A longitudinal biosocial study of cortisol and peer influence on the development of adolescent antisocial behavior

Tilburg University

A longitudinal biosocial study of cortisol and peer influence on the development of

adolescent antisocial behavior

Platje, E.; Vermeiren, R.; Raine, A.; Doreleijers, Th.A.H.; Keijsers, L.; Branje, S.T.J.; Popma,

A.; van Lier, P.A.C.; Koot, H.M.; Meeus, W.H.J.; Jansen, L.M.C.

Published in:

Psychoneuroendocrinology

DOI:

10.1016/j.psyneuen.2013.07.006

Publication date:

2013

Document Version

Publisher's PDF, also known as Version of record

Link to publication in Tilburg University Research Portal

Citation for published version (APA):

Platje, E., Vermeiren, R., Raine, A., Doreleijers, T. A. H., Keijsers, L., Branje, S. T. J., Popma, A., van Lier, P. A.

C., Koot, H. M., Meeus, W. H. J., & Jansen, L. M. C. (2013). A longitudinal biosocial study of cortisol and peer

influence on the development of adolescent antisocial behavior. Psychoneuroendocrinology, 38(11), 2770-2779.

https://doi.org/10.1016/j.psyneuen.2013.07.006

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain

• You may freely distribute the URL identifying the publication in the public portal

Take down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

A

longitudinal

biosocial

study

of

cortisol

and

peer

influence

on

the

development

of

adolescent

antisocial

behavior

E.

Platje

*

,

R.R.J.M.

Vermeiren

,

A.

Raine

,

T.A.H.

Doreleijers

,

L.G.M.T.

Keijsers

,

S.J.T.

Branje

,

A.

Popma

,

P.A.C.

van

Lier

,

H.M.

Koot

,

W.H.J.

Meeus

,

L.M.C.

Jansen

a

VUUniversityMedicalCenter,DepartmentofChildandAdolescentPsychiatry,Amsterdam,TheNetherlands b

Curium,LeidenUniversityMedicalCenter,Leiden,TheNetherlands

c_{Departments of Criminology, Psychiatry, and Psychology, Jerry Lee Center of Criminology, University of} Pennsylvania,Philadelphia,USA

d_{UtrechtUniversity,ResearchCenterAdolescentDevelopment,Utrecht,TheNetherlands} e

VUUniversity,DepartmentofDevelopmentalPsychology,Amsterdam,TheNetherlands f_{TilburgUniversity,TilburgSchoolofBehavioralandSocialSciences,Tilburg,TheNetherlands}

Received8February2013;receivedinrevisedform9July2013;accepted10July2013 Psychoneuroendocrinology(2013)38,2770—2779 KEYWORDS HPAaxis; Cortisol; Aggression; Rule-breaking; Peerinfluences; Longitudinal

Summary Itisincreasinglyrecognizedthatinordertounderstandthecomplexphenomenonof antisocialbehavior,interrelationsbetweenbiologicalandsocialriskfactorsshouldbetakeninto account.Inthecurrentstudy,thisbiosocialapproachwasappliedtoexaminethemediatingroleof deviant peersin longitudinal associations linking the level of hypothalamic-pituitary-adrenal (HPA)axisactivitytoaggressionandrule-breaking.

Participantswere425boysandgirlsfromthegeneralpopulation,whowereassessedyearlyat ages15,16,and17.AsameasureofHPAaxisactivity,cortisolwasassessedatawakening,30,and 60minlater(thecortisolawakeningresponse,CAR).Participants,aswellastheirbestfriend, reportedontheirownaggressiveandrule-breakingbehavior,therebyallowingtoassess bidirec-tionalinfluenceswithinfriendships.

Aggression was only predicted by a decreased cortisol level at awakening, and not by aggressivebehavioroftheirfriend.Decreasedlevelsofcortisolatawakeningpredicted adoles-cents’rule-breaking,whichsubsequentlypredictedincreasedrule-breakingoftheirbestfriend. Thelatterwasonlyfoundforadolescentswhochangedfriends,ascomparedtoadolescentswith thesamefriendineveryyear.Genderdifferenceswerenotfound.

* Correspondingauthorat:PObox303,1115ZGDuivendrecht,TheNetherlands.Tel.:+31208901545;fax:+31207745690. E-mailaddress:e.platje@debascule.com(E.Platje).

Availableonlineatwww.sciencedirect.com

j o ur nal h o m ep a ge:w ww.e ls e vi e r.c o m /l o c at e/ p s y ne ue n

1.

Introduction

Itisincreasinglyrecognizedthatinordertounderstandthe complex phenomenonof antisocialbehavior, interrelations betweenbiological and socialrisk factors should betaken intoaccount(Bassarath,2001;DodgeandPettit,2003;Raine, 2002;Susman,2006).Afrequentlyhypothesizedand exam-ined biological risk factor for antisocial behavior, is a decreasedlevelofhypothalamic-pituitary-adrenal(HPA)axis activity(e.g.McBurnett etal.,2000;Popmaetal.,2007). AssociationsbetweenadecreasedlevelofHPAaxisactivity andantisocialbehaviorhavebeenconfirmed,butless con-sistently for adolescent than for childhood samples (see reviewbyAlinketal.,2008).Atthesametime, in adoles-cence deviant peer influence becomes a major social risk factor for antisocial behavior (Brown, 2004; Gardner and Steinberg,2005).Moreover,thereareindicationsthat devi-antpeersmaymediatetheassociationbetweenthelevelof HPAaxisactivityandantisocialbehavior(Raineetal.,2005; Yanovitzky,2005).Tocapturethedevelopmentalchangesin HPAaxisactivitylevels,peerrelations,andantisocial beha-vior,whicharecharacteristic foradolescence,longitudinal studiesarerequired.Therefore,thecurrentstudyfocusedon themediatingroleofpeerinfluencesinlongitudinal associa-tionslinkingdecreasedlevelsofHPAaxisactivitytoantisocial behavior.

Abiologicalperspectiveonthedevelopmentofantisocial behaviorisofferedbythelowarousaltheories(Raine,1993; Zuckermanand Neeb, 1979).Low arousalis consideredto constitute a negative physiological state, which could be increased (i.e., normalized) by seeking sensation through antisocial behavior (Zuckermanand Neeb, 1979). Alterna-tively,lowarousalmightreflectfearlessness,asaresultof whichyoungstersmaynotfearthenegativeconsequencesof antisocial behavior (Raine, 1993). Although the exact mechanismsareunknown,ithasbeenposedintheoretical models that low arousal may have resulted from genetic vulnerabilities or early life adversities. The amygdala is considered to link such early stressors to dysfunctions in arousal(Susman, 2006;vanGoozenetal.,2007). TheHPA axisisoneofthemajorphysiologicalstresssystems,andlow arousal can be operationalized as low levels of HPA axis activity.Asameasure ofHPAaxisactivity,salivarycortisol levels are often assessed. For instance, McBurnett et al. (2000)foundthatpersistentaggressioninschoolagedboys was associatedwith lower daytime cortisol levels.Popma et al. (2007) specifically studied the cortisol awakening response (CAR) in adolescent boys, and reported that the levelof theCAR, but nottheresponse to awakening,was decreasedinantisocialboyscomparedtonormalcontrols.In ameta-analysis,however,associationsbetweenthelevelof HPAaxisactivityandantisocialbehaviorwerenotfoundin adolescentsamples(Alinketal.,2008).

