Professional Documents
Culture Documents
hypoxemia in men.
Teare4, Sean A. Martin1, 2, 3, Gary A. Wittert1, 2, 3, R. Doug McEvoy4, Peter G. Catcheside4, Robert J.
Adams1, 3, 4.
t
*
Co-first authors
ip
cr
1
The Health Observatory, Adelaide Medical School, University of Adelaide, The Queen Elizabeth
us
Hospital Campus, Woodville, South Australia, Australia.
2
Freemasons Foundation Centre for Men’s Health, Adelaide Medical School, University of
an
Adelaide, Adelaide, South Australia, Australia.
3
M
South Australian Health and Medical Research Institute, Adelaide, South Australia, Australia.
4
Adelaide Institute for Sleep Health, a Flinders Centre of Research Excellence, College of
ed
Medicine and Public Health, Flinders University, Bedford Park, South Australia, Australia.
5
NeuroSleep - NHMRC Centre of Research Excellence, and Centre for Sleep and Chronobiology
pt
(CIRUS), Woolcock Institute of Medical Research, University of Sydney, New South Wales,
ce
Australia.
6
School of Psychology, Faculty of Science, Brain and Mind Centre and Charles Perkins Centre,
Ac
Corresponding authors: Sarah Appleton, The Health Observatory, The University of Adelaide,
Queen Elizabeth Hospital, Woodville Rd, Woodville, SA, 5011, Australia. Email:
Andrew Vakulin, Flinders Medical Centre (RGH C-Block), GPO Box 2100, Adelaide 5001, South
© Sleep Research Society 2019. Published by Oxford University Press on behalf of the Sleep Research
Society. All rights reserved. For permissions, please e-mail journals.permissions@oup.com.
Abstract
Study objectives: Quantitative EEG measures of sleep may identify vulnerability to obstructive
sleep apnea (OSA) sequelae, however, small clinical studies of sleep microarchitecture in OSA
during rapid eye movement (REM) and non-REM (NREM) sleep and OSA severity among a large
t
ip
Methods: All-night EEG (F4-M1) recordings from full in-home polysomnography (Embletta
X100) in 664 men with no prior OSA diagnosis (age ≥40) were processed following exclusion of
cr
artefacts. Power spectral analysis included non-REM and REM sleep computed absolute EEG
us
power for delta, theta, alpha, sigma and beta frequency ranges, total power (0.5-32 Hz) and EEG
an
slowing ratio.
Results: Apnea hypopnea index (AHI) ≥10/h was present in 51.2% (severe OSA [AHI ≥30/h]
M
11.6%). In mixed effects regressions, AHI was positively associated with EEG slowing ratio and
EEG power across all frequency bands in REM sleep (all p<0.05); and with beta power during
ed
NREM sleep (p=0.04). Similar associations were observed with oxygen desaturation index (3%).
pt
Percentage total sleep time with oxygen saturation <90% was only significantly associated with
increased delta, theta and alpha EEG power in REM sleep. No associations with subjective
ce
with increased EEG power and EEG slowing predominantly in REM sleep, independent of
insomnia. Further study is required to assess if REM EEG slowing related to nocturnal hypoxemia
is more sensitive than standard PSG indices or sleepiness in predicting cognitive decline.
Keywords: quantitative EEG analysis, power spectral analysis, sleep disordered breathing,
during REM and NREM sleep (frequency range 0.5-32Hz). To date, quantitative analysis of sleep
EEG is limited to small, clinical samples and generalisability of these results to the broader
t
ip
population is unknown. During REM sleep, increasing OSA severity was associated with increases
cr
in power across all frequency bands and EEG slowing. Although speculative, REM EEG slowing
may provide a new biomarker of cognitive impairment in OSA. The utility of nocturnal hypoxemia-
us
related REM EEG slowing to identify patients at risk of cognitive decline and adverse
an
cardiometabolic outcomes requires study.
M
ed
pt
ce
Ac
3
Introduction
The prevalence of obstructive sleep apnea (OSA) is increasing in line with that of obesity and has
been reported to be present in up to 50% of men aged over 40 years.1,2 However, within the large
such as apnea hypopnea index (AHI) and oxygenation desaturation index (ODI), correlate
t
ip
performance.3,4
cr
us
Recordings of brain electrical activity using electroencephalography (EEG) are routinely acquired
during overnight sleep studies in the assessment of OSA. However scoring is typically limited to
an
assessment of sleep stages and micro-arousals, and potentially useful phenotypic information may
M
be missed. Quantitative EEG (qEEG) analysis of sleep oscillations using power spectral analysis
(PSA) may offer more sensitive markers for identifying different patient phenotypes at risk of
ed
adverse health, performance, and safety outcomes.5-7 In addition, EEG measures may be useful in
insights into the trait-like characteristics in individuals and represent a unique “fingerprint” of brain
ce
activity.8 Moreover, EEG changes are also state-dependent and sensitive to the demands of prior
Nine small controlled studies have previously assessed sleep microarchitecture in OSA using PSA6.
