Brigitta Kakuszi, Bálint Szuromi, István Bitter, Pál Czobor with an accelerated decline? Last in, ﬁrst out - delayed brain maturation Attention deﬁcit hyperactivity disorder:

(1)

with an accelerated decline?

Brigitta Kakuszi, Bálint Szuromi, István Bitter, Pál Czobor

^∗

SemmelweisUniversity,DepartmentofPsychiatryandPsychotherapy,Balassau.6.,Budapest1083, Hungary

Received 24September2019;receivedinrevisedform9March2020;accepted13March2020

KEYWORDS ADHD;

Neuro-development;

Maturational-delay;

NoGoP3;

EEG;

Age-relateddecline

Abstract

WhileearlyneurodevelopmentalprocessesduringtheemergenceofADHDinchildhoodreceived considerableattention,theneurobiologicalmechanismsthatunderliethechangesinADHDin adulthoodremainlargelyunaddressed.Wewantedtodelineateneurodevelopmentalchanges inadultADHDusinganelectrophysiologicalmeasure,thefronto-centralNoGoP3event-related potential(ERP),whichisanimportantneurophysiologicalindexofbrainfunctioninginADHD, andbiomarkerfor responseinhibitionandaging.ERPs wereobtained from45ADHD and41 healthysubjectsusinga128-channelBioSemirecording-system,applyingemotionally-valenced andneutral stimuliin aresponse inhibition task. Our resultsindicated that ADHD subjects manifesteddelayeddevelopmentalP3-trajectoryinyoung-adulthoodascomparedtocontrols;

theyalsoshowedP3reductionacrossallemotionalvalences,andthereductionwasmostpro- nouncedatyoungerages.ThedifferencesinP3diminishedbymid-adulthood,andstartedto increaseagainatmoreadvancedages.Thus,similartostructural-MRIindices,developmental braindifferencesinthefronto-centralNoGoP3inADHDlargelynormalizeinyoung-adulthood.

However,areductionofP3occursagain startingfrommid-adulthood.Asthefronto-central NoGoP3reflectsthefunctioningofthefrontalareas(whichshowdelayedmaturationinADHD), ourfindingsareconsistentwiththe‘‘lastin,firstout’’hypothesis,whichreferstoamirroring patternofbraindevelopmentandaging,andpositsthatbrainregionsthatdeveloprelatively latedegeneraterelativelyearlywithage.Thus,ADHDmaynotonlybeassociatedwithdelayed neurodevelopment,butalsowithaprematureage-relateddeterioration,atleastinsomemea- suresofelectrophysiologicalfunctioning.

ThisisanopenaccessarticleundertheCCBY-NC-NDlicense.

(http://creativecommons.org/licenses/by-nc-nd/4.0/)

∗Correspondingauthor.

E-mailaddress:czobor.pal@med.semmelweis-univ.hu (P.Czobor).

https://doi.org/10.1016/j.euroneuro.2020.03.011

(http://creativecommons.org/licenses/by-nc-nd/4.0/)

(2)

1. Introduction

While early neurodevelopmental alterations during the emergenceofADHDinchildhoodreceivedconsiderableat- tentionintheliterature(Shawetal.,2006;Spronketal., 2008; Vaidya,2012),the neurobiological mechanismsthat underliethe changesinADHD inadulthood remainlargely unaddressed. With respect to early neurodevelopmental processes, longitudinal structural-MRI data indicate that children withADHD undergo delayed maturation in many brainregions,especiallyinthefrontalcortex(Gogtayetal., 2002;Shawetal.,2007).Asigniﬁcant portionofthismat- urational delay, however, is being made up by thebegin- ningofyoung-adulthood(Shawetal.,2007).Alargecross- sectional investigation of structural-MRI data - termed as

’mega-analysis’bytheauthors(Hoogmanetal.,2017)-of 1723ADHDand1523controlsubjects(agerange=4–63years) providedstrongsupportforthe"brainmaturationdelaythe- oryfor ADHD".Sinceitis conceivablethat -throughbrain plasticityorexperience-dependentsynapticreinforcement (Caseyetal.,2005;Whitfordetal.,2007)-thesechanges in brainmorphology duringchildhoodand adolescencein- ducefunctionalalterations,itwouldbeimportanttoinves- tigatepost-adolescentneurophysiologicalchangesinADHD as compared to healthy individuals. In terms of func- tionalbrainalterations,arecentlypublishedcomprehensive overview of developmentaltrajectories of ADHD over the lifespan(Frankeetal.,2018)identiﬁedonesmall,longitu- dinalEEGstudy(Doehnertetal.,2013)whichreportedthat onlyERPimpairmentsrelatedtoresponsepreparationwere associatedwithADHDinadulthood.However,thestudywas small(ADHDn=11/controln=12)andthefollow-upofsub- jects(childrenwithameanage=10.9years)didnotextend beyondyoungadulthood(meanage=21.9years).

