Scoring System Based on Post-Transplant Complications in Patients after Allogeneic Hematopoietic Cell Transplantation for Myelodysplastic Syndrome: A Study from the SFGM-TC

(1)

Original

article

Scoring

System

Based

on

Post-Transplant

Complications

in

Patients

after

Allogeneic

Hematopoietic

Cell

Transplantation

for

Myelodysplastic

Syndrome:

A

Study

from

the

SFGM-TC

Alexis

Caulier

a

,

Elodie

Drumez

b

,

Jordan

Gauthier

c

,

Marie

Robin

d

,

Didier

Blaise

e

,

Yves

Beguin

f

,

Mauricette

Michallet

g

,

Patrice

Chevallier

h

,

Jacques-Olivier

Bay

i

,

Stéphane

Vigouroux

j

,

Yohan

Desbrosses

k

,

Jérôme

Cornillon

l

,

Stéphanie

Nguyen

m

,

Charles

Dauriac

n

,

Régis

Peffault

de

Latour

d

,

Bruno

Lioure

o

,

Pierre-Simon

Rohrlich

p

,

Martin

Carré

q

_,

_Jean-Henri

_Bourhis

r

_,

_Anne

_Huynh

s

_,

_Felipe

_Suarez

t

_,

_Federico

_Garnier

u

_,

Alain

Duhamel

b

,

Ibrahim

Yakoub-Agha

c,

*

a_{Hématologie,}_Centre_Hospitalier_{Universitaire}_(CHU)_Sud,_Amiens,_France

b_Univ._Lille,_CHU_Lille,_EA₂₆₉₄_–_Santé_publique:_{épidémiologie}_et_qualité_des_soins,_Unité_de_{biostatistique,}_F-59000_Lille,_France c

CHUdeLille,LIRIC,INSERMU995,UniversitédeLille,59000Lille,France

d

Hématologie-Transplantation,AP-HP,HôpitalSaintLouis,UniversitéParis7,Paris,France

e

Hématologie,InstitutPaoli-Calmettes,Marseille,France

f

Hématologie,UniversityofLiège,Belgium

g_{Hématologie,}_CHU_Lyon_Sud,_Lyon,_France h_{Hématologie,}_CHU,_Nantes,_France i

Hématologie,CHU,ClermontFerrand,France

j

Hématologie,CHU,Bordeaux,France

k

Hématologie,CHU,Besançon,France

l

Hématologie,InstitutdeCancérologiedelaLoire,Saint-Etienne,France

m

Hématologie,HôpitaldelaPitié-Salpêtrière,UniversitéParis6,Paris,France

n_{Hématologie,}_CHU,_Rennes,_France o_{Hématologie,}_CHU,_Strasbourg,_France p

Hématologie,CHU,Nice,France

q

Hématologie,CHU,Grenoble,France

r

InstitutGustaveRoussy,Villejuif,France

s

Hématologie,CHUPurpan,Toulouse,France

t

Hématologieadulte,AP-HP,HôpitalNecker,UniversitéParis5,Paris,France

u_Agence_de_la_{biomédecine,}_Saint_Denis,_France

ARTICLE INFO Articlehistory: Received27July2018 Accepted20August2018 Availableonline8September2018 Keywords:

myelodysplasticsyndrome prognosticscoringsystem earlypost-transplantcomplications allogeneichematopoieticcell transplantation

ABSTRACT

Purpose: We developed a prognosticscoring system to evaluate the prognosis of myelodysplastic

syndrome(MDS)patientssurvivingmorethan100daysallogeneichematopoieticcelltransplantation

after(allo-HCT).

Patientsandmethods:Weperformedalandmarkanalysisonaderivationcohortof393casestoidentify

prognosticfactorsfor3-yearoverallsurvival.Potentialpredictorvariablesincludeddemographicand

clinicaldata,transplantationmodalitiesandearlypost-transplantcomplications.Thescoringsystemwas

testedagainstavalidationcohortwhichincluded391patients.

Results:Complicationsoccurringbeforeday100suchasrelapse[HR=6.7;95%CI,4.5-10.0](4points),lack

ofplateletrecovery[HR,3.6;95%CI,2.2-5.8](2points),grade-IIacuteGVHD[HR=1.7;95%CI,1.2-2.5](1

point)andgrade-III/IV[HR=2.6;95%CI,1.8-3.8](2points)weretheonlyindependentpredictorsof

3-yearOS.

The3-yearOSassociatedwithlow(0),intermediate(1-3)andhigh(4)riskscoreswasrespectively70%,

46%and6%.Themodelperformedconsistentlyinbothcohorts,withgoodcalibration.

Conclusion:Thispost-transplantscoringsystemisapowerfulpredictorofoutcomeafterallo-HCTfor

*Correspondingauthorat:UAMallogreffesdeCSH,CHRULille,F-59037LilleCEDEX,France. E-mailaddress:ibrahim.yakoubagha@chru-lille.fr(I.Yakoub-Agha).

http://dx.doi.org/10.1016/j.retram.2018.08.003

Available

online

at

ScienceDirect

(2)

MDS,andcanprovideusefulguidanceforclinicians.Additionalstudiesarerequiredtoevaluatethis

scoringsystemforotherhematologicmalignancies.

INTRODUCTION

Allogeneichematopoieticcell transplantation (allo-HCT)is a therapeuticoptionforhigh-riskhematologicaldisordersincluding myelodysplastic syndrome (MDS) [1,2]. Nevertheless, this ap-proach is still associatedwith potentiallylife-threatening com-plicationssuchasconditioning-relatedtoxicity,graft-versus-host disease(GVHD),poorgraftfunctionandrelapse[3,4].

