Predicting
success
of
vagus
nerve
stimulation
(VNS)
from
EEG
symmetry
Janneke
Hilderink
a,*
,
Marleen
C.
Tjepkema-Cloostermans
a,
Anita
Geertsema
b,
Janita
Glastra-Zwiers
b,
Cecile
C.
de
Vos
a,caDepartmentofNeurologyandClinicalNeurophysiologyandDepartmentofNeurosurgery,MedischSpectrumTwente,Enschede,TheNetherlands b
SEINCentreofExcellenceforEpilepsyandSleepMedicine,Zwolle,TheNetherlands
c
MIRAInstituteforBiomedicalEngineeringandTechnicalMedicine,UniversityofTwente,Enschede,TheNetherlands
ARTICLE INFO Articlehistory: Received2January2017 Accepted31March2017 Keywords: Epilepsy Quantitativeelectroencephalography (qEEG)
Brainsymmetryindex(BSI) Vagusnervestimulation(VNS)
ABSTRACT
Purpose:Vagusnerve stimulation(VNS) hasshowntobeaneffectivetreatmentfordrugresistant epilepsy,withachievingmorethan50%seizurereductioninonethirdofthetreatedpatients.Inorderto predictwhichpatientswillprofitfromVNS,wepreviouslyfoundthatalowpairwisederivedBrain SymmetryIndex(pdBSI)couldpotentiallypredictgoodresponderstoVNStreatment.Thesefindings howeverhavetobevalidatedbeforetheycanbegeneralized.
Methods:39patients(age18–68years)withmedicallyintractableepilepsywhowerereferredforan implantedVNSsystemwereincluded.RoutineEEGregistrations,recordedbeforeimplantation,were analyzed.Artefact-freeepochswitheyesopenandeyesclosedwerequantitativelyanalyzed.ThepdBSI wastestedforrelationwithVNSoutcomeoneyearaftersurgery.
Results:Twenty-threepatients(59%)obtainedareductioninseizurefrequency,ofwhomten(26%)hada reductionofatleast50%(goodresponders)andthirteen(33%)areductionoflessthan50%(moderate responders).Sixteenpatientswithoutseizurereductionaredefinedasnon-responders.Nosignificant differenceswerefoundinthepdBSIofgoodresponders(mean0.27),moderateresponders(mean0.26) andnon-responders(mean0.25)(p>0.05).Besidesseizurereduction,manypatients(56%)reported additionalpositiveeffectsofVNSintermsofseizureduration,seizureintensityand/orpostictalrecovery. Conclusion:EEGfeaturesthatcorrelatewithVNStherapyoutcomemayenablebetterpatientselection andpreventunnecessaryVNSsurgery.Contrarytoearlierfindings,thisvalidationstudysuggeststhat pdBSImightnotbehelpfultopredictVNStherapyoutcome.
©2017BritishEpilepsyAssociation.PublishedbyElsevierLtd.Allrightsreserved.
1.Introduction
Vagusnerve stimulation(VNS) hasshown tobean effective treatment for drug resistant epilepsy in numerous patients. However,long-termstudiesshowedthatagoodresponse(>50% seizurereduction)isonlyachievedin20–55%ofthepatients[1–3], which meansthat a substantial number of patientsonly show moderateorevennoresponsetoVNStreatment.Determiningthe successofVNS is importanttocounselpatientsand givethem information about the expected seizure reduction. Potential respondersmight notneedtotryotherkindsof therapybefore theyreceiveaneffectiveVNSsystemandontheotherhand,alow
likelihood to respondcouldprevent someone fromundergoing surgeryandhavinganexpensiveVNSsystemimplantedwhileonly minimaleffectswillbeobtained.Despitethegrowingapplication ofVNS,itisstillnotpossibletopredictwhichpatientsrespondto whatextenttoVNStherapy.Moststudiesthatattempttopredict the success of VNS are based upon patient characteristics[4], epilepsysyndrome[5]orlocalizationoftheseizurefocus[2,6,7].A meta-analysisbyEnglotetal.onpredictorsof responsetoVNS therapy,showedthatyoungpatients(<6years)respondslightly better in terms of seizure reduction compared to adults [2]. However,goodpredictorsofefficacyofVNStherapyforindividual patientsarestillelusive.