An importantelectrophysiologicalmarkerof brainfunc- tioning in ADHD is the fronto-central P3 ERP, a positive- going potential between approximately 300 to 450 msec post-stimulus,which iswidelyviewedasameasureofbe- havioral response inhibition (BRI) (Fallgatter etal., 2005; Knezevicand Marinkovic,2017;RandallandSmith, 2011), andof aging(Rossinietal.,2007). Itistypicallyobserved afterNoGotrialsinresponseinhibitionparadigms,andmay reflecttheactivityofmultiplecorticallocations,including thefrontalcortex,pre-supplementarymotorarea,thecin- gulateregionsandtheinferiorparietallobe(Albertetal., 2013; Tammetal., 2002;Vara etal.,2014).BRI is acore executive function (EF); its impairmentleads tobehavior dyscontrol, distractibility and deficits in sustained attention, which are considered to be key features of ADHD (Wodka etal., 2007).Diminished ability tosuppress inap- propriateprepotentresponsestostimulihasbeenobserved inADHD(LipszycandSchachar,2010;Wodkaetal.,2007), which hasbeen viewedbysome authors asa "disorderof deficient self-regulation"(DSR) (Barkley, 1997). BRI under- goesa prolonged maturation:in typicallydeveloping (TD) subjects,itsdevelopmentstartsattheageof6years,and lasts at least until the age of 20 years (Jonkman, 2006; Tammetal.,2002).

In line withthe developmental changesof BRI, P3also shows a protracted developmentalpath in TDchildren. It cannot yet be observed in 6–7-year oldchildren (while it is present in 9–10 year olds), which hasbeen interpreted

as a sign of immature response inhibition processing in earlychildhood(Jonkman,2006).Anotherstudyalsofound thatthe NoGo P3develops relatively late,around age 10 (Okazakietal.,2004).AsystematicreviewbyvanDinteren etal.2014identiﬁedaquadratictrajectoryfromchildhood tooldagesforthefronto-centralP3amplitude:asteadyin- creasefrompreschool-ageuntilamaximum ataroundthe third-decade;andasubsequentplateau,followedbyade- clinefromtheﬁfth-decade.

Astwoimportantsourcesofthefronto-centralP3arethe frontal cortexand the cingulateareas, and its amplitude correlateswiththefrontalgraymattervolume(Fordetal., 2004), a parallel can be drawn between the gradual in- crease of P3 amplitude with age and the prolonged development of the frontal cortex. Importantly, the NoGo P3, as a neural marker of BRI, exhibits altered develop- mentinADHD:diminishedNoGoP3isconsistentlyreported in ADHD as compared to TD children between 7 and 14- yearsofage(Fallgatteretal.,2004;Fallgatteretal.,2005; Wiersemaetal.,2006).

Prior studies of NoGo P3 ERP almost exclusively used BRI tasks which ignored affective contexts, and did not examinewhetherBRIisaffectedbythestimuli’semotional valence(see,e.g.,Albertetal.foranexceptioninhealthy subjects(Albertetal.,2013)).ThesestudiesconsideredEFs aspurelycognitive skills(Tsermentseli andPoland, 2016).

However, more recent theoretical models of EFs make a distinction between cool cognitive EFs (i.e., primarily thoseencompassed in the earlier conceptualization, such as attention, working memory, planning and inhibition) and “hot” EFs which involve emotions and motivation (ZelazoandCarlson,2012).Thistheoreticalframeworkhas crucialimplicationsforthedevelopmentalresearchinADHD becausecurrentconceptualizationsconsiderDeﬁcientEmo- tional Self-Regulation (DESR) as a core feature of ADHD (Barkley, 1997; Castellanos et al., 2006; Surman et al., 2013).Nonetheless,wearenotawareofanydevelopmen- taldataontheP3ERPinADHDduringadulthoodthatwere obtainedinanemotionalresponseinhibitiontask.

Toﬁllintheaboveknowledgegaps,inthecurrentstudy wefocusedonthefronto-centralP3ERPtoinvestigatethe neurobiologicalunderpinningsofdevelopmentalchangesin ADHDduringadulthood.Usingaquadraticregressionmodel- ingapproach,wewanted(a)todelineatethedevelopmen- tal trajectory of P3ERPin ADHD ascomparedto healthy controls; and (b) to examine whether the developmental changes are associated with changes in response inhibition.Weadoptedanemotionalresponseinhibitionparadigm basedona Go/NoGo task usingemotionally-valenced and neutral pictures from the International Affective Picture System(IAPS)(Langetal.,2008).

2. Experimentalprocedures 2.1. Studysample

Patients who met DSM-IV criteria for adult ADHD (mean age=30.4years, range:18–59years) and had no history of neuro- logicalillnesswererecruitedforthestudyattheDepartmentof Psychiatryand Psychotherapy, Semmelweis University,Budapest, Hungary,whichprovidesoutpatientserviceforpatientswithadult

(3)

Fig.1 Taskandprocedure.Stimuliwerepresentedcentrallyevery1400msecfor800msec(inter-stimulus-interval=600msec).

Subjects wereaskedtopushabuttonassoonaspossibleupontheappearanceofthestimuluspictures(Gotrials);they were, however, askedtowithholdresponse incase apicturewasrepeated(NoGotrials). 240stimuliwere shownintwo blocks.The probabilityofGoandNoGotrialswas0.85and0.15,respectively.

