Epilepsy & Behavior 14 (2009) 202–209

Contents lists available at ScienceDirect

Epilepsy & Behavior

journal homepage: www.elsevier .com/locate /yebeh

Patient-reported cognitive side effects of antiepileptic drugs: Predictorsand comparison of all commonly used antiepileptic drugs

Hiba Arif a, Richard Buchsbaum b, David Weintraub a, Joanna Pierro a,Stanley R. Resor Jr. a, Lawrence J. Hirsch a,*

a Comprehensive Epilepsy Center, Department of Neurology, College of Physicians and Surgeons, Columbia University, Box NI-135, 710 West 168th Street, New York, NY, USAb Department of Biostatistics, Mailman School of Public Health, Columbia University, New York, NY, USA

a r t i c l e i n f o

Article history:Received 4 April 2008Revised 14 October 2008Accepted 15 October 2008Available online 17 December 2008

Keywords:Antiepileptic drugCognitionTolerability

1525-5050/$ - see front matter � 2008 Elsevier Inc. Adoi:10.1016/j.yebeh.2008.10.017

* Corresponding author. Fax: +1 212 305 1450.E-mail address: ljh3@columbia.edu (L.J. Hirsch).

a b s t r a c t

Subjective cognitive side effects (CSEs) are common in patients taking antiepileptic drugs (AEDs). Theobjective of this study was to predict which patients are at risk for CSEs, and compare the CSE profilesof all commonly used AEDs. In this nonrandomized retrospective study, medical records of 1694 adultoutpatients with epilepsy seen at our center over a 5-year period who had taken one or more AEDs wereexamined. Non-AED predictors of CSEs were investigated, and rates of AED-related CSEs were comparedin 1189 patients (546 on monotherapy) newly started on an AED at our center. The average rate of AED-related intolerable CSEs (leading to dosage change or discontinuation) was 12.8%. On multivariate anal-ysis, no significant non-AED predictors of CSEs were found. Significantly more intolerable CSEs wereattributed to topiramate (21.5% of 130 patients) than to most other AEDs, including carbamazepine(9.9%), gabapentin (7.3%), levetiracetam (10.4%), lamotrigine (8.9%), oxcarbazepine (11.6%), and valproate(8.3%). CSE rates with zonisamide (14.9%) were significantly higher than those for gabapentin and lamo-trigine. After exclusion of CSEs during the first 8 weeks of therapy, rates of CSEs were lower, but relativedifferences remained unchanged. In monotherapy, significantly more intolerable CSEs occurred withtopiramate (11.1% of 18 patients) than with carbamazepine or valproate, and both phenytoin and zonisa-mide were associated with more CSEs than valproate. From this study, it can be concluded that intoler-able patient-reported CSEs are most common with topiramate, followed by zonisamide, phenytoin, andoxcarbazepine. They are least likely to be reported with gabapentin, valproate, lamotrigine, carbamaze-pine, and levetiracetam.

� 2008 Elsevier Inc. All rights reserved.

1. Introduction tion to mood), so it is worthwhile to further explore factors that

Cognitive side effects (CSEs) are commonly seen in patientsundergoing chronic antiepileptic drug (AED) therapy [1,2]. Formany patients, they may be more debilitating than the actual sei-zures themselves and, thus, contribute to a worse quality of life [3].These side effects are well known among the older AEDs [4,5], andrecent studies have begun to focus on the effects of the newer AEDs[6,7].

Although the causes of subjective and objective cognitiveimpairment in patients with epilepsy have not been fully explored,at least four factors are clearly involved: mood (particularly forsubjective CSEs as studied here), underlying etiology of epilepsy,the effects of the seizures themselves, and the central side effectsof AED therapy [8,9]. Most often, these factors are related and con-tribute in varying degrees to the cognitive profile of the individualpatient. Of these, side effects associated with AED therapy may beone of the few potentially preventable tolerability issues (in addi-

ll rights reserved.

predict cognitive impairment.Studies of AED effects on cognition have been conducted mostly

in healthy volunteers and patients with epilepsy, migraine, orbipolar disorder. Most studies in healthy volunteers were unableto investigate adverse effects on cognition occurring with long-term AED use. Carbamazepine and valproate have been demon-strated to be superior to phenytoin in some studies [10–16], butneither carbamazepine nor valproic acid is entirely free of negativecognitive effects, as reported in many other studies [12,17–21].These negative effects were more apparent when carbamazepineand valproic acid were compared with the newer AEDs [22]. Theonly new AEDs known to cause significant CSEs are topiramateand zonisamide: both have diffuse cognitive effects, as well as spe-cific effects on language and memory [17,23–26].

Still, the relative incidence of these side effects is not wellknown, primarily because of the dearth of large-scale studies com-paring many or all of the newer AEDs. The objective of the presentstudy was to determine the relative prevalence and predictors ofsubjective CSEs in a large population of adult outpatients withepilepsy taking any of the commonly used AEDs.

H. Arif et al. / Epilepsy & Behavior 14 (2009) 202–209 203

2. Methods

We examined the charts of 1694 adult (P16 years old) outpa-tients seen by one of 13 epileptologists at the Columbia Compre-hensive Epilepsy Center between January 1, 2000 and January 1,2005, including all 1189 adult patients who had been newlystarted on an AED (carbamazepine [CBZ], clobazam [CLB], felbam-ate [FBM], gabapentin [GBP], levetiracetam [LEV], lamotrigine[LTG], oxcarbazepine [OXC], phenobarbital [PB], phenytoin [PHT],primidone [PRM], tiagabine [TGB], topiramate [TPM], vigabatrin[VBG], valproic acid [VPA], or zonisamide [ZNS]). For each patient,an AED was classified as ‘‘newly started” if it was initiated for thefirst time at our center, during the 5-year study period.

We reviewed patient demographics (Table 1), past medical andpsychiatric history, clinical tolerability, comedications and dos-ages, laboratory test results, side effects, and efficacy. Charts werereviewed for documentation of CSEs developing with the use ofany AED, documented either by patients on a symptom checklistcompleted at each visit or in physician notes. Attribution of CSEsto a given AED was determined by physician notes. CSEs were cat-egorized as one of the following: language problems includingaphasia, anomia or word-finding trouble, poor concentration, poormemory, psychomotor or cognitive slowing, and confusion/disori-entation. When any adverse effect occurred, it was determinedwhether this led to a decrease in dose or discontinuation of theAED; if either was the case, the AED was considered ‘‘intolerable”for that regimen.

