Altered K-complex morphology during sustained inspiratory airflow limitation is associated with next-day lapses in vigilance in obstructive sleep apnea

Abstract

Study Objectives

Determine if changes in K-complexes associated with sustained inspiratory airflow limitation (SIFL) during N2 sleep are associated with next-day vigilance and objective sleepiness.

Methods

Data from thirty subjects with moderate-to-severe obstructive sleep apnea who completed three in-lab polysomnograms: diagnostic, on therapeutic continuous positive airway pressure (CPAP), and on suboptimal CPAP (4 cmH₂O below optimal titrated CPAP level) were analyzed. Four 20-min psychomotor vigilance tests (PVT) were performed after each PSG, every 2 h. Changes in the proportion of spontaneous K-complexes and spectral characteristics surrounding K-complexes were evaluated for K-complexes associated with both delta (∆SWAK), alpha (∆αK) frequencies.

Results

Suboptimal CPAP induced SIFL (14.7 (20.9) vs 2.9 (9.2); %total sleep time, p < 0.001) with a small increase in apnea–hypopnea index (AHI3A: 6.5 (7.7) vs 1.9 (2.3); p < 0.01) versus optimal CPAP. K-complex density (num./min of stage N2) was higher on suboptimal CPAP (0.97 ± 0.7 vs 0.65±0.5, #/min, mean ± SD, p < 0.01) above and beyond the effect of age, sex, AHI3A, and duration of SIFL. A decrease in ∆SWAK with suboptimal CPAP was associated with increased PVT lapses and explained 17% of additional variance in PVT lapses. Within-night during suboptimal CPAP K-complexes appeared to alternate between promoting sleep and as arousal surrogates. Electroencephalographic changes were not associated with objective sleepiness.

Conclusions

Sustained inspiratory airflow limitation is associated with altered K-complex morphology including the increased occurrence of K-complexes with bursts of alpha as arousal surrogates. These findings suggest that sustained inspiratory flow limitation may be associated with nonvisible sleep fragmentation and contribute to increased lapses in vigilance.

Statement of Significance.

Sustained inspiratory airflow limitation is thought to cause nonvisible sleep fragmentation. However, it is not clear why, despite this sleep fragmentation, some subjects show no daytime symptoms. In this study, we demonstrate that the morphology of K-complexes, hallmark of nonrapid eye movement sleep, appears to shift from promoting sleep (i.e. bursts of delta) to an arousal surrogate (i.e. bursts of alpha) in presence of sustained inspiratory airflow limitation. We find that this altered morphology is associated with next-day lapses in vigilance suggesting that examining morphological features of K-complexes can help explain the large inter-individual variability seen in the relationship between sustained inspiratory airflow limitation and daytime symptoms.

Introduction

Sustained inspiratory airflow limitation is a commonly observed phenotype of obstructive sleep apnea (OSA). Airflow limitation is characterized primarily by an elevated upper airway resistance and is visible as flattening on the inspiratory flow/time tracing for multiple breaths, i.e. for a prolonged period of time [1, 2]. The flattening of the inspiratory flow/time tracing can be observed during a routine nocturnal polysomnogram (PSG) using a nasal cannula/pressure transducer system [3]. On the spectrum of severity of respiratory disturbances in sleep, sustained inspiratory flow limitation (SIFL) falls on the lower end of severity below hypopneas and apneas [4]. Sustained inspiratory airflow limitation is associated with long-term adverse health consequences. In particular, individuals whose primary physiology consists of sustained inspiratory airflow limitation, sometimes labelled as having the upper airway resistance syndrome (UARS), have an increased risk of developing high blood pressure, irritable bowel syndrome, and daytime sleepiness [5–13].

Apneas and hypopneas almost always result in acute detectable physiological changes such as electroencephalographic (EEG) arousal and oxygen desaturation. These usually alter macroscopic sleep architecture and produce sleep fragmentation [14]. In contrast, episodes of sustained inspiratory airflow limitation elicit subtle changes in EEG that may not be visually discernible on a routine nocturnal PSG [15, 16]. In particular, the two hallmark consequences of upper airway obstruction, arousal, and oxygen desaturations, are often absent during sustained inspiratory airflow limitation events. It has been suggested that EEG changes that are not discernible visually occur, and that these changes may underlie unrefreshing sleep, particularly in subjects whose primary form of respiratory disturbance is SIFL [17].

