absolute EEG power characteristics were also found when using quantitative criteria (Pillai’s Trace = 0.06, F(7, 322) = 3.182, p = .003; see Table 3), with one-way ANCOVAs showing nominal differences in alpha, sigma, beta1, and beta2. Differences in beta1 and beta2 were statistically significant after the Hochberg correction, with the insomnia group experiencing higher activity on average than the control group for both beta1 (Cohen’s d = 0.35) and beta2 (Cohen’s d = 0.37).

ORP characteristics

The MANCOVA showed significant differences in ORP characteristics based on qualitative criteria (Pillai’s Trace = 0.07, F(8, 423) = 4.083, p < .0001). Subsequent one-way ANCOVAs on the eight dependent variables revealed significant results for wake ORP, REM ORP, overall ORP, and ORP-9 (see Table 4). Differences in REM and overall ORP and ORP-9 remained significant after the Hochberg correction. The insomnia group had lighter sleep depth based on REM ORP (Cohen’s d = 0.24), overall ORP (Cohen’s d = 0.25), and less sleep drive after arousal based on ORP-9 (Cohen’s d = 0.29) (see Table 4 and Figure 2).

The MANCOVA using a quantitative group definition was not statistically significant (Pillai’s Trace = 0.06, F(8, 256) = 1.921, p > .05) (see Figure 2).

Discussion

This research highlights significant differences in sleep physiology between individuals with and without insomnia, using qualitative and quantitative criteria. Our findings emphasize that these criteria have partially overlapping but also distinct patterns of sleep EEG features. We found that qualitative criteria captured the group differences between insomnia and control groups in terms of ORP values, which index sleep depth. Although the quantitative criterion did not reach statistical significance, the model using qualitative criteria was statistically significant. This suggests that, despite the minimal overall difference between the two criteria, the qualitative approach may be more effective in distinguishing between the groups in this context. These results demonstrate the importance of sleep quality in evaluating insomnia. In other words, ORP effectively reflects sleep impairments, with shallower sleep being perceived as lower quality compared with quantitative sleep aspects. These findings are further supported by comparing qualitative and quantitative criteria on ORP values. Lastly, increased ORP values and beta activity may serve as reliable biomarkers of cortical hyperarousal. This study marks the first case-control study to examine two different group definitions, symptom severity derived from individuals’ subjective perspective and symptom severity based on quantified criteria, with sleep physiological features quantified across different layers—macro PSG sleep architecture, micro PSG sleep architectures, and ORP. This supports the operationalization of insomnia as a disorder of poor sleep quality and persistent heightened neural activity, potentially interfering with the ability to achieve and maintain deep sleep.

Demographic characteristics

We observed a relatively similar distribution of men and women in our sample of individuals without OSA (AHI < 5). Given that the sample consisted of patients undergoing sleep studies for suspected OSA at sleep disorders centers, it might have been suspected that our sample would have more males, given that OSA is generally more prevalent among men. Indeed, the sample overall consisted of more males, but once those with AHI > 5 were excluded, the remaining sample had more of a balance. Even after excluding individuals with OSA, the overall sample in this study may include a higher proportion of men compared with typical insomnia populations [25]. This suggests that the demographic characteristics of this sample may differ from those observed in general insomnia cohorts.

Macro PSG sleep parameters related to insomnia symptoms

These results suggest that longer subjective SOL is a more sensitive indicator of subjective sleep complaints, while results for subjective WASO were more variable. Interestingly, the impact of perceived sleep continuity disturbances may vary, potentially influenced by factors such as age [26], which our findings show to have a more pronounced effect on WASO. Although subjective SOL and WASO are both measures of sleep continuity, this result suggests that they might reflect distinct aspects of insomnia. Harvey [27] pointed out that cognitive arousal is often more pronounced before than during sleep and thus may be more likely to impact SOL than WASO. Furthermore, a recent study by Maltezos et al. [28] showed that SOL and WASO misperceptions differ depending on the methods used in insomnia patients; for instance, SOL is over-reported in the insomnia group at their home, but WASO is stable at their home and over-reported evaluated at the laboratory. Additionally, Kalmbach et al. [29] found that high levels of nocturnal cognitive arousal were linked to longer SOL, while a connection was not stable between nocturnal cognitive arousal and WASO, suggesting that the impact of cognitive arousal may differ across these measures of sleep disruption.