A key issue is managing patients with post-transplant complications.Inaddition,thereisaneedforaclinicaltoolthat providesmorecost-effectiveuseofmedicalresourcesby stream-liningtheselectionprocessofthosepatientswhowilltrulybene_ﬁt fromcurativecare,includingintensivecare.

While most published scoring systems have focused on prognosticvariablesdeterminedbeforeallo-HCT[5–7],thereare limited data regarding post-transplant prognostic evaluation, particularlywithinaspeciﬁcdiseasetype.

Wehypothesizedthatamorecomplexmodelincludingearly post-transplant events might be a more accurate predictor of. survivalforthosepatientswhohavealreadybeentransplanted. Therefore, wedeveloped a prognostic scoringsystem based on homogenouscohortsof transplantedMDSpatientswhichtakes intoaccountdiseasecharacteristicsandtransplantationmodalities aswellasearlypost-transplantcomplications.

PATIENTSANDMETHODS

This multicenter study was approved by the board of the FrancophoneSocietyofBoneMarrowTransplantationandCellular Therapy(SFGM-TC), andwas conductedin accordancewiththe DeclarationofHelsinki.Writtenconsenttousemedicallyrelevant dataforresearchpurposeswasobtainedfromeachpatientand donorbeforetransplant.Toensureconﬁdentiality,eachcasewas anonymizedbyassignmentofarandomidentiﬁcationnumber. DataSource

The SFGM-TC database(ProMISe) was used to retrievedata frompatientswhounderwentallo-HCTforMDS.Inadditiontoan independentmonitoringdouble-check,thequalityofthedatawas controlledviaacomputerizedsearchfordiscrepancyerrorsand vigorouson-sitedataveriﬁcationofeachﬁle.HLAmatchingwas cross-checkedwiththedataoftheFrenchBoneMarrowDonor Registryaspreviouslydescribed[8,9].

PatientSelection

Twodistinct cohortsof patientswereused for this study, a derivationcohortandavalidationcohort,withmatchinginclusion andexclusioncriteria.

The derivation dataset came from the derivation cohort of patients(n=461)treatedforMDSwithallo-HCTfromJanuary1999 toDecember2009.

Similarly,thevalidationdatasetincludedalldatafrompatients (n=451) treated for MDS with allo-HCT from January 2010 to December2013.

For inter-cohort homogeneity purposes, we only included patientswho received a ﬁrst allo-HCT. Stem cell sources were eitherbonemarroworperipheralbloodstemcells(PBSC)froma siblingorHLA-matched(so-called10/10)unrelateddonor.Those

who received allo-HCT from an HLA-mismatched or haplo-identicaldonorwereexcluded.Nopatientreceivedanex-vivo T-cell depleted graft. Patients with chronic myelo-monocytic leukemiawerealsoexcluded.

Toavoidthecompetingriskofearlydeath,patientswhodied beforeday100wereexcludedfromthederivation(n=68,14.8%) andvalidation(n=60,13.3%)cohorts.

STATISTICALANALYSIS

Patientsanddonorcharacteristics,transplantationmodalities and early complications that occurred within 100 days post-transplantweretakenintoaccount.Inclusionofavariableinthe predictive modelwasbasedonclinicalrelevance.Thefollowing variables were chosen: age at time of transplant; IPSS and cytogeneticscore[10,11];diseasestatusattransplant(responder vs. non responder) according to IWG 2006 criteria [12]; bone marrowblastcountat transplant(<5% vs.5%), HLAmatching (siblingvs.HLA-matchedunrelateddonor);sexmismatchdefined asmalepatientreceivinggraftfromafemale;CMVpositivestatus for donor and recipient; use of anti-thymocyteglobulin (ATG); source of stem cells (bone marrow vs. PBSC); intensity of the conditioningregimenasdefinedbyBacigalupoetal[13];gradeof acute GVHD (0/I, II, III/IV) within 100 days post-transplant accordingtostandardcriteria[14];earlyrelapsebeforeday100 post-transplant; lack of platelet recovery [15]. The latter was definedbyoccurrenceoftransfusion-independentplateletcount 20G/L forthe7consecutive daysleadinguptoday100 post-transplantevenifplateletcountmayhavethendecreasedatsome point.

Quantitative variables are expressed as means (standard deviation) if the distribution was normal and as medians (interquartile range)ifthedistributionisotherwise.Categorical variablesareexpressedasfrequenciesandpercentages.Normality of distribution was checked graphically and with use of the Shapiro–Wilktest.UsingtheKaplan-Meiermethod,weestimated OSfrom100daysaftertransplant,thelandmarktimepoint.

Before developing the multivariable prognostic model with Cox’s proportional hazard regression model, the log-linearity assumptionforagewasassessedusingMartingaleresidualplots andtherestrictedcubicsplinefunctions[16].Sincenoevidenceof anon-log-linearrelationshipwasshown,agewasintroducedasa linearterminthemultivariablemodel.Theproportionalhazards assumptionwasassessedforeachcandidatepredictorbyplotting theSchoenfeldresidualsagainsttherankofsurvivaltime[17].

IrrespectiveoftheirunivariateassociationwithOS,allvariables wereenteredintothemultivariableprognosticregressionmodel. Missingdataforindividualvariablesrangedfrom0to9.9%(1.7%of missing data points)leading to86.0% of patientswith nodata missing data(ie.completedata)inthemultivariableprognostic regression model. To prevent case deletion in multivariable analysis,missingdatawereimputedbysimpleimputationusing a regression-switching approach (chained equations using all variables; death status and survival times introduced after logarithmic transformation) with predictive mean-matching methodforthefollowing:continuousvariables;logisticregression modelforbinaryvariables;ordinallogisticregressionforordered categoricalvariables[18].

After a backward selection procedure, variables with low prognosticvalue(P>0.05)werenotkeptintheﬁnalCoxmodel.

