Sex differences in objective measures of sleep in post-traumatic stress disorder and healthy control subjects

Authors


Summary

A growing literature shows prominent sex effects for risk for post-traumatic stress disorder and associated medical comorbid burden. Previous research indicates that post-traumatic stress disorder is associated with reduced slow wave sleep, which may have implications for overall health, and abnormalities in rapid eye movement sleep, which have been implicated in specific post-traumatic stress disorder symptoms, but most research has been conducted in male subjects. We therefore sought to compare objective measures of sleep in male and female post-traumatic stress disorder subjects with age- and sex-matched control subjects. We used a cross-sectional, 2 × 2 design (post-traumatic stress disorder/control × female/male) involving83 medically healthy, non-medicated adults aged 19–39 years in the inpatient sleep laboratory. Visual electroencephalographic analysis demonstrated that post-traumatic stress disorder was associated with lower slow wave sleep duration (F(3,82) = 7.63, = 0.007) and slow wave sleep percentage (F(3,82) = 6.11, = 0.016). There was also a group × sex interaction effect for rapid eye movement sleep duration (F(3,82) = 4.08, = 0.047) and rapid eye movement sleep percentage (F(3,82) = 4.30, = 0.041), explained by greater rapid eye movement sleep in post-traumatic stress disorder females compared to control females, a difference not seen in male subjects. Quantitative electroencephalography analysis demonstrated that post-traumatic stress disorder was associated with lower energy in the delta spectrum (F(3,82) = 6.79, = 0.011) in non-rapid eye movement sleep. Slow wave sleep and delta findings were more pronounced in males. Removal of post-traumatic stress disorder subjects with comorbid major depressive disorder, who had greater post-traumatic stress disorder severity, strengthened delta effects but reduced rapid eye movement effects to non-significance. These findings support previous evidence that post-traumatic stress disorder is associated with impairment in the homeostatic function of sleep, especially in men with the disorder. These findings suggest that group × sex interaction effects on rapid eye movement may occur with more severe post-traumatic stress disorder or with post-traumatic stress disorder comorbid with major depressive disorder.

Introduction

Subjective sleep disturbance characterized by non-restorative sleep, frequent awakenings and nightmares related to traumatic life events are core features of post-traumatic stress disorder (PTSD). Despite the prominence of these symptoms in the distress experience of PTSD patients, the underlying neurophysiology of disturbed sleep in PTSD is poorly understood (Germain, 2013; Spoormaker and Montgomery, 2008). To date, research has produced conflicting findings regarding differences in objective measures of sleep in PTSD subjects and healthy controls. Furthermore, because most studies of sleep in PTSD have been conducted in male samples, very little is known about objective sleep disturbance in women with PTSD (Kobayashi et al., 2007).

A meta-analysis by Kobayashi et al. (2007) provides a synthesis of the available information on visually scored objective sleep disturbance in PTSD. Based on a review of 20 polysomnographic studies of PTSD+ and PTSD subjects conducted between 1966 and 2006, the authors determined that PTSD was associated with a statistically significant increase in Stage 1 sleep and a statistically significant decrease in slow wave sleep (SWS).

There is scant research on quantitative analysis of sleep electroencephalography (EEG) in PTSD and controls. However, a handful of studies have found that PTSD is associated with a decline in delta sleep, at least in male subjects (Neylan et al., 2003, 2006; Otte et al., 2007; Woodward et al., 2000). For example, in a sample of male PTSD subjects and age-matched controls studied under controlled laboratory conditions, Neylan and colleagues demonstrated a significant decrease in delta sleep in PTSD subjects compared to controls. Using similar methods, Otte and colleagues found a numerical but non-significant decrease in delta sleep in female PTSD subjects compared to controls (Otte et al., 2007). In contrast, a small in-home study examining delta sleep in PTSD in a mixed gender, but predominantly male, sample found increased delta sleep in PTSD (Germain et al., 2006). Because total slow wave (delta) activity is associated with the homeostatic recovery function for the organism and is demonstrated increasingly to have a pronounced role in glucose metabolism and other fundamental biological processes, deficiencies in delta sleep over the course of the night in PTSD subjects may have important health consequences (Scheen et al., 1996; Tasali et al., 2008).

