doi:10.1093/brain/awu294
Brain 2014: 137; 3213–3222
| 3213
BRAIN A JOURNAL OF NEUROLOGY
Complete remission of childhood-onset epilepsy: stability and prediction over two decades Anne T. Berg,1,2 Karen Rychlik,3 Susan R. Levy4,5 and Francine M. Testa4,5 1 Ann and Robert H. Lurie Children’s Hospital of Chicago, Epilepsy Centre, Department of Paediatrics, Chicago, IL, 60611, USA 2 Ann and Robert H. Lurie Children’s Hospital of Chicago, Northwestern Memorial Feinberg School of Medicine, Department of Paediatrics, Chicago, IL, 60611, USA 3 Ann and Robert H. Lurie Children’s Hospital of Chicago, Biostatistics Research Core, Stanley Manne Children’s Research Institute, Chicago, IL, 60611, USA 4 Department of Neurology, Yale School of Medicine, New Haven, CT, 06510, USA 5 Department of Paediatrics, Yale School of Medicine, New Haven, CT, 06510, USA
The ultimate seizure outcome of childhood epilepsy is complete resolution of all seizures without further treatment. How often this happens and how well it can be predicted early in the course of epilepsy could be valuable in helping families understand the nature of childhood epilepsy and what to expect over time. In the Connecticut study of epilepsy, a prospective cohort of 613 children with newly-diagnosed epilepsy (onset age 0–15 years), complete remission, 55 years both seizure-free and medication-free, was examined as a proxy of complete seizure resolution. Predictors at initial diagnosis were tested. Information about seizure outcomes within 2 years and from 2–5 years after diagnosis was sequentially added in a proportional hazards model. The predictive value of the models was determined with logistic regression. Five hundred and sixteen subjects were followed 510 years. Three hundred and twenty-eight (63%) achieved complete remission; 23 relapsed. The relapse rate was 8.2 per 1000 person-years and decreased over time: 10.7, 6.7, and 0 during first 5 years, the next 5 years, and then 410 years after complete remission (P = 0.06 for trend). Six participants regained complete remission; 311 (60%) were in complete remission at last contact. Baseline factors predicting against complete remission at last contact included onset age 510 years (hazard ratio = 0.55, P = 0.0009) and early school or developmental problems (hazard ratio = 0.74, P = 0.01). Factors predicting for complete remission were uncomplicated epilepsy presentation (hazard ratio = 2.23, P 5 0.0001), focal self-limited epilepsy syndrome (hazard ratio = 2.13, P 5 0.0001), and uncharacterized epilepsy (hazard ratio = 1.61, P = 0.04). Remission (hazard ratio = 1.95, P 5 0.0001) and pharmaco-resistance (hazard ratio = 0.33, P 5 0.0001) by 2 years respectfully predicted in favour and against complete remission. From 2 to 5 years after diagnosis, relapse (hazard ratio = 0.21, P 5 0.0001) and late pharmaco-resistance (hazard ratio = 0.21, P = 0.008) decreased and late remission (hazard ratio = 2.40, P 5 0.0001) increased chances of entering complete remission. The overall accuracy of the models increased from 72% (baseline information only), to 77% and 85% with addition of 2-year and 5-year outcomes. Relapses after complete remission are rare making this an acceptable proxy for complete seizure resolution. Complete remission after nearly 20 years is reasonably well predicted within 5 years of initial diagnosis.
Received June 27, 2014. Revised August 22, 2014. Accepted August 28, 2014. Advance Access publication October 22, 2014 ß The Author (2014). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: [email protected]
Downloaded from by guest on December 16, 2014
Correspondence to: Anne T. Berg, Ph.D., Research Professor, Ann and Robert H. Lurie Children’s Hospital of Chicago, Northwestern University Feinberg School of Medicine, 225 East Chicago Avenue, Box 29, Chicago, IL 60611-2605, USA E-mail: [email protected]
3214
| Brain 2014: 137; 3213–3222
A. T. Berg et al.
Keywords: epidemiology; prognosis; refractory epilepsy; remission; children
Introduction
Materials and methods The Connecticut Study of Epilepsy invited families of all children with newly diagnosed epilepsy in 1993–97 through the offices of 16 of 17 paediatric neurologists practicing in Connecticut at the time. In the USA, paediatricians are primary care physicians and typically refer to paediatric neurologists when epilepsy is suspected. Details have been previously presented (Berg et al., 1999). Participants were followed prospectively for up to 20 years. Contact with parents and later young adult participants every 3–4 months and biannual review of the medical records allowed for detailed recording of seizure occurrence and seizure medication usage. Follow-up was concluded over a period of several months, terminating in February 2014. Complete remission was defined as achieving a 5-year seizure-free and 5-year medication-free period. Any subsequent seizure for any reason was considered a relapse. Complete remission at last contact was defined based on the date of last seizure and date seizure medications were completely stopped. Eight participants achieved complete remission but then several years later began taking seizure medications for other conditions (e.g. migraine). Their follow-up was censored at the time they initiated the medication and their final outcome was designated as complete remission without relapse. Prognostic variables considered included those used in previous analyses of shorter-term seizure outcomes in this cohort (Berg et al.,
Downloaded from by guest on December 16, 2014
The ultimate desired seizure outcome of paediatric epilepsy is the complete resolution of seizures without further need for medications. How to know the future by predicting remission based upon early clinical factors and early seizure course is of tremendous interest and has been the focus of considerable research efforts (Sillanpa¨a¨ and Schmidt, 2006, 2009a, b; Camfield and Camfield, 2010, 2014; Geerts et al., 2010; Wirrell et al., 2011). Remission, however, is a relative concept. Inevitably, a variety of different markers of good prognosis have been used ranging from 1 year seizure-free at last contact (Sillanpa¨a¨ and Schmidt, 2009b; Wirrell et al., 2011) to 5 years seizure-free without treatment for some period of time (Geerts et al., 2010). Recently, The International League Against Epilepsy suggested a definition of ‘resolution’ of epilepsy to include seizure remission for 10 years and at least 5 years off seizure medications (Fisher et al., 2014). This was based, in part, on a definition of complete remission (5 years seizure-free and 5 years medication-free) used in a previous analysis from a subgroup of the Connecticut Study of Epilepsy (Berg et al., 2011). Here we present data regarding the occurrence of complete remission in the full community-based Connecticut cohort and the stability of this outcome up to almost 15 years later. We also develop models of complete remission at the end of the study in which we evaluate the predictive value of clinical variables from initial diagnosis and the additional predictive value of seizure outcomes during the first 5 years after diagnosis.
2001, 2004, 2006, 2009a, 2011) and, as much as possible, those reported in the analyses from other well-known longitudinal paediatric cohorts (Camfield and Camfield, 2003, 2008, 2014; Arts et al., 2004; Sillanpa¨a¨ and Schmidt, 2009b; Geerts et al., 2010, 2012; Dhamija et al., 2011; Wirrell et al., 2011, 2012). Baseline prognostic variables were factors that were or could have been obtained as part of the initial evaluation (e.g. relevant imaging abnormality). This included grouped age at onset (55, 5–9, 510 years at first seizure), gender, history of febrile seizures, family history of epilepsy, any episode of status epilepticus (seizure 530 min) before the diagnosis of epilepsy, initial seizure frequency, slowing on the initial EEG, abnormal initial neurological exam (per the diagnosing physician’s report), and imaging results (abnormal and likely relevant versus normal or incidental/non-causative findings) as previously described (Berg et al., 2009a). Early developmental or school problem was defined as receipt of any school system or similar services before or at the time of epilepsy diagnosis or a Vineland composite score 525%. The Vineland Adaptive Behaviour Scales were administered for preschool-aged children at the time of initial interview for the study. Over half of the participants had research neuropsychological testing 9 years into the study. Based on this and all other information from the medical records and interviews, a determination was made about the diagnosis of intellectual disability (Berg et al., 2008). Types of epilepsy were grouped into non-syndromic with focal features (not fitting any well-defined epilepsy syndromes), encephalopathic (largely consisting of West, Lennox-Gastaut, and Doose syndromes), self-limited-focal (mostly benign epilepsy with central-temporal spikes), genetic generalized (mostly childhood and juvenile absence and juvenile myoclonic epilepsy), and a group of poorly characterized epilepsy not fitting into the above categories. Because some infants and young children develop West or Lennox-Gastaut syndromes as an evolution from a non-syndromic presentation, we did not include those children in the encephalopathic epilepsy group at baseline but did at 2 years. Finally, we designated the overall presentation as uncomplicated if the child had normal neurological and imaging exams, absence of clear intellectual disability, and no identified insult or condition to which the epilepsy might be reasonably attributed. Complicated presentations included the presence of one or more of these factors. Seizure outcome at 2 years after diagnosis were designated as early remission (in 1-year remission at 2 years), early pharmaco-resistance (failed trials of two appropriate seizure medications by 2 years), and neither in remission or pharmaco-resistant. We created three additional indicators for the seizure outcomes occurring between 2 and 5 years: relapse after first remission, failure of a second medication (late pharmaco-resistance), and attainment of a first 1-year remission (late remission). Because the purpose of the analyses is to examine outcomes over an extended period of time, we excluded participants followed 510 years. Analysis of relapse following complete remission was based on calculating the person-years from attaining complete remission to date of relapse or of last contact if no relapse. Poisson regression was used to determine the rates of relapse during the first 5 years, second 5 years, and 410 years after achieving complete remission. The ‘rule-of-three’ was used to calculate upper confidence limits for 0-numerator results (Hanley and Lippman-Hand, 1983). Data were analysed in SAS 9.3. Bivariate comparisons were performed with t-tests and chi-square tests as appropriate to the data.
Complete remission of childhood epilepsy Proportional hazards models were used to identify independent predictors of complete remission at last contact. We considered first only those factors that were ascertainable at baseline. The non-syndromic group was used as the comparison against which each other epilepsy type group was tested. We then added 2-year seizure outcomes to that model and reassessed contributions of the original baseline variables. Finally, we added the 5-year seizure outcome variables. Sequential models were created for the full group and separately for participants with complicated and uncomplicated presentations. We used logistic regression to determine the sensitivity and specificity and overall predictive value of the three sequential models and to determine the area under the curve, a metric of predictive accuracy, for each model. Finally, because the decision to stop medications may be influenced by the same prognostic factors under study (e.g. family history, early seizure history, etc.), we compared our models of complete remission at last contact to those for 5-year and for 10-year remission at last contact regardless of medication. Consent was obtained in accordance with the Declaration of Helsinki. Parental written consent and child’s assent (when appropriate) was obtained at initial entry into the study. Young adult consent was obtained upon reaching the age of majority. All procedures were approved by the Institutional Review Boards of all institutions involved in this research.
Of 613 children originally recruited, 14 died before completing 10 years of follow-up. Excluding those who died before 10 years, 516/600 (86%) were followed for 510 years. Excluding subjects who died during the first decade after diagnosis, participants followed for 510 versus 510 years were highly comparable with respect to demographic features, types of epilepsy, and seizure outcomes in the first 2 years of follow-up (Table 1). In 516 patients followed for 510 years, the median length of follow-up was 17 years [interquartile range (IQR) 16–18] for a total of 8856 person-years of follow-up.
