In the baseline experimental night, lights off was between 10:00 p.m. and 12:00 a.m. (midnight) and wake time between 6:00 a.m. and 8:00 a.m., depending on each participant's habitual sleep-wake cycle. After the baseline recording, sleepwalkers were instructed to go about their regular daytime activities but were forbidden from taking naps. They returned to the laboratory in the evening for the sleep deprivation protocol and remained under constant supervision for the night. Control subjects were submitted to the same protocol but remained in the laboratory under a constant routine condition for the diurnal period, as they were healthy controls for another sleep study as well. Recovery sleep was scheduled the following morning for all participants, 1 h after their previous wake time (i.e., following 25 h of wakefulness). Participants were prohibited from consuming alcohol, caffeine, or other stimulating substances the day prior to and during all laboratory procedures. Further details on this sleep deprivation protocol have been published elsewhere (10).

PSG recordings were conducted on a 32-channel Grass polygraph (sensitivity at 7 μV, bandpass at 0.3–100 Hz; Grass Instruments, Quincy, MA). Signals were relayed and digitized at a sampling rate of 256 Hz, and digitally filtered with an upper cut-off frequency of 100 Hz using a commercial software (Harmonie, Stellate Systems, Montreal, Canada). EEG recordings and electrodes placement were performed according to the international 10–20 system with a linked-ear reference and included electro-oculograms, submental electromyography, surface electromyography of the bilateral anterior tibialis, and continuous recording of 3-lead EKG (standard DI, DII and DIII). Respiration was also monitored using an oronasal canula and a thoracoabdominal plethysmograph, whereas oxygen saturation was recorded with a finger pulse oximeter. Sleep stages were scored according to standard criteria (27). A sleep cycle was defined as a NREM sleep period lasting at least 15 min followed by an REM sleep period lasting at least 5 min, except for the first REM episode which could have been shorter. Sleep time over the first two sleep cycles was calculated from sleep onset to the end of the second cycle. Sleep efficiency was the total sleep time over the first two cycles divided by the total sleep period from lights off to the end of the second sleep cycle. Slow-wave activity was averaged on C3 and C4 during the first two cycles and assessed as previously described (25).

Slow-wave sleep has been viewed as a stable and robust recording condition to assess HRV. In fact, the reduced inputs from external stimuli, limited bodily movements, being in a supine position, and stable and regular respiratory cycles that characterize slow-wave sleep are viewed as representing a unique opportunity to measure HRV indexes with few confounding factors (14).

The EKG was analyzed in a two-step process to obtain frequency-domain indexes of the HRV during slow-wave sleep, or N3 sleep stage. 5 min EKG segments with stationary signals were first selected blind by two of the authors (G.S. and A.A.B.) from slow-wave sleep during both baseline and recovery sleep. The 5 min length for selected segments was chosen as this duration was suggested to be the ideal time window to measure HRV during sleep to avoid artifactual or interruptive events and to facilitate standardization across studies (11). Selected segments were free of EEG arousals, movement artifacts such as leg movements during sleep, and respiratory events during and 40 s before and after each segment. Segments were still selected if a maximum of two N2 20-s epochs were present. However, no segments were selected containing epochs of REM sleep, and no segments were selected 5 min away from any REM sleep epoch. All appropriate 5 min segments corresponding to these criteria were selected from the first two sleep cycles. If more than one segment met the criteria within the same cycle, all were selected provided the segments were separated by 15 min or more in order to obtain a wider sampling of slow-wave sleep. Whenever more than one EKG segment was selected from the same sleep cycle, values for cardiac variables were averaged into a single score for that cycle. Each participant thus had a single score for each HRV variable for each cycle for both baseline and recovery sleep.

