Healthy minds 0–100 years: Optimising the use of European brain imaging cohorts (“Lifebrain”)

(1)

Review

Healthy minds 0 – 100 years: Optimising the use of European brain imaging cohorts ( “ Lifebrain ” )

Kristine B. Walhovd

^a

, Anders M. Fjell

^a

, René Westerhausen

^a

, Lars Nyberg

^b

,

Klaus P. Ebmeier

^c,

*, Ulman Lindenberger

^d

, David Bartrés-Faz

^e

, William F.C. Baaré

^f

, Hartwig R. Siebner

^f

, Richard Henson

^g

, Christian A. Drevon

^h

,

Gun Peggy Strømstad Knudsen

ⁱ

, Isabelle Budin Ljøsne

ⁱ

, Brenda W.J.H. Penninx

^j

,

Paolo Ghisletta

^k,l

, Ole Rogeberg

^m

, Lorraine Tyler

ⁿ

, Lars Bertram

^o

, Lifebrain Consortium

¹

aUniversityofOsloCentreforLifespanChangesinBrainandCognition(UiO),DepartmentofPsychology,HaraldSchelderupsHus,Forskningsveien3A,N-0373 Oslo,Norway

bCentreforFunctionalBrainImaging(Umeå),UmeåUniversitet,SE-90187,Umeå,Sweden

cUniversityofOxfordWellcomeCentreforIntegrativeNeuroimaging&DepartmentofPsychiatry(UOXF),UniversityofOxford,WarnefordHospital, OxfordOX37JX,UK

dMax-PlanckInstituteforHumanDevelopmentCentreforLifespanPsychology(MPIB),Lentzeallee94,D-14195Berlin,Germany

eUniversityofBarcelonaBrainStimulationLab(UB),FacultatdeMedicina,CampusClínic,C/.Casanova,143,AlaNord,5aplanta,S-08036,Barcelona,Spain

fRegionHovedstaden(RegionH),DanishResearchCentreforMagneticResonance,CentreforFunctionalandDiagnosticImagingandResearch,Copenhagen UniversityHospitalHvidovre,Section714,KettegardAllé30,DK-2650,Hvidovre,Denmark

gMedicalResearchCouncilCognitionandBrainScienceUnit(MRC),UniversityofCambridge,15ChaucerRoad,Cambridge,CB27EF,UK

hVitasAS(AnalyticalServices),Gaustadalléen21,N-0349Oslo,Norway

iNorwegianInstituteofPublicHealthOslo(NIPH),POBox4404Nydalen,N-0403Oslo,Norway

jVUUniversityMedicalCentre(VUmc),POBox7057,NL-1007MBAmsterdam,TheNetherlands

kUniversityofGeneva(UNIGE),FacultyofPsychologyandEducationalSciences,ResearchgroupofMethodologyandDataAnalysis,SandrineAmstutz, UniMail,4eétage,BdduPont-d'Arve40,1205Genève,Switzerland

lSwissDistanceLearningUniversity,Brig,Switzerland

mRagnarFrischCentreforEconomicResearch(Frisch),Gaustadalleen21,N-0349Oslo,Norway

nUniversityofCambridgeDepartmentofPsychology(UCAM),DowningStreet,CambridgeCB23EB,UK

oUniversityofLübeckInterdisciplinaryPlatformforGenomeAnalytics(LIGA-UzL),UniversityofLübeck,Maria-Goeppert-Str.1(MFC1),23562D-Lübeck, Germany

ARTICLE INFO Articlehistory:

Received27August2017

Receivedinrevisedform4October2017 Accepted5October2017

Availableonline12February2018 Keywords:

Psychiatricdisorder Dementia

Economicsandpsychiatry Epidemiology

MRI

Neurodevelopment

ABSTRACT

Themainobjectiveof“Lifebrain”istoidentifythedeterminantsofbrain,cognitiveandmental(BCM) health at different stages of life. By integrating, harmonising and enriching major European neuroimagingstudiesacrossthelifespan,wewillmergeﬁne-grainedBCMhealthmeasuresofmore than5000individuals.Longitudinalbrainimaging,geneticandhealthdataareavailableforamajorpart, aswellascognitiveandmentalhealthmeasuresforthebroadercohorts,exceeding27,000examinations intotal.Bylinkingthesedatatootherdatabasesandbiobanks,includingbirthregistries,nationaland regionalarchives,andbyenrichingthemwithanewonlinedatacollectionandnovelmeasures,wewill addresstheriskfactorsandprotectivefactorsofBCMhealth.Wewillidentifypathwaysthroughwhich riskandprotectivefactorsworkandtheirmoderators.ExploitingexistingEuropeaninfrastructuresand initiatives,wehopetomakemajorconceptual,methodologicalandanalyticalcontributionstowards largeintegrativecohortsandtheirefﬁcientexploitation.WewillthusprovidenovelinformationonBCM healthmaintenance,aswellastheonsetandcourseofBCMdisorders.Thiswilllayafoundationfor earlierdiagnosisofbraindisorders,aberrantdevelopmentanddeclineofBCMhealth,andtranslateinto

DOIoforiginalarticle:http://dx.doi.org/10.1016/j.eurpsy.2017.10.005

* Correspondingauthor.

E-mailaddresses:[email protected](K.B. Walhovd),[email protected](L.Nyberg),[email protected](K.P.Ebmeier),

[email protected](U.Lindenberger),[email protected](D.Bartrés-Faz),[email protected](W.F.C. Baaré),[email protected](R.Henson), [email protected](C.A. Drevon),[email protected](G.P.StrømstadKnudsen),[email protected](B.W.J.H. Penninx),[email protected] (P.Ghisletta),[email protected](O.Rogeberg),[email protected](L.Tyler),[email protected](L.Bertram).

1 See4.LifebrainConsortium.

http://dx.doi.org/10.1016/j.eurpsy.2017.12.006

ContentslistsavailableatScienceDirect

European Psychiatry

j o u r n a lh o m e p a g e : h t t p : / / w w w . e u r o p s y -j o u r n a l . c o m

(2)

futurepreventiveandtherapeuticstrategies.Aimingtoimproveclinicalpracticeandpublichealthwe willworkwithstakeholdersandhealthauthorities,andthusprovidetheevidencebaseforprevention andintervention.

