UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)

UvA-DARE (Digital Academic Repository)

(Un-)certainties in radiotherapy of rectal cancer

Nijkamp, J.A.

Link to publication

Citation for published version (APA):Nijkamp, J. A. (2012). (Un-)certainties in radiotherapy of rectal cancer. Amsterdam: Nederlands Kanker Instituut- Antoni van Leeuwenhoek Ziekenhuis.

General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, statingyour reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Askthe Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,The Netherlands. You will be contacted as soon as possible.

Download date: 17 Apr 2020

InternationalJournalofRadiationOncology*Biology*PhysicsVolume79,Issue2,Februari2011,Pages481-489

CliftonD.Fuller1,2,3,JasperNijkamp4,JoopC.Duppen4,CoenR.N.Rasch4,CharlesR.Thomas3,SamuelJ.Wang3,PaulOkunieff5,WilliamE.Jones1,

DanielBaseman1,ShilpenPatel6,CarloG.N.Demandante1,7,AnnaM.Harris1,BenjaminD.Smith7,8,AlanW.Katz5,CamilleMcGann8,JenniferL.Harper10,

DanielT.Chang11,StephenSmalley12,DavidT.Marshall10,KarynA.Goodman13,Niko Papanikolaou1,2,andLisaKachnic14onbehalfoftheRadiationOncology

CommitteeoftheSouth-WestOncologyGroup

Prospective randomized double-blind pilot study of site-specific consensus

atlas implementation for rectal cancer target volume delineation in the

cooperative group setting

1Department of Radiation Oncology and 2Graduate Division of Radiological Sciences/Department of Radiology, The University of Texas Health Science Center at San Antonio, San Antonio, TX, USA.

3Department of Radiation Medicine, Oregon Health & Science University, Portland, OR, USA. 4Department of Radiation Oncology, The Netherlands Cancer Institute,

Antoni van Leeuwenhoek Hospital, Amsterdam, The Netherlands. 5Department of Radiation Oncology, University of Rochester, Rochester, NY, USA.

6Department of Radiation Oncology, University of Washington Medical Center, Seattle, WA, USA. 7Radiation Oncology Flight, Wilford Hall Medical Center, Lackland Air Force Base. San Antonio, TX, USA.

8Department of Radiation Oncology, MD Anderson Cancer Center, Houston, TX, USA. 9Department of Radiation Oncology, Brooke Army Medical Center, San Antonio, TX, USA.

10Department of Radiation Oncology, Medical University of South Carolina, Charleston, SC, USA. 11Department of Radiation Oncology, Stanford University, Stanford, CA, USA.

12Department of Radiation Oncology, University of Kansas, Kansas City, KS, USA; Radiation Oncology Center of Olathe, Olathe, KS, USA.

13Department of Radiation Oncology, Memorial Sloan-Kettering Cancer Center, New York, NY, USA. 14Department of Radiation Oncology, Boston University Medical Center, and Massachusetts General Hospital Cancer Center, Boston, MA, USA.

3

44 |

Chapter 3

Abstract

PurposeVariations in target volumedelineation represent a significant hurdle in clinical trialsinvolving conformal radiotherapy.We sought to determine the effect of a consensusguideline-basedvisualatlasoncontouringthetargetvolumes.

Material and methodsArepresentativecasewascontoured(Scan1)by14physicianobserversandareferenceexpertwithandwithouttargetvolumedelineationinstructionsderivedfromaproposedrectal cancer clinical trial involving conformal radiotherapy. The gross tumor volume(GTV),andtwoclinicaltargetvolumes(CTVA,includingtheinternaliliac,presacral,andperirectalnodes,andCTVB,which includedtheexternal iliacnodes)werecontoured.Theobserverswererandomlyassignedtoreceipt(GroupA)ornonreceipt(GroupB)ofaconsensusguidelineandatlasforanorectalcancersandtheninstructedtore-contourthesamecase/images (Scan2).Observervariationwasanalyzedvolumetricallyusingtheconformationnumber(CN,whereCN=1equalstotalagreement).

ResultsOf14evaluablecontoursets(1expertand7GroupAand6GroupBobservers),greateragreementwasfoundfortheGTV(meanCN,0.75)thanfortheCTVs(meanCN,0.46–0.65).AtlasexposureforGroupAledtosignificantlyincreasedinter-observeragreementforCTVA(meaninitialCN,0.68,afteratlasuse,0.76;p=0.03)andincreasedagreementwiththeexpertreference(initialmeanCN,0.58;afteratlasuse,0.69;p=0.02).FortheGTVandCTVB,neithertheinter-observernortheexpertagreementwasalteredafteratlasexposure.

