Abstract
BACKGROUND:
Sleep disturbance plays a significant role in cognitive impairment following traumatic brain injury (TBI).
OBJECTIVES:
To summarize recent findings that examine sleep disturbance and cognition in TBI.
METHODS:
Epidemiological information on sleep disorders in people with TBI is presented. A simple introduction to the role of sleep in normal cognition provides context for the literature on clinical populations. Current theory on the mechanisms underlying cognitive problems in people with sleep disorder is briefly described. Findings on the relationship between sleep disorder and cognitive problems in TBI is examined in more detail.
RESULTS:
Consistent reports of an association between sleep duration and cognition include several studies noting positive associations (shorter sleep duration accompanies cognitive impairment) and others observing negative associations (longer sleep duration accompanies cognitive problems). Both insomnia and hypersomnolence are forms of sleep disturbance that disrupt key mental processes such as memory consolidation. Obstructive sleep apnea, cerebral structural abnormalities, neurochemical changes and psychiatric pathology are implicated.
CONCLUSIONS:
Additional information is needed on how severity of injury impacts sleep and cognition. Hypothesized mechanisms underlying the effects of sleep on cognition in TBI should be empirically tested. Further, discrepancies between objective and subjective measures of sleep and cognition must be explored.
Introduction
There are several comprehensive summaries of recent literature on sleep and TBI (e.g., Barshikar & Bell, 2017, Grima et al., 2017 and Sandsmark et al. 2017), which show a high prevalence of sleep problems in people with brain injury. Another excellent review is a recent meta-analysis of self-report and objective measures of sleep which showed over 50% of people with TBI experience some form of sleep disturbance (Mathias et al., 2012). One of the prospective studies (Castriotta et al., 2007) included in that meta-analysis used objective measuressuch as polysomnography (PSG) and multiple sleep latency testing (MSLT) which showed particularly high prevalence of obstructive sleep apnea and excessive daytime sleepiness (both over 20%). Compared to controls, people with TBI sleep 1.2 hours more per day as measured by actigraphy at 6 months after injury (Imbach, Valko, Li, et al., 2015). The same sleep duration was noted at 18 months after injury (Imbach, Buchele, & Valko, 2016). In both studies by Imbach and colleagues, subjective self-reports underestimated the duration of sleep compared to objective actigraphic measurements. Similarly, both studies showed people with TBI gave subjective self-reports of excess daytime sleepiness that were substantially lower than objective measurement of sleepiness with MSLT. Finally, patients with severe TBI had more sleep per day than mild TBI patients but mean sleep latencies on MSLT did not differ between TBI patients with low and high severity trauma. (Imbach, Valko, Li et al., 2015).
Background: Sleep and cognition
In order to understand the mechanisms by which sleep disturbance affects cognition in people with TBI, it is necessary to briefly review of the role of sleep in cognition among the neurologically intact. In healthy samples, sleep has been found to play a crucial role in different forms of new learning. Declarative memory, which includes the recollection of experiences (episodic memory) and facts (semantic memory), appears to be mediated by the hippocampus, while nondeclarative memories which include procedural skill learning (such as playing the piano), behavioral conditioning (including both classical and operant) and priming, rely more on circuits running through the motor cortex and basal ganglia. (Moscovitch et al., 2005). Both declarative and non-declarative memory involve neural processes that are affected by sleep.
In declarative learning, disturbance of sleep can interfere with encoding of new information (Mander et al., 2011). But sleep is particularly important for new memory consolidation, in which a fragile memory trace is stabilized for efficient retrieval and future use. It is this process of memory consolidation that is most frequently examined in research on sleep and cognition. Early research emphasized that during new learning, sleep plays a passive role; recall is enhanced by the absence of stimuli that could interfere with consolidation (Jenkins & Dalenbach, 1924). Later work identified specific aspects of sleep that play an active role in the process of consolidation. Plihal and Born (1997) showed that recall at waking was stronger for a group that learned word associations before 3 hours of slow wave sleep than it was for another group who learned them before 3 hours of REM sleep. Later it was demonstrated that this effect is also seen in individuals who take a daytime nap (Payne et al., 2015). Reactivation and storage of new learning is associated with several electrophysiological phenomena that occur during slow wave sleep, including slow oscillations (0.5 to 4 Hz) and bursts of high frequency hippocampal activity known as sharp-wave ripples (200 Hz) (Rosanova et al., 2005; Siapas and Wilson, 1998). The role of the former phenomenon in memory consolidation during sleep was reported by Marshall et al. (2006) who showed that use of transcranial direct current stimulation to induce slow oscillations during slow wave sleep resulted in improved recall for word pairs compared to a sham stimulation condition.
