Focal epilepsy impacts rapid eye movement sleep microstructure

Abstract

Study Objectives

Whereas there is plenty of evidence on the influence of epileptic activity on non-rapid eye movement (NREM) sleep macro- and micro-structure, data on the impact of epilepsy on rapid eye movement (REM) sleep remains sparse. Using high-density electroencephalography (HD-EEG), we assessed global and focal disturbances of sawtooth waves (STW) as cortically generated sleep oscillations of REM sleep in patients with focal epilepsy.

Methods

Twenty-two patients with drug-resistant focal epilepsy (13 females; mean age, 32.6 ± 10.7 years; 12 temporal lobe epilepsy) and 12 healthy controls (3 females; 24.0 ± 3.2 years) underwent combined overnight HD-EEG and polysomnography. STW rate, duration, frequency, power, spatial extent, IED rates and sleep homeostatic properties were analyzed.

Results

STW rate and duration were reduced in patients with focal epilepsy compared to healthy controls (rate: 0.64/min ± 0.46 vs. 1.12/min ± 0.41, p = .005, d = −0.98; duration: 3.60 s ± 0.76 vs. 4.57 ± 1.00, p = .003, d = −1.01). Not surprisingly given the fronto-central maximum of STW, the reductions were driven by extratemporal lobe epilepsy patients (rate: 0.45/min ± 0.31 vs. 1.12/min ± 0.41, p = .0004, d = −1.35; duration: 3.49 s ± 0.92 vs. 4.57 ± 1.00, p = .017, d = −0.99) and were more pronounced in the first vs. the last sleep cycle (rate first cycle patients vs. controls: 0.60/min ± 0.49 vs. 1.10/min ± 0.55, p = .016, d = −0.90, rate last cycle patients vs. controls: 0.67/min ± 0.51 vs. 0.99/min ± 0.49, p = .11, d = −0.62; duration first cycle patients vs. controls: 3.60s ± 0.76 vs. 4.57 ± 1.00, p = .003, d = −1.01, duration last cycle patients vs. controls: 3.66s ± 0.84 vs. 4.51 ± 1.26, p = .039, d = −0.80). There was no regional decrease of STWs in the region with the epileptic focus vs. the contralateral side (all p > .05).

Conclusion

Patients with focal epilepsy and in particular extratemporal lobe epilepsy show a global reduction of STW activity in REM sleep. This may suggest that epilepsy impacts cortically generated sleep oscillations even in REM sleep when epileptic activity is low.

Graphical Abstract

Graphical Abstract.

Statement of significance.

Despite sparsity of epileptic activity in REM sleep, sawtooth wave activity as a marker of REM microstructure is impaired in patients with various focal epilepsies or epilepsy syndromes. The decrease was especially pronounced in extratemporal lobe patients and in the first compared to the last sleep cycle. This may suggest that epilepsy may have an effect that is independent to that of epileptic activity. Future studies are warranted to investigate potential functional consequences of impaired REM sleep microstructure.

Introduction

Epilepsy and sleep have multiple reciprocal interactions [1], with epileptic activity comprised of both seizures and interictal epileptiform discharges (IEDs) being facilitated during nonrapid eye movement (NREM) sleep, and suppressed in rapid eye movement (REM) sleep [2]. Especially phasic REM sleep, characterized by REMs and muscular twitches, seems to be protective against epileptic activity, with IEDs being even further reduced in phasic compared to tonic REM sleep [3, 4]. There are differences in the sleep macrostructure between epilepsy patients and healthy controls; these are likely explained by a disruption caused by epileptic activity [5, 6]. In contrast to these findings about sleep macrostructure, sleep microstructure is less well studied [7]. As one parameter of sleep microstructure, sleep oscillations reflect physiological neuronal activity within cortical-subcortical networks and underlie several functions such as learning, cognition, and sleep stability [8–10]. There is increasing evidence in NREM sleep that these physiological processes are disturbed by epileptic activity: patients with epilepsy showed reductions in NREM sleep oscillations; reductions in these latter are further associated with lower cognitive and memory performance [11–15]. However, in contrast to these findings in NREM sleep microstructure, studies focusing on the impact of epilepsy on REM sleep microstructure are sparse. Despite lower rates of epileptic activity in REM sleep [2, 16], sleep oscillations generated on the cortex may be nevertheless altered in epilepsy because of a constant dysfunctional cortex especially in the brain region with the epileptic focus [17]. Therefore, the present study aimed to analyze sleep oscillations in REM sleep in patients with epilepsy.

Hallmarks of REM sleep are rapid eye movements and sawtooth waves (STWs) as REM sleep specific oscillations [18]. STWs are distinct trains of regular EEG waves (20–100 μV) with a triangular shape at 2–5 Hz with a peak at 2–3 Hz, appearing as slow bursts in contrast to the low voltage fast activity of REM sleep often registered shortly before a burst of REMs [19–21]. The highest amplitude is over the frontocentral regions [22, 23]. Using high-density electroencephalography (HD-EEG), potential generators of STWs were located in the frontocentral and occipitotemporal brain areas [24]. Intracranial EEG showed in addition posterior parietal and insular areas as another source of STWs [25]. Furthermore, the involved brain regions during STW activity showed high spatiotemporal heterogeneity suggesting that STWs are under local cortical regulation [25].