Sustained inspiratory airflow limitation is associated with perturbations in sleep EEG. An increase in alpha frequency in nonrapid eye movement (NREM) sleep has been observed in UARS subjects as compared to matched healthy controls [18]. Further, delta frequency was also observed to be higher in subjects with UARS who also complained of chronic fatigue than healthy controls [17]. Alpha-delta sleep, the co-occurrence of alpha frequency bands and delta frequency bands in NREM sleep has been reported to be a common phenomenon in subjects with predominantly SIFL [19]. Chervin et al. observed that esophageal pressure swings were highly correlated with changes in EEG spectral characteristics [16, 20]. Notably, several of these studies observed that periods of EEG surrounding SIFL often did not show evidence of an American Academy of Sleep Medicine (AASM)-defined cortical arousal.

In addition to broad-spectrum changes in EEG accompanying sustained inspiratory airflow limitation, measures of sleep EEG microstructure such as cyclic alternating pattern (CAP) and stage N2 K-complexes have also been investigated. The presence of these CAP events indicates sleep fragmentation and instability. In subjects with predominantly inspiratory flow limitation, the rate at which CAP events occur has been reported to be significantly higher than in healthy subjects [21]. More recently, spontaneous K-complexes during sleep stage N2 were systematically examined during airflow limitation [22]. Using transient reductions in therapeutic CPAP level during stable N2 sleep to experimentally induce short runs of mild airflow limitation in OSA subjects, it was noted that K-complexes were elicited at a higher rate during this mild airflow limitation. Whether the increase in K-complexes was associated with next-day outcomes was not determined at the time.

K-complexes have been thought to have dual roles [23]: either as surrogates of cortical arousal [24] or as a sleep protective/promotion phenomenon [25–27]. These roles are not necessarily mutually exclusive. The dual role of K-complexes is primarily assessed through a spectral lens, specifically using alpha and delta frequency bands. It is thought that alpha frequency related K-complexes are associated with the arousal phenomenon [28], whereas delta frequency related K-complexes are associated with a sleep protection/promotion role [22, 29, 30]. Notably, K-complexes are part of several EEG features that have been studied in UARS patients: the A1 phase of CAP includes K-complexes; and the EEG delta band includes power due to K-complexes. Whether changes in morphological features of K-complexes due to SIFL are related to daytime outcomes, however, remains unclear.

Accordingly, the primary objective of this study was to determine potential changes in K-complex morphology during prolonged experimentally induced inspiratory flow limitation with suboptimal CPAP and whether these changes are associated with next-day consequences of vigilance and objective sleepiness. We hypothesized that changes in K-complex morphology associated with suboptimal CPAP would be related to poor vigilance and sleepiness. Given the dual role of K-complexes, we also investigated changes in EEG spectral characteristics surrounding K-complexes, namely delta and alpha bands, with suboptimal CPAP.

Methods

Data presented in this paper arise from a secondary analysis of a parent study examining the relationship between OSA and daytime functioning. The parent study has been described previously [27, 31]. All subjects signed informed consent documents and the protocol for the parent study was approved by the NYU IRB. The analysis of data obtained from the parent study was also approved by the Mount Sinai IRB. The timeline of the visits is shown in Figure 1. For the present study, we included all subjects who were diagnosed with moderate-severe OSA and had data available on both the optimal CPAP and the suboptimal (subop) CPAP visits. After three months of CPAP use at optimal titrated pressure, subjects were studied with an in-lab PSG at this pressure (optimal CPAP visit). Approximately 30 days after the optimal CPAP visit, subjects had their CPAP levels lowered by 4 cmH₂O and after 2 weeks at this lower pressure, subjects were studied with an in-lab PSG at the lowered pressure (suboptimal CPAP visit). At home, subjects used a custom CPAP machine (Fisher & Paykel Healthcare, Auckland, New Zealand), which provided continuous monitoring of their CPAP pressures and thus allowed objective quantification of the CPAP levels that were used. The 2-week subop CPAP period was intended to ensure the presence of sustained flow limitation for a sustained period prior to the study, mimicking UARS or mild OSA.

Figure 1.

Partial timeline of parent study examining the relationship between OSA and daytime function. Data used in this study consists of polysomnographic data available on diagnostic, optimal CPAP, and suboptimal CPAP (subop CPAP) visits. The average time between optimal CPAP and subop CPAP visits was approximately 30 days.

Polysomnography

All studies consisted of full in-lab nocturnal PSG (Sandman sleep system, Embla systems Inc., Broomfield, CO) performed according to AASM guidelines. Recordings of frontal, central and occipital EEG, electrooculogram, and submental electromyogram (EMG) were used to monitor sleep metrics. Leg movements were monitored with an anterior tibialis EMG. A unipolar electrocardiogram was used for cardiac monitoring. Oxygen saturation was monitored with a pulse oximeter (Masimo). Chest wall and abdominal movement were monitored with piezoelectric strain gauges. Sleep position was monitored with a multiposition switch. Respiratory airflow was recorded directly from the CPAP machine.

