Hyunmi Choi (Department Neurology) and Gary Heiman (Department of Genetics, Rutgers University) are the authors responsible for statistical analyses.
Address correspondence to Hyunmi Choi, M.D., M.S., Columbia Comprehensive Epilepsy Center, The Neurological Institute, 710 West 168th Street, Box 210, New York, NY 10032, U.S.A. E-mail: email@example.com
Purpose: To investigate the cumulative probabilities of ≥12 month seizure remission and seizure relapse following remission, and to test the associations of clinical characteristics with these two study end points in a prevalence cohort of intractable adult epilepsy patients during medical management.
Methods: A retrospective cohort study of intractable epilepsy patients seen in 2001 at a single center was conducted. Kaplan–Meier analysis was used to estimate the cumulative probabilities of seizure remission and subsequent seizure relapse. Cox proportional hazards models were used to estimate the association (1) between clinical factors and ≥12 month seizure remission and (2) between clinical factors and seizure relapse following remission.
Results: One hundred eighty-seven subjects met the eligibility criteria for intractable epilepsy. The estimate of probability of remission was about 4% per year. Seizure remission was temporary for some individuals, as 5 out of 20 subjects with remission ultimately relapsed. No clinical factors predicted the likelihood of achieving ≥12 month seizure remission or subsequent seizure relapse.
Discussion: Some people with intractable epilepsy achieve ≥12 month seizure remission during medical treatment. Remission, however, is only temporary for some individuals. We were unable to identify clear predictors for remission.
Prognosis after temporal lobe epilepsy surgery is well documented in the literature. The probability of becoming free of disabling seizures after epilepsy surgery is estimated to be 58% to 67% in those who have anteromesial temporal lobe resections (Wiebe et al., 2001; Engel et al., 2003). Similarly, prognosis among newly diagnosed epilepsy patients is also well established. Approximately half of all newly diagnosed epilepsy patients become seizure free after a first antiepileptic medication trial (Kwan & Brodie, 2000). While there are several studies addressing prognosis in children with intractable epilepsy, only a few studies have evaluated the prognosis among adult patients who have chronic and intractable epilepsy (Selwa et al., 2003; Callaghan et al., 2007; Luciano & Shorvon, 2007). These studies show (1) that during every year of follow-up approximately 5% of intractable adult patients experience a minimum of 6-month seizure remission (Callaghan et al., 2007) or (2) that 11% of patients that have previously failed five or more medications can achieve a 12 month seizure remission with the next AED (Luciano & Shorvon, 2007). Although these two studies provide hopeful information for adults with intractable epilepsy, it remains unclear whether the seizure remissions in these patients are sustained or only temporary.
Our primary aim was to investigate the cumulative probabilities of (1) ≥12 month complete seizure remission, and (2) subsequent seizure relapse in those who experienced seizure remission during medical management in a prevalence cohort with long-standing and intractable epilepsy who may or may not be epilepsy surgery candidates. Our secondary aim was to investigate the associations of various clinical characteristics with (1) ≥12 month complete seizure remission, and (2) subsequent seizure relapse in those who experienced seizure remission.
Sample and procedure
Between 2005 and 2006, we performed a retrospective cohort study of epilepsy patients seen in 2001 at the Columbia Comprehensive Epilepsy Center and followed subsequently. Our study population included all adult patients seen in 2001 at our tertiary care center who met criteria for intractable epilepsy. We identified subjects by searching our electronic practice management system for anyone who had an appointment in 2001. We then screened their outpatient medical charts to determine eligibility. To be eligible for our cohort study, potential subjects had to meet all the following criteria of intractable epilepsy: recurrent seizures after ≥2 adequate trials of antiepileptic medications (AEDs), average seizure frequency of ≥1 seizure per month for 3 consecutive months prior to index date, and be 18 years or older. The index date was the date of first visit in 2001 in which they met inclusion criteria. We excluded those with a diagnosis of nonepileptic psychogenic seizures.
One attending epileptologist (HC) and two epilepsy fellows (DP, JC) reviewed the medical charts and collected information on a priori–defined predictor variables (Table 2).
Table 2. Association between clinical characteristics and achieving ≥12 months of seizure remission
(N = 187)
aAge of onset analyzed using categories: ages 0–9, 10–20, and >20.
bHR could not be calculated because of lack of convergence.