Analyses of delta sleep in PTSD raise a complementary question: is PTSD associated with an increase in power in the higher frequency bands, which may be indicators of brain hyperarousal during sleep? The reports that have examined this question have not indicated pronounced effects of PTSD on rapid eye movement (REM), non-REM (NREM) or overall beta power (Cohen et al., 2013; Germain et al., 2006; Woodward et al., 2000). These reports indicate that the neurophysiological underpinnings of heightened subjective arousal during sleep in PTSD remains elusive, despite preliminary research examining this topic.

Existing findings, based on predominantly male samples, highlight that information on objective measures of sleep in female subjects with PTSD is severely lacking. In the study by Kobayashi and colleagues described previously, a comparison of weighted average effect sizes in male samples versus mixed-sex samples demonstrated that PTSD-associated decrements in objective sleep quality were more pronounced in males than in females (Kobayashi et al., 2007). These findings are consistent with Neylan and colleagues' findings regarding delta sleep: female subjects with PTSD demonstrated a numerically, but not statistically, significant decline in delta sleep compared to control subjects (Otte et al., 2007). The delta sleep finding was significant in males only (Neylan et al., 2003). Because sleep is considered increasingly to be central to the pathophysiology of PTSD, and women are at greater risk of PTSD than males even when controlling for trauma exposure, the examination of sleep in women with PTSD is essential.

Our current analyses take advantage of a 2 × 2 design (PTSD/control × female/male) including age- and sex-matched medically healthy, unmedicated young adults studied in a controlled laboratory environment to compare EEG measures of sleep across groups and between sexes. In this study, three nights of polysomnography with nocturnal blood sampling (night 1 =  adaptation, nights 2–3 = pre- and post-metyrapone administration) were conducted at the University of California, San Francisco. This report focuses on the night 2 data during which our subjects were medication-free. Our primary hypothesis, confirming previous research, was that PTSD is associated with greater visually scored slow wave sleep and cumulative NREM delta power (i.e. delta energy), and that this finding would be most pronounced in males. We also predicted that PTSD would be associated with more Stage 1 sleep. Exploratory analyses examined overnight power in all frequency bands in both NREM and REM sleep, although based on recent findings we did not predict that PTSD would be associated with higher power in the higher frequency bands. Additionally, exploratory analyses of spectral power were performed to test for group (PTSD versus control) × sex interactions.

Methods

The study sample was comprised of 40 individuals with current chronic PTSD [53% female; mean age 30.63, standard deviation (SD) = 6.63] and 43 control subjects (49% female; mean age 30.39, SD = 8.15). This sample was drawn from a larger sample of 93 participants. Data from 10 participants were excluded due to poor quality sleep EEG recordings. Chronic PTSD was defined by fulfilment of DSM-IV criteria for chronic PTSD on the clinician-administered PTSD Scale (CAPS) and a CAPS score >40. Control subjects had no lifetime or current history of a PTSD diagnosis. The sleep of female subjects was measured during the follicular phase of the menstrual cycle. Exclusion criteria included history of traumatic brain injury, presence of neurological disorders or systemic illness; use of psychiatric, anticonvulsant, antihypertensive or sympathomimetic, steroidal, statin or other prescription medications; obesity [defined as body mass index (BMI) >30]; alcohol abuse or dependence in the previous 2 years; substance abuse or dependence in the previous year; any psychiatric disorder with psychotic features; bipolar disorder or obsessive compulsive disorder; and pregnancy. Exclusion criteria for control subjects included a lifetime history of major depressive disorder or panic disorder. This research was approved by the Committee on Human Research at the University of California. All participants provided written informed consent before participating in any study procedures.

Measures

The CAPS was used to assess current and lifetime PTSD. The CAPS assesses the frequency and intensity of PTSD symptoms corresponding to the re-experiencing, avoidance and hyperarousal symptoms described in the DSM-IV diagnostic criteria (Blake et al., 1995). Diagnosis of PTSD was based on symptoms experienced in the previous month which were associated with the participant's self-identified worst traumatic event.

The structured Clinical Interview for DSM-III-R (SCID) was used to diagnose all other psychiatric disorders, including major depressive disorder (MDD) (Spitzer et al., 1992).

