Electroencephalography characteristics related to risk of sudden unexpected death in epilepsy in patients with Dravet syndrome

Objective To investigate the quantitative electroencephalography (EEG) features associated with a high risk of sudden unexpected death in epilepsy (SUDEP) in patients with Dravet syndrome (DS). Methods Patients with DS and healthy controls (HCs) who underwent EEG were included in the study. EEG signals were recorded using a 21 channel digital EEG system, and pre-processed data were analyzed to identify quantitative EEG features associated with a high SUDEP risk. To assess the risk of SUDEP, SUDEP-7 scores were used. Results A total of 64 patients with DS [38 males and 26 females, aged: 128.51 ± 75.50 months (range: 23–380 months)], and 13 HCs [7 males and 6 females, aged: 95.46 ± 86.48 months (range: 13–263 months)] were included. For the absolute band power, the theta power was significantly higher in the high-SUDEP group than in the low-SUDEP group in the central brain region. For the relative band power, the theta power was also significantly higher in the high-SUDEP group than in the low-SUDEP group in the central and occipital brain regions. The alpha power was significantly lower in the high-SUDEP group than in the low-SUDEP group in the central and parietal brain regions. Conclusion Patients with high SUDEP-7 scores have different EEG features from those with low SUDEP-7 scores, suggesting that EEG may be used as a biomarker of SUDEP in DS. Significance Early intervention in patients with DS at a high risk of SUDEP can reduce mortality and morbidity. Patients with high theta band powers warrant high-level supervision.


Introduction
Dravet syndrome (DS) is an SCN1A mutation-related, infantileonset epilepsy syndrome, characterized by a distinctive seizure history: prolonged febrile or afebrile seizures beginning in the first year of life, followed by subsequent multiple seizure types.Patients with DS experience developmental regression during early childhood, mostly because of frequent pharmaco-resistant seizures (1)(2)(3).Early diagnosis of DS is critical to avoid anticonvulsants that may aggravate seizures and increase morbidity (4).
Premature mortality is a leading cause of fear among caregivers (5)(6)(7).Up to 15% of patients with DS die during early childhood or adolescence (1), and most of them experience SUDEP (1), which is defined as sudden, unexpected, witnessed or unwitnessed, non-traumatic, and non-drowning death in patients with epilepsy (8).SUDEP accounts for 7.5%-17% of all deaths of epilepsy (9, 10), but SUDEP rates are particularly high in patients with DS attributing to 20% of all deaths (11).The mean age of SUDEP is as young as 4.6 years in patients with DS (6), emphasizing the high risk of mortality.
SUDEP is a diagnosis of exclusion (12), as the definitive postmortem signs or biomarkers of SUDEP have not yet been identified (13,14).Indirect evidence has linked SUDEP to seizure-induced apnea, pulmonary edema, dysregulation of cerebral circulation, and cardiac arrhythmias (9, 10, 15), which may occur secondary to hormonal or metabolic changes or autonomic discharges (9, 15, 16), but the exact mechanisms remain unclear.
Electroencephalography (EEG) is a critical tool that can show the clinical status of patients, as well as ictal changes (17,18) and age-related changes in patients with DS (19).Recently, few studies have analyzed the risk of SUDEP using EEG in patients with various epilepsies.One recent multi-center study has suggested that machine learning-driven models may be used to quantify SUDEP risk in patients with epilepsy (20).Another study used the data of a group of patients with drug-resistant epilepsy to compare the EEG and ECG data of 21 patients with definite or probable SUDEP, and it reported an increased autonomic stimulation associated with seizures in patients with SUDEP (21).However, these studies included patients with various epilepsies, and the results were largely dependent on the ECG data (20,21).
In the present study, we performed a quantitative EEG analysis to identify the EEG features related to a high risk of SUDEP.We hypothesized that the EEG features would be different between patients with high SUDEP scores and the others with lower SUDEP scores.