Complete remission and relapse rates A total of 328 participants achieved complete remission. The average duration of follow-up after achieving complete remission was 8.6 years (standard deviation 3.7). Of the 328, 265 were followed for 55 years and 138 participants were followed for 510 years after attaining complete remission. Twenty-three (7.0%) individuals experienced relapse for an overall relapse rate of 8.2 (4.8, 11.5) per 1000 person-years of follow-up. This rate was slightly higher in the first 5 years than in the second 5 years after complete remission, 10.7 [95% confidence interval (CI) 5.5–15.9] versus 6.7 (95% CI 1.7–11.6). No relapse occurred more than 10 years after attaining complete remission. We can be 95% confident that the rate of relapse after 10 years is 411.2/1000 person-years (Hanley and Lippman-Hand, 1983). These data suggested a decreasing risk of relapse over time since achieving complete remission (P = 0.06 for trend, Table 2). For 18 of 23 individuals who relapsed, the relapse seemed to be spontaneous (with no apparent explanation). Of these, 12 had more than one
| 3215
spontaneous seizure whereas the others had only one seizure and no others during the remainder of the follow-up. Three individuals acknowledged use of illicit drugs in proximity to their seizure relapse, and two others had acute provoking factors (eclampsia and shunt malfunction). Six individuals who relapsed after achieving complete remission experienced a second complete remission episode. Only one went on to experience pharmaco-resistance (Kwan et al., 2010). At the end of the study, 311 (60.3%) of those followed for 510 years were in complete remission. The median time to complete remission at last contact was 8.1 years (IQR 6.9–9.7). At the end of the study, 374 (72.5%) and 303 (58.7%) were in 5- and 10-year remission regardless of treatment. Most of the individual baseline prognostic factors that we examined had strong bivariate associations with complete remission at last contact, as did all of the 2- and 5-year seizure outcomes (Table 3). Nineteen of 20 participants who underwent surgery for treatment of epilepsy or removal of a brain tumour were followed for 510 years. Six, all of whom had undergone tumour resection, were in complete remission. Two others were seizure-free for 5 years but on medications, and the remaining 11 were seizure-free for 55 years at last contact. Details of the surgeries were previously reported (Berg et al., 2009).
Multivariable analyses Sequential multivariable model starting with baseline information and then adding 2- and 5-year seizure outcomes were created for the overall, uncomplicated, and complicated presentation groups (Table 4).
Overall In the first proportional hazards model of baseline clinical factors, focal self-limited epilepsy syndromes [hazard ratio (HR) = 2.13, P 5 0.0001], uncharacterized epilepsies (HR = 1.61, P = 0.04), and an uncomplicated epilepsy presentation (HR = 2.23, P 5 0.0001) were independently associated with an increased chance of attaining complete remission, whereas age at onset 510 years (HR = 0.55, P = 0.0009) and early school or developmental problems (HR = 0.74, P = 0.01) were associated with a lower chance. Early remission by 2 years was associated with an increased chance of complete remission (HR = 1.95, P 5 0.0001), while early pharmaco-resistance (HR = 0.33, P 5 0.0001) predicted against. Finally, both late relapse (HR = 0.21, P 5 0.0001) and late pharmaco-resistance (HR = 0.21, P = 0.005) weighed against complete remission. Late remission (HR = 2.40, P 5 0.0001), however, weighed in favour of complete remission. The reasons associated with relapses occurring between 2 and 5 years were not differentially associated with attaining complete remission: 11/37 (29.7%) for a spontaneous relapse with good medication adherence; 4/5 (80%) in patients who had never taken seizure medication; 2/6 (33%) for relapses following non-adherence of treatment; 4/9 (44.4%) for relapses during medication tapering; 14/46 (30.4%) for relapses that occurred after completely stopping medications; 1/4 (25%) for relapses in the context of a significant illness such as the flu; and 2/7 (28.6%) for relapses associated with
Downloaded from by guest on December 16, 2014
Results
Brain 2014: 137; 3213–3222
3216
| Brain 2014: 137; 3213–3222
A. T. Berg et al.
Table 1 Comparison of survivors followed for 510 years versus those followed for 510 years Feature
Followed 510 years n = 83
Gender Female 44 (53%) Male 39 (47%) Age at onset of epilepsy 55 years 32 (39%) 5–10 years 34 (41%) 410 years 17 (20%) Early seizure outcome at 2 years after diagnosis* 1-year remission by 2 years post diagnosis 34 (48%) Pharmaco-resistant by 2 years post diagnosis 9 (13%) Unclear, neither pharmaco-resistant or in remission 28 (39%) Type of epilepsy Encephalopathic 6 (7%) Non-syndromic with focal features 42 (51%) Self-limited epilepsy with focal features 9 (11%) Generalized genetic epilepsy 18 (22%) Unclear, not categorized 8 (10%) Presence of presumed cause or abnormal neurological exam None 67 (81%) Present 16 (19%)
Followed 510 years n = 516
Pvalue
248 (48%) 268 (52%)
0.40
237 (46%) 191 (37%) 88 (17%)
0.42
283 (54%) 59 (11%) 174 (33%)
0.63
46 (9%) 261 (51%) 62 (12%) 113 (22%) 31 (6%)
0.78
404 (78%) 112 (22%)
0.65
*12 children lost to follow-up before 2 years are excluded from these counts.
Time period since achieving complete remission
Number contributing person-years to the time epoch
Relapse, n
Person years
Rate/1000 person-years
95% CI
Overall 0–5 years 5–10 years 410 years
327 327 265 138
23 16 7 0
2813.5 1496.6 1048.7 268.1
8.2 10.7 6.7 0
4.8–11.5 5.5–15.9 1.7–11.6 0–11.2
P = 0.06 for the trend in decreasing rate from one time epoch to the next.
other factors (e.g. outgrowing medication dosage, severe sleep deprivation, or excessive stress).
Uncomplicated presentation In the uncomplicated group, age at onset 510 years (HR = 0.63, P = 0.01), early school or developmental problems (HR = 0.73, P = 0.01), and family history of epilepsy (HR = 0.69, P = 0.003) independently weighed against complete remission. Self-limited epilepsy syndromes (HR = 2.21, P 5 0.0001) and uncharacterized forms of epilepsy (HR = 1.71, P = 0.02) were associated with an increased chance of a good outcome. These factors remained in the predictive model with the addition of early remission, which weighed in favour (HR = 1.87, P 5 0.0001), and early pharmacoresistance (HR = 0.29, P = 0.0004), which weighed against complete remission. With the exception of early school or developmental problems, all of the above factors remained in the model once late remission (HR = 2.14, P = 0.0004), relapse by 5 years (HR = 0.2, P 5 0.0001), and late pharmaco-resistance (HR = 0.29, P = 0.03) were added.
Complicated presentation Older age at onset (HR = 0.22, P = 0.04), intellectual disability (HR = 0.39, P = 0.005), and a relevant imaging abnormality (HR = 0.49, P = 0.03) were independently associated with lower rates of attaining complete remission. At 2 years, early remission (HR = 2.53, P = 0.008) favoured attaining complete remission. We combined late pharmaco-resistance and relapse between 2 and 5 years as they were too intercorrelated to test separately. Together, they strongly weighed against attaining complete remission (HR = 0.11, P 5 0.0001). Late remission, however, strongly weighed in favour of a good outcome (HR = 7.43, P = 0.0003).
Predictive values of the proportional hazards models The areas under the curve for the overall group were 72%, 77%, and 85% at baseline, 2, and 5 years, respectively. The comparable figures were 69%, 75%, and 84% for the uncomplicated group and 74%,
Downloaded from by guest on December 16, 2014
Table 2 Rate of relapse after attaining complete remission (5 years seizure-free and 5 years drug-free) overall and by 5-year epochs
Complete remission of childhood epilepsy
Brain 2014: 137; 3213–3222
| 3217
Table 3 Bivariate associations for complete remission at last contact and baseline, 2-year, and 5-year variables Factor (n, %)
Total of 516
P-value
Bivariate hazard ratio (95% CI)
93 (39.2%) 60 (31.4%) 52 (59.1%)
144 (60.8%) 131 (68.6%) 36 (40.9%)
50.0001
1.0 1.24 (0.97–1.57) 0.56 (0.39–0.81)
0.08 0.002
106 (42.7%) 99 (36.9%)
142 (57.3%) 169 (63.1%)
0.18
1.09 (0.87–1.37)
0.44
74 (66.1%) 131 (32.4%)
38 (33.9%) 272 (67.6%)
50.0001
1.0 2.84 (2.02–4.00)
50.0001
161 (35.5%) 44 (69.8%)
292 (64.5%) 19 (30.2%)
50.0001
1.0 0.34 (0.21–0.54)
50.0001
147 (35.5%) 47 (73.4%) 11 (28.9%)
267 (64.5%) 17 (26.6%) 27 (71.1%)
50.0001
1.0 0.29 (0.18–0.47)
50.0001
147 (33.6%) 58 (74.4%)
291 (66.4%) 20 (25.6%)
50.0001
1.0 0.27 (0.17–0.42)
50.0001
88 (31.8%) 117 (49.0%)
189 (68.2%) 122 (51%)
50.0001
1.0 0.59 (0.47–0.74)
50.0001
180 (38.8%) 22 (47.8%) 3 (50%)
284 (61.2%) 24 (52.2%) 3 (50%)
0.23
1.0 0.72 (0.48–1.10)
0.13
107 (37.4%) 95 (42.4%) 3 (50%)
179 (62.6%) 129 (57.6%) 3 (50%)
0.25
1.0 0.83 (0.66–1.04)
0.10
179 (39.8%) 24 (38.7%) 2 (50%)
271 (60.2%) 38 (61.3%)
0.87
1.0 0.95 (0.68–1.34)
0.79
182 (38.6%) 23 (52.3%)
290 (61.4%) 21 (47.7%)
0.07
1.0 0.62 (0.40–0.96)
0.03
1.0 (1.0–1.0)
0.21
37 91 32 12 15 18
(36.6%) (38.9%) (38.1%) (38.7%) (51.7%) (48.7%)
64 143 52 19 14 19
(63.4%) (61.1%) (61.9%) (61.3%) (48.3%) (51.4%)
0.10
117 27 3 48 10
(44.3%) (58.7%) (4.8%) (42.5%) (32.3%)
147 19 59 65 21
(55.7%) (41.3%) (95.2%) (57.5%) (67.7%)
50.0001
154 (36.3%) 51 (55.4%)
270 (63.7%) 41 (44.6%)
0.0007
181 (38.6%)
288 (61.4%)
0.10
1.0 0.70 2.79 1.06 1.64
(0.43–1.13) (2.01–3.80) (0.79–1.42) (1.04–2.59)
1.0 0.61 ( 0.43–0.84)
P-value
0.14 50.0001 0.70 0.03
0.003
1.0 (continued)
Downloaded from by guest on December 16, 2014
Age at onset 55 years 237 (46%) 5–10 years 191 (37.0%) 410 years 88 (17.1%) Gender Female 248 (48.1%) Male 268 (51.9%) Presentation Complicated 112 (21.7%) Uncomplicated 404 (78.3%) Neurological exam Normal 453 (87.8%) Abnormal 63 (12.2%) Neuroimaging Normal/incidental 414 (80.2%) Abnormal/relevant 64 (12.4%) No imaging 38 (7.4%) Intellectual disability Absent 438 (84.9%) Present 78 (15.1%) Early school/developmental problem Absent 277 (53.7%) Present 239 (46.3%) Family history of epilepsy, first degree relative Absent 464 (89.9%) Present 46 (8.9%) Unknown 6 (1.2%) Any family history of epilepsy Absent 286 (55.4%) Present 224 (43.4%) Unknown 6 (1.1%) History of febrile seizures Absent 450 (87.2%) Present 62 (12.0%) Unknown 2 (50%) Status Epilepticus by the time of diagnosis Absent 472 (91.5%) Present 44 (8.5%) Initial seizure frequency 51/month 101 (19.6%) 1234 (45.4%) 1084 (16.3%) 5031 (6.0%) 100 29 (5.6%) 5200/month 37 (7.2%) Type of epilepsy Non-syndromic – focal 264 (51.2%) Encephalopathic 46 (8.9%) Self-limited – focal 62 (12.0%) Genetic generalized 113 (21.9%) Not further specified 31 (6.0%) Any EEG slowing Absent 424 (82.2%) Present 92 (17.8%) Focal slowing Absent 469 (90.9%)
In complete remission (n = 311, 60.3%)
Not in complete remission (n = 205, 39.7%)
