By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Anne Germain, PhD, Associate Professor of Psychiatry and Psychology, University of Pittsburgh, 3811 O’Hara Street, Room E-1118, Pittsburgh, PA 15213, USA. Tel.: 412-246-6413; fax: 412-246-5300; e-mail: email@example.com
Sleep disturbances are a hallmark feature of post-traumatic stress disorder (PTSD), and associated with poor clinical outcomes. Few studies have examined sleep quantitative electroencephalography (qEEG), a technique able to detect subtle differences that polysomnography does not capture. We hypothesized that greater high-frequency qEEG would reflect ‘hyperarousal’ in combat veterans with PTSD (n =16) compared to veterans without PTSD (n =13). EEG power in traditional EEG frequency bands was computed for artifact-free sleep epochs across an entire night. Correlations were performed between qEEG and ratings of PTSD symptoms and combat exposure. The groups did not differ significantly in whole-night qEEG measures for either rapid eye movement (REM) or non-REM (NREM) sleep. Non-significant medium effect sizes suggest less REM beta (opposite to our hypothesis), less REM and NREM sigma and more NREM gamma in combat veterans with PTSD. Positive correlations were found between combat exposure and NREM beta (PTSD group only), and REM and NREM sigma (non-PTSD group only). Results did not support global hyperarousal in PTSD as indexed by increased beta qEEG activity. The correlation of sigma activity with combat exposure in those without PTSD and the non-significant trend towards less sigma activity during both REM and NREM sleep in combat veterans with PTSD suggests that differential information processing during sleep may characterize combat-exposed military veterans with and without PTSD.
Sleep disturbances have been called the hallmark feature of post-traumatic stress disorder (PTSD) (Ross et al., 1989). Insomnia and nightmares are highly prevalent in PTSD (Lewis et al., 2009), associated with worse clinical outcomes (depression, suicidality, substance use, poorer perceived physical health) (Krakow et al., 2000; Nishith et al., 2001), and are often resistant to PTSD treatments (Zayfert and Deviva, 2004). Furthermore, insomnia and rapid eye movement (REM) sleep fragmentation early after a trauma are associated with future development of PTSD (Mellman et al., 2002). Identifying objective markers of sleep disturbances in PTSD may have important clinical implications for the treatment of this pervasive disorder.
Traditional polysomnography (PSG) has not been able to capture objective, PTSD-specific sleep disturbances (Kobayashi et al., 2007). Quantitative EEG (qEEG) is another sensitive objective measure obtained via spectral analysis of EEG that examines the microarchitecture of EEG signals. For example, high-frequency beta activity (16–32 Hz) and gamma activity (32–50 Hz) have been used as putative indices of central arousal during sleep in the study of insomnia (Merica et al., 1998; Perlis et al., 2001). Thus, qEEG affords the opportunity to examine subtle and functionally important differences in central activity within sleep states that may provide insights into the central mechanisms underlying sleep complaints in PTSD.
There have been relatively few studies using sleep qEEG in PTSD. Compared to controls, Vietnam veterans with chronic PTSD (and high rates of psychiatric comorbidity) had a higher ratio of rapid eye movement/non-rapid eye movement (REM/NREM) beta power, reduced low-frequency qEEG activity during slow wave sleep (SWS) only, and a positive correlation between NREM sigma (12–15.8 Hz) and hyperarousal symptoms (Woodward et al., 2000). A small ecological study using in-home PSG found PTSD to be associated with increased beta activity (20–32 Hz) and increased delta activity (0.5–4 Hz) (Germain et al., 2006).
In contrast to these studies, a prospective study following 10 subjects involved in a life-threatening event (five of whom subsequently developed PTSD), REM sleep beta power was correlated negatively to nightmare severity and overall PTSD symptom severity 6 weeks later (Mellman et al., 2007).
