ContentslistsavailableatScienceDirect
NeuroImage
journalhomepage:www.elsevier.com/locate/neuroimage
Immediate
effects
of
a
single
session
of
physical
exercise
on
cognition
and
cerebral
blood
flow:
A
randomized
controlled
study
of
older
adults
Gaia Olivo
a,∗, Jonna Nilsson
a,b, Benjamín Garzón
a,c, Alexander Lebedev
a,d, Anders Wåhlin
e,f,
Olga Tarassova
b, Maria Ekblom
b,g, Martin Lövdén
a,ca Aging Research Center (ARC), Department of Neurobiology, Care Sciences and Society (NVS), Karolinska Institutet, Tomtebodavägen 18A, 171 65 Stockholm, Sweden b The Swedish School of Sport and Health Sciences, Stockholm, Sweden
c Department of Psychology, University of Gothenburg, Gothenburg, Sweden d Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden e Department of Radiation Sciences, Umeå University, Umeå, Sweden
f Umeå Center for Functional Brain Imaging (UFBI), Umeå University, Umeå, Sweden g Department of Neuroscience, Karolinska Institute, Stockhom, Sweden
a
r
t
i
c
l
e
i
n
f
o
Keywords: Physical exercise Working memory Cerebral blood flow Cerebral perfusion Physical activity ASL
a
b
s
t
r
a
c
t
Background:Regularphysicalactivityisbeneficialforcognitiveperformanceinolderage.Asingleboutofaerobic physicalexercisecantransientlyimprovecognitiveperformance.Researchershaveadvancedimprovementsin cerebralcirculationasamediatoroflong-termeffectsofaerobicphysicalexerciseoncognition,buttheimmediate effectsofexerciseoncognitionandcerebralperfusionarenotwellcharacterizedandtheeffectsinolderadults arelargelyunknown.
Methods: Forty-nineolderadultswererandomizedtoa30-minaerobicexerciseatmoderateintensityor re-laxation.Groupswerematchedonageandcardiovascularfitness(VO2max).AverageGreyMatterBloodFlow (GMBF),measuredbyapulsedarterial-spinlabeling(pASL)magneticresonanceimaging(MRI)acquisition,and workingmemoryperformance,measuredbyfigurativen-backtaskswithincreasingloadswereassessedbefore and7minafterexercising/resting.
Results: Accuracyonthen-backtaskincreasedfrombeforetoafterexercising/restingregardlessofthetypeof activity.GMBFdecreasedafterexercise,relativetothecontrol(resting)group.Intheexercisegroup,higher n-backperformanceafterexercisewasassociatedwithlowerGMBFintherighthippocampus,leftmedialfrontal cortexandrightorbitofrontalcortex,andhighercardiovascularfitnesswasassociatedwithlowerGMBF.
Conclusion:ThedecreaseofGMBFreportedinyoungeradultsshortlyafterexercisealsooccursinolderadultsand relatestocardiovascularfitness,potentiallysupportingthelinkbetweencardiovascularfitnessand cerebrovas-cularreactivityinolderage.
1. Introduction
Physical activityis associated with better cognitive performance (Costaetal.,2019;Liu-Ambroseetal.,2018;MoreauandChou,2019) andreducedriskfordementia(Liu-Ambroseetal.,2018),depression (Kandolaetal.,2019)andstroke(Krameretal.,2019).Asinglebout ofphysicalexerciseatmoderateintensity(3–6metabolicequivalents (METs; i.e. requiringthree tosix times as much energy per minute asresting)canimprovecognitiveperformanceimmediatelyafterand atleast up until30minafterexercising(Chang etal.,2012). These acuteeffectsofphysicalexerciseareparticularlypronouncedon work-ingmemory (WM)performance (Weng et al., 2015; Pontifex et al., 2009;Vossetal.,2020;Wheeleretal.,2019)andexecutivefunctions
∗Correspondingauthor.
E-mailaddress:gaia.olivo@ki.se(G.Olivo).
(Wheeleretal.,2019),asindicatedbyinterventionstudiesinyounger adultsandrandomizedcontrolledtrialsinolderadults.Theimmediate exercise-inducedeffectoncerebralbloodflowmightplaya preponder-antroleinboostingcognitiveperformanceafterasingleboutofphysical activity(Davenportetal.,2012),contributingtoaugmentingthesupply ofexcitatoryneurotransmittersinkeybrainregionsshortlyafter exer-cise(McMorrisandHale,2012;Vecchioetal.,2018;Szalewskaetal., 2017),thatmaythenenhancecognitionbyinfluencingcorticalarousal (Vecchioetal.,2018;Szalewskaetal.,2017).Therelationshipbetween exercise and cognition followsan inverted U-shape, with moderate-intensityexercisebeingthemosteffectiveininducingshort-and long-termpositiveoutcomesoncognition(Changetal.,2012).Aerobic ex-ercisehasbeenindicatedasthemostefficient,possiblyexerting
benefi-https://doi.org/10.1016/j.neuroimage.2020.117500
Received16June2020;Receivedinrevisedform15September2020;Accepted21October2020 Availableonline24October2020
cialeffectsonexecutivefunctionsandWMalreadyafterinterventionsof lessthan16weeks(Cabraletal.,2019),withsmalltomoderateeffects (Sandersetal.,2019).
Although the positive effects exerted on cognition are observed acrosstheentirelifespan,exerciseseemstobemoreefficientin boost-ingcognitiveperformanceinchildrenandolderadults,whenexecutive functionsarerespectivelydevelopingordeclining(Liu-Ambroseetal., 2018;AudiffrenandAndre,2019).Arecentmeta-analysisof random-izedcontrolledtrialsinadultshashighlightedtheimportanceofagein moderatingtheeffectofregularphysicalactivityinterventions,stressing thatpositiveeffectsoncognitionareparticularlypronouncedinolder adultsover65 years(RathoreandLom, 2017).Fronto-parietalbrain regions,keycontributorstoexecutivefunctions,areinfactparticularly vulnerabletoaging,andmightthereforebemoresensitivetotheimpact ofmoderatorssuchasexercise(AudiffrenandAndre,2019).Moreover, associationareassuchasthemedialfrontalregionsandthelimbic cor-tices,whicharealsoinvolvedinexecutivefunctions,attention,learning andmemory,arethemostaffectedbythedeclineinrestingcerebral bloodflowoccurringwithaging(Bentourkiaetal.,2000;Chenetal., 2011).
Researchershavesuggestedthatangiogenesisandimprovementsin cerebralcirculationmaybecrucialforthelong-termbeneficialeffects ofphysicalexerciseoncognition(Vecchioetal.,2018;Tyndalletal., 2018).Cerebralbloodflowdeclineswithage,asconsequenceofarterial stiffness,lowerbrainengagementandactivityatrestandpossibly synap-ticloss(Bentourkiaetal., 2000).Cerebrovascularreserve,definedas theabilityofthecerebralbloodvesselstodilateinresponsetoachange inmetabolicdemand,isalsoimpairedinaging(Tyndalletal.,2018). Regularexercisemaythushelpcounteractthedetrimentalimpactof agingonbrainhealthandcirculation(BarnesandCorkery,2018),by increasingglobalcerebralbloodflow,particularlybyenhancingblood flowvelocitythrougharteriessupplyingtheneocortex,and/orby im-provingcardiovascularreactivity(Wheeleretal.,2019;Tarumietal., 2015;Brugniaux etal., 2014).However,therelationshipof cerebral andperipheral vascular function with cognitivefunction is still de-bated (Tyndall etal., 2018; Tarumietal., 2015; Lucasetal., 2012; Youngetal.,2015).Importantly,suchrelationshipseemstobemediated byseveralfactors,notleastthecognitivehealthstatusoftheindividual (Stillmanetal.,2020).Positiveeffectsof long-termphysical exercise oncognitionhavebeenreportedinhealthyolderadultsaswellasin olderadultswithmildcognitiveimpairment(MCI),thoughwith differ-entunderlyingmodificationsinthecerebralbloodflow(Alfinietal., 2019).Indeed,adultswithMCIhadahigherbloodflowintheinsula comparedwithhealthyadults,thatdecreasedtocontrollevelsafter12 weeksofexercise(Alfinietal.,2019).Healthyolderadultsshowed in-steadincreasedbloodflowintherightanteriorcingulatecortexafterthe exerciseprogram(Alfinietal.,2019).Ontheotherhand,olderadults with already developed mild-to-moderate Alzheimer disease showed noimpactofphysicalactivityoncerebralbloodflowaftera16-week moderate-to-highintensityexerciseprogram(vanderKleijetal.,2018). Long-terminterventionsfeaturingrepeatedsessionsofphysicalactivity arehoweverlikelytoengageavarietyofcompensatoryand plasticity-basedbrainmechanismsandarenotdirectlyinformativeofthe path-waysbehindtheeffectsofphysicalexercise(BassoandSuzuki,2017). Thus,theneedforamorein-depthinvestigationoftheeffectsof sin-glesessionsofphysicalexercise,particularlyifaidedbythecomparison witharestingcontrolgroupmatchedforbaselinecardiovascularfitness, hasbeenhighlighted(Stimpsonetal.,2018).
