Cellwww

UNCORRECTEDPROOFEBioMedicinexxx(2019)xxx-xxxContentslistsavailableatScienceDirectEBioMedicinejournalhomepage:www.
elsevier.
comDonor-derivedcell-freeDNApredictsallograftfailureandmortalityafterlungtransplantationSeanAgbor-Enoha,b,c,YanWangd,IlkerTunca,c,MoonKyooJanga,c,AndrewDavisa,c,IwijnDeVlamincke,HelenLuikartf,PaliD.
Shaha,b,IrinaTimoftea,d,AnneW.
Browna,g,ArgitMarishtaa,c,KennethBhattia,c,SashaGorhama,c,UlgenFidelia,c,JenniferWylief,DavidGrimmf,NatalieGoodwinf,YanqinYanga,c,KapilPatelf,JunZhua,c,AldoIaconoa,d,JonathanB.
Orensa,b,StevenD.
Nathana,g,CharlesMarboea,h,GeraldJ.
Berrya,f,StephenR.
Quakei,KiranKhushf,,HannahA.
Valantinea,c,aGenomicResearchAllianceforTransplantation(GRAfT),10CenterDrive,7S261,Bethesda,MD20982,UnitedStatesbDivisionofPulmonaryandCriticalCareMedicine,TheJohnsHopkinsSchoolofMedicine,1830EastMonumentStreet,Baltimore,MD,UnitedStatescDivisionofIntramuralResearch,NationalHeart,LungandBloodInstitute,10CenterDrive,7S261,Bethesda,MD20982,UnitedStatesdUniversityofMarylandMedicalCenter,Baltimore,MD,UnitedStateseMeinigSchoolofBiomedicalEngineering,CornellUniversity,Ithaca,NY,UnitedStatesfStanfordUniversitySchoolofMedicine,PaloAlto,CA,UnitedStatesgInovaFairfaxHospital,Fairfax,VA,UnitedStateshDepartmentofPathology,NewYorkPresbyterianUniversityHospitalofCornellandColumbia,NY,NewYork,USAiDepartmentofBioengineering,StanfordUniversity,PaloAlto,CA,USAARTICLEINFOArticlehistory:Received18October2018Receivedinrevisedform13December2018Accepted14December2018AvailableonlinexxxABSTRACTBackground:Allograftfailureiscommoninlung-transplantrecipientsandleadstopoorout-comesincludingearlydeath.
Noreliableclinicaltoolsexisttoidentifypatientsathighriskforallograftfailure.
Thisstudytestedtheuseofdonor-derivedcell-freeDNA(%ddcfDNA)asasensitivemarkerofearlygraftinjurytopredictimpendingallograftfailure.
Methods:Thismulticenter,prospectivecohortstudyenrolled106subjectswhounderwentlungtransplantationandmonitoredthemaftertransplantationforthedevelopmentofallograftfail-ure(definedasseverechroniclungallograftdysfunction[CLAD],retransplantation,and/ordeathfromrespiratoryfailure).
Plasmasampleswerecollectedseriallyinthefirstthreemonthsfollowingtransplantationandassayedfor%ddcfDNAbyshotgunsequencing.
WecomputedtheaveragelevelsofddcfDNAoverthreemonthsforeachpatient(avddDNA)anddetermineditsrelationshiptoallograftfailureusingCox-regressionanalysis.
Findings:avddDNAwashighlyvariableamongsubjects:medianvalueswere3·6%,1·6%and0·7%fortheupper,middle,andlowtertiles,respectively(range0·1%–9·9%).
Comparedtosubjectsinthelowandmiddletertiles,thosewithavddDNAintheuppertertilehada6·6-foldhigherriskofdevelopingallograftfailure(95%confidenceinterval1·6–19·9,p=0·007),lowerpeakFEV1values,andmorefrequent%ddcfDNAelevationsthatwerenotclinicallyde-tectable.
Interpretation:Lungtransplantpatientswithearlyunresolvingallograftinjurymeasuredvia%ddcfDNAareatriskofsubsequentallograftinjury,whichisoftenclinicallysilent,andpro-gressestoallograftfailure.
Correspondenceto:H.
A.
ValantineLaboratoryofOrganTransplantGenomics,GenomicResearchAllianceforTransplantation(GRAfT),NationalInstitutesofHealth,50CenterDr.
Rm4249,Bethesda,MD20892,USA.
Correspondenceto:K.
KhushDivisionofCardiovascularMedicine,StanfordUniversitySchoolofMedicine,300PasteurDrive,Stanford,CA94305-5406,USA.
Emailaddress:hannah.
valantine@nih.
gov,hvalantine@Stanford.
edu(H.
A.
Valantine)https://doi.
org/10.
1016/j.
ebiom.
2018.
12.
0292352-3964/2018.
UNCORRECTEDPROOF2EBioMedicinexxx(2019)xxx-xxxFund:NationalInstitutesofHealth.
2018.
ResearchincontextEvidencebeforethisstudyLungtransplantpatientshavetheshortestsurvivalofanyothersolidorgantransplantationprimarilyduetoahighinci-denceofchronicrejection(alsocalledchroniclungallograftdysfunction–CLAD).
Severaltherapieshavebeenattemptedbutaregenerallyineffective.
TheclinicalcourseofCLADisthereforeprogressivewithirreversibleallograftinjurythatultimatelyleadstoallograftfailure.
PerhapsinterventionsatearlierstagesbeforeallograftinjurybecomesirreversiblemaydelayorevenpreventthedevelopmentofCLADandimprovelungtransplantoutcomes.
To-date,noreliableclinicalpredic-tivebiomarkerexists.
Severalbiomarkershavebeenproposedbutremainlimitedforoneormoreofthefollowingreasons:requirementofinvasiveproceduressuchasbronchoscopytoobtainsamples,poorspecificityorsensitivity,and/ordetectionCLADwithsignificanttime-lagtotheirreversibleclinicalmanifestations.
Thisstudyproposesanon-invasivebloodtestasapotentialpredictivebiomarker.
OuroverarchinghypothesisisthatallograftinjuryearlyaftertransplantationpredictiveofCLADandotherpooroutcomes.
Twoclinicalobservationssupportthishypothesis.
First,earlypost-transplantcompli-cationslikeprimarygraftdysfunctionshowastrongrelationshipwithCLADsuggestingthatallograftinjuryearlyaftertransplantationisaprecursorofCLAD.
Second,lungtransplantpatientsundergorigorousmonitoringwithbronchoscopiesalongwithtransbronchialbiopsy,spirometryandothertestingtodetectandtreatacutecomplicationswiththegoalofpre-ventingCLAD.
Yet,CLADstilloccursanalarminglyhighrateleadingustosuspecttheexistenceofallograftinjurythatisundetectableclinicallyandbymonitoringtools.
Testingthishypothesisrequirequantificationofallograftinjuryearlyaftertransplantation.
Unfortunately,thelimitationsofavailableclinicaltoolsmakethemunreliabletoquantitateearlyallograftinjury.
Histopathology,thecurrentgoalstandard,issemiquantitativeatbestandlimitedbylowsensitivity,invasivenessandhighvariability.
Spirometry,anothermonitoringtool,islimitedbylowsensitivityowingtolargepulmonaryreserve.
Recently,ourgroupintroducedasensitivegenomicbloodtestthatreliablyquantitatesallograftinjuryfrominfection,acuterejectionandothercomplications.
Thistesttakesadvantageofthewidegenomicdifferencebetweentransplantdonorsandrecipients,aswellasthesensitivityofgenomesequencingtoidentifyandquantifycirculatingdonor-derivedcell-freeDNA–ddcfDNA.
Thetestisbroadlyapplicableacrosstransplantationandhasbeenusedtodetectacuterejection.
Inthisstudy,weleveragethesensitivityof%ddcfDNAtoquantitateallograftinjury(bothclinically-detectedandclinically-silent)intheearlypost-transplantperiodanddetermineitsrelationshiptoallograftfailure(CLADordeath).
AddedvalueofthisstudyWemonitoredlungtransplantpatientsforthedevelopmentofCLADordeath.
Theirseriallycollectedplasmasamplesintheearlypost-transplantperiodwereusedtoquantify%ddcfDNA.
Theaverage%ddcfDNAintheearlypost-transplantperiodof3-months,whichwedesignateasavddDNA,wasvariablebetweenpatients.
LevelsofavddDNAcorrelatedwithearlypost-transplantclinicalriskfactorslikeolderageandprimarygraftdysfunction.
PatientswithhighavddDNA,signi-fyinghighearlypost-transplantallograftinjury,showed6·6-foldhigherhazardofdevelopingCLADordeaththanthosewithloweravddDNAlevels.
