Onset of intractability and its course over time: The Dutch study of epilepsy in childhood


Address correspondence to Ada Geerts, Department of Neurology Ee 2293, Erasmus MC, PO Box 1738, 3000 DR Rotterdam, The Netherlands. E-mail: mail@adageerts.nl


Purpose:  Intractability in epilepsy is difficult to define, and little is known about its onset, course, and duration. We investigated these aspects (as well as the occurrence of intractability) during long-term follow-up in patients with epilepsy, focusing on possible explanations for the variation in time of onset and duration of intractability.

Methods:  After diagnosis, 453 patients with childhood-onset epilepsy had a 5-year follow-up with regular visits and data collection. Ten years later they received a questionnaire with items concerning epilepsy, which was completed by 413 patients resulting in a mean follow-up of 15 years. Intractability during the first 5 years was compared with that in the last year of follow-up. Intractability was defined as having no 3-month remission during a 1-year period despite adequate medical treatment.

Key Findings: At least 12.1% of the cohort had a period of intractability during the 15-year follow-up, and 8.5% were intractable in the final year. Of the patients with idiopathic etiology 4.3% had a period of intractability versus 17.0% for those with cryptogenic, and 22.6% for those with remote symptomatic etiology (p < 0.001). Other risk factors at baseline were younger age at first seizure, generalized cryptogenic/symptomatic or localization-related symptomatic epilepsy, mental retardation, and febrile convulsions before enrollment. The cumulative risk of a period of intractability was 6.1% (95% confidence interval [CI] 3.7–8.5) at 2 years follow-up and 8.2% (95% CI 5.4–11.0) at 5 years. The mean time to onset of intractability during the first 5 years of follow-up was 1.6 (95% CI 1.3–2.0; median 1.0) years and the mean duration of intractability during these 5 years was 3.3 (95% CI 2.8–3.8; median 3.6) years. Fifteen patients were intractable only during the first 5 years of follow-up (group A), and 19 subjects were intractable both during the first 5 years and the last year of follow-up (group B). Compared with group A, group B had shorter remission and a longer time to intractability during the first 5 years and more were intractable in the fifth year of follow-up. Sixteen other patients had a late onset of intractability after 5 years of follow-up, sometimes after long periods of remission (group C). No significant differences in baseline characteristics were found among groups A, B, and C, but slightly more children in groups B and C became mentally retarded during the follow-up. In all groups, antiepileptic drugs were of little use in preventing and ending intractability.

Significance:  There is a large unpredictable variation in time of onset, course, and duration of intractability, with a higher chance of final intractability after a poor course during the first 5 years of follow-up. The natural course of epilepsy probably best explains the variable course of intractability. The effect of medication seems to be minor.

Recent studies indicate that the course of childhood-onset epilepsy may vary during long-term follow-up. Periods of remission and periods with frequent seizures can interchange (Berg et al., 2004, 2006; Sillanpää & Schmidt, 2006a; Berg et al., 2009; Geerts et al., 2010). The end of an episode with frequent seizures could result from the effectiveness of treatment, or from the self-limiting character of some types of epilepsy. Conversely, persistent seizures (or so-called intractability) starting late in the course of epilepsy could result from poor adherence with the prescribed treatment, from development of drug resistance or tolerance, or from an ongoing process of epileptogenesis that is insensitive to treatment (Loscher & Schmidt, 2006; Abou-Khalil, 2007; Brown et al., 2009; Vervloet et al., 2011).

Studying the course of intractability, especially its time of onset and the interchanging pattern with periods of remission, may enable one to find explanations for the development, discontinuation, or chronicity of intractability, and to distinguish the natural course of epilepsy from treatment effects.

Between 1988 and 1992, we recruited 494 consecutive children who presented at the hospital with new-onset epilepsy. Results of the clinical course at 2 and 5 years after diagnosis have been published (Arts et al., 1999, 2004) as well as the results of 413 patients with a mean 15-year follow-up (Geerts et al., 2010); this latter publication presented risk factors for final intractability (nonidiopathic etiology, febrile seizures before enrollment, no 3-month remission in the first 6 months of follow-up, and having a eriod of intractability during the first 5 years of follow-up).

This report focuses on the course of intractability during follow-up and possible reasons for its variation. A comparison is made between intractability in the first 5 years and intractability in the final year.


The methods of this study have been reported in detail elsewhere (Geerts et al., 2010). There follows a brief overview.


Between 1988 and 1992, four hospitals in The Netherlands recruited 494 children aged 1 month to 16 years with newly diagnosed epilepsy. The ethical committees of the participating hospitals approved the study. Written informed consent was provided by all parents and children aged ≥12 years. This Dutch Study of Epilepsy in Childhood had a recruitment rate approximating 75% of the expected annual incidence in the referral area. Excluded from the study were children with seizures as a result of brain tumors or other progressive diseases diagnosed at the time of recruitment.