ADHD. All included patients fulﬁlledthe DSM-IV criteriafor the combinedsubtypeofADHD(i.e.,had>6symptomsofthetotalof 9 symptomsofInattentionand Hyperactivity/Impulsivity,respectively). Recruitment was done bystaff of the outpatientclinic.

Diagnosiswasconﬁrmedthroughsemi-structuredinterviewbythe treatingphysician.Healthycontrols(HC)wererecruitedfromthe community throughfriendsandacquaintancesofthestaff ofthe University.HCshad amean ageof31.3years(range:18–59years), and had no history of psychiatric disease or currentpsychiatric comorbidity. They wereindividuallymatchedto patientson age (±5years),genderandeducationlevel.Additionalexclusioncrite- riaforparticipantsinthecontrolgroupwereanypresentorpast neurologicaldisorderandhistoryofheadinjurywithlossofcon- sciousness.Thestudycompliedwiththeethicalstandardsofthe DeclarationofHelsinki,andreceivedapprovalfromthelocalEthi- calCommittee.AllparticipantsgavewrittenInformedConsentfor thestudy.

2.2. Descriptionofmeasures

The Conners’ Adult ADHD Rating Scale (CAARS;66-item version) was used to characterize ADHD symptom severity across core psychopathological domains of ADHD: Inattention, Hyperactiv- ity, Impulsivity and Problems with Self-Concept(Conners, 1999; Erhardtetal.,1999).TheAdultSelf-ReportScalesymptomCheck- list(Adleretal.,2006)wasusedtodelineateADHDsymptomsand toestablishADHDsubtype.ThetotalscoreontheSymptomCheck List 90R(SCL-90R), a self-report scale was used to measure the severityongeneraldomainsofpsychopathology(Derogatis,1994).

Thescalewasdesignedprimarilytoassesssymptompatternsina broadspectrumofpopulations,ranging fromnon-patienthealthy subjects to individuals with psychiatric disorders(Derogatis and Cleary, 1977). The scale was administered by study person- nel including psychologists and researchassistants trained inits administration.

2.3. Stimuliandprocedure

Participantswereseatedinadimlylitroom,approximately100cm fromthemonitor.TheInternationalAffectivePictureSystem(IAPS) (Langetal.,2008),asetofimageswithneutral,positive,andneg- ativevalences,wereusedasstimuli,andpresentedinrandomse- quenceusingthePresentation13.0software(NeurobehavioralSys-

tems, Inc.,Albany,Calif.).The stimuliwerepresentedcentrally every1400msecfor800msec(interstimulus-interval=600msec).

Atotalof480stimuliwerepresentedintwoblocks,eachconsisting of240stimuli.Subjectswereaskedtopushabuttonassoonaspos- sibleuponappearanceofthepictures(Gotrials);theywereasked towithholdresponseifapicturewasrepeated(NoGotrials).The probabilityofGoandNoGotrialswas0.85and0.15,respectively.

ThestimulustaskandprocedureareillustratedinFig.1.

2.4. EEGrecordingandpre-processing

High-densityEEGswererecordedusinga128-channelBioSemiAc- tiveTwosystemwithanaveragereference,atadigitizationrateof 1024Hz,applyingaband-passﬁlterof0.5–70Hz.Weappliedastan- dardBioSemi128-electrodeheadcapsystem(onlinesource:http://

www.biosemi.com/pics/cap_128_layout_medium.jpg). Data were storedandanalyzedoff-lineusingtheElectromagneticSourceSig- nal Imaging (EMSE)Suite as well as theStatistical Analysis Sys- tem (SAS 9.4) software. EEG was re-referenced off-line to the common average potential and ﬁltered between 0.5 and 70Hz using zero-phase shiftforward and reverse IIRButterworth-ﬁlter.

Additionally, the signal was filtered using the 48–52Hz Parks- McClellan stop-band Notch filter. The notch filter was used to remove any potential electric-interference from the 50Hz line.

Artifacts dueto blinks and eyemovements were removedman- ually and with the electrooculography artifact removal proce- dure. Epoch selection for the analyses was conducted manu- ally, as well as applying automatic artifact rejection criteria.

Stimulus-lockeddataweresegmented intoepochsof1000msec, including200msecbeforestimulus and800msecafterstimulus.

Segmentswithactivityexceeding±100μVwereexcludedfromfur- theranalysis.Thethresholdcut-off was50fortherequiredmini- mumnumberofusablesegmentsfortheERPanalyses.Onlycorrect trialswereincludedinthecurrentanalyses.Thestimulus-locked segmentswerebaseline-correctedusinga200msecpre-response window,andaveragedtoobtaintheERP-waveformsforeachsub- jectandeachemotionalvalence.

2.5. Statisticalanalyses

2.5.1. Behavioraldata

WeusedAnalysisofCovariancewithrepeated-measures(rANCOVA) to investigate behavioral data, including commission-error rates

(4)

Table1 Basicdescriptivestatisticsofthestudypopulation.

