Is the underlying cause of epilepsy a major prognostic factor for recurrence?

Epilepsy, defined as the recurrence of nonprovoked epileptic seizures, is one of the most common neurologic disorders. The prevalence of active epilepsy is approximately 6 per 1000 and 3% to 5% of the general population experience one or more seizures at some time during their lives.^1,2 In many patients, seizures are well controlled with antiepileptic drugs (AEDs), but approximately 25% to 30% continue to experience recurrent seizures despite optimal therapy.³ These patients have medically refractory epilepsy. Several prognostic factors have been identified, such as seizure type, neurologic and psychiatric deficits, electroencephalographic abnormalities, and postnatally acquired brain lesions.⁴

Studies have shown that one of the most important risk factors for refractory epilepsy is related to the type of epilepsy. Patients with complex partial seizures are reported to have a higher seizure recurrence rate than patients with generalized tonic-clonic seizures.^5-7 One study confirmed that the rate of seizure-free patients is significantly higher in cases of idiopathic generalized epilepsy than in cases of cryptogenic or symptomatic partial epilepsy.⁸ Thus, there is evidence that partial epilepsies carry a poor prognosis. Nevertheless, in patients with partial epilepsy, the role of brain abnormalities in influencing the outcome remains unknown. In these patients, most studies have found a higher risk of seizure recurrence associated with a clearly identifiable cause such as a tumor.⁹ Some authors have suggested that when cerebral lesions are associated with epilepsy, their nature may have a strong prognostic influence, but to our knowledge this point has not previously been investigated.^10,11

Using a cross-sectional study on a large sample of adult epileptic patients, we assessed 1) the number of patients with uncontrolled seizures according to the type of epilepsy based on the International League Against Epilepsy (ILAE) classification,¹² and 2) the prognostic role of brain abnormalities and of the localization of the epileptogenic zone in patients with partial epilepsy.

Methods. This observational hospital-based study in adult epileptic outpatients (16 years of age or older) was performed from November 1990 through November 1997 at the Epilepsy Unit of La Salpêtrière Hospital in Paris. All the patients referred to our center (a tertiary referral center) during this period were included in the study and followed up. Our center tends to receive referrals of intractable patients, and only a small proportion (8%) of patients were seen at the time of their first seizure.

Patients. Each neurologist registered the information for every patient on a case-record form at the time of the first visit. Data updates were recorded on the same form at each subsequent visit. Collected data included age; gender; age at onset of epilepsy; family history of epilepsy; patient history; seizure type, according to the ILAE Classification of Epileptic Seizures¹³; EEG and imaging data; epilepsy syndrome, according to the revised ILAE Classification of Epilepsies and Epileptic Syndromes¹²; AED treatment; and seizure control.

Classification of patients according to epilepsy type. Partial and generalized epilepsies were defined by electroclinical criteria of the ILAE Classification of Epilepsies and Epileptic Syndromes. A diagnosis of symptomatic epilepsy was made when an etiologic factor (e.g., definite brain trauma or anoxia) was clearly recognizable or when radiologic or clinical signs of brain damage consistent with the type of epilepsy were detected. When the cause was hidden, cryptogenic epilepsy (i.e., epilepsy presumed to be symptomatic but whose etiology remains unknown) was diagnosed. When MRI demonstrated hippocampal sclerosis (HS) in a patient with electroclinical evidence of temporal lobe epilepsy (TLE), the patient was classified as having cryptogenic partial epilepsy associated with isolated HS. In patients with partial epilepsy, the location of the epileptogenic zone was defined based on clinical ictal semiology, interictal EEG, and in some cases ictal EEG, and MRI data. Only a few underwent continuous EEG-video recording and ¹⁸fluorodeoxyglucose (FDG)-PET for presurgical evaluation. Partial epilepsies were classified according the ILAE into temporal, frontal, parietal, occipital lobe, and multilobar epilepsies.¹² More precise localization was available only for patients who underwent surgery and has not been analyzed in this study. When the data were incongruent or insufficient, patients were classified as having partial epilepsy of undetermined localization.

Patients without unequivocal features of focal or generalized seizures (e.g., patients with only nocturnal generalized tonic-clonic seizures) were classified as having epilepsy undetermined, whether focal or generalized. Patients with isolated seizures, acute symptomatic seizures or situation-related seizures according to the ILAE guidelines for epidemiologic studies,¹⁴ or insufficient data were excluded from the analysis.