Standard sleep and respiratory scoring were performed according to AASM criteria (AASM Scoring Manual, version 2.5, 2018). From these measures, absolute and relative amounts of each sleep stage, total sleep time (TST), sleep efficiency, and wake after sleep onset (WASO) were extracted. Arousals were scored based on AASM criteria with an arousal index calculated as a number of arousals per hour of sleep. Apneas were defined relative to baseline as a >90% decrease in airflow for >10 s. Inspiratory flow limitation was identified manually as per the recent ATS guidelines [32]. Sustained flow limitation (SIFL) events were marked visually based on a run of inspiratory flow limitation for more than 2 min, with shorter periods labeled as hypopneas if criteria were met. Candidate hypopneas were initially identified from the airflow signal as a visible reduction (typically > 30% decrease and accompanied by inspiratory flow limitation) for >10 s, and then labeled if they were linked to O₂ desaturation and/or EEG arousal. The apnea-hypopnea index (AHI3A) was then calculated as the sum of all apneas and hypopneas with either 3% desaturation and/or EEG arousal.

Assessment of objective daytime function

Daytime function was assessed objectively using the psychomotor vigilance test (PVT) and the multiple sleep latency test (MSLT). Four 20-min PVT’s (Ambulatory Monitoring Inc., Ardsley, NY) were performed 2 h apart, beginning in the morning after each PSG. The primary variable from each 20-min PVT trial was the number of lapses (reaction time > 500 ms). We transformed the mean number of lapses from the four tests (PVT lapses = sqrt(lapses) + sqrt(lapses + 1) and used it as a measure of vigilance (PVT lapses). Objective daytime sleepiness was assessed using the MSLT. Subjects were given the opportunity to fall asleep during a 20-min period four times a day at 2-h intervals and the tests were administered 30 min after the PVT. Sleep latency on each 20-min test was calculated as the elapsed time from lights out to the first epoch scored as sleep and latency were averaged over the four tests.

Quantitative EEG analyses

All quantitative EEG analyses were restricted to the EEG derived from the frontal region (Fz/A2), as K-complexes, delta waves are maximally expressed in this region as compared to central/parietal brain regions. EEG was sampled at 128 Hz. Spontaneous K-complexes were detected automatically over the entire PSG using a previously validated automated method DETOKS [33] written in MATLAB (version R2020a, Mathworks, Natick, MA). K-complex detection was carried out blinded to the CPAP condition (optimal vs subop). K-complexes associated with EEG arousals were discarded. For the spectral analyses on NREM sleep periods, we used a modified Welch periodogram with overlapping (50%) 10-s segments. The segments were smoothed using the Hamming window. In addition, spectral analyses were also conducted on 1-s EEG segments prior to and following a DETOKS-detected stage N2 K-complex. For these 1-s prior and post-K-complex segments the same Welch periodogram was used, however, the window size was fixed at 1-s (128 samples with a 50% overlap between the windowed segments). The spectrum of the EEG segments was decomposed into delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), sigma (12–15 Hz), and beta (15–30 Hz) bands. For each frequency band, we calculated the average percent change pre- to post-K-complex across the EEG segments. The average percent change in delta pre- to post-K-complex is denoted by ∆SWAK whereas the average percent change in alpha pre- to post-K-complex is denoted by ∆αK. In addition to these EEG metrics, we also quantified (1) K-complex density (number of K-complexes/min of stage N2 sleep), (2) K-complex amplitude (peak-to-peak), and (3) inter K-complex interval, i.e. time between consecutive K-complexes during a continuous run of stage N2 sleep.

Statistical analyses

All statistical analyses were carried out using IBM SPSS Statistics Version 24 and MATLAB Version R2020a. The statistical significance level was set at p < 0.05 (two-tailed) for all tests. Normality of data was tested using the Shapiro–Wilk test. Spearman’s rank correlation was used to assess partial correlations and the resulting rho value was reported. Comparisons across conditions for demographics, sleep macrostructure, degree of SIFL, and daytime sleepiness were carried out using two-sided paired t-tests, with corrections for multiple comparisons (Bonferroni). In cases of non-normal data, or missing data, we used Wilcoxon signed-rank tests with listwise deletion for missing data. To assess the dichotomous relationship between ∆SWAK and ∆αK, we used linear regression with covariates like age, sex, and condition type. Hierarchical linear regression models were used to assess the relationship between sleep EEG K-complex morphological features and vigilance as well as objective sleepiness with covariates of age, sex, AHI3A, and duration of SIFL.