HRV was investigated using CardioLab (Fondazione S. Maugeri, Italy). Detection of R-wave peaks representing each heartbeat was first automatically detected and instances of arrhythmia or signal artifacts subsequently removed. Results from the software's automatic signal detection were then verified and confirmed by visual inspection. Frequency-domain components of the HRV assessed were absolute LF and HF bands, the LF/HF ratio, and normalized LF and HF bands (nLF, nHF) on total power. Frequency-domain components were quantified with an autoregressive decomposition algorithm to compute spectral peak powers, central frequencies, and classified into LF (0.04–0.15 Hz) and HF (0.15–0.4 Hz) bands. Because normalized measures are mathematically redundant and equal carriers of information with the LF/HF ratio when the denominator of their normalization is LF+HF, we opted for normalizing the bands on total power (from very low frequency and very high frequency) (28). The normalization was made to assess the relative contribution of the LF and HF bands to the total power and to remove the very large inter-subject variability that typically characterizes measures of spectral HRV (28).

Of note, although frequency-domain variables should be interpreted with caution as described above, for the purpose of this study, a shift in the LF/HF ratio or normalized bands will be referred to as a “predominance.”

Statistics were performed with SPSS version 25 (IBM, Armonk, NY). Group (sleepwalkers, controls) ^* Condition (baseline, recovery) ANOVAs were performed for sleep architecture and all HRV variables during slow-wave sleep in both the first cycle as well as the second sleep cycle, as a secondary goal was to evaluate whether HRV changes extend to the night's second period of slow-wave sleep. HRV variables in statistical models were transformed using natural logarithm to normalize the distribution as is commonly done for these metrics (23). Group ^* Condition ANOVAs were also performed for sleep architecture characteristics in the first and second sleep cycles grouped together. Significant interactions were decomposed using post-hoc comparisons. P-values <0.05 were considered significant. A false-discovery rate (FDR) correction was applied to Group ^* Condition ANOVAs for interactions with HRV variables.

A series of post-hoc analyses were performed to better characterize the association between HRV variables, markers of disrupted slow-wave sleep and elevated arousals, and sleepwalking episodes. First, correlations between interactions terms (Group ^* HRV variables during recovery slow-wave sleep of cycle 1) and sleep architecture and episodes were conducted to evaluate if groups differed in their associations with HRV and sleep characteristics. Second, we correlated interaction terms (slow-wave activity ^* HRV variables during recovery slow-wave sleep of cycle 1) with sleep architecture and episodes in sleepwalkers to assess whether HRV variables and slow-wave activity showed interactions in their association with sleep parameters associated with sleepwalking. Finally, we re-ran Group ^* Condition interactions while adjusting for slow-wave activity.

During baseline PSG recordings, 67% of sleepwalkers experienced a somnambulistic episode, with 50% of sleepwalkers having more than one. During recovery sleep, all sleepwalkers experienced a somnambulistic episode, with 75% of sleepwalkers having more than one. As expected, no somnambulistic episodes were observed in the control group during baseline or recovery sleep.

Sleepwalkers and controls differed on a few sleep variables assessed during the first two sleep cycles. Table 1 presents the means and standard deviations of sleep variables as well as summaries from the Group ^* Condition ANOVAs. A Group ^* Condition interaction was observed for slow-wave activity (F = 4.9, p < 0.05). Post-hoc comparisons revealed that sleepwalkers had significantly lower slow-wave activity than did the controls both at baseline and during recovery sleep (p < 0.05). In fact, only controls showed a significant increase in slow-wave activity from baseline to recovery sleep (p < 0.001). We also observed a Group ^* Condition interaction for the percentage of REM sleep (F = 4.3, p < 0.05), but post-hoc comparisons did not reveal significant effects.

Main group effects were also observed. Compared to controls' values at baseline and during recovery sleep, sleepwalkers showed lower sleep efficiency (F = 11.6, p < 0.01), lower percentage of N3 sleep (F = 4.7, p < 0.05), more arousals during N3 sleep (F = 6.4, p < 0.05), and a higher percentage and duration of N2 sleep (F = 6.4, F = 6.1; p < 0.05).