However,theimmediateeffectsofsinglesessionsofphysical exer-ciseoncerebrovascularandcognitivefunctionarenotwelldescribed sofar,andtheeffectsinthehealthyolderpopulationarelargely un-known(RathoreandLom,2017;BassoandSuzuki,2017).Thepaucity ofstudiesontheimmediateeffectsofexerciseinolderagemighthave contributedtothelimitedpossibilityformeta-analyticevidence con-cerningthecognitiveeffects(orlackthereof)ofsinglesessionsof
phys-icalactivitysofar(RathoreandLom,2017;BassoandSuzuki,2017). Moreover, littleisknownabout theimmediateeffectof physical ex-ercise on cerebrovascular responses and their relation to cognition. Similar to cognition, the cerebrovascular response tophysical exer-cisealsodependsontheintensityofexercise.Theincreaseincerebral bloodflowinfactmirrorstheintensityofthephysicalexertionupto 60%ofmaximaloxygenuptake(OgohandAinslie,2009;Queridoand Sheel,2007).Beyondthatpoint,exercise-inducedhyperventilationand sympathetically-mediatedprotectivemechanismstriggercerebral vaso-constriction, causinga plateau or evena reductionin cerebral flow (Ogoh andAinslie,2009; QueridoandSheel,2007).Independent of theintensityofexertion,thepost-exercisehypotensionand consequen-tialdropincardiacoutputleadtoaquickdecreaseincerebralblood flow,detectableafewsecondsafterthecessationoftheexercisebout (QueridoandSheel,2007).Atleastinyoungadults,uptofortyminutes mightbenecessaryforthegreymatterbloodflowtoresumeits base-linelevels(MacIntoshetal.,2014),followingaphysiologicalre-boost andreactivevasodilationinducedbycerebralhypoxiaandbythe nor-malizationofthecardiacoutput(QueridoandSheel,2007).However, findingsarestillconflicting.Increasesofapproximately20%inglobal cerebralbloodflow comparedwithbaseline aftermoderateintensity exercisehavealsobeenreportedinyoungadults,lastingupto30min afterexercise(Smithetal.,2010).Localincreasesby10–12%inblood flowhavealsobeenreportedtooccurinthehippocampusofyoungand middle-agedadultsat15,40and60minafterasingleboutof moder-ateintensityphysicalexercise(Steventonetal.,2020).Worthnoting, allthesestudieshavemeasuredcerebralbloodflowofindividualsina supinepositioninthescanner,indicatingthatperfusionchangescanbe detectedevenwhenlayingdown,whentheimpactofcardiacoutput andvascularpressureonthebrainisnotascriticalasintheupright position. Moreover, theexercise-related regulation of cerebralblood flowinolderageisstillunclear.Olderadultshaveelevatedhypocapnic responses(Stefanidisetal.,2019)(i.e.theyaremore proneto react-ingwithcerebralvasoconstrictioninresponsetohyperventilation)and lowerpartialpressureofcarbondioxide(PaCO2)comparedwithyoung adults(BrazandFisher,2016).Cerebrovascularreactivityalsoseemsto impactcognitiondifferentlyinyoungerandolderage(Catchloveetal., 2018).
Toourknowledge,nostudieshaveexaminedimmediate exercise-relatedchangesoncerebralbloodflowandcognitioninolderadults. Giventhepotentialapplicationsofexerciseastreatmentaidformild cognitive impairment and dementia prevention (Alfini et al., 2019; Gedaetal.,2010),itisofgreatimportancetounderstanditsimpact oncerebralperfusioninolderadults.Tobridgethisknowledgegap,we thereforeinvestigatedtheeffectsofasinglesessionofmoderate inten-sityphysicalexerciseoncerebralbloodflowandWMperformancein older adults.Participants, matchedonage andcardiovascular fitness (VO2max),wererandomizedtoundergo30-minofexercisingor rest-ing. Wehypothesizedthatexercisewould improveWMperformance andincreaseaveragegreymatterbloodflow(GMBF)comparedwith resting,andthattheincreaseinGMBFwouldpositivelycorrelatewith theimprovementincognitiveperformance.Wealsohypothesizedthat increasesinpulsepressurewouldbepositivelyassociatedwithincreases inGMBF.Thestudywaspre-registeredintheOpenScienceFoundation database(osf.io/ve2zj).
2. Methods
2.1. Ethics statement
Thestudywasapprovedbytheethicscommitteeofregion Stock-holminSweden(RegionalaEtikprövningsnämndeniStockholm,Dnr: 2018/1340-31/2)andcompliedwiththeprinciplesoftheDeclaration ofHelsinki.
Fig.1. Flow-chartoftheexperimentalprotocol.Theflow-chartreportsthe se-quentialsessionsoftheexperimentalprotocols.Thenumberofindividuals par-ticipatingineachsessionandthenumberofdrop-outsarenoted.
2.2. Subjects
Participantswererecruitedviaadvertisementsinlocalnewspapers andflyers inthe Stockholmareabetween September andDecember 2018,seekingforoldervolunteers,agedbetween65and75years,fora studyinvestigatingtheeffectsofexerciseandrestingonbrainfunction. Tworesearch assistantswereresponsibleforthe recruitment.A pre-liminarycheckforeligibilityaccordingtostudycriteriawasperformed viatelephoneinterviewoftheinterestedparticipants.Inclusioncriteria weretheavailabilitytoattendallsessions,adequatehearing,normal orcorrectedvision,fluentSwedish(requiredtounderstandthe instruc-tions),adequatemobilityinordertoensuretheycouldliestillinthe magneticresonanceimaging(MRI)scannerforthewholeexperimental session,right-handedness,andweightbelow120kg.Exclusioncriteria were:previousorcurrentneurologicaldiseases,includingParkinson’s Diseases,dementiaandepilepsy;ascorebelow26ontheMini-Mental StateExamination(MMSE)(Folsteinetal.,1975);previousorcurrent cardiovasculardiseases,thoughbloodpressureupto200/100mmHg wasallowed;historyofbraindamageorstroke;uncontrolledmetabolic diseases;typeIorpharmacologically-treatedtypeIIdiabetes(typeII diabeteswasallowediftreatedonlywithdietary,non-pharmacological interventions);currentcancer,unlessmorethanoneyearhadpassed sincethetreatmentend;psychiatricillness,exceptforahistoryofmild tomoderatedepressionand/oranxietyallowed;historyofheadtrauma resultinginlossofconsciousness;previousparticipationinstudies in-volvingcognitivetests;presenceofmetalinthebody;claustrophobia; soundsensitivity(duetoMRInoise); neuromotoror muscoloskeletal dysfunctionslimitingorimpedethefitnesstest;cardiovascular-active medicationspotentiallyaffectingmaximalandsubmaximalfitnesstests; ongoinginfections;chestpain.
Individualsfulfillingthestudycriteriawereinvitedtoattendan in-troductionmeetingattheAgingResearchCenter(KarolinskaInstitute, Stockholm).Theprotocolconsistedofthreesessions:introductory meet-ing,fitnesscharacterizationsession,andexperimentalsession(Fig.1).
Ofthe73individualswhohadinitiallyexpressedtheirinterestinthe study,50joinedtheexperimentalsession(Fig.1).Onewasexcluded duetoMRIsignssuggestiveofpast,undiagnosedstroke.
Samplesizewasdeterminedbystatisticalpoweranalyses(G∗Power).
Themeta-analysisof McMorrisandHale(McMorrisandHale,2012) indicatedaneffectofmoderateexercise oncognitiveperformanceof 0.5SD,butthisislikelyanoverestimationduetopublicationbias. In-stead,webasedouranalysesonaconservativeestimateofhalfofthis effect(0.25),withsmallereffectsthanthisalsobeingoflimited inter-estinthiscontext.Assuming measurementreliabilityandcorrelation amongrepeatedmeasuresof0.8forourcognitiveperformance mea-sures(Lebedevetal.,2018) andanalphalevelof0.05,weneededa totalsamplesizeof54(27pergroup)todetectthiseffectwithapower of0.8.Thefinalsamplesizewas51individuals.Atapost-hocpower analysiswiththesameabove-statedparameters,thepoweractualpower resultedtobe0.79.Moredetailscanbefoundinthepreregistration. 2.3. Behavioral and demographic measures
Duringtheintroductorymeeting,participantsreceiveddetailed in-formation concerning thestudy.Relaxation exerciseshave alsobeen reported to exert positive effects on cognition (Ma et al., 2017; Ferreiraetal.,2015;Tangetal.,2007;Gardetal.,2014;Galvinetal., 2006;Siponkoskietal.,2019),andtheparticipantswerenotinformed aboutthestudyhypothesis(i.e.whethermoderateintensityphysical ex-ercisewasexpectedtobesuperiortorestinginimprovingcognitionand brainmarkers),withtheaimofminimizingtheinfluenceofexpectations onthecognitiveperformancemeasurements.Nexttheysignedthe in-formedconsentform.Allparticipantsfilledinquestionnairesfocused ondemographicinformation,aswellastheEdinburghHandedness In-ventorytoconfirmright-handednessandtheMMSE.Habitualphysical activitywasassessedwiththeInternationalPhysicalActivity Question-naire(IPAQ)(Craigetal.,2003).Participantswerealsoaskedtorate theirmemory,toratetheirmemorycomparedtowhentheywere20 yearsold,andtoratetheiragreementwiththefollowingsentence:“I thinkwehaveacertainmentalabilitythatcannotbechangedtoany greaterdegree”.Bloodpressurewasalsomeasured.Instructionsonhow toperformthen-backtasktobedoneinthemainexperimentalsession werealsohandedout,andparticipantscouldtryitoutfor5min. 2.4. Fitness characterization
Priortothefitnesscharacterization,asecondhealthscreeningwas performedviatelephoneinterviewbyaphysicianatÅstrand Labora-toryattheSwedishSchoolofSportandHealthSciences,GIH.The fit-nesscharacterizationwasperformedatGIH.Participantsperformeda submaximalincrementaltestonacycleergometerandamaximal incre-mentalrunningtestonatreadmilltoassesstheirVO2maxandmaximal heartrate(Nilssonetal.,2020).Theheartrate,O2andCO2-data ac-quiredduringthesubmaximalcyclingwereusedtoidentifyaworkrate andheartrateatwhichtheVO2reachedaround60%ofthatrecorded duringthemaximalrunningtest.Thisworkrateisexpectedtoresultin aperceivedlevelofexertionfrom13to15ontheBorgRPEscaleand aheartrateofaround60–70%ofmaximalindividualheartrate.The submaximaltestonthebikewasnotusedtoestimateVO2max,butto identifytheindividualworkrateandheartrateduringbikingatwhich theVO2reachedaround60%oftheVO2maxrecordedduringthe maxi-malrunningtest.Thisindividualworkrateandheartratewasthenused duringbikingintheexerciseinterventionoftheexperimentalsession. ThiswasnecessarysincewedidnotrecordVO2duringtheexperimental sessions.