ConclusionsConsensusguidelineatlas implementationresultedinadetectabledifferencein inter-observeragreementandagreaterapproximationofexpertvolumesfortheCTVAbutnotfortheGTVorCTVBinthespecifiedcase.VisualatlasinclusionshouldbeconsideredasafeatureinfutureclinicaltrialsincorporatingconformalRT.

| 45

Consensus atlas implementation for rectal cancer target volume delineation

IntroductionInter-observerdifferences in targetvolumedelineationareademonstratedsourceofpotentialtreatmentvariabilityinthecontextofclinicaltrialsthatincorporateconformalradiotherapy (RT) approaches [1, 2]. Recent publications have suggested that targetdelineationconsensusdocumentationishighlydesirableforclinicaltrials[3]andthatspecific instructionaloreducational interventionsmightaffordameasurableeffect intermsofphysiciancontouring[4,5]. AsapartofeffortstoimproveRTimplementationfortheSouthwestOncologyGroup(SWOG) trials and consistent with its focus on quality improvement in cooperativestudies, theSWOGRadiationOncologyCommitteeauthorized thepresent studyasapilotprojecttoachievethefollowingprimaryspecificaims:thefeasibilityofcentralizedtargetvolumedelineationevaluationasapre-trialadjuncttoaSWOG-sponsoredstudy(SWOGS0713),andthedeterminationoftheeffectofimplementationofaconsensusanatomicatlasontargetvolumevariability.

Material and methodsThis prospective institutional review board-exempt study was conducted under theauspicesof theUniversityofTexasHealthScienceCenteratSanAntonio institutionalreviewboard.Thepresentstudywasdesignedasadouble-blind,randomizedhypothesis-generatingpilotstudy(Fig.3.1).Statisticalpowerforagreementanalysiswasestimatedforanon–Bonferroni-correctedpaired-measuresWilcoxontest (assumingaminimumasymptoticrelativeefficiencyof≥0.863comparedwithapairedttest),withaspecified1-βof0.7andαof≤0.05, resulting inaminimal requisitesamplesizeof6observers(radiationoncologists)pergroup,calculatedusingtheG*Power3statisticalsoftware[6].Goalenrolmentwas10–12observerspercohort.

Fig. 3.1:Studydesign

46 |

Chapter 3

The participating radiation oncologists (observers) were recruited from SWOG-participating institutions. Those who indicated interest were sent the studydocumentation,which included a standardized case report, description of the targetvolumes tobe contoured, and a compact disc (CD) containing 3-mmaxial computedtomography (CT) images derived from the Digital Imaging and Communications inMedicine(DICOM)fileofthestandardizedcasestudy’ssimulationCTscan.ThevolumesweretobecontouredtwiceusingtheBigBrothertargetdelineationsoftwareprogram.‘‘Big Brother’’ is a custom target volume delineation evaluation software platformdevelopedatTheNetherlandsCancerInstitute[7,8].Itconsistsofauserinterfacewithtargetdelineationfeaturescommontomostcommercialtreatmentplanningsystems[9]andcollectsawidearrayofvolumetricandtargetdelineationdataunobtrusivelyduringthecontouringsession. TheincludedcasestudydepictedthehistoryandclinicalfindingsfromananonymizedpatientwithStageT3N0M0adenocarcinomaoftherectumwithinstructionsmodelledon a SWOG protocol in development at that time that included detailed directivesregarding3-dimensionalconformalRTandintensity-modulatedRTplandesign(SWOGS0713:APhase IIStudyofOxaliplatin,Capecitabine,CetuximabandRadiation inPre-operative Therapy of Rectal Cancer; ClinicalTrials.gov Identifier NCT00686166), withtheterminologymodifiedtofitthenomenclatureestablishedinthethen-unpublishedRadiationTherapyOncologyGroup(RTOG)consensusguidelinesfortargetdelineationin anorectal cancers [10]. The observers were asked to contour the structures aslisted in Table 3.1. The axial CT imageswere extractedusing a singleDigital ImagingandCommunicationsinMedicinedataset;identicalcopiesofthereconstructed(axial,sagittal,andcoronalviews)werethendesignatedasScan1andScan2. One-half of the distributed CDs contained an automated HTML link, which, aftersubmissionofthefirstcontouringsession(Scan1)andthesubsequentelectronicsurvey,directeduserstoaprepublicationversionoftheRTOGconsensusguidelinesfortargetvolumesinanorectalcancer[10]andinstructionstore-contourtheexactsameaxialCTimagesasecondtime(Scan2),withthesamecasepresentation,instructions,andtargetdefinitions,using theRTOGconsensusguidelinevisualatlasasaguide (GroupA).AllotherCDscontainedHTMLpop-updirectionstore-contourthesamevolumesontheidenticalCTsimulation-deriveddataset(Scan2),usingthesameaforementionedcasedata/instructionsaspreviously(GroupB).Thus,GroupBdidnotreceiveconsensusatlasguidance for re-contouring thecase.TheCDswithandwithout theHTML link to theconsensusatlaswererandomlyshuffledbeforelabelinganddeliverytotheparticipants;thestudypersonnelandphysicianobserverswerebothunawareofwhichCDhadbeendistributedtoeachparticipantuntilelectronicsurveycompletion. Aftercompletionofthegrosstumorvolume(GTV)andclinicaltargetvolume(CTV)delineationonScan1,theobserverssubmittedthecasebye-mailandweredirectedtoanelectronicsurvey.Subsequently,theparticipantswereprovidedwithinstructionstore-contourthecasewithorwithouttheassistanceofananatomicallyspecificconsensusatlas.TherecentlypublishedRTOGconsensusatlas[10]wasusedinprepublicationform(availablefrom:www.rtog.org/pdf_document/AnorectalContouringGuidelines.pdf).

| 47

Consensus atlas implementation for rectal cancer target volume delineation

Inaddition,oneofthemembersinvolvedinthedevelopmentoftheRTOGconsensusguidelineswasaskedtodelineateScans1and2.Thisobserver(L.K.)wasdesignatedasthe ‘‘referenceexpert,’’withhercontours servingas thede facto referencestandardagainstwhichtocomparetheobserver-derivedcontours.Duringthestudyperiod,onlythe reference expert user had any previous knowledge of this atlas; thus, the studyparticipantsrepresentedatabularasawithregardtotheconsensusguidelines.

