CONFIDENTIAL: Not for dissemination, duplication, or citation
1
Practice advisory: Timing of antiepileptic medication withdrawal in seizure-free patients 1
with epilepsy (update of practice parameter) 2
3
Report of the Guideline Development, Dissemination, and Implementation Subcommittee of the 4
American Academy of Neurology 5
6
David Gloss, MD, MPH&TM1; Kimberly Pargeon, MD, MA2; Jay Varma, MD3; Jackie French, 7
MD4; Cynthia Harden, MD5 8
9
1. Department of Neurology, Charleston Area Medical Center, Charleston, WV 10
2. Department of Neurology, New York Downtown Hospital, NY 11
3. Department of Neurology, Barrow Neurological Institute, Phoenix, AZ 12
4. Department of Neurology, New York University Langone Comprehensive Epilepsy Center, 13
New York 14
5. Department of Neurology, Mount Sinai Health System, New York, NY 15
16
17
CONFIDENTIAL: Not for dissemination, duplication, or citation
2
DISCLOSURE 1
Dr. Gloss: Served as an evidence-based medicine methodology consultant for the American 2
Academy of Neurology until January 2018, and continues to serve as a volunteer methodologist, 3
and has served as an associate editor (risk of bias classification) for Neurology. 4
5
Dr. Pargeon: Has no disclosures to report. 6
7
Dr. Varma: Has no disclosures to report. 8
9
Dr. French: Has served as a consultant for Acorda, Biotie, Eisai Medical Research, 10
GlaxoSmithKline, Impax, Johnson & Johnson, Lewis County General Hospital, Marinus, 11
Novartis, Pfizer, Sunovion, SK Life Science, Supernus Pharmaceuticals, UCB, Upsher-Smith, 12
and Vertex; has received grants from Eisai Medical Research, the US Epilepsy Research 13
Foundation, the Epilepsy Study Consortium the Epilepsy Therapy Project of the Epilepsy 14
Foundation, Lundbeck, Pfizer, and UCB; and is president of the Epilepsy Study Consortium. All 15
consulting is done on behalf of the Consortium, and fees are paid to the Consortium. New York 16
University receives salary support from the Consortium. 17
18
Dr. Harden: Has received royalties from Wiley and UpToDate, and has served as a contributing 19
editor for Epilepsy Currents. 20
21
22
CONFIDENTIAL: Not for dissemination, duplication, or citation
3
ABSTRACT 1
Objective: To assess the evidence for medication withdrawal among epilepsy patients who are 2
seizure-free. 3
Methods: A 4-member panel evaluated the available evidence. The panel reviewed abstracts 4
from the 2148 identified articles, and deemed 154 possibly relevant. Two panelists working 5
independently of one another reviewed the full text of each article and selected articles for full 6
data extraction. Case reports, meta-analyses, and editorials were excluded. 7
Results: There was 15 included studies; 1 of which was Class I. 8
Recommendations: Voting after round one: 12 of 16 recommendations received a Level. Four 9
recommendations need to be re-voted on. This section will be updated after voting is complete. 10
11
12
CONFIDENTIAL: Not for dissemination, duplication, or citation
4
INTRODUCTION 1
Epilepsy is a common disease of the brain and accounts for approximately 1% of the global 2
burden of disease (Murray 1994, for more information about global burden of disease, see 3
http://www.who.int/healthinfo/global_burden_disease/about/en/). In the United States alone, an 4
estimated 70,000–200,000 adults per year will present with a first unprovoked seizure (Hauser 5
1993, Hauser 2008). The purpose of prescribing an antiepileptic drug (AED) is to render patients 6
with epilepsy seizure-free, a task that is accomplished approximately two-thirds of the time 7
(Kwan 2000, Mohanraj 2006). In these instances of success, there is the inevitable question 8
concerning if and when patients should be weaned from the AEDs. Among patients who are 9
seizure-free, it is unclear if they are truly disease-free. According to its latest definition, epilepsy 10
is not considered resolved until a patient is seizure-free for at least 10 years and off anti-seizure 11
medications for at least the last 5 years (Fisher 2014). 12
13
This practice advisory is an update to the 1996 American Academy of Neurology (AAN) 14
practice parameter on the same topic(Quality Standards Subcommittee, 1996), which 15
recommended that after assessing the risks and benefits to both patient and society from a 16
recurrent seizure, the discontinuation of antiepileptic drugs may be considered by the physician 17
and informed patient or parent/guardian if the patient meets the following profile: 18
Seizure-free 2–5 years on AEDs (mean 3.5 years) 19
Single type of partial seizure (simple partial or complex partial or secondary generalized 20
tonic-clonic seizure [GTCS]) or single type of primary generalized seizures; 21
Normal neurological examination/normal IQ 22
EEG normalized with treatment 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
5
A Cochrane review addressed this question in children without generalized seizures but was 1
unable to address this issue in adults or in children with generalized seizures (Sirven 2012). This 2
review recommended treatment until the child was seizure-free for at least 2 years before 3
considering withdrawing medications, particularly if there were partial (i.e., focal) seizures or 4
EEG abnormalities. 5
6
The panel for this practice advisory examined questions similar to the Cochrane review but 7
allowed broader inclusion criteria and used the AAN methodology for crafting 8
recommendations. The aim of the practice advisory is to provide information that will be 9
important for practitioners who care for epilepsy patients. 10
11
Questions 12
For patients with epilepsy on AED medication, who have been seizure-free for at least 12 13
months, does stopping AEDs, compared to not stopping: 14
1. increase the risk of recurrence, and are there risk factors for seizure recurrence? 15
2. increase the risk of status epilepticus? 16
3. reduce medication-related side effects? 17
4. change quality of life? 18
5. change the risk of mortality? 19
6. change any of the above risks based on the speed of medication withdrawal? 20
21
Children and adult data were analyzed separately because the literature for 1 group may differ in 22
quality and there are biological differences in the risk of recurrence. We also distinguished 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
6
partial (i.e., focal) from generalized seizures because of differences in the underlying mechanism 1
causing epilepsy. 2
3
Inclusion criteria 4
The author panel included all studies that examined AED weaning in patients with epilepsy who 5
have been seizure-free for at least 12 months and contained data on the risk of seizure 6
recurrence, status epilepticus, medication side effects, quality of life data, or mortality. We 7
included randomized and nonrandomized studies and prospective and retrospective case series 8
published after 1991 that had a control group. We chose to examine studies after 1991 because 9
the previous practice parameter did not contain literature beyond this point.(Quality Standards 10
Subcommittee 1996). 11
12
The following were excluded: 13
1. any articles with fewer than 30 patients; the small sample size would necessarily lead to 14
