Presented in preliminary form at the annual meeting of the American Epilepsy Society, December 2000, Los Angeles, California, U.S.A.
Two-Year Remission and Subsequent Relapse in Children with Newly Diagnosed Epilepsy
Article first published online: 11 JAN 2002
Volume 42, Issue 12, pages 1553–1562, December 2001
How to Cite
Berg, A. T., Shinnar, S., Levy, S. R., Testa, F. M., Smith-Rapaport, S., Beckerman, B. and Ebrahimi, N. (2001), Two-Year Remission and Subsequent Relapse in Children with Newly Diagnosed Epilepsy. Epilepsia, 42: 1553–1562. doi: 10.1046/j.1528-1157.2001.21101.x
- Issue published online: 11 JAN 2002
- Article first published online: 11 JAN 2002
- Revision accepted October 8, 2001.
- Seizure control;
Summary: Purpose: Although remission is the ultimate measure of seizure control in epilepsy, and epilepsy syndrome should largely determine this outcome, little is known about the relative importance of syndrome versus other factors traditionally examined as predictors of remission or of relapse after remission. The purpose of this study was to examine remission and relapse with respect to the epilepsy syndrome and other factors traditionally considered with respect to seizure outcome.
Methods: A prospectively identified cohort of 613 children with newly diagnosed epilepsy was assembled and is actively being followed to determine seizure outcomes. Epilepsy syndrome and etiology were classified at diagnosis and again 2 years later. Remission was defined as 2 years completely seizure-free, and relapse as the recurrence of seizures after remission. Multivariable analysis was performed with the Cox proportional hazards model.
Results: Five hundred ninety-four of the original 613 children were followed ≥2 years (median follow-up, 5 years). Remission occurred in 442 (74%), of whom 107 (24%) relapsed. On multivariable analysis, idiopathic generalized syndromes and age at onset between 5 and 9 years were associated with a substantially increased remission rate, whereas remote symptomatic etiology, family history of epilepsy, seizure frequency, and slowing on the initial EEG were associated with a decreased likelihood of attaining remission. Young onset age (<1 year) and seizure type were not important after adjustment for these predictors. Relapses occurred more often in association with focal slowing on the initial EEG and with juvenile myoclonic epilepsy. Benign rolandic epilepsy and age at onset <1 year were associated with markedly lower risks of relapse. About one fourth of relapses were apparently spontaneous while the child was taking medication with good compliance, and more than half occurred in children who were tapering or had fully stopped medication.
Conclusions: A large proportion of children with epilepsy remit. Symptomatic etiology, family history, EEG slowing, and initial seizure frequency negatively influence, and age 5–9 years and idiopathic generalized epilepsy positively influence the probability of entering remission. Factors that most influence relapse tend to be different from those that influence remission.
Remission is the ultimate measure of seizure control in epilepsy. In theory, the epilepsy syndrome should largely determine outcome, especially in pediatric-onset epilepsy in which the diversity in the types of syndromes is much greater than in adult-onset epilepsy. Several studies have examined remission in individuals with epilepsy (1–9). Some of these studies were conducted before the current International League Against Epilepsy (ILAE) classification of epileptic syndromes and thus relied on factors that are components of syndromes: etiology, age at onset, EEG characteristics, and seizure type. It is unclear, however, whether and how knowledge of the syndrome explains what is currently understood about the factors traditionally studied. In addition, remission does not always endure (1,8,9), and there is relatively little information about which patients relapse after remission and why.
The goal of this report is to identify the determinants of remission and relapse in children with newly diagnosed pediatric epilepsy. We emphasize the ILAE classification of the epilepsies.
Data are from an ongoing prospective community-based study. Children with newly diagnosed epilepsy were recruited from 16 of the 17 child neurologists practicing in Connecticut during 1993 through 1997 as well as from some selected pediatricians and adult neurologists who indicated that they occasionally cared for children with epilepsy without referral to a child neurologist. Methods for active surveillance were put in place in each practice to identify all eligible children. Our study is not strictly population based; however, we recruited from academic centers, private practices, and community clinics in an effort to assemble a cohort representative of children with epilepsy within the state. To be eligible, a child had to have experienced the first unprovoked seizure between ages 1 month and 15 years, have at least two unprovoked seizures on separate days, and be newly diagnosed during the recruitment period by a participating physician. We obtained written parental consent and, as appropriate, written or oral assent of the child. During follow-up, we reenrolled and obtained informed consent from study participants themselves when they reached the age of majority. The Institutional Review Boards of all participating institutions approved all procedures. Full details of the methods have been previously presented (10,11).
