Correlation between scalp high‐frequency oscillations and prognosis in patients with benign epilepsy of childhood with centrotemporal spikes

Abstract Aims The study aimed to explore whether high‐frequency oscillations (HFOs) can predict seizure risk and atypical manifestations of benign epilepsy of childhood with centrotemporal spikes (BECTS). Methods We recruited 60 patients and divided them into three groups: (1) seizure‐free BECTS, (2) active typical BECTS, and (3) active atypical forms of BECTS. Electroencephalogram was used to record the number, location, average amplitude, and duration of spikes, and spike ripples were analyzed using time‐frequency technology. Multivariable logistic regression analysis was used to investigate independent predictive factors for prognosis. Results The number of sleep spike ripples, rather than spikes, was an independent risk factor for the active period of the disease (odds ratio [OR] = 4.714, p = 0.003) and atypical forms of BECTS (OR = 1.455, p = 0.049); the optimal thresholds for the spike ripple rate were >0 (area under the curve [AUC] = 0.885, sensitivity = 96.15%, specificity = 73.33%) and >0.6/min (AUC = 0.936, sensitivity = 84.21%, specificity = 96.15%), respectively. Furthermore, in typical BECTS, the spike ripple rate showed significant negative correlations with time since the last seizure (ρ = −0.409, p = 0.009) and age (ρ = −0.379, p = 0.016), while the spike rate did not. Conclusion Spike ripple was a marker for distinguishing typical and atypical forms of BECTS and reflected the risk of seizure recurrence better than the spike alone. The present findings might assist clinicians in BECTS treatment.


| INTRODUC TI ON
Benign epilepsy of childhood with centrotemporal spikes (BECTS) is the most common childhood epilepsy syndrome remitting in late adolescence. 1,2 Electroencephalogram (EEG) shows high-amplitude centrotemporal spike waves activated during sleep 2 in patients with BECTS. Seizure frequency is usually low, but some children suffer from frequent seizures and require continuous treatment. 3 A minority of patients might have common manifestations and EEG features at the early stage but more severe seizures and neuropsychological impairment later, which are called the atypical forms of BECTS. 4,5 Previous investigations on the prognosis of BECTS focused more on spikes, but the prediction accuracy of spikes was low. 6 High-frequency oscillation (HFO) is an EEG indicator that mainly consists of ripples (80-250 Hz) and fast ripples (250-500 Hz), which are more closely related to epileptogenesis than spikes. [7][8][9] HFOs were first recorded with a microelectrode 10 and then with an invasive intracranial electrode clinically. 11 In 2010, Kobayashi et al. 12 originally described HFOs in scalp EEG of continuous spike-waves during slow-wave sleep (CSWS). We also previously reported HFOs in eight patients with BECTS in a cognition-related study. 13 Notably, scalp ripples have been confirmed to be related to epileptic activity in children with West syndrome 14 and respond to antiepileptic drugs (AEDs) better than spikes in BECTS. 15 More ripples might indicate more frequent seizures in patients with rolandic spikes. 16,17 Moreover, accumulating evidence indicates that scalprecorded HFOs are more prominent in atypical and symptomatic rolandic epilepsy and CSWS than in typical BECTS, 8,15,18 and ripples co-occurring with spikes have greater pathological significance than ripples alone. [16][17][18] A few studies have identified scalp-recorded HFO as a new biomarker; however, these studies have several limitations due to sample capacity and some clinical confounding factors. 8,[16][17][18] Moreover, there is debate about its cut-off value for distinguishing between typical and atypical forms of BECTS and estimating disease activity. 16,17 Accordingly, this study aimed to evaluate the relationship of spikes and spike ripples with the outcomes of BECTS and identify EEG biomarkers to predict seizure risk and atypical manifestations.