1.Background

Neurodevelopmental and degenerative changes, addiction disordersandotherpsychiatricproblemsarebothinfluencedby andmirroredinbrainchangesthatoccurthroughoutlife.Amajor challengeis todeterminewhich age-relatedchanges are detri- mental and which enhance cognitive and mental health. The potential economic benefits of an improved understandingare large, with total costs of brain disorders in Europe in 2010 estimated at s798 billion [1]. Throughout life, our genetic dispositionsinteract continuously with environmental, societal, occupationalandlifestylefactorstoinfluencebrainstructureand function.Suchchanges,fromtheearlieststagesoflifetooldestage, aremappedindetail inEuropean longitudinalstudies,utilizing Magnetic ResonanceImaging (MRI). MRI yieldshigh-resolution imagesofvariationsinbrainmacrostructure,microstructureand function, which can be compared with measurable changes in cognitive function and mental health. However, since MRI is expensive and time-consuming, the number of participants includedin suchstudiestendstobelow.Thismakesithard to disentangletheroleofthemanyfactorsthatcaninfluencebrain, cognition and mental health at different stages of life. While forming a precondition for a possible personalised medicine approach,suchindividualvariationsneedfirsttobeestablished.

Forinstance,age-speciﬁcmechanismsnecessitatealargenumber ofparticipantsatallstagesoflife,andsex-speciﬁceffectsfurther halve the sample sizes, thus narrowing degrees of freedom availableforanalyses.

1.1.Overallaimandobjectives

This EU Horizon 2020project “Lifebrain” aims to maximise theexploitation ofbrain imagingcohorts by bringingtogether studies onhow differences and changesin brain age relate to cognitive function and mental health. This will be done by integrating and standardizing data and results from 11 large predominantlylongitudinalEuropeansamplesfrom7countries [2–12](Table1).

Thiswillyieldadatabaseofﬁne-grainedBCMhealthmeasures for morethan 5.000 individual participants.Longitudinal brain imagingdataareavailableforamajorportion,aswellascognitive andmentalhealthmeasuresforbroadercohorts,exceeding27.000 examinations.Theprojectisacollaborativeinitiativeinvolvinga small and medium-sized enterprise (SME), several of Europe’s majorbrainresearchcentres,aswellasstakeholdersforefﬁcient exploitationofresults(Fig.1).

Lifebrainincludesfoursub-objectives:

IntegrationofdataacrossexistingmajorlongitudinalEuropean neuroimaging studies of age changes in brain, cognition and mental health, including genetic, epigenetic, lifestyle, and medical registry information for thousands of individuals, furtherenrichedwithhealthoutcomesandbiomarkers.

Development and standardization of measures and methods acrossthesemajor Europeanstudiesof agechanges inbrain, cognitionandmentalhealth.

Provisionofnovelinformationonbrain,cognitionandmental health maintenance, onset and courseof diseases and health inequalities,toyieldtheevidencebasefordevelopmentofpolicy

strategiesforpreventionandintervention,therebyaddressing healthinequalities.

Communication and implementation of new knowledge, exploiting the integrative cohorts in age-speciﬁc prevention andtreatmenttooptimisebrain,cognitionandmentalhealth, improvingclinicalpracticeandhealthpolicy.

1.2.Vision

Personalizedhealth carerequiresfundamentalknowledgeof riskfactorsandprotectivefactors,aswellasthepathwaysthrough which they work at different ages. Extrapolating from known effectsofcertainrisksandinterventions[13,14],amultifactorial andpersonalisedapproachcouldidentifymodiﬁableenvironmen- talfactorsthatpromotecognitivedevelopmentinchildhoodand adolescence,fostermaintenance ofcognitive functionsintolate adulthood, delayonset of dementia, reduceneed for care, and improve working ability through prevention and intervention programs. Cognitive and mental health disorders are a serious burden forindividuals aswellas societies [15,16]. Within 5–10 years, we hopeknowledge established in Lifebrain will enable policy makers and health care systems to implement low- thresholdstrategiesforindividualpreventionbymodiﬁablelife- stylefactors,aswellasnon-pharmacologicalinterventions.Ina Europeanperspective,thesecouldhaveenormousconsequences for individualwell-being, work abilities,and for thetotal costs related to increased health care needs and reduced working capabilitiesinolderadultsduringthecomingdecades.

2.Methods

2.1.Concept

Anewapproachtomodelbrain,cognitiveandmentalhealthis needed that differs in fundamental ways from previous approaches:itshouldbedimensional,focusedonlifespanrather thanspecificphasesofdevelopmentorage,andbasedonsystems- vulnerability and resilience, rather than simple cause-effect relationships. We argue, (1) factors that affect cognitive and mental health will often vary along a continuum across the population,(2)risksandbenefitsaccumulateovertime,andwill not be coincident with the ageat which their effects become apparent, and (3) the effects of these factors will vary across individuals asa function of theirgenotype.We aimtoidentify importantcausalfactors,toimproveourunderstandingofhow theseaffectbrainhealthatdifferentagesandindifferentpeople, and to identify beneficial and cost-effective interventions. The projectwillproceedthroughdistinctbuttightlyinteractingphases (Fig.2).

Wewill furthercombinea largepopulation-basedapproach with an in-depth neurocognitive approach in order to clarify mechanisms, and how these translate into speciﬁc cognitive functions. For instance, functional variation with age and sex acrosscountriesmaybeduetoindividualsinsomeregionsofthe world having experienced better conditions in childhood and adulthood,relatingtonutrition,education,diseaseexposureand physicalandsocialactivitypatterns[17,18].However,thepathways and mechanisms by which these broad factors work, remain

(3)

unknown.Thisknowledgeispivotalfordevelopmentofefﬁcient healthpoliciesandtargetedprevention.

2.2.Identifyingriskandprotectivefactors

Toidentifyriskfactorsforcognitiveandmentaldisorders,wewill use approaches previously developed to group participants

according totheir trajectoriesofchange [19,20]. Factorssuch as higher education, physical activity, female sex, carrying the met-alleleoftheCOMTgene,andnotlivingalonewereassociated with maintenance ofmemoryfunction;lowereducation,unem- ployment,beingmale,andcarryingaAPOE

e

⁴^allele^were^risk^factors

fordecline[20]. Wehaverecentlyemployedstructuralequation modelling (SEM)trees asa methodto uncoverheterogeneity in Fig.1.Lifebrain’svision,mission,overallaimandmainobjectives,andtheirrelationships.

Table1

CentralfeaturesofthestudiesthatwillfeedintotheLifebraindatabase.

Cohort Description N Obs. MRI/

Obs.*

CM DNA**

BarcelonaSpain(UB) Collectionofcrosssectionalandlongitudinalstudies(threetimepoints)incognitively normaladults(38–90yrs.),MRI,lifestyleactivitiesandcognitivereservemeasured

261 467 261/

415 All x BASE-IGermanyMPIB,UzL TheBerlinAgingStudy(BASEI)isanextensivemultidisciplinaryinvestigation(70–100+

yrs.)with14testsessionsonmentalandphysicalhealth,psychologicalfunctioning,and SES,witheighttimepoints

516 1402 All x

BASE-IIGermanyMPIB,UzL BASE-IIinvestigatesthephysical,cognitive,andsocialconditionsthatleadtosuccessful ageinginapopulation-basedlongitudinalstudy(1600participants60–80yrs.,600 participants20–35yrs.),withathoroughphysicalhealthexamination,multipletestsof cognitiveabilities,andcompletequestionnaires.Twotimepoints.

