Temporal dynamics of awakenings from slow‐wave sleep in non‐rapid eye movement parasomnia

Non‐rapid eye movement parasomnia disorders, also called disorders of arousal, are characterized by abnormal nocturnal behaviours, such as confusional arousals or sleep walking. Their pathophysiology is not yet fully understood, and objective diagnostic criteria are lacking. It is known, however, that behavioural episodes occur mostly in the beginning of the night, after an increase in slow‐wave activity during slow‐wave sleep. A better understanding of the prospect of such episodes may lead to new insights in the underlying mechanisms and eventually facilitate objective diagnosis. We investigated temporal dynamics of transitions from slow‐wave sleep of 52 patients and 79 controls. Within the patient group, behavioural and non‐behavioural N3 awakenings were distinguished. Patients showed a higher probability to wake up after an N3 bout ended than controls, and this probability increased with N3 bout duration. Bouts longer than 15 min resulted in an awakening in 73% and 34% of the time in patients and controls, respectively. Behavioural episodes reduced over sleep cycles due to a reduction in N3 sleep and a reducing ratio between behavioural and non‐behavioural awakenings. In the first two cycles, N3 bouts prior to non‐behavioural awakenings were significantly shorter than N3 bouts advancing behavioural awakenings in patients, and N3 awakenings in controls. Our findings provide insights in the timing and prospect of both behavioural and non‐behavioural awakenings from N3, which may result in prediction and potentially prevention of behavioural episodes. This work, moreover, leads to a more complete characterization of a prototypical hypnogram of parasomnias, which could facilitate diagnosis.


| INTRODUCTION
Patients with non-rapid eye movement (NREM) parasomnia disorders, also called disorders of arousal, exhibit nocturnal behavioural manifestations arising from slow-wave sleep (SWS; Broughton, 1968).The International Classification of Sleep Disorders (ICSD-3) describes these manifestations as recurrent episodes of incomplete awakening with absent or inappropriate responsiveness, limited or no cognition or dream report, and partial or complete amnesia (American Academy of Sleep Medicine, 2014).Based on the type of behaviours, different sub-disorders are distinguished, with confusional arousals (CA), sleep walking (SW; somnambulism) and sleep terrors (ST) being the most prevalent (Idir et al., 2022).
Diagnosis is mostly based on the history of behavioural events at home, sometimes in combination with a clinical polysomnography (PSG).The latter can confirm the suspicion for a NREM parasomnia based on the presence of behavioural manifestations from SWS, but is mostly used to rule out other disorders like sleep-related hypermotor epilepsy (SHE) or rapid eye movement (REM)-behaviour disorder (American Academy of Sleep Medicine, 2014).Because behavioural episodes are not always caught during a single night of PSG, Pilon et al. (2008) formulated a protocol based on sleep deprivation and auditory stimuli to trigger episodes in predisposed adults.However, this protocol requires additional labour-intensive procedures for applying auditory stimuli.In a research setting it has limitations because it does not contribute to enhanced understanding of the underlying pathophysiology of spontaneous events, while such an understanding is vital to eventually formulate diagnosis protocols based on a normal PSG.
Parasomnia episodes have been reported to occur more frequently in the beginning of the night (American Academy of Sleep Medicine, 2014), but it is generally unknown which factors underpin this observation.The large amount of SWS in the beginning of the night could be a cause, but it does not necessarily have to be the only one.Pressman (2007) also observed that purely the presence of more SWS is not the only factor that induces behavioural episodes.An increased probability for an awakening from SWS in the beginning of the night could also explain this phenomenon, for example.
Apart from effects over the course of the night, temporal effects preceding parasomnia episodes have also been found.Behavioural episodes were observed to be preceded by a gradual increase in slow-wave activity (SWA) in frontal and central areas on a timescale of tenths of seconds to minutes prior to the episode, followed by a steep increase within the last seconds (Camaioni et al., 2021;Cataldi et al., 2022;Desjardins et al., 2017;Espa et al., 2000;Guilleminault et al., 2001;Jaar et al., 2010).Moreover, these behavioural awakenings were found to be preceded by higher SWA than non-behavioural awakenings from the same sleep stage (Espa et al., 2000;Perrault et al., 2014).It has, however, not been investigated how the duration of a SWS period affects behavioural episodes.