For this analysis, each AED regimen (and not an office visit) con-stituted a separate data point. For each patient, a regimen was de-

Table 1Patient demographics by AED (n = 1694)

AEDAE n % Male Mean age(years)

Epilepsy type, %

Focal Primary generalized Symptomat

CBZb 609 48 41 80 10 4CLB 177 53 46 93 2 3FBM 89 32 39 67 20 10GBP 362 46 45 85 6 4LEV 642 46 42 74 17 5LTG 1056 45 43 75 16 3OXC 243 46 42 87 4 3PB 127 38 45 72 18 5PHT 438 51 45 74 14 3PRM 50 54 44 60 22 12TGB 41 34 41 85 3 5TPM 247 36 38 63 20 11VGB 45 64 41 78 2 18VPA 378 48 40 44 42 8ZNS 226 36 37 67 20 7Overall 1694 45 42 74 16 5

a Mean ± SD.b CBZ, carbamazepine; CLB, clobazam; FBM, felbamate; GBP, gabapentin; LEV, levetira

PRM, primidone; TGB, tiagabine; TPM, topiramate; VGB, vigabatrin; VPA, valproate; ZNS

Table 2Non-AED predictors of AED-related cognitive side effects (CSEs) in patients with epilepsya

Attribute Numberof patientswith attribute

Number of patient–drugcombinationswith attribute

Ccw

Female gender 917 2554Localization-related epilepsy 1182 3428Central nervous system infection 82 290 1Chronic obstructive pulmonary disease 90 236 1Other comorbid illness 428 1270 1Static encephalopathy 120 449Age Continuous variable

a Overall (n patients = 1694). Predictors tested: 77 variables including sex, age, wepsychiatric history. Only variables with P < 0.1 in univariate analysis are shown here; a

fined as a specific combination of AEDs or AED dosages. Each timean AED or dose was changed for any reason (either during an officevisit or via telephone), this was recorded as a new regimen for thatpatient. There were an average of 8.1 regimens per patient. Foreach AED a patient was taking, if a CSE was ever reported andattributed to that AED, CSE occurrence was recorded as ‘‘positive”for that AED.

There was only one rater/data enterer per patient record, and noformal interrater agreement testing was conducted; however,there were regular spotchecks of random patients entered by alldata enterers, both by the physician investigators alone and as agroup, to maintain uniform practices. Furthermore, there weremany automated data checks in place to maintain accuracy, suchas making sure that seizure types were compatible with the epi-lepsy syndrome, that the subsequent dose was lower after beingconsidered ‘‘intolerable,” that all fields required to assess AED tol-erability and efficacy were completed, and so forth. In addition, weroutinely reviewed all ‘‘outlier” data points (those that fell morethan 2 SD from the mean) for serum levels and doses.

We first investigated the predictors of CSEs in all 1694 patients.Seventy-seven variables, including demographics, medical andpsychiatric history, and epilepsy details (risk factors, epilepsyand seizure types, syndromes and lobe involved), were tested aspotential predictors of CSEs (see Supplementary Table 1 for a listof all tested variables, and Table 2 for those that were significantin univariate analysis). We then compared the rates of AED-relatedCSEs for all AEDs newly started in 10 or more patients (total n ofpatients newly started on an AED at our center = 1189). In addition,we compared the rates of CSEs in patients taking a given AED in

Mean weight(kg)

Maximum dose,mg/daya

Median duration,months (max)

ic generalized Unclear

6 76 945 ± 650 39 (172)2 77 17 ± 11.5 25 (81)2 69 2566 ± 1320 27 (65)5 74 1710 ± 540 25 (106)3 75 1865 ± 1020 28 (80)6 74 424 ± 255 35 (104)5 76 1210 ± 145 8 (72)5 76 95 ± 38 32 (248)7 76 295 ± 68 39 (255)6 74 780 ± 425 35 (134)5 73 30 ± 22 15 (42)6 78 344 ± 186 19 (77)2 70 2434 ± 480 24 (56)5 75 1345 ± 1099 33 (170)3 75 270 ± 45 16 (74)5 75 35 (115)

cetam; LTG, lamotrigine; OXC, oxcarbazepine; PB, phenobarbital; PHT, phenytoin;, zonisamide.

SEs in patient–drugombinationsith attribute

CSEs in patient–drugcombinationswithout attribute

P, univariate Odds ratio (95% CI),univariate

9.8% 7.6% 0.03 1.006 (1.001–1.92)8.9% 8.4% 0.06 1.29 (1.1–2.8)0.7% 8.6% 0.09 1.54 (0.88–1.91)4.8% 8.4% 0.05 1.63 (1.23–3.01)0.4% 8.2% 0.006 1.44 (0.85–2.93)4.2% 9.2% 0.077 0.6 (0.34–1.29)

0.083 1.006 (0.82–1.53)

ight, seizure type, epilepsy syndrome, epilepsy risk factors, medical history, andll these variables were nonsignificant in multivariate analysis.

Table 3Pairwise comparison of AED-attributed cognitive side effects (CSEs) leading to intolerability (dose change or discontinuation) in adults with epilepsy newly started on an AED(n = 1189)a

AED n Mean time to occurrence ofintolerable CSE, months

Rate ofCSEs

CBZb CLB FBM GBP LEV LTG OXC PB PHT TGB TPM VGB VPA ZNS

CBZ 111 6.9 9.9%CLB 68 7.1 13.2% 0.49FBM 29 6.8 6.9% 0.62 0.37GBP 179 10.1 7.3% 0.43 0.14 0.94LEV 546 6.8 10.4% 0.87 0.48 0.54 0.21LTG 617 5.9 8.9% 0.74 0.25 0.71 0.49 0.38OXC 173 9.0 11.6% 0.66 0.72 0.46 0.17 0.68 0.29PB 23 6.3 13.0% 0.65 0.98 0.46 0.33 0.69 0.50 0.84PHT 96 6.1 14.6% 0.30 0.81 0.28 0.05 0.23 0.08 0.47 0.85TGB 22 16.1 4.5% 0.42 0.26 0.72 0.64 0.37 0.48 0.32 0.32 0.20TPM 130 4.4 21.5% 0.01 0.15 0.07 <0.001 <0.001 <0.001 0.02 0.35 0.18 0.06VGB 12 NA 0.0% 0.25 0.18 0.35 0.33 0.24 0.28 0.21 0.19 0.16 0.45 0.07VPA 109 6.2 8.3% 0.67 0.29 0.81 0.76 0.49 0.82 0.37 0.47 0.15 0.55 <0.001 0.30ZNS 208 5.1 14.9% 0.21 0.73 0.24 0.02 0.09 0.01 0.34 0.81 0.94 0.18 0.12 0.15 0.09

a Results are shown for AEDs newly started in 10 or more patients. Significant P values (60.05) are in boldface.b CBZ, carbamazepine; CLB, clobazam; FBM, felbamate; GBP, gabapentin; LEV, levetiracetam; LTG, lamotrigine; OXC, oxcarbazepine; PB, phenobarbital; PHT, phenytoin;

TGB, tiagabine; TPM, topiramate; VGB, vigabatrin; VPA, valproate; ZNS, zonisamide.