3218
| Brain 2014: 137; 3213–3222
A. T. Berg et al.
Table 3 Continued Factor (n, %)
Total of 516
Not in complete remission (n = 205, 39.7%)
In complete remission (n = 311, 60.3%)
Present 47 (9.1%) 24 (51.1%) 23 Other slowing Absent 468 (90.7%) 175 (37.4%) 293 Present 48 (9.3%) 30 (62.5%) 18 Appearance of West or Lennox-Gastaut syndrome or other encephalopathy At onset 46 (8.9%) 27 (58.7%) 19 By 2 years 8 (1.6%) 6 (75%) 2 None 462 (89.5%) 172 (37.2%) 290 Seizure outcome at 2-years In 1 + year remission 283 (54.8%) 80 (28.3%) 203 Pharmaco-resistant 59 (11.4%) 48 (81.4%) 11 Uncertain 174 (33.7%) 77 (44.3%) 97 Relapse between 2 and 5 years* None 374 (76.3%) 101 (27.0%) 273 Present 116 (23.7%) 78 (67.2%) 38 Failure of a second AED between 2 and 5 years Absent 492 (95.4%) 184 (37.4%) 308 Present 24 (4.7%) 21 (87.5%) 3
P-value
(48.9%)
Bivariate hazard ratio (95% CI)
P-value
0.74 (0.49–1.14)
0.17
1.0 0.49 (0.30–0.79)
0.003
(62.6%) (37.5%)
0.0007
(41.3%) (25%) (62.8%)
0.005
(71.7%) (18.6%) (55.8%)
50.0001
1.96 (1.55–2.49) 0.29 (0.16–0.57) 1.0
(73.0%) (32.8%)
50.0001
1.0 0.31 (0.22–0.44)
(62.6%) (12.5%)
50.0001
1.0 0.13 (0.04–0.41)
50.0001 50.0001
50.0001 0.0005
Bivariate hazards ratios with 95% CI from a Cox proportional hazards models are provided. AED = anti-epileptic drug.
Models and terms A. Overall (n = 516) Baseline variables Age at onset 510 years Focal self-limited epilepsy syndrome Uncharacterized epilepsy Uncomplicated epilepsy presentation Early school problem or low Vineland With addition of seizure outcomes at 2 years Age at onset 510 years Focal self-limited epilepsy syndrome Uncharacterized epilepsy Uncomplicated epilepsy presentation Early remission Early pharmaco-resistance With addition of seizure outcomes from 2–5 years Age at onset 510 years Focal self-limited epilepsy syndrome Uncomplicated epilepsy presentation Early remission Early pharmaco-resistance Late remission Relapse between 2 and 5 years Late pharmaco-resistance B. Uncomplicated initial presentation (n = 411) Baseline variables Age 510 at onset Self-limited focal epilepsy Uncharacterized epilepsy Early developmental or school problem
Hazard ratio (95% CI)
P-value
Area under the curve (%)
0.55 2.13 1.62 2.22 0.74
(0.39–0.79) (1.58–2.87) (1.03–2.53) (1.55–3.21) (0.58–0.94)
0.0009 50.0001 0.04 50.0001 0.01
72 (68, 76)
0.52 2.07 1.59 2.03 1.97 0.33
(0.36–0.73) (1.53–2.80) (1.01–2.49) (1.43–2.88) (1.55–2.50) (0.20–0.57)
0.0002 50.0001 0.04 50.0001 50.0001 50.0001
77 (73, 81)
0.59 1.60 1.68 3.99 0.32 2.35 0.21 0.21
(0.41–0.83) (1.19–2.16) (1.18–2.38) (2.82–5.64) (0.19–0.54) (1.59–3.48) (0.15–0.30) (0.07–0.67)
0.003 0.002 0.004 50.0001 50.0001 50.0001 50.0001 0.008
85 (82, 89)
0.63 2.21 1.71 0.73
(0.44–0.91) (1.63–3.00) (1.08–2.72) (0.56–0.94)
0.01 50.0001 0.02 0.01
69 (64, 74)
(continued)
Downloaded from by guest on December 16, 2014
Table 4 Multivariable Cox proportional hazards models and area under the curve for sequential models to predict complete remission at last contact
Complete remission of childhood epilepsy
Brain 2014: 137; 3213–3222
| 3219
Table 4 Continued Models and terms
0.69 (0.54–0.88)
P-value
Area under the curve (%)
0.003
0.57 2.13 1.66 0.74 1.87 0.29
(0.40–0.83) (1.57–2.89) (1.04–2.64) (0.58–0.94) (1.45–2.42) (0.15–0.57)
0.003 50.0001 0.03 0.01 50.0001 0.0003
75 (70, 80)
0.66 1.76 1.74 0.68 3.53 0.26 2.14 0.20 0.29
(0.45–0.95) (1.29–2.39) (1.09–2.78) (0.53–0.86) (2.45–5.08) (0.13–0.51) (1.41–3.26) (0.13–0.29) (0.09–0.91)
0.02 0.0003 0.02 0.002 50.0001 50.0001 0.0004 50.0001 0.03
84 (80, 88)
74 (65, 83) 0.22 (0.05–0.94) 0.39 (0.21–0.75) 0.49 (0.26–0.94)
0.04 0.005 0.03
0.24 0.46 0.63 2.53
(0.06–1.03) (0.24–0.88) (0.32–1.24) (1.28–5.02)
0.06 0.02 0.18 0.008
0.30 0.37 0.41 14.35 0.11 7.46
(0.07–1.27) (0.18–0.73) (0.21–0.81) (5.11–40.22) (0.05–0.27) (2.51–22.22)
0.10 0.004 0.01 50.0001 50.0001 0.0003
78 (69, 88)
87 (81, 94)
*Kept in model for parsimony because it was retained in preceding and following models.
78%, and 87% for the complicated group, respectively. Most of these changes from baseline to 2–5 years represented significant incremental improvements in the accuracy of the models (Figure 1).
Remission regardless of medication To explore whether factors such as family history of epilepsy or early school/developmental problems might be associated with use of medication rather than remission itself, we re-ran all analyses with 5- and 10-year remission as outcomes, regardless of seizure medication use. Only uncharacterized epilepsy syndrome and early school/developmental problem were no longer associated with 5-year remission, overall and in the uncomplicated group. Models for the complicated group were not substantially altered. These patterns were mirrored in analyses of 10-year remission (Supplementary Table 1).
Discussion Over the course of nearly two decades, almost two-thirds of children with newly diagnosed epilepsy achieve complete remission.
Our findings suggest that this is a meaningful proxy for the resolution of epilepsy. Although there is a small residual risk of relapse (51% per year), this is the strongest marker of enduring remission that we are aware of to date. The proportion in remission is comparable to what has been reported in other cohorts followed for longer periods of time but based on varying and less stringent definitions of remission (Sillanpa¨a¨ et al., 1998, 2012; Geerts et al., 2010; Wirrell et al., 2011; Camfield and Camfield, 2014). Complete remission, as we have defined it, has a low risk of later relapse and as such is an excellent marker of seizure resolution. While no relapses were recorded 410 years after attaining complete remission, our data cannot rule out a residual relapse rate of 1% per year, 10 times the population rate for new incident seizures (Olafsson et al., 2005). In terms of anticipatory guidance and prognostication, our findings confirm the unique position of benign epilepsy with centraltemporal spikes and related epilepsies in determining long-term resolution of seizures (Loiseau et al., 1983, 1988; Bouma et al., 1997). They also highlight the importance of a clear neurological insult or related condition and of early seizure outcomes in
Downloaded from by guest on December 16, 2014
Any family history of epilepsy At 2 years Age 510 at onset Self-limited focal epilepsy Uncharacterized epilepsy Any family history of epilepsy Early remission Early pharmaco-resistance At 5 years Age 510 at onset Self-limited focal epilepsy Uncharacterized epilepsy Any family history of epilepsy Early remission Early pharmaco-resistance Late remission Relapse between 2 and 5 years Failure of second AED between 2 and 5 years C. Complicated presentation (n = 112) Age 510 years Intellectual disability Relevant imaging abnormality At 2 years Age 510 years Intellectual disability Relevant imaging abnormality* Early remission At 5 years Age 510 years* Intellectual disability Relevant imaging abnormality Early remission Relapse or late pharmaco-resistance Late remission
Hazard ratio (95% CI)
3220
| Brain 2014: 137; 3213–3222
A. T. Berg et al.
Figure 1 Receiver operator curves for predication of complete remission at last contact based on initial clinical information, with addition of 2-year seizure outcomes, and with addition of 5-year seizure outcomes: (A) Overall (n = 516), (B) uncomplicated epilepsy presentation (n = 404), and (C) complicated epilepsy presentation (n = 112).
predictive of remission without medication. Our finding that most factors, especially early seizure outcomes, that predicted complete remission at last contact also predicts 5- and 10-year remission (regardless of medications) suggests that the factors we identified are indeed predictive of seizure resolution. Over the last several years, many seizure medications have been found useful in the treatment of a variety of other disorders. In fact, we deliberately censored follow-up for a handful of patients in whom complete remission was ‘broken’ because medications used to treat seizures were initiated for non-seizure indications (e.g. migraine, other pain, psychiatric and behavioural indications). Whether resolution of seizures without treatment realistically can be studied today given the multiple purposes for which seizure medications are used is another concern and creates questions that might be better addressed in other settings. There has, in the past, been debate, especially among epidemiologists, about the value of recognizing epilepsy syndromes (Olafsson et al., 2005). At this point in time, there can be no doubt that these syndromes are of considerable value, although not just for predicting outcome. They are critical for pursuing causes, genetic counselling, selecting treatment, management, and anticipatory guidance. Each tends to occur in a small number of children. The individual contributions of each syndrome to understanding complete remission in a heterogeneous cohort such as ours and those published by others cannot be reasonably studied in a single cohort-wide analysis. By combining into broad groups (e.g. encephalopathic, generalized genetic, etc.), we are able to consider only the most dramatic impacts with respect to complete remission, particularly that of benign epilepsy with centraltemporal spikes. Furthermore, we have considered only one outcome, complete remission of epilepsy. Of those who do not attain this outcome, there is a range of outcomes from well-controlled, to moderately controlled, to relentlessly refractory seizures. To understand fully the contribution of individual syndromes to seizure prognosis requires a detailed analysis within sufficiently large homogeneous samples and consideration of this broader range of outcomes. All studies have limitations and the greatest concern with ours is that two decades of follow-up may not be adequate to reveal the longer-term prognosis of childhood epilepsies. While that may be true for some purposes, it is not clear that a longer follow-up
Downloaded from by guest on December 16, 2014
predicting later outcomes (Sillanpa¨a¨ et al., 1998, 2012; Camfield and Camfield, 2003, 2008, 2010; Geelhoed et al., 2005; Sillanpa¨a¨ and Schmidt, 2009a, b; Geerts et al., 2010, 2012; Dhamija et al., 2011; Wirrell et al., 2011). Area under the curve is a common metric used to summarize the trade-off between sensitivity and specificity in predictive models (Eng, 2005). According to some recommended interpretations of area under the curve, the overall predictive accuracy based on the initial clinical variables was only fair (70–80%) (Kaiser-Permanente, 2009). With seizure outcomes through 5 years, predictive accuracy was good (80–90%) but not excellent (490%). Excellent prediction of seizure outcome is elusive. For example, Geelhoed et al. (2005) reported only slightly less predictive accuracy in models of 2-year seizure outcome and also noted the importance of early (6-month) outcomes relative to initial clinical (baseline) information. Given the types of information available in a large-scale study, this may be the best that can be expected, especially in an ‘omnibus’ analysis that includes such heterogeneous epilepsies. Analyses limited to more homogeneous patient subgroups might provide more precise information, although limited and relevant to the subgroup only. Of particular interest was the dramatic decrease in attaining complete remission if any relapse occurred after remission in the first 5 years. The factors associated with relapse (e.g. planned discontinuation of medication or non-adherence versus spontaneous remissions) did not seem to influence complete remission. While our data cannot definitively support the notion that maximizing efforts to limit relapses early will improve long-term outcomes, they are certainly consistent with the observations and the interpretations presented by Modi and colleagues (2014) who found that poor early adherence to medication schedules heralded a poorer chance of seizure control several years later. Although about a third who relapsed in our cohort still achieved complete remission, this is substantially less than the overall average, and it is information that needs to be communicated carefully to families. There is the insurmountable problem, in this and virtually any study; remission without medication requires that medications be stopped. The reasons for continuing medication may not necessarily be related to the patient’s chances of remaining seizure-free off medication and thus could, potentially, seem to be negatively
Complete remission of childhood epilepsy
Acknowledgements We thank the paediatric neurologists of Connecticut who contributed to this study. We extend a special recognition to Eugene Shapiro and Shlomo Shinnar for their contributions, to Barry Russman who opened doors and made the study possible right from the very beginning, and to the research associates, Christina Rios, Charles Hurst, and Maria Consolatore as well as Susan Smith and Barbara Beckerman whose dedication was an essential ingredient for success. The families who generously gave their time and energy to this study made this all possible and worthwhile.