Two self-report instruments, the Pittsburgh Sleep Quality Index (PSQI) and the PSQI-addendum for PTSD were used to assess subjective sleep quality (Buysse et al., 1989; Germain et al., 2005). The PSQI-addendum assesses the frequency of seven disruptive nocturnal behaviours. A score of 4 was shown to yield a sensitivity of 94%, specificity of 82% and positive predictive value of 93% for discriminating those with PTSD from those without PTSD (Germain et al., 2005).

Polysomnographic measurement

Polysomnography recordings were obtained with ambulatory polysomnography (Nihon Kohden Trackit Ambulatory Recording System, Foothill Ranch, CA, USA). The parameters recorded included an electroencephalogram (EEG) at leads C3, C4, O1 and O2, left and right electro-oculograms (EOG), submental electromyogram (EMG), bilateral anterior tibialis EMGs and electrocardiogram (ECG) in accordance with standardized guidelines (Rechtschaffen and Kales, 1968). Electrode impedance was set at <5 kΩ at the start of the recording. The EEG and EOG leads were referenced to linked mastoids. Raw EEG signals were filtered and amplified, then digitized at 256 Hz and recorded on a removable hard disk in European Data Format (EDF) file format. The low- and high-frequency hardware filters on the recorder were single pole analogue filters with 3 db points at 0.5 and 100 Hz. Pass Plus was utilized for both visual scoring and quantitative EEG analysis of the digitized polysomnography data.

Visual sleep scoring

Visual scoring was conducted by one of the authors, a highly experienced registered polysomnography technician, who classified all 30-s epochs in every sleep record as wake, Stages 1, 2, 3, REM or movement using current American Association of Sleep Medicine (AASM) criteria. Sleep onset was defined as the first minute of 8 consecutive minutes of Stage 2 sleep with no more than two intervening minutes of Stage 1 sleep or minutes awake. Total sleep time was defined by time spent in epochs scored as NREM Stages 1–3 and stage REM. Wake after sleep onset (WASO) was defined as the time spent in epochs scored as wake between sleep onset and final awakening. An awakening was defined by EEG arousals lasting 15 s or longer. We were unable to report sleep latency because we did not acquire time of ‘lights out’.

Power spectral analysis

Pass Plus (Delta Software, University City, MO, USA) analytical software was used to measure sleep activity in all frequency bands delta through gamma from the C3 electrode by power spectral analysis (PSA). The C4 electrode was used if there was excessive artefact. A limitation of Pass Plus is that artefact removal is accomplished by removal of whole epochs tagged with artefact. This has the potential to introduce additional confounds, given the removal of typically longer bouts of uncontaminated EEG. Therefore, epochs were tagged for slow and fast artefact for additional analyses. Primary analyses were conducted with all epochs and then checked for the impact of removal of epochs with slow and fast artefact. Removal of fast artefact (for bandwidths alpha and above) and slow artefact (for bandwidths delta and theta) did not impact our findings significantly in NREM sleep. In REM sleep, artefact removal altered our findings in only one respect: a statistically significant group × sex interaction on REM delta energy emerged, driven by lower REM delta energy in PTSD+ males compared to control males with the absence of such a difference in females. Because of the small amount of delta activity in REM sleep and because the artefact-removal mechanism of Pass Plus removes entire epochs tagged with artefact, we were not surprised that removal of slow artefact had a statistically significant effect on delta findings in REM sleep and are concerned that it yielded a spurious finding. All results are therefore reported without removal of epochs containing artefact. Pass Plus applied a 5 μV smoothing constant to eliminate spurious waves caused by electrical jitter. PSA was conducted on all epochs of NREM and REM sleep. Epochs scored as wake were not included in these analyses. Pass Plus was used to perform Fast-fourier transformation analysis on 4.0-s Welch tapered windows with 2-s overlap, yielding 15 windows per 30-s epoch. Power spectra for Delta (1–4 Hz) were analysed to address our primary hypothesis with respect to delta sleep. Theta (4–8 Hz), alpha (8–12 Hz), sigma (12–15 Hz), beta1 (15–23 Hz), beta2 (23–30 Hz) and gamma (30–50 Hz) bands were analysed for secondary analyses.