Delineation agreement analysisAll delineations were first analyzed visually (Fig. 3.2), and any protocol deviationsfrom thedelivered instructionswere identifiedby a reviewof all axial contours. Thetotal volumeencompassed in cubic centimeters for all structureswas calculatedandtabulated.AstatisticalcomparisonofthevolumedifferentialsbetweenScans1and2wasperformedforeachstructureforGroupAandB,respectively. Thebaselineinter-observervariationfortheSWOGprotocolwasderivedfromthedelineationsonScan1fromallobservers,exceptforthereferenceexpert.Thebaselineintra-observervariationwasderivedfromacomparisonofthevolumeofScans1and2 fromGroupB.Theeffectof theatlason inter-observervariationwasquantifiedbycomparingtheinter-observervariationforScans1and2fromGroupA.Foracomparisonwithin cohorts, a compositemedian delineationwas calculated for each group. Themediandelineationrepresented50%coverageoftheisosurfacebytheobservers,suchthat each voxel insidewas designated by ≥50%of the observers andwas calculatedforGTV,CTVA(internal iliac,presacral,andperirectalnodes),andCTVB(external iliacnodes).TheCTVCstructurewasnotdesignatedintheinstructionsasanecessaryvolumebe contoured for this clinical case and was, therefore, not analyzed. For volumetricagreement analysis for Group A, the common volume was first calculated betweeneitherthemedianorexpertcontour(V1)andtheobservercontour(V2).Subsequently,asamodificationoftheconceptintroducedbyFeuvretetal.[11]andvan’tRietetal.

Table 3.1:TargetvolumedefinitionsandinstructionsStructures Definition/Instructions

GTV Includesprimary tumorandanypelvicnode thought tobe involvedgrosslywithmetastaticdiseaseAssessmentofgrosslyinvolvednodaldiseasecanbemadeaccordingtocomputedtomographyscanforthisstudy

CTV Consists of CTVA,CTVB, andCTVC; should includeGTVand followingnodalgroups:perirectalnodes,presacralnodes,andinternaliliacandcommoniliacnodesbelowL5–Sjunction

CTVA For this study, definedas nodal regions thatwould regularly be treated forrectalcancer(i.e.,internaliliac,presacral,andperirectalnodes)

CTVB Forthisstudy,definedasexternaliliacnodalregionCTVC Forthisstudy,definedasinguinalnodalregion

Specificinstructions

ContourstructuresdenotedGTV,CTVA,CTVB,and(optionally)CTVC,accordingtoprecedingtargetvolumedefinitionsDo NOT add ANY margin to account for positional variability. It will beassumedthatastandardizedplanningtargetvolumeexpansionof0.7–1.0cmomnidirectionallywillbeappliedusingtheGTV/CTVstructuresgenerated

48 |

Chapter 3

[12],aconformationnumber(CN)wasderivedasCN=CV2/(V1XV2).DifferencesintheCNvaluesforthetargetstructures(e.g.,GTV,CTVA,CTVB)forScans1and2forGroupA,usingboth the referenceexpertand thegroupmediandelineation isosurfaceasacomparator,werecalculatedandformallyassessedforstatisticalsignificanceusingthepaired-measuresWilcoxontest. ForGroupB,theintra-observerCNvalueswerecalculatedusingtheaforementionedvan’tRietformula[12];thecommonvolumewascalculatedbetweeneitherScan1(V1)orScan2(V2).

Post hoc exploratory contour surface variability analysis After completion of the planned volumetric analysis, surface distance analysis wasperformed to identify the regional delineation variation within the CTVA volume,usingvirtualvolumeunfolding,aspreviouslypublished[13,14].Inbrief,forthesurfacedistance variation calculation, the reference structure (median or expert) was firstresampledto100equidistantpointsperdelineatedslice.Second,foreachpointonthereferencestructure,thedistance,perpendiculartothesurface,totheobserver-derivedcontourwascalculated.Fortheobservervariationanalysis, thestandarddeviationofallobserverswascalculatedforeachpointonthereferencestructure.Forcomparisonwiththeexpert,thegroupmedianwascalculatedbytakingthemedianofthedistancesforeachpointontheexpert-derivedreferencestructuretotheperpendicularsurfaceofeveryobserver-derivedcontour.Regionaldifferentialsinsurfacevariationwerethenexploredgraphicallyandnumerically.