Thesubmaximaltestwasperformedon acycleergometer(model 828E,Monark,Varberg,Sweden).Theprotocolstartedwith4minutes cyclingat astandardworkrate(resistanceof0.5kp, 60rpm), after whichtheresistancewasincreasedinstepsof0.5–1kp,untilreaching
60–80%ofindividualmaximalcapacityandarateofperceivedexertion (RPE)ofaround16(andnotabove16).
Themaximalincrementalrunningtestonatreadmillwasusedto di-rectlymeasureVO2max.Participantswarmedupfor5–10minbefore startingthetest.Thetreadmillwasinitiallysettoaninclineof1° and acomfortablespeed(aroundRPE12–13).Inclinedegreeand/orspeed wereincreasedeveryminuteuntilvolitionalexhaustion.Participants woreasafetyharnesstopreventfallingandincreasetheirconfidence inreachingexhaustion.VO2maxwasmeasuredusingacomputerized metabolicsystem(JaegerOxyconPro,Hoechberg,Germany). Accord-ingtopreviousstudies(Bjorkmanetal.,2016;Ekblom-Baketal.,2014), thefulfilmentofthreeoutoffiveofthefollowingcriteriawasrequired toacceptthevalidityoftheVO2maxmeasurement:(a)VO2was lev-elingoff despiteanincreaseinspeedordecline,(b)RPEexceeded16, (c)arespiratoryexchangeratioexceeded1.1,(d)maximalHRwithin ±15beatsperminutefromage-predictedmaximalHRand(e)awork withtimeabove6minuteswasperformed.Thehighest30secondsof registeredvaluesofVO2werereferredtoasVO2max.
2.5. Randomization and blinding
Beforetheexperimentalsession,participantswererandomized(1:1) totwogroups:restingandexercisingcondition.AgeandVO2maxwere usedasstratifiers.Therandomizationwasconductedusinglabel shuf-flingwithpost-hocnon-parametrictestsforthestratifiers,asdescribed byLebedevetal.(2020).Thecodefortherandomization procedure waswrittenbyALandusedbyALandMVH(seniorlabmanager;see acknowledgments)toallocatetheparticipantstotheexerciseorresting condition.Thesubjectswerenotawareofwhichoftheparadigms rep-resentedcontrolandactiveconditions.Noblindingwasrequiredforthe staff collectingthedataforthisstudy.
2.6. Experimental session protocol
Noblindingwasrequiredforthisstudy.Theexperimentalsession wasperformedat theSiemens Prisma3-Teslafacilityatthe Karolin-skaUniversityHospitalinHuddinge.Priortothescanning,participants wereallowedtopracticetheWMtaskfor10min.MRIacquisitionwas carriedout atbaseline,andafter30minofexercise orresting. Sub-jectsrandomizedtotheexercisegroupcycledonastationarybikefor 30min,startingattheintensityestimatedfromthefitness characteri-zation.Subjectsrandomlyallocatedtotherestingconditionwereasked tolaydownandrelaxfor30min,whilelisteningtorelaxingmusicwith water-soundsinthebackground.Heartratewasmeasuredwitha chest-strapheartratemonitorthroughouttheperiodbetweenthescanning.
Thepost-testMRIscanningstartedsevenminutesaftertheendof the;thistimewasnecessaryforplacing,inarelaxedmanner,the sub-jectbackinthescanner.Systolicanddiastolicbloodpressurewas mea-suredatfourdifferenttime-points:beforepre-testscanning;after pre-testscanning(beforeexercising/restingfor30min);afterphysical exer-cise/resting(beforethepost-testscanning);andafterpost-testscanning. Theimagingprotocolconsistedofgreymatterbloodflowassessmentvia awhole-brain3Dpulsedarterial-spinlabeling(pASL) sequence, task-fMRIacquiredduringtheperformanceofthen-backtask;anatomical assessmentviaT1-weightedmagnetizationpreparedgradient-echo se-quence(MPRAGE).AroutineclinicalT2-weightedstructuralMRIwas alsoacquiredatpre-testfortheneuroradiologicalassessmentofthe par-ticipants.Forthecurrentstudyandaccordingtopreregistration,only thepASLandtheMPRAGEsequenceswereincludedintheanalyses. 2.7. Adverse events
Onesubjectexhibitedunstableheart-beatafterthemaximalfitness test,andwasbroughttothehospitaltoseekmedicalattention. This subjectwasexcludedfromtherandomizationprocessandthusfromthe
study. Nootheradverse eventsoccurred duringthephysical activity procedures.
2.8. . N-Back task
N-backperformancewasmeasuredbeforeandaftertheexerciseor restingsession.Thetaskconsistedofafiguraln-backtasktoassessWM performance.Three blockswithincreasingcognitiveload(1-back, 2-back,3-back)wereadministeredinanalternatefashionforthreetimes, witha 4-sfixationscreen betweenblocks indicatingtherulefor the forthcominglevel.Eachblocklastedoneminute.Theperformanceon thetaskwasexpressedintermsofaccuracy,calculatedas:
(correcthits+correctrejections)∕totalstimuli;
Accuracyforthe1-back,2-back,and3-backwascalculated,along withaverageaccuracyacrossloads.Reactiontimewasalsorecorded. Priortopre-testimaging,ashortpracticeofthen-backtaskwas per-formedwhenthesubjectwasinthescanner.Duetoamalfunctioningof thekeyboard,then-backdataofnineparticipantscouldnotberecorded. 2.9. MRI acquisition
Greymatterbloodflowwasrecordedviaawhole-brain3Dpulsed arterial-spinlabeling (pASL)sequence(scanduration≈5min). Arte-rialspinlabelingmeasurescerebralbloodflowbymagnetically label-ingarterialwaterandusingitasanendogenoustracer.Thefollowing parameters wereusedfor thepASL acquisition:repetitiontime(TR) 4000ms;timetoecho(TE)12ms:inversiontime(TI)2000ms;field ofview(FOV)128×128;1.9×1.9×4.5mm3;slicethickness4.5mm;
32slices.T1-weightedmagnetizationpreparedgradient-echosequence (MPRAGE)wasusedtoassessbrainstructure(scanduration≈5min), withthefollowingparameters:TR2300ms;TE2.01ms;flipangle:9°; FOV240 × 256;voxelsixe:1mm3;slicethickness1mm;208slices.
2.10. ASL images pre-processing
Priortopre-processing,ASLdatawerecheckedforoutlierson move-ment.Rootmeansquaresignalchange(DVARS)values,definedasthe rootmeansquareintensitydifferenceofpairedvolumes(volumeNto volume N + 1) (Power et al., 2012),wereused tocheck for move-ment.Data-pointwithDVARSvaluesexceedingthethresholdof75th percentile+1.5times theInterQuartileRange(IQR)wereexcluded. Thenumberofdata-pointswithexcessivemovementrangedbetween0 and15.Subjectswithmorethan30%ofthedata-pointsexceedingthis thresholdwereremoved.Thisledtotheexclusionofthreeindividuals, ofwhichonesubjectbelongedtotheexercisegroupandtwosubjects belongedtotherestinggroup.
ASLdatawerepre-processed withFSL(FMRIBSoftwareLibrary), byusingtheoxford_aslcommandincludedintheBASIL(Bayesian In-ference forArterial Spin labeling MRI)toolbox toprocess ASLdata (Chappelletal., 2009).Structuraldatawerepre-processed according thefsl_anatpipeline,consistingofthefollowingsteps:(1)theimages arereorientedtomatchthestandardMontrealNeurologicalInstitute (MNI)orientation;(2)theimagesarethencroppedand(3)bias-field correctionisapplied,usingtheFAST(FMRIB’sAutomatedSegmentation Tool)algorithm(Zhangetal.,2001);(4)theimagesarelinearly (us-ingFLIRT-FMRIB’sLinearImageRegistrationTool)(Jenkinsonetal., 2002)andnon-linearly(usingFNIRT-FMRIB’sNon-linearImage Reg-istrationTool)registeredtotheMNIspace;(5)brainextractionis per-formedwiththeBetExtractionToolbox(BET)(Smith,2002)and(6) tissue-typesegmentationiscarriedoutwiththeFASTalgorithmfor cor-ticalstructuresandwithFIRST(FMRIB’sIntegrated Registrationand SegmentationTool)forsubcorticalstructures.Thepre-processed struc-turalimageswerefedintotheoxford_aslcommandinordertoachieve theregistrationoftheASLdatatotheMNIstandardspace.