Characteristics ADHD(n=45) HealthyControl(n=41) Teststatistic^a p-value

Categoricalvariables(%) Chi-squaretest

Male 71.1% 75.6% 0.2 0.64

Educationlevel(highschoolgraduate) 40% 46.3% 1.2 0.56

Continuousvariables:mean(SD) ANOVA(F)

Age(years) 30.4(10.9) 31.3(11.4) 0.2 0.69

ConnersAdultADHDRatingScale(CAARS)

Totalscore 121.6(23.6) 50.3(25.7) 165.1 <0.0001

CAARSSubscales:Inattention/MemoryProblem 24.5(6.6) 8.5(6.6) 116.8 <0.0001

Hyperactivity/Restlessness 20.2(6.4) 11.1(5.7) 43.8 <0.0001

Impulsivity/Emot.Lability 18.2(7.3) 8.8(4.5) 45.1 <0.0001

ProblemswithSelf-Concept 10.8(5.0) 4.2(4.2) 39.1 <0.0001

aANOVAforcontinuousvariables,Chi²⁻testforcategoricalvariables.

and reactiontimes.Commission-errorrate(%) andreaction time (RT, msec)wereappliedinseparate analysesasdependent vari- ables. Studygroup (ADHD, HC; used asa between-subjects fac- torandemotionalvalence(positive,negative,neutral;,usedasa within-subjectfactor)weretheindependentvariables.Agewasap- pliedasacontinuousregressor.Inordertoexaminepotentialnon- linearchangesovertime,agewasincludedintheanalysisbothas a linearanda quadraticterm.Interactionbetween studygroup, emotionalvalenceandagewerealsotested.

2.5.2. ERPmeasures

Based on literature, the deﬁnition of frontal P3 component time-window included the post-stimulus epoch of 300–

450 msec (Knezevic and Marinkovic, 2017). Similar to our earlier work (DeSanctis et al., 2013), thestatistical analyses were based on random-regression hierarchical linear modeling (HLM;

(BrykA.S.andRaudenbushS.W.,1992;Gibbonsetal.,1988)).Re- peatedmeasurementsoftheERPamplitude(inmicrovolts)inthe frontalregions fortheP3 ERP(regionof interestincludedelec- trodeFzandtheadjacentelectrodessurroundingFz)inthecom- ponent time-window served as dependent variable.Study group (between-subjectsfactor),andemotionalvalence(within-subject factor)weretheprincipalindependentvariablesofinterest.Time (samplingpointinthecomponentwindow,relativetostimuluson- set)wasalsoincludedintheanalysisasawithin-subjectfactor.A ﬁrst-orderautoregressivemovingaveragecorrelationmatrixamong thesamplingpointswasspeciﬁedintheHLMmodel.Genderserved asacovariate.Similartotheanalysesofbehavioraldata,agewas usedasa continuousregressor.Interactionbetweenstudygroup, emotionalvalenceandage(usedinlinearandquadraticform)was investigatedintheHLM.Theinteractioneffectbetweengroupand agetestedwhetherthestudygroupshadadifferentdevelopmental trajectorywithage.

3. Results

3.1. Demographiccharacteristics

PatientswithADHD(N=45)andHCs(N=41)hadnosignif- icant differencein age(Table1).The proportionofmales wassomewhathigher(75.6%)amongtheHCthanamongthe ADHDsubjects(71.1%); thedifferencewasnotsigniﬁcant.

Therewasnosigniﬁcantdifferenceintheeducationlevel, with morethan 40% of the sample in both groups attain- ingcollegeorhighereducation.Thetwogroupssigniﬁcantly differedonCAARSpsychopathologicalsymptomdimensions:

asexpected,patientshadhigherseverityonallsymptomdi- mensions.ThepatientsamplemetthecriteriafortheADHD combinedsubtype.

Amongthe ADHDpatients, 12(26.6%)of the45patients had comorbidity according to DSM-IV. In the majority of cases(10[22.2%]of45),comorbiditiesfellin theaffective categories.Speciﬁcally,affective comorbiditiesincluded8 cases withdepression in their medical history, while dys- thymiaandbipolardisorderoccurredin 1patient,respectively.Additionalcomorbiditieswereobsessive-compulsive disorder in one patient, and substance (cannabis) use in onepatient.Approximatelyhalfofthe45patients(n=23, 51.1%)receivedstimulantmedication.Patientstakingstim- ulanttreatmentwereoff medicationforatleast24hbefore testing.

3.2. Behavioralmeasures

Withrespect to commission-errors, theanalysis indicated asigniﬁcant maineffectof group(F1,85=4.72,p=0.032), and a signiﬁcant linear (but no quadratic) effect of age on the commission-error rates (F_1,85=8.75, p=0.0041).

Post-hoc analyses showed that ADHD patients made more commission-errorsthanHCs,andthatthecommission-error ratedecreasedwithage.Themaineffectofemotionalva- lence,anditsinteractionwithgroupandagewerenotsignificant (p > 0.1 for all respectiveterms in therANCOVA analyses). The developmental trajectory of commission- errorrates,determinedonthebasisofthelinearregression estimates,isshowninFig.2(top).To delineatetheabove resultsintermsoftheobserveddata,wedichotomizedthe studygroupsintoyoung-adult(≤30years)andmiddle-aged (>30years)groups,andprovidedsummarystatisticsforboth agegroups (Table 2,left). In the ADHDgroup, amongthe middle-agedsubjects theproportion of commission-errors wassignificantly(p<0.05)reducedforallthreeemotional valencesascomparedtotheyoungersubjects.Among the HCs,thereductionwassignificantonlyfortheneutralstim- uli.