WepreviouslyshowedthatquantifyingEEGasymmetryusing the pairwise derived Brain Symmetry Index (pdBSI) could potentiallypredictwhichpatientswillbenefitfromVNStreatment. Itwasobservedthatnon-respondersshowsignificantlyhigherEEG
* Correspondingauthorat:MedischSpectrumTwente,P.O.Box50.000,7500KA Enschede,TheNetherlands.
E-mailaddress:j.hilderink@mst.nl(J.Hilderink).
http://dx.doi.org/10.1016/j.seizure.2017.03.020
1059-1311/©2017BritishEpilepsyAssociation.PublishedbyElsevierLtd.Allrightsreserved.
ContentslistsavailableatScienceDirect
Seizure
asymmetry,reflectedinhigher pdBSIvalues, ascompared with responders[9].Thesefindingshoweverrequirevalidationinnew patientgroupsbeforetheycanbegeneralized.Wehavetherefore conducted this prospective study in adult epilepsy patients withoutcognitiveimpairment;expectingtoconfirmthepredictive valueofEEGsymmetry,asdefinedbypdBSI.
2.Materials/methods 2.1.Patients
Wehaveincludedotherwisehealthyepilepsypatients(age>18 years)whowerereferredbySEINCentreofExcellenceforEpilepsy andSleepMedicine(EpilepsyCentre)toMedischSpectrumTwente hospitalforimplantation ofa VNSsystem.Allpatientssuffered frommedicallyintractable(generalizedorlocalized)epilepsywith varyingfocuslocations.Patientshadtokeepseizurediariesforat leastsixmonthspriortoimplantation andduringtheoneyear followupperiodofthestudy.
2.2.VNSoutcome
After implantation, patientshave regular follow upvisits at theirEpilepsyCentre.Duringthesevisits,stimulationparameters areoptimized for the individualpatient and the effect of VNS therapyismonitored.ThesuccessofVNSwasdeterminedbythe amountof seizure reduction due tothetherapy. Patients were groupedasgood(>50%reduction),moderate(<50%reduction)and non-responders (no reduction) based on seizure diary data provided by the treating nurse practitioner from the Epilepsy Centre.
Inaddition,patientsareaskedtofillinaquestionnaireabout the effects of VNS and their satisfaction with the device approximately one year after implantation. The questionnaire addressesvariousparameterssuchasseizurefrequency,duration and intensity, postictal recovery, patient satisfaction, and side-effects. Patients indicate how much various parameters have improved or worsened upon VNS therapy on a 7-point scale (ClinicalGlobalImpression–Improvementscale,CGI-I),where1 meansverymuchimprovedand7meansverymuchworsened. Data from questionnaires was used to compare patient's own perceptionwiththedatafromtheEpilepsyCentreandisusedto make a second classification, where CGI-I scores 1–2=good responder, score 3=moderate responder, scores 4–7= non-re-sponder.
2.3.EEGanalysis
Thirty-minute routine EEG registrations were made several weeksbeforeVNSimplantation.Electrodeswereplacedconform theinternational10–20system,usinganelectrocapandsignalwas recorded using a BrainLab EEG recording system (OSG BVBA, Belgium) with a sampling frequency of 250Hz. During the registration,thepatientswerecomfortablylyingdowninaquiet, shielded room. Artefact-free epochs with eyes open and eyes closed were selected for quantitative analysis. Selected epochs werefilteredwithabandpassfilterbetween0.5and30Hz.Epochs of500sampleswith50%overlapwereFouriertransformedwith pwelchinMATLAB(TheMathworks,Inc.,USA)usingaHamming window.
Brain symmetry was quantifiedusing the pair-wise derived BrainSymmetry Index(pdBSI),which was describedpreviously
[9–11].Briefly,thepdBSIevaluatesasymmetrybycalculatingthe powerperfrequencycoefficientalong homologousEEGchannel pairs.LowpdBSIvaluesrepresentsymmetricEEGactivity,whereas higherpdBSIvalues indicatehigher asymmetryof theEEG.For
each patient, pdBSI values were determined for four different frequencybands:delta(0.5–4Hz),theta(4–8Hz),alpha(8–12Hz) andbeta(12–30Hz).EEGsymmetry,definedbypdBSI,wastested forrelationwitheffectofVNStherapyafteroneyear.