Of the original cohort, 453 subjects had a 5-year follow-up with regular visits and data collection. Ten years later, they received a simple questionnaire (that did not burden the subjects too much) with items addressing epilepsy, such as date of the very last seizure, use of antiepileptic drugs (AEDs) at last contact and, if applicable, date of AED discontinuation. When subjects reported seizures in the last year of follow-up, we contacted their treating physician or checked their medical records to determine intractability in the final year. This means that only the information covering the first 5 years and the last year of follow-up was sufficiently detailed to investigate intractability. Nonresponders were contacted again and the final response rate based on the original 494 patients was 84%. Of the 413 respondents, 47.2% was male. Mean age at epilepsy onset was 5.5 years (median 5.1, range 0.0–15.5) years. Mean follow-up was 14.8 years (median 14.8, range 11.6–17.5) years.


All seizures and epilepsies were classified at enrollment and, if necessary, revised after 2 years using the seizure, etiology, and epilepsy classifications of the International League Against Epilepsy (Gastaut, 1969; ILAE, 1981, 1989, 1993). Etiology at enrollment was idiopathic in 50.8%, cryptogenic in 21.3%, and remote symptomatic including mental retardation (intelligence quotient [IQ] <70) in 27.8%. At the end of the present study, the etiologic classifications were checked again, as some patients proved to have neurologic brain-related morbidities.


Intractability: no remission exceeding 3 months (at least one seizure per 3 month) during a minimum period of 1 year of observation despite adequate treatment (Arts et al., 2004). Adequate treatment: optimal choice and use of at least two AEDs, either alone or in combination. AED intolerance or allergy was not considered as failure. To make the analysis pragmatic, we calculated the duration of intractability as the interval between the first and last seizure of a period with at least one seizure per 3 month. Early onset of intractability: onset of intractability within the first 5 years of follow-up. Late onset of intractability: onset after the first 5 years of follow-up. Fast response to medication: 6 months of remission starting within 2 months after initiation of AED. Longest remission: maximum interval between seizures. Terminal remission: interval between the very last seizure and the end of follow-up. A 5-year terminal remission was reached by 71% of the cohort.


To investigate possible differences between groups, Chi-square statistics were used in case of categorical variables and one-way analysis of variance (ANOVA) in case of continuous variables. For the risk of intractability we used Kaplan-Meier survival statistics. Log rank (Mantel-Cox) was used to test for equality of the survival distributions for the different types of epilepsy.


Entire follow-up

Figure 1 shows how many subjects were intractable during the 15-year follow-up and at which points in time. At least 50 patients had a period of intractability (12.1%): in 34 subjects it started during the first 5 years of follow-up (8.2% with early onset) and in 16 other subjects it started later and was present during the last year of follow-up (3.9% with late onset). In the last year of follow-up a total of 35 patients were intractable (8.5% of 413). The mean follow-up of these 35 patients did not differ from those that were not intractable in the final year (14.6 vs. 14.8 years, ANOVA: F = 0.698, d.f. = 1, p = 0.40), neither did the median follow-up (14.5 [range 13.0–16.8] years versus 14.9 [range 11.6–17.5] years), independent samples median test: p = 0.22).

Figure 1.

Flowchart of patients with or without intractability during a mean follow-up of 15 years (TR, terminal remission).

Another 31 patients with frequent seizures during the first 5 years were not considered intractable during that particular period despite their seizure pattern, because of insufficient compliance (n = 6), or inadequate treatment (n = 25), according to our definition of intractability (Fig. 1). The same is true for four patients with inadequate treatment in the last year of follow-up and one patient with insufficient compliance in that year. For all the patients with inadequate treatment (no optimal choice and use of at least two AEDs) there was a plausible reason.

Between the first 5 years and the last year of follow-up (intervening years: Fig. 1), 283 patients (13 + 270) had definitely not been intractable during this period as could be derived from the date of their last seizure or AED discontinuation. The remaining 76 patients (15 + 61) may have had a period of intractability in these intervening years. This means that the proportion of patients with at least one period of intractability during the entire 15-year follow-up ranges from a minimum of 12.1% definite intractable cases (34 + 3 + 13 = 50 of 413) to a possible maximum of 30.5% (34 + 3 + 13 + 15 + 61 = 126 of 413).