Statistical analysis

Data screening procedures demonstrated that all visually scored sleep parameters, with the exception of WASO, were distributed normally. The WASO variable was natural log-transformed for analysis of variance (anova). Distributions of quantitative EEG measures were all right-skewed. These variables were natural log-transformed, resulting in normalization of their distributions. Our primary hypotheses regarding SWS, Stage 1 sleep and total delta energy (μV2 s) were tested using anova. Given the known associations between MDD and our primary outcome variable, delta energy, as well as recent findings from our laboratory indicating that two or more categories of childhood trauma accounted for the effects of PTSD on leucocyte telomere length, a marker of biological aging (O'Donovan et al., 2011), we also examined the effects of MDD and childhood trauma history on delta energy.

Results

Demographic data and clinical characteristics

The demographic and clinical characteristics of the sample are presented in Table 1. There were no significant differences in sex distribution between PTSD and control subjects, nor were there significant differences in age or education across all four groups. Male and female PTSD subjects did not differ in terms of CAPS scores, rates of current major depressive disorder or history of childhood trauma defined by the presence of two or more categories of childhood trauma compared to one or none. Thirteen control subjects reported a lifetime history of a criterion A1 event, but all had current CAPS scores of zero and none had a lifetime history of PTSD. As per the exclusion criteria, no control subjects met criteria for current MDD. Additionally, none of the control subjects reported a history of two or more categories of childhood trauma.

Table 1. Demographic and clinical characteristics of the sample
 Female (= 42)Male (= 41)Contrast math formula =0.112 = 0.74
ControlPTSD+ControlPTSD+
= 21= 21= 22= 19
  1. SD, standard deviation.

  2. a

    These three subjects endorsed Hispanic ethnicity but did not select a racial descriptor. Seven additional subjects endorsed Hispanic ethnicity, in addition to a racial category of Caucasian or African American race yielding a total of 10 subjects self-identifying as Hispanic in this sample.

  3. b

    Control subjects had clinician-administered post-traumatic stress disorder (PTSD) Scale (CAPS) scores of zero or had an absence of criterion A events; t-test compares mean CAPS score between male and female PTSD subjects only.

  4. c

    Absence of current major depressive disorder (MDD) was required for inclusion into the control group. Fisher's exact test compared MDD frequency between male and female PTSD subjects only.

  5. d

    Childhood trauma exposure was defined, based on findings from our prior research, by exposure to two or more categories of childhood trauma. Three control subjects reported a history of one category of childhood trauma. Chi-square test compared frequency of childhood trauma between male and female PTSD subjects only.

Age (mean, SD)30.57 (7.67)30.10 (6.96)30.23 (8.76)31.21 (6.39)

F(3,82)= 0.09

= 0.97

Education (years) (mean, SD)15.6 (2.03)15.6 (1.68)15.5 (2.08)14.6 (2.40)

F(3, 80) =0.95

P = 0.42

Race/ethnicity    

math formula = 17.56

= 0.13

Caucasian (n)16131711 
African American (n)0213 
Asian/Asian American (n)3240 
Other (n)1403 
Hispanic, race unknown (n) a1002 
Clinical characteristics
CAPS (mean, SD)b58.86 (17.32)52.79 (13.81)

t(38) = −1.22

= 0.23

Current MDD, n (%)c03 (14)05 (26)

Fisher's exact

= 0.44

Childhood trauma, n (%)d08 (42)010 (50)

math formula = 0.24

= 0.62

Subjective sleep quality

There were significant differences in subjective sleep reports across groups (see Table 2). There was a main effect of group on total PSQI score, with both male and female PTSD subjects demonstrating significantly higher scores than control subjects. PTSD subjects also scored significantly higher than control subjects on the PSQI-addendum. There was also a main effect for gender, with women reporting more PTSD-related sleep disturbance on the PSQI-addendum than their male counterparts.

Table 2. Subjective sleep scores and visually scored sleep parameters by sex and group [post-traumatic stress disorder (PTSD)+ versus controls]a
 FemaleMaleF-test group effectF-test sex effectF-test group × sex effect
ControlPTSD+ControlPTSD+
Mean (SD)Mean (SD)Mean (SD)Mean (SD)F (P-value)F (P-value)F (P-value)
  1. SWS, slow wave sleep; PSQI, pittsburgh sleep quality index; SD, standard deviation.

