文档介绍：该【phase i clinical trials in adoptive t‐cell therapies 2021 sean m. devlin 】是由【妙玉】上传分享，文档一共【20】页，该文档可以免费在线阅读，需要了解更多关于【phase i clinical trials in adoptive t‐cell therapies 2021 sean m. devlin 】的内容，可以使用淘豆网的站内搜索功能，选择自己适合的文档，以下文字是截取该文章内的部分文字，如需要获得完整电子版，请下载此文档到您的设备，方便您编辑和打印。:.
Received:2July2020|Accepted:1March2021
DOI:
ORIGINALARTICLE
PhaseIclinicaltrialsinadoptiveT-celltherapies
SeanM. Devlin1Alexia Iasonos1John O’Quigley2
||
1DepartmentofEpidemiologyand
Biostatistics,MemorialSloanKetteringAbstract
CancerCenter,NewYork,NY,USAWedevelopthreeapproachestophaseIdosefindingde-
2DepartmentofStatisticalScience,
UniversityCollegeLondon,London,UK
addressaveryparticulardifficultyinthisclinicalsetting:
Correspondenceaninabilitytofullyadministerthedoseallocatedtosome
,-
CancerCenter,NewYork,NY,USA.
Email:******@
Fundinginformationislargelysimilar,andallofferanadvantageoverthecur-
NIH,Grant/AwardNumber:P30CA008748
bytheoreticalstudy,andweprovidesomenewresultson
thisapproach.
1|INTRODUCTION
Adoptivecelltherapyisatherapeuticareainoncologythathasseenconsiderabledevelopmentoverthelast
antigenreceptor(CAR).TheCARfocusesthespecificityoftheTcells,suchthatoncetheyareinfused
backintothepatient,theT-cellsidentifythetumourcellantigens,resultinginboththeeliminationofthe
tumourcellsandtheproliferationofCARandnon-CART-cells(June&Sadelain,2018).
TheCARselectedforacancerpatientdependsontheunderlyingmalignancyanditsassociated
tumour-,asCD19
isexpressedoncertainB-cell-associatedleukaemiasandlymphomas,andmultipleproductshavenow
gainedFDAapproval(June&Sadelain,2018).AdditionalCARtargetsarenowunderdevelopment
inotherbloodandsolidtumours,suchasthebb2121CARformultiplemyelomaandtheIL13Rα2
CARforglioblastoma(Brownet al.,2016;Rajeet al.,2019).TherearealsonewgenerationsofCAR
productsthataddco-,theseequateto
manyCART-cellclinicaltrialsinearlydrugdevelopment.
Unliketypicalchemotherapyagents,CARtherapyinvolvesacomplexandindividualizedmanu-
-cells,apatientundergoesleukapheresis,and
theTcellsarethenselected,activatedandtransducedwithatransgene(Feinset al.,2019).Thesub-
sequentstepexpandstheproductexvivotoachieveasetnumberofviableT-cells,whichdefinesthe
‘doselevel’,thisdoselevelcannotalwaysbeachieved
;00:1–©2021RoyalStatisticalSociety|1:.
2|DEVLINEtaL.
,earlyphaseclinicaltrialsneededtospecifyawindowor
rangearoundthetargetdoseandincludeevaluabilitycriteriaifthatwindow/rangeisnotachieved.
Forexample,anearlyphasetrialofCARTcellsindiffuselargeB-celllymphomasetatargetdoseof
5×108viablecellsbutadditionallysetaminimumallowabledoseof1×108;patientswhoreceived
,themediandosereceivedamongevaluable
×108viableT-cells(range:1to5×108cells)(Schusteret al.,2019).
Oncegenerated,-cellsinteractwiththe
underlyingtumourandactivate,aninflammatoryresponseoccursinsomepatients,deemedcytokine
releasesyndrome(CRS),whichclinicallymanifestsasflu-likesymptomsandcanleadtohypotension
andhypoxia(June&Sadelain,2018;Parket al.,2016).Typicallyeitherconcurrentorsubsequentto
CRS,immuneeffectorcell-associatedneurotoxicitysyndrome(ICANS)canoccurresultinginde-
creasedalertnessinthepatientbutultimatelymayprogress(Neelapu,2019).BothCRSandICANS
areassociatedwithT-cellexpansion,andworsetoxicityhasbeenobservedwithhigherinfuseddose
-limitingtoxicities(DLT)definition
inphaseIprotocolswhenidentifyingthemaximumtolerateddose(MTD)forsubsequentstudies.