To date, a comprehensive study focusing on the distribution of STWs and the potential influence of epileptic activity on STW structure is missing. There is only one single study in temporal lobe epilepsy patients showing reduced STWs compared to healthy controls [26]. If this finding is generalizable to all types of focal epilepsies is unknown. There is also no knowledge if these reductions are global or focal showing a clear relationship with the localization of the epileptic focus. HD-EEG provides a high spatial resolution and offers the possibility to explore the distribution of STWs in patients with focal epilepsy in specific brain areas. Furthermore, it allows the comparison of activity in the lobe of the focus (focus proximity) and outside of the lobe of the focus (distant from focus). To explore REM microstructure and the potential impact of epilepsy on cortically generated oscillations, we aimed to provide a mapping of STWs in adult patients with focal temporal and extratemporal lobe epilepsy by comparing full night HD-EEG recordings obtained in patients and in healthy controls.

In our primary hypothesis, we aimed to explore the differences in rate, duration, frequency, power and extent of STW activity between focal epilepsy patients and healthy controls. We assume that STWs as cortically generated sleep oscillations may be impacted by epilepsy and therefore be altered in patients with focal epilepsy compared to healthy controls. As exploratory endpoints, we aimed to analyze the anatomical extent of the potential alterations of STW activity within epilepsy patients, compare the distribution of STWs across the night (first vs. last sleep cycle), and assess the correlation of the IED index (IED rate per minute) during REM sleep with rate, duration, frequency, power and extent of STW activity. We hypothesized that i) STW activity is reduced in patients with focal epilepsy compared to healthy controls, ii) within patients with focal epilepsy, STW activity is decreased in the brain region with the epileptic focus, iii) the reduction is stronger in the beginning of the night than in the end of the night, like that seen for other sleep phenomena, and iv) the IED index during REM sleep is correlated negatively with STW activity.

Methods

Study population

Patients with a diagnosis of a unilateral drug-resistant focal epilepsy or epilepsy syndrome who underwent combined PSG and HD-EEG at the Montreal Neurological Institute and Hospital between January 2019 and July 2021 were included in this study. From a total of 25 patients fulfilling these criteria, 22 patients (13 females; mean age, 33.5 ± 11.3 years (18–53)) were selected for the analysis. The reasons for exclusion of three patients were previous brain surgery (n = 1), ≥10 electrodes with artifacts (n = 1), and interference of epileptic activity with sleep scoring (n = 1). We decided to use HD-EEG to be able to examine not only global but also focal disturbances of STW activity. Patients were grouped in temporal lobe (n = 12, 54.5%) and extratemporal lobe epilepsy patients (n = 10). Temporal lobe epilepsy patients had their focus in the temporolateral or mesiotemporal lobe. Patients with an extratemporal focus or temporal plus epilepsy (temporal plus another region) were allocated to the extratemporal lobe epilepsy group. The epileptic focus was defined based on phase 1 or 2 presurgical evaluation with long-term video EEG monitoring, anatomical 3T magnetic resonance imaging (MRI), positron emission tomography and neuropsychological evaluation, or stereo-electroencephalography where applicable. Table 1 shows the demographic characteristics of the patient group. To compare the findings obtained in epilepsy patients with a control group, 12 healthy controls (3 females; mean age, 24.0 ± 3.2 years (20–29)) were recruited at the Neuroscience Center of Queen’s University. Healthy controls had no sleep and neurological disorders as assessed with a screening questionnaire. All study participants provided written informed consent in agreement with the Research Ethics Board at the Montreal Neurological Institute and Hospital and Queen’s University.

Table 1.