Morethanhalfofthesepatientsshowednoclinicalcomplicationovertheearlypost-transplantperiod.
PatientswithhighavddDNAalsoshowedlowerlungfunctionthanthosewithlowavddDNAlevels.
Atamolecularlevel,theycontinuedtoshowhigh%ddcfDNAlevelsbeyondtheearlypost-transplantperiod.
Surprisingly,onlyone-thirdofthehigh%ddcfDNAlevelsweredetectedclinicallyorbymonitoringtools,therestwereclinicallyunrecognized.
ImplicationsofalltheavailableevidenceWeshowthatavddDNAisapotentialpredictivebiomarkerforCLADanddeath.
Ifvalidated,thisbloodtestcouldbeusedtonon-invasivelyriskstratifypatientsforCLADasearlyastheinitialthree-monthsoftransplantation.
Withclinicalversionsofthetestsoonbecomingavailable,suchamarkerwouldthereforeprovideearlytime-pointstointervene,per-hapspreventingirreversibleclinicalmanifestationsandCLADsets-in.
Wealsoshowthatmostinjuryinlungtransplantpatientsgounrecognizedclinicallyorbyexistingmonitoringtoolsandthereforegountreated.
IdentifyingthetriggersofthispreviouslyunrecognizedallograftinjurymayopennewavenuestointerveneandpreventthedevelopmentofCLADandotherpooroutcomes.
Thiswarrantsfurtherinvestigation.
Together,ourresultsindicatethatcumulativeinjuryearlyafUNCORRECTEDPROOFEBioMedicinexxx(2019)xxx-xxx3tertransplantationpredictssubsequentallograftinjuryandisaprecursorfordownstreampooroutcomes.
Thesefindingsarepotentiallyapplicationacrosstransplantationsincetheassayisequallyreliableintheseothertransplantsettings.
1.
IntroductionAllograftfailureoccursinlung-transplantrecipientsathigherratesthaninothersolid-organtransplantrecipients[1–3].
Thisdifferenceisprimarilyduetothefrequentdevelopmentofchroniclungallograftdysfunction(CLAD),whichwhensevere,nearlyalwaysleadstodeath[1].
CLADhasseveralclinicalphenotypes,themostpredominantbeingbronchiolitisobliteranssyndrome(BOS)whichcauseslungobstruction,andrestrictiveallograftsyndrome(RAS)whichrestrictslungexpansionandfunction[1,2,4].
Allograftfailureisthoughttoresultfromaccumulationofinjuryfromvarioussources,includingprimary-graftdysfunc-tion(PGD),antibody-mediatedrejection(AMR),acutecellularrejection(ACR),infections,andotherscauses[1,5–7].
Theseriskfactorsalsopredisposetoallograftfailureinothersolidorganstransplantation.
PGDoccursintheearlypost-transplantperiod,anditsstrongassociationwithallograftfailuresuggestthatearlypost-transplantallograftinjuryisariskfactorforsubsequentdevelopmentofCLADandallograftfailure.
TheriskofallograftfailureisstrongestforpatientswithPGDthatpersistsbeyond72haftertransplantation,makingpersistentPGDapotentialproxyforongoingallograftinjury.
Wethereforehypothesizethatpersistentinjuryintheearlypost-transplantperiodisaprecursorofongoingallograftinjuryandallograftfailure.
Asensitive,anddirectquantitativemethodthatcanaccuratelyquantifyearlyallograftinjurymaythereforepredictallograftfailureandwouldenableearlierandmoreaccurateidentificationofpatientsmostatriskofsevereCLAD,allo-graftfailureandprematuredeath.
Thesepatientsmaybenefitfrommoreclosermonitoring.
Unfortunately,currentclinicaltools,spirometryandhistopathologyontransbronchialbiopsiesaresemi-quantitative,invasive,and/orarelimitedbypoorsensitivity[8]makingthemunreliabletoassessearlypost-transplantallograftinjury.
Advancesingenomesequencingtechnologiesnowofferanopportunitytoovercomethelimitationsofsensitivityassociatedwithcurrentclinicalmethods.
However,toourknowledge,nopriorstudyhasassessedwhetherddcfDNAismarkerofchronic,ongoingallograftinjurythatisultimatelyirreversible.
Inthetransplantationsetting,wheredonorandrecipienthavedifferentgenomes,thesemethodscanbeappliedtodiscriminateandquantitateplasmadonor-derivedcell-freeDNA(ddcfDNA)asamarkerofallograftinjury.
CellfreeDNA(cfDNA)isreleasedbydyingcellsintothebloodstreamandisthereforeadirectmeasureoforganinjury,correlatingwithdiseaseseverityordiseasestateinvariousclinicalscenariosincludingsepsis[9],trauma[10],andcancer[11].
Cell-freeDNAhasashorthalf-lifeof15min[12]andthusprovidesassessmentsoforganinjuryatahightemporalresolution.
Detectionof%ddcfDNAusingunbiasedwhole-genomesequencingishighlyreproducible[13],andthismethodhasbeenshowntoaccuratelyquantifymicrobialnucleicacids,enablingconcurrentmonitoringofinfectiouscomplicationsfollowingtransplantation[14]andchangesinmicrobiomearchitecture.
TheddcfDNAissensitivetodetectacuterejectioninlung[14],heart[15],liver[16]andkidney[17]transplants.
Followinglungtransplantation,%ddcfDNAvalues≥1%havebeenshowntodetectsevereacuteallograftrejectionwith100%sensitivity[14],andthismayrepresentaclinicallyrelevant%ddcfDNAthreshold.
Thishighsensitivityprovidesanopportunitytoassessearlypost-transplantinjuryanddetermineitsrelationshipwithdownstreamallograftinjuryandallograftfailure.
Inthisstudy,weexaminedthefollowingresearchquestions:1.
Whatrecipientanddonorfactorsarerelatedto%ddcfDNA2.
how%ddcfDNAisrelatedtodownstreamallograftinjuryandallograftfailure2.
Methods2.
1.
StudydesignSubjectsawaitinglungtransplantationintwoongoingcohortstudieswereincluded.
Thefirststudy,GenomeTransplantDy-namics(GTD)study(NCT01985412),commencedin2010andisasingle-centerstudyatStanfordUniversityHospital,Cali-fornia,investigatingthetestcharacteristicsof%ddcfDNAtodetectacuterejection.
ThesecondstudyistheGenomicResearchAllianceforTransplantation(GRAfT)study(NCT02423070),whichcommencedrecruitmentin2015atthree-centers(theJohnsHopkinsHospitalandtheUniversityofMarylandMedicalCenter,Maryland,andInovaFairfaxHospital,Virginia).
Inbothstud-ies,subjectswereatleast18yearsofageandenrolledwhileawaitinglungtransplantation.
Subjectsweremonitoredprospectivelyaftertransplantation.
Weexcludedsubjectsforwhomnoplasmasampleswereavailablefor%ddcfDNAassay.
108subjectshadsufficientsamplestocomputeavddDNAandwereincludedtodeterminetherelationshipbetweencovariatesandavddDNA.
Twoofthesesubjectsdiedwithin3monthsandwereexcludedforoutcomeregressionanalyses,leaving106subjects.
ThestudydesignissummarizedinFig.
1.
TheenrollmentdetailsfortheinstitutionsinvolvedinthestudyareincludedinSupplementaryTable1a.
ThestudydesignwasapprovedbytheInstitutionalReviewBoardateachinstitution.
ThepatientmonitoringplanisdetailedintheSupplementaryMethodssection,andtheimmunosuppressionregimenisoutlinedinSupplementaryTable1b.
UNCORRECTEDPROOF4EBioMedicinexxx(2019)xxx-xxxFig.
1.
Studydesign.
Subjectswereexcludediftheydiedwithinthreemonthsoftransplantationordidnotprovideplasmasamplesfor%ddcfDNAassessment.
106subjectswereincluded.
Plasmasamplescollectedwithintheinitialthreemonthsfollowingtransplantationwereassessedfor%ddcfDNAbyshotgunsequenc-ing.
Theaverage%ddcfDNA(avddDNA)wascalculatedasthepredictivemarker,anditsrelationshiptotheprimaryoutcomewasassessed.
Theprimaryoutcomewaslung-allograftfailure(acompositeoutcomeincludingdeathfromrespiratorycauses,retransplantationand/orseverechroniclungallograftdysfunction-CLADdefinedbyISHLTcriteria1).
Thesecondaryoutcomewasall-causemortality.
2.
2.