Sleep also plays a critical role in consolidation of procedural learning, a form of nondeclarative memory. In procedural learning, consolidation allows tasks to become more automatic, requiring less effort. Walker et al. (2002) showed that typing a key sequence with the non-dominant hand was faster and more accurate in a group when training and testing were separated by a period of stage 2 of non-REM sleep. In contrast, REM sleep appears to involve neural reprocessing associated with consolidation of learning of tasks requiring more effort and focus (Peigneux et al., 2003). Both REM and slow wave non-REM sleep enhanced consolidation of a visual discrimination tasks (Stickgold et al., 2000). Even a 60-minute nap was associated with enhanced performance as long as it included periods of both SWS and REM sleep (Mednick et al., 2003).
In addition to understanding the role of sleep underlying mechanisms of normal memory formation, it is necessary to understand the relationship between sleep disorders and abnormal cognition in order to provide context for the literature on sleep and cognition in TBI. According to a review of studies of sleep and cognition in older adults, mixed findings leave it undetermined whether insomnia is a risk factor for cognitive decline and dementia (Yaffe et al., 2014). In contrast there is better consensus that poor sleep quality as measured by subjective self-report measures is associated with increased risk of cognitive decline or dementia in older adults (Jelcic et al., 2002; Potvin et al., 2012; Elwood et al., 2011 and Sterniczuk et al., 2013). Objective measures such as actigraphy show a similar relationship between sleep quality and cognitive outcomes. Actigraphic measures of sleep efficiency, sleep onset latency and duration of waking after sleep onset were all associated with reduced cognitive functioning (Blackwell, et al. 2006, 2011).
The relationship between sleep duration and cognition is more complicated due to disconcordance between subjective and objective measures of sleep. Blackwell et al. (2011) reported in a cross-sectional study that cognition was not associated with actigraphic estimates of total sleep time but reduced cognitive function was associated with self-report of long sleep duration. These authors reported that after adjusting for time awake after sleep onset, the association decreased, suggesting that self-report of long total sleep time might be inaccurate due to confounding effects of time in bed but not asleep. The shape of the relationship is also inconsistent between studies. In contrast to the inverse linear relationship between cognitive function and objective sleep duration reported by Blackwell et al. (2011), a large (N = 28,670) cross-sectional study of older adults found a curvilinear association between cognitive function and self-reported sleep duration (Xu et al., 2011). The analysis obtained in this sample showed more impaired cognitive function was observed in individuals who self-reported long or short sleep duration but not among those with intermediate duration (7–8 hours). Several prospective studies of cognition and self-reported sleep duration showed similar results (Potvin et al., 2012; Ferrie et al., 2011 and Virta et al., 2013). Yaffe et al. (2014) described a pattern of U-shaped associations in which cognitive impairment was observed in older adults who self-reported either long or short sleep duration. Yaffe et al. called for prospective studies using objective measures of sleep duration since such measures differ from self-report especially in populations with cognitive impairment or functional disability (Van Den Berg et al., 2008).
To summarize, sleep disturbance is highly prevalent in people with TBI. Sleep plays a crucial role in formation of different type of memory (declarative and non-declarative) and in different stages of memory formation (initial encoding and consolidation). Sleep problems appear to be related to cognitive impairment in other populations, although the relationship is complex, since there is discordance between objective and subjective measures of sleep-related variables. A further complication is that the shape of the relationship of sleep duration and cognition varies between studies that show an inverse linear association (longer duration of sleep is seen with lower cognitive performance) and other studies that found a curvilinear relationship (lower cognitive performance is associated with too much or too little sleep but intermediate duration accompanies stronger cognition). Readers are directed to the excellent review by Chambers (2017) for additional information on sleep and cognition.