It isin adolescence whenaffiliation with deviantpeers becomes an important social risk factor for developing

antisocial behavior (Brown, 2004; Gardner and Steinberg, 2005; Hartup and Stevens, 1997). Peer influences are dynamicand bidirectional (Dishionand Owen, 2002;Popp etal.,2008):adolescentsselectfriendswhoare similarto themselvesinbehaviorandattitudes(selection),andfriends becomemoresimilartooneanotherovertime(socialization) (BrechwaldandPrinstein,2011;Kandel,1978).Imbalanceor dissimilaritybetweenmutualfriends’behaviorandattitudes islikelytoresultinendingthefriendshipandseekingmore similarfriends, orto stay friendsandmodifying theirown behaviortothatofthefriend(Kandel,1978).Selectionand socializationarenotmutuallyexclusive,butcancoexistand enhance oneanother. For instance, antisocial adolescents mayselect friends showing more antisocial behaviorthan themselves,whichcanexacerbatetheirownantisocial beha-vior(Gattietal.,2005;Thornberryetal.,1993).However, friends also tend to overestimate the similarity between their behaviors, that is, an adolescent may feel his/her friendsareequally antisocialashe/she is,whereasinfact thefriendsmaybelessantisocial(Aseltine,1995).To over-comethisoverestimatingofthesimilarities,andprovidean accurateviewof thefriends’ antisocialbehavior, thebest friendsreportedontheirownbehaviorinthecurrentstudy. Antisocialfriendshipsmaymediateassociationsbetween thelevelofHPAaxisactivityandantisocialbehavior.Ithasfor instancebeen shownthatsensation seeking,as associated withloweredlevelsofHPAaxisactivity(cf.thelowarousal theory, seeabove), is also associated with affiliating with deviantfriends(Yanovitzky,2005).Thesedeviantfriendsin turnmayinfluencetheadolescenttowardbehaving antiso-cially(Moffitt,1993;Thornberry etal.,1994).Also, it has beenshownthatpersistentantisocialyouthshow neurocog-nitiveimpairments comparedto adolescence limited anti-social youth (Raine et al., 2005). This could imply that biologicalriskfactors,includingdecreasedlevelsofHPAaxis activity,maybespecificforpersistentantisocialyouths.As they are already involved in deviant behaviors,in adoles-cencetheyare morelikelyto selectantisocialfriends and influence others into antisocial behavior (Moffitt, 1993). Hence,two paths maybepresentlinking the levelof HPA axisactivitytodeviantfriends:(1)influencesofdecreased levelsofHPAaxisactivitymayoperateviadeviantfriends, andalso(2)lowerlevelsofHPAaxisactivitymayfirstleadto adolescentantisocial behavior,which makes these adoles-centsmorelikelytohaveantisocialfriends.Bothpathwaysof influencewillbetestedinthisstudy.

Toclarifywhenandhowtheinfluenceof friendscomes intoplay,andtocomparethetwopaths,alongitudinaldesign isrequired. To the best of ourknowledge, the onlystudy which has investigated HPA axis activity levels and peer influences, was cross-sectional in nature (Dorn et al., 2009).Dornetal.foundthatchildrenwithdisruptive beha-viordisordersshowedlowestlevelsofHPAaxisactivityifthey hadfriendswhoshowedlowlevelsofantisocialbehavior.As these children already showed antisocial behavior, these Thesefindingssuggestthatinterrelationsbetweenbiologicalandsocialriskfactorsaredifferent forthedevelopmentofaggressionversusrule-breaking.Furthermore,decreasedlevelsofHPAaxis activitymayrepresentasusceptibilitytoselectingdeviantpeers.

findings indicate that theirbehaviorwas not theresult of peer influence. However, these children were aged 6—11 years,anddeviantpeersdonotbecomeamajorriskfactor forantisocialbehavioruntiladolescence(Brown,2004; Gard-nerandSteinberg,2005).Tofurther investigatethese pro-mising findings, and incorporate the dynamics and bidirectionalityofpeerinfluencestowardantisocialbehavior inyouthswithdecreasedlevelsofHPAaxisactivity,a long-itudinaldesignwasappliedinthecurrentstudy.

Furthermore, the relative influence of biological and socialriskfactorsmaydifferbythetypeofantisocial beha-vior. Aggression and rule-breaking are two main types of behavioroftenrecognizedwithinadolescentantisocial beha-vior(e.g.Achenbachetal.,1989;Burt,2012).Bothtypesare thoughttoresultfrombiologicalaswellassocialriskfactors (Moffitt,1993;Raineetal.,2005).However,thereare indi-cations that aggression may be more strongly related to decreasedlevelsofHPAaxisactivity(Burt,2012;McBurnett etal.,2000;Platjeetal.,2013a,b),whereasrule-breaking maybemorestronglyrelatedtoaffiliationwithandinfluence ofdeviantpeers (Barnowetal.,2005;Reitz etal.,2007). Therefore,bothtypesofantisocialbehaviorwereassessedin thisstudy.

Forthesereasons,inthecurrentstudy,theroleofdeviant peerinfluencesinlongitudinalassociationslinkingthelevel of HPA axis activity to aggression and rule-breaking was investigated in a general population sample of both boys andgirls.Becausewithinfriendshipsbidirectionalinfluences canoccur,whichmayincreaseantisocialbehavior,two indir-ect paths wereexamined: (1)do lower levels of HPAaxis activitypredict higherlevels of antisocial behavior ofthe bestfriend,whichinturnpredictshigherlevelsofadolescent antisocialbehavior,(2)dolowerlevels ofHPAaxisactivity predicthigherlevelsofadolescentantisocialbehavior,which inturn predicts higherlevelsof antisocial behaviorofthe bestfriend?Thiswasexaminedinalargesampleofboysand girls,whoparticipatedinthreeannualassessmentsatages 15,16and17.AsameasureofHPAaxisactivity,theCARwas assessed,andspecifiedintwoways;firstlyascortisollevels atawakening,andsecondlyastheresponseincortisollevels toawakening.Theresponsetoawakeninghasbeenshownto be influenced by situational factors (Fries et al., 2009; Hellhammer et al., 2007), therefore associations are expectedtobestrongerforthecortisollevelatawakening. Thebestfriendsoftheadolescentsalsoparticipated, report-ingontheirownaggressiveandrule-breakingbehaviorover theyears.The bestfriendcould changefromyeartoyear, andstabilityofthefriendshipwasexaminedtoaccountfor selectionandsocializationeffects.Asthesampleconsisted ofbothboysandgirls,andgenderdifferencesmaybepresent inantisocialbehavior(Moffitt,2001)and/orHPAaxisactivity (e.g.Friesetal.,2009),genderwastakenintoaccountinall analyses.

2.

