BACKGROUND■ ClassifyingCINVriskbychemotherapytypeandcorrespondingemetogenicityhasprovided

a valuable framework for the development of international prophylactic antiemeticguidelines.1,2

■ However,antiemeticguidelinerecommendationsbasedsolelyonchemotherapyemetogenicitymaypotentiallyundertreatpatientsathigherindividualemeticrisk,asconsiderableworkhasrevealedthatanumberofpatient-relatedvariablesmayalsocontributetoemeticrisk.

■ Between20%to40%ofcancerpatientsreceivingchemotherapyfailtoachievecompletecontrolofnauseaandvomiting(N&V).

■ RecentCINVriskmodelsweredevelopedbyDranitsarisetal3-5andMolassiotisetal6withthe goal of prospectively identifying patient risk factors to support the establishment ofpersonalizedCINVmanagement.

■ Though valuable, each of these models was developed from a limited sample size. Bycombiningthisdata,thiscurrentstudyutilizesthelargestCINVdatasetassembledoutsideoftheclinicaltrialsetting.Itencompassesapproximately1200patientswithavarietyofcancertypeswhoreceiveduptofourcyclesofchemotherapy.

METHODSOBjECTivE■ Toidentifyriskfactorsassociatedwithgrade≥2CINVincancerpatientsreceivingoutpatient

anticancertherapy■ Todeveloparepeatedmeasurespredictionmodelforgrade≥2CINVthatwillallowthe

identificationofhigh-riskpatientspriortoeachcycleoftherapy■ Todevelopanumericalriskalgorithmsystemthatcanbeusedforeachpatientforindividualized

riskassessmentDATA SOURCES AND OUTCOMES ■ Datawaspooledfromfournon-interventionalprospectivestudiesconductedinCanada3,5,7,8

andonestudyintheUnitedKingdom.6■ Adult patients with a variety of solid tumors scheduled to receive up to four cycles of

chemotherapyofvaryinglevelsofemetogenicity(lowtohigh)participatedinthesestudies.■ AllstudyparticipantscompletedadailypatientdiaryorutilizedtheMASCCAntiemesisTool

(MAT)eachchemotherapycycle,recordingepisodesofvomiting,severityofnausea,anduseofnon-prescriptionrescueantiemeticsfromtimeofchemotherapyinitiationthroughoutthefollowing5days.

■ Patientdemographicandbaselineclinicalcharacteristicswerecollectedalongwithadditionalbaselinedatasuchashistoryofnausea/vomiting,expectationofCINV,pre-chemotherapyanxiety,andhoursofsleepthenightpriortochemotherapy.

■ Thedependentvariableforthepooleddatabasemodelwas≥grade2CINVfromthefirstdayofchemotherapythroughDay5accordingtotheNationalCancer InstituteCommonToxicityCriteria(NCICTC).9

–Vomitingwasdefinedas≥3episodesin24hours. – Nausea was defined as either oral intake decreased without significant weight loss,

dehydration or malnutrition (based on NCI CTC) or at least mild nausea (based on a4-pointLikertscale).

STATiSTiCAl ANAlySiS ■ Thedatabaseswerecombinedandpredictorvariablesrecodedtoensureconsistency.■ To identify thesetof factorswith the largestpotentialcontribution topresenceofCINV,

thosewithap-valueof0.25orlessinasimplelogisticregressionwereretainedforfurtherconsideration.

■ Generalizedestimatingequations(GEE)forarepeatedmeasuresanalysiswerethenusedtodevelopthefinalriskmodelusingabackwardseliminationprocesswithapresetalphaat<0.05.

■ ThegoodnessoffitofthefinalmodelwasthenassessedwiththeHosmer-Lemeshowtest.■ Modelcalibrationwasevaluatedbyestimatingasmoothcalibrationlinebetweentheobserved

andpredictedoutcomes.

■ Internal validation of the final regression coefficients was done using nonparametricbootstrapping.

