In clinical epidemiology, prognosis refers to the prediction of course and long-term outcome of a disease. Prognosis reflects the continuing relationship between predictors and outcomes in a defined population with a defined disease condition (Hemingway, 2006). Studying prognosis helps us understand the course and outcome, as well as the public health impact of the condition (Hesdorfter & Logroscino, 2003). In this review, we first discuss the methodologic aspects and challenges in interpreting prognosis studies in epilepsy in general, followed by a critical review of the prognosis of idiopathic generalized epilepsy (IGE) and its subsyndromes based on the published literature.
Prognosis describes the trajectory and long-term outcome of a condition. Most studies indicate a better prognosis in idiopathic generalized epilepsy (IGE) in comparison with other epilepsy syndromes. Studies looking at the long-term outcome of different IGE syndromes are relatively scant. Childhood absence epilepsy appears to have a higher rate of remission compared to juvenile absence epilepsy. In absence epilepsies, development of myoclonus and generalized tonic–clonic seizures predicts lower likelihood of remission. Although most patients with juvenile myoclonic epilepsy (JME) achieve remission on antiepileptic drug therapy, <20% appear to remain in remission without treatment. Data on the prognosis of other IGE syndromes are scarce. There are contradictory findings reported on the value of electroencephalography as a predictor of prognosis. Comparisons are made difficult by study heterogeneity, particularly in methodology and diagnostic criteria.
In any prognosis study the key components are case ascertainment, defining predictors, and outcome measures of prognosis.
The natural history of epilepsy spans from the biologic onset of the condition to a prognostic endpoint such as death, remission, or intractability. In epidemiologic studies, patients may be captured at different stages, starting from the first seizure to the prognostic endpoint (Fig. 1). In this paradigm, two types of study cohorts can be identified. The incident cohort ideally consists of new-onset cases. It is widely believed this is the best population for prognosis studies to limit selection bias (Laupacis et al., 1994). It is also important to recognize that new-onset epilepsy and newly diagnosed epilepsy are not the same (Thurman et al., 2011).
On the contrary, both new and old cases are included in the prevalent cohort. Because these two cohorts represent patients at different stages of the condition, one could perceive that different subsets of these two cohorts may have different prognoses.
Both cohorts are drawn either from the community, primary care, or hospitals ranging from secondary care to tertiary centers. Such a population is usually structured in a pyramidal fashion (Fig. 2). The geometry of the pyramid depends on several factors, primarily related to the health care system access and patient referral habits. When interpreting results of prognosis studies, it should be borne in mind that prognosis will differ at each level of case ascertainment in the pyramid (Hemingway, 2006).
In this double dichotomous model (incident vs. prevalent and community-based vs. hospital-based), bias could occur at several levels as highlighted in Table 1 (Berg & Shinnar, 1994; Hesdorfter & Logroscino, 2003).
|Design: Prospective vs. retrospective|
|Cohort: Incident vs. new onset vs. prevalent cases|
|Setting: Community vs. tertiary center|
|Duration of follow-up and attrition|
|Evolution of definitions: Classification systems, diagnostic criteria, operational groups, inclusion and exclusion criteria for research purposes|
The diagnostic accuracy, definitions, inclusion, and exclusion criteria as well as classifications are likely to affect the outcome of prognosis studies. The diagnostic criteria of IGE and syndromes have been evolving over the last two decades. Hence, the prognostic outcomes in different studies are likely to reflect the era in which the study was conducted and the inclusion and exclusion criteria used. The diagnosis of epilepsy is subject to interobserver variations depending on the degree of expertise (van Donselaar et al., 2006). Classification bias may also influence prognostic observations.
Study design is likely to have an impact on the results. Prospective studies can be designed to acquire more specific data to control for potential confounders and avoid bias. Attrition and the length of follow-up will also influence the final outcome (MacDonald, 2001). The remission rate in epilepsy depends on the length of follow-up, explaining variable prognosis at different stages of follow-up (MacDonald, 2001).
In prognosis studies, the predictors could be defined in the form of a triangle of interacting causes consisting of environmental factors (e.g., socioeconomic conditions), host factors (e.g., demographics, psychosocial), and disease factors (e.g., pathophysiology; Hemingway, 2006; Fig. 3).
Most commonly studied predictors of epilepsy prognosis include etiology, age of onset, seizure type and frequency, electroencephalography (EEG) findings, temporal pattern of seizures, epilepsy syndrome, early response to antiepileptic drug (AED) treatment, and family history of epilepsy (Table S1).
Outcome measures of prognosis
Prognostic outcomes are variable depending on the study population and can generally be classified as fatal events, nonfatal events, patient-centered (patient-reported) events, and wider burden (Hemingway, 2006; Table 2 and Fig. 3). In epilepsy prognosis studies, the most commonly measured outcomes are mortality, remission, intractability, and social outcomes such as education, employment, and marriage (Hesdorfter & Logroscino, 2003).
|Fatal events||Mortality rate, case fatality rate, standardized mortality ratio, proportionate mortality ratio|
|Nonfatal events||Remission, relapse, intractability, evolution into a different syndrome, injuries, cognition, behavior|
|Patient-centered events||Quality of life, functional status, symptoms|
|Wider burden||Sickness leave, employment, educational status, disability allowance, driving|
Remission refers to seizure freedom over a defined period of time. A recent International League Against Epilepsy (ILAE) commission report has defined this as 2 or 5 years of seizure freedom off antiepileptic drugs (AEDs; Thurman et al., 2011).
