INTRODUCTION
Obstructive sleep apnea (OSA) is a prevailing sleep-related breathing disorder, characterized by recurring instances where breathing halts during sleep. Among the various symptoms associated with OSA, excessive daytime sleepiness (EDS) emerges as a prominent hallmark, attributable to the disruptive effects of sleep fragmentation and intermittent hypoxia inherent to the negative health condition [1]. Individuals suffering from EDS often find themselves battling drowsiness and struggling to maintain wakefulness throughout the day, consequently impeding their cognitive acuity and overall functionality. This, in turn, exerts a detrimental impact on daily life, encompassing cognitive functions, mood regulation, and both physical and psychological well-being [2]. Consequently, quality of life gets impaired and EDS becomes a significant public health concern. Moreover, the persistence of EDS into the morning hours upon awakening is a common complaint among patients with OSA, resulting from the fragmented sleep architecture marked by frequent arousals and micro awakenings triggered by respiratory disturbances throughout the night [3]. Importantly, the severity of daytime sleepiness often parallelly correlates with OSA severity. EDS is associated decreasing daytime brain activity and cognitive performance in a previous report [4].
Quantitative electroencephalography (qEEG) is an invaluable modality for assessing neural activity patterns in the brain. Previous investigations have evidenced discernible alterations in qEEG profiles, particularly during different sleep stages in individuals diagnosed with OSA [5,6].
A recent study has examined qEEG power spectral analysis in patients with OSA and found that increased delta and theta power, coupled with decreased alpha and beta power, were associated with OSA severity. Importantly, these changes were more pronounced in patients with OSA who also reported experiencing EDS. The authors suggest that it may reflect normal sleep architecture disruption and contribute to EDS development in patients with OSA [5]. Another evidence evaluated the relationship between OSA severity and qEEG during sleep measures in a population-based cohort and indicated that increased delta and theta power, along with decreased alpha and beta power, were associated with higher apnea–hypopnea index (AHI) values. Moreover, these changes were more pronounced in patients with OSA who reported experiencing EDS [7]. However, they have not specifically focused on qEEG analyses particularly after awakening.
Accordingly, this investigation aims to elucidate the disparities in quantitatively analyzed spectral characteristics of electroencephalography (EEG) upon awakening after overnight sleep in patients with moderate to severe OSA experiencing EDS, compared to that in age- and sex-matched healthy participants.
METHODS
This was a prospective non-invasive observational and crosssectional study. The study design was approved by the Institutional Review Board of Dongtan Sacred Heart Hospital of Hallym University (No. 2021-10-009). All participants provided written informed consent.
Participants
Participants over the age of 19 years with clinical suspicion of OSA who visited the sleep center of a single university hospital (Hallym University Dongtan Sacred Heart Hospital) from November 2021 to November 2023 were recruited. A total of 17 treatment-naive patients with moderate to severe sleep apnea and self-reported EDS were enrolled. The inclusion criteria for OSA and EDS required an AHI of ≥15 and an Epworth Sleepiness Scale (ESS) of ≥11 [8], respectively. Additionally, patients previously diagnosed with any neurological or brain parenchymal disease (such as stroke, epilepsy, dementia, or head trauma) or those receiving medication involving hypnotics or sedatives were excluded from the qEEG analysis.
To compare the EEG of patients with OSA and healthy participants, we selected data of age- and sex-matched healthy adults (n=17) from the big data collected named the normative database (NormDB) during developing the EEG spectral analysis software iSyncBrain® (iMediSync, Inc., Seoul, Republic of Korea) [9]. The NormDB includes standardized data for 4.5–81 year-old 1,289 individuals (553 men and 736 women) with strict criteria of healthy conditions, which was collected between 2014 and 2019 at the Korean EEG Center at Seoul National University. We used part of NormDB to compare with age- and sex-matched controls from a community-based population recruited by a structured telephonic interview with a questionnaire designed to exclude participants who have any history of medical (such as cardiac, pulmonary, and renal disease, or metabolic syndrome), neurological (such as stroke, epilepsy, or head trauma), or psychological illness (such as behavioral or conduct disorders), including sleep disorders. This approach aligns with standard practices in similar studies and ensures the reliability of our control group selection [9]. However, there is a lack of objective tests to exclude patients with specific sleep disorder such as OSA, hypersomnia, or other sleep disorders.