As reviewed recently by D’Rozario et al.,6 and Puskas et al.,11 only 4 small studies have examined
brain activity with PSA in NREM and REM sleep in OSA and controls, across a range of EEG
frequencies (0.5-32 Hz).12-15 Reported abnormalities during NREM sleep in OSA include decreased
spindle frequency activity in the sigma EEG power range and fewer sleep spindle events,12,14,16,17
4
and increased fast frequency activity (beta EEG power).12,16 Significant deficits in slow wave
activity (SWA, delta EEG power), the putative marker of homeostatic sleep need and depth, have
also been shown in OSA compared to controls in some12,13,16 but not all studies14,18,19 of NREM
slow to fast frequency EEG activity) over frontal, central, and parietal regions in severe untreated
OSA compared to controls that was unrelated to hypoxemia, objective sleepiness and inconsistently
t
related to AHI.20, 21
ip
cr
Insomnia, in which cortical hyperarousal is proposed to be integral to the disorder, is associated
us
with increased beta and sigma EEG power in NREM sleep.22 Given that insomnia may co-occur in
an
over 30% of patients with OSA,23 the presence of insomnia symptoms may account for some faster
No studies in large population-based samples are available to determine how representative findings
of small clinical studies may be to the broader population, or to determine the relationship of sleep
pt
qEEG measures with OSA disease severity metrics and daytime sleepiness. Therefore, the aim of
ce
this study was to examine the relationship between sleep microarchitecture and disease severity
metrics and subjective sleepiness in a large sample of community dwelling men (n=664) without a
Ac
prior diagnosis of OSA. We also sought to determine the effect of OSA on sleep microarchitecture
5
Methods
Study participants
randomly selected community dwelling men aged at least 35 years at baseline (2002-2006), of
largely Australian or European descent (96%) as previously described,24 and a sub-study of home-
t
based polysomnography occurred in 2010-11 in men without a previous OSA diagnosis as
ip
previously described.1 Briefly, 837 participants underwent successful 8-channel in-home unattended
cr
polysomnography (Embletta X100, Embla Systems, Broomfield, CO, USA) to measure EEG, EOG,
us
chin/submental EMG, nasal pressure, thoracic and abdominal effort, oximetry, and body position.
The EEG recording was obtained from 1 referential derivation F4-M1 lead site/s, plus left and right
an
electrooculogram (EOG), with a sampling rate of 200 Hz. Trained staff set-up and attached the
sleep study equipment, administered the Epworth Sleepiness Scale (ESS) and Pittsburgh Sleep
M
Quality Index questionnaire, obtained anthropometric measures and recorded current medications.
ed
Approval for the MAILES study was obtained from the Human Research Ethics Committees of the
North West Adelaide Health Service (approval number: 2010054). All participants provided written
pt
informed consent.
ce
Ac
All PSGs were manually scored, according to 2007 American Academy of Sleep Medicine (AASM)
alternate criteria as recommended by the AASM and by an expert panel of the Australasian Sleep
Association for use in prospective epidemiological studies.25 OSA was defined as an apnea
hypopnea index (AHI) ≥10/hour of sleep, with further categorisation into mild:10-19/hour,
moderate:20-29/hour, and severe: ≥30/hour, consistent with the findings of Ruehland et al. that an
AHI of 10/h using the alternate AASM definition is equivalent to an AHI of 5/h scored by the
desaturation index (ODI 3%) and percentage of sleep time with oxygen saturation <90% (TST90).
6
EEG data processing
Data pre-processing: Detailed descriptions of quantitative EEG analysis have been previously
described elsewhere.27,28 Synchronised European data format (EDF) and sleep stage files were
the 837 men who underwent sleep studies, PSG were of adequate quality for quantitative EEG
analysis in 734 men. An algorithm identified artefactual EEG data over consecutive non-
t
ip
overlapping 5-s epochs based on previously validated artefact detection threshold parameters.28
cr
Contaminated 5-s epochs were subsequently excluded from qEEG analysis. Verification of the
automated artefact scoring accuracy was performed in 10% of PSGs randomly selected for manual
us
review of over-read with excellent agreement shown.
an
EEG Power Spectral Analysis
M
After artefactual epochs were rejected, power spectra were obtained using a standard fast Fourier
ed
transform (FFT) with a rectangular weighting window,29 for each non-overlapping 5-s epoch of
EEG for the frontal derivation (F4-M1). We calculated absolute spectral power (µV2) in the delta,
pt
theta, alpha, sigma and beta bands defined as EEG activity in each of the frequency bands of 0.5–
ce
4.5, 4.5–8, 8–12, 12-15 and 15–32 Hz respectively, as well as total EEG power (0.5 – 32Hz). The
EEG power for each sleep-staged 30-s epoch of the PSG recording was calculated by averaging
Ac
data from up to six artefact-free 5-s epochs of EEG that comprised that 30-s recording segment. The
weighted average spectral power within the defined frequency bands was then computed for NREM
sleep (stages N2 and N3) and for REM sleep. A global measure of EEG slowing was also calculated
MAILES clinic assessments24 included determination of smoking, and alcohol use by self-report;
and depressive symptoms with the Center for Epidemiological Studies Depression scale (score ≥16)
7
or Beck Depression Inventory-1a (score ≥13). The SF-36 health-related quality of life
questionnaire was administered and the Physical Component Summary (PCS) and Mental
Component Summary (MCS) scales were calculated so as to generate a mean of 50 with a SD of 10.
glucose analysis. Insomnia symptoms during the past month23 were identified based on the
International Classification for Sleep Disorders-3 criteria30 by a response of ‘three or more times a
t
week’ to the following: “How often have you had trouble sleeping because 1) You cannot get to
ip
sleep within 30 mins? or 2) You wake up in the middle of the night or early morning?” Men were
cr
classified as having insomnia symptoms if they reported difficulty initiating or maintaining sleep
us
and clinically significant fatigue (SF-36 Vitality scale score one standard deviation below the
mean). This definition more closely meets current insomnia criteria,30 but without the 3 month
an
duration criterion for chronic insomnia. Sleep-disrupting medications recorded on the night of the
M
sleep study included opiates, anti-psychotics, anti-epileptics, anti-depressants, nicotine replacement
Statistical analysis
pt
Analyses were conducted in 664 men after exclusion of 70 (9.5%) participants reporting the use of
ce
one or more sleep disrupting medications including benzodiazepine medications (n=25, 3.4%). Data
Ac
were analyzed using IBM SPSS version 24.0 (IBM Corporation, Armonk, NY, USA) and R version
3.4.4 31 using the nlme package. Differences in characteristics of the population in relation to OSA
severity were determined by logistic regression analysis for dichotomous variables or one-way
analysis of variance with pairwise comparisons between severity group means compared to AHI
<10 using a Bonferroni correction for multiple comparisons for continuous variables (e.g. sleep
architecture). Median values (interquartile ranges) of qEEG variables by AHI and ODI severity
categories are presented due to the non-normality of the qEEG variables. Mann-Whitney U-tests
8
determined significant differences across categories of ODI 3% (< compared to ≥16/h32) or TST90
(comparison threshold set at the 75th percentile, equating to 4%), and significant daytime sleepiness
(ESS <11 compared to ≥11). Associations between OSA measures (AHI, ODI and TST90) and
each OSA measure two qEEG models are constructed, model 1 with qEEG power spectra as the
outcome, and model 2 with two summary qEEG measures (total power and slowing ratio) as the
t
outcome. After visual inspection of residual distributions all qEEG outcomes were log transformed.