Methods

2.1. Participants

Participantswere425adolescents(239boys,186girls)taking partinthreeannualassessments;atages15,16and17years. TheywererecruitedfromtheRADAR(ResearchonAdolescent

Development And Relationships) study. RADAR is a Dutch populationbased cohortstudy,withover-sampling(50%) of boysandgirlswithaborderline-clinicalscoreonthe externa-lizing scaleoftheTeacher’s Report Form(TRF, Achenbach, 1991a) at age 11. All participants and their parents have providedwritteninformedconsentandreceiveda reimburse-ment for their participation. The RADAR study has been approvedbytheresponsiblemedicalethicscommittee,and wasconductedinaccordancewiththeDeclarationofHelsinki. Thisstudyisbasedondatafromthethird(2008)tothefifth wave(2010)ofRADAR,inthispaperreferredtoasages15—17. Ofthe425participants,379adolescents(89.2%) partici-patedintheHPAaxismeasurementsatanyyear.Participants intheHPAaxismeasurementsdidnotdifferinage,gender, pubertalstatus,BMI,nicotineuseatage16and17,oralcohol useatage15and16(t-testsandx2,allps>.05)fromthose who didnot participate,yet participants moreoftenused alcoholatage17(x2(1)=5.016,p=.05)andnicotineatage 15 (x2(1)=4.483,p=.05).After exclusion on the basis of samplingerrors,technicalproblemsinthelab,orstatistical outliers(seebelow),for362adolescents(332atage15,283 at age 16 and 254 at age 17) HPAdata was available for analyses.

Adolescents were asked to invite their best friend to participate in the study, andat any year, for 407 (95.8%) adolescentstheirbestfriendparticipated.Atage15for387 adolescentsthebestfriendparticipated,for381atage16, and361atage17.Theydidnotdifferfromparticipantsfor whomnofriendparticipatedongender,pubertalstatus,BMI, nicotineoralcoholuse(t-testsandx2,allps>.05),butwere on average 3 months younger (t(423)=3.719, p001). Reciprocityinthesefriendshipswasexamined,andthelarge majority(91.6%atage15,98.4%atage16,and98.3%atage 17)ofbestfriendsmentionedtheadolescentasafriend.

Adolescents were specifically instructed to invite their bestfriend,andcouldthereforeinviteanother bestfriend fromyearto year.Allbest friendsweregiventheirownID numberinthestudy,enablingtheexaminationofstabilityof thefriendship.Atoneorbothintervals,150(38.8%) adoles-centschangedfriends.

Thenumberofparticipantsfluctuatedperyear,with417 participants at age 15, 405 at age 16 and 389 at age 17. Attritionwaslowoverthethreeyears,16(3.4%)droppedout atage16,and19(4.7%)atage17.Drop-outwas not asso-ciated with gender (x2_{(1)=1.899, p=.18), but drop-outs} wereon average 3months older(t(423)=5.027,p001). To estimate thepattern of missing values, Little’s Missing Completelyat Random(MCAR)test(Little,1988)was con-ducted. Although this very stringent test was significant (x2(2376)=2609.032,p=.001),thex2/dfratioof1.10 indi-cateda good fit between sample scoreswith and without imputation(Bollen,1989).Participantswithpartiallymissing datacouldthusbeincludedintheanalyses.Thefinalmodels wereranon the425adolescentsparticipatingat anyyear, applying afull-informationmaximum likelihoodestimation (EndersandBandalos,2001).

2.2. Aggressionandrule-breaking

Antisocialbehaviorwasassessedbymeansofthe externaliz-ingscalesoftheYouthSelfReport(YSR,Achenbach,1991b).

GoodreliabilityandvalidityhavebeenreportedfortheDutch YSRversion(Verhulstetal.,1997).Theadolescentsandtheir bestfriendseachreportedontheirownbehavior.Withinthe externalizingdimension,sub-scalesdifferentiateaggression and rule-breaking behavior. The aggression subscale con-sisted of 19 items (a’s ranging from .86 to .87) assessing physicalactsagainstpersonsorthings(i.e.,fighting,being crueltoothers).Therule-breakingsub-scaleconsistedof11 items(a’srangingfrom.70to.73)assessingbehaviorssuchas truancyandstealing.Itemsarescoredonathree-pointscale (0=nottrue,1=somewhattrue,2=verytrueoroftentrue). Missingswerehandledaccordingto theYSRmanual guide-lines;nomorethan2missingswereallowedandthesewere replacedbythemeanofthesub-scale.

2.3. HPAaxisassessment

Cortisol was measured in saliva. Saliva samples were col-lected by passive drooling, immediately after awakening (Cort0),and30min(Cort30)and60min(Cort60)later.These three samples constitute the CortisolAwakening Response (CAR,Pruessneretal.,1997).Cortisolsamplingtookplacein February and March of each consecutive year, as soon as possibleafterassessingaggressionandrule-breaking. Parti-cipantswerefirstgivendetailedverbalandwritten informa-tionregardingcortisol measurements. Subsequently,saliva sampling was planned for a suitable morning on a regular weekday.Thefirstsample(atawakening)wasplannedbefore 8 a.m., while taking into consideration the participant’s normalschedule. Sampling timeswereset andwritten on adetailedinstructionform.

Participants wereinstructedto rinse theirmouths with waterbeforesampling,andnottoeat,drinkmilkorjuice, smokeorbrushtheirteethbeforecompletingCort60.They wererequestedto reporttheexact samplingtimeson the instructionformonthedayofsampling,andalsotoreportif mistakesweremadeinanyoftheaboveinstructions.After collection,participantswereaskedto storethesamples in therefrigeratorandsendthembymailtotheresearchcenter thesameday.

At the research center, all samples were checked for correctness ofsampling. When necessary, e.g.whenCort0 wassampledafter8:00a.m.orsamplingtimeofCort30or Cort60wasover15minlate,ormistakesweremadeinanyof the other instructions, participants were asked to collect newsalivasamples,andanewsamplingdaywasscheduled. Atage15,28participantscollectednewsalivasamples,20at age16,and15atage17.If,despitethis,participantshadstill notsampledcorrectly,theincorrectsampleswereexcluded. Intotal39samples(4samplesatage15,15atage16,and20 atage17)wereexcludedforincorrectsampling.

Salivawasstoreduncentrifugedat208Cuntilanalysis. Salivarycortisollevelswereanalyzedusing electrochemilu-minescenceimmunoassayECLIA(E170Roche,Switzerland). Thelower detectionlimitwas0.5nmol/l,andmean intra-assayandinter-assaycoefficientsofvariationwere respec-tively3.4%and12.2%.Duetotechnicalproblemsinthelab (i.e.in84%ofthesamplestoolittlesalivawaspresentand6% contaminatedsamples),133samplescouldnotbeassayed. Participantswithsamplesthatcouldnotbeassayeddidnot differinage,gender,pubertalstatus,BMI,nicotineuseatage

15oralcoholuse(t-testsandx2,allps>.05),yetmoreoften used nicotine at age 16 (x2(1)=10.286, p=.003) and 17 (x2(1)=4.920,p=.034).