The oxford_aslcommand automatically performs a seriesof pre-processingstepsonASLdata,inanautomatedpipeline.First,motion correctionisperformed.Atissueperfusionmapisthenproduced,and registeredtotheMNIstandardspacebyapplyingtheparameters gen-eratedduringtheregistrationofthestructuralimageandstoredina transformationmatrix.Thecalibratedperfusionmapisthengenerated usingtheprotondensityweightedimageandsettingthecerebrospinal fluid(CSF)asareferencetocalculatetheequilibriummagnetizationof bloodwithinaCSFmask,automaticallygeneratedfromthe segmenta-tionofthestructuralimageandnormalizedtotheMNIspace.Aquality checkwas performedvisuallytoensurethegoodqualityof the pre-processing.Thecalibratedimageswereusedforsubsequentstatistical analyses,astheyexpressgreymatterbloodflowasml/100g/min.Prior tostatisticalanalyses,theimagesweresmoothedwithan8mmFWHM gaussiankernel.
2.11. Statistical analyses
Toassessstabilityofthemeasurementovertime,wecalculated pre-topost-testcorrelationsforaveragen-backaccuracy,bloodflow,and movement(meanDVARS)separatelywithineachgroup.Thedatawere notnormallydistributed,thusSpearman´srhowasusedforcorrelation analyses.Thethresholdforsignificancewassetat p <.05.Belowwe describethepreregisteredhypotheses(Hi)andresearchquestions(RQi)
andthestatisticalanalyses(confirmatoryandexploratory,respectively) usedtoaddressthem.
2.12. H1. Working memory performance will improve more after exercise than after resting
Statistical analyses were performed with SPSS (Statistical Pack-age for Social Science), v26 ( https://www.ibm.com/analytics/spss-statistics-software).Ninepeoplehadmissingdataon then-back per-formance.Then-back subsamplethusconsistedof20peoplein each group.
Box plots were used to check for the presence of outliers on group∗session∗loadcombinations.Outliersweredefinedasmildfor
val-uesbetween1.5and3IQR,andasextremeifoutside3IQR.Five data-pointsreachedthecut-off forbeingextremeoutliers,whilesixadditional data-pointswerecategorizedasmildoutliers.Theseledtothe identifi-cationoftwoextremeoutliersandtwomildoutliers,allbelongingto therestinggroup.Foreachoutlier,correspondingdata-pointsonboth sessionswereremoved,asourinterestlaidinperformancechangeover time.Extremeoutlierswereexcluded,whilemildoutlierswereretained astheycanbereflectiveofthenormalpopulationvariability,whenthe sampleisnotbigenoughtoberepresentativeofthewholepopulation. Nonetheless,theanalyseswererunwithandwithoutmildoutliersto confirmthatresultswerenotdrivenbythepresenceofoutliers.Asthe presenceofmildoutliersdidnotimpacttheanalysis,theywereretained forsubsequentcorrelationanalyses.
Ageneralizedlinearmixedmodel(GLMM)withabinomial distribu-tionandlogitlinkfunctionwasusedtoanalysethetaskperformance. Thebinomialdistributionfunctionspecifiesthenumberoftimesthat aneventoccursinasequenceof n independenttrials.Thenumberof correctanswers(correcthits+correctrejection)over90trials(foreach load)wasenteredasdependentvariable.Group,timeandloadwere enteredasfixedfactors;thegrouprepresentedthebetween-subject fac-tor,whiletimeandloadwerewithin-subjectfactors.Thesubjects fac-torrepresentedtherandomfactor.Arandominterceptmodelwas pre-ferredovertherandomcoefficientmodelgiventhelownumberof mea-surementsforeachindividual(twotime-pointsonly)(Wright,2017).A fullfactorialmodelwascarriedouttotestformaineffects,two-ways andthree-waysinteractionsbetweenthefactors:group,timeandload. Theresidualmethodwasusedforthedenominatordegreesoffreedom estimation,andtherobustestimationwaschosen forcoefficient esti-mates.Thecriticaleffectstotestourhypothesiswerethegroup∗time
andgroup∗time∗loadinteractionterms.Thethresholdforsignificance
wassetatp <.05.AGLMM,withthesamefactorsandalinear distribu-tion,wascarriedouttoanalysereactiontime(TableS4,Supplementary materials).Thiswasanexploratoryanalysis,notreportedinthe prereg-istrationofthecurrentstudy.
Effect sized were calculated on the statistically significant pairwise comparisons using Cohen’s d for paired observations (d =(M1−M2)/
√ (S2
1+S22)/2)),where M isthemeanand S isthe
stan-darddeviation.Effectsizesareregardedas(FunderandOzer,2019): verysmall(0.05),verysmallfortheexplanationofsingleeventsbut po-tentiallyconsequentialinthenot-very-longrun;small(0.1)atthelevel ofsingleeventsbutpotentiallymoreultimatelyconsequential;medium (0.2)withsomeexplanatoryandpracticaluseevenintheshortrunand thereforeevenmoreimportant;large(0.3)andpotentiallypowerfulin boththeshortandthelongrun.
2.13. H2. GMBF will increase after exercise, compared with resting Mean DVARSwascalculatedfor eachsubject,andwas testedfor maineffectsofgroup,session,andgroup∗session,toensurethat
move-mentdidnotdifferbetweengroupsandsessions.MeanGMBFwas au-tomaticallycalculatedaspartoftheoutputoftheoxford_aslandwas importedinSPSSforfurtheranalyses.MeanGMBFwastestedfor nor-malityofthedistributionwithShapiroWilk’stest.Thedatawerenot normallydistributed(p =.008),andthedistributionwasright-skewed. AGLMMwithagammadistributionwithloglinkfunctionwasused fortheanalysis.Groupandsessionwereenteredasfixedfactors;the grouprepresentedthebetween-subjectfactor,whilesessionwassetas within-subject factor. Subjectidentity was enteredas randomfactor (randominterceptmodel).Maineffectsoftheinterventionandsession, andtheeffectofthegroup∗sessioninteractionweretested.The
inter-actioneffectwascriticaltotestourhypothesis,asweexpectedGMBF toincreasemoreintheexercisegroupthanintherestinggroup.The residualmethodwasusedforthedenominatordegreesoffreedom esti-mation,andtherobustestimationwaschosenforcoefficientestimates. Thethresholdforsignificancewassetat p <.05.Effectsizeswere cal-culatedasdescribedabove.
2.14. RQ1. What is the regional distribution of exercise-induced GMBF changes?
Voxel-wise analyses on ASL data were carried out in SPM 12. A flexible factorial design was used totest for the main effects of group(between-groupfactors)andsession(within-groupfactor),and group∗sessioninteractiononGMBF(criticaleffect).Apreliminary
un-correctedthresholdof p <.001wasapplied(Wooetal.,2014).Voxels survivingsuchthresholdwerefurthercorrectedforfamily-wiseerror (FWE)rateatclusterlevelwithathresholdofp < .05(Wooetal.,2014). Theanalysiswasrestrictedtogreymatter.
2.15. H3. Changes in mean GMBF will correlate with working memory performance changes more strongly in the exercise group than in the resting group
Accordingtoour hypotheses,changes in then-back performance would correlatewithGMBF changes.N-back accuracywasaveraged acrossblocks,andthedifferencefrompre-topost-testwascalculated. Thesubjectswhowereextremeoutliersononeormoreofthe measure-mentswereexcluded.Additionally,twoextremeoutliersontheaverage accuracy,bothbelongingtotherestinggroup,werealsoexcluded.The changeinGMBFwasalsocomputed.Theanalyseswererunintwosteps. Asafirststep,andaccordingtothepre-registration(osf.io/ve2zj),we testedforassociationsbetweenn-backperformancechangeandGMBF change,andforangroup∗GMBFchangeinteractioneffect.Weuseda
generalizedlinearmodel,withthechangeinn-backperformancesetas dependentvariableandthechangesinGMBFandgroupsetasfactors.In
thismodel,theinteractioneffecttestsforbetween-groupsdifferencesin theassociationbetweentwovariables(n-backperformanceandGMBF) andrepresentedthecriticaleffectfortestingourhypothesis,aswe ex-pectedtheassociationtobestrongerintheexercisethanintheresting group.
Asasecondstep,weperformedGLMManalysestofurtherexplore whethertheassociationbetweenthevariableschangeddifferentlyover timeinthetwogroups.Tothispurpose,andsimilarlytotheprevious analysis,weusedageneralizedlinearmodelwherethenumberof cor-rectanswerswasusedastargetvariable,andthenumberoftrialswas enteredasdenominator.Afullfactorialmodelincludingmaineffects, two-waysandthree-waysinteractiontermsforthefactorsgroup, ses-sionandGMBFwascarriedout.Thethresholdforsignificancewasset at p < .05.
2.16. RQ2. Is the correlation between GMBF and working-memory performance regionally localized?
Correlationsbetweenmeann-backperformanceandGMBFwere as-sessedseparatelyineachgroupateachsessionwithlinearregression modelsinSPM12.Apreliminaryuncorrectedthresholdofp <.001was applied(Wooetal.,2014).Voxelssurvivingsuchthresholdwere fur-thercorrectedforfamily-wiseerror(FWE)rateatclusterlevelwitha thresholdofp <.05(Wooetal.,2014).Voxel-wiseGMBFwassetas de-pendentvariable;meann-backperformancewassetassoleindependent variable.