| Patients
We recruited 60 patients who visited the Department of Neurology, Xuanwu Hospital, Capital Medical University, between July 2018 and August 2022. We reviewed the following clinical information: the age at seizure onset, seizure types, seizure frequencies, time since the last seizure, and medication conditions. Based on the clinical manifestations, we divided these children into three groups: (1) typical seizure-free BECTS, (2) typical BECTS with active disease, and (3) active atypical forms of BECTS.
The diagnosis of typical BECTS followed the 1989 ILAE criteria: (1) focal seizures, which may be secondary to the general tonic-clonic seizures; (2) seizures usually occurring during sleep; (3) interictal EEG showing normal background and centrotemporal spike waves activated by sleep; and (4) normal neurological examination. The following criteria were used to define the patients with atypical features 1 : (1) early onset of typical BECTS and transition to more serious seizure forms, such as negative myoclonus and atypical absences; (2) transient oromotor dysfunction, such as hypersalivation or difficulties in articulation and pronunciation; (3) EEG showing a spike-wave index ≥50% during non-rapid eye movement (NREM) sleep, and the patients had cognitive abnormalities. Children with other severe systemic diseases or magnetic resonance imaging abnormalities were excluded from this study. Patients with active disease were defined as having had at least one seizure within the last 1 year, while seizure-free patients had no seizure for at least 1 year, as described previously. 17 Twelve months of seizure freedom indicates a low risk of recurrence in the majority of children with epilepsy. 19 The study protocol was reviewed and approved by the Ethics Committee of Xuanwu Hospital, Capital Medical University, China.
Informed consent was obtained from all the participants and their guardians.

| EEG acquisition
Scalp EEGs were recorded using a Bio-logic and Nihon-Kohden Neurofax system with a 10-20 electrode placement system at a sampling rate >500 Hz. All children underwent video-EEG monitoring for 4 h to record awake and sleep periods lasting at least 30 min.

| EEG analysis
We randomly selected EEG segments with artifacts free at stage 2 or 3 of NREM sleep for 10 min and an awake state for 5 min. We scored the following: (1) sleep and awake spike rate; (2) location of spikes, especially the frontal EEG focus; and (3) sleep spike ripple rate. We also chose 60 spikes (minimum of 20) in each segment to measure the amplitudes and durations of the prominent negative wave with a cursor and computed their average values. 20,21 As described previously, 16,17 we analyzed the EEG data using an average montage. Spikes were marked in the EEG traces using a low frequency of 0.53 Hz (10 s/page), while ripples were identified in the temporal expansion (2 s/page) of the EEG with a low-cut filter at 80 Hz. 15,18 Ripples were defined as events of 80-250 Hz with at least four consecutive oscillations standing out in morphology and amplitude from the background in the 80 Hz high-pass-filtered signal. 22 If a ripple occurred 50 ms before or after the spike peak, the event was marked as a spike ripple. 23,24 One reviewer marked the ripples, which were subsequently checked and discussed with a second reviewer. We counted the moment of the event regardless of the number of channels involved. 16,25 The EEG segments containing visually inspected HFOs were further subjected to time-frequency analysis using a wavelet transform. On the spectrogram of each inspected HFO, only a "spectral island" in the frequency range of 80-200 Hz was defined as a true HFO 12,15,18 (Figure 1).
The analysis was performed using MATLAB R2020a (Mathworks Inc.).

| Statistical analysis
We analyzed the distribution of continuous variables using the Shapiro-Wilk test. Then, to compare the three groups, we used oneway ANOVA or the Kruskal-Wallis H test, followed by Bonferroni statistical tests to avoid cumulation of type-I errors. Categorical variables of clinical or EEG data were analyzed using the chi-squared test. To adjust for confounding factors, the variables having a p-value less than 0.1 were selected in multivariate logistic regression analysis. 26 A receiver operating characteristic (ROC) curve was used to evaluate the predictive value of the independent predictor, and the results were reported as areas under the curves (AUCs). Moreover, we estimated the correlation of spikes and spike ripples with age and the time to last seizure using partial correlation analysis.
Statistical significance was set at a p-value less than 0.05. All the statistical analyses were performed using SPSS for Windows, ver-

| Demographic and clinical data
Sixty children (33 boys and 27 girls) met the inclusion criteria: 15 with typical seizure-free BECTS, 26 with active typical BECTS, and 19 with active atypical forms of BECTS. The age at seizure onset was lower in patients with atypical forms of BECTS than in those with typical BECTS (p = 0.004). Children in the seizure-free group were marginally older than those in the active group but the difference was insignificant (p = 0.057) ( Table 1).

