Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Ann Neurol. Author manuscript; available in PMC 2010 May 1.
Published in final edited form as:
PMCID: PMC2818062

Remission of epilepsy after 2 drug failures in children: a prospective study



Determine the probability of ≥1 year remission after failure of a second drug in children prospectively followed from initial diagnosis of epilepsy and then from time of second drug failure. Identify prognostic factors for remission after second drug failure.


Of 613 children, 128 failed 2 drugs, had a trial of at least a third drug (median=3) and were followed ≥1 year (median= 10.1 years) since second drug failure. Product-limit and proportional hazards techniques were used to analyze predictors of any 1 year remission (Rem1) and 1 and 3-year remission at last contact (Rem1/3-LC).


73 (57%) had a remission. Repeated remissions and relapses were common. Only 48 (37.5%) achieved Rem1-LC and 28 (23%) Rem3-LC. Idiopathic epilepsy (Rem1: rate ratio(RR)= 3.64, p<0.0001; Rem1-LC: RR=2.57, p=0.008) and seizure frequency (Rem1: RR= 0.76, p=0.003; Rem1-LC: RR=0.82, p=0.04 per increase in category) were the most robust predictors. Symptomatic cause was the only correlate of Rem3-LC. Remission before second drug failure did not predict remission after second drug failure.


Remission after second drug failure is common but often temporary.. Children who have failed two appropriate drugs should be carefully evaluated to maximize therapy and, possibly considered for more aggressive treatments.

Pharmacoresistance in epilepsy is a serious problem, yet it is often only studied years after it first occurs, frequently in the context of epilepsy surgery evaluations or randomized trials of new drugs. Even in most available cohort studies, intractability is only defined at an arbitrary point in time (e.g. at last contact or five years after diagnosis) 13.

Several studies have attempted to study the risk and predictors of pharmcoresistance; however, only two (including ours) have prospectively applied definitions of pharmacoresistance and identified pharmacoresistance as it occurred 4, 5. Early prospective identification of intractability, as soon as it can be recognized, is increasingly important in light of the growing interest in the earlier use of more invasive therapies, particularly surgery when medications fail to work, 69.

At the same time, early identification of intractability runs the risk of being overly sensitive and perhaps including individuals whose epilepsy may respond to another drug or resolve on its own. Consequently, before any early definition of intractability is employed, especially for purposes such as surgical intervention, the seizure course once patients meet specified criteria should be examined. Several studies, including our own, have documented brief, sometimes temporary remissions in patients who have met various criteria for intractable epilepsy 4, 5, 1014.

We have previously reported on the prediction 15 and time course 4 of intractability in this cohort. Here we focus on the long-term outcomes in those cohort members who met minimum criteria for pharmacoresistance, failure of adequate trials of two appropriate antiepileptic drugs (AED).


Data are from a prospective study of children (age 1month – 16 years) with newly diagnosed epilepsy who were recruited from the offices of 16 of the 17 practicing pediatric neurologists in the State of Connecticut from 1993–1997. Detailed methods have been previously published 4, 16.

This analysis focuses on those individuals whose seizures did not come under control (remission for ≥1 year) after two trials of AED therapy appropriate for the seizure and epilepsy type and pushed to maximum tolerated or recommended limits and taken as prescribed. AEDs discontinued because of side-effects or other concerns without adequate evidence to determine efficacy for seizure control were not considered as failures. Information about drug doses, levels, compliance, growth spurts, and seizures was obtained through ongoing review of medical records and telephone calls with parents. All of this information was taken into account when evaluating AED failure. In these analyses, subjects were further limited to those followed at least one year after the second drug failure and who also received an informative trial of at least one further AED. As a simplifying convention, a drug added to a failed drug regimen which was then associated with a period of remission ≥1 year was credited with the success. The failed drugs were often stopped, allowing a clearer test of the added drug’s effect; however, the treating neurologists had sole control over how the AEDs were used and many patients received multiple drugs simultaneously.

