Termination of seizure clusters is related to the duration of focal seizures

Transcription

1 FULL-LENGTH ORIGINAL RESEARCH Termination of seizure clusters is related to the duration of focal seizures * Victor Ferastraoaru, Andreas Schulze-Bonhage, *Richard B. Lipton, Matthias D umpelmann, * Alan D. Legatt, Julie Blumberg, and * Sheryl R. Haut SUMMARY Dr. Victor Ferastraoaru is Assistant Professor of Clinical Neurology at Albert Einstein College of Medicine/Montefiore Medical Center. Objective: Clustered seizures are characterized by shorter than usual interseizure intervals and pose increased morbidity risk. This study examines the characteristics of seizures that cluster, with special attention to the final seizure in a cluster. Methods: This is a retrospective analysis of long-term inpatient monitoring data from the EPILEPSIAE project. Patients underwent presurgical evaluation from 2002 to Seizure clusters were defined by the occurrence of at least two consecutive seizures with interseizure intervals of <4 h. Other definitions of seizure clustering were examined in a sensitivity analysis. Seizures were classified into three contextually defined groups: isolated seizures (not meeting clustering criteria), terminal seizure (last seizure in a cluster), and intracluster seizures (any other seizures within a cluster). Seizure characteristics were compared among the three groups in terms of duration, type (focal seizures remaining restricted to one hemisphere vs. evolving bilaterally), seizure origin, and localization concordance among pairs of consecutive seizures. Results: Among 92 subjects, 77 (83%) had at least one seizure cluster. The intracluster seizures were significantly shorter than the last seizure in a cluster (p = 0.011), whereas the last seizure in a cluster resembled the isolated seizures in terms of duration. Although focal only (unilateral), seizures were shorter than seizures that evolved bilaterally and there was no correlation between the seizure type and the seizure position in relation to a cluster (p = 0.762). Frontal and temporal lobe seizures were more likely to cluster compared with other localizations (p = 0.009). Seizure pairs that are part of a cluster were more likely to have a concordant origin than were isolated seizures. Results were similar for the 2 h definition of clustering, but not for the 8 h definition of clustering. Significance: We demonstrated that intracluster seizures are short relative to isolated seizures and terminal seizures. Frontal and temporal lobe seizures are more likely to cluster. KEY WORDS: Seizure cluster, Repetitive seizures, Seizure duration, Clustering effect, Seizure localization. Accepted March 2, 2016; Early View publication March 31, *Department of Neurology, Albert Einstein College of Medicine and Montefiore Medical Center, Bronx, New York, U.S.A.; Comprehensive Epilepsy Management Center, Montefiore Medical Center, Bronx, New York, U.S.A.; Epilepsy Center, University Medical Center Freiburg, Freiburg, Germany; and Department of Neuropediatrics and Muscle Disorders, University Medical Center Freiburg, Freiburg, Germany Address correspondence to Victor Ferastraoaru, Montefiore Medical Center, Department of Neurology, th St, Bronx, NY 10467, U.S.A. vferastr@montefiore.org Wiley Periodicals, Inc International League Against Epilepsy When seizures occur in close temporal proximity they are said to cluster. Operational definitions of seizure clusters have varied. 1 3 Clustering is an important phenomenon because of the associated morbidity and mortality risk, because of the potential impact on seizure localization during epilepsy monitoring, 4 7 and because of the opportunity for improving treatment. In patients whose seizures usually cluster, the initial seizures may provide an opportunity for short-term preventive intervention. 1 Clarifying the 889