To assess the effect of SIFL on K-complex morphology, we fit four linear mixed-effects model for (1) K-complex density, (2) ∆SWAK, (3) ∆αK, and (4) inter-K-complex interval with fixed effects for condition (optimal CPAP no-SIFL and subop CPAP SIFL) and potentially correlated random effect for intercept and condition grouped by subject (“y ~ AHI3A + duration of SIFL + condition + (condition | subject)”). These linear models take into account possible correlations among subjects between-nights. All linear models consisted of the same covariates of AHI3A and duration of SIFL. Note that since this was a within-subjects study, we did not include the covariates of age and gender. All linear mixed-effects models used an unstructured covariance matrix for repeated effects (condition) and a variance-components covariance structure for random effects (intercept and slope) estimated using the restricted maximum likelihood method. The same linear mixed-effects models were used to assess the effect of SIFL within-night on subop CPAP, i.e. comparing periods of SIFL with periods of no SIFL stage N2 on subop CPAP.

Results

Thirty subjects had nocturnal PSG data on both optimal CPAP and suboptimal CPAP visits, as well as a diagnosis. Table 1 describes the demographics of the subjects. As expected, subop CPAP compared to optimal CPAP resulted in a significant increase in AHI3A (6.7(7.7) vs 1.9(2.3); median(iqr)). However, arousal index was not significantly different on subop CPAP as compared to optimal CPAP (see Table 1). While respiratory disturbances were greater on subop CPAP as compared to optimal CPAP, sleep architecture on subop CPAP was not significantly different from that on optimal CPAP but was less fragmented than on diagnostic PSG (see Table 1). There was a tendency for an increase in lapses in vigilance on the PVT after subop CPAP, however, it was not statistically significant. Lapses in vigilance were not related to subjective or objective measures of sleepiness across either of the two visits (Supplement Table 1).

Table 1.

Subject characteristics

	Diagnostic N = 30	Optimal CPAP N = 30	Subop CPAP N = 30
Demographics
Age (years)	48 ± 11	–	–
Gender	18M/12F	–	–
BMI (kg/m2)	37.5 (10.6)	–	–
AHI3A (#/h)	62.9 ± 31.4	1.9 (2.3)*	6.5 (7.7) *,**
Arousal index (#/h)	45.4 ± 25.1	12.4 ± 5.2*	10.2 (9.1) *
CPAP (cmH2O)	–	12 (4)	8 (4)**
CPAP adherence (h)		5.6 ± 1.4	– †
Sleep architecture
TST (min)	422.4 ± 76.2	444.7 (59.3)	431.2 (64.3)
WASO (%)	15.7 (12.1)	9.1 (6.9) *	9.4 (10.4) *
SE (%)	78.2 ± 11.5	85.8 (11.7) *	85.1 (11.2) *
Stage N1 (%)	32.8 ± 15.9	14.7 (5.6) *	16.9 (8.8) *
Stage N2 (%)	44.2 ± 12.7	49.1 (7.9)	47.5 (9.7)
SWS (%)	0.5 (4.4)	13.9 (10.7) *	15.1 (12.1) *
REM (%)	14.7 ± 6.4	22.1 (6.1) *	21.4 (6.8) *
Daytime sleepiness
PVT lapses	6.5 (6.9)	3.7 (2.4)	4.7 (6.5)
ESS	13.4 ± 5.9	10.7 (4.6)	10.3 (5.8)
Mean MSLT (min)	7.7 ± 4.7	7.6 (3.5)	7.9 (3.9)

Normally distributed variables are reported as mean ± SD, whereas non-normally distributed variables are reported as median (interquartile range). BMI, body mass index; AHI3A, apnea–hypopnea index; OI, obstructive index; TST, total sleep time; WASO, wake after sleep onset; SE, sleep efficiency; SWS, slow-wave sleep; REM, rapid eye movement; PVT, psychomotor vigilance test; ESS, Epworth sleepiness scale; MSLT, multiple sleep latency test.

* Compared to diagnostic (p < 0.05).

^** Compared to optimal CPAP (p < 0.05).

^† Adherence data was unavailable during 2 weeks of suboptimal CPAP therapy. CPAP adherence data was not available in 11/30 subjects.

Subop CPAP successfully induced SIFL: a significantly higher amount of SIFL was observed overnight (see Figure 2, a, p < 0.001) with increased SIFL observed during stage N2 (see Figure 2, b, p < 0.001) on subop CPAP. On subop CPAP 25/30 subjects had SIFL during stage N2, 16 had SIFL during stage N3 (1.2(29.1), % of time in stage N3; median(iqr)), and 23 had SIFL during REM (5.41(18.2), % of time in REM; median (iqr)), whereas six subjects had no SIFL. However, the presence of apneas and hypopneas was observed in these six subjects. On subop CPAP, periods of SIFL during stage N2 did not necessarily end in an AASM-defined cortical arousal (events with arousals = 1 (2), median (iqr)) and periods of SIFL were generally not associated with oxygen desaturations (oxygen desaturation was less than 2% in 92% of events).

