Interobserver variability in histological diagnosis of serrated colorectal polyps

w w w . j c o l . o r g . b r

Journal

of

Coloproctology

Original

Article

Interobserver

variability

in

histological

diagnosis

of

serrated

colorectal

polyps

Rosimeri

Kuhl

Svoboda

Baldin

,

Raul

Alberto

Anselmi

Júnior

,

Marina

Azevedo

,

Ana

Paula

Martins

Sebastião

,

Mário

Montemor

,

Luiz

Fernando

Tullio

,

Luiz

Felipe

de

Paula

Soares

,

Lúcia

de

Noronha

b_{PontifíciaUniversidadeCatólicadoParaná(PUCPR),Curitiba,PR,Brazil} c_{UniversidadePositivo,Curitiba,PR,Brazil}

d_{UniversidadeEstadualdePontaGrossa,PontaGrossa,PR,Brazil} e_{ServiceofEndoscopy,HospitalSantaCruz,Curitiba,PR,Brazil}

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received31March2015 Accepted8June2015

Availableonline2October2015

Keywords: Colon Polyps Serrated

a

b

s

t

r

a

c

t

Objectives: Tocomparetheinterobservervariabilityinthediagnosticofserratedand non-serrated adenomatouslesionsandhyperplasticpolyps ofcolonbetweentwogroupsof pathologists.

Methods:310colorectalpolypswerestudied,withhistologicaldiagnosesestablishedbya groupofpathologistscomprisingthreegeneralpathologistsforinitialdiagnosis,andtwo gastrointestinalpathologistsforexpertdiagnosis.

Results:High interobservervariability wasobservedinthediagnosisofserrated polyps, when comparingtheinitial diagnosiswiththeexpert diagnosis(kappa=0.102).For the majorityofbothtraditionalserratedadenomasandsessileserratedadenomas(27/31),a diagnosisofhyperplasticpolypswasestablishedattheinitialdiagnosis.

Conclusions: Pooragreementwasobservedinthediagnosisofserratedpolypsbetweenthe twogroupsofpathologists.Theaccuracyinthediagnosisoftheselesionsisessentialfor thepreventionofcolorectalcancer.

©2015SociedadeBrasileiradeColoproctologia.PublishedbyElsevierEditoraLtda.All rightsreserved.

Variabilidade

interobservador

no

diagnóstico

histológico

de

pólipos

colorretais

serrilhados

r

e

s

u

m

o

Objetivo:Comparar a variabilidade interobservador dos diagnósticosdas lesões adeno-matosasserrilhadasenãoserrilhadasepóliposhiperplásicosdocólonentredoiscomitê depatologistas.

∗ _{Correspondingauthor.}

E-mails:rosimeribaldin@gmail.com,rksbaldin@ufpr.br(R.K.S.Baldin). http://dx.doi.org/10.1016/j.jcol.2015.06.008

Métodos: Foramestudados310póliposcolorretais,diagnosticadoshistologicamenteporum comitêdepatologia,compostoportrêspatologistasgeraisparaodiagnósticoinicialepor doispatologistasgastrointestinaisparaodiagnósticodosespecialistas.

Resultados: Houvealta variabilidadeinterobservador no diagnóstico dospólipos serril-hados, ao seremcomparadosodiagnósticoinicial comodiagnósticodos especialistas (kappa=0,102).Amaioriadaslesõesadenomatosasserrilhadassésseisetradicionais(27/31) foidiagnosticadapelodiagnósticoinicialcomopóliposhiperplásicos.

Conclusões: Houvebaixaconcordâncianodiagnósticodospóliposserrilhadoscolorretais entreosdoiscomitêsdepatologistas.Aacuráciadessesdiagnósticoséfundamentalparaa prevenc¸ãodocarcinomacolorretal.

©2015SociedadeBrasileiradeColoproctologia.PublicadoporElsevierEditoraLtda. Todososdireitosreservados.

Introduction

Serratedpolyps(SP),whichrepresent20%ofcolorectalpolyps, maybeprecursorsof15–20%ofcolorectalcancers.1

ThehistologicaldiagnosisofSPhasbeenimprovedsince 1990,fromthenewconceptsofLongacreandFenoglio-Preiser untilthepresentday,beingnaturalthatdifficultiesarisedue tochangeswhichhadoccurred;however,thelackof preci-sionjeopardizesthepatientfollow-upand undervaluesthe method.2

Thecurrentclassificationandnomenclatureofcolorectal SPrecommendedbytheWorldHealthOrganization3 subdi-videcolorectalserratedpolyps(SP)intothreegroups:

• Hyperplasticpolyps(HP)

• Adenomas/sessileserratedpolyps(A/SSP) - Withcytologicaldysplasia

- Withoutcytologicaldysplasia

• Traditionalserratedadenomas(TSA)

Atfirst,hyperplasticpolyps(HP),consideredasbenign con-ditions,didnotrequire follow-up.Usuallytheendoscopists didnotbothertocarryoutresectionofminorHP-likeinjuries, thankstotheabsenceoftheriskofmalignancy.Butcurrently, inthefaceoftheknownriskofmalignanttransformationof adenomas/serratedsessilepolyps(A/SSP),especiallythosein therightcolonandinthefaceofdifficultiesof differentiat-ingbetweencolorectalSPthroughcolonoscopy,removaland histologicalexaminationoftheselesionsaremandatory.4–6

Unknown factorslead tochangesin thelocation ofthe cryptalproliferativezone,aswellasincrypts’ anchoringin theselesions,withvariantsofhistologicalappearance.7_What

primarilyallowsustodistinguishbetweenthedifferent cat-egoriesofpolypsaresomearchitecturalchangesresultingof theseproliferativeabnormalities.

Accordingtosomeauthors,8_{avariationoccursatthe}

loca-tionoftheproliferativezone,whichusuallyislocatedinthe lowerthirdofcrypts,andthematurecellsarepushedtoward theintestinallumen.

InHPpatients,theproliferativezoneismaintainedinthe lowerportionofthecrypt,occupyingmorethanhalfofthe lengthofthecrypt.Cellscontinuetomaturetowardthe sur-face;butduetoalateapoptosis,thesecellstendtopileup, formingserratedstructures.

InA/PSS,theproliferativezonemovestowardthelateral aspectofthecrypt.Thisallowsthatmaturecellsmoveboth toward the intestinal lumen and to the base ofthe crypt, whichbecomesenlargedanddistorted,becausethe muscu-larismucosaeblocksthecryptalgrowth.

Incasesoftraditionalserratedadenoma(TSA),the proli-ferativezoneislocatedinsmallectopiccryptsthatdevelop atthesideoftheoriginalcrypt.Thesemini-cryptsdevelopa complexpatternofgrowth.

Objective

Tocomparetheinterobservervariabilityofdiagnosesoflarge bowelserratedandnon-serratedadenomatouspolypsandof HP,inparticularintheSPgroup(HP,A/PSSandTSA)between twogroupsofpathologists.

Methods

ThecaseswereselectedfromfilesofreportsfromtheService ofPathologyofHospitalSantaCruz,Curitiba-Pr.Histological slidesfromcollectedpolypsbycolonoscopyduring2008and stainedwithhematoxylin–eosin(H&E)wereused.

This projectwas approved bythe Ethics Committeefor ResearchinHumanBeingsofHC-UFPR,accordingtothetasks setoutinResolutionCNS466/2012andinOperationalNorm No.001/2013ofCNSunderopinionNo.820,432.

Wecollectedclinicaldata,including patientage,gender, polyplocationinthecolon(right,leftorsigmoid),presenceof singleormultiplepolyps,associationwithcancer,polypsize atendoscopy,polypappearance(sessileorpedunculated), ini-tialpathologicaldiagnosis(bytheDepartmentofPathologyof HospitalSantaCruz,composedofthreegeneralpathologists) andexpertdiagnosis(twogastrointestinalpathologists).

Inclusioncriteria(eligiblecases):

- polyps collected by colonoscopyprocedure performed in patientsattheServiceofEndoscopy,SantaCruzHospital, in2008.

- serratedandnon-serratedadenomatouscolorectalpolyps andHP.

- absenceofsamplefragmentation. - absenceoffulgurationartifacts. - non-scarcityofsample.

310 slides (polyps) 507 slides (polyps) 452 tests included 407 excluded (there were no polyps) 859 colorectal exams 197 slides excluded (insufficient material,

with artifacts, etc.)

Fig.1–Sequentialflowchartofcollectionofcasesof colorectalpolyps.

- absenceofinflammatorypolyps.

- absenceofdiagnosticdisagreementamonggeneral pathol-ogists.

- absenceofdiagnosticdisagreementamonggastrointestinal pathologists.

Thesequenceofcollectionofcasesfromthebeginningtill attainingthestudy“n”canbeseeninFig.1.

Areinterpretationoftheslideswascarriedoutbytwo doc-torswithgastrointestinalpathologyspecializationblindedfor thediagnosisinitiallyestablishedbygeneralpathologists.