2.4.Statistics
Statisticalanalyseswereperformedusingt-testswhen(normal distribution) and Mann–Whitney Utests (non-normal distribu-tion)withaconfidenceintervalof95%.
3.Results
3.1.Patientcharacteristics
Between March 2011 and January 2015, 39 patients had a routineEEGrecordedafewweekspriortoimplantationofaVNS systemandgaveinformedconsenttoanalyzeEEGcharacteristics andlookforrelationwithVNSeffects.Patientcharacteristicsare summarized in Table 1. Patients were not considered surgical candidatesandtheirintellectualabilityvaried,howevernoneof thepatientswasseverelycognitivelyimpaired.Forall39patients, dataprovidedbythenursepractitioner(follow-uptimeonaverage 14months,range8–24months) andpdBSIvaluesareavailable. Fiveoutof39patientsdidnotsendbackthequestionnaireabout theirperceptionofeffectsandsatisfactionwiththeVNStherapyso thisinformationisonlyavailablefor34patients(follow-uptimeon average14months,range6–36months).
3.2.EffectsofVNS 3.2.1.Seizurereduction
Basedondataprovidedbythenursepractitioner,twenty-three patientsobtainedareductioninseizurefrequency,ofwhomten hadareductionofatleast50%(goodresponders)andthirteena reduction of less than 50% (moderate responders). The other sixteenpatientsdidnotshowanyreductioninseizurefrequency andweredefinedasnon-responderstoVNStherapy.Neitherthe patient'sagenorthetypeofepilepsycorrelatedwiththeseizure reductionobtainedwithVNSandthereforetheseparameterscould notpredicttheeffectofVNS(Table1).
Besides seizure reduction, other positive effects of VNS treatmentwerereportedbythenursepractitioner.Outofsixteen patientswhoweredefinedasnon-responders,sevenpatientsstill experiencedotherpositiveeffectsofVNS.Also,themajorityofthe goodresponders(7outof10)andmoderateresponders(8outof 13)showedadditionalpositiveeffectsintermsofseizureduration, seizureintensityand/orpostictalrecovery.
3.2.2.Patientperception
Inadditiontothedataprovidedbythenursepractitioner,VNS outcome was determined using patient questionnaires. Ten patients indicated that their seizure frequency has improved very much or much (CGI-I score 1 or 2). Fourteen patients mentionedasmall improvementin seizurefrequency(score3) andten patients indicated thatthe seizurefrequency hadnot changed (score 4). No patient indicated worsening of seizure frequency. The patient's general impression regarding seizure frequency only partly corresponded with the percentages of seizurereductionthatwereprovidedbythenurse practitioner (Table1).
All patients who indicated that their seizure frequency has (very)muchimprovedalsoexperienced(some)improvementin seizureintensityand/orpostictalrecovery.Majorityofthefourteen patientswhoreportedminimalimprovementinseizurefrequency also report little or no improvement in seizure intensity and
recovery after seizures. However, three of these patients mentioned much improvement in intensity, recovery, or both. Thenon-responders,reportedlittleornoimprovementintermsof seizureintensityandpostictalrecovery.
To assess their overall satisfaction with the VNS therapy, patientswereaskedhowsatisfiedtheywereonascaleof1(very satisfied)to7 (verydissatisfied).Thistimethewholescalewas usedbythepatientsandonaveragetheyscored3.4.Asexpected, theextentofsatisfactionisstronglycorrelatedwiththeeffectsof thestimulator.Inthegoodrespondergroup(accordingtothenurse
practitioner)theaveragesatisfactionis2.1,whereasthemoderate andnon-respondersscore3.2and4.2,respectively.
3.2.3.Sideeffects
Mostpatients(33outof39)experienceside-effectsfromVNS treatment, independently of theeffectivenessof VNS(Table 2). Somepatientsexperiencemorethanonetypeofside-effect. Side-effects that werementioned occur only during stimulation. In general, comparable occurrence and type of side effects are experiencedbyallthreegroups.
Table2
Reportedsideeffects.Numbersindicatenumberofpatientspergroup(good,moderate,non-responder)reportingtherespectiveside-effect.Somepatientsexperience multipletypesofside-effects.