EEG spectral features related to insomnia symptoms

Our results also demonstrate that the overall pattern of EEG spectral features across different insomnia criteria supports increased beta activity during sleep as an insomnia biomarker, which has been interpreted as an indicator of cortical hyperarousal [13, 14, 30]. More specifically, power in both the beta1 and beta2 ranges indicate cortical arousal across the night, which could contribute to reduced sleep quality. The beta EEG effects were seen to some extent in both qualitative and quantitative case definitions, suggesting that hyperarousal is relevant irrespective of how insomnia is defined. Although increased beta activity is typically associated with cortical hyperarousal, a meta-analysis paper on EEG spectral analysis in insomnia has suggested that different features depend on NREM or stage R sleep [13]. For example, insomnia exhibited increased alpha and sigma power during stage R sleep and decreased delta power and increased levels of theta, alpha, and sigma power during NREM sleep [13].

ORP features related to insomnia symptoms

The novel use of the ORP as a biomarker for sleep depth revealed that individuals with insomnia tend to experience lighter and more fragmented sleep, a finding consistent with ORP studies in insomnia populations [20]. This was also evident from the higher ORP values across all sleep stages in the insomnia group compared with controls. Specifically, REM and whole-night ORP, and ORP-9 showed consistently significant group differences using both definitions, suggesting that the insomnia group might experience more difficulty returning to deeper sleep after arousal and spend more time in lighter stages of sleep, indicating hyperarousal. Additionally, previous research with the Wake Electroencephalographic Similarity Index, which is used to measure hyperarousal features such as ORP, has indicated an elevation across all sleep stages except N3 [31]. The insomnia group had a higher likelihood of waking from all sleep stages, indicating that individuals with insomnia exhibit more wake-like brain activity during these stages. This supports the observation that higher ORP values indicate more wake-like activity during these stages. Furthermore, REM ORP has been regarded as highly reproducible and has robust features to detect insomnia [20], which aligns with our findings. However, previous research suggested that the insomnia group tends to have low sleep pressure, characterized by a high wake ORP, and that their NREM sleep is lighter compared with individuals without insomnia [20, 32]. However, in our research, wake ORP and NREM ORP did not show significant differences between groups. This may be because the entire group consisted of people who visited sleep disorders centers, suggesting that even the control group might have included people with other sleep disturbances or lower sleep quality. As a result, the differences between the two groups may have made it difficult to detect significant group differences.

Clinical implications

From a clinical perspective, the implications of these findings are considerable. Results suggest that ORP may be an innovative assessment method that better reflects aspects of sleep quality. Additionally, it might be important to assess insomnia by inquiring about SOL and WASO and determining the extent to which each symptom is problematic for the patient rather than the quantity itself. Future studies should explore the distinct mechanisms underlying SOL and WASO to better understand the qualitative aspects of each symptom and how they impact the patient beyond quantitative measures. Additionally, research should investigate how these qualitative aspects influence the relationship between subjective and objective sleep measures, allowing evaluations to be more effectively tailored to the individual needs and complexities of insomnia.

PSG is not a standard diagnostic tool for assessing insomnia in clinical practice, even though it provides detailed and objective measurements of sleep architecture, which are crucial for assessing sleep quality in insomnia [33]. This is largely due to the lack of consistent abnormalities research from PSG [10] and limited diagnostic utility in uncomplicated cases of insomnia. However, the results of our study suggest that PSG could offer additional value in specific clinical contexts. For instance, it captures detailed sleep architecture data, revealing markers of hyperarousal, such as increased beta power during sleep and high ORP values in the insomnia group, which may enhance our understanding of the pathophysiological mechanisms of insomnia, thus guiding more personalized treatment approaches. In addition, our exploration of qualitative and quantitative case definitions of insomnia underscores the potential for identifying distinct phenotypes of insomnia with PSG and other objective measurements, as misperception of sleep in insomnia can vary significantly among individuals, and PSG can help identify subtypes based on these discrepancies [34]. By leveraging both qualitative and quantitative frameworks with objective data, future research may advance the development of a phenotype-based classification of insomnia, enabling more precise and effective interventions.