Assessment of the sleep factors and polysomnography
All patients completed self-reported questionnaires covering sociodemographic characteristics, height, weight, sleep habits (sleep duration, time-in-bed at night during weekdays and weekends), sleep-related profiles, including sleep quality, insomnia, habitual snoring, and comorbid medical conditions.
The participants were subjected to overnight polysomnography in a controlled laboratory setting. Subsequently, EEG sessions were conducted within 30 min of awakening with participants in an eyes-closed condition to ensure consistency and minimize variability due to different states of alertness and external stimuli. EEG data were captured using conventional electrodes arranged as per the International 10–20 system at positions. These signals were sampled at 250 Hz and digitally filtered to limit the frequency range between 0.3 and 70 Hz. Continuous, artifact-free segments of 5-min EEG were identified and extracted by two neurologists, GWN and HJI, for further analytical assessment. Sleep analysis of overnight polysomnography were evaluated based on the polysomnography data using the American Academy of Sleep Medicine’s standard criteria [10]. This included calculating the AHI, which accounts for episodes of significant airflow reduction, over 90% for apneas and over 30% for hypopneas lasting more than 10 s, coupled with at least 3% drop in oxygen saturation. These metrics were scored by accredited technicians specializing in sleep study assessments.
QEEG analysis
QEEG processing and group analyses, including de-noising with advanced mixture-independent component analysis (am-ICA), extracting features at both sensor and source levels, and generating topographic (topomap) images were conducted using iSyncBrain®, a cloud-based, artificial intelligence EEG analysis platform (https://isyncbrain.com/) [11]. EEG pre-processing was performed to denoise all data and minimize the effects of artifacts. During the first stage of EEG pre-processing, the signals were sampled at 250 Hz and filtered with a 1.0–45.5 Hz band-pass filter. The EEG were then passed through a notch filter in preparation for downstream processing. Finally, artifacts identified via electromyogram, cardiac signal, body moving, and electrooculogram were removed to yield cleaned qEEG normative data. To enhance the performance of iSyncBrain®, we added an extra denoising process, which included using 1.0–45.5 Hz band-pass filter, referencing using common average reference. The detailed processing and artifact removal methods have been previously outlined [12].
ast Fourier transform was used to calculate the spectral band power with a Hanning window (4-s sliding window with 50% overlap). The spectral frequency bands were categorized into eight categories: δ (1–4 Hz), θ (4–8 Hz), α1 (8–10 Hz), α2 (10–12 Hz), β1 (12–15 Hz), β2 (15–20 Hz), β3 (20–30 Hz), and γ (30–45 Hz) [13,14]. The absolute and relative power of each frequency band were calculated. In addition to visual inspection, artifacts caused by eye movements, muscle contractions, and cardiac signals were isolated and removed using amICA [15]. Source reconstructions were performed with the standardized low resolution brain electromagnetic tomography (sLORETA) using the Colin Head model and network nodes and edges were defined as 68 regions of interest based on the Desikan–Killiany atlas [16], and the inter-nodal iCoherence was evaluated accordingly. A network is a mathematical representation of a real-world complex system and defined by a collection of nodes (vertices) and links (edges) between pairs of nodes. Each region of interest (ROI) from sLORETA analysis is considered a node [17]. The 68 ROIs can be classified into five regions: frontal, central, temporal, parietal, and occipital. The frontal lobe includes the frontal pole, superior frontal, rostral middle frontal, caudal middle frontal, pars opercularis, pars orbitalis, pars triangularis, medial orbitofrontal, lateral orbitofrontal, precentral, rostral anterior cingulate, and caudal anterior cingulate. The topographic map images were generated with iSyncBrain, while the 3D and brain network images were created using MATLAB (MATLAB R2017b, Mathworks, Natick, MA, USA). The color scale on the images represents the comparative power differences between the studied group and a reference group. For example, the topomap illustrates the differences between the patients and healthy participants (Fig. 1A). Here, colors shifting toward the red spectrum suggest that the patients (the group of interest) exhibited relatively lower power than the healthy participants. Bluish tones indicate that the OSA group, which is the focus of this comparison, showed relatively higher power than the control group. The presence of red in the p value color scale indicates statistically significant differences (p<0.05), while green signifies non-significant findings (p>0.05). The methodologies for these analyses have been detailed by Baik et al. [18].