Patient demographics

Nr	Age, Sex	Handed-ness	Epileptic Focus	EEG Findings	MRI findings	Medication at day of HD-EEG	Daytime seizures (time to PSG start, h)	Generator	SEEG	Surgery, Outcome >1year	Pathology
1	26, F	R	L mesio-temporal	Interictal: T3, T9, FT9, FT7, F9>T5, F7 (frequent) or F10, T10, T4, F8, Zy2, P10 *Ictal: L posterior temporal (T3, T5, P9)	Normal	Lamotrigine 550mg Lacosamide 500 mg Perampanel 8 mg	-	Deep	X	RFTC, Ia
2	18, F	R	R temporo-occipital	Interictal: PO8>P12, T6, P10, O2, PO4 *Ictal: R temporo-occipital (T6, P10, O2)	Normal	Carbamazepine 1000 mg Perampanel 8 mg	-	Inter-mediate	X	R temporo-occipital resection, n.a.	FCD IIa
3	41, F	R	R mesio-temporal	Interictal: F8, T4, F10, T10, FT8, FT10, FT12 (frequent) or F7, T3, F9, T9, FT9, FT7 *Ictal: R temporal (F8, T4, F10, T10) or bitemporal (F8, T4, F10, T10 and F7, T3, F9, T9)	R MTS	Lacosamide 200 mg	1 FIA (-6.3h)	Deep	-	R anterior temporal lobectomy, Ia	Hippocampal sclerosis
4	25, F	L	L mesio-temporal	Interictal: F7, F9, T9, FT9, F11, FT11 *Ictal: L temporal (F7, T3, F9, T9, Zy1)	L MTS	Lamotrigine 100 mg	-	Deep	-	-
5	51, F	R	R mesio-temporal	Interictal: F8, T4, F10, T10, FT10, AF8, F12 *iEEG: R temporal (F8, T4, F10, T10, Zy2)	R hippocampal atrophy	Levetiracetam 3000 mg	2 FIA (-8.5h, -2.1h)	Deep	-	R anterior temporal lobectomy, Ia	Hippocampal sclerosis
6	20, M	R	L dorsolateral prefrontal	Interictal: F3, C3, F5, FC5, FC3, FC1 *Ictal: L fronto-temporal (F3, C3>F7, T3, F9, T9)	L frontocentral ischemic encephalo-malacic lesion	Lacosamide 400 mg Lamotrigine 175 mg	-	Super-ficial	-	L dorsolateral frontal resection, Ia	Gliosis, FCD IIa
7	53, F	R	Left temporal	Interictal: F7, T3, F9, T9, Zy1 *Ictal: F7, T3, F9, T9, Zy1	L MTS	Carbamazepine 400 mg Lacosamide 100mg	-	Deep	-	-
8	23, M	R	R parietal operculum/posterior insula	Interictal: C4, P4, Pz, CPz, P2, P6, CP2, CP4, CP6 *Ictal: R centro-parietal (C4, P4, Pz)	R parietal operculum/ posterior insula FCD	Carbamazepine 1600 mg Levetiracetam 1000 mg	-	Intermediate	-	-
9	24, M	R	R frontopolar	Interictal: Fp2, SO2, AF8, AF4, AFz, F4, Fz>Fp1, F3, AF7, AF3 *Ictal: R frontopolar (SO2, Fp2)	R frontopolar FCD	Carbamazepine 800 mg	-	Super-ficial	-	-
10	43, F	R	R parietal	Interictal: TP8, FT12, FT10, P10, T4, T6, T10 *Ictal: R hemisphere	Normal	Lacosamide 200 mg Levetiracetam 1500mg Carbamazepine 1100 mg	-	Super-ficial	X	-
11	40, F	R	L posterior temporal	Interictal: F7, T3, T5, F9, T9, Zy1 *Ictal: L posterior temporal region (T5, T3, P9)	Normal	Lacosamide 100 mg	5 FIA (-8.7h, -6.8h, -5.2h, -5.0h, -2.8h)	Deep	-	-
12	43, M	L	R mesio-temporal	Interictal: F8, T4, F10, FT8, FT10 *Ictal: R temporal (F10, T10, F8, T4, Zy2)	R MTS	Lamotrigine 75 mg Carbamazepine 400 mg	2 FA (-5.5h, -3.6h)	Deep	-	-
13	35, M	R	L mesio-temporal	Interictal: F7, F9, FT9, F11, FT11> Fp1, SO1 *Ictal: L temporal region (F7, T3, Zy1)	L temporal lobe atrophy affecting mesial structures	Clobazam 10 mg	1 FA (-10.2h)	Deep	-	-
14	43, F	R	R insula	Interictal: spike negative *Ictal: R posterior temporal region (T6, P4, P10)	Severe bilateral frontal atrophy	Oxcarbazepine 2100 mg Clobazam 30 mg	-	Deep	X	-
15	51, M	R	L mesio-temporal	Interictal: spike negative *Ictal: L temporal (F9, T9, Zy1>F7, T3)	L mesio-temporal FCD or low-grade tumor	Lacosamide 400 mg	-	Deep	-	-
16	40, F	R	R frontal operculum	Interictal: Fp2, SO2, F10, AF8, Zy2, Fpz *Ictal: no EEG changes1	R prefrontal cavernoma	Eslicarbazepine 1200 mg Lamotrigine 350 mg Clobazam 30mg Primidone 125 mg	-	Super-ficial	X	R fronto-operculum resection, Ia	FCD IIa
17	21, F	R	R mesio-temporal	Interictal: spike negative *Ictal: R hemispheric	Normal	Lamotrigine 400 mg Clobazam 10 mg	-	Deep	X	R anterior temporal lobectomy, n.a.	FCD I
18	24, M	Ambi-dextrous	L mesio-temporal	Interictal: spike negative *Ictal: L Temporal (F7, T3, F9, T9)	L mesio-temporal cystic lesion	Levetiracetam 3000 mg Lamotrigine 375 mg Lacosamide 400mg	-	Deep	-	-
19	19, F	R	R temporo-occipital	Interictal: T4, T6, TP8, TP10 > C4, P4 *Ictal: R temporo-occipital (O2>T6, P10, P4)	R lingual encephalo-malacic lesion	Lacosamide 200 mg Levetiracetam 1000 mg	-	Deep	-	R occipital resection, n.a.	Mild FCD
20	32, M	R	R posterior temporal/ posterior insula	Interictal: T4, T6, P12, PO8, TP8 *Ictal: R posterior temporal (T4, T6, T10, P10)	R temporal neocortex atrophy & atrophy/ agenesis of the right piriform	Lacosamide 250 mg Brivaracetam 100 mg Eslicarbazepine 800 mg	-	Inter-mediate	X	-
21	27, F	L	R latero-occipital	Interictal: P10, T6, O2> PO8, TP10, TP8, TP12, F8, T4 *Ictal: R posterior temporal (T6, P10)	Normal	Levetiracetam 3250 mg Lacosamide 400 mg	1 FA (-2.9h)	Super-ficial	X	-
22	40, M	R	L mesio-temporal	Interictal: FT11, FT9, T9, P9, TP9, P11, TP11, P9, TP9, T9 *Ictal: L temporal (F7, T3, F9, T9)	L fusiform lesion	Levetiracetam 2000 mg Carbamazepine 1200 mg	-	Deep	X	-