2.4. Control variables

Allcontrolvariableswereassessedyearlythoughself-report atthesametimesasaggressionandrule-breaking.As phy-sicaldevelopment,substanceuseandstressfulexperiences usearerelatedto developmentofHPAaxisactivity(Platje etal.,2013a,b;Trickettetal.,2010)andmaybeassociated withantisocialbehavior,wetookthesevariablesintoaccount intheanalyses.Physicaldevelopmentwasassessedas pub-ertaldevelopmentandthebodymassindex.Pubertal devel-opment was measured by a modification of the Pubertal DevelopmentScale (PDS, Petersenet al.,1988) consisting of seven questions regarding physical development, i.e. growthspurt,axillaryhair,pubarche,menarche,thelarche, voicechangeandfacialhair.Itwasassessedonlyatages15 and16,asitwasexpectedthatatage17thelargemajority wouldbefullymatured.Atage15,41.9%scoredin/overlate pubertalrange,atage16thiswas73.6%.Thebodymassindex was calculated from self-reported height and weight as weightinkg/(lengthinm)2.Substanceusewasassessedas nicotineandalcoholuse(Monshouweretal.,2008).Alcohol useoverthelastfourweekswasassessedbymeansofa six-optionquestion,rangingfrom‘‘none’’to‘‘daily’’.Nicotine usewasassessedbyanine-optionquestionrangingfrom‘‘I have never smoked’’ to ‘‘I smoke every day’’. Stressful experiencesinthepastyear,suchassexualassault,physical assault,andbeingthreatenedwithviolence,wereassessed withaquestionnairebasedon theInternationalCrime Vic-tims Survey (ICVS; Nieuwbeerta, 2002), and specified by perpetrator(parent=2,someoneelse=1,not=0). 2.5. Statistical analyses

Cortisolvalues over 3SDabove themean were defined as outliersandexcluded(28samples).TheCARwasdefinedas the cortisol level at awakening (CARlevel) and the cortisol response to awakening (CARresponse) per year, with Latent GrowthModeling (LGM;e.g. Kline,2005) within Mplus6.0 (Muthe´n andMuthe´n, 2007)with maximumlikelihood esti-mation(Satorra andBentler,1994).Foreach year,cortisol levelsatawakening,and30and60minlater,wereusedas indicatorstoestimatethelatentintercept(i.e.thelevelof cortisolatawakening—CARlevel)andslope(i.e.changesin cortisolfrom awakeningthrough 30and60minafter awa-kening— CARresponse) factors in LGM.The CARlevelandthe CARresponse werenormally distributed. Of the control vari-ables,onlysubstanceusewasassociatedwithpredictorsas wellasoutcomevariables.TheCARwasthereforecontrolled forsubstanceuseeffects,byaddingalcoholandnicotineuse inthemodelfortheCARlevelandCARresponseastime-varying ordinal covariates on the cortisol levels at awakening for eachyear.

levelandresponse,andbestfriend’santisocialbehavior.The 1-and2-yearstabilityeffectsovertheyearsforallvariables wereadded,andwithin-yearcorrelationsbetweenthe vari-ables, as longitudinal associations were of interest. Two annualintervalswereavailabletoassesslongitudinaleffects (ages15—16,andages16—17).Sixlongitudinalcross-lagged effectswereestimatedperannualinterval,from:(1)CARlevel to adolescent behavior one year later, (2) CARresponse to adolescent behavior one year later, (3) CARlevel to best friend’s behavior one year later, (4) CARresponse to best friend’sbehavior oneyear later,(5) adolescent’s behavior tobestfriend’sbehavioroneyearlaterand(6)bestfriend’s behaviortoadolescent’sbehavioroneyearlater.Notethat paths5and6 arebidirectional, theotherpaths werealso tested for bidirectionality, but non-significant Wald tests indicatedthatthese could beremovedto keepthemodel asparsimoniousaspossible(foraggression:Wald(8)=1.976, p=.98; for rule-breaking: Wald(8)=6.979, p=.54). The cross-laggedpathswerefoundtobetime-invariant,as con-strainingtheeffectsofage15—16,andfromage16—17,tobe estimatedthesamedidnotworsenmodelfit(foraggression: Wald(6)=6.812,p=.34;forrule-breaking:Wald(6)=7.237, p=.30).Thismodification wasthereforeretained,tokeep themodelasparsimoniousaspossible.Finally,themediating role of peer influences was modeled by estimating two indirectpaths:oneindirectpathfromCARlevelviathe ado-lescent’sbehaviortothebestfriend’sbehavior,andonewith anindirectpathfromCARlevelviathebestfriend’sbehaviorto theadolescent’sbehavior.Totestwhetherdifferentmodels forboysandgirls,orstableandchangingfriendshipswouldbe warranted, the models as describedabove werealso per-formed as multi-group models by gender, and stability of friendship.Differencesinthecrosspathsdueto genderor stabilityoffriendshipweretestedbyconstrainingthepaths tobeequal,andevaluatingadecreaseinmodelfitwith chi-square difference tests. Significant chi-square tests would indicate thatthe paths weresignificantly different. Ifthe pathsdidnotdifferforboysandgirls,genderwascontrolled forbyaddinggendertothemodelasacovariateatage15.

3.

Results

InTable1descriptivestatisticsareshown.Itcanbeseenthat theCARlevelwasnotcorrelatedtoaggressionor rule-break-ing,andaweakpositivecorrelationwasfoundbetweenthe

CARresponseandadolescentrule-breaking.Thelevelof aggres-sionorrule-breakingwasmoderatelypositivelycorrelatedto aggressionorrule-breakingofthebestfriend.These correla-tionswerecomparableforage16andage17.

Adolescentswhochangedbestfriends,showedmore rule-breakingbehavioratage15,thanadolescentswhokeptthe same best friend (t(382)=2.091, p=.04). No effects of stabilityoffriendshiponrule-breakingbehaviorwerefound atage16or17,onaggressivebehavioroftheadolescentor aggressionorrule-breakingbehaviorofthebestfriendatany time(t-tests,allps>.05).

First,itwasexaminedwhethergenderdifferenceswere presentinlongitudinalinterrelationsbetweentheCAR, anti-socialbehaviorofthebestfriend,andadolescentantisocial behavior.Therefore, structural equationmodels were per-formed for aggression and rule-breaking with gender as groupingvariable inmulti-group models.Overall thecross paths didnot differ between boys or girls, for aggression (Dx2(6)=2.945, p=.82) or rule-breaking (Dx2(6)=2.758, p=.84).

ToexaminelongitudinalinterrelationsbetweentheCAR, antisocialbehaviorofthebestfriend,andadolescent anti-socialbehavior,structuralequationmodelsweretestedfor aggression and rule-breaking respectively. All variables showedsignificantstabilityovertime,exceptfortheCAR re-sponse.ResultsofthecrosspathsbetweentheCAR,antisocial behaviorofthebestfriend,andadolescentantisocial beha-viorareshowninTable2.

As can be seen in Fig. 1 and Table 2, higher levels of aggressivebehaviorwerepredictedbyalowerCARlevel,over andabovetheeffectofaggressivebehaviorintheprevious year. Aggressive behavior of the adolescent was not pre-dictedbythelevelofaggressionofthebestfriendor vice versa.

Higherlevelsofrule-breakingbehavioroftheadolescent werepredictedbyalowerCARlevel,overandabovetheeffect ofrule-breakingbehavioroftheadolescentintheprevious year. Also, more rule-breaking behaviorof the adolescent predictedincreasedrule-breakingofthebestfriendoneyear later,overandabovetheeffectrule-breakingbehaviorofthe best friendinthepreviousyear.Rule-breaking behaviorof theadolescentwasnotpredictedbythelevelof rule-break-ingofthebestfriend(seeTable2).

Basedonthenotionthatbidirectionalinfluenceswithin friendships toward aggression or rule-breaking may occur, two indirect paths were examined: (1) from CARlevel to

Table1 Descriptivestatistics.

