A systematic review of variables associated with sleep paralysis.

Sleep paralysis is a relatively common but under-researched phenomenon. While the causes are unknown, a number of studies have investigated potential risk factors. In this article, we conducted a systematic review on the available literature regarding variables associated with both the frequency and intensity of sleep paralysis episodes. A total of 42 studies met the inclusion criteria. For each study, sample size, study site, sex and age of participants, sleep paralysis measure, and results of analyses looking at the relationship(s) between sleep paralysis and associated variable(s) were extracted. A large number of variables were associated with sleep paralysis and a number of themes emerged. These were: substance use, stress and trauma, genetic influences, physical illness, personality, intelligence, anomalous beliefs, sleep problems and disorders (both in terms of subjective sleep quality and objective sleep disruption), symptoms of psychiatric illness in non-clinical samples (particularly anxiety symptoms), and psychiatric disorders. Sleep paralysis appears to be particularly prevalent in post-traumatic stress disorder, and to a less degree, panic disorder. Limitations of the current literature, directions for future research, and implications for clinical practice are discussed.


Introduction
Sleep paralysis involves a period of time at either sleep onset or upon awakening from sleep during which voluntary muscle movements are inhibited. Ocular and respiratory movements remain unaltered and perception of the immediate environment is clear [1]. These episodes are frequently associated with a variety of hallucinations, such as a sense of an evil presence (known as intruder hallucinations), pressure felt on the chest (incubus hallucinations), and illusory feelings of movement (vestibular-motor (V-M) hallucinations) [2]. Sleep paralysis is a global phenomenon, with terms for sleep paralysis existing in over 100 cultures [3]. In many places, sleep paralysis experiences are interwoven with a culture's folklore [4,5]. Episodes of sleep paralysis have been suggested as an explanation for supposed paranormal phenomena such as witchcraft [6], demonic assault [7], and space alien abduction [8,9]. Fear and distress are typically associated with episodes [2], though feelings of bliss are sometimes reported [10].
A review of lifetime prevalence rates of sleep paralysis in the general population estimated prevalence to be approximately 8%, though individual study estimates greatly vary from 2-60% [11]. The lack of a 'gold standard' measure of sleep paralysis is likely part of the reason for this [3] and the problem is amplified by the fact that the precise phrasing used to ask about sleep paralysis has been shown to affect the reported prevalence rate [12]. Averaging over multiple studies, no effects of age have been found and sex differences show mixed results [11].
Finally, there are slightly higher lifetime prevalence rates of sleep paralysis in non-Caucasian compared to Caucasian groups [11].

M A N U S C R I P T
A C C E P T E D ACCEPTED MANUSCRIPT 4 In addition to limitations with the measures used to assess sleep paralysis, another outstanding problem is the lack of consistency with terminology. Sleep paralysis is a common symptom of narcolepsy, a neurological disorder. Narcolepsy is characterised by excessive daytime sleepiness, cataplexy (sudden, brief, bilateral losses of muscle tone in response to strong emotions such as laughter or anger [13]) and disturbed nocturnal sleeping patterns [14]. Therefore, the term isolated sleep paralysis is preferred when sleep paralysis is present in the absence of a diagnosis of narcolepsy [3]. Some authors also use the term fearful isolated sleep paralysis to indicate cases where episodes are causing clinically significant fear and/or distress [3]. Finally, when episodes occur repeatedly, the term recurrent (fearful) isolated sleep paralysis can be used, though there is no agreement on exactly how often episodes need to occur in order to be considered recurrent [15,16].
The causes of sleep paralysis are likely to be multifactorial [17]. It is therefore important for clinicians and researchers to understand the factors that may influence the frequency and intensity of episodes. Here, we systematically review the available literature regarding variables associated with sleep paralysis. It is hoped that this review will provide a resource to both clinicians, who shall have a better understanding of this common experience, and also researchers interested in formulating new research questions. were used: "sleep paralysis", "isolated sleep paralysis", "parasomnia not otherwise specified", "hypnagogic", "hypnopompic", "parasomnia", "sensed presence", and "incubus".

Methods
There was no restriction made on the age of articles included in this review. For all identified articles published since 2000, reference lists were also scanned to see whether we had missed any articles suitable for inclusion in this review. Furthermore, manual searches were conducted of all journals containing more than 5 of the studies identified via the database searches. Where this happened, only articles published since 2000 were scanned for reasons of feasibility. These steps allowed us to reduce the possibility of non-indexed studies being missed.
For inclusion in this systematic review, studies were required to meet the following criteria: 1. Results are presented as a full original research paper published in a peer-reviewed journal.
2. Sleep paralysis is clearly defined, and it is clear that sleep paralysis has been identified in the sample, and is not another phenomenon (e.g. night terrors, hypnagogic/hypnopompic hallucinations with no paralysis, nightmares).
3. The presence of sleep paralysis in the sample was measured either via a continuous measure of frequency or a binary sleep paralysis present/absent measure.
4. An association has been explored between sleep paralysis and other variable(s). Either a comparison between groups (e.g. sleep paralysis vs no sleep paralysis) or an association between sleep paralysis frequency/intensity and the variable(s) under study. 5. Single case studies were excluded. 6. Review, commentary, or opinion articles not including any original data were excluded (but reference lists from such papers were checked, in line with the search strategy). 7. Only English language articles were included.