These findings played an essential role in defining the questions in studies of people with TBI. Studies in this population took two broad directions in conceptualizing the aspects of cognition that were believed to be affected by sleep. One group of studies examined correlates of neuropsychological performance, as measured by formal psychometric instruments and the other examined correlates of functional disability, especially as measured by clinician ratings.
Cognition and sleep in TBI
Neuropsychological performance
Clinchot et al. (1998) reported that a sample of 86 consecutive patients admitted for inpatient rehabilitation for TBI, underwent a phone interview one year after their injury on sleep quality, concurrent functioning (e.g., persistent symptoms, employment status), alcohol use and medications. Those with severe injuries were less likely to have problems with sleep. Of neuropsychological variables examined, immediate memory was the strongest predictor of sleep difficulty at follow-up. People with average or above average immediate memory at discharge were more likely to have sleep problems one year post-injury. Those with attention or delayed memory in the average to above average range were also more likely to have problems with sleep, although other cognitive domains were not related to sleep difficulties. None of the cognitive variables were associated with initiation or maintenance of sleep although sleep duration trended toward significance (those with impaired immediate memory were more likely to sleep longer than before their injury). Concurrent complaints of headache were significantly associated with sleep disturbance. Those who were working or going to school were less likely to have problems with sleep maintenance. Use of medications (seizure meds or antidepressants) at follow-up was not associated with sleep disturbance. In multiple logistic regression, the best demographic predictors of sleep disturbance were gender, age and alcohol abuse: difficulties were more likely in older females with an alcohol abuse history. When incorporating all the variables, age and immediate memory comprised the best predictive model: older persons and those with average or better immediate memory were more likely to be having problems with sleep at follow-up. The authors suggested the association between severe injury and self-reported good sleep may reflect underreporting of sleep disturbance by those with severe cognitive impairment. Although the effects of alcohol abuse were examined, the effects of other forms of comorbid psychopathology (e.g., depression) were not.
Castriotta et al. (2007) described a multicenter study that collected PSG, MSLT, neuropsychological data and self-report of mood and of sleep quality for 87 individuals who had sustained a TBI 3 months or more before examination. MSLT was used to classify subjects into sleepy vs. non-sleepy groups. Neuropsychological performance was compared for 60 non-sleepy subjects and 20 sleepy subjects. The results of these group comparisons disclosed: 1) sleepy subjects’ fastest reaction times were significantly slower than the non-sleepy subjects, 2) sleepy subjects trended toward having a slower average slow reaction times, and 3) sleepy subjects rated better quality of life than non-sleepy subjects. Groups did not differ on self-reported sleepiness. Diagnoses based on PSG and MSLT findings included obstructive sleep apnea in 23% of subjects, posttraumatic hypersomnia in 11% of subjects, narcolepsy in 6% and periodic limb movements in sleep in 7% of subjects. In total, 40 subjects were diagnosed with a sleep disorder and were compared against 47 subjects without a sleep disorder. Groups differed in body mass index (sleep disordered subjects were heavier). Group comparisons on neuropsychological data showed: 1) sleep-disordered subjects’ fastest reaction times were significantly slower than the non-sleep-disordered subjects, 2) sleep-disordered subjects demonstrated significantly slower reaction times, and 3) sleep-disordered subjects made more lapses than the non-sleep-disordered group. Groups did not differ in self-reported mood. There was a trend toward higher sleep-related quality of life in the sleep-disordered group. That is, in spite of objective evidence of poor vigilance, there was a trend towards sleepy TBI subjects and TBI subjects with a sleep disorder diagnosis to actually report better sleep related quality of life than those that did not carry a diagnosis and/or were not sleepy. The authors suggest it was highly likely that poor awareness resulted in subjects over-reporting self-perceived quality of life and underreporting mood changes and subjective sleepiness. They argued this conclusion was supported by the low correlation between objective sleepiness and subjective sleepiness, and by the lack of group differences of self-reported mood variables such as fatigue and vigor. Due to missing data, the authors could not directly address the role of injury severity in explaining awareness and accuracy of self-reported sleep disturbance.