Table 4Pairwise comparison of AED-attributed cognitive side effects (CSEs) leading to intolerability (dose change or discontinuation) in adults with epilepsy newly started on an AED inmonotherapy (n = 546)a

AED n Mean time to occurrence of intolerable CSE, months Rate of CSEs CBZ GBP LEV LTG OXC PHT TPM VPA ZNS

CBZ 68 5.8 1.5%GBP 39 7.9 2.6% 0.69LEV 107 6.4 5.6% 0.17 0.45LTG 270 6.8 3.7% 0.35 0.72 0.41OXC 64 9.0 7.8% 0.08 0.27 0.57 0.15PHT 57 6.1 8.8% 0.06 0.22 0.44 0.10 0.85TPM 18 4.8 11.1% 0.05 0.18 0.38 0.13 0.66 0.77VPA 46 NA 0.0% 0.41 0.27 0.10 0.18 0.05 0.04 0.02ZNS 44 5.1 9.1% 0.06 0.21 0.43 0.11 0.81 0.96 0.81 0.04

a Results are shown for AEDs newly started in 10 or more patients. Significant P values (60.05) are in boldface.

204 H. Arif et al. / Epilepsy & Behavior 14 (2009) 202–209

monotherapy (n = 546). The rates of CSEs were compared usingseveral measures. For both polytherapy (Table 3) and monotherapy(Table 4), we compared the rates of CSEs that led to either a dosagechange or discontinuation of a given AED, which we consideredevidence of ‘‘intolerability.” For patients on polytherapy, we com-pared the rates of intolerable CSEs excluding those that occurredduring the first 8 weeks of treatment (to exclude the main titrationperiod) (see Supplementary Table 2). We also compared the inci-dences of CSEs attributed to AEDs, including those that did not leadto any change in dosage (see Supplementary Tables 3 and 5). Final-ly, we compared the rates of CSEs that led to discontinuation of anAED (see Supplementary Tables 4 and 6). For all of the aforemen-tioned comparisons, a given patient may have been included inmore than one AED group. As a result, the rates of CSEs betweenAEDs may have been correlated.

2.1. Statistical analysis

To determine which factors influenced the overall rate of CSEs, amultivariate binary logistic regression analysis was performed. Thedependent variable (whether a patient ever experienced CSEs ornot) was dichotomous, and the independent (predictor) variableswere a mix of dichotomous and continuous variables. The predic-tor variables included past medical and psychiatric history, pres-ence or absence of static encephalopathy, age, and epilepsy riskfactors. Each predictor variable was entered into a univariate bin-ary logistic regression analysis. Variables that predicted CSEs witha degree of significance of P < 0.1 were entered into a multivariate

binary logistic regression analysis. In the multivariate analysis, sig-nificance was set at P < (0.05/number of tested variables), accord-ing to the Bonferroni correction.

To compare the rate of CSEs between AEDs, we performed a ser-ies of pairwise v2 analyses without Bonferroni correction to avoidthe occurrence of a type II error. Each drug was compared witheach of the other AEDs in a two-by-two comparison. When v2 anal-yses included expected values <5, P values were calculated usingFisher’s exact test. To avoid type I error, we interpreted the resultsin terms of consistent patterns seen in the relative rates of CSEsattributed to the AEDs.

3. Results

Demographics are summarized in Table 1. The primary resultsof this study are illustrated in Figs. 1 and 2 and Tables 3 and 4.Overall 320 of 1694 (18.9%) patients ever experienced AED-relatedCSEs.

In univariate analysis, the following were significant predictors(P < 0.1) of AED-related CSEs (Table 2): older age, female gender,localization-related epilepsy, and presence of central nervous sys-tem infection, chronic obstructive pulmonary disease, and ‘‘othercomorbid conditions.” In the same analysis, the presence/historyof static encephalopathy was negatively correlated with AED-re-lated CSEs. Because there were seven variables with P < 0.1 in uni-variate analysis of the predictors of CSEs, the P value forsignificance in multivariate analysis was set at <0.05/7 = 0.0071.None of these predictors were found to be significant in the multi-

0%

5%

10%

15%

20%

25%

CBZ(111)

CLB(68)

FBM(29)

GBP(179)

LEV(546)

LTG(617)

OXC(173)

PB(23)

PHT(96)

TGB(22)

TPM(130)

VGB(12)

VPA(109)

ZNS(208)

Antiepileptic Drug

% o

f pat

ient

s

IntolerabilityDiscontinuation

Fig. 1. Comparison of AED-attributed cognitive side effects (CSEs) causing intolerability in adults with epilepsy newly started on an AED. Only results for AEDs started in P10patients are given here; refer to Table 3 for complete results including exact rates and statistical comparisons of CSEs. n = 1189. CBZ, carbamazepine; CLB, clobazam; FBM,felbamate; GBP, gabapentin; LEV, levetiracetam; LTG, lamotrigine; OXC, oxcarbazepine; PB, phenobarbital; PHT, phenytoin; TGB, tiagabine; TPM, topiramate; VGB, vigabatrin;VPA, valproate; ZNS, zonisamide.