OutcomemeasuresTheprimaryoutcomeofthestudywasdefinedasthetimefromtransplantationtofirstdetectionofallograftfailure,repre-sentedbyanyofthethreeendpoints:severeCLAD(asdefinedbytheInternationalSocietyforHeartandLungTransplantationcriteria[1]),retransplantation,anddeathfromrespiratorycauses.
Thesethreeendpointsrepresentthemostsevereclinicalcom-plicationsthatarisefromafailinglungallograft.
AdditionalinformationontherationaleforselectingthiscompositeprimaryoutcomesanddefinitionofCLADisprovidedinSupplementaryMethods.
Thesecondaryoutcomesweredefinedasthetimefromtransplantationtoall-causemortalityandCLAD-freesurvival.
Alloutcomeswereadjudicatedbytransplantphysiciansblindedto%ddcfDNAmeasurements.
2.
3.
ClinicalcovariablesandmeasurementsClinicalcovariateswererecordedthroughoutthestudy.
Beforetransplantation,donorandrecipientdemographics,smokinghistory,cytomegalovirusserologystatusandothervariableswererecorded,includingdonorcauseofdeathandhistoryofchesttrauma;andrecipientreasonfortransplantationandlungallocationscore.
Intheimmediatepost-transplantperiod,arterialbloodgasandchestx-raydataonDay3aftertransplantationwererecordedtodefinedPGDstatus.
Useofinductiontherapywasalsodocumented.
Inthepost-transplantationperiod,serialdataobtainedaspartofroutineclinicalcareincludingspirometry,histopathology,donor-specificantibodies,microbiologicaltests,chestradiographsandotherlaboratorytestwererecordedtode-finetheprimaryandsecondaryendpointsaswellasclinicalcomplicationsofACR,AMRorinfections.
Themethodsandsched-ulesforcollectingthesedata,aswellasdefinitionsofthesevariablesisdetailedintheSupplementaryMethods.
UNCORRECTEDPROOFEBioMedicinexxx(2019)xxx-xxx52.
4.
%ddcfDNAmeasurementPlasmasampleswerecollectedserially(SupplementaryFig.
1)and%ddcfDNAwasmeasuredusingapreviouslydescribedshotgun-sequencingmethod[14].
Insummary,genomicDNAisolatedfromdonorandrecipientpre-transplantbloodwasgeno-typed,andthedataforthetwosampleswerecomparedtoidentifysingle-nucleotidepolymorphisms(SNPs).
Followingthetrans-plant,cfDNAwasisolatedfromplasmasamplestogenerateaDNAlibraryforshotgunsequencing.
ThecfDNAsequencereadswerethensurveyedforthepresenceofdonorandrecipientSNPsand%ddcfDNAwascalculatedasthepercentageofdonorSNPstototal(recipientanddonor)SNPs[14,15].
Numberofsamplesanalyzedforeachcohort,aswellassequencingdatacharacteris-ticsarerepresentedinSupplementaryTable2.
2.
5.
Calculatingthree-monthaverage%ddcfDNA(avddDNA)Theearlypost-transplantationdecaykinetics%ddcfDNAwereanalyzedseparatelyfordouble-andsingle-lungtransplantsus-ingpreviouslydescribedmethodthatshowedatwo-steplogarithmicdecay[14].
Sincemedian%ddcfDNAvaluesforsingle-lungtransplantswerehalfthemedianvaluesfordouble-lungtransplants,wemultipliedthe%ddcfDNAbytwotocorrectforsin-gle-lungtransplantation.
Tocomputeaverage%ddcfDNAwithintheinitialthree-monthperiod,weplottedindividual%ddcfDNAversustimeandcalculatedtheareaunderthecurve(AUC)betweendays14and90(inclusive).
Plasmasamplesfromdays0–13wereanalyzed,butthe%ddcfDNAvalueswereexcludedfromourcalculationstolimitanycontaminationbycfDNAreleasedduringtransplantsurgery.
Wethuscalculatedtheaverage%ddcfDNA(avddDNA)bydividingtheAUCof%ddcfDNAby77(torepresentthe13to90-dayperiod).
MoredetailisprovidedintheSupplementaryMethods.
2.
6.
StatisticalanalysisDescriptiveanalyseswereconducted.
Thedistributionofthecontinuousvariablesfordonorandrecipientfactors(e.
g.
recipi-entage,donorage,lungallocationscore(LAS),ischemictime,numberofrejectionepisodes,numberofHLAmismatches)werechecked.
Sinceeachfellwithinthelimitofthecutofffornormality,werepresentthemeanacrossthethreetertilesofavddDNA.
Forcategoricalvariables,werepresentthefrequency(Table1).
Wealsocomparedthemeanofcontinuousvariableswithinde-pendentsamplet-testsandfrequenciesofcategoricalvariableswithChi-squareorFisher'sExacttests(ifanycellsizecome(severePGD)regardlessofthep-values.
Wealsoincludestudycohort(GTDvs.
GRAfT)sincethetwocohortsshoweddifferentratesoftheprimaryandsecondaryoutcomes.
TheunivariateandmultivariateanalysesarepresentedinTable2.
Theprincipalanalysiswasatime-to-eventanalysis,orsurvivalwithavddDNAandotherfactors,aspredictorandallograftfailureasoutcome(Table3).
Survivalanalysiswasrepeatedseparatelyforeachofthetwosecondaryoutcomes–CLAD-freesurvivalandall-causemortality(SupplementaryTable3).
Forallthesubjects,thetimefromtransplantationtolastclinicalfollowupvisitrangesfrom0.
47–63months.
Sincethepredictorwasmeasuredovertheinitial3monthsoftransplantation,subjectswithfollow-upof3monthsorlesswereexcluded.
Theprimaryoutcome,allograftfailure,isdefinedassuchifthesubjecthadatleastoneofthreeconditions:severeCLAD,re-transplantationordeathfromrespiratorycauses.
Thesurvivaltimewascalculatedfromthedateoftransplantationuntiltheearliestdateoutofthethreeconditions(forsubjectswithevents)orlastclinicalfollowup(forcensoredsubjects).
Thesubjectswerecensorediftheydidnothavetheeventbythelastclinicalfollowup.
Therearetwosecondaryoutcomes.
The1stoutcomeisdefinedifthesubjecthadatleastoneofthetwoconditions:anygradeofCLADordeathfromanycause.
Thesurvivaltimewascalculatedfromthedateoftransplantationuntiltheearliestdateoutofthetwoconditionsorthelastclinicalfollow-up.
Notmeetinganyofthetwooutcomes(anygradeCLADordeath)definesCLAD-freesurvivalorbeingaliveandCLAD-free,andwascensoredifsubjectsdidnothavetheeventbythelastclinicalfol-low-up.
The2ndoutcomeisall-causemortalitydefinedasdeathofanycause.
Thesurvivaltimewascalculatedfromthedateoftransplantationuntildeathorlastclinicalfollow-upvisit.
Thesubjectswerecensorediftheydidnothavetheevent(death)bythelastclinicalfollow-up.
Foreachsurvivaloutcome,Kaplan-MeiermethodwasusedtoestimatethesurvivalfunctionbythethreetertilesofavddDNA.
Log-ranktestswereusedtotestthedifferencesinsurvivalbythetertiles.
TheproportionalhazardassumptionwastestedonthebasisofSchoenfeldresiduals.
Theproportionalhazardassumptionwasheldforeachoutcome.
UnivariateCoxregressionmodelwasfirstconductedtoes-timatetheHazardRatio(HR)ofeachoutcomeofinterestinrelationtoavddDNAandothervariables,includingavddDNAasaUNCORRECTEDPROOF6EBioMedicinexxx(2019)xxx-xxxTable1DonorandrecipientcovariatesbyavddDNAtertiles(n=108).