■ Ariskscoringalgorithmwasthenderivedfromfinalmodelcoefficients.■ Thepredictiveaccuracyoftheriskalgorithmsystemwasdeterminedviaareceiveroperator

characteristic(ROC)curveanalysis.

RESUlTS■ Atotalof1198patientsreceivingatotalof4197cyclesofchemotherapywereincludedin

theanalysis.

PATiENT AND TREATMENT CHARACTERiSTiCS ■ Patientdemographics,clinicalcharacteristics,baselineexperienceswithpotentialtoinfluence

CINVrisk,andchemotherapy/antiemeticspecificsareshowninTable 1.■ Seventy-fivepercentofpatientsinthispooleddatasetwerefemaleswithamedianageof

58years;mosthadearlystagediseasewithbreastcancerbeingthepredominantcancertypeinabouthalfofpatients(Table 1).

CiNv OUTCOMES DATA ■ Consideringallcycles,morethanhalfofpatients(61.1%)experiencedeithernauseaand/or

vomitingduringthe5daysfollowingchemotherapy,with42.2%ofpatientswith≥grade2CINV.

■ Figure 1showtheproportionsofpatientsexperiencing≥grade2CINVbycycle.

ASSESSMENT Of CiNv RiSK fACTORS ■ AnumberofvariableswereshowntohaveasignificantassociationwithCINV(Table 2).

RiSK PREDiCTiON AlGORiTHM ■ Ariskscoringalgorithmwasderivedfromthefinalmodelcoefficients(Table 3).

■ The ROC analysis indicated good predictive accuracy with an area under the curve of0.71(95%CI:0.69–0.73) (Figure 2).

■ Patientswithatotalscore≥16wouldbeconsideredathighriskofCINV(Table 4).

■ Moving from one scoring category to another increases CINV risk by 80% (OR = 1.80;p<0.001)(Figure 3).

■ Tomaximizemodelsensitivity,patientswithCINVprobability≥43.7%wouldbeconsidered“highrisk”.

–Therefore,apatientreceivingalowemetogenic(LEC)regimenwouldbereclassifiedasmoderatelyemetogenic(MEC)iftheircalculatedCINVprobabilitywere≥43.7%.

–Similarly,apatientreceivingaMECregimenwouldbereclassifiedashighlyemetogenic(HEC)iftheircalculatedCINVprobabilitywere≥43.7%.

■ Limitationsofthepredictiontoolinclude: –TheROCanalysissuggestedthatthereisadditionalroomtoimprovetheaccuracyofthe

predictiontool. –Onlyreadilymeasurablevariableswereconsideredbythemodel;hence,notallthevariability

wasaccountedfor(egpharmacoeconomicfactors). –Themodelsdonotreplaceclinicaljudgementbutprovideadditionalinformationtosupport

medicaldecision-making.

CONClUSiONS ■ ThisrobustpooleddatabaseevaluatedriskfactorsassociatedwiththedevelopmentofCINV

inapproximately1200patientsreceivingalmost4200cyclesofchemotherapy.■ Asidefromchemotherapyemetogenicity,anumberofpatient-relatedriskfactorswereshown

tobestronglypredictiveofCINVfollowingchemotherapy.■ Thereexistsaneedforgreaterawarenessoftherolethatpatient-relatedfactorsmayplayin

increasingemeticrisk,apartfromtypeofchemotherapy.Thismaybeparticularlyrelevantinthemoderatelyemetogenic(MEC)settingwhichencompassesabroadrangeofchemotherapiesknowntoelicitemesisin30%-90%ofpatients.

■ ACINVriskassessmenttoolhasbeendevelopedbasedonthisstudytoassistphysiciansincomprehensivelyassessingpatients’riskofdevelopingCINV(thiswillbeavailablesoonatwww.cinvrisk.org).