Prognosis of IGE as Reflected in Epilepsy Prognosis Studies
Insights into the prognosis of IGE in comparison to other epilepsies are best reflected in large-scale prognosis studies on epilepsy. Given below is a summary of major population-based as well as hospital-based studies.
The Rochester Project is a prospective study involving an incident cohort of all ages. Remission was defined as 5-year seizure freedom. As it was established prior to the current ILAE classification of epilepsy syndromes (ILAE, 1989), IGE does not appear as a discrete category in the original data analysis. However, it is still possible to extract some information on IGE from subgroup analysis based on etiology, and seizure type. In the classification, etiology was defined as idiopathic, secondary (defined as central nervous system lesions acquired postnatally), and major neurologic dysfunction of uncertain cause (Annegers et al., 1979). Both focal and generalized epilepsies were likely to constitute the idiopathic group demonstrating a 20-year remission rate of 74% in comparison to 46% in the group with neurologic dysfunction. The probability of achieving remission at 10 years after diagnosis without medication was higher in the idiopathic group compared to the secondary group (36% vs. 20%). With regard to seizure type, the probability of 20-year remission was higher for generalized tonic–clonic seizures (80%) and absences (85%) compared to complex partial seizures (65%), which, however, was not specified for myoclonic seizures (Annegers et al., 1979). These data provide indirect evidence suggestive of better prognosis of the idiopathic epilepsies in comparison to the symptomatic and focal epilepsies. The authors provided a reclassification of syndromes in a subsequent paper, which, however, did not carry data on prognosis (Zarrelli et al., 1999).
A population-based, long-term, prospective study on the prognosis of childhood epilepsy has been reported from Finland (Sillanpaa et al., 1998; Sillanpaa & Schmidt, 2006, 2009). The cohort consisted of both incident (61%) and prevalent (39%) cases. Multivariate analysis has revealed etiology, type of seizure, response to AEDs, and initial frequency of seizures as significant predictors of remission. Idiopathic and cryptogenic epilepsy had a significantly higher rate of remission (92% and 68%, respectively) as compared to remote symptomatic epilepsy (45%; Sillanpaa et al., 1998). Compared to the Rochester study, this is a higher rate of remission. Inclusion of only childhood cases, having a mix of incident and prevalent cases, as well as longer follow-up could account for this discrepancy. Subsequently, the researchers published the analysis of the incident cohort (Sillanpaa & Schmidt, 2006, 2009). Sixty-seven percent of the total cohort achieved terminal remission at the end of the mean follow-up of 37 years (Sillanpaa & Schmidt, 2006). One-year terminal remission was achieved by 76%. On univariate analysis, seizure frequency of less than weekly (before or during treatment) and idiopathic/cryptogenic etiology were found to be associated with entering remission. However, on multivariate analysis, only seizure frequency of less than weekly emerged as a significant association (Sillanpaa & Schmidt, 2009).
Another prospective, population-based study involving newly diagnosed epilepsy patients, the National General Practice Study of Epilepsy from the United Kingdom used a classification of etiology and seizure types along the lines adopted by the Rochester Project (Sander et al., 1990). At 9-year follow-up, 5-year terminal remission was achieved by 54% of the total cohort with definite epilepsy, 69% with idiopathic epilepsy and 61% with remote symptomatic epilepsy (Cockerell et al., 1997). This study failed to establish etiology and seizure type as significant prognostic factors (MacDonald et al., 2000). Lack of magnetic resonance imaging (MRI) and EEG in some cases, which would affect classification as well as differential mortality in patients with severe pathology, might account for this discrepancy. The fact that the final diagnosis was made by a group of experts based on retrospective medical records rather than direct patient interviews also limits the diagnostic precision.
The Nova Scotia cohort is a population-based, incident cohort of childhood epilepsy studied prospectively. Potential cases were identified from the EEG laboratory. This study was classified as a population-based study on the basis that all children in the community who had a seizure would have an EEG. In the early analysis, patients were classified into three groups based on seizure type: absences, partial seizures + primary generalized tonic–clonic, and secondarily generalized epilepsies (Camfield et al., 2002). The remission in each group was found to be 65%, 52%, and 33%, respectively (Camfield & Camfield, 2003). Remission was defined as seizure freedom off medication at the end of follow-up, but the duration of seizure freedom was not predefined. According to the seizure classification, patients with IGE were likely to be spread across all three groups. However, the researchers published data on IGE prognosis subsequently, which will be discussed later in this article.