ip
In all models fixed effects included three way interactions between sleep stage, outcome type
cr
(either model 1: wave type or model 2: total power v slowing ratio) and other fixed effects (OSA
us
metric and potential covariates). In the adjusted models the covariates included age, BMI, fasting
glucose, alcohol use, smoking, pulse pressure (mean systolic pressure - mean diastolic pressure) and
an
use of at least one sleep disrupting medication, insomnia, depressive symptoms and ESS scores.
M
Random intercepts were included per individual with an assumed exponential correlation structure
for the power spectra models, and a compound symmetry correlation structure for the total power
ed
and slowing ratio models. Missing covariate data (pulse pressure 3%; ESS 1.4%; smoking 2.0%;
glucose 2.7%; depressive symptoms 5.3%) were imputed using cohort means. Akaike's information
pt
criterion (AIC) is used to compare the predictive ability of models with different fixed effects (OSA
ce
measures). The model with the lowest AIC is considered the most predictive of a particular
outcome. Of note, absolute AIC values are not informative, and AIC cannot be compared across
Ac
outcomes. We report cohort standard deviation adjusted restricted maximum likelihood estimates
for the OSA measures and selected covariates (age and ESS), and use full maximum likelihood
9
Results
≥30/h). Mean (SD) total sleep time and REM sleep time were 374 (56) and 56 (21) minutes
t
the sample categorised by AHI severity are presented in Table 1. Mean total sleep time and sleep
ip
efficiency did not differ across OSA categories. However, men with severe OSA spent significantly
cr
less time in REM sleep and more time in N1 sleep. ESS scores did not differ across OSA
us
categories, but depressive symptoms increased and quality of life scores decreased in men with
severe OSA compared to men with AHI <10 (p<0.05). Insomnia was present in 72 (10.9) men and
an
did not significantly vary across OSA categories.
M
In covariate unadjusted analyses, AHI was significantly associated with increases in EEG power
pt
across all frequency bands and the EEG slowing ratio during REM sleep (Table 2). In NREM sleep
ce
stage specific analyses, a significant association of AHI and NREM beta power was only evident
during stage 2 sleep (Figure 1, Table S1). Covariate adjustment did not qualitatively change any of
Ac
the associations between EEG measures and AHI (Table 2, Table S1).
In covariate adjusted analyses, TST90 showed significant positive associations with delta, theta and
alpha EEG power during REM sleep while no associations with the REM EEG slowing ratio or
NREM sleep EEG power were seen (Table 3). In contrast, in unadjusted analyses, consistent with
10
the findings for AHI, ODI 3% was significantly associated with increased EEG power across all
frequency bands and the EEG slowing ratio during REM sleep, and with increased beta EEG power
during NREM sleep (Table S2). In covariate adjusted analyses, these associations persisted with
Relationship between EEG power spectra and clinical categories of AHI and nocturnal hypoxemia
t
ip
Median (interquartile range) values of absolute qEEG power are also presented in relation to
cr
clinical categories of AHI (Table 4), TST90 (Table 3) and ODI 3% (Table S2) and unadjusted
us
associations were generally consistent with those observed when indices were treated as continuous
measures. Compared to AHI <10/h, severe OSA (AHI≥ 30/h) was associated with increases in EEG
an
power across all frequency bands during REM sleep (Table 4) and with increased beta EEG power
during NREM sleep. TST90 ≥75th percentile and ODI ≥16/h were significantly associated with
M
increased power in delta and theta EEG power during REM sleep and ODI ≥16/h was additionally
ed
associated with alpha and sigma EEG power. Of the hypoxemia metrics however, only ODI ≥16/h
was significantly associated with increased beta power during NREM sleep.
pt
ce
There were no detectable associations between EEG power spectra in either NREM or REM sleep
with excessive daytime sleepiness when sleepiness was considered as a categorical variable (ESS
scores <11 compared to ≥11) or a continuous ESS scores (Table 5). Covariate adjustment did not
11
Predictive value of OSA metrics
The AIC of the AHI model was -7567.2, -7562.8 for the ODI model and -7515.6 for TST90 model
indicating that AHI explains more variance in qEEG than TST90, while ODI explained a similar
including AHI, ODI and TST90 were -1631.1, -1626.4, and -1606.0 respectively, suggesting that
t
ip
cr
Discussion
us
This is the first study to examine quantitative EEG sleep microarchitecture in NREM and REM
sleep across the night in relation to sleep apnea severity and daytime sleepiness while accounting
an
for insomnia in a large population-based sample of men (n=664). Increasing OSA severity (AHI)
M
was significantly and consistently associated with higher EEG power in all frequency bands and
increased EEG slowing during REM sleep, and with higher beta EEG power during NREM sleep.