Limitations

Our research comes with several limitations. First, it is important to note that our participants were composed of people who visited sleep disorders centers, which suggests that they might have underlying sleep complaints, even if they do not have a sleep disorder diagnosis. Therefore, any conclusions and generalizability drawn from this data should be approached cautiously, considering the potential influence of these underlying issues. Second, while we intentionally defined insomnia using both qualitative and quantitative criteria in this study to raise questions about the current diagnostic framework for insomnia and its limitations, it is a limitation that the case definitions were determined by self-report, rather than clinician assessment. When defining insomnia, symptom-based subjective severity and frequency, perceived poor sleep continuity, daytime dysfunction, and chronicity for more than 3 months were considered, though. Third, although it is clear that medication use affects insomnia symptoms, sleep patterns, and EEG activity [35, 36], we did not exclude participants on medications and did not adjust medication effects; this information was not reliably captured in the SAGIC cohort. Lastly, a key limitation of our study is the lack of an intervention, such as cognitive behavioral therapy for insomnia, to assess pre- and posttreatment changes. Previous research found the therapeutic of cognitive behavioral therapy to induce neurophysiological changes linked to improved sleep [37], suggesting that future studies incorporating such interventions could provide deeper insights into treatment effects.

Conclusion

In conclusion, our findings indicate that the insomnia group exhibited significant increases in general ORP values based on the qualitative criteria, while beta band power increased in the insomnia group across both the qualitative and quantitative criteria, suggesting continuous cortical hyperarousal. Additionally, the qualitative group showed longer SOL, potentially related to general sleep complaints. In contrast, WASO was influenced by perceived sleep continuity. These findings support the hypothesis that insomnia may stem from hyperarousal. Methodologically, our results advocate for both using qualitative and quantitative criteria in clinical assessment. Future research should expand ORP distribution studies, employ a broader array of analytical methods to explore EEG spectral variations, and evaluate both qualitative and quantitative criteria across different insomnia subtypes.

Acknowledgments

We thank Bruno Saconi for his help in performing the kappa analysis.

Funding

This research was supported by the National Institutes of Health (NIH) under multiple grants, including P01 HL160471-01A1 and R01 HL173043 from the National Heart, Lung, and Blood Institute (NHLBI), and R01 NR018836 from the National Institute of Nursing Research (NINR).

Conflict of interest statement

Dr Thomas Penzel reports relationships with Bayer, Bioproject, Cerebra, Idorsia, Jazz Pharmaceuticals, Sleepimage, Lowenstein Medical, Philips, and the National Sleep Foundation, including consulting or advisory roles and speaking and lecture fees. Dr Bhajan Singh has received speaking and lecture fees from SomnoMed Australia. Dr Peter A. Cistulli reports relationships with ResMed and SomnoMed, including consulting or advisory roles and funding grants, as well as consulting or advisory roles with Signifier Medical Technologies, Bayer, and Sunrise Medical. Dr Richard J. Schwab reports funding grants from ResMed, Inspire, and CryOSA, consulting or advisory roles with Eli Lilly, and patents with royalties paid to UpToDate and Merck Manual. Dr Magdy Younes is the inventor of ORP, holds a patent on the method, and receives royalties from Cerebra Health for the licensed technology. Other authors declare that they have no known competing financial interests that could have appeared to influence the work reported in this paper. Dr Richard J. Schwab reports his position on the Medical Advisory Board for eXciteOSA. Other authors declare that they have no known personal relationships or other nonfinancial interests that could have appeared to influence the work reported in this paper.

Data Availability

Data can be made available upon request and approval from the SAGIC investigators.

References