Regarding main condition effects, when compared to sleepwalkers' and controls' baseline, recovery sleep revealed a significantly shorter sleep latency (F = 50.2, p < 0.001), shorter sleep time during the first two sleep cycles (F = 10.5, p < 0.01), lower percentage and duration of N2 sleep (F = 12.4, F = 17.6; p < 0.01), higher percentage of N3 sleep (F = 19.2, p < 0.001), and a lower arousal index (F = 7.0, p < 0.01).

Further details on the sleep characteristics and behavioral episodes in this patient cohort have been published elsewhere (10, 25, 26).

Significant Group ^* Condition interactions were observed during the first sleep cycle for nLF and nHF as well as for the LF/HF ratio (Table 2). Post-hoc comparisons revealed that during the first cycle of recovery slow-wave sleep, sleepwalkers showed a lower LF/HF ratio and higher nHF compared to controls, although no group difference was observed at baseline. When compared to their slow-wave sleep at baseline, sleepwalkers' recovery slow-wave sleep presented lower nLF and LF/HF ratio as well as higher nHF during the first cycle. These significant interactions and post-hoc tests are presented in Figures 1, 2. No significant effect was found between the controls' baseline and their recovery slow-wave sleep.

No main effects were observed for nLF, nHF or the LF/HF ratio (see Table 2). No significant interactions or main effects were observed for total power, absolute LF and absolute HF. Although interactions were non-significant for absolute measures during the first sleep cycle of slow-wave sleep, the direction of effects was consistent with the significant normalized measures, with medium effect sizes (Table 2, Figure 2). The absence of statistically significant differences in the absolute measures may be due, in part, to large inter-subject variability as evidenced by the large standard deviations in these measures.

During the first slow-wave recovery sleep cycle, group by HRV variables interactions were observed when correlating with N3 awakenings (p < 0.05), but not for N3 sleep stage duration, N3 micro-arousals or number of sleepwalking episodes. A lower LF/HF ratio during recovery slow-wave sleep of sleepwalkers correlated with an increase in the number of N3 awakenings from baseline to recovery sleep (delta between recovery and baseline, r = −0.58, p < 0.05), while this was not observed in controls. No significant associations were observed between HRV variables with the number of sleepwalking episodes during recovery slow-wave sleep. However, we could not investigate whether HRV variables correlated with the episode occurrence since all sleepwalkers experienced at least one episode during their recovery sleep. On the other hand, EEG slow-wave activity and nLF in sleepwalkers' first recovery slow-wave sleep cycle interacted together in their association with the increase of episodes from baseline to recovery sleep (p < 0.05): lower nLF was more strongly associated with the increased number of episodes from baseline to recovery in sleepwalkers that presented with lower levels of slow-wave activity.

When slow-wave activity was entered as a covariate in the Group ^* Conditions interactions, the interactions for the LF/HF ratio and nLF remained significant, whereas the interaction for nHF was no longer statistically significant.

None of the Group ^* Condition effects remained significant during the second sleep cycle. This finding suggests that changes in sleepwalkers' autonomic modulation occur mostly during their first cycle slow-wave sleep, similar to their other atypical sleep characteristics (increased arousals, diminished slow-wave activity) (1, 29).

Overall, the interactions suggest that sleepwalkers show an autonomic modulation that is parasympathetic-dominant during the first cycle recovery slow-wave sleep compared to controls as well as to their own baseline.

As previously noted in other cohorts (30), sleepwalkers in the present study showed abnormal slow-wave sleep characterized by lower slow-wave activity and more arousals from slow-wave sleep when compared to controls. These findings support the view that sleepwalking is associated with a difficulty in maintaining consolidated slow-wave sleep (1). Our novel finding of lower LF/HF ratio suggesting an autonomic modulation that is parasympathetic predominant also points to an altered regulation of slow-wave sleep. The only previous study of autonomic function in 10 sleepwalkers vs. 10 controls found no group differences in frequency-domain metrics during participants' overall sleep (i.e., without examining specific sleep stages, nor sleep deprivation) or during wakefulness (19). These findings are consistent with those from the present study since we did not observe HRV change at baseline, suggesting that sleepwalkers may exhibit abnormal autonomic regulation as a function of homeostatic sleep pressure following extended sleep deprivation. In healthy controls, and as observed in the present study, sleep deprivation increases homeostatic sleep pressure, and thus, slow-wave sleep and slow-wave activity (17). This pattern was also previously reported in sleepwalkers after 38 h of sleep deprivation (31). However, using a 25 h sleep deprivation protocol, we did not observe a significant increase in slow-wave activity during sleepwalkers' recovery slow-wave sleep, which also suggests poorer consolidated slow-wave sleep in this population.