(3)

Theresultsofsigniﬁcantpredictorswerereportedashazardratios (HRs)and95%conﬁdenceintervals(CI).Theproportionalhazard assumptionwasalsoassessedbyplottingtheSchoenfeldresiduals fortheprognosticindexderivedfromtheselectedmodelagainst

the rank of survival time. The performance of this model was assessedthroughitscalibrationanddiscrimination.Inparticular, calibration(i.e.thepredicted-to-observedsurvivalagreement)was evaluatedbycomparingthepredictedmeansurvivalcurvestothe

Table1

Patientanddiseasecharacteristicsattransplantandearlypost-transplantcomplicationsobservedinderivationandvalidationcohorts. Derivationcohort

(n=393)

Validationcohort (n=391) Characteristicsofthetransplant

Recipientagea_,_years,_mean_SD _52.0_10.8 _56.6_9.9 Recipientsex,n(%) Male Female 244(62.1) 149(37.9) 230(58.8) 161(41.2) Sexmismatcha,b

90(22.9) 83(21.2) FAB/WHOcategory,n(%) RA/RARS/RCMD RAEB-1 RAEB-2 RAEB-t/AML unclassiﬁed 93(23.7) 12(3.1) 253(64.4) 32(8.1) 3(0.7) 57(14.6) 2(0.5) 285(72.9) 9(2.3) 38(9.7) IPSSscorea ,n(%) Low/intermediate-1 Intermediate-2/high 204(51.9) 189(48.1) 138(37.0) 235(63.0) IPSSCytogeneticriskscore,n(%)

Favorable Intermediate Highrisk 213(55.3) 90(23.4) 82(21.30) 188(48.8) 81(21.1) 116(30.1) Pretransplantationtherapy,n(%)

Hypomethylatingagentsonly Chemotherapyonly Both None 43(12.0) 154(43.0) 14(3.9) 147(41.1) 156(39.9) 89(22.8) 59(15.1) 87(22.2) Diseasestatusattransplanta

,n(%) Responder Noresponder 214(54.5) 179(45.5) 206(52.7) 185(47.3) Bonemarrowblastcounta_,_(%)

<5% 5% 156(39.7) 237(60.3) 119(34.5) 226(65.5) HLAmatchinga ,n(%) No Yes 247(62.8) 146(37.2) 203(51.9) 188(48.1) DonorCMVstatusa ,n(%) Positive Negative 196(49.9) 197(50.1) 172(44.3) 216(55.7) RecipientCMVstatusa ,n(%) Positive Negative 213(54.2) 180(45.8) 199(50.9) 192(49.1) Useofanti-lymphocyteseruma

,n(%) No Yes 212(53.9) 181(46.1) 90(23.0) 301(77.0) Stem-cellsourcea_,_n(%) Marrow PBSC 119(30.3) 274(69.7) 59(15.1) 332(84.9) Conditioningregimena ,n(%) RIC MAC 145(36.9) 248(63.1) 109(27.9) 282(72.1) Totalbodyirradiation,n(%)

No Yes 254(64.8) 129(35.2) 326(83.8) 63(16.2) Earlypost-transplantcomplications

GradeofacuteGVHDa ,n(%) 0/I II III/IV 243(61.8) 92(23.4) 58(14.8) 293(74.9) 60(15.4) 38(9.7) Lackofplateletrecoverybeforeday100a

,n(%) No Yes 368(93.6) 25(6.4) 365(93.6) 25(6.4) Relapsebeforeday100a

,n(%) No Yes 358(91.1) 35(8.9) 354(90.8) 36(9.2)

Abbreviations:RA,refractoryanemia;RARS,RAwithringsiderobalsts;RCMD,Refractorycytopeniawithmultilineagedysplasia;RAEB,RAwithexcessofblasts;RAEB-t,REAB intransformation<30%ofmarrowblasts;HLA,humanleukocyteantigen;CMV,cytomegalovirus;PBSC,peripheralbloodstemcells;ATG,anti-thymocyteglobulin;TBI,total bodyirradiation;GVHD,graft-versus-hostdisease;d100,day100.

a

Indicatesselectedvariablesforentranceintothemultivariableprognosticregressionmodel.

b _Deﬁned

(4)

Kaplan-Meiersurvivalcurvesofthreespeciﬁcriskgroupswhich weredeterminedbytheprognosticindex’sdistribution.

Discrimination was assessed via the Harrell’s c-index of agreement [16], which indicates to what extent the model distinguishesbetweenpatientswhowilldiefromthosewhowill survive.Thisc-statisticisexpectedtorangefrom0.60to0.85for survivaldata[19].Thepredictedoverallsurvivalprobabilitieswere derivedfromthebaselinesurvivalestimatesatthemeanvaluesof selectedvariables.Toaccountforthereportedoverestimationof regressioncoefﬁcientsinprognosismodelsderivedfrom multivari-ableregressionanalysis[16],weperformedaninternalvalidationby usingbootstrap resamplingwith200repetitionsto estimatethe shrinkagefactorandthec-statisticcorrectedforover-optimism.

For clinical purposes, a point-scoring model (e.g. a risk assessmenttool)wascreatedbyassigningonepointtothelowest hazardratio(theconstantfactorthatreﬂect1pointinﬁnalpoint system) and by weighing every other one to the nearest approximationof the ratio of its HR withconstant factor. The predictiveaccuracyofthepoint-scoringmodelwasassessedbythe previouslymentionedmethod.

For the external validation, calibration and discrimination performanceswereassessedfor continuousand discreet point-score models in the validation dataset. The predicted survival probabilitiescalculatedwithinthevalidationdatasetcame from thecoefﬁcientestimates(afterapplyingtheshrinkagefactor)and thebaselinesurvivalestimatefromthederivationdataset.