MeansandSDs Correlationsatage15

Age15 Age16 Age17 1 2 3 4 5

behaviorofthebestfriend,tobehavioroftheadolescent, and (2) from CARlevel to behavior of the adolescent, to behavior of the best friend (see Table 2). For aggressive behavior, both indirect paths were not significant. For rule-breaking behavior, however,the second indirect path wassignificant:alowerCARlevelatage15predictedincreased adolescent rule-breaking at age 16, which subsequently predictedincreasedrule-breakingofthebestfriendatage 17(seeTable2).

Additionally,as150adolescentschangedfriends,itwas examined whetherthe effectswoulddiffer bystability of friendship.Foraggression,overallthecrosspathsdidnotdiffer between changing or stable friendships (Dx2(6)=6.790, p=.34).Forrule-breakinghowever,thecrosspathsdiddiffer by stability of friendship (Dx2(6)=13.046, p=.04). The effectsasdescribedabovewereonlypresentforadolescents whochangedfriends.AscanbeseeninTable3andFig.2,

higherlevelsofrule-breakingbehavioroftheadolescentwere predictedbyalowerCARlevel,overandabovetheeffectof rule-breakingbehavioroftheadolescentinthepreviousyear. Higherlevelsofrule-breakingbehaviorofthebestfriendwere predictedbyrule-breakingbehavioroftheadolescent. Rule-breakingbehavioroftheadolescentwasnotpredictedbythe levelof rule-breaking of thebest friend. Also, an indirect effect was found: a lower CARlevel at age 15 predicted increasedadolescentrule-breakingat age16,which subse-quentlypredictedincreasedrule-breakingofthebestfriendat age17.

4.

Discussion

Inthecurrentstudy,themediatingroleofantisocialbehavior ofbestfriendsinlongitudinalassociationslinkingthelevelof HPAaxisactivityto rule-breaking andaggressivebehavior, Table2 Resultsofthestructuralequationmodelsforaggressionandrule-breaking.

Directeffects Aggression Rule-breaking

b b

CARlevelage15 !ASBadolescentage16 0.051* 0.072**

CARresponseage15 0.036 0.013

ASBbestfriendage15 0.015 0.036

CARlevelage16 !ASBadolescentage17 0.068* 0.090**

CARresponseage16 0.046 0.015

ASBbestfriendage16 0.015 0.036

CARlevelage15 !ASBbestfriendage16 0.037 0.006

CARresponseage15 0.013 0.035

ASBadolescentage15 0.028 0.100***

CARlevelage16 !ASBbestfriendage17 0.049 0.007

CARresponseage16 0.016 0.041

ASBadolescentage16 0.030 0.098***

Indirecteffects

CARlevelage15!ASBbestfriendage16!ASBadolescentage17 0.001 0.000 CARlevelage15!ASBadolescentage16!ASBbestfriendage17 0.005 0.008* Note.Correlationsbetweenvariableswithinyearsandstabilityofvariablesoveryearswerealsoincludedinthemodels.

*_p.05. **_p.01. ***_p.001.

wasinvestigated.Resultsrevealedthatadecreasedlevelof HPAaxisactivitypredictedadolescentrule-breaking,which subsequentlypredictedincreasedrule-breakingofthebest friend. This was only found for adolescents who changed friends,andnotforthosewithstablebestfriends. Aggres-sion,ontheotherhand,wasonlypredictedbyadecreased levelofHPAaxisactivityandnotassociatedwithaggressive behaviorof friends. These effects were present over and abovethepredictionbypriorantisocialbehavior.

Thefindingsonrule-breakingbehaviorindicatethat ado-lescents with a decreased level of HPA axis activity may changetheirfriendships towardmorerule-breaking peers. Inlinewiththelowarousaltheories,adecreasedlevelofHPA axis activity predicted antisocial behavior (Raine, 1993; ZuckermanandNeeb, 1979). Decreased levels ofHPA axis activity may be specific for persistent antisocial youths (Raine et al., 2005), who in adolescence, already being involvedindeviantbehaviors,aremorelikelytoalsoselect

friendswhoareantisocial(Moffitt,1993).Itmayalsoindicate thatnon-deviantadolescentsrejectfriendswhoare becom-ing increasinglydeviant. Nevertheless, the current results pointtoaselectioneffectratherthanasocializationeffect asitwaspresentforthosewhochangedfriendsonly.Thisisin linewithKnechtetal.(2010)whoalsofoundevidencefora selectioneffectindeviantfriendships,butnotfora socia-lization effect. However, scholarslargelyagree thatthese processescoexist(HaynieandOsgood,2005;Kandel,1978). Attentionisthereforewarranted,asbyaffiliatingwithother antisocialyouthsinadolescence,antisocialbehaviorwithin thesefriendshipsislikelytoexacerbate(Gattietal.,2005). Aggression and rule-breaking are not only different expressions of antisocialbehavior, thecurrent resultsalso pointtoadifferentbiosocialinterplayassociatedwiththese behaviors. This is in accordance with a meta-analysis on heritabilityof aggressionandrule-breaking,whichshowed that whereas aggressionwas largelyinfluenced bygenetic Table3 Resultsofthestructuralequationmodelsforrule-breaking,bystabilityoffriendship.

Directeffects Rule-breaking Changingfriends Samefriends

b b

CARlevelage15 !ASBadolescentage16 0.133* 0.044

CARresponseage15 0.007 0.035

ASBbestfriendage15 0.079 0.011

CARlevelage16 !ASBadolescentage17 0.119* 0.065

CARresponseage16 0.008 0.039

ASBbestfriendage16 0.072 0.012

CARlevelage15 !ASBbestfriendage16 0.075 0.043

CARresponseage15 0.044 0.030

ASBadolescentage15 0.244*** _0.044

CARlevelage16 !ASBbestfriendage17 0.071 0.063

CARresponseage16 0.054 0.033

ASBadolescentage16 0.219*** _0.046

Indirecteffects

CARlevelage15!ASBbestfriendage16!ASBadolescentage17 0.016 0.002 CARlevelage15!ASBadolescentage16!ASBbestfriendage17 0.087* 0.006 Note.Correlationsbetweenvariableswithinyearsandstabilityofvariablesoveryearswerealsoincludedinthemodels.

* _p.05. ***_p.001.

Figure2 Cross-laggedpanelmodel predicting rule-breaking behaviorof theadolescent and thebestfriend, by rule-breaking behaviorofrespectively thebestfriendand theadolescent,aswellasCARleveland CARresponse.Modelfor changingfriendships. Mediation(indirecteffect)isdepictedbythedashedarrow.Note.Modelfit:X2_{(76)=125.286;RMSEA=0.058;CFI=0.959,TLI=0.916.} Onlysignificanteffectsaredisplayed.*p.05,***p.001.

factors,withlittleenvironmentalinfluences,theinfluenceof genetic factors was significantly smallerfor rule-breaking, whereenvironmentalinfluencehadanimportantrole(Burt, 2009).Also,anotherneurobiologicalparameterforlow arou-sal,lowrestingheartrate,hasbeenfoundtopredict aggres-sion, but not rule-breaking (Raine et al., 1997), whereas deviant peers have been found to be closely related to rule-breaking,andmuchlessto aggression(Barnowetal., 2005). Together these results suggest, that although both aggression and rule-breaking result from an interplay between biological and social risk factors, the balance may differ by behavior. Aggression appears to be more stronglyassociatedwithbiologicalrisk,whereas rule-break-ingappearstobeassociatedwithbothbiologicalandsocial riskfactors.