0%

2%

4%

6%

8%

10%

12%

CBZ(68)

GBP(39)

LEV(107)

LTG(270)

OXC(64)

PHT(57)

TPM(18)

VPA(46)

ZNS(44)

Antiepileptic Drug

% o

f pat

ient

s

IntolerabilityDiscontinuation

Fig. 2. Comparison of AED-attributed cognitive side effects (CSEs) causing intoler-ability in adults with epilepsy newly started on an AED in monotherapy. Onlyresults for AEDs started in monotherapy in P10 patients are given here; refer toTable 4 for complete results including exact rates statistical comparisons of CSEs.n = 546. CBZ, carbamazepine; GBP, gabapentin; LEV, levetiracetam; LTG, lamotri-gine; OXC, oxcarbazepine; PHT, phenytoin; TPM, topiramate; VPA, valproate; ZNS,zonisamide.

H. Arif et al. / Epilepsy & Behavior 14 (2009) 202–209 205

variate analysis. All of the variables that were significant in the uni-variate analysis are listed in Table 2. See Supplementary Table 1 fora complete list of tested predictors.

3.1. Relative rates of CSEs attributed to all newly started AEDs(n = 1189)

We compared relative rates of CSEs attributed to each of the 14AEDs that were newly started in 10 or more patients treated at ourcenter; results are illustrated in Figs. 1 and 2 and summarized inTables 3 and 4, Supplementary Tables 2–6. One patient may havebeen newly started on more than one AED and hence may havebeen included in more than one group when comparing relativerates of CSEs for all the AEDs. Rates of intolerability and discontin-uation due to AED-related CSEs are illustrated in Fig. 1 (Fig. 2 formonotherapy). Refer to Table 3 for rates of intolerable CSEs (Table4 for monotherapy), Supplementary Table 3 for overall incidencerates of CSEs (even those not leading to dose change) (see Supple-mentary Table 5 for monotherapy), and Supplementary Table 4 forrates of CSEs causing AED discontinuation (Supplementary Table 6for monotherapy). The average rate of AED-related CSEs for a givenAED was 16.6%, with 12.8% leading to dosage change or discontin-uation (intolerability). The exact P values derived from the pair-wise comparison of the intolerable CSE rates of each AED to all

other AEDs are listed in Table 3 for polytherapy and Table 4 formonotherapy; Supplementary Tables 3 and 4 report pairwise com-parisons of the incidence of CSEs (even if no dose change was re-quired) and those that required AED discontinuation,respectively, in patients on AED polytherapy (Supplementary Ta-bles 5 and 6 report the same for monotherapy).

The highest rate of intolerable CSEs was attributed to TPM(n = 130, 21.5% intolerability); this rate of CSE intolerability wassignificantly higher (P 6 0.05) compared with the intolerabilityrates of most other AEDs, including CBZ (n = 111, intolerabil-ity = 9.9%), GBP (n = 179, intolerability = 7.3%), LEV (n = 546, intol-erability = 10.4%), LTG (n = 617, intolerability = 8.9%), OXC(n = 173, intolerability = 11.6%), and VPA (n = 109, intolerability =8.3%) (see Table 3 for exact P values of drug–drug pairwise compar-isons of CSE intolerability rates). ZNS (n = 208, 14.9% intolerability)had a significantly higher rate of intolerable CSEs compared withGBP (7.3%) and LTG (8.9%).

To ascertain whether the differential rates of intolerability toCSEs attributed to various AEDs were affected by varying titrationrates, we compared the rates of intolerable CSEs after excludingCSEs that occurred during the first 8 weeks of treatment withany AED (Supplementary Table 2). In this analysis, rates of CSEswere lower, but relative differences remained unchanged.

Rates of intolerable CSEs were lower in patients on mono-therapy (Table 4) than in those on polytherapy. No intolerable CSEswere attributed to VPA in monotherapy (n = 46). The highest rateof intolerable CSEs attributed to AEDs in monotherapy was seenwith TPM (11.1% of 18 patients), and was significantly higher thanthat of CBZ (1.5%, n = 68) or VPA (0.0%, n = 46).

Further significant differences between the AEDs becameapparent in other subgroup analyses of patients on monotherapy:in the comparison of CSE incidence rates (not necessarily requiringdose change or discontinuation) in monotherapy (SupplementaryTable 5), CBZ (2.9%) and VPA (2.2%) were both significantly betterthan PHT (15.8%) and OXC (17.2%), and LTG (7.8%) was significantlybetter than OXC (17.2%). In the comparison of CSE rates leading todiscontinuation (Supplementary Table 6) in patients on mono-therapy, CBZ (0.0%) and LTG (0.7%) were both significantly betterthan ZNS (6.8%) and PHT (7.0%), whereas LTG (0.7%) was signifi-cantly better than OXC (4.7%) (Supplementary Table 6).

4. Discussion

The main findings of this study were as follows: No significantnonmedication predictors were identified as being associated with

206 H. Arif et al. / Epilepsy & Behavior 14 (2009) 202–209

a higher rate of subjective CSEs. The cognitive side effect profiles ofthe AEDs appear to differ from each other based on the results ofour study. CSEs were most common with TPM, followed by ZNSand PHT (Fig. 1 and Table 3). In pairwise comparisons of CSEs lead-ing to intolerability (dose change or discontinuation), TPM was sig-nificantly worse than CBZ, LTG, LEV, OXC, PHT, and VPA, whereasZNS was worse than GBP and LTG (Table 3). When CSEs occurringduring the titration period were excluded, the results were identi-cal (Supplementary Table 2). In a similar analysis limited to pa-tients on monotherapy, TPM remained significantly worse thanCBZ and VPA, whereas ZNS and PHT were both worse than VPA (Ta-ble 4). It should be noted that even with TPM, the AED with thehighest rate of CSEs, fewer than 14% of patients had to discontinuethe medication due to CSEs.

Although no predictors of CSEs were identified in multivariateanalysis, there were several variables that appeared to correlateweakly but significantly with the occurrence of CSEs in the univar-iate analysis, and these should be taken into account in futuretrials.

Cognitive side effects related to TPM treatment have beenwidely documented in a number of controlled trials, retrospectivestudies, meta-analyses, and postmarketing reports [22,24,27–42].