ed
The observed associations of ODI were largely consistent with those for AHI. In contrast, TST90, a
measure of intermittent hypoxemia that was not highly correlated with AHI, was positively
pt
correlated with slow but not fast frequency EEG activity in REM sleep, and no associations were
ce
Our results regarding AHI are concordant with findings in small samples where EEG slowing in
frontal, central and parietal regions during REM sleep20,21 and increased beta EEG power during
NREM sleep12,16 in central and occipital regions have been previously reported in untreated OSA
patients compared to controls. However, we did not find significantly reduced spindle frequency
12
Previous studies have not specifically accounted for the effects of insomnia which is frequently
comorbid with OSA.23 Our findings were independent of insomnia, however future studies are
required to confirm these findings and to examine if there are differential effects of treatment on
Sleep EEG measures were not significantly related to subjectively-rated daytime sleepiness (ESS
t
ip
scores). Previous studies have also reported no relationship between qEEG measures and subjective
cr
sleepiness (ESS and Karolinska Sleepiness Scale)27 or objective sleepiness measured by the
multiple sleep latency test (MSLT).20,21,27 These previous studies also demonstrated no significant
us
relationships between EEG measures and nocturnal hypoxemia.20,21,27 In contrast, in a study of 10
an
severe OSA syndrome patients responsive to CPAP, shorter sleep latency on the MSLT was related
Hypoxemia during sleep (ODI and TST90) was positively related to slow frequency EEG activity
(delta and theta EEG power) in REM sleep, a finding that may be important in OSA related
pt
neurobehavioral sequelae. While worse sleep nocturnal hypoxemia has been associated with
ce
increased slow frequency wake EEG in some studies,21,33 our findings are in contrast to previous
small clinical studies of sleep EEG.13,15,18,20,21 Further studies are required to determine effects of
Ac
CPAP on sleep EEG microarchitecture and very limited evidence to date suggest that CPAP reduces
EEG slowing.6 A lack of association of daytime sleepiness with sleep EEG measures in our study
supports the observations in these prior studies that changes in EEG activity are inconsistently
Our findings support those studies showing no significant effect of OSA on NREM slow
waves.14,18,19 Some small studies do however report that in clinical samples of OSA patients, EEG
13
slowing is also evident during NREM sleep.34 Discrepancies in findings may be explained by small
samples, differing sample characteristics and inconsistent quantitative EEG methodologies across
studies. Previous studies recruited clinical OSA populations with potentially different clinical
t
ip
The consistent associations between OSA severity and sleep EEG measures predominantly in REM
cr
sleep with limited impact upon NREM EEG power is notable. These findings may be related to an
us
hypoxemia, hypercapnia and oxidative stress) that occur during REM sleep which tend to be longer,
an
with greater oxygen desaturations and increases in heart rate and blood pressure than those during
NREM sleep.35 Furthermore, associations have been reported between REM AHI, executive
M
Pathological similarities between OSA and Alzheimer’s disease and putative mechanisms linking
pt
the disorders have been recently reviewed37,38 and of relevance, REM sleep disturbance occurs in
ce
Alzheimer’s disease.39 EEG slowing during REM sleep has been observed in individuals with
memory difficulties and mild cognitive impairment,40 and in patients with mild to moderate
Ac
Alzheimer’s disease.41 This is thought to reflect possible degeneration of neuronal systems that
modulate cortical activation during wake and REM sleep.41 Furthermore, neurobehavioral deficits
in patients with severe OSA are not fully reversed by treatment.42,43 Although speculative as we did
not assess cognitive function, the EEG slowing associated with hypoxemia in REM sleep seen in
our study may indicate that changes in sleep EEG microarchitecture can be used to identify those at
risk of cognitive decline in OSA. Supporting this hypothesis, some of our previous work shows
14
greater EEG delta activity in REM associated with worse driving simulator performance in patients
not allow for topographical assessment of regional EEG power during sleep and how this relates to
OSA severity. A high density EEG investigation of sleep by Jones et al.13 in a middle aged, non-
t
ip
clinical sample showed no significant differences in absolute global EEG power between OSA
cr
(n=9) and control subjects (n=9) in NREM and REM sleep despite a posterior (parietal cortex)
regional reduction of sleep EEG power in subjects with OSA that was evident across all stages of
us
sleep and all frequency bands but most prominently in cycle 3 for slow wave activity. Furthermore,
an
specific topographical changes in sleep spindle frequency across the night, have also been
thalamo-cortical dysfunction.44 One possible reason that we did not detect reduced frontal spindle
frequency activity during NREM sleep as previously reported12-14, 16,17 may be related to the use of
ed
The strengths of our study include the unprecedented large sample of community-living men in
which to conduct quantitative EEG analysis, identification of insomnia symptoms, sleep stage
Ac
specific data and the ability to account for important confounding variables that impact sleep
structure and cortical activity. The majority of PSGs were of adequate quality for quantitative EEG
analysis and provide the largest examination to date of the relationship between cortical brain
activity and OSA. This sleep sub-study is part of a specific men’s health study, The Men Androgen
Inflammation Lifestyle Environment and Stress (MAILES) study, and we do not have data available
on women. Our study population of men only largely reflects the previous cited studies where the
samples have been comprised either exclusively or largely of men. Therefore, our findings in men
15
may not be generalizable to women but at the same time, were not influenced by previously
described sex differences in spectral power densities45-47 Daytime sleepiness was subjectively
assessed with the ESS, however, objective tests (e.g. MSLT or maintenance of wakefulness test)
studies while correlations between ESS scores and MSLT times range from poor to strong.49 The
t
ip
cr
In conclusion, in a large sample of community dwelling men, higher AHI was associated with
increased EEG power across all frequency bands and greater EEG slowing during REM sleep and
us
these associations were not modified by the presence of insomnia symptoms. In contrast,
an
associations between OSA severity and EEG measures during NREM sleep were limited to
increased beta power only. Our results add evidence to support the importance of the consideration
M
of REM sleep in human health as we have demonstrated in this cohort.50 Although we did not study
vigilance or attention in this study, we have previously shown that increased EEG power during
ed
sleep was associated with impaired performance in OSA, whereas traditional OSA metrics had no
pt
predictive value.27 This suggests that quantitative analysis of the sleep EEG derived from routine
sleep studies may be more related to and informative regarding OSA pathophysiology than self-
ce
reported sleepiness. Furthermore given recent evidence51 showing qEEG measures maybe more
Ac
sensitive than standard PSG indices to detect women at risk of cognitive decline, our findings
highlight the potential of qEEG to provide insight into identifying susceptibility to adverse health
outcomes.