VariablesTotalLowtertileMiddletertileUppertertileRecipientcovariatesAgeMean(SD)54·61(15·11)51·56(14·26)55·06(15·48)57·22(15·44)Sexn(%)Male58(54)18(50)18(50)22(61)Female50(46)18(50)18(50)14(39)Racen(%)Non-Caucasian19(18)8(24)3(9)8(23)Caucasian85(82)26(76)32(91)27(77)ObesityBMI≥30Kg/m2n(%)No90(84)31(89)30(83)29(81)Yes17(16)4(11)6(17)7(19)Smokinghistoryn(%)Never77(71)27(75)26(72)24(67)Past31(29)9(25)10(28)12(33)Transplantationreasonn(%)COPD22(20)5(14)9(25)8(22)CF18(17)9(25)4(11)5(14)ILD52(48)13(36)20(56)19(53)PAH3(3)1(3)1(3)1(3)Sarcoidosis/others13(12)8(22)2(6)3(8)LASmean(SD)47·10(16·71)41·19(10·92)48·69(17·88)51·12(18·90)DonorcovariatesAgemean(SD)36·36(15·09)39·75(15·18)35·67(15·89)33·67(13·90)Sexn(%)Male65(60)22(61)22(61)21(58)Female43(40)14(39)14(39)15(42)Racen(%)Non-Caucasian28(27)12(35)6(17)10(29)Caucasian75(73)22(65)29(83)24(71)ObesityBMI≥30Kg/m2n(%)No78(74)28(78)25(74)25(69)Yes28(26)8(22)9(26)11(31)Smokinghistoryn(%)Never98(95)33(97)33(97)32(91)Past5(5)1(3)1(3)3(9)Chesttrauman(%)No103(98)35(97)34(100)34(97)Yes2(2)1(3)01(3)Causeofdeathn(%)HeadtraumaorGSW44(41)12(33)16(44)16(46)CVA42(39)18(50)12(33)12(34)Anoxiaoroverdose18(17)5(14)7(19)6(17)Others3(3)1(3)1(3)1(3)MismatchcovariatesSexn(%)Match71(66)24(67)22(61)25(69)Mismatch37(34)12(33)14(39)11(31)Racen(%)Match67(62)17(47)26(72)24(67)Mismatch41(38)19(53)10(28)12(33)HLAmismatchmean(SD)9·70(2·04)9·27(1·49)9·95(2·36)9·90(2·20)Sensitizedrecipientsn(%)41(37)14(37)17(50)9(25)CMVstatusn(%)D+R+40(42)14(44)14(42)12(39)D+R22(23)10(31)7(21)5(16)DR+19(20)5(16)5(15)9(29)DR15(16)3(9)7(21)5(16)Peri-transplantcovariatesAverageischemictimemean(SD)276·17(62·02)282·14(53·26)265·58(67·14)281·74(64·56)Typeoftransplantationn(%)Bilateral84(79)30(86)30(83)24(67)Single23(21)5(14)6(17)12(33)Inductionn(%)No70(65)24(69)26(72)20(56)Yes37(35)11(31)10(28)16(44)PGDgrade3n(%)No78(78)27(82)27(79)24(73)Yes22(22)6(18)7(21)9(27)UNCORRECTEDPROOFEBioMedicinexxx(2019)xxx-xxx7Table1(Continued)VariablesTotalLowtertileMiddletertileUppertertileDSAwithin3months43(41)13(35)13(38)17(50)Acuterejectionepisodesmean(SD)1·81(1·92)1·89(2·34)1·22(1·17)2·31(2·07)Studycohortn(%)GRAfT54(50)16(44)16(44)22(61)GTD54(50)20(56)20(56)14(39)COPD=chronicobstructivelungdisease,CF=cysticfibrosis,ILD=interstitiallungdisease,PAH=pulmonaryarterialhypertension,LAS=lungallocationscoreobtainedclosesttotransplantation·GSW=gunshotwound,overdose=deathfromdrugoverdose,HLA=humanleukocyteantigenCMVstatus=cytomegalovirusIgGstatus,D+=donorpositiveforCMVIgG,D-=donornegativeforCMVIgG,R+=recipientpositiveforCMVIgG,R-=recipientnegativeforCMVIgG,PGDgrade3=Primarygraftdysfunctiongrade3definedbyInternationalSocietyofHeartandLunggradingcriteria,DSA=donor-specificantibodies.
continuousvariablefirstandthenasacategoricalvariable,separately.
Forthemultivariateanalysis,wefirstincludedavddDNAasacontinuousvariabletoestimatetheadjustedHazardRatios(adjustedHRs,model1),adjustingforthecovariatesbasedonthestatisticalsignificanceifthepvaluewascomeusingavddDNAasacategoricalvariablebasedontertiles(upper-,middle-,andlow-tertileavddDNA)withthelow-ter-tileasthereferencegroup(model2).
StatisticalanalyseswereperformedusingGraphPadPrism7graphing(GraphPadSoftware)[20]andtheSTATA12·0statisticalpackage(StataCorpLLC)[21].
2.
7.
PoweranalysesTheprincipalanalysisassessedtherelationshipofavddDNAtoallograftfailure.
Thepriorisamplesizecalculationsuggestedthat,giventhestandarddeviationofavddDNAas1·9,asamplesizeof106willallowustohaveenoughpower(>0·80)todetectasignificantregressioncoefficientforavddDNAiftheregressioncoefficientis>0·2,assumingthattheoverallfailureinthesur-vivalanalysesas50%.
2.
8.
ddcfDNAandclinicalassessmentsbeyondthreemonthsToassessallograftinjurybeyondtheinitialthreemonthsoftransplantation,weusedbothclinical(spirometry,histopathol-ogy,microbiology)andmolecular(%ddcfDNA)tools.
MeanFEV1and%ddcfDNAforeachavddDNAtertilewerecalculatedandcomparedatarbitrarythree-monthintervals;valueswithintwoweeksoneithersideofeacharbitraryintervalwereincluded.
Basedonpreviouslypublisheddatafordetectingacuterejection[14]wefocusedonepisodesinwhich%ddcfDNAwas≥1%todeterminewhetherelevated%ddcfDNAlevelscoincidedwithclinicalindicatorsofallograftinjury.
TheseclinicalindicatorsincludedACR,AMR,and/orclinicalinfectionandwereconsideredepisodesiftheywereseparatedbyatleast2weeks.
Wetime-matchedtheepisodesofelevated%ddcfDNAtoclinicaldataandassignedthemasclinicallydetectableiftheywerewithintwoweeksofaclinicaleventorclinicallysilentifoutsideofthistimerange.
ClinicaldefinitionsofAMR,ACR,andclinicalinfectionsaresummarizedintheSupplementaryMethods.
3.
Results3.
1.
ParticipantcharacteristicsandclinicaloutcomesOfthe190subjectsapproachedforthisstudy,20subjectsdeclinedparticipation,tendiedwithinthreemonthsoftransplan-tationandafurther52subjectslackedplasmasamplesforavddDNAmeasurements,leavingafinalcohortof108subjectsforanalysis–54fromGTDand54fromGRAfT.
Theanalyzedcohortof108subjectshadanaverageageattransplantationof54·1years.
Interstitiallungdiseasewasthemostcommonreasonfortransplantation(46·3%)followedbycysticfibrosis(20·4%).
Diseaseseverityindexdefinedbylungallocationscoreattransplantationshowedanaverageof47·3(Table1).
ComparisonsoftheanalyticandexcludedcohortsandoftheGTDandGRAfTcohortsareshowninSupplementaryTable1candd.
Medianfollow-uptimewas36·3months.
Twoofthe108subjectsdiedwithin3-monthswereexcludedleaving106subjectsforfurtheranalysis.
Ofthe106subjects,43·5%(n=46)reachedtheprimaryoutcome(deathfromrespiratorycauses=27,CLADbutnorespiratorydeathorre-transplantation=14,andre-transplantation=5),41·7%(n=45)reachedthesecondaryoutcomeofall-causemortality,and41·5%ofthecohortwerealiveandfreefromCLAD.
Causesofdeathwererespiratoryfailure(27/45),followedbysepsis(14/45),andothers(4/45:coloncancer,myocardialinfarction,pulmonaryembolism,post-transplantlympho-proliferativedisorder).
Ofthe18subjectswhodiedfromnon-respiratorycauses,themajority(14/18)hadconcurrentacuterejec-tionorevidenceofsevereCLADbythetimeofdeath.
UNCORRECTEDPROOF8EBioMedicinexxx(2019)xxx-xxxTable2RelationshipbetweenrecipientanddonorcovariatesandavddDNA(n=108).
