Since the formal characterization of sleep stages, there have been reports that seizures may preferentially occur in certain phases of sleep. Through ascending cholinergic connections from the brainstem, rapid eye movement (REM) sleep is physiologically characterized by low voltage fast activity on the electroencephalogram, REMs, and muscle atonia. Multiple independent studies confirm that, in REM sleep, there is a strikingly low proportion of seizures (~1% or less). We review a total of 42 distinct conventional and intracranial studies in the literature which comprised a net of 1458 patients. Indexed to duration, we found that REM sleep was the most protective stage of sleep against focal seizures, generalized seizures, focal interictal discharges, and two particular epilepsy syndromes. REM sleep had an additional protective effect compared to wakefulness with an average 7.83 times fewer focal seizures, 3.25 times fewer generalized seizures, and 1.11 times fewer focal interictal discharges. In further studies REM sleep has also demonstrated utility in localizing epileptogenic foci with potential translation into postsurgical seizure freedom. Based on emerging connectivity data in sleep, we hypothesize that the influence of REM sleep on seizures is due to a desynchronized EEG pattern which reflects important connectivity differences unique to this sleep stage.
A bidirectional relationship between epilepsy and sleep has been observed since the time of Hippocrates [
Based on a wealth of animal and human data accumulated since the discovery of REM sleep in 1953 [
Of the neurons whose spontaneous firing rate is highest in REM sleep, some exhibit a spontaneous bursting depolarization pattern due to a “low threshold spike” inward calcium current [ low voltage fast electroencephalographic (EEG) activity, rapid eye movements (REMs), muscle atonia.
Low voltage fast activity on the EEG is due largely to depolarization of the thalamus by cholinergic MRF neurons [
Rapid eye movements are heralded by discharges, known as pontogeniculooccipital (PGO) waves, from a dorsorostral subpopulation of cholinergic PRF neurons which project to the occipital lobe via the lateral geniculate nucleus (LGN). The presence of PGO waves precedes REMs by 3–5 waves and low voltage fast activity by 30–60 seconds [
Muscle atonia is partly the result of neurons in the dorsolateral PRF [
Collectively the LDT, PPT, PRF, and MRF are known as the “REM-on” neurons. In contrast, there are populations of “REM-off” neurons mainly in the serotonergic midline raphe nuclei and noradrenergic locus coeruleus (LC) [
The REM-on and REM-off neurons mutually antagonize each other. The model first proposed by McCarley and Hobson in 1975 [
With respect to anatomical and functional correlation of REM-on neurons in the model, there exists a positive feedback connection between LDT/PPT and PRF/MRF neurons [
The reciprocal interaction model provides one method of explaining the 90-minute alternations between 30-minute REM sleep periods and NREM sleep periods over the course of a usual night. In order to account for the first shorter REM episode, which typically occurs 70–120 (on average 90) minutes after sleep onset, subsequent versions of the model have included a “limit cycle” modification [
Furthermore, the hormone orexin (also known as hypocretin), which is secreted by neurons in the lateral hypothalamus, additionally fine-tunes transitions into and out of REM sleep by diurnally gating REM sleep over the course of the entire sleep-wake cycle [
Initial studies on the frequency of interictal and ictal events during REM sleep were largely anecdotal and consisted primarily of case reports. Studies were heterogeneous in terms of seizure/epilepsy classification (e.g., waking epilepsy, definitely symptomatic epilepsy), patient population (e.g., severity of epilepsy, use of antiepileptic drugs), use of the EEG (e.g., 10–20 system, montages, method of detecting abnormalities, inclusion of benign variants as abnormal features), use of the polysomnogram (PSG) (e.g., use of electromyography, definition of wakefulness and sleep stages), and outcome measures of both interictal and ictal events. Gradually, however, the methodology for recording and scoring became more standardized and this permitted comparison.
In this review, the total number of events in wakefulness and each sleep stage was extracted for each study examined. Rates of interictal and ictal events in wakefulness and each sleep stage were also extracted. If rates were not explicitly provided, then they were calculated by dividing the number of events by the duration of wakefulness and/or each sleep stage when available. To facilitate comparison, each rate was then divided by the rate for REM sleep in order to determine an “indexed” rate. In averaging these indexed rates, each study was not treated equally. Rather, a “weighted mean” was produced by weighting each study based on the number of patients contained with each.