M A N U S C R I P T
A C C E P T E D ACCEPTED MANUSCRIPT 6 In total, 35 articles were identified that matched all the inclusion criteria. One article was found by searching article reference lists, and 1 was obtained through personal correspondence.
Furthermore, an additional 5 articles that also matched all the inclusion criteria were recommended by a reviewer. Therefore, a total of 42 articles were included in this report. The process for selecting studies is displayed in the supplementary materials. From the included studies, the following data were extracted: Sample size and study site, sex and age of participants, sleep paralysis measure, and results of analyses looking at the relationship(s) between sleep paralysis and associated variable(s).

Results
The articles included in the review can be found in Table 1. The studies identified come from a wide range of different research groups, providing a highly international and cross-cultural sample.
Studies looking at associations between sleep paralysis and other variables are found in Table 2. Associations between sleep paralysis and variables specifically related to sleep-related factors are displayed in Table 3, with associations with other sleep disorders shown in Table 4. Table 5 shows associations between sleep paralysis and symptoms of psychiatric illness. Finally, Table 6 shows associations with psychiatric disorders and medication. Overall, the majority of studies looked at sleep paralysis frequency alone, with fewer studies examining variables associated with the intensity and/or vividness of sleep paralysis and associated hallucinations.

Demographics
Relationships between demographic variables and sleep paralysis are shown in Table 2.
Generally, studies that have investigated age differences in sleep paralysis prevalence have found no significant effect of age [17][18][19][20][21]. Similarly for sex differences the majority of studies found no significant effects [17][18][19][22][23][24][25][26][27]. Two large-scale surveys of Asian adolescents did find significant sex differences, with higher prevalence in females though the reported differences were very small. Munezawa and colleagues found an 8.2% prevalence for males and 8.4% for females. Ma et al. found a prevalence of 6.1% for males and 7.4% for females [24,28]. In a study of Hmong immigrants, higher odds of experiencing sleep paralysis were found in male participants (odds ratio (OR) = 1.61), though the exact prevalence rates for both sexes were not reported [21]. Finally, a study in an American sample found that males were more likely to have experienced lifetime isolated sleep paralysis than females, but there were no differences in lifetime fearful or recurrent fearful isolated sleep paralysis episodes [29]. With regards to ethnicity, evidence is also mixed. One study found African American individuals experienced a higher incidence of sleep paralysis compared with Caucasians [30]. A second study found that non-Caucasians had a higher prevalence of fearful and recurrent fearful isolated sleep paralysis compared to Caucasians [29]. Two other studies however found no significant effect of ethnicity [23,31].
Presence of sleep paralysis was shown to be higher in rural compared to urban areas in a survey of Chinese adolescents [28]. Having a higher amount of available money had a small but significant association with the presence of sleep paralysis (10.9% of participants with ¥5000 or more available to them experienced sleep paralysis compared with 7.3% who had less than ¥5000 available) in one Japanese sample, as did regularly eating breakfast, with those who ate M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 8 breakfast everyday reporting a lower incidence of sleep paralysis (7.5%) compared to those who ate breakfast occasionally (10.8%) [24], though findings are inconsistent [21,32]. It is unclear why these variables should be associated with sleep paralysis. Some research suggests that food timing may be important in the synchronisation of internal circadian clocks [33]. It is possible that by eating breakfast only occasionally, circadian clocks are disrupted leading to an increased incidence of sleep paralysis.

Drinking alcohol, smoking, and substance use
Relationships between substance-use variables and sleep paralysis are shown in Table 2. It is unclear whether sleep paralysis is associated with general substance use. In two large nationwide samples in China (N = 11,754) and Japan (N = 90,081) it was found that those who reported drinking at least one alcoholic drink per day over the last month were significantly more likely (9-12%) than others (6-7%) to report experiencing sleep paralysis [24,28]. However, in a large (N = 862) UK sample where self-reported weekly alcohol consumption in terms of units of alcohol was recorded, alcohol intake over the past week did not predict sleep paralysis independently of anxiety symptoms, depressed mood, threatening events, and sleep quality [17].
Evidence for the association between smoking behaviour and sleep paralysis is also mixed -with increased odds of experiencing sleep paralysis in those who smoked (defined as at least one cigarette a day) (15%) compared with those who did not (8%) [24]. This relationship was not found in two other studies [17,28]. Of note, none of the studies on this topic have provided a quantitative estimate of amount of nicotine consumed. Given the negative impact caffeine can have on sleep [34], it was surprising that there was no evidence of a significant association between caffeine intake and sleep paralysis [17,28].