We hypothesized that variables related to timing both across and within sleep cycles could play a role in the occurrence of parasomnia episodes.To investigate this hypothesis, we analysed per sleep cycle the probabilities of being in SWS, ending a SWS period, and subsequently transitioning to a behavioural or a non-behavioural awakening.Our probabilistic approach enabled to objectively separate factors that may underpin a behavioural episode.Additionally, we analysed the relation between N3 bout durations and the prospect of waking up, either with or without a behavioural manifestation.The SOMNIA database includes patients from a heterogeneously sleep-disordered population that were scheduled for a routine diagnostic video-PSG, performed according to the recommendations of the American Academy of Sleep Medicine (AASM; Troester et al., 2023).We included n = 52 patients with an ICSD-3 diagnosis of at least one of the following: CA; SW; or ST.In practice, patients are often diagnosed with a combination of these sub-disorders, which are often believed to exhibit the same disorder on a continuum, rather than being three independent disorders (Derry et al., 2009).For that reason we considered CA, SW and ST together, instead of analysing them separately.Exclusion criteria were: (1) age < 18 years; (2) medication that disturbs sleep structure (e.g.hypnotics, antipsychotics, antidepressants); (3) presence of neurological or psychiatric disorders; and (4) other severe sleep-disturbing comorbidities.Minor comorbidities such as mild periodic limb movements were present in 16 subjects, and were checked to not affect sleep structure or (behavioural) awakenings from SWS. Summarizing statistics of our patient cohort can be found in Table 1.

| Healthy controls
Healthy controls were selected from the Healthbed study (van Meulen et al., 2023), which includes one clinical video-PSG recording per subject, recorded at Sleep Medicine Center Kempenhaeghe (Heeze, the Netherlands) according to the AASM guidelines (Troester et al., 2023).The Healthbed study protocol was approved by the medical ethics committee of Maxima Medical Center (Veldhoven, the Netherlands, W17.128), and the data analysis protocol for our study was approved by the institutional review board of Sleep Medicine Center Kempenhaeghe (November 2019).All patients included in the Healthbed study provided written informed consent.We included all subjects aged between 18 and 50 years (n = 79).Summarizing statistics can be found in Table 1, which shows that the statistics were very similar in both cohorts.

Hypnogram
Certified sleep experts from Sleep Medicine Center Kempenhaeghe manually annotated all PSG recordings with AASM sleep stages per 30 s (Troester et al., 2023).

Behavioural versus non-behavioural awakenings
All N3-W transitions in the hypnograms of the patient cohort were accompanied by a detailed description of the behavioural manifestations visible on the video, if any.This was done by certified sleep experts from Sleep Medicine Center Kempenhaeghe.We subsequently categorized each description in a behavioural and a nonbehavioural category.N3-W episodes without any visible action were included in the non-behavioural category.As sleep-related movements have also been reported in healthy controls (Loddo et al., 2018), we also included mild sleep-related movements that are not specific to NREM parasomnias in the non-behavioural category, examples being: opening of eyes, turning in bed, touching of the face, and stretching out.Episodes categorized as behavioural events included at least one of the following actions in various complexities: tilting the head, watching around or looking at something, body movements other than only turning in the bed, any form of talking, and reactions linked to emotion (fear, panic, laughter, surprise, confusion).
The motor events in the aforementioned behaviours all fall into one of the three sub-categories of motor events in NREM parasomnias as categorized by Loddo et al. (2018), being simple arousal movements (SAMs), rising arousal movements (RAMs), and complex arousal with ambulatory movements (CAMs).

Sleep cycles
Sleep cycle detection was done manually, but supported by an automated algorithm to provide an initial prediction.The algorithm used as a base criterion that at least 15 min of NREM sleep was followed by at least 1 min of consecutive REM sleep.It was noted that two (or sometimes three) N3 bouts in the beginning of the night were not always separated by a REM episode.We thus relaxed the REM criterion in the first 2 hr after sleep onset.If no REM sleep was scored here, and the subject showed N3 bouts (duration ≥ 1 epoch) with a scored awakening in between, then this awakening was considered the end of a sleep cycle as well.Additional sleep cycles detected in this way were required to take at least 45 min to prevent detection of many cycles when multiple N3-W transitions occurred.