While increased central arousal—as indexed by increased high-frequency qEEG activity—has been postulated in PTSD, the evidence from the three available studies summarized above is mixed. In this study, PSG and qEEG were used to compare the macro- and microstructure of REM sleep and NREM sleep in combat veterans of Operation Enduring Freedom (OEF) and/or Operation Iraqi Freedom (OIF) with and without a current diagnosis of PTSD. We hypothesized that, in this young sample, free of psychotropic medications and psychiatric comorbidity, combat veterans with PTSD would show increased beta and gamma EEG power as a reflection of central ‘hyperarousal’ during both REM sleep and NREM sleep. Given that the qEEG profiles during sleep have not been characterized extensively in this population, we performed the same analyses on other frequency bands (delta, theta, alpha, sigma, gamma) for exploratory purposes. The correlations between qEEG activity and clinical variables relating to PTSD and combat exposure were also explored.
Data for the current analyses were drawn from three separate studies (MH083035, PR054093, PT073961; Principle Investigator: Germain) conducted at the University of Pittsburgh School of Medicine Neuroscience Clinical and Translational Research Center (UL1RR024153; Principle Investigator: Reis). All studies employed the same clinical assessments, self-report questionnaires, PSG methods and EEG measurements and analysis approaches. All studies were approved by the University of Pittsburgh Institutional Review Board, and all participants provided written informed consent.
Participants were combat-exposed OEF/OIF veterans between the ages of 18 and 50 years, who did (n =16) or did not (n =13) meet diagnostic criteria for PTSD. Documentation of military service was obtained at this first visit from all potential participants by provision of DD Form 214, a form issued by the Department of Defense upon a military service member’s separation from active military duty. Active-duty personnel were asked to provide a valid military identification form. Participants who were deemed medically healthy after a comprehensive physical examination supplemented by routine blood work and a urine drug screen were eligible for inclusion in these studies.
Exclusion criteria included unstable medical condition, suicidality or recent hospitalization for suicidality, psychotic or bipolar disorder, untreated severe major depressive disorder, current substance abuse or dependence disorder (past 3 months), restless leg syndrome, apnea–hypopnea index > 15 and pregnancy or breast feeding. Participants retained for the present study were free of psychotropic medications known to affect sleep [e.g. antidepressants, benzodiazepines or other gamma-aminobutyric acid (GABA)-ergic medications, antipsychotics]. In addition, participants were not included if they were currently receiving treatment for traumatic brain injury or concussive symptoms at the time of the study.
Participants completed baseline sleep assessments, including the Pittsburgh Sleep Quality Index (PSQI) (Buysse et al., 1989) and the Insomnia Severity Index (ISI) (Bastien et al., 2001). Disruptive nocturnal behaviors often seen in PTSD (trauma-related nightmares, nocturnal intrusive memories, distressing dreams not related to the trauma, sleep terrors, nocturnal panic attacks, dream enactment behaviors and other complex motor behaviors) were assessed with the PSQI addendum for PTSD (PSQI-A) (Germain et al., 2005). Psychiatric diagnosis was determined using the Structured Clinical Interview for the DSM-IV (SCID) (Spitzer et al., 1994) by certified assessors. The presence and severity of PTSD was assessed using the clinician-administered PTSD scale (CAPS) (Blake et al., 1995), the gold standard for the assessment of PTSD. Self-report measures of clinical symptoms included the PTSD Checklist (PCL) (Forbes et al., 2001) and the Beck Depression Index (BDI) (Beck et al., 1961). Exposure to combat was measured with the Combat Exposure Scale (CES) (Keane et al., 1989). As an estimate of time elapsed since combat-related traumas, the approximate number of months between the baseline assessment and both the earliest and the most recent significant traumas reported on the CAPS were recorded.
All subjects underwent 1 night of PSG to screen for sleep disorders (such as sleep apnea) followed by a 1-night baseline PSG for sleep staging and quantitative analysis. Bedtime and rise times were determined individually to match the participant’s habitual sleep schedule closely. PSG was conducted using Grass Telefactor M15 bipolar Neurodata amplifiers and Stellate-Harmonie collection software. The recording montage consisted of bilateral central (C3, C4) EEG leads referenced to A1 + A2; right and left electro-oculogram referenced to A1 + A2; and bipolar submentalis electromyogram (EMG). On the screening night, additional channels were used to monitor sleep-related breathing (nasal–oral thermistors, inductance plethysmography, fingertip oximetry, V2 EKG) and periodic limb movements (bilateral anterior tibialis EMG). EEG recordings used a high-frequency filter of 100 Hz, a low-frequency filter of 0.3 Hz and a 60-Hz notch filter. Sleep stages were scored in 20-s epochs, according to the Rechtschaffen and Kales criteria. REM density was calculated using an automated algorithm (Doman et al., 1995).