Funding This study was fully funded by a grant from the National Institutes of Health – the National Institute of Neurological Disorders and Stroke–R37-NS31146.
Supplementary material Supplementary material is available at Brain online.
References Arts WF, Brouwer O, Peters A, Stroink H, Peeters E, Schmitz P, et al. Course and prognosis of childhood epilepsy: 5-year follow-up of the Dutch study of epilepsy in childhood. Brain 2004; 127: 1774–84.
| 3221
Berg AT, Langfitt JT, Testa FM, Levy SR, DiMario F, Westerveld M, et al. Global cognitive function in children with epilepsy: a community-based study. Epilepsia 2008; 49: 608–14. Berg AT, Levy SR, Testa FM, D’Souza R. Remission of epilepsy after two drug failures in children: a prospective study. Ann Neurol 2009a; 65: 510–9. Berg AT, Lin J, Ebrahimi N, Testa FM, Levy SR, Shinnar S. Modeling remission and relapse in pediatric epilepsy: application of a Markov process. Epilepsy Res 2004; 60: 31–40. Berg AT, Mathern GW, Bronen RA, Fulbright RK, DiMario F, Testa FM, et al. Frequency, prognosis and surgical treatment of structural abnormalities seen with magnetic resonance imaging in childhood epilepsy. Brain 2009b; 132: 2785–97. Berg AT, Shinnar S, Levy SR, Testa FM. Newly diagnosed epilepsy in children: presentation at diagnosis. Epilepsia 1999; 40: 445–52. Berg AT, Shinnar S, Levy SR, Testa FM, Smith-Rapaport S, Beckerman B, et al. Defining early seizure outcomes in pediatric epilepsy: the good, the bad and the in-between. Epilepsy Res 2001; 43: 75–84. Berg AT, Vickrey BG, Testa FM, Levy SR, Shinnar S, DiMario F, et al. How long does it take for epilepsy to become intractable? A prospective investigation. Ann Neurol 2006; 60: 73–9. Berg AT, Testa FM, Levy SR. Complete remission in nonsyndromic childhood-onset epilepsy. Ann Neurol 2011; 70: 566–73. Bouma PA, Bovenkerk AC, Westendorp RG, Brouwer OF. The course of benign partial epilepsy of childhood with centrotemporal spikes: a meta-analysis. Neurology 1997; 48: 430–7. Camfield P, Camfield C. Idiopathic generalized epilepsy with generalized tonic-clonic seizures (IGE-GTC): A population-based cohort with 420 year follow up for medical and social outcome. Epil Behav 2010; 18: 61–3. 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. p. 113–26. Camfield C, Camfield P. Twenty years after childhood-onset symptomatic generalized epilepsy the social outcome is usually dependency or death: a population-based study. Dev Med Child Neurol 2008; 50: 859–63. Camfield CS, Camfield PR. The adult seizure and social outcomes of children with partial complex seizures. Brain 2014; 136(pt 2): 593–600. Dhamija R, Moseley BD, Cascino GD, Wirrell EC. A population-based study of long-term outcome of epilepsy in childhood with a focal or hemispheric lesion on neuroimaging. Epilepsia 2011; 52: 1522–6. Elterman RD, Shields WD, Mansfield KA, Nakagawa J. Group tUISVS. Randomized trial of vigabatrin in patients with infantile spasms. Neurology 2001; 57: 1416–21. Eng J. Receiver operating characteristic analysis: a primer1. Acad Radiol 2005; 12: 909–16. First Seizure Trial Group. Randomized clinical trial of the efficacy of antiepileptic drugs in reducing the risk of relapse after a first unprovoked tonic-clonic seizure. Neurology 1993; 43: 478–83. Fisher RS, Acevedo C, Arzimanoglou A, Bogacz A, Cross JH, Elger CE, et al. ILAE Official Report: A practical clinical definition of epilepsy. Epilepsia 2014; 55: 475–82. Geelhoed M, Boerrigter AO, Camfield P, Geerts AT, Arts W, Smith B, et al. The accuracy of outcome prediction models for childhood-onset epilepsy. Epilepsia 2005; 46: 1526–32. Geerts A, Arts WF, Stroink H, Peeters E, Brouwer O, Peters B, et al. Course and outcome of childhood epilepsy: a 15-year follow-up of the Dutch Study of Epilepsy in Childhood. Epilepsia 2010; 51: 1189–97. Geerts A, Brouwer O, Stroink H, van Donselaar C, Peters B, Peeters E, et al. Onset of intractability and its course over time: the Dutch study of epilepsy in childhood. Epilepsia 2012; 53: 741–51. Hanley JA, Lippman-Hand A. If nothing goes wrong, is everything all right? Interpreting zero numerators. JAMA 1983; 249: 1743–5. Kaiser-Permanente. Receiver Operating Characteritic (ROC) Curves: assessing the predictive properties of a test statistic-decision theory. Kaiser Permanente Center for Health Research. Portland, OR, USA, 2009.
Downloaded from by guest on December 16, 2014
would provide more precise information. Treatment options have expanded tremendously since the mid-1990s when the Connecticut cohort was recruited. Diagnostic technologies have also changed. In our cohort, 85% of participants had an MRI for either clinical or research purposes; however, this was mostly with 1.5 T machines. Whether today’s advanced imaging technologies and genetic testing could provide more precise prognostic information is a question of great interest. There are many seizure outcomes that are studied, usually for different purposes. The choice should be determined by the question under investigation and the time scale of the study. For example, treatment trials of infantile spasms might reasonably target a 1-week spasm-free period as a meaningful marker of therapeutic success (Elterman et al., 2001). A randomized trial of AEDs for newly diagnosed epilepsy might target 1-year seizure freedom (First Seizure Trial Group, 1993; Marson et al., 2005, 2007a, b). Such trials are informative about the impact of the treatment on seizures. They provide limited insight, however, into the long-term likelihood that epilepsy will fully resolve, be a life-long, chronic disorder (managed to a greater or lesser extent), or be relentlessly intractable. For long-term studies, perhaps complete resolution of seizures would be of greatest interest. To advance our understanding of the long-term, perhaps life-long, prognosis of epilepsy with regard to seizure resolution, better prediction of shorter-term outcomes that reliably predict longer-term outcomes may be the most efficient way to move forward.
Brain 2014: 137; 3213–3222
3222
| Brain 2014: 137; 3213–3222
Kwan P, Arzimanoglou A, Berg AT, Brodie MJ, Hauser WA, Mathern G, et al. Definition of drug resistant epilepsy: consensus proposal by the ad hoc task force of the ILAE commission on therapeutic strategies. Epilepsia 2010; 51: 1069–77. Loiseau P, Duche´ B, Cordova S, Dartigues JF, Cohadon S. Prognosis of benign childhood epilepsy with centrotemporal spikes: a foliow-up study of 168 patients. Epilepsia 1988; 29: 229–35. Loiseau P, Pestre M, Dartigues JF, Commenges D, Barberger-Gateau C, Cohadon S. Long-term prognosis in two forms of childhood epilepsy: typical absence seizures and epilepsy with rolandic (centrotemporal) EEG foci. Ann Neurol 1983; 13: 642–8. Marson A, Jacoby A, Johnson A, Kim L, Gamble C, Chadwick D. Immediate versus deferred antiepileptic drug treatment for early epilepsy and single seizures: a randomised controlled trial. Lancet 2005; 365: 2007–13. Marson AG, Al-Kharusi AM, Alwaidh M, Appleton R, Baker GA, Chadwick DW, et al. The SANAD study of effectiveness of valproate, lamotrigine, or topiramate for generalised and unclassifiable epilepsy: an unblinded randomised controlled trial. Lancet 2007a; 369: 1016–26. Marson AG, Al-Kharusi AM, Alwaidh M, Appleton R, Baker GA, Chadwick DW, et al. The SANAD study of effectiveness of carbamazepine, gabapentin, lamotrigine, oxcarbazepine, or topiramate for treatment of partial epilepsy: an unblinded randomised controlled trial. Lancet 2007b; 369: 1000–15. Modi AC, Rausch JR, Glauser TA. Early pediatric antiepileptic drug nonadherence is related to lower long-term seizure freedom. Neurology 2014; 82: 671–3.
A. T. Berg et al. Olafsson E, Ludvigsson P, Hesdorffer D, Kjartansson O, Hauser WA, Gudmundsson G. Incidence of unprovoked seizures and epilepsy in Iceland and assessment of the epilepsy syndrome classification: a prospective study. Lancet Neurol 2005; 4: 627–34. Sillanpa¨a¨ M, Jalava M, Kaleva O, Shinnar S. Long-term prognosis of seizures with onset in childhood. N Engl J Med 1998; 338: 1715–22. Sillanpa¨a¨ M, Saarinen M, Schmidt D. Long-term risks following first remission in childhood-onset epilepsy. A population-based study. Epil Behav 2012; 25: 145–9. Sillanpa¨a¨ M, Schmidt D. Delayed time to first remission identifies poor long-term drug response of childhood-onset epilepsy: a prospective population-based study. Epilepsy Behav 2009a; 16: 507–11. Sillanpa¨a¨ M, Schmidt D. Early seizure frequency and aetiology predict long-term medical outcome in childhood-onset epilepsy. Brain 2009b; 132: 989–98. Sillanpa¨a¨ M, Schmidt D. Natural history of treated childhood-onset epilepsy: prospective, long-term population-based study. Brain 2006; 129: 617–24. Wirrell E, Wong-Kisiel L, Mandrekar J, Nickels K. Predictors and course of medically intractable epilepsy in young children presenting before 36 months of age: a retrospective, population-based study. Epilepsia 2012; 53: 1563–9. Wirrell EC, Grossardt BR, So EL, Nickels KC. A population-based study of long-term outcomes of cryptogenic focal epilepsy in childhood: cryptogenic epilepsy is probably not symptomatic epilepsy. Epilepsia 2011; 52: 738–45.