Thefindingsasdescribedabovewereonlypresentforthe CARlevel,thatis,thelevelofHPAaxisactivityatawakening, and not for the response to awakening. This could be explained bythe finding thatthe CARresponse did notshow stabilityovertheyears.Indeed,theresponsetoawakening hasbeenshowntobeinfluencedbysituationalfactors(Fries etal., 2009; Hellhammeret al.,2007), andmatures over adolescence(Platjeetal.,2013a,b).ThelevelofHPAaxis activity may be a better reflection of trait-like HPA axis activity(Hellhammeretal.,2007),whichfacilitatesfinding longitudinalassociationswithbehavior.Anotherexplanation maybethat,becauseofthecontinuedmaturationin adoles-cence, the response to awakening was often negative. A negative response may need a different interpretation, whichcouldalsoexplaintheabsenceofanassociationwith antisocialbehavior.

Inordertoobtainalargervarianceinantisocialbehavior, youth with teacher-reported borderline clinical scores on externalizingbehavioratage11wereover-sampled (50%). Atages15—17only11—14%scoredintheborderlineclinical rangeaccordingto self-report,whichisonlyslightlyhigher thanwhatwouldbeexpectedwithoutover-sampling( Achen-bach,1991a).Thissamplethuslargelyreflectgeneral popu-lationadolescentantisocialbehavior,andresultscannotbe generalized to clinicalor referred samples. Noteworthy is that equal numbers of boys and girls were over-sampled, resultinginrelativelymoregirlsshowinghighlevelsof anti-socialbehaviorthangenerallyfoundinthegeneral popula-tion. Due to this equal gender distribution, the level of antisocialbehavioringirlsissimilartothatinboys.

Althoughinvestigatinggenderdifferenceswasnotanaim ofthisstudyandthiswasnotexaminedin-depth,genderwas taken intoaccount. As overall gender differences did not becomeapparent,thissuggeststhattheroleofdeviantpeer influencesinlongitudinalassociationslinkingthelevelofHPA axis activity to aggressionand rule-breaking is similar for boysandgirls.However,furtherresearchisneededto con-firmthis,especiallyasgirlshavebeenhypothesizedtoshow differential interpersonal expressions of HPA reactivity to stress,markedby‘‘tendandbefriend’’responses,asopposed to‘‘fightorflight’’responses(Tayloretal.,2000).

Animportant avenuefor future longitudinalresearch is the causal directionality of associations betweenHPA axis activity andantisocial behavior.This isimportant because the HPAaxis isconsidered to be oneof the mainsystems involvedinadaptationtotheenvironment(McEwen,2004). Behavingantisociallyfrequently couldtheoreticallyleadto

habituation,whichmayresultin(further)decreasedHPAaxis activity(vanGoozenetal.,2007).Itmayalsoleadtomore extremeforms of sensationseeking, and potentially more severe antisocial behavior. The body could continuously adapttoarousalasaresultofsensationseeking,andrequires increasingly more arousal to achieve the original effects. AlthougheffectsfromantisocialbehaviortoHPAaxisactivity werenottheaimofthecurrentstudy,wedidexaminesuch reverseassociations,whichwerenotfound.Thismay indi-catethatsensationseekingthroughantisocialbehaviorwas successful,withouthabituationeffects.Perhapsinyounger samples, or with a shorter assessment interval, possible bidirectionaleffectsmayberevealed.

Thereare somemethodologicallimitationsofthestudy thatshouldbenoted.First,eachyeartheCARwasassessedin salivasampledathomeononedayonly.Correctingfor day-to-dayvariationwasthereforenotpossible.Previousstudies havehoweverreportedthattheCARshowsmediumtohigh stability across days (Edwards et al., 2001; Kudielka and Wust, 2010; Roisman et al., 2009; Wust et al., 2000). Althoughwe took all possible precautionsin the sampling procedure,amongwhichself-reportofexactsamplingtimes, directly monitoring participant’s compliance to the CAR assessment was not possible. However, self-reported sam-plingtimeshavebeenfoundto bepreferabletoautomatic timerecording(Kraemeretal.,2006) andsampling ofthe CARathomewaspreviouslyfoundnottodifferfromsampling in a controlled laboratory environment (Wilhelm et al., 2007).Second,informationon medicationusewas unavail-able,andcouldnotbecontrolledforintheanalyses. How-ever,psychostimulantia,whicharemostfrequentlyusedby antisocialadolescents,havepreviouslybeenreportednotto be associated with differences in salivary cortisol in the morning(Hibeletal.,2007).Third,useoforalcontraceptives wasnotcontrolledfor,asthisinformationwasonlyavailable forcirca60% ofthegirls,andno effectson theCARwere previously found in this sample (Platje et al., 2013a,b). AnotherstudyshowedthatgirlsusingOCdisplayedaslightly bluntedresponse,buttheleveloftheCARwasnotdifferent fromthatoffree-cyclinggirls(Boumaetal.,2009).Asthe findingsinthecurrentstudywerefoundonthelevelofthe CAR,potentialeffectsofOCuseareexpectedtobeminimal. Fourth, this study was performed in a general population sample,andalthoughyouths withscores intheborderline clinicalrangeatage11wereover-sampled,thiswas effec-tiveonlytoalimitedextend,resultsthereforereflect nor-mative levels of antisocial behavior. As such, the results cannot be generalized to e.g. clinic-referred youths with disruptivebehavior disordersor severedelinquent popula-tions.Asespeciallyfor themostseverely disturbedyouths theirneurobiologicaldeficitsareexpectedtointeract cumu-latively with their adverse social environment (Moffitt, 1993),biosocialstudiesare particularlywarrantedinthese youths,inordertoprovidetoolstointerveneinthisadverse process.

friends, future research should investigate whether the same principle holds true in groups of antisocial youths. Furthermore,theseresultsaddtotheincreasingevidence ofdifferentpathwaystoaggressionandrule-breaking.This implies that eventually, intervention and preventionmay alsoneedtobedirectedatbehaviorspecificfactorsinorder tobe successful.Therefore,moreresearch isrequired to elucidate the possible different etiologies for aggression andrule-breaking,andprovidetools for behaviorspecific interventions.

Role

of

the

funding

sources

The RADAR study has been financially supported by main grants from the Netherlands Organisation for Scientific Research (GB-MAGW 480-03-005, GB-MAGW 480-08-006, GB-MAGW 400-05-212), and Stichting Achmea Slachtoffer enSamenleving(SASS), and various other grantsfrom the Netherlands Organisation for Scientific Research, the VU UniversityAmsterdam andUtrechtUniversity. Thisproject hasbeensupportedbytheFoundation‘‘DeDrieLichten’’in theNetherlands.Thefundingsourceshadnofurtherrolein studydesign,collection,analysisandinterpretationofdata, norin thewriting of the manuscript or in thedecision to submitthepaperforpublication.

Conflict

of

interest

Allauthorsdeclarethattheyhavenoconflictofinterest.