2.17. H4. Changes in pulse pressure will correlate with GMBF changes, more strongly in the exercise group than in the resting group
WehadalsohypothesizedthatGMBFchangeswouldbeassociated withpre-topost-testchangesinpulse.Theanalyseswererunfollowing thesamestepsasdetailedabove.Wefirsttestedtheassociationbetween GMBFchange(normallydistributed)andpulsechangeandgroup∗pulse
change(criticaleffect).Pulsepressurechangewascalculatedasthe dif-ferencebetweenthepulsepressuremeasuredatbaseline,beforethestart oftheexperimentalsession(firstmeasurement),andafterthe exercis-ingorrestingsession,beforethesecondMRIacquisition(third measure-ment).Asasecondstep,weperformedGLMManalysestofurthertest forassociationsbetweenGMBFandpulsebyusingafullfactorialmodel includingmaineffects,two-waysandthree-waysinteractiontermsfor thefactors:group,sessionandpulse.Thethresholdforsignificancewas setat p <.05.
2.18. RQ3. Is VO2 max associated with mean GMBF?
Tobetterelucidatethemechanismsunderlyingourfindingsrelated toexercise-induced effectsonGMBF(seebelow),wedecidedtoalso exploretherelationshipbetweenGMBFandcardiovascularfitness,as measuredbytheVO2max(hypothesisnotpreregistered).Tothis pur-pose,andsimilarlyastowhatwasdetailedabove,wefirsttestedthe associationbetweenGMBFchangeandVO2maxandgroup∗VO2max
(criticaleffect).WethenusedGLMManalysestofurthertestfor associ-ationsbetweenGMBFandsession∗group∗VO2max.Thethresholdfor
significancewassetat p <.05.
3. Results
3.1. Sample descriptives
Theexerciseandrestinggroupswereequivalentregardingage,sex, physicalactivityreports(IPAQ),MMSEscore,bloodpressureand sub-jectivememoryevaluation(Table1).Thesubgroupwhosen-backdata wereavailablewascomparabletothetotalsample,bothbeforeand af-tertheexclusionoftheoutliers(TableS1).Pulsepressurewasmeasured atfourtime-points:(1)beforethefirstMRI;(2)afterthefirstMRIand beforetheexercisingorrestingsession;(3)aftertheexerciseorresting
Table1
Samplecharacterizationatpre-test.
Exercise Resting (mean (SD)) (Mean (SD))
N 24 25
Age (years) 69.6 (2.8) 70.7 (3.1)
VO2 max (ml/min/kg) 31.4 (5.5) 32.3 (5.6)
MMSE 28.9 (1.2) 28.4 (1.0)
IPAQ (MET/min/week) 2731.0 (1442.0) 3268.2 (2027.0) Systolic blood pressure (mmHg) 136.1 (16.4) 140.8 (16.7) Diastolic blood pressure (mmHg) 87.8 (11.1) 86.5 (9.7) Memory evaluation 3.6 (0.7) 3.6 (0.7) Memory vs 20 years old 3.1 (0.9) 3.4 (0.7) Plasticity belief 2.1 (0.9) 2. (1.2)
% %
Sex (females; males) 50.0; 50.0 40.0; 60.0
Marital status (married) 66.7 56
sessionandbeforethesecondMRI;(4)afterthesecondMRI(Fig.2;Fig. S4,Supplementarymaterial).Inbothgroups,pulsepressureshoweda statisticallysignificantincreaseafterthefirstMRI(2),probablydueto stressinrelationtotheexperimentalsession(Trappetal.,2014).After resting(3),pulsepressureintherestinggrouphadreducedandreached thebaselinelevels(1),asexpectedfollowingthephysiological recov-eryofbloodpressureindicesafterstress,aidedbytherestingsession (Kaushiketal.,2006;Chafinetal.,2004;Santaellaetal.,2006).Onthe otherhand,pulsepressurewasstillhigherthanthebaselineinthe exer-cisegroup,asexpectedasresultofthephysicalexercise(Sharmanetal., 2005).Pulsepressurethenincreasedagainintherestinggroupduring thesecondimagingsession(4),whereitreachedthesamelevelsasafter thefirstMRI(2);ontheopposite,pulsepressuredecreasedinthe exer-cisegroup(4)aspartofthepost-exerciserecoveryphase(Sharmanetal., 2005),goingbacktothebaselinelevels(1).
3.2. Pre- to post-test correlations
Pre-topost-testcorrelationsshowedoverallagoodstabilityofthe measurementsovertime.N-backaccuracyatpre-testshoweda statis-ticallysignificantcorrelationwithaccuracyatpost-testinboththe ex-ercise(Spearman’sRho=0.636)andintherestinggroup(Spearman’s Rho=0.901).Astatisticallysignificantpre-topost-testcorrelationon GMBFwasalsopresentinbothgroups(Spearman’sRho=0.710inthe exercisegroupandSpearman’sRho=0.869intherestinggroup).Mean DVARSalsoshowedastatisticallysignificantcorrelationfrompre-to post-test(Spearman’sRho=0.734intheexercisegroupandSpearman’s Rho=0.475intherestinggroup).Pre-topost-testcorrelationsare rep-resentedinFigs.S1–S3,inSupplementarymaterials.
3.3. H1. Working memory performance will improve more after exercise than after resting
Intotal,234data-pointswereincludedintheanalysis.Statistically significantmaineffectsofsessionandloadwerefoundonn-back per-formance,whilethemaineffectofgroupwasnotstatisticallysignificant (Table2;Figs.2and3).Astatisticallysignificantimprovementin per-formancewasobserved frompre-testtopost-test,though theoverall differencein accuracywasrather small(1.4%gaininaccuracyfrom pre-topost-test;Cohen’s d =0.042).Asexpected,theperformance de-creasedwithincreasingdifficultyloadregardlessofsession. Concern-ingtheinteractionterms,astatisticallysignificantsession∗loadeffect
was detected, indicating thatthe improvementin performancefrom pre- topost-testwas limitedtothe1-back task(Cohen’s d = 0.333; mediumeffect).Nostatisticallysignificanteffectsofthegroup∗session,
intervention∗load,or group∗session∗loadinteractionswere found,
al-thoughatrendforabetterperformanceonthe3-backtaskforthe exer-cisegroupcomparedwiththerestinggroupwasfoundatbothpre-and post-test.However,theconfidenceintervals(C.I.)relativetothe
post-Fig.2. Grouplevelpre-topost-testchangesonn-backperformanceatdifferentloads.Theboxplotsrepresentthen-backperformanceatdifferentcognitiveloads (1-back,2-back,3-back)intheexercise(blue)andresting(red)groups,before(pre-test)andafter(post-test)theexercisingorrestingsession.Thetaskaccuracy decreasedwithincreasingload(p<.001)andwasoverallhigheratpost-testcomparedwithpre-test(p=.012;1.4%increaseinaccuracy).Theeffectofthe group∗session∗loadwasnotstatisticallysignificant(p=.647).Circlesrepresentthemildoutliers,definedasvalueslyingbetween1.5and3interquartilesrange (IQR).
Table2
Effectsofgroup,sessionandloadonn-backperformance.
C.E. S.E. t Adj. sig. 95% C.I. (min, max) GROUP, F (1222) = 0.349; p = .555 Ex vs Rel 0.009 0.015 0.583 0.561 − 0.020, 0.038 † SESSION, F (1222) = 5.560; p = .011 Post-test vs Pre-test † 0.014 0.005 2.546 0.012 0.003, 0.024 † LOAD, F (2222) = 114.1; p < .001 1- vs 2-back † 0.172 0.011 15.896 < 0.001 0.151, 0.193 1- vs 3-back † 0.225 0.008 26.491 < 0.001 0.208, 0.241 2- vs 3-back † 0.053 0.009 5.608 < 0.001 0.034, 0.071 GROUP∗ SESSION, F (1222) = 0.105; p = .746 Pre-test, Ex vs Res 0.007 0.016 0.448 0.655 − 0.025, 0.040 Post-test, Ex vs Res 0.010 0.015 0.645 0.520 − 0.020, 0.039 GROUP∗ LOAD, F (2222) = 0.436; p = .647 1-back, Ex vs Res − 0.001 0.005 − 0.156 0.876 − 0.011, 0.009 2-back, Ex vs Res 0.022 0.023 0.976 0.330 − 0.023, 0.068 3-back, Ex vs Res † 0.044 0.017 2.543 0.012 − 0.079, − 0.010 † SESSION ∗ LOAD, F (2222) = 3.361; p = .036
1-back, Post- vs Pre-test † 0.006 0.002 2.549 0.011 0.001, 0.010 2-back, Post- vs Pre-test − 0.001 0.007 − 0.123 0.903 − 0.015, 0.013 3-back, Post- vs Pre-test 0.009 0.007 1.195 0.233 − 0.006, 0.023 GROUP∗ LOAD ∗ SESSION, F
(2222) = 0.766; p = .466
Pre-test, 1-back, Ex vs Res − 0.002 0.006 − 0.382 0.703 − 0.015, 0.010 Pre-test, 2-back, Ex vs Res 0.021 0.023 0.909 0.364 − 0.024, 0.066 Pre-test, 3-back, Ex vs Res † 0.052 0.019 2.727 0.007 0.014, 0.089 Post-test, 1-back, Ex vs Res < 0.001 0.005 0.051 0.959 − 0.009, 0.010 Post-test, 2-back, Ex vs Res 0.024 0.025 0.960 0.338 − 0.025, 0.073 Post-test, 3-back, Ex vs Res † 0.037 0.019 1.996 0.047 0.0005, 0.074 C.E.,contrastestimate;S.E.,standarderror;df,degreesoffreedom;C.I.,confidenceintervals; †,statisticallysignificant;Ex,exercise;Res,resting.
testcomparisonswereveryclosetoencompassingzero(95%C.I.(min, max)=0.0005,0.074),suggestingthatotherfactorsmighthave influ-encedthebaselinemeasurements,followedbyaregressiontothemean atpost-test.Whenexcludingmildoutliersfromtheanalysis,theresults weremostlyunchanged(TableS2),thoughthetrendforadifferenceon 3-backperformanceatpost-testwasnolongerobservable.Overall,our hypothesisthatn-backperformancewouldimprovemoreafterexercise thanafterrestingwasnotsupportedbyourfindings.