For statistical evaluation, SP were divided into two groups: the first composed by HP and the second by ser-ratedadenomatouslesions(SAL)comprisingA/PSSandTSA, sincetheselastconditionsclearlydemonstratepotentialfor malignancy.9

Statisticalanalysis

Toevaluatetheagreementbetweendiagnosesofpaired eval-uators, Cohen’s Kappa coefficient was estimated.10 _{For a}

comparisonoftheevaluatorsregardingthelikelihoodof diag-nosis,thebinomialtestwasconsidered.p-Values<0.05were consideredstatisticallysignificant.

Table1–Clinicalandendoscopicdataofcolorectal polypsinthisstudy.

Polyps,total(n) 310 Totalnumberof patients(orof colonoscopy procedures) 240

Ageofpolypcarriers Minimum19years Maximum91years Median60years

Gender 129women

181men Aspectofthepolyp 281sessile(90.6%)

29withpedicle(9.4%) Localization 135ontherightside(rightand

transversecolon)

175ontheleftside(leftcolon andsigmoid)

Singleormultiple 113singlepolyps 197multiplepolyps Size 2–30mm >10mm(22%) 6–9mm(11%) 4–5mm(43%) 2–3mm(24%)

Results

310 casesof colorectal polypsfrom a total of240 patients (somewithmorethanonepolyp)wereconsideredeligiblefor thestudy.Ofthetotalnumberofcases,129polypsinwomen and181inmenwereobserved.Themedianagewas60years, rangingfrom19to91yearsofage.

Astotheendoscopicformofthepolyp,29polyps(9.4%) wereclassifiedbytheendoscopistaspedicledpolyps,and281 (90.6%)assessilepolyps.Endoscopically,thepolypsizeranged from2to30mminitslargestdiameter,withameanof6.5mm perpolyp,withthefollowingdistribution:polyps≥10mm,22% ofcases;6–9mm,11%;4–5mm,43%;and2–3mm,24%.Polyps <2mmwerenotremoved.

As for the location in the colon, 135 polyps were located on the right (ascending and transverse colon) and 175 on the left (descending and sigmoid colon). At endoscopy,113singlepolypsand 197multiplepolypswere found. Thirteen casesof polyp-associated carcinoma were diagnosed.

Table1showspartiallytheresultofdatatabulationof clin-icalandendoscopicdataobtained.

Table2showshowwasthedistributionofdiagnosesbythe pathologycommittee(initialdiagnosis)andby gastrointesti-nalpathologists(expertdiagnosis).

Serratedpolyps

The31SALdiagnosedbygastrointestinal pathologistswere previouslydistributedbythepathologycommitteeasfollows: 27HP(87.1%of31polyps),1TA-LGD(3.2%of31polyps)and3 SAL(9.7%of31polyps).

Fig.2showshowthepathologycommitteediagnosedthose serratedpolypsdiagnosedbygastrointestinalpathologists.

Table2–Histologicaldiagnosisof310colorectalpolyps inthisstudy,establishedbyagroupofpathologyandby expertsingastrointestinalpathology.

InitialDx(group) ExpertDx

HP 87(28.1%) 46(14.8%) SAL 3(0.9%) 31(10%) TA-LGD 172(55.5%) 142(45.8%) TA-HGD 9(2.9%) 5(1.6%) TVA-LGD 22(7.1%) 46(14.8%) TVA-HGD 16(5.2%) 4(1.3%) VA-LGD 0(0%) 25(8.1%) VA-HGD 1(0.3%) 11(3.5%)

HP,hyperplastic polyp; SAL,serratedadenomatous lesion (ade-noma/sessile serratedpolypandtraditional serratedadenoma); TA-LGD,tubularadenomawithlow-gradedysplasia;TA-HGD, tubu-laradenomawithhigh-gradedysplasia;TVA-LGD,tubule-villous adenomawithlow-gradedysplasia;TVA-HGD,tubule-villous ade-nomawithhigh-gradedysplasia;VA-LGD,villousadenomawith low-gradedysplasia; VA-HGD,villousadenoma with high-grade dysplasia;initialDx(group),initialdiagnosismadebythepathology committee;expertDx,diagnosismadebytwoexpertpathologists ingastrointestinalpathology;SALvalueswerehighlighted.

Evaluationofagreementbetweeninitialandexpert diagnoses

We tested the null hypothesis of non-correlation between initialdiagnosticevaluationsandofthediagnosis ofexpert pathologistsversusanalternative hypothesis ofagreement. Table3presentstheresultsobtainedinthestudy.