Participants
This study was performed at Severance Children's Hospital.We enrolled patients who were diagnosed with DS between 1 January, 2012 and 31 March, 2021.DS was diagnosed according to the following criteria: (1) febrile or afebrile, generalized or unilateral, or clonic or tonic-clonic seizures that occur in the first year of life in an otherwise healthy infant; (2) the development of drug-resistant myoclonus, atypical absences, and focal seizures; (3) a developmental delay within the second year of life; (4) the subsequent development of cognitive impairment or other neuropsychiatric and behavioral disorders; and (5) an identified SCN1A gene mutation.Only patients who recently underwent EEG after 1 January, 2020 were included.EEGs were performed as routine clinical care to evaluate seizures and the background EEG activity.The healthy controls (HCs) group included patients who came to the neurologic outpatient clinic due to non-epileptic events such as dizziness or headache.Based on the SUDEP-7 inventory score, patients with DS were then divided into low-(score: 0-3), mid-(score: 4-6), and high-SUDEP (score 7-9) subgroups.This study was approved by the institutional review board (4-2021-0377).Informed consent was waived because of the retrospective nature of the study, and we used anonymous clinical data.

SUDEP-7 score
The SUDEP-7 inventory was assembled from the large prospective cohort study of SUDEP reported by Walczak et al. (22).The core risk factors identified by Walczak et al. ( 22) were consolidated into a seven-item inventory.Risk factors with low odds ratios (0-2) were not included.The risk factor "any seizures, average per month" was consolidated into two core risk factors: any seizures in the last year or more than 50 seizures per month.The weighting for each risk factor was determined by the natural log of the odds ratio rounded to the nearest integer.The weighted SUDEP-7 inventory was scored from 0 to 12.The seven items were shown in Table 1.

EEG acquisition and analysis
The patients lay on an examination bed in a room with ambient noise blocked.EEGs were recorded for at least 30 min.The EEG signals were recorded using a 21-channel digital EEG system (Xltek, Natus Medical Incorporated, San Carlos, California or Telefactor Aurora ® EEG machine, Grass-Telefactor, Melbourne, Australia).The electrodes were attached according to the international 10-20 system.Data were recorded using a sampling rate of 200 or higher with filter settings of 1-70 Hz.Epochs with too many artefacts were removed from the recorded data by visual inspection.The pre-processed EEG data was divided into multiple epochs of a length of 2 s.Power spectral  analysis was used to compress the rhythmic information of the brain wave signals.In the power spectral analysis, the periodogram function in MATLAB R2020a (MathWorks, Natick, MA, United States) was used to calculate the power spectral density of each epoch.The spectral absolute and relative powers were then averaged according to randomly selected 30 epochs.

Statistical analysis
Independent t-tests were used to compare the demographic data.A multivariate analysis of covariance (MANCOVA) was used to compare the absolute and relative EEG band powers between patients with DS and HCs.Age was controlled for as a covariate.
The MANCOVA was used to compare the absolute and relative EEG band powers among the SUDEP subgroups (low-, mid-, and high-SUDEP) of patients with DS. p-values were adjusted for age.Statistical analyses were performed using SPSS 21 (SPSS, Inc., Chicago, IL, United States).There was a difference in age between patients with DS and HCs (p = 0.021), and other demographics are no different.

Results
Demographic data of the DS patients and the HCs are reported in Table 2.

Differences in EEG between the DS and HCs groups
The multivariate analysis of covariance (MANCOVA) was applied to the absolute and relative EEG power between patients with DS and HCs.Age was controlled as a covariate.

Differences in EEG between the low-SUDEP and high-SUDEP group
The MANCOVA was applied to the absolute and relative EEG power between the low-SUDEP and high-SUDEP groups.Age was controlled for as a covariate.
For the absolute band power, the theta power was significantly higher in the high-SUDEP group than in the low-SUDEP group in the central brain regions [23.60 (11.30; 44.02) vs. 47.16 (19.37; 61.79), p = 0.042].All the comparison results of the absolute EEG band power between the high-SUDEP and low-SUDEP groups are presented in Table 5.
Figure 1 presents topographic maps.(A) Differences of relative power in theta band between Dravet syndrome (DS) and healthy