In the present study, the effect of a 25 h sleep deprivation, which results in daytime recovery sleep, differed between controls and sleepwalkers on HRV metrics. Whereas, sleep deprivation was associated with HRV metrics suggestive of a parasympathetic predominance in sleepwalkers' recovery slow-wave sleep, this association was not observed in controls. Although deeper slow-wave sleep is normally characterized by a predominance of parasympathetic modulation with elevated HF (13–16, 32), results from one study suggest that extended wakefulness is associated with elevated sympathetic activity during recovery sleep despite longer slow-wave sleep (17), which could explain our findings in controls. These varying results in healthy controls may stem from a mixture of effects from circadian, homeostatic and extended wakefulness factors. That said, while the effect of sleep deprivation on autonomic modulation during recovery sleep, and especially slow-wave sleep, remains unexplored in the literature, our current findings suggest that the effect differs between sleepwalkers and healthy controls.

Why sleepwalkers present different patterns of autonomic modulation during recovery slow-wave sleep than do controls remains unclear. Since structures of the central nervous system that regulate sleep and autonomic function are closely interrelated (11, 12), this difference in autonomic function in sleepwalkers might be the consequence of abnormal functioning of central neural regulatory systems. Moreover, given that a genetic predisposition likely plays an important role in the development of somnambulism (33–36), genetic factors could underlie the differential autonomic modulation observed in sleepwalkers after sleep deprivation. In fact, it was suggested that certain genes can affect the interaction between autonomic and slow-wave sleep regulation (17). It is possible that genetic contributions to sleepwalking include genes associated with autonomic functioning.

Whereas, sleepwalkers' overall slow-wave sleep showed HRV metrics suggestive of parasympathetic predominance in our study, sleepwalkers' slow-wave sleep prior to behavioral episodes or arousals was previously associated with heart rate markers suggesting sympathetic activation. In one study, increased HRV total spectral power was found to increase during the 5 min preceding episodes in 10 sleepwalkers in a normal sleep condition (19). Another study under normal sleep conditions showed that arousals during slow-wave sleep in 38 predisposed individuals were associated with an abrupt heart rate acceleration (18). A similar finding was found for motor arousals during slow-wave sleep associated with sleepwalking or sleep terrors in 20 participants, with arousals being preceded by an abrupt heart rate acceleration (20). Overall, our present findings suggest that this activation prior to episode occurrence does not appear to extend to entire periods of slow-wave sleep.

Taken together, these previous findings on heart rate acceleration as well as the findings suggestive of parasympathetic dominance observed in the present study during sleepwalkers' recovery slow-wave sleep point to altered autonomic modulation that may be a marker of the disorder's underlying pathophysiology, or could play a more direct role in the occurrence of somnambulistic episodes. Indeed, we observed that HRV variables suggestive of parasympathetic dominance were associated with increased slow-wave sleep awakenings and interacted with slow-wave activity in their association with increased frequency of episodes from baseline to recovery sleep. Although the directionality of the association between autonomic modulation and sleep-related arousals remains subject to debate, some findings show that normal sleep-related arousals and EEG changes are preceded by autonomic fluctuations (11), suggesting that autonomic modulation may play a role in the initiation of arousals. In fact, an interaction between parasympathetic predominance during slow-wave sleep and transitions from sleep to wakefulness characterized by sudden sympathetic activation may contribute to behavioral manifestations associated with sleepwalking. This pattern is similar to what is observed with slow-wave activity in sleepwalkers. Specifically, although sleepwalkers present with decreased slow-wave activity (30), as seen in the present study, somnambulistic episodes are often characterized by relatively abrupt increases in slow oscillations and slow delta frequency seen immediately prior to episodes' occurrence (25). The dissociation between parasympathetic dominance during slow-wave sleep and potentially abrupt sympathetic activation during arousals could signal the concomitant presence of features associated with both sleep and wakefulness during sleepwalkers' abnormal arousal reactions. Consistent with this line of reasoning, a study of EEG functional connectivity in sleepwalkers found that somnambulistic episodes are preceded by brain processes suggestive of the coexistence of arousal and sleep (37). The coexistence of arousal and sleep mechanisms in disorders of arousals were also suggested by studies using high density EEG source modeling, EEG current density imaging, single-photon emission computed tomography, and intracerebral EEG (38–41).