The analysisof RTs showed a linearmain effect of age (F1,85 = 13.7, p=0.0004), with no quadratic or interac- tioneffects.Post-hocanalysesindicatedthatinbothstudy groups RTs increased with age. The estimated RT trajec-

(5)

Fig.2 Commission-errorrate(toppanel)andreactiontime(bottompanel)asafunctionofageintheADHDandthehealthycontrol group,determinedonthebasisofthelinearregressionestimatesobtainedfromtherepeatedmeasuresANCOVAmodel.Theshaded bands depictthestandarderroroftheestimates.For commission-errors,theanalysis indicatedasignificant (p=0.032)global effectofgroup,withahighererrorrateamongADHD subjectsirrespectiveofemotionalvalence;andasignificant(p=0.0041) reductionoferrorratewithage.Themaineffectofemotionalvalence,anditsinteractionwithgroupandagewerenotsignificant.

Forreactiontime,theanalysisshowedasignificant(p=0.0004)linearagemaineffect,indicatingincreasedreactiontimeswith ageinbothgroups.Thedifferenceinreactiontimebetweenthetwogroupswasnotsignificant,andtheage-relatedchangesdid notshowasignificantinteractionwithemotionalvalenceorstudygroup.

(6)

Table2 CommissionerrorratesandreactiontimesbyagegroupamongADHDandhealthycontrol(HC)subjects^a. Studygroupsbyage CommissionError(%)mean(SE) ReactionTime(msec)mean(SE)

POS NEU NEG POS NEU NEG

ADHD≤30yrsN=28 37%(4) 38%(3) 36%(4) 416.1(13.4) 410.5(12.8) 426.7(13.8) ADHD>30yrsN=17 18%(5) 22%(4) 18%(5) 444.2(16.8) 435.0(16.2) 447.8(17.4) Diff.(≤30vs.>30yrs) 19%(6) 16%(6) 18%(6) −27.4(2.14) −24.5(2.06) −21.1(2.2) F,p^b 9.11,0.0043 6.24,0.0164 8.72,0.0051 1.62,0.2095 1.42,0.2406 0.90,0.3490 Control≤30yrsN=27 20%(3) 19%(2) 17%(2) 413.4(9.8) 410.2(10.1) 416.9(10.8) Control>30yrsN=14 15%(4) 8%(3) 9%(4) 470.5(13.6) 467.1(14.0) 477.9(14.9) Diff.(<30vs.>30yrs) 5%(5) 11%(4) 8%(5) −57.1(16.8) −56.9(1.73) −61(1.84) F,p^b 0.74,0.3949 7.07,0.0113 2.65,0.1114 11.51,0.0016 10.82,0.0021 10.99,0.0020

aStudygroups(ADHD,HC)weredichotomizedbyage(≤30years[youngadults]vs.>30years[middle-agedsubjects]).

bANOVAwasappliedforbothstudygroupstoexaminedifferencesincommissionerrorsratesandreactiontimesbetweenthetwoage groups(youngadultsvs.middleagedsubjects).Theanalyseswereconductedseparatelyforthepositive,neutralandnegativestimuli.

Errorrate(%)andreactiontimewasappliedasdependentvariable.Agegroup(≤30yearsvs.>30years)wasusedasanindependent variable.Signiﬁcant(p<0.05)differencesbetweenagegroupsarehighlighted.

Fig.3 GroupaveragedNoGoERPresponsestonegative,neutralandpositivepicturesintheemotionalNoGoparadigm,broken downbystudygroup(ADHD,Control).ShadedareasindicatethetimewindowofinterestfortheP3ERPcomponent.Asshownbythe Figure,theP3componentwasprecededbyalargenegative–goingwave,thefrontalN2,whichshiftedthecomponentdownward (i.e., innegativedirection).Random-regression Hierarchicallinear model (HLM)analyses showed asigniﬁcantly diminished P3 amplitudeintheADHDgroupforeachofthethreeemotionalvalences(negativep=0.0043;neutralp=0.038;positivep=0.023).

tory with age, determined on the basis of the regression estimates, is shown in Fig.2(bottom). Similartothe commission-errors, we alsocharacterized the aboveﬁnd- ings in terms of the observed data by subdividing the study groups intoyoung-adult andmiddle-aged subgroups (Table 2, right). By mid-adulthood, the RTs increased nu- mericallyinboth theADHDandtheHCgroupfor allemo- tionalvalences.Theincrease,however,obtainedstatistical signiﬁcanceonlyfortheHCs.

3.3. P3amplitude:groupdifferenceand developmentaltrajectories

Group-averaged NoGo ERPstonegative, neutral andposi- tivepicturesareshown inFig.3separatelyfor bothstudy groups.ThetimewindowofinterestfortheP3component is marked asa shaded area.As shown by the ﬁgure, the P3component wasshiftedin negativedirection by apre- cedinglarge negative–goingwave,thefrontalN2,whichis typicallyenhanced invisual tasks(Folsteinand Van Petten C.,2008;Zordanetal.,2008),especiallyunderexperimen-

tal conditions that involve considerable conﬂict process- ingeffort(Donkers andvan Boxtel, 2004; Gajewski et al., 2008; Nieuwenhuis et al., 2003). HLM analysis indicated a main effect of group with a larger P3 among the controls(F_1,85 =6.15,p=0.015),andaninteraction between groupandemotionalvalence(F1,85= 4.04,p=0.019).The two-way interaction of group with the linear or with the quadraticeffectofageontheP3amplitudewasalsosignif- icant(linearF1,85=5.08,p=0.027,quadraticF1,85=5.24, p=0.0247).The three-wayinteraction ofagebygroup by emotionalvalencewasnotsigniﬁcant(p>0.1).