Specific epilepsy syndromes were identified at initial diagnosis according to published criteria 17, 18 and revised throughout follow-up as new information (e.g. a more informative EEG or MRI) became available, or as the syndrome evolved (e.g. West to Lennox-Gastaut). A simplified grouping of syndromes (based on syndromic diagnosis as determined nine years after initial study entry) was used for these analyses: traditional idiopathic syndromes17, epileptic encephalopathies/secondary generalized epilepsies, and other focal or undetermined epilepsies. If a child at any point in the course of his epilepsy expressed a syndrome considered to be an epileptic encephalopathy (e.g. West syndrome evolving to focal epilepsy) the syndrome was still grouped with the epileptic encephalopathies.

Data analysis

Chi-square and t-tests were used as appropriate for bivariate comparisons. Survival analysis with the product-limit methods and associated Kaplan-Meier curves was used for examining the occurrence of remission (≥1 year) ever and 1 and 3-year remission at last contact. The Cox proportional hazards model was employed for multivariable analyses 19.


Because one of the purposes in performing this analysis was to provide information relevant to surgical decisions,,individuals who had resective or disconnection surgical procedures were censored at the time of surgery. Their post-surgical outcomes are described.


All procedures were approved by the Institutional Reviews Boards of all institutions involved. Written informed assent of the child and written permission of the parent were obtained. Written informed consent of subjects when they attained the age of majority was obtained.


As of 8–2008, a total of 140 subjects had failed trials of at least 2 different AEDs considered appropriate for their seizures and type of epilepsy. Twelve were excluded from analysis for having less then 1 year of follow-up since second drug failure (N=2), no additional drug trials after second drug failure (N=8), or both (N=2). Descriptive information for the total group and those included versus excluded is provided in table 1.

Table 1
Characteristics of subjects who failed at least two AEDs overall and by inclusion or exclusion in the analyses.

The 128 included study subjects have been followed a median 11.8 years since initial diagnosis of epilepsy (maximum 15.4y, inter-quartile range (IQR) =11.0 to 13.1). Since second AED failure, they have been followed a median of 10.1 years (maximum 14 years, IQR=6.7 to 11.7). Ten subjects have died. Of the other 118, 106 (90%) are actively followed.

Because children were enrolled and initially treated in 1993–1997, the choice of first drugs was strongly weighted toward the drugs considered first-line at the time (table 2). The use of newer agents as second and third drugs reflects the increasing availability of newer drugs towards the end of the recruitment period and, to a certain extent, the tendency of some individuals to have a delayed expression of their intractability and not to be treated with these drugs until later 4, 5.

Table 2
First two drugs failed, most common combinations of first two AEDs failed, and third drugs failed for those failing another AED by grouped type of epilepsy.

Responses to additional drug trials and other treatments

All subjects included in these analyses had tried 1 to 15 additional different AEDs (median=3). For each of the additional drugs tried, only a small proportion became seizure free for at least one year and continued seizure-free at last contact or had periods of nonresponsiveness to the drug mixed in with at least one period seizure-free lasting ≥1 year (table 3).

Table 3
Results for additional drugs tried after failure of the second AED

Remission after second drug failure (figure 1)

Figure 1
Overall cumulative probability of any remission and of 1 and 3 year remission at last contact.

Seventy-three (57%) individuals experienced at least one period of remission ≥1 year. The cumulative probability (with 95% confidence intervals (CI)) of entering remission 2, 5, 10 and 12 years after second drug failure was 27.3% (19.4, 35.1%), 47.1% (38.1, 56.2%), 64.5% (54.8, 74.2%), and 68.6 (58.5, 78.9%) (figure 1). The median time to first remission was 2.3 years (range 1 to 11.3; inter-quartile range (IQR) = 1.4 to 4.5 years). Relapses occurred in 50 (68%) of those who had a 1-year remission. Repeated remissions and relapses were common; 37 individuals had a second remission period, of whom 19 relapsed. Of those 19, 15 remitted for a third time with 11 subsequently relapsing. Of those 11, four regained remission (1 after relapsing then remitting for a fifth time).