Discussion
SUDEP is a fatal condition that can occur frequently in patients with DS.However, biomarker for predicting SUDEP remains to be investigated.This study demonstrates that patients with high SUDEP-7 scores have different EEG features from those with low SUDEP-7 scores, suggesting that EEG may be used as a biomarker of SUDEP in DS.
The most significant finding of this study was the increase in the theta powers in patients with high SUDEP-7 scores.At the same time, the relative alpha powers of EEG signals decreased in the same group.In previous studies, the 1-30 Hz-frequency range and theta activity (4-7 Hz) showed a consistent relationship with epilepsy.Topographically a diffuse increase in theta activity characterizes the broad spectrum of genetic-developmental, non-lesional childhood epilepsies (23, 24), idiopathic generalized epilepsies (25), and focal epilepsies with dissimilar aetiology (26,27) compared to the theta activity in healthy controls.Previous studies have reported a diffuse theta, and delta slowing in DS (17,28).Modifications of theta and delta rhythms have been reported in SCN1a knock-out mice (29).
Notably, the alpha band powers decreased in DS patients with high SUDEP scores.Alpha-band oscillations play an important role in information processing.Alpha band rhythm increases attention by inhibiting task-irrelevant processes (30-33).In epilepsy, a slower alpha rhythm is associated with poorer seizure control (34).These findings suggest that the dysregulation of alpha activities, which represent cognitive activities, is related to a high SUDEP risk score.Intellectual disability has been suggested as a factor associated with a high SUDEP risk, but data about this hypothesis is still limited (35).Our findings correlate with these previous findings, and demonstrate that alpha is decreased in high-SUDEP group using a quantitative EEG analysis.Our findings suggest that an increased theta and alpha activity may be used to predict an increased risk of SUDEP.
In this study, we also compared the EEG findings of patients with DS to those of HCs.This study showed that absolute beta activities increased diffusely in patients with DS, in which occipital delta activities decreased in patients with DS.The relative alpha and gamma powers of EEG signals decreased in the same group.An increase in diffuse beta activity might be associated with anti-seizure drugs.Sedative benzodiazepines increase beta activity (36).Clobazam, an oral 1,5-benzodiazepine, is the first-line treatment drug for DS (37).
Delta is usually considered the slowest EEG frequency band (38).Delta oscillations are frequently observed in pathologic conditions, including coma (39-41) and Lennox-Gastaut syndrome (42).Increased delta activity can be observed as an ictal (43) or postictal activity (28).However, we found a decrease in delta activity in DS that might be correlated with their age.It is known that the rate of slowing increases with age in DS, suggesting that delta activity increases with the duration of epilepsy (17).Previous findings showed that SUDEP risk is increased in correlation with the duration of epilepsy (44).
In contrast to delta oscillations which usually represent pathologic conditions, gamma waves are the fastest brain waves which mainly occur when brain is highly alert and conscious.Gamma activity represents the finely-tuned inhibitory inter-neuron network (45), specifically γ-aminobutyric acid type A (GABAA) receptor-induced inhibitory postsynaptic currents (46)(47)(48)(49).As the functional connectivity of the brain is modulated by the inhibitory inter-neuron network, dysregulation and reduction of gamma activity are observed in pathological conditions such as Alzheimer's disease (50).A normal gamma frequency represents cognitive functions including sensory processing (51, 52), recognition, and memory (53).Modifications of gamma oscillations have been reported in mouse models of Alzheimer's disease (in which NaV1.1 expression is reduced, leading to epileptiform activities) (54).
This study had some limitations.As our center was a tertiary referral epilepsy center, only patients with severely drug-resistant epilepsy were included.This could have caused a selection bias.In addition, we used a SUDEP-7 inventory score in our investigation instead of the actual event because SUDEP is not a   factors for SUDEP (55)(56)(57).Also, a broad range of age (2-31 years old) was considered.In future research, it may be beneficial to compare subpopulations within different age ranges to highlight potential differences during specific life stages.This could help use this data for early intervention strategies before clinical deterioration and seizures occur.

Conclusion
Here, we identified the quantitative EEG findings that correlate with a high SUDEP-7 score.Patients with high theta band powers, and the low relative alpha band powers warrant high-level supervision.

FIGURE 1
FIGURE 1 Topographic maps.(A) Differences of relative power in theta band between Dravet syndrome (DS) and healthy controls (HCs) groups (DS−HCs); (B) differences of relative power in alpha band (DS−HCs); (C) differences of relative power in theta band between high-SUDEP and low-SUDEP groups (high-SUDEP−low-SUDEP); (D) differences of relative power in alpha band (high-SUDEP−low-SUDEP).

TABLE 3
Comparison of the absolute electroencephalography (EEG), power between patients with Dravet syndrome (DS) and healthy controls (HCs).

TABLE 4
Comparison of the relative EEG power between patients with Dravet syndrome (DS) and healthy controls (HCs).
condition that can be confirmed.However, we still think this study is valuable because this seven-item weighted inventory derived from a prospective SUDEP study (22, 55) is a validated, well-known tool which has been used frequently to identify risk

TABLE 5
Comparison of the absolute EEG power among low-SUDEP, mid-SUDEP, and high-SUDEP groups in patients with Dravet syndrome.