Post-hocanalysis of theinteraction between group and emotional valence showed a signiﬁcantly lower P3ampli- tudein theADHD groupfor thenegativepicturesascom- paredtothepositive(t=−3.44,df=85,p=0.0009)orneu- tralones (t= −2.40, df=85, p=0.0187). The difference between positive and neutral pictures did not reach sig- niﬁcance (t=−1.62, df=85, p=0.11). In the HC group, therewere nodifferencesby emotional valence (p > 0.1 inallpairwisecomparisons).Post-hocanalysisoftheinter- action of group withthe linear andthe quadratic effects of age revealed that in the ADHD group both the linear

(7)

Fig.4 NoGoP3amplitudeasafunctionofageintheADHDandhealthycontrol(HC)groupscombinedforallstimuluspictures.HLM analysisindicatedamaineffectofgroup(F_1,85=6.15,p=0.015),andaninteractionofgroupwithboththelinearandthequadratic effectofageontheP3amplitude(linearF_1,85=5.08,p=0.027,quadraticF_1,85=5.24,p=0.0247).IntheHCgroupneitherthe linearnorthequadraticregressioneffectsobtainedsigniﬁcance,indicatingnosystematicchangeinP3amplitudewithage.Inthe ADHD group,boththelinearandthequadraticeffectwassigniﬁcant(p<0.05),indicatingacurvilinearrelationship withage.

ThedisplayedtrajectoriesrepresenttheestimatedregressionfunctionsofP3amplitudewithagebasedjointlyonthelinearand quadratictermsfromtheHLM.Theshadedbandsdepictthestandarderroroftheestimates.

andthequadraticeffectsreachedsigniﬁcance(lineareffect t=2.47,df=85,p=0.0156);quadraticeffectt=−2.23, df = 85, p=0.0286), indicating a curvilinear relationship of P3 amplitude with age. In the HC group, neither the linear nor the quadratic effectwas signiﬁcant (p > 0.1), indicating no systematic change in P3 amplitude with age.

Fig.4showstheregressionestimatesfortheNoGoP3am- plitudeasafunctionofage(i.e.,theestimatedquadratic regressionfunctions),combinedacrossallstimuluspictures.

ThedevelopmentalpathoftheP3amplitudewasmarkedly differentinpatientsandcontrols:patientswithADHDevi- dencedasigniﬁcantamplitudereductionatyoungeragesas comparedtoHCs.Theamplitudereductiondiminishedand lost signiﬁcancebymid-adulthood,butthenstarted toin- crease again at higherages.Even though HCs exhibiteda slightincreasewithage,theabovementionedgroupdiffer- enceswereprimarilydrivenbythesubstantiallymorepro- nouncedage-relatedchangesintheADHDgroup.

3.4. Subsidiaryanalyses

In subsidiaryanalyses, we investigatedwhethertheinclu- sion of clinicallyrelevant covariates, suchasthe severity

of psychopathology, comorbidities and medication status, had an impact on our principal electrophysiological find- ings. Fortheseverityof psychopathologywe usedthe to- talscoreontheCAARSscale,whileadichotomousvariable wasapplied for the use of stimulant medication (yes/no) andcomorbidities.OurresultsareshowninOnlineTable1. The rows of the Table depict the results of the addition of the covariates. The table shows that the group effect and the linear and quadratic interaction terms re- mained similarto those obtained from the analyses with no covariate in the model. Specifically, after the introduction of the covariates in the analyses, there was no change in termsof statistical significance withrespectto the group effect and the linear and quadratic age inter- actions. Furthermore, the covariate-effects did not ob- tainstatisticalsignificance inthe analyses.We alsoinves- tigated whethertherewasa differencein the severityof psychopathology inthe ADHD sample between the young- adultandmiddle-agedsubgroups.TheANOVAanalysesindi- catednodifferenceintheCAARS totalscore(F_1,44=0.12, p=0.73): the mean total score was 120.6(SD=24.1) and 123.2(SD=23.4) in the young-adult and middle-aged sub- group, respectively. Furthermore, the two subgroups did not differ in any of the CAARS subscales (p > 0.1 in all analyses).

(8)

4. Discussion

Our principal ﬁndings suggest that patients with ADHD manifestadelayeddevelopmentaltrajectoryofthefronto- central NoGo P3 in young-adulthood relative to HCs. In general,thedevelopmentaltrajectoryofP3wasmarkedly differentinthepatientsandcontrols. Whilepatients with ADHD showed a signiﬁcant P3 reduction across all emo- tionalvalences,thereductionwasthemostpronouncedat youngerages.TheagedifferenceinP3diminishedbymid- adulthood;thisdiminutioneffect,however,waslimitedto a certainage range,beyond which the differencestarted toincreaseagainathigherages.