Sideeffect Goodresp.(#8/10) Moderateresp.(#12/13) Non-resp.(#13/16) Total(#33/39)
Hoarseness 7 8 8 23 70%
Throatdiscomfort 3 2 6 11 33%
Voicealterations 2 5 1 8 24%
Shortnessofbreath(duringexercise) 1 2 3 6 18%
Tinglingsensationinneck/shoulder/arm 2 0 1 3 9%
Painfuljaw/teeth 0 0 2 2 6%
Coughing 0 1 0 1 3%
Snoring 1 0 0 1 3%
Table1
PatientcharacteristicsandeffectofVNS.EffectofVNSreductionisbasedonthedatafromseizurediariesprovidedbythetreatingnursepractitioner.Patient'sperceptionof seizurereductionandpatient'soverallsatisfactionwiththeVNSdevicearebasedontheClinicalGlobalImpression–Improvement(CGI-I)scores.(NA:informationisnot available).
No. Sex Age(years) Focuslocation EffectVNS(%reduction) Patientperceptionofreduction Patientoverallsatisfaction
1 M 68 Multifocal >50 1 1 2 F 37 Multifocal >50 2 1 3 F 27 Rtemporal >50 1 1 4 M 45 Multifocal >50 4 4 5 F 46 Multifocal >50 3 2 6 F 40 Rtemporal >50 3 3 7 M 19 Multifocal >50 2 3 8 F 23 Multifocal >50 3 4 9 F 27 Ltemporal >50 2 2 10 M 62 Rtemporal >50 NA NA 11 F 31 Frontal <50 2 4 12 M 21 Rtemporal <50 2 3 13 F 54 Multifocal <50 4 6 14 F 25 Multifocal <50 3 3 15 F 49 Rtemporal <50 4 3 16 M 60 Ltemporal <50 3 3 17 F 35 Rtemporal <50 3 3 18 M 65 Multifocal <50 4 4 19 F 59 Multifocal <50 2 2 20 F 54 Rtemporal <50 3 2 21 F 22 Ltemporal <50 3 2 22 F 26 Multifocal <50 NA NA 23 F 20 Multifocal <50 NA NA 24 F 51 Ltemporal 0 3 3 25 F 48 Rtemporal 0 4 7 26 F 52 Multifocal 0 2 3 27 M 51 Multifocal 0 2 3 28 M 47 Multifocal 0 3 4 29 F 42 Frontal 0 4 6 30 F 18 Multifocal 0 3 3 31 M 31 Multifocal 0 4 6 32 M 20 Frontal 0 4 NA 33 F 19 Multifocal 0 3 5 34 M 62 Multifocal 0 3 3 35 F 24 Multifocal 0 3 4 36 M 32 Frontal 0 4 5 37 M 35 Multifocal 0 4 3 38 F 28 Multifocal 0 NA NA 39 M 46 Multifocal 0 NA NA
3.3.PairwisederivedBrainSymmetryIndex(pdBSI) 3.3.1.EEG
Artefact-free EEG signal of interictal episodes was used to determinethepdBSI.Epochsof10switheyesopenandeyesclosed wereselected.Foreachpatient–foreyesopenonaverage54sand foreyesclosedonaverage57sof–EEGsignalwasavailablefor quantitativeanalysis.Nostatisticallysignificantdifferenceswere observedbetweenpdBSIvaluesobtainedfromEEGperiodswith eyesopenandeyesclosed(Fig.1).
3.3.2.RelationpdBSIandseizurereduction
Patientsweregroupedasgood,moderateandnon-responders basedonseizurereductionasindicatedbythenursepractitioner. NosignificantdifferenceswereobservedbetweenpdBSIvaluesin delta,theta,alphaandbetabands,forgoodresponderscompared with moderate responders and non-responders (Fig. 1A).
Additionally,patientswereclassifiedbasedontheirownreported seizure reduction. Even though this is a partially different grouping,itresultsinonlyslightlydifferentaveragepdBSIvalues forallfrequencybandsandagainnosignificantdifferenceswere observed(Fig.1B).Similarly,patientsclassifiedbasedontheirown overallsatisfactionwithVNS,doesalsonotshowanysignificant differences in pdBSI values between good, moderate and non-responders(Fig.1C).