2200 2527 415/

772 All x

BetulaSwedenUmU Populationbased(25–100yrs.),startedin1988,theseventhwaveisongoing.10400test sessions,with5850longitudinalobservations.Allwerehealthyandnon-dementedat baseline,butmorethan400havedevelopedsomeformofdementia.Very

comprehensiveexaminationsonhealth,biomarkers,genotyping,neuropsychologyand brainimaging.TimepointsﬁveandsixarepartofLifebrain.

376 707 376/

707 All x

Cognitionandplasticitythroughthe lifespan(LCBC),Norway–UiO

CollectionofstudiesbyLCBCaswellascollaborations,well-screened(4–90+yrs.), comprehensiveneuropsychology,MRI,lifestyle,health,biomarkers,andothermeasures.

SomeparticipantspresentwithADorotherdementiasordecline.Threetimepoints.

1677 2533 1638/

2494 All x

WhitehallIIMRI-substudyUKUOXF Non-industrialcivilservantsfromLondon(35–55yrs.atinclusion),cognitiveandhealth databackto1985,withclinicalexaminationevery5years,yieldingacomprehensive databaseofinformationthroughthemultiplefollowups,includingarangeofhealthand life-stylerelatedparameters,biomarkers,earlypredictorsandgenetics

800 800 775 All x

CambridgeCentreforAgeingand Neuroscience(Cam-CAN),UK–MRC

CambridgeCentreforAgeingandNeuroscience,epidemiological,behavioural,and neuroimaging,populationrepresentativesample(18–88yrs.),measurehealth,lifestyle, SES,andpsychiatry.Threetimepoints.

2690 2927 633/

870 All x

CambridgeCentreforAttention, LearningandMemory(CALM),UK– MRC

Children(5–16)referredfordifﬁcultieswithattention,learning,andmemory.

Comprehensiveneuropsychologicalassessment,incl.personalityandexecutive difﬁculties,andMRIaccordingtoCamCANprotocols.Follow-upplanned

482 482 227 All x

DanishDevelopmentalstudies(HUBU), Denmark–REGIONH

Brainandbehaviouraldevelopmentinchildrenandadolescents,withupto12time points.MRI,cognition,Socioemotionalfunctioning,personality,lifeevents,lifestyle, biomarkers.First10assessmentswereperformedwith6monthsintervals

94 817 94/

817 All x

Danishageingstudies(LISA)Denmark– REGIONH

LiveActiveSuccessfulAgeing,physicalactivityandinterventionandageingstudy, community-based.Threetimepoints.Initially,onlybaselinedatawillbepartof Lifebrain.

450 450 390 All x

NetherlandsStudyofDepressionand Anxiety(NESDA),Netherlands– VUMC

NetherlandsStudyofDepressionandAnxiety,participants(agerange18–65atinclusion) withandwithoutsymptoms,measurepsychopathology,health,demographicsandSES, psychosocialfunction,cognition.Sixtimepoints.MRIobtainedontimepointsone,three andsix.

2981 14860 301/

632 All x

Total 12527 27972 5140/

8099

All 1800

N:Numberofuniqueindividuals.

Obs:Totalnumberofobservations(oneforeveryindividualforeachtimepointinlongitudinalstudies.

CM:Cognitivetestsandmentalhealthmeasures.MRIincludes,T1,T2,DTI,RSI,resting-statefMRI,task-relatedfMRI.Proportionofparticipantswithneurodegenerative diseasevaries.

*NumberofuniquesubjectswithMRI/totalnumberofMRIsessionse.g.includingfollow-up;**DNA:thegenotypingvariessomeacrossstudies.ManyhaveGWAS,including allinNESDA,allinBaseIand300inBASEII,200inLCBC,2200inBetula,andGWASfor2000willbeperformedaspartofLifebrain.Cam-CANhascollectedsalivasamples andDNAextractionandGWASisapartofLifebrain.AllresearchparticipantsdescribedinTable1.3willbeincludedinstatisticalanalysesbasedonexistingdata.All participantsthat(a)canbereached,(b)arefromsampleswhererelevantethicalapprovalisgiven,(c)forwhomrelevantdatadonotalreadyexist,and(d)areabletogivefull informedconsent,willbeaskedfordriedbloodspotsamples,buccalswabsamplesandonlinetesting.

(4)

empiricaldata,togetherwithpredictorsandpotentialinteractions amongthem,explainingthisheterogeneity[19].Wewillusesimilar approachestostratifylargernumbersofparticipantsaccordingto importantcognitive,psychiatricandbrainhealthmeasures,andtest theeffects ofarangeofriskand protectivefactors,includinggenetics, epigenetics,socio-demographic,environmental,andlifestyle.This partoftheprojectwillallowidentiﬁcationofgeneralriskfactorsfor cognitive and mental health. Some of the relations will be correlationalinnature; otherscanbemodelledlongitudinallyto shedlightoncauseandeffect.Pathwaysfortheobservedrelations, andhowtheyaremoderatedbyanumberofotherfactors,willbe speciﬁcallytargeted(seebelow).

Wewillalsoassesslongtermeffectsofcandidatevariablesfrom earlylife.Theearliestofthesearemarkersofpre-andneonatal health.Abnormalcognitiveandbehaviouraloutcomes,suchasin ADHDandschizophrenia,havebeenlinkedtoearlydevelopment, includingfactorsrelatedtofoetalgrowthand adversity[21,22].

Lowbirth weight, as a marker of adverse intrauterinecircum- stances,hasbeenassociatedwitharangeofdiseasesandreduced function in daily life [23,24]. Some factorsare associated with cognitivedeclineandincreasedAlzheimer’sdisease(AD)risk,e.g.

coronaryheart disease,hypertension,and type2 diabetes[25].

Anatomicalstudiesshowthatlowbirthweight,prematurebirth, andprenatalsubstanceexposurecanaffectcorticalandspeciﬁcally fronto-striatal development, impacting attention and executive function [26–28]. These ﬁndings underscorethe importance of neonatal characteristics for brain development and individual functional differences along a continuum from normality to pathology[26].