ConcerningthediminutionofP3alterationinADHDwith age,ourﬁndingsparalleltheresultsofseveralMRIstudies that examined age-relatedchanges. Forexample, a large longitudinalstudyin223ADHDand223normally-developing children showedthatthepeakof cortical-thicknessmatu- ration was delayed in ADHD by an average of 3 years; in frontalandtemporalareas,thematurationwasdelayedby upto5years(Shawetal.,2007).Arecentlongitudinalstudy bythesameresearch-groupextendedthepreviousﬁndings by demonstrating that the maturation of cortical surface areawasalsodelayedinchildrenwithADHD(Shawetal., 2012). Dataonpotential developmentalchangesin adulthood were yielded by two MRI meta-analyses (Frodl and Skokauskas, 2012; Nakao et al., 2011), which included cross-sectional data fromchildren and adults(from sepa- ratestudies).Thesemeta-analysesprovidedconsistentev- idence that the age-related alterations in structural-MRI measures diminish in adult ADHD. A recently published structural-MRI study (Onnink et al., 2014) which examined volumetric measures in a large cross-sectional sam- pleofadultADHDpatientsconcludedthat“developmental braindifferencesin ADHDlargelynormalizein adulthood”

(Onninketal.,2014).

ConsistentwiththeaboveMRIdata,ourresults suggest that a normalization of developmental brain differences in ADHD in adulthood may also occur in terms of neuro- physiologicalmeasures.Asthevolumeofthefrontalcortex wasshowntocorrelatewiththeP3amplitude(Fordetal., 2004),thisﬁndingisinlinewiththeassumptionofdelayed developmental maturationof thefrontal cortex in ADHD.

Furthermore, this P3age-effect is consistent, at least in part,withHalperinandSchulz’s(H-S)neurodevelopmental model of ADHD persistence (Halperin and Schulz, 2006), which suggeststhatpartiallydistinct neural andcognitive mechanismsareinvolvedinthecauseofandrecoveryfrom ADHD.

The model assumes that the degree to which the de- velopmentof theprefrontal cortexis abletocompensate for early neural deﬁcits accounts for the improvement of symptoms frequently seen in ADHD with age. We inves- tigated the improvement of a core executive function in ADHD, the BRI, in relation with age, as indexed by the commission-errors. We found that both study groups ex- hibited better performance with increasing age in terms ofcommission-errors,buttheimprovementwasmorepro- nouncedintheADHDsample.Additionally,wefoundapro- longation of RTs with age in both groups, which, in con- trasttothecommission-errors,wasmoresubstantialamong HCs.

WhilepartiallyconsistentwiththeH-Smodel,theﬁnding thattheP3diminutioneffectwasrestrictedtoacertainage rangepointsbeyondthecompensatoryprocessesconsidered inthemodel.Inparticular,withrespecttotheneurodevel- opmentaltimecourseofP3wefoundadeclineofP3ampli- tudeintheADHDgroup,whichstartedfrommid-adulthood.

Asthefronto-central NoGoP3isconsidered toreflectthe functioningofthefrontalareas(whichshowadelayedmat- urationinADHD),ourfindingsareconsistentwiththe‘‘last in,firstout’’hypothesis (Douaudetal.,2014;Raz, 2005), which refers to a mirroring pattern of development and aging of the human brain. As stated by (Douaud et al., 2014), “thesequence of events associated withbrain decline should present itself in reverse order to the series ofeventsrelatedtobraindevelopment,withbrainregions thought to develop relatively late - at both ontogenetic andphylogeneticlevel– alsodegeneratingrelativelyearly"

(Hilletal.,2010; Razetal.,2005;Reisbergetal.,2002).

Hence,ourﬁndingsraisethepossibilitythatADHD– atleast in patients whose symptoms persist into adulthood - may notonlybeassociatedwithadelayedneuro-developmental trajectory,butalsowithaprematureage-relateddeterioration,atleastinsomeareasofneurophysiologicalfunction- ing.

Although impairment of “hot executive” functions was postulated as a key deficit in ADHD (Barkley, 1997; Surman et al., 2013), previous studies used emotionally- neutralstimuliinNoGoBRItaskstoinvestigatethefronto- centralP3(e.g.,(Fallgatter etal.,2004;Fallgatter etal., 2005;Wiersemaetal.,2006)).Inthecurrentstudy,applyingemotionally-neutralaswellasvalencedstimuli,wede- tectedasignificantgroupbyemotionalvalenceinteraction, whichrevealedaneffectofemotionalvalenceintheADHD group:the P3reduction wassignificantlygreater for neg- ativethan for neutralor positive stimuli. Among HCs,we foundnooveralldifferencebyemotionalvalenceacrossall ages.

ThereductionofP3inADHDandthemodulationof the P3 reduction by negative emotional valence may reﬂect the impairment of hot EFs. These ﬁndings can be interpreted in thecontext of cognitive-energetic model (CEM) (Sergeant, 2005), which postulates that ADHD causes de- fectsacross threelevels of hierarchy, includingattention atlevel1(withcomputationalaspectsandmotoroutput);

energeticpools atlevel 2(effort,arousaland activation);

and,the executive/management system (level 3). We as- sume that the overall reduction of P3 in ADHD is associated with a top level deﬁcit in executive/inhibitorycon- trol(whichwouldbelinkedtoimpairmentsin“cool” EFs).