imprecision of any results and would be underpowered to detect underlying changes, 15
2. articles that were or would be graded as Class IV, 16
3. articles including > 20%, or an unknown amount, of patients who have had a single 17
seizure, rather than epilepsy, 18
4. reviews, systematic reviews, or editorials. 19
20
Outcome criteria 21
We addressed the following outcome measures, measured at 12 months or more: 22
CONFIDENTIAL: Not for dissemination, duplication, or citation
7
1. seizure relapse. These data were divided between children and adults and between those 1
with electroclinical syndromes and epilepsy surgery 2
2. risk factors for either higher or lower rates of seizure recurrence that give odds ratios at 3
the same time points as measured by the chance of seizure freedom 4
3. quality of life data available at the same time points, compared to those who withdrew 5
from medications and those who remained on medications 6
4. occurrence of status epilepticus, compared to those who withdrew from medications and 7
those who remained on medications 8
5. mortality after medication withdrawal, compared to those who withdrew from 9
medications and those who remained on medications 10
11
Variables: 12
We did not prespecify the variables to be considered; instead, we accepted anything the included 13
papers found. We hoped it would be more likely to produce meaningful evidence without 14
prespecified limitations. For any question with 10 or more trials, a funnel plot was preplanned to 15
be performed and looked for publication bias. 16
17
DESCRIPTION OF THE ANALYTIC PROCESS 18
The mission of the AAN Guideline Development, Dissemination, and Implementation (GDDI) 19
Subcommittee is to publish evidence-based recommendations for the diagnosis and treatment of 20
neurologic disorders (see appendices 1 and 2 for a listing of the mission and members of the 21
AAN GDDI). The subcommittee appointed a 5–member panel with special expertise in epilepsy, 22
in AAN guideline development, or both. The panel followed the methods described in the 2011 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
8
AAN guideline development process manual, as amended to include the new practice advisory 1
publication type, an updated classification scheme for therapeutic studies, and a change in the 2
order of steps for document development. A methodologist searched MEDLINE, CINAHL, 3
EBSCO, CENTRAL for relevant articles published between January of 1991 and April of 2013 4
(see appendix 3 for search strategies). 5
6
The original search yielded 2,148 abstracts, which were reviewed for inclusion by at least 2 7
panel members. 154 abstracts seemed promising, and the full text papers were reviewed for 8
inclusion. During full text review, there was evidence for an additional question regarding the 9
speed of withdrawal (#6), so it was added to the list of questions since it was germane to the 10
topic. Of those 154 papers, 13 were rated by 2 authors to be above Class IV and meet inclusion 11
criteria. When articles appeared to include subsets of patients who were incorporated in previous 12
publications, all known information (so long as the paper was rated above Class IV) was 13
included to provide the most comprehensive information possible. This stipulation added 2 more 14
papers. When assigning confidence in evidence, studies with multiple papers examining the same 15
or parts of the same cohorts were counted once. Two panel members rated each of these articles 16
according to the applicable AAN classification of evidence schemes (therapeutic or prognostic; 17
see appendix 4), with 6 explicit clarifications. 18
19
The guideline panel made the following a priori decisions: 20
1. Any articles which address specific electroclinical syndromes were considered separately 21
as electroclinical syndromes have a known time-course. 22
CONFIDENTIAL: Not for dissemination, duplication, or citation
9
2. Articles addressing epilepsy surgery patients were considered separately as they have 1
substantially different characteristics and histories than typical seizure-free patients. 2
3. Unblinded cohorts would be rated Class II, so long as it was explicitly shown to have a 3
careful matching of covariates (as was allowed in the 2004 process manual if the study 4
followed the other characteristics of Class II). This decision was made because there 5
were amendments to the current process in progress that would allow unblinded cohorts 6
to be Class II under specific narrow circumstances. 7
4. For open-label cohort studies, seizure diaries or an explicit statement that patients 8
determined outcome would be enough to qualify for Class III if other characteristics for 9
Class III were fulfilled (as was done in the VNS update, Morris 2013). This decision was 10
based on the fact that seizure recurrence is not necessarily objective (such as auras 11
without motor component); however, seizure diaries represent an outcome determined 12
independently from the investigators. 13
5. Due to the subjectivity of the material, and the broad range of physician and patient 14
preference, the author panel decided that if there were only very low confidence in 15
evidence in non-counseling recommendations, the articles would not go through the 16
modified GRADE/modified Delphi process and the question would be given a U 17
recommendation for insufficient evidence; otherwise, panel authors would be creating a 18
consensus-only recommendation. 19
20
There was 1 post hoc decision imposed by the GDDI subcommittee after a review of the initial 21
manuscript, in which patient-reported GTCS confirmed by a clinical coordinator would be 22
enough to qualify for Class III as this would be considered adequate for a reliable patient-23
CONFIDENTIAL: Not for dissemination, duplication, or citation
10
reported outcome. Appendix 5 describes the process followed to determine the strength of 1
confidence in the evidence. 2
3
The strength of recommendations (A, B, C, and U; appendix 6) was determined by several 4
factors: the strength of the evidence based on the number of Class I, II, and III studies; the AAN 5
modified GRADE process; and the modified Delphi process. The modified GRADE process was 6
performed with 2 of the coauthors (D.G. and J.V.) and reviewed by the panel. 7
8
ANALYSIS OF EVIDENCE 9
For patients with epilepsy on AED medication, who have been seizure-free for at least 12 10
months, does stopping AEDs, compared to not stopping increase the risk of recurrence, and 11
are there risk factors for seizure recurrence? 12
13
Adults 14
One Class I study (Lossius 2008) and 3 Class III studies (Specchio 2002, MRC 1991, Chadwick 15
1996) examined this question. The Class I study (Lossius 2008) enrolled 160 adults who had 16
been seizure-free for 2 years. Participants were randomized to AED withdrawal (with placebo) or 17
AED continuance. The risk of relapse (i.e., risk of failure of seizure-freedom) during the 12 18
months of the study was not significantly different between groups: 15% of the withdrawal group 19
vs 7% of the non-withdrawal group (RR 2.46, 95% CI 0.85–7.08, p = 0.95) experienced seizures 20
following medication withdrawal. The follow-up portion of the study, which looked at a median 21