Some clinical trials cited paresthesias as being the leading causefor discontinuation of TPM, even at the lower doses that are usuallyused for migraine [43,44], whereas in other trials the occurrence ofCSEs was responsible for the highest rate of discontinuation[31,34,37]. Martin et al. [33] compared cognitive effects of TPMwith those of GBP and LTG in 17 healthy young adult volunteers.Tests of attention, psychomotor speed, language, memory, andmood were examined. Subjects taking TPM showed declines inattention and letter and category word fluency when tested3 hours after large initial doses, and also at 2 and 4 weeks. Subjectson GBP or LTG did not show any significant declines at any point.The authors did, however, acknowledge that the dose titrationschedules for TPM and LTG were more rapid than that normallyused in a clinical setting. Moreover, their results addressed side ef-fects only in the first month of TPM use. Meador et al. [39] com-pared the cognitive and behavioral effects of LTG and TPM in 47healthy adults using a double-blind, randomized crossover designwith two 12-week treatment periods. Direct comparison of thetwo AEDs revealed significantly better performance on 33 (80%)variables (spanning both objective cognitive and subjective behav-ioral measures) for LTG, but none for TPM. Even after adjustmentfor blood levels (by excluding subjects with TPM levels >8 lg/mL), performance was still better on 19 (46%) variables for patientstaking LTG, but on none for those taking TPM. Bootsma et al. retro-spectively compared patient-perceived cognitive problems at 6, 12,and 18 months of treatment in 402 patients prescribed either TPM(n = 260) or LEV (n = 142) at an epilepsy center [40]. Drug retentionat 18 months was significantly lower for those taking TPM (46%)than for those taking LEV (61%). Cognitive complaints accountedfor a significant number of drug discontinuations, and the high fre-quency of cognitive complaints in the first period of TPM treatmentappeared to be different from that for LEV (P = 0.042). In theremaining patients, the difference in neurocognitive complaintswas not statistically significant.

In other studies, CSEs have been more commonly encounteredwith ZNS [26,45] and PHT [10–16], and less often with GBP, LTG,LEV [33,40,41,46–51], VGB [52–54], and VPA and CBZ [12,17,18].Data are inconclusive with respect to OXC [55] and TGB [56,57],although our results, particularly those in monotherapy, suggestthat OXC has significant CSEs, at least when self-reported. In fact,contrary to prior studies [58], significantly more CSEs were re-ported with OXC monotherapy than with CBZ monotherapy(17.2% vs 2.9% overall incidence, 7.8% vs 1.5% intolerability, and4.7% vs 0.0% leading to discontinuation).

Meador et al. [20,59], in a direct comparison of 1-month expo-sure to PHT and CBZ, found that both CBZ and PHT impaired perfor-mance on tasks of response inhibition and memory compared withnondrug conditions, but the effects were not clinically significant.However, a subsequent publication using the same data reporteda decline in verbal ‘‘paragraph memory” of approximately 15% withPHT and CBZ compared with the nondrug condition [60]. Giventhat the task administered here (reading a story to a patient andasking him or her to recall it immediately and after 20 min) reflectsthe verbal memory demands of routine daily life, this 15% reduc-tion may be clinically significant. In another study of healthy vol-unteers, Meador et al. [60] found that subjects taking CBZperformed more poorly than subjects taking LTG and subjects offAEDs on measures of attention, cognitive speed, memory, andgraphomotor coding. In another double-blinded crossover studyof 35 healthy volunteers receiving either GBP (2400 mg/day) orCBZ (mean 731 mg/day) over 5 weeks, subjects on GBP performedsignificantly better on 8 of 17 cognitive target measures, as com-pared with those taking CBZ [61].

Overall the findings in the literature are consistent with ourfindings in the present study. In our study, CSE incidence rateswere comparable in patients on CBZ and LTG (11.7 and 11.8%,respectively), and both of these rates were significantly lower(P < 0.05) than the rate with TPM (26.2%) (see Supplementary Table3); similarly the rates of CSEs requiring dose change or discontin-uation with LTG use (8.9%, P < 0.05 when comparing both TPMand ZNS) and CBZ use (9.9%, P < 0.05 were significantly lower thanthat with TPM [21.5%] but not significantly different from the ratesfor other AEDs (see Table 3).

In our study, the most commonly encountered specific CSEsattributed to any AED were decreased memory and concentration(see Table 5). There were no obvious differences in the type of CSEsbetween AEDs, and no AED was seen to be associated specificallywith a certain kind of CSE. Although TPM has been associated withlanguage disturbances and anomia [45,62], we were unable to con-firm this: most of the subjective CSEs seen with TPM were psycho-motor slowing or decreased memory.

The primary limitations of our study stem from its retrospectivenature, from self-reporting, and from physician bias. The determi-nation of CSEs was based on a combination of patient reports andphysician notes. There was no systematic evaluation of the occur-rence of cognitive deficits (with neuropsychological testing orstandardized screening tools like the Quality of Life in Epilepsyinventory and the Adverse Event Profile), and 13 different physi-cians were involved. In an attempt to minimize or control for thesepotential biases, we performed multiple secondary analyses, suchas examining overall reports of CSEs whether or not these led todosage changes; comparing the rates of intolerability (dosechange/discontinuation) due to CSEs in patients on both poly-therapy and monotherapy; analyzing incidences and rates of CSEsleading to AED discontinuation; and comparing rates of CSEs in pa-tients while taking (but not necessarily attributed to) a particularAED (data not shown). To minimize the effect of recall bias regard-ing side effects of AEDs used in the past, in the drug comparisonanalysis we included only patients who had been newly startedon the given AEDs at our center. There is also a chance that CSEswere overreported: patients are routinely warned as part of the in-formed consent process of AED side effects (e.g., potential for cog-nitive effects with TPM). It may be difficult to tell whether patientswho were provided with the expectation of experiencing CSEs aremore likely than those who were not to report them.

Rates of CSEs with use of AEDs prescribed in monotherapy werealso analyzed to control for CSEs occurring by interaction withother AEDs. We were unable to control for concurrent non-AEDmedications and their potential to cause CSEs as a result of druginteractions. The relative results (comparing one AED with any

Table 5AED-attributed specific cognitive side effects (CSEs) in adults with epilepsy newly started on any AED: Incidence and intolerability (n = 1189)a

AED n Aphasia/anomia/word-finding difficulty

Decreased concentration Decreased memory Psychomotor/cognitiveslowing

Confusion/disorientation

Incidence(%)

Intolerability(%)

Incidence(%)

Intolerability(%)

Incidence(%)

Intolerability(%)

Incidence(%)

Intolerability(%)

Incidence(%)

Intolerability(%)