  2. a

    Bold value indicates statistically significant effects.

  3. b

    T-test indicates a statistically significant difference between PTSD+ females and control females.

  4. c

    T-test indicates statistically significant difference between PTSD+ males and control males.

  5. d

    T-test indicates statistically significant difference between PTSD+ females and PTSD+ males.

  6. e

    T-test indicates statistically significant difference between control females and control males.

  7. f

    Analysis of variance (anova) was performed on log-transformed wake after sleep onset (WASO) due to right-skewed distribution.

PSQI2.64 (1.45)b11.01 (3.04)b2.77 (1.72)c10.26 (3.46)c 203.80 (<0.0001) 0.30 (0.584)0.63 (0.431)
PSQI-addendum0.57 (0.81)b7.57 (2.64)bd0.32 (0.65)c5.53 (3.41)cd 164.69 (<0.0001) 5.84 (0.018) 3.55 (0.063)
TST (min)431.10 (53.03)427.17 (60.54)d403.66 (54.96)374.61 (49.64)d1.87 (0.175) 11.02 (0.001) 1.09 (0.300)
Stage 1 (min)16.57 (8.67)17.30 (9.12)14.18 (8.34)15.52 (8.55)0.31 (0.581)1.18 (0.280)0.03 (0.869)
Stage 1 (%)3.87 (2.07)4.12 (2.21)3.58 (2.23)4.18 (2.36)0.78 (0.380)0.06 (0.810)0.13 (0.719)
Stage 2 (min)258.88 (53.68)254.42 (53.83)236.80 (42.39)238.21 (32.74)3.47 (0.066)0.02 (0.883)0.08 (0.776)
Stage 2 (%)59.76 (7.63)59.62 (9.88)58.68 (7.16)c63.99 (7.90)c2.05 (0.156)0.83 (0.364)2.28 (0.135)
SWS (min)64.40 (29.17)48.67 (33.76)52.70 (25.02)c33.08 (27.94)c 7.63 (0.007) 4.54 (0.036) 0.09 (0.762)
SWS (%)15.38 (7.27)11.68 (8.34)13.21 (6.41)c8.91 (1.69)c 6.11 (0.016) 2.33 (0.131)0.04 (0.852)
REM (min)91.24 (26.33)106.76 (33.85)99.98 (29.10)87.76 (35.43)0.06 (0.810)0.56 (0.458) 4.08 (0.047)
REM (%)20.99 (4.73)be24.57 (5.97)b24.53 (5.67)e22.92 (6.38)0.62 (0.433)0.57 (0.454) 4.30 (0.041)
WASO (min) f55.81 (35.53)65.81 (53.52)54.02 (53.79)71.26 (48.28)1.66 (0.202)0.20 (0.656)0.83 (0.364)
Sleep maintenance0.89 (0.01)0.87 (0.02)0.89 (0.02)0.85 (0.02)2.36 (0.129)0.23 (0.634)0.27 (0.604)
Number of awakenings12.86 (4.86)15.29 (4.90)12.41 (5.21)13.89 (4.65)3.27 (0.074)0.72 (0.400)0.19 (0.664)

Visually scored sleep measures

Findings from the analysis of visually scored sleep EEG are presented in Table 2. Consistent with our hypothesis, PTSD subjects demonstrated significantly lower SWS duration compared to control subjects (F(3,82) = 7.63, = 0.007). The proportion of SWS to TST was also significantly smaller in PTSD subjects as a group compared to controls (F(3,82) = 6.11, = 0.016). Pairwise comparisons looking at males and females separately demonstrate that differences between PTSD subjects and controls were more pronounced in males than females. Contrary to expectations, there was no group effect for Stage 1 sleep. There were also no significant main or interaction effects of group and sex on Stage 2 sleep or WASO. Females displayed significantly greater total sleep time compared to males, an effect that was accounted for primarily by a statistically significant difference in TST between PTSD females and PTSD males (F(3,82) = 11.02, = 0.001). Finally, our analyses revealed a statistically significant group × sex interaction effect on REM sleep, both in terms of absolute duration of REM sleep and percentage of REM sleep. This was explained by a difference in direction of the difference between PTSD subjects and controls by sex: females with PTSD demonstrated statistically significantly greater REM sleep compared to controls, whereas males with PTSD displayed non-statistically significantly lower REM sleep compared to controls.