16
Financial Support: This study was funded by a National Health and Medical Research Council of
Australia Project Grant (627227) 2010-12. Financial support for the conduct of sleep studies (2011)
and quantitative EEG analysis (2016) was obtained from the ResMed Foundation, California, USA.
t
ip
cr
us
an
M
ed
pt
ce
Ac
17
References
t
meta-review. Respirology 2013;18:61-70.
ip
4. Quan SF, Chan CS, Dement WC, et al. The association between obstructive sleep apnea and
cr
neurocognitive performance--the Apnea Positive Pressure Long-term Efficacy Study (APPLES).
Sleep 2011;34:303-14B.
us
5. Vakulin A, Catcheside PG, Baulk SD, et al. Individual variability and predictors of driving
simulator impairment in patients with obstructive sleep apnea. J Clin Sleep Med 2014;10:647-55.
an
6. D'Rozario AL, Cross NE, Vakulin A, et al. Quantitative electroencephalogram measures in
adult obstructive sleep apnea - Potential biomarkers of neurobehavioural functioning. Sleep Med
Rev 2017;36:29-42.
M
7. Landolt H-P, Dijk D-J. Chapter 30 - Genetics and Genomic Basis of Sleep in Healthy
Humans - In: Kryger M, Roth T, Dement WC, eds. Principles and Practice of Sleep Medicine
ed
2004;430:78-81.
10. Dijk DJ. Regulation and functional correlates of slow wave sleep. J Clin Sleep Med
Ac
2009;5:S6-15.
11. Puskas S, Kozak N, Sulina D, Csiba L, Magyar MT. Quantitative EEG in obstructive sleep
apnea syndrome: a review of the literature. Rev Neurosci 2017;28:265-70.
12. Guilleminault C, Do Kim Y, Chowdhuri S, Horita M, Ohayon M, Kushida C. Sleep and
daytime sleepiness in upper airway resistance syndrome compared to obstructive sleep apnoea
syndrome. Eur Respir J 2001;17:838-47.
13. Jones SG, Riedner BA, Smith RF, et al. Regional reductions in sleep
electroencephalography power in obstructive sleep apnea: a high-density EEG study. Sleep
2014;37:399-407.
18
14. Ondze B, Espa F, Dauvilliers Y, Billiard M, Besset A. Sleep architecture, slow wave activity
and sleep spindles in mild sleep disordered breathing. Clin Neurophysiol 2003;114:867-74.
15. Wang D, Piper AJ, Yee BJ, et al. Hypercapnia is a key correlate of EEG activation and
daytime sleepiness in hypercapnic sleep disordered breathing patients. J Clin Sleep Med
t
Neural Plast 2016;2016:7328725.
ip
18. Heinzer R, Gaudreau H, Decary A, et al. Slow-wave activity in sleep apnea patients before
cr
and after continuous positive airway pressure treatment: contribution to daytime sleepiness. Chest
2001;119:1807-13.
us
19. Himanen SL, Virkkala J, Huupponen E, Hasan J. Spindle frequency remains slow in sleep
apnea patients throughout the night. Sleep Medicine 2003;4:229-34.
an
20. Morisson F, Decary A, Petit D, Lavigne G, Malo J, Montplaisir J. Daytime sleepiness and
EEG spectral analysis in apneic patients before and after treatment with continuous positive airway
M
pressure. Chest 2001;119:45-52.
21. Morisson F, Lavigne G, Petit D, Nielsen T, Malo J, Montplaisir J. Spectral analysis of
wakefulness and REM sleep EEG in patients with sleep apnoea syndrome. Eur Respir J
ed
1998;11:1135-40.
22. Feige B, Baglioni C, Spiegelhalder K, Hirscher V, Nissen C, Riemann D. The
pt
23. Lang C, Appleton S, Vakulin A, et al. Comorbid OSA and insomnia increases depression
prevalence and severity in men Respirology 2017;22:1407-15.
Ac
24. Grant JF, Martin SA, Taylor AW, et al. Cohort Profile: The Men Androgen Inflammation
Lifestyle Environment and Stress (MAILES) Study Int J of Epidemiol 2014;43:1040-53.
25. ASTA/ASA. Commentary on AASM Manual for the Scoring of Sleep and Associated
Events. December 2010. http://www.sleep.org.au/information/sleep-documents/astaasa-
commentary-on-aasm-manual-for-scoring-of-sleep-associated-events-december-2010 (accessed
5/12/2018).
26. Ruehland WR, Rochford PD, O'Donoghue FJ, Pierce RJ, Singh P, Thornton AT. The new
AASM criteria for scoring hypopneas: impact on the apnea hypopnea index. Sleep 2009;32:150-7.
19
27. Vakulin A, D'Rozario A, Kim JW, et al. Quantitative sleep EEG and polysomnographic
predictors of driving simulator performance in obstructive sleep apnea. Clin Neurophysiol
2016;127:1428-35.