Intractability during the first 5 years of follow-up

During these years, the mean time to onset of intractability was 1.6 (95% confidence interval [CI] 1.3–2.0, median 1.0) years, and its mean duration was 3.3 (95% CI 2.8–3.8, median 3.6) years. The cumulative risk of intractability was 6.1% (95% CI 3.7–8.5) at 2 years of follow-up, and 8.2% (95% CI 5.4–11.0) at 5 years. Figure 2A,B shows the Kaplan-Meier curves for several types of epilepsy according to the classifications of the ILAE and of Berg et al. (ILAE, 1989; Berg et al. 2006). Using the ILAE classification, there was a significant difference between the curves in Fig. 2A (Log rank [Mantel-Cox] χ2 = 33.78, d.f. = 6, p < 0.0001). Those with West syndrome or Lennox-Gastaut syndrome or epilepsy with myoclonic absences or myoclonic-astatic seizures (generalized cryptogenic and/or symptomatic epilepsies) had the highest risk of intractability in the first 5 years, whereas none of the patients with localization-related idiopathic etiology had a period of intractability. Using the Berg classification, the difference between the three curves in Fig. 2B was also significant (log rank [Mantel-Cox] χ2 = 27.24, d.f. = 3, p < 0.0001). However, the difference between the focal (localization-related cryptogenic and symptomatic groups) and catastrophic (West syndrome or Lennox-Gastaut syndrome, epilepsy with myoclonic-astatic seizures or myoclonic absences, epilepsies and syndromes undetermined whether focal or generalized) curves was only marginally significant (log rank [Mantel-Cox] χ2 = 3.44, d.f. = 1, p = 0.064). Of the focal group, 11.7% developed intractability during the first 5 years versus 21.7% of the catastrophic group. Time to onset of intractability during the first 5 years of follow-up did not differ between patients with focal epilepsy and those with catastrophic epilepsy (independent samples Mann-Whitney U-test: p = 0.149).

Figure 2.

(A,B) Time course and cumulative risk of intractability during the first 5 years of follow-up for different types of epilepsy. (A) Gen crypto/symp, generalized cryptogenic, and/or symptomatic epilepsies and syndromes (West syndrome; Lennox-Gastaut syndrome, epilepsy with myoclonic-astatic seizures or with myoclonic absences); loc rel symp, localization-related symptomatic epilepsies and syndromes; gen symp, generalized symptomatic epilepsies and syndromes; loc rel crypto, localization-related cryptogenic epilepsies and syndromes; gen idiop, generalized idiopathic epilepsies and syndromes; loc rel idiop, localization-related idiopathic epilepsies and syndromes. None of the rest group (n = 16) became intractable in these years (not in figure). (B) Catastrophic, all cryptogenic and symptomatic generalized epilepsies, and epilepsies with both focal and generalized features; focal, cryptogenic and symptomatic focal; idiopathic, partial, and generalized forms of idiopathic epilepsy.

Risk factors

Of the patients with idiopathic etiology, 4.3% had a period of intractability either during the first 5 years and/or last year of follow-up, whereas this was 17.0% for those with cryptogenic and 22.6% for those with remote symptomatic etiology (p < 0.001). Other risk factors at baseline were younger age at first seizure, type of epilepsy (generalized cryptogenic/symptomatic, localization-related symptomatic), mental retardation at enrollment, and febrile convulsions before enrollment (Table 1).

Table 1.   Baseline characteristics and variables collected during follow-up in relation to the three intractability groups. The last three columns present the statistical significance of possible differences found in the pairwise comparison of the groups with intractability
 Intractability during 15 yearsa
(n = 50)
Intractable in the first 5 years only
A (n = 15)
Intractable in the first five & last year of follow-up
B (n = 19)
Intractable in the last year of follow-up only
C (n = 16)
  1. LR, localization-related; TR5, terminal remission at 5 years of follow-up; intract., intractability; int., interval; –, not possible or not meaningful to calculate.

  2. a Either during the first 5 years and/or during the last year of follow-up.

  3. b Later reclassified to nonidiopathic epilepsy.

  4. c In the 1st 5 years of follow-up.

  5. d n = 12.

  6. e n = 14.

  7. f n = 18.

  8. #p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001.