Figure 2.

Boxplots showing amount of sustained inspiratory flow limitation (SIFL) represented as (a) % of total sleep time (TST) and (b) % of time in stage N2 across the two conditions: optimal CPAP and suboptimal CPAP (subop CPAP). Dark colored patches indicate one standard deviation, whereas the light-colored patches indicate 95% confidence intervals. Solid lines indicate the mean and dashed lines indicate the median. Note that jitter is added for visual clarity (*** p < 0.001).

K-complex morphology during SIFL on subop CPAP

We assessed changes in sleep EEG within-night, i.e. comparisons within the night on subop CPAP, associated with SIFL. In particular, we compared sleep EEG during periods of stage N2 SIFL with sleep EEG during periods of stage N2 normal breathing (no-SIFL). Periods of sleep EEG during apneas and hypopneas were excluded for this analysis. Within-night, K-complex density was significantly higher during SIFL (Figure 3, A, F(1,18.5) = 4.4, p = 0.05). While ∆SWAK appeared to be lower during SIFL, it was not significant (Figure 3, B, F(1,24.1) = 1.2, p = 0.3). On the other hand, ∆αK was significantly higher during SIFL (Figure 3, C, F(1, 35.83) = 12.1, p < 0.001). Furthermore, K-complexes were closer to each other in time during SIFL as compared to no-SIFL, measured by the inter-K-complex interval (52.8 ± 29.6 s vs 216.3 ± 61.8 s, F(1, 24.3) = 9.7, p < 0.01).

Figure 3.

Changes in K-complex morphological features within-night during periods with (SIFL) and without SIFL (no SIFL). (A) K-complex density (#/min. of stage N2), (B) average ∆SWAK (% change in delta pre- to post-K-complex). (C) Average ∆αK (% change in alpha pre- to post-K-complex). Solid black lines indicate the mean (*p < 0.05; ***p < 0.001).

Change in K-complex morphology with suboptimal CPAP and its relation to daytime function

Across the two nights, we observed a significantly higher K-complex density overnight on subop CPAP as compared to optimal CPAP (Figure 4, A, F(1,29.3) = 5.1, p = 0.03). Although ∆SWAK was lower on subop CPAP as compared to optimal CPAP, however, the difference was not statistically significant (Figure 4, B, F(1,31.1) = 0.4, p = 0.6). Similarly, subop CPAP was not associated with a significant change in either ∆αK (Figure 4, C, F(1,31.3) = 0.06, p = 0.8) or the inter-K-complex interval (105.5 (176.3) s vs 157.7 (135.3) s, F(1,32.1) = 0.63, p = 0.4).

Figure 4.

Changes in K-complex morphological features across two CPAP conditions. (A) K-complex density (#/min. of stage N2), (B) average ∆SWAK (% change in delta pre- to post-K-complex), and (C) average ∆αK (% change in alpha pre- to post-K-complex) across nights (optimal CPAP and subop CPAP). Solid black lines indicate the mean (*p < 0.05).

Investigating the relationship between the changes in K-complex morphology associated with subop CPAP and daytime function, we found that the increase in K-complex density with subop CPAP was not associated with the change in PVT lapses (see Table 2). However, the decrease in ∆SWAK with subop CPAP was significantly associated with increase in PVT lapses on subop CPAP (Table 2) above and beyond the effect of age, sex, change in AHI3A (change from optimal CPAP to subop CPAP), and the duration of SIFL on subop CPAP. Change in ∆SWAK explained an additional 17% of the variance in change in PVT lapses. However, the overall model which included the covariates change in AHI3A and duration of SIFL was not significant (F(5,22) = 1.95, p = 0.13). Changes in K-complex morphological features of ∆αK and inter-K-complex interval with subop CPAP were not associated with the change in PVT lapses (Table 2). CPAP adherence during the 3-month optimal treatment period was not associated with changes in K-complex density (rho = 0.33, p = 0.21), ∆SWAK (rho = –0.17, p = 0.5), or ∆αK (rho = 0.06, p = 0.8). There was a trend toward significance in the relationship between CPAP adherence and PVT lapses on subop treatment (rho = 0.41, p = 0.11). K-complex morphological features, including those that were related to PVT lapses, were not associated with objective sleepiness (mean MSLT times; data not shown).