CBZ 111 1.8 1.8 5.4 4.5 5.4 4.5 4.5 3.6 0.9 0.9CLB 68 2.9 2.9 5.9 4.4 2.9 2.9 5.9 4.4 1.5 0FBM 29 0 0 3.4 0 6.9 3.4 3.4 3.4 0 0GBP 179 0.6 0.6 4.5 2.8 5.6 3.4 2.8 1.7 1.1 0.6LEV 546 1.5 1.3 3.5 3.1 5.3 3.5 4.6 4.4 1.3 1.3LTG 617 1.6 1.1 3.4 2.4 6 4.1 2.6 2.3 0.8 0.8OXC 173 2.3 1.7 5.2 2.9 11 6.9 2.3 2.3 1.7 1.7PB 23 4.3 4.3 0 0 8.7 8.7 0 0 0 0PHT 96 3.1 3.1 5.2 5.2 8.3 5.2 3.1 3.1 1 1PRM 8 0 0 12.5 12.5 12.5 12.5 0 0 12.5 12.5TGB 22 0 0 0 0 0 0 0 0 4.5 4.5TPM 130 5.4 3.8 3.8 2.3 7.7 6.9 13.8 11.5 3.1 3.1VGB 12 0 0 0 0 0 0 0 0 0 0VPA 109 2.8 1.8 1.8 1.8 3.7 2.8 5.5 5.5 1.8 1.8ZNS 208 2.4 1.9 4.8 3.8 5.8 5.8 6.7 5.8 1.9 1.4Averageb 1189 2.4 (28) 1.9 (22) 5.2 (62) 3.9 (46) 8.3 (99) 5.9 (70) 4.7 (56) 4.3 (51) 2.0 (24) 1.8 (21)

a Intolerability defined as side effect leading to dose reduction or drug discontinuation.b Percent (n).

H. Arif et al. / Epilepsy & Behavior 14 (2009) 202–209 207

other) are largely similar in the monotherapy analyses, which re-vealed the highest rates with TPM use. The only differences arethat PHT and OXC (and ZNS in one analysis) were associated withhigher rates of CSEs in monotherapy, reaching significance in somecomparisons. Similarly CBZ, LTG, and VPA were significantly betterthan some other AEDs (not only TPM) in some of the monotherapyanalyses.

Another limitation is the lack of data on titration rates.Although practice varied somewhat from physician to physician,most AEDs were begun at a rate equal to or slower than the recom-mended titration rates. Results remained virtually identical whenrepeating our analyses after excluding CSEs that occurred in thefirst 8 weeks of treatment (the estimated titration period) (Supple-mentary Table 2).

Multiple prior studies have found depression to be a commonaccompaniment of subjective cognitive dysfunction; we were,however, unable to confirm this in the current study. All psychiat-ric comorbid conditions (including depression, psychosis, aggres-sion, and suicide attempt) were included in the predictoranalysis, but were not found to be significant predictors of CSEseven in the univariate analysis. Formal neuropsychological testingmay help to more objectively elucidate the extent of cognitivechanges with individual AEDs and help to determine whetherthese changes are due to the medications themselves or to otherfactors, such as mood. The results of another recent study fromour database that investigated predictors and rates of AED-relatedpsychiatric side effects also did not seem to suggest such a correla-tion: in that study, psychiatric side effects (including depression)were most commonly attributed to LEV (15.7% incidence, signifi-cantly higher than the average rate of 8.4%), whereas TPM andZNS were both associated with intermediate rates of psychiatricside effects (6.3 and 9.9%, respectively, both nonsignificant) [63].However, the limitations of subjective reporting of cognitive sideeffects may also apply to that of psychiatric side effects, whichmay have precluded the detection of significant associations be-tween mood and cognition that may otherwise become evidentby using standardized measures of cognition and psychiatricsymptoms. For the same reason, we could not account for peoplewho were unaware of their CSEs even though they might besignificant.

Nevertheless, the method of ascertainment of CSEs in the cur-rent study is similar to the imperfect manner in which CSEs aredetermined in routine daily care of patients with epilepsy; thus,

the results should be directly applicable to clinical practice. Evenin the face of these limitations, we believe that any CSEs bother-some enough for the patient to report to their physician were infact documented and thus reported in this study. We were ableto track the occurrence of CSEs for all commonly used AEDs overlong durations of treatment (at least a year and much longer formost commonly used AEDs) and compare the rates with eachother. Although the occurrence of CSEs may be related to severalfactors (including polytherapy, mood, and etiology of epilepsy,among others), the significantly lower rates of CSEs seen with AEDssuch as LTG, LEV, and GBP at least tells us which AEDs are relativelyfree of these adverse effects, even among patients on polytherapy,and should therefore be considered for the treatment of patientswho at a higher risk of developing CSEs, who are already sufferingfrom cognitive complaints, or who feel it is vital to minimize thechance of developing them.

In summary, for many patients with epilepsy, subjective cogni-tive impairment may be more debilitating than the actual seizuresthemselves. It is therefore important to be aware of the possiblerole of AEDs in cognitive impairment. With this information, phy-sicians may be able to better assess the risk/benefit ratio of variousAEDs in patients with epilepsy. Our results suggest that the effectsof TPM on cognition are greater than the effects of most otherAEDs, though only a randomized, double-blinded trial could an-swer that definitively. For patients who wish to minimize their riskof subjective cognitive dysfunction, TPM should be lowest on thelist, followed by ZNS and PHT, whereas LTG, LEV, GBP, VPA, andCBZ should be higher on the list. However, even on TPM, the vastmajority of patients (>85%) will not experience CSEs that requirediscontinuation of the medication.

Role of the funding source

The funding sources had no role in study design, data collection,data analysis, data interpretation, or writing of the article. Allauthors had full access to all the data in the study, and had finalresponsibility for the decision to submit for publication.

Conflict of interest statement

Over the past 7 years, the AED Database has been funded byAbbott, Elan, GlaxoSmithKline, Novartis, Ortho–McNeil, Pfizer,and UCB Pharma.

208 H. Arif et al. / Epilepsy & Behavior 14 (2009) 202–209

Dr. Arif has received funds for travel to academic meetings fromGlaxoSmithKline.

Mr. Buchsbaum, Mr. Weintraub, and Ms. Pierro have no relevantdisclosures.

Dr. Resor has received consulting fees from Medpoint and Ab-bott Pharmaceuticals.

Dr. Hirsch has received honoraria for speaking from Glaxo-SmithKline and UCB Pharma; and has provided consultation forJazz Pharmaceuticals.