3.4. H2. GMBF will increase after exercise, compared with resting Aftertheremovalofmotionoutliers,eachgroupconsistedof23 in-dividuals,fora totalof92 data-points.Nostatisticallysignificant ef-fectsofgroup,sessionorgroup∗sessionwerefoundonmovement(mean
DVARS).Astatisticallysignificanteffectofthegroup∗session
interac-tiononGMBFwasfound(Table3,Fig.4).Inparticular,reducedGMBF wasfoundintheexercisegroupatpost-testcomparedtopre-test (Co-hen’sd =0.302;mediumeffect),whilenostatisticallysignificantpre-to post-testdifferenceswereobservedintherestinggroup.Moreover,no between-groupsdifferencesweredetectedeitheratpre-testorpost-test. Noeffectsofgrouporsessionwerefound.Thesefindingswere conflict-ingwithourinitialhypothesisthatGMBFwouldincreaseafterexercise, comparedwithresting.
3.5. RQ1. What is the regional localization of GMBF changes?
At theexploratory,voxel-wiseanalysis,no statisticallysignificant effectsofgroup,grouporofthegroup∗sessioninteractionwerefound.
Fig.3. Subject-levelpre-topost-testchangesonn-backperformanceatdifferentloads.Thefigurerepresentsthechangesonn-backperformanceatdifferentcognitive loads(1-back,2-back,3-back)intheexercise(upperpanel)andresting(lowerpanel)groups,before(pre-test)andafter(post-test)thegroup.Thegroup∗load∗session interactionwasnotstatisticallysignificant(p=.466).
Table3
EffectsofgroupandsessiononGMBF.
C.E. S.E. t Adj. sig. C.I. 95% (min, max) GROUP, F (1,88) = 0.009; p = .924 Exercise vs Resting 0.157 01.647 0.096 0.924 − 3.116, 3.341 SESSION, F (1,88) = 1.687; p = .197 Post-test vs Pre-test − 0.71 0.543 − 1.308 0.194 − 1.788, 0.368 † GROUP ∗ SESSION, F (1,88) = 6.574; p = .012
Exercise, post- vs pre-test † − 2.122 0.784 − 2.706 0.008 − 3.681, − 0.564 Resting, post- vs pre-test 0.688 0.734 0.937 0.351 − 0.771, 2.146 C.E.,contrastestimate;S.E.,standarderror;df,degreesoffreedom;C.I.,confidenceintervals; †,statisticallysignificant.
Fig.4. Effectofthegroup∗sessioninteractionongreymatterbloodflow.TheboxplotsrepresentGMBFintheexercise(left)andresting(right)groups,before (pre-test)andafter(post-test)theexerciseorrestingsessions.GMBFhadastatisticallysignificantdecreaseatpost-testcomparedwithpre-testintheexercisegroup (p=.008),butnotintherestinggroup(p=.351).Circlesrepresentthemildoutliers,definedasvalueslyingbetween1.5and3interquartilesrange(IQR).
Table4
AssociationbetweenGMBFandn-backperformance.
Coeff. S.E. t Sig. C.I. 95% (min, max)
† Group ∗ GMBF, F(1,60) = 16.674; p ≤ .001
Exercise vs Resting † − 0.043 0.0099 − 4.284 < 0.001 − 0.062, − 0.023
† Session ∗ GMBF, F(1,60) = 4.229; p = .044
Post-test vs pre-test † − 0.016 0.0069 − 2.331 0.023 − 0.030, − 0.002 Group∗ session ∗ GMBF, F(1,60) = 2.759; p = .102
Exercise group, post- vs pre-test † − 0.017 0.0068 − 2.439 0.020 − 0.030, − 0.003 Resting group, post-test vs pre-test − 0.001 0.0052 − 0.219 0.829 − 0.012, 0.010 S.E.,standarderror;df,degreesoffreedom;C.I.,confidenceintervals;†,statisticallysignificant.
Table5
AssociationbetweenVO2maxandGMBF.
Coeff. S.E. t Sig. C.I. 95% (min, max)
† Group ∗ VO2max, F(1,84) = 4.024; p = .048
Exercise vs Resting † − 0.027 0.0136 − 1.993 0.049 − 0.054, − 0.001
† Session ∗ VO2max, F(1,84) = 4.079; p = .047
Post-test vs pre-test − 0.010 0.0074 − 1.331 0.187 − 0.024, 0.005 Group∗ session ∗ VO2max, F(1,84) = 2.595; p = .111
Exercise group, post- vs pre-test † − 0.022 0.0098 − 2.248 0.030 − 0.042, − 0.002 Resting group, post-test vs pre-test − 0.003 0.0073 − 0.350 0.728 − 0.017, 0.012 S.E.,standarderror;df,degreesoffreedom;C.I.,confidenceintervals.
Theclustershowingastrongerinteractioneffectwaslocatedintheleft inferiorparietalgyrus(IPL),extendingtothemedialprefrontalcortex (pFWE-corr=0.393,clusterextent:179voxels,MNIcoordinates:−60, −56,12).
3.6. H3. Changes in mean GMBF will correlate with working memory performance changes more strongly in the exercise group than in the resting group
Thepre- topost-testchangeon average n-backperformance was notcorrelatedwiththepre-topost-testdifferenceonGMBF,norwas aneffectofthegroup∗GMBFdifferencefoundstatisticallysignificant.
In the GLMM, a statistically significant group∗GMBF interaction
ef-fect wasfound, indicatinganoveralldifferencein theregression ef-fectbetweengroups(Table4).Inparticular,GMBF wasinversely as-sociated with n-back performance in the exercise group, and posi-tivelyassociatedin therestinggroup (Fig.5a).A statistically signif-icanteffect of thesession∗GMBF was alsofound,while theeffect of
thegroup∗session∗GMBF interactionwas not statistically significant;
nonetheless,inthepost-hoctests,astatisticallysignificantchangein theassociationbetween GMBF andn-back performancefrompre-to post-testwasdetectedin theexercisegroup (Fig.5b)butnot inthe restinggroup(Fig.5c).Thus,thebetween-groupdifferenceobservedin thepreviousgroup∗GMBFinteractionseemstobemostlydrivenbythe
exercise-inducedinverseassociationbetweenGMBFandperformanceat post-test.OurhypothesisthatchangesinmeanGMBFwouldcorrelate withworkingmemoryperformancemorestronglyintheexercisegroup thanin therestinggroupwerethuspartly supportedbythedata, al-thoughweexpectedtheassociationtobepositiveratherthannegative. 3.7. RQ2. Is the correlation between GMBF and working-memory performance specifically localized?
Astatistically significantnegativecorrelation wasfoundbetween the meann-back performanceand post-testGMBF in theright hip-pocampus(clusterextent=1488voxels;pFWE-corr=0.001; F =62.3; MNIcoordinates(x,y,z)=32,−38,−6)andleftmedialfrontalcortex (clusterextent=733voxels;pFWE-corr=0.017; F =35.3;MNI coordi-nates(x,y,z)=−8,26,−14),largelyextendingtothecontralateral or-bitofrontalcortexintheexercisegrouponly(Table5;Fig.5d).
3.8. H4. Changes in pulse pressure will correlate with GMBF changes, more strongly in the exercise group than in the resting group
Noeffectsofeitherpulsedifferenceorgroup∗pulsepressure
interac-tiononGMBFpre-topost-testchangewerefound.Atthemixedmodel analysis,nostatisticallysignificanteffectsofpulsepressure∗group,pulse
pressure∗session nor pulse pressure∗group∗session interactions were
foundonGMBF.Thedatadidnotsupportourhypothesisthatchanges inpulsepressurewouldcorrelatewithGMBFchangesmorestronglyin theexercisegroupcomparedwiththerestinggroup.
3.9. RQ3. Is VO2 max associated with mean GMBF?
NoeffectsofeitherVO2maxorgroup∗VO2maxinteractionwere
foundonthepre-topost-testGMBFchange.AttheGLMM,a statisti-callysignificanteffectofthegroup∗VO2maxinteractiononGMBFwas
found;inparticular,astrongerinverseassociationbetweenGMBFand VO2maxwasobservedintheexercisegroupcomparedwiththe rest-inggroup(Table5).Astatisticallysignificanteffectofthesession∗VO2
max wasalsodetected;however,post-hoctestswerenotstatistically significant,likelyduetothesmalldifferenceintheassociationbetween pre-andpost-test.Ontheotherhand,despitetheinteractionbetween group∗session∗VO2max was not statistically significant, the post-hoc
testdidnonethelessshowastatisticallysignificantdifferencebetween groupsinthepre-topost-testregressioneffect(Table5).Inparticular, theregressioncoefficientchangedfrompre-topost-testintheexercise group,whileitremainedstableintherestinggroup(Fig.6).