Addingtothispicture,anumberofcommonvariantsinrisk genesforpsychiatricdisorderswererecentlyfoundpredictiveof brainstructureatbirth[29].Forinstance,neonatescarryingAPOE

e

^4, ^the^major^genetic^risk ^factor ^for^AD, ^were^reported^to^have

reducedvolumesoftemporal cortex,similartothatreportedin older adults [29]. For variants of the fat mass and obesity- associatedgene(FTO),which hasbeenrelatedtoreducedbrain volumesinhealthyageingandriskofAD[30,31]anddepression [32], smaller brain volumes were recently shown also in adolescents[33].Suchﬁndings shouldincrease ourattentionto researchingandoptimizingearlyinﬂuencestoprovideresources and knowledgeonhealth and diseasedeterminants, onset and courseofdiseasesandpublichealth.Effortstopostponedeclineor disease may be futile if they are only targeted at old age. By including the full lifespan, we will investigate how early life variables,including genetic and epigenetic variants, exert their effects.

In addition tothe very early predictors related to pre- and perinatalconditions,wewillalsofocusonindicatorsinchildhood andyoungadulthood.Forinstance,intelligencescoresatage11are remarkablygoodpredictorsforcognitiveperformance atage90 years[34],andneuroticismandintroversionpredictdevelopment ofdepressionlaterinlife[35].Furthermore,cardiovascularﬁtness andcognitiveperformanceinearlyadulthoodincreasedtheriskof

early-onsetdementiaandmildcognitiveimpairment(MCI)more than four decades later [36]. Current levels of self-reported physicalactivityhavebeenrelatedtolesscorticalatrophyinthe prefrontalcortexacrossa3.5-yearperiodintheadultlifespan[37].

Thismaysuggestthatolderadultsshouldexercisemoretoprevent cortical atrophy.Alternatively, themore physically active older adults may also have been the more physically active and ﬁt youngeradults.Conversely,certainbraincharacteristicsorchanges mayaffectpeople’sphysicalactivity[38].

Itispivotaltoexaminetowhatextentthelate-lifeassociation canbeexplainedbyearlyphysicalﬁtness,currentphysicalactivity, andtheirrelationship,orifbothexertuniqueinﬂuences.People withbodymassindex(BMI)>25,showedarelationshipbetween BMIandbrainatrophy[37].However,forolderadults,higherBMI maynotconferthesameriskasforyoungerormiddle-agedadults, and may even be protective, conferring a smaller risk of cardiovascular complications [39]. Similarly, certain genetic variations, such as the APOE

e

⁴ ^allele, ^reduce ^brain ^plasticity

andincreasebrainvulnerability[40].Thus,wemightexpectless stabilityof cognitivefunction[20] anda highervulnerabilityto declineinthefaceofotherriskfactorsforthosewiththeriskallele.

Equally, brain-derived neurotrophic factor (BNDF) is linked to neuronal growth and differentiation, and thus contributes to memoryandlearning,leadingtodifferentialcognitivetrajectories [41]andtomentalhealthdifferences[42].Theseﬁndingsarein line with the broader “resource modulation” hypothesis [43], accordingtowhich theeffects of commongeneticvariation on cognitive performance increase from early to late adulthood, reﬂectingthenon-linearassociationbetweenbrainresourcesand performance(forreview,see[44]).

In Lifebrain, using integrated cohorts representing diverse European social models over a large age range, and with the additionaluseofUSdatabases,wewilltestnationaldifferencesin riskandprotectivefactorsinBCMhealth.

A critical aspect of Lifebrain is the unique longitudinal neuroimagingdesign insomeofitscohorts.Datafromtheﬁrst sixtime-pointsoftheHUBU-project,forexample,showthatthe developmentaltrajectoriesin hippocampalvolumearenotwell predictedfromonlycross-sectionalobservationsusingageneral- ized additive model (left panel of Fig. 3), compared to when estimated from longitudinal information (with a generalized additive MIXEDmodel, rightpanel),which takes actualchange into account. Thesame phenomenon is frequently observedin ageing,whereunavoidablerecruitmentbiasmayleadtounder- estimationoftrueageeffects.

Insummary,wewillanalysetheentirelifespan,includingearly lifecharacteristics,andindividualfactors,aswellasthelevelofthe individualand broadersociety,and identifyrisk and protective factorsforbrain,cognitionandmentalhealth.Wewillachievethis bycombiningpopulation-based,cross-sectionalandlongitudinal cohortswithgeneralindicatorsofhealth,togetherwithin-depth detailedandtargetedcognitiveanalysesacrossEuropeanandUS studies.

Fig.2.VariableandanalysisstructurewithinLifebrain.

(5)

2.3.Identifyingpathwaysandmoderators

Our longitudinal design permits identiﬁcation of pathways fromriskand protectivefactorstocognitive andmental health problemsandtheirmoderatorsthataffectthestrengthordirection ofrelationships.Wewillexaminehowbroaderriskfactorsworkat a biological and behavioural level (pathways), and evaluate interactionsbetweenidentiﬁedriskfactors,intermediatevariables aspartofpathways,andalongwithotherimportantvariables,e.g.

sex[45,46].Sexhasbeenassociatedwithdifferencesintheriskfor developingseveraldisorders[45–48].Observedsexdifferencesin behaviour,cognition,emotions,and measuresofbrainstructure andfunction,andtheirdevelopment [45,49,50] areoftensubtle andinconsistent,asmanystudiesinvestigatelimitedagerangesor lacklongitudinalobservations[51].Moreover,theeffectofsexmay differacrossthelifespan[20,37,52,53].Improvedlivingconditions inEuropeandlessgender-restrictededucationareassociatedwith increased sex differences in some cognitive functions, suchas episodicmemory,favouringwomen,andalsodecreasesofothers, suchasnumeracy[18].Itisinterestingwhetherpathwaysatthe levelofbrainandotherbiomarkerscanbeidentifiedforsexasa moderatorofcognitiveandmentalhealth overtime,andacross physicalandsocialenvironments,includingcountries.Variablesof interest include fatty acids, BMI, vitamins, cholesterol, blood pressure,geneticandepigeneticfactors,andMRImeasuresofthe brain.OutcomemeasuresinLifebrainwillincludemanyobjective healthvariables,andrefinedcognitiveandbrainimagingassess- ments, mental health measures, as well as social and mental factors.Thiswillbeusedtoaddressthepathwaysandmechanisms foridentifiedriskandprotectivefactors.Essentialfattyacidsare important for development of cognitive function in normal pregnancies[54],aswellasforseverelyprematureinfants[55].

BMIandcertainbloodmarkersare,forexample,relatedtoless(e.g.

DHA, vitamin D)or more (e.g. cholesterol)brain atrophy [37].