VariablesCoefficientUnadjustedregressioncoefficientpCoefficientAdjustedregressioncoefficientp95%CI95%CIRecipientcovariatesAge0·030–0·050·0230·020·05–0·070·973SexMaleReferenceFemale0·4311·17–0·310·256RaceNon-CaucasianReferenceCaucasian0·771·75–0·210·1210·371·81–1·080·611ObesityBMI≥30Kg/m2NoReferenceYes1·290·3–2·280·0111·200·17–2·580·085SmokinghistoryNeverReferencePast0·780·03–1·590·0580·151·51–1·810·857TransplantationreasonCOPDReferenceCF1·082·29–0·140·0830·79--2·78–1·210·432ILD0·251·22–0·730·6170·481·82–0·860·476PAH0·362·72–2·200·7600·753·81–2·300·621Sarcoidosis/others0·892·23–0·450·1930·992·89–0·920·303LAS0·030·01–0·060·0050·030·002–0·060·038DonorcovariatesAge0·030·05–00·0390·010·05–0·020·357SexMaleReferenceFemale0·340·42–1·090·376RaceNon-CaucasianReferenceCaucasian0·120·98–0·740·789ObesityBMI≥30Kg/m2NoReferenceYes0·090·77–0·950·837SmokinghistoryNeverReferencePast0·031·71–1·650·968ChesttraumaNoReferenceYes1·231·37–3·830·350CauseofdeathHeadtraumaorGSWReferenceCVA0·791·86–0·280·1470·2751·63–1·080·685Anoxiaoroverdose0·781·61–0·050·0651·413·34–0·520·149Others0·352·63–1·940·7650·323·10–2·450·816MismatchcovariatesSexMatchReferenceMismatch0·060·84–0·720·882RaceMatchReferenceMismatch0·230·54–0·990·558SensitizedrecipientNoReferenceYes0·020·21–0·240·880HLAmismatch0·561·32–0·210·1530·831·81–0·140·091CMVstatusD+R+ReferenceDR+0·661·67–0·360·2020·431·86–0·550·510D+R1·000·07–2·060·0670·251·01–1·510·697DR0·091·06–1·250·8730·141·79–1·140·851Peri-transplantcovariatesAverageischemictime0·000·01–0·010·498TypeoftransplantationBilateralReferenceSingle0·570·33–1·470·2160·211·40–0·970·719InductionNoReferenceUNCORRECTEDPROOFEBioMedicinexxx(2019)xxx-xxx9Table2(Continued)VariablesCoefficientUnadjustedregressioncoefficientpCoefficientAdjustedregressioncoefficientp95%CI95%CIYes0·550·23–1·320·1671·182·84–5·200·559PGDgrade3NoReferenceYes1·090·2–1·970·0170·311·06–1·690·647DSAwithin3months0·020·17–0·210·834Acuterejectionepisodes0·590·13–1·300·1070·610·39–1·600·225StudycohortGRAfTReferenceGTD0·681·41–0·050·0670·134·26–4,510·954UnadjustedandadjustedlinearregressionmodelswereperformedinrelationtoavddDNA.
Theadjustedmodelincludedvariableswithp-valuecomparabletothefirstgroupbythreemonthsaftertransplantation.
AthirdgroupwithavddDNAintheuppertertile(median3·6%,range=2·2%–9·8%)showedevenslower%ddcfDNAdecay(Fig.
2c,solidline;SupplementaryFig.
2,Subjects11–15)withpersistentlyelevated%ddcfDNAlevelscomparedtotheothertwogroups.
SubjectswithavddDNAintheuppertertileshadhigherLASscoreandnumberofacuterejectionepisodes(Table1)thanthemiddleorlowavddDNAgroups.
3.
3.
RelationshipbetweendonorandrecipientcovariatesandavddDNASubjectsintheupperavddDNAshowedthefollowingpre-transplantcharacteristicscomparedtosubjectsinthelowavddDNAtertile:olderage(57.
2vs.
51.
6years),higherLASscore(51.
1vs.
41.
2),higherfrequencyofprevioussmoking(33%vs.
29%),andhigherfrequencyofCMVDR+status(29%vs.
16%).
Interstitiallungdiseasewasmorecommon(53%vs.
36%)whilecysticfibrosiswaslesscommon(14%vs.
25%)asreasonsfortransplantationfortheupperthanlowavddDNAtertile.
Attransplanta-tion,subjectsintheupperavddDNAtertilereceivedmoresinglelungtransplantation(33%vs.
21%)andinductiontherapy(44vs.
31%)thansubjectsinthelowavddDNAtertile.
Aftertransplantation,theydevelopedPGDgrade3(27%vs.
18%)andDSA(50%vs.
37%)morefrequently,andalsoshowedmoreepisodesofacuterejection(2.
3vs.
1.
9episodes)thansubjectsinthelowavddDNAtertile(Table1).
WenextperformedlinearregressionstoidentifyfactorsthatcorrelatewithhigheravddDNAlevels.
Pre-transplantfactorsin-cludingrecipientage,obesity,LASscore,aswellasdonoragecorrelatedwithhigheravddDNAlevelsonunivariateanalysis.
OnlyrecipientLASwaspositivelyassociatedwithhigheravddDNAaftermultivariateanalysis(Table2).
Intheearlypost-trans-plantationperiod,useofinductiontherapyorischemictimedidnotcorrelatewithavddDNAlevels.
ThepresenceofPGDgrade3at72h[22]correlatedhigheravddDNAlevelsthannograde3PGD(Table2)onunivariatebutnotonmultivariateanalysis.
UNCORRECTEDPROOF10EBioMedicinexxx(2019)xxx-xxxTable3RelationshipbetweenavddDNA,donor,andrecipientcovariatesandallograftfailure(n=106).
VariablesHRUnadjustedhazardratio(HR)pHRModel1adjustedHRHRModel2adjustedHR95%CI95%CIp95%CIpContinuousavddDNA1·341·16–1·55comeafterBonferroniadjustment.
Afteradjustingforlungmassbydoubling%ddcfDNAvaluesforsingle-lungtransplantations,single-anddouble-lungtransplan-tationsshowedcomparableavddDNAlevels.
3.
4.
Relationshipbetweenaverage%ddcfDNA(avddDNA),clinicalcovariatesandoutcomesToestablishanylinksbetweenavddDNA,donorcovariatesorrecipientcovariatesandtheprimaryorsecondaryoutcomesofthestudy,weconductedCoxmodels,includingavddDNAasacontinuousorcategoricalvariable(avddDNAtertiles)foroutcomeseparately.
Intheunadjustedunivariateanalysis,bothavddDNAandavddDNAtertileweresignificantlyrelatedtotheprimaryoutcome(ps≤0.
001foravddDNAanduppertertilevs.
lowtertile).
ThesignificanceheldafterBonferroniadjustment(pscome:a1%increaseinavddDNAincreasedtheriskofallograftfailure1·4-fold(95%CI1·1–1·5,p=0·015).
SubjectsintheupperavddDNAtertilehada6·6-foldgreaterriskofallograftfailurecomparedtothesubjectsinthelowavddDNAtertile(HR=6·6;95%CI1·6–19·9,p=0·007,Fig.
3a,Table3).
Theirmediantimetodevelopallograftfailurewas25monthscomparedto42and45monthsforsubjectsinthemiddleandlowavddDNAtertilesrespectively(SupplementaryTable4).
OnmultivariateanalysiswithcategoricalavddDNA,thefollowingfactorsalsoincreasedtheriskofreachingtheprimaryendpoint:non-Caucasianrace,PGDgrade3,singlelungtransplant,CMVD+Rstatus,highernumberacuterejectionepisodes(Table3).
Donorcauseofdeathfromanoxiaordrugoverdosereducedtheriskoftheprimaryoutcome.
TheprimaryoutcomewashigherintheGTDthanintheGRAfTcohortonunivariateanalysis,butnotafteradjustingfordonorandrecipientcovariates.
MultivariateanalysiswithcontinuousavddDNAispresentedinTable3.
ThesecondaryoutcomeofCLAD-freesurvival,definedasnoCLADornodeath,showedastrongassociationtoavddDNA.
A1%increaseinavddDNAincreasedtheriskofCLAD/deathby1·5(95%CI1·2–1·9,p=0·001).
SubjectsintheupperavddDNAtertileshoweda7·8-foldhigherriskofCLAD/deathcomparedtothesubjectsintheavddDNAlowtertile(95%CI2·2–27·7,p=0·001,Fig.
3b,SupplementaryTable3a).
Theirmediantimetoreachthisendpointwas14monthscomparedto36and32monthsforsubjectsinthemiddleandlowavddDNAtertilesrespectively.
TheavddDNAwasequallypredictiveofCLAD-freesurvivalwhentheGRAfTandGTDcohortswereanalyzedseparately(GRAfT,logrankchisquare=7·4,p=0·006,GTD,logrankchisquare=10·9,p=0·001).
Femalesex,non-Caucasianrace,recipientstransplantedforILD,highernumberofepisodesofacuterejectionandbeingintheGTDcohortalsoincreasedtheriskofCLAD/deathonmultivariateanalysisthatincludedavddDNAascategoricalvariable.
Withchronicobstructivelungdiseaseasreference,cysticfibrosiswastheonlyreasonfortransplantationassociatedwiththeprimaryoutcome.