Spectral analysis (qEEG)
Methods for power spectral analysis have been published previously. Briefly, EEG signals were digitized at a rate of 256 Hz. The raw digitized data were band-limited to 64 Hz using a low-pass finite impulse response filter, then decimated to 128 Hz for quantitative analyses. Using a locally developed automated algorithm, 4-s EEG epochs were rejected from analysis when certain threshold criteria were met bilaterally in the EOG channel, indicating an eye-movement artifact. High-frequency EEG artifacts, such as muscle twitches, were identified and excluded in 4-s bins with a previously validated and published algorithm that uses a moving window threshold (Brunner et al., 1996). This algorithm excludes 4-s epochs where power in the frequency range of 26.25–32 Hz exceeds the power in adjacent epochs by a factor of 4 or greater.
Power spectral analysis was used to quantify the frequency content of the sleep EEG from 0.50 to 50 Hz. The frequency bands of interest were defined as: delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), sigma (12–16 Hz), beta (16–32 Hz) and gamma (32–50 Hz). Non-overlapping 4-s epochs were weighted with a Hamming window, and periodograms were then computed for these epochs using the Fast Fourier transform (FFT). EEG spectra for each artifact-free 4-s epoch were then aligned with 20-s visually scored sleep stage data to exclude epochs scored as awake.
Variable distributions were first inspected for normality, and transformations were performed when appropriate to normalize the distribution prior to comparative analyses. The two study groups were compared on demographic variables, sleep and psychiatric clinical scales, and standard PSG variables using independent t-tests, using the Satterthwaite method in the case of unequal variances. Quantitative EEG absolute power within the frequency bands of interest was subjected to natural log transformations before statistical analysis. One participant was found to be a clear outlier for spectral profile and spectral analyses were first conducted without this (non-PTSD) participant’s data, and subsequently repeated while including this participant. Independent t-tests and Cohen’s d effect sizes were calculated for group differences on whole-night qEEG data. Spearman’s rho correlations were performed between spectral data and clinical variables related to PTSD symptomatology and combat exposure (CAPS, CAPS subscales, PCL, CES) within each group. Variables were considered significant at P <0.05 (two-tailed).
Our sample of 29 combat veterans consisted of 28 men (96.6%) and one woman (3.4%), of whom 25 were non-Hispanic Caucasian (86.2%), two were African American (6.9%), one was Hispanic Caucasian (3.4%) and one was Asian/Pacific Islander (3.4%). The mean age for the whole group was 28.4 years [standard deviation (SD)=4.4, range 23–40]. During combat, 17 participants (58.6%) reported exposure to blast, fire or explosion, seven participants (24.1%) reported a history of closed head injury, and three of them also endorsed a loss of consciousness. Two participants had been diagnosed previously with mild traumatic brain injury (one diagnosed during high school), but none had concussive symptoms or were being treated for any sequelae of traumatic brain injury at the time of participation.
Clinical ratings and self-report measures of trauma exposure, PTSD and other psychiatric symptoms and sleep symptoms for the PTSD and non-PTSD groups are presented in Table 1. Significant differences are seen between the groups on PTSD scales (CAPS total and its subscales, PCL total and PSQI-A) in the expected direction. The PTSD group also had significantly higher scores on the Beck Depression inventory—although both group means were well below clinically significant scores (Beck et al., 1961)—as well as more severe insomnia symptoms as determined by the ISI. The two groups did not differ on subjective sleep quality (PSQI), age, combat exposure (CES) or time since trauma exposure (neither first nor most recent trauma).
*Except when indicated by: †PTSD n =15, non-PTSD n =6; ‡PTSD n =14, non-PTSD n =13.
§Satterthwaite method reported due to unequal variances.
¶PTSD n =16, non-PTSD n =12.