The output of the algorithm was manually inspected by two authors, and in case of doubt a third author was consulted.This resulted in a manual update in 30% of the nights, which mostly concerned the start of the second sleep cycle when a REM bout was missing after the first cycle.Sleep cycles were defined prior to all analyses, and not updated afterwards anymore.Figure 1 shows a hypnogram of one of the patients, with annotations for sleep cycles and N3-W transitions.Appendix A.1 provides for both cohorts the number of transitions from N3, split per sleep cycle.

| Probabilistic analyses
To further characterize transitions from N3, we took a probabilistic approach.
Such decomposition is typically read from right to left.That is, part I denotes the probability that the N3 bout ends (i.e.transitioning to a non-N3 sleep stage while having been in N3), and part II is the probability of transitioning to W B , given the N3 bout had ended.
Lastly, the probability of being in N3, defined as the proportion of epochs scored as N3, also affects the probability for an N3-W B transition.The following holds: where p N3 ½ is the probability of

| Duration of SWS periods
Next to looking at differences across sleep cycles, intra-cycle effects were also investigated.We focused on the bout length of SWS periods, defined as the number of minutes a patient consecutively  2).This effect was present in all cycles (Figure 3d, top).An opposite, but nonsignificant, effect was found for N3-N2 transitions per night, and per hour of SWS (Table 2), which also generalized to most cycles (Figure 3d, bottom).
Figure 3 visualizes the probabilities as introduced in Equations ( 1) and ( 2 In the first cycle, patients had a higher probability of ending an N3 bout (0.07 [0.05-0.12]versus 0.05 [0.03-0.08];U = 271, p < 0.01; Figure 3b), and had a significantly lower probability for being in N3 (0.28 [0.19-0.39]versus 0.39 [0.27-0.51];U = 1459, p < 0.01; Figure 3a).These effects reduced in the second cycle to nonsignificant differences, and disappeared later in the night.Both cohorts showed a decreasing probability of being in N3 over cycles (Figure 3a), and a slight trend of an increasing probability of ending an N3 bout over cycles (Figure 3b).

| Duration of SWS periods
Figure 4 shows the duration of N3 periods prior to an awakening (either with or without a behaviour in the patient group).In sleep cycle 3 and later, no significant differences were found (H = 0.27; p = 0.87).
However, in the first two cycles, the N3 bouts that preceded a nonbehavioural awakening in the patient group were significantly shorter than the bouts prior to a behavioural awakening in the same group (5.8 [2.1-9.5]versus 10.5 [5.5-17.5];U = 1600, p < 0.001), and shorter than the N3 bouts that preceded an awakening in the control group (18.0 [5.5-30.0];U = 1150, p < 0.001).Bouts preceding behavioural episodes in patients were, on average, shorter than the N3 periods prior to awakenings in the control group, but the difference was not significant after Bonferroni correction (U = 3755; p = 0.02).For completeness, Figure A1 in Appendix A.3 shows the same plot for the N3 bouts that preceded transitions to N2.No significant differences were found between cohorts.
Figure 5 shows the probability of an awakening, given the N3 bout ended after a certain duration.N3 bouts of less than 5 min were T A B L E 2 Median and interquartile range of the number of N3-W and N3-N2 transitions over subjects per night, and per hour spent in N3.
followed by an awakening in 17% and 6% of the time in patients and controls, respectively.A transition to N2 after such short N3 periods was more common (Figure A2 in Appendix A.3). Longer N3 bouts were more often followed by an awakening in both cohorts.
Bouts that lasted at least 15 min resulted in an awakening in 73% of the time in the patient cohort, which was the case for 34% of these bouts in the control group.
In fact, the probability for an awakening grew steadily with N3 bout duration in the patient cohort (Figure 5

| DISCUSSION
We investigated the temporal dynamics of awakenings from SWS in NREM parasomnia patients and compared those with healthy controls.Patients showed more awakenings from SWS per night, and per hour of SWS, which is in line with earlier observations (Espa et al., 2000;Gaudreau et al., 2000;Lopez et al., 2014).This higher prevalence of N3 awakenings generalized to all sleep cycles, and was in particular caused by a higher probability of waking up when the N3 bout ended.
The difference in N3-W prevalence was most predominant in the first sleep cycle, where N3 instability played an additional role, as patients showed a significantly larger probability to end an N3 bout.
This most likely caused the lower probability of being in N3 that patients showed in this cycle, which was defined as the fraction of the cycle that was scored as N3.The fact that patients spent a lower proportion of their first cycle in N3 sleep may explain the lower cycleaverage SWA that has earlier been reported for the first cycle in NREM parasomnia patients, compared with controls (Espa et al., 2000;Gaudreau et al., 2000;Guilleminault et al., 2001Guilleminault et al., , 2006)).