| Spikes and ripples in three groups
There were significant differences among these three groups in the awake spike rate (p < 0.001), spike amplitude (p = 0.004), spike duration (p = 0.009), number of spikes (p < 0.001), and spike ripples (p < 0.001) during the NREM sleep stages. Moreover, compared with children with seizure-free BECTS, those with active BECTS had a significantly higher number of spike ripples (p = 0.008); however, there was no significant difference in the spike rate during the awake (p = 0.240) and sleep (p = 0.462) periods. In addition, the awake spike rate (p = 0.009), number of spikes (p = 0.001), and spike ripples (p < 0.001) during the NREM sleep were significantly higher in patients with atypical forms of BECTS than in those with active BECTS (Figure 2). However, there was no significant difference in spike amplitude and duration between seizure-free and active BECTS or between typical and atypical forms of BECTS ( Table 1).
The frontal focus was most common in atypical forms of BECTS, but there was no significant difference in the frequency of frontal focus among different groups (p = 0.715).
Multivariate analysis was carried out using the aforementioned variables with p-values less than 0.1. Spike ripple was regarded as an independent predictor of seizure activity and atypical forms of BECTS, with adjusted ORs of 4.714 and 1.455, respectively (Tables 2).
3.2.3 | Correlation of spikes and spike ripples with age and the time since the last seizure The time to the last seizure was longer, and the age of children was older; thus, these factors were included in the partial correlations.
In typical BECTS, after controlling for the time since the last sei-  Table 3).
In atypical forms of BECTS, we did not find a similar correlation between these indicators (Table S1).

| DISCUSS ION
Most childhood-onset BECTS will remit in adolescence, but some children with BECTS may suffer from frequent seizures or atypical presentations over time. Lack of treatment or premature withdrawal of drugs might lead to seizure recurrence and neurocognitive impairment. 13 However, AEDs may cause adverse reaction. 27 Thus, it is important to estimate the atypical course and possibility of seizure recurrence in patients with BECTS. In this study, our findings suggest that scalp spike ripple may be not only a predictor for atypical forms of BECTS but also an indicator to reflect the conditions of seizure control and spike ripples may have age-dependent changes.

| Typical and atypical forms of BECTS
Previous studies 16,28,29 found that the age at seizure onset was younger, and the rates of sleep and waking spikes were higher in chil-  29 In addition, Kanemura et al. 29,33,34 suggested that in atypical BECTS, the frontal focus was sustained for an extended period in sleep EEGs associated with brain maturation. We discovered a higher proportion of frontal focus in children with atypical forms of BECTS, but no significant difference, probably due to the lack of follow-up.
Spike ripple in scalp sleep EEGs was regarded as a factor for distinguishing between typical and atypical forms of BECTS. In 2016, van Klink et al. 16 recruited eight patients with typical rolandic epilepsy and eight patients with atypical or symptomatic rolandic epilepsy to analyze their EEGs. They found that the presence of more than five spike ripples in EEGs of 10 min was a predictor for atypical or symptomatic epilepsy. The EEGs they analyzed included those during awake or sleep periods; notably, spikes and ripples were activated during sleep. 2,35 In our investigations, we included a larger sample size of children with typical and atypical forms of BECTS and analyzed the data using a multivariate logistic regression model to exclude the disturbance of confounding factors. We found that spike ripple rates >0.6/min might indicate atypical presentations. Based on these findings, we speculate that the disturbance of pathological HFOs to immature brain networks might change the TA B L E 1 Patient characteristics in three groups. Abbreviations: BECTS, benign epilepsy of childhood with centrotemporal spikes; p 1 , difference between active and seizure-free BECTS; p 2 , difference between typical and atypical forms of BECTS. Childhood epilepsy with centrotemporal spikes is a clinical spectrum consisting of syndromes with various severities. Previous studies 8,12 confirmed that HFOs were more prominent in CSWS than in BECTS. Based on the findings and our present findings, we speculate that the number of HFOs might increase gradually from BECTS to atypical BECTS and CSWS. In the early course, it is difficult to distinguish typical and atypical forms of BECTS using traditional EEGs.