If individual sleep stages were not separated, then the same rate was used for each constituent stage (e.g., a combined N1/N2 rate was used individually as a rate for N1 and a rate for N2). With respect to numbers of events, this was divided equally among the constituent stages (e.g., a total of 33 seizures for N1/N2/N3 were counted as 11 for each stage). Formerly stage III and stage IV sleep were combined into stage N3 for analysis. Depending on the study, the definition of wakefulness may have included wake periods after sleep onset (WASO), nocturnal awakenings, morning awakenings, and/or samples of fully alert daytime wakefulness. As studies were divergent, statistical significance could not be calculated.
A total of 542 patients with a collective 1990 seizures over 9 studies [
Raw sum of focal seizures.
The percentage of focal seizures during REM sleep over total recording time was extremely low (1%) over all these studies. However, because these studies did not provide specific durations, the length of recording may have led to artificial overinflation or underinflation of data. To address this issue, Table
Relative focal seizure rates*.
Paper/sleep stage | W | N1 | N2 | N3 | REM |
---|---|---|---|---|---|
Minecan et al. 2002 [ |
0.00 | 6.00 | 7.00 | 5.00 | 1.00 |
Crespel et al. 1998 [ |
133.42 | 14.59 | 14.59 | 14.59 | 1.00 |
Crespel et al. 1998 [ |
55.08 | 1.67 | 1.67 | 1.67 | 1.00 |
Terzano et al. 1991 [ |
0.00 | 5.52 | 2.16 | 3.77 | 1.00 |
| |||||
Weighted mean | 7.83 | 87.25 | 67.84 | 50.78 | 1.00 |
Relative to REM sleep, the focal seizure rate was 7.83 times higher in wakefulness, 87 times higher in stage N1 sleep, 68 times higher in stage N2 sleep, and 51 times higher in stage N3 sleep. These data imply that focal seizures were most frequent in NREM sleep, intermediate in wakefulness, and lowest in REM sleep. However, the increased rate in wakefulness was highly variable with the weighted mean being powered by a single study [
A total of 256 patients with idiopathic generalized epilepsy were included among 7–9 studies [
Table
Relative generalized discharge rates.
Paper/sleep stage | W | N1 | N2 | N3 | REM |
---|---|---|---|---|---|
Halász et al. 2002 [ |
8.14 | 14.53 | 10.47 | 3.49 | 1.00 |
Parrino et al. 2001 [ |
3.50 | 3.50 | 3.50 | 1.00 | |
Horita et al. 1991 [ |
1.43 | 3.54 | 0.75 | 0.00 | 1.00 |
Autret et al. 1987 [ |
1.37 | 1.50 | 1.50 | 2.25 | 1.00 |
Autret et al. 1987 [ |
2.30 | 3.66 | 3.66 | 4.91 | 1.00 |
Touchon 1982 [ |
5.05 | 3.26 | 0.10 | 1.00 | |
Kellaway et al. 1980 [ |
1.68 | 5.63 | 5.63 | 5.63 | 1.00 |
Sato et al. 1973 [ |
3.32 | 16.04 | 43.45 | 1.00 | |
Ross et al. 1966 [ |
4.94 | 3.12 | 3.94 | 11.06 | 1.00 |
| |||||
Weighted mean | 3.25 | 3.10 | 3.13 | 6.59 | 1.00 |
% patients with maximal generalized discharges per state.
Paper/sleep stage | W (%) | N1 (%) | N2 (%) | N3 (%) | REM (%) |
---|---|---|---|---|---|
Horita et al. 1991 [ |
0.0 | 100.0 | 0.0 | 0.0 | 0.0 |
Sato et al. 1973 [ |
0.0 | 8.3 | 91.7 | 0.0 | |
Ross et al. 1966 [ |
23.1 | 7.7 | 7.7 | 61.5 | 0.0 |
| |||||
Weighted mean | 22.8 | 14.8 | 5.9 | 56.4 | 0.0 |
Benign epilepsy of childhood with rolandic spikes (BECRS) is the best studied focal syndrome in terms of discharge frequency in REM sleep. A total of 110 patients aged 3–16 were examined by Billiard et al. [
Table
Relative rolandic discharge rates.