Controls showed a larger probability for N3-N2 transitions for all sleep cycles but overnight, and corrected for hours of SWS, the effect was found not significant (Table 2).This suggests that the N3-W transitions in patients do not fully replace N3-N2 transitions, but happen additionally.
The probability for an N3-W transition reduced over cycles in both cohorts, for which the reduction in probability of N3 over cycles was found to be the largest contributor.In patients, after the first two cycles a slight decrease was also visible in the probability for waking The probability of being in N3 reduced over cycles in both cohorts (medians and interquartile range shown), but in the first cycle it was significantly lower in the patient group.(b) The probability of ending an N3 period was significantly higher in patients than controls in the first cycle, and showed an increasing trend over cycles.(c) After an N3 bout had ended, patients showed a higher probability to wake up (top), and a slightly lower probability to transition to N2 (bottom).The ratio of behavioural versus non-behavioural awakenings in the patient group decreased over cycles (top).(d) The combination of the factors displayed in (a-c) caused a higher probability of having an N3-W transition for patients in all cycles compared with controls, with a declining occurrence over cycles (top).*Significance level of α = 0.05/5.

F I G U R E 4
In the first two cycles, non-behavioural awakenings in the patient group were preceded by significantly shorter N3 bouts than behavioural awakenings in the patient cohort and nonbehavioural awakenings in controls (left).No significant group differences were found later in the night (right).*Significance at p < 0.001.
up after N3 ended.Not only the total number of N3-W transitions, but also the number of the behavioural N3-awakenings in the patient group decreased over cycles.This was mainly attributable to both the decrease in N3, and a decrease in the ratio of behavioural SWS awakenings versus non-behavioural awakenings.
Despite this decline of parasomnia episodes on the group level, some (but a minority) of the patients experienced all behavioural episodes only in the second part of the night, which is in line with earlier findings (Zucconi et al., 1995).The ICSD-3 (American Academy of Sleep Medicine, 2014), which states that behavioural episodes mainly occur in the first third of the night, should thus be taken with care for diagnosis purposes when conclusions are drawn based on a single night of PSG.Zadra et al. (2008) also investigated the ratio of SW episodes to non-behavioural arousals from SWS, and found that this ratio increased after sleep deprivation.Our findings in combination with the latter suggest that this ratio seems to relate positively to sleep pressure, which is high after sleep deprivation and in the beginning of the night.Sleep pressure is reflected in high SWA, which in turn was found to be higher prior to behavioural episodes compared with nonbehavioural awakenings from the same sleep stage (Espa et al., 2000;Perrault et al., 2014).Also, sleep deprivation protocols, which are known to incur a slow-wave rebound, have been found effective in provoking behavioural episodes in parasomnia patients (Januszko et al., 2016;Pilon et al., 2008).The decrement of SWA throughout the night may thus explain the reduced ratio between behavioural and non-behavioural awakenings that we found overnight (Espa et al., 2000;Gaudreau et al., 2000;Guilleminault et al., 2001Guilleminault et al., , 2006)).
In the patient cohort, we found a positive relation between the probability for an awakening from N3 and the duration of the preceding N3 bout.In the first two cycles this was purely caused by behavioural awakenings, while both types of SWS awakenings contributed to this effect later in the night.Previous works showed that SWA increased during the N3 bout in NREM parasomnia patients, until the moment of (partially) waking up (Desjardins et al., 2017;Espa et al., 2000;Guilleminault et al., 2001;Jaar et al., 2010).The latter could imply that the longer the duration of a SWS period, the higher the build-up of SWA.As such, the SWA towards the end of a long N3 bout may be higher than in short bouts.If this is indeed the case, the hypothesis of Halász andcolleagues (2004, 2022) The probability that the N3 bout was followed by an awakening increased with bout duration in the patient group (top).In the beginning of the night mainly the behavioural awakenings contributed to this effect (bottom left), while later in the night the difference between both types of N3 awakenings was less pronounced (bottom right).Nobili et al., 2011;Terzaghi et al., 2009Terzaghi et al., , 2012;;Zadra et al., 2013).In later sleep cycles, SWA is known to have a lower amplitude (Gaudreau et al., 2000).We might thus speculate that in that period of the night, an arousal might more easily invoke a full awakening, irrespective of the time spent in N3.This could explain why much smaller or no differences were found between behavioural episodes and full awakenings in cycle 3 and later, and why the ratio of behavioural versus non-behavioural awakenings from N3 reduced after the third cycle.