Parents were interviewed in detail when their children first entered the study. Medical records were reviewed to determine etiology, syndrome, and seizure types (12). The criteria of the International League Against Epilepsy (ILAE) were used (13–16). Etiology (15) is partially determined by syndrome. Idiopathic is used in reference to those epilepsies that meet the criteria for one of the localization-related or generalized idiopathic syndromes. “Remote symptomatic” refers to epilepsy that occurs in an individual who has a condition associated with an increased risk of epilepsy (e.g., cerebral palsy, history of bacterial meningitis, tuberous sclerosis). One can have a condition associated with an increased risk of epilepsy but clinically still have an idiopathic syndrome. In these cases (n = 2 at diagnosis), we classified the syndrome as idiopathic and the etiology as remote symptomatic. “Cryptogenic” refers to epilepsy that has no identified remote symptomatic cause and does not meet the criteria for an idiopathic syndrome.
Our protocol required reevaluating syndromes, seizure types, and etiology after 2 years of follow-up. Because of subsequent information accumulated during the first 2 years after initial diagnosis, several children's syndromes, seizures, and etiology were assessed as being different from their initial classification either because a syndrome evolved or subsequent information permitted a different classification (17).
For predicting remission, we used the classification of syndrome, etiology, and seizures at the time of initial diagnosis. By the time children reached a 2-year remission, our assessment of these factors had changed. Thus we considered risk of relapse as a function of our best understanding of the syndrome, seizure types, and etiology once the child had achieved a 2-year remission. Other prognostic factors that we considered have been defined previously (11).
Study personnel telephoned parents every 3 months to ascertain additional seizures that children may have had since the last contact and to find out about changes in medication. Medical records were reviewed every 6 months. Discrepancies between parental report and what was recorded in the medical record were resolved by speaking to the parent, the physician, or both.
There is no single, accepted duration for defining remission, although the ILAE guidelines for epidemiologic studies refer to 5 years (15). Shorter periods of remission represent significant milestones with decisions about stopping medications, driving, and other restrictions being made after much shorter seizure-free periods (18–20). We chose 2 years because that is the typical period of time that many physicians consider necessary for medication discontinuation (21). We note that this definition was made independent of our definition of intractable epilepsy (two or more drugs failures, and one or more seizures/month for ≥18 months). Children could enter remission and later become intractable or become intractable and later remit. There also were children who did not meet the criteria for intractability and who never entered a 2-year remission (22).
A relapse was defined as the occurrence of any seizures after a 2-year remission had been achieved.
Bivariate analyses were performed with t tests and χ2 tests as appropriate. Multivariable analyses were performed with the Cox model (23). The model yields a rate ratio (RR) that quantifies the direction and magnitude of an effect of a variable after adjustment for related variables. In the analysis of relapse, stopping medications was treated as a time-dependent covariate, and its effect counted only once tapering of medications was initiated (24). In addition, we allowed the effect of tapering medication to change over time, as evidence from other sources indicates that the effect should be greatest in the year immediately after tapering (25). This approach allows, for example, an individual's rate to be increased for a period after tapering medications and then return to its previous level. The analyses focused on factors that have been examined as predictors of remission in previous studies, syndromes, and also factors that have been discussed as being potentially important to the outcome of epilepsy. These include status epilepticus, number of seizures, family history, febrile seizures, and seizure frequency.
In all, 613 children were recruited into the study. Half the children were boys, and the median age at onset was 5.3 years. A minimum of 2 years of follow-up was obtained on 594 (96.9%). This is one more than previously reported (17), as one child who was lost for a period was subsequently recontacted. The median follow-up as of April 2001 is 5.3 years (range, 2–8 years). About 80% of the children were treated, starting at the time of initial diagnosis. After a year, 90% were treated. Use of treatment varied substantially by epilepsy syndrome (26).