ResultsEightSWOGinstitutionshadat leastoneusersubmittingcontours,aswellasasinglenon–SWOG-affiliatedparticipant.Ofthe26observersdirectlyaskedtoparticipate,15submittedcontoursetpairs,ofwhich14weretechnicallyevaluable(1expert,7inGroupA,and6inGroupB).Thenon-evaluablecontoursetconsistedofnon-connected,non-overlappingcontoursthatprecludedreadyanalysiswiththecohortatlarge. All14remainingobserversdelineatedtheGTVandCTVAonScans1and2.AlthoughtheCTVBwasmandatoryinthespecificdelineationinstructions,itwasonlydelineatedby 11 of the 14 observers. The CTVC, which should not have been delineated, wascontouredby2observersonbothScan1andScan2,by1observeronScan1only,andby1observeronScan2only.For1observer inGroupAand4observers inGroupB,majordeviationsfromthedelineationprotocol(e.g.,theGTVwasnotencompassedbytheCTVA)werevisibleonaxialslicereview.ForanadditionalobserverinGroupA,theCTVBcoveredtheinternaliliacvesselsinsteadoftheexternaliliaconbothScan1andScan2.Forthe5observersforwhomtheCTVAdidnotfullycovertheGTV,theCTVAwasmanuallyeditedsuchthattheobserver-contouredGTVwasencompassedforthevolumeanalysis;apreliminarystatisticalevaluationevidencedminimalalterationofthevolumetricstatisticsbythismodification.

| 49

Consensus atlas implementation for rectal cancer target volume delineation

BetweenScans1 and2,only the increase in the volumeof theCTVB inGroupAapproachedstatisticalsignificance(p=0.06;Table3.2).InGroupB,thenumberofCTVBslicescoveredbyallobserversdecreasedfrom14to3axialCTslices,andtheaveragedelineatednumberofaxialCTslicescontouredonlydecreasedfrom20to16slices.ThemedianGTVdelineatedonScan1forallobservershadavolumeof74cm3.TheaverageCNforthebaselineinter-observervariationoftheGTVwas0.75(range,0.60–0.81).ThemedianCTVAhadavolumeof709cm3andaCNof0.65(range,0.47–0.75),indicatingcomparativelygreater inter-observerdisagreement forCTVAcomparedwith theGTV.ForCTVB,withamedianvolumeof70cm3,evenlessinter-observeragreementcouldbefound,withanaverageCNof0.46(range,0.24–0.70).

Fig. 3.2:Overviewof largest and smallest clinicaltargetvolumeAdelineations (white)andmedianclinicaltargetvolumeAdelineation(black)fromallobserversonScan1.

Fig. 3.3:Conformationnumber(CN)forobservers inGroupAbefore(Scan1)andafterintroductionofatlas(Scan2)comparedwithmediansurface(a)andexpertdelineation(b).CN=1indicatescompletevolumetricagreement.Boxplotdisplaysinterquartilerange(i.e.,25th to 75th percentiles), with horizontal line dem-onstratinggroupmedian;whiskers includeallpointswithin 1.5 times inter-quartile range from 25th and75th percentiles. Pre- and postatlasmean CNs con-nectedbydiagonalblueline.

50 |

Chapter 3

AtlasexposureledtoastatisticallysignificantincreaseinvolumetricagreementonCTVA between observers (Fig. 3.3a) andwith the expert (Fig. 3.3b), asmeasured byCN.Theaverageinter-observerCN(i.e.,agreementwiththemediansurface)increasedfrom0.68(range,0.41–0.78)onScan1to0.76(range,0.57–0.87)onScan2(p=0.031,pairedWilcoxonsignedranktest;Fig.3a).TheaverageCN,comparedwiththeexpert,increasedfrom0.58beforetheatlas(range,0.42–0.70)to0.69aftertheatlas(range,0.58–0.78, p=0.016; Fig. 3.3b). For the CTVB, however, neither the inter-observervariation(meanCN,0.39[range,0.26–0.67]vs.meanCN,0.45[range,0.13–0.68];p=.4)

Fig. 3.4:Inter-observervariationchangeswithintroductionofatlas.(Left)Inter-observervariation(standarddeviation)inGroupAbeforeintroductionofatlasshownforanterior,sagittal,andposteriorview. (Right) Inter-observervariation insamegroupshownafterintroductionofatlas.Groupstandarddeviationofdistancetoexpertsurfaceshownascolorscaleonright(valuesincentimetersfromexpertsurface).

| 51

Consensus atlas implementation for rectal cancer target volume delineation

northeagreementwiththeexpert(meanCN,0.31[range,0.16–0.49]vs.meanCN,0.30 [range,0.11–0.44])wasaltered toastatistically significant degree after atlasexposure(p=0.8). Because the atlas only affectedthe observer variation for the CTVA, theexploratory post hoc surface distancevariation analysis was limited to CTVA(Figs.3.4and3.5).Totranslatethesurfacemaps into numbers, first the referencestructure (median/expert CTVA) wasdivided into the anterior, lateral, andposteriorregionsandsubdividedintotheupperandlowersub-regionsatthelevelofthecoccyxtip.Foreachofthesixregions,the standard deviation value covering5–95% of the regional surface distancedifference was taken to characterize theminimal and maximal regional variation(Table3.3), althoughno formal statisticalcomparison of the regional sub-volumeswas performed. Visual inspection (Figs.4and5)showedthatthe introductionofthe atlas resulted in modification of thesurface distance between the observersandexpertprimarily in limitedregionsofthe CTVA, rather than the CTVA volumeglobally. Modification of the targetvolumes was most notably localized totheupper-anteriorregionadjacenttothebladder,lower-posterior,andlateralCTVA(data not shown); however, statisticalsignificancewasnotformallyassessed.Foralldefinedregions,except for theupper-posterior and upper-lateral, the upper95% confidence interval of the inter-