2 890 V. Ferastraoaru et al. Key Points The terminal seizures of a cluster are longer in duration than intracluster seizures and similar in duration to the isolated seizures Seizures of frontal and temporal lobe onset are more likely to cluster than those from other regions Seizures within 4 h of a prior seizure may not represent independent seizures for the purpose of surgical localization (cluster effect) International EEG data collaborations present the opportunity to explore important aspects of neurophysiology neurophysiologic basis for clustering may facilitate the identification and assessment of appropriate treatments. Seizures are not randomly distributed in time. Sometimes seizures tend to occur in groups, suggesting a self-triggering capacity (inherent capacity of seizures to trigger seizures). 2 Proposed mechanisms for seizure initiation and termination include local changes in neuronal excitability, perturbation of large networks, and effects on distant systems. 8,9 This study uses epilepsy monitoring data to clarify the characteristics of seizure clustering. We hypothesized that seizure clusters reflect a failure of inhibitory mechanisms, in part due to seizures being too short to activate inhibitory mechanisms. Based on this hypothesis, we predicted that in comparison to isolated seizures, the early seizures of a cluster would be relatively short and by virtue of their brevity, fail to activate inhibitory mechanisms. In addition, we hypothesized that the terminal seizure of a cluster would be of longer duration, equivalent in length to an isolated seizure, and therefore able to activate inhibitory mechanisms terminating a cluster. Methods Study data This study is a retrospective analysis of long-term electroencephalographic (EEG) monitoring data for patients admitted to University Medical Center Freiburg between February 2002 and November 2009 included in the EPILEPSIAE (Evolving Platform for Improving Living Expectation of Patients Suffering from IctAl Events) project. EPILEPSIAE is a European research project that has as its output the European Database on Epilepsy, one of the largest available databanks for patients with epilepsy The database includes only patients who underwent presurgical evaluation. It contains metadata on surface and intracranial EEG recordings and clinical annotations of all seizures. For each patient, the clinical and electrographic seizure onset and offset, also the electrographic spatial propagation, the seizure semiology, and type (e.g., focal with or without dyscognitive features or evolving bilaterally) are marked by experienced electrophysiologists and verified by a second expert in each case. The seizure durations were calculated as the interval from the electrographic onset to electrographic offset. The information available also includes extensive clinical information about patients (e.g., patient s age, brain imaging findings, epilepsy etiology, seizure frequency, and antiepileptic medication withdrawal during monitoring). This study was approved by the institutional review boards, and informed consent was obtained from each patient. The data analyzed were deidentified; patient specific identifiers had been removed. Seizure clustering Several definitions of seizure clustering have been proposed. 1,13 For this study, we defined clustering as at least two consecutive seizures with interseizure intervals of <4 h. Seizures that did not meet the clustering criteria were defined as isolated seizures. The last seizure in a cluster was defined as the terminal seizure, and any other seizure during a cluster was defined as an intracluster seizure, including the first seizure of a cluster (Fig. 1). Because the optimal definition of clustering is uncertain, and to assess the robustness of our findings, as a sensitivity analysis we also examined interseizure intervals of 2 and 8 h. Some previous papers required three consecutive seizures for a cluster. 1 To be consistent with previous papers derived from inpatient monitoring data looking at the cluster effect, 6 we required at least two consecutive seizures within the time frames. Statistical analysis Both seizure duration and the logarithmic transform for seizure duration were compared for intracluster seizures versus terminal and isolated seizures across groups. Seizure durations were compared using a Mann-Whitney U test and median test for nonparametric data. The seizure duration data was then normalized using logarithmic transform (18), and the logarithmic transform data were analyzed using independent sample test (t-test) for confirmation. For descriptive purposes, the groups information is displayed for calculated means at that juncture. The seizure type (focal, i.e., remaining restricted to one hemisphere vs. evolving bilaterally), localization of seizure onset (both by lobe and temporal vs. extratemporal), and the concordance among pairs of consecutive seizures regarding seizure origin were compared for the intracluster and terminal/isolated seizures for each case definition applying Fisher s exact test. To determine how well the model fits the data, a logistic regression was completed using IBM SPSS Statistics Version (IBM Corp.).