In light of the evidence that the P3amplitude decreases withincreasedarousal(Polich,2007),wethinkthatthead- ditionalP3reduction inADHD by negativestimulimay be viewedaslinkedtothe alterations in theenergetic pool, includingarousal(atlevel2intheCEM).

According to the low arousal theory of the disorder (Barry etal., 2003a; Barry et al., 2003b), ADHD patients arecharacterizedby hypo-arousal,which hasbeen shown tobeassociatedwitharelativeup-regulationofthephasic arousal-response(SikstromandSoderlund,2007).Thismay lead to hypersensitivity to emotionally-salient environ- mentalstimuli,suchasthenegativepicturesinourstudy.

The behavioral ﬁnding that commission-error rates are

(9)

(LittmanandTakacs,2017).

Severalissuesandstudylimitationsmeritcomment.First, the response inhibition task we used involves a working memory (WM) component, as it requires a “same” vs.

“different” decision based on the immediately preceding stimulus. It is therefore conceivable that WM alterations presented a confound in our investigation. Further studies thatsystematicallymanipulate theWM demandinthe task are needed to clarify this issue. Second, the possi- bility of premature deterioration in patients with ADHD in certainelectrophysiologicalindicesof functioning,such asthe age-relatedP3reduction,maybe seen atvariance withsomeof theearlierﬁndings.Forexample,theafore- mentioned mega-analysis of structural-MRI data indicated a later onset of volume decrease in the nucleus accum- bensandputamenintheADHDascomparedtothecontrol group (Hoogmanet al., 2017). However,one has to keep in mindthatvariouscorticaland subcorticalbrainregions have different developmental trajectories (Coupe et al., 2017;Frankeetal.,2018;Storsveetal.,2014).Third,our study included only adults with ADHD, and therefore did not coverthe criticaldevelopmentaltrajectory fromado- lescencetoadulthood.Furthermore,itwascross-sectional, whichraisestheissueofpotentialconfoundingintermsof causalrelationshipsandpossible age-cohorteffects.How- ever, our goalwasnottoidentifycausalassociations,but todescribedevelopmentaltrajectorieswithage.Sincewe continuously sampled an age-rangerather than using dis- tinct age-cohorts, andapplied aregression approach, the age-cohort effects seem less likely.Nonetheless, we con- cur with the conclusions of a recent review of lifespan changesinADHDthat“aseverelackofknowledgeonlifes- panaspectsinADHD stillexists”,andlarge-scaleresearch efforts would be essential to overcome the “knowledge gaps through appropriately granular longitudinal studies”

(Frankeetal.,2018).

Finally,approximatelyhalfofthepatientsreceivedstim- ulantmedication,andaboutone-fourthhadcomorbidities.

However,theinclusionofmedicationandcomorbiditysta- tus inthe analysesdidnotinﬂuence theprincipalresults.

Finally,sincewefocusedonadultpatients,whosesymptom severity metthediagnostic criteria(i.e.,their ADHDper- sisted),ourresultspertaintothisgroupofpatients.Hence, ourﬁndingsmaynotgeneralizetopatientswhosesymptoms donotreachthediagnosticthresholdin adulthood.These patientsmayevidenceamorecompleteandearliernarrow- ing of the age-developmentalgap than thosewhoremain

2014;Raz,2005),whichreferstoamirroringpatternofde- velopment and agingof the human brain.Hence, our re- sultsraisethepossibilitythatADHDmaynotonlybeasso- ciatedwithadelayedneuro-developmentaltrajectory,but alsowithaprematureage-relateddeterioration,atleastin someareasofneurophysiologicalfunctioning.

Role of the funding source

Thisstudy wassupportedby theHungarianBrainResearch Program(2017-1.2.1-NKP-2017-0002).The HungarianBrain ResearchProgramhadnofurtherroleinstudydesign;inthe collection,analysisandinterpretationofdata;inthewrit- ingofthereport;andin thedecisiontosubmitthe paper forpublication.

Contributors

AuthorsKakuszi,CzoborandBitterdesignedstudy.Author Kakusziperformedthedatacollectionandpreprocessingof the EEGdata under the supervision of authors Bitterand Czobor.AuthorsKakusziandCzobormanagedtheliterature searchesandanalyses. AuthorSzuromiparticipated inthe clinicaldatacollection.AuthorsCzoborandKakusziunder- took the statistical analysis. Authors Kakuszi and Czobor wrote the first draft of the manuscript. All authors con- tributedtotheinterpretationofthefindings,andhaveap- provedthefinalmanuscript.

Conﬂict of interest

AuthorsB.Kakuszi,B.Szuromi,I.BitterandP.Czoborde- clarethat theyhave noconﬂictofinterest withregard to thisstudy.

Acknowledgments

The authors would like to acknowledge the help of the teamoftheOutpatientClinicforAdultswithADHD,atthe SemmelweisUniversity,inBudapest,Hungaryfortheirpar- ticipationintherecruitmentandclinicalexaminationofthe subjects.