^*Ictal data were recorded during the stay in the monitoring unit, no seizure was registered during the night of the HD-EEG recording.

F = female, M = male, R = right, L = left, MTS = Mesial Temporal Sclerosis, FCD = Focal Cortical Dysplasia, RFTC = radiofrequency thermocoagulation, n.a. = not applicable (follow up duration < 1 year), FIA = focal impaired awareness, FA = focal awareness.

¹This patient did not have significant ictal EEG changes,and the focus was found to be superficial based on SEEG exploration, histopathology, and outcome. In order to see epileptic activity at the scalp level in the EEG, an area >8-10 cm² of activation is required even if the generator is superficial [43]. SEEG confirmed at the moment of the clinical manifestation that the ictal activity was very focal and only involved a few electrode contacts.

HD-EEG overnight recordings

HD-EEG data were collected in epilepsy patients during their hospitalization for presurgical evaluation in the epilepsy monitoring unit at the Montreal Neurological Hospital. Recordings were performed with the Nihon Koden system (Tokyo, Japan) using 83 glued electrodes placed according to the 10–10 EEG system with a sampling rate of 1000 Hz. No seizures were recorded during the HD-EEG overnight recordings. HD-EEG in healthy controls was performed with a 256-channel EGI system (Electrical Geodesic Inc., EGI, now Magstim EGI, Eden Prairie, MN, USA) at a sampling rate of 1000 Hz. The 10-10 system was approximated from the full montage [27]. To allow a regional comparison of STW activity, electrodes were grouped into 11 regions: five regions per hemisphere (frontal, central, temporal, parietal, occipital) and the midline (Figure 1A).

Figure 1.

(A) Electrode grouping of 83 electrodes; (B) Example of STW marking in a subset of bipolar channels spacing the parasagittal head region in patient #17. Please note the occurrence of STWs (blue box) shortly prior to a burst of REMs (red box). Phasic control segment (dashed line) and tonic control segment (dotted line) were selected with the same duration as the STW marking. The composition of the final control piece is weighted according to the proportions of phasic and tonic REM sleep during the STW segment in this case 70% tonic and 30% phasic.

Sleep scoring, STW, REM, and IED marking

Sleep was manually scored according to American Academy of Sleep Medicine (AASM) criteria in 30-s epochs displaying the EEG in a mastoid referential montage, electrooculography (EOG) and chin electromyography (EMG) [28]. We selected all REM epochs available for each subject. To study the effect of sleep homeostasis, we compared the first and last sleep cycle. Arousals and artifacts were manually marked and excluded from further analysis.

Although not mandatory for scoring REM sleep according to AASM criteria [28], STWs are characteristic sleep oscillations of REM sleep and visible in the scalp EEG. STWs were marked according to the criteria of bursts of consecutive surface-positive 2–5 Hz frontocentral waves with an amplitude of 20–100 µV [22] in a bipolar montage (Figure 1B). The term STWs will be used throughout the manuscript referring to bursts of STWs. We increased the duration criterion from “≥3 consecutive waves” to “≥2s duration burst of consecutive waves” to improve specificity of our detections [25]. Initial markings were performed by a single sleep expert. Subsequently, the first 10 markings per subject were reviewed by another expert. Overall, the agreement between both scorers was 88.1%, highlighting the reliability of the markings. STW rates were calculated as bursts of STWs per minute.

To specify phasic and tonic REM sleep and to choose adequately matched control pieces for the STW markings, REMs were marked manually on the EOG derivation blinded to the information of the STW marking. Bursts of REMs were marked, when they fulfilled the criterion of “≥3s of sharp onset eye movements clearly standing out of the background EOG” [4, 25]. Bursts of REMs refer to the rate of marked REMs per minute. The duration of bursts of REMs are reported in seconds. IEDs were marked at their peak across all channels by a board-certified epileptologist using a bipolar montage.

Control segments

To define the relative power increase during the STW markings, a background was defined using control segments that were selected close in time and with the same duration as each STW marking [25]. Each variable associated to a control segment consisted of a weighted average of a segment in tonic REM sleep and another segment in phasic REM sleep (Figure 1B). Both segments in tonic and phasic REM sleep preceded the considered STW segment by a minimum of 2 s between STW segment onset and control segment offset. The weights were chosen to respect the proportion of tonic and phasic sleep present in the STW segment. Furthermore, the control segments were chosen outside marked artifacts and arousals.