TheCohen’skappacoefficientwasestimatedat0.102.The statisticaltestresultsindicatednon-rejectionoftheabsence ofanagreementhypothesisbetweenthetwogroupsof evalu-ators(p=0.151).Thus,ourstudydoesnotrejectthehypothesis thatthenon-agreementhasbeenaproductofchance.

Additionally,wetestedanullhypothesisfortheprobability thatadiagnosis ofSALisequalforbothgroups of evalua-torsversusanalternativehypothesisofdifferentprobabilities.

0 5 10 15 20 25 30 27 1 3 Dx Initial LASs AT-DBG PH n.º of cases

Fig.2–Initialdiagnosisof31casesofserratedcolorectal polypsevaluatedbygastrointestinalpathologistsasbeing serratedadenomatouslesions.InitialDx,diagnosticofthe pathologygroup;SALS,serratedadenomatouslesions (adenoma/sessileserratedpolypandtraditionalserrated adenoma);TA-LGD,tubularadenomawithlow-grade dysplasia;HP,hyperplasticpolyp.

Table3–ConcordanceofdiagnosticofSALbetweenthe initialdiagnosisandthatofexperts.

ExpertDx InitialDx Total

SAL OtherDx

SAL 2(0.65%) 29(9.35%) 31(10%)

OtherDx 1(0.32%) 278(89.68%) 279(90%) Total 3(0.97%) 307(99.03%) 310(100%) Dx,diagnosis;SAL,serratedadenomatouslesion.

The test results indicated rejection of the null hypothesis (p<0.001).Thus,thereisevidencethattheexpertgroup diag-nosishasahigherprobabilityofobtainingaclassificationof SAL(10%ofsamplecases)incomparisonwiththeinitial diag-nosis(0.97%ofsamplecases).

Discussion

Therecognitionoftheexistenceofaserratedcarcinogenesis pathwaymakesitimportanttostandardizethehistological diagnosisofSP,asthefollow-upinpatientswithapolypwill dependonitshistologicclassification.1

Todescribetheagreementbetweenobservers,weusedthe statisticindexkappa(k),whichusesmathematicalcoefficients to adjust the agreement to chance. In this index, a value closeto0indicatesanagreementsimilartothatwhichwould be expectedbychance.k values <0.21,0.21–0.40,0.41–0.60, 0.61–0.80and>0.80representpoor,weak,moderate,strong, andverystronginterobserveragreement,respectively.11

Inareviewof20histologicalsectionsofcolorectalpolyps performed for 20 randomlychosen general pathologists, a correctdiagnosisofadenomawasidentifiedin94%ofthe read-ings;ontheotherhand,foracorrectdiagnosisofcaseswith high-gradedysplasia,only47%ofthereadingswerecorrect. HPwascorrectlydiagnosedin75%ofcases.Theauthorsofthat reviewconcludedfortheoccurrenceofdiscrepancymainlyin thediagnosisofhigh-gradedysplasia.12

From the colonic SPgroup, agreater interobserver vari-abilityisexpected,asthisisarelativelynewconceptamong pathologists,andwhichhasbeenperfectedovertime.Inthis study,theagreementofadiagnosisofserratedpolypsbetween thepathologycommittee(initialdiagnosis)andexpert pathol-ogistsresultedinakappavalue=0.102(poor),showingavery lowcorrelationbetweenthesetwogroups.AdiagnosisofLSA wasestablishedin0.97%ofcasesbythecommittee(first diag-nosis)and in10%ofcasesbygastrointestinal pathologists, showingthatthelatterweremorelikelytoestablisha diagno-sisofSALversusgeneralpathologists.

In 2009, investigators carried out an assessment of 40 proximal colonicpolyps(size≥5mm)removed in2001and originallyinterpretedasHPbygeneralpathologistsat Indi-ana University, and later revised by three gastrointestinal pathologists.13 _{In this review, A/PSS diagnoses were}

estab-lished in 85%, 43% and 30% of these polyps by each gastrointestinal pathologist(kappa=0.16). Theauthors con-cluded that many polyps diagnosed as being hyperplastic conditions in 2001 were considered as A/SSP by gastroin-testinal pathologists in2007. Butit was alsonoted alarge

interobserver variation (little agreement on the diagnosis) amongpathologists,similartothevarietyfoundinourstudy: from 31 SALdiagnosed bygastrointestinal pathologists, 27 (87.1%)wereinitiallyclassifiedasHPbythepathology com-mittee.