Power spectral analysis of EEG

Findings from quantitative analysis of sleep EEG are presented in Tables 3 (NREM sleep) and 4 (REM sleep). Values represent absolute spectral energy, which is cumulative spectral power in NREM and REM sleep, respectively, over the duration of the sleep period. As hypothesized, there was a significant effect of PTSD on NREM delta activity (i.e. delta energy) (see Table 3, F(3,82) = 6.79, = 0.011). There were no other significant main effects of PTSD on NREM energy in the higher frequency bands. We found main effects for sex on NREM energy in all frequency bands delta to beta1. When NREM sleep time was taken into account (i.e. when absolute power, as opposed to energy, was examined), gender effects disappeared and the main group effect for delta remained nearly significant (= 0.071).

Table 3. Quantitative electroencephalograph (EEG) sleep measures by sex and group (PTSD+ versus controls): NREM sleepa
 FemaleMaleF-test group effectF-test sex effectF-test group × sex effect
ControlPTSD+ControlPTSD+
Mean (SD)Mean (SD)Mean (SD)Men (SD)F (P-value)F (P-value)F (P-value)
  1. SD, standard deviation.

  2. a

    Units of energy [absolute spectral power over the duration of non-rapid eye movement (NREM) sleep] are μV2 s per 100. Raw energy values must be multiplied by 100 to reflect actual energy and are shown as presented by Pass Plus software to enhance readability. Bold value indicates statistically significant effects.

  3. b

    T-test indicates statistically significant difference between post-traumatic stress disorder PTSD+ females and PTSD+ males.

  4. c

    T-test indicates statistically significant difference between PTSD+ males and control males.

  5. d

    Group effect for delta becomes a trend (= 0.071) when NREM time is controlled for in regression analysis.

  6. e

    Gender effects disappear when NREM time is controlled for in regression analyses.

Delta (1–4 Hz)38 948 (18 877)31 049 (15 929)30 198 (13 733)22 799 (10 423)   
Log delta  b c b,c 6.79 (0.011) d 7.60 (0.007) e 0.15 (0.697)
Theta (4–8 Hz)7839 (4168)6833 (2463)5822 (2132)5427 (2973)   
Log theta  b   b 1.38 (0.244) 7.97 (0.006) e 0.11 (0.740)
Alpha (8–12 Hz)3648 (1699)3300 (1531)2793 (1226)2383 (1183)   
Log alpha  b   b 1.92 (0.170) 8.97 (0.004) e 0.15 (0.701)
Sigma (12–15 Hz)1894 (1119)2069 (1757)1799 (853)1254 (765)   
Log sigma  b c b , c 2.88 (0.094)3.90 (0.052)e3.04 (0.085)
Beta1 (15–23 Hz)911 (323)947 (328)815 (325)797 (366)   
Log beta1    0.01 (0.920) 4.63 (0.035) e 0.29 (0.592)
Beta2 (23–30 Hz)363 (121)374 (132)373 (240)337 (169)   
Log beta2    0.03 (0.873)1.25 (0.267)0.08 (0.784)
Gamma (30–50 Hz)574 (191)552 (282)633 (482)548 (370)   
Log gamma    0.45 (0.504)0.19 (0.666)0.14 (0.712)
Table 4. Quantitative electroencephalograph (EEG) sleep measures by sex and group (PTSD+ versus controls): REM sleepa
 FemaleMaleF-test group effectF-test sex effectF-test group × sex effect
ControlPTSD+ControlPTSD+
Mean (SD)Men (SD)Mean (SD)Mean (SD)F (P-value)F (P-value)F (P-value)
  1. SD, standard deviation.

  2. a

    Units of energy [absolute spectral power over the duration of rapid eye movement (REM) sleep] are μV2 s per 100. Raw energy values must be multiplied by 100 to reflect actual energy and are shown as presented by Pass Plus software to enhance readability.

  3. b

    T-test indicates statistically significant difference between post-traumatic stress disorder PTSD+ females and PTSD+ males.

  4. c

    T-test indicates statistically significant difference between PTSD+ males and control males.