of 49 months of follow-up, was Class IV; therefore, these results were not reviewed. The period 22
CONFIDENTIAL: Not for dissemination, duplication, or citation
11
of follow-up in the Class I portion of the study was too short to be certain of the generalizability 1
of the paper due to the surprisingly low numbers which were not seen in other papers. 2
3
One Class III study was a nonrandomized, mostly adult, cohort trial where participants and their 4
caregivers decided whether or not to withdraw medication (Specchio 2002) The participants of 5
the study were epilepsy patients who were seizure-free for at least 2 years and on stable 6
monotherapy. At 60 months, the chance of remaining seizure-free was 68% (95% CI 62%–74%) 7
among participants who continued treatment and 48% (95% CI 38%–57%) among participants 8
who did not. During the 60 months of follow-up, after multivariate adjustment, the hazard ratio 9
(HR) for seizure relapse of participants in whom medication was withdrawn was 2.9 (95% CI 10
1.8–4.6). 11
12
One Class III study randomized 1,013 participants who were seizure-free for at least 2 years and 13
were taking AEDs to continued AED treatment vs slow withdrawal (MRC 1991). These data 14
were not separable between GTCSs and other seizure types, but it was decided to include the 15
study because a majority of recurrences were GTCSs (74%), and thus considered objective. The 16
study primarily included adults (median age was 26–27 and medians were presented by group; 17
however, 25% of participants were 16–17 at enrollment). Data were not presented separately 18
between adults and children; therefore, the study was used to inform the adult question. The 19
baseline groups were not equivalent. For example, those with a history of attempted AED 20
withdrawal (OR 0.6, 95% 0.5–0.8), as well as those with a driving license, were less likely to be 21
randomized (OR 0.13, 95% CI 0.1–0.18), and those with special schooling were more likely to 22
be randomized (OR 5.4, 95% CI 3.5–9.4). More than 99% of participants were followed until 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
12
death or within 6 months of 1989. In the withdrawal group, there were 221 recurrences 1
(participants who had a seizure during the period of the study) out of 510 participants (43%), 2
whereas there were 133 recurrences out of 503 participants with continued therapy (26%), 3
yielding an odds ratio (OR) of 2.12 (95% CI 1.63–2.77, p < 0.0001). 4
5
There was another Class III study that analyzed recurrences (Chadwick 1996), with 168 6
recurrences out of 221 participants in the withdrawal group, and 100 recurrences out of 133 7
participants in the medication continuation group having multiple recurrences. There was not a 8
statistically significant difference in risk of more than 1 seizure if there was a single recurrence 9
(OR 1.04, 95% CI 0.63–1.72, p < 0.86). The study concluded that although the risk of seizures 10
increases in the first 1–2 years, there is no evidence for a difference in long-term outcome. 11
12
Conclusions 13
There is insufficient evidence to support or refute that in adults who have been seizure-free for 2 14
years, the risk of recurrence in the first year of weaning has a point estimate of 2.5 times greater 15
recurrence in those who taper vs those who do not (15% vs 7%), although this difference is not 16
significantly different. The strength of evidence was downgraded due to imprecision, and a 17
second downgrade occurred due to a lack of generalizability (very low confidence). In the long 18
term (24–60 months), the risk of recurrence is possibly significantly higher among those who 19
taper medications (low confidence). 20
21
Children 22
CONFIDENTIAL: Not for dissemination, duplication, or citation
13
There were 2 Class II and 2 Class III studies that addressed this question. In 1 Class II study, 57 1
patients (mean age 9.5) who had seizure control for at least 2 years were tapered either at 2 years 2
or 4 years (Serra 2005). The Kaplan-Meier survival curve did not demonstrate significant 3
differences over the 54 months of follow-up. In the other Class II study, 149 participants (mean 4
age of 11 years) who had been seizure-free for at least 18 months were randomized into tapering 5
at 2 years or 4 years (Tennison 1994). Kaplan-Meier survival curves over the 300 weeks of 6
follow-up were not significantly different between the 2 groups. In the Class III study, 7
participants who had been seizure-free for 18 months were randomized to have medication 8
tapered immediately or to wait an additional 6 months (taper at 24 months; mean age taper at 18 9
months was 6.7 years old, 24 months was 5.8 years old) (Gabremariam 1999). There was no 10
significant difference for recurrence risk during the follow-up period, although the average 11
duration of follow-up was different for the 2 groups. During the period of follow-up, 12 of 41 12
(29%) who tapered at 18 months had recurrence during their 38 months of mean follow-up; 14 of 13
39 (36%) who tapered at 24 months had recurrence during their 24 months of mean follow-up, 14
yielding an RR of 0.815 (95% CI 0.432–1.537). A second Class III study of 238 participants 15
(mean age of 8.8 years), who were randomized to treatment for 1 or 3 years; both groups were 16
followed for 5 years (Braathan 1996). After correcting for multiple comparisons (Bonferroni 17
correction), there were no significant differences in either the percent of participants seizure-free 18
during the last 6 months of observation (72% vs 84%, RR 0.857, 95% OR 0.740–0.991, p > 19
0.05), or the percent of participants seizure-free during the entire follow-up (32% vs 41%, RR 20
0.775, 95% CI 0.539–1.115, p > 0.05). In this trial, 8% of participants continued to have some 21
seizures with treatment, and 2% became medication resistant, which they defined as more than 1 22
seizure per month. 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
14
1
Conclusions 2
There is probably not a significant difference in tapering medication at 2 years vs 4 years, in 3
pediatric patients who are seizure-free for 18–24 months (moderate confidence in evidence). 4
There is insufficient evidence whether there is a significant difference in the risk of recurrence if 5
tapering medications at 18 vs 24 months in pediatric patients (very low confidence). 6
7
Electroclinical syndromes 8
Electroclinical syndromes are ones with some or all of the following characteristics: specified 9
age range of onset, specific developmental changes, specific physical characteristics, specific 10
provoking/triggering factors for their seizures, or specific EEG features (Berg 2010). 11
There were no included studies above Class IV addressing this question. 12
13
Epilepsy surgery patients 14
A single Class III study addressed this question (Kerling 2009). This study did not demonstrate 15
significant differences in remaining seizure-free between participants who had undergone 16
surgery and were seizure-free for at least a year who had undergone medication taper vs those 17
who remained on medications at both 1 and 4 years (31/34 [91%] vs 20/26 [77%] remaining 18