STW analysis

STWs were characterized by their rate (STWs per minute), duration (s), frequency (Hz), relative power increase compared to the background, and extent (%) calculated as the percentage of channels involved in the STW activity. To determine the frequency of the STW, first a Fast Fourier Transform was performed for each channel during the marked STW segments zero-padded to a length of 20 s to allow for STW segments of long duration. The same procedure was followed for the control segments, and the difference between the STW magnitude and the background (control segments) magnitude was computed. The STW frequency was defined then as the average among channels of the frequency showing the maximum magnitude difference in the 2–5 Hz band between STW and background, weighted by its magnitude. That is, more weight was given to the channels with a higher activity increase during STWs compared to control segments.

The power of the STW segments was defined as the difference of power in the 2–5 Hz band between the STW segment and the weighted control segments, averaged among all channels. We used the relative power increase (in times) to perform the statistical analysis. Finally, the extent of the STWs was defined as the percentage of channels showing a power increase of at least 50% compared to the background.

Statistics

Data were tested for normal distribution with the Anderson–Darling test, and reported as mean ± standard deviation (SD) in case of normally distributed data, or median and range otherwise. Effect sizes were calculated using Cohen’s Delta for normally distributed data and Cliff’s Delta for nonnormally distributed data. Statistical analyses were performed using Matlab 2020a. To correct for multiple testing, a Bonferroni correction was applied for testing our primary hypothesis to compare rate, duration, frequency, power and extent of STW activity between focal epilepsy patients and healthy controls. Therefore, after correction for five tests, a p-value < 0.01 was considered to indicate statistical significance. For the exploratory endpoints, a p-value < 0.05 was considered to indicate statistical significance.

Results

REM macrostructure

A total of 2079 min of REM sleep were available for epilepsy patients, with a median per subject (range) of 86 min (16–228.5 min), and a total of 805 min for healthy controls, with a median of 71.5 min (28.5–101 min). Epilepsy patients had a median of 4 (1–7) sleep cycles, and healthy controls 3.5 (1–5) sleep cycles. Standard sleep parameters of patients and controls are provided in Table 2. Average absolute background power and average absolute power during STWs as well as relative power of STWs and the variability of all 34 subjects are presented in Figure 2. One patient did not show STWs in the overnight recording and in another patient, STWs were only registered in the first sleep cycle. In one patient and two control subjects, only one sleep cycle was observed in the overnight recording.

Table 2.

Sleep macrostructure of patients with epilepsy in comparison to the healthy control group

Sleep parameter	Patients with epilepsy (n = 22)	Healthy controls (n = 12)	Difference patients (n = 22) vs. controls (n = 12)	Temporal lobe epilepsy patients (n = 12)	Extratemporal lobe epilepsy patients (n = 10)
TIB (min)	494.6 ± 115.2	529.1 ± 33.7	p =0.32	434.3 ± 96.7	566.8 ± 94.3
TST (min)	430.9 ± 116.2	448.4 ± 47.3	p = 0.61	370.8 ± 94.0	501.3 ± 102.0
Sleep efficiency (% TIB)	86.4 ± 7.2	84.8 ± 7.6	p = 0.55	85.1±7.9	88.0 ± 6.2
Sleep onset latency (min)	12.3 [4.5; 43.5]	6.7 [3.0; 23.5]	p = 0.008, d = .52	9.4 [4.5; 31.0]	15.0 [7.0; 43.5]
REM sleep latency (min)	147.6 [56.0; 364.5]	174.5 [65.5; 394.0]	p = 0.07	108.8 [56.0; 364.5]	162.5 [80.5; 301.3]
N1 (% TIB)	6.7 ± 3.4	8.5 ± 5.6	p = 0.27	7.5 ± 3.6	5.8 ± 3.1
N2 (% TIB)	38.1 ± 10.2	43.7 ± 8.8	p = 0.12	37.2 ± 11.7	39.2 ± 8.4
N3 (% TIB)	22.6 ± 6.5	20.0 ± 6.0	p = 0.26	22.7 ± 6.9	22.5 ± 6.4
REM (% TIB)	18.9 ± 6.9	12.7 ± 4.1	p= 0.008, d = .57	17.7 ± 6.5	20.5 ± 7.5
WASO (% TIB)	9.3 ± 7.2	11.3 ± 7.3	p = 0.44	10.8 ± 7.8	7.6 ± 6.4

Abbreviations: TIB = time in bed; TST = total sleep time; WASO = wake after sleep onset.

Note: Decreased amount of REM sleep in healthy controls may be explained by inconvenience to sleep with 256-channel EEG.

Figure 2.

(A) Absolute power during STWs averaged over all events from all 34 subjects, (B) average absolute power of the background (weighted combination of control segments), (C) relative power of STWs (increase with respect to background, in times) averaged among all events, (D) and its variability (D). Please note that variability is higher in the frontal channels.