Statisticaltestingwasdoneatthetwo-tailedαlevelof0.05. DatawereanalyzedusingtheSASsoftwarepackage,release9.4 (SASInstitute,Cary,NC).

RESULTS Database

Thederivationcohortincluded393patientsandthevalidation cohort included 391 patients. The median follow-up from transplantation respectively for the derivation and validation cohortwas3.8years(range,0.3to11.8years)and2.9years(range, 0.4to5.5years).Table1reportsbaselinecharacteristicsandearly post-transplantcomplicationsofbothcohorts.Meanrecipientage was52.010.8yearsinderivationcohort,and56.69.9yearsin validation cohort. In the derivation and validation cohort, respectively 37.9% and 41.2% of recipients were female. Main differences were higher incidence of intermediate2/high IPSS score, broader use of ATG, increased pre-transplant treatment usinghypomethylatingagents,andless-intensiveconditioningin thevalidationcohort.Inthederivationcohort, 173deathsoccurred, and145deathsoccurredinthevalidationcohort.Mean3-yearOS was56%forthederivationcohortand63%forthevalidationcohort. PrognosticScoringModel

Univariateassociationsofclinicallyrelevantpredictorswith 3-yearOSin thederivationcohort areavailable inSupplementary TableS1.

After a stepwise-backward selection procedure, the only independentriskfactorsof3-yearOSwerethefollowing:

1GradeofacuteGVHD(HRforgradeII:1.7;95%CI:1.2-2.5;HRfor III/IV:2.6;95%CI:1.8to3.8using0/Iasreference).

2Lackofplateletrecoverybeforeday100(HR:3.6;95%CI,2.2 -5.8).

3Relapsebeforeday100(HR:6.7;95%CI:4.5-10.0).

SeeTable1forobservedoccurrenceofthesethree complica-tionsamongpatientsinbothcohorts.

Internalvalidation

Therewasnodeviationfromproportionalhazardassumptions for theprognosticindexderivedfromtheselectedmodel.After correctionforoveroptimism,thediscriminationoftheprognostic scoringmodelwas0.67(95%CI:0.63-0.71)withashrinkagefactor of0.903.Calibrationat1to3yearsfromtransplantationwasgood inallgroups,implyingthatthepredictedsurvivalforeachgroup wereclosetotheobservedones(Fig.1,Table3,andSupplementary Table S2). Formulae and coefﬁcients to estimate 1 to 3-year survivalprobabilityaredetailedintheappendix.

ExternalValidation

Thediscriminationvalueinthevalidationdatasetwas0.65(95% CI: 0.61-0.69). After the shrinkage of the coefﬁcients, the calibration stayed satisfactory despite a slight underestimation observedinthehighestriskgroup:21.9vs9.7for3-yearOS(see SupplementaryTableS2).

Post-TransplantScoringSystem(PTSS)

Asshown inTable2,grade ofacuteGVHD,lack of platelet recovery and relapse within 100 days after transplant were respectivelyassigned1,2and4points.Thepointscoremodel rangedfrom0-8,discriminatinglow(0),intermediate(1-3), andhigh-risk(4-8)patients,accordingtosurvivalprognosis. Theobserved3-yearOSaftertransplantationinpatients with low,intermediateandhighscoreswasaccordingly70%(95%CI: 63-76%),46%(95%CI,38-55%)and6%(95%CI,2-16%).Inboth datasets, discrimination and calibration were satisfactory (Table3).

HowtoUsethePTSS:ProposalofanAlgorithm

Patients requiring intense care due to early post-transplant complicationsarealmostsystematicallytransferred[15].However,

Fig.1.Calibrationofsurvivalprobabilities.Thecontinuousprognosticmodelinthe derivationcohorthighlightedlow(0points),intermediate(1-3points)andhigh(> 3points)riskscoreaccordingtocorrespondingsurvivalprobability.Originofthe landmarkanalysisisday100.

(5)

thedecisiontotransferthosedevelopingcomplicationsafterday 100post-transplantismoredifﬁculttomake.

AsshowninFig.2,thePTSStogetherwithotherscoringsystems suchastheSepsis-RelatedOrganFailureAssessment(SOFA)[20] canfacilitatethe decision-making processbefore transferring a patient to the ICU. According to the expected 3-year survival observed with the PTSS, patients with a low score should be transferredregardless oftheircurrent complications,and those withahighscorearelesslikelytobenefitfromintensivecare.Prior tomakingafinaldecisionregardingspecificICUcare,patientswith anintermediate PTSS score must undergo additional screening based on standard ICU scores [20–23]. Therefore, they can be transferredtotheICUiftheirSOFAscore,forinstance,isA,BorC (lessthan10points).

DISCUSSION

Toourknowledge,thisistheﬁrststudytoproposeaclinicaltool generatedfromvariablesevaluatedafterallo-HCT.Thistoolallows forearly post-transplantprognosticassessment inpatientswith MDS.

Several approaches have been used to attempt to predict patientoutcomeafterallo-HCT.Oneofthesescoringapproachesis toapplywell-establisheddisease-specificprognosticscoresatthe timeoftransplant[1,5,6],whichposessignificantlimitations.By definition,patientseligibleforallo-HCTareusuallyclassifiedas “high-risk”perthegivenscoringsystem.While thesescoresare

helpfulfordecision-makingintermsoftransplant,theyoverlook keytransplant-relatedvariables[6,7].

Themainbodyofliteratureinvestigatingprognosticfactors assessedafterhematopoieticcelltransplantwaspublishedby the intensivist community. While of interest for patients admittedtotheICU, suchstudiesfailtoincludekey prognos-ticators such as disease characteristics and transplantation modalities[20,24–26].ForMDSinparticular,otherauthorshave combineddisease-speciﬁcscores(R-IPSS)withothervariables assessedatthetimeoftransplant(HCT-CI)toassessprognosis afterallo-HCT[5].