seizure-free, RR 1.185, 95% CI 0.937–1.499, p > 0.05). 19
20
Conclusion 21
CONFIDENTIAL: Not for dissemination, duplication, or citation
15
There is insufficient evidence to support or refute that the rate of seizure recurrence with 1
medication withdrawal at 1 vs 4 years is not significantly different than maintaining patients on 2
medication (very low confidence). 3
4
For patients with epilepsy on AED medication, who have been seizure-free for at least 12 5
months, does stopping AEDs, compared to not stopping, increase the risk of status 6
epilepticus? 7
There were 1 Class I study and 4 Class III cohort studies that addressed this question. The 8
previously discussed Class I study (Lossius 2008), did not find any significant predictors for the 9
risk of status epilepticus (after Bonferroni correction); the study authors looked at age, gender, 10
age at epilepsy onset, partial vs generalized epilepsy, MRI findings, duration of seizure-freedom, 11
specific AEDs, and a normal neurological examination, in trying to find predictors for status 12
epilepticus. 13
14
Of the Class III studies, 1 study was in a mixed cohort of mostly adults (Specchio 2002). In this 15
paper, the following factors were significant for increased risk of recurrence: 2 years of 16
remission at study entry OR 2.6 (95% CI 1.5-4.8) and abnormal psychiatric examination OR 2.1 17
(95% CI 1.3–3.6). Duration of active disease, epilepsy syndrome, and abnormal CT/MRI were 18
not significant factors for an increased risk of recurrence. 19
20
One Class III Medical Research Council (MRC) study looked at whether there was a difference 21
among antiepileptic drugs (Chadwick 1999). This study found that monotherapy with valproate 22
(HR 1.97, 95% CI 1.29–3.0), phenobarbitone and primidone (HR 3.55, 95% CI 1.24–10.2), and 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
16
phenytoin (HR, 3.02 95% CI 1.84–4.97) were associated with a significant risk for seizure 1
recurrence with medication withdrawal, while this was not true for carbamazepine (HR 1.32, 2
95% CI 0.85–2.0). A second Class III MRC study created a risk index for recurrence (MRC 3
1993). Using a Cox proportional hazards regression, they found 7 prognostic factors for 4
increased risk of seizures: age 16 and older (RR 1.75, 95% CI 1.30–2.35), use of > 1 AED (risk 5
ratio [RR] 1.83, 95% CI 1.40–2.39), a history of seizures after starting an antiepileptic drug (RR 6
1.56, 95% CI 1.19–2.04), a history of tonic-clonic seizures (RR 1.56, 95% CI 1.09–2.22), history 7
of myoclonic seizures (RR 1.84, 95% CI 1.13–3.01), and an abnormal EEG in the last year (RR 8
1.32, 95% CI 1.01–1.73). 9
10
There were 3 Class III prognostic studies of children. One Class III study (Gabremariam 1999) 11
found that an abnormal EEG at discontinuation was associated with seizure recurrence in 12
children (RR 6.21, 95% CI 5.62–68.5). An abnormal EEG was defined as one with spikes, sharp 13
waves, paroxysmal slowing, or non-paroxysmal abnormalities. There were 2 Class III prognostic 14
studies of the same cohort (Braathan 1997, Andersson 1997). The first (Braathan 1997) 15
considered more than 20 factors; after Boferonni correction, there were no statistically 16
significant factors. The second (Andersson 1997) looked more specifically at EEG and found 17
that among both groups of children (1-year and 3-year withdrawal), there was a higher risk of 18
recurrence in children with epileptiform activity on the EEG (24 of 51 with epileptiform activity 19
(47%) vs 31 of 94 without epileptiform activity (33%), yielding an OR of 2.87 (95% CI 1.35–20
6.11, p = 0.006). 21
22
Conclusions 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
17
There is insufficient evidence to support or refute that a variety of risk factors demonstrate 1
change in chance seizure recurrence (very low confidence). An epileptiform EEG possibly 2
increases the risk of recurrence in children (low confidence). Neither of the included studies 3
specify what kind of EEG was performed (e.g., length of study, sleep deprived), so we are unable 4
to comment as to what kind of EEG is needed. 5
6
For patients with epilepsy on AED medication, who have been seizure-free for at least 12 7
months, does stopping AEDs, compared to not stopping, reduce medication-related side 8
effects? Is there quality of life data with medication withdrawal? 9
There was 1 Class I study which addressed this question (Lossius 2008). They did not find 10
significant differences in 3 measures of quality of life, including the quality of life in epilepsy 11
inventory-89 (mean score difference 0.3, 95% CI -1.55 to 2.07), between participants who 12
withdrew medications and those who did not. The lack of results in quality of life indicators may 13
be, in part, because among the 15 neuropsychiatric measures tested, they only found significant 14
differences (after Bonferroni correction) in lexical choice reaction time (mean difference -24.42 15
(-40.11 to -9.14, p = 0.045). For lexical choice reaction time, the mean may not be the best 16
measure and could possibly be misleading given that this distribution typically has a significant 17
tail and there was no test shown fitting to the Gaussian distribution (Hauk 2012). There were 18
subsets of the same cohort reviewed for other neuropsychiatric measures, and when 95% Cis 19
were constructed, there were no convincingly significant differences (Hessen 2006, Hessen 20
2007). 21
22
Conclusion 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
18
In adults who are seizure-free, medication weaning possibly does not change in quality of life 1
(low confidence). These data were imprecise, so the confidence in evidence was downgraded. 2
3
For patients with epilepsy on AED medication, who have been seizure-free for at least 12 4
months, does stopping AEDs, compared to not stopping, change quality of life? Is there 5
any information about the risk of status epilepticus? 6
No participants in the medication withdrawal arm of the 1-year adult Class I randomized 7
controlled trial had status epilepticus (Lossius 2008). Most studies did not specifically mention 8
any information on this. 9
10
Conclusion 11
Medication withdrawal possibly may not increase the risk of status epilepticus in adults. 12
Confidence in the evidence is also lowered due to imprecision (low confidence). 13
14
For patients with epilepsy on AED medication, who have been seizure-free for at least 12 15
months, does stopping AEDs, compared to not stopping, change the risk of mortality? 16
Only 2 studies specifically mentioned mortality. During the 1-year adult Class I trial of 17
medication withdrawal, there were no deaths (Lossius 2008). During the 6 years of follow-up of 18
the Class III MRC cohort study (MRC 1991), 2 participants who continued therapy died, most 19
likely from a seizure. 20
21
Conclusion 22
CONFIDENTIAL: Not for dissemination, duplication, or citation
19
There is insufficient evidence to support or refute that medication withdrawal may not cause an 1
increase in the risk of mortality in adults (very low confidence). 2
3
For patients with epilepsy on AED medication, who have been seizure-free for at least 12 4