Downloaded from by guest on December 16, 2014
Brain 2014: 137; 3213–3222
| 3213
BRAIN A JOURNAL OF NEUROLOGY
Complete remission of childhood-onset epilepsy: stability and prediction over two decades Anne T. Berg,1,2 Karen Rychlik,3 Susan R. Levy4,5 and Francine M. Testa4,5 1 Ann and Robert H. Lurie Children’s Hospital of Chicago, Epilepsy Centre, Department of Paediatrics, Chicago, IL, 60611, USA 2 Ann and Robert H. Lurie Children’s Hospital of Chicago, Northwestern Memorial Feinberg School of Medicine, Department of Paediatrics, Chicago, IL, 60611, USA 3 Ann and Robert H. Lurie Children’s Hospital of Chicago, Biostatistics Research Core, Stanley Manne Children’s Research Institute, Chicago, IL, 60611, USA 4 Department of Neurology, Yale School of Medicine, New Haven, CT, 06510, USA 5 Department of Paediatrics, Yale School of Medicine, New Haven, CT, 06510, USA
The ultimate seizure outcome of childhood epilepsy is complete resolution of all seizures without further treatment. How often this happens and how well it can be predicted early in the course of epilepsy could be valuable in helping families understand the nature of childhood epilepsy and what to expect over time. In the Connecticut study of epilepsy, a prospective cohort of 613 children with newly-diagnosed epilepsy (onset age 0–15 years), complete remission, 55 years both seizure-free and medication-free, was examined as a proxy of complete seizure resolution. Predictors at initial diagnosis were tested. Information about seizure outcomes within 2 years and from 2–5 years after diagnosis was sequentially added in a proportional hazards model. The predictive value of the models was determined with logistic regression. Five hundred and sixteen subjects were followed 510 years. Three hundred and twenty-eight (63%) achieved complete remission; 23 relapsed. The relapse rate was 8.2 per 1000 person-years and decreased over time: 10.7, 6.7, and 0 during first 5 years, the next 5 years, and then 410 years after complete remission (P = 0.06 for trend). Six participants regained complete remission; 311 (60%) were in complete remission at last contact. Baseline factors predicting against complete remission at last contact included onset age 510 years (hazard ratio = 0.55, P = 0.0009) and early school or developmental problems (hazard ratio = 0.74, P = 0.01). Factors predicting for complete remission were uncomplicated epilepsy presentation (hazard ratio = 2.23, P 5 0.0001), focal self-limited epilepsy syndrome (hazard ratio = 2.13, P 5 0.0001), and uncharacterized epilepsy (hazard ratio = 1.61, P = 0.04). Remission (hazard ratio = 1.95, P 5 0.0001) and pharmaco-resistance (hazard ratio = 0.33, P 5 0.0001) by 2 years respectfully predicted in favour and against complete remission. From 2 to 5 years after diagnosis, relapse (hazard ratio = 0.21, P 5 0.0001) and late pharmaco-resistance (hazard ratio = 0.21, P = 0.008) decreased and late remission (hazard ratio = 2.40, P 5 0.0001) increased chances of entering complete remission. The overall accuracy of the models increased from 72% (baseline information only), to 77% and 85% with addition of 2-year and 5-year outcomes. Relapses after complete remission are rare making this an acceptable proxy for complete seizure resolution. Complete remission after nearly 20 years is reasonably well predicted within 5 years of initial diagnosis.
Received June 27, 2014. Revised August 22, 2014. Accepted August 28, 2014. Advance Access publication October 22, 2014 ß The Author (2014). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: [email protected]
Downloaded from by guest on December 16, 2014
Correspondence to: Anne T. Berg, Ph.D., Research Professor, Ann and Robert H. Lurie Children’s Hospital of Chicago, Northwestern University Feinberg School of Medicine, 225 East Chicago Avenue, Box 29, Chicago, IL 60611-2605, USA E-mail: [email protected]
3214
| Brain 2014: 137; 3213–3222
A. T. Berg et al.
Keywords: epidemiology; prognosis; refractory epilepsy; remission; children
Introduction
Materials and methods The Connecticut Study of Epilepsy invited families of all children with newly diagnosed epilepsy in 1993–97 through the offices of 16 of 17 paediatric neurologists practicing in Connecticut at the time. In the USA, paediatricians are primary care physicians and typically refer to paediatric neurologists when epilepsy is suspected. Details have been previously presented (Berg et al., 1999). Participants were followed prospectively for up to 20 years. Contact with parents and later young adult participants every 3–4 months and biannual review of the medical records allowed for detailed recording of seizure occurrence and seizure medication usage. Follow-up was concluded over a period of several months, terminating in February 2014. Complete remission was defined as achieving a 5-year seizure-free and 5-year medication-free period. Any subsequent seizure for any reason was considered a relapse. Complete remission at last contact was defined based on the date of last seizure and date seizure medications were completely stopped. Eight participants achieved complete remission but then several years later began taking seizure medications for other conditions (e.g. migraine). Their follow-up was censored at the time they initiated the medication and their final outcome was designated as complete remission without relapse. Prognostic variables considered included those used in previous analyses of shorter-term seizure outcomes in this cohort (Berg et al.,
Downloaded from by guest on December 16, 2014
The ultimate desired seizure outcome of paediatric epilepsy is the complete resolution of seizures without further need for medications. How to know the future by predicting remission based upon early clinical factors and early seizure course is of tremendous interest and has been the focus of considerable research efforts (Sillanpa¨a¨ and Schmidt, 2006, 2009a, b; Camfield and Camfield, 2010, 2014; Geerts et al., 2010; Wirrell et al., 2011). Remission, however, is a relative concept. Inevitably, a variety of different markers of good prognosis have been used ranging from 1 year seizure-free at last contact (Sillanpa¨a¨ and Schmidt, 2009b; Wirrell et al., 2011) to 5 years seizure-free without treatment for some period of time (Geerts et al., 2010). Recently, The International League Against Epilepsy suggested a definition of ‘resolution’ of epilepsy to include seizure remission for 10 years and at least 5 years off seizure medications (Fisher et al., 2014). This was based, in part, on a definition of complete remission (5 years seizure-free and 5 years medication-free) used in a previous analysis from a subgroup of the Connecticut Study of Epilepsy (Berg et al., 2011). Here we present data regarding the occurrence of complete remission in the full community-based Connecticut cohort and the stability of this outcome up to almost 15 years later. We also develop models of complete remission at the end of the study in which we evaluate the predictive value of clinical variables from initial diagnosis and the additional predictive value of seizure outcomes during the first 5 years after diagnosis.
2001, 2004, 2006, 2009a, 2011) and, as much as possible, those reported in the analyses from other well-known longitudinal paediatric cohorts (Camfield and Camfield, 2003, 2008, 2014; Arts et al., 2004; Sillanpa¨a¨ and Schmidt, 2009b; Geerts et al., 2010, 2012; Dhamija et al., 2011; Wirrell et al., 2011, 2012). Baseline prognostic variables were factors that were or could have been obtained as part of the initial evaluation (e.g. relevant imaging abnormality). This included grouped age at onset (55, 5–9, 510 years at first seizure), gender, history of febrile seizures, family history of epilepsy, any episode of status epilepticus (seizure 530 min) before the diagnosis of epilepsy, initial seizure frequency, slowing on the initial EEG, abnormal initial neurological exam (per the diagnosing physician’s report), and imaging results (abnormal and likely relevant versus normal or incidental/non-causative findings) as previously described (Berg et al., 2009a). Early developmental or school problem was defined as receipt of any school system or similar services before or at the time of epilepsy diagnosis or a Vineland composite score 525%. The Vineland Adaptive Behaviour Scales were administered for preschool-aged children at the time of initial interview for the study. Over half of the participants had research neuropsychological testing 9 years into the study. Based on this and all other information from the medical records and interviews, a determination was made about the diagnosis of intellectual disability (Berg et al., 2008). Types of epilepsy were grouped into non-syndromic with focal features (not fitting any well-defined epilepsy syndromes), encephalopathic (largely consisting of West, Lennox-Gastaut, and Doose syndromes), self-limited-focal (mostly benign epilepsy with central-temporal spikes), genetic generalized (mostly childhood and juvenile absence and juvenile myoclonic epilepsy), and a group of poorly characterized epilepsy not fitting into the above categories. Because some infants and young children develop West or Lennox-Gastaut syndromes as an evolution from a non-syndromic presentation, we did not include those children in the encephalopathic epilepsy group at baseline but did at 2 years. Finally, we designated the overall presentation as uncomplicated if the child had normal neurological and imaging exams, absence of clear intellectual disability, and no identified insult or condition to which the epilepsy might be reasonably attributed. Complicated presentations included the presence of one or more of these factors. Seizure outcome at 2 years after diagnosis were designated as early remission (in 1-year remission at 2 years), early pharmaco-resistance (failed trials of two appropriate seizure medications by 2 years), and neither in remission or pharmaco-resistant. We created three additional indicators for the seizure outcomes occurring between 2 and 5 years: relapse after first remission, failure of a second medication (late pharmaco-resistance), and attainment of a first 1-year remission (late remission). Because the purpose of the analyses is to examine outcomes over an extended period of time, we excluded participants followed 510 years. Analysis of relapse following complete remission was based on calculating the person-years from attaining complete remission to date of relapse or of last contact if no relapse. Poisson regression was used to determine the rates of relapse during the first 5 years, second 5 years, and 410 years after achieving complete remission. The ‘rule-of-three’ was used to calculate upper confidence limits for 0-numerator results (Hanley and Lippman-Hand, 1983). Data were analysed in SAS 9.3. Bivariate comparisons were performed with t-tests and chi-square tests as appropriate to the data.
Complete remission of childhood epilepsy Proportional hazards models were used to identify independent predictors of complete remission at last contact. We considered first only those factors that were ascertainable at baseline. The non-syndromic group was used as the comparison against which each other epilepsy type group was tested. We then added 2-year seizure outcomes to that model and reassessed contributions of the original baseline variables. Finally, we added the 5-year seizure outcome variables. Sequential models were created for the full group and separately for participants with complicated and uncomplicated presentations. We used logistic regression to determine the sensitivity and specificity and overall predictive value of the three sequential models and to determine the area under the curve, a metric of predictive accuracy, for each model. Finally, because the decision to stop medications may be influenced by the same prognostic factors under study (e.g. family history, early seizure history, etc.), we compared our models of complete remission at last contact to those for 5-year and for 10-year remission at last contact regardless of medication. Consent was obtained in accordance with the Declaration of Helsinki. Parental written consent and child’s assent (when appropriate) was obtained at initial entry into the study. Young adult consent was obtained upon reaching the age of majority. All procedures were approved by the Institutional Review Boards of all institutions involved in this research.
Of 613 children originally recruited, 14 died before completing 10 years of follow-up. Excluding those who died before 10 years, 516/600 (86%) were followed for 510 years. Excluding subjects who died during the first decade after diagnosis, participants followed for 510 versus 510 years were highly comparable with respect to demographic features, types of epilepsy, and seizure outcomes in the first 2 years of follow-up (Table 1). In 516 patients followed for 510 years, the median length of follow-up was 17 years [interquartile range (IQR) 16–18] for a total of 8856 person-years of follow-up.