To date, 442 (74%) children have achieved a 2-year remission. Median time to achieve remission was 2.3 years, with a range of 2 (the minimum possible) to 6 years. The probability of a 2-year remission (with 95% confidence intervals) at 24, 30, 36, 48, 60, 72, and 84 months after diagnosis was 7% (5–9%), 40% (36–44%), 50% (46–54%), 66% (62–70%), 73% (69–77%), 81% (77–85%), and 84% (80–88%) (Fig. 1).
Predictors of attaining remission
The idiopathic syndrome groups, both partial and generalized, experienced the highest remission rates, and the symptomatic and cryptogenic generalized (including West and Lennox–Gastaut syndrome), the lowest (Table 1). Although the idiopathic generalized syndromes overall had a high remission rate (>80%), only 58% of those with juvenile myoclonic epilepsy remitted. Several other factors were associated with the probability of entering remission (Table 2). Remission was highest in those with onset at ages 5–9 years, and lowest in those younger than 1 year. Children with idiopathic etiology had the highest remission rates, and those with remote symptomatic, the lowest. Absence seizures also carried a relatively good prognosis. Factors significantly associated with a decreased probability of remission were slowing on the initial EEG, family history of epilepsy, myoclonic seizures, status epilepticus, and initial seizure frequency. Neither prior febrile seizures before the onset of epilepsy nor the number of seizures before diagnosis was significantly associated with attaining remission. Treatment was initiated at the time of initial diagnosis in 340 (77%) of 442 who eventually attained remission and in 126 (83%) of 152 who did not attain remission (p = 0.12). The decision to treat was not randomized.
|Syndrome groupinga||No.||No. (%) ever in 2-year remission|
|Partial epilepsies||345||269 (78)|
|Idiopathic partialb||57||52 (91)|
|Benign rolandic||55||50 (91)|
|Benign occipital||2||2 (100)|
|Symptomatic partial||189||135 (71)|
|Cryptogenic partial||99||82 (83)|
|Generalized epilepsies||176||128 (73)|
|Idiopathic generalized (IGE)||124||106 (85)|
|Benign myoclonic epilepsy in infancy||1||1 (100)|
|Childhood absence||73||65 (89)|
|Juvenile absence||15||13 (87)|
|Juvenile myoclonic||12||7 (58)|
|With GTCS on awakening||2||1 (50)|
|Other IGE/unclassified IGE||18||16 (89)|
|With specific modes of precipitation||3||3 (100)|
|Symptomatic/cryptogenic generalized||43||20 (47)|
|Not further classified||3||1 (33)|
|With myoclonic absence||2||2 (100)|
|Symptomatic generalized||9||2 (22)|
|Nonspecific etiology||4||1 (25)|
|With specific etiology||5||1 (20)|
|Epilepsies of undetermined onset||73||45 (62)|
|Undetermined with both focal and generalized features||5||3 (60)|
|Undetermined without unequivocal focal or generalized features||68||42 (62)|
|Prognostic factor||No.||No. (%) ever in 2-yr remission||p Value|
|Remote symptomatic||108||56 (52)|
|Both partial and gen'd||19||13 (68)|
|Specific seizure typesb|
|Typical absence||95||81 (85)||0.008|
|Atypical absence||15||8 (53)||0.08|
|Generalized onset tonic–clonic||129||98 (76)||0.65|
|Complex partial||254||191 (75)||0.70|
|Simple partial||74||58 (78)||0.40|
|Secondarily generalized||173||132 (76)||0.50|
|None before diagnosis||541||409 (76)||0.04|
|1+ episodes||53||33 (62)|
|Febrile seizure history (missing 4)|
|Family history of epilepsy (first degree only) (missing 7)|
|EEG (5 missing/results unknown)|
|Epileptiform abnormality present||417||311 (75)||0.92|
|Focal spikes||294||215 (73)||0.66|
|Generalized spike and wave||163||131 (80)||0.24|
|Any slowing present||106||66 (62)||0.001|
|Focal slowing only||58||37 (64)||0.03|
|Initial seizure frequency|
|Number of seizures before diagnosis|
On multivariable analysis (Table 3), idiopathic generalized syndromes as a group were associated with a higher likelihood of attaining 2-year remission. Age at onset between 5 and 9 years also was marginally associated with a better chance of remission (p = 0.04). Log of initial seizure frequency (higher seizure frequency), remote symptomatic etiology, slowing on the initial EEG, and a family history of epilepsy were all associated with a decreased likelihood of 2-year remission. In the multivariable model, the idiopathic partial epilepsies had an associated RR of 1.0 [95% confidence interval (CI), 0.72–1.40]. We explored why this was the case by entering individual terms into the model and examining their effect on the estimate for idiopathic partial epilepsy. It appeared that the factors characteristic of idiopathic partial epilepsy, especially low seizure frequency and age at onset between 5 and 9 years, are important predictors of achieving remission regardless of the specific syndrome.