 Age at onset: mean years (median; range)*5.1 (3.4; 0.2–13.5)4.1 (2.7; 0–15.3)4.1 (3.1; 0.1–11.6)nsnsns
 Type of epilepsy***   nsnsns
  Generalized idiopathic (n = 177)8 (4.5%)3 (1.7%)2 (1.1%)3 (1.7%)b   
  Generalized symptomatic (n = 28)4 (14.3%)2 (7.1%)2 (7.1%)0   
  Generalized cryptogenic/symptomatic (n = 32)13 (40.6%)3 (9.4%)6 (18.8%)4 (12.5%)   
  LR idiopathic (n = 23)0000   
  LR symptomatic (n = 59)13 (22.0%)5 (8.5%)4 (6.8%)4 (6.8%)   
  LR cryptogenic (n = 78)12 (15.4%)2 (2.6%)5 (6.4%)5 (6.4%)   
  Rest (n = 16)0000   
 Type of epilepsy (Berg et al., 2006)***   nsnsns
  Idiopathic (n = 200)8 (4.0%)3 (1.5%)2 (1.0%)3 (1.5%)   
  Focal (n = 137)25 (18.2%)7 (5.1%)9 (6.6%)9 (6.6%)   
  Catastrophic (n = 60)17 (28.3%)5 (8.3%)8 (13.3%)4 (6.7%)   
  Rest (n = 16)0000   
 Etiology at enrollment***   nsnsns
  Idiopathic (n = 210)9 (4.3%)3 (1.4%)3 (1.4%)3 (1.4%)   
  Remote symptomatic (n = 115)26 (22.6%)9 (7.8%)9 (7.8%)8 (7.0%)   
  Cryptogenic (n = 88)15 (17.0%)3 (3.4%)7 (8.0%)5 (5.7%)   
 Mentally retarded at enrollment***   nsnsns
  Yes (n = 86)20 (23.3%)7 (8.1%)7 (8.1%)6 (7.0%)   
  No (n = 327)30 (9.2%)8 (2.4%)12 (3.7%)10 (3.1%)   
 Febrile convulsions before enrollment*   nsnsns
  Yes (n = 42)9 (21.4%)2 (4.8%)5 (11.9%)2 (4.8%)   
  No/unknown (n = 371)41 (11.1%)13 (3.5%)14 (3.8%)14 (3.8%)   
 Status epilepticus before enrollmentns   nsnsns
  Yes (n = 31)3 (9.7%)01 (3.2%)2 (6.5%)   
  No (n = 382)47 (12.3%)15 (3.9%)18 (4.7%)14 (3.7%)   
 Positive family history ns   nsnsns
  Positive for epilepsy (n = 55)8 (14.5%)1 (1.8%)5 (9.1%)2 (3.6%)   
 Negative/unknown (n = 358)42 (11.7%)14 (3.9%)14 (3.9%)14 (3.9%)   
 EEG at enrollmentns   nsnsns
  Normal (n = 101)12 (11.9%)3 (3.0%)6 (5.9%)3 (3.0%)   
  Epileptiform (n = 244)30 (12.3%)9 (3.7%)12 (4.9%)9 (3.7%)   
  Other abnormalities (n = 68)8 (11.8%)3 (4.4%)1 (1.5%)4 (5.9%)   
During follow-up       
 Number of seizures <AED (59 subjects no AEDs)#   nsnsns
  10 or less (n = 167)16 (9.6%)6 (3.6%)5 (3.0%)5 (3.0%)   
  >10 (n = 169)31 (18.3%)7 (4.1%)14 (8.3%)10 (5.9%)   
  Unknown (n = 18)3 (16.7%)2 (11.1%)01 (5.6%)   
 >25 Seizures in first 6 months of follow-up***   nsns#
  Yes (n = 150)29 (19.3%)10 (6.7%)13 (8.7%)6 (4.0%)   
  No (n = 263)21 (8.0%)5 (1.9%)6 (2.3%)10 (3.8%)   
 Longest remission: mean years (median; range)c***1.1 (0.6; 0.1–3.4)d0.6 (0.4; 0.0–2.0)2.2 (1.6; 0.7–4.7)#****
 TR5: mean years (median; range)***0.7 (0.1; 0.0–3.4)e0.2 (0; 0.0–2.0)f0.9 (0.1; 0.0–4.7)#ns#
 Int. intake-first AED: mean months (median; range)ns0.5 (0.03; 0–5.1)1.9 (0.16; 0–19.7)2.0 (0.2; 0–16.9)nsnsns
 Int. intake-second AED: mean months (median; range)3.8 (1.4; 0–16)9.4 (4.2; 0–43)13.1 (3.9; 0–56.7)dns#ns
 Fast response to AED (59 subjects no AEDs)***   nsnsns
  Yes (n = 150)6 (4.0%)2 (1.3%)1 (0.7%)3 (2.0%)   
  No (n = 204)44 (21.6%)13 (6.4%)18 (8.8%)13 (6.4%)   
 Time to intract.: mean years (median; range)c1.2 (1.0; 1.0–2.3)2.0 (1.3; 1.0–5.0)Unknown*
 Duration of intract. mean years (median; range)c3.1 (3.2; 1.0–4.9)3.4 (3.9; 1.0–5.0)Unknownns
 Number of AEDs: mean (median; range)c***5.1 (5.0; 2–9)4.7 (4.0; 2–8)2.4 (2.0; 1–4)ns******
 Polytherapy: (59 subjects no AEDs)c***   ns*****
  Yes (n = 90)40 (44.4%)14 (15.6%)19 (21.1%)7 (7.8%)   
  No (n = 264)10 (3.8%)1 (0.4%)09 (3.4%)   
 Intractable during the 5th year (n = 19)4 (21.1%)15 (78.9%)0**
 Status epilepticus during follow-up***   nsnsns
  Yes (n = 5)4 (80%)1 (20.0%)2 (40.0%)1 (20.0%)   
  No (n = 408)46 (11.3%)14 (3.4%)17 (4.2%)15 (3.7%)   
 Mentally retarded during follow-up***   nsnsns
  Yes (n = 32)10 (31.3%)1 (3.1%)5 (15.6%)4 (12.5%)   
  No (n = 381)40 (10.5%)14 (3.7%)14 (3.7%)12 (3.1%)   
 Etiology at end of follow-up***      
  Idiopathic (n = 192)6 (3.1%)2 (1.0%)2 (1.0%)2 (1.0%)nsnsns
  Remote symptomatic (n = 151)36 (23.8%)11 (7.3%)14 (9.3%)11 (7.3%)nsnsns
  Cryptogenic (n = 70)8 (11.4%)2 (2.9%)3 (4.3%)3 (4.3%)nsnsns