Mahmood et al. (2004) reported on an archival sample of 87 TBI patients. A hierarchical multiple regression showed that poor performance on neuropsychological tests was associated with self-reported sleep disturbance after controlling for sex and injury severity. Measures sensitive to higher-order reasoning and speed of information processing had strong positive associations with self-reported sleep disturbance but a block design measure that is less sensitive to executive functioning had a strong negative association with self-reported sleep disturbance. In an analysis of severity of injury and aspects of self-reported sleep disturbance, it was determined that compared to those with moderate and severe TBI, patients with mild TBI endorsed more severe problems in subjective sleep quality, self-reported sleep duration and specific sleep disturbance. Exploratory analyses of subscales on a self-report measure showed severe TBI patients reported shorter sleep onset and longer sleep duration. The authors observed that differences between mild and severe TBI in self-reported sleep variables have been reported by several other studies (e.g., Clinchot et al., 1998; Beetar et al., 1996, Ouellet et al., 2006). They did not examine the effects of comorbid psychopathology.
Makley et al. (2009) reported objectively measured sleep efficiency predicted emergence from posttraumatic amnesia (PTA) in nine individuals with moderate to severe TBI on an inpatient rehabilitation unit. Sleep efficiency was measured by actigraphy, which showed the proportion of time sleeping relative to the total time in bed. Presence and severity of posttraumatic amnesia was assessed by serial admission of a brief mental status exam. For those with ongoing posttraumatic amnesia on admission to inpatient rehabilitation, for each 10-unit increase in sleep efficiency there was a 1-unit increase in the mental status exam score, which was significantly different from those who cleared posttraumatic amnesia prior to admission, in whom sleep efficiency was not related to mental status exam results. Self-reported satisfaction with life did not differ between those in PTA and those who cleared PTA prior to admission. The authors did not explore the extent to which cognitive impairment affected reliability of self-report or the extent to which results were affected by comorbid psychopathology.
Bloomfield et al. (2009) recruited 44 individuals with history of TBI ranging from mild to severe to undergo neuropsychological testing of attention, to complete self-report questionnaires of mood and sleep and to undergo objective actigraphic measurement of sleep for one week. A group of 15 individuals with good sleep were identified through agreement of objective and self-report measures. Another group of 11 individuals with poor sleep were identified through the same method, while 18 individuals could not be classified due to disagreement between sleep measures. The Good Sleep and Poor Sleep groups differed on errors of commission on a sustained attention task, reflecting impairment in the latter group. A positive association was noted between depression and commission errors for the Good Sleep group but not the Poor Sleep group. Groups did not differ significantly on other measures of attention or on perceived failures of sustained attention. The lack of association between self-report of attention problems and objective measures of attention was considered consistent with past reports of reduced insight in severe TBI, although the authors did not specify that the association was any stronger for their subjects with mild TBI. They noted that a substantial portion of their sample gave self-report of sleep that was discordant with objective measures of sleep, which may reflect inaccuracy due to cognitive difficulties. They also offered an alternative theory that insomnia symptoms are heterogeneous and consequently require multi-modal assessment. They specified that individuals with mild injury may be more vulnerable to psychiatric pathology, contributing to insomnia, although they did not conduct analyses of injury severity in their sample to confirm this hypothesis.
Zollman and Larson (2012) reported on a sample of 24 outpatients with TBI and insomnia who participated in a clinical trial of acupuncture. At pre-intervention baseline, divided attention was positively associated with perception of insomnia. Similarly, post-intervention divided attention was associated with perception of insomnia measured at follow-up. Although Total Sleep Time as measured by actigraphy did not improve from preintervention to post, self-reported perception of sleep improved in the treatment group alone. This improvement in perception of sleep was sustained for at least 1 month after cessation of acupuncture. This suggests that, in addition to providing equal efficacy in sleep time achieved, acupuncture offered a sustained benefit in perception of sleep time/quality, a benefit not seen in those undergoing conventional treatment for insomnia. This is significant because others have proposed that distorted perception of sleep (i.e., tendency to underestimate sleep time or quality) plays a key role in maintaining insomnia, and that correcting that distortion can help resolve complaints of insomnia (Tang & Harvey, 2004). Limited information was available on severity of injury so the authors did not examine whether reliability of self-report was compromised in those with severe injury and whether depression impacted perception of sleep quality in those with mild injury, as suggested in previous research.