Statistical analysis
We performed two group comparisons for power spectra and network analysis: patients with OSA and age- and sexmatched healthy participants. For each spectral frequency band, the absolute and relative power within that band was subjected to Mann–Whitney tests across channels, comparing the patients with OSA to healthy participants. Adjustments for multiple comparisons were made using the Bonferroni correction for each spectral frequency band across five cerebral regions (frontal, central, temporal, parietal, and occipital regions) by 5 (40 tests). The reported p value for the difference in delta band power between the patient and control groups was corrected for these multiple comparisons. Statistical analysis was automated using the iSyncBrain® program and MATLAB.
RESULTS
Participants
Seventeen patients diagnosed with moderate to severe OSA and experienced EDS were included the analysis. The participants were mostly middle-aged males (16 males, 57.2±8.0 years). Their average BMI was 28.4±4.3kg/m2, which was classified as overweight and they suffered from severe EDS (average ESS 14.4±3.2).
Polysomnographic variables
Based upon polysomnography results, the average AHI was 38.1±20.5 per hour and more than half (10 patients, 58.8%) patients had severe OSA (AHI more than 30). Table 1 shows the overall polysomnographic findings of the participants. Sleep efficiency was 86.8%±9.0%. A nadir of arterial oxygen saturation (maximum desaturation) was 80.6%±6.4%. Sleep latency was 10.1±9.5 min and rapid eye movement (REM) latency was 114.3±51.6min. The percent of total sleep time spent in sleep stages N1, N2, N3 (slow-wave sleep), and REM were 36.5%±20.6%, 45.9%±17.3%, 6.8%±8.3%, and 10.8%±5.4%, respectively. The arousal index and periodic leg movements in sleep index were 40.0±21.0 and 11.4±22.0/h, respectively.
Difference in quantitative EEG after awakening between OSA group vs. normal control group
We performed an absolute and relative sensor-level analysis between the patients with OSA and age- and sex-matched healthy adults. In the qEEG topographic analysis, the absolute delta band power distribution was distinctly different between the OSA and normal control groups. Fig. 1A demonstrates a topomap of comparison of absolute spectral delta power band. Specifically, absolute delta band activity in the frontal lobe regions in individuals with moderate to severe OSA increased when compared to that in the healthy adults (p<0.05). A comparative spectral plot in each brain region delineated in Fig. 1B illustrates the difference in delta band power between the patient and control groups. This analysis encompasses the overall higher average delta band power and its standard deviation in the patients with OSA than those in the healthy adults in each brain region but was statistically significantly higher only in frontal region (3.78±5.53 vs. 3.22±0.98 μV2, p=0.03). Fig. 1C shows the comparative plot between groups according to 19 electrode channels, but the sum of the average absolute delta spectral power in the frontal electrodes (Fp1, Fp2, F7, F3, F4, F8) showed a significant difference compared to that in the control group, rather than each electrode channel.