■ Theclinicalapplicationofthisriskassessmentpredictiontoolshouldenhancepatientcarebyoptimizingtheuseoftheantiemeticsinaproactivemanner.

identifying Cancer Patients at High Risk for Chemotherapy-induced Nausea and vomiting (CiNv): The Development of a Prediction Tool George Dranitsaris1, Alex Molasiotis2, Mark Clemons1, Eric Roeland3, Lee Schwartzberg4, David Warr5, Karin Jordan6, Pascale Dielenseger7, Matti Aapro8 1Ottawa Hospital Regional Cancer Centre, Ottawa, Canada; 2Hong Kong Polytechnic University, Hong Kong; 3University of California San Diego Moores Cancer Center, la jolla, CA, USA; 4The West Clinic, Memphis, TN, USA; 5Princess Margaret Cancer Center, Toronto, Canada; 6University of Halle, Halle, Germany; 7Gustave Roussy Cancer Campus, villejuif, france; 8 iMO Clinique de Genolier, Genolier, Switzerland

REFERENCES1. HeskethPJ,BohlkeK,LymanGH,etal(2016)Antiemetics:AmericanSocietyofClinicalOncologyfocusedguidelineupdate.JClinOncol34(4):381-386.2.MASCC/ESMOAntiemeticGuideline2016.Availablefrom:http://www.mascc.org/assets/Guidelines-tools/mascc_antiemetic_guidelines_english_2016_v.1.1.pdf.3.DranitsarisG,etal.Identifyingpatientsathighriskfornauseaandvomitingafterchemotherapy:Developmentofapracticalpredictiontool.I.Acutenauseaandvomiting.JSupportOncol2009;7:W1-W8.4.PetrellaT,etal.Identifyingpatientsathighriskfornauseaandvomitingafterchemotherapy:Developmentofapracticalpredictiontool.II.Delayednauseaandvomiting.JSupportOncol2009;7:W9-W16.5.Dranitsaris,G,etal.Prospectivevalidationofapredictiontoolforidentifyingpatientsathighriskforchemotherapy-inducednauseaandvomiting.JSupportOncol.2013Mar;11(1):14-21.6.MolassiotisA,etal.Developmentandpreliminaryvalidationofriskpredictionmodelforchemotherapy-inducednauseaandvomiting.SupportCareCancer2013;21(10):2759-67.7.BouganimN,etal.Prospectivevalidationofriskpredictionindexesforacuteanddelayedchemotherapy-inducednauseaandvomitingCurrOncol2012;19(6):e414-21.8.ClemonsM,etal.Riskmodel-guidedantiemeticprophylaxisvsphysicianchoiceinpatientsreceivingchemotherapyforearlystagebreastcancer.JAMAOncology2016;2(2):225-231.9.NationalCancerInstitute,CommonTerminologyCriteriaforAdverseEventsv4.0NCI,NIH,DHHS.May29,2009.NIHpublication#09-7473.

ACKNoWLEDGEMENTS: ThisprojectissupportedbyHelsinnHealthcare.

European Society for Medical Oncology 7-11 October 2016Copenhagen, Denmark

Poster: 1438PD

Table 1. Patient and Treatment Characteristics from Pooled CINV Studies

Patient Characteristic Total Number of Patients = 1198 %ofpatients(no.ofpatients)

Femalegender 74.6%(894)Medianpatientage(range) 58(19-100)

Type of cancerBreast 55.5%(665)Gastrointestinal 14.4%(172)Genitourinary 1.8%(21)Gynecological 5.7%(68)Lung 8.1%(97)Other 13.2%(158)Missing 1.4%(17)

Earlystage(vs.metastatic) 73.4%(879)

Historyofmotionsickness 26.7%(320)Historyofmorningsicknessduringapregnancy 37.5%(449)Dailyalcoholintake 24.5%(294)

Patient/Treatment Characteristic Total Number of Cycles = 4197 %ofcycles(no.ofcycles)

Median number of cycles (range) 2(1to11)

Anticipatory nausea and vomiting 27.9%(1174)

Sleep the night before chemotherapyLessthan5hours 12.8%(538)Fivetosevenhours 56.0%(2351)Eighttoninehours 26.8%(1125)Tenhoursofmore 2.9%(123)Missing 1.4%(60)