History of surgery (n = 12)
Status epilepticus (n = 16)
Age of onseta
Mental retardation (n = 36)
Febrile seizure (n = 25)
Etiology (n = 80)
MTS (n = 20)
Duration of epilepsy >10 year (n = 152)
Number of failed AED >5 (n = 90)
Study end points
Remission criteria: The primary outcome of the study was achieving a ≥12 month seizure remission. Remission was defined as being free of any seizures by self-report. Patients were followed for the occurrence of ≥12 month seizure remission in all subjects during medical treatment. When the beginning date of seizure freedom was not noted in the chart, the first clinic visit in which no seizure was reported was considered the start date of remission.
Relapse criteria: The secondary outcome of the study was the occurrence of seizure relapse after experiencing a ≥12 month seizure remission. Relapse was defined as the occurrence of a seizure after having at least 1 year of seizure freedom. When the exact date of seizure relapse was not noted in the chart, the first clinic visit in which seizure was reported was considered the date of relapse.
Variables tested for association with study end point included the following: history of prior epilepsy surgery; status epilepticus; age of onset; epilepsy classification; mental retardation; febrile seizure; etiology; mesial temporal lobe sclerosis; duration of epilepsy (>10 years); and number of failed AEDs (>5 AEDs). Age of onset was categorized as follows: <10 years of age, 10–20 years of age, >20 years of age. Epilepsy syndromes were categorized as idiopathic generalized epilepsy (IGE), symptomatic generalized epilepsy (SGE), localization-related epilepsy (LRE), and other. The “other” category consisted of those subjects that had more than one epilepsy syndrome or unclassifiable epilepsy syndrome. Mental retardation included those subjects that had Full Scale IQ <70 on neuropsychological testing or those that did not have neuropsychological testing but were diagnosed with mental retardation based on clinical assessment. The categories for etiology included trauma, encephalitis, stroke, vascular malformation, migration disorder, tumor, and other. Number of failed AEDs included all medications, which the subject failed before and during the study period.
Kaplan–Meier estimates analysis
Seizure remission: Kaplan–Meier analysis was used to estimate the cumulative probability of ≥12 month seizure remission during medical treatment: Time to event was defined as the duration between index date and 365 days after the start date seizure remission. Observations were censored (1) on the last clinic visit if subjects did not have primary study outcome by their last clinic visit or (2) on the date of surgery if they had epilepsy surgery. Of note, there were 33 individuals who went on to have their epilepsy surgery after their index date. These subjects were censored on the date of their epilepsy surgery.
Relapse: Kaplan–Meier analysis was used to estimate the cumulative probability of seizure relapse among those who had ≥12 month seizure remission. Time to event was defined as the duration between the beginning of ≥12 month seizure remission and the date of seizure relapse. Observations were censored on the last clinic visit if subjects did not have seizure relapse by their last clinic visit.
Cox proportional hazards analysis
We fitted Cox proportional hazards models to the data to estimate the association (1) between clinical factors and ≥12 month seizure remission, and (2) between clinical factors and seizure relapse following remission. These associations were tested in subjects only during medical treatment. Each predictor variable was analyzed independently. Hazard ratios (HR) and 95% confidence interval values are reported.
For evaluation of interrater reliability, reviewers JC and HC both reviewed 10 randomly chosen charts and reviewers DP and HC both reviewed 10 other randomly chosen charts. We evaluated agreement between reviewers JC and HC and reviewers DP and HC on key variables using the kappa statistic (κ), which measures agreement between two independent reviewers after excluding agreement expected by chance (Landis & Koch, 1977). Variables that were compared included the following: the presence of ≥12 month seizure remission, date of remission, date of relapse, epilepsy type, history of status epilepticus, and the presence of mesial temporal lobe sclerosis.
All analyses were conducted with Stata statistical software (College Station, TX, U.S.A.), except for interrater reliability evaluation, which was conducted with SPSS statistical software (Chicago, IL, U.S.A.).
Out of 1308 patients screened, 187 subjects (with a mean age of 41 years and mean duration of epilepsy of 25.6 years) met our eligibility criteria (Table 1). The 1121 individuals who did not meet eligibility criteria included several subgroups: 31 (2.7%) with <2AED trial, 163 (14.5%) with no follow-up visit after 2001; 27 (2.4%) with nonepileptic events; 55 (4.9%) with diagnosis other than epilepsy; 47 (4%) with unclear seizure frequency; 798 (71.1%) with seizure frequency not meeting our inclusion seizure frequency criteria. Figure 1 shows the number of subjects that met our study end points by the epilepsy syndrome classification.
Table 1. Demographics of cohort
Other epilepsy syndrome: 1 subject with both LRE and SGE, 1 subject with both LRE and IGE, and 1 subject with an unclassifiable epilepsy type.