28. D'Rozario AL, Dungan GC, 2nd, Banks S, et al. An automated algorithm to identify and
t
30. Sateia MJ. International classification of sleep disorders-third edition: Highlights and
ip
modifications. Chest 2014;146:1387-94.
cr
31. R Core Team. R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. https://www.R-project.org/. 2018.
us
32. Chung F, Liao P, Elsaid H, Islam S, Shapiro CM, Sun Y. Oxygen Desaturation Index from
Nocturnal Oximetry: A Sensitive and Specific Tool to Detect Sleep-Disordered Breathing in
an
Surgical Patients. Anesthesia & Analgesia 2012;114:993-1000.
33. Xiromeritis AG, Hatziefthimiou AA, Hadjigeorgiou GM, Gourgoulianis KI,
M
Anagnostopoulou DN, Angelopoulos NV. Quantitative spectral analysis of vigilance EEG in
patients with obstructive sleep apnoea syndrome: EEG mapping in OSAS patients. Sleep Breath
2011;15:121-8.
ed
34. Wang D, Thomas RJ, Yee BJ, Grunstein RR. Hypercapnia is more important than hypoxia
in the neuro-outcomes of sleep-disordered breathing. J App Physiol 2016;120:1484.
pt
35. Findley LJ, Wilhoit SC, Suratt PM. Apnea duration and hypoxemia during REM sleep in
patients with obstructive sleep apnea. Chest 1985;87:432-6.
ce
36. Duffy SL, Lagopoulos J, Terpening Z, et al. Association of Anterior Cingulate Glutathione
with Sleep Apnea in Older Adults At-Risk for Dementia. Sleep 2016;39:899-906.
Ac
20
40. Brayet P, Petit D, Frauscher B, et al. Quantitative EEG of Rapid-Eye-Movement Sleep: A
Marker of Amnestic Mild Cognitive Impairment. Clin EEG Neurosci 2016;47:134-41.
41. Petit D, Lorrain D, Gauthier S, Montplaisir J. Regional spectral analysis of the REM sleep
EEG in mild to moderate Alzheimer's disease. Neurobiol Aging 1993;14:141-5.
t
44. Schönwald SV, Carvalho DZ, de Santa-Helena EL, Lemke N, L Gerhardt GJ. Topography-
ip
specific spindle frequency changes in Obstructive Sleep Apnea. BMC Neuroscience 2012;13:89.
cr
45. Latta F, Leproult R, Tasali E, Hofmann E, Van Cauter E. Sex Differences in Delta and
Alpha EEG Activities in Healthy Older Adults. Sleep 2005;28:1525-34.
us
46. Luca G, Haba Rubio J, Andries D, et al. Age and gender variations of sleep in subjects
without sleep disorders. Ann Med 2015;47:482-91.
an
47. Svetnik V, Snyder ES, Ma J, Tao P, Lines C, Herring WJ. EEG spectral analysis of NREM
sleep in a large sample of patients with insomnia and good sleepers: effects of age, sex and part of
M
the night. J Sleep Res 2017;26:92-104.
48. Johns MW. Sensitivity and specificity of the multiple sleep latency test (MSLT), the
maintenance of wakefulness test and the epworth sleepiness scale: failure of the MSLT as a gold
ed
obstructive sleep apnea during rapid eye movement (REM) sleep. Chest 2016;150:495–505.
51. Djonlagic I, Aeschbach D, Harrison SL, et al. Associations between quantitative sleep EEG
Ac
21
Table 1. Characteristics of the population (n=664) categorised by OSA severity.
Mean (SD)
Age, years 58.9 (10.5) 61.9 (10.7) 63.2* (10.4) 63.6* (11.7)
t
Body mass index, kg/m2 27.7 (3.7) 28.4 (4.0) 29.2* (3.9) 30.7* (4.9)
ip
Pulse pressure, mmHg 49.0 (12.5) 51.6 (13.2) 53.9* (15.5) 53.0 (14.9)
cr
Glucose, mmol/L 5.4 (1.2) 5.4 (1.3) 5.4 (1.3) 5.8 (2.5)
us
% (n)
Current smokers 16.9 (54) 10.6 (18) 19.3 (17) 8.2 (6)
an
High risk alcohol use 3.7 (12) 7.5 (13) 5.6 (5) 2.6 (2)
Oxygen desaturation index ≥16/hr 0 (0) 13.3 (23) 73.3 (66) 94.8 (73)
Total sleep time oxygen desaturation 11.1 (36) 19.7* (34) 40.0* (36) 53.2* (41)
ed
<90% ≥4.0%
Total sleep time, min 374 (60) 374 (53) 378 (52) 369 (51)
ce
Sleep efficiency, % 78.9 (9.9) 79.7 (9.9) 78.4 (9.7) 76.8 (10.8)
Stage 1 (N1), % 13.7 (5.4) 13.7 (5.7) 17.1* (6.2) 17.5* (7.4)
Ac
Stage 2 (N2), % 56.0 (8.2) 53.9 (8.7) 53.6 (8.3) 53.2 (9.4)
Stage 3 (N3), % 15.1 (8.3) 17.2 (8.9) 14.4 (8.1) 15.5 (10.2)
NREM, mins 316 (51) 316 (47) 321 (46) 318 (47)
Daytime dysfunction
22
Epworth Sleepiness Scale, mean (SD) 5.8 (3.7) 5.9 (3.8) 6.3 (3.7) 6.1 (3.6)
Physical component summary scale 51.6 (7.2) 50.9 (8.2) 49.5 (9.5) 47.6* (10.3)
Insomnia symptoms [% (n)] 10.2 (33) 12.1 (21) 8.9 (8) 13.0 (10)
Depressive symptoms [% (n)] 4.8 (15) 9.2 (15) 10.6 (9) 11.3* (8)
t
ip
*
p <0.05 compared to AHI <10.