4. Discussion
Weassessedtheeffectsofacutemoderateintensityphysicalexercise onWMperformanceandGMBFin23olderindividuals,comparedwith 23subjectswhorelaxedbetweenthemeasurements.Confirmatory anal-yseswerecarriedouttoconfirmourhypothesesconcerning exercise-inducedchangesonn-backperformanceandGMBF.Overall,cognitive performancedecreasedwithincreasingdifficultyloadregardlessof ses-sionandgroup,andslightlyincreasedfrompre-testtopost-test, inde-pendentofgroup.GMBFdecreasedfrompre-topost-testinthe exer-cisegroup,butnotintherestinggroup.Correlationanalysesbetween GMBFandaveragen-backperformance,andbetweenpulsepressureand GMBF,consistedofaconfirmatoryparttestingforassociationsbetween
Fig. 5. Association between GMBF and n-backperformance. Thefigure representsthe between-groupsassociations between n-back performanceandthemean(a,b,c,d)or voxel-wise(e)GMBF.Theupperpanel(a)showsthe differentassociationofaveragen-back perfor-mancewithmeanGMBFbetweentheexercise group (coefficient:−0.023, p=.008) andthe restinggroup(coefficient:0.017,p=.003). Co-efficientswerecalculatedseparatelyforeach group using a GLMM. Panel (b) shows the changeintheassociationbetweenaverage n-backperformancewithmeanGMBFfrom pre-(B=0.002, p=.178) topost-test (B=−0.001,
p=243),calculatedseparatelyineachgroup andsessionusinglinearregression.The mid-dlepanel(c,d)showsthechangeinthe as-sociationfrompre-topost-testin(c)the ex-ercisegroup(B<−0.001,p=.989atpre-test;
B=−0.004,p=.013atpost-test)andin(d)the restinggroup(B=0.004,p=0.105atpre-test;
B=0.002,p=.251atpost-test),respectively, calculatedseparatelyineachgroupandsession using linearregression. The lowerpanel (e) showsvoxel-wisecorrelationsbetweenaverage n-backperformanceandGMBFatpost-testin theexercisegroup,intherighthippocampus andleftmedialfrontalgyruslargelyextending totherightorbitofrontalcortex.
pre-topost-testchangesonthevariables,andofanexploratory(not preregistered)parttestingtheimpactofthegroupontheassociation betweenthevariables.Theconfirmatoryanalysesshowedno statisti-callysignificant associationsbetween the pre-topost-testdifference inn-backperformanceandthedifferenceinGMBF.However,the ex-ploratoryanalysesshowedthatintheexercisegroup,lowerGMBFwas associatedwithhighern-backperformance,whileapositiveassociation wasfoundintherestinggroup.Thispatternwaslikelydrivenbythe inverseassociationbetweenGMBFandn-backscoresemerginginthe exercisegroupatpost-test,thatwasnotobservedintherestinggroup, wherenochangeintheassociationstrengthwasobserved.Regional as-sociationswerefoundintheexercisegroupbetweenlowerGMBFinthe righthippocampus,leftmedialfrontalcortexandrightorbitofrontal
cor-texandhigheraveragen-backperformance.Anadditionalexploratory analysisrevealedanegativeassociationbetweenVO2maxandGMBF thatwasstrongerintheexercisegroupatpost-testcomparedwiththe restinggroup.
4.1. Working memory performance
Contrarytoourhypothesis,exercisedidnotinfluencethelevelof performanceinastatisticallysignificantway,incontrastwithsome pre-viousreportsofimmediatepositiveeffectsofphysicalexerciseon exec-utivefunctionsandWM(VolkersandScherder,2014;Langloisetal., 2013;Alvesetal.,2012).Ourstatisticalpowerishoweverlowandthe uncertaintiesaroundtheestimatesarehigh. Wearethushesitant to
Fig.6. AssociationbetweenVO2maxandGMBF.Thefigureshowsthepre-topost-testchangeintheassociationbetweenVO2maxandGMBF(a)intheexercise group(B=−0.653,p=.009atpre-test;B=−0.920,p<.001atpost-test)and(b)intherestinggroup(B=−0.256,p=.246atpre-test;B=−0.286,p=.213atpost-test). Coefficientswerecalculatedseparatelyineachgroupandsessionusinglinearregression.
maketoomuchoutthisnullfinding,althoughwenotethatdespitethe largebodyofevidenceinsomereportsinfavorofthepositivecognitive effectsofacuteexercise,theevidenceis stillconsideredinconclusive (Youngetal.,2015;Angevarenetal.,2008;GasquoineandChen,2020). Recentmeta-analysesofrandomizedcontrolledtrialshavearguedthat acuteburstsofphysicalactivityhaveminoreffectsonWMperformance atbest(effectsize:0.15;95%CI:−0.33,0.63)(RathoreandLom,2017), andthatthesuggestedbeneficialeffectsofaerobicphysicalactivityare notsupportedbyavailableevidenceincognitivelyhealthyolderadults (Youngetal.,2015).However,thestudiesincludedinthemeta-analysis hadconsiderableheterogeneityintermsofinstrumentsusedtoassess WMperformance, unequal samplesizes, andage of theparticipants (RathoreandLom,2017).Age,inparticular,moderatestheeffectsof atleastchronicphysicalactivityinterventions,withsmallpositive ef-fects(effectsize:0.324;95%CI:0.185,0.463)onlyobservablein indi-vidualsover65years(RathoreandLom,2017).However,olderadults areunderrepresentedininvestigationsofimmediateeffectsofphysical activityonWMperformance.Thelackofstudiesinvestigatingthe im-mediateeffectsofphysicalactivityinolderadultsmighthavemasked anypotentialeffectofacutephysicalexercise.Anotherpotential expla-nationfornullfindingsisthatthegainsincognitiveperformanceare notuniformacrossindividuals,beingratherinfluencedbytheir base-lineWMperformance(SibleyandBeilock,2007;Yamazakietal.,2018). Inparticular,cross-overdesignstudieshaveindicatedthatindividuals startingwithlowerWMseemtogainthemost intermsofcognitive performanceafteranacuteboutofmoderateintensityphysicalactivity (SibleyandBeilock,2007;Yamazakietal.,2018).Accordingly,several studiesreportingimprovementsincognitivefunctionsafteraphysical trainingprogramhavebeencarriedoutinolderindividualsaffectedby mildormoderatecognitiveimpairment(VolkersandScherder,2014; Parketal.,2019).
Apivotalroleintheinconsistentresultsacrossstudiesmightalsobe playedbythelengthoftheexerciseprotocol(RathoreandLom,2017). Infact,the majorityof studieshave focusedon long-termprograms (RathoreandLom,2017).Whiletheseseemtobemoreefficientthan acutephysicalexercisesessionsinexertingbeneficialeffectson execu-tivefunctionsandcognition,thedatarelativetoacutephysicalexercise arenotsufficientforthisclaim(RathoreandLom,2017).Concerning acuteexercisesession,aerobicmoderateintensityphysicalexercise last-ingatleast20minhasbeensuggestedtoexertthemostprominenteffect onexecutivefunction(Changetal.,2012),particularlyinolderadults (Ludygaetal.,2016).However,themoderatorroleplayedbythe
in-tensityanddurationoftheexercisesession(RathoreandLom,2017), thetimeofdaywhenthetestiscarriedout,aswellasbytheageand fitnesslevelof theparticipants,hastobenoted(Changetal.,2012; Ludygaetal.,2016).Insuchregard,ourstudyprotocolwasdesignedin ordertominimizetheconfoundingeffectsofthesefactors.Our partic-ipantswerematchedforageandcardiovascularfitnessandtherewere nodifferencesinthetimeofday(morningvsafternoon)whenthe par-ticipantsinthetwogroupsweretested.Wecanthereforeatleastrule outthattheseknownconfoundersmighthaveimpactedonthelackof effectofphysicalexerciseoncognitiveperformanceinourstudy. How-ever,wecannotexcludethatthetimingofthepost-testassessmentof cognitivefunctionmighthaveplayedarole.Infact,onestudyhas re-portedthatthebeneficialeffectsoncognitiveperformancestarttobe evidentfewminutesaftertheexercise,reachingtheirpeak11–20min afterandthenstartingtosubside(Changetal.,2012).Whilethe post-testWMtaskstartedaround12minaftertheendofthephysicalexercise (Fig.2)andwasthuswithintheexpectedpeakwindow,thepaucityof evidenceconcerningthetimingofthecognitiveeffectswarrantsfor fur-therinvestigationonthematter.