Inthisstudy,wecomputedacomprehensivesetofdataintoa multivariableCoxmodelencompassingpatientcharacteristicsas well as detailed transplantation modalities. In contrast with previouslyreportedmodels[7–11],wealsoincludedcrucialearly post-transplantcomplications usinga landmark analysisatday 100.Notably,weconsideredIPSSscoreratherthanIPSS-Revised scoreforbaselinediseasecharacteristics,asrecommendedbythe EuropeanLeukemiaNet[27]. Inaddition,IPSS-Rdo notseemto

Table2

Scoringsystemofmortalityriskaccordingtoearlypost-transplantcomplications. Predictors Points

0 1 2 4

GradeofacuteGVHD

Lackofplateletrecoverybeforeday100 Relapsebeforeday100

0/I No No II -III/IV Yes -Yes Riskgroups Scorea

Low Intermediate High 0 1-3 4-8 GVHD=graft-versus-hostdisease. a

Score:additionpointsofthe3predictorsdescribedabove.

Table3

Calibrationanddiscriminationofpoint-scoremodel.

Derivationcohort Validationcohort

Death/N Observed Predicted Death/N Observed Predicted 1-yearsurvival Riskgroup Low(0) 37/207 0.820 0.832 31/244 0.871 0.812 Intermediate(1to3) 38/142 0.727 0.686 36/109 0.669 0.655 High(4to8) 37/44 0.159 0.202 23/38 0.395 0.250 2-yearsurvival Riskgroup Low(0) 52/207 0.744 0.745 61/244 0.738 0.715 Intermediate(1to3) 63/142 0.544 0.550 45/109 0.576 0.510 High(4to8) 39/44 0.109 0.087 27/38 0.283 0.114 3-yearsurvival Riskgroup Low(0) 60/207 0.697 0.692 68/244 0.697 0.658 Intermediate(1to3) 72/142 0.463 0.475 48/109 0.537 0.434 High(4to8) 41/44 0.055 0.050 29/38 0.219 0.069 C-statistic(95%CI) 0.67(0.63-0.71)a 0.65(0.61-0.69) ObservedvalueswerecalculatedusingKaplan–Meierestimates.

Inthederivationdataset,predictedestimatesineachriskgroupwerecalculatedasthemeanpredictedprobabilitiesbytheCoxregressionmodel(usingscoreascontinuous predictor).Inthevalidationdataset,predictedestimatesuseshrinkagefactorbasedonbootstrapvalidationinthederivationdataset(seemethodandappendixformore information).

a_C-statistic_corrected_for_{over-optimism.}

Fig.2. ProposalofanalgorithmforpatientsrequiringtransfertoICUafterday100 post-transplant.(*examplesofICUscoringsystems20,24,25

(6)

change the outcome after allo-HCT compared to IPSS [28]. Interestingly,thesepost-transplantcomplicationswerefoundto betheonlyindependentriskfactorsassociatedwithdecreased 3-yearOS,whereaspre-transplantparameterswerenotfoundtobe independentriskfactors.Althoughpre-transplantfactorssuchas diseasecharacteristicsandtransplantationmodalitiesplayamajor role inpost-transplantsurvival, earlypost-transplant complica-tionsappearedtooverridetheprognosticimpactofpre-transplant parameters.

Strikingly,lackofplateletrecoveryatday100wasastrongand independent predictor of low OS. The prognostic impact of impairedplateletrecoveryhasalreadybeendescribedbyothers [29,30], but these studies observed heterogeneous cohorts in termsofdiseasetypeandtransplantmodalities.Therefore,thisis thefirstclearreportofthenegativeandindependentimpactof poorplateletrecoveryonlong-termsurvival.Onlyafewpatients whodidnotrecovertheirplateletcountbeforeday100showed earlyrelapse:6/25inderivationcohort;4/25invalidationcohort. Regardingthis, absence of durable platelet recovery could also reflect poor graft function, when not reflecting viral/bacterial infectionsorlinkedtoGVHD[4,31,32].Predictably,relapsebefore day100 along withgrade 2-4acute GVHD adverselyimpacted survivalinourmodel[33].

Inlightoftherisingnumberofpatientsundergoingallo-HCT, the decision-making process of whether a patient should be treatedwithcurativeorpalliativeintentoughttobeguidedby robuststatisticalmodels.Despitetherecentchangesinallo-HCT management,expectedandobservedratesoverlappedwithour prognostic model when applied to a more recent cohort of patients(ie.validationcohort).Infact,olderpatientscouldstill beneﬁtfromallo-HCT,withmoreseverediseasehighlightedbya higher-risk IPSS score. The recent growing use of PBSC as a transplantsourceand,therefore,theprophylacticuseofATGfor GVHD [34] partly explain these differences. Interestingly, we observedthatmorepatientsreceivedamyeloablative condition-ing(MAC)regimeninthevalidationcohort,despiteofincreased ageoftransplantation.

For daily use, the PTSS, a clinical risk assessment tool was elaborated, stratifying patients with a score according to the followinglevels of mortality risk:low (0),moderate(1-3) and high(4).

The PTSS score takes into account scores relative to the followingvariables:acuteGVHDgrade;lackofplateletrecovery beforeday 100; relapse before day100. This proposed clinical scoringtoolcouldprovideforamorecost-effectiveuseofmedical resources because it aims to more appropriately select those patients who will truly beneﬁt from curative care, including intensivecare.

Dramatically,earlyrelapsesystematicallydrivespatienttothe high-riskgroupasopposedtotheoccurrenceofacuteGvHD,which more weakly affects prognosis. Nevertheless, a prospective evaluation is needed to prove the cost-effectiveness of such a strategy.

