(a, b) Plot proportion of sleep stages by time for 90 min leading to and 30 min following a verified arousal from sleep. Blue line corresponds to non-REM sleep, red to REM sleep, and green to waking. The median clock time of plotted arousals is also indicated. Data aggregated over 140 arousals from 33 participants. Note marked increase in REM sleep leading up to arousals associated dream mentation versus those without
(a–c) Plot proportion of sleep stages by time leading to and following a verified arousal from sleep that included a dream report. Color coding as before. Data aggregated over 67 arousals. Note marked increase in REM sleep leading up to arousals associated with more extensive dream mentation
Plot proportion of sleep stages by time leading to and following arousals attributed to external stimuli. Color coding as before. Data aggregated over 14 arousals. Note marked increase in waking prior to these arousals (with the exception of the immediately preceding 15 min per criteria)
(a–c) Plot proportion of sleep stages by time leading to and following a verified arousal from sleep categorized according to absence/presence/severity of distressing mentation. Color coding as before. Data aggregated over 130 arousals. (Ten arousals were not clearly categorizable on this dimension.) Note increase in REM sleep leading up to arousals associated with more distressing dream mentation
Plot proportion of sleep stages by time leading to and following arousals attributed to an interoceptive stimulus (bladder distention). Color coding as before. Data aggregated over 38 arousals. Note predominance of non-REM sleep prior to these arousals
A second set of analyses compared participants’ sleep HRs over symptomatic versus non-symptomatic arousals (See Fig. 20.6). Surprisingly, HRs tended to be lower prior to symptomatic arousals than non-symptomatic arousals; however, when aggregated over the full 90-min pre-arousal epoch, this difference was not statistically significant (t(12) = 1.07, p = 0.305). This result was not modified by adjustment for the circadian influence on heart rate. (This adjustment was performed by regressing all pre-arousal HRs against arousal clock time and then residualizing each individual pre-arousal HR by reference to the HR predicted for that time by the model. The linear regression model was highly significant, F(1138) = 32.6, p < 0.001, with pre-arousal HR lowering at a rate of 2.18 BPM per hour of sleep.) Not surprisingly, HRs were higher after symptomatic than after non-symptomatic arousals; however, this difference was again not significant (t(12) = 1.07, p = 0.305). The interaction of arousal type and time of measurement (pre- vs post-) was statistically significant (F(1,12) = 8.12, p = 0.015) as was the main effect of time (F(1,12) = 5.26, p = 0.041) with post-arousal HRs being higher than pre-arousal.
Plot mean HR by time leading to and following arousals associated with or without distressing mentation. Data are from 13 participants who provided examples of both types along with adequate sleep ECG. Within-subject within-category data were averaged and then quantified for statistical analysis (see text). Red and blue lines plot mean HRs associated with symptomatic and non-symptomatic, respectively
The sleep preceding different classes of nocturnal arousals in this small mixed sample of persons with PTSD manifested a familiar relationship between the presence and volume of reportable dream mentation and the preponderance of REM sleep leading up to the arousal. An association between longer dream reports and a preponderance of REM versus other stages of sleep was recently reconfirmed in a large naturalistically acquired sample by Stickgold et al. . Increasingly distressful dream content was also associated with increasing proportions of REM sleep. Again, it is noteworthy that though the most disturbing arousals containing life-threat content exhibited the highest preponderance of REM sleep , a minority of such arousals were preceded by non-REM sleep. This latter observation supports the conclusion of Wittman et al.  that trauma-related nightmares are not exclusive to REM sleep. Nocturnal arousals associated with external stimuli or interoceptive alarms were not preceded by increased proportions of REM sleep. Instead, arousals attributed to external stimuli were preceded by increased waking, while those attributed to an interoceptive alarm were preceded by relatively high amounts of NREM sleep. It seems entirely plausible that reports of external triggering of arousal would be associated with sleep interrupted by waking. As well, the relative paucity of REM sleep leading up to interoceptive alarms is compatible with the fact that the nucleus locus coeruleus (LC), which mediates such internal alarms, is quiescent during REM [16, 17]. This is because interoceptive alarms such as bladder distention are routed through the LC  and so should be gated by LC quiescence.
The results of this study are limited by the small size of the corpus of criterial mid-sleep arousals. In particular, this prevented comparisons of symptomatic versus non-symptomatic arousals within subjects and within sleep stages, especially within REM-preceded arousals, of objective indices of EEG spectral composition c.f.  and EOG activity c.f [20, 21]. Nevertheless, it is noteworthy that no prodromal elevation of sleep HR was observed in advance of symptomatic versus non-symptomatic arousals evaluated within subjects given the relatively larger HR excursions following the former. Adjusting for circadian effects on HR did not modify this result. This observation contrasts with the earlier reports of Fisher et al.  suggesting that nightmares are anticipated by autonomic arousal and instability. (It is unknown if Fisher’s subjects met criteria for PTSD as the diagnosis had not yet been incorporated into psychiatric nosology). Any effort to understand this result in light of our knowledge of waking intrusive or reexperiencing symptoms confronts our lack of knowledge regarding their autonomic correlates, as well. While ambulatory HR has been recorded in persons with PTSD [22–26] with reference to certain discrete events such as cigarettes smoked and episodes of hostility, this work has not extended to examinations of the psychophysiological concomitants of phasic symptoms such as intrusive thoughts, flashbacks, and trauma-cue-elicited distress. Hence, it is altogether unknown whether such events are preceded by prodromal autonomic arousals , as has recently been found for panic attacks .