Complete remission and relapse rates A total of 328 participants achieved complete remission. The average duration of follow-up after achieving complete remission was 8.6 years (standard deviation 3.7). Of the 328, 265 were followed for 55 years and 138 participants were followed for 510 years after attaining complete remission. Twenty-three (7.0%) individuals experienced relapse for an overall relapse rate of 8.2 (4.8, 11.5) per 1000 person-years of follow-up. This rate was slightly higher in the first 5 years than in the second 5 years after complete remission, 10.7 [95% confidence interval (CI) 5.5–15.9] versus 6.7 (95% CI 1.7–11.6). No relapse occurred more than 10 years after attaining complete remission. We can be 95% confident that the rate of relapse after 10 years is 411.2/1000 person-years (Hanley and Lippman-Hand, 1983). These data suggested a decreasing risk of relapse over time since achieving complete remission (P = 0.06 for trend, Table 2). For 18 of 23 individuals who relapsed, the relapse seemed to be spontaneous (with no apparent explanation). Of these, 12 had more than one
| 3215
spontaneous seizure whereas the others had only one seizure and no others during the remainder of the follow-up. Three individuals acknowledged use of illicit drugs in proximity to their seizure relapse, and two others had acute provoking factors (eclampsia and shunt malfunction). Six individuals who relapsed after achieving complete remission experienced a second complete remission episode. Only one went on to experience pharmaco-resistance (Kwan et al., 2010). At the end of the study, 311 (60.3%) of those followed for 510 years were in complete remission. The median time to complete remission at last contact was 8.1 years (IQR 6.9–9.7). At the end of the study, 374 (72.5%) and 303 (58.7%) were in 5- and 10-year remission regardless of treatment. Most of the individual baseline prognostic factors that we examined had strong bivariate associations with complete remission at last contact, as did all of the 2- and 5-year seizure outcomes (Table 3). Nineteen of 20 participants who underwent surgery for treatment of epilepsy or removal of a brain tumour were followed for 510 years. Six, all of whom had undergone tumour resection, were in complete remission. Two others were seizure-free for 5 years but on medications, and the remaining 11 were seizure-free for 55 years at last contact. Details of the surgeries were previously reported (Berg et al., 2009).
Multivariable analyses Sequential multivariable model starting with baseline information and then adding 2- and 5-year seizure outcomes were created for the overall, uncomplicated, and complicated presentation groups (Table 4).
Overall In the first proportional hazards model of baseline clinical factors, focal self-limited epilepsy syndromes [hazard ratio (HR) = 2.13, P 5 0.0001], uncharacterized epilepsies (HR = 1.61, P = 0.04), and an uncomplicated epilepsy presentation (HR = 2.23, P 5 0.0001) were independently associated with an increased chance of attaining complete remission, whereas age at onset 510 years (HR = 0.55, P = 0.0009) and early school or developmental problems (HR = 0.74, P = 0.01) were associated with a lower chance. Early remission by 2 years was associated with an increased chance of complete remission (HR = 1.95, P 5 0.0001), while early pharmaco-resistance (HR = 0.33, P 5 0.0001) predicted against. Finally, both late relapse (HR = 0.21, P 5 0.0001) and late pharmaco-resistance (HR = 0.21, P = 0.005) weighed against complete remission. Late remission (HR = 2.40, P 5 0.0001), however, weighed in favour of complete remission. The reasons associated with relapses occurring between 2 and 5 years were not differentially associated with attaining complete remission: 11/37 (29.7%) for a spontaneous relapse with good medication adherence; 4/5 (80%) in patients who had never taken seizure medication; 2/6 (33%) for relapses following non-adherence of treatment; 4/9 (44.4%) for relapses during medication tapering; 14/46 (30.4%) for relapses that occurred after completely stopping medications; 1/4 (25%) for relapses in the context of a significant illness such as the flu; and 2/7 (28.6%) for relapses associated with
Downloaded from by guest on December 16, 2014
Results
Brain 2014: 137; 3213–3222
3216
| Brain 2014: 137; 3213–3222
A. T. Berg et al.
Table 1 Comparison of survivors followed for 510 years versus those followed for 510 years Feature
Followed 510 years n = 83
Gender Female 44 (53%) Male 39 (47%) Age at onset of epilepsy 55 years 32 (39%) 5–10 years 34 (41%) 410 years 17 (20%) Early seizure outcome at 2 years after diagnosis* 1-year remission by 2 years post diagnosis 34 (48%) Pharmaco-resistant by 2 years post diagnosis 9 (13%) Unclear, neither pharmaco-resistant or in remission 28 (39%) Type of epilepsy Encephalopathic 6 (7%) Non-syndromic with focal features 42 (51%) Self-limited epilepsy with focal features 9 (11%) Generalized genetic epilepsy 18 (22%) Unclear, not categorized 8 (10%) Presence of presumed cause or abnormal neurological exam None 67 (81%) Present 16 (19%)
Followed 510 years n = 516
Pvalue
248 (48%) 268 (52%)
0.40
237 (46%) 191 (37%) 88 (17%)
0.42
283 (54%) 59 (11%) 174 (33%)
0.63
46 (9%) 261 (51%) 62 (12%) 113 (22%) 31 (6%)
0.78
404 (78%) 112 (22%)
0.65
*12 children lost to follow-up before 2 years are excluded from these counts.
Time period since achieving complete remission
Number contributing person-years to the time epoch
Relapse, n
Person years
Rate/1000 person-years
95% CI
Overall 0–5 years 5–10 years 410 years
327 327 265 138
23 16 7 0
2813.5 1496.6 1048.7 268.1
8.2 10.7 6.7 0
4.8–11.5 5.5–15.9 1.7–11.6 0–11.2
P = 0.06 for the trend in decreasing rate from one time epoch to the next.
other factors (e.g. outgrowing medication dosage, severe sleep deprivation, or excessive stress).
Uncomplicated presentation In the uncomplicated group, age at onset 510 years (HR = 0.63, P = 0.01), early school or developmental problems (HR = 0.73, P = 0.01), and family history of epilepsy (HR = 0.69, P = 0.003) independently weighed against complete remission. Self-limited epilepsy syndromes (HR = 2.21, P 5 0.0001) and uncharacterized forms of epilepsy (HR = 1.71, P = 0.02) were associated with an increased chance of a good outcome. These factors remained in the predictive model with the addition of early remission, which weighed in favour (HR = 1.87, P 5 0.0001), and early pharmacoresistance (HR = 0.29, P = 0.0004), which weighed against complete remission. With the exception of early school or developmental problems, all of the above factors remained in the model once late remission (HR = 2.14, P = 0.0004), relapse by 5 years (HR = 0.2, P 5 0.0001), and late pharmaco-resistance (HR = 0.29, P = 0.03) were added.
Complicated presentation Older age at onset (HR = 0.22, P = 0.04), intellectual disability (HR = 0.39, P = 0.005), and a relevant imaging abnormality (HR = 0.49, P = 0.03) were independently associated with lower rates of attaining complete remission. At 2 years, early remission (HR = 2.53, P = 0.008) favoured attaining complete remission. We combined late pharmaco-resistance and relapse between 2 and 5 years as they were too intercorrelated to test separately. Together, they strongly weighed against attaining complete remission (HR = 0.11, P 5 0.0001). Late remission, however, strongly weighed in favour of a good outcome (HR = 7.43, P = 0.0003).
Predictive values of the proportional hazards models The areas under the curve for the overall group were 72%, 77%, and 85% at baseline, 2, and 5 years, respectively. The comparable figures were 69%, 75%, and 84% for the uncomplicated group and 74%,
Downloaded from by guest on December 16, 2014
Table 2 Rate of relapse after attaining complete remission (5 years seizure-free and 5 years drug-free) overall and by 5-year epochs
Complete remission of childhood epilepsy
Brain 2014: 137; 3213–3222
| 3217
Table 3 Bivariate associations for complete remission at last contact and baseline, 2-year, and 5-year variables Factor (n, %)
Total of 516
P-value
Bivariate hazard ratio (95% CI)
93 (39.2%) 60 (31.4%) 52 (59.1%)
144 (60.8%) 131 (68.6%) 36 (40.9%)
50.0001
1.0 1.24 (0.97–1.57) 0.56 (0.39–0.81)
0.08 0.002
106 (42.7%) 99 (36.9%)
142 (57.3%) 169 (63.1%)
0.18
1.09 (0.87–1.37)
0.44
74 (66.1%) 131 (32.4%)
38 (33.9%) 272 (67.6%)
50.0001
1.0 2.84 (2.02–4.00)
50.0001
161 (35.5%) 44 (69.8%)
292 (64.5%) 19 (30.2%)
50.0001
1.0 0.34 (0.21–0.54)
50.0001
147 (35.5%) 47 (73.4%) 11 (28.9%)
267 (64.5%) 17 (26.6%) 27 (71.1%)
50.0001
1.0 0.29 (0.18–0.47)
50.0001
147 (33.6%) 58 (74.4%)
291 (66.4%) 20 (25.6%)
50.0001
1.0 0.27 (0.17–0.42)
50.0001
88 (31.8%) 117 (49.0%)
189 (68.2%) 122 (51%)
50.0001
1.0 0.59 (0.47–0.74)
50.0001
180 (38.8%) 22 (47.8%) 3 (50%)
284 (61.2%) 24 (52.2%) 3 (50%)
0.23
1.0 0.72 (0.48–1.10)
0.13
107 (37.4%) 95 (42.4%) 3 (50%)
179 (62.6%) 129 (57.6%) 3 (50%)
0.25
1.0 0.83 (0.66–1.04)
0.10
179 (39.8%) 24 (38.7%) 2 (50%)
271 (60.2%) 38 (61.3%)
0.87
1.0 0.95 (0.68–1.34)
0.79
182 (38.6%) 23 (52.3%)
290 (61.4%) 21 (47.7%)
0.07
1.0 0.62 (0.40–0.96)
0.03
1.0 (1.0–1.0)
0.21
37 91 32 12 15 18
(36.6%) (38.9%) (38.1%) (38.7%) (51.7%) (48.7%)
64 143 52 19 14 19
(63.4%) (61.1%) (61.9%) (61.3%) (48.3%) (51.4%)
0.10
117 27 3 48 10
(44.3%) (58.7%) (4.8%) (42.5%) (32.3%)
147 19 59 65 21
(55.7%) (41.3%) (95.2%) (57.5%) (67.7%)
50.0001
154 (36.3%) 51 (55.4%)
270 (63.7%) 41 (44.6%)
0.0007
181 (38.6%)
288 (61.4%)
0.10
1.0 0.70 2.79 1.06 1.64
(0.43–1.13) (2.01–3.80) (0.79–1.42) (1.04–2.59)
1.0 0.61 ( 0.43–0.84)
P-value
0.14 50.0001 0.70 0.03
0.003
1.0 (continued)
Downloaded from by guest on December 16, 2014
Age at onset 55 years 237 (46%) 5–10 years 191 (37.0%) 410 years 88 (17.1%) Gender Female 248 (48.1%) Male 268 (51.9%) Presentation Complicated 112 (21.7%) Uncomplicated 404 (78.3%) Neurological exam Normal 453 (87.8%) Abnormal 63 (12.2%) Neuroimaging Normal/incidental 414 (80.2%) Abnormal/relevant 64 (12.4%) No imaging 38 (7.4%) Intellectual disability Absent 438 (84.9%) Present 78 (15.1%) Early school/developmental problem Absent 277 (53.7%) Present 239 (46.3%) Family history of epilepsy, first degree relative Absent 464 (89.9%) Present 46 (8.9%) Unknown 6 (1.2%) Any family history of epilepsy Absent 286 (55.4%) Present 224 (43.4%) Unknown 6 (1.1%) History of febrile seizures Absent 450 (87.2%) Present 62 (12.0%) Unknown 2 (50%) Status Epilepticus by the time of diagnosis Absent 472 (91.5%) Present 44 (8.5%) Initial seizure frequency 51/month 101 (19.6%) 1234 (45.4%) 1084 (16.3%) 5031 (6.0%) 100 29 (5.6%) 5200/month 37 (7.2%) Type of epilepsy Non-syndromic – focal 264 (51.2%) Encephalopathic 46 (8.9%) Self-limited – focal 62 (12.0%) Genetic generalized 113 (21.9%) Not further specified 31 (6.0%) Any EEG slowing Absent 424 (82.2%) Present 92 (17.8%) Focal slowing Absent 469 (90.9%)
In complete remission (n = 311, 60.3%)
Not in complete remission (n = 205, 39.7%)