months, does stopping AEDs, compared to not stopping, change any of the above risks 5
based on the speed of medication withdrawal? 6
Two Class II studies specifically addressed this question in children. In 1 Class II study, 57 7
patients had medication tapered either at 25% every 10 days or 25% every 2 months (Serra 8
2005). Kaplan-Meier survival curves over 54 months of follow-up were not significantly 9
different between groups. In the other Class II study, medications were tapered at a rate of 25% 10
every 2 weeks or 25% every 2 months (Tennison 1994). No significant differences were present 11
in the Kaplan-Meier survival curves across the 300 weeks of follow-up. 12
13
Conclusion 14
Withdrawal of medication at a rate of 25% every 10 days to 2 weeks is probably not significantly 15
different than withdrawal at a rate of 25% every 2 months, in children who are seizure-free, in > 16
4 years of follow-up (moderate confidence). 17
18
PRACTICE RECOMMENDATIONS 19
Adults 20
Rationale 21
There is low confidence that in the long term (24–60 months), the risk of recurrence is 22
significantly higher among those who taper medications (EVID). Once epilepsy is masked, it is 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
20
unknowable if you have epilepsy or not (PRIN). Patients should be allowed to be part of the 1
medical decision-making process, especially when there is clinical equipoise (PRIN). 2
3
While there is evidence for the predictive power of EEGs in children (EVID), there is no 4
evidence above Class IV in adults. Moreover, the evidence in children cannot be used as related 5
evidence in adults as it is based on Class III data. The same applies for the small chance of 6
medication resistance seen after medication withdrawal. These differences may relate to 7
biological differences between children and adults (PRIN). 8
9
To enable patients to make decisions, all the clinically relevant information should be made 10
available (PRIN). Low-quality evidence shows no difference in quality of life between patients 11
with well-controlled epilepsy who stop or continue their medications (EVID). Factors 12
contributing to quality of life are likely to be highly individual and may include ease of 13
medication administration (e.g., dose frequency), experience of side effects, seizure recurrence, 14
and comorbidities (PRIN). In the 1 year of 1 trial, there were no deaths, and in the 6 years of the 15
other trial, the only deaths that occurred were in the patients that continued their medications 16
(EVID). There does not seem to be an increased risk of status epilepticus in patients who are 17
seizure-free for 2 years who withdrew their AEDs; however, the risk of status epilepticus may be 18
small in the cohorts of the study, and there may not be enough patients and time to detect a 19
difference (EVID). There is no evidence that 2 years has special significance (EVID). 20
21
Recommendations 22
CONFIDENTIAL: Not for dissemination, duplication, or citation
21
In adults who are seizure-free for at least 2 years, there [may/should/must] be a discussion 1
between the clinician and the patient and their caregivers, if any, about the risks and benefits of 2
medication withdrawal, which specifically includes and documents: 3
1. there is higher seizure recurrence in patients who had medication withdrawal, and 4
2. that if seizures recur during or after withdrawal, there is a small chance they will no 5
longer respond to medications (recommendation level not yet determined). 6
This recommendation is meant to be an expectation of the physician if appropriate for the 7
patient. 8
9
Clinicians [may/should/must] discuss with patients that medication withdrawal can be 10
contemplated because there is a statistically significant difference between withdrawal and no 11
withdrawal, noting that there is significant uncertainty with the evidence (recommendation level 12
not yet determined). 13
14
Counseling should include discussion that there is not strong evidence regarding changes in the 15
risk of mortality and status epilepticus, and, as such, these risks have not been excluded by the 16
evidence (Level B). 17
18
Clinicians must counsel that recurrent seizures put people at risk for status epilepticus and death 19
(Level A). 20
Clinicians should explore contributors to individual patients’ quality of life as part of shared 21
decision making regarding AED discontinuation (Level B). 22
23
CONFIDENTIAL: Not for dissemination, duplication, or citation
22
Clinicians may discuss with seizure-free patients that there is not strong prognostic evidence that 1
EEG or imaging study abnormalities are predictive of recurrence in patients (Level C). 2
3
Clinicians [may/should/must] discuss with epilepsy surgery patients that there is not enough data 4
to base decisions on medication withdrawal (recommendation level not yet determined). 5
6
Children 7
Rationale 8
There does not appear to be a statistically significant difference when medication is withdrawn in 9
pediatric patients who have been seizure free for 2 years vs 4 years when patients have been 10
seizure-free for 18–24 months during the first 4–6 years of follow-up (EVID). While the cohorts 11
were broad, they do not include large numbers of children with electroclinical syndromes. When 12
there is not significant difference between treatment and lack of treatment over long periods of 13
time, lack of treatment may be the preferred option (PRIN). There is a small risk of becoming 14
medication-resistant with medication withdrawal (EVID). 15
16
There is low confidence in evidence that an epileptiform EEG increases the risk of recurrence in 17
children (EVID). 18
19
Patients and families of children who are seizure-free and contemplating medication withdrawal 20
would want any information about the withdrawal process that is available (PRIN). The evidence 21
suggests there is not a significant difference is weaning 25% every 10 days to 2 weeks vs 2 22
months (EVID). 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
23
1
Recommendations 2
In children who are seizure-free for at least 18–24 months, who do not have an electroclinical 3
syndrome suggesting otherwise, there should be a discussion about the risks and benefits of 4
medication withdrawal, which specifically includes and documents that if seizures recur during 5
either withdrawal or after withdrawal, there is a small chance they will no longer respond to 6
medication (Level B). This recommendation is meant to be an expectation for the physician, if it 7
is appropriate. 8
9
Clinicians should discuss with patients that medication withdrawal can be considered because 10
withdrawal of AEDs does not clearly increase risk of recurrence (Level B). 11
12
Counseling should include discussion that there is not good evidence regarding changes in the 13
risk of mortality and status epilepticus, and, as such, these risks have not been excluded by the 14
evidence (Level B). 15
16
Clinicians should counsel that recurrent seizures put people at risk for status epilepticus and 17
death (Level B). 18
19
Clinicians should explore contributors to individual patients’ quality of life as part of shared 20