MultivariateanalysisthatincludesavddDNAascontinuousvariablesisrepresentedinSup-plementaryTable3a.
All-causemortalityalsoshowedastrongassociationwithavddDNA;a1%increaseinavddDNAincreasedtheriskofdeathfromanycauseby1·5-fold(95%CI1·2–1·8,p=0·001,SupplementaryTable3b).
SubjectsintheupperavddDNAtertileshowedofa3·9-foldhigherriskofall-causemortalitycomparedtothesubjectsinthelowavddDNAtertile(95%CI1·4–10·1,p=0·006,SupplementaryTable3b).
ThemediansurvivalfortheupperavddDNAtertilewas31months,comparedto42and55monthsforthemiddleandlowavddDNAtertilesrespectively(SupplementaryTable4).
InadditiontoavddDNA,PGDgrade3andbeingintheGTDcohortincreasedtheriskall-causemortalityonmultivariateanalysis(SupplementaryTable3).
MultivariateanalysisthatincludesavddDNAascontinuousvariablesisrepresentedinSupplementaryTable3b.
UNCORRECTEDPROOF12EBioMedicinexxx(2019)xxx-xxxFig.
2.
Calculationofthepredictivebiomarker:average%ddcfDNA(avddDNA).
(A)%ddcfDNAvs.
timecurveforarepresentativepatientusedtocalculatethepredictivemeasure,avddDNA,whichistheareaunderthecurve(AUC)dividedby77days(theintervalduringdays14-90,inclusive).
(B)FrequencydistributionofavddDNAvalues.
Thecohortwasdividedintoupper,middleandlowtertilesbasedonavddDNAvalues.
Thex-axisbreaksrepresenttheavddDNAtertiles.
(C)Individualsubjects%ddcfDNAvaluesovertimeseparatedbyavddDNAtertilesofupper(blackdots,blacksolidline),middle(reddots,redsolidline)andlow(bluedots,bluesolidline)tertiles.
Trendsof%ddcfDNAforarepresentativesubjectineachavddDNAisrepresentedbysolidlines.
Plotsfor15individualsubjects,fiverepresentativesforeachgroup,areshowninSupplementaryFig.
3.
3.
5.
RelationshipbetweenavddDNA,PGD,andoutcomesWefurtherassessedtherelationshipofavddDNAandallograftfailureinsubjectswithPGDgrade3,themoststudiedearlyriskfactorforallograftfailure.
Twenty-oneof106subjectshadPGDgrade3,ofthese,sevensubjectswereintheavddDNAuppertertile,eightsubjectswereinthemiddletertile,andsixwereinthelowtertile.
Subjectswithgrade3PGDintheuppertertilewereathigherriskofallograftfailure(5/7or71·4%)thanthoseinthemiddle(3/8or37·5%)andlowtertiles(1/6or16·6%).
3.
6.
%ddcfDNAtrendsbeyondthreemonthsaftertransplantationTofurtherinvestigateourhypothesisthatearlyunresolvingallograftinjurysetsthestageforsubsequentallograftinjury,westudied%ddcfDNAlevelsbeyondtheearlypost-transplantationperiodfrom3to15months.
Subjectsinthelowtertileforavd-dDNAshowedcontinuousstable%ddcfDNAlevelsuntil12monthsaftertransplantation,afterwhichweobservedaslowsec-ondaryincrease(Fig.
4a).
Bycontrast,subjectsintheupperavddDNAtertileshowedpersistentlyhigher%ddcfDNAlevelscom-paredtoothertertiles.
Their%ddcfDNAlevelsnadiratsixmonthspost-transplantationandthereafter,themedian%ddcfDNAincreasedfortheremainderofthestudyperiod.
3.
7.
ClinicalmanifestationsofelevatedddcfDNAepisodesWenextinvestigatedwhethertheepisodesofelevated%ddcfDNAlevels(%ddcfDNA≥1%)manifestedclinically.
Wematchedelevated%ddcfDNAepisodestoclinically-diagnosedepisodesandfoundthattwo-thirdsofelevated%ddcfDNAUNCORRECTEDPROOFEBioMedicinexxx(2019)xxx-xxx13Fig.
3.
Relationshipofaverage%ddcfDNA(avddDNA)andoutcomes.
Kaplan-Meiercurvesfortimetoallograftfailure(A)andCLAD-freesurvival(B)fortheupper-,middle-,andlowavddDNAtertiles.
ThetablebelowthecurvesshowsthenumberofsubjectsatriskineachavddDNAtertileatdifferenttimepoints.
P-UNCORRECTEDPROOF14EBioMedicinexxx(2019)xxx-xxxvaluesdeterminedbylog-ranktest.
Cox-regressionanalysesofavddDNAinrelationshiptotheprimaryandsecondaryendpointsarepresentedinTable3andSupplementaryTable3aFig.
4.
Molecular(%ddcfDNA)andclinical(FEV1)measurementsofallograftinjurybeyondtheearlypost-transplantationperiod(a)Median%ddcfDNAtrendsfortheupper-,middle-,andlowtertilesofavddDNAupto18monthspost-transplantation(b)Episodesofelevated%ddcfDNA(definedas1%%ddcfDNA)weretime-matchedtoclinicaleventsofacuterejection(AMRorACR)orclinicalinfectionsforeachavddDNAtertile.
Thosewhichcoincidedwithaclinicaleventwereclassifiedas"clinically-detected"(greyfilled),andthoseepisodesthatdidnotcoincidewithaclinicaleventwereclassifiedas"clinically-silent"(nofill).
(c)Thenumberofclinically-detectedandclinically-silentepisodesperfive%ddcfDNAmeasurementsisshown(averageforeachtertile).
Clinically-detected(greyfilled)andclinically-silent(nofilled)elevated%ddcfDNAepisodesarerepresentedseparately.
Errorbarsrepresentstandarderror.
P-valuesweredeterminedusingStudentt-test.
(d)AverageFEV1overtimeforeachavddDNAtertile.
ErrorBarsrepresentstandarderror.
episodes(n=138)wereclinicallysilent(Fig.
4b)andonlycoincidedwithathirdoftheelevated%ddcfDNAepisodesACR(n=21),AMR(n=22),clinicalinfection(n=33).
Theremainingclinically-silentandclinicallyapparent%ddcfDNAepisodeswerepredominantinsubjectsintheavddDNAuppertertilecomparedtotheothertwotertiles(Fig.
4c).
Ofthe33clinicalinfec-tionepisodeswithelevated%ddcfDNA,respiratoryvirusesweremorecommonlydetected(n=19)thanbacterial(n=8)orfungalinfections(n=6).
Wealsoconsidermicrobialcolonizersaspotentialtriggersforclinically-silent%ddcfDNAelevations.
46·2%ofthecohortharboredanorganismclassifiedasacolonizer.
%ddcfDNAattimepointswithcolonizersweresimilartonon-rejectioncontroltime-points.
However,whenCLAD-associatedcolonizers(pseudomonas,staphylococcus)wereconsideredseparately,weob-servedhigher%ddcfDNAlevelsthannon-rejectiontime-points(1·1%vs.
0·3%,p=0·056).
3.
8.
RelationshipofavddDNAleveltolungfunctionWeanalyzed3899spirogramstostudylungfunctionovertimefortheupper,middle,andlowavddDNAtertiles.
SubjectsintheuppertertilehadamedianpeakFEV1of74%ofpredictedvalues,occurringonaverage,sixmonthsaftertransplantation.
Thereafter,themedianFEV1ofthistertileshowedasustaineddecline,ultimatelyleadingtoCLAD.
Overthesametimeperiod,theFEV1ofsubjectsinthemiddleandlowavddDNAtertilescontinuedtoincrease:to80%at12monthsinthemiddletertiles,and84%at18monthsinthelowtertiles(Fig.
4d).
UNCORRECTEDPROOFEBioMedicinexxx(2019)xxx-xxx154.
DiscussionInthisproof-of-conceptstudy,weusedanovelgenomic-derivedbiomarker,%ddcfDNA,tocharacterizepost-transplantationtrendsoflung-allograftinjurythatleadtoallograftfailureanddeath.
Ourresultsdemonstratethattheaverage%ddcfDNAintheearlypost-transplantationperiod(avddDNA)correlatedwiththedevelopmentofallograftfailureandall-causemortality(Fig.
3).
Adoptionofanynewpredictivebiomarkerrequiresanevaluationofitscorrelationwithandbenefitcomparedtoexistingclin-icalmeasures.