Stress and trauma
Relationships between stress and trauma, and sleep paralysis are shown in Table 2. A confirmed or unconfirmed history of childhood sexual abuse (CSA) was found to be significantly related to frequency of sleep paralysis episodes [35]. This study defined a confirmed case of CSA as either having a person other than the victim and/or abuser confirm the abuse, or that other people confronted the abuser, or that the abuser was charged in connection with the abuse. An unconfirmed history of CSA was defined as having vivid recollections of abuse in the absence of anyone being aware of it/no one being confronted or charged as the abuser. The frequency and intensity of intruder and incubus hallucinations were significantly greater in both CSA groups as compared to those who did not report sexual abuse [35]. No differences in terms of V-M hallucinations were found between groups. Another study found the reported prevalence of sleep paralysis did not differ significantly between groups of participants who reported remembering their CSA (47%) with those who believed they had experienced CSA but possessed no autobiographical memories of it (44%). Both groups were significantly higher than a control group who reported not having experienced CSA (13%) [36].
Other experiences of threatening/traumatic events also appear to be related to sleep paralysis. In a sample of Hmong immigrants living in the USA, stressful experiences during the Vietnam war (e.g. "I was exposed to chemical warfare", "I lost family, close relatives or friends") were related to increased odds of experiencing sleep paralysis [21]. General experiences with potentially traumatic events (such as assault, death of a loved one, disasters, etc.) were found to be related to sleep paralysis in terms of the occurrence of a traumatic event [37]. Also a link was found between increasing numbers of traumatic events experienced and sleep paralysis [17,37]. Relatedly, self-report levels of life-stress showed similar associations M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 10 with sleep paralysis [19,38,39]. In studies of Nigerian doctors and nurses, the authors did not find a relationship between sleep paralysis and threatening events [26,27]. This may be due to a smaller sample size or could be because of the specific demographic targeted.

Hereditary factors
Relationships between hereditary factors and sleep paralysis are shown in Table 2. A familial association has been reported for sleep paralysis [25,40,41]. One study by us used a twin modelling approach to disentangle genetic from environmental effects. There appeared to be moderate genetic influences (estimated at 53%) on variation in presence of sleep paralysis [17].
A number of specific genes involved in circadian cycles were also examined and specific polymorphisms of the PER2 gene were associated with sleep paralysis although the association was no longer significant when adjusting for multiple testing [17]. Another study employing a very small sample (N = 44) found no association between sleep paralysis and narcolepsy genes HLA DR2 or HLA DQ1 [41].

Physical health
Relationships between physical health and sleep paralysis are shown in Table 2. General physical health problems appear to be associated with sleep paralysis when using general physical health quality of life scales [20,42,43], though this was not found in one study [18].
Little research however has linked any specific physical health problems with sleep paralysis.
For example, body mass index (BMI) was not found to be related to whether someone reported having experienced sleep paralysis during their lifetime [32] or to vary with frequency of episodes [20]. However, in another study, higher BMI was related to both lifetime and recurrent episodes of fearful isolated sleep paralysis but not related to lifetime non-fearful isolated sleep M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 11 paralysis [29]. In an experimental study, participants who experienced sleep paralysis in a laboratory environment showed higher diastolic and lower systolic blood pressure in SOREMPs preceding a sleep paralysis episode compared to SOREMPs without sleep paralysis [42]. Finally, one study showed that complaints of chronic pain were linked to the presence of sleep paralysis experiences [21].

Personality, intelligence, and anomalous beliefs
Relationships with sleep paralysis are shown in Table 2. Measures of personality traits, such as the Eysenck personality questionnaire, do not appear to be related to sleep paralysis frequency [22,26,27,38]. Levels of waking state dissociative experiences, involving depersonalisation, derealisation, and amnesia, were found to be related to both sleep paralysis frequency [36] and the frequency/intensity of all three hallucination types [35]. The degree to which individuals become absorbed in their mental fantasy has been linked to hallucination frequency and intensity [35], but another study did not find absorption to be related to overall episode frequency [36]. It should be noted that these studies were conducted in groups containing individuals who had experienced CSA. However, another study in a more general sample found dissociative experiences to be linked to sleep paralysis frequency, and also to intruder and V-M hallucinations [39].
In a university sample, a composite measure of "imaginativeness" that comprised scales of absorption, fantasy proneness, magical thinking, imagery vividness, paranormal and mystic beliefs, perceptual aberration, and unusual sensory experiences, was related to sleep paralysis frequency and intensity [22]. It was unclear which of these components was driving the relationship. In the same sample, those reporting sleep paralysis compared to others showed greater hypnotisability assessed both subjectively and objectively [22]. Sensory imagery M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 12 vividness was found to not be related to sleep paralysis frequency, but significantly predicted the intensity of all three hallucination types [39]. Frequency of daydreams was not related to sleep paralysis, but daydreaming style, particularly positive constructive daydreams, was related to V-M hallucinations [39].
One study looked at intelligence in relation to sleep paralysis and found a significant negative association between intelligence quotient (IQ) and reports of lifetime isolated sleep paralysis, fearful isolated sleep paralysis, and fearful recurrent isolated sleep paralysis [29].
When levels of paranormal belief have been examined, mixed evidence has been found, with some studies finding no relationship [19,22], whilst others report links [39]. Paranormal beliefs have also been related to intruder and V-M hallucinations during sleep paralysis [39]. Levels of belief in conspiracy theories have been found not to relate to sleep paralysis [39].