CARphaseIstudiestypicallyinvestigatefourtofivedoselevels,andthemajorityoftrialsusethe
3 + 3designoramodificationofthe3 + 3designtoidentifyasafedose(Parket al.,2018;Rajeet al.,
2019).Auniquechallenge,however,inthesephaseIstudiesishowtohandlethescenariowhenthe
exvivoexpansionoftheT-
,such
thatpatientstreatedwith50–80%ofthedosearedeemedtohavereceivedthefullassigneddose(Lee
et al.,2015;Parket al.,2018;Rajeet al.,2019).Forfurtherreductionsinthegeneratedproduct,pro-
oftheseapproachesincludethefollowing:
‘inevalu-
able’aspartofthedoseescalationdesignandreplacewithanewpatientforthatspecific
cohort,
,or
(Parket al.,2018;Rajeet al.,2019)(examples:
NCT01583686;NCT00924326;NCT02215967).
Asinvestigatorsattempttofittheclinicalscenarioathandintoastandarddoseescalationdesign,
thesubsequentmodificationstothestudydesignandparticipantevaluabilitycriteriapresentpresent
,acoreassumptionisthatthetox-
icityassociatedwith50–80%oftheassignedcelldoseisthesameasifthepatienthadreceived100%
,asanexample,onestudyofCART-cellsshowedthatpatientswhoreceived
doseslowerthantheassigneddosehadlesshigh-gradeCRS(Bachanovaet al.,2019).Therefore,if
thisassumptionisnotmetinthephaseIstudy,adoselevelwithanunacceptableleveloftoxicitycould
,patientstreatedwiththeCART-cellsbut
deemedinevaluableinthedoseescalationdesignisacompletelossoftoxicityinformationandre-
placingthepatientinthetrialcomesataconsiderablefinancialcostandadelayinthestudyduration.
WhiletheproductioncostsofaCARproductduringdrugtestingisnotpubliclyavailable,thecostof
oneapprovedproductforleukaemiais$475,000,whichreflectsinpartthehighexpenseoftheman-
,replacingpatientsinthedoseescalationalgorithmisnotaviableoption.
Asanillustrationoftheproblem,thetoprowofFigure1showsasimulatedphaseItrialwhere
patientsareassignedtooneoffivedoselevels;however,somepatientsreceiveonlyafractionofthe:.
DEVLINEtaL.|3
observedinatrialandaccuratelyestimatethelocationoftheMTDamongtheassigneddoselevels.
Tothisend,thismanuscriptpresentsthreesolutionstoaccountfortheinformationobtainedfrom
ofO’Quigleyet al.(1990),andmodifyittoallowthetoxicityinformationforapatientwhoreceived
allowsallpatientstocontributeinformationtotheidentificationoftheMTDlocation.
|Examplesfromtwopublishedclinicaltrials
TwophaseIstudieswererecentlypublishedthatexaminedthesafetyofCART-
firststudyreportedthesafetyofCD19-specificCARTcellsinpatientswithacutelymphoblasticleu-
kaemia(Parket al.,2018).Thestudyexaminedthreedifferentdoselevels(1×107,3×107,1×108)
withonefallbacklevel(3×106)
utilizeda3 + 
receivedlessthan50%oftheplanneddosewouldbetreatedbutwouldberemovedandreplacedas
partoftheDLT-evaluablepopulation().
ThesecondphaseIstudyreportedthesafetyofB-cellmaturationantigen(BCMA)-targetingCAR
Tcellsforpatientswithmultiplemyeloma(Rajeet al.,2019).Thestudywasdesignedtoevaluatethe
safetyoffivedifferentdoses(5×107,15×107,45×107,80×107,120×107)withonefallbackdose
(×107).Thestudysimilarlyuseda3 + 
evaluableiftheyreceivedatleast80%ofthecelldose;whenlessthan80%ofthedosewasgenerated,
patientswouldstillreceivethecelldosebutwouldbereplacedforMTDdetermination.