Female gender—n (% of total)
41 yrs (±12.2)
Follow-up time in years—Mean (SD)
Duration of epilepsy—Mean (SD)
25.6 yrs (±14.7)
Epilepsy classification—n (% of total)
Idiopathic generalized (IGE)
Symptomatic generalized (SGE)
Number of failed AEDs—Mean (SD)
Complete seizure remission
Out of 187 subjects, 20 subjects achieved a ≥12 months of complete seizure remission. To ensure that these individuals were not previously in seizure remission, which was merely interrupted by 3 months of seizure exacerbation, leading to inclusion in our study, we examined their seizure frequency for 12 months preceding the start of their monthly seizure for 3 months (inclusion criteria) for a total of 15 months prior to index date. We found that none of the 20 individuals had had a sustained period (>3 months) of seizure freedom in the preceding 12 months prior to meeting our inclusion criteria. In fact, 14 out of 20 individuals who experienced remission during the study had had monthly seizures for the 15 months prior to index date. The remaining six individuals all had multiple clusters of consecutive monthly seizures in the preceding 15 months prior to index date, each clusters ranging between 3 to 9 months. The mean duration of follow-up of the 20 subjects achieving remission was 3.9 years (SD ± 1.1 years). The estimated cumulative probability of achieving a ≥12 month seizure remission was 0.6% at year 1, 4% at year 2, 8% at year 3, 13% at year 4, and 18% at year 5 of follow-up (Fig. 2).
Cox proportional hazard ratios
The association of history of epilepsy surgery, status epilepticus, age of onset, epilepsy syndrome classification, mental retardation, febrile seizure, etiology (present or not), MTS, duration of epilepsy (>10 years vs. ≤10 years), and number of failed AEDs (>5 vs. ≤5) with ≥12 month seizure remission were examined (Table 2). None of these variables reached a statistical significance, including the types of epilepsy. None of the individuals with IGE (n = 8) had a remission during follow-up, whereas 18 out of 158 (11%) subjects with LRE and 1 out of 20 (5%) subjects with SGE experienced remission. These proportion did not differ significantly from one another [IGE vs. LRE (Fisher exact p-value = 0.601) and IGE vs. SGE (Fisher exact p-value = 1.0)]. Although none of the clinical variables reached a statistical significance, the presence of prior status epilepticus, mental retardation, longer duration of epile- psy (>10 years vs. ≤10 years), having failed more AEDs (>5 failed AEDs vs. ≤5) were associated with point estimates consistent with a decreased likelihood of remission.
Seizure relapse after complete seizure remission
Among those that had ≥12 month seizure remission during medical treatment (n = 20), five individuals relapsed during the study period. The mean duration of follow-up of the five subjects who relapsed was 3.5 years (SD ± 1.4 years). The estimated cumulative probability of seizure relapse was 33% at year 2 and 44% at year 3 (from the start time of their remission) (Fig. 3).
Cox proportional hazard ratios
The associations of these same clinical factors with seizure relapse were also examined. None of these variables reached a statistical significance (Table 3).
Table 3. Association between clinical characteristics and experiencing subsequent seizure relapse (in the 20 subjects that achieved ≥12 months of seizure remission)
(N = 20)
a HR could not be calculated because of lack of convergence.
bAge of onset analyzed using categories: ages 0–9, 10–20, and >20.
History of surgery (n = 2)
Status epilepticus (n = 0)
Age of onsetb
Mental retardation (n = 4)
Febrile seizure (n = 3)
Etiology (n = 9)
MTS (n = 2)
Duration of epilepsy >10 year (n = 15)
Number of failed AED >5 (n = 9)
Agreement between reviewers JC and HC and between reviewers DP and HC was perfect for all seven variables that were examined (k = 1, p-value < 0.002).
Based on our definition of intractability, we collected a prevalence cohort of 187 adults with intractable epilepsy, identified from a pool of 1,308 patients seen at our tertiary-care epilepsy center in 2001. This group, as evidenced by their demographic characteristics shown in Table 1, represents the most intractable of cases that are generally assumed to have poor prognoses in terms of complete seizure remission. In our prevalence cohort, we found that approximately 4% of adults with long-standing and intractable epilepsy experience 12 months or more of complete seizure remission by second year of follow-up, with additional 4% of subjects per year experiencing this outcome during our study follow-up.