Forty-eight (38%) patients were in remission ≥1 year at last contact, 20 of whom were no longer taking AEDs. The proportion entering and remaining in 1-year remission at 2, 5, 10 and 12 years after a second drug failure was 8.8% (3.9, 13.8%), 18.6% (11.5, 25.6%), 40.1% (30.2, 49.9%), 47.3% (36.2, 58.3) (figure 1). Of those who failed 0, 1–2, 3–5, 6–9, and 9–15 trials of additional drugs, 71%, 54%, 21%, 24%, and 0% were in 1-yr remission at last contact. For those in 1-year remission at last contact, the median time to achieve that remission was 5.3 years (range 1.1 to 12.7, IQR=2.3 to 8.1).

Only 28 (22%) subjects were ≥3 years seizure-free at last contact. . The proportion entering and remaining in 3-year remission at 5, 10 and 12 years after a second drug failure was 10.0% (4.4, 15.6%), 24.5% (15.5, 33.5%), and 26.3% (16.8, 35.7%). Of those who failed 0, 1–2, 3–5, 6–9, and 9–15 trials of additional drugs, 48%, 38%, 9%, 6%, and 0% were in 3-yr remission at last contact

Thirteen patients who had epilepsy surgery were censored at the time of surgery. Five of them entered remission and were seizure-free at last contact. If their post-surgical time and remissions were included, the figures above would remain essentially unchanged. In addition, vagal nerve or brain stimulators were used in 14, and the ketogenic diet in 24. Observations were not censored for these therapies. Had they been, the results would not have been appreciably different.

Predictors of remission after a second drug failure

We selected potential predictive factors based on their associations with seizure outcomes in other studies and settings (Table 4). In a multivariable proportional hazards model, idiopathic syndromes and lower frequency of days with seizures remained as independent predictors of entering remission (Table 5). After adjustment for these factors, the RR for underlying symptomatic cause was 0.63 (p=0.08). Epileptic encephalopathies/other secondary generalized epilepsies and remission prior to second drug failure did not further contribute to predicting remission. For patients with focal epilepsy (N=70), symptomatic cause (p=0.04) and seizure frequency (p=0.03) were independently associated with remission. Remission prior to second drug failure (p=0.35, figure 2a) was not substantially associated with subsequent remission in this subset.

Figure 2
Cumulative probability of (a) any remission after second AED failure (b) 1-year remission as of last contact, and (c) 3-year remission at last contact as a function of remission prior to second AED failure.
Table 4
Bivariate association of any remission and remission at last contact with potential prognostic factors
Table 5
Results of multivariable proportional hazards models

For 1-year remission at last contact, similar associations were found. Idiopathic forms of epilepsy and prior seizure frequency were the only independent predictors (Table 5). Remission prior to second drug failure (figure 2b), symptomatic cause, and the epileptic encephalopathy/secondary generalized epilepsies were not associated with remission at last contact. Within the focal epilepsies, only prior seizure frequency attained marginal significance (p=0.02) on its own.

By contrast, only symptomatic cause was associated (negatively) with 3-year remission (RR=0.33, p=0.02). Exploratory analyses suggested a higher risk of relapse after remission associated with idiopathic versus other forms of epilepsy (57% vs. 40%, p=0.25). Prior seizure frequency was strongly correlated with relapse after remission, from lowest to highest seizure frequency category, 56%, 52%, 39%, 24% (p=0.005, trend).

Any remission >1 year prior to second drug failure was not predictive of later remission (figure 3c). Prior remissions lasted an average of 3.1 years (range 1.2 to 8.5). The duration of prior remission was not correlated with remission at last contact.