Acknowledgements

DataoftheRADARstudywereused.RADARhasbeen finan-ciallysupportedbymaingrantsfromtheNetherlands Orga-nisationforScientificResearch,StichtingAchmeaSlachtoffer enSamenleving(SASS),andtheVUUniversityAmsterdamand UtrechtUniversity.Thisprojecthasbeensupportedbythe Foundation‘‘DeDrie Lichten’’in theNetherlands.We sin-cerelythankallparticipatingfamiliesandschoolsfortaking partinthisstudy.

References

Achenbach,T.M.,1991a.Manual for the Teachers Report Form and 1991 Profile. University of Vermont, Department of Psychiatry, Burlington, VT.

Achenbach,T.M.,1991b.Manual for the Youth Self-Report and 1991 Profile.UniversityofVermont,Department ofPsychiatry, Bur-lington,VT.

Achenbach,T.M.,Conners,C.K.,Quay,H.C.,Verhulst,F.C.,Howell, C.T.,1989. Replication ofempirically derivedsyndromesas a basis for taxonomy of child/adolescent psychopathology. J. Abnorm.ChildPsychol.17,299—323.

Alink,L.R.A.,vanIjzendoorn,M.H.,Bakermans-Kranenburg,M.J., Mesman,J.,Juffer,F.,Koot,H.M.,2008.Cortisoland externaliz-ingbehaviorinchildrenandadolescents: mixedmeta-analytic evidencefortheinverserelationofbasalcortisol andcortisol reactivity with externalizing behavior. Dev. Psychobiol. 50, 427—450.

AseltineJr., R.H., 1995. A reconsideration ofparental and peer influences onadolescent deviance. J. Health Soc. Behav.36, 103—121.

Barnow,S.,Lucht,M.,Freyberger,H.J.,2005.Correlatesof aggres-siveanddelinquentconductproblemsinadolescence.Aggressive Behav.31,24—39.

Bassarath,L.,2001.Conductdisorder:abiopsychosocialreview.Can. J.Psychiatry46,609—616.

Bollen,K.,1989.StructuralEquationModelingwithLatentVariables. Wiley, New York.

Bouma,E.M.,Riese,H.,Ormel,J.,Verhulst,F.C.,Oldehinkel,A.J., 2009. Adolescents’ cortisol responses to awakening and social stress: effects of gender, menstrual phase and oral contracep-tives.TheTRAILSstudy.Psychoneuroendocrinology34,884—893. Brechwald,W.A.,Prinstein,M.J.,2011.Beyondhomophily:adecade ofadvances inunderstandingpeerinfluenceprocesses.J.Res. Adolesc.21,166—179.

Brown,B.B.,2004.Adolescents’relationshipswithpeers. Handbook ofAdolescentPsychology,Vol.2.pp.363—394.

Burt,S.A.,2009.Aretheremeaningfuletiologicaldifferenceswithin antisocialbehavior?Resultsofameta-analysis. Clin.Psychol. Rev.29,163—178.

Burt,S.A.,2012.Howdoweoptimallyconceptualizethe heteroge-neity within antisocial behavior? An argument for aggressive versus non-aggressive behavioral dimensions. Clin. Psychol. Rev.32,263—279.

Dishion,T.J.,Owen,L.D.,2002.Alongitudinalanalysisoffriendships and substance use: bidirectional influence from adolescence to adulthood. Dev. Psychol. 38, 480.

Dodge, K.A., Pettit, G.S.,2003. A biopsychosocial model of the development of chronic conduct problems in adolescence. Dev. Psychol.39,349—371.

Dorn,L.D.,Kolko,D.J.,Susman,E.J.,Huang,B.,Stein,H.,Music,E., Bukstein, O.G., 2009. Salivary gonadal and adrenal hormone differencesinboysandgirlswithandwithoutdisruptivebehavior disorders:contextualvariants.Biol.Psychol.81,31—39. Edwards,S.,Clow,A.,Evans,P.,Hucklebridge,F.,2001.Exploration

oftheawakeningcortisolresponseinrelationtodiurnalcortisol secretoryactivity.LifeSci.68,2093—2103.

Enders,C.K.,Bandalos,D.L.,2001.Therelativeperformanceoffull informationmaximumlikelihoodestimationformissingdatain structuralequationmodels.Struct.Equ.Model.8,430—457. Fries,E.,Dettenborn,L.,Kirschbaum,C.,2009.Thecortisol

awak-eningresponse(CAR):factsandfuturedirections.Int.J. Psycho-physiol. 72, 67—73.

Gardner,M.,Steinberg,L.,2005.Peer influence on risk taking, risk preference, and risky decision making in adolescence and adult-hood: an experimental study. Dev. Psychol. 41, 625—635. Gatti,U.,Tremblay,R.E.,Vitaro,F.,McDuff,P.,2005.Youthgangs,

delinquencyanddruguse:atestoftheselection,facilitation, and enhancementhypotheses. J.ChildPsychol.Psychiatr. 46, 1178—1190.

Hartup,W.W.,Stevens,N.,1997.Friendshipsandadaptationinthe lifecourse.Psychol.Bull.121,355.

Haynie,D.L., Osgood,D.W.,2005.Reconsideringpeersand delin-quency:howdopeersmatter? Soc.Forces84,1109—1130. Hellhammer, J.,Fries,E.,Schweisthal,O.W.,Schlotz,W.,Stone,

A.A.,Hagemann,D.,2007.Severaldailymeasurementsare nec-essarytoreliablyassessthecortisolriseafterawakening: state-andtraitcomponents.Psychoneuroendocrinology32,80—86. Hibel,L.C.,Granger,D.A.,Cicchetti,D.,Rogosch,F.,2007.Salivary

biomarker levels and diurnal variation: associations with medica-tions prescribed to control children’s problem behavior. Child Dev. 78, 927—937.

Kandel,D.B.,1978.Homophily, selection, and socialization in ado-lescentfriendships.Am.J.Sociol.427—436.

Kline, R.B., 2005. Principles and Practice ofStructural Equation Modeling, 2nded.TheGuildfordPress,NewYork,NY. Knecht,A.,Snijders,T.A.B.,Baerveldt,C.,Steglich,C.E.G.,Raub,

W.,2010.Friendshipand delinquency:selectionandinfluence processesinearlyadolescence.Soc.Dev.19,494—514.

Kraemer, H.C., Giese-Davis, J., Yutsis, M., O’Hara, R.,Neri, E., Gallagher-Thompson,D.,Taylor,C.B.,Spiegel,D.,2006.Design decisionstooptimizereliabilityofdaytimecortisolslopesinan olderpopulation.Am.J.Geriatr.Psychiatry14,325—333. Kudielka,B.M.,Wust,S.,2010.Humanmodelsinacuteandchronic

stress:assessingdeterminantsofindividual hypothalamus—pitui-tary—adrenal axis activity and reactivity. Stress 13, 1—14. Little, R.J.A.,1988. A test of missing completely at random for

multivariate data with missing values. J. Am. Stat. Assoc. 1198—1202.

McBurnett,K.,Lahey, B.B., Rathouz,P.J.,Loeber, R.,2000. Low salivarycortisol andpersistentaggressioninboysreferredfor disruptivebehavior.Arch.Gen.Psychiatry57,38—43.

McEwen,B.S.,2004.Protectionanddamagefromacuteandchronic stress: allostasisand allostatic overloadand relevanceto the pathophysiology ofpsychiatric disorders. Ann. N.Y. Acad.Sci. 1032,1—7.