The absence of autonomic activation prior to symptomatic arousals in PTSD is reminiscent of another finding from our laboratory that self-reported nightmare symptom severity is associated with reduced all-night movement time in Vietnam combat veterans with PTSD studied in the laboratory . While “thrashing movements during sleep”  are reported by veterans with PTSD, objective verification of such episodes and their association with nightmares has been limited. Evidence of autonomic and motoric quiescence in association with nightmares has prompted us to revisit a basic question: how do other animals living in dangerous environments, or with histories of exposure to acute stress, sleep? If PTSD is broadly understood as a “failure of recovery” from a state adapted to conditions of extreme threat , how might nightmares function as a component of such adaptation?
Humans are not unique in their need to achieve sleep despite the current or recent experience of life threat. The impacts of acute and chronic stress on sleep in laboratory species have been covered in Chap. 12. Here, we will consider ethological investigations of animals sleeping in the wild, work that includes the seminal studies of Zepelin and Rechtschaffen  and Allison and Ciccetti , the fieldwork of Anderson , and a recent series of papers by Lima, Lesku, and colleagues [34–38] (see also ). First and foremost, a critical finding across these studies is that sleep in mammals and birds is sensitive to predation threat. That sensitivity is expressed through two main strategies, first, concealment  and, second, the preservation of arousability through the de-emphasis of sleep states associated with elevated arousal thresholds, slow wave sleep (SWS), and REM sleep. Into this broad framework, aspects of the sleep of humans unrecovered from traumatization already fit fairly neatly. The Kobayashi et al.  meta-analysis concluded that sleep in PTSD is associated with both an increase in stage N1 sleep and a reduction in SWS (or stage N3), adaptations which would result in lower arousal thresholds throughout the night. To this complex may be added the impact of nightmares, per se, which are, by definition, dreams that awaken the sleeper. It is also noteworthy that the DSM-V (307.47) criteria for primary nightmares includes the following language, “On waking from the nightmare, the person becomes oriented and alert” . Rapid orientation distinguishes nightmares from sleep terrors which are typically associated with impaired consciousness and impaired memory for the event. If nightmare-related awakenings are also behaviorally subtle and devoid of autonomic activation, this would be in line with the conserved strategy of concealment. While concealment during sleep is not a concept with which modern humans are familiar, it is worth considering that hominins ground-slept without benefit of protective shelter, within close range of large nocturnal predators, from at least 2.6 MYA up to approximately 380,000 years ago. (This period is bracketed by the emergence of H. erectus, who was devoid of any arboreal skeletal adaptations, and the Terra Amata Shelter in southern France, among the earliest sites with evidence of shelter construction, probably built by H. heidelbergensis.)
An alternative framework within which formulates this discussion has been variously referred to as the predator imminence or defensive cascade model [42, 43]. In the terminology of this framework, our strategy was to look for evidence of “circa-strike” psychophysiology during sleep in persons with trauma-related nightmares. In fact, our own data may be more suggestive of “post-encounter” psychophysiology. While highly contrastive behaviorally, both circa-strike (aka fight/flight) and post-encounter defensive modes are orchestrated by the amygdala operating through specific effector subsystems, especially, in the latter case, the bed nucleus of the stria terminalis and the periaqueductal gray . The post-encounter pattern combines “freezing ” or “movement suppression ”  with focused attention directed toward the threat. The presumed goal of post-encounter behavior is the avoidance of detection by a predator combined with sustained vigilance to the latter’s location, movement, gaze, etc. The freezing component has a long history of use as an index of fear in the animal literature but has been little studied in humans because it has, until recently, been difficult to operationalize in the laboratory [46–49]. Importantly for the model we are proposing, freezing is associated, at least acutely, with a HR deceleration sometimes termed “fear bradycardia ” [45, 50] instead of the phasic and/or chronic HR accelerations we expected to observe during sleep in PTSD.
Implicit in the above discussion is the proposition that the pre-encounter mode of defensive behavior, in contrast to circa-strike, is somehow compatible with sleep. A recent study by Cano et al.  suggests that such a proposition is not far-fetched. In that study, the investigators devised an experimental paradigm that resulted in simultaneous co-activation of sleep-promoting and arousal-promoting subsystems in the brainstem, midbrain, and forebrain. The paradigm involved placing a male rat in a cage previously occupied by another male with no intervening cleaning/deodorization. By instituting a large number of experimental controls and surveying a wide range of brain regions using the intermediate-early gene c-Fos to index neuronal activation, these investigators documented, in the cage-exchange condition, concurrent activation of sleep-promoting regions such as the median and ventrolateral preoptic nuclei of the hypothalamus and arousal-promoting regions such as the locus coeruleus (LC), tuberomammillary nucleus (TMN), and extended amygdala . In some sleep-promoting regions, cage-exchange animals exhibited substantially more activation than controls. Nevertheless, in these animals, cerebral cortex also exhibited increased c-Fos induction, especially the infralimbic and anterior cingulate cortices, regions of the rat brain thought to correspond, respectively, to subgenual and dorsal anterior cingulate cortices in humans , both implicated in PTSD [53, 54]. Cano and colleagues aimed to provide a model of “stress-induced insomnia ” rather than of PTSD-related sleep disturbance. Nevertheless, viewed from the perspective of defensive adaptation, the Cano et al. model demonstrates that downgraded versions of sleep and arousal can coexist in the rat brain, perhaps representing a “solution” to the conundrum faced by all animals sleeping under threat, the simultaneous requirements to sleep and remain vigilant.