Table 2.

Hierarchical regression analyses testing the association between vigilance and change in K-complex morphology (change from optimal CPAP to subop CPAP, N = 25)

Dependent variable	Models	Predictors	β	95% CI for B	P value#	Adj. R2	ΔR2
Change in PVT lapses	Model 1	Age	–0.20	(–0.15, 0.06)	0.39	–0.01	–
		Sex	–0.31	(–4.13, 0.53)	0.13
		AHI3A	–0.23	(–0.19, 0.05)	0.25
		Duration of SIFL	–0.01	(–0.02, 0.02)	0.9
	Model 2: Model 1 + one EEG metric	Change in K-complex density	–0.2	(–4.1, 1.2)	0.28	0.04	0.04
		Change in ∆SWAK	–0.43	(–0.3, –0.01)	0.03	0.15	0.17*
		Change in ∆αK	0.03	(–0.06, 0.08)	0.87	–0.03	0.01
		Change in inter-K-complex interval	0.14	(–0.04, 0.08)	0.54	0.04	0.02

Note that confidence interval (CI) and significance for intercept are not shown. 95% CIs are shown for the unstandardized coefficient B and not β. ∆R²: change in R² from the model with covariates only. Significant change in R² is highlighted in bold.

^# p value for each predictor.

* p < 0.05.

Relationship between alpha and delta bursts following K-complexes

We observed an inverse relationship between ∆SWAK and ∆αK (Figure 5, A; rho = –0.54, p < 0.001) such that an increase in ∆SWAK was associated with a decrease in ∆αK across 3,196 K-complexes assessed in this study. The covariates age, sex, and AHI3A did not alter this association suggesting that intrinsically each K-complex was almost always associated with either a delta burst or an alpha burst.

Figure 5.

Relationship between ∆SWAK and ∆αK. ∆SWAK(%) and ∆αK(%) are shown for each K-complex analyzed in this study (N = 3,196). Each dot represents a single K-complex and the different colors represent the different subjects.

To further illustrate this inverse relationship between delta bursts and alpha bursts following K-complexes, particularly during SIFL, we show grand mean profiles of K-complexes during SIFL on subop CPAP and the associated spectrograms for two subjects: one subject (Figure 6, A and C) who had a high ∆αK (%) and low ∆SWAK (%) overnight and another subject (Figure 6, B and D) who had a high ∆SWAK (%) and low ∆αK (%) overnight. The grand mean profiles indicate the dichotomous behavior of subjects on subop CPAP such that subjects may either exhibit K-complexes that are almost always associated with alpha or K-complexes that are almost always associated with delta.

Figure 6.

Grand mean K-complex waveforms on subop CPAP for two example subjects. Subject 1 (A and C) demonstrates a burst of alpha activity (8–12 Hz) following spontaneous K-complexes. Subject 2 (B and D) demonstrates a burst of delta activity (1–4 Hz) following spontaneous K-complexes. Subject 1 has a high ∆αK (171 %) and a low ∆SWAK (4.2%), whereas subject 2 has a high ∆SWAK (12 %) and low ∆αK (0.15%)

Discussion

The present study provides further evidence that SIFL alters sleep EEG that is indiscernible visually and that these alterations are associated with impaired next-day vigilance. In addition to an increased K-complex elicitation rate, we observed a significant decrease in delta bursts accompanied by a significant increase in alpha bursts following K-complexes during SIFL. The decrease in delta bursts following K-complexes was associated with next-day lapses in vigilance and explained an additional 17% variance in lapses in vigilance. Our results indicate that SIFL alters spectral features of stage N2 K-complexes.

Our observation of increased K-complex density on suboptimal CPAP is consistent with previous findings of altered sleep microstructure associated with respiratory disturbances. Indeed, Nyugen et al. observed an increase of about 40% in K-complex density when flow limitation was induced using transient reductions in the therapeutic CPAP level [22]. It should be noted that in their study, as in ours, flow limitation was not sufficient to cause AASM-defined cortical arousals. Further evidence, of an increase in K-complex density with an increase in respiratory disturbances comes from studies that examine sleep EEG in untreated OSA [27, 34, 35]. If we consider SIFL to be on one end of the spectrum of obstructive sleep-disordered breathing, with the other end being apneas and hypopneas, it is plausible to assume that the highest K-complex density would be observed in the most severe form of OSA. Indeed, data from the parent study previously published, from which data for this study was derived, demonstrated that the highest K-complex density was observed in untreated OSA (diagnostic as well as after two-night CPAP withdrawal) with the lowest on optimal CPAP [27].