According to recommendations from the SFGM-TC, patients requiringintensivecareduring theearlypost-transplantperiod shouldbetransferredtotheICUregardlessoftheircomplication severity[15].

ThePTSScanbehelpfulinmakingthedecisiontoproceedto intensivecareforpatientssurvivingmorethan100daysafter allo-HCT.Indeed,itseemsdifﬁculttorefusetotransferapatientwitha lowPTSS-score(estimated3-yearOSof70%).Incontrast,with6% ofestimated3-yearOS,patientswithahighscorearelesslikelyto beneﬁtfromICUadmission.Screeningsystemsbasedonstandard ICUscores20,24,25_would_be_more_helpful_when_it_comes_to_patients with intermediate PTSS-score (estimated 3-year OS of 46%). Nevertheless,suchhypothesisshouldbefurthervalidatedinan

independentprospectivecohortofpatientsrequiringtreatmentin intensivecaresettings.

Oneofthestrengthsofthisstudywastheclear-cutdeﬁnitionof thepredictorvariables,whichrelevanceareinlinewithprevious studies. An additional strength is the simplicity of use of our prognostic assessment tool, suitable for clinical practice and bridging the gap between scores typically used by onco-hematologistsandthoseappliedintheICU.

Due toprospective data collection in theProMISedatabase, some variables were not included in the model, such as comorbidity evaluation,developmentof infection, acutekidney injury,orsinusoidalocclusionsyndrome.However,occurrenceof such complications and severe comorbidities were considered beforehand with regards to conditioning intensity, which was includedinourmultivariablemodel.

Finally,wewouldliketomentionthatanincreasingamountof attentionisbeingfocusedonmoleculardatainMDS.Duetothe inclusionyears,suchdatawerenotavailable inourdatabase.If moleculardataaresuggestedtoimprovetheriskstratiﬁcationof MDSpatients[35],suchanapproachhasyettobeprospectively validatedinhematopoietictransplantationprocedures.Givenour aim todevelopa simple and usable clinicalscore, addingsuch speciﬁc data did not appear relevant to us. Extending our prognosticscore topatientswho underwent allo-HCTforother hematological disorders, malignant or not, or other types of transplantisanencouragingpotentialnextstepthatwillideally requireprospectivevalidationinhomogenouscohorts.

CONCLUSION

Inthisstudy,wecreatedandvalidatedtheﬁrstprognosticscore basedonearlypost-transplantcomplicationstoquicklyandsimply estimatethesurvivalprobabilityofmyelodysplasticpatientswho survivedmorethan100daysafterallo-HCT.Ourﬁndingssupport the robustness, the reliability and the reproducibility of this scoringsystem.Thepotentialextensionofourscoringsystem,to patients withother hematologicmalignancies willbe the next phaseforthePTSSwepropose.Thisnextstepwillrequirefurther prospectiveevaluation.

AuthorsContribution

AC,ED,ADandIYAdesignedthestudy,AC,ED,AD,IYAreviewed thedata,analyzedresults,andmadetheﬁgures.Allauthorswrote andapprovedthemanuscript.

Conﬂictofinterest

Allauthorsdeclarenocompetingﬁnancialinterests.

Thisstudywas presentedinpartattheAmericanSociety of Hematology(SanDiego,2016).

ResearchSupport

This study was supported in part by a research grant from AssociationCapucine.

Acknowledgments

TheauthorswouldliketothankMathildeCuvelierforcollecting and checking data on-site. Special thanks to Nicole Raus, the SFGM-TCdatamanager.Theauthorswishtoacknowledgeallour medical teams for their dedication and involvement in daily patient care. A special thanks toDr. LauraRavasi for her help writingandproofreadingthisarticle.IbrahimYakoub-Aghawould liketothankAssociationCapucinefortheirsupportinhisresearch.

(7)

AppendixA.Calculatingthe1,2and3-yearoverallsurvival usingthecontinuousprognosticmodel

Step1:CalculateS

S=[-1.2706*(Plateletrecovery)+0.5251*(acuteGVHDII)+ 0.9592*(acuteGVHDIII/IV)+1.9011*(Relapse)]*0.9027

Where

(1)Plateletrecovery=

a 0.0636ifthepatientrecoverstheirplateletcountbeforeday 100

b 0.9364ifthepatientdoesn_’trecovertheirplateletcount beforeday100

(2)AcuteGVHDII=

a 0.7659ifthepatienthasgradeIIacuteGVHD

b 0.2341ifthepatientdoesnothavegradeIIacuteGVHD (3)AcuteGVHDIII/IV=

a 0.8524ifthepatienthasgradeI/IVacuteGVHD

b 0.1476ifthepatientdoesnothavegradeI/IVacuteGVHD (4)Relapse=

a 0.9109ifthepatientrelapsesbeforeday100 b 0.0891ifthepatientdoesn’trelapsebeforeday100 (5) 1.2706,0.5251,0.9592and 1.9011=regression coefﬁcients

estimatedforthederivationdataset

(6)0.9027 = the shrinkage factor estimated using bootstrap validationforthederivationdataset

Step2:CalculatesurvivalprobabilityusingS 1-yearsurvivalprobability=100*0.7374exp(S) 2-yearsurvivalprobability=100*0.6130exp(S) 3-yearsurvivalprobability=100*0.5426exp(S)

Where0.7374,0.6130and 0.5426arethesurvivalrate atthe meanvaluesofpredictorsinderivationdataset.

Example:apatientwhorecoversplateletcount,relapsesbefore 100daysandhasgrade0acuteGVHD.