observersurfacestandarddeviationwasreducedby0.2–0.8cmaftertheintroductionoftheatlas.AsthedatainTable3.4demonstrate,>1cmofsurfacevariationwasobservedformultiple regions before atlas implementation for all users, and although reducedafteratlasadministration,>1cmwasstillneededtocover95%ofthesurfacevariationintheCTVsub-regions.

Fig. 3.5:Intra-observervariation.Intra-observervaria-tion (standard deviation) shown for anterior, sagittal,andposteriorviewsforobserverswhodelineatedcasetwice without any atlas exposure (Group B). Groupstandard deviation of distance between equivalent-pointsforScans1and2foreachusershownascolorscaleonright(incentimeters).

52 |

Chapter 3

Regardingintra-observervariation,theabsolutevolumeofallrespectivestructurescontouredwasessentiallyequivalent(Table3.2).AcomparisonbetweenthedelineationsonScans1and2inGroupByieldedanaverageCNof0.80(range,0.75–0.82),0.68(range,0.47–0.89),and0.54(range,0.16–0.72)fortheGTV,CTVA,andCTVB,respectively.Theregionalintra-observervariabilityisillustratedgraphicallyinFig.3.5.

DiscussionDespite the well-known consequences of geometric inaccuracy in target volumedelineation[15–17],inter-observervariabilityintargetdefinitionhasbeendemonstratedin a host of studies and at various anatomic sites [18]. Simply put, ‘‘inter-observervariabilityinthedefinitionofGTVandCTVisamajor,forsometumorlocationsprobablythelargest,factorcontributingtotheglobaluncertaintyinradiationtreatmentplanning’’[18]. Consequently, efforts to implement solutions to possible sources of variability/error in the target volumedelineationprocesshave continued. These solutionshave

Table 3.2: Selected volumetric and axial slice measuresReferenceexpert GroupA(atlas) GroupB(control)

Parameter Scan 1 Scan 2 Scan 1 Scan 2 p-value Scan 1 Scan 2 p-valueMeanvolume(cm3)

GTV 68 78 78±15.2 78±15.3 1.0 68±9.7 68±6.9 0.9CTVA 784 820 800 ± 276 809 ± 172 0.7 590 ± 208 642 ± 251 0.4CTVB 67 91 77 ± 62 100 ± 78 0.06 71 ± 26 51 ±32 0.3

Meandelineatedlength(no.ofaxialslices)GTV 12 12 19±8.3 19±8.3 16±7.4 16±7.4CTVA 19 21 44±6.8 41±5.9 40±6.3 39±6.2CTVB 19 21 13±7.5 15±8.5 20±4.7 16±7.7

AxialCTslicescoveredbyallobserversGTV 12 12 1 12 12 12CTVA 45 43 32 31 31 33CTVB 19 21 9 11 14 3

Table 3.3:CTVAinterobserversurfacedistance,expressedas95%confidenceintervalofgroupstandarddeviationfrommediangroupisosurface,byregionalsubdivision,beforeandafteratlasexposureVariable Anterior(cm) Posterior(cm) Lateral(cm)

Allobservers,Scan1(i.e.,initialtotalinterobservervariation)Upper 0.6-1.4 0.2-0.7 0.3-1.1Lower 0.6-1.2 0.3-1.2 0.4-1.1

GroupA,Scan1(i.e.,initialinterobservervariation)Upper 0.5-1.4 0.2-0.7 0.2-1.0Lower 0.7-1.3 0.4-1.4 0.5-1.4

GroupA,Scan2(i.e.,postatlasinterobservervariation)Upper 0.5-1.1 0.1-0.7 0.2-0.9Lower 0.6-1.1 0.2-0.6 0.3-0.8