Paper/sleep stage | W | N1 | N2 | N3 | REM |
---|---|---|---|---|---|
Billiard et al. 1990 [ |
0.54 | 1.19 | 1.19 | 1.19 | 1.00 |
Dalla Bernardina et al. 1982 [ |
0.22 | 0.97 | 1.08 | 1.28 | 1.00 |
| |||||
Weighted mean | 0.27 | 1.00 | 1.10 | 1.27 | 1.00 |
In addition to focal syndromes, a few case reports have also explored the impact of REM sleep on the epileptic encephalopathies. In 1981, Billiard et al. [
In 1982, Tassinari [
A total of 214 patients were included among 7–10 different studies [
Table
Relative focal interictal discharge rates.
Paper/sleep stage | W | N1 | N2 | N3 | REM |
---|---|---|---|---|---|
Clemens et al. 2005 [ |
1.52 | 2.50 | 1.85 | 2.67 | 1.00 |
Clemens et al. 2003 [ |
0.41 | 1.56 | 1.48 | 2.46 | 1.00 |
Ferillo et al. 2000 [ |
0.91 | 1.09 | 1.81 | 1.00 | |
Malow et al. 1998 [ |
2.45 | 3.79 | 7.39 | 1.00 | |
Malow et al. 1997 [ |
2.38 | 4.50 | 7.13 | 1.00 | |
Billiard et al. 1981 [ |
0.84 | 1.68 | 1.68 | 1.68 | 1.00 |
Billiard et al. 1981 [ |
0.70 | 1.43 | 1.43 | 1.43 | 1.00 |
Autret et al. 1987 [ |
0.44 | 1.18 | 1.18 | 1.15 | 1.00 |
Rossi et al. 1984 [ |
1.17 | 1.33 | 1.33 | 1.18 | 1.00 |
Montplaisir et al. 1982 [ |
1.21 | 2.43 | 1.00 | ||
| |||||
Weighted mean | 1.11 | 1.75 | 1.69 | 2.46 | 1.00 |
Although REM sleep had the lowest rate of focal interictal discharges overall, this did not mean that each individual patient necessarily had lower rates of discharges in REM sleep. As Table
% patients with maximal focal discharges per state.
Paper/sleep stage | W (%) | N1 (%) | N2 (%) | N3 (%) | REM (%) |
---|---|---|---|---|---|
Clemens et al. 2005 [ |
11.1 | 11.1 | 0.0 | 66.7 | 11.1 |
Ferrillo et al. 2000 [ |
19.4 | 11.1 | 61.1 | 8.3 | |
Sammaritano et al. 1991 [ |
2.5 | 3.8 | 3.8 | 77.5 | 12.5 |
| |||||
Weighted mean | 5.9 | 8.7 | 5.1 | 69.5 | 10.9 |
In intracranial depth recordings, Rossi et al. [
Like seizure frequency in REM sleep, the impact of REM sleep on the distribution of interictal and ictal phenomena is controversial. The most powerful argument for clinically useful localization of an epileptogenic focus in REM sleep comes from a subset of tuberous sclerosis patients and a single temporal lobe epilepsy patient in studies by Ochi et al. [
Other studies have explored the localizing ability of REM sleep in relation to a “final” localization based on general concordance of all available data. 100% of unilateral temporal lobe patients with REM-lateralized interictal discharges or seizures were lateralized to the same hemisphere as the “final” localization. For NREM sleep, the concordance rate of interictal discharges and seizures was 100% and 94%, respectively. For wakefulness, it was a respective 88% and 94%. In patients where discharges were bitemporal, REM localization agreed with the final localization 100% of the time (compared to 81% in NREM and 100% in wakefulness). In an intracranial study by Lieb et al. [
However, there remains controversy regarding the localizing value of interictal discharges. For example, in Lieb’s study [
As previously discussed, the observed desynchronized EEG of REM sleep is the result of cholinergic MRF neurons depolarizing the thalamus which allows transmission of information to the cortex. Like REM sleep, the EEG of wakefulness is also desynchronized because cholinergic activity is likewise present.
In contrast, cholinergic neurons are less active in NREM sleep with the least activity occurring in deep slow-wave sleep (i.e., stage N3) [
To summarize, REM sleep and wakefulness represent states of maximal cortical synchrony, stages N1 and N2 sleep are states of intermediate synchrony, and stage N3 sleep represents a state of maximal cortical desynchrony.
