Address correspondence and reprint requests to Dr. S. R. Haut at Department of Neurology, Montefiore Medical Center, 111 East 210th St., Bronx, NY 10467-2490, U.S.A. E-mail: firstname.lastname@example.org
Summary: Purpose: To identify risk factors associated with seizure clustering during epilepsy monitoring and to assess the effect of clustering on localization of the epileptogenic zone.
Methods: Patients undergoing presurgical epilepsy monitoring at Montefiore Medical Center or Yale–New Haven Hospital were recruited. Seizure clustering was defined as three or more seizures within 24 h. Risk factors for seizure clustering were examined by using regression analysis. The effect of clustering on localization was examined by Student's t test evaluation of mean interseizure interval for consecutive concordant and discordant seizures.
Results: Of 91 patients, clustering was present in 56 (61.5%). Variables significantly associated with clustering included a history of seizure clustering at home (p = 0.0003) and the presence of mesial temporal sclerosis (MTS) on magnetic resonance imaging (MRI; p = 0.0172). Clustering was present in nine of 10 patients with more than one seizure-onset zone. Ictal EEG localization was not associated with clustering, nor was rapid medication withdrawal. Mean interseizure intervals were not significantly different for concordant and discordant seizures.
Conclusions: Seizure clustering during epilepsy monitoring is common. Risk factors include history of clustering at home, MTS on MRI, and possibly more than one seizure focus. In this study, seizures that occurred in clusters had as important a role in localization as did nonclustered seizures.
It has long been noted that seizures occasionally occur in “clusters,” or “flurries”(1–4). Seizure clustering occurs frequently during inpatient CCTV-EEG monitoring for presurgical evaluation, which may be related to an increased risk of clustering in the population of patients with medically intractable epilepsy (5,6), or to medication reductions commonly performed to precipitate seizures.
Seizure clustering occurring during monitoring potentially poses both neurophysiologic and clinical complications. Seizures occurring in a cluster either may represent truly independent seizures or may reflect increased irritability or decreased inhibition at a particular region (7,8). Thus clustering of seizures may interfere with identification of the predominant epileptogenic zone. Additionally, seizure clustering may pose an increased risk to patient safety during monitoring. The identification of patients who are at high risk for clustered seizures may serve to decrease these potential confounders or complications of presurgical monitoring.
The purposes of this study were to identify risk factors associated with seizure clustering in patients undergoing inpatient epilepsy monitoring for presurgical evaluation and to examine the effects of clustering on localization of the epileptogenic zone.
MATERIALS AND METHODS
All patients with medically intractable epilepsy who underwent inpatient CCTV-EEG monitoring for presurgical evaluation at the Montefiore Medical Center (MMC) or Yale–New Haven Hospital between January 1998 and March 1999 were recruited prospectively. Patients whose seizures did not have both clinical and electrographic manifestations were excluded, as were patients who did not have seizures recorded during the monitoring admission. If patients were monitored more than once, the more remote admissions were excluded, and if both extracranial and intracranial monitoring were performed, extracranial recordings were excluded in preference to intracranial data.
On admission, clinical history obtained included current age, age at seizure onset, typical seizure frequency, and maximal seizure frequency at home. Three or more seizures within a 24-h period at home was considered to represent seizure clustering by history.
Magnetic resonance imaging (MRI) results were obtained, when available. All MRIs were interpreted by the neuroradiology departments of either MMC or Yale–New Haven Hospital.
Admission doses and levels of all anticonvulsants (AEDs), as well as changes in doses with dates, were recorded. Rapid medication adjustment was defined as a >50% reduction in dosage of an AED within 24 h.
Date, time, and seizure-onset localization of all seizures occurring during the monitoring period were recorded. Seizure onset was defined as a sustained rhythmic change in the EEG, at a frequency of >2 Hz, accompanied by clinically typical seizure activity and clearly distinguished from background EEG, as previously described (9). All seizure onsets were reviewed for localization according to the definition provided. Final single localization required all localized seizures to show the same localization and to allow determination of a single lobe and region of that lobe to be defined as the region from where the seizures arose. Final bilateral or multifocal unilateral localization required all localized seizures to show one of two separate localizations, each of which was specific to a single lobe and region of that lobe. When a minority of recorded seizures did not show adequate localization, the localization from the others was considered acceptable. When a majority of recorded seizures did not show adequate localization, localization was considered indeterminate. Interictal activity was not considered in this localization determination.
Seizure clustering: We previously defined seizure clustering as three or more seizures in 24 h (8). The analysis for this study used this definition and also examined the continuous variable of “maximal seizure frequency in any 24-h period” as a further indicator of seizure clustering.
Concordant and discordant seizures: Localizations for each set of two consecutive seizures were compared. Those with the same ictal localization were termed “concordant,” and those with different ictal localizations were termed “discordant.”