From CAP analyses, it was reported that 60% of the behavioural events occurred during a CAP cycle, and all of those events happened during phase A (Zucconi et al., 1995).Moreover, sub-phase A1 in a CAP cycle relates closely to hyper-synchronous delta activity (Parrino et al., 2006(Parrino et al., , 2012)), which in turn has often (but not unambiguously) been linked to behavioural episodes in NREM parasomnias (Camaioni et al., 2021).Now the question arises whether the prevalence of behavioural events may not only be influenced by time in N3 and the sleep cycle in the night, but whether an interaction effect may exist with being in a CAP cycle or not, and the specific CAP phase.This question opens new avenues for future research.
In our analyses, we defined an N3 bout as a consecutive period of scored N3 epochs.In practice it could have included arousals that did not induce a sleep stage switch in the scored hypnogram.As a consequence, we have excluded behavioural events that arose from an arousal without awakening in N3 (this was only reported twice in our cohort).Note, however, that arousals that lasted more than 15 s were always scored as Wake (Troester et al., 2023), and thus included in the analyses.By redefining the end of an N3 bout as the moment either a non-N3 stage was scored, or an arousal happened while the scoring remained N3, could make N3 duration a better proxy for consolidation of SWS, and could strengthen the relation we found between N3 duration and the probability for behavioural episodes.
However, arousal annotations are not made for every clinical PSG night as it is not strictly recommended by the AASM, and is timeconsuming.So possible findings based on arousal annotations might teach us more about NREM parasomnia disorders, but might be more difficult to directly apply in clinical practice for diagnoses purposes.
Behavioural episodes in NREM parasomnia patients not strictly occur around N3-W transitions only, and may also (although less frequently) originate from N2, or result in N1 or an arousal with behaviour (as discussed above).To simplify our model, we only included N3-W transitions, but the presented concept can be extended to other transitions with behavioural manifestations, especially when a large-scale data set is available that contains a multitude of those and/or contains arousal annotations.Such analyses could result in a more complete view and understanding of NREM parasomnias in the future.
Our data included both spontaneous and triggered awakenings from N3, which is in line with earlier observations (Derry et al., 2009).
Triggers had either internal causes (e.g. a periodic leg movement) or they were unintended external triggers (e.g. a sound in the hallway).
We did not exclude triggered events, as Cataldi et al. (2022) found similar spatio-temporal dynamics in the electroencephalogram between provoked and spontaneous episodes.Moreover, it cannot be determined with certainty that spontaneous events were truly spontaneous, and that they did not get triggered by internal mechanisms that were simply not measured or measurable by the PSG.
A gold-standard for sleep cycles is unavailable, which may have caused subjectivity in our labelling.We aimed to standardize the procedure as much as possible by using automation to provide an initial indication.Yet, the sleep cycle detector we used was not designed to be self-contained and should not be used without a manual confirmation of its predictions.
Despite efforts to find suitable categorizations of the behaviours in parasomnia episodes, no standardized procedure exists today (Arnulf et al., 2014;Derry et al., 2009;Joncas et al., 2002;Loddo et al., 2018Loddo et al., , 2019)).Loddo et al. (2018) made a detailed categorization of motor events in NREM parasomnias, by categorizing events as SAMs, RAMs and CAMs.Nevertheless, we decided to binarize episodes (non-behavioural versus behavioural), rather than subcategorizing the behavioural events in the three proposed subcategories.Binarizing prevented small sample sizes for the RAM and CAM categories.Moreover, it resulted in high agreement between authors, removing additional subjectivity that could be introduced by using a more complex scale.Nevertheless, it is of interest for future studies whether a distinguishment of the behavioural events in more fine-grained sub-categories could shed further light on the timing and prevalence of behavioural events in NREM parasomnias.The PSGs of the control group were scored according to the standard clinical AASM recommendations.Because no behaviours were reported, we did not classify the individual N3-W events separately in this group, and annotated all N3-W transitions as non-behavioural.