3 891 Termination of Seizure Clusters Figure 1. Illustration of seizure type definition. Epilepsia ILAE Results Study sample Data from 96 patients were available, consisting of a total of 1,282 seizures during 856 total days of monitoring. Thirty-one of the subjects (32%) underwent intracranial recordings, and 65 underwent scalp recording. There were 19 patients (20%) who were 18-yearsold or younger. The demographic and baseline characteristics of the initial population are summarized in Table 1. Eligibility criteria were applied both to patients and seizures. Of 96 patients, 4 were deemed ineligible and excluded from this analysis. Patients had to have at least three seizures; a patient with only two seizures was excluded. Patients with more than 50 seizures (3 SD [standard deviations] above the mean) were ineligible; three patients were excluded on this basis (these included extremely brief interseizure intervals, which suggested a lack of discrete events). The study sample comprised 92 subjects experiencing 3 31 seizures each. Among eligible patients, some seizures were deemed ineligible. Seventy-four of the seizures occurred on 38 days when the rescue medications were delivered. These observations were excluded from the analysis, as the rescue medications could alter the seizure recurrence/clustering rate. Six seizures had a duration of 30 min or longer and were excluded, as these seizures met criteria for status epilepticus. Four seizures that were shorter than 4 s were also eliminated from examination. The remaining 996 seizures were eligible for analysis (Fig. 2). If seizures were randomly distributed in time, they should follow a Poisson distribution with a mean number equal to the variance. We tested the hypothesis that the seizures were randomly distributed in time, following a Poisson distribution. The variance of the number of seizures per each 4 h interval was compared to the mean seizures per each 4 h interval (in the Poisson model, the mean and the variance are equal). A variance greater than the mean indicates a clustered pattern. None of the patients with clustered seizures had seizures following Poisson distribution during the monitoring. Table 1. Demographic data and seizure characteristics for the initial population Value (mean standard deviation) Age (years) Monitoring time (days) Seizures during monitoring (number) Seizure foci per patient (number) Seizure duration (seconds) 189 1,611 Time to first seizure (days) Value (percent) Seizure localization Temporal 61.8 (33.5 left and 28.3 right side) Extratemporal 28.9 (22.7 frontal, 2.4 central, 2.4 occipital, and 1.4 parietal) Nonlocalizable 9.3 Seizure classification Focal 20.3 Focal with dyscognitive features 35.2 Focal evolving bilaterally 6.6 Unclassified 37.9 Relative frequency of clustering The distribution of seizure clusters is indicated (Fig. 2). Overall, 77 (83%) of the 92 subjects met the criteria for 4-h clustering (at least two seizures separated by an interseizure interval of 4 h or less). The subjects with seizure clusters had between 1 and 7 clusters, with 2.25 (1.33) [mean ( SD)] clusters per person on average (median 2, interquartile range [IQR] 2). There were 422 isolated seizures and 174 clusters, with an average of 3.2 seizures per cluster. In this population, 47 patients (51%) had three or more seizures in 24 h or less, 44 patients (48%) had three or more seizures in 8 h or less, and 33 patients (36%) had three or more seizures in 4 h or less. Seizure duration The average seizure duration was 84 s (86s). There were 656 focal only (unilateral) seizures, 61 seizures evolving bilaterally, and 279 seizures with unclear classification. The mean duration for focal only (unilateral) seizures was 90 s (86 s) and for secondarily generalized seizures it was 144 s (99 s) (p < 0.001). The seizure duration for each

4 892 V. Ferastraoaru et al. localization is presented in Figure 2. Shortest duration seizures were the seizures with frontal lobe origin (mean s [SD]); the temporal lobe seizures were longer (mean s) (p < 0.001). Seizure duration did not differ significantly for intracranial versus scalp recordings (90 70 vs s, p = 0.107). Seizure data were compared for the terminal seizure in a cluster and the intracluster seizures (any other seizure of a cluster). There was a significant difference in seizure duration, the intracluster seizures being shorter than the last seizure in a cluster (mean duration 78 s vs. 95 s for the 4-h definition, p = 0.011) (Table 2). All these findings were confirmed for the logarithmic transform of seizure durations. Next, we compared the intracluster seizures with the terminal cluster seizures and with the isolated seizures separately. There was a statistically significant difference in seizure duration, with the duration being shorter for the intracluster seizures than for both the terminal seizures and the isolated seizures (Table 2). These findings were confirmed by showing no statistically significant difference between terminal and isolated seizures. In terms of duration, the last seizure in a cluster more closely resembles the isolated seizures than it resembles the intracluster seizures. To confirm that these findings were not influenced by within-person effects, the outliers in terms of seizure clusters were eliminated and the analysis was repeated (there was one subject with seven seizure clusters; the other subjects had between one and five clusters). Similar results were obtained: intracluster seizures were shorter than the terminal and isolated seizures (78 s vs. 93 s vs. 86 s, respectively, p = 0.036). The preceding findings were similarly significant for a 2-h definition of clustering, but not for an 8-h definition of clustering. Further analysis was undertaken to evaluate whether the duration of the initial seizure in a cluster was a determinant factor in the subsequent evolution of the cluster. The average seizure duration for the first seizure in each cluster was determined (173 seizures could be analyzed, with an average duration of 79.3 s 65), and these were divided into relatively short (lower than average) versus long durations. The number of clustered seizures following the initial seizure was determined (range 1 18). The short first seizures in a cluster were followed on average by 2.5 seizures (median 2) and the long seizures by 1.8 seizures (median 1) (p = 0.027). Seizure localization The overall seizure localization distribution for the study population is presented in Figure 2. Regarding clustering, Figure 2. Data flow. Epilepsia ILAE Table 2. Analysis of seizure duration based on seizure position in relation to a cluster Comparison of Comparison of intracluster intracluster seizures Mean standard deviation (seconds) seizures versus terminal versus terminal cluster cluster seizures versus isolated seizures Definition Intracluster seizures Terminal cluster seizures Isolated seizures p-value p-value ISI 4 h ISI 2 h ISI 8 h Significant difference for 2- and 4-h definitions, but not for the 8-h definition of clustering. ISI, interseizure interval.