Considering sleep EEG on optimal CPAP, i.e. sleep EEG during normal breathing, as baseline, the baseline role of K-complexes appears to be that of sleep promotion rather than as a surrogate of arousal, as on average K-complexes appeared to be followed by more bursts of delta activity than alpha activity. This baseline behavior was also apparent during periods of normal breathing on subop CPAP night. During SIFL, however, K-complexes were on average associated with more alpha bursts than delta. Further, the K-complexes were closer in time to each other during SIFL. The shift in activity following K-complexes during SIFL suggests that the role of K-complexes alternated between the promotion of sleep during the absence of SIFL, i.e. when there were predominantly bursts of delta following K-complexes, and as an arousal surrogate during SIFL, i.e. when there were predominantly bursts of alpha. We observe that within the same night, K-complexes can alternate between the two roles. As an arousal surrogate, K-complexes during SIFL were followed by more alpha activity and less delta activity, but did not alter macroscopic sleep architecture, i.e. time spent in different sleep stages. However, the low delta activity following K-complexes was associated with increased lapses in vigilance. It should be noted though that there were significant inter-individual differences: not all subjects appeared to exhibit K-complexes with increased alpha bursts during SIFL.

Our results demonstrate that a large interindividual variability exists in the response to SIFL. Subjects appeared to respond differently to the presence of SIFL induced by the lowering of their titrated therapeutic CPAP pressure. In particular, ~70% of the subjects had a reduction in ∆SWAK and an increase in ∆αK during SIFL. On the other hand, the remaining ~30% subjects responded in the opposite direction, i.e. an increase in ∆SWAK and a decrease in ∆αK. While there were some potential differences in terms of the PVT lapses in these two groups, such that the larger group of subjects appeared to be less vigilant on optimal CPAP (4.1 (1.6) vs 3.4 (2.9), median(iqr)) we did not conduct formal statistical analyses as the sample sizes were vastly different and small. However, our results suggest that including K-complex morphological features in a future case-control study could potentially aid in answering the long-standing question of why some subjects with SIFL exhibit no daytime symptoms whereas others do.

As in the case of respiratory disturbances, where SIFL is considered typically to be of the milder form, it could be argued that sleep EEG activity following K-complexes also lies along a spectrum, with microarousals and/or awakenings on a possibly “severe” end and short bursts of alpha activity on the “milder” end. Sleep fragmentation associated with microarousals and/or awakenings following K-complexes is well-established and can be seen with a more severe form of untreated OSA [14]. In our study, we observe that the milder form of respiratory disturbance, i.e. SIFL, is associated with a milder form of arousals. This milder form of arousal is not sufficient to cause visible sleep fragmentation or other visible physiological changes such as microarousals and/or awakenings or hypoxia. However, we observe that it is associated with impaired vigilance the next day. In addition to lapses in vigilance, our study also examined the relationship between EEG changes and objective sleepiness using the MSLT. None of the changes in EEG measures due to SIFL were associated with mean MSLT times, suggesting that SIFL may affect vigilance more acutely than sleep propensity. It should be noted that the subjects mean MSLT times on optimal CPAP visit indicated that they were on average sleepy and their sleepiness did not change with suboptimal CPAP.

The sleep promotion role of K-complexes, when they are associated with delta bursts, has been studied previously. In insomnia, Forget et al. observed that in comparison to good sleepers, insomnia subjects had lower delta power following stage N2 spontaneous K-complexes [26]. When studying transitions from stage N2 to slow-wave sleep, De Gennaro et al. observed that K-complexes were more prevalent as compared to the transition from stage N2 to REM sleep [25]. In comparison, K-complexes that were followed by bursts of alpha were observed to contribute to the presence of nonrestorative sleep, particularly in subjects with restless legs syndrome [28]. K-complexes which are primarily followed by alpha bursts are also observed in subjects with chronic fatigue syndrome/myalgic encephalomyelitis (CFS/ME) [36–38]. In addition, several studies suggest that K-complexes followed by alpha bursts are commonly observed in patients with UARS who complain of unrefreshing sleep, whose primary disease form is that of flow limitation [17, 18, 21]. These studies measured either CAP phases, which consisted of alpha bursts surrounding K-complexes, or the alpha-delta sleep, which is described as the presence of alpha activity over delta waves in NREM. While CAP and alpha-delta sleep metrics are not direct measures of alpha activity following K-complexes, their significant overlap allows us to draw the same conclusion.