Given the remarkable resemblances found between NREM parasomnias and SHE, the disorders have been hypothesized to be a continuum of one and the other, rather than being fully distinct (Mutti et al., 2020;Halász et al., 2022).As such, differential diagnosis is often challenging, despite various efforts that have already been made to distinguish these disorders (Derry et al., 2006(Derry et al., , 2009;;Loddo et al., 2020;Montini et al., 2021;Pani et al., 2021).Following upon this work, it may, therefore, be of interest to investigate how the temporal dynamics of behavioural events as investigated in this work translate to a cohort of patients with SHE.
A trend in sleep medicine is seen towards the use of wearable sensors that facilitate multi-night ambulatory measurements (Arnal et al., 2020;Nakamura et al., 2017;Schneider et al., 2021)  The positive relation we found between the prospect of waking up from N3 and the duration of the N3 period opens new research avenues that investigate whether it will be possible to predict behavioural episodes on a minute scale.Forecasts at such a high temporal resolution could lead to automated waking protocols, which may be especially useful for patients that tend to harm themselves or others, and for whom these episodes cannot sufficiently be prevented with medication due to negative side-effects or unresponsiveness.

| CONCLUSION
We found clear differences in transition probabilities from SWS between NREM parasomnia patients and controls.Moreover, in the patient cohort, the behavioural episodes reduced over cycles, which was found to be attributable to both a reduction in N3 over cycles, and a reducing percentage of SWS awakenings that resulted in behavioural exhibitions.Differences between behavioural and nonbehavioural N3 awakenings were mainly found in the first two cycles, where the duration of the preceding N3 period differed between both types of awakenings.Figure A2 shows the probability of transitioning to N2, given the N3 bout had ended after a certain duration.N3 bouts shorter than 5 min transitioned with high probability to N2.In the patient cohort, this probability decreased the longer the N3 bout took, independent of the cycle.In controls this effect was also visible in the beginning of the night, but much less pronounced.
In the second half of the night, the probability for transitioning to N2 from N3 decreased from almost 1 to 0.4 in controls when the N3 bout took longer than 5 min, but for longer bouts the probability remained constant or even seemed to slightly increase again.
F I U R E A 1 Duration of consecutive N3 epochs followed by N2 in cycles 1 and 2 (left) or in the rest of the night (right).
F I G U R E A 2 Probability for a transition to N2, given the N3 bout ended, split for the first two cycles and the rest of the night.
Data 2.1.1 | NREM parasomnia patients We used data from the Sleep and OSA Monitoring with Non-Invasive Applications (SOMNIA) database, collected at Sleep Medicine Center Kempenhaeghe, Heeze, the Netherlands (van Gilst et al., 2019).The SOMNIA study protocol was approved by the medical ethical committee of the Maxima Medical Center (Veldhoven, the Netherlands, N16.074).All patients included in the SOMNIA study provided written informed consent.The data analysis protocol for this study was approved by the institutional review board of Sleep Medicine Center Kempenhaeghe (March 2021).
We quantified the number of N3-W transitions in the hypnogram per night and per hour of SWS.For the parasomnia patients, N3-W T A B L E 1 Demographics and sleep parameters (mean ± standard deviations, and range for age) for both cohorts.CA (n = 13), SW (n = 14), ST (n = 4), CA + SW (n = 13), CA + ST (n = 5), SW + ST (n = 2), CA + SW + ST (n = 1) - transitions were split into awakenings with behaviour (W B ) and without behaviour (W NB ).The control group only exhibited nonbehavioural N3 awakenings, denoted with N3-W.Transitions to N2 are most likely from N3, while transitions to N1 and REM are less common (Appendix A.1).As such, we considered all N3-N2 transitions in both cohorts as well.To compare cohorts, non-parametric two-sided Mann-Whitney U-tests were used, including a Bonferroni correction to correct for multiple comparisons, setting a significance level of α = 0.05/4.