Sleep-related factors
Relationships between sleep-related variables and sleep paralysis are shown in Table 3.
Symptoms of automatic behaviour, the spontaneous production of purposeless verbal or motor behaviour without conscious control, an auxiliary symptom of narcolepsy [44], were found to predict sleep paralysis independently of age, physical health, sleep problems, mental health, and psychiatric medication [20]. In a specific sample of 747 Hmong immigrants living in the USA, sleep paralysis was higher in individuals who had a relative who had suffered sudden unexpected nocturnal death syndrome (SUNDS) (29% of the sample reported having a relative die of SUNDS) [21] as compared to those who did not have a relative die of SUNDS.
A number of 'anomalous' sleep experiences appear to be more common in people who also experience sleep paralysis. Experiencing nightmares appears to be common in those with M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 13 sleep paralysis, as an association has been found in a number of studies [22,24,32]. Responses to whether participants have ever experienced a lucid dream, which is a dream in which the dreamer, whilst dreaming, is aware that they are in a dream [45], showed no relationship with sleep paralysis frequency in two studies [22,46]. However, in another study that used a continuous measure of lucid dreaming frequency, a significant association with sleep paralysis was found [39]. This was particularly true of sleep paralysis episodes featuring V-M hallucinations [39]. Hypnopompic (sleep-offset) hallucinations were a significant predictor of sleep paralysis, though it is possible that these hallucinations were not occurring during sleep paralysis [20]. Exploding head syndrome (hearing loud noises such as explosions at wake-sleep or sleep-wake transitions) is more common in individuals reporting episodes of fearful isolated sleep paralysis than in those not reporting sleep paralysis (37% vs 14%) [47]. Out-of-body experiences (OBE), often associated with REM sleep [48,49], were reported to be associated with sleep paralysis although the possibility that the OBE was a hallucination experienced during sleep paralysis cannot be ruled out [22].
Two experimental studies were able to induce episodes of sleep paralysis in a laboratory environment [42,50]. Sleep can be distinguished into different stages. Currently a broad distinction is made between rapid eye movement (REM) and non-rapid eye movement (NREM) sleep. Within NREM, sleep is further divided into stage 1, stage 2, and stage 3. A typical sleep cycle comprises of the individual progressing from stage 1 to stage 2 to stage 3, and then back to stage 1 before entering a period of REM sleep. In a study of 16 participants, whom had all selfreported having sleep paralysis at least twice, it was shown that sleep-onset REM periods A second study using 13 participants, again selected based on self-reporting at least two lifetime episodes of sleep paralysis, induced a total of eight episodes of sleep paralysis from 184 sleep interruptions (4%) using a multi-phasic sleep wake schedule [42]. Six of the eight episodes occurred following a SOREMP whilst no episodes occurred following a regular REM period.
The other two sleep paralysis episodes occurred immediately after sleep onset and, even though PSG data suggested a REM period was imminent, activity could not be scored as REM in the conventional manner [42]. Participants who had episodes of sleep paralysis in the laboratory showed increased amounts of stage 1 sleep, compared to participants who did not experience sleep paralysis. Furthermore, poorer performance on a vigilance task (longer reaction times) was found before a SOREMP with sleep paralysis. More generally, poorer vigilance task performance, and increased subjective sleepiness were found in the sleep paralysis group [42]. It should be noted in both of these studies, SOREMPs were defined as REM periods occurring within 25 minutes of sleep onset. This differs from the American Academy of Sleep Medicine definition, which defines a SOREMP as REM sleep occurring within 15 minutes of sleep onset [1]. Related to sleep disruption, shift work was related to a higher self-reported prevalence of sleep paralysis (48% vs 36%) [51].