Acknowledgments

This study was approved by the appropriate institutional ethicscommittee and has therefore been performed in accordance withthe ethical standards laid down in the 1964 Declaration of Helsinki.Waiver of informed consent was obtained from our institutionalreview board for all patients included in this study.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.yebeh.2008.10.017.

References

[1] Meador KJ. Newer anticonvulsants: dosing strategies and cognition in treatingpatients with mood disorders and epilepsy. J Clin Psychiatry 2003;64(Suppl.8):30–4.

[2] Meador KJ. Cognitive and memory effects of the new antiepileptic drugs.Epilepsy Res 2006;68:63–7.

[3] Gilliam F. Optimizing health outcomes in active epilepsy. Neurology2002;58(8, Suppl. 5):S9–20.

[4] Aldenkamp AP, Vermeulen J. Phenytoin and carbamazepine: differential effectson cognitive function. Seizure 1995;4:95–104.

[5] Meador KJ. Cognitive side effects of antiepileptic drugs. Can J Neurol Sci1994;21:S12–6.

[6] Aldenkamp AP, Baker G. A systematic review of the effects of lamotrigine oncognitive function and quality of life. Epilepsy Behav 2001;2:85–91.

[7] Aldenkamp AP, De Krom M, Reijs R. Newer antiepileptic drugs and cognitiveissues. Epilepsia 2003;44(Suppl. 4):21–9.

[8] Aldenkamp AP. Antiepileptic drug treatment and epileptic seizures: effects oncognitive function. In: Trimble MSB, editor. The neuropsychiatry ofepilepsy. New York: Cambridge Univ. Press; 2002. p. 256–67.

[9] Ortinski P, Meador KJ. Cognitive side effects of antiepileptic drugs. EpilepsyBehav 2004;5(Suppl. 1):S60–5.

[10] Andrewes DG, Bullen JG, Tomlinson L, Elwes RD, Reynolds EH. A comparativestudy of the cognitive effects of phenytoin and carbamazepine in new referralswith epilepsy. Epilepsia 1986;27:128–34.

[11] Duncan JS, Shorvon SD, Trimble MR. Effects of removal of phenytoin,carbamazepine, and valproate on cognitive function. Epilepsia1990;31:584–91.

[12] Gallassi R, Morreale A, Lorusso S, Procaccianti G, Lugaresi E, Baruzzi A.Carbamazepine and phenytoin: comparison of cognitive effects in epilepticpatients during monotherapy and withdrawal. Arch Neurol 1988;45:892–4.

[13] Gillham RA, Williams N, Wiedmann KD, Butler E, Larkin JG, Brodie MJ.Cognitive function in adult epileptic patients established on anticonvulsantmonotherapy. Epilepsy Res 1990;7:219–25.

[14] O’Dougherty M, Wright FS, Cox S, Walson P. Carbamazepine plasmaconcentration: relationship to cognitive impairment. Arch Neurol1987;44:863–7.

[15] Prevey ML, Mattson RH, Cramer JA. Improvement in cognitive functioning andmood state after conversion to valproate monotherapy. Neurology1989;39:1640–1.

[16] Thompson P, Huppert F, Trimble M. Anticonvulsant drugs, cognitive functionand memory. Acta Neurol Scand Suppl 1980;80:75–81.

[17] Goldberg JF, Burdick KE. Cognitive side effects of anticonvulsants. J ClinPsychiatry 2001;62(Suppl. 14):27–33.

[18] Jokeit H, Kramer G, Ebner A. Do antiepileptic drugs accelerate forgetting?Epilepsy Behav 2005;6:430–2.

[19] Dodrill CB, Troupin AS. Neuropsychological effects of carbamazepine andphenytoin: a reanalysis. Neurology 1991;41:141–3.

[20] Meador KJ, Loring DW, Allen ME, et al. Comparative cognitive effects ofcarbamazepine and phenytoin in healthy adults. Neurology 1991;41:1537–40.

[21] Meador KJ, Loring DW, Huh K, Gallagher BB, King DW. Comparative cognitiveeffects of anticonvulsants. Neurology 1990;40(3, Pt. 1):391–4.

[22] Meador KJ, Loring DW, Hulihan JF, Kamin M, Karim R. Differential cognitiveand behavioral effects of topiramate and valproate. Neurology2003;60:1483–8.

[23] De Araujo Filho GM, Pascalicchio TF, Lin K, Sousa PS, Yacubian EM.Neuropsychiatric profiles of patients with juvenile myoclonic epilepsytreated with valproate or topiramate. Epilepsy Behav 2006;8:606–9.

[24] Burton LA, Harden C. Effect of topiramate on attention. Epilepsy Res1997;27:29–32.

[25] Gomer B, Wagner K, Frings L, et al. The influence of antiepileptic drugs oncognition: a comparison of levetiracetam with topiramate. Epilepsy Behav2007;10:486–94.

[26] Park SP, Hwang YH, Lee HW, Suh CK, Kwon SH, Lee BI. Long-term cognitive andmood effects of zonisamide monotherapy in epilepsy patients. Epilepsy Behav2008;12:102–8.

[27] Aldenkamp AP, Baker G, Mulder OG, et al. A multicenter, randomized clinicalstudy to evaluate the effect on cognitive function of topiramate compared withvalproate as add-on therapy to carbamazepine in patients with partial-onsetseizures. Epilepsia 2000;41:1167–78.

[28] Fritz N, Glogau S, Hoffmann J, Rademacher M, Elger CE, Helmstaedter C.Efficacy and cognitive side effects of tiagabine and topiramate in patients withepilepsy. Epilepsy Behav 2005;6:373–81.

[29] Ketter TA, Post RM, Theodore WH. Positive and negative psychiatric effects ofantiepileptic drugs in patients with seizure disorders. Neurology 1999;53(5,Suppl. 2):S53–67.

[30] Kockelmann E, Elger CE, Helmstaedter C. Cognitive profile of topiramate ascompared with lamotrigine in epilepsy patients on antiepileptic drugpolytherapy: relationships to blood serum levels and comedication. EpilepsyBehav 2004;5:716–21.

[31] Lainez MJ, Pascual J, Pascual AM, Santonja JM, Ponz A, Salvador A. Topiramatein the prophylactic treatment of cluster headache. Headache 2003;43:784–9.

[32] Lee S, Sziklas V, Andermann F, et al. The effects of adjunctive topiramate oncognitive function in patients with epilepsy. Epilepsia 2003;44:339–47.