Classification of patients with partial epilepsy according to brain abnormalities. In all patients who had partial epilepsy, imaging data were obtained with a standardized MRI protocol, as previously described,^15-17 except when MRI scan was contraindicated (in which case CT scanning was performed) or when patients were lost to follow-up. Brain abnormalities were classified as cerebral cortical dysgenesis (abnormalities of gyration, heterotopia, tuberous sclerosis, focal cortical dysplasia, dysembryoplastic neuroepithelial tumor); HS; scars (postcerebrovascular accident, post-traumatic brain injury, miscellaneous); tumors (glial tumors, meningioma); vascular malformations (cavernous angioma, arteriovenous malformations); and miscellaneous abnormalities (other neurologic or systemic diseases). The association of HS with another lesion was described as dual pathology.¹⁸

Assessment of seizure control. Seizure control was assessed based on the year preceding the patient's last visit and was subdivided into four groups: easy-to-control seizure (seizure-free patients using monotherapy at usual daily dosage); difficult-to-control seizure (seizure-free patients using polytherapy with high daily dosage and often multiple AED trials); not seizure free; and seizure control not assessable (e.g., insufficient follow-up, poor compliance).⁸ Only unprovoked seizures were considered when evaluating seizure control.

Statistical analysis. Data were stored in an electronic database and analysis was based on all the information available at the time of the patient's last visit. They were first assessed by examining the univariate relation between each variable (characteristics of patients and epilepsy) and seizure control. For categoric variables, the comparison of percentages according to treatment response (seizure control) was made using chi-square analysis or Fisher's exact test if chi-square was not a valid test. For continuous variables (duration of the disease and age at onset), the distributions were tested with Shapiro-Wilk statistic, and all were not normal. Comparison of means between groups was therefore performed using Mann-Whitney rank sum test and Kruskall-Wallis test to compare more than two groups. To determine the relative important of the predictor variables on the treatment response, multivariate analysis using logistic regression was performed. A stepwise selection of the variables was used. Age at onset and duration of epilepsy in years were entered in the model as continuous variables. The different categories of nominal variables were entered in the model as dichotomous (dummy) variables. The categories with the most favorable outcome were taken as reference level; all the other categories were included in the model. Odds ratios and their confidence intervals (CIs) were calculated. The α-to-enter and α-to-exit were set, respectively, at 0.15 and 0.20. To assess the predictive ability of the model, concordance rate between predicted and observed responses was calculated. The goodness-of-fit of the logistic regression model was assessed using likelihood ratio test and Hosmer and Lemeshow test.¹⁹

The statistical software SAS²⁰ (version 6.11/UNIX; proc freq, proc univariate, proc npar1way, proc logistic) was used for statistical analysis.

Results. Among the 2,500 patients included in the database, 2,200 patients (49.8% male and 50.2% female) of at least 16 years of age with definite epilepsy were registered in the study. Median age was 33 years (interquantile, 27 to 43) and median age at epilepsy onset was 15 years (interquantile, 8 to 23).

Type of epilepsy. Partial epilepsy was the type most frequently observed (62% of all the patients; table 1). Generalized epilepsy was found in 22% of the patients, and 16% of the patients had epilepsy undetermined, whether focal or generalized. Of the patients with generalized epilepsy, 92% had idiopathic epilepsy. Of the 1,369 patients with partial epilepsy, 572 had cryptogenic epilepsy (pure cryptogenic, 334 patients; partial epilepsy associated with isolated HS, 238 patients); 576 had symptomatic epilepsy; and 221 could not be classified because of suboptimal MRI data. Idiopathic partial epilepsy (benign childhood epilepsy with centrotemporal spike, childhood epilepsy with occipital paroxysms, and primary reading epilepsy) was a rare condition in this study conducted in adults. The imaging data in patients with partial epilepsy are shown in table 2. In patients with partial epilepsy, MRI examinations were noncontributive in 29% (normal or nonspecific abnormalities); HS was the most frequent abnormality (isolated HS, 21% of patients; HS associated with another lesion, 4% of the patients). In 806 patients with partial epilepsy (i.e., 59% of patients with partial epilepsy), there was a sufficiently characteristic clinical onset and EEG focus to attempt seizure localization (table 3).

Seizure control. Complete seizure control was not achieved in most patients. Of the 2,200 patients, 1,696 were assessable for seizure control and received an AED treatment (101 did not receive an AED and 403 had unassessable seizure control). Of these 1,696 patients, 55% had persistent seizures despite AED treatment and 45% were seizure free (follow-up, 1 to 7 years). Seizure-free patients were separated into two groups: patients with easy-to-control seizure (25%) and patients with difficult-to-control seizure (20%).