Our data present a long-term view of seizure outcomes after a second drug failure, a criterion for pharmacoresistance that has been adopted (with or without further specifications) in this and several other studies 1, 35, 12. More than half of patients subsequently experienced significant periods of seizure freedom. We found this was often not a lasting remission. Although it is common to refer to remission at last contact as “terminal,” in fact, we do not know what the future outcomes will be in those last known to be in remission. Those who were last in 1-year remission took many years to reach that point. Fewer were in a longer-term (≥3 year) remission.

Some of the newer drugs may explain some of the high remission rates that we saw. Schiller and Najjar clearly demonstrated in patients ≥12 years old that after each successive drug failure, a proportion of previously pharmacoresistant patients would become seizure-free with the introduction of the next drug 20. We too found a large proportion that attained remission, some after repeated drug failures. Whether remission was in response to a new drug or for reasons related to the natural history of the disease, or to other unidentified factors we do not know.

Although our evidence does not suggest that the newer drugs were dramatically more successful at controlling seizures than were the older drugs, these data must be viewed in context. The newer drugs were just being released when the children in our study were first being treated. This is also true of some the other studies cited above. By the time a child received some of these newer drugs, he had often failed multiple trials of the other drugs. Thus, due to secular trends, those who received newer drugs were somewhat selected for being particularly difficult to treat, and the number of previous drug failures tends to predict response to the next therapeutic agent tried 10, 11, 20.

The findings of brief remission after meeting criteria for pharmacoresistance has been reported by ourselves and others in prospective 4, 5 and retrospective 10, 11, 13, 14 studies. The Dutch investigators discussed the variability over time in seizure outcomes as well 3. While these brief remissions may encourage optimism, in fact, they are often not enduring. To the extent they are, they appear to take many years to achieve. Subjects with infrequent seizures were more likely to experience remission. They were also more likely to relapse after remission. Possibly, in these patients, AEDs were partially effective and reduced their break-through seizure rate to such a low point that they experienced 1-year seizure-free periods, thus appearing to be well-controlled when, in fact, they were not fully controlled. There are no standards or guidelines in the literature for handling this situation. Potentially, such patients would have to be seizure-free for many years before drug effectiveness could be reliably determined.

We also note that remission period before the appearance of pharmacoresistance has been reported previously 4, 25. Our findings suggest that remission before a second drug failure did not substantially improve the likelihood of attaining remission after a second drug failure. Thus, prior remission probably should not be a major consideration in management decisions after a second AED failure.

The recent studies of this issue 10, 11 were quite different from ours. We focused on an exclusively pediatric population in which the forms of epilepsy are quite diverse. Interestingly, we did not find that the most aggressive pediatric syndromes (epileptic encephalopathies) behaved differently than focal epilepsies once they became intractable. Instead, it appeared that the seizure frequency prior to second drug failure was a more reliable determinant of later 1-year remission. This was also associated with early appearance of intractability in this cohort 15. The finding of a relatively better seizure prognosis for those with traditional idiopathic syndromes is not surprising and others have reported similar results 10, 11, 20. These developmental syndromes are often self-limiting and the seizure propensity resolves on its own; however, some in our study have clearly not fully resolved at this point. Of these, all had idiopathic generalized syndromes, which occasionally may persist long after childhood or for life. The recent studies of Callaghan 11 and Luciano 10 were quite different from our own both in methods and purpose. These studies were both retrospective and sampled prevalent cases, most with epilepsy ≥10 years duration when they entered the observation periods reported in those articles. Prevalence sampling is known for over-sampling cases with the longest course of the disorder (milder for diseases that are rapidly fatal and severe for diseases that tend to remit, or are perhaps inoperable as with epilepsy). Ours by contrast involved incident cases of epilepsy prospectively identified and followed from initial diagnosis and, for this analysis, from date of second AED failure. This is the preferred method for assessing prognosis as the observation period begins at a defined moment, as soon as the second drug has failed. Our study directly addresses the question of prognosis after second AED failure, a point identified as a time when a comprehensive epilepsy evaluation should be obtained both for children and adults 6, 8.