Perhaps the most descriptive term applied to the post-encounter state is “attentive immobility ” . In the rat, attentive immobility has been associated with elevated gamma-band EEG similar to both active wake and REM sleep . Gamma-band EEG is normally much lower in amplitude in NREM; however, Cano et al. found amplification of gamma-band EEG in NREM sleep in cage-exchange rats versus controls, an effect which could be reversed either by lesion of LC (by ibotenic acid) or inhibition of the TMN (by the H3 autoreceptor agonist immepip). Thus, in their preparation, NREM gamma-band EEG appeared to be a specific index of arousal-promoting systems co-activated during sleep.
The Cano et al. findings suggest that gamma-band EEG might provide positive signs of PTSD-related sleep adaptations where sleep HR has not. A direct translation to PTSD-related insomnia would predict the observation of elevated gamma in NREM or SWS. In normals, gamma-band EEG has also been reported elevated in REM sleep, especially when rapid eye movements are present . A possible association with REM density is attractive, here, as this microstructural feature of sleep has been reported to be elevated in PTSD in the studies of Ross and colleagues [20, 58] and confirmed in the Kobayashi meta-analysis . It must be acknowledged that recording gamma-band EEG from the human scalp is more difficult than recording it from the dura mater in laboratory animals. Gamma-band EEG represents only 1% of total EEG power at the scalp. Under baseline conditions, scalp gamma-band EEG is contaminated by both scalp EMG [59, 60] and extraocular EMG [61, 62]. Intracranial recordings in epileptics suggest that the cortical sources of gamma-band activity are spatially restricted and may require dense electrode arrays to resolve optimally [63–65]. Gamma-band EEG may also be primarily event-related . Of particular interest, here, is the evidence that in sleep, gamma-band EEG is phase-locked to the “UP” states within delta-band EEG . These “UP” states should be uncorrelated temporally with potential EMG contaminations. While, to date, this delta phase-locked gamma has been recorded only intracranially, statistical aggregation over space and time employing source-modeled slow waves at the scalp should increase the effective signal-to-noise of this process , as would any amplification of gamma related to coactive arousal.
Unfortunately, the researcher employing gamma-band EEG in this context would still face the low-to-zero appearance rate of PTSD nightmares in the sleep laboratory. One untried approach to this challenge would be to exploit prazosin withdrawal. Many PTSD patients experience reductions in nightmares when treated with prazosin [69–71]; however, because of this drug’s very short half-life (2–3 h), nightmares can reemerge quickly if it is withdrawn (Raskind, personal communication). In principal, a prazosin withdrawal design could render the laboratory study of nightmares feasible. Needless to say, strong justifications would need to be in place to support the ethicality of such a study, including, the above discussion would suggest, a high degree of certainty that gamma-band EEG can be recorded from the human scalp during sleep. But what of ethical justifications driven by morbidity? If many PTSD patients’ nightmares can be attenuated by prazosin, or by behavioral interventions [71–73], why continue to pursue a deeper understanding of the pathophysiology of these events? First, of course, these treatments do not work for everyone [74–77]. Second, we have no strong explanations for why these treatments work at all. In particular, it is remarkable that prazosin , an alpha-adrenergic receptor antagonist, has exhibited therapeutic impact on PTSD symptoms that emerge preferentially from REM sleep, an arousal state characterized by an absence of adrenergic modulation . Third, recent data have recast nightmares in a more serious light. There is now excellent evidence that nightmare complaints are independently associated with suicidal ideation in medical patients  and college students , with both suicidal ideation  and self-harm in adolescents , with suicide attempts in psychiatric outpatients [82–84], and with completed suicides in the Finnish population . The weight of evidence suggests that the risk for suicidality conferred by nightmares is independent of both insomnia and depression [86, 87]. The risk of suicidality is known to be elevated in PTSD independent of depression (reviewed in ) and to be high in military personnel with mental disorders . Studies are now underway to assess the role of trauma-related nightmares in suicidality in military samples (Pigeon, personal communication). The association of nightmares with suicidality mandates that we push ahead in our efforts to better understand both the psychophysiology and neurobiology of these events, to more aggressively screen for them , and to treat them [70, 91–94].
First MB, Spitzer RL, Gibbon M, Williams JBW. Structured clinical interview for the DSM-IV axis-I disorders. New York: Biometrics Research Department, New York State Psychiatric Institute; 1995.