Prognostic factors. Significant differences in seizure control were found in patients according to the different epileptic syndromes (p < 0.001). Patients with generalized idiopathic epilepsy usually had easy-to-treat epilepsy (82% of patients seizure free), whereas patients with partial or generalized epilepsy, whether symptomatic or cryptogenic, usually experience drug-resistant epilepsy (11% to 45% of patients seizure free; figure 1).

The longer the duration of epilepsy, the more difficult the seizure control (p < 0.001). When considering all patients together, median duration of epilepsy was 11 years in patients with easy-to-control seizure, 19 years in those with difficult-to-control seizure, and 21.5 years in patients with persistent seizures. Similar statistically significant (p < 0.001) results were observed in patients with partial epilepsy, in whom mean duration was 9 years, 19 years, and 21 years, respectively (table 4) and also in patients with generalized epilepsy (mean duration: 13 years, 20 years, and 21 years, respectively).

The level of seizure control differed significantly according to age at onset. When considering all patients, median age at onset was 17 years in patients with easy-to-control seizure, 14 years in those with difficult-to-control seizure, and 12 years in those with persistent seizures (p < 0.001).

In those patients, no differences were found between patients with difficult-to-control seizures and patients with persistent seizures. In patients with partial epilepsy, median age at onset in these three groups was also significantly different (20, 14.5, and 13 years, respectively; p < 0.001). Similarly, median age at onset was significantly different in patients with generalized epilepsy (16, 13, and 10 years, respectively; p < 0.001). In these patients, no differences were found between patients with difficult-to-control seizures and patients with persistent seizures.

The location of the epileptogenic zone was not a major prognostic factor and was associated with only slight differences between the lobar subgroups. No significant differences were found between extra-TLE subgroups (33%, 35%, and 37% of patients seizure free in parietal, occipital, and frontal lobe epilepsies, respectively; p = 0.954). The only significant difference in seizure control was found between the TLE and extra-TLE subgroups (20% and 36% of patients seizure free, respectively; p < 0.001; figure 2). No significant difference was found between patients with TLE without HS (n = 235) and those with extra-TLE (n = 214, p = 0.18).

In contrast, brain abnormalities associated with partial epilepsy were found to have a great influence on the rate of intractability (figure 3). First, intractable epilepsy was more frequent in patients with MRI abnormalities than in patients without MRI-detected lesions (25% and 42% of patients seizure free, respectively; p < 0.001). In patients with MRI-detected lesions, a poorer outcome was observed in cases of dual pathology (i.e., HS associated with another lesion; all but one of the patients with dual pathology had refractory partial epilepsy). The presence of HS was a major prognostic factor because only 10% of the 206 TLE patients with HS were seizure free, whereas 31% of the 235 TLE patients without HS were seizure free (p < 0.001; figure 4). Most patients with cortical dysgenesis also had refractory partial epilepsy (24% of patients seizure free). The most favorable outcome was seen in stroke sequelae, vascular malformation, and tumor (54%, 50%, and 46% of patients seizure free, respectively; p = 0.809, NS).

A logistic regression was made in patients with symptomatic and cryptogenic epilepsy. The outcome variable was easy-to-control seizure or not easily controllable. The dummy variables of brain abnormalities were cerebral dysgenesis, HS, dual pathology, and miscellaneous scars. The epilepsy syndrome was a dichotomous variable: symptomatic or not, cryptogenic epilepsy was the reference group. The localizations of epileptic focus associated with the worst outcome (temporal and multilobar) were transformed to dichotomous variables. The logistic regression results (table 5) confirmed that the most important prognostic factor was the presence of an MRI-detected brain abnormality-particularly HS, dual pathology, and cerebral dysgenesis. Duration of epilepsy and age at onset also had an important prognosis value: for each additional 10 years duration of epilepsy, the probability of refractory epilepsy was multiplied by 1.3. The estimated odds ratio for each decrease of 10 years in age at onset was 1.2. The concordance between predicted and observed responses was good (73.6%). The Hosmer and Lemeshow goodness-of-fit test was not significant.

Discussion. Overall, 55% of the patients studied had persistent seizures despite AED treatment. This compares with 10% to 40% commonly reported.^6,10,21 There are two likely explanations. First, our study was hospital-based in a neurologic university center and included a higher percentage of refractory patients. Second, the definition of seizure-free patients used in our study was more stringent than in other studies. Intractable epilepsy has been defined as more than one seizure over a 2-year period.²² Others have postulated that freedom from seizure should be the ultimate goal.^23,24 We used the criterion of 1 year without seizure. The results that we obtained were close to those of Avanzini et al.,⁸ who used a similar definition and found that 46% of the patients observed by a network of epilepsy centers had persistent seizure.