decision-making regarding AED discontinuation (Level B). 21
22
CONFIDENTIAL: Not for dissemination, duplication, or citation
24
If there is agreement between the physician and patient’s family to pursue consideration of AED 1
medication withdrawal, an EEG should be ordered [Level B]. 2
3
If the EEG does not show epileptiform activity, medication withdrawal [may/should/must] be 4
offered, at a rate no faster than 25% every 10–14 days. 5
6
Areas in which there is no low to moderate risk of bias evidence 7
Clinicians [may/should/must] take into account the known natural history of the specific 8
electroclinical syndrome when counseling about medication withdrawal (recommendation level 9
not yet determined). 10
11
There is neither evidence for, nor against, tapering medication for seizure-free patients with 12
structural etiologies (epilepsy surgery patients) (Level U). 13
14
RECOMMENDATIONS FOR FUTURE RESEARCH 15
There is only a single low risk of bias study in this area, and that one is inadequately powered as 16
well as not clearly generalizable. Future areas of study include the many gaps shown by this 17
guideline. High quality studies are needed which answer: 18
1. Is there a significant difference in medication withdrawal versus remaining on 19
medications in adults? 20
2. Is an EEG a relevant prognostic factor in medication withdrawal in adults? 21
3. Does modern MRI imaging have prognostic relevance? 22
CONFIDENTIAL: Not for dissemination, duplication, or citation
25
4. Are there other prognostic factors which help decide who should or should not have 1
medication withdrawn? 2
5. Is there a certain speed of medication withdrawal in adults which should be 3
recommended? 4
6. What happens to people who recur: are there additional risks for them? 5
Ideally a registry could be formed, which could have cohorts large enough to look at and 6
examine if there are additional risks of status epilepticus or mortality. 7
8
9
10
CONFIDENTIAL: Not for dissemination, duplication, or citation
26
APPENDICES 1
Appendix 1: AAN GDDI mission 2
The mission of the GDDI is to develop, disseminate, and implement evidence-based systematic 3
reviews and clinical practice guidelines related to the causation, diagnosis, treatment, and 4
prognosis of neurologic disorders. The GDDI is committed to using the most rigorous methods 5
available within its budget, in collaboration with other available AAN resources, to most 6
efficiently accomplish this mission. 7
8
CONFIDENTIAL: Not for dissemination, duplication, or citation
27
Appendix 2: AAN GDDI Subcommittee members 1
The AAN has structured its subcommittee overseeing guideline development in several ways in 2
recent years. The GDDI was first formed in 2014; it existed under a previous name and structure 3
when this guideline project was inaugurated. At the time this guideline was approved to advance 4
beyond subcommittee development, the subcommittee was constituted as below. 5
6
Cynthia Harden, MD (Chair); Steven R. Messé, MD (Co-Vice-Chair); Sonja Potrebic, MD, PhD 7
(Co-Vice-Chair); Stephen Ashwal, MD; Lori L. Billinghurst, MD; Brian Callaghan, MD; 8
Gregory S. Day, MD, MSc; Diane Donley, MD; Richard M. Dubinsky, MD, MPH; Jeffrey 9
Fletcher, MD; Gary S. Gronseth, MD (Senior Evidence-based Medicine Methodology Expert); 10
Michael Haboubi, DO; John J. Halperin, MD; Yolanda Holler-Managan, MD; Annette M. 11
Langer-Gould, MD, PhD; Nicole Licking, DO; Mia T. Minen, MD; Pushpa Narayanaswami, 12
MBBS, DM; Maryam Oskoui, MD; Alejandro A. Rabinstein, MD; Alexander Rae-Grant, MD; 13
Kevin Sheth, MD; Kelly Sullivan, PhD; Eric J. Ashman, MD (Ex-Officio); Jacqueline French, 14
MD (Ex-Officio, Guideline Process Historian) 15
16
17
CONFIDENTIAL: Not for dissemination, duplication, or citation
28
Appendix 3: Complete search strategy 1
1. The 17 articles included in the original guideline’s table will be examined for possible 2
inclusion (AAN 1996). 3
2. Both the included and the excluded articles from the Cochrane review will be examined for 4
possible inclusion (Sirven 2012). 5
3. We will search from 1991 onward with the following criteria for MEDLINE (adapted from 6
Sirven 2012, being more inclusive): 7
1. exp animals/ not humans.sh. 8
2. not 1 9
3. exp Epilepsy/ 10
4. Seizures/ 11
5. (epilep$ or seizure$ or convuls$).tw. 12
6. exp Anticonvulsants/ 13
7. (anticonvulsant$ or antiepilep$).tw. 14
8. 3 or 4 or 5 or 6 or 7 15
9. (relapse or recurrence).ti,ab. 16
10. (remission and epilep$).ti,ab. 17
11. (discontinu$ or withdrawal).ti,ab. 18
12. (prognosis and epilep$).ti,ab. 19
13. 9 or 10 or 11 or 12 20
14. 2 and 8 and 13 21
22
CONFIDENTIAL: Not for dissemination, duplication, or citation
29
We will search CENTRAL, DARE, and CINAHL with analogous searches. Article references 1
will be considered for inclusion. Additional articles that experts are aware of will also be 2
considered. 3
4
Two panel members will review the abstracts from the searches. Any disagreement will be 5
resolved by mutual agreement. The selected abstracts will then be reviewed in full text format to 6
determine if they have any of the outcomes we are looking for. 7
8
The papers which have been selected for possible inclusion, as well as the papers from the old 9
practice parameter, and both the included and excluded papers from the Cochrane review will be 10
graded independently by 2 panel members. Any disagreement in grading will be resolved by 11
mutual agreement. 12
13
CONFIDENTIAL: Not for dissemination, duplication, or citation
30
Appendix 4. AAN rules for classification of evidence for risk of bias 1
2
Prognostic accuracy scheme 3
Class I 4
A cohort study of a broad spectrum of persons at risk for developing the outcome (e.g., target 5
disease, work status). The outcome is defined by an acceptable reference standard for case 6
definition. The outcome is objective or measured by an observer who is masked to the presence 7
of the risk factor. Study results allow calculation of measures of prognostic accuracy. 8
Class II 9
A case-control study of a broad spectrum of persons with the condition compared with a broad 10
spectrum of controls, or a cohort study of a broad spectrum of persons at risk for the outcome 11
(e.g., target disease, work status) where the data were collected retrospectively. The outcome is 12
defined by an acceptable reference standard for case definition. The outcome is objective or 13
measured by an observer who is masked to the presence of the risk factor. Study results allow 14
calculation of measures of prognostic accuracy. 15
Class III 16
A case-control study or a cohort study where either the persons with the condition or the controls 17
are of a narrow spectrum where the data were collected retrospectively. The outcome is defined 18
by an acceptable reference standard for case definition. The outcome is objective or measured by 19
an observer who did not determine the presence of the risk factor. Study results allow calculation 20
of measures of a prognostic accuracy. 21
Class IV 22
Studies not meeting Class I, II, or III criteria, including consensus, expert opinion, or a case 23
report. 24
CONFIDENTIAL: Not for dissemination, duplication, or citation
31
1
Therapeutic scheme 2
Class I 3