| 53

Consensus atlas implementation for rectal cancer target volume delineation

includedoptimizationofimaginginputs[19–22],instructionalprotocolmodification[5,23,24],integrationofspecifictrainingprograms[25,26],developmentofsoftwaretools[27–31],and implementationofstandardizedguidelines[32–37]fordistinctanatomicsubsites.Forclinicaltrials,thesituationispotentiallymorevexing,becausetoensureadequatetreatmentuniformitybetweencomparisoncohortsnecessitatescomparativelyincreased attention to both protocol construction and enrollee plan review, costingsignificanttimeintermsofresourcesfortheprimaryinvestigators. In terms of feasibility, the study was readily completed (total study duration, 5months).Ofthe26invitedSWOGinstitutions,12(46%)confirmedanintenttoparticipate;however,only8(31%)hadresultantsubmissions.Nonetheless,ourfindingssuggestthatareasonablypoweredtargetdelineationtrialmightbeimplementedwithamodicumofcooperativegroupresourceallocationintimelymannerandthatsuchastudywouldbebothtechnicallyandlogisticallyfeasible. TheanalysisoftheresultantdataalludestothedifficultyinexecutingclinicaltrialsintheconformalRTera.Thehighproportionofmajorprotocoldeviationswasconsistentwiththat inpreviousreports.Thesubstantialvariationfromtheexpertreferenceandmediancontoursurfacesobserved forallusersbefore the intervention (Figs.3.4and3.5 and Table 3.3) suggests that efforts to furtherminimize inter-observer variabilityareimperative.AsthedatainTable3.4demonstrate,substantialinter-observersurfacedeviation was observed for multiple CTVA sub-regions before atlas implementation.Afteratlasadministration,areductionof0.3,0.6,and0.8cmwasachievedfortheupper-anterior, lower-lateral, and lower-posterior CTVA sub-region upper limit of standarddeviationfromthemedianisosurface.Although>1cmwouldstillbeneededtocover95%ofallcontouringvariability,theachievedreductionswouldresultinadecrementin the required planning target volume expansion margins. However, an additionalreductionofvariationisdesired,becausetheplanningtargetvolumemarginsrequiredto encompass the residual variation in target delineation would limit the practicaladvantagesofintensity-modulatedRTcomparedwithconventionalRT. Severallimitationstothepresentpilotstudyareevident.Thesamplesizewaslimited,andonlyasinglecasewascontoured.Theuseofareferenceexpert’scontoursasthede facto reference standardpoints to the fact that the ‘‘ground truth’’ in contouringtheCTVremainsambiguous(Table3.2;notethevariationwithinthereferenceexpertuser’s sequential contours). Some variance in the studymight be attributable to theinstructions,whichweredistinctfromstandardclinicalpractice(e.g.,theexternaliliacnodes are not typically contoured for T3 rectal cases). Our invitationwas limited toSWOGinstitutes,creatingapotentialsamplingbiasandthatonlyinterestedobserversparticipatedcreatedanavenueforselectionbias.Nonetheless,ourdatasuggest thatinclusionofavisualatlasinadditiontowritteninstructionscanimproveconformanceto a reference expert’s contours (Fig. 3.3a) and reduce inter-observer variability to astatistically detectable degree (Fig. 3.3b). However, our data also suggest substantialresidualvariabilityinrectaltargetvolumedelineation,evenafteratlasuse(Tables3.2and3.3).

54 |

Chapter 3

Theresultsofthepresentstudyareconsistentwiththosefrompreviousinvestigationsofeducationalinterventionsandconsensusguidelineapplicationincontouringstudies.Recently, Bekelman et al. [25] demonstrated improvement in contour quality aftera directed teaching intervention, echoing previous work by Tai et al. [26] showingincreasedprotocolcomplianceafterasite-specificeducationalexperience.Withregardto consensus guideline application as an avenue toward target variability reduction,Dimopoulosetal.[32]reportedastudyinwhich19cervicalcancercaseswerecontouredusingtheGroupeEuropéendeCuriethérapieandtheEuropeanSocietyforTherapeuticRadiologyandOncology(GEC-ESTRO)guidelinesby2observers[11,21],witharesultantbetween-userconformityindexintherange0.6–0.7fortargetvolumes,roughlyconsistentwiththeCNs inthepresentseries.Likewise,Wongetal. [38]recentlydemonstrated,usingatest-retestsequence,thatimprovedconsistencyinseromacontouringcouldbeobservedafterexposuretoconsensusguidelines.Intheclinicaltrialsetting,itislikelythat‘‘trial-specific’’atlasesshouldbeusedaccordingtopatternsoffailuredata(asperRoelsetal.[39])or,possibly,afterapilotcontouringtrialsimilartothepresentstudy.For instance, the Radiation Therapy Oncology Group anorectal consensus guidelinesstipulate coverage ‘‘extending CTVA 1cm into the posterior bladder, to account forday-to-day variation in bladder position’’ [10]. This incorporationofmotion intoCTVgeneration,ratherthantheplanningtargetvolumeexpansion,representsaconceptualbreakwithInternationalCommissiononRadiationUnitsandMeasurements6240andotherguidelines[39],inwhichtheposteriorbladderwallwouldnotbecontoured.NousersinGroupAincludedsignificantportionsoftheposteriorbladderbeforeusingtheatlas,althoughmostdidsoafteratlasexposure(incompliancewiththepresentedatlas[10]andconsistentwiththereferenceexpert). Future studies are required to ascertain whether the observed effects of atlasadministrationaretransferabletootheranatomicsiteswithpotentiallymorecomplicatedanatomic relationships [5, 24]. The SWOG Radiation Therapy Committee intends tosuggest building target delineation studies into clinical trial protocol development/qualityassuranceprocesses.Aspectsofthisdatasetmightalsobeintegratedintothedesignof educationalmaterials for aproposedDutch cooperative group rectal studyworkshop.We plan to use portions of this data set to construct composite modelsaccountingforrectalmotionandsetupvariability[14,41],aswellasthedevelopmentofnovelsoftwarestrategiesforevaluation[42]andminimizationoftargetdelineationvariance.