Three intractability groups

We distinguished three groups of subjects with intractability: 15 subjects with intractability during the first 5 years only (group A), 19 subjects with intractability during both the first 5 years and the final year (group B), and 16 subjects with intractability during the final year but not during the first 5 years (group C). The mean follow-up of group A, B, and C, and the group without intractability, showed no significant differences (14.8, 14.4, 14.9, and 14.8 years, respectively, ANOVA: F = 0.648, d.f. = 3, p = 0.59); neither did the median follow-up (14.7 [range 13.3–16.8] years, 14.4 [range 13.0–16.4] years, 14.9 [range 13.2–16.8] years, and 14.9 [range 11.6–17.5] years, independent samples median test: p = 0.33). A summary of the individual course of epilepsy and treatment effects on the 15 patients in group A is presented in Table S1. Eleven patients were seizure free in the last year of follow-up; seven of them even had a 5-year terminal remission. For five patients of group A, we found a plausible explanation for the ending of their intractability, that is, seizures were abolished or the frequency was reduced after the introduction of a new AED in four patients and after a successful temporal lobectomy in one patient (Table 2).

Table 2.   Possible explanations for temporary, continuing, or late onset of intractability
Group A (n = 15)
Group B (n = 19)
Late onset
Group C (n = 16)
Effect of a new AED4  
Effect of a new AED, but not enough to be seizure free in the last year1  
Effect of temporal lobectomy1  
No treatment effect916 
Three-year remission while on AEDs but relapsed 1 
Parietal arachnoid cyst: AED no effect. Five-year remission after drain removal. Relapse after 2 years without AEDs. Reinstitution of AEDs no effect 1 
Stop AED after remission: AED failed after relapse 14
Stop AED after remission: AED successful after relapse, but later on intractable  4
Stop AED after remission: AED effect unknown after relapse and later on intractable  1
Deteriorated without an obvious reason  7

In group B (with intractability both during the first 5 years and the final year), 16 of the 19 patients used AEDs during the entire follow-up and had no substantial periods of remission and 50% had daily to monthly seizures. All patients in group B were considered to be drug resistant (Table 2).

Table S2 describes the individual course of epilepsy for the 16 patients with late onset of intractability (group C). Seven had remissions in the first 5 years, ranging from 2.4 to 4.7 years. Nine patients even discontinued AEDs after remission during the first 5 years of follow-up, but all had recurrences and restarted medication (Table 2); in four of them reinstitution of treatment failed to regain remission. This was the only possible explanation we found for the late onset of intractability in group C. At the end of follow-up, 8 of the 16 patients of group C were drug-resistant for about 3–8 years.

To unravel why there was such a large variation in time of onset and duration of intractability, we focused on differences between these three groups that might explain this variation. The three groups did not differ in baseline characteristics (Table 1), indicating that the risk for patients to fall into any of the three groups was similar. However, during follow-up, patients of group B had a significantly longer time to intractability (Table 1, Fig. 3) and, consequently, more were intractable in the fifth year of follow-up and had a shorter terminal remission; they also had a slightly shorter longest remission during the first 5 years compared with those of group A. Comparison of group A with the two groups that were intractable in the last year of follow-up (groups B + C) showed a tendency for a higher proportion of patients in the latter two groups to develop mental retardation during follow-up: one of 15 (6.7%) versus 9 of 35 (25.7%) (95% CI of difference 2.9–35.2, p < 0.1). The significant differences found in the comparisons of group A and B with group C were all inherent to the poor results of groups A and B during the first 5 years.

Figure 3.