Worth noting,relaxationexerciseshave alsobeenreportedto ex-ertpositiveeffectsoncognition(Maetal.,2017;Ferreiraetal.,2015; Tangetal.,2007;Gardetal.,2014;Galvinetal.,2006;Siponkoskietal., 2019).Althoughourcontrolgroupdidnotperformanyactiverelaxation exercise(i.e.sustainingattentionalfocusontheirbody,breath,imagery orelse),theywerestillinstructedtorestandexposedtosoothingsounds. Thismayhavecontributedtoimprovetheircognitiveperformance, lim-itingthepowertodetectpotentialbeneficialeffectsoftheexercisebout. 4.2. Grey matter blood flow
GMBF wasdecreased frompre- topost-testin theexercisegroup only.Theliteratureis quiteconsistentinsuggestingthatincreasesin cerebralbloodflow occurduringexercise(QueridoandSheel,2007; Steventonetal.,2020;Kleinloogetal.,2019;Chaddock-Heymanetal., 2016; Jorisetal.,2018),despitesomescatteredconflictingevidence (vanderKleijetal.,2018;QueridoandSheel,2007).Thisdiscrepancy mightbepartlyduetothedifferentprotocolsofphysicalexerciseusedin differentstudies.Theincreaseincerebralbloodflowinfactmirrorsthe intensityofthephysicalexertionupto60%ofmaximaloxygenuptake (OgohandAinslie,2009;QueridoandSheel,2007);athigherphysical exertion,ontheotherhand,aplateauorevenareductionisreached, due tohyperventilation-inducedcerebralvasoconstriction(Ogoh and
Ainslie,2009;QueridoandSheel,2007).Infact,thehypocapniainduced byhyperventilation,coupledtothesympathetically-mediated protec-tivemechanismsopposingexcessiveincreasesincerebralbloodflow, triggerscerebralvasoconstrictionathighphysicalexertion(Queridoand Sheel,2007).Evenatmoderateintensity,alreadyafewsecondsafterthe cessationoftheexercisebout,thecerebralbloodflowstartstodecrease duetopost-exercisehypotension(QueridoandSheel,2007).Inaddition tothepost-exercisedropincardiacoutput,thisleadstoatransient de-creaseinGMBFimmediatelyafterexercise(QueridoandSheel,2007). AccordingtootherpreviousstudiesusingASLtoassessbloodflowina supineposition,thistransientdecreaseinGMBFisstillobservableup totenminutesafterthecompletionofasingleboutofaerobicphysical exerciseinyoungadults,resumingbaselinelevelsfortyminutes post-exercise(MacIntoshetal.,2014).Infact,thedecreaseincerebral per-fusionandtheresulting hypoxia,aswellasthenormalizationofthe cardiacoutputafterexercisecessation,inducealaterre-boostinGMBF (QueridoandSheel,2007;MacIntoshetal.,2014).Ourfindings indi-catethatthetransientdecreaseonGMBFshortlyafteranacuteboutof moderateintensityaerobicexerciseismaintainedinolderage.
Itisunlikelythatanysustainedeffectofswitchingfromastanding positiontoalyingpositionmighthaveinfluencedthebloodflow mea-surementsinthescanner,asthetimeallowedbetweenchangingbody postureandGMBFmeasurementsshouldhavebeenenoughforthe au-toregulationofbloodflowtooccur(atleasttwominutesforpreparing thesubjectinthescannerpluslocalizeracquisition).Studiesmeasuring MCA-velocityhaveinfactindicatedthataroundtwominutesmaybe sufficientforautoregulatorymechanisms(Haubrichetal.,2004). 4.3. Associations between task performance, GMBF, and cardiovascular variables
Whilethepre-topost-testdifferenceinGMBFdidnotcorrelatewith theimprovementonthen-backperformance,thegroupsexhibited dif-ferentcorrelationalpatternsbetween GMBFandaveragen-back per-formanceovertime.Consistentlywithpreviousliterature,intheresting grouphigherGMBFwasassociatedwithbettern-backperformance. Pre-viousstudiesperformedinolderadultsusingASL,infact,havereported cross-sectionalpositiveassociationsbetweenglobalcerebralbloodflow andexecutivefunctions(Leeuwisetal.,2018),andGMBFhasbeen re-portedtoalsopredictlongitudinalchangesonfluidintelligence(DeVis etal.,2018).Ontheotherhand,lowerGMBFwasassociatedwithhigher n-backperformanceintheexercisegroup.Thisfindingmaybeascribed tobe theinverseassociation betweenGMBFandn-backemergingin theexercisegroupatpost-test.LowerGMBFwasthusassociatedwith highertaskaccuracyafterexercise,contrarytoourexpectations.
MostofthepreviousresearchcarriedoutwithASLreporting pos-itiveassociations between cerebral bloodflow andcognitive perfor-mancewasbasedonlongerphysicaltraininginterventions,lasting sev-eral weeks (Chapman et al., 2013). Whena single boutof physical activityisperformed,therelationshipbetweencognitiveperformance andcerebralperfusionseemstoprogressivelyuncoupleduringexercise (Lucasetal.,2012).Indeed,theexercise-inducedtemporaldynamics ofcerebralperfusionchangesmighthelpuselucidatetheobserved in-versedrelationshipbetweenn-backperformanceandGMBFafter exer-cise.Asmentionedabove,we measuredGMBFwithin thetimeframe forthetransientpost-exercisereductionincerebralbloodflow.Ifwe assumethatlowerGMBF inthis contextisindicativeof higher cere-brovascularreactivity,assupported bythepresence of the intercur-rent negativecorrelationbetween GMBFandVO2 max foundatthe exploratoryanalysis,itsassociationwithtaskperformanceisthenless surprising.In fact,higher cardiovascularfitness wasassociated with lowerGMBF,andtheassociationwasstrongerafterexercisecompared withresting.NegativeassociationsbetweenVO2maxandGMBFhave beenreportedbeforeinyoungadults(Fosteretal.,2020;Furbyetal., 2019),andpossiblyinhealthyolderadults(Intzandtetal.,2019).The strengtheningoftheassociationafterexerciseisconsistentwiththe
no-tionthatawell-functioningcerebrovascularreactivity,asassessedwith thecombineduseofDopplerandPET(CO2),isprotectiveagainst cere-brovasculardiseasesassociatedwithageing(Murrelletal.,2013). Cere-bralvasodilatorresponsestohypercapnia,asmeasuredbytranscranial Doppler,arealsoassociatedwithhighercardiovascularfitnessandVO2 max inhealthyolderadults(Barnesetal.,2013).Inthis framework, wemightexpectindividualswithlowerGMBFafterexercisetohave experiencedahigherboost inflowfollowinginitial vasoconstriction. Ifwe hadmeasuredGMBF afewminutes later,we mighthavethen beenabletoobserveareversedassociationofGMBFwithtask perfor-mance,asreportedinpreviousliterature.Supportingthishypothesis, wehavealsofoundbetterpost-testn-backperformancetobe unexpect-edlyassociatedwithlower,ratherthanhigher,post-testGMBFinthe exercisegroupinthehippocampalandorbitofrontalregions,both well-knowntobepositively relatedwithbetterWM(Barbeyetal.,2011; Loprinzietal.,2019;Nissimetal.,2016) andexercise-relatedeffects on cognitiveperformance (Basso andSuzuki,2017; Chapman etal., 2016;DenOudenetal.,2018;Firthetal.,2018;Ericksonetal.,2011; Brockettetal.,2015).Nonetheless,itmustbenotedthatwedidnothave adirectmeasureofcerebrovascularreactivity,andonlyearly measure-mentsonGMBFwereavailable.Thoughbasedonwell-established liter-atureonexercise-relatedcerebrovascularreactivity(Lucasetal.,2012; Fosteretal.,2020;Furbyetal.,2019;Intzandtetal.,2019;Murrelletal., 2013;Barnesetal.,2013),theproposedexplanationisthustobe con-sideredspeculative.
Wehadhypothesizedthatpulsepressurewouldalsobepositively relatedtotheGMBF.Infact,whilepersistentlyelevatedpulsepressure hasbeen relatedwitharterialstiffness (Safaretal., 2011) and cere-brovascular events(Thorin-Trescases etal., 2018),when maintained within thephysiological rangepulsepressure contributestoan ade-quatecerebrovascularperfusion.Therelationshipbetweenblood pres-sureandbrainhealthisindeedquitecomplexinoldage(Shangetal., 2016;Forteetal.,2019;McDadeetal.,2016),whenlowsystolicblood pressurecanbeasdetrimentalforthebrain(Foster-Dingleyetal.,2015; Mulleretal.,2010)andcognition(Shangetal.,2016;Forteetal.,2019; McDadeetal.,2016;Mosselloetal.,2015)ashighbloodpressure. In-deed,individualswithAlzheimer’sDiseasehavelowerpulsepressure andcerebralhypoperfusioncomparedwithcontrols(Roheretal.,2012). Moreover, low pulsepressure,particularlywhencombined withlow cerebralbloodflow,hasbeenassociated withcortical(Mulleretal., 2010)andsubcortical(Foster-Dingleyetal.,2015)atrophy.The rela-tionship betweenpulsepressure andGMBF inour samplewas how-eversomewhatstrongerintherestinggroup,ratherthaninthe exer-cisegroupasweexpected.ConsideringtheGMBFreductionobserved intheexercisegrouppost-test,itisnotsurprisingthatsuchrelationship wouldbeweakerinthisgroup.Giventhecomplexityoftherelationship betweenpulsepressureandGMBF,furtherstudieswillneedtofurther elucidatethisassociation.
4.4. Limitations
Somelimitationsinourstudyhavetobeacknowledged.Firstofall, thelimitedsamplesizedidnotallowforastratificationofthe partici-pantsaccordingtotheirbaselineWMcapacity,whichhasbeenindicated tomoderatethegainsincognitiveperformanceinducedbyphysical ex-ercise(SibleyandBeilock,2007;Yamazakietal.,2018).Ourprotocol didnotincludeadelayedmeasurementofGMBF,thuspreventingusto fullydescribethetrajectoryofbloodflowchangesafterexercise. More-over,thesignal-to-noiseratioisinherentlylowforASLacquisitions,as thesignalfromthelabeledinflowingbloodconstitutesonly0.5−1.5% ofthefulltissuesignal.Furthermore,arterialstiffnesswasnotevaluated (Petcharunpaisanetal.,2010).Concerningtheexerciseprotocol, mea-surementsofbloodgasesduringexercisewerenotacquired,potentially allowingfor toohighintensitiesofexercise beingreached. Also, de-hydrationisknowntoaffectcerebralbloodflowmeasurementsduring exercise(Trangmaretal.,2015).Ourparticipantswereallowedtodrink