In Fig. 2A, the patient group (G1) have higher absolute delta wave values in the rostral middle frontal, caudal middle frontal, and pars opercularis regions within the frontal lobe on both the left and right sides compared to the normal control group (G2) among 68 ROIs. Additionally, the patients with severe OSA accompanied by EDS show decreased connections between the frontal lobe and other lobes compared to the healthy participants.
In Fig. 2B, the statistical significance of the differences between the OSA group (G1) and the normal control group (G2) is shown. It particularly highlights that the connectivity between the rostral middle frontal and pars orbitalis ROIs within the right frontal lobe is statistically significantly higher in OSA group (G1) compared to that in the normal control group (G2).
DISCUSSION
EDS is the one of leading complaints among patients with OSA, particularly adults. This has long been recognized and the most extensively studied symptom [19,20]. Intermittent hypoxia and the resultant sleep fragmentation in OSA disrupts the sleep architecture, specifically diminishing the proportion of restorative slow-wave and REM sleep. Such disturbances contribute to neurocognitive function impairment, such as deficits in attention, memory, and executive function, thereby underpinning that the observed daytime somnolence EDS in patients with OSA is of paramount clinical importance due to its profound impact on quality of life, cognitive functioning, and daily performance [20]. Sleepiness impairs the ability to focus and sustain attention, leading to difficulties in processing information and poor performance on tasks that require continuous monitoring. It reduces social productivity, increases workplace accidents as well as elevates the risk of motor vehicle accidents beyond psychological physical health effect [21]. Although not all patients with OSA suffer from EDS, it frequently manifests executive functioning deficits in even children [22]. In aspects of cognitive and behavioral effects of OSA, dysfunction of prefrontal regions of the brain cortex has been associated with executive dysfunction including behavioral inhibition, self-regulation of arousal, working memory, and contextual memory [23].
In this study, we report that patients with moderate to severe OSA experiencing EDS exhibit changes in sleep architecture that align with the clinical features of OSA [24]. Patients with moderate to severe OSA experiencing EDS showed short sleep latency, increased WASO, and altered sleep structure compared to normal sleep parameters, which is similar to the findings of our study [25]. Specifically, the percent of total sleep time spent in stage N1 and N2 sleep increased, alongside a reduction in slow-wave (stage N3) and REM sleep.
The second main finding of this study was the significant difference in qEEG, with increased absolute delta activity in the frontal area on awakening after night sleep compared with healthy participants. Our findings showed that sleep architecture disruption due to oxygen discontinuation, a characteristic of OSA, can cause EDS and contribute to changes in brain activity, particularly in the delta frequency band.
Delta waves, which are slow oscillations with 0.5–4.0 Hz frequencies, are predominantly associated with slow-wave sleep during non-REM stage N3 [26]. Delta activities during sleep are considered synaptic plasticity effectors [27]. Since they are typically associated with stage N3 sleep, the presence of delta activity during alert states, such as upon awakening, is unusual under normal circumstances. Although we could not perform additional analyses correlating delta power with other clinical or polysomnography parameters due to lack of information for the control group, previous researches support the clinical relevance of our findings. Studies indicate that delta power increases after prolonged wakefulness and decreases as sleep deepens [28]. In patients with OSA, increased delta power during REM sleep is linked to disrupted sleep architecture and daytime dysfunction, particularly in those experiencing EDS [29]. These suggest the association with delta power and sleep homeostatic drive consequent to non-restorative sleep consistent with the results of a previous report [30].
The implication of delta activity might arise from frequent arousals during sleep due to apnea events, which can prevent the sustained presence of delta waves, leading to insufficient stage N3 sleep. This results in non-restorative sleep, as the benefits of delta activity (physical and cognitive restoration) are not adequately achieved. The presence of delta activity during awakening can also signify several physiological and pathological states and has potential implications for cognitive impairment and neural pathology. A previous study has explored the relationship between delta power during sleep and neuropsychological performance in patients with OSA [5]. However, some evidences have demonstrated the beneficial role of delta power during task. Delta oscillations during mental task inhibited interferences that could improve task performance, potentially by modulating the activity of networks that should remain inactive to complete the task; low frequency delta activity may reflect waking performance in prefrontal cortex-specific neuropsychological tests [31,32].