VitaminDhasreceivedmuchattentionlately,asthelinkbetween vitaminD deficiencyandrisk of mentaland neurodegenerative diseases has been established [32,56], and insufficiency or deficiencyofvitaminDiscommoninolderclinicalsamples.Little is knownabout the direction of the relationship, and whether similarrelationshipscanbefoundamonggroupsofnon-demented elderly[57].HigherlevelsofvitaminDappeartoberelatedtoless degenerationof majorwhite mattertracts inthebrain,evenin cognitivelyhealthyelderlyparticipants.VitaminDmaystimulate productionofneurotrophic,antioxidativeandanti-inflammatory factors,reduceriskforcardiovascularandcerebrovasculardisease and eveninfluence amyloid phagocytosis andclearance [56]. If

these relationships between brain change and nutrients like vitamin D can be replicated across larger and international samples,suchmechanismswillprovideafoundationfortargeted interventions and randomized controlled trials also in normal ageing.

Wedo notknowhowpotentiallymodiﬁablebehaviours,e.g.

eating habits and physical activity, affect blood nutrient and vitamin levels. Alsounknown are effects of modifying medical factorssuchasBMIandhypertension,howgeneticfactorsinteract, andtheimpactofearly-lifecognitivefunction.Thereisreasonto expect that established genetic variants, e.g. FTO [58], may inﬂuence the relation betweenfood intake and blood nutrient andcardiovascularmarkers,andalsotherelationofsuchmarkers with brain atrophy. Furthermore, APOE

e

⁴ ^may ^reduce ^brain

plasticity, and increase vulnerability to negative impact on cognitive function[40], aswellasmental health [59]. APOE

e

⁴

carriersalsohavemorepronouncedsleepdisturbances[60] and poorsleephabitshavebeenrelatedtobrainatrophy [61],poor whitematterintegrity[62] andamyloidbetadeposition[63] in agedindividuals.Positivelifestylefactorssuchasphysicalactivity [64], healthy diet [65] and high education levels [66] may be beneﬁcialfor APOE

e

⁴ ^carriers, ^by^lowering^amyloid^deposition

[67] orpromoting‘maintenance'ofbrainmetabolism[68].Such observationsareexamplesofconsideringdistinctlifestylesinthe contextofasingleorcombinationofgeneticvariant(s),inorderto developtailoredstrategiesforinterventioninsubgroupsofolder adults. We will measure blood-based nutritional factors, food intake,genetics,andepigenetics,andrelatethemwithmeasuresof brainandcognitivefunction.Ourcombinationofsamplesisvitalto obtainsufﬁcientstatisticalpower.

Therich informationinourlargelongitudinaldatabaseswill allow us to examine individual differences in the onset and magnitudeofcognitivechange,andidentifygeneticandlifestyle factorsthatpredictpreservedordecliningcognition.IntheBetula study [20], there were three distinct ageing trajectories for episodicmemory(Fig.4).

Good episodic memoryin old agewas alsoassociated with preservedhippocampalfunction[69].Parkinson'sdiseasepatients with MCI havereduced fronto-striatalperformance of working memory[70].Itislikelythatthoseolderparticipantsdisplaying fronto-striatal/workingmemoryimpairmentwillbedistinctfrom those showinghippocampus/episodicmemoryimpairment, and thatdistinctgeneticandlifestylepredictorsarerelevantforeach phenotype.Thesamemayapplytotheseparationofverbaland visuo-spatial memory [71]. Identification of specific cognitive phenotypesmaysuggestthatdifferentkindsofinterventionswill Fig.3. Developmentaltrajectoriesforhippocampalvolumeareverydifferentwhenestimatedfromcross-sectionalobservationsonlybyageneralizedadditivemodel(left paneloffigurebelow)comparedwithageneralizedadditiveMIXEDmodel(rightpanel)takingactualchangeintoaccount[10,11].Hippocampalvolume[mm³]wasmeasured fromT1-weightedMRIimages,segmentedbyFreeSurferfromthelefthemisphere.Thefitcurveswerebasedon510scansfrom94participants(meanage11.2years,7.5–15.4 years).

(6)

be effective. In addition to the broad set of existing genetic, epigenetic,health,demographicandlifestylevariablesavailablein the Lifebrain database, we will conduct additional analyses to enrichthedatabasewith,e.g.telomerelength[72]andlongitudi- nalbrainimaging[73].

In summary, we will identify variables and pathways with impactonneurocognitivefunctionandmentalhealththroughout lifetobeusedintargetedinterventionandprevention.

2.4.Standardizationandharmonisation

Standardizationofmeasuresacrosssitesposthoc, aswellas throughenrichmentofexistingcohorts,isenactedintwoseparate workpackages (WP). Partners have experience with multi-site neuroimaging,cognitionandmentalhealthstudies.Wewillutilize proceduresdevelopedanddescribedbytheBioSHaREprojectfor dataharmonisation,integrationandfederateddataanalyses[74].

Measurescan alsobeharmonized acrosslatentvariables:Since existingdata havenot beencollectedwithidenticalscales and measures(e.g.differentversionsofneuropsychologicaltestsand differentMRIscanners),site-specificdifferencesareexpectedfor somemeasures.Toallowmeasurementstobemademoreresistant tomeasurementerror,and toabstractfromindividualscalesto underlyingconstructs,ouranalyseswillthereforebebasedalsoto somedegreeonlatentfactormodels.We willemploystandard psychometricapproaches todetermineconstructvalidityacross sites.We will rely on thestandardized itemsof the newdata collectionphase. Ifstandardizeditemsarecollectedasa partial reassessment of site-specific items from the original data collection, this allows for the evaluation of convergent and divergent construct validity. We will test the convergence of latentfactorsfromsite-specificandstandardizedinstrumentsto determine statistical overlap of hypothetical constructs across sites. This is also an effective means of dealing with missing data.Suchanapproachwillallowmergingofdataacrosssites,and will probably be beneficial in multi-site studies outside the collaboration.

2.5.Linkagewithregistrydata

ForsomeoftheLifebraindata,linkagetoregistrydatamaybe possible.Indeed, some have already been linked by individual consent–e.g.theMRIcohortsfromtheNorwegianMotherand ChildStudy(MoBa)arelinked,andNorwegianadultcohortsalso

have consentedtolinktheir MRI,cognitive, mental health and genetic data with the Medical Birth registry, as well as Army conscriptiondata.Furtherconsentforindividualisedlinkagewith population registries will be pursued especially for some Scandinaviansamples.

2.6.Criticalmeasures

2.6.1.Magneticresonanceimaging(MRI)

MRIatallsitesincludesT1-weightedscansformorphometric analyses.Inaddition,manycohortshavediffusion-weightedscans forstructuralconnectivityandresting-stateBOLD-weightedscans forfunctionalconnectivityanalyses.Diffusiontensorimaging(DTI) inparticularissensitivetomicrostructuralbraintissueproperties, and is a promising biomarker related to development, ageing, disease,andcognition[75].