REM microstructure–rapid eye movements (REMs)

Rates of bursts of REMs and their duration were not different between epilepsy patients and controls (rate: 1.13/min ± 0.38 vs. 1.10/min ± 0.38, p = 0.80, d = 0.09, duration: 7.53s ± 1.55 vs. 7.98 s ± 1.61, p = 0.4, d = −0.29). In all subjects, bursts of REMs were shorter in the first compared to the last REM cycle (6.22 s ± 1.71 vs. 7.88 s ± 2.10, p = 0.0012, d = −0.80). This was especially prevalent in patients with epilepsy (6.05 s ± 1.29 vs. 7.87 s ± 1.80, p = 0.0002, d = −1.01) whereas in controls, the duration did not differ between the first and last REM cycle (6.56 s ± 2.43 vs. 7.89 s ± 2.74, p = 0.30, d = −0.51). There was no difference in all subjects in number of bursts of REMs in the first and last cycle (0.94/min ± 0.53 vs. 1.16/min ± 0.47, p = 0.06, d = −0.44). The percentage of phasic REM sleep during all REM epochs was not different in epilepsy patients compared to controls (14.7% ± 6.7 vs. 15.2% ± 7.9, p = 0.85). There was a lower percentage of phasic REM sleep in the first compared to the last sleep cycle (10.3% ± 7.8 vs. 15.5% ± 7.5, p = 0.008, d = −0.64).

REM microstructure–sawtooth waves (STWs)

Epilepsy patients vs. healthy controls

In epilepsy patients, reduced STW rates were found compared to healthy controls (0.64/min ± 0.46 vs. 1.12/min ± 0.41, p = 0.005, d = −0.98) (Figure 3A). Furthermore, duration of STW bursts was decreased in epilepsy patients compared to controls (3.60 s ± 0.76 vs. 4.57 ± 1.00, p = 0.003, d = −1.01) (Figure 3B). Whereas frequency of STWs did not differ between epilepsy patients and controls (2.95 Hz ± 0.29 vs. 2.96 Hz ± 0.11, p = 0.90, d = −0.04), frequency was more variable in epilepsy patients than in controls (p = 0.002) (Figure 3C). Power and extent of STWs were not different in epilepsy patients versus healthy controls (power: 1.78 [1.16–3.01] vs. 1.77 [1.36–2.10], p = 0.51, d = 0.14; extent: 33% ± 14 vs. 29% ± 8, p = 0.34, d = 0.35) (Figure 3D). STW parameters did not differ between patients with daytime seizures (n = 6; three patients with focal awareness and three patients with focal impaired awareness seizures) compared to patients without daytime seizures on the day prior to the sleep recording (n = 16; all ps > 0.05).

Figure 3.

(A) Mean STW rate/min, (B) mean STW duration (s), (C) mean STW frequency (Hz) and (D) median extent (%) of STWs in epilepsy patients and healthy controls. Each dot represents one subject (blue = temporal lobe epilepsy, green = extratemporal lobe epilepsy, red = healthy control).

Epilepsy patients – Regional analysis

Comparing the region with the epileptic focus to the homologous contralateral side, frequency (2.96 Hz ± 0.34 vs. 2.89 Hz ± 0.26, p = 0.08, d = −0.21), power (1.49 [1.15–2.26] vs. 1.46 [0.84–2.85], p = 0.49, d = 0.12) and extent (20% ± 22 vs. 20% ± 14, p = 0.96, d = 0.01) did not differ. There was also no regional difference when selecting patients with an epileptic focus partly or fully in the frontocentroparietal areas (n = 5) all with p > 0.05.

Temporal lobe epilepsy patients vs. extratemporal lobe epilepsy patients

Of 22 patients with focal epilepsy, 12 patients had temporal lobe epilepsy and 10 had extratemporal lobe epilepsy. STW rate, duration, frequency, power and extent for temporal lobe and extratemporal lobe epilepsy patients as well as the control group are presented in Table 3. Extratemporal lobe epilepsy patients showed stronger reductions in STW rates compared to controls than temporal lobe epilepsy patients. The effect of the decrease in duration showed a similar effect size for temporal and extratemporal lobe epilepsy patients.

Table 3.

STW rate, duration, frequency, power and extent for temporal lobe and extratemporal lobe epilepsy patients in comparison to the healthy control group

	Temporal lobe epilepsy (TLE) (n=12)	Extratemporal lobe epilepsy (ExtraTLE) (n=10)	Control group (n=12)	TLE vs control	ExtraTLE vs. control
STW rate	0.80/min ± 0.51	0.45/min ± 0.31	1.12/min ± 0.41	p = 0.10, d = −0.67	p = 0.0004, d = −1.35
STW duration	3.69s ± 0.62	3.49 s ± 0.92	4.57s ± 1.00	p = 0.016, d = −0.95	p = 0.017, d = −0.99
STW frequency	2.88 Hz ± 0.24	3.03 Hz ± 0.32	2.96 Hz ± 0.11	p = 0.30, d = −0.43	p = 0.50, d = 0.30
STW power	1.82 [1.16–2.34]	1.69 [1.29–3.01]	1.77 [1.36–2.10]	p = 0.40, d = 0.21	p = 0.82, d = 0.07
STW extent	33% ± 13	33% ± 16	29% ± 8	p = 0.37, d = 0.38	p = 0.42, d = 0.36