Wiseman-Hakes et al. (2013) reported that in an outpatient sample of 12 people with remote history of TBI who underwent 4–13 months of pharmacological and behavioral treatment for sleep / wake disorders, self-reported gains in sleep were accompanied by gains in neuropsychological measures of language and speed of verbal processing. Subjects endorsed improvement in depression and across all domains assessed by a self-rating measure of sleep quality, mood, fatigue, naps, attention, memory and language processing. Although improvement in cognition coincided with improvement in sleep, the lack of a placebo group prevented examination of the influence of non-specific factors on outcome. The authors highlighted the heterogeneity of their sample, which included 10 people with moderate-severe injury and 2 with mild injury. Due to the small size of their sample, they were not able to conduct formal analysis of whether severity of injury played a role in their findings. They also did not report on concordance between objective and subjective measures of sleep and of cognition or on the effects of comorbid psychopathology.
Mantua et al. (2015) reported on a sample of 56 young adults, which included 26 with self-reported history of concussion and 30 without such a history. Half of each diagnostic group was assigned to a sleep condition and half was assigned to a no-sleep condition. A 2X2 ANOVA was conducted to examine effects of diagnosis (concussion vs. no-concussion) and sleep (sleep vs. no-sleep) on memory. All subjects completed verbal associates learning trials and returned after 12 hours to complete a delayed recall trial. During the intervening delay, sleep groups from both diagnostic groups underwent PSG. There was not a main effect for diagnosis: those with a history of concussion did not differ from the no-concussion group on retention of new learning. There was a main effect for sleep: the combined sleep groups demonstrated stronger retention than the combined no-sleep groups. There was no significant interaction of diagnosis and sleep on memory. The concussion sleep group had a greater proportion of the sleep period in Slow Wave Sleep (SWS) than did the non-TBI sleep group. Although memory was not significantly associated with SWS, it was associated with proportion of the sleep period in non-REM sleep (including non-REM Stage 2 and Slow Wave Sleep). The authors concluded that sleep composition is altered following mild TBI but sleep deficits are not associated with insufficient sleep-dependent memory consolidation. The authors noted that findings may not generalize to moderate-severe TBI, consistent with others who concluded that the effects of sleep on cognition in mild TBI may be different than that in moderate-severe injury. They did not examine self-report of sleep or cognition and did not examine the effects of comorbid psychopathology.
Holcomb et al. (2016) described a multicenter study that examined 106 individuals receiving inpatient rehabilitation for moderate to severe TBI. Over a period of three weeks, they underwent weekly observer ratings of sleep-wake cycle disturbance (SWCD). Fifty-six subjects were classified as positive for SWCD and 50 were classified as SWCD negative. A brief formal exam measured severe impairment of orientation, basic attention, visual recognition memory, comprehension (yes/no responding) and vigilance. Individual Growth Curve analysis yielded a continuous dimension of cognition over time described as the individual growth trajectory. The analysis showed a significant group by time interaction in which the cognitive recovery trajectory for the SWCD positive group significantly differed from that of the negative group, the latter of which improved at a more rapid rate. The authors conclude that those with sleep dysfunction could be prone to different and potentially poorer courses of cognitive recovery. They noted this finding is consistent with animal models of recovery from stroke in which experimentally-injured rats that were deprived of sleep were found to have significantly lower amounts of pathophysiological indices of brain repair (e.g., axonal sprouting, synaptogenesis) than an injured control group (Zunzunegu et al., 2011). They also noted their findings are consonant with Zunzunegu’s report that sleep-deprived injured rats performed significantly worse on behavioral learning tasks compared to controls who were allowed to sleep during recovery. They did not report on subjects’ perception of their cognitive status or on how psychiatric pathology may have covaried with recovery.