The literature regarding seizure remission among intractable epilepsy patients is heavily weighted toward the pediatric population (Huttenlocher & Hapke, 1990; Riikonen, 2001; Berg et al., 2006; Sillanpaa & Schmidt, 2006; Camfield & Camfield, 2007), which makes counseling the adult patients with intractable epilepsy difficult. In these pediatric studies, the seizure remission occurred in about 20% to 36% of patients or at a rate of about 4% per year. In contrast to pediatric epilepsy literature, very few studies have examined seizure remission among adults with intractable epilepsy. A recent, large retrospective study quantified the effect of adding an unused AED to intractable adult epilepsy patients (Luciano & Shorvon, 2007). Among 155 adult patients with intractable epilepsy, defined as seizure occurring at least once per month, they found that 43 (28%) of 155 adult patients with intractable epilepsy became seizure free for a year or more by a drug introduction during a mean follow-up of 18 months. Another recent study (Callaghan et al., 2007), in which 246 adult patients with intractable epilepsy were followed for 3 years, found that approximately 5% of patients per year experienced a 6-month terminal remission (i.e., seizure free at last follow-up).
The ability to define and predict which patient characteristics lead to seizure remission in those with intractable epilepsy would allow physicians to inform such patients about what to expect in the future. One study found that individuals treated previously with less than five drugs (as opposed to more than five drugs), those with idiopathic epilepsy (as opposed to cryptogenic or symptomatic epilepsy), and those with a duration of epilepsy of less than 10 years were more likely to be rendered seizure free by the new drug (Luciano & Shorvon, 2007). Another study found that prior history of status epilepticus, longer duration of intractability (i.e., ≥10 years), increased number of failed drug therapies (i.e., ≥6 AEDs), and presence of mental retardation were associated with diminished likelihood of experiencing a 6-month terminal remission (Callaghan et al., 2007).
Although we found similar direction of risk for some variables as found in the two recent studies, none of the clinical variables reached a statistical significance in our study. One reason for the lack of significant associations in our study could be due to inadequate power. Another reason might be that no significant relationships exist between various clinical factors and seizure remission.
It is important to note whether the seizure remission period was immediately preceded by an AED trial. We found that 17 out of 20 individuals with ≥12 months of seizure remission had undergone a medication change immediately preceding the start of their remission (within the preceding 3 months). Most of the medication changes were additions of a new antiepileptic medication, although 2 of 17 individuals had increases in the dose of their medication. Three individuals out of 20 (15%) experienced remission for no clear reason. These three individuals probably represent a small number of pharmacoresistant epilepsy patients that inexplicably experience certain periods of seizure remission, perhaps reflecting fluctuation in their disease course.
Our study provides additional information to those provided by the two recent studies. In contrast to the study by Luciano and Shorvon, we accounted for variable follow-up time of study subjects and examined approximate yearly probability of seizure remission by employing the survival analysis technique. In contrast to the two recent studies (Callaghan et al., 2007; Luciano & Shorvon, 2007), we examined the probability of seizure relapse occurring in those with remission. Five out of 20 (25%) subjects experiencing remission ultimately relapsed during follow-up, although the estimated cumulative probability of relapse, factoring in duration of follow-up, was 44% at last follow-up. This finding provides a more complete picture of outcomes in these patients.
Our study needs to be interpreted in light of few limitations. As our aim was to examine the probability of seizure remission in adult patients with chronic and intractable epilepsy, our study cohort was inclusive of various epilepsy syndromes and etiologies. This might limit counseling individual patients with specific epilepsy syndromes. Second, we utilized a retrospective cohort study design, because of speed and efficiency associated with this design compared to a prospective design. One of the inherent weaknesses of retrospective study design is that source document from which data are abstracted does not include all of the information that may be important for research questions. Third, we relied on self-reported seizure frequency by patients, documented in the medical charts, in order to measure our study end points. As previously demonstrated, a proportion of patients may be unaware of their seizures (Gotman, 1990; Blum et al., 1996; Tatum et al., 2001; Heo et al., 2006). Therefore, possible underreporting of seizures, arising from lack of awareness of one's own seizures, may have resulted in overestimation of our seizure remission outcome.
Although an ideal study design to examine long-term prognosis is to assemble an incidence cohort of patients, entering the study at a uniform point in time in their condition (i.e., as patients become intractable), our study findings provide evidence that seizure remission can occur in a small proportion of adult patients with chronic and intractable epilepsies. However, remission is only temporary for many. Future prospective cohort studies of incidence cases of intractable epilepsy are needed, which should prospectively identify individuals when they become intractable and follow them over time to examine their prognosis.
This research was supported by National Institute of Health grant K12 RR017648 (to HC). The funding organization had no role in the design and conduct of the study, collection, management, analysis or interpretation of data, or in preparation of this manuscript. The funding organization has not reviewed this manuscript.
Conflict 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. The authors report no conflicts of interest.