Time course and cumulative risk of intractability during the first 5 years of follow-up for the group with temporary intractability and for the group both intractable during the first 5 years and the last year of follow-up. Test of equality of survival distributions: log rank (Mantel-Cox) χ2 = 6.06, d.f. = 1, p = 0.014. Group A: only intractable during first 5 years; group B: intractable during both the first 5 years and the last year.


The present study was a pragmatic investigation. Our main interest was the time of onset of intractability, its course over time, and the variables that might influence it. The definition of intractability itself was not the issue; any other definition could have been used and probably would have yielded similar results.

Unfortunately, our results and the resulting conclusions are limited by the lack of data on the onset of intractability in the period after the fifth year of follow-up. Nevertheless, we think that the data provide valuable insight into the onset and course of intractability during the long-term follow-up of our cohort, even though intractability in some years could not be established. However, in our opinion, the latter is not a significant drawback. In fact, if all subjects with possible intractability in the intervening years had been intractable, our conclusions about a fluctuating course and temporary intractability would have been strengthened. The main outcome of this prospective study is that at least 12% of our cohort with childhood-onset epilepsy experienced a period of intractability during a 15-year follow-up since diagnosis, and 8.5% were intractable in the final year. The majority became intractable within 5 years after epilepsy onset, but approximately 50% of the patients were no longer intractable at the last contact; some even reached a 5-year terminal remission. Other patients had a late onset of intractability after 5 years of follow-up, and in some, intractability was interrupted by long periods of remission. This has also been reported by others (Takenaka et al., 2000; Sillanpää & Schmidt, 2006a). These findings show that time of onset, course, and duration of intractability vary considerably between subjects. According to the present study, these aspects of intractability are hardly predictable, even though intractability itself can be predicted to a certain extent. The risk factors for intractability found in the present study (younger age at onset, remote symptomatic etiology, type of epilepsy, and mental retardation) have also been reported by others (Huttenlocher & Hapke, 1990; Berg et al., 1996; Casetta et al., 1999; Kwong et al., 2003; Altunbasak et al., 2007). However, in the present study none of these risk factors were associated with a temporary, continuing, or late onset of intractability. In practice this means that a treating physician can only estimate the risk of intractability for an individual patient, but not the time of onset or duration of intractability, and the sustainability of remission following intractability. Especially for surgery candidates, it would have been valuable if these aspects of intractability could have been predicted.

Intractability as a temporary phenomenon

The finding that intractability can be followed by remission has also been reported by others (Berg et al., 2006, 2009; Callaghan et al., 2007, 2011; Choi et al., 2008, 2011), and some state that the natural history of intractable seizures in children follows a gradual line of slow improvement, especially in patients with normal intelligence (Huttenlocher & Hapke, 1990). In our opinion this might particularly apply to children with a period of intractability early in the course of their epilepsy.

In a prevalence cohort study of adults with intractable epilepsy no clear predictors for remission were found (Choi et al., 2008, 2011), whereas in a similar study, developmental delay, symptomatic generalized epilepsy, longer duration of intractability, and number of failed AEDs were negatively associated with remission, although in multivariate analysis only the number of failed AEDs remained significant (Callaghan et al., 2011). In a prospective childhood-onset study, idiopathic epilepsy and lower seizure frequency were positively associated with remission (Berg et al., 2009). In the present study no clear predictors were found for temporary intractability versus ongoing or repeating intractability. However, although the numbers were small, there were some significant differences between subjects with temporary intractability and those intractable again at the end, indicating that the epilepsy of the latter patients was probably more severe from the beginning and deteriorated further during follow-up.

It has been shown that remission after intractability started after medication changes in 28–76% of adult subjects, although the relapse rate was high (Luciano & Shorvon, 2007; Callaghan et al., 2011; Choi et al., 2011). In our cohort, we found a medication effect in 27% (4 of 15) of patients with temporary intractability, with 2 of the 4 subsequently entering terminal remission for at least 10 years. In our cohort, apart from the patients with a positive effect of AEDs or surgery, and the patient not considered intractable in the final year, we think that the improvement resulted from the natural course of the patients’ epilepsy.