4.Discussion
To prevent patientswho willnot respond toVNStreatment from undergoing unnecessary surgery, it is necessary to find predictors of the expected success of VNS. Low pdBSI values showedpromisingresultsinpredictinggoodresponderstoVNS treatmentinapreviousstudy[9].However,wehavenotbeenable toreproducethepreviousfindingsinthiscurrentstudy.Although
Fig.1.Differencesbetweengoodresponders,moderateresponders,andnon-respondersinpdBSIforthedelta,theta,alphaandbetabandandaveragedoverallfrequencies, forEEGperiodswitheyesopenandeyesclosed.Classificationgood,moderateandnon-responderswasbasedon:(A)dataprovidedbythetreatingnursepractitioner,(B) patient'sperceptionofseizurereduction,and(C)patient'soverallsatisfactionwiththetherapy.Errorbarsrepresentthestandarddeviation.
thestudypopulationofthiscurrentstudywaslargerthaninour previousstudy(n=39versusn=19),thenumberofpatientsisstill relatively small, and again the study population consists of a heterogeneousgroupofpatientsthatarenotsuitableorwillingto undergoepilepsysurgery.Previouslywebasedouranalysesonly ontheinformationprovidedbynursepractitioners,whereasinthis study we also included two measures of the patients’ own perceptionofVNSoutcome, andevaluatedpdBSIvaluesof EEG recordingwitheyesopenaswellaswitheyesclosed.Nevertheless, nosignificantdifferencesinpdBSIvalues wereobservedwithin thesevariousconditionsoroutcomemeasures.
Thisprospectivestudyenableda long-termfollow-upof the samepatient population, eliminating theinfluenceof potential earlydropoutsthatmaybemissedinretrospectivestudies.Inour study, most patientsexperience moderate(33%) togood (26%) seizurereductionuponVNStherapy.Studiesinyoungerpatients haveshownabetterresponsetoVNStherapy[2,7,8].However,our studyincludedonlyadults.Eventhoughonly26%ofthepatients hadstrongseizurereductionuponVNStreatment(dataprovided by nurse practitioner), mostpatients (62%) indicated that they wereingeneral(somewhattovery)satisfiedwiththedevice.This mayberelatedtootherpositiveeffectsmanypatientsexperience besides seizure reduction, suchas reduced seizure intensityor durationand/orshorterpostictalrecoverytimes,alsoof impor-tancetothepatient.Inaddition,potentialreductionofnocturnal seizures, which may not be noticed and thus not reflected in reduced seizure frequency, may result in an improved sleep quality,increased daytimealertnessand a betteroverallhealth condition[12]. On the other hand, patientswho showseizure reduction upon VNS treatment may experience negative side-effectsthatcouldinfluencetheirperceptionofseizurereduction and decreasetheir overallsatisfaction withthetherapy. In our study,nearlyall patients(85%)experienced some,mostlymild, side-effects.Thispercentageis higherthandescribedbyothers, however,thetypeofreportedside-effectsissimilar[13,14].
Patients were defined as either good, moderate or non-responders based onthe amount of their seizure reduction as providedbynursepractitioners,basedonseizurediarieskeptby the patients themselves, which is still the standard, but has neverthelesslimitedaccuracy[15].Hence,thediscrepancy with thepatients’ownperceptionofseizurereductionaswellastheir satisfaction with the therapy, when asked one year after implantation. We found that seizure reduction as experienced bypatients(CGI-Iscale) differedsubstantially fromtheseizure reduction provided by the nurse practitioner. Patients’ own perceptioncouldbe bothbetteror worse thanwhat thenurse practitioners’ data suggested and 18 patients were classified differently. Possibly, this classification is influenced by other factorslikeside-effectsorgeneral satisfactionwiththetherapy, eventhoughwehavetriedtopreventthisbyaskingaboutthese aspectsinseparatequestions.Nevertheless,patients’perceptionof seizurereductionisaskedonlyonceafteroneyearasascoreona 7-pointscale,whichisadifferentmeasurethantheinformation based on daily seizure diaries. Moreover, interpretation of the 7-pointscalemayvaryamongstpatients.Onepatientmayscorea 30%seizurereductionas‘verymuchimproved’whereasanother patientmayscore thesame percentageof seizure reductionas ‘minimally improved’, probably coloured by the impact that (varioustypesof)seizureshaveonthepatient'sdailylife.
Incontrasttoourearlierfindings,thecurrentdataindicatethat EEGsymmetryquantifiedaspdBSIvaluesmightnotcorrelatewith patientsatisfactionorwithseizurereductionduetoVNStherapy andcanthereforenotbeusedasapredictoroftheeffectofVNS. Newstudiesneedandwillbeconductedtocontinuethesearchfor biomarkers to reliably predict the effects of VNS therapy and enablebetterpatientselectionandpreventunnecessarysurgical procedures.