3218
| Brain 2014: 137; 3213–3222
A. T. Berg et al.
Table 3 Continued Factor (n, %)
Total of 516
Not in complete remission (n = 205, 39.7%)
In complete remission (n = 311, 60.3%)
Present 47 (9.1%) 24 (51.1%) 23 Other slowing Absent 468 (90.7%) 175 (37.4%) 293 Present 48 (9.3%) 30 (62.5%) 18 Appearance of West or Lennox-Gastaut syndrome or other encephalopathy At onset 46 (8.9%) 27 (58.7%) 19 By 2 years 8 (1.6%) 6 (75%) 2 None 462 (89.5%) 172 (37.2%) 290 Seizure outcome at 2-years In 1 + year remission 283 (54.8%) 80 (28.3%) 203 Pharmaco-resistant 59 (11.4%) 48 (81.4%) 11 Uncertain 174 (33.7%) 77 (44.3%) 97 Relapse between 2 and 5 years* None 374 (76.3%) 101 (27.0%) 273 Present 116 (23.7%) 78 (67.2%) 38 Failure of a second AED between 2 and 5 years Absent 492 (95.4%) 184 (37.4%) 308 Present 24 (4.7%) 21 (87.5%) 3
P-value
(48.9%)
Bivariate hazard ratio (95% CI)
P-value
0.74 (0.49–1.14)
0.17
1.0 0.49 (0.30–0.79)
0.003
(62.6%) (37.5%)
0.0007
(41.3%) (25%) (62.8%)
0.005
(71.7%) (18.6%) (55.8%)
50.0001
1.96 (1.55–2.49) 0.29 (0.16–0.57) 1.0
(73.0%) (32.8%)
50.0001
1.0 0.31 (0.22–0.44)
(62.6%) (12.5%)
50.0001
1.0 0.13 (0.04–0.41)
50.0001 50.0001
50.0001 0.0005
Bivariate hazards ratios with 95% CI from a Cox proportional hazards models are provided. AED = anti-epileptic drug.
Models and terms A. Overall (n = 516) Baseline variables Age at onset 510 years Focal self-limited epilepsy syndrome Uncharacterized epilepsy Uncomplicated epilepsy presentation Early school problem or low Vineland With addition of seizure outcomes at 2 years Age at onset 510 years Focal self-limited epilepsy syndrome Uncharacterized epilepsy Uncomplicated epilepsy presentation Early remission Early pharmaco-resistance With addition of seizure outcomes from 2–5 years Age at onset 510 years Focal self-limited epilepsy syndrome Uncomplicated epilepsy presentation Early remission Early pharmaco-resistance Late remission Relapse between 2 and 5 years Late pharmaco-resistance B. Uncomplicated initial presentation (n = 411) Baseline variables Age 510 at onset Self-limited focal epilepsy Uncharacterized epilepsy Early developmental or school problem
Hazard ratio (95% CI)
P-value
Area under the curve (%)
0.55 2.13 1.62 2.22 0.74
(0.39–0.79) (1.58–2.87) (1.03–2.53) (1.55–3.21) (0.58–0.94)
0.0009 50.0001 0.04 50.0001 0.01
72 (68, 76)
0.52 2.07 1.59 2.03 1.97 0.33
(0.36–0.73) (1.53–2.80) (1.01–2.49) (1.43–2.88) (1.55–2.50) (0.20–0.57)
0.0002 50.0001 0.04 50.0001 50.0001 50.0001
77 (73, 81)
0.59 1.60 1.68 3.99 0.32 2.35 0.21 0.21
(0.41–0.83) (1.19–2.16) (1.18–2.38) (2.82–5.64) (0.19–0.54) (1.59–3.48) (0.15–0.30) (0.07–0.67)
0.003 0.002 0.004 50.0001 50.0001 50.0001 50.0001 0.008
85 (82, 89)
0.63 2.21 1.71 0.73
(0.44–0.91) (1.63–3.00) (1.08–2.72) (0.56–0.94)
0.01 50.0001 0.02 0.01
69 (64, 74)
(continued)
Downloaded from by guest on December 16, 2014
Table 4 Multivariable Cox proportional hazards models and area under the curve for sequential models to predict complete remission at last contact
Complete remission of childhood epilepsy
Brain 2014: 137; 3213–3222
| 3219
Table 4 Continued Models and terms
0.69 (0.54–0.88)
P-value
Area under the curve (%)
0.003
0.57 2.13 1.66 0.74 1.87 0.29
(0.40–0.83) (1.57–2.89) (1.04–2.64) (0.58–0.94) (1.45–2.42) (0.15–0.57)
0.003 50.0001 0.03 0.01 50.0001 0.0003
75 (70, 80)
0.66 1.76 1.74 0.68 3.53 0.26 2.14 0.20 0.29
(0.45–0.95) (1.29–2.39) (1.09–2.78) (0.53–0.86) (2.45–5.08) (0.13–0.51) (1.41–3.26) (0.13–0.29) (0.09–0.91)
0.02 0.0003 0.02 0.002 50.0001 50.0001 0.0004 50.0001 0.03
84 (80, 88)
74 (65, 83) 0.22 (0.05–0.94) 0.39 (0.21–0.75) 0.49 (0.26–0.94)
0.04 0.005 0.03
0.24 0.46 0.63 2.53
(0.06–1.03) (0.24–0.88) (0.32–1.24) (1.28–5.02)
0.06 0.02 0.18 0.008
0.30 0.37 0.41 14.35 0.11 7.46
(0.07–1.27) (0.18–0.73) (0.21–0.81) (5.11–40.22) (0.05–0.27) (2.51–22.22)
0.10 0.004 0.01 50.0001 50.0001 0.0003
78 (69, 88)
87 (81, 94)
*Kept in model for parsimony because it was retained in preceding and following models.
78%, and 87% for the complicated group, respectively. Most of these changes from baseline to 2–5 years represented significant incremental improvements in the accuracy of the models (Figure 1).
Remission regardless of medication To explore whether factors such as family history of epilepsy or early school/developmental problems might be associated with use of medication rather than remission itself, we re-ran all analyses with 5- and 10-year remission as outcomes, regardless of seizure medication use. Only uncharacterized epilepsy syndrome and early school/developmental problem were no longer associated with 5-year remission, overall and in the uncomplicated group. Models for the complicated group were not substantially altered. These patterns were mirrored in analyses of 10-year remission (Supplementary Table 1).
Discussion Over the course of nearly two decades, almost two-thirds of children with newly diagnosed epilepsy achieve complete remission.
Our findings suggest that this is a meaningful proxy for the resolution of epilepsy. Although there is a small residual risk of relapse (51% per year), this is the strongest marker of enduring remission that we are aware of to date. The proportion in remission is comparable to what has been reported in other cohorts followed for longer periods of time but based on varying and less stringent definitions of remission (Sillanpa¨a¨ et al., 1998, 2012; Geerts et al., 2010; Wirrell et al., 2011; Camfield and Camfield, 2014). Complete remission, as we have defined it, has a low risk of later relapse and as such is an excellent marker of seizure resolution. While no relapses were recorded 410 years after attaining complete remission, our data cannot rule out a residual relapse rate of 1% per year, 10 times the population rate for new incident seizures (Olafsson et al., 2005). In terms of anticipatory guidance and prognostication, our findings confirm the unique position of benign epilepsy with centraltemporal spikes and related epilepsies in determining long-term resolution of seizures (Loiseau et al., 1983, 1988; Bouma et al., 1997). They also highlight the importance of a clear neurological insult or related condition and of early seizure outcomes in
Downloaded from by guest on December 16, 2014
Any family history of epilepsy At 2 years Age 510 at onset Self-limited focal epilepsy Uncharacterized epilepsy Any family history of epilepsy Early remission Early pharmaco-resistance At 5 years Age 510 at onset Self-limited focal epilepsy Uncharacterized epilepsy Any family history of epilepsy Early remission Early pharmaco-resistance Late remission Relapse between 2 and 5 years Failure of second AED between 2 and 5 years C. Complicated presentation (n = 112) Age 510 years Intellectual disability Relevant imaging abnormality At 2 years Age 510 years Intellectual disability Relevant imaging abnormality* Early remission At 5 years Age 510 years* Intellectual disability Relevant imaging abnormality Early remission Relapse or late pharmaco-resistance Late remission
Hazard ratio (95% CI)
3220
| Brain 2014: 137; 3213–3222
A. T. Berg et al.
Figure 1 Receiver operator curves for predication of complete remission at last contact based on initial clinical information, with addition of 2-year seizure outcomes, and with addition of 5-year seizure outcomes: (A) Overall (n = 516), (B) uncomplicated epilepsy presentation (n = 404), and (C) complicated epilepsy presentation (n = 112).
predictive of remission without medication. Our finding that most factors, especially early seizure outcomes, that predicted complete remission at last contact also predicts 5- and 10-year remission (regardless of medications) suggests that the factors we identified are indeed predictive of seizure resolution. Over the last several years, many seizure medications have been found useful in the treatment of a variety of other disorders. In fact, we deliberately censored follow-up for a handful of patients in whom complete remission was ‘broken’ because medications used to treat seizures were initiated for non-seizure indications (e.g. migraine, other pain, psychiatric and behavioural indications). Whether resolution of seizures without treatment realistically can be studied today given the multiple purposes for which seizure medications are used is another concern and creates questions that might be better addressed in other settings. There has, in the past, been debate, especially among epidemiologists, about the value of recognizing epilepsy syndromes (Olafsson et al., 2005). At this point in time, there can be no doubt that these syndromes are of considerable value, although not just for predicting outcome. They are critical for pursuing causes, genetic counselling, selecting treatment, management, and anticipatory guidance. Each tends to occur in a small number of children. The individual contributions of each syndrome to understanding complete remission in a heterogeneous cohort such as ours and those published by others cannot be reasonably studied in a single cohort-wide analysis. By combining into broad groups (e.g. encephalopathic, generalized genetic, etc.), we are able to consider only the most dramatic impacts with respect to complete remission, particularly that of benign epilepsy with centraltemporal spikes. Furthermore, we have considered only one outcome, complete remission of epilepsy. Of those who do not attain this outcome, there is a range of outcomes from well-controlled, to moderately controlled, to relentlessly refractory seizures. To understand fully the contribution of individual syndromes to seizure prognosis requires a detailed analysis within sufficiently large homogeneous samples and consideration of this broader range of outcomes. All studies have limitations and the greatest concern with ours is that two decades of follow-up may not be adequate to reveal the longer-term prognosis of childhood epilepsies. While that may be true for some purposes, it is not clear that a longer follow-up
Downloaded from by guest on December 16, 2014
predicting later outcomes (Sillanpa¨a¨ et al., 1998, 2012; Camfield and Camfield, 2003, 2008, 2010; Geelhoed et al., 2005; Sillanpa¨a¨ and Schmidt, 2009a, b; Geerts et al., 2010, 2012; Dhamija et al., 2011; Wirrell et al., 2011). Area under the curve is a common metric used to summarize the trade-off between sensitivity and specificity in predictive models (Eng, 2005). According to some recommended interpretations of area under the curve, the overall predictive accuracy based on the initial clinical variables was only fair (70–80%) (Kaiser-Permanente, 2009). With seizure outcomes through 5 years, predictive accuracy was good (80–90%) but not excellent (490%). Excellent prediction of seizure outcome is elusive. For example, Geelhoed et al. (2005) reported only slightly less predictive accuracy in models of 2-year seizure outcome and also noted the importance of early (6-month) outcomes relative to initial clinical (baseline) information. Given the types of information available in a large-scale study, this may be the best that can be expected, especially in an ‘omnibus’ analysis that includes such heterogeneous epilepsies. Analyses limited to more homogeneous patient subgroups might provide more precise information, although limited and relevant to the subgroup only. Of particular interest was the dramatic decrease in attaining complete remission if any relapse occurred after remission in the first 5 years. The factors associated with relapse (e.g. planned discontinuation of medication or non-adherence versus spontaneous remissions) did not seem to influence complete remission. While our data cannot definitively support the notion that maximizing efforts to limit relapses early will improve long-term outcomes, they are certainly consistent with the observations and the interpretations presented by Modi and colleagues (2014) who found that poor early adherence to medication schedules heralded a poorer chance of seizure control several years later. Although about a third who relapsed in our cohort still achieved complete remission, this is substantially less than the overall average, and it is information that needs to be communicated carefully to families. There is the insurmountable problem, in this and virtually any study; remission without medication requires that medications be stopped. The reasons for continuing medication may not necessarily be related to the patient’s chances of remaining seizure-free off medication and thus could, potentially, seem to be negatively
Complete remission of childhood epilepsy
Acknowledgements We thank the paediatric neurologists of Connecticut who contributed to this study. We extend a special recognition to Eugene Shapiro and Shlomo Shinnar for their contributions, to Barry Russman who opened doors and made the study possible right from the very beginning, and to the research associates, Christina Rios, Charles Hurst, and Maria Consolatore as well as Susan Smith and Barbara Beckerman whose dedication was an essential ingredient for success. The families who generously gave their time and energy to this study made this all possible and worthwhile.