2.6.2.Neuropsychology

Extensive cognitive tests are available for all cohorts, with validatedandreliablemeasures.Wefocus especiallyongeneral cognitiveabilitiesandepisodicmemory,inadditiontocognitive rating scales and measures of daily life function. With some variation across cohorts, cognitive measures will also include workingmemory,executivefunctionandprocessingspeed.

2.6.3.Geneticsandepigenetics

SalivaorbloodforextractionofDNAhavebeencollectedforall buttwoofthesamplesincludedinLifebrain.Inaddition,wewill examineabout2000furtherindividuals.

2.6.4.Driedbloodspots(DBS)

DBSwillbecollectedatmostsitesandanalysedbyVitas(WP2 andWP3).Prioritywillbegiventonutrients,suchasessentialfatty acids,cholesterol, andvitamins.Vitaswilldevelop andvalidate newbiomarkersonDBStailoredforLifebrain.Candidatesfornew biomarkersarespeciﬁcproteins,andfromtheVitascomprehen- sive UPLC-HR-TOF-MS lipidomics platform, a panel of lipid markerswithpotentialtopredictpreclinicaltransitiontoclinical stagesofAlzheimer'sdisease.

2.6.5.Lifestyleandhealthmeasures

Formostparticipants,informationaboutlifestyleandgeneral health is available, such as blood pressure, recordings of medicationandotherindices ofcardiovasculardiseaseandBMI, and willbeenrichedin aharmonizedwaythroughonline data collection.Dataonsleepingproblemsexistformostsubjectsand willbeenrichedthroughonlinedatacollection.

2.6.6.Mentalhealth

Standardizedmeasuresofsymptomsofdepressionandanxiety existformostparticipantsandwillbeenrichedinaharmonized waythroughonlinedatacollection.

2.7.Investigations

Ourapproachcallsforsophisticatedmodelling,andexpertsin longitudinalstatisticalmodelling,bioinformatics,andgeneticsare includedin thecoregroup,as wellasrenowned neuroimaging experts.Aspeciﬁctaskisdedicatedtobuildanintegratedmulti- modal processing stream for Lifebrain, based on combining existingtools(FreeSurferandFSL)withcustommadeprocedures andnewdevelopments.

2.7.1.Neuroimagingpreprocessingandstatistics

FreeSurfer (FS) will be used for quantiﬁcation of cortical thickness,volumeandlocalarealizationcontinuouslyacrossthe Fig.4.Bytrackingmemoryperformanceover15years,factoringinattrition,we

wereabletoshowthat18%of1558participantsupheldgoodmemoryfunctionin ageing,while13%declined[20].Thememoryscorewasacompositebasedon5 episodic-memorytasks(maxscore=76).

(7)

brainsurface,and volumesof arange ofsubcortical structures, includinghippocampus.FSsegmentsimagesatsub-voxelresolu- tion and detects ﬁne-graded effects, validated by manual segmentationsandhistology.Allcorticalmetricswillbesampled inthesamesurface-basedtemplate,yieldinguniquepossibilities for multi-modal integration by combining a range of different metricswhileensuringspatialcoherence.

ForDTIwewillcombinetwoapproaches.Wewillusetract- basedspatialstatistics (TBSS)developedintheFMRIBCentreat Oxford for whole-WM comparisons. This allows voxel-based comparisonsthatarerobusttoanatomicaldifferencesandpartial volumeeffects.ForquantiﬁcationofchangesinmajorWMtracts, we usea newly developed automated and robust probabilistic reconstructionscheme–Tracula(TRActsConstrainedbyUnderLy- ingAnatomy)[76].Itisespeciallysuitedforlongitudinalanalyses, withhighsensitivityandnobiasinchangeestimates.

2.7.2.Cross-sitestandardizationofneuroimagingdata

TheMRIscomefromdifferentscanners,whichwillaffectthe absolute measurement values. The applicants have extensive experience with large multi-centre neuroimaging, including ENIGMA [77], ADNI, e.g. [78,79], PING [27,28], and the Oslo Multi-Sample Aging Study [80]. The segmentation procedures usedinthisprojectarenotbiasedbyscannerplatformandseem nottoaffectthestrengthofrelationshipstoneuropsychological scores[81].Atlas-basednormalizationtoincreaserobustnessand accuracyofthesegmentationsacrossscannerplatforms[82],and normalizing analyses for differences in grey matter – white mattercontrastwillincreasesensitivityinmulti-sitestudies[83].

Also,thesizeofeachsubsamplewill belargeenough toallow statisticaladjustmentbyscanner.InternalvalidationintheOslo Multi-SampleAgingStudy,showedaminordecreaseinsensitivi- ty by pooling together data from six different samples and scanners was accompanied by a large increase in power. In summary,wearewellsuitedtotacklethechallengesinherentin anymulti-sitestudy,andbeneﬁtfromthemanifoldincreasesin sample size. As a feasibility check, we compared the age- trajectoriesof hippocampalvolumefromthe Cam-CANsample (n=651)andanLCBCsubsample(n=1100)andobservedhighly comparabletrajectories.

2.7.3.Geneticandepigeneticanalyses

Datagenerationforgenome-andepigenome-wideassociation study (GWAS and EWAS) analyses will be coordinated by University of Lübeck, using state-of-the-art high-throughput genome technologies for microarray-based profiling of DNA sequence(forGWAS)andmethylationstates(forEWAS).Several datasetshavegenome-wideSNPdataavailable,andinadditionwe willselect2000individualsfromatleastfourdifferentdatasetsfor denovocollectionofbuccalswabs,allowingustodetermineboth genome-wideSNPgenotypeandDNAmethylationprofilesusing the “Global Screening Array” and “Infinium Methylation EPIC” array, respectively (both Illumina Inc.). Buccal swabs will be collected by participants at home using the Catch-All Sample CollectionSwabs(Epibio,Inc.)followingstandardizedcollection protocols. Genome-wide data will serve several purposes in Lifebrain:First,theexistinggenotypeandnewlygeneratedDNA methylationdatacanbeusedinthecontextofGWASandEWAS analyses, to identify novel (and confirm previously reported) genetic/epigeneticdeterminants–andtheirinteraction–ofthe cognitiveandimagingtraitsofinterest.Second,thegenome-wide genotype data allow to precisely assess and correct for subtle differencesinpopulationsubstructurewithinandacrosscohorts allowingtooptimisedataanalysisproceduresandinferencesof non-geneticvariables.