5 893 Termination of Seizure Clusters seizure localization (temporal vs. extratemporal) was not a contributing factor to seizure position in a cluster. Subgroup analysis demonstrates that temporal and frontal seizures are more likely to cluster when compared with other localizations (e.g., occipital, parietal, or central). For example, 59% of the frontal and temporal seizures and only 35% of the occipital, parietal, and central seizures were part of a cluster (p < 0.001). Frontal and temporal seizures were similar in terms of clustering (83 of 136 frontal seizures clustered and 408 of 686 temporal seizures clustered, p = 0.736). Occipital, parietal, and central seizures were similar in terms of clustering (12 of 24 occipital seizures, 5 of 16 parietal, and 7 of 27 central seizures clustered, p = 0.24). Seizure type Of the total 996 seizures, 65.9% (656 seizures) were focal only, 6.1% (61 seizures) evolved bilaterally, and 28% (279 seizures) could not be classified. When we compared the last seizure in a cluster with the intracluster seizures and with the isolated seizures, we found no correlation between the seizure type (focal vs. evolving bilaterally) and the seizure position in a cluster, across all three definitions for clustering. Although focal seizures were shorter in duration than the seizures that evolved bilaterally, there was no statistically significant difference between the group of clustered seizures and the group of isolated seizures regarding the proportion of seizures evolving bilaterally for any of the clustering criteria. For example, for the 4-h definition, 22 (8.2%) of 268 intracluster seizures evolved bilaterally, compared to 39 (8.8%) of 449 terminal and isolated seizures (p = 0.825). Clustering effect Seizure concordance for site of origin was measured in this population through a separate analysis. There were 33 patients with two or more seizure foci, with a total of 357 seizures that met the initial inclusion criteria in the study sample. For each pair of consecutive seizures, the seizures were considered concordant if both seizures had the same localization. If the seizures had different localizations within the brain, the seizures were considered discordant. The seizure pairs for which one or both seizures were nonlocalizable were excluded from analysis. In total there were 239 seizure pairs eligible for this analysis. First, the rates of concordance for clustered versus isolated seizures were compared. Of the clustered seizures pairs, 76% were concordant, whereas only 56% of the isolated seizures were concordant. Second, the rates of clustering for concordant versus discordant seizures were compared. Using the same clustering definition, 65% of concordant seizures clustered, whereas only 42% of discordant seizures were part of a cluster (p = 0.001). Similar results were demonstrated for the 2-h definition for clustering (Table 3). There were no statistically significant differences for the 8-h definition. Data modeling Clustering was analyzed using regression analysis for the characteristics of the seizures tested, including localization, classification, and duration/logarithmic transform of seizure duration. Variables that remained significant in the models, for example, for the 4-h definition, included localization (frontal and temporal seizures are more likely to cluster, p = 0.009) and logarithmic transform of seizure duration (significant for predicting intracluster seizures vs. terminal and isolated seizures, p = 0.04). Variables that were not determinants for seizure position with respect to a cluster included seizure type (focal vs. evolving bilaterally, p = 0.762). Discussion This study uses epilepsy monitoring data to characterize the duration of isolated seizures, intracluster seizures, and the seizures that terminate a seizure cluster. Osorio et al. 2 had suggested that at least some seizures have an inherent self-triggering capacity. Seizure clusters may be an expression of this self-triggering capacity. When clusters do not occur, we hypothesized that postictal inhibitory mechanisms overcame the self-triggering capacity. If longer duration seizures are required to activate inhibitory mechanisms, we predicted that intracluster seizures that do not sufficiently activate the inhibitory mechanisms would be of shorter duration and that isolated seizures and terminal Table 3. Analysis of seizure localization concordance among pairs of consecutive seizures a,b Percent of concordant seizure localization Percent of seizure clustering Definition Clustered seizures Isolated seizures Concordant seizures Discordant seizures p-value ISI 4 h 76% 56% 65% 42% ISI 2 h 77% 59% 44% 25% ISI 8 h 72% 63% 76% 68% Significant for 2- and 4-h definitions of clustering. ISI, interseizure interval. a Analysis is restricted to the 239 seizure pairs that occurred in patients with more than one seizure focus. b If two consecutive seizures have the same localization, they are classified as concordant, if they have different localization they are classified as discordant.