Step1:CalculateS=[ 1.2706*(0.0636)+0.5251*( 0.2341)+ 0.9592*( 0.1476)+1.9011*(0.9109)]*0.9027=1.2515

Step2:CalculatesurvivalprobabilityusingS

1-yearsurvivalprobability=100*0.7374exp(1.2515)₌_34.5% 2-yearsurvivalprobability=100*0.6130exp(1.2515)₌_18.1% 3-yearsurvivalprobability=100*0.5426exp(1.2515)=11.8% AppendixB.Supplementarydata

Supplementarymaterialrelatedtothisarticlecanbefound,in the online version, at doi:https://doi.org/10.1016/j. retram.2018.08.003.

REFERENCES

[1]Garcia-ManeroG..Myelodysplasticsyndromes:2015Updateondiagnosis, risk-stratiﬁcationandmanagement.AmJHematol.2015;90(September (9)):831–41.

[2]Yakoub-AghaI,deLaSalmonièreP,RibaudP,SuttonL,WattelE,KuentzM,etal. Allogeneicbonemarrowtransplantationfortherapy-relatedmyelodysplastic syndromeandacutemyeloidleukemia:along-termstudyof70 patients-reportoftheFrenchsocietyofbonemarrowtransplantation.JClinOncolOffJ AmSocClinOncol.2000;18(March(5)):963–71.

[3]GooleyTA,ChienJW,PergamSA,HingoraniS,SorrorML,BoeckhM,etal. Reducedmortalityafterallogeneichematopoietic-celltransplantation.NEngl JMed.2010;363(November(22)):2091–101.

[4]HoranJT,LoganBR,Agovi-JohnsonM-A,LazarusHM,BacigalupoAA,BallenKK, etal.Reducingtheriskfortransplantation-relatedmortalityafterallogeneic hematopoieticcelltransplantation:howmuchprogresshasbeenmade?JClin OncolOffJAmSocClinOncol.2011;29(March(7)):805–13.

[5]Della PortaMG,AlessandrinoEP,BacigalupoA,vanLintMT,MalcovatiL, PascuttoC,etal.Predictivefactorsfortheoutcomeofallogeneic

transplantationinpatientswithMDSstratiﬁedaccordingtotherevisedIPSS-R. Blood2014;123(April(15)):2333–42.

[6]GreenbergPL,TuechlerH,SchanzJ,SanzG,Garcia-ManeroG,SoléF,etal. Revisedinternationalprognosticscoringsystemformyelodysplastic syndromes.Blood2012;120(September(12)):2454–65.

[7]ShafferBC,AhnKW,HuZ-H,NishihoriT,MaloneAK,ValcarcelD,etal.Scoring SystemPrognosticofOutcomeinPatientsUndergoingAllogeneic HematopoieticCellTransplantationforMyelodysplasticSyndrome.JClin Oncol.2016;34(June(16)):1864–71.

[8]DamajG,DuhamelA,RobinM,BeguinY,MichalletM,MohtyM,etal.Impactof azacitidinebeforeallogeneicstem-celltransplantationformyelodysplastic syndromes:astudybytheSociétéFrançaisedeGreffedeMoelleetde Thérapie-CellulaireandtheGroupe-FrancophonedesMyélodysplasies.JClin OncolOffJAmSocClinOncol.2012;30(December(36)):4533–40. [9]GuièzeR, Damaj G, PereiraB, RobinM,ChevallierP, MichalletM,etal.

ManagementofMyelodysplasticSyndromeRelapsingafterAllogeneic HematopoieticStemCellTransplantation:AStudybytheFrenchSocietyof BoneMarrowTransplantationandCellTherapies.BiolBloodMarrow TransplantJAmSocBloodMarrowTransplant2016;22(February(2)):240–7. [10]GauthierJ,DamajG,LangloisC,RobinM,MichalletM,ChevallierP,etal. ContributionofRevisedInternationalPrognosticScoringSystemCytogenetics toPredictOutcomeAfterAllogeneicStemCellTransplantationfor MyelodysplasticSyndromes:AStudyFromtheFrenchSocietyofBoneMarrow TransplantationandCellularTherapy.Transplantation.2015;99(August (8)):1672–80.

[11]GreenbergP,CoxC,LeBeauMM,FenauxP,MorelP,SanzG,etal.International scoringsystemforevaluatingprognosisinmyelodysplasticsyndromes.Blood 1997;89(March(6)):2079–88.

[12]ChesonBD,GreenbergPL,BennettJM,LowenbergB,WijermansPW,NimerSD, etal.ClinicalapplicationandproposalformodiﬁcationoftheInternational WorkingGroup(IWG)responsecriteriainmyelodysplasia.Blood2006;108 (July(2)):419–25.

[13]BacigalupoA,BallenK,RizzoD,GiraltS,LazarusH,HoV,etal.Deﬁningthe intensityofconditioningregimens:workingdeﬁnitions.BiolBloodMarrow TransplantJAmSocBloodMarrowTransplant.2009;15(December (12)):1628–33.

[14]PrzepiorkaD,WeisdorfD,MartinP,KlingemannHG,BeattyP,HowsJ,etal. 1994ConsensusConferenceonAcuteGVHDGrading.BoneMarrowTransplant 1995;15(June(6)):825–8.

[15]MoreauA-S,BourhisJ-H,ContentinN,CouturierM-A,DelageJ,DumesnilC, etal.Transferofallogeneicstemcelltransplantrecipientstotheintensivecare unit:GuidelinesfromtheFrancophonesocietyofmarrowtransplantationand cellulartherapy(SFGM-TC).BullCancer(Paris).2016;103(November(11S)): S220–8.

[16]HarrellFE, Lee KL,Mark DB. Multivariable prognosticmodels:issues in developingmodels,evaluatingassumptionsandadequacy,andmeasuringand reducingerrors.StatMed.1996;15(February(4)):361–87.

[17]Schoenfeld D. Partial Residuals for TheProportional Hazards Regression Model.Biometrika.1982;69(April(1)):239.