Subjective sleep quality, as assessed via measures such as the Pittsburgh Sleep Quality
Index [52], was found to be related to sleep paralysis in multiple studies [17,18,24,28,32,39].
Excessive daytime sleepiness was associated with sleep paralysis in two studies [18,24], with another two finding no association [20,25]. Particular aspects of 'sleep hygiene' (which refers to habits and practices that are conducive to sleeping well on a regular basis) were associated with greater odds of reporting sleep paralysis. Specifically, excessively short (< 6 hours) [24,28,32] or long (> 9 hours) [24,28] sleep duration and napping [32], especially long naps (> 2 hours) [24,28], were associated with increased odds of sleep paralysis. Long self-reported sleep latency (> 30 minutes) and difficulty initiating sleep were related to an increased likelihood of reporting sleep paralysis [24,32]. Bedtimes also showed some relationship to sleep paralysis. With a bedtime of 10pm as the reference group, the odds of having ever experienced sleep paralysis (compared with never having experienced sleep paralysis) were significantly reduced in participants reporting a bedtime between 10pm and midnight, and significantly increased in participants reporting a bedtime later than midnight [24,28] Sleep paralysis episodes are most likely to occur when the individual is lying in the supine position (on the back) [53,54]. Despite this, it was shown that people who had reported experiencing at least one episode of sleep paralysis, as compared to those who had not, were no more likely to fall asleep lying in the supine position [53].
Distribution of sleep paralysis events show a weakly bi-modal distribution with 28% of episodes occurring within an hour after bedtime and 26% occurring five or more hours following bedtime [55]. Perhaps unsurprisingly, hypnagogic (when falling asleep) episodes of sleep paralysis were most prominent in the first hour after bedtime, hypnomesic (sometime during sleep) episodes most common between 1-3 hours after bedtime, and hypnopompic episodes were most common between 6-12 hours post bedtime [55]. Overall, hypnomesic episodes were most common [55].

Sleep disorders
Associations between sleep paralysis and other sleep disorders are shown in Table 4. In a large population-level study, insomnia disorder was not associated with increased prevalence of sleep paralysis, however non-restorative sleep, a common symptom of insomnia [1], was associated with increased sleep paralysis rates (33% vs 13%) [20]. In a further study, disrupted sleep (defined as waking up repeatedly during the night) was a significant predictor of sleep paralysis, with those reporting disrupted sleep more than five nights per month being three times more likely to experience sleep paralysis compared to those reporting less than five nights of disrupted sleep per month [56]. One study reported increased sleep paralysis in obstructive sleep apnoea (OSA) patients compared to controls (20% vs 5%) [57], however no independent association was found in a separate study [20]. In studies of narcoleptic populations, patients had a higher sleep paralysis prevalence than controls (59% vs 15%) [46]. Another study showed differing sleep paralysis prevalence rates in narcolepsy patients with varying concentrations of cerebral spinal fluid (CSF) hypocretin 1 [58], which is a biomarker for narcolepsy [59]. Those with below 200 pg/ml concentration exhibited more frequent episodes of sleep paralysis (52%) than those with a normal concentration of CSF hypocretin-1 (33%). Despite the increased frequency of episodes, there were no significant differences in episode severity between groups [58]. In a different study also with narcolepsy patients, those who also experienced idiopathic hypersomnia showed no significant differences in sleep paralysis frequency compared to narcolepsy without hypersomnia [60]. Prevalence of sleep paralysis did not differ between two groups of idiopathic hypersomnia patients, when comparing patients who either did or did not have a long sleep time [61].
Using minimal International Classification of Sleep Disorders 90 (ICSD-90) criteria [62], and a large population-level sample [20], those with sleep paralysis, as compared to those without, showed higher rates of a number of dyssomnias (disorders of getting to sleep and staying asleep) and parasomnias (disorders of abnormal/unusual behaviours during sleep).
Ohayon and colleagues found that no disorder other than nocturnal leg cramps predicted sleep paralysis frequency in a multiple predictor model [20,22].

Symptoms of psychiatric illness
Relationships between symptoms of psychiatric illness and sleep paralysis are shown in Table 5.
In non-clinical samples, multiple studies have examined the links between sleep paralysis and general mental health problems. The latter have been indexed by scales such as the mental health short-form, which gives an index of a person's quality of mental health. These studies have shown that poorer general mental health is significantly associated with frequency of sleep paralysis [18,22,24,25,28,42], although two studies of Nigerian health workers did not find a relationship [26,27].
Mirroring the clinical findings, anxiety-related symptoms appear to be associated with sleep paralysis [17,38,39], though this was not found in all studies [56]. Furthermore, anxiety sensitivity (fear of the physical symptoms of anxiety) showed a small but significant relationship with the presence of sleep paralysis in one study [19]. A second showed anxiety sensitivity to be significantly correlated with lifetime and recurrent fearful sleep paralysis [29]. Social anxiety symptoms also appear to be related to sleep paralysis, with a study finding participants who experienced sleep paralysis with a 'felt presence' hallucination showing higher levels of social anxiety than participants who experienced sleep paralysis without hallucinations [63]. Felt presence hallucinations during sleep paralysis have also been related to the amount of distress associated with an episode of sleep paralysis. The level of distress related to sleep paralysis episodes was associated with dysfunctional social imagery [64].
Higher scores on self-report measures of depressed mood have been associated with sleep paralysis frequency [29,36,37]. It has also been shown that this relationship remains significant after anxiety symptoms have been controlled for [56]. However, other multivariate studies assessing multiple predictors (such as depressed mood, anxiety symptoms, alcohol intake, threatening events, sleep quality, dissociative experiences, daydreaming, and sensory imagery) did not find depressed mood to independently predict sleep paralysis [17,39].