Delta (1–4 Hz)2655 (1847)3358 (2249)5174 (11 884)2439 (2391)   
Log delta    0.38 (0.541)0.56 (0.456)3.13 (0.081)
Theta (4–8 Hz)1208 (658)1149 (461)1981 (4557)805 (627)   
Log theta  b c b , c 2.75 (0.101)2.39 (0.126)2.80 (0.098)
Alpha (8–12 Hz)528 (321)605 (321)497 (267)417 (300)   
Log alpha  b   b 0.38 (0.539)2.15 (0.147)2.84 (0.096)
Sigma (12–15 Hz)162 (79)205 (107)183 (92)170 (155)   
Log sigma    0.03 (0.866)0.09 (0.767)3.17 (0.079)
Beta1 (15–23 Hz)267 (150)351 (184)312 (187)290 (230)   
Log beta1    0.16 (0.688)0.00 (0.997)2.88 (0.094)
Beta2 (23–30 Hz)130 (86)184 (136)138 (97)136 (110)   
Log beta2    0.44 (0.509)0.16 (0.693)1.77 (0.188)
Gamma (30–50 Hz)189 (161)260 (235)213 (181)170 (145)   
Log gamma    0.06 (0.814)0.04 (0.842)1.91 (0.171)

The role of depression and childhood trauma

There was no significant correlation between current major depression and history of childhood trauma (defined as 2 or more versus 1 or no categories of childhood trauma) and delta energy in bivariate analysis, therefore these variables were not included in further analyses. Post-hoc analyses examining delta and REM effects after removal of the five males and three females with current major depression resulted in a strengthening of the NREM delta effect (= 7.80; = 0.0067) and a weakening of the group × gender interaction effect on REM duration (= 1.90; = 0.172) and REM percentage (= 2.17, = 0.146).

Relationship of objective sleep variables to CAPS total and subscale scores

Among PTSD subjects, there was no significant correlation between NREM delta energy and either total or subscale scores on the CAPS. Post-hoc bivariate correlations examining the relationship between REM sleep time and CAPS total and subscale scores revealed no significant correlations. Post-hoc correlations between REM sleep time and CAPS total and subscale scores within sexes were not statistically significant.

Discussion

These data provide additional evidence that PTSD is associated with a decrease in visually and quantitatively measured slow wave sleep. Quantitative sleep EEG research is providing strong evidence that delta sleep is essential for a variety of biological functions, including cognitive performance and glucose metabolism (Scheen et al., 1996; Tasali et al., 2008). Existing research also indicates that individuals with PTSD are at greater risk of for multiple medical problems such as cardiovascular disease, obesity, diabetes and dementia, which raises the possibility that disrupted delta sleep plays a mechanistic role in producing these negative health outcomes (Boscarino, 2004; Kobayashi et al., 2007; Yaffe et al., 2010). Consistent with our laboratory's previous work, these data also indicate that this effect is more robust in men compared to women. This raises unexamined questions about sex differences in the health and functional consequences of PTSD.

Contrary to our hypothesis, we did not identify group differences in Stage 1 sleep although, in males, we found an increase in Stage 2 sleep in PTSD subjects compared to controls. All subjects in this study slept with intravenous catheters on all three nights of the overall study, which may have increased light sleep in all subjects. However, Stage 1 sleep was not elevated in the sample and there were no significant differences across groups in Stage 1 sleep, in sleep maintenance or in number of awakenings across groups, all of which provide indirect evidence that sleep was not disrupted excessively by catheterization.

Consistent with recent findings, these data provide further support that current sleep EEG methods have not identified the neurophysiological correlates of subjective hyperarousal in PTSD. Our data indicate no statistically significant group differences in all-night energy in the higher frequency bands, including beta bands, in either REM or NREM, although group differences in REM theta approached significance. Several alternatives may account for this finding. Previous research has demonstrated increased hyperarousal in PTSD in the context of trauma cues, but less consistently in their absence (Orr et al., 2002; Pitman et al., 1999). It may be that PTSD subjects differ from controls only in the context of trauma reminders. Alternatively, current EEG approaches may be too coarse to identify effectively the neurophysiological correlates of central arousal in PTSD. More refined approaches, including topographical approaches that take a finer look at regional brain activity, may be an important next step in identifying regional foci of brain arousal in sleep in PTSD (Marzano et al., 2008).