2.8.Dataintegrationandstatisticalconsiderations 2.8.1.Statisticalmodellingofchange

Thecomplexlongitudinalnatureofthedataposessignificant statistical challenges in order to fully exploit the potential of combined data. To this end, Lifebrain has dedicated a task to developmentofnewstatisticaltools.Theconsortiumwilladdress statisticalchallenges inthree steps:(1)comparative analysisof datasetsandresearchdesigns;(2)developmentandapplicationof statisticaltools;(3)toolrefinementandmodelselection:ForStep 1,thedifferentresearchdesignsrepresentedinLifebrainwillbe compared to obtain an overview of their relative strengths, including differences in statistical power to detect effects of interest.Comparativeanalysiswillinformdataanalysisstrategies andfutureexpansionofdatasets.Wewillintroduceastatistical toolthatpermitsresearcherstocomputeeffectsizesthatallowfor unequal and person-specific measurement intervals, non-linear change, and selectiveattrition. For Step2,wewillpromotethe applicationofthreeinterrelatedsetsofstatisticaltools:multivariate and dynamic variants of longitudinal structural equation modelling (SEM); classification and regression trees (CART);

generalized additive mixed modelling (GAMM). Multivariate dynamicSEMiswellsuitedtoidentifylead-lagrelationsamong constructsrepresentingbrainfunctionsandstructures,cognitive performance,andhealthoutcomes.CARTandrelateddatamining techniques help to uncover classes of individuals with similar profilesandidentifyrelevantpredictorsofclassmembershipeven withverylargenumbersofinteracting predictors.GAMMoffers powerfultestsofnonlineareffectsonunivariatecriterionvariables, includinginteractionswithothercovariates.Wewillintroducetwo newtoolsthatcombinethebenefitsofSEMandCART:(i)SEMtrees [19]fordiscoveringformerlyundetectedsubgroupsthatdifferin SEMparameters;(ii)SEMforests[84]foridentifyingvariablesthat excelinpredictingindividualdifferencesinsuchparametersacross manypredictors.Wewillalsocombinethebenefitsofeachofthe threeapproaches byproposingsimultaneous estimationtechni- ques for generalized multi-level models and evaluate the conditionsunderwhichtheysurpassexistingstepwiseprocedures.

InStep3,wewillreﬁnethetoolsmadeavailableinSteps1and2 basedonfeedbackfromresearchthroughouttheproject.Further- more, wewill tackle theimportant problemofidentifying and selectingthosemodelsthatbestsummarizedataacrosssites.We will adapt split-sample schemes and information-theoretic measurestocontrolformultipletestingandavoidoveroptimistic inferencesduetodouble-dipping.Thegoalistoidentifymodels thatoptimallyrepresenttheconsolidatedﬁndingsoftheconsor- tium,andhencecontributetotheorydevelopmentandgeneraliz- ability.Membersofthegrouphavedevelopedstatisticaltoolsfor estimatingthestatisticalpowertodetectindividualdifferencesin change in the context of longitudinal studies analysed with structuralequationormulti-levelmodels,suchaslatentgrowth curve models [85–89]. Use of these tools will permit group members tocheckwhetherthe powertodetectmoderators of changewithagivendatasetisadequate,andinformfuturedesign decisions,suchasthespacingofmeasurements,orrecruitmentof newcohorts.In general,weexpectthelargedatabasesofhigh- quality, well-validated and in-depth measuresincludedin Life- braintoyieldexcellentstatisticalpower.

2.8.2.BIGdata–storage,transferandprocessingsystem

Sharingof brainimaging data between largecohort studies acrossEuropeforintegrativeandcomparativeanalysesisacoreof Lifebrain research.Lifebrainwillestablishaninternationalbrain image sharingand analysisplatform.For this, wewillbuildon expertise gained through existing collaborative data platforms amongtheparticipants(i.e.,theDementiasPlatformUKImaging

(8)

Informatics,DPUK-II;UOXF;UKBiobank,MRC;andtheMaxPlanck BrainImagingLibrary,OpenBILD,MPIB).

Integratedanalysesofthedataareenvisionedintwoforms:(a) fullyintegratedpre-processingandanalysisofrawimagingdata (mega-analysis);(b)statisticalintegrationoflocalpre-processing andanalysis(meta-analysis).Afullyintegratedanalysisofrawdata wouldprobablyachievethebest possibleintegration.However, formalandtechnicalconstraintsmightpreventsharingrawdata fromsomesites,sotheintegrationoflocallyprocesseddata(e.g., sharedintabulatedform)willbeconsidered.

Atleastforthemoment,theethicalpermissionofthedifferent participatingcentresisrestrictedtosharingwithintheLifebrain consortium.Asthesepermissionsrequiretobespeciﬁcandare slightlydifferentfor differentcountries, somedatasets willbe availabletoexternalapplicantsasaconditionoffunding(e.g.UK MRC),whileforothersaposthocethicalapprovalforsharingis requiredandwillbetiedtospeciﬁcconditions.

2.9.Uptakeofresearchoutputs

Lifebrainwilldevelopmechanismsandtoolstoengagevarious stakeholdersand bringtheirviewsandprioritiestotheproject.

Scientiﬁcexchangewilltakeplace withrelevantpolicymakers, nationaldecisionmakers,healthcareproviders,patientorganiza- tions,cohortparticipantsandresearcherstosupporttheuptakeof projectoutputs. These outputs will be translated into speciﬁc guidelinesandrecommendationsthatwillbeactivelydisseminat- edaccordingtotheconsortium’sdissemination,exploitationand communicationplan

3.Conclusions

Byintegratingandstandardizingmajorlongitudinalstudiesof brain,cognitionandmentalhealth,Lifebrainaimstomaximisethe potentialof Europeanbrainimaging cohorts.Thiswillfacilitate identification of the key determinants of cognitive and mental healthacrossallages,frombirthtooldage.Ourlifespanfocusfits withthenovellife-coursemodelofriskrecentlypublishedbythe LancetCommissiononDementiaprevention,interventionandcare [90].While wedo notspecificallytargetold ageand dementia, identificationoftheriskfactorsandprotectivefactorsatallstages oflifewillbecriticaltoenablefuturepreventionofcognitiveand mentaldisorders.

Acknowledgement

ThisresearchisfundedbytheEUHorizon2020Grant:‘Healthy minds0–100years:OptimisingtheuseofEuropeanbrainimaging cohorts (“Lifebrain”)’. Grant agreement number: 732592. Call:

Societalchallenges:Health,demographicchangeandwell-being.

AppendixA.