Late-onset intractability

More than 30% of our intractable patients had a late onset of intractability, starting 5 years or more after diagnosis; this has also been reported by others (Berg et al., 2003; Berg, 2004; Sillanpää & Schmidt, 2006a). Three years or more after diagnosis, Berg et al. (2006) found a late-onset intractability in 32% of children meeting their stringent definition of intractability. This occurred even after long periods of remission, a result that was confirmed in the present study. In line with this, Sillanpää et al. found a worsening course in 14% of their cohort, defined as a relapse occurring after a 5-year remission without regaining a final 5-year terminal remission (Sillanpää & Schmidt, 2006a). So far, no clear explanation for the occurrence of late-onset intractability is available, although a poorer adherence to medication during follow-up might be one (Sabaté, 2003). In addition, AED failure after restart has been reported as a possible reason (Takenaka et al., 2000; Schmidt & Loscher, 2005; Sillanpää & Schmidt, 2006b). Two studies found that reinstitution of medication that worked for years failed to achieve control in one of three or four relapsed patients (Bouma et al., 2002; Sillanpää & Schmidt, 2006b). In our study this occurred in five patients. During ongoing treatment, mechanisms such as developing drug tolerance or resistance might lead to deterioration or intractability. However, it remains uncertain whether late-onset intractability or a worsening course results from these mechanisms, or from the progressive nature of some types of epilepsy (Takenaka et al., 2000; Abou-Khalil, 2007). In view of this, we doubt that late-onset intractability is a temporary phenomenon, like part of the early-onset intractability, but rather that it might be severe and persistent.

One study found evidence of a later onset of intractability in patients with focal epilepsy compared with those with catastrophic epilepsy (Berg et al., 2006). In our cohort, we found that slightly more patients with catastrophic epilepsy developed intractability during the first 5 years of follow-up (mostly during the first year) compared with those with focal epilepsy, but that there was no significant difference in the time to intractability. In addition, during the entire follow-up, the difference in intractability between these two groups was not significant (focal 18.2% vs. catastrophic 28.3%, Pearson’s χ2: 2.53, d.f. = 1, p = 0.112), unless they would have been temporarily intractable 5 years or more after enrollment. We conclude that our results do not convincingly support the findings of Berg et al., although the trend pointed in the same direction. In the study of Berg et al., more subjects with catastrophic epilepsy developed intractability during the first 5 years of follow-up than in our study (stringent definition: approximately 50.% vs. 21.7%, respectively), although for those with focal epilepsy the numbers were the same (approximately 10% vs. 11.7%). In our cohort, the difference between the catastrophic and focal group was only marginally significant.

Drug-resistant epilepsy according to the ILAE commission

Recently the ILAE proposed the following definition for drug-resistant epilepsy (Kwan et al., 2010): failure of adequate trials of two tolerated, appropriately chosen and used AED schedules to achieve sustained seizure freedom. In response to this, and to make our article as up-to-date as possible, we applied this definition to our cohort in addition to our definition at two points in time: at 5 years after enrollment and at last contact.

A comparison of both definitions shows that there is very little difference (Table 3). The overall conclusion is more or less the same; in both cases intractability or drug-resistance can be temporary, continuing, or delayed. However, the ILAE definition labels more patients as being drug-resistant. Certainly there is little doubt about treatment failure when a person still has seizures, however, we consider that in terms of severity of epilepsy, the ILAE definition may be too broad in labeling patients with only sporadic seizures as having intractable epilepsy. On the other hand, our definition may not label some highly refractory patients with brief remissions as intractable, when in fact they are. Having used the ILAE definition, we think that it is difficult to apply in a large cohort study, as the history of each individual patient has to be judged carefully, and the intraobserver and interobserver variability may be high. It should be stressed that it is not our intention to decide which definition is best or to promote our definition. We merely used our definition as a tool to study the course of intractability. In future cohort studies, we advocate the use of one worldwide accepted definition to make comparisons between studies easier. We consider the proposed definition of the ILAE as a good step forward.

Table 3.   Comparison of the proposed ILAE definition of drug resistance and our definition of intractability applied to 413 subjects
ILAE definition (Kwan et al., 2010)Intractability definition as used in the present studyIntractability definition applied to the fifth and last year of follow-up
Drug resistant at the end of the fifth year of follow-up36 (8.7%)A period of intractability during the first 5 years of follow-up34 (8.2%)Intractable during the fifth year of follow-up18 (4.4%)
Drug resistant at last contact48 (11.6%)Intractable during the last year of follow-up35 (8.5%)Intractable during the last year of follow-up35 (8.5%)
Only drug resistant at the end of the fifth year of follow-up12 (2.9%)Only a period of intractability during the first 5 years of follow-up15 (3.6%)Only intractable during the fifth year of follow-up4 (1.0%)
Drug resistant at the end of the fifth year of follow-up and at last contact24 (5.8%)A period of intractability during the first 5 years of follow-up and in the last year of follow-up19 (4.6%)Intractable during the fifth year and last year of follow-up14 (3.4%)
Only drug resistant at last contact24 (5.8%)Only intractable during the last year of follow-up16 (3.9%)Only intractable during the last year of follow-up16 (3.9%)


Considering the high proportion of patients with temporary intractability, one could argue that the time span used in our definition of intractability was too short, leading to the inclusion of too many subjects not fundamentally intractable. Table 4 shows our attempt to predict intractability in the final year of the 15-year follow-up with intractability in the first 5 years, while varying the period of continuous seizures. However, a longer period of continuous seizures in the first 5 years led to a decreased sensitivity of the prediction, whereas specificity remained almost the same. In this case, fewer subjects being intractable in the final year of follow-up would have been identified. Consequently, a longer period of continuous seizures as part of the definition of intractability does not solve the problem. Moreover, of the 14 patients with intractability lasting 4–5 years, two reached a 5-year and two a 1-year terminal remission. Therefore, determining how long a patient has to have frequent seizures to be labeled intractable remains arbitrary.