Sleep homeostatic properties – first vs. last sleep cycle

In the first sleep cycle, STW rates were reduced in epilepsy patients compared to healthy controls (0.60/min ± 0.49 vs. 1.10/min ± 0.55, p = 0.016, d = −0.90) (Figure 4A). In the last sleep cycle, the STW rates did not differ between epilepsy patients and controls (0.67/min ± 0.51 vs. 0.99/min ± 0.49, p = 0.11, d = −0.62). Durations of STWs were decreased in epilepsy patients compared to controls in the first and last sleep cycle with a stronger effect in the first sleep cycle (first: 3.60 s ± 0.76 vs. 4.57 ± 1.00, p = 0.003, d = −1.01; last: 3.66 s ± 0.84 vs. 4.51 ± 1.26, p = 0.039, d = −0.80) (Figure 4B). Frequency, power and extent of STWs did not differ between epilepsy patients and controls in the first and last sleep cycle (all p > 0.05) (Figure 4C,D).

Figure 4.

(A) Mean STW rate/min, (B) mean STW duration (s), (C) mean STW frequency (Hz), and (D) median extent (%) of STWs in epilepsy patients and healthy controls in the first and last sleep cycle. Each dot represents one subject (blue = temporal lobe epilepsy, green = extratemporal lobe epilepsy, red = healthy control).

Within epilepsy patients, no differences were found between the first and last sleep cycle in regard to STW rates (0.60/min ± 0.49 vs. 0.65/min ± 0.51, p = 0.43, d = −0.14), duration (3.66s ± 0.84 vs. 3.75 s ± 0.81, p = 0.61, d = −0.1), frequency (2.96 Hz ± 0.33 vs. 2.93 Hz ± 0.29, p = 0.68, d = 0.09), power (1.72 [0.75–4.04] vs. 1.74 dB [0.94–2.38], p = 0.98, d = −0.01) and extent (32% ± 20 vs. 31% ± 16, p = 0.79, d = −0.05).

Within controls, no differences were found between the first and last sleep cycle regarding STW rates (1.10/min ± 0.55 vs. 0.99/min ± 0.49, p = 0.59, d = 0.21), duration (4.51 s ± 1.26 vs. 4.27 s ± 1.15, p = 0.57, d = 0.20), and power (1.49 [0.98–1.99] vs. 1.77 [1.36–2.61], p = −0.08, d = −0.48). Frequency was higher in the first compared to the last sleep cycle (3.11 Hz ± 0.19 vs. 2.89 Hz ± 0.15, p = 0.004, d = 1.07), and the extent was lower in the first sleep cycle compared to the last sleep cycle (21% ± 12 vs. 32% ± 10, p = 0.030, d = −0.85).

Correlation with IED rate

All 22 patients showed IEDs during REM sleep. The median IED rate per patient was 0.25/min (range: 0.025–14.3/min). In our cohort, the IED rate was not significantly correlated with the STW rate (r = −0.11), duration (r = −0.06), frequency (r = 0.02), power (r = −0.12), and extent (r = −0.29) all with p > 0.05. This was also the case when only selecting patients with a superficial or intermediate focus (n = 8) all with p > 0.05.

Discussion

Using combined HD-EEG and polysomnography, we explored STW activity as typical marker of REM sleep microstructure in patients with various focal epilepsies or epilepsy syndromes and a healthy control group. Our main findings are that (i) global STW rate and duration are reduced in patients with epilepsy in comparison to healthy controls, while phasic REM sleep is not (ii) this effect is stronger in the first sleep cycle than the last sleep cycle and, (iii) there was no local reduction between the region with the epileptic focus and the homologous contralateral side, suggesting an effect exceeding the lobe of the epileptic focus.

STW rates and duration are reduced in epilepsy

This is the first work showing a reduction of STW rate and duration in patients with various focal epilepsies or epilepsy syndromes in comparison to healthy controls. In a study in 16 patients with temporal lobe epilepsy using standard polysomnography, global STW rate, duration and frequency were reduced compared to controls [26]. Our work extended this finding by including patients with various focal epilepsies and epilepsy syndromes and using 83-channel HD-EEG to be able to additionally trace focal disturbances. The effect sizes of the decrease in STW activity (d = −0.80–1.35) are similar to those found for sleep spindle reduction in patients with epilepsy compared to healthy controls (range: d = −0.51 to d = −1.21) [12, 15, 29]. The rate of bursts of REMs did not differ between patients with focal epilepsy and controls. No differences in REM density were also found between patients with drug-resistant and controlled epilepsy [30]. Among others, the brainstem plays an important role in REM generation [31]. However, STWs are generated in the cortex and are therefore not only influenced by the brainstem. Therefore, epilepsy might only affect cortically generated REM sleep oscillations such as STWs and not brainstem generated bursts of REMs. The impairment of the REM microstructure may potentially have an influence on dreaming as one sleep-related cognitive process [32]. STW activity may drive multifocal faster activities [24, 25] and a reduction could therefore lead to interference with dreaming or REM sleep memory functions [32, 33].