Although a first remission is achieved fairly rapidly in some and the cumulative proportion experiencing remission is quite high, the proportion attaining a 3-year remission as of last contact is substantially lower. It takes several years to achieve this longer-term remission, during which time, the individual must live with all of the consequences of uncontrolled epilepsy including the consequences of the seizures, the side-effects and costs of medications, and the increased risk of mortality 21. Our finding that relatively few subjects rapidly attained seizure control emphasizes the recommendations of the ILAE’s subcommission for pediatric epilepsy surgery: children whose seizures are not controlled after reasonable trials of two or three agents should be evaluated at a comprehensive epilepsy center 8.

Referral for a comprehensive evaluation may not be enough though. In the last decade we have seen an explosion in the number of pharmacologic and other therapeutic options available for the treatment of epilepsy. Two recent surveys of experts in Europe 22 and the US 23, indicated that, although there was excellent consensus on the initial approach to treatment of epilepsy in children and very good agreement on the second therapeutic choice if the first failed, after that, there was increasing divergence regarding what constitutes appropriate third, fourth, fifth and subsequent therapies 24. Thus we have the therapies, but not a strong evidence base or consensus to guide their most effective and rational use.

Strengths of our study include that it is prospective, community-based, and with considerable follow-up. We also prospectively assessed AED failure and distinguished failure from discontinuation due to adverse events prior to achieving an informative trial and from other reasons a trial might not be adequate, distinctions not always clearly made by other studies. Limitations include a limited sample size for complex multivariable analyses and somewhat limited power to provide definitive identification or exclusion of prognostic factors. We encourage others to examine these issues in larger prospective studies and to extend prospective investigations into adult populations as well.

How best to define pharmacoresistance is still a matter for debate 2628. Increasingly, the failure of two appropriate drugs pushed to maximum limits is considered a basic threshold which should generally trigger specific actions, referral for comprehensive evaluation. Our data show that many patients who pass this threshold have periods of seizure-freedom, but substantial and perhaps lasting remission, to the extent we can measure it over the course of 10 years, is more elusive. The implications of this varying course are profound for defining drug refractoriness, understanding treatment response, recommendations for surgical evaluation, and design of drug trials 26.


This work was supported by a grant from the National Institutes of Health, National Institute of Neurologic Disorders and Stroke, R37-NS-31146.

We are very grateful to all the physicians in Connecticut who have made it possible for us to recruit and follow their patients all these years. We also would like to thank Eugene Shapiro who provided essential administrative help throughout and Shlomo Shinnar and Francis DiMario who participated in other phases of this study. This study was made possible by the generous help of the many families who have participated over the course of the last 15 years.