The identification of a group × sex interaction for REM sleep raises intriguing questions about the role of REM sleep in PTSD and the potential sex differences in that role. Previous studies were not designed to have a balanced sample of healthy age-matched men and women that would allow detection of group × sex interactions. Previous studies have not found differences in REM duration between PTSD and control subjects, although several studies have suggested that REM sleep disruption may be correlated with PTSD (Breslau et al., 2004; Habukawa et al., 2007; Mellman et al., 2007). The latter findings have led some researchers to speculate that REM sleep disruption in PTSD may reflect a disturbance in the processing of traumatic experiences (Habukawa et al., 2007; Mellman et al., 2007). Conversely, increased REM sleep has also been shown to be correlated with greater consolidation of emotionally salient memories, the mechanisms and benefits of which, in PTSD and healthy subjects, remain unclear (Baran et al., 2012; Nishida et al., 2009). Our findings indicate that sex differences should be considered in studies that examine REM sleep-dependent ‘processing’ of emotionally salient events.

Adding complexity to this group × sex interaction on REM duration and REM percentage, we found that removal of subjects with current MDD in post-hoc exploratory analyses reduced the effects to non-significance. This was in contrast to the effects on NREM delta energy, which were strengthened. Existing research suggests strongly that MDD is associated with decreased REM latency and increased REM density (Palagini et al., 2013; Pillai et al., 2011) and some studies provide evidence for increased REM duration in MDD (Palagini et al., 2013). However, existing research does not demonstrate a group × sex interaction effect on REM duration in MDD such as was found in our study. There are therefore no a priori reasons to expect MDD, per se, to account for this reduction in effect. Given the high degree of symptom overlap between PTSD and MDD, and evidence that similar biological abnormalities, such as increased release of corticotrophin-releasing factor, may contribute to sleep disturbance in both disorders, it is possible that overlapping features of MDD and PTSD contribute highly to the group × sex interaction we found. It is also possible that the REM interaction effect is highly sensitive to loss of severe PTSD cases, as may have been the case in this sample, and our previous research indicates strongly that comorbid MDD is more common as CAPS scores increase. In this sample, specifically, mean CAPS score in MDD-negative subjects was 53.9, while mean CAPS score in MDD-positive subjects was 64.3 (= 0.099). There is increasing evidence that sex differences need to be considered in neuroscience research (Cahill, 2006) and our findings highlight that further biological research on sex differences is crucial to better understand the unique and shared relationships of traumatic stress and depressive symptoms to sleep disturbance and psychopathology.

Despite the many strengths inherent in this study's design, several limitations have to be considered. As stated earlier, all participants were catheterized during sleep laboratory admissions. Another limitation of this study is that our findings are based on data from a single night of polysomnography recordings. Finally, generalization of our results to the overall PTSD population is limited by the nature of the sample, consisting of healthy, non-medicated subjects. However, we consider these sample features to be a considerable strength of this study, because it enables us to gain information about sleep in men and women with PTSD without the potential confounding effects of medications and other factors potentially affecting sleep biology.

In conclusion, the current findings support previous evidence that PTSD is associated with impairment in the homeostatic function of sleep, especially in men with the disorder. Interaction effects of sex and PTSD status on REM sleep raise intriguing questions about the role of REM sleep in sexual dimorphism in PTSD pathophysiology. Replication of our findings with balanced samples of men and women are needed to understand the specific role of REM sleep in men and women with PTSD.

Acknowledgements

This project was supported by grants from the National Institute for Mental Health (TCN: 5R01MH073978-04, 5R34MH077667-03), the Mental Illness Research and Education Clinical Center of the US Veterans Health Administration and the Clinical Research Center of the National Center for Advancing Translational Sciences, National Institutes of Health, through UCSF-CTSI Grant Number UL1 RR024131. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH. This material is the result of work supported with resources and the use of facilities at the Veterans Administration Medical Center, San Francisco, California. We are grateful to Ian Campbell PhD for his scientific input and to Maryann Lenoci for her logistical and technical support.

Conflict of Interest

None of the authors have any financial interests or conflicts of interest to report.

Ancillary