[18]vanBuurenS,Groothuis-OudshoornK.mice:MultivariateImputation by ChainedEquationsinR.JStatSoftw[Internet]. [cited2016Mar9];45(3). Availablefrom:.2011.http://www.jstatsoft.org/v45/i03/.

[19]RoystonP,MoonsKGM,AltmanDG,VergouweY.Prognosisandprognostic research:Developingaprognosticmodel.BMJ2009;338:b604.

[20]NeumannF,LobitzO,FenkR,BrunsI,KösteringM,SteinerS,etal.The sepsis-relatedOrganFailureAssessment(SOFA)scoreispredictiveforsurvivalof patientsadmittedtotheintensivecareunitfollowingallogeneicbloodstem celltransplantation.AnnHematol.2008;87(April(4)):299–304.

[21]AzoulayE,PèneF,DarmonM,LenglinéE,BenoitD,SoaresM,etal.Managing criticallyIllhematologypatients:Timetothinkdifferently.BloodRev.2015;29 (November(6)):359–67.

[22]BayraktarUD, Shpall EJ,Liu P, Ciurea SO,Rondon G, deLima M, etal. Hematopoieticcelltransplantation-speciﬁccomorbidityindexpredicts inpatientmortalityandsurvivalinpatientswhoreceivedallogeneic transplantationadmittedtotheintensivecareunit.JClinOncolOffJAmSoc ClinOncol2013;31(November(33)):4207–14.

[23]TownsendWM,HolroydA,PearceR,MackinnonS,NaikP,GoldstoneAH,etal. Improvedintensivecareunitsurvivalforcriticallyillallogeneic

haematopoieticstemcelltransplantrecipientsfollowingreducedintensity conditioning.BrJHaematol2013;161(May(4)):578–86.

[24]AzoulayE,MokartD,PèneF,LambertJ,KouatchetA,MayauxJ,etal.Outcomes ofcriticallyillpatientswithhematologicmalignancies:prospective multicenterdatafromFranceandBelgium–agroupederechercherespiratoire enréanimationonco-hématologiquestudy.JClinOncolOffJAmSocClin Oncol.2013;31(August(22)):2810–8.

[25]MoreauA-S,SeguinA,LemialeV,Yakoub-AghaI,GirardieP,RobriquetL,etal. Survivalandprognosticfactorsofallogeneichematopoieticstemcell transplantrecipientsadmittedtointensivecareunit.LeukLymphoma2014;55 (June(6)):1417–20.

[26]PèneF,AubronC,AzoulayE,BlotF,ThiéryG,RaynardB,etal.Outcomeof criticallyillallogeneichematopoieticstem-celltransplantationrecipients:a reappraisalofindicationsfororganfailuresupports.JClinOncolOffJAmSoc ClinOncol.2006;24(February(4)):643–9.

[27]MalcovatiL,Hellström-LindbergE,BowenD,AdèsL,CermakJ,DelCañizoC, etal.Diagnosisandtreatmentofprimarymyelodysplasticsyndromesin adults:recommendationsfromtheEuropeanLeukemiaNet.Blood2013;122 (October(17)):2943–64.

(8)

[28]AlzahraniM,PowerM,AbouMouradY,BarnettM,BroadyR,ForrestD,etal. ImprovingRevisedInternationalPrognosticScoringSystemPre-Allogeneic StemCellTransplantationDoesNotTranslateIntoBetterPost-Transplantation OutcomesforPatientswithMyelodysplasticSyndromes:ASingle-Center Experience.BiolBloodMarrowTransplantJAmSocBloodMarrowTransplant. 2018;24(June(6)):1209–15.

[29]KimDH,SohnSK,JeonSB,BaekJH,KimJG,LeeNY,etal.Prognosticsigniﬁcance ofplateletrecoverypatternafterallogeneicHLA-identicalsibling

transplantationanditsassociationwithsevereacuteGVHD.BoneMarrow Transplant.2006;37(January(1)):101–8.

[30]RamírezP,BrunsteinCG,MillerB,DeforT,WeisdorfD.Delayedplateletrecovery afterallogeneictransplantation:apredictorofincreasedtreatment-related mortalityandpoorersurvival.BoneMarrowTransplant.2011;46(July(7)):981–6. [31] DominiettoA,RaiolaAM,vanLintMT,LamparelliT,GualandiF,BerissoG,etal. Factorsinﬂuencinghaematologicalrecoveryafterallogeneichaemopoietic stemcelltransplants:graft-versus-hostdisease,donortype,cytomegalovirus infectionsandcelldose.BrJHaematol.2001;112(January(1)):219–27.

[32]OlssonRF,LoganBR,ChaudhuryS,ZhuX,AkpekG,BolwellBJ,etal.Primary graftfailureaftermyeloablativeallogeneichematopoieticcell

transplantationforhematologicmalignancies.Leukemia.2015;29(August (8)):1754–62.

[33]Matsumura-KimotoY,InamotoY,TajimaK,KawajiriA,TanakaT,HirakawaT, etal.Associationofcumulativesteroiddosewithriskofinfectionafter treatmentforsevereacutegraft-versus-hostdisease.BiolBloodMarrow TransplantJAmSocBloodMarrowTransplant.2016;8(March).

[34]DuléryR, MohtyM,Duhamel A,Robin M,BeguinY,MichalletM,etal. AntithymocyteGlobulinbeforeAllogeneicStemCellTransplantationfor ProgressiveMyelodysplasticSyndrome:AStudyfromtheFrenchSocietyof BoneMarrowTransplantationandCellularTherapy.BiolBloodMarrow Transplant.2014;20(May(5)):646–54.

[35]HaferlachT,NagataY,GrossmannV,OkunoY,Bacher U,NagaeG,etal. Landscapeofgeneticlesionsin944patientswithmyelodysplasticsyndromes. Leukemia2014;28(February(2)):241–7.