NREM- non-rapid eye movement sleep, REM- rapid eye movement sleep.
cr
Depressive symptoms were identified with the Center for Epidemiological Studies Depression scale
us
(score ≥16) or Beck Depression Inventory-1a (score ≥13).
an
M
ed
pt
ce
Ac
23
Table 2. Unadjusted and covariate adjusted mixed effects regression estimated associations of
Unadjusted Adjusted*
NREM
t
Theta
ip
-0.003 (-0.018, 0.013) 0.75 -0.001 (-0.016, 0.015) 0.93
Alpha -0.004 (-0.02, 0.011) 0.59 -0.002 (-0.017, 0.014) 0.84
cr
Sigma 0.001 (-0.015, 0.016) 0.92 0.007 (-0.009, 0.023) 0.38
us
Beta 0.014 (-0.001, 0.03) 0.060 0.015 (-0.001, 0.031) 0.060
Total power -0.010 (-0.027, 0.0071) 0.25 -0.003 (-0.02, 0.015) 0.77
an
Slowing ratio -0.014 (-0.031, 0.0027) 0.090 -0.009 (-0.026, 0.009) 0.32
REM
M
Total power 0.041 (0.024, 0.057) <0.001 0.045 (0.028, 0.062) <0.001
Slowing ratio 0.023 (0.007, 0.04) 0.005 0.031 (0.013, 0.048) <0.001
Ac
Standardized β coefficients are presented and estimates represent change in log transformed EEG power.
Delta: 0.5-4 Hz; Theta: 4.5-8Hz; Alpha: 8-12 Hz; Sigma: 12-15 Hz; Beta: 15-32 Hz; Total power: 0.5-32Hz.
*
Multivariable mixed effects regression models were adjusted for age, BMI, fasting glucose, alcohol use,
smoking, pulse pressure (mean systolic pressure - mean diastolic pressure), insomnia, depression and ESS
scores.
24
Table 3. Median (interquartile range) sleep EEG spectral power by percentage total sleep
time with oxygen saturation <90% (TST90) and mixed effects regression estimated
percentile percentile
t
(n=516) (n=147)
ip
NREM
cr
Delta 279 (192, 407) 264 (192, 387) -0.002 (-0.017, 0.014) 0.84 0.008 (-0.008, 0.024) 0.31
Theta 26 (21, 35) 27 (20, 35) 0.007 (-0.009, 0.022) 0.39 0.011 (-0.005, 0.027) 0.17
us
Alpha 16 (12, 22) 16 (12, 22) -0.0001 (-0.016, 0.015) 0.99 0.003 (-0.013, 0.019) 0.68
Sigma 6.2 (4.7, 8.2) 5.9 (4.4, 7.7) -0.007 (-0.022, 0.009) 0.40 -0.0002 (-0.016, 0.016) 0.98
an
Beta 10.1 (8.1, 12.5) 10.4 (8.3, 13.1) 0.008 (-0.007, 0.024) 0.28 0.009 (-0.008, 0.024) 0.29
Total 340 (244, 493) 331 (246, 458) -0.0006 (-0.017, 0.016) 0.95 0.008 (-0.010, 0.025) 0.39
M
Slowing 9.0 (7.0, 12.4) 9.7 (6.6, 12.9) -0.003 (-0.02, 0.014) 0.73 0.004 (-0.014, 0.021) 0.69
ratio
ed
REM
Delta 82 (65, 113) 97* (69, 137) 0.028 (0.012, 0.043) <0.001 0.033 (0.017, 0.049) <0.001
pt
Theta 14 (11, 19) 15* (12, 22) 0.020 (0.004, 0.035) 0.01 0.021 (0.005, 0.037) 0.008
ce
Alpha 8.2 (6.3, 11.0) 8.9 (6.6, 12.2) 0.018 (0.002, 0.033) 0.02 0.016 (0.0004, 0.032) 0.040
Sigma 3.4 (2.7, 4.6) 3.6 (3.0, 4.8) 0.015 (-0.0007, 0.03) 0.06 0.013 (-0.003, 0.029) 0.11
Ac
Beta 9.3 (7.4, 12.4) 9.7 (7.7, 13.1) 0.012 (-0.003, 0.028) 0.11 0.014 (-0.002, 0.03) 0.09
Total 120 (98, 157) 135* (105, 188) 0.025 (0.008, 0.042) 0.003 0.028 (0.011, 0.046) 0.001
Slowing 4.5 (3.5, 6.4) 5.0* (3.7, 6.9) 0.011 (-0.006, 0.028) 0.18 0.016 (-0.002, 0.034) 0.07
ratio
*
p<0.05 for ≥75th percentile (TST90 ≥ 4.01%) compared to <75th percentile TST90
Standardized β coefficients are presented and estimates represent change in log transformed EEG power.
25
†
Multivariable mixed effects regression models were adjusted for age, BMI, fasting glucose, alcohol use,
smoking, pulse pressure (mean systolic pressure - mean diastolic pressure), insomnia, depression and ESS.
Delta: 0.5-4 Hz; Theta: 4.5-8Hz; Alpha: 8-12 Hz; Sigma: 12-15 Hz; Beta: 15-32 Hz; Total power: 0.5-32Hz.