|Factor||Rate ratio||95% CI||p Value|
|Associated with a decreased chance of remission|
|Log initial seizure frequency||0.90a||0.86–0.95||0.0001|
|Family history of epilepsy||0.58||0.42–0.81||0.001|
|Remote symptomatic etiology||0.63||0.47–0.84||0.002|
|Any slowing on initial EEG||0.71||0.54–0.93||0.01|
|Associated with an increased chance of remission|
|Idiopathic generalized epilepsy||1.65||1.28–2.12||0.0001|
|Age at onset, 5–9 yr||1.23||1.01–1.50||0.04|
After adjustment for statistically significant predictors of remission, the residual associations for symptomatic generalized epilepsies as a group (RR = 0.50; CI, 0.12–2.06), West syndrome (RR = 0.65; CI, 0.31–1.36), Lennox–Gastaut syndrome (RR = 1.23; CI, 0.17–9.03), and juvenile myoclonic epilepsy (RR = 0.58; CI, 0.27–1.27) were such that we cannot preclude that some may have important influences on remission. These groups are too small for their effects to be definitively appreciated in this context.
Sixty-eight children in our cohort met the study criteria for intractable epilepsy (11). Of those who met the criteria and who have been followed up ≥2 additional years, only four (7.1%) of 56 have gone on to experience a 2-year remission.
After achieving a 2-year remission, 107 (24.2%) children experienced a relapse. The risk of relapse (and 95% confidence interval) at 6, 12, 24, 36, and 72 months after attaining two-year remission was 11% (8–14%), 16% (13–20%), 22% (18–26%), 26% (21–30%), and 31% (25–37%). Relapses occurred at a median of 6 months after remission (range, 0–5 years; Fig. 2). Risk of relapse varied by syndrome (Table 4). Idiopathic partial epilepsy had a very low and juvenile myoclonic epilepsy a very high risk. Although some syndromes or groups of syndromes had a lower likelihood of achieving remission, once they achieved remission, their risk of relapse was about average. Age at onset younger than 1 year was associated with a relatively lower risk of subsequent relapse. Status epilepticus, symptomatic etiology, and focal slowing on the initial EEG were all associated with an increased risk of relapse. Seizure type did not appear to be related to relapse (Table 5). In addition, there was no difference between those who did and did not relapse with respect to the number of seizures before diagnosis, the number of seizures between diagnosis and the time of attaining remission, or the time to attain remission (data not shown).