Beaulieu-Bonneau et al. (2017) compared 22 people with moderate-severe TBI and 22 neurologically-intact controls on neuropsychological abilities, self-reported fatigue, self-reported sleepiness, objective measures of sleep (PSG) and performance on a driving simulator. They found that people with TBI scored worse on neuropsychological tests, had more difficulty with driving, reported more fatigue and showed an association between decreased Total Sleep Time and decreased speed on a trail-making test, although sleep was not associated with performance on the driving simulator. In both controls and people with TBI, poorer attention on a continuous performance task was observed in those subjects who reported a larger increase in sleepiness after completing that task. Interestingly, people with TBI declined in processing speed across trials on that task, although that decline was not correlated with increased self-reported fatigue, suggesting that objective and subjective measures of fatigue measured independent constructs. They did not assess the effects of comorbid depression on sleep or cognition.
Functional disability
Duclos (2014) reported that lack of a 24-hour sleep-wake cycle during a 10-day period of acute hospital care was associated with lower clinician ratings of ADLs. Actigraphy data assessing consolidation of the rest-activity cycle were collected during the entire period of inpatient hospitalization of 16 patients with moderate to severe TBI who were enrolled within 4 to 52 days of injury. Lower ratios of activity counts measured over during daytime hours to those measured during nighttime hours were significantly associated with an occupational therapist’s rating of severity of disability. Lower day/night activity ratios were also associated with increased length of stay in intensive care and length of hospitalization. Patients with rest-activity cycle consolidation were more likely to have cleared posttraumatic amnesia at the time of discharge from inpatient care. The authors acknowledged the lack of polysomnography data in their sample prevented them from concluding that all episodes of absence of activity were due to sleep, but they expressed confidence the absence of rest-activity consolidation is associated with fragmentation of sleep and wake episodes. They noted it is difficult to establish that sleep consolidation has a causal relationship with cognitive recovery. While it is possible that sleep fragmentation impedes neural recovery and consolidation of new learning, it is also possible that structural brain damage affects both memory and sleep and recovery from both occurs simultaneously although the course of improvement in one is not necessarily contingent on the improvement of the other. They observed these findings were consistent with previous reports that decreased sleep is associated with decreased performance of activities of daily living as measured by clinician rating (e.g., Worthington & Melia 2006). They did not include self-report measures and did not explore the role of comorbid depression on sleep or cognitive performance.
Lequerica et al. (2017) reported on 237 people who received inpatient rehabilitation for moderate-severe TBI who were examined at 1, 2 and 5 years post-injury. Groups were identified based on those who gave a self-report of resolved fatigue versus those with unresolved fatigue. Group comparisons were reported on self-report measures of sleep quality, depression and community integration and clinician ratings of functional independence, disability and depression across the two follow-up intervals (Year 1 to Year 2 and Year 2 to Year 5). Persistence of self-reported fatigue from 1–2 years post-injury was associated with self-reported depression, self-reported sleep quality and with clinician-rated disability, but not with clinician-rated functional independence. Persistence of self-reported fatigue from 2–5 years post-injury was associated with self-reported community integration. The authors did not report on discrepancies between self-report and clinician ratings. They also did not examine the extent to which depression covaried with sleep quality or disability.
Mechanisms
Other articles in the present issue will address possible mechanisms underlying sleep problems in people with TBI. Although it is beyond the scope of the present article to survey those mechanisms in any detail, those that have been hypothesized to have implications for cognition will be noted briefly along with notes about needs for further study. Obstructive sleep apnea, neurochemical changes, psychiatric pathology and cerebral structural abnormalities have all been identified as factors relevant to sleep and cognitive impairment in TBI.
Obstructive sleep apnea (OSA) was examined by Wilde et al. (2007), who examined a sample of 35 people who were at least 3 months post TBI. Obstructive sleep apnea (OSA) was confirmed by nocturnal PSG in 19 subjects who did not differ from the other 16 in age, education or time post injury. The OSA group performed significantly worse on measures of delayed recall and retention of verbal and visual information and had more lapses on a vigilance task. A limitation of the study was the lack of information about injury severity for over half the sample. It is unknown whether injury severity has a role in the relationship between OSA and cognition in people with TBI.