ConclusionsTheadditionofavisualatlasandconsensustreatmentguidelinestoawrittenprotocolincreasedCTVdelineationconformancewiththeexpert-derivedcontoursandincreasedcontouragreementamongtheparticipantsfortheCTVA,butnottheGTVorCTVB,fortheincludedrectalcancercase.Thedetectedinter-observer(bothwithandwithouttheatlas)andintra-observervariationincontouringtargetstructureswassubstantial.Visualatlas-basedsupplementarytargetvolumespecificationmaterialsshouldbeconsideredforclinicaltrialsinvolvingconformalRTapproaches.

| 55

Consensus atlas implementation for rectal cancer target volume delineation

References[1] FoppianoF,FiorinoC,FrezzaG,etal.Theimpactofcontouringuncertaintyonrectal3Ddose-volume

data: Results of a dummy run in amulticenter trial (AIROPROS01-02). Int J RadiatOncol Biol Phys2003;57:573–579

[2] Yamazaki H, Nishiyama K, Tanaka E, et al. Dummy run for a phase II multi-institute trial ofchemoradiotherapy for unresectable pancreatic cancer: Inter-observer variance in contourdelineation.AnticancerRes2007;27:2965–2971

[3] LiXA,TaiA,ArthurDW,etal.Variabilityoftargetandnormalstructuredelineationforbreastcancerradiotherapy: An RTOG multi-institutional and multiobserver study. Int J Radiat Oncol Biol Phys2009;73:944–951

[4] MitchellDM,PerryL,SmithS,etal.Assessingtheeffectofacontouringprotocolonpostprostatectomyradiotherapy clinical target volumes and interphysician variation. Int J Radiat Oncol Biol Phys2009;75:990–993

[5] SenanS,vanSornsendeKosteJ,SamsonM,etal.Evaluationofatargetcontouringprotocolfor3Dconformalradiotherapyinnon-smallcelllungcancer.RadiotherOncol1999;53:247–255

[6] FaulF,ErdfelderE,LangAG,etal.G*Power3:Aflexiblestatisticalpoweranalysisprogramforthesocial,behavioral,andbiomedicalsciences.BehavResMethods2007;39:175–191

[7] Steenbakkers R, Duppen J, Fitton I, et al. Observer variation in target volume delineation of lungcancerrelatedtoradiationoncologist–computer interaction:A ‘‘BigBrother’’evaluation.RadiotherOncol2005;77:182–190

[8] SteenbakkersRJ,DuppenJC,FittonI,etal.ReductionofobservervariationusingmatchedCT-PETforlungcancerdelineation:Athree-dimensionalanalysis.IntJRadiatOncolBiolPhys2006;64:435–448

[9] Duppen JC,SteenbakkersRJHM,Fitton I,etal.Delineationsoftware for teaching in targetvolumedefinition.RadiotherOncol2004;73:S437–S438

[10] MyersonRJ,GarofaloMC,Naqa IE, et al. Electiveclinical target volumes for conformal therapy inanorectalcancer:ARadiationTherapyOncologyGroupconsensuspanelcontouringatlas.IntJRadiatOncolBiolPhys2008;74:824–830

[11] Feuvret L,Noel G,Mazeron JJ, et al. Conformity index: A review. Int J Radiat Oncol Biol Phys2006;64:333–342

[12] van’t Riet A, Mak AC, Moerland MA, et al. A conformation number to quantify the degree ofconformalityinbrachytherapyandexternalbeamirradiation:Applicationtotheprostate.IntJRadiatOncolBiolPhys1997;37:731–736

[13] HoogemanM,vanHerkM,deBoisJ,etal.Quantificationoflocalrectalwalldisplacementsbyvirtualrectumunfolding.RadiotherOncol2004;70:21–30

[14] Nijkamp J, de Jong R, Sonke JJ, et al. Target volume shape variation during hypo-fractionatedpreoperativeirradiationofrectalcancerpatients.RadiotherOncol2009;92:202–209

[15] Jeanneret-SozziW,MoeckliR,ValleyJ,etal.Thereasonsfordiscrepanciesintargetvolumedelineation.StrahlentherOnkol2006;182:450–457

[16] LogueJP,SharrockCL,CowanRA,etal.Clinicalvariabilityoftargetvolumedescriptioninconformalradiotherapyplanning.IntJRadiatOncolBiolPhys1998;41:929–931

[17] NjehCF.Tumordelineation:Theweakestlinkinthesearchforoccuracyinradiotherapy.JMedPhys2008;33:136–140

56 |

Chapter 3

[18] WeissE,HessC.Theimpactofgrosstumorvolume(GTV)andclinicaltargetvolume(CTV)definitionon the total accuracy in radiotherapy theoretical aspects and practical experiences. StrahlentherOnkol2003;179:21–30

[19] BassiMC,TurriL,SacchettiG,etal.FDG-PET/CTimagingforstagingandtargetvolumedelineationinpreoperativeconformalradiotherapyofrectalcancer.IntJRadiatOncolBiolPhys2008;70:1423–1426

[20] CattaneoG,ReniM,RizzoG,etal.Targetdelineationinpost-operativeradiotherapyofbraingliomas:Inter-observervariabilityandimpactofimageregistrationofMR(pre-operative)imagesontreatmentplanningCTscans.RadiotherOncol2005;75:217–223