1. Dlugos D, Sammel M, Strom B, Farrar J. Response to first drug trial predicts outcome in childhood temporal lobe epilepsy. Neurology. 2001;57:2259–2264. [PubMed]
2. Camfield P, Camfield C. Nova Scotia pediatric epilepsy study. In: Jallon P, Berg A, Dulac O, Hauser A, editors. Prognosis of epilepsies. Montrouge France: John Libbey, Eurotext; 2003. pp. 113–126.
3. Arts WFM, Brouwer OF, Peters ACB, et al. Course and prognosis of childhood epilepsy: 5-year follow-up of the Dutch study of epilepsy in childhood. Brain. 2004;127:1774–1784. [PubMed]
4. Berg AT, Vickrey BG, Testa FM, et al. How long does it take epilepsy to become intractable? A prospective investigation. Ann Neurol. 2006;60:73–79. [PubMed]
5. Spooner CG, Berkovic SF, Mitchell LA, et al. New onset temporal lobe epilepsy in children: lesion on MRI predicts poor seizure outcome. Neurology. 2006;67:2147–2153. [PubMed]
6. Engel J. The goal of epilepsy therapy: no seizures, no side effects, as soon as possible. CNS Spectrums. 2004;9:95–97. [PubMed]
7. Mathern GW, Pretorius JK, TL B. Influence of the type of initial precipitating injury and at what age it occurs on course and outcome in patients with temporal lobe seizures. J Neurosurg. 1995;82:220–227. [PubMed]
8. Cross JH, Jaykar P, Nordli D, et al. Propose criteria for referral and evaluation of children for epilepsy surgery: Recommendations of the Subcomission for Pediatric Epilepsy Surgery. Epilepsia. 2006;47:953–959. [PubMed]
9. Weiner HL, Carlson C, Ridgway EB, et al. Epilepsy Surgery in Young Children With Tuberous Sclerosis: Results of a Novel Approach. Pediatrics. 2006;117:1494–1502. [PubMed]
10. Luciano AL, Shorvon SD. Results of treatment changes in patients with apparently drug-resistant chronic epilepsy. Ann Neurol. 2007;62:375–381. [PubMed]
11. Callaghan BC, Anand K, Hesdorffer D, et al. Likelihood of seizure remission in an adult population with refractory epilepsy. Ann Neurol. 2007;62:382–389. [PubMed]
12. Huttenlocher PR, Hapke RJ. A follow-up study of intractable seizures in childhood. Ann Neurol. 1990;28:699–705. [PubMed]
13. Takenaka J, Kosaburo, Watanabe K, et al. Transient remission in intractable localization-related epilepsy. Pediatr Neurol. 2000;23:328–331. [PubMed]
14. Berg AT, Levy SR, Novotny EJ, Shinnar S. Predictors of intractable epilepsy in childhood: a case-control study. Epilepsia. 1996;37:24–30. [PubMed]
15. Berg AT, Shinnar S, Levy SR, et al. Early development of intractable epilepsy in children: a prospective study. Neurology. 2001;56:1445–1452. [PubMed]
16. Berg AT, Shinnar S, Levy SR, Testa FM. Newly diagnosed epilepsy in children: presentation at diagnosis. Epilepsia. 1999;40:445–452. [PubMed]
17. Commission on Classification and Terminology of the International League Against Epilepsy. Proposal for revised classification of epilepsies and epileptic syndromes. Epilepsia. 1989;30:389–399. [PubMed]
18. Roger J, Bureau M, Dravet C, et al. Epileptic syndromes in infancy, childhood, and adolescence. 2nd ed. London: John Libbey; 1992.
19. SAS. Cary, NC: SAS Institute Inc.; 1990.
20. Schiller Y, Najjar Y. Quantifying the response to antiepileptic drugs: effect of past treatment history. Neurology. 2008;70:54–65. [PubMed]
21. Sperling MR. The consequences of uncontrolled epilepsy. CNS Spectrums. 2004;9:106–109. [PubMed]
22. Wheless JW, Clarke DF, Arzimanoglou A, Carpenter D. Treatment of pediatric epilepsy: European expert opinion, 2007. Epileptic Disord. 2007;9:353–412. [PubMed]
23. Wheless JW, Clarke DF, Carpenter D. Treatment of pediatric epilepsy: expert opinion, 2005. J Child Neurol. 2005;20:S1–S56. [PubMed]
24. Berg AT. Our work is still cut out for us … Epileptic Disord. 2008;10 [PubMed]
25. Berg AT, Langfitt J, Shinnar S, et al. How long does it take for partial epilepsy to become intractable? Neurology. 2003;60:186–190. [PubMed]
26. French JA. Refractory epilepsy: one size does not fit all. Epilepsy Curr. 2006;6:177–180. [PMC free article] [PubMed]
27. Arzimanoglou A, Ryvlin P, et al. Toward a clinically meaningful definitoin of drug resistance. In: Kahane P, Berg A, Loscher W, et al., editors. Drug resistant epilepsies. Montrouge, France: John Libbey Eurotext; 2008. pp. 1–6.
28. Berg AT, et al. The risks, correlates, and temporal patterns of intractable epilepsy. In: Kahane P, Berg A, Loscher W, et al., editors. Drug resistant epilepsies. Montrouge, France: John Libbey Eurotext; 2008. pp. 7–16.