Furtherdetailsaboutevaluabilityanddoseescalationareprovidedinthestudyprotocols;both
studiesprovidedtheirstudyprotocolassupplementalmaterialatthetimeofpublication.
2|METHODOLOGICALAPPROACH
Wedividethissectionintotwoparts:asubsectionthatdescribesthemaingoalsandtheoverallap-
proachandasubsectionthatfocusesonhowtoadaptthisapproachtothespecificproblemsraisedby
dosefindingforadoptivecelltherapies.
|Maingoalandoverallapproach
Ourgoalistoidentifythedoselevelamongkorderedlevels,d1,d2,…,dk,thathasatoxicityrate
,pairsofrandomvariables(Xj,Yj)
areobserved,wherethebinaryvariableYjisthedrug-relatedtoxicityforindividualj,equalto1if
adose-limitingtoxicity(DLT)isobservedaftertheindividualistreatedatXj=xj,ideallyoneof
,xj,andwere-
∈{d1,d2,…,dk}.Forcompleteness,weaddtothisset
thedosed0=,theprobabilitiesoftoxicity,Pr(Yj=1 Xj=xj),satisfy:
0<R(d1)<,…,<R(dk)<1,translatingtherequirementthatthefunctionR(·)isinvertibleonthe
interval(0,1).Morethanonedefinitionofthemaximumtolerateddose,dM( )ispossibleandthemost
commononeisthedoseclosesttosometarget,θ,specifically,:.
4|DEVLINEtaL.
AssignedReceived
DoseDose
00800
DLT
NoDLT
4008400
00200
DoseDose
1002100
5050
1510152015101520
AccruedPatientAccruedPatient
PosteriorProbabilityofTrueMTDPosteriorProbabilityofTrueMTD
First6PatientsFirst12Patients
.
.
EstimatedProbabilityofMTEstimatedProbabilityofMT
DoseLevelDoseLevel
PosteriorProbabilityofTrueMTDPosteriorProbabilityofTrueMTD
First16PatientsAll20Patients
.
.
EstimatedProbabilityofMTEstimatedProbabilityofMT
DoseLevelDoseLevel:.
DEVLINEtaL.|5
-limitingtoxicities
(DLTs)
probabilitiesthateachdoselevelisthetrueMTDafterthefirst6,12,16,and20thpatientaccruedusingFDLA-CRM.
Thetruetoxicityrateisshownatthetopofeachposteriorprobabilityplot(ToxicityScenario2)
dM( )=argmin R(di)− .(1)
i
ThefunctionR(di),i=1,…,k,isnotknownandwereplaceitbyaworkingmodel (di,a)whereais
(x, a)torespectcertainminimalproperties,describedinseveralpapers,and
,supposewehaveconsistentestimates (di,⋅),ofR(di)then,asa
consequencewehavearunningestimateofdM( )givenby:
d̃M( )=argmin (di,⋅)− .(2)
i
Thisisthebasicideaunderlyingmodel-baseddesigns,theCRMinparticular,wheretheoverallgoalofthe
studyandthegoalofprovidingthebesttreatmentforeachenteredpatientarereflectedinthesetwoequations.
Whilewefollowthisbasicidea,itturnsoutthatwecanrelaxtheassumptionthatourmodelgener-
atestheobservations—astrongassumptionthatwouldprovideconsistency—aslongastheconditions
.
TheobjectivethenistoestimatetheMTDasaccuratelyaspossiblewhilemaximizingthenumber
={(x ,y ), =1,…,j}denotethedose-
dosethattheyweremeanttoreceive,thelikelihoodcanbewrittenas
j
L(a:Ω)= y (x,a){1− (x,a)}(1−y ).(3)
j
=1
SomeauthorspreferaBayesianapproach,bringingonboardapriordensityg(a),,the
posteriordistributionfortheparametera,