LifebrainConsortium(http://www.lifebrain.uio.no/about/)PIs:

KristineWalhovd(UiO–https://www.oslobrains.no/),LarsNyberg (Umeå–http://www.ufbi.umu.se/english),KlausEbmeier(UOXF– https://www.psych.ox.ac.uk/research/neurobiology-of-ageing), Ulman Lindenberger (MPIB – https://www.mpib-berlin.mpg.de/

en/research/lifespan-psychology),DavidBartrés-Faz(UB–http://

www.ub.edu/bbslab/), William Baaré (RegionH – http://www.

drcmr.dk/),RichardHenson(MRC–http://www.mrc-cbu.cam.ac.

uk/),ChristianDrevon(Vitas–http://www.vitas.no/),GunPeggy StrømstadKnudsen(NIPH–https://www.fhi.no/),BrendaPenninx (VUmc – https://www.vumc.com/), Paolo Ghisletta (UNIGE – http://www.unige.ch/fapse/mad/), OleRogeberg(Frisch–http://

www.frisch.uio.no/),LorraineTyler(UCAM–http://www.psychol.

cam.ac.uk/), Lars Bertram (LIGA UzL – http://www.liga.uni- luebeck.de/)

Research Associates: Anders M. Fjell, René Westerhausen, AthanasiaMonikaMowinckel, IngeAmlien(UiO),MicaelAnder- sson,SaraPudas,MikaelStiernstedt(Umeå),HeidiJohansen-Berg, EnikÅZsoldos,SanaSuri,ClaireE.Sexton(UOXF),SimoneKühn, AndreasBrandmaier,YlvaKöhncke,SandraDüzel(MPIB),Cristina Solé-Padullés, Barbara Segura Fabregas (UB), Kathrine Skak Madsen, Louise Baruël Johansen, Olga Rigina, Hartwig Siebner, MichaelKjær,EllenGarde,ErikLykkeMortensen(RegionH),Rogier Kievit (MRC), Thomas E. Gundersen, Tonje Fossheim (Vitas), IsabelleBudinLjøsne(NIPH), LauraNawijn(VUmc),SezenCekic (UNIGE),ChristinaM.Lill(LIGAUzL).

ProjectAdministration:BarbaraBodorkosFriedman(UiO).

References

[1]GustavssonA,SvenssonM,JacobiF,AllgulanderC,AlonsoJ,BeghiE,etal.Cost ofdisordersofthebraininEurope2010.EurNeuropsychopharmacol2011;21 (10):718–79.

[2]Habib R, Nyberg L, Nilsson LG. Cognitive and non-cognitive factors contributingtothelongitudinalidentiﬁcationofsuccessfulolderadultsin thebetulastudy.NeuropsycholDevCognBAgingNeuropsycholCogn2007;14 (3):257–73.

[3]NevalainenN,RiklundK,AnderssonM,AxelssonJ,OgrenM,LovdenM,etal.

COBRA:aprospectivemultimodalimagingstudyofdopamine,brainstructure andfunction,andcognition.BrainRes2015;1612:83–103.

[4]FilippiniN,ZsoldosE,HaapakoskiR,SextonCE,MahmoodA,AllanCL,etal.

Studyprotocol:theWhitehallIIimagingsub-study.BMCPsychiatry 2014;14:159.

[5]GathercoleSE,WoolgarF,TeamC,KievitRA,AstleD,ManlyT,etal.How commonareWMdeﬁcitsinchildrenwithdifﬁcultiesinreadingand mathematics?JApplResMemCogn2016;5(4):384–94.

[6]TaylorJR,WilliamsN,CusackR, AuerT,ShaftoMA, DixonM,etal. The CambridgeCentreforAgeingandNeuroscience(Cam-CAN)datarepository:

structuralandfunctionalMRI,MEG,andcognitivedatafromacross-sectional adultlifespansample.Neuroimage2017;144(PtB):262–9.

[7]HulurG,DreweliesJ,EibichP,DuzelS,DemuthI,GhislettaP,etal.Cohort differencesinpsychosocialfunctionover20years:currentolderadultsfeel lesslonelyandlessdependentonexternalcircumstances.Gerontology 2016;62(3):354–61.

[8]FjellAM,SneveMH,GrydelandH,StorsveAB,AmlienIK,YendikiA,etal.

Relationshipbetweenstructuralandfunctionalconnectivitychangeacrossthe adultlifespan:alongitudinalinvestigation.HumBrainMapp2017;38(1):561–

73.

[9]Dijkstra-KerstenSMA,Sitnikova K,TerluinB, PenninxB, TwiskJWR,van MarwijkHWJ,etal.Longitudinalassociationsofmultiplephysicalsymptoms withrecurrenceofdepressiveandanxietydisorders.JPsychosomRes 2017;97:96–101.

[10]MadsenKS,BaareWF,SkimmingeA,VestergaardM,SiebnerHR,JerniganTL.

Brainmicrostructuralcorrelatesofvisuospatialchoicereactiontimein children.Neuroimage2011;58(4):1090–100.

[11]RamsoyTZ,LiptrotMG,SkimmingeA,LundTE,SidarosK,ChristensenMS,etal.

Healthyagingattenuatestask-relatedspecializationinthehumanmedial temporallobe.NeurobiolAging2012;33(9):1874–89.

[12]ManeruC,JunqueC,Salgado-Pineda P,Serra-GrabulosaJM,Bartres-FazD, Ramirez-RuizB,etal.Corpuscallosumatrophyinadolescentswith antecedentsofmoderateperinatalasphyxia.BrainInj2003;17(11):1003–9.

[13]EngvigA,FjellAM,WestlyeLT,MobergetT,SundsethO,LarsenVA,etal.Effects ofmemorytrainingoncorticalthicknessintheelderly.Neuroimage2010;52 (4):1667–76.

[14]NganduT,LehtisaloJ,SolomonA,LevalahtiE,AhtiluotoS,AntikainenR,etal.A 2yearmultidomaininterventionofdiet,exercise,cognitivetraining,and vascularriskmonitoringversuscontroltopreventcognitivedeclineinat-risk elderlypeople(FINGER):arandomisedcontrolledtrial.Lancet2015;385 (9984):2255–63.

[15]BeekmanAT,CopelandJR,PrinceMJ.Reviewofcommunityprevalenceof depressioninlaterlife.BrJPsychiatry1999;174:307–11.

[16]Schweizer S, Kievit RA, Emery T, Cam CAN, Henson RN. Symptoms of depressioninalargehealthypopulationcohortarerelatedtosubjective memorycomplaintsandmemoryperformanceinnegativecontexts.Psychol Med2017;1–11.

[17]SkirbekkV,LoichingerE,WeberD.Variationincognitivefunctioningasa reﬁnedapproachtocomparingagingacrosscountries.ProcNatlAcadSciUSA 2012;109(3):770–4.

[18]WeberD,SkirbekkV,FreundI,HerlitzA.Thechangingfaceofcognitivegender differencesinEurope.ProcNatlAcadSciUSA2014;111(32):11673–8.

[19]Brandmaier AM, von Oertzen T, McArdle JJ, Lindenberger U. Structural equationmodeltrees.PsycholMethods2013;18(1):71–86.