Microstructure of cortical activity during sleep reflects respiratory events and state of daytime vigilance

The study was conducted in the Department of Otorhinolaryngology, Head and Neck Surgery, of the Friedrich-Alexander University Erlangen-Nürnberg (FAU), Germany, between March 2016 and November 2016, following approval by the local Ethics Committee (# 323_16 Bc). All 31 participating OSA patients (26 male, 5 female; median age 45 years, lower and upper quartiles: 37 and 57 years), were recruited by the Department of Otorhinolaryngology, Head and Neck Surgery. Inclusion criteria for the study were male and female patients aged between 21 and 80 years with untreated OSA according to the third International Classification of Sleep Disorders (ICSD-3; American Academy of Sleep [13]). Exclusion criteria were a positive history of misuse of sedatives, alcohol, or addictive drugs, or other untreated sleep disorders. All patients were asked to fill out the Epworth Sleepiness Scale (ESS) before the first night of cardiorespiratory diagnostics. In the morning after the cardiorespiratory PSG, objective assessment of vigilance was carried out. In the subsequent night, CPAP titration was performed. Data analysis was carried out during time in bed (TIB) of the cardiorespiratory diagnostic night, accumulating to a total recording time of 276 h and 12 min. In 18 out of 31 patients, CPAP titration was performed in the subsequent night, summing up to a total sleep time of 153 h and 48 min. Patients used the CPAP mask during the complete time in bed.

PSG was carried out using the 33-channel cardiorespiratory SOMNOscreen diagnostic system (SOMNOmedics, Randersacker, Germany). Technical implementation of PSG followed the American Academy of Sleep Medicine (AASM) recommendations using standardized procedures including an EEG (F4-M1, C4-M1, O2-M1), right and left electrooculogram (EOG), electromyogram (EMG) of the mentalis and tibialis muscles, a nasal respiratory flow sensor (nasal pressure cannula), thoracic and abdominal respiratory effort sensors (induction plethysmography), body position sensors, pulse oximetry, snoring microphone, a one-channel electrocardiogram (ECG), and infrared video recording [7]. For EEG recording, impedances were kept below 5 kΩ and data were sampled at 256 Hz with high- and low-pass filters at 0.2 and 35 Hz, respectively. Minimal digital resolution was 16 bits per sample. ECG signal elimination was performed for M1. The sleep stages (wake; N1; N2; N3; rapid eye movement, REM) were analyzed and scored visually in 30 s epochs, and associated events were scored according to the AASM criteria (version 2.1, 2014). Scoring was conducted by a highly trained sleep specialist accredited by the German Sleep Society (DGSM) [1, 7]. Thereby, typical artifacts were marked and removed subsequently for further analysis and processing steps [16].

In the night following cardiorespiratory PSG, CPAP titration was performed according to the clinical guidelines of the Positive Airway Pressure Titration Task Force of the American Academy of Sleep Medicine [10].

Computerized evaluation of sleepiness was performed using the Pupillographic Sleepiness Test (PSTxsIII, AMTech Pupilknowlogy GmbH, Dossenheim, Germany). Spontaneous and involuntary pupil movements in darkness are hereby used to measure and evaluate central nervous activation objectively over a period of 11 min. Prior to measurement, subjects had to wear infrared goggles for about 2 min to adjust to the dark. The patients were seated on a chair and instructed to focus on a light-emitting diode in front of them. The pupillary unrest index (PUI) was calculated using an electronic pupillometer with video goggles and infrared illumination with the aid of customized software. The sampling rate was 25 Hz. Details of this methodology have been validated and published elsewhere [11, 18, 19].

Neurophysiological recordings, e. g., EEG, provide measures of neuronal activity evolving over time. Therefore, such data can be seen as a time series of amplitude values that can be broken down to one single value in a given time interval by RMS analysis. These single values recorded across the different recording channels during the same time bin reflect a spatial pattern of neuronal activity present at that time. The complete method has been described previously elsewhere [8, 9] but will be summarized here: The normalized RMS values of each single EEG channel were calculated for each 30 s epoch (synchronized to the sleep stage analysis). The RMS amplitudes of the three recording channels correspond to a three-dimensional vector for each 30 s interval, whereas successive vectors form a trajectory in three-dimensional space. Each point was labeled with the corresponding sleep stage (cluster labels). A newly developed method was then applied to statistically compare these clusters of data points in n-dimensional space. It is based on calculating the so-called discrimination value, taking into account all intra- and inter-cluster distances between each pair of data points in n-dimensional space. These distances are a measure of how similar or dissimilar two points (i. e., EEG activation patterns) in n-dimensional space are. The more similar they are, the smaller their distance [8, 9]. The intra-cluster distance within each sleep stage corresponds to the variance of neuronal patterns within this stage. The inter-cluster distance between the sleep stages corresponds to the dissimilarity of the neuronal activity when comparing these conditions.