**PTSD n =16, non-PTSD n =11.
−0.65, 27.0, 0.52
Time since earliest major trauma (months)†
0.64, 19, 0.53
Time since most recent major trauma (months)†
0.06, 19, 0.96
Combat Exposure Scale‡
−0.60, 25, 0.56
7.29, 27, 0.001
Sleep symptoms only§
3.41, 19.0, 0.003
Daytime symptoms only
7.32, 27, <0.001
4.61, 27, <0.001
4.36, 16.7, <0.001
5.90, 27, <0.001
3.83, 24.2, 0.001
PSQI-A for PTSD
2.47, 27, 0.02
1.66, 27, 0.11
2.29, 25, 0.03
Beck Depression Index (BDI)
2.58, 27, 0.01
PSG findings are presented in Table 2. There were no statistically significant differences in PSG between the two groups, including no differences in REM density, number of REM periods or length of REM periods.
Table 2. Mean (and standard deviations) for polysomnographic (PSG) parameters for the whole group and for veterans with post-traumatic stress disorder (PTSD) and without PTSD (non-PTSD)
Whole group n =29
PTSD n =16
Non-PTSD n =13
t-statistic, df, P-value
REM, rapid eye movement.
*Natural log transformation Ln performed before t-tests. Mean values [standard deviation (SD)] reported for original units.
†Square root transformation performed before t-tests. Mean values (SD) reported for original units.
‡Natural log transformation Ln(100−n−1) performed before t-tests. Mean values (SD) reported for original units.
§Satterthwaite method used due to unequal variances.
¶Natural log transformation Ln(n +0.1) performed before t-tests. Mean values (SD) reported for original units.
Sleep latency (min)*
−0.90, 27, 0.38
Total sleep time (min)
0.27, 27, 0.79
Wake time after sleep onset (min)†
0.86, 27, 0.40
−0.35, 27, 0.73
1.43, 27, 0.16
−1.56, 19.8, 0.14
0.44, 27, 0.66
1.18, 23.1, 0.25
0.15, 27, 0.88
REM time (min)
0.26, 27, 0.80
REM latency (min)*
0.54, 27, 0.59
REM density (counts per min)†
0.84, 27, 0.41
Average REM period duration (min)§
0.78, 21.7, 0.46
Number of REM periods
−0.27, 27, 0.79
Average NREM period duration (min)
−0.19, 27, 0.85
Number of NREM periods
−0.05, 27, 0.96
There were no statistically significant differences in whole-night REM or NREM spectral power between the PTSD and non-PTSD groups in any frequency band (Fig. 1). However, medium effect sizes detected in the whole-night spectral data suggest that PTSD group had less REM beta activity (Cohen’s d =0.38), less REM sigma (Cohen’s d =0.37), less NREM sigma (Cohen’s d =0.51) and more NREM gamma (Cohen’s d =0.44) compared to the non-PTSD group.
Correlations between activity bands and clinical variables related to PTSD symptomatology are shown in Table 3. In the PTSD group, combat exposure ratings were correlated positively to NREM beta activity, and no other correlations were found in the PTSD group. In the non-PTSD group, combat exposure ratings were correlated positively to REM sigma, NREM sigma and NREM alpha. Also in the non-PTSD group, PTSD avoidance symptoms were correlated negatively with NREM sigma and NREM alpha; self-reported PTSD symptoms were correlated negatively with NREM alpha.
Table 3. Correlations between quantitative electroencephalography (qEEG) activity bands and clinical variables within diagnostic groups*
Correlations tested were between rapid eye movement (REM) and non-REM (NREM) qEEG frequency bands and: post-traumatic stress disorder (PTSD) symptoms [clinician-administered PTSD Scale (CAPS); total score and hyperarousal, re-experiencing, and avoidance subscales); self-rated PTSD symptoms [PTSD Checklist (PCL)]; and combat exposure scale (CES).
*Correlations were no longer significant after Bonferroni correction for multiple comparisons: alpha = 0.05/30(=5 frequency bands × 6 clinical variables) = 0.002.