When neurons exhibit asynchronous discharging behaviour at baseline, there is a reduced opportunity for spatial and temporal summation of any additional spontaneous depolarization [
The reduced opportunity for spatial and temporal summation of such abnormal depolarizations may account for the results contained within Table
Another possible mechanism by which cortical desynchrony may account for this disparity in focal discharge rates is through the emergence of regional antiepileptic “microrhythms”. In contrast to the uniform global cortical synchrony of stage N3 sleep, there usually are distinct regional rhythms in more desynchronized states. For example, a posterior dominant alpha rhythm often exists in wakefulness. Even during the intermediately synchronized EEG of stage N2 sleep, there are regional sleep spindles located in the frontocentral regions bilaterally which, by definition, disappear with the onset of slow-wave sleep. Furthermore, a recent study has commented on “islands of hyperconnectivity” in REM sleep [
While the regional rhythms mentioned above are not known to be antiepileptic, another rhythm has been described, the hippocampal theta rhythm, which is also present in the desynchronized states of REM sleep and wakefulness [
However, the opposite may also be true and there could exist proepileptic regional rhythms in certain individuals. This may account for the interindividual variability in Table
Similar to focal interictal discharges, focal seizures are also hypothesized to arise from the “paroxysmal depolarizing shift” [
Like secondarily generalized seizures, primary generalized epilepsy also involves spread via long pathways but through a mechanism distinct from the paroxysmal depolarizing shift. Both primary generalized ictal and interictal phenomena have been hypothesized to be the result of abnormally synchronized and reverberating thalamocortical networks [
Tables
Namely, the lower likelihood for spatial and temporal summation of aberrant spontaneous depolarizations in the cortically desynchronized states of REM sleep and wakefulness reduces the chance of spread along “short” pathways to surrounding neurons in focal epilepsy and “long” thalamocortical and association pathways in generalized epilepsy. Furthermore, should a desynchronized cortical milieu permit the emergence of regional antiepileptic microrhythms, this would present a further impediment to the spread of any aberrant depolarization.
However, unlike focal interictal discharges, the potential presence of proepileptic regional rhythms in certain individuals would not be expected to impact a primary generalized phenomenon. This is consistent with the results of Table
Returning to Tables
Serotonergic neurons primarily located in the raphe nuclei, noradrenergic neurons in the locus coeruleus, and histaminergic neurons from the tuberomammillary nucleus demonstrate maximal firing rates in wakefulness and lowest firing rates during REM sleep [
An fMRI study [
Because a loss of connectivity precludes the presence of synchrony, strategic losses of brain connectivity in REM sleep compared to wakefulness might explain any extra antiepileptic effect of REM sleep. As previously discussed, the greater the degree of desynchronization, the less likely the spatial and temporal summation of any aberrant spontaneous depolarization which would allow “spread” along “short” or “long” pathways in the brain.
From the aforementioned case reports on the epileptic encephalopathies, REM sleep has been noted to usually have an antiepileptic effect. As an encephalopathy is, by definition, a spread-out and diffuse process, the reduced potential for such spread in the desynchronized environment of REM sleep may explain this observed antiepileptic effect.
In contrast, Table
From the reviewed studies involving multifocal (i.e., tuberous sclerosis [
However, there are also clearly described instances of false localization in the literature [
Sixty years after the discovery of REM sleep, a wealth of literature has commented on the effect of REM sleep on seizures. In our review, we have demonstrated that, compared to NREM sleep, REM sleep has a strong antiepileptic effect against focal interictal discharges, focal seizures, and generalized seizures. We also found that REM sleep has an additional antiepileptic effect compared to wakefulness against focal and generalized seizures.
While cases of false localization have been described, REM sleep has been demonstrated to have promise in helping localize epileptogenic foci with possible translation into postsurgical seizure freedom. The potential selective localizing value of REM sleep may argue for the use of dedicated sleep recordings in the presurgical evaluation of epilepsy.
Finally, we hypothesize that the impact of REM sleep on epilepsy is due to a maximally desynchronized EEG pattern which reduces the likelihood of spatial and temporal summation of aberrant depolarizations. Although at first glance similar to wakefulness, recent connectivity studies demonstrate a further strategic loss of connectivity in REM sleep which we hypothesize accounts for its unique antiepileptic influence on seizures.
Dr. Pavlova has received research support from the Harvard Catalyst (Grant no. UL1 RR 02758) in 2010-2011.