Wefoundthattraditionalriskfactorsofpooroutcomesincludingolderrecipients,priorsmokinghistory,CMVD+Rstatus,severePGD,numberofrejectionepisodesandothervariables(Table1)weremorecommoninsubjectsintheupperthanloweravddDNAtertiles.
Wealsocorrelatedourgenomicmarkerstospirometry,aclinicalmeasureofallograftfunction.
WefoundthatsubjectswithhighavddDNAlevelssubsequentlyshowedlowerlungfunctionandmoreclinicalcomplications(acuterejectionorinfections)thansubjectswithloweravddDNAlevels.
Atamolecularlevel,thesesubjectsshowmorefrequentelevated%ddcfDNAlevelsbeyondtheearlypost-transplantationperiod,suggestingthatearlyunresolvingallograftinjurysetsthestageforfurtherallograftinjuryanddysfunction.
Onlyone-thirdoftheseelevated%ddcfDNAepisodeswereassociatedwithacuterejectionorclinicalinfection.
Theremainderwerenotcoincidenttoanysignsdetectablebyhistopathology,spirometry,clinicalexaminationorbyanyotherclinicaltests.
Theseepisodesofclinically-silentelevationsin%ddcfDNAcouldrepresentearlydetectionofinjurythatprogressestopathologicallyovertchanges.
OurrecentanalysisofsubjectswithAMRsupportsthishypothesis;weobservedasustainedrisein%ddcfDNAforseveralweekstomonthsbeforeclinicalorhistologicalmanifestationsofAMRbecameapparent[23].
Similarly,inthecurrentstudy,subjectswithavddDNAintheuppertertileshowedanearlyandlowerpeakFEV1levels,(Fig.
4)andagreaterriskofprogressiontoallograftfailure.
Thus,theseclinicallysilenteventsdetectedby%ddcfDNAandnotbypreviouslydescribedclinicaltoolscouldofferearlierdetectionofpathologicalchanges.
ThismolecularassaythereforeoffersearliertimepointstopotentiallyintervenewithtreatmentslikeextracorporealphototherapyandotherstoreduceorpreventsubsequentCLADandallograftfailure.
Potentialbiologicaltriggersoftheseclinically-silentincreasesin%dd-cfDNAmayincludealloantibodies[23],acidreflux[24],andoccultinfection;thesewarrantfurtherinvestigation.
Thecurrentstudyindicatesthatwecouldstratifysubjectsintotertilesofrisk,evensubjectswithprimaryGraftDysfunction(PGD)[22],themoststudiedearlyriskindicatorofallograftfailure.
PGDhastheadvantageofassigningriskallograftriskwithin72hoftransplantation,asopposedtoavddDNA,whichiscomputedoverthreemonths.
However,useofavddDNAenabledbetterstratificationofthePGDsubjectsfortheriskofallograftfailurethanuseofPGDalone.
Weacknowledgethatthelackofcorrela-tionmaybeduetothesmallsamplesizeofourstudyinrelationtopriorstudies[5].
Further,ourstudyevaluatedPGDonDay3andnotDay1or2,andthereforepotentiallyunderestimatedtheincidenceofPGD.
TheinteractionbetweenPGDandavddDNAdeservescarefulanalysisinalargersamplesize.
Withmorepowerfulcomputingnowavailable,weanticipatethatfuturestudieswillusemachinelearningtoolstoderiveequallypredictiveavddDNAwithlesstimelag.
Inourpatientcohort,respiratoryfailurewasthepredominantcauseofdeath.
Thestrongcorrelationobservedbetweenavd-dDNAandoursecondaryoutcomeofall-causemortalitythereforeaddssupporttoourcentralhypothesisthatearlyandunresolv-ingallograftinjuryisanearlyindicatorofallograftfailure.
TherelationshipbetweenavddDNAandall-causemortalityinothercohortswhererespiratoryfailureisalesspredominantcauseofdeath[25]deservesfurtherevaluation.
However,therelationshipweobservedbetweenavddDNAandallograftfailureislikelytobesimilareveninthelattercohort.
LimitationsofourstudyincludetheinabilitytoassessCLADinsubjectswhoweretoounwelltoundergospirometryassess-ments.
However,thesesubjectsrapidlyprogressedtodeathandtheiradverseoutcomewasthereforecapturedinourcompositeendpoint.
Also,weobservedlowersurvivalinourcohortcomparedtosomelargeregistries[25].
ThelowsurvivalwasprimarilyinGTDcohortduetoahighincidenceofCLADsecondarytoantibody-mediatedrejection.
However,avddDNAwaspredictiveintheGRAfTandGTDcohortswithdifferentmortalityratessuggestingthehighmortalityisnotalimitationtogeneralizabilityofourfindings.
Further,variabilityinnumberofsamplesanalyzedperpatientmayintroducebiasincomputingavddDNA.
Thevariablenumberofsamplesperpatientwasdueprimarilytodifferencesinthenumberclinically-indicatedbronchoscopiesandthereforereflectusualpost-transplantcareratherthansamplingbias.
Nonetheless,suchdifferencemayintroduceanalyticbias.
Toreducepotentialbias,weselectedtheAUCmethodtocomputeavddDNA.
Thisapproachnormalizesfornumberofsampleanalyzedandfixedtheintervalofinterestto3monthsforallpatients.
Further,onlysubjectswithatleast3samplesanalyzedwithintheintervalofinterestwereincluded.
Futurestudiesshouldaddresstheselimitations,validateourfindings,andconcurrentlyemploymachinelearningalgorithmtocomputeavddDNAmeasurementwithlesstimelagthanthethreemonthsusedinthisstudy.
Suchfuturestudywouldrequirealargercohortsizeandmorefrequent%ddcfDNAmeasurements.
Theearliermeasuresmayenableassessmentsoftheriskofallograftfailureintheinitial3monthsaftertransplantation.
Thisriskwasnotassessedinouranalysiswheresubjectswhodiedwithin3monthsoftransplantationwereexcludedfromthestudy.
Ifvalidated,avddDNAwouldenablethestratificationsubjectsatriskofallograftfailureinroutineclinicalpractice.
Routineclinicaluseofthe%ddcfDNAtestishighlyfeasiblesincethetestisreproducibleandtransferrable[13].
Further,commercialversionsofthetestwillsoonbeavailableforclinicaluse.
Usingthistesttoidentifyhigh-risksubjectsearlycouldallowtailoredmonitoring,manipulationoftheirimmunosuppressionregimensorothertherapiesinthehopeofdelayingorevenavoidingallograftfailure.
UNCORRECTEDPROOF16EBioMedicinexxx(2019)xxx-xxxInsummary,fromourproof-of-conceptstudy,weproposeavddDNAasapredictivemarkerforallograftfailureandprema-turedeath.
Toourknowledge,thisisthefirstreportofamethodtodetectandquantifyclinicallysilent,butlikelypathological,eventsprecedingallograftfailure.
Theseunderlyingevents,uncoveredthroughouranalysisof%ddcfDNA,mayhelptoexplaintheunacceptablyhighrateofCLADthatremainstheAchillesheeloflungtransplantation.
Ourfindingsrequirefurthervalidationinlargercohorts,butneverthelesslaythefoundationforfutureclinicalstudies.
Mechanisticstudieswillalsobeneeded,toin-vestigatethepathologicalimportanceofclinically-silentallograftinjury.
Futurestudiesmayalsoevaluateourhypothesisinothersolidorganstransplantation,inwhichassessmentsof%ddcfDNAhaveshownsimilarperformancecharacteristicsfordetectingacutecomplications[13,15–17].
FundingTheGenomeTransplantDynamicsStudy(NCT01985412)isfundedbytheNationalInstitutesofHealth,GrantRC4AI092673.
TheGenomicResearchAllianceforTransplantationStudy(NCT02423070)isfundedbytheDivisionofIntra-muralResearchoftheNationalHeart,Lung,andBloodInstitute.
Fundershadnoroleinstudydesign,datacollection,dataanaly-sis,interpretation,writingofthereport.
Author'scontributionsSA-E,KKKandHA-Econceivedthestudyanddesignedanalyses.
UF,IT,PDS,HL,SDN,JBO,AWB,AIrecruitedandmon-itoredpatients.
IT,PDS,SDN,JBO,AWB,KP,AIadjudicatedforoutcomes.
GJBandCMperformedconsensushistopathologyreads.
AD,JW,HL,UF,NG,DG,AM,KBcollectedbiologicalsamplesandclinicaldataandprepareddatatables.
KKK,SRQ,IDVmeasuredddcfDNAforGTD.
SG,MK,JZ,ITmeasuredddcfDNAforGRAfT.
SA-EandYWperformedstatisticalanalysisandcompileresults.