Psychiatric disorders and medication use
Results focusing on sleep paralysis in relation to psychiatric illness and medication are displayed in Table 6. Patients with a diagnosis of post-traumatic stress disorder (PTSD) in American, Cambodian, and Chinese samples all showed higher prevalence of sleep paralysis (between 65-100%) compared to healthy controls (20-25%) [65][66][67]. In a study focusing on fearful isolated sleep paralysis in patients with panic attacks, individuals who also were diagnosed with PTSD were significantly more likely to experience recurrent fearful isolated sleep paralysis compared with patients with panic attacks but without PTSD [29]. No differences were found with regards to lifetime isolated sleep paralysis in those with and without PTSD. Significant positive correlations between PTSD symptoms and intruder, incubus, and vestibular-motor (V-M) hallucination frequency and intensity during sleep paralysis episodes were found in one study [35]. This suggests that PTSD is associated with both the frequency and intensity of sleep paralysis episodes, though a separate study did not find a link between PTSD and fear/distress experienced during sleep paralysis [19]. In a sample of 142 African Americans, whilst both current (25% vs 20%) and lifetime (56% vs 48%) PTSD diagnosis rates were numerically higher in participants with sleep paralysis, this study did not find statistically significant differences [37].
Significantly higher prevalence rates of sleep paralysis in patients with panic disorder were also found. Lifetime prevalence rates of 56-59% were reported for panic disorder patients which was significantly higher than those without panic disorder [30,67], whose prevalence was approximately 19% [67], though not all studies found this difference [37]. Another study found 31% of participants who had experienced sleep paralysis met diagnostic criteria for clinical or subclinical panic disorder, whilst none of those who did not experience sleep paralysis met diagnostic criteria [19]. In a study comprising African American and Caucasian participants, patients with panic disorder were significantly more likely to experience recurrent sleep paralysis (defined as 4+ episodes per year) (36%) as compared to patients with other anxiety disorders, such as obsessive-compulsive disorder, social phobia, and generalised anxiety disorder (11%) [30]. In a further study, individuals with frequent sleep paralysis were more likely to have experienced panic attacks than those with less frequent/no sleep paralysis, though exact prevalence rates were not reported [40]. In other studies focusing on panic attacks, sleep paralysis was more common in individuals with panic attacks (32-49%) compared to those without (10-15%) [31,37].
Studies that have considered anxiety disorders more generally have shown that participants with self-reported anxiety disorders were significantly more likely to experience sleep paralysis than those without [31]. However, other studies have failed to find anxiety disorders to be a significant predictor of sleep paralysis [20,29]. One study found slightly higher prevalence rates for patients with panic disorder (21%) and social phobia (22%) compared with generalised anxiety disorder (16%), though differences were not significant (this may be due to the low sample size, N = 62). This study did however find a significant association between the presence of sleep paralysis and a co-morbid anxiety disorder (35%), compared to being diagnosed with a single anxiety disorder (11%) [23]. Similarly, another study found that the number of fearful sleep paralysis episodes was positively correlated with the total number of anxiety and mood disorder diagnoses [29].
There is less evidence for an association between sleep paralysis and depression.
Specifically, sleep paralysis prevalence was not significantly higher in a group meeting diagnostic criteria for depression (31%) compared with a group who did not meet the criteria (28%) [37]. In a study of outpatients with anxiety disorders, a comorbid depression diagnosis was not associated with an increased prevalence in sleep paralysis compared to patients without comorbid depression [23]. In a population level survey of over 4000 individuals, depression (as assessed by DSM-IV criteria) was not significantly related to sleep paralysis in a multiple predictor model (including age, physical health, sleep problems, bipolar disorder, and psychiatric medication) [20]. One study however did find a significant correlation between depression and recurrent fearful sleep paralysis, with both lifetime and lifetime fearful sleep paralysis showing no relationship to depression [29]. There has been just one study looking at bipolar disorder and sleep paralysis. Here is was found that bipolar disorder occurred more frequently in a severe sleep paralysis group, defined here as at least one episode per week (19%), compared with a no sleep paralysis group (2%) [20]. Sleep paralysis has been associated with a lifetime, but not current, diagnosis of alcohol/substance abuse disorder [37].
There is little evidence regarding the association between psychiatric medication use and sleep paralysis (see Table 6). Whilst one study found higher use of anxiolytic medication in individuals with weekly sleep paralysis as compared to those without [20], the reasons for this  [20,23,29,56] and hypnotic medications [20] was also found not to be related to sleep paralysis frequency.