PTSD group (n =14)
NREM beta: CES
Non-PTSD group (n = 12)
REM sigma: CAPS avoidance
REM sigma: CES
NREM sigma: CES
NREM alpha: CAPS avoidance
NREM alpha: PCL (n =11)
NREM alpha: CES
The results presented above are after a spectral outlier was removed due to strong suspicion of artifactual data. Nevertheless, we reproduced the analyses detailed above after inclusion of this subject to determine if findings would change. There were still no whole-night differences in any band. Including this subject, REM sigma activity, NREM sigma activity and NREM gamma activity continued to show medium effect sizes, but REM beta activity no longer did (Cohen’s d =0.27). Including this subject in the correlational analyses did not change the magnitude or significance of previously reported correlations, except for the correlations between NREM alpha and combat exposure or NREM sigma and combat exposure.
In this study, we compared medication- and comorbidity free combat-exposed veterans with and without PTSD, during both REM sleep and NREM sleep, using quantitative EEG in addition to polysomnography. Except for expected differences in PTSD-related symptom severity and insomnia severity, the two groups of veterans did not differ on measures of depression or overall sleep quality. There were no differences between groups in PSG measures and the previously reported increases in REM density (Ross et al., 1994), number of REM periods or length of REM periods (Mellman et al., 2002) in PTSD were not replicated in this sample. REM sleep changes may be moderated by other factors, such a psychiatric comorbidity, past history of addictive disorders and medication (Kobayashi et al., 2007).
We hypothesized that the ‘hyperarousal’ that characterizes PTSD would be indexed during both REM and NREM sleep by increased beta and gamma qEEG power compared to the non-PTSD group. Contrary to this hypothesis, there were no statistically significant whole-night differences in beta or gamma power between the two groups. However, a medium effect size suggested less REM beta in the PTSD group, a finding in the opposite direction of our hypothesis. The only clinical variable showing a positive correlation with NREM beta activity was combat exposure in the PTSD group, the directionality of which is conceptually in line with our initial hypothesis. Similarly, a medium effect size was detected for the gamma band, where the PTSD group showed more gamma in NREM sleep that the non-PTSD group. However, further analysis showed that the groups did not differ on the PSQI, nor was gamma correlated to either sleep quality or insomnia severity (data not shown). Therefore, in young, medication-free combat veterans with and without PTSD uncomplicated by concurrent psychiatric comorbidity, neither beta activity nor gamma activity appear to be indices of hyperarousal during sleep in PTSD. A recent prospective small study showed less REM beta soon after trauma in those who later developed PTSD (Mellman et al., 2007), and the moderate effect size showing less REM beta in our PTSD group is consistent with this prior finding.
In primary insomnia, increased beta and gamma activity have been conceptualized as indicative of increased central arousal (Merica et al., 1998; Perlis et al., 2001). The current findings raise the possibility that the physiological basis of increased arousal as indexed by beta or gamma activity may arise from different underlying systems in primary insomnia and in PTSD. For instance, hyperarousal in primary insomnia may be related to increased alertness, conditioned arousal related to sleep cues (Perlis et al., 2001), sleep state dysregulation and/or a greater proportion of neuronal columns remaining active, while others have transitioned to a sleep state (Krueger and Obal, 1993). In PTSD, hyperarousal may reflect a failure of sleep-related emotional regulation or memory processing and/or increased autonomic and limbic activity (Germain et al., 2008; Nielsen and Levin, 2007) that may not be captured by this same indices. Probing the correlates of qEEG activity bands using sleep neuroimaging methods may clarify the underpinnings of these indices in healthy and clinical samples.