6 894 V. Ferastraoaru et al. seizures would be of longer duration. As predicted, we showed that the last seizure of a cluster was longer in duration than the intracluster seizures and similar in duration to the isolated seizures. These characteristics may help elucidate the influence of one seizure on the probability of next, setting the stage for predicting seizure recurrence and developing acute treatment strategies for seizure clusters. Prior research has examined the duration of seizures in relation to their spatial and temporal distribution, but not directly in relation to clustering Research on epileptogenesis models suggests two distinct processes One possibility is that transient postictal refractoriness decreases the probability of seizure recurrence in proximity to an initial seizure. Under this hypothesis, clustering may result from inadequate postictal inhibition. A second possibility is that, over time, seizures persistently or permanently lower seizure threshold. 23 Our study suggests that a mechanism related to refractoriness may outlast the immediate postictal phase, which in classical neurophysiology is considered as refractory. 25 Inadequate postictal inhibition could be related to the development of status epilepticus. In this study, we excluded status epilepticus, defined by a seizure duration of >30 min. As it has been reported previously, patients who experience seizure clusters may be at risk for status epilepticus and vice versa. 7 This remains an important area for future studies, to explore the relationship, if any, between status epilepticus and clustered seizures. Temporal and frontal seizures were more likely to cluster when compared with other localizations (e.g., occipital, parietal, or central). This finding has not been uniform across prior studies. Differences in inhibitory mechanisms by cortical region could contribute to these findings. 8 Oakley et al. found that the interseizure interval between extratemporal seizures was significantly shorter than between temporal seizures and did not find any significant effect of interseizure interval on localization among patients within the extratemporal seizure group. 5 A different study reported that specific EEG localization was not found to be significantly associated with seizure clustering. 27 Bergey et al. analyzed data from presurgical evaluation with intracranial electrodes and found that complex partial seizures lasted longer for mesial temporal lobe epilepsy (median 106 s) than for neocortical extratemporal lobe epilepsy (median 78 s), and that more than half of the patients in each group had relatively similar duration of all their seizures, suggesting intrinsic seizure dynamics for focal epilepsy. 14 Although the focal seizures were shorter than the seizures evolving bilaterally, the seizure type was not associated with seizure clustering in our study sample. We expected that a larger proportion of terminal and isolated seizures would evolve bilaterally ( generalized ), in comparison with the intracluster seizure group. Known mechanisms act on local neurons or networks to terminate seizures, and additional mechanisms limit the spread of seizures. 8 In our sample, the mechanisms that act distantly and involve both hemispheres were not more likely to terminate the clusters than the intrahemispheric/local mechanisms. Prior studies have addressed the seizure duration versus seizure type question for patients with epilepsy, but not in the context of specifics regarding seizure clustering. Theodore et al. 28 reported on the duration of complex partial seizures (mean 128 s) and the tonic clonic phase of the generalized seizures (mean 68 s; 83% of the time preceded by partial seizure phase). 29 Sperling et al. reported mean seizure duration for subclinical seizures as 47 s, for simple partial seizures as 50 s, for complex partial seizures as 123 s, and for generalized seizures as 126 s, 20 and in a different study, reported the duration of simple partial seizures as 28 s, complex partial seizures as 78 s, and for secondarily generalized seizures as 130 s. 16 When comparing consecutive seizure pairs, previously described cluster effect was present for seizures separated by 2- and 4-h interseizure intervals. Prior studies have described a cluster effect, in which seizures separated by interseizure intervals <8 h were more likely to arise from the same localization than seizures separated by longer durations, although this finding has not been uniformly replicated. 