As previously suggested,¹¹ the type of epilepsy and epileptic syndrome plays an important role in determining prognosis. In general agreement with previous studies,^25,26 82% of the patients in our study who had idiopathic generalized epilepsy were seizure free, suggesting a favorable prognosis for these syndromes. In partial epilepsies, many factors have been reported as predictive of intractable epilepsy, such as acquired brain damage, neurologic deficit, mental retardation, early age at onset, seizure type, combined seizure types, high seizure frequency, long duration of epilepsy, psychotic disturbance, and EEG abnormalities.^5,6,21,27,28 Among prognostic factors for drug resistance, however, the nature of the underlying cause of epilepsy has seldom been examined because of a lack of information before the advent of MRI. Several studies have reported that symptomatic epilepsies are associated with a higher risk of drug resistance^8,22 or a higher risk of seizure recurrence after a first seizure.²⁷ Nevertheless, in these studies the underlying cause of epilepsy was not precisely determined. In our study, 71% of the imaging examinations detected brain abnormalities in patients with partial epilepsy. This high percentage of structural brain abnormalities compared with percentages reported in previous studies^8,29-31 is probably related to the systematic MRI examination performed in all patients with partial seizures using a specific protocol¹⁷ and to the severity of the epilepsies seen in our center. It is well established that identification of subtle brain abnormalities (e.g., HS or some types of cortical dysgenesis) requires MRI examinations with thin slices, special sequences, and specific planes of image acquisition.^32-34 Using this technique, we found a high rate of patients with partial epilepsy and MRI-detected HS (21% of the patients with partial epilepsy demonstrated isolated HS and 4% HS associated with another lesion). This high rate of HS is common in refractory TLE patients before surgical treatment^32,35 but few studies have been performed in TLE patients who have varying severity.¹⁷

In our study, the prognosis of partial epilepsy was found to be more closely related to the type of lesion than to the lobar localization of the epileptogenic zone. Nevertheless, as previously mentioned by Manford et al.,²⁹ localization of the zone is not easy in most cases of partial seizures in outpatients, and we proposed a lobar localization in only 59% of the patients. Only a few patients in our study underwent continuous video-EEG monitoring. Thus, localization of the epileptic focus was limited to a lobar classification, which did not allow definite conclusions regarding the organization of the epileptogenic zone and its relation to intractability. TLE was the most frequent partial epilepsy (66% of the patients with localized partial epilepsy) and the most drug-resistant form (20% of patients seizure free). In the other types of partial epilepsies (i.e., frontal lobe, occipital lobe, and parietal lobe), the proportion of seizure-free patients was similar (33% to 37% of patients seizure free). Thirty-one percent of patients with TLE and without MRI-detected HS were seizure free and thus carried a similar rate of drug resistance to extra-TLE patients. Only 10% of the patients with TLE and HS were seizure free, demonstrating the importance of HS as a major prognostic factor. When considering all patients with partial epilepsy, we found that several lesions (e.g., poststroke, vascular malformation, tumors) were associated with a relatively high number of seizure-free patients, and some lesions (HS, cerebral dysgenesis) were associated with highly drug-resistant epilepsy. Previous studies focusing on specific brain lesions associated with partial epilepsy have demonstrated that the risk of seizure recurrence differs according to the type of CNS lesion. Patients with cerebral scars, such as poststroke³⁶ and post-traumatic lesions,³⁷ and patients with vascular malformation^38-40 have been associated with a more favorable prognosis. Conversely, patients with cortical dysplasia often have drug-resistant epilepsies.^41,42 We did not stratify patients by type of cerebral cortical dysgenesis because of the few patients in this category. Further studies are needed to investigate the AED response to the various types of cerebral cortical dysgenesis.

The presence of HS was the main risk factor for poor outcome, as demonstrated by multivariate analysis, and only 11% of such patients with partial epilepsy were seizure free. The worst prognostic factor is the presence of a dual pathology (i.e., the association of HS and another lesion, a relatively rare condition in patients with TLE).⁴¹

We suggest that the lesions associated with partial epilepsy are of great value for determining prognosis and should be considered in the future revised International Classification of Epilepsies and Epileptic Syndromes.

Acknowledgments

The authors thank the other neurologists who took part in this study: Profs. C. Pierrot-Deseilligny and D. Laplane, and Drs. N. Abou Jaoude, S. Dupont, D. Hasboun, and M. Levasseur.