SA-Ewroteinitialmanuscriptdraft,allauthorsreviewedthemanuscriptsandrevisions.
Uncitedreferences[26–29]Acknowledgement1.
NationalCancerInstituteGenomicCoreforgenotyping2.
AlisonDavis,PhD,NationalInstitutesofHealthforeditorialreview3.
ThefollowingOrganprocurementorganizationsforprocurementofdonorwholebloodsamples;WashingtonRegionalTrans-plantConsortium,LifeNetHealth,LivingLegacyFoundation4.
GenomicsandComputationCoresofDivisionofIntramuralResearchofNationalHeart,LungandBloodInstituteforsequenc-ingandcomputationalsupport.
5.
AdjudicationcommitteeforPGDErrolBush(JohnsHopkins),PaliShah(JohnsHopkins),JoshuaDiamond(UniversityofPennsylvania),KapilPatel(Stanford),AnneBrown(InovaFairfax).
AppendixA.
SupplementarydataSupplementarydatatothisarticlecanbefoundonlineathttps://doi.
org/10.
1016/j.
ebiom.
2018.
12.
029.
References[1]K.
C.
Meyer,G.
Raghu,G.
M.
Verleden,etal.
,AninternationalISHLT/ATS/ERSclinicalpracticeguideline:diagnosisandmanagementofbronchiolitisobliteranssyndrome,EurRespirJ44(6)(2014)1479–1503.
[2]J.
P.
Woodrow,O.
A.
Shlobin,S.
D.
Barnett,N.
Burton,S.
D.
Nathan,Comparisonofbronchiolitisobliteranssyndrometootherformsofchroniclungallo-graftdysfunctionafterlungtransplantation,JHeartLungTransplant29(10)(2010)1159–1164.
[3]A.
Sulania,S.
Sachdeva,D.
Jha,G.
Kaur,R.
Sachdeva,Organdonationandtransplantation:anupdatedoverview,MAMCJMedSci2(1)(2016)18–27.
[4]S.
E.
Verleden,J.
L.
Todd,M.
Sato,etal.
,ImpactofCLADphenotypeonsurvivalafterlungretransplantation:amulticenterstudy,AmJTransplant15(8)(2015)2223–2230.
[5]J.
M.
Diamond,J.
C.
Lee,S.
M.
Kawut,etal.
,Clinicalriskfactorsforprimarygraftdysfunctionafterlungtransplantation,AmJRespirCritCareMed187(5)(2013)527–534.
[6]V.
N.
Lama,S.
Murray,R.
J.
Lonigro,etal.
,CourseofFEV(1)afteronsetofbronchiolitisobliteranssyndromeinlungtransplantrecipients,AmJRespirCritCareMed175(11)(2007)1192–1198.
[7]L.
J.
Lobo,P.
G.
Noone,Respiratoryinfectionsinpatientswithcysticfibrosisundergoinglungtransplantation,LancetRespirMed2(1)(2014)73–82.
[8]S.
M.
Bhorade,A.
N.
Husain,C.
Liao,etal.
,Interobservervariabilityingradingtransbronchiallungbiopsyspecimensafterlungtransplantation,Chest143(6)(2013)1717–1724.
[9]D.
J.
Dwidevi,L.
J.
Toltl,L.
L.
Swystun,Prognosticutilityandcharacterizationofcell-freeDNAinpatientswithseveresepsis,CritCare16(2012).
[10]M.
Ggenur,J.
Burcharth,I.
Ggenur,Theroleoftotalcell-freeDNAinpredictingoutcomesamongtraumapatientsintheintensivecareunit:asystem-aticreview,CritCare21(1)(2017)14.
UNCORRECTEDPROOFEBioMedicinexxx(2019)xxx-xxx17[11]Z.
Wang,Y.
Cheng,T.
An,etal.
,DetectionofEGFRmutationsinplasmacirculatingtumourDNAasaselectioncriterionforfirst-linegefitinibtreatmentinpatientswithadvancedlungadenocarcinoma(BENEFIT):aphase2,single-arm,multicentreclinicaltrial,LancetRespirMed6(9)(2018)681–690.
[12]Y.
M.
D.
Lo,J.
Zhang,T.
N.
Leung,T.
K.
Lau,A.
M.
Z.
Chang,N.
M.
Hjelm,RapidClearanceofFetalDNAfromMaternalPlasma,AmJHumanGenet64(1)(1999)218–224.
[13]S.
Agbor-Enoh,I.
Tunc,I.
DeVlaminck,etal.
,Applyingrigorandreproducibilitystandardstoassaydonor-derivedcell-freeDNAasanon-invasivemethodfordetectionofacuterejectionandgraftinjuryafterhearttransplantation,JHeartLungTransplant36(9)(2017)1004–1012.
[14]I.
DeVlaminck,L.
Martin,M.
Kertesz,etal.
,Noninvasivemonitoringofinfectionandrejectionafterlungtransplantation,ProcNatlAcadSciUSA112(43)(2015)13336–13341.
[15]I.
DeVlaminck,H.
A.
Valantine,T.
M.
Snyder,etal.
,Circulatingcell-freeDNAenablesnoninvasivediagnosisofhearttransplantrejection,SciTranslMed6(241)(2014)241ra77.
[16]E.
Schütz,A.
Fischer,J.
Beck,etal.
,Graft-derivedcell-freeDNA,anoninvasiveearlyrejectionandgraftdamagemarkerinlivertransplantation:aprospective,observational,multicentercohortstudy,PLoSMed14(4)(2017),e1002286.
[17]R.
D.
Bloom,J.
S.
Bromberg,E.
D.
Poggio,etal.
,Cell-freeDNAandactiverejectioninkidneyallografts,JAmSocNephrol28(7)(2017)2221–2232.
[18]J.
Hair,W.
C.
Black,B.
J.
Babin,R.
E.
Anderson,Multivariatedataanalysis7ed.
UpperSaddleRiver,PearsonEducationalInternational,NewJersey,2010.
[19]D.
W.
Hosmer,S.
Lemeshow,Appliedlogisticregression,Wiley,NewYork,2000.
[20]G.
Prism,GraphPadPrism.
version7,GraphPadSoftware,LaJolla,CA,2008.
[21]StataCorp,Statastatisticalsoftware,In:12R,editor.
Collegestation,StataCorpLp,TX,2011.
[22]J.
D.
Christie,M.
Carby,R.
Bag,etal.
,ReportoftheISHLTWorkingGrouponprimaryLungGraftDysfunctionpartII:definition.
AconsensusstatementoftheInternationalSocietyforHeartandLungTransplantation,JHeartLungTransplant24(10)(2005)1454–1459.
[23]S.
Agbor-Enoh,A.
M.
Jackson,I.
Tunc,etal.
,Latemanifestationofalloantibody-associatedinjuryandclinicalpulmonaryantibody-mediatedrejection:evidencefromcell-freeDNAanalysis,JHeartLungTransplant37(7)(2018)925–932.
[24]S.
M.
Palmer,A.
P.
Miralles,D.
N.
Howell,S.
R.
Brazer,V.
F.
Tapson,R.
D.
Davis,Gastroesophagealrefluxasareversiblecauseofallograftdysfunctionaf-terlungtransplantation,Chest118(4)(2000)1214–1217.
[25]R.
D.
Yusen,L.
B.
Edwards,A.
Y.
Kucheryavaya,etal.
,Theregistryoftheinternationalsocietyforheartandlungtransplantation:thirty-secondofficialadultlungandheart-lungtransplantationreport—2015;FocusTheme:earlyGraftfailure,JHeartLungTransplant34(10)(2015)1264–1277.
[26]J.
Wanger,J.
L.
Clausen,A.
Coates,etal.
,Standardisationofthemeasurementoflungvolumes,EurRespirJ26(3)(2005)511–522.
[27]S.
Stewart,M.
C.
Fishbein,G.
I.
Snell,etal.
,Revisionofthe1996workingformulationforthestandardizationofnomenclatureinthediagnosisoflungre-jection,JHeartLungTransplant26(12)(2007)1229–1242.
[28]D.
J.
Levine,A.
R.
Glanville,C.
Aboyoun,etal.
,Antibody-mediatedrejectionofthelung:aconsensusreportoftheInternationalSocietyforHeartandLungTransplantation,JHeartLungTransplant35(4)(2016)397–406.
[29]S.
Husain,M.
L.
Mooney,L.
Danziger-Isakov,etal.
,A2010workingformulationforthestandardizationofdefinitionsofinfectionsincardiothoracictransplantrecipients,JHeartLungTransplant30(4)(2011)361–374.