It remains unclear whether altered autonomic modulation during recovery slow-wave sleep is only a marker of abnormal neural circuitry in predisposed individuals or if it plays an active role in precipitating somnambulistic episodes. Some psychotropic medications, most notably zolpidem, are known to facilitate sleepwalking in individuals with no past history of the disorder (42). During sleep, zolpidem has been shown to promote the reduction of blood pressure (43), a normal phenomenon known as “dipping” that is viewed as being a consequence of lower sympathetic outflow during sleep (12, 44). Moreover, zolpidem reduces levels of epinephrine and norepinephrine (43), two neurotransmitters secreted following sympathetic activation. This suggests that a potential shift of the sympathovagal balance observed with zolpidem could inform sleepwalking pathophysiology. Similarly, sleepwalking can arise through the use of beta blockers (42), whose main mechanism of action is competitive beta adrenergic receptor antagonists and thus, block sympathetic effects. More recently, methylphenidate, a stimulant that increases the sympathovagal balance toward sympathetic modulation (45), has been used to treat two patients presenting with a history of chronic and severe sleepwalking (46). Taken together, these findings suggest that autonomic modulation, and especially parasympathetic dominance during slow-wave sleep, might play a role in the pathophysiology of sleepwalking. Whether or not therapeutic avenues known to affect autonomic modulation could also impact sleepwalking remains to be investigated.

Sleepwalkers were allowed to leave the laboratory setting for the day after their baseline sleep assessment whereas controls remained at the laboratory in constant routine. Thus, sleepwalkers may have experienced greater daytime activity levels than controls before returning to the lab for the sleep deprivation protocol. However, greater levels of daytime activity would have been expected to be associated with increased slow-wave activity (47, 48), which was not observed in sleepwalkers, suggesting that our findings are not accounted for by different daytime activity levels. Because recovery sleep took place during the daytime and that daytime sleep is not habitual for our patients, we do not know the extent to which this circadian phase of usually low sleep propensity could have contributed to our findings. A third limitation was the absence of a respiration signal coupling that could have affected HRV variables. However, slow-wave sleep has been hypothesized to be an optimal recording condition to measure HRV because of stable breathing patterns (14).

Moreover, as stated earlier, the LF/HF ratio has been challenged as a marker of the sympathovagal balance (21, 23, 49), and thus, further studies using sympathetic and parasympathetic blockade might be needed to confirm our findings. Here, we have interpreted the direction of our HRV findings in sleepwalkers' recovery slow-wave sleep during the first cycle as a marker of parasympathetic predominance. However, given the complex physiological factors affecting the LF band, our findings could be interpreted in another direction: The LF band has been suggested to be more closely related to the neural modulation rather than effect on cardiac function (21), and thus, might be of central origin, which is highly relevant in the study of sleepwalking pathophysiology. In fact, lower nLF and LF/HF in sleepwalkers' recovery slow-wave sleep could represent a reduced baroreflex sensitivity and control over cardiac function (21), which is normally the greatest during slow-wave sleep (11). This interpretation also supports an abnormal autonomic modulation of slow-wave sleep in sleepwalkers, especially during their first cycle of recovery sleep.