[21] RaschC,BarillotI,RemeijerP,etal.DefinitionoftheprostateinCTandMRI:Amulti-observerstudy.IntJRadiatOncolBiolPhys1999;43:57–66

[22] Rasch C,Keus R, Pameijer FA, et al. The potential impact of CT-MRI matching on tumor volumedelineationinadvancedheadandneckcancer.IntJRadiatOncolBiolPhys1997;39:841–848

[23] Coles C. Quantitative assessment of inter-clinician variability of target volume delineation formedulloblastoma:QualityassurancefortheSIOPPNET4trialprotocol.RadiotherOncol2003;69:189–194

[24] VorwerkH,BeckmannG,BremerM,etal.Thedelineationoftargetvolumesforradiotherapyoflungcancerpatients.RadiotherOncol2009;91:455–460

[25] Bekelman JE, Wolden S, Lee N. Head-and-neck target delineation among radiation oncologyresidentsafterateachingintervention:Aprospective,blindedpilotstudy.IntJRadiatOncolBiolPhys2009;73:416–423

[26] Tai P, Van Dyk J, Battista J, et al. Improving the consistency in cervical esophageal target volumedefinitionbyspecialtraining.IntJRadiatOncolBiolPhys2002;53:766–774

[27] ChaoKS,BhideS,ChenH,etal.Reduceinvariationandimproveefficiencyoftargetvolumedelineationbyacomputer-assistedsystemusingadeformableimageregistrationapproach.IntJRadiatOncolBiolPhys2007;68:1512–1521

[28] CiernikIF,HuserM,BurgerC,etal.Automatedfunctionalimage-guidedradiationtreatmentplanningforrectalcancer.IntJRadiatOncolBiolPhys2005;62:893–900

[29] Pasquier D, Lacornerie T, VermandelM, et al. Automatic segmentation of pelvic structures frommagneticresonanceimagesforprostatecancerradiotherapy.IntJRadiatOncolBiolPhys2007;68:592–600

[30] van SornsendeKoste JR, Senan S,UnderbergRW,et al.Useof CD-ROM-based tool for analyzingcontouringvariationsininvolved-fieldradiotherapyforStageIIINSCLC.IntJRadiatOncolBiolPhys2005;63:334–339

[31] YushkevichPA,PivenJ,HazlettHC,etal.User-guided3Dactivecontoursegmentationofanatomicalstructures:Significantlyimprovedefficiencyandreliability.Neuroimage2006;31:1116–1128

[32] Dimopoulos JC, Vos VD, BergerD, et al. Inter-observer comparison of target delineation forMRI-assisted cervical cancer brachytherapy: Application of the GYN GEC-ESTRO recommendations.RadiotherOncol2008;89:164–171

[33] GregoireV.CT-baseddelineationoflymphnodelevelsandrelatedCTVsinthenode-negativeneck:DAHANCA,EORTC,GORTEC,NCIC,RTOGconsensusguidelines.RadiotherOncol2003;69:227–236

[34] LawtonC,Michalski J, Elnaqa I, etal.RTOGGU radiationoncology specialists reachconsensusonpelviclymphnodevolumesforhigh-riskprostatecancer.IntJRadiatOncolBiolPhys2008;74:383–387

| 57

Consensus atlas implementation for rectal cancer target volume delineation

[35] PoortmansP,BossiA,VandeputteK,etal.Guidelinesfortargetvolumedefinitioninpost-operativeradiotherapyforprostatecancer,onbehalfoftheEORTCRadiationOncologyGroup.RadiotherOncol2007;84:121–127

[36] Sidhom MA, Kneebone AB, Lehman M, et al. Post-prostatectomy radiation therapy: consensusguidelinesoftheAustralianandNewZealandRadiationOncologyGenito-UrinaryGroup.RadiotherOncol2008;88:10–19

[37] SmallW,Mell LK,AndersonP,etal.Consensusguidelines fordelineationofclinical targetvolumeforintensity-modulatedpelvicradiotherapyinpostoperativetreatmentofendometrialandcervicalcancer.IntJRadiatOncolBiolPhys2008;71:428–434

[38] WongEK,TruongPT,KaderHA,etal.Consistencyinseromacontouringforpartialbreastradiotherapy:Impactofguidelines.IntJRadiatOncolBiolPhys2006;66:372–376

[39] RoelsS,DuthoyW,HaustermansK,etal.Definitionanddelineationoftheclinicaltargetvolumeforrectalcancer.IntJRadiatOncolBiolPhys2006;65:1129–1142

[40] International Commission on Radiation Units and Measurements. ICRU report 62: Prescribing,recording, and reportingphotonbeam therapy. Bethesda: International Commission onRadiationUnitsandMeasurements;1999

[41] IppolitoE,MertensI,HaustermansK,etal.IGRTinrectalcancer.ActaOncol2008;47:1317–1324[42] JenaR,KirkbyNF,BurtonKE,etal.Anovelalgorithmforthemorphometricassessmentofradiotherapy

treatmentplanningvolumes.BrJRadiol2010;83:44-51