Seizure clustering as 3 or more seizures within 24 h
The association between seizure clustering and each dichotomous variable was tested for significance by using Fisher's exact test. The association between seizure clustering and continuous variables was tested for significance by using either Student's t test, if assumptions were met, or Wilcoxon rank sum test.
Variables for which the associations with seizure clustering yielded p values <0.20 were entered into the initial multiple logistic regression model. A backward stepwise elimination procedure was used, with attention to the proportion of missing values for each variable. Variables with significance levels <0.05 were retained in the final model (10).
Maximal seizures in 24 hours as a continuous variable
The association between maximal seizures in 24 h and each dichotomous variable was tested for significance by using either a Student's t test for independent samples, if assumptions were met, or Wilcoxon rank sum test. The association between maximal seizures in 24 h and continuous variables was tested for significance by using Pearson or Spearman correlation coefficients, depending on whether assumptions were met.
Variables for which the associations with maximal seizures in 24 h yielded p values <0.20 were entered into the initial multiple linear regression model. A backward stepwise elimination procedure was used, with attention to the proportion of missing values for each variable. Variables with significance levels <0.05 were retained in the final model. A transformation of scale was performed on the dependent variable the better to approximate assumptions (10).
Impact of interseizure interval on localization
Patients with more than one seizure onset were identified, and seizures were classified as concordant or discordant, as described earlier. Interseizure intervals were calculated, and the mean interseizure interval for concordant seizures was compared with the mean interseizure interval for discordant seizures by use of a Student's t test.
Ninety three patients were entered into the study. Five of the 93 patients had both extracranial and intracranial monitoring during the study period, and the intracranial monitorings were preferentially included. Two patients were excluded, as they had no seizures during the monitoring period, leaving 91 patients. Of these 91, 66 (72.5%) underwent extracranial monitoring, and 25 (27.5%) underwent intracranial monitoring. Gender distribution was 52% female, 48% male patients; mean age, 33 ± 12.2 years. Distribution of duration of monitoring is presented in Fig. 1.
Temporal distribution of seizures
The distribution of maximal seizures within 24 h is presented in Fig. 2. Thirty-five (38.5%) patients had a maximum of one to two seizures in a 24-h period, whereas 56 (61.5%) patients met the definition of seizure clustering by experiencing three or more seizures in a 24-h period during the monitoring. Twenty-six (28.5%) patients had five or more seizures within a 24-h period.
For 69 patients in whom localization was determined, 48 had unilateral temporal onset, eight had bitemporal independent onset, 11 had extratemporal unifocal onset, and two had extratemporal multifocal onset. Seizure localization was hemispheric or indeterminate in 22 (24.2%) of the patients.
Risk factors associated with seizure clustering
Variables examined by bivariate analysis included age at first seizure, average seizure frequency per month, history of seizure clustering at home, history of status epilepticus, seizure localization, more than one seizure-onset zone, presence of mesial temporal sclerosis (MTS) on MRI, rapid medication discontinuation, and discontinuation of each specific medication. Variables included in the initial multiple regression model were age at first seizure, average seizure frequency per month, history of seizure clustering at home, and presence of MTS on MRI.
Variables that remained significant in the model included a history of seizure clustering at home (p = 0.0003) and the presence of MTS on MRI (p = 0.0172).
Risk factors for greater maximal seizures within 24 h
Variables were examined by bivariate analysis as described earlier. Variables included in the initial multiple regression model were age at first seizure, history of seizure clustering at home, presence of MTS on MRI, more than one seizure localization, and rapid medication withdrawal before the 24-h period with maximal seizures.
Variables that remained significant in the model included a history of seizure clustering at home (p = 0.0001) and the presence of MTS on MRI (p = 0.04).
Seizure clustering and localization
Of patients with temporal onset, 64% had three or more seizures within 24 h, with a subset of 20.4% who had five or more seizures within 24 h. Similarly, 62.5% of the patients with frontal lobe onset had three or more seizures within 24 h, with a subset of 28.6% who had five or more seizures within 24 h, and this difference was not statistically significant. Thus specific EEG localization was not found to be significantly associated with seizure clustering.
Single versus multiple localizations
Of the 10 patients who had more than one localization, 90% had maximal seizure frequency of three or more, with a subset of 33% whose maximal seizure frequency was five or more. In contrast, for patients with a single localization, 56.4% had a maximal seizure frequency of three or more, with a subset of 24.2% whose maximal seizure frequency was five or more. Mean average for maximal seizures in 24 h was higher for patients with multiple localizations, which approached but did not reach statistical significance (p = 0.09).
Localization by MRI
Of patients with MTS, 84% had experienced three or more seizures within 24 h, with a subset of 25.7% who had experienced five or more seizures within 24 h. MTS on MRI was significantly associated with seizure clustering (p = 0.0172) and greater maximal seizures in 24 h (p = 0.04) in final regression models.