1,5,6,30 In the current study, concordant seizure localization was demonstrated for 2 and 4 h, but not for an 8-h interseizure interval. The increased probability of clustering at one focus in patients with multiple foci supports a role for local effects as a determinant of recurrent seizures. It is consistent with the hypothesis that clustering arises as a response to a proepileptogenic environment, induced by local and general factors that lower seizure threshold. There are limitations to our study. The inpatient scalp and intracranial EEG monitoring in the epilepsy unit represent the standard approach for confirming epilepsy diagnosis and localizing seizures. But this process is often different from the natural occurrence of seizures outside the hospital (e.g., at home), as antiepileptic drugs are frequently tapered in the inpatient setting (including in our study population) and the hospital setting introduces unique stressors. 4 In our study, depending on seizure frequency and severity, antiepileptic medications were tapered on an individualized basis and gradually. Further analysis will explore what role medication taper plays in occurrence and evolution of seizures and seizure clusters. In addition, there is selection bias patients admitted to the epilepsy monitoring unit typically have medically refractory epilepsy and therefore may not represent the general population of patients with epilepsy. These factors limit the generalizability of our findings, but not the validity of our observations. Inpatient monitoring, on the other hand, has major advantages in terms of seizure documentation, and it is required to accurately assess seizure duration and the other seizure parameters analyzed here. Based on these observations, we suggest that interseizure intervals of 4 h may be the appropriate definition for seizure clustering in the setting of epilepsy monitoring.

7 895 Termination of Seizure Clusters Recognizing clusters of seizures with short duration is not only important with regard to their diagnostic, localizing validity, but may also serve as an indicator for interventions to avoid morbidity associated with seizure clustering. Conclusion This study analyzes the characteristics of individual clustered seizures and the intrinsic features of the seizures that terminate a cluster. Our novel finding is that the terminal seizures of a cluster are longer in duration than the intracluster seizures and similar in duration to the isolated seizures. Further studies are necessary to address the relationship between specific seizure characteristics and inhibitory input in relation to the termination of a seizure cluster. Seizures of frontal and temporal lobe onset were more likely to cluster than those from other regions. The previously described clustering effect was present in this study but only for seizures separated by 4 h or less. This suggests a modification to the previously described definition of a seizure cluster in the epilepsy monitoring unit. These findings expand our understanding of clustering characteristics and mechanisms. Acknowledgments Neurophysiologic and clinical data analyzed in this paper were gathered as part of EU project EPILEPSIAE (Grant ). Disclosure of Conflicts of Interest A. Schulze-Bonhage has received honoraria for advice and presentations from Cerbomed, Desitin, Eisai, Precisis, and UCB. R. B. Lipton receives research support from the National Institutes of Health (NIH):PO1 AG (Program Director), RO1AG (Investigator), RO1AG A2 (Investigator), RO1AG (Investigator), RO1AG12101 (Investigator): and the National Headache Foundation. He serves on the editorial boards of Neurology and as senior advisor to Headache. In addition, he has reviewed for the National Institute on Aging and National Insitute of Neurological Disorders and Stroke, holds stock options in eneura Therapeutics; and serves as consultant, advisory board member, or has received honoraria from Alder, Allergan, American Headache Society, Autonomic Technologies, Avanir, Boehringer-Ingelheim, Boston Scientific, Bristol-Myers Squibb, Colucid, Dr. Reddy s, Electrocore, Eli Lilly, eneura Therapeutics, Informa, Merck, Novartis, Pfizer, Teva, and Vedanta. He receives royalties from Wolff s