Rather than gross changes in sleep architecture or in patterns of heightened central arousal, our exploratory findings raise the possibility that sleep disturbances in PTSD may reflect alterations in sleep-related cognitive processes, as suggested by results in the sigma band, which were the most consistent during both REM and NREM sleep. There were medium effect sizes between groups suggesting that participants with PTSD had less REM sleep sigma and less NREM sleep sigma compared to participants without PTSD. In participants without PTSD (but not in those with PTSD), REM sleep and NREM sleep sigma activity was correlated positively with combat exposure ratings, and NREM sigma was correlated negatively with PTSD avoidance symptoms. Sigma (12–16 Hz) includes sleep spindles (12–14 Hz spikes during NREM lasting a few seconds). Sleep spindles have been associated with memory processing, including both consolidation and integration (Fogel and Smith, 2011; Tamminen et al., 2010). PTSD has been conceptualized as having at its core a failure of emotional memory processing during sleep, leading to the persistence of overly emotionally charged and vividly episodic intrusive memories, flashbacks and nightmares (re-experiencing symptoms) (Stickgold, 2002). While it is unclear what role sigma band activity may have in emotional memory processing in the context of PTSD, and although the correlations did not survive Bonferroni corrections, the observed group difference in the sigma activity band and the observed correlations between sigma and combat exposure raise the possibility that sigma activity may be a marker of resilience to combat exposure. The latter provides a testable hypothesis that deserves further investigation using well-controlled experimental paradigms to elucidate the potential neural and clinical correlates of sigma activity in the context of trauma exposure and/or PTSD and other stress-related outcomes.
Although whole-night REM sleep and NREM sleep alpha activity did not differ between the two groups of participants, NREM alpha activity was correlated positively to combat exposure and negatively to PTSD avoidance symptoms and self-report PTSD symptoms. While there is a less theoretical basis for alpha activity as a marker of information processing, it appears to be acting in a similar manner to sigma in this study.
Finally, delta or theta activity in whole-night REM sleep or NREM sleep did not differ in participants with and without PTSD. In addition, no significant correlations were observed between delta or theta activity and PTSD symptom severity or combat exposure. The absence of difference may reflect the relatively small (albeit significant) group differences on subjective sleep measures and on PSG sleep measures.
This study has several strengths relative to previous studies that have examined qEEG in PTSD. Our sample is relatively young and healthy, with less chronic PTSD not complicated by comorbid medical, sleep or psychiatric conditions, or by the use of psychotropic medications. The two groups were similar on measures of combat exposure, overall sleep quality and depression symptom severity. In addition, we were able to perform spectral analysis on whole-night sleep REM sleep and NREM sleep periods.
Nevertheless, some limitations must be acknowledged. The absence of a non-combat exposed group does not allow for determining how combat exposure alone may affect sleep, which may limit the generalizability of our results to civilian populations. Our sample may be biased due to the fact that our participants were seeking participation in a research study. The artifact rejection process, while necessary in order to remove contamination from eye movements and non-physiological activity, also limits the REM sleep samples to the tonic phase of REM sleep, even in the absence of group differences on REM density, a measure of phasic REM sleep activity. In addition, the restricted EEG montage employed prevented further spatial analysis on qEEG activity bands. The detection of medium effect sizes also suggests that group differences may be captured more robustly in larger samples of combat veterans.
The use of high-density EEG in future studies and the enhanced spatial resolution it provides may be necessary to capture the microstructural EEG signature that may best characterize PTSD or, alternatively, resilience to trauma. In a related manner, integrating EEG recordings methods with sleep neuroimaging techniques will be helpful in identifying the neural substrates and functional correlates of qEEG activity bands in healthy and clinical samples. This type of approach would allow for investigating whether physiological markers of central arousal, such as beta and gamma activity, may capture altered activity more effectively in specific neural circuits that are involved in PTSD relative to other conditions such as insomnia. Multi-modal measurement methods are also necessary to identify novel pathways for the treatment of both the sleep disruptions and daytime consequences of PTSD, and to identify potential biomarkers of resilience or risks of maladaptive stress reactions.
The authors gratefully acknowledge the contributions of the staff working on the Veterans Sleep Study Program and of the Neuroscience Clinical and Translational Research Center for their valuable and expert assistance in conducting these studies. Spectral analysis programs were written by Ray Vasko PhD and Rob Seres, with additional technical assistance from David Cashmere and Jim Havstad. This study was supported by the US Department of Defense (PR054093, PT073961; Principle Investigator: Germain), and the National Institutes of Health (MH083035, Principle Investigator: Germain; RR024153, Principle Investigator: Reis; HL082610, Principle Investigator: Buysse).
The authors do not have any conflicts of interest to report.