|Syndrome grouping*||No. total after 2-year reassessment of syndromes and etiology||No. (%) achieving 2-year remission||No. (%) relapse|
|Total||594||441 (74)||107 (24)|
|Partial epilepsies||354||274 (77)||65 (24)|
|Idiopathic partialb||62||55 (93)||4 (7)|
|Benign rolandic||59||52 (88)||3 (6)|
|Benign occipital||3||3 (100)||1 (33)|
|Symptomatic partial||205||147 (72)||44 (30)|
|Cryptogenic partial||87||72 (83)||17 (24)|
|Generalized epilepsies||191||132 (69)||32 (24)|
|Idiopathic generalized (IGE)||133||111 (83)||28 (25)|
|Benign myoclonic epilepsy in infancy||2||2 (100)||0 (0)|
|Childhood absence||75||66 (88)||15 (23)|
|Juvenile absence||17||14 (82)||3 (21)|
|Juvenile myoclonic||15||8 (53)||5 (63)|
|With GTCS on awakening||2||1 (50)||1 (100)|
|Other IGE/unclassified IGE||18||17 (94)||3 (19)|
|With specific modes of precipitation||4||3 (75)||1 (33)|
|Symptomatic/cryptogenic generalized||49||20 (41)||4 (20)|
|Not further classified||3||1 (33)||—|
|West||17||9 (53)||1 (11)|
|Lennox–Gastaut||19||1 (5)||1 (100)|
|Doose||8||7 (88)||1 (14)|
|With myoclonic absence||2||2 (100)||1 (50)|
|Symptomatic generalized||9||1 (11)||0 (0)|
|With nonspecific etiology||3||0 (0)||—|
|With specific etiology||6||1 (17)||0 (0)|
|Epilepsies of undetermined onset||49||36 (73)||10 (28)|
|Undetermined with both focal and generalized features||2||1 (50)||1 (100)|
|Undetermined without unequivocal focal or generalized features||47||35 (74)||9 (26)|
|No. who attained remission||Relapse no. (%)||p Value|
|Age at onset (yr)|
|Remote symptomatic||59||23 (39)|
|Both partial and gen'd||18||2 (11)|
|Specific seizure typesb|
|Typical absence||84||18 (21)||0.50|
|Atypical absence||9||2 (22)||0.88|
|Generalized onset tonic–clonic||100||22 (22)||0.55|
|Complex partial||195||54 (28)||0.14|
|Simple partial||62||12 (19)||0.34|
|Secondarily generalized||126||41 (30)||0.06|
|None before 2-yr remission||399||91 (23)||0.04|
|1+ episodes||43||16 (37)|
|Febrile seizure history|
|Family history of epilepsy (first degree only)|
|Initial EEG (2 missing)b|
|Focal spikes||210||48 (23)||0.84|
|Generalized spike and wave||131||35 (27)||0.62|
|Any present||66||21 (32)||0.11|
|Focal only||37||13 (35)||0.09|
|Initial seizure frequency|
|Medication in relation to 2-year remission and subsequent relapsed|
|Never treated||46||3 (7)||<0.001|
|Treatment tapered before 2-yr remission achieved||181||42 (23)|
|Treatment tapered after achieving 2-yr remission||88||42 (48)|
|Treated with no taper after achieving 2-yr remission||127||20 (16)|
On multivariable analysis (Table 6), juvenile myoclonic epilepsy and focal slowing on the initial EEG were associated with an increased risk of relapse, whereas benign rolandic epilepsy, age at onset younger than 1 year, and log of initial seizure frequency were associated with a decreased rate of relapse. In addition, children who achieved a 2-year remission and had never taken medication were very unlikely to relapse. After statistical adjustment for these factors, symptomatic etiology had an RR of 1.53 (CI, 0.94–2.48; p = 0.09).
|Factor||Rate ratio||95% CI||p Value|
|Associated with decreased risk of relapse|
|Log initial seizure frequency||0.86||0.81–0.97||0.01|
|Idiopathic partial epilepsy||0.25||0.09–0.71||0.009|
|Onset <1 yr old||0.25||0.08–0.80||0.02|
|Never taken AEDs||0.24||0.07–0.78||0.02|
|Associated with an increased risk of relapse|
|Juvenile myoclonic epilepsy||2.67||1.09–6.66||0.03|
|Focal slowing on initial EEG||2.13||1.19–3.87||0.01|
There was a nonstatistically significant increased risk of relapse during the year immediately after a child began to taper medications.
Reasons for relapse
More than half of the relapses occurred in children who were tapering or had completely tapered their medications or who had never taken medication. Relapses associated with documented noncompliance or in the context of an illness were few. Approximately one fourth of the relapses occurred apparently spontaneously while the child was taking medication with good compliance (as best determined; Table 7). There were no trends to suggest that certain syndromes were more prone to spontaneous relapses than were others.
|Possible explanation for relapse||No. (%)|
|Taking AEDs, apparently unexplained breakthrough||31 (29)|
|Never treated with AEDs, apparently unexplained breakthrough||3 (3)|
|Noncompliance documented by parent or physician||5 (5)|
|Tapering off medicationsa||57 (53)|
|While tapering off AED||7 (6)|
|Completely tapered AED||50 (47)|
|Other (includes combinations of above)||7 (6)|
Outcome after relapse
Of the 107 children who relapsed, 50 have been followed for at least an additional 2 years since their relapse. Of those, 29 (58%) have attained another 2-year remission, although six have since gone on to relapse yet again. Two children met the study criteria for intractability after relapsing from a 2-year remission.