Table 4.   Intractability in the first 5 years as a predictor of intractability in the final year of follow-up depending on the length of the period of intractability during the first 5 years
Length of the period of intractability during the first 5 yearsIntractable in the first 5 yearsIntractable in the final yearSensitivity, %Specificity, %
One-year period382057.195.2
18-month period291542.996.3
Two-year period241337.197.1
Three-year period181131.498.1

Apart from the time span in our definition of intractability, using only two AEDs could be a matter of discussion. Other investigators use or recommend three AEDs (Camfield & Camfield, 1996; Kwong et al., 2003; Malik et al., 2008), although there is ample evidence that a third AED has little or no effect (Arts et al., 2004; Mohanraj & Brodie, 2006). Moreover, all patients who were intractable in the final year and 31 of the 34 patients with early onset had used at least three AEDs during follow-up. Considering this, we think that the number of AEDs in our definition is not an issue.

Furthermore, there might be discussion about subjects in our study having an intractable seizure pattern in the last year of follow-up while using only one AED. After a long period of intractability, it may be best for a patient to stop medication or to reduce polytherapy in case none of the therapies is really effective. Many physicians would probably have labeled these subjects as being intractable; however, in the present study we applied our definition strictly, and by doing so may have underreported the occurrence of intractability.

The course of intractability was studied during a period of impressive maturation of the child’s brain. The fact that many patients had temporary intractability might be the result of these processes. On the other hand, the phenomenon of late-onset intractability might result from a continuing process of epileptogenesis.

We studied the course of intractability using an objective approach and did not address the subjective severity of intractability, although we realize that this is also important, especially for patients just reaching adulthood.

In this study, incorrect diagnoses of intractability may have been made as a result of noncompliance, incorrect medication, or incorrect diagnosis or classification of epilepsy. We attempted to exclude these errors to the best of our ability. During the first 5 years, regular plasma concentrations of AEDs were taken, giving us an impression of compliance. We did not have this information for the patients in the last year of follow-up, unless it was indicated in their medical records. The Dutch Study of Epilepsy in Children was a pragmatic study, conducted with experienced physicians working according to their usual clinical practice.

To our knowledge, only a few cohort studies on childhood-onset epilepsy have aimed to identify possible reasons (other than clinical features) for the onset or ending of intractability, which makes the present study uncommon. Only a few studies reported that remission after intractability and relapse occurred in conjunction with changes in medication (Luciano & Shorvon, 2007; Callaghan et al., 2011; Choi et al., 2011). Although this is a delicate topic we think that the success or failure of treatment has to be considered at all times in future research investigating intractability.

The fact that we found no risk factors for temporary, continuing, or late-onset intractability might be due to the small numbers, even though our cohort is one of the largest with childhood-onset epilepsy. Only meta-analyses using multiple cohorts might be able to detect these risk factors.


At least 12% of patients with childhood-onset epilepsy will have a period of intractability during a 15-year follow-up. Onset, course, and duration of intractability can fluctuate to a great extent, and periods of remission and intractability can alternate with each other. Intractability can be temporary (especially in case of early onset), but can also begin many years after diagnosis, and in that case, is more likely to persist. These findings are in line with other investigations.

Apart from the risk factors for intractability, it is difficult to predict the time of onset of intractability and whether it will be temporary. Due to the relatively low frequency of intractability, meta-analyses are almost certainly needed for this.

This study did not find a clear explanation for the variation in intractability. The most probable explanation is the natural course of the epilepsy itself, as intractability developed in some despite an early and successful start of treatment, and in only 16% could medication be held responsible for the termination or late onset of intractability. However, the effect of medication may be different in patients who did not have a period of intractability. More research is needed on the actual role of AEDs on the process of epileptogenesis; the question remains whether AEDs modify this process or merely refine it.

Despite recent reports by the ILAE commission, a solid and all-encompassing definition of intractability, also taking into account the severity of sporadically occurring seizures, remains to be established.


This work was financially supported by the Dutch National Epilepsy Fund [grant number 05-06]. They had no role in the design and conduct of the study, data collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. The corresponding author had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.


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.