Neurochemical pathology was included in an exhaustive survey of TBI-related sleep problems by Sandsmark et al. (2017). They summarized TBI-related chemical pathophysiology involving neurotransmitter/neuropeptide systems that result in sleep problems and may also result in cognitive dysfunction. Whether it is sleep that mediates that cognitive dysfunction is unclear.
Comorbid psychiatric pathology probably has a role in sleep and cognitive impairment in TBI. As noted earlier, Bloomfield et al. (2009) reported the counterintuitive finding that psychiatric pathology was associated with cognitive impairment among TBI patients with healthy sleep, but not in those with sleep problems. The role of depression was also implicated in a study by Gosselin et al. (2009). They noted that concussed athletes endorse greater sleep problems although they did not differ from uninjured athletes in PSG-derived sleep variables and that the concussed athletes group also reported more depressive symptoms on a questionnaire, even after sleep-related items were excluded. Further, they observed that concussed athletes gave lower subjective ratings of their cognitive performance than did controls, although objective neuropsychological testing showed no differences between groups. Yet, the role of comorbid psychiatric pathology is not sufficient in itself to explain the effect of sleep disturbance on disability. In a sample of people with mild TBI, Mollayeva et al. (2016) found that persistent sleep-wake disturbances and insomnia is associated with global disability after controlling for depression, anxietyand pain.
Structural abnormalities of potential significance were studied by Valko, Gavrilov and Yamamoto (2016), who reported that postmortem examinations showed people with severe TBI had fewer noradrenergic neurons in the locus coeruleus and fewer serotonergic neurons in the dorsal raphe nucleus. Since neurons in both of these regions play important roles in promoting wakefulness, their loss in TBI may explain abnormalities in sleep in that population, although their finding that structural changes were relatively mild is inconsistent with the severity of associated sleep changes. Similarly, Hou et al. (2013) noted the majority of cases of TBI do not have an obvious structural abnormality that would explain their sleep-wake dysfunction. The possibility that sleep impairment due to such pathophysiology exerts direct or indirect effects on cognition has not been expressly addressed and requires furtherinvestigation.
Conclusion: Methodological questions and future directions
An association between sleep disturbance and cognition is consistently reported in people with TBI, although confidence about that association is tempered by variable data about the nature of the relationship. Since cognitive impairment is observed in some with hypersomnolence and in others with insomnia, simple linear models will fail to explain the relationship between sleep duration and cognitive performance. Similarly, discrepancies between subjective and objective measures of sleep and cognition raise questions about the validity of self-report. Alternately, those discrepancies may indicate a need for more sophisticated models that explain how subjective satisfaction with sleep may be distinct from objectively-confirmed sleep pathology, each one of which may have a different relationship with cognition. Such models must provide an account of the role of depression, which can influence perception of quality of both sleep and cognition.
Durant (2015) made an important observation about the next step for research on sleep and cognition in TBI in his response to Mantua et al. (2015). He noted that while early studies of this topic appropriately made use of the tools that were available, including self-report measures and actigraphy, those instruments are not strongly associated with the objective measures of sleep provided by polysomnography, underlining the strong need for the latter in future studies. Similarly, he observed that severity of injury exerts an influence and must be explicitly examined. Several years before, a literature review of 33 studies of sleep and TBI (Orff, 2009) identified 21 that included mixed samples of people with mild and moderate-severe injuries, 10 restricted to mild TBI, one with severe and one that did not specify. The author of that review suggested a limitation of those studies was that they had not adequately examined how injury severity affects the prevalence and character of sleep disturbance. Our capacity to develop accurate models of the relationship between sleep and cognition in TBI will be largely contingent on our ability to accurately characterize our samples and explore how that relationship covaries with injury severity.
Conflict of interest
The author has no interests to declare.
Footnotes
Acknowledgments
The assistance of Nalini Mahajan and Jane Aruin of the Medical Library at Marianjoy Rehabilitation Hospital is gratefully acknowledged.