Anxiety before chemotherapyNone 29.5%(1237)Mild 21.8%(916)Moderate 18.1%(760)High 4.9%(205)Missing 25.7%(1079)

Type of ChemotherapyPlatinum-based 27.4%(1151)Anthracycline-based 52.8%(2217)Singleagenttaxanes 7.4%(310)Other 12.4%(519)

Figure 1. Prevalence of CINV (0-120 h) by Cycle of Chemotherapy

Perc

enta

ge o

f Pat

ient

s

1

50

30

20

0

60

2 3 4

10

Cycle

52

4440.4 39.2

5 6

39.335.540

Table 2. Predictive Factors for Nausea and Vomiting from Day 0 to Day 5

Predictive Factor1 odds Ratio2 Impact on Risk

Agelessthan60 1.41 ↑ by41%

PatientexpectstodevelopCINV 1.41 ↑by41%

Sleeplessthan7hours 1.34 ↑by34%

Historyofmorningsickness 1.30 ↑by30%

Platinum-oranthracycline-basedchemotherapy 1.94 ↑by94%

NorVinpriorcycle 5.17 ↑by5.17times

Useofnon-prescribedantiemeticsathome 2.70 ↑by2.7times

Cycle number (vs. cycle 1)

Cycle2 0.17 ↓by83%≥Cycle3 0.15 ↓by85%

Dependentvariable:≥grade2CINVfromday0to51Thesevariableswere retained in thefinalmodelusingabackwardseliminationprocesswithp<0.05as thecutoff to retain.2Anoddsratiooflessthanonemeanslowerriskandgreaterthanonemeansincreasedrisk.

Figure 2. Percent risk of CINV vs risk score

CIN

V Pr

eval

ence

(%)

Risk score

80

20

0

60

5 10 15 20

40

AUC of RoC curve for:internal validation(0.69; 95%Ci 0.67-0.70)

Table 3. CINV Prediction Tool

Start at base score of 10

Patientage Ifpatientage<60 +1

Expectation IfpatientexpectstohaveCINV +1

Sleep Ifpatientslept<7hoursthenightbeforechemotherapy +1

Morningsickness Ifpatienthasahistoryofmorningsickness +1

Chemotherapy Ifpatientisabouttoreceiveplatinumoranthracyclinechemotherapy +2

PriorCINV Ifpatienthadnauseaorvomitinginthepriorcycle +5

Antiemeticuseathome Ifnon-prescriptionantiemeticsareusedathome +3

Cycle If2ndcycleofchemotherapy -5 If≥3rdcycle -6

Association between score and risk of CINV: oR = 1.18, p < 0.001

Figure 3. Percent Risk of CINV vs Risk Category Score

Risk category 4 (score ≥ 16) would be considered “high risk”

CIN

V ri

sk (%

)

Risk category

20

0

60

1 2 3 4

40

80

Figure 3. Percent Risk of CINV vs Risk Category Score

Risk category 4 (score ≥ 16) would be considered “high risk”

0

20

40

60

80

1 2 3 4 5 6 7

CIN

V r

isk

(%)

Risk category

5 6 7

Table 4. Accuracy of the Prediction Model

Score CINV Sensitivity Specificity Likelihoodcut point incidence* ratio

<8 12.5% 100% 0% 1.0

≥8to<12 13.6% 99.8% 1.2% 1.01

≥12to<16 23.1% 97.9% 10.7% 1.10

≥16to<20 43.7% 87.4% 38.4% 1.42

≥20to<24 57.6% 51.2% 75.7% 2.11

≥24to<28 72.8% 18.8% 94.8% 3.60

≥28 87.9% 2.1% 99.8% 9.08

*Asobservedinthepatientsample.

CopiesofthisposterobtainedthroughQR(QuickResponse)codeareforpersonaluseonlyand

maynotbereproducedwithoutwrittenpermissionoftheauthors.

Ifquestions,pleasecontactGeorgeDranitsarisatgeorge.dranitsaris@gmail.com