A randomized controlled clinical trial of the intervention of interest with masked or objective 4
outcome assessment, in a representative population. Relevant baseline characteristics are presented 5
and substantially equivalent between treatment groups, or there is appropriate statistical adjustment 6
for differences. 7
The following are also required: 8
a. concealed allocation 9
b. no more than 2 primary outcomes specified 10
c. exclusion/inclusion criteria clearly defined 11
d. adequate accounting for dropouts (with at least 80% of enrolled subjects completing the study) and 12
crossovers with numbers sufficiently low to have minimal potential for bias. 13
e. For noninferiority or equivalence trials claiming to prove efficacy for one or both drugs, the 14
following are also required*: 15
i. The authors explicitly state the clinically meaningful difference to be excluded by 16
defining the threshold for equivalence or noninferiority. 17
ii. The standard treatment used in the study is substantially similar to that used in 18
previous studies establishing efficacy of the standard treatment (e.g., for a drug, the 19
mode of administration, dose, and dosage adjustments are similar to those previously 20
shown to be effective). 21
iii. The inclusion and exclusion criteria for patient selection and the outcomes of patients 22
on the standard treatment are comparable to those of previous studies establishing 23
efficacy of the standard treatment. 24
CONFIDENTIAL: Not for dissemination, duplication, or citation
32
iv. The interpretation of the study results is based upon a per-protocol analysis that 1
accounts for dropouts or crossovers. 2
f. For crossover trials, both period and carryover effects examined and statistical adjustments 3
performed, if appropriate 4
Class II 5
An RCT of the intervention of interest in a representative population with masked or objective 6
outcome assessment that lacks one criteria a–e above (see Class I) or a prospective matched cohort 7
study with masked or objective outcome assessment in a representative population that meets be 8
above (see Class I). (Alternatively, a randomized crossover trial missing 1 of the following 2 9
characteristics: period and carryover effects described or baseline characteristics of treatment order 10
groups presented.) All relevant baseline characteristics are presented and substantially equivalent 11
among treatment groups, or there is appropriate statistical adjustment for differences. 12
Class III 13
All other controlled trials (including studies with external controls such as well-defined natural 14
history controls). (Alternatively, a crossover trial missing both of the following 2 criteria: period and 15
carryover effects described or baseline characteristics of treatment order groups presented.) A 16
description of major confounding differences between treatment groups that could affect outcome.** 17
Outcome assessment is masked, objective, or performed by someone who is not a member of the 18
treatment team. 19
Class IV 20
Studies that (1) did not include patients with the disease, (2) did not include patients receiving 21
different interventions, (3) had undefined or unaccepted interventions or outcomes measures, or (4) 22
had no measures of effectiveness or statistical precision presented or calculable. 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
33
*Note that numbers 1–3 in Class Ie are required for Class II in equivalence trials. If any 1 of the 3 is 1
missing, the class is automatically downgraded to Class III. 2
**Objective outcome measurement: an outcome measure that is unlikely to be affected by an 3
observer’s (patient, treating physician, investigator) expectation or bias (e.g., blood tests, 4
administrative outcome data). 5
6
CONFIDENTIAL: Not for dissemination, duplication, or citation
34
Appendix 5. Rules for determining confidence in evidence 1
Modal modifiers used to indicate the final confidence in evidence in the conclusions 2
o High confidence: highly likely or highly probable 3
o Moderate confidence: likely or probable 4
o Low confidence: possibly 5
o Very low confidence: insufficient evidence 6
Initial rating of confidence in the evidence for each intervention outcome pair 7
o High: requires 2 or more Class I studies 8
o Moderate: requires 1 Class I study or 2 or more Class II studies 9
o Low: requires 1 Class II study or 2 or more Class III studies 10
o Very low: requires only 1 Class III study or 1 or more Class IV studies 11
Factors that could result in downgrading confidence by 1 or more levels 12
o Consistency 13
o Precision 14
o Directness 15
o Publication bias 16
o Biological plausibility 17
Factors that could result in downgrading confidence by 1 or more levels or upgrading 18
confidence by 1 level 19
o Magnitude of effect 20
o Dose response relationship 21
o Direction of bias 22
23
24
CONFIDENTIAL: Not for dissemination, duplication, or citation
35
Appendix 6. Steps and rules for formulating recommendations 1
2
Constructing the recommendation and its rationale 3
4
Rationale for recommendation summarized in the rationale includes 3 categories of 5
premises 6
Evidence-based conclusions for the systematic review 7
Stipulated axiomatic principles of care 8
Strong evidence from related conditions not systematically reviewed 9
10
Actionable recommendations include the following mandatory elements 11
The patient population that is the subject of the recommendation 12
The person performing the action of the recommendation statement 13
The specific action to be performed 14
The expected outcome to be attained 15
16
Assigning a level of obligation 17
18
Modal modifiers used to indicate the final level of obligation (LOO) 19
Level A: Must 20
Level B: Should 21
Level C: May 22
Level U: No recommendation supported 23
24
LOO assigned by eliciting panel members’ judgments regarding multiple domains, using 25
a modified Delphi process. Goal is to attain consensus after a maximum of 3 rounds of 26
voting. Consensus is defined by: 27
> 80% agreement on dichotomous judgments 28
>80% agreement, within 1 point for ordinal judgments 29
If consensus obtained, LOO assigned at the median. If not obtained, LOO 30
assigned at the 10th percentile 31
32
Three steps used to assign final LOO 33
34
1. Initial LOO determined by the cogency of the deductive inference supporting the 35
recommendation on the basis of ratings within 4 domains. Initial LOO anchored 36
to lowest LOO supported by any domain. 37
Confidence in evidence. LOO anchored to confidence in evidence 38
determined by modified form of the Grading of Recommendations 39
Assessment, Development and Evaluation process 40
Level A: High confidence 41
Level B: Moderate confidence 42
Level C: Low confidence 43
Level U: Very low confidence 44
CONFIDENTIAL: Not for dissemination, duplication, or citation
36
Soundness of inference assuming all premises are true. LOO anchored to 1
proportion of panel members convinced of soundness of the inference 2
Level A: 100% 3
Level B: ≥ 80% to < 100% 4
Level C: ≥ 50% to < 80% 5
Level U or R: < 50% 6
Acceptance of axiomatic principles: LOO anchored to proportion of panel 7
members who accept principles 8