Consistent with the dual role of K-complexes, we found that K-complexes on both optimal and suboptimal CPAP were either associated with delta bursts or alpha bursts. For a few K-complexes there appeared to be both alpha and delta bursts, but the magnitude (change in relative spectral power from pre-K-complex to post-K-complex) was always substantially higher (approximately 2–3 times greater) for one of them, delta or alpha, and not both. Support for this finding comes from a recent study by Latreille et al. where intracranial recordings were examined during scalp K-complexes in subjects with drug-resistant epilepsy [23]. It was observed that K-complexes followed by an increased delta power and/or decreased power in beta to gamma bands were associated with the supramarginal, anterior part of the superior temporal, middle frontal and middle cingulate gyri. On the other hand, K-complexes followed by increased power in theta to alpha bands, representing an arousal-related response, were associated with the angular, postcentral, and middle temporal gyri. Their results suggested that the sleep-promoting or arousal promoting response of K-complexes was region-specific.

There are several limitations to our study. The present study was a secondary analysis of a parent study examining OSA and daytime function. As such, all limitations associated with secondary analysis of existing data (power, bias, etc.) carry forward to this study. Subop CPAP, delivered as a reduction in optimal CPAP level (lower by 4 cmH₂O) in our study appeared to successfully induce SIFL without disrupting sleep architecture. While this standardized reduction avoids any investigator bias, it is likely that a subject-specific reduction in CPAP level, i.e. dropping the pressure until SIFL was visible, would have resulted in higher levels of SIFL overnight. Another noteworthy limitation of our study was the absence of a measure of snoring. During the in-lab subop CPAP visit, snoring was not monitored. It cannot be concluded with certainty whether the findings in this study were affected by the presence of snoring. We note, however, that, while an analysis of the data during the diagnostic visit, which monitored snoring, could answer this question, the subop CPAP model represents a cleaner experiment: apneas and hypopneas were relatively infrequent, and the primary respiratory disturbance was SIFL. It could be argued that the EEG changes analyzed in this study present a rather confined view of the effects of SIFL on EEG, as primarily stage N2 sleep EEG microarchitecture was assessed. However, we note that several studies examining prolonged periods of flow limitation as well as subjects with UARS syndrome have suggested that the primary effects of flow limitation appear to be in NREM [39], in particular stage N2 [7, 17, 18, 21]. It should be noted that in other analyses of K-complexes, they are usually grouped as either evoked or spontaneous, where evoked K-complexes represent K-complexes elicited due to a stimulus. This grouping allows a better comparison of K-complexes across studies. Although K-complexes during SIFL can be termed evoked K-complexes, it is more likely that in our study K-complexes during SIFL are a mix of evoked and spontaneous K-complexes and as such comparisons of the results of this study to prior studies examining K-complex morphology should be tended with care. Our data suggested that subjects with high adherence to CPAP during the 3-month optimal treatment period had greater lapses in vigilance on subop CPAP. Due to the small sample size of our data, we did not have sufficient power to investigate whether CPAP adherence mediated the observed EEG changes on subop CPAP. Lastly, we note that there appeared to be two distinct groups in our data: subjects whose sleepiness improved after three-months of CPAP treatment and subjects whose sleepiness did not improve. Theoretically, four distinct groups can be formed based on sleepiness and lapses in vigilance as done by Prasad et al. [40], the small sample size of our dataset does not allow us the statistical power to analyze EEG changes in these groups separately.

In summary, our study further adds to the evidence that SIFL is associated with changes in stage N2 sleep EEG microarchitecture. We find that SIFL is associated with an increase in the number of K-complexes which are closer to each other in time. Using within-night comparisons on suboptimal CPAP as well as across nights comparisons on optimal CPAP, we find that K-complexes elicited during SIFL are on average associated with increased alpha bursts. These changes in spectral characteristics of stage N2 sleep EEG K-complexes during SIFL appear to explain additional variability observed in lapses in vigilance in subjects with OSA. Further studies are warranted to assess whether the examining stage N2 sleep EEG changes associated with sustained flow limitation reduce the variability in vigilance and sleepiness in treatment-naïve subjects with OSA.

Disclosure Statement

Financial Disclosure: Dr Rapoport has received research support for grants and clinical trials from Fisher & Paykel Healthcare and consults for Fisher & Paykel Healthcare. In addition, he holds multiple U.S. and foreign patents covering techniques and analysis algorithms for the diagnosis of OSA and techniques for administering CPAP. Several of these have been licensed to Fisher & Paykel Healthcare. Dr Ayappa has received research support for grants and clinical trials from Fisher & Paykel Healthcare. She also holds multiple U.S. and foreign patents covering techniques and analysis algorithms for the diagnosis of OSA and techniques for administering CPAP. Several of these have been licensed to Fisher & Paykel Healthcare and Advanced Brain Monitoring. The other authors have indicated no financial conflicts of interest.

Non-financial Disclosure: none.