About three fourths of children with newly diagnosed epilepsy achieve a substantial period of remission soon after initial diagnosis. The single most important factor in previous studies of remission has been symptomatic versus nonsymptomatic etiology (1,4,6–9). Our results tend to support this distinction for remission. One study reported that the long-term (30+ years) outcome for idiopathic etiology was excellent in comparison to cryptogenic etiology, which was still substantially better than symptomatic epilepsy (9). We found this in part for the idiopathic generalized epilepsies but not for the idiopathic partial syndromes. Data from case series (27,28), however, demonstrate that benign rolandic epilepsy has a nearly 100% remission rate by midadolescence. It will require several years before all of the children in our cohort have reached that age.
It is not always possible to determine whether other specific syndromes provided additional prognostic information for either remission or relapse. Several syndromes are relatively rare. Juvenile myoclonic epilepsy and West syndrome, as well as the group of symptomatic generalized syndromes, were all associated with a lower probability of remission even after adjustment for all of the statistically significant factors. They simply did not achieve statistical significance in this study. In the case of the nonidiopathic generalized syndromes, much of the effect with respect to remission was explained by remote symptomatic etiology. Of this group, 20% with symptomatic causes and 65% with cryptogenic etiology remitted.
In theory, the epilepsy syndrome should provide reliable information about long-term outcome that can guide treatment, management, and other decisions. One concern is that the diagnosis of syndromes is sometimes made only after the outcome is known (28). Thus the syndrome may be determined by its eventual outcome and not the outcome predicted by the early knowledge of the syndrome. The approach we have taken of prospectively classifying the syndrome at initial diagnosis can provide more accurate information on the prognostic significance of prospectively diagnosed syndromes. A full appreciation of the prognostic value of the syndromic classification will require prolonged follow-up.
Age at onset is especially important in studies that have included a broad age range from childhood through adulthood (1,2,5,7). These studies often find childhood onset to be associated with a better prognosis. Although our study cannot address differential prognosis of childhood- versus adult-onset epilepsy, these earlier findings may be due to the large number of children with idiopathic syndromes, which rarely if ever have their onset during adulthood, and the relatively greater proportion of adults with remote symptomatic causes.
Within the pediatric age range, very young age at onset is sometimes found to be associated with a poor prognosis (4,29,30). The distribution of specific syndromes and etiologies is very different by age. Specifically, the cryptogenic and symptomatic generalized syndromes occur largely during the first year of life, and there is very little idiopathic epilepsy in this age group. There also is more remote symptomatic epilepsy in the very young onset group. Once the effects of etiology are factored out, most of the differences disappear, and very young age at onset has about the same prognosis for remission as do older ages, and an extremely good prognosis once remission is attained.
We noted previously that the syndrome assessed at diagnosis is not always the same as that assessed 2 years later, once additional information has accumulated (17). Because the question of relapse after 2-year remission would not arise until ≥2 years after diagnosis, by which time the additional information would have modified the initial diagnosis, we chose to focus on the reclassified syndromes (and etiology) at 2 years in our analysis of predictors of relapse. Most of the changes that occurred involved partially classified localization-related or unclassified syndrome. The results of the same analysis using the initial classification of syndromes were highly comparable to what was obtained with the 2-year reclassification.
Despite scattered and somewhat inconsistent reports about the possible importance of seizure type in understanding remission (1,2,5,7,9), our analysis suggests that most if not all of the association between seizure type and outcome can be explained by syndrome and etiology.
Likewise, specific EEG abnormalities (e.g., 3-Hz generalized spike and wave, centrotemporal spikes consistent with benign rolandic epilepsy, or hypsarrhythmia) are largely synonymous with specific syndromes. Conversely, slowing generally raises concern about underlying brain dysfunction. Slowing on the EEG has not been explored in remission studies, although it has been examined with respect to other outcomes (11,31,32) and found to be associated with a poor prognosis. The collected findings to date suggest that slowing on the EEG is a marker for poor outcome.