26
Table 4. Median (interquartile range) EEG spectral power in NREM and REM sleep in
NREM
t
Delta 270 (200, 418) 296 (195, 410) 241 (177, 338) 275 (172, 426) 275 (192, 404)
ip
Theta 26.5 (20.8, 34.1) 27.0 (20.5, 34.2) 25.6 (18.9, 35.4) 25.7 (20.4, 38.3) 26.2 (20.4, 34.6)
cr
Alpha 16.2 (12.5, 21.3) 16.1 (12.0, 21.0) 17.1 (11.9, 24.1) 16.1 (12.0, 20.4) 16.2 (12.2, 21.5)
us
Sigma 6.2 (4.6, 8.1) 6.1 (4.6, 7.7) 6.1 (4.7, 8.7) 6.1 (4.5, 8.0) 6.2 (4.6, 8.1)
Beta 9.9 (7.9, 12.3) 10.2 (7.9, 12.9) 10.0 (8.2, 12.8) 10.8 (8.6, 13.5) 10.2 (8.2, 12.7)
an
Total power 333 (251, 502) 351 (249, 504) 305 (217, 414) 342 (227, 470) 338 (245, 488)
EEG slowing 9.3 (7.1, 12.6) 9.7 (7.3, 13.1) 8.2 (6.2, 11.5) 8.7 (6.4, 12.7) 9.1 (6.9, 12.5)
M
ratio
REM
ed
Delta 82 (66, 110) 85 (67, 117) 79 (63, 107) 105 (71, 182) 84 (66, 116)
Theta 14.0 (11.3, 18.5) 14.4 (11.7, 18.9) 14.2 (10.8, 20.0) 17.1 (12.7, 22.9) 14.4 (11.4, 19.2)
pt
Alpha 8.1 (6.4, 10.7) 6.3 (8.3, 11.2) 8.8 (6.3, 12.0) 9.2 (6.9, 13.1) 8.3 (6.4, 11.2)
ce
Sigma 3.3 (2.7, 4.3) 3.6 (2.8, 4.7) 3.5 (2.8, 4.8) 4.0 (3.1, 5.2) 3.5 (2.8, 4.6)
Beta 9.1 (7.4, 12.0) 9.9 (7.2, 12.6) 9.5 (7.4, 12.5) 10.5 (8.2, 14.4) 9.4 (7.4, 12.5)
Ac
Total power 121 (98, 150) 124 (100, 161) 117 (98, 154) 149 (109, 231) 122 (99, 162)
Slowing ratio 4.5 (3.5, 6.7) 4.7 (3.7, 6.2) 4.4 (3.3, 6.2) 5.0 (3.6, 8.0) 4.6 (3.5, 6.5)
Delta: 0.5-4 Hz; Theta:4.5-8Hz; Alpha:8-12 Hz; Sigma: 12-15 Hz; Beta: 15-32 Hz; Total power:0.5-32Hz.
27
Table 5. Median (interquartile range) sleep EEG spectral power by excessive daytime
sleepiness by the Epworth Sleepiness Scale (ESS) and mixed effects regression estimated
(n=579) (n=75)
t
NREM
ip
Delta 275 (190, 405) 296 (208, 387) 0.005 (-0.011, 0.020) 0.53 -0.001 (-0.017, 0.014) 0.88
cr
Theta 26 (20, 34) 29 (22, 37) 0.015 (-0.001, 0.030) 0.06 0.009 (-0.006, 0.025) 0.23
Alpha 16 (12, 21) 18 (12, 23) 0.015 (-0.0004, 0.031) 0.05 0.009 (-0.007, 0.024) 0.26
us
Sigma 6.1 (4.6, 8.1) 6.2 (4.3, 8.0) 0.010 (-0.006, 0.025) 0.22 -0.002 (-0.017, 0.014) 0.85
Beta 10.3 (8.2, 12.7) 9.4 (7.7, 12.8) -0.003 (-0.019, 0.012) 0.66 -0.011 (-0.027, 0.005) 0.16
an
Total 335 (244, 489) 374 (257, 479) 0.006 (-0.011, 0.023) 0.47 -0.0001 (-0.017, 0.017) 0.99
power
M
Slowing 9.0 (6.9, 12.6) 9.6 (7.4, 11.7) -0.003 (-0.02, 0.014) 0.73 -0.002 (-0.019, 0.016) 0.85
ratio
ed
REM
Delta 84 (66, 117) 82 (68, 109) -0.010 (-0.025, 0.006) 0.22 -0.01 (-0.026, 0.006) 0.19
pt
Theta 14 (11, 19) 16 (12, 21) 0.011 (-0.005, 0.026) 0.16 0.0061 (-0.010, 0.022) 0.44
ce
Alpha 8.3 (6.3, 11.0) 9.0 (7.2, 12.6) 0.014 (-0.001, 0.03) 0.07 0.010 (-0.005, 0.026) 0.19
Sigma 3.4 (2.8, 4.6) 3.5 (2.8, 5.1) 0.007 (-0.008, 0.023) 0.34 0.0006 (-0.015, 0.016) 0.93
Ac
Beta 9.3 (7.4, 12.6) 9.2 (7.2, 11.8) -0.0004 (-0.016, 0.015) 0.96 -0.007 (-0.022, 0.009) 0.40
Total 121 (99, 162) 124 (101, 161) -0.005 (-0.022, 0.012) 0.56 -0.007 (-0.024, 0.01) 0.43
power
Slowing 4.6 (3.5, 6.5) 4.6 (3.6, 6.3) -0.014 (-0.031, 0.003) 0.11 -0.010 (-0.027, 0.008) 0.26
ratio
*
Multivariable mixed effects regression models were adjusted for age, AHI, BMI, fasting glucose, alcohol
use, smoking, pulse pressure (mean systolic pressure - mean diastolic pressure), insomnia, and depression.
Standardized β coefficients are presented and estimates represent change in log transformed EEG power.
28
Figure 1. Association of apnea hypopnea index (AHI) with NREM sleep stage and REM sleep EEG
stage 3 sleep.
Delta: 0.5-4 Hz; Theta:4.5-8Hz; Alpha:8-12 Hz; Sigma: 12-15 Hz; Beta: 15-32 Hz; Total
t
ip
power:0.5-32Hz.
cr
Standardized estimates for multivariable model presented in Supplementary Table 1.
us
an
M
ed
pt
ce
Ac
29