Moffitt,T.E.,1993.Adolescence-limitedandlife-course-persistent antisocialbehavior: a developmental taxonomy. Psychol.Rev. 100,674—701.

Moffitt,T.E.,2001.SexDifferencesinAntisocialBehaviour:Conduct Disorder,Delinquency,andViolenceintheDunedinLongitudinal Study.CambridgeUniversityPress,Cambridge.

Monshouwer,K.,Verdurmen,J.,VanDorsselaer,S.,Smit,E.,Gorter, A.,Vollebergh,W.,2008.Youth and risky behaviour. Core data 2007. In: Smoking, Drinking Drug Use and Gambling in Pupils over 10 years old. Trimbos-instituut, Utrecht, The Netherlands. Muthe´n, L.K., Muthe´n, B.O.,2007. Mplus User’s Guide, 4th ed.

Muthe´n&Muthe´n,LosAngeles.

Nieuwbeerta,P.,2002.CrimeVictimizationinComparative Perspec-tive.BoomJuridischeuitgevers,DenHaag,TheNetherlands. Petersen,A.C.,Crockett,L.,Richards,M.,Boxer,A.,1988.A

self-reportmeasureofpubertalstatus:reliability,validity,andinitial norms.J.YouthAdolesc.17,117—133.

Platje, E.,Jansen, L.M.C.,Raine,A., Branje,S.J.T.,Doreleijers, T.A.H.,DeVries-Bouw,M.,Popma,A.,vanLier,P.A.C.,Koot,H.M., Meeus,W.H.J.,Vermeiren,R.R.J.M.,2013a.Longitudinal asso-ciationsinadolescencebetweencortisolandpersistent aggres-siveorrule-breakingbehavior.Biol.Psychol.93,132—137. Platje,E.,Vermeiren,R.R.J.M.,Branje,S.J.T.,Doreleijers,T.A.H.,

Meeus,W.H.J.,Koot,H.M.,Frijns,T.,vanLier,P.A.C.,Jansen, L.M.C., 2013b. Long-term stability of the cortisol awakening response over adolescence. Psychoneuroendocrinology 38, 271—280.

Popma,A.,Doreleijers,T.A.H.,Jansen,L.M.C.,vanGoozen,S.H.M., vanEngeland,H.,Vermeiren,R.,2007.Thediurnalcortisolcycle indelinquentmaleadolescentsandnormalcontrols. Neuropsy-chopharmacology32,1622—1628.

Popp, D., Laursen, B., Kerr, M., Stattin, H., Burk, W.K., 2008. Modelinghomophilyovertimewithanactor-partner interdepen-dencemodel.Dev.Psychol.44,1028.

Pruessner,J.C.,Wolf,O.T.,Hellhammer,D.H.,Buske-Kirschbaum, A.,vonAuer,K.,Jobst,S.,Kaspers,F.,Kirschbaum,C.,1997. Freecortisollevelsafterawakening:areliablebiological mark-erfortheassessmentofadrenocorticalactivity.LifeSci.61, 2539—2549.

Raine,A.,1993.ThePsychopathologyofCrime:CriminalBehavioras aClinicalDisorder.AcademicPress,Inc.,SanDiego,CA,USA.

Raine,A.,2002.Biosocialstudiesofantisocialandviolentbehavior inchildrenandadults:areview.J.Abnorm.ChildPsychol.30, 311—326.

Raine,A.,Moffitt,T.E.,Caspi,A.,Loeber,R.,Stouthamer-Loeber,M., Lynam,D.,2005.Neurocognitiveimpairmentsinboysonthe life-coursepersistentantisocialpath.J.Abnorm.Psychol.114,38—49. Raine,A.,Venables,P.H.,Mednick,S.A.,1997.Low resting heart rate at age 3 years predisposes to aggression at age 11 years: evidence from the Mauritius Child Health Project. J. Am. Acad. Child Adolesc. Psychiatry 36, 1457—1464.

Reitz,E.,Prinzie,P.,Dekovic¸,M.,Buist,K.L.,2007.Theroleofpeer contactsinthe relationshipbetween parentalknowledge and adolescents’ externalizing behaviors: a latent growth curve modelingapproach.J.YouthAdolesc.36,623—634.

Roisman,G.I.,Susman,E.,Barnett-Walker,K.,Booth-LaForce,C., Owen,M.T.,Belsky,J.,Bradley,R.H.,Houts,R.,Steinberg,L., NICHDEarlyChildCareResearchNetwork,2009.Earlyfamilyand child-careantecedents ofawakeningcortisol levels in adoles-cence.ChildDev.80,907—920.

Satorra,A.,Bentler,P.M.,1994. Correctionstoteststatisticsand standarderrorsincovariancestructureanalysis.In:vonEye,A., Clogg, C.C. (Eds.), Latent Variable Analysis: Applications for DevelopmentalResearch.SagePublishers,ThousandOaks,CA, pp.399—419.

Susman,E.J.,2006.Psychobiology of persistent antisocial behavior: stress, early vulnerabilities and the attenuation hypothesis. Neu-rosci. Biobehav. Rev. 30, 376—389.

Taylor,S.E.,Klein,L.C.,Lewis,B.P.,Gruenewald,T.L.,Gurung,R.A., Updegraff, J.A., 2000. Biobehavioral responses to stress in females: tend-and-befriend, not fight-or-flight. Psychol. Rev. 107,411—429.

Thornberry,T.P.,Krohn,M.D.,Lizotte,A.J.,Chard-Wierschem,D., 1993.Theroleofjuvenilegangsinfacilitatingdelinquent behav-ior.J.Res.CrimeDelinquen.30,55—87.

Thornberry,T.P.,Lizotte,A.J.,Krohn,M.D.,Farnworth,M.,Jang, S.J.,1994.Delinquentpeers,beliefs,anddelinquentbehavior:a longitudinaltestofinteractionaltheory.Criminology32,47—83. Trickett,P.K.,Noll,J.G.,Susman,E.J.,Shenk,C.E.,Putnam,F.W., 2010.Attenuationofcortisolacrossdevelopmentforvictimsof sexualabuse.Dev.Psychopathol.22,165—175.

vanGoozen,S.H.M.,Fairchild,G.,Snoek,H.,Harold,G.T.,2007.The evidence for a neurobiological model of childhood antisocial behavior. Psychol. Bull. 133, 149—182.

Verhulst,F.C.,VanderEnde,J.,Koot,J.M.,1997.Handleiding voor de Youth Self Report. Afdeling Kinder- en Jeugdpsychiatrie, Sophia Kinderziekenhuis/Academisch Ziekenhuis Rotterdam/ ErasmusUniversiteitRotterdam,Rotterdam.

Wilhelm,I.,Born,J.,Kudielka,B.M.,Schlotz,W.,Wust,S.,2007.Is thecortisolawakeningrisea responsetoawakening? Psycho-neuroendocrinology32,358—366.

Wust,S.,Wolf,J.,Hellhammer,D.H.,Federenko,I.,Schommer,N., Kirschbaum,C.,2000.Thecortisolawakeningresponse—normal valuesandconfounds.NoiseHealth2,79—88.

Yanovitzky,I.,2005.Sensationseekingandadolescentdruguse:the mediatingroleofassociationwith deviantpeersand pro-drug discussions.HealthCommun.17,67—89.