Acknowledgments
WewouldliketoacknowledgetheDepartmentofNeurosurgery andespeciallyM.J.DriesseMDandK.KhoMDwhoimplantedthe VNSdevices,fortheircontinuoussupportandtime.Inaddition,we wouldliketothanktheDepartmentofClinicalNeurophysiology fortheregistrationoftheEEGs.
Thisresearchdidnotreceiveanyspecificgrantfromfunding agenciesinthepublic,commercial,ornot-for-profitsectors.
WeconfirmthatwehavereadtheJournal'spositiononissues involved in ethical publication and affirm that this report is consistentwiththoseguidelines.
References
[1]ArdeschJJ,BuschmanHP,Wagener-SchimmelLJ,vanderAaHE,HagemanG. Vagusnervestimulationformedicallyrefractoryepilepsy:along-term follow-upstudy.Seizure2007;16:579–85.
[2]EnglotDJ,ChangEF,AugusteKI.Vagusnervestimulationforepilepsy:a meta-analysisofefficacyandpredictorsofresponse.JNeurosurg2011;115:1248–55. [3]ShahwanA,BaileyC,MaxinerW,HarveyAS.Vagusnervestimulationfor refractoryepilepsyinchildren:moretoVNSthanseizurefrequencyreduction. Epilepsia2009;50:1220–8.
[4]GhaemiK,ElsharkawyAE,SchulzR,HoppeM,PolsterT,PannekH,etal.Vagus nervestimulation:outcomeandpredictorsofseizurefreedominlong-term follow-up.Seizure2010;19:264–8.
[5]LabarD.Vagusnervestimulationfor1yearin269patientsonunchanged antiepilepticdrugs.Seizure2004;13:392–8.
[6]JanszkyJ,HoppeM,BehneF,TuxhornI,PannekHW,EbnerA.Vagusnerve stimulation:predictorsofseizurefreedom.JNeurolNeurosurgPsychiatry 2005;76:384–9.
[7]MarrasCE,ChiesaV,DeBenedictisA,FranziniA,RizziM,VillaniF,etal.Vagus nerve stimulation in refractory epilepsy: new indications and outcome assessment.EpilepsyBehav2013;28:374–8.
[8]LagaeL,VerstrepenA,NadaA,VanLoonJ,TheysT,CeulemansB,etal.Vagus nervestimulationinchildrenwithdrug-resistantepilepsy:ageat implanta-tionandshorterdurationofepilepsyaspredictorsofbetterefficacy?Epileptic Disord2015;17:308–14.
[9]deVosCC,MelchingL,vanSchoonhovenJ,ArdeschJJ,deWeerdAW,van LambalgenHC,etal.Predictingsuccessofvagusnervestimulation(VNS)from interictalEEG.Seizure2011;20:541–5.
[10]Sheorajpanday RV, Nagels G, Weeren AJ, van Putten MJ, De Deyn PP. Reproducibility and clinicalrelevanceof quantitative EEG parameters in cerebralischemia:abasicapproach.ClinNeurophysiol2009;120:845–55. [11]van Putten MJ. Extended BSI for continuous EEG monitoring in carotid
endarterectomy.ClinNeurophysiol2006;117:2661–6.
[12]RizzoP,BeelkeM,DeCarliF,CanovaroP,NobiliL,RobertA,etal.Chronicvagus nervestimulationimprovesalertnessandreducesrapideyemovementsleep inpatientsaffectedbyrefractoryepilepsy.Sleep2003;26:607–11. [13]GooneratneIK,GreenAL,DuganP,SenA,FranziniA,AzizT,etal.Comparing
neurostimulationtechnologiesinrefractoryfocal-onsetepilepsy. JNeurol NeurosurgPsychiatry2016;87:1174–82.
[14]KubaR,BrazdilM,KalinaM,ProchazkaT,HovorkaJ,NezadalT,etal.Vagus nervestimulation:longitudinalfollow-upofpatientstreatedfor 5years. Seizure2009;18:269–74.
[15]HoppeC,PoepelA,ElgerCE.Epilepsy:accuracyofpatientseizurecounts.Arch Neurol2007;64:1595–9.