Figure 2 illustrates the possible transitions from N3, again with a distinction between behavioural awakenings (W B ) and nonbehavioural awakenings (W NB ).The vertical bar in probabilistic notation indicates a conditional probability, that is, p W B jN3 ½ denotes the probability of transitioning to a behavioural awakening in the next epoch, given that the current epoch is scored as N3.To get insight into the contributing factors of the given probability, it may be decomposed as follows: spent in N3 before transitioning to another sleep stage.Durations of the N3 bouts that preceded an awakening (either behavioural or nonbehavioural in the patient cohort), or a transition to N2, were F I G U R E 1 A hypnogram from the patient cohort.Dashed lines in the bottom indicate starts of sleep cycles, and stars denote N3-W transitions.F I G U R E 2 State switching diagram of sleep stage transitions from N3.We considered transitions to N2 and Wake (split in behavioural awakenings W B and non-behavioural awakenings W NB ).Each transition (i.e.arrow) is characterized by a conditional probability, for example, p W B jN3 ½ , which denotes the probability for a behavioural awakening, given the previous epoch was scored as N3.compared between both cohorts.Data from cycles 1 and 2 were grouped, and compared with data from cycle 3 and later.For the N3-W transitions, differences were tested using a non-parametric Kruskal-Wallis test with a significance level of α = 0.05.For the post hoc tests, pairwise two-sided Mann-Whitney U-tests were used with a Bonferroni correction, resulting in α = 0.05/3.For the N3-N2 transitions, a pairwise two-sided Mann-Whitney U-test was directly used to compare cohorts.We further investigated whether p Wjend N3 ½ and p N2jend N3 ½ depended on the duration of the preceding N3 bout.These analyses were again done for cycles 1 and 2 versus cycle 3 and later in the night.Bootstrapping (with 200 iterations) was used to create confidence intervals around the computed probabilities.To this end, a set of all N3 bouts within each cohort was created.This set was subsequently resampled 200 times with replacement, to create 200 (slightly varying) sets of N3 bouts.For each of these sets, the bouts were grouped based on their durations, and the conditional probabilities (i.e. p Wjend N3 ½ and p N2jend N3 ½ ) were computed.Grouping based on durations was done per 5 min, to ensure that multiple bouts were present per group.We report medians and 95% confidence intervals across the 200 sets.If not all of the bootstrap iterations contained at least one N3 bout of a certain duration within a cohort, statistics were not computed for that N3 duration in that cohort, preventing unreliable results due to lack of data.3 | RESULTS 3.1 | Transition dynamics from SWS Patients showed a significantly higher number of N3-W transitions per night compared to controls (4.5 [3.0-6.0]versus 2.0 [1.0-3.0];U = 3483, p < 0.001), also when corrected for hours of SWS (3.0 [2.1-4.5]versus 1.2 [0.6-1.7];U = 3523, p < 0.001; Table ), which decompose the probability for N3-W transitions and N3-N2 transitions.In the patient cohort, the probability of an awakening given that the N3 bout had ended was on average 0.36 in the first two sleep cycles, and reduced to 0.29 in later cycles.This probability ranged from 0.10 to 0.21 across cycles in the control group(Figure 3c, top).The ratio of behavioural versus non-behavioural awakenings from N3 in the patient cohort decreased from (on average) 70.3% in the first three cycles, to 58.1% and 22.2% in cycles 4 and 5, respectively(Figure 3c, top; Table A1 in Appendix A.1).
, top).This relation was more modest in the control group in the beginning of the night (Figure 5, left), and disappeared later in the night (Figure 5, right).In the first two sleep cycles of the patient cohort, only the behavioural awakenings contributed to the positive relation between N3 bout duration and probability of an awakening (Figure 5, bottom left).The probability of a non-behavioural N3-awakening remained more constant throughout the N3 bout.In the second part of the night, the prospect for both type of awakenings increased more similarly with N3 bout duration (Figure 5, bottom right).
may explain the positive relation we found between N3 duration and the probability for a behavioural awakening.They formulated that sleep mechanisms and arousal mechanisms are opposing forces that are continuously battling.When the arousal mechanism "wins", a full (non-behavioural) awakening happens.On the other hand, when both mechanisms continue to conflict due to high sleep pressure (i.e.high SWA) in certain brain areas, a partial awakening might occur, causing behavioural exhibitions(Bassetti et al., 2000;Castelnovo et al., 2022; prevent parasomnia episodes.Generalizing conclusions to multi-night recordings from wearable devices should, however, be done with caution, as (possibly automatic) sleep staging based on these data might result in less precise scoring of sleep instability and wake intrusions.
Figure A1 shows the duration of the N3 bouts that preceded a transition to N2 in both cohorts and in both parts of the night.No significant differences were found between the patient group and the controls in the first two cycles (2.0 [0.54.5] versus 1.5 [0.5-8.0];U = 65,777, p = 0.57) or in cycle 3 and later (1.25 [0.5-3.5]versus 1.0 [0.5-4.0]);U = 34,121, p = 0.54).