Though these inconsistent interpretation of delta activity on cognitive performance, the alterations in delta activity observed in OSA have significant implications for neurocognitive functioning [27]. The changes of EEG pattern in the brain frontal lobe in our study, which are essential for decision-making and executive functions, warrants further investigation to elucidate the clinical significance of this altered delta band activity [33].
Strength and limitation
This study uses spectral analysis techniques to visualize and compare the brain impacts of EDS in the patients with moderate to severe OSA versus that in healthy participants. To the best of our knowledge, this is the first study to analyze the qEEG data of patients with OSA experiencing EDS, specifically measured upon awakening after night-time sleep. However, this study has some limitations.
First, one of the primary limitations is the small sample size, which can significantly affect the power of the findings. In addition, analysis for five brain regions categorized by each lobe and the statistical application of the Bonferroni correction is conservative and can increase the risk of type II errors. The alternative correction methods, such as the Benjamini–Hochberg procedure, can be considered but resulted in non-significant p values. A future study involving a larger population is needed to achieve adequate statistical power and enhance the generalizability of the findings.
Second, the current study identifies spectral differences in EEG between patients with moderate to severe OSA experiencing EDS and healthy participants but fails to distinguish whether these differences are due to OSA or EDS. This ambiguity complicates the interpretation of how each condition specifically alters brain function. Third, comparing the qEEG of patients with moderate to severe OSA or EDS with those of healthy controls who are not in an immediate post-arousal state presents another significant limitation. This can lead to inaccurate comparisons and conclusions about the specific impacts of OSA experiencing EDS on brain activity. Standardize EEG comparison conditions using EEG data from similar arousal states across all groups (immediate post-arousal if possible) will allow for more accurate and meaningful comparisons. We also acknowledge that consolidating brain regions or integrating spectral frequency bands (such as combining alpha 1 and 2 into alpha, or beta 1, 2, and 3 into beta) could offer statistical advantages. However, the current analysis program does not support merging or simplifying brain regions or frequency bands. Fourth, the age and sex-matched healthy participants from the big database (NormDB) were screened for comorbid disorders using questionnaires owing to the lack of objective test or tools to exclude individuals with specific sleep disorders such as OSA and hypersomnia. This lack of specific exclusion measures could affect the reliability of the control group data. Lastly, to assess the impacts of EDS on the brain, specifically in terms of cognitive function related to specific spectral power, a detailed analysis including cognitive assessments is necessary. Incorporating comprehensive cognitive function tests linked to EEG data will enhance our understanding of how moderate to severe OSA and EDS influence cognitive impairment in daily life, potentially leading to improved treatment approaches. This multifaceted approach by combining qEEG and detailed neuropsychological evaluations would provide insights into the mechanisms underlying the cognitive deficits associated with OSA and EDS, potentially leading to more targeted and effective interventions to manage these debilitating symptoms.
Overall, this is a well-designed and executed pilot study that provides valuable insights into the neurophysiological underpinnings of EDS in patients with OSA. The findings have the potential to inform future research and clinical interventions in this area.
Conclusion
This study demonstrates the spectral impacts of moderate to severe OSA with EDS utilizing qEEG measured immediately upon awakening, as well as alterations in sleep architecture. By demonstrating the distinct differences in the qEEG profiles between participants with severe OSA accompanied by EDS and healthy participants, this research enhances our understanding of the neurophysiological bases of EDS within this clinical framework. Further research with larger sample sizes is necessary to confirm these preliminary observations and help develop targeted interventions aimed at managing EDS in patients with OSA.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation 
Print 