Family history influences the development of epilepsy (33–36). Its association with outcome once epilepsy has developed is less clear. The idiopathic epilepsies have a presumed genetic basis. Interestingly, exploratory analyses suggested that the effect of family history was strongest in the generalized idiopathic syndromes. The significance of such findings will be better addressed by the ongoing and future efforts in molecular genetics under way by several groups worldwide.
Our findings for treatment must be interpreted cautiously, as none of the treatment factors was randomized. Many children are not immediately treated on initial diagnosis (26,37). There are substantial differences between those who are and are not treated. Those selected for no treatment are most likely those perceived to have the best chances for immediate remission or, in the case of idiopathic partial epilepsy, for ultimate remission regardless of short-term outcomes. Consequently, it is not surprising that children who achieved remission without treatment tended to stay in remission.
Those who tapered medications after a 2-year remission had an excellent outcome, better than those who continued medication; however, many intervening clinical variables that we were not able to include in our statistical model may have been involved in the decision to stop medications. Data from a large randomized trial (25) demonstrated that tapering AEDs after ≥2 seizure-free years substantially increased the risk of relapse for 1–2 years after tapering, but that afterward, the risk paralleled that in the group randomized to continue treatment. This study also demonstrated that patients randomized to continue medication still relapsed but at a lower rate. Our nonrandomized results did suggest that, once adjustments were made for other important prognostic factors, there was a period of increased risk of relapse immediately after tapering of medications. Ultimately, those who continued instead of stopped medications most likely had very different underlying risks of relapse.
Many children who relapsed in this study reattained remission, although some relapsed again. Some of this was in conjunction with stopping and restarting medications. About a fourth of relapses had no apparent explanation. These findings underscore the undulating course of pediatric epilepsy during the first several years. We suspect that there are some children whose first several years of epilepsy will be marked by widely spaced, very occasional seizures regardless of treatment and whose outcome will be appreciated only in the very long term.
Acknowledgment: This study was funded by a grant from the National Institutes of Health, NINDS RO1 NS31146.
We are especially grateful to the physicians in Connecticut who referred their patients to this study, Drs. Robert Cerciello, Francis Dimario, Barry Russman, Michelle Kleiman, Carol Leicher, Edwin Zalneraitis, Philip Brunquell, Laura Ment, Edward Novotny, Bennet Shaywitz, S. Nallainathan, Alok Bhargava, Martin Kreminitzer, Barbara Coughlin, Harriet Fellows, Jack Finkelstein, Daniel Moalli, Louise Resor, Owen Erlich, Bernard Giserman, John Monroe, Lawrence Rifkin, and Murray Engel. We also thank Drs. Edward Novotny and Francis DiMario for reinterpreting selected EEGs for the study. Dr. Eugene Shapiro has kindly facilitated many administrative issues for us. We also thank the research associates, Lynnette Bates, Joann Gehrels, and Kris Engel, for their dedicated work on this project and Wuthikrai Uayingsak for his exceptional programming expertise. This study would not be possible without the generous participation of the many parents and their children.
- 16Roger J, Bureau M, Dravet C, Dreifuss FE, Perret A, Wolf P. Epileptic syndromes in infancy, childhood, and adolescence. London: John Libbey, 1992.
- 19Shinnar S, O'Dell C. Treating childhood seizures: when and for how long? In: Shinnar S, Amir N, Branski D, eds. Childhood Seizures. Vol. 6. Basel: Karger, 1995:100–110.
- 21Chadwick DW, Berg AT. Stopping treatment for epilepsy. In: Schmidt D, Schachter SC, eds. Epilepsy: Problem solving in clinical practice. London, UK: Martin Dunitz, 2000:221–229.
- 23Klein JP, Moeschberger ML. Survival analysis techniques for censored and truncated data. New York: Springer-Verlag, 1997.
- 24Allisons PD. Survival analysis using the SAS system. Care, NC: SAS Institute Inc., 1995.
- 25Medical Research Council Antiepileptic Drug Withdrawal Study Group. Randomised study of antiepileptic drug withdrawal in patients in remission. Lancet 1991;337:1175–1180.