Seizure clustering and medications
Most patients were receiving more than one AED on admission. The distribution of medications is presented in Table 1. Overall, 53.8% of patients underwent rapid withdrawal of at least one AED. Seizure clustering was not significantly associated with rapid medication withdrawal or with withdrawal or use of any particular medication.
Table 1. Distribution of most common medications
Patients were taking multiple medications, therefore totals >100%.
The impact of interseizure interval on seizure localization
Ten (11%) of 91 patients had more than one seizure-onset localization identified during monitoring, seven during intracranial monitoring and three during extracranial monitoring. These 10 patients had a total of 82 seizures. All consecutive seizures were included in the analysis, regardless of interseizure interval. Sixty-five percent of consecutive seizures were concordant for ictal onset, and 35% were discordant for ictal onset. Mean interseizure interval was 1,045 min for concordant seizures and 822 min for discordant seizures, which was not significantly different (p = 0.58). In subgroup analysis of the seven patients who underwent intracranial monitoring, mean interseizure interval was 755 min for concordant seizures and 852 min for discordant seizures, which again was not statistically significant (p = 0.84).
Distribution of seizure clustering and localization was examined to assess the effect of clustering on identification of the predominant focus. This analysis was limited, as the degree of seizure clustering was not uniform across the patients and because an equal number of seizure clusters contained a discordant seizure as did not. Overall examination of the distribution of concordant and discordant seizures led to the conclusion that identification of the predominant focus was not significantly affected by seizure clustering in these patients.
Risk factors for seizure clustering during epilepsy monitoring have not previously been identified. Seizure clustering has been anecdotally associated with extratemporal epilepsy, although this association has not been clearly demonstrated. Manford (11) analyzed frontal and temporal epilepsy seizure patterns and found that seizure frequency and timing were not consistently different between groups with different localizations. In Bauer's seizure clustering group (12), the percentage of women was significantly higher than men, and there was an association of clustering with longer duration of epilepsy.
In this study, we demonstrated that age, gender, and age at epilepsy onset were not associated with seizure clustering. Not surprisingly, patients with a history of seizure clustering at home had a significant risk of seizure clustering during monitoring. Although no association was demonstrated between specific EEG localization and seizure clustering, the presence of MTS on MRI was associated with clustering. An association between seizure clustering and more than one seizure localization approached but did not reach statistical significance. Interestingly, no association was demonstrated between frontal lobe onset and clustering. The relatively small number of patients with frontal lobe onset in this study may have limited our ability to demonstrate such an association. Overall, rapid medication taper, which is commonly performed during presurgical evaluation, or withdrawal or use of any specific AED, was not associated with seizure clustering. Our study, however, did not address the specific temporal relation between dates of withdrawal and dates of seizures, or the specific combinations of AEDs and seizure clustering. Although a history of status epilepticus has been previously reported to be associated with seizure clustering (6), such an association was not demonstrated in this study.
The temporal distribution of seizures during inpatient presurgical evaluation and the resulting impact on seizure localization has been only minimally addressed. Todorov et al. (7) discussed seizure clustering during monitoring and hypothesized that clustered seizures might tend to arise from the same focus. Blum (13) noted that seizures tended to cluster in groups, with seizures from a secondary focus tending to occur consecutively rather than randomly, but this was without statistical significance. We previously conducted a retrospective analysis of seizures in patients with bitemporal independent onsets and reported a “cluster effect,” in that seizures occurring after an interseizure interval of <8 h had a greater likelihood of arising from the same side as the previous seizure than did those with longer interseizure intervals (8).
In the current prospective study, we did not find evidence of this cluster effect. Of the seizure pairs in patients with more than one seizure focus, there was not a significant difference in mean interseizure interval between concordant and discordant seizure pairs. This aspect of the study may have been limited because of a small number of patients with more than one seizure focus. Identification of the predominant focus does not appear to have been affected by seizure clustering in these patients.
We conclude that seizure clustering during inpatient epilepsy monitoring is common, and patients at a higher risk include those with a history of clustering at home, MTS on MRI, and possibly those with more than one ictal focus. No association between clustering and frontal lobe epilepsy or rapid medication withdrawal was demonstrated. Whereas seizures that occur widely spaced in time have been considered to be optimal for ictal localization, in this study, seizures that occurred in clusters had as important a role in localization as did nonclustered seizures.
Acknowledgment: We gratefully acknowledge the support of the Epilepsy Foundation of America and the assistance of the members of the epilepsy departments at Montefiore Medical Center and Yale–New Haven Hospital. Funding supported by Epilepsy Foundation of America Junior Investigator Award 1998–1999 to Sheryl Haut, M.D.
Presented, in part, at the Annual Meeting of the American Epilepsy Society, Orlando, Florida, December 5–11, 1999.