The processed data were analyzed with Statistica 8 (StatSoft, Inc., Tulsa, OK, USA). For the datasets of the 31 patients obtained before CPAP therapy, multiple linear regression analyses were performed with the ESS, PUI, and AHI data and the different neuronal pattern cluster radii and distances independently. The results of these regression analyses are summarized in Table 1 and 2. Note that in the ESS analyses no significant correlations were found and the results are not shown in detail. Total sleep time, AHI, and number of epochs of each sleep stage before and after CPAP therapy were analyzed with paired Wilcoxon tests in the 18 patients for whom measurements were obtained during CPAP titration. For comparison of the neuronal patterns pre- and post-CPAP therapy, we calculated the individual change of intra- and inter-cluster distances during each sleep stage (CPAP night value minus untreated night value) and the percentage of patients showing either an increase or decrease in these parameters.

The number of nocturnal respiratory events per hour (median AHI 19.0; lower and upper quartiles 10.4 and 36.5) was obtained from 31 patients with OSA (26 male; median age 45years; median body mass index, BMI: 27.3 kg/m²) and correlated with the cortical EEG patterns recorded during sleep. Five significant positive linear correlations (Table 1) of sleep stage EEG patterns with the AHI were found (N2_intra, N3_intra, wake–N2, wake–N3, N2–N3) as well as three tendencies for positive linear correlations (wake_intra, wake-REM, wake–N1). These correlations indicate that an increase in AHI (i. e., the severity of the OSA) and pathologically changed cortical EEG patterns occur together. These EEG pattern changes are most prominent in intra- and inter-cluster distances (i. e., the heterogeneity within a pattern and the dissimilarity between patterns) of sleep stages N2 and N3 and the wake stage.

Subjective daytime sleepiness was obtained by the ESS questionnaire in all 31 patients and correlated with the cortical EEG patterns recorded during sleep. The results of the multiple linear regression analyses showed no correlation of subjective sleepiness with any of the recorded neuronal patterns.

Objective daytime sleepiness was measured in the form of PUI. The results of the multiple linear regression analyses of PUI with the cortical EEG patterns are given in Table 2. In two cases we found the objective sleepiness to be significantly negatively correlated with the intra-cluster distance of the wake state (wake_intra) and the inter-cluster distance between the patterns of wake and sleep stage N3 (wake–N3), respectively. Additionally, in two other cases we found tendencies for negative linear correlations of the PUI with the inter-cluster distance between wake and N1 (wake–N1) or wake and N2 (wake–N2) patterns (Fig. 1a, b; Table 2). In other words, a decrease in patients’ vigilance (increased PUI) leads to homogenization of wake stage simultaneously with a shift of the wake stage towards the sleep stages N1, N2, N3, and REM (Fig. 1c).

In 18 out of 31 patients a CPAP titration was performed during the subsequent night. Median total sleep time was not changed significantly with CPAP therapy compared to pre therapy (TST_pre = 8.9h, interquartile range [IQR] 8.3–9.7h; TSTCPAP = 8.5h, IQR 7.9–9.4h; paired Wilcoxon test, p = 0.08). CPAP therapy significantly reduced AHI (Fig. 2a; paired Wilcoxon test, p < 0.001; AHI_CPAP = 8.3 [4.9, 17.2]) and significantly changed the number of wake and sleep stages (Fig. 2b).

From the results of the AHI–EEG pattern correlations we expected the pathologically increased intra- and inter-cluster distances to be reduced back to normal when the AHI decreases (Fig. 3). The real effect of CPAP therapy on cortical EEG patterns during the different sleep stages is depicted in Fig. 4. The paired change of the median value (CPAP night minus untreated night) of inter- and intra-cluster distances between all combinations of sleep and wake stages is given in Fig. 4a. Note that only decreases of inter- and intra-cluster distances were found (blue colors), and that this was true for most sleep stage combinations tested. In other words, with CPAP therapy, cortical EEG patterns during sleep show less variation and therefore a higher homogeneity. This was particularly true for the intra-cluster distances of all sleep stages which showed reduced heterogeneity of EEG patterns during CPAP therapy. With respect to inter-cluster distances, EEG patterns associated with the different sleep stages became less different during CPAP therapy as well.

Fig. 4b gives the percentage of patients showing a decrease in the cortical EEG patterns. Most patients (up to 80%) showed a decrease in inter- and intra-cluster distance in N1_intra, REM–N1, REM–N2, and N1–N2. Taken together, these data show, that CPAP therapy not only has a positive effect on breathing but also affects brain activity during sleep. The cortical EEG patterns during the different sleep stages become more homogeneous already during the first night of therapy, thereby presumably resembling the EEG patterns of healthy subjects.