Level A: 100% 9
Level B: ≥ 80% to < 100% 10
Level C: ≥ 50% to < 80% 11
Level U or R: < 50% 12
Belief that evidence cited from rerated conditions is strong: LOO anchored 13
to proportion of panel members who believe the related evidence is strong 14
Level B: ≥ 80% to 100% (recommendations dependent on 15
inferences from nonsystematically reviewed evidence cannot be 16
anchored to a Level A LOO) 17
Level C: ≥ 50% to < 80% 18
Level U or R: < 50% 19
20
2. LOO is modified mandatorily on the basis of the judged magnitude of benefit 21
relative to harm expected to be derived from complying with the recommendation 22
Magnitude relative to harm rated on 4-point ordinal scale 23
Large benefit relative to harm: benefit judged large, harm judged 24
none 25
Moderate benefit relative to harm: benefit judged large, harm 26
judged minimal; or benefit judged moderate, harm judged none 27
Small benefit relative to harm: benefit judged large, harm judged 28
moderate; or benefit judged moderate, harm judged minimal; or 29
benefit judged small, harm judged none 30
Benefit to harm judged too close to call: benefit and harm judged 31
to be substantially similar 32
Regardless of cogency of the recommendation the LOO can be no higher 33
than that supported by the rating of the magnitude of benefit relative to 34
harm 35
Level A: large benefit relative to harm 36
Level B: moderate benefit relative to harm 37
Level C: small benefit relative to harm 38
Level U: too close to call 39
LOO can be increased by one grade if LOO corresponding to benefit 40
relative to harm greater than LOO corresponding to the cogency of the 41
recommendation 42
43
3. LOO optionally downgraded on the basis of the following domains 44
CONFIDENTIAL: Not for dissemination, duplication, or citation
37
Importance of the outcome: critical, important, mildly important, not 1
important 2
Expected variation in patient preferences: none, minimal, moderate, large 3
Financial burden relative to benefit expected: none, minimal, moderate, 4
large 5
Availability of intervention: universal, usually, sometimes, limited 6
7
CONFIDENTIAL: Not for dissemination, duplication, or citation
38
REFERENCES 1
AAN. A guideline for discontinuing antiepileptic drugs in seizure-free patients-Summary 2
statement. Neurology 1996; 47: 600-602. 3
Andersson T, Braahan G, Persson A, Theorell K. A Comparison Between One and Three Years 4
of Treatment in Uncomplicated Childhood Epilepsy: A prospective Study. II. The EEG as 5
Predictor of Outcome After Withdrawal of Treatment. Epilepsia 1997; 38(2): 225-232. 6
Berg AT, Berkovic SF, Brodie MJ, et al Revised terminology and concepts for organization of 7
seizure and epilepsies: Report of the ILAE Commission n Classification and Terminology, 2005-8
2009. Epilepsia 2010; 51(4): 676-685. 9
Braathan G, Andersson T, Gyljie, et al. Comparison Between One and Three Years of Treatment 10
in Uncomplicated Childhood Epilepsy: A Prospective Study. I. Outcome in Different Seizure 11
Types. Epilepsia 1996; 822-832. 12
Braathan G, Melander H. Early discontinuation of treatment in children with uncomplicated 13
epilepsy: A prospective study with a model for prediction of outcome. Epilepsia 38 (5): 561-569. 14
Chadwick D, representing the MRC Antiepileptic Drug Withdrawal Study Group. Does 15
withdrawal of different antiepileptic drugs have different effects on seizure recurrence? Further 16
results of the MRC Antiepileptic Drug Withdrawal Study. Brain 1999; 122: 441-448. 17
Chadwick D, Taylor J, Johnson T. on behalf of the MRC Antiepileptic Drug Withdrawal Group. 18
Epilepsia 37 (11): 1043-1050. 19
Fisher RS, Acevedo C, Arzimanoglou A, Bogacz A, Cross JH, Elger CE, et al. A practical 20
clinical definition of epilepsy. Epilepsia 2014; 55(4):475-483. 21
Gabremariam A, Megesha W, Enqusilassie F. Disconitnuing anti-epileptic medication(s) in 22
epileptic children: 18 versus 24 months. Ann Trop Med Ped 1999; 19: 93-99. 23
CONFIDENTIAL: Not for dissemination, duplication, or citation
39
Hauk O, Coutout C, Holden A, Chen Y. The time-course of single-word reading: Evidence from 1
fast behavior and brain responses. Neuroimage 2012; 60 (1-2) 2
: 1462-1477. 3
Hauser WA, Beghi E. First seizure definitions and worldwide incidence and mortality. Epilepsia 4
2008; 49(s1): s8-s12. 5
Hauser WA, Annegers JF, Kurland LT. The incidence of epilepsy and unprovoked seizures in 6
Rochester, Minnesota, 1935- 1984. Epilepsia 34: 453-468. 7
Hessen E, Lossius MI, Reinvang I, Gjerstad. Influence of Major Antiepileptic Drugs on 8
Attention, Raction Time, and Speed of Information Processing: Results from a Randomized, 9
Double-Blind, Placebo-controlled Withdrawal Study of Seizure-free Epilepsy Patients Receiving 10
Monotherapy. Epilepsia 2006; 47(12): 2038-45. 11
Hessen E, Lossius MI, Reinvaug I, Gjerstad L. Slight improvement in mood and irritability after 12
antiepileptic drug withdrawal: A controlled study in patients on monotherapy. Epilepsy & Behav 13
2007; 10: 449-455. 14
Kerling F, Pauli E, Blumcke I, Buchfelder M, Stefan H. Drug withdrawal after successful 15
epilepsy surgery: How safe is it? Epilepsy & Behav 2009; 15: 476-480. 16
Kwan P, Brodie MJ. Early Identification of refractory epilepsy. New England Journal of 17
Medicine 2000;342(5): 314–9. 18
Mohanraj R, Brodie MJ. Diagnosing refractory epilepsy: response to sequential treatment 19
schedules. European Journal of Neurology 2006;13(3):277–82. 20
Medical Research Council Antiepileptic Drug Withdrawal Study Group. Randomised study of 21
antiepileptic drug withdrawal in patients in remission. The Lancet 1991; 337 (8751): 1175-80. 22
CONFIDENTIAL: Not for dissemination, duplication, or citation
40
Medical Research Council Antiepileptic Drug Withdrawal Study Group. Prognostic index for 1
recurrence of seizures after remission of epilepsy. BMJ 1993; 306: 1374 - 8. 2
Morris GL, Gloss D, Buchhalter J, Mack KJ, Nickels K, Harden C. Evidence-based guideline 3
update: Vagus Nerve Stimulation for the treatment of epilepsy: report of the guideline 4
development subcommittee of the American Academy of Neurology. Neurology 2013; 81(16): 5
1453-9. 6
Murray CJL, Lopez JD. Global Comparative Assessment in the Health Sector. 1st Edition. 7
Geneva: World Health Organization, 1994. 8
Quality Standards Subcomiitee of the American Academy of Neurology. Practice Parameter: A 9
guideline for discontinuing antiepileptic drugs in seizure-free patients-Summary Statement. 10
Neurology 1996: 47: 600-602. 11
Serra JG, Montenegro MA, Geurreiro MM. Antiepileptic Drug Withdrawal in Childhood: Does 12
the Duration of Tapering Matter for Seizure Recurrence? J Child Neurol 2005; 20: 624-6. 13
Sirven J, Sperling MR, Wingerchuk DM. Early versus late antiepileptic drug withdrawal for 14
people with epilepsy in remission. Cochrane Database of Systematic Reviews 2012, Issue 3. Art. 15
No.: CD001902. DOI: 10.1002/14651858.CD001902. 16
Specchio LM, Tramacere L, La Neve A, Beghi E. Discontinuing antiepileptic drug in patients 17
who are seizure free on monotherapy. J Neurol Neurosurg Psychiatry 2002; 72: 22-25. 18
Tennison M, Greenwood R, Lewis D, Thorn M. Discontinuing antiepileptic drugs in children 19
with epilepsy. NEJM 1994; 330 (20): 1407-10. 20