BECTS with atypical presentations (n = 19) p-Value
To a certain extent, comparing the rate of spikes between the two entities mentioned above is unreliable, but the spike ripple is a more accurate indicator.

| Seizure activity of BECTS
Previous cohort studies [36][37][38] confirmed that seizure recurrence of BECTS was associated with higher spike rates in the awake or sleep period and more frontal focus sustained for a long period. Our present findings indicated similar results, but the difference was not statistically significant. Interictal epileptic discharge has been demonstrated to be the primary biomarker for predicting remission and recurrence of BECTS, but its accuracy is low. 6 Interestingly, according to earlier researches, spike rates in BECTS always increase in the 6 and 12 months after onset and then decrease over time, 37 which might explain the weak correlation between seizure activity and the number of spikes. 17 Therefore, when a patient experiences an improvement in seizures, the sleep spike rates might increase.
The spike ripple rate is an independent indicator of seizure control. The relationship between HFOs and disease activity has long been discussed in studies. In the last century, scholars explored the relationship in kainic rats and confirmed that the earlier the HFO occurrence, the shorter the latent period of seizure occurrence. 39 A study of intracranial recordings later found that HFOs were more sensitive to AEDs reduction than spikes. 9   and specificity of 73.33%. Previous finding 16 showed that children with rolandic spikes but without epilepsy usually had less than two ripples in the EEGs of 10 min, which was slightly different with our results, probably because these children the study recruited had developmental problems. HFOs might reflect pathologically interconnected clusters of principal neurons discharged in common. 40 Thus, spikes co-occurring with ripples are more closely related to epileptogenesis and may provide greater reliability for predicting active BECTS.
In addition, Kobayashi et al. 18 and Kramer et al. 17 showed that in patients with BECTS, the time since the last seizure was shorter in those with more spike-related ripples than in those with spikes alone. We reached a similar conclusion and confirmed the strong relationship between spike ripples and disease activity. Moreover, we also found that spike ripples rather than spikes decreased with age in patients with BECTS; the finding is consistent with that in the study by Ohuchi et al. 8  Therefore, previous studies on scalp ripples in patients with epilepsy focused on the ripple overlapping the spike, 16,17 which is associated with epileptogenic properties. 8 In other words, it is reasonable to primarily analyze ripples co-occurring with spikes.
In this study, most of the patients recruited had taken AEDs. We did not find a significant difference in the use of AEDs between the three groups. Therefore, medication might not sufficiently explain the differences in spikes and ripples among the groups. Certainly, in the future, we could design a prospective study to distinguish the drug effect from the remission of the disease.
Nevertheless, our study has some limitations. Prior research have discussed the effects of development and aging on HFOs.
Somatosensory evoked HFOs were enhanced in both healthy children and aged subjects. 45,46 However, the relationship between age and scalp HFOs in epileptic patients is still controversial. 43,47 Future studies could include more patients with various types of epilepsy to explore the changes in physiological and pathological scalp HFOs with age.

| CON CLUS ION
Considering the variation in the severity of BECTS and the potential adverse reactions to AEDs, it is difficult for clinicians to determine whether to use AEDs and when to withdraw the drugs. This study indicates that spike ripple is a non-invasive biomarker for estimating the risk of seizure recurrence and atypical presentations. The present findings would help prevent unnecessary or insufficient treatment and assist drug screening.

CO N FLI C T O F I NTER E S T S TATEM ENT
None declared. TA B L E 3 Partial correlation of spikes and spike ripples with age and the time since the last seizure in BECTS.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data are not publicly available due to privacy restrictions. The data that support the findings of this study are available from the corresponding author upon reasonable request.