Discussion
This review set out to investigate variables associated with episodes of sleep paralysis. Whilst strengths in the current literature include a wide range of variables investigated and a diverse array of samples, plus a number of relatively large-scale studies, limitations currently exist that should be addressed in future research.
As described in the introduction, a great deal of variation in terms of measures to assess sleep paralysis was noted (as shown in Table 1). Furthermore, very few studies reviewed provided any evidence regarding the psychometric validity/reliability of their chosen measure.
The Waterloo unusual sleep experiences questionnaire (WUSEQ) has been shown to have a reliable three-factor structure [2,10] and internal reliability (as assessed by Cronbach's alpha) was shown to be good in one study that reported it [39]. Furthermore, research to date has relied on retrospective reporting. As many studies assessed lifetime prevalence of sleep paralysis, the reliability of such measures is questionable. Some studies using a prospective measure of sleep paralysis have found a similar factor structure of hallucinations to when retrospective measures are used [68,69]. However, no studies to date have assessed associated variables using a prospective design. This is an important area for future research as it will allow for a more accurate assessment of sleep paralysis episodes and their predictors. Again, studies investigating these issues would be useful in better refining our understanding of which factors precipitate the occurrence of episodes. For example, one prediction could be that anxiety is related to sleep paralysis via an indirect effect of poorer sleep quality [70], which in this instance would be acting as a mediator. Despite these limitations, the research reviewed here has provided important insights into variables associated with sleep paralysis. For example, rates of sleep paralysis do appear to be elevated in a number of sleep disorders, especially narcolepsy [46,58]. Studies also suggested an increased prevalence in obstructive sleep apnea patients [57], nocturnal leg cramps [20], as well M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 24 as those suffering from insomnia symptoms [20,56]. Interestingly, one study that was not included in this review because of a lack of a control group found a sleep paralysis prevalence rate of 14% in patients with Kleine-Levine syndrome [71] (a sleep disorder characterised by episodes of severe hypersomnolence in association with cognitive, psychiatric, and behavioural disturbances [1]). Without a control group it is hard to say whether this prevalence is increased relative to controls. In the only study to assess the prevalence of multiple sleep disorders in those with or without SP, few links were found [20]. However, it should be noted that diagnoses were made by a trained computer system and not by a sleep medicine clinician. In the clinical setting, sleep paralysis often occurs within the context of disturbed sleep. Preliminary experimental work into the link between SOREMPs and sleep paralysis, and self-report measures of poor sleep quality, support this idea [17,18,24,28,32,39,42,50].
Certain psychiatric groups showed higher rates of sleep paralysis compared to controls, especially patients with PTSD [29,[65][66][67]. Whether sleep paralysis is the cause or the consequence of PTSD is unclear, however some studies focused on patients who had experienced trauma during the Cambodian civil war suggest that sleep paralysis appears as a symptom of the PTSD [65,66]. Qualitative evidence also suggests that the hallmark symptom of 'flashbacks' present in PTSD may actually manifest as hallucinations occurring during sleep paralysis [65,66]. It is also now well-documented that PTSD patients often suffer from 'disruptive nocturnal behaviours' (DNB) such as trauma-related nightmares [72]. Future research should consider DNBs alongside sleep paralysis to ascertain whether these two sleep phenomena are closely linked in this illness. Anxiety disorders generally, and panic disorder in particular, also appear to be associated with a higher prevalence of sleep paralysis [19,23,30,31,40,67], with preliminary evidence suggesting that socially anxious individuals who also experience felt M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
25 presence hallucinations during wakefulness are more likely to experience sensed presence hallucinations during sleep paralysis, compared to socially anxious individuals who do not experience felt presence hallucinations during wakefulness [63]. These findings are especially important for clinicians, and suggest that experience of sleep paralysis should be more routinely assessed within certain psychiatric groups. No studies reviewed looked at associations between sleep paralysis and schizophrenia. However one study estimated a prevalence of 15% in schizophrenia patients, though they did not have a control group [73]. This prevalence does not seem abnormally different to prevalence estimates in healthy samples [11], though differential diagnosis between these two conditions is an important consideration [74].