As expected, given the suppressive effect of REM sleep on epileptic activity [2, 3], IED rates were low in patients with epilepsy. We did not find a correlation between STW activity and IED rates. Maybe not epileptic activity itself, but the underlying neuropathology of epilepsy is responsible for the decrease in STW activity. Patients with focal epilepsy show a broad spectrum of structural lesions and widespread network abnormalities [17] which might be responsible for the alteration of physiological sleep oscillations. In temporal lobe epilepsy, focal impaired awareness seizures during the day and particularly when occurring at night-time during sleep were shown to have an impact on sleep structure and the amount of REM sleep [34]. Six of 22 patients of our cohort showed daytime seizures before the sleep recording. We did not find a difference in STW activity reduction between both patient groups. However, different to Bazil et al. [34], we did not perform within patient comparisons, and hence it might still be possible that daytime seizures impact STW activity in an individual subject.

STW rates and duration are globally but not focally reduced in patients with epilepsy

While global STW rates and duration were decreased between patients with epilepsy and healthy controls, there were no regional differences in STW activity in the region with the epileptic focus compared to the homologous contralateral side. This points to a more global reduction in STW activity instead of a local decrease determined by the epileptic focus as shown in our previous work for NREM sleep transients [15]. Furthermore, epilepsy was shown to have remote effects: IEDs in the hippocampus were coupled with spindles in remote brain regions outside of the epileptic network [35]. Therefore, even regions outside the epileptic focus may be altered in STW activity. Additionally, we found particularly strong effect sizes for extratemporal lobe patients possibly due to the findings that generators of STWs are mostly extratemporal [24, 25] and the maximum of STWs is over the frontocentral regions [22, 23].

Reduction of STW activity is more pronounced in early than late REM sleep.

The impairment in REM microstructure had a stronger effect in the first versus last REM sleep cycle. In human sleep, REM sleep episodes become progressively prolonged over the night with usually a short episode during the first sleep cycle [36] which was also the case for our data. However, STW rates were calculated in relation to the duration of the first and last REM period and therefore not affected by different lengths of REM sleep episodes. Yet, it is known in patients with epilepsy that macro- und microstructure of NREM sleep is altered [6, 7] and the duration of REM episodes is also correlated to the duration of subsequent NREM episodes [37]. Further, REM sleep episodes with preceding awakenings and those ending in wakefulness were shown to be of longer duration [38]. Especially in the beginning of the night, epileptic activity was found to be related to higher amplitude slow waves and pathological high frequency oscillations were more prevalent than at the end of the night [39, 40]. Therefore, the REM episodes in the beginning of the night may show higher disturbance in microstructure due to higher epileptic activity in early NREM sleep. Finally, there was a lower percentage of phasic REM sleep in the first vs. the last cycle which may have contributed to lower STW rates in the beginning of the night.

There are some limitations to this study. The data were collected with scalp HD-EEG. Intracranial EEG provides high temporal resolution but is limited in spatial sampling, whereas HD-EEG covers large parts of the brain and allows a regional investigation [41]. Albeit we cannot exclude that the low correlations of IEDs and STW activity might be due to the fact that IEDs are generated in small cortical areas <6cm² or originate from a deep generator and are therefore not visible in the scalp EEG [42, 43], we found identical results when only analysing the subgroup of patients with a superficial generator. Furthermore, patients and controls showed similar sleep macrostructure and in contrast to previous findings [5], REM sleep was increased in patients. This could be because controls had a different sleep environment and recording system (256 electrodes) than patients which might have even led to an underestimation of the reduction in STW activity. Finally, we cannot rule out that there was an effect of antiseizure medication on STW activity because there are no data available about a potential influence of medication on the physiology of STWs. However, the effect of the reduction in STW activity was shown in a study population with different focal epilepsy syndromes on various medication enhancing generalizability to all adult patients with focal epilepsy.

Conclusion

Our study revealed an impaired REM sleep microstructure in patients with various focal epilepsies or epilepsy syndromes. Despite epileptic activity is low in REM sleep, patients showed a reduction of STWs as cortically generated sleep oscillations. The decrease in STW activity was more pronounced in extratemporal than in temporal lobe patients and was stronger in the first vs. the last sleep cycle of the night. Future studies are warranted to investigate potential functional consequences of impaired REM microstructure in people with epilepsy

Institution where work was performed: Montreal Neurological Institute and Hospital

Disclosure Statement

Financial disclosure: none

Non-financial disclosure: none

Data availability statement

The data underlying this article will be shared on reasonable request to the corresponding author.

Funding

This work was funded by project grants of the Canadian Institutes of Health Research (PJT-175056 to B.F. and FDN-143208 to J.G.) and the Hewitt Foundation to B.F.. K.S. is supported by a Postdoctoral Fellowship of the Savoy Epilepsy Foundation. B.F. is supported by a salary award (“Chercheur-boursier clinicien Senior”) of the Fonds de Recherche du Québec – Santé 2021-2025.

Conflict of interest statement: The authors report no conflict of interest.