Funding This study was fully funded by a grant from the National Institutes of Health – the National Institute of Neurological Disorders and Stroke–R37-NS31146.
Supplementary material Supplementary material is available at Brain online.
References Arts WF, Brouwer O, Peters A, Stroink H, Peeters E, Schmitz P, et al. Course and prognosis of childhood epilepsy: 5-year follow-up of the Dutch study of epilepsy in childhood. Brain 2004; 127: 1774–84.
| 3221
Berg AT, Langfitt JT, Testa FM, Levy SR, DiMario F, Westerveld M, et al. Global cognitive function in children with epilepsy: a community-based study. Epilepsia 2008; 49: 608–14. Berg AT, Levy SR, Testa FM, D’Souza R. Remission of epilepsy after two drug failures in children: a prospective study. Ann Neurol 2009a; 65: 510–9. Berg AT, Lin J, Ebrahimi N, Testa FM, Levy SR, Shinnar S. Modeling remission and relapse in pediatric epilepsy: application of a Markov process. Epilepsy Res 2004; 60: 31–40. Berg AT, Mathern GW, Bronen RA, Fulbright RK, DiMario F, Testa FM, et al. Frequency, prognosis and surgical treatment of structural abnormalities seen with magnetic resonance imaging in childhood epilepsy. Brain 2009b; 132: 2785–97. Berg AT, Shinnar S, Levy SR, Testa FM. Newly diagnosed epilepsy in children: presentation at diagnosis. Epilepsia 1999; 40: 445–52. Berg AT, Shinnar S, Levy SR, Testa FM, Smith-Rapaport S, Beckerman B, et al. Defining early seizure outcomes in pediatric epilepsy: the good, the bad and the in-between. Epilepsy Res 2001; 43: 75–84. Berg AT, Vickrey BG, Testa FM, Levy SR, Shinnar S, DiMario F, et al. How long does it take for epilepsy to become intractable? A prospective investigation. Ann Neurol 2006; 60: 73–9. Berg AT, Testa FM, Levy SR. Complete remission in nonsyndromic childhood-onset epilepsy. Ann Neurol 2011; 70: 566–73. Bouma PA, Bovenkerk AC, Westendorp RG, Brouwer OF. The course of benign partial epilepsy of childhood with centrotemporal spikes: a meta-analysis. Neurology 1997; 48: 430–7. Camfield P, Camfield C. Idiopathic generalized epilepsy with generalized tonic-clonic seizures (IGE-GTC): A population-based cohort with 420 year follow up for medical and social outcome. Epil Behav 2010; 18: 61–3. 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. p. 113–26. Camfield C, Camfield P. Twenty years after childhood-onset symptomatic generalized epilepsy the social outcome is usually dependency or death: a population-based study. Dev Med Child Neurol 2008; 50: 859–63. Camfield CS, Camfield PR. The adult seizure and social outcomes of children with partial complex seizures. Brain 2014; 136(pt 2): 593–600. Dhamija R, Moseley BD, Cascino GD, Wirrell EC. A population-based study of long-term outcome of epilepsy in childhood with a focal or hemispheric lesion on neuroimaging. Epilepsia 2011; 52: 1522–6. Elterman RD, Shields WD, Mansfield KA, Nakagawa J. Group tUISVS. Randomized trial of vigabatrin in patients with infantile spasms. Neurology 2001; 57: 1416–21. Eng J. Receiver operating characteristic analysis: a primer1. Acad Radiol 2005; 12: 909–16. First Seizure Trial Group. Randomized clinical trial of the efficacy of antiepileptic drugs in reducing the risk of relapse after a first unprovoked tonic-clonic seizure. Neurology 1993; 43: 478–83. Fisher RS, Acevedo C, Arzimanoglou A, Bogacz A, Cross JH, Elger CE, et al. ILAE Official Report: A practical clinical definition of epilepsy. Epilepsia 2014; 55: 475–82. Geelhoed M, Boerrigter AO, Camfield P, Geerts AT, Arts W, Smith B, et al. The accuracy of outcome prediction models for childhood-onset epilepsy. Epilepsia 2005; 46: 1526–32. Geerts A, Arts WF, Stroink H, Peeters E, Brouwer O, Peters B, et al. Course and outcome of childhood epilepsy: a 15-year follow-up of the Dutch Study of Epilepsy in Childhood. Epilepsia 2010; 51: 1189–97. Geerts A, Brouwer O, Stroink H, van Donselaar C, Peters B, Peeters E, et al. Onset of intractability and its course over time: the Dutch study of epilepsy in childhood. Epilepsia 2012; 53: 741–51. Hanley JA, Lippman-Hand A. If nothing goes wrong, is everything all right? Interpreting zero numerators. JAMA 1983; 249: 1743–5. Kaiser-Permanente. Receiver Operating Characteritic (ROC) Curves: assessing the predictive properties of a test statistic-decision theory. Kaiser Permanente Center for Health Research. Portland, OR, USA, 2009.
Downloaded from by guest on December 16, 2014
would provide more precise information. Treatment options have expanded tremendously since the mid-1990s when the Connecticut cohort was recruited. Diagnostic technologies have also changed. In our cohort, 85% of participants had an MRI for either clinical or research purposes; however, this was mostly with 1.5 T machines. Whether today’s advanced imaging technologies and genetic testing could provide more precise prognostic information is a question of great interest. There are many seizure outcomes that are studied, usually for different purposes. The choice should be determined by the question under investigation and the time scale of the study. For example, treatment trials of infantile spasms might reasonably target a 1-week spasm-free period as a meaningful marker of therapeutic success (Elterman et al., 2001). A randomized trial of AEDs for newly diagnosed epilepsy might target 1-year seizure freedom (First Seizure Trial Group, 1993; Marson et al., 2005, 2007a, b). Such trials are informative about the impact of the treatment on seizures. They provide limited insight, however, into the long-term likelihood that epilepsy will fully resolve, be a life-long, chronic disorder (managed to a greater or lesser extent), or be relentlessly intractable. For long-term studies, perhaps complete resolution of seizures would be of greatest interest. To advance our understanding of the long-term, perhaps life-long, prognosis of epilepsy with regard to seizure resolution, better prediction of shorter-term outcomes that reliably predict longer-term outcomes may be the most efficient way to move forward.
Brain 2014: 137; 3213–3222
3222
| Brain 2014: 137; 3213–3222
Kwan P, Arzimanoglou A, Berg AT, Brodie MJ, Hauser WA, Mathern G, et al. Definition of drug resistant epilepsy: consensus proposal by the ad hoc task force of the ILAE commission on therapeutic strategies. Epilepsia 2010; 51: 1069–77. Loiseau P, Duche´ B, Cordova S, Dartigues JF, Cohadon S. Prognosis of benign childhood epilepsy with centrotemporal spikes: a foliow-up study of 168 patients. Epilepsia 1988; 29: 229–35. Loiseau P, Pestre M, Dartigues JF, Commenges D, Barberger-Gateau C, Cohadon S. Long-term prognosis in two forms of childhood epilepsy: typical absence seizures and epilepsy with rolandic (centrotemporal) EEG foci. Ann Neurol 1983; 13: 642–8. Marson A, Jacoby A, Johnson A, Kim L, Gamble C, Chadwick D. Immediate versus deferred antiepileptic drug treatment for early epilepsy and single seizures: a randomised controlled trial. Lancet 2005; 365: 2007–13. Marson AG, Al-Kharusi AM, Alwaidh M, Appleton R, Baker GA, Chadwick DW, et al. The SANAD study of effectiveness of valproate, lamotrigine, or topiramate for generalised and unclassifiable epilepsy: an unblinded randomised controlled trial. Lancet 2007a; 369: 1016–26. Marson AG, Al-Kharusi AM, Alwaidh M, Appleton R, Baker GA, Chadwick DW, et al. The SANAD study of effectiveness of carbamazepine, gabapentin, lamotrigine, oxcarbazepine, or topiramate for treatment of partial epilepsy: an unblinded randomised controlled trial. Lancet 2007b; 369: 1000–15. Modi AC, Rausch JR, Glauser TA. Early pediatric antiepileptic drug nonadherence is related to lower long-term seizure freedom. Neurology 2014; 82: 671–3.
A. T. Berg et al. Olafsson E, Ludvigsson P, Hesdorffer D, Kjartansson O, Hauser WA, Gudmundsson G. Incidence of unprovoked seizures and epilepsy in Iceland and assessment of the epilepsy syndrome classification: a prospective study. Lancet Neurol 2005; 4: 627–34. Sillanpa¨a¨ M, Jalava M, Kaleva O, Shinnar S. Long-term prognosis of seizures with onset in childhood. N Engl J Med 1998; 338: 1715–22. Sillanpa¨a¨ M, Saarinen M, Schmidt D. Long-term risks following first remission in childhood-onset epilepsy. A population-based study. Epil Behav 2012; 25: 145–9. Sillanpa¨a¨ M, Schmidt D. Delayed time to first remission identifies poor long-term drug response of childhood-onset epilepsy: a prospective population-based study. Epilepsy Behav 2009a; 16: 507–11. Sillanpa¨a¨ M, Schmidt D. Early seizure frequency and aetiology predict long-term medical outcome in childhood-onset epilepsy. Brain 2009b; 132: 989–98. Sillanpa¨a¨ M, Schmidt D. Natural history of treated childhood-onset epilepsy: prospective, long-term population-based study. Brain 2006; 129: 617–24. Wirrell E, Wong-Kisiel L, Mandrekar J, Nickels K. Predictors and course of medically intractable epilepsy in young children presenting before 36 months of age: a retrospective, population-based study. Epilepsia 2012; 53: 1563–9. Wirrell EC, Grossardt BR, So EL, Nickels KC. A population-based study of long-term outcomes of cryptogenic focal epilepsy in childhood: cryptogenic epilepsy is probably not symptomatic epilepsy. Epilepsia 2011; 52: 738–45.
Downloaded from by guest on December 16, 2014