The Connecticut cohort consists of incident cases of childhood epilepsy being followed up prospectively. Cases were recruited from most, but not all, community clinics, academic centers, and private practices in the community. Hence, this is not a strictly population-based cohort. The authors used ILAE syndromic classification (ILAE, 1989). With a median follow-up of 5.3 years, the 2-year remission of the total cohort was found to be 74%. IGE demonstrated a higher remission (85%) than symptomatic/cryptogenic epilepsies (47%). Remission of childhood absence epilepsy (CAE) (89%) and juvenile absence epilepsy (JAE) (87%) was higher than JME (58%). On multivariable analysis, higher initial seizure frequency, family history of epilepsy, remote symptomatic etiology, and EEG slowing emerged as factors significantly associated with a lower chance of remission. IGE was associated with a higher likelihood of remission (Berg et al., 2001). Detailed assessment, syndromic classification, and prospective follow-up are the major strengths of the study.
The Dutch Study of Epilepsy in Childhood is a hospital-based, prospective, incident cohort study on the prognosis of childhood epilepsy (Geerts et al., 2010). In the cohort, 5-year terminal remission was achieved by 70.9% (61.9% off AEDS, 9% on AEDs). Remission of CAE (82%) and JAE (75%) was higher than JME (57.1%). Similar to most studies cited in preceding text, idiopathic etiology was found to be a significant predictor of good prognosis. The duration of active epilepsy and mortality were significantly higher in the remote symptomatic group compared to idiopathic (Geerts et al., 2010). The researchers also studied intractability as an outcome. Intractability was defined as no seizure freedom >3 months on treatment during a minimum of 1 year observation. Intractability was experienced by 12.1% of the cohort during the 15-year follow-up period. Those with idiopathic epilepsies had a significantly lower rate of refractoriness compared to cryptogenic and remote symptomatic etiologies (Geerts et al., 2012). The size of the cohort (413), long prospective follow-up (mean 14.8%), detailed assessment, and ILAE classification by an expert panel are the main strengths of the study. As for drawbacks, being a hospital-based study, there is a bias in case ascertainment.
A retrospective study involving both adults and children with newly diagnosed epilepsy from two Italian centers, reported remission (defined as ≥2 year seizure freedom after initial AED therapy) in 62.5% (Del Felice et al., 2010). IGE constituted 20.4% of the cohort. The predictors of early remission (defined as 2-year seizure control immediately after treatment commenced) versus late remission (2-year seizure control achieved at least 24 months after AEDs commenced) were studied. Only the interaction between seizure type and number of seizures before commencing treatment emerged as a predictor of late remission. Etiology and epilepsy syndrome were not found to be significant predictors of late remission, which contradicts findings from population-based studies. The tertiary referral center bias, retrospective study design, and underrepresentation of children (only 10.5%) may account for this discrepancy and lower remission.
As shown earlier, comparisons are difficult due to heterogeneity of studies. However, in general, studies demonstrate that the majority of patients with epilepsy, particularly with childhood epilepsy, enter long-term remission. Most of the studies show that idiopathic etiology is a predictor of good prognosis in terms of remission and intractability, providing indirect evidence that IGE has better prognosis in comparison to symptomatic epilepsies.
Cognitive and behavioral outcomes
Cognitive and behavioral outcomes of prognosis are gradually gaining greater emphasis among researchers. It has been recognized that children with epilepsy have poor social, cognitive, and behavioral outcomes. This aspect becomes relevant in IGE, as it is predominantly a childhood-onset condition. School children with newly diagnosed epilepsy have been found to have more cognitive and behavioral problems than their healthy classmates (Oostrom et al., 2003a,b). The cohort had a mix of both IGE (24.6%) and cryptogenic epilepsies. Hermann et al. (2006) reported similar findings from a study involving children with new-onset IGE and focal epilepsy in comparison to healthy controls. In terms of academic problems, they did not find any significant difference between children with IGE and focal epilepsy. From a prospective, controlled study Oostrom et al. (2005) demonstrated that measures of learning, memory span, attention, and behavior were worse among children diagnosed with epilepsy compared to health controls. The sample consisted of both idiopathic and cryptogenic epilepsies. The etiology, seizure remission, or AED use did not have an impact on the outcome. A prospective study evaluated the impact of neurobehavioral comorbidities (attention-deficit/hyperactivity disorder, academic problems) on the cognitive development of children with newly diagnosed epilepsy in comparison to health controls. It was found that the presence of neurobehavioral comorbidity at the time of epilepsy diagnosis was associated with significantly worse cognitive outcome subsequently (Hermann et al., 2008).
The underpinning mechanisms for poor cognitive and behavioral outcomes in childhood onset epilepsy (including IGE) are not entirely clear. Researchers have hypothesised the influence by epileptogenesis, preexisting neurodevelopmental abnormalities, genetic susceptibility, and environmental factors (Hermann et al., 2012).
The Long-Term Outcome of IGE in General
In the literature, some epidemiologic studies analyze the prognosis of IGE as a group, whereas a few focus on different IGE syndromes. The most commonly studied outcome measure has been remission. However, the definition of remission is highly variable between the studies.
The reported remission rate of IGE ranges from 64–82% (Mohanraj & Brodie, 2007; Kharazmi et al., 2010; Szaflarski et al., 2010). The lower rate of 64% was reported in a cross-sectional study from a specialist epilepsy clinic (Mohanraj & Brodie, 2007). Overrepresentation of refractory cases due to selection bias as well as underrepresentation of children with good prognosis could account for the lower remission.
Among different studies, predictors of poor outcome include history of febrile seizures (Mohanraj & Brodie, 2007), age of onset <5 years, “atypical” presentation defined as atypical absences, myoclonic epilepsies, generalized tonic–clonic seizures with onset <3 and >20 years (Nicolson et al., 2004), and EEG asymmetries (focal slowing, focal epileptiform discharges, asymmetric generalized spike-wave discharges; Szaflarski et al., 2010).
The heterogeneity between the three studies needs to be taken into consideration in the final interpretation. All three studies are from tertiary centers. However, due to its comprehensive referral system, the Finnish study (Sillanpaa et al., 1998) is likely to represent a wider spectrum of severity and a mix of incident as well as prevalent cases. The study by Szaflarski et al. (2010) used 3-month remission, whereas the other two studies analyzed 1-year remission.
Prognosis of Epilepsies with Absence Seizures
Studies done before the era of ILAE syndromic classification provide prognosis of epilepsies with absences without a specific syndromic diagnosis.
A meta-analysis found remission rates ranging from 0.21 to 0.89 with an average of 0.59 (Bouma et al., 1996). Patients who experienced only typical absence seizures achieved a higher rate of remission (range 0.15–1, average 0.7) compared to those who developed generalized tonic–clonic seizures (range 0–0.84, average 0.35). However, the studies were heterogeneous in terms of demographics, inclusion criteria, definitions, and methodology including statistical methods, which may explain the wide range of the results.
Predictors of remission
Predictors of prognosis are also variable and reflect the heterogeneity of the studies. However, notable predictors of good prognosis include normal intelligence and normal neurologic examination, whereas development of generalized tonic–clonic seizures appears to be a predictor of bad prognosis (Currier et al., 1963; Gibberd, 1966; Sato et al., 1976, 1983). Late onset of absence seizures (>8 years), electrographic photosensitivity, and lack of posterior delta rhythm seem to predispose to development of “grand mal” seizures (Loiseau et al., 1983).
Prognosis of Childhood Absence Epilepsy
The diagnostic criteria of CAE have been a focus of debate and discussion. The ILAE (1989) defined CAE as a syndrome with a peak onset age of 6–7 years, characterized by very frequent absences accompanied by 3-Hz, bilateral, synchronous, symmetrical, generalized spike-wave discharges with a normal background on the EEG. However, some authorities such as Loiseau and Panayiotopoulos have considered this definition too broad and have proposed more stringent criteria (Hirsch & Panayiotopoulos, 2005; Panayiotopoulos, 2005). When comparing results from different studies, it should be borne in mind that the criteria used in the particular study have an impact on the outcome data.
A wide range in remission exists among published studies due to variations in diagnostic criteria, age at last assessment, follow-up duration, therapy, and inclusion of absence epilepsies other than CAE (Hirsch & Panayiotopoulos, 2005). Experts believe that if strict diagnostic criteria are used, CAE carries an excellent prognosis and successful AED withdrawal is possible in those who are in remission with a normalized EEG (Hirsch & Panayiotopoulos, 2005; Panayiotopoulos, 2005).
The reported remission rate in CAE ranges from 56–84% (Wirrell et al., 1996; Bartolomei et al., 1997; Trinka et al., 2004; Grosso et al., 2005; Callenbach et al., 2009). Heterogeneity of the studies would account for this wide range. Both the lowest (Trinka et al., 2004), and the highest (Callenbach et al., 2009) remissions were reported in studies from tertiary centers. However, only the latter was a prospective study. It is possible that patients in remission might have been underrepresented in the retrospective study (“convenient sample bias”), contributing to a lower remission rate.
Inclusion criteria and classification have a strong influence on the outcome as demonstrated in a retrospective study (Grosso et al., 2005). The researchers classified patients with CAE into two groups based on the ILAE classification (group 1; ILAE, 1989) and more stringent criteria proposed by Panayiotopoulos (2005; group 2). The group 2 recorded higher rate of terminal remission defined as ≥1 year seizure freedom off AEDs (82% vs. 51%), lower generalized tonic–clonic seizures (8% vs. 30%), and no relapses on AED discontinuation (0% vs. 22%; Grosso et al., 2005).
Other seizure-related outcomes
The percentage of patients who go on to develop generalized tonic–clonic seizures ranges from 8–69% among the studies (Wirrell et al., 1996; Bartolomei et al., 1997; Trinka et al., 2004; Grosso et al., 2005; Callenbach et al., 2009; Vierck et al., 2010). Family history of generalized tonic–clonic seizures and late onset (>8 years) absence seizures seem to predict development of generalized tonic–clonic seizures (Loiseau et al., 1983; Vierck et al., 2010). Whether this observation relates to syndrome evolution or earlier age of onset of a seizure type seen typically as part of more refractory syndrome such as JME remains contentious. In a retrospective study, it was demonstrated that 15% of patients with CAE developed JME later (Wirrell et al., 1996).
Predictors of remission
Most studies have reported the development of myoclonus (Wirrell et al., 1996; Trinka et al., 2004; Grosso et al., 2005), and generalized tonic–clonic seizures (Trinka et al., 2004; Grosso et al., 2005) as predictors of bad prognosis. Other poor prognostic features include absence status, late onset (>8 years), family history of generalized seizures, background slowing of EEG (Wirrell et al., 1996), and atypical EEG features (irregular discharges, “lead-in” discharges; Grosso et al., 2005). On the contrary, EEG features, generalized tonic–clonic seizures, or family history of epilepsy did not influence the final outcome in a prospective study (Callenbach et al., 2009).
Contrary to common belief that CAE is a benign disorder, several studies have shown that children with CAE have poor psychosocial outcomes (Hertoft, 1963; Olsson & Campenhausen, 1993; Wirrell et al., 1997). Outcome data are more useful when compared with reference groups such as other IGE syndromes, people without epilepsy, and patients diagnosed with another chronic disease. In one of the early studies published in 1963, it was noted that among a cohort with “petit mal epilepsy,” 20% had intelligent quotient <90, 20% had substantial social difficulties, and 14% were surviving on welfare benefits (Hertoft, 1963). In another study involving children with typical absence seizures, almost one third were found to be having social maladjustment (Loiseau et al., 1983). However, both studies did not provide comparisons with a reference group from the general population, making it difficult to draw firm conclusions from the data. Poor social and educational performances as well as psychosocial problems were significantly more common among patients diagnosed with CAE (38%) compared to JAE (15%) in a retrospective study (Bartolomei et al., 1997). A population-based study found that young adults with absence epilepsy not in remission were significantly more employed in manual work compared to the reference group without epilepsy (Olsson & Campenhausen, 1993). Wirrell et al. (1997) compared psychosocial outcomes between young adults diagnosed with typical absence epilepsy and juvenile rheumatoid arthritis. Those with absence epilepsy recorded significantly worse outcomes in terms of academic/social achievements and behavioral aspects.
Prognosis of Juvenile Absence Epilepsy
The 1989 ILAE definition of JAE is based mainly on the age of onset and seizure frequency. Compared to CAE, the onset of JAE is around puberty and absences occur less frequently in a sporadic manner. Panayiotopoulos (2005) has proposed some modifications to this definition, with particular reference to excluding those patients with absences accompanied by marked myoclonus, extremely mild or clinically undetectable impairment of consciousness, and consistent sensory precipitation.
Experts consider JAE to be a life-long condition, although seizures are usually well controlled with AED therapy (Wirrell, 2003; Panayiotopoulos, 2005). Absence seizures tend to become less severe with age (Panayiotopoulos, 2005).
Prognosis studies in JAE are scarce. The percentage of patients who achieve remission ranges from 37–62% (Loiseau et al., 1995; Bartolomei et al., 1997; Trinka et al., 2004; Tovia et al., 2006; Aiguabella Macau et al., 2011). All the cited studies were retrospective and hospital/clinic-based. Three studies adopted ILAE criteria. The study by Tovia et al. (2006) did not specify diagnostic criteria while Loiseau et al. (1995) used more strict criteria. Remission was defined as 2-year seizure freedom in three studies (Loiseau et al., 1995; Trinka et al., 2004; Aiguabella Macau et al., 2011), and not specified in the other two. The most important difference between the studies is the length of follow-up. The study with the longest follow-up (mean 30 years) reported the highest remission, (Trinka et al., 2004). Late remissions are known to occur in this condition; hence the length of follow-up is a vital factor in the interpretation of remission (Loiseau et al., 1995).
There is not much data on AED withdrawal in JAE. In one retrospective study involving 21 patients, AED withdrawal was unsuccessful when attempted on eight patients in remission (Aiguabella Macau et al., 2011).
Predictors of remission
Predictors of outcome are not well delineated in most studies. Trinka et al. (2004) found development of myoclonus and generalized tonic–clonic seizures as predictors of lower remission in the combined group of CAE and JAE. In another cohort of JAE and CAE, those who responded to initial appropriate AED therapy were significantly more likely to achieve terminal remission and less likely to progress to JME (Wirrell et al., 2001), which raises the possibility of pharmacogenomic predictors. On the contrary, a different study did not find any significant predictors of outcome (Bartolomei et al., 1997).
Other seizure-related outcomes
Among the studies the proportion with generalized tonic–clonic seizures ranges from 47–95% (Loiseau et al., 1995; Bartolomei et al., 1997; Trinka et al., 2004; Tovia et al., 2006; Aiguabella Macau et al., 2011). Notably, the highest frequency of 95% was reported in the study with the longest follow-up (Trinka et al., 2004). Most often generalized tonic–clonic seizures occur after the onset of absence seizures (Panayiotopoulos, 2005), hence the length of follow-up is important in this respect. Significantly higher number of patients were found to have coexistent generalized tonic–clonic seizures in JAE compared to CAE in two studies (JAE 81% and 95% vs. CAE 38% and 69%; Bartolomei et al., 1997; Trinka et al., 2004). On follow up, significantly more patients evolved to develop JME among JAE (19%) compared to CAE (5%; Trinka et al., 2004).
Prognosis of Juvenile Myoclonic Epilepsy
JME is defined as an epilepsy syndrome appearing around the time of puberty with bilateral myoclonic jerks predominantly in the arms, usually occurring shortly after awakening. Generalized tonic–clonic seizures and photosensitivity are often associated with the syndrome, but absences are infrequent and less common (ILAE, 1989; Thomas et al., 2005). Historically, underdiagnosis has been a major pitfall (Genton & Gelisse, 2001).
There is general belief that patients with JME need life-long AED therapy, and withdrawal of treatment may not be appropriate in seizure-free patients due to high risk of relapse (Panayiotopoulos, 2005; Schmidt, 2011). However, this is an area where evidence is insufficient and further research is needed.
Janz (1985) reported that 75% of his patients achieved 2-year seizure freedom on medication. However, remission rates in JME appear to vary according to the seizure type among different studies. The highest remission has been reported for generalized tonic–clonic seizures (64.6–81.4%; Kleveland & Engelsen, 1998; Baykan et al., 2008). All three seizure types (myoclonus, generalized tonic–clonic seizures, absences) in combination and myoclonus alone account for the lowest rates (44.2% and 64.6%; Kleveland & Engelsen, 1998; Baykan et al., 2008), whereas the combination of myoclonus and generalized tonic–clonic seizures lies in between (27–65.1%; Kleveland & Engelsen, 1998; Baykan et al., 2008). In another study, after a mean follow-up of 26 years, 78% were seizure free, although the seizure types were not specified (Camfield & Camfield, 2009). The follow-up duration, methodology, and diagnostic criteria vary among the studies, which make comparisons difficult.
Remission off AED therapy is an outcome measure of great practical importance. Unfortunately, data are scarce. In his original work, Janz (1985) reported that 91% patients relapsed on AED dosage reduction or withdrawal after being in remission for 2 years. Only a minority (7–8%) achieved remission without antiepileptic medication in two retrospective, hospital-based studies (Kleveland & Engelsen, 1998; Baykan et al., 2008). This could be an underestimate due to convenient sample bias. On the contrary, a population-based, prospective study found total seizure remission in 17% and “partial” remission (presence of only myoclonus) in 13% after a mean follow-up of 25 years. The main drawback of the study is small sample size (24; Camfield & Camfield, 2009).
The chronological evolution of myoclonic seizures was observed in a long-term study. Myoclonic seizures tend to subside with time and achieve remission or become mild in the fourth decade of life (Baykan et al., 2008).
Predictors of remission
Predictors are not well delineated among the published studies. Further dissections of JME into subsyndromes appear to have a considerable impact on prognosis. Based on a single, large-scale, multicenter study involving 257 probands and families, JME was subclassified into four phenotypic subsyndromes as; classic JME, CAE persisting as JME, JME with adolescent onset pyknoleptic absences, and JME with astatic seizures (Martinez-Juarez et al., 2006). The best outcome was seen in classic JME with 58% achieving freedom from all seizure types on AED therapy and 5% off medication. CAE persisting as JME demonstrated the worst outcome with only 7% free from all seizure types and 66% free from only generalized tonic–clonic seizures. None was in remission without AED therapy except in the classic JME group. Although the sample size was large (222), there was case ascertainment bias, as the patients were selected from a large number of affected families and followed-up in specialist clinics.
Response to AED therapy
The general perception is that complete seizure control is achieved in 80–90% with appropriate AED therapy (Panayiotopoulos et al., 1994; Thomas et al., 2005). Pseudoresistance due to lifestyle issues and poor compliance is seen in some (9.7–16.7%), whereas true resistance to AED therapy is reported in around 16% (Gelisse et al., 2001; Baykan et al., 2008). Coexistent psychiatric disorders are predictors of true resistance (Gelisse et al., 2001; Baykan et al., 2008). In another study, higher frequencies of reflex seizure traits (praxis and language induced) and personality disorders as well as higher anxiety scores were found to be associated with patients with poor seizure control compared to those who were seizure free (Guaranha et al., 2011).
Certain seizure type combinations (generalized tonic–clonic seizures + myoclonus + absences and generalized tonic–clonic seizures + myoclonus) also appear to be associated with poor response to AEDs (Matsuoka, 1992; Gelisse et al., 2001). There is conflicting data on EEG as a predictor. One study found poor response to AED therapy was associated with focal epileptiform discharges, less photosensitivity, and more neuropsychological EEG activation (Matsuoka, 1992), whereas another study failed to establish such an association (Baykan et al., 2008). Variations in EEG methodology could be, at least partially, responsible for this discrepancy.
In a retrospective study, the presence of structural brain abnormalities on neuroimaging was not found to influence the seizure control (Gelisse et al., 2000).
Despite good seizure control, social outcomes are not always favorable in JME. In a small population-based study of JME, only 69% were employed and 74% reported at least one major unfavorable social outcome. Almost half of the cohort experienced behavioral problems at school, whereas 80% of the pregnancies were unplanned and without a stable relationship. Although there were no comparisons with a reference group, no relationship was found between seizure control and social outcomes (Camfield & Camfield, 2009).
Prognosis of Other IGE Syndromes
There are limited data available on the prognosis of other IGE syndromes. The long-term outcome of IGE with generalized tonic–clonic seizures was reported in a population-based study with an average follow-up of 22.2 years. The data reveals a fairly benign outcome in terms of seizure control. In the cohort of 40 patients, 92% became seizure free and 75% reached terminal remission off AED therapy. None developed intractable epilepsy. The syndrome remained stable throughout the follow-up period and only one evolved into a different syndrome (JME). However, despite good seizure control, 76% had poor social outcomes in relation to education, employment, relationships, financial success, substance abuse, comorbidities, and social activities (Camfield & Camfield, 2010).
Another study reported the long-term outcome of two patients diagnosed with eyelid myoclonia with absences. Both continued to experience absences as well as eyelid myoclonia despite treatment with multiple antiepileptic drugs during the follow-up period of 7 and 13 years (Siren et al., 2002).
Evolution of Epilepsy Syndromes
Evolution from one epilepsy syndrome to another is sometimes seen in the natural history. Among IGE syndromes, evolution from CAE and JAE to JME has been reported (Wirrell et al., 1996; Trinka et al., 2004; Martinez-Juarez et al., 2006). Martinez-Juarez et al. (2006) described CAE persisting as JME to be a phenotypic subsyndrome of JME. Whether this phenomenon represents true evolution or merely a subset of the phenotypic spectrum is a matter of interpretation depending on the diagnostic criteria. Potential errors in the diagnosis and classification due to misinterpretation of initially available data may also compromise the interpretation of results.
EEG as a Predictor of Prognosis
Although the use of EEG in the diagnosis and classification of epilepsy is well established, its value as a predictor of prognosis in IGE is less well defined with studies reporting variable and conflicting results. Furthermore, only a small number of studies have evaluated this aspect (Table 3).
|References||Design||Cohort||Syndrome||N||Age onset range years (mean)||Statistics||Outcome||EEG features studied||EEG poor prognostic features|
|Bartolomei et al. (1997)||R||Hospital||CAE, JAE||80||10–26 (9.14)||UV, MV||Response to AEDs||FA, PS, PD, POS||POS in sleep in CAE|
|Baykan et al. (2008)||R||Hospital||JME||48||(14.4 ± 2.9)||UV||Remission||FA, PS, SWDHV, NRX||None|
|Benjamin et al. (2011)||R||Hospital||CAE, JME, IGE||21||2–20||UV||Response to AEDs||GSWD frequency||GSWD frequency <3.2 Hz|
|Callenbach et al. (2009)||P||Hospital||CAE||47||1–9.7 (5.4)||UV||Remission||BG, ED||None|
|Currier et al. (1963)||R||Hospital||Petit mal||32||3–17||NS||Remission||GSWD frequency/duration/amplitude, EEG seizure||None|
|Del Felice et al. (2010)||R||Hospital||IGE, focal, single seizure||352||3–84 (31.5)||UV, MV||Remission||ED, slow activity||Abnormal EEG (only in UV analysis)|
|Gibberd (1966)||R||Hospital||Petit mal||139||NS||NS||Remission||BG, GSWD||None|
|Grosso et al. (2005)||R||Hospital||CAE||119||NS||UV, MV||Remission||Typical GSWD, atypical (PS, abnormal BG, irregular GSWD, “lead-in,” fixation-off||Atypical EEG|
|Hedstrom and Olsson (1991)||R||Community||Absence epilepsy||97||<15||UV||Development of GTCS||BG, duration of GSWD, PD, POS, FA, irregular GSWD||Lack of PD, GSWD duration <10 s, FA|
|Kamel et al. (2010)||R||Hospital||IGE, focal||34||NS||UV||Remission||ED||Presence of ED|
|Loiseau et al. (1983)||R||Hospital||Absence epilepsy||90||NS||UV||Remission, development of GTCS||Regularity of GSWD, PD, PS, absence seizure on EEG||PS, lack of PD|
|Matsuoka (1992)||R||Hospital||JME||32||NS||NS||Response to AEDs||NS||Focal ED, lack of PS, ED on neuropsychological activation|
|Mohanraj and Brodie (2007)||P||Hospital||IGE||103||5–51||UV||Remission||ED||None|
|Nicolson et al. (2004)||R||Hospital||IGE||962||NS||UV||Remission||GSWD, PS, FA||None|
|Sato et al. (1976)||P||Hospital||Absence epilepsy||52||5.3–24.3 (10.5)||UV, MV||Remission||BG normal or abnormal||None|
|Sinclair and Unwala (2007)||R||Hospital||Absence epilepsy||119||NS||NS||Remission||Typical vs. atypical (BG slow, FA, irregular GSWD)||None|
|Szaflarski et al. (2010)||R||Hospital||IGE||267||NS||UV||Remission||Symmetry of GSWD, focal ED, focal slow||Focal ED, focal slow, asymmetric GSWD|
|Verrotti et al. (2009)||P||Hospital||IGE||63||6.7–20.8||UV||Remission||PS||None|
|Vierck et al. (2010)||R||Hospital||CAE||115||(6.5)||UV, MV||Development of GTCS||FA, POS||None|
|Wirrell et al. (1996)||R||Population||CAE||72||1–14 (5.7)||UV, MV||Remission||GSWD, BG, PS, SWDHV||BG slowing|
Many variables affect the yield of EEG abnormalities in IGE (Seneviratne et al., 2012). Therefore, heterogeneity of studies, particularly in methodology, may account for the conflicting results reported on the prognostic value of EEG in IGE.
Difficulties in Interpretation and Comparison of Current Data
In general, large community-based studies are suggestive of better prognosis in idiopathic epilepsies compared to other epilepsies (Annegers et al., 1979; Sillanpaa et al., 1998). However, results from prognosis studies on IGE itself are variable and at times contradictory. Heterogeneity of studies could be a major factor responsible for such inconsistency. This could arise at different levels as highlighted below.
Inclusion/exclusion criteria and definitions
How cases, predictors, and outcomes are defined will affect the results. The study cohorts will differ according to the diagnostic criteria used. One good example is the case of epilepsies with typical absences. The different prognoses in CAE with ILAE and Panayiotopoulos criteria have already being alluded to in a previous section. Similarly, there are gray cases in the border zone between CAE and JAE, and how these cases are classified is not uniform. How remission is defined varies among studies, which obviously has a bearing on the results.
Prospective versus retrospective studies
Due to its planned nature, prospective studies are more likely to yield more robust results. On the contrary, retrospective studies are relatively easy to conduct, but are more likely to have missing information. Inability to standardize methodology is a major drawback in retrospective studies.
Incident versus prevalent cohorts
Incident and prevalent cases represent patients at different stages of the natural history with different prognoses. Hence, comparison of incident with prevalent cohorts can yield misleading conclusions.
Standardization of methodology
Comparison of studies would be difficult if the methodology was not standardized. This is particularly evident in the use of EEG as a predictor of prognosis.
Length of follow-up
The length of follow-up could affect the results. Some patients may enter remission after several years. Hence, in the same cohort, a longer follow-up is likely to yield a higher remission rate.
Selection of outcome measures
There are several outcome measures of prognosis as shown in Table 2. However, most studies have focused on only remission and mortality. Other outcomes such as quality of life, social, cognitive, and behavioral outcomes have received relatively less emphasis. Reporting remission alone from a study does not provide the complete picture of prognosis.
The differences in statistical methods employed should also be taken into consideration when interpreting the results. Both univariate and multivariate analyses have been used with different outcomes. Multivariate model would be more appropriate for defining predictors when multiple factors are involved.
Overfitting refers to a potential problem in multivariate analysis yielding unreliable results when the number of outcomes is too few compared to the number of independent variables. As a rule of thumb, for each variable there should be 10 or more outcome events (Concato et al., 1993). Cox Proportional Hazards regression analysis becomes unreliable when the relative risk of the variable changes with time (“nonproportional hazard”; Concato et al., 1993).
Conclusions, Suggestions, and Future Directions
Prognosis studies are essential for us to understand the effect of IGE on the patient, family, and society, with this knowledge ultimately leading to improved management decisions. Prognostic outcomes appear to vary among different electroclinical syndromes of IGE; yet no firm conclusions can be drawn on predictors due to methodological heterogeneity. In general, seizure remission appears to be more favorable in IGE compared to other epilepsies. However, psychosocial, cognitive, and behavioral outcomes of childhood epilepsies (including IGE) appear to be worse in comparison to healthy controls.
Although some studies have yielded conflicting results, general trends can be observed, which enable us to make suggestions on management of IGE. The small number of studies available indicates that in JME <20% achieve remission without treatment suggesting caution when AED withdrawal is contemplated. Some authors indicate that JAE needs lifelong AED therapy. Available data, although scant, indicate that in JAE, similar to CAE, occurrence of generalized tonic–clonic seizures and myoclonus predicts poor outcome in terms of seizure remission. On the contrary, most patients with CAE, particularly when strict diagnostic criteria are employed, enter remission by puberty. Hence, AED withdrawal may be appropriate in those typical cases in terminal remission. With conflicting data, no clear recommendations can be made with regard to the use of EEG as a predictor of prognosis. Prospective studies involving incident cases are needed for us to obtain a clearer understanding of the natural history of IGE. These studies need to look beyond seizure remission at the emerging psychosocial, cognitive, and behavioral aspects of the condition, both as prognostic outcomes and predictors to guide management.
Dr Seneviratne reports no disclosures. Professor Cook has received speaker honoraria from UCB Pharma and Sanofi Australia and travel honoraria from UCB Pharma and SciGen. He has received research grants from National Health and Medical Research Council (Australia), and Australian Research Council. He has also received a Science, Technology, and Innovation Grant from the State Government of Victoria, Australia. A/Professor D’Souza has received travel, investigator-initiated, and speaker honoraria from UCB Pharma, educational grants from Novartis Pharmaceuticals and Pfizer Pharmaceuticals, educational, travel and fellowship grants from GSK Neurology Australia, and honoraria from SciGen Pharmaceuticals. 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.