There is still very little work into successful interventions for reducing and/or improving outcomes for individuals suffering from severe cases of sleep paralysis. We did not find clear evidence that sleep paralysis frequency was associated with any psychiatric medications [20,23,29,56]. However, it should be noted that none of the studies on this topic had a primary aim of investigating the efficacy of any particular medication. No studies reviewed systematically investigated whether a specific medication led to a reduction in sleep paralysis episodes. Rather, these studies looked to see whether taking medications predicted sleep paralysis in a cross-sectional design, with minimal information regarding why the medication was being taken, how long the medication was being taken, and what the exact dosages were. It is currently unclear whether these substances are or are not effective in reducing sleep paralysis symptoms, as suitable studies to test these hypotheses are yet to be performed. This is a clear area for future research. Research into treatment options for narcolepsy have not yielded many insights into potential treatments for sleep paralysis. A 2007 report by the AASM [75] suggested sodium oxybate as an effective for treatment of cataplexy, daytime sleepiness, and disrupted M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
26 sleep due to narcolepsy, but evidence for its effectiveness in treating sleep paralysis was mixed and based on lower quality evidence. One double-blind, placebo-controlled trial looking at sodium oxybate for treating narcolepsy found changes in sleep paralysis did not differ from placebo [76]. A follow-up study by the same group did however find a 6g dosage sodium oxybate to lead to a reduction in episodes [77]. With regards to selective serotonin reuptake inhibitors (SSRIs) and other antidepressant medication, the report concluded that this recommendation is based on anecdotal clinical experience [75]. Many clinicians do prescribe antidepressant medications (such as SSRIs) to those with narcolepsy to help alleviate both symptoms of cataplexy and sleep paralysis. The hypothesised mechanism of action is to enhance postural muscle tone during REM sleep, the stage of sleep most associated with sleep paralysis [74]. Despite this reasonable theoretical link between SSRIs and sleep paralysis, more studies need to be conducted in the future on this important topic. It has been noted that sensory stimulation may help end sleep paralysis episodes, but may be even more unpleasant than the sleep paralysis itself [78].
This review suggests a couple of areas where potential interventions could be targeted.
As factors such as anxiety and stress appear to be linked with sleep paralysis [17,19,38,39,43], techniques aiming to reduce levels of these factors may also help in alleviating sleep paralysis episodes. Cognitive behavioural therapy (CBT) has been shown to be useful in reducing levels of anxiety [79] and coping with stress [80], and has also been shown to be effective in various sleep disorders, particularly insomnia [81,82]. A CBT for isolated sleep paralysis manual now exists as a promising therapy [3], however systematic evidence for its effectiveness is still lacking.
Poor subjective sleep quality was shown to be consistently related to sleep paralysis [17,18,24,28,32,39], suggesting that efforts to improve the quality of an individual's sleep could M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
27 help in reducing sleep paralysis episodes. The importance of sleep hygiene in various parasomnias, including sleep paralysis, has been suggested before [83]. Sleep hygiene education is intended to improve sleep initiation and maintenance, through interventions such as adjusting bedroom temperature, avoiding alcohol close to bedtime, and going to bed and waking up at similar times each day [3,84]. Based on the evidence reviewed here, future work into sleep hygiene education and sleep paralysis is warranted [3].
Very little experimental work into the underlying neurophysiology of sleep paralysis has been performed. It does appear however that episodes are linked with sleep disruption and are particularly associated with SOREMPs [42,50]. Preliminary genetic evidence also suggests a link with altered circadian rhythms [17]. At a neurobiological level, the paralysis associated with REM sleep is believed to be regulated by the GABA and glycine neurotransmitter systems, which are important in the inhibition of motor neurons during REM sleep, so contributing to the muscle atonia [85]. What is currently unknown is why some people regain waking levels of consciousness during REM sleep resulting in an episode of sleep paralysis. Sleep paralysis has been suggested to be a dissociated state, featuring both REM and waking type brain activity [42,50,[86][87][88], though see [89]. Understanding how and why this unique state of consciousness arises is an important topic for future research, not only because it will further our understanding of sleep paralysis, but also because heighted levels of awareness during REM sleep have been suggested as a potential way of studying the emergence of consciousness [90][91][92].

Practice points
• Sleep paralysis should be considered by clinicians when diagnosing sleep disorders, especially in narcolepsy and those exhibiting symptoms of insomnia.
• Sleep paralysis should be assessed more frequently in certain psychiatric groups, particularly those presenting with PTSD and anxiety disorders.      Body mass index [29] Significantly correlated with lifetime (r = .21) and recurrent (r = .23) fearful SP [20] No association [32]     Sleep starts [20] Prevalence of sleep starts in individuals with SP vs those without SP, 7.1% vs 1.5%, ns in a multiple predictor model SP = sleep paralysis, OR = odds ratio, CI = 95% confidence intervals, M = mean, SD = standard deviation, ns = not significant. Article reference numbers are shown in square brackets.

61
Diagnosis of an anxiety disorder [31] -SP more frequent in those with an anxiety disorder vs no anxiety disorder [29] -correlation between SP and interview measures of anxiety ns [20] -Ns in a multiple predictor model Co-morbid anxiety disorder