Headache, 8th Edition, Oxford Press University, S.R. Haut served as consultant for Acorda, Upsher-Smith, Neurelis, and Xeris. The remaining authors have no conflicts of interest. We confirm that we have read the Journal s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. Preliminary results were presented at the 68th Annual American Epilepsy Society meeting. References 1. Haut SR. Seizure clusters: characteristics and treatment. Curr Opin Neurol 2015;28: Osorio I, Frei MG, Sornette D, et al. Pharmaco-resistant seizures: self-triggering capacity, scale-free properties and predictability? Eur J Neurosci 2009;30: Martinez C, Sullivan T, Hauser WA. Prevalence of acute repetitive seizures (ARS) in the United Kingdom. Epilepsy Res 2009;87: Rose AB, McCabe PH, Gilliam FG, et al. Occurrence of seizure clusters and status epilepticus during inpatient video-eeg monitoring. Neurology 2003;60: Kim W, Miller JW, Drane DL, et al. Is there a clustering effect on electroencephalographic seizure localization? Epilepsy Res 2014;108: Haut SR, Legatt AD, O Dell C, et al. Seizure Lateralization During EEG Monitoring in Patients with Bilateral Foci: the Cluster Effect. Epilepsia 1997;38: Haut SR, Shinnar S, Moshe SL. Seizure clustering: risks and outcomes. Epilepsia 2005;46: Lado FA, Moshe SL. How do seizures stop? Epilepsia 2008;49: D umpelmann M, Jacobs J, Schulze-Bonhage A. Temporal and spatial characteristics of high frequency oscillations as a new biomarker in epilepsy. Epilepsia 2015;56: Klatt J, Feldwisch-Drentrup H, Ihle M, et al. The EPILEPSIAE database: an extensive electroencephalography database of epilepsy patients. Epilepsia 2012;53: Wagenaar JB, Worrell GA, Zachary I, et al. Collaborating and Sharing Data in Epilepsy Research. J Clin Neurophysiol 2015;32: Schulze-Bonhage A, Feldwisch-Drentrup H, Ihle M. The role of highquality EEG databases in the improvement and assessment of seizure prediction methods. Epilepsy Behav 2011;22(Suppl. 1):S88 S Fisher RS, Bartfeld E, Cramer JA. Use of an online epilepsy diary to characterize repetitive seizures. Epilepsy Behav 2015;47: Afra P, Jouny CC, Bergey GK. Duration of complex partial seizures: an intracranial EEG study. Epilepsia 2008;49: Holtkamp M, Sharan A, Sperling MR. Intracranial EEG in predicting surgical outcome in frontal lobe epilepsy. Epilepsia 2012;53: Jenssen S, Gracely EJ, Sperling MR. How long do most seizures last? A systematic comparison of seizures recorded in the epilepsy monitoring unit. Epilepsia 2006;47: Ng M, Pavlova M. Why are seizures rare in rapid eye movement sleep? Review of the frequency of seizures in different sleep stages. Epilepsy Res Treat 2013;2013: Sanchez Fernandez I, Klehm J, An S, et al. Comparison of risk factors for pediatric convulsive status epilepticus when defined as seizures >/= 5 min versus seizures >/= 30 min. Seizure 2014;23: Shinnar S, Berg AT, Moshe SL, et al. How long do new-onset seizures in children last? Ann Neurol 2001;49: Zangaladze A, Nei M, Liporace JD, et al. Characteristics and clinical significance of subclinical seizures. Epilepsia 2008;49: Giblin KA, Blumenfeld H. Is epilepsy a preventable disorder? New evidence from animal models Neuroscientist 2010;16: Fisher RS, Scharfman HE, decurtis M. How can we identify ictal and interictal abnormal activity? Adv Exp Med Biol 2014;813: Loscher W, Kohling R. Functional, metabolic, and synaptic changes after seizures as potential targets for antiepileptic therapy. Epilepsy Behav 2010;19: Nutt DJ, Cowen PJ, Green AR. Studies on the post-ictal rise in seizure threshold. Eur J Pharmacol 1981;71: Mares P, Kubova H. Developmental patterns of postictal refractoriness and potentiation akin to cortical stimulation. Epilepsia 2015;56:e10 e Williams PA, White AM, Clark S, et al. Development of spontaneous recurrent seizures after kainate-induced status epilepticus. J Neurosci 2009;29: Haut SR, Swick C, Freeman K, et al. Seizure Clustering during Epilepsy Monitoring. Epilepsia 2002;43: Theodore WH, Porter RJ, Penry JK. Complex partial seizures: clinical characteristics and differential diagnosis. Neurology 1983;33: Theodore WH, Porte RJ, Albert P, et al. The secondarily generalized tonic-clonic seizure: a videotape analysis. Neurology 1994;44: Choi EJ, Kang JK, Lee SA. Effect of interseizure interval on seizure lateralization in patients with bilateral seizure foci. Seizure 2006;15: