What predicts enduring intractability in children who appear medically intractable in the first 2 years after diagnosis?




In a population-based retrospective cohort of children with newly diagnosed epilepsy, to determine (1) what proportion meet criteria for early medical intractability, and (2) predictors of enduring intractability.


Children with newly diagnosed epilepsy between 1980 and 2009 while resident in Olmsted County, MN, and followed >36 months, were stratified into groups based on both early medical intractability (“apparent” medical intractability in the first 2 years) and enduring intractability (persisting intractability at final follow-up or having undergone surgery for intractable epilepsy), and variables predicting these outcomes were evaluated.

Key Findings

Three hundred eighty-one children were included, representing 81% of our cohort with newly diagnosed epilepsy. Seventy five (19.7%) had early medical intractability, and predictors of this outcome on multivariable analysis were neuroimaging abnormality (risk ratio, 2.70; p = 0.0004), abnormal neurologic examination at diagnosis (risk ratio, 1.87; p = 0.015), and mode of onset (association was significant for focal vs. generalized onset [risk ratio, 0.25; p < 0.0001] but not unknown vs. generalized onset [p = 0.065]). After a median follow-up of 11.7 years, 49% remained intractable, 8% had rare seizures (≤ every 6 months), and the remainder were seizure-free. The only factor predicting enduring intractability on multivariable analysis was neuroimaging abnormality (risk ratio, 7.0; p = 0.0006).


Although a significant minority of children with early medical intractability ultimately achieved seizure control without surgery, those with an abnormal imaging study did poorly. For this subgroup, early surgical intervention is strongly advised to limit comorbidities of ongoing, intractable seizures. Conversely, a cautious approach is suggested for those with normal imaging, as most will remit with time.

Population-based studies in pediatric epilepsy have shown a relatively favorable long-term outcome, with nearly two thirds achieving seizure freedom for >3–5 years at last follow-up (Sillanpaa, 1993; Camfield et al., 1996). However, 8–22% have persistent seizures, despite trials of multiple antiepileptic drugs (AEDs), and epilepsy is then defined as medically intractable.

Medically intractable epilepsy leads to difficulties that extend far beyond uncontrolled seizures, including progressive intellectual disability, psychosocial morbidity, injury, and poor quality of life (Hermann et al., 2008). Complete seizure control is the best mechanism to limit such comorbidities. In select cases of medically intractable epilepsy, with an identified, surgically remediable focus, targeted resection is a viable option. Although earlier surgery offers the benefit of quicker seizure control, and possible reduction of associated comorbidities, resective brain surgery is not without risk. Hence, accurate and early prediction of medical intractability is crucial to make the correct management choice.

In adults, several studies have suggested that outcome of epilepsy can be predicted with reasonable accuracy early in the course (Elwes et al., 1984; Reynolds, 1987; Kwan & Brodie, 2000). Most patients achieve seizure freedom early, and remission rate is much reduced if seizures persist longer than 1–2 years after AED initiation (Reynolds, 1987). However, accurate, early prediction of medical intractability appears more difficult in children (Berg et al., 2001b; Geelhoed et al., 2005). Furthermore, there is debate in the literature as to how often children with medically intractable epilepsy will achieve seizure freedom without surgical intervention (Huttenlocher & Hapke, 1990; Callaghan et al., 2007).

We established a population-based cohort of new-onset pediatric epilepsy in Olmsted County, Minnesota, diagnosed from 1980 to 2009. The primary goals of this study were (1) to determine what proportion of children meet criteria for early medical intractability (“apparently” intractable epilepsy during the first 2 years after diagnosis), and (2) to evaluate long-term seizure outcome in these children. Specifically, we assessed what proportion with early medical intractability in the first 2 years had enduring intractability, defined as persistence of medically intractable seizures at final follow-up or undergoing epilepsy surgery for medically intractable seizures. In addition, we evaluated predictors of both early medical intractability and enduring intractability.


Case identification

Cases were ascertained retrospectively by screening the complete diagnostic indices of the Rochester Epidemiology Project, which include inpatient diagnoses, as well as diagnoses at the time of outpatient and emergency room visits at all medical care facilities in Olmsted County, Minnesota (Melton, 1996). Charts were screened using a diagnostic rubric, which included all seizure and convulsion diagnosis codes, and all identified charts were reviewed by a pediatric epileptologist. We identified children aged 1 month through 17 years with new-onset epilepsy diagnosed while residing in Olmsted County, Minnesota, between 1980 and 2009. Date of epilepsy diagnosis was defined as the date the child or teen was first given the diagnosis of epilepsy by a physician. Subjects were included in this study if they were followed for longer than 3 years after initiation of antiepileptic medication. This duration of follow-up was chosen to ensure a minimum of 1 year follow-up after the diagnosis of early medical intractability. Ethical approval for this study was received from the Mayo Clinic Institutional Review Board.


Epilepsy was defined as a predisposition to unprovoked seizures. Although most subjects had two or more unprovoked seizures, patients with a single unprovoked seizure who were determined to be at higher risk of seizure recurrence and commenced on prophylactic AED treatment were also included. An abnormal neurodevelopmental examination, focal abnormality on brain imaging, initial presentation in status epilepticus, or specific electroencephalography (EEG) findings (epileptiform discharge, intermittent rhythmic focal delta activity) were considered indicative of a higher risk of recurrence. Patients who were treated after a single seizure, but who lacked any of the preceding features were excluded. We included children who had two afebrile seizures occurring within 24 h, as these children likely have epilepsy (Camfield & Camfield, 2000).

Children presenting with acute symptomatic seizures alone, defined as “seizures at the time of a systemic insult or in close temporal association with an acute neurological insult” (Beghi et al., 2010) were excluded. Similarly, children who had only febrile seizures were excluded. Children with neonatal seizures were included only if their seizures recurred after 1 month of age.

For each patient, epilepsy was classified using the new International League Against Epilepsy (ILAE) Commission on Classification and Terminology 2005–2009 Report (Berg et al., 2010). Each subject was classified independently by two pediatric epileptologists, and disagreements were resolved by further discussion. Factors considered in the classification included seizure type(s) based on descriptive semiologies from the medical record, EEG and neuroimaging findings, cognitive function, and for some specific syndromes, age at onset. Each case was classified based on mode of onset (generalized/bilateral cortical or subcortical, focal/networks limited to one hemisphere, unknown—which included spasms), etiology (genetic, structural/metabolic, or unknown) and syndrome or constellation, if applicable. Etiology was determined after careful review of genetic, metabolic, and neuroradiologic data obtained by chart review. Genetic etiology was defined as cases where the epilepsy, as best understood, is the direct result of a known or presumed genetic defect, in which seizures are the core symptom of the disorder and furthermore, there is virtually no knowledge to support specific environmental influences as causes of or contributors to these forms of epilepsy. This category included the idiopathic generalized epilepsies as well as other documented genetic causes including Dravet syndrome. Structural etiology was defined as a documented structural abnormality of the brain on neuroimaging. Clinical entities that may be genetically inherited but that result in epilepsy because of structural brain change, such as tuberous sclerosis, were classified as structural. Metabolic etiology was defined if there was a documented metabolic disturbance which is known to be associated with a substantially elevated risk of developing epilepsy, regardless of whether this was genetically inherited. Unknown etiology was used if the cause was unknown and included all types of epilepsies with normal imaging and no documented genetic, metabolic, or immune etiology. This category was also used for the idiopathic focal epilepsies including benign epilepsy of childhood with centrotemporal spikes and Panayiotopoulos syndrome, where there may be some genetic contribution to the epilepsy, but current evidence does not suggest that genetic factors are paramount.

Early medical intractability was defined as a seizure frequency of more than every 6 months and failure of two or more antiepileptic medications for lack of efficacy within the first 2 years after epilepsy diagnosis. Enduring intractability was defined as either (1) persisting seizures and failure of two or more AEDs at final follow-up, or (2) having undergone epilepsy surgery for intractable seizures.

Data obtained from chart abstraction

Data variables abstracted from the medical charts are shown in Table 1. Magnetic resonance imaging (MRI) was not routinely used until 1984. Intellectual function was assessed by either formal neuropsychological testing (if available) or best clinical judgment by the reviewer (based on developmental milestones and academic achievement recorded in the patient history). Intellectual disability was defined as an estimated or measured developmental quotient of <70.

Table 1. Data abstracted from medical records
  1. NICU, neonatal intensive care unit.

DemographicAge at follow-up
HistoryPrenatal complications
Gestational age
Perinatal problems (NICU <7 days, NICU ≥7 days)
Postnatal brain injury (head injury, meningitis, encephalitis, ischemic brain injury, other)
Febrile seizures (prolonged, focal, clustering)
Epilepsy detailsSeizure type(s) and number/frequency at diagnosis
Age at onset of epilepsy
History of status epilepticus at diagnosis or ever
Family history of epilepsy in first-degree relatives
Neurologic examination

Normal/abnormal and type of abnormality

Intellectual functioning at initial diagnosis and final follow-up

Neuroimaging findingsMagnetic resonance imaging – type of abnormality and location
Computerized tomography – type of abnormality and location
EEG findingsEpileptiform discharge (presence, location) and background slowing (focal, generalized) at diagnosis, within the first 2 years of diagnosis and at final follow-up
OutcomeSeizure type(s) and frequency at 1 and 2 years after initial diagnosis and at final follow-up

Number of AEDs currently used and number of AEDs failed for lack of efficacy at 1 and 2 years after initial diagnosis, and at final follow-up

Ketogenic diet

Epilepsy surgery – resective or palliative (vagal nerve stimulator, callosotomy

Data analysis

What proportion of children meet criteria for early medical intractability?

Subjects meeting criteria for early medical intractability were reported as a percentage of the total cohort. The cohort was stratified into two groups, based on presence of early medical intractability. Clinical variables either defined at diagnosis or early in the course of epilepsy were selected and examined by chi-square analysis for categorical variables and t-test for continuous variables. Those factors showing statistical univariate association with early medical intractability (p < 0.05) were then considered as candidates for multivariable analysis.

Factors examined included age at epilepsy onset, sex, presence of intellectual disability, neurologic examination abnormality, perinatal complications (defined as remaining in hospital longer than 7 days at birth, excluding for maternal reasons alone), prior febrile seizure, prior atypical febrile seizure, number of total partial and generalized tonic–clonic seizures prior to AED treatment, mode of onset (generalized, focal, or spasms), etiology (genetic, structural/metabolic, or unknown), history of status epilepticus (febrile or afebrile) at presentation, presence of neuroimaging abnormality, presence and type of epileptiform discharge on initial EEG, presence of focal or generalized background slowing on initial EEG, and positive family history of afebrile seizures in first-degree relatives.

What is the long-term seizure outcome in children with early medical intractability?

Only children who met criteria for early medical intractability were selected for this part of the analysis. Long-term seizure outcome, defined by seizure frequency and medication use at last follow-up, were reported. Outcomes were stratified into four groups:

  1. Medically intractable at final follow-up: defined as ongoing seizures greater than every 6 months and failure of two or more AEDs for lack of efficacy.
  2. Ongoing seizures in the year preceding final follow-up but not medically intractable.
  3. Seizure-free in the year preceding final follow-up but continuing on AEDs.
  4. Seizure-free in the year preceding final follow-up and off AEDs.

The proportion of children who underwent epilepsy surgery for medically intractable seizures over the course of their epilepsy was noted, including type of surgical procedure and long-term seizure outcome. For children who achieved seizure freedom without resective surgery, the type of treatment resulting in seizure freedom (medication, ketogenic diet, vagal nerve stimulation) was noted.

What factors predict enduring intractability in children with early medical intractability?

Again, only children who met criteria for early medical intractability were selected for this part of the analysis. The cohort was then stratified into two groups for further analysis:

  1. Children who achieved seizure freedom for the year preceding final follow-up, on or off AEDs, without surgical intervention.
  2. Children who were not seizure-free in the year preceding final follow-up, or who had undergone surgical intervention for medically intractable seizures (enduring intractability).

Clinical variables were evaluated and those showing statistical association in univariable analysis were then considered as candidates for a multivariable analysis based on approach by Spiegelman and Hertzmark (2005). Associations were summarized as risk ratios and 95% confidence intervals using Genmod procedure from SAS version 9.1.3 (SAS Inc, Cary, NC, U.S.A.).


Four hundred sixty-eight children with new-onset epilepsy were identified over the study period. Of these, 381 (81%) were followed for at least 3 years after initial diagnosis and were included in the study. The study group consisted of 204 males (54%) and median age at onset of epilepsy was 6.0 years (25th–75th %ile, 2.6–10.8). Median duration of follow-up was 10.3 years (25th–75th %ile, 5.8–16.0). Of our cohort of 381 children, 100 (26%) were diagnosed between 1980 and 1989, 142 (37%) between 1990 and 1999, and 139 (37%) between 2000 and 2009.

What proportion of children met criteria for early medical intractability?

Seventy-five children (19.7%) met criteria for early medical intractability. Significant predictors of this outcome on univariate analysis are shown in Table 2, and include neuroimaging abnormality, abnormal neurologic examination, etiology and generalized background slowing on initial EEG (all p < 0.001), multifocal epileptiform discharge or hypsarrhythmia on initial EEG (p = 0.001), younger age at onset (p = 0.004), status epilepticus at or prior to initial diagnosis (p = 0.011), and intellectual disability at time of diagnosis (p = 0.043). In addition, subjects diagnosed in the 1980s were significantly more likely to have early medical intractability (31/100, 31%) than those diagnosed in the 1990s (24/142, 17%) or the 2000s (21/139, 15.1%), (p = 0.005).

Table 2. Predictors of early medical intractability
VariableEarly medical intractability (%)No early intractability (%)p-Value
  1. a

    Two had both generalized and focal modes of onset.

  2. SD, standard deviation.

Age at onset in years   
Mean (SD)5.5 (5.0)7.4 (5.1)0.004
Male gender39/75 (52.0)165/306 (53.9)0.77
Intellectual disability at diagnosis (4 unknown)33/72 (45.8)101/305 (33.1)0.043
Abnormal neurological exam36/75 (48.0)64/306 (20.9)<0.001
Perinatal complications (3 unknown)19/74 (25.7)51/304 (16.8)0.08
Neonatal seizures (5 unknown)7/73 (9.5)16/303 (5.3)0.17
Febrile seizure (2 unknown)13/75 (17.3)45/304 (14.8)0.59
Complex febrile seizure (2 unknown)9/75 (12.0)22/304 (7.2)0.17
Any status epilepticus at presentation18/75 (24.0)38/306 (12.4)0.011
Mode of onseta   
Generalized25 (34.7)68 (23.4)0.075
Focal42 (58.3)217 (74.8)
Unknown including spasms6 (8.2)21 (6.7)
Genetic17 (22.7)71 (23.2)<0.001
Structural/metabolic35 (46.7)71 (23.2)
Unknown23 (30.7)164 (53.6)
Number of generalized tonic–clonic seizures prior to treatment   
Mean (SD)0.8 (1.2)1.2 (2.3)0.16
Number of focal seizures prior to treatment   
Mean (SD)6.8 (18.3)4.8 (13.6)0.30
First degree family history of epilepsy (11 unknown)9/72 (12.5)50/298 (16.8)0.37
Abnormal neuroimaging35/75 (46.7)63/306 (20.6)<0.001
Epileptiform discharge on initial EEG (5 unknown as no EEG done)   
Any discharge63/74 (85.1)208/302 (68.9)0.005
Multifocal discharge or hypsarrhythmia25/74 (33.8)49/302 (16.2)0.001
Generalized discharge alone19/74 (25.7)61/302 (20.2)0.30
Background slowing on initial EEG (5 unknown as no EEG done)   
Generalized slowing35/74 (47.3)78/302 (25.8)<0.001
Focal slowing14/74 (18.9)50/302 (16.6)0.63

Three variables were found to be predictive of early medical intractability on multivariable analysis including neuroimaging abnormality (risk ratio 2.70 [95% confidence interval 1.56–4.69], p = 0.0004), abnormal neurologic examination (risk ratio 1.87 [95% confidence interval 1.13–3.10], p = 0.015), and mode of onset (p < 0.05). For mode of onset, the association was significant for focal compared to generalized (risk ratio 0.35, [95% confidence interval 0.22–0.57], p < 0.0001), meaning that focal mode of onset was less likely to be associated with early intractability. However, the association was not significant for unknown compared to generalized mode of onset (p = 0.065). For this multivariable model, the area under the curve is 0.73, which shows the model has good discrimination for development of early medical intractability (Hosmer & Lemeshow, 2000).

What is the long-term seizure outcome in children with early medical intractability?

The median follow-up duration for the 75 children who met criteria for early medical intractability was 11.7 years (25th–75th %ile, 5.9–19.1 years). Of children with early medical intractability, 64 (85%) had undergone MRI of the brain, and 10 (13%) had computerized tomography (CT) alone. The one child who was not imaged had refractory childhood absence epilepsy. Of those undergoing MRI, 31 studies (48%) were abnormal. Of those undergoing CT alone, 4 of 10 studies were abnormal. Of the 35 children with early medical intractability who had neuroimaging abnormalities, 20 (57%) were potential candidates for surgical resection, as they had either a single focal or hemispheric lesion (n = 16), tuberous sclerosis (n = 3), or mesial temporal sclerosis with periventricular leukomalacia (n = 1).

At final follow-up, 34 (45.3%) remained medically intractable, 7 (9.3%) still had seizures in the preceding year but were not intractable (i.e., their seizure frequency was not greater than every 6 months), 17 (22.7%) were seizure-free on medication, and 17 (22.7%) were seizure-free and off antiepileptic medication.

Sixteen children underwent epilepsy surgery after a median of 22.5 months after onset of intractability (25th–75th %ile, 9.3–63.0). Of these, 7 (44%) achieved seizure freedom at final follow-up. Of those achieving seizure freedom, 5/7 (71%) were able to discontinue AEDs. There was no correlation between latency to epilepsy surgery and final seizure freedom. Table 3 summarizes the surgery types performed and outcomes.

Table 3. Surgical procedures and outcomes
Surgery typeNumber of childrenSeizure-free off AEDs No. (%)Seizure-free on AEDs No. (%)Not seizure-free, not intractable No.Remains intractable No. (%)
Lesionectomy32 (67%)001 (33)
Temporal lobectomy22 (100)000
Extratemporal resection40004 (100)
Frontotemporal resection10001 (100)
Hemispherectomy402 (50)02 (50)
Callosotomy21 (50)001

Although children diagnosed in the 1980s had higher rates of early medical intractability, the proportion with enduring intractability (those who either remained medically intractable at final follow-up or had undergone epilepsy surgery) did not differ significantly based on decade of diagnosis (p = 0.32). Similarly, the proportion of children who underwent epilepsy surgery did not differ based on decade at diagnosis (p = 0.51).

Twenty (27%) children with early medical intractability had a defined electroclinical syndrome, and, of these seizure freedom without surgical intervention occurred in 12 (60%). This outcome was more likely to be seen in the idiopathic generalized epilepsies (9/12, 75%), including two of two with childhood absence epilepsy, two of three with juvenile absence epilepsy, one of one with juvenile myoclonic epilepsy, zero of two with idiopathic generalized epilepsy with generalized tonic–clonic seizures (GTCS) alone, two of two with genetic epilepsy with febrile seizures plus, and two of two with myoclonic atonic epilepsy). Outcome was less favorable with other syndromes, where seizure freedom without surgery was seen in only three (38%) of eight, including one of one with atypical benign partial epilepsy with continuous spike-wave in sleep (CSWS), 0/1 with Lennox-Gastaut, zero of one with Ohtahara, and three of five with West syndrome.

Of the 27 children who achieved seizure freedom without epilepsy surgery, 13 became seizure-free with addition of another AED (7 valproate, 2 lamotrigine, and 1 each levetiracetam, ethosuximide, zonisamide, and oxcarbazepine), 2 with withdrawal of carbamazepine (both of whom had generalized epilepsy), 2 had childhood syndromes that were outgrown (1 myoclonic atonic epilepsy of Doose, 1 atypical benign partial epilepsy with CSWS), 1 became seizure-free on the ketogenic diet, and 9 became seizure-free without any change in medication or addition of new therapies.

What factors predict for enduring intractability in children with early medical intractability?

Of children with early medical intractability, 48 had ongoing seizures, or had undergone epilepsy surgery at final follow-up, whereas 27 achieved seizure freedom without surgical intervention (Table 4). Risk factors for enduring intractability without surgery included mode of onset, etiology and neuroimaging abnormality (all p < 0.001), lack of generalized discharge alone on initial EEG (p = 0.001), focal slowing on initial EEG (p = 0.011), neonatal seizures (p = 0.033), lack of preceding febrile seizures (p = 0.035), and higher number of focal seizures prior to treatment (p = 0.049). There was a nonsignificant trend for intellectual disability at diagnosis to correlate with lack of seizure control (p = 0.07). Only neuroimaging abnormality was found to predict for enduring intractability in those with early medical intractability on multivariable analysis (risk ratio 7.0 [95% confidence interval 2.30–21.27], p = 0.0006) (Table 5). For this multivariable model, the area under the curve is 0.78, which shows that the model has very good discrimination (Hosmer & Lemeshow, 2000).

Table 4. Predictors of enduring intractability in children with early medical intractability?
VariableSeizure-free at last follow-up without surgical intervention (N = 27)Not seizure-free at last follow-up and/or surgical intervention (N = 48)p- Value
  1. a

    2 had both generalized and focal modes of onset.

Age at onset in years [Mean ± SD]5.55 ± 4.145.47 ± 5.490.94
Gender (male)14/27 (51.9)25/48 (52.1)0.99
Intellectual disability at diagnosis13/27 (48.1)13/48 (27.1)0.07
Abnormal neurologic exam10/27 (37.0)26/48 (54.2)0.15
Perinatal complications (1 unknown)5/27 (18.5)14/47 (29.8)0.29
Neonatal seizures (2 unknown)0/27 (0.0)7/46 (15.2)0.033
Febrile seizure8/27 (29.6)5/48 (8.6)0.035
Complex febrile seizure5/27 (18.5)4/48 (8.33)0.19
Any status epilepticus at presentation4/27 (14.8)14/48 (29.2)0.16
Mode of onseta   
Generalized16/25 (64.0)9/48 (18.8)<0.001
Focal6/25 (24.0)36/48 (75.0)
Unknown including spasms3/25 (12.0)3/48 (6.3)
Genetic13/27 (48.1)4/48 (8.3)<0.001
Structural/metabolic3/27 (11.1)32/48 (66.7)
Unknown11/27 (40.7)13/48 (27.1)
No. of generalized tonic–clonic seizures before AEDs [mean ± SD]0.8 ± 1.10.7 ± 1.30.81
No. of focal seizures before AEDs [mean ± SD]1.3 ± 4.310.0 ± 22.20.049
First-degree family history of epilepsy (3 unknown)5/27 (18.5)4/45 (8.9)0.23
Abnormal neuroimaging3/27 (11.1)32/48 (66.7)<0.001
Epileptiform discharge on initial EEG (1 unknown as no EEG done)   
Any discharge24/27 (88.9)39/47 (83.0)0.49
Multifocal discharge/hypsarrhythmia4/27 (14.8)12/47 (25.5)0.28
Generalized only discharge13/27 (48.1)6/47 (12.8)0.001
Background slowing on initial EEG (1 unknown as no EEG done)   
Focal slowing1/27 (3.7)13/47 (27.7)0.011
Generalized slowing13/27 (48.1)22/47 (46.8)0.91
Table 5. Probability of achieving seizure freedom without surgical intervention in patients with early medical intractability
NeuroimagingProbability95% lower95% upper

The neuroimaging findings of both children who did and did not achieve seizure freedom at final follow-up without surgery are summarized in Table 6.

Table 6. Summary of abnormal neuroimaging findings in 35 children with early medical intractability
FindingNo. with enduring intractabilityNo. who achieved seizure freedom at final follow-up without surgery
  1. a

    Also had associated mesial temporal sclerosis.

Remote ischemia or atrophy  
Focal, extratemporal20
Malformations of cortical development  
Focal, extratemporal10
Mesial temporal sclerosis alone30
Mesial temporal sclerosis with periventricular leukomalacia10
Tuberous sclerosis30
Low grade tumor20
Vascular malformation  
Hemispheric – Sturge-Weber10
Focal, temporal10
Focal, extratemporal10
Hypothalamic hamartoma10
Isolated agenesis of corpus callosum01
Calcified granuloma10
Focal inflammatory lesion in basal ganglia, not further defined10


Nearly 20% of children in our population-based, incidence cohort of newly diagnosed epilepsy met criteria for early medical intractability, failing two or more AEDs for lack of efficacy and having seizures more often than every 6 months within the first 2 years after diagnosis. Our incidence of early medical intractability is somewhat higher than two similar prior studies (Berg et al., 2001a; Geerts et al., 2012). Consistent in the definition of medical intractability in all three studies was lack of efficacy of two or more AEDs; however, both prior studies required a higher seizure frequency to meet intractability criteria. In a prospective cohort study of 595 children with new-onset epilepsy, 10% met criteria for intractability, which required an average of more than one seizure/month × 18 months, and no more than three consecutive months seizure-free (Berg et al., 2001a). In the Dutch study of epilepsy, 8.2% of children had a period of intractability, defined as no remission exceeding 3 months during a minimum 1-year observation period during the first 5 years after diagnosis (Geerts et al., 2012).

We found three significant predictors of early medical intractability on multivariable analysis: neuroimaging abnormality, abnormal neurological examination at diagnosis, and mode of onset. Neuroimaging abnormalities have been reported to correlate with poorer outcome (Semah et al., 1998; Ko & Holmes, 1999; Berg et al., 2001a; Geerts et al., 2012), but type of abnormality may be significant. High rates of medical intractability are reported with hippocampal sclerosis (67–89%) and cortical dysplasia (76–83%) compared to encephalomalacia (Semah et al., 1998). Prior results from a study assessing the impact of neuroradiologic abnormalities on course of epilepsy from a subgroup of our cohort suggested higher rates of intractability with malformations of cortical development or mesial temporal sclerosis compared to encephalomalacia (Dhamija et al., 2011). Abnormal neurologic examination at the time of diagnosis most likely reflects a congenital or early acquired brain lesion, and has been shown in numerous studies to correlate with intractability (Brorson & Wranne, 1987; Sillanpaa et al., 1995; Ko & Holmes, 1999; Camfield & Camfield, 2003). Mode of onset was significant, with focal mode of onset being associated with lower rates of early medical intractability than generalized mode of onset. Seizures with a focal mode of onset have been reported to be predictive of medical intractability in several, predominantly adult studies of new-onset epilepsy (Elwes et al., 1984; Cockerell et al., 1997), presumably as they are commonly associated with structural lesions in this age group. However, in the Connecticut study of new-onset epilepsy in children, mode of onset was not predictive of poor early outcome (Berg et al., 2001c). In the Dutch study of early prognosis of childhood epilepsy, only simple partial, as opposed to complex partial seizures, were predictive of worse outcome (Arts et al., 1999). We recently reported that more than half of nonidiopathic focal epilepsy in our population-based cohort of new-onset epilepsy in children has no known cause (Wirrell et al., 2011). The outcome for these children was favorable, with one third having a very benign disorder, remaining seizure-free after initiation of medication, or in those who were untreated, after their second afebrile seizure, and 81% ultimately achieving seizure freedom. In children, seizures with generalized modes of onset are seen with some of the early onset epileptic encephalopathies such as myoclonic-atonic epilepsy or Lennox-Gastaut syndrome, or with specific chromosomal or metabolic etiologies, all of which tend to be more pharmacoresistent, as well as the idiopathic generalized epilepsies, which are usually pharmacoresponsive.

Although outcome in our cohort of children with early medical intractability was concerning, after long-term follow-up, just over one third achieved seizure freedom with medical treatment alone, and an additional 9.3% became seizure-free after successful epilepsy surgery. A small number of studies have reported on long-term outcome of medically intractable epilepsy. Although Callaghan found that only 11% continuing medical treatment achieved a 6-month terminal remission and remained seizure-free at follow-up (Callaghan et al., 2007), Huttenlocher reported a somewhat brighter picture, particularly if cognition was normal (Huttenlocher & Hapke, 1990). Seizure control was achieved in 4% of children with intractable epilepsy who had normal cognitive function per year, but in only 1.5% of those with low IQ. Similar to our results, Berg et al. found that after two AEDs had failed for lack of efficacy, 37.5% achieved a 1 year remission at last contact, after median follow-up of 11.8 years. Significant predictors of this outcome included idiopathic forms of epilepsy and lower prior seizure frequency (Berg et al. 2009). In the Dutch Study of Epilepsy, of the 34 subjects who experienced early medical intractability, which was defined as no remission exceeding 3 months during the first 5 years despite optimal use of at least two AEDs, only 19 (56%) continued to be medically intractable at final follow-up (Geerts et al., 2012). Subjects who remained intractable at final follow-up had a significantly longer time to intractability and tended to have a higher rate of development of mental retardation during follow-up. Finally, further work from Carpay et al. (1998) found that the probability of achieving good seizure control dropped with successive numbers of failed AEDs, with 51% achieving good outcome after one failed AED, 29% after two failed AEDs, and only 10% after three failed AEDs.

We found only one significant factor on multivariable analysis that predicted for enduring intractability in children with early medical intractability. Of patients with a neuroimaging abnormality only 8.6% achieved ultimate seizure control without surgery. This outcome was achieved in 20% of children with remote ischemic lesions, versus no child with malformation of cortical development, tuberous sclerosis, mesial temporal sclerosis, or other focal lesions.

We did not address in detail specific electroclinical syndromes, as the numbers of patients with each of these syndromes was small. However, similar to the results of Berg et al. (2009), we found that children with idiopathic generalized epilepsy who had early medical intractability overall still had a favorable long-term outcome, with 75% achieving ultimate seizure freedom.

Several previous studies have found accurate prediction of outcome in pediatric epilepsy to be difficult. Using a statistical model to predict outcome based on two large prospective cohort studies, Geelhoed et al. (2005) noted prediction was incorrect in nearly one third of patients. The difficulty in accurately predicting outcome raises the question whether one should “wait it out” to see if a child will spontaneously remit before offering more invasive treatment, such as surgical resection. Our data suggest that this approach is not warranted in children with early medical intractability and a neuroimaging abnormality, as most will not achieve control of seizures without surgery. Numerous studies have shown that focal MRI abnormalities are one of the most significant predictors of seizure freedom after epilepsy surgery (Tonini et al., 2004; Cossu et al., 2008). Conversely, a more cautious approach should be taken in children with normal imaging, particularly in the absence of other adverse predictors on univariate analysis.

Our study has a number of strengths. The Rochester Epidemiology Project allowed for complete identification of a population-based incidence cohort of children with new-onset epilepsy, and provided access to all inpatient and outpatient records in the community. Therefore, our study cohort was not contaminated by selection bias. Secondly, our patients had ready access to high-quality health care. Most were seen by a child neurologist either at the time of diagnosis or shortly thereafter, and underwent neuroimaging with magnetic resonance imaging. Thirdly, many of our patients were followed into their adult years, allowing a clearer understanding of their long-term prognosis.

Several limitations also exist. Many of our subjects did not benefit from the significant neuroimaging and genetic advances made in the last two decades, and, as such, it is likely that some were erroneously classified as unknown etiology, given this limitation.

In summary, we found that a significant minority of children who appeared medically intractable in the first 2 years after epilepsy onset ultimately achieved seizure control without surgery. However, those who had an abnormal imaging study were very unlikely to become seizure-free with medical therapy alone, and early surgical intervention is strongly advised to limit comorbidities of ongoing, intractable seizures. Greater caution is advised for those with normal imaging, as most of these children ultimately remitted without surgery.


This study was supported by a CR20 Research award from the Mayo Foundation, and made possible by the Rochester Epidemiology Project (Grant # R01-AG034676 from the National Institute on Aging).


None of the authors has any conflict of interest to disclose. 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.