[33] Martin R, Kuzniecky R, Ho S, et al. Cognitive effects of topiramate, gabapentin,and lamotrigine in healthy young adults. Neurology 1999;52:321–7.

[34] Mei D, Capuano A, Vollono C, et al. Topiramate in migraine prophylaxis: arandomised double-blind versus placebo study. Neurol Sci 2004;25:245–50.

[35] Privitera M, Fincham R, Penry J, et al., for the Topiramate YE Study Group.Topiramate placebo-controlled dose-ranging trial in refractory partial epilepsyusing 600-, 800-, and 1,000-mg daily dosages. Neurology 1996;46:1678–83.

[36] Salinsky MC, Storzbach D, Spencer DC, Oken BS, Landry T, Dodrill CB. Effects oftopiramate and gabapentin on cognitive abilities in healthy volunteers.Neurology 2005;64:792–8.

[37] Tatum WOT, French JA, Faught E, et al. Postmarketing experience withtopiramate and cognition. Epilepsia 2001;42:1134–40.

[38] Thompson PJ, Baxendale SA, Duncan JS, Sander JW. Effects of topiramate oncognitive function. J Neurol Neurosurg Psychiatry 2000;69:636–41.

[39] Meador KJ, Loring DW, Vahle VJ, et al. Cognitive and behavioral effects oflamotrigine and topiramate in healthy volunteers. Neurology2005;64:2108–14.

[40] Bootsma HP, Aldenkamp AP, Diepman L, et al. The effect of antiepileptic drugson cognition: patient perceived cognitive problems of topiramate versuslevetiracetam in clinical practice. Epilepsia 2006;47(Suppl. 2):24–7.

[41] Cramer JA, Fisher R, Ben-Menachem E, French J, Mattson RH. New antiepilepticdrugs: comparison of key clinical trials. Epilepsia 1999;40:590–600.

[42] Van Passel L, Arif H, Hirsch LJ. Topiramate for the treatment of epilepsy andother nervous system disorders. Expert Rev Neurother 2006;6:19–31.

[43] Brandes JL, Saper JR, Diamond M, et al. Topiramate for migraine prevention: arandomized controlled trial. JAMA 2004;291:965–73.

[44] Silberstein SD, Neto W, Schmitt J, Jacobs D. Topiramate in migraineprevention: results of a large controlled trial. Arch Neurol 2004;61:490–5.

[45] Ojemann LM, Ojemann GA, Dodrill CB, Crawford CA, Holmes MD, Dudley DL.Language disturbances as side effects of topiramate and zonisamide therapy.Epilepsy Behav 2001;2:579–84.

[46] Leach JP, Girvan J, Paul A, Brodie MJ. Gabapentin and cognition: a double blind,dose ranging, placebo controlled study in refractory epilepsy. J NeurolNeurosurg Psychiatry 1997;62:372–6.

[47] Marciani MG, Stanzione P, Mattia D, et al. Lamotrigine add-on therapy in focalepilepsy: electroencephalographic and neuropsychological evaluation. ClinNeuropharmacol 1998;21:41–7.

[48] Meador K. Assessing cognitive effects of a new AED without the bias of practiceeffects. Epilepsia 1997;38(Suppl. 3):60.

[49] Neyens LG, Alpherts WC, Aldenkamp AP. Cognitive effects of a new pyrrolidinederivative (levetiracetam) in patients with epilepsy. ProgNeuropsychopharmacol Biol Psychiatry 1995;19:411–9.

[50] Wong IC, Lhatoo SD. Adverse reactions to new anticonvulsant drugs. Drug Saf2000;23:35–56.

[51] Blum D, Meador K, Biton V, et al. Cognitive effects of lamotrigine comparedwith topiramate in patients with epilepsy. Neurology 2006;67:400–6.

[52] Kalviainen R, Aikia M, Saukkonen AM, Mervaala E, Riekkinen Sr PJ. Vigabatrinvs carbamazepine monotherapy in patients with newly diagnosed epilepsy: arandomized, controlled study. Arch Neurol 1995;52:989–96.

[53] Chadwick D, For the Vigabatrin European Monotherapy Study Group. Safetyand efficacy of vigabatrin and carbamazepine in newly diagnosed epilepsy: amulticentre randomised double-blind study. Lancet 1999;354:13–9.

[54] Ylinen A, Kalviainen R, Riekkinen Sr PJ. Long-term efficacy and cognitive effectsof vigabatrin. Acta Neurol Scand Suppl 1995;162:47–50.

[55] Aikia M, Kalviainen R, Sivenius J, Halonen T, Riekkinen PJ. Cognitive effects ofoxcarbazepine and phenytoin monotherapy in newly diagnosed epilepsy: oneyear follow-up. Epilepsy Res 1992;11:199–203.

H. Arif et al. / Epilepsy & Behavior 14 (2009) 202–209 209

[56] Dodrill CB, Arnett JL, Sommerville KW, Shu V. Cognitive and quality of lifeeffects of differing dosages of tiagabine in epilepsy. Neurology1997;48:1025–31.

[57] Kalviainen R, Aikia M, Mervaala E, Saukkonen AM, Pitkanen A, Riekkinen Sr PJ.Long-term cognitive and EEG effects of tiagabine in drug-resistant partialepilepsy. Epilepsy Res 1996;25:291–7.

[58] Dam M, Ekberg R, Loyning Y, Waltimo O, Jakobsen K. A double-blind studycomparing oxcarbazepine and carbamazepine in patients with newlydiagnosed, previously untreated epilepsy. Epilepsy Res 1989;3:70–6.

[59] Meador KJ, Loring DW, Abney OL, et al. Effects of carbamazepine andphenytoin on EEG and memory in healthy adults. Epilepsia 1993;34:153–7.

[60] Motamedi GK, Meador KJ. Antiepileptic drugs and memory. Epilepsy Behav2004;5:435–9.

[61] Meador KJ, Loring DW, Ray PG, et al. Differential cognitive effects ofcarbamazepine and gabapentin. Epilepsia 1999;40:1279–85.

[62] Mula M, Trimble MR, Thompson P, Sander JW. Topiramate and word-finding difficulties in patients with epilepsy. Neurology2003;60:1104–7.

[63] Weintraub D, Buchsbaum R, Resor Jr SR, Hirsch LJ. Psychiatric and behavioralside effects of the newer antiepileptic drugs in adults with epilepsy. EpilepsyBehav 2007;10:105–10.