Otherresearchers14_{evaluated60casesofpolypsdiagnosed}

by4gastrointestinalpathologists,withclassificationof colo-rectalpolypsin5categories:serratedadenoma,hyperplastic polyp,conventional adenoma, mixed polyp,and other ser-ratedpolyps.Acompleteconcordanceoccurredonlyin40% ofcases.Thekvalueforobserverswas0.49.Thekvaluefor thediagnosisofserratedadenomaversusallotherpolypswas 0.38.Theauthorsconcludedthatgastrointestinalpathologists achievedamoderateagreementinthediagnosisof colorec-talserratedpolyps,withweakagreementinthediagnosisof serratedadenomas.

Someauthors1 _{appliedan onlinequestionnaire with20}

colorectal polyps, containing 3 images of each lesion, to 168pathologists, inorder totest the diagnostic variability, especiallyindistinguishingthetwoprecursorlesionsofthe serratedpathwayforcolorectalcancer,TSAandA/PSS, com-paredtoHPandtubular-villousadenomas(TVA).Thecorrect responsesforA/PSS (54%)andTSA(44%)were significantly lowerthanthoseresponsesforTVA(90%)andHP(80%).The authorsmentionthatTSAcouldbeconfusedwithTVAinthose casesinwhichtheserratedarchitecturewaslesspronounced, andinthefaceofaprominentnuclearpseudo-stratification andhyperchromasia.

Some considerations can be proposed with respect to the growing curve of diagnoses of A/PSS by the pathol-ogy committee in relation to that of gastrointestinal pathologists:

- Thediagnosisestablishedbythecommitteewascarriedout earlier(2008),andtheexpertdiagnosiswasmorerecently established(2013)andthereforemoreupdated.

- Thegreaterexperienceanddeeperunderstandingofcriteria forthediagnosisofserratedpolypsamonggastrointestinal pathologists.

Therefore,wecanconcludethat,betweenthetwogroups studied,theinterobservervariabilityinthediagnosisof colo-rectalserratedpolypswashigh,possiblyduetodifferencesin thedatesofdiagnoses,suspicionlevel,andabetterknowledge ofmorphologicalcharacteristics bygastrointestinal pathol-ogists. The relevance of this issue is important, because the follow-updepends solelyon thehistological diagnosis. Increasingthe qualityofpreventionofcolorectalcancer in the community can be achieved with a more consistent

diagnosisofthesepolyps.Thevalueofthisstudyisthesense ofself-criticismthatboostedtheirauthors.

Conflicts

of

interest

Theauthorsdeclarenoconflictsofinterest.

r

e

f

e

r

e

n

c

e

s

1.GlatzK,PrittB,GlatzD,HartmannA,O’BrienMJ,BlaszykH. Internet-basedassessmentofobservervariabilityinthe diagnosisofserratedcolorectalpolyps.AmJClinPathol. 2007;127:938–45.

2.FoucarE.Erroridentificationasurgicalpathologydilemma. AmJSurgPathol.1998;22:1–5.

3.RexDK,AhnenDJ,BaronJA,BattsKP,BurkeCA,BurtRW,etal. Serratedlesionsofthecolorectum:reviewand

recommendationsfromanexpertpanel.AmJGastroenterol. 2012;107:1315–29.

4.LiebermanDA,PrindivilleS,WeissDG,WillettW,VA CooperativeStudyGroup380.Riskfactorsforadvanced colonicneoplasiaandhyperplasticpolypsinasymptomatic individuals.JAMA.2003;290:2959–67.

5.JassJR,BakerK,ZlobecI,HiguchiT,BarkerM,BuchananD, etal.Advancedcolorectalpolypswiththemolecularand morphologicalfeaturesofserratedpolypsandadenomas: conceptofa‘fusion’pathwaytocolorectalcancer. Histopathology.2006;49:121–31.

6.O’BrienMJ.Hyperplasticandserratedpolypsofthe colorectum.GastroenterolClinNAm.2007;36:947–68. 7.HiguchiT,JassJR.Myapproachtoserratedpolypsofthe

colorectum.JClinPathol.2004;57:682–6.

8.SnoverDC.Uptadeontheserratedpathwaytocolorectal carcinoma.HumPathol.2011;42:1–10.

9.ChristopheR,DavidGH,IanSB,BarbaraAL,VickiLJW. Serratedpolypsofthelargeintestine:currentunderstanding ofdiagnosis,pathogenesis,andclinicalmanagement.J Gastroenterol.2013;48:287–302.

10.LandisJR,KochGG.Themeasurementofobserveragreement forcategoricaldata.Biometrics.1977;33:159–74.

11.SimJ,WrightCC.Thekappastatisticinreliabilitystudies: use,interpretation,andsamplesizerequirements.PhysTher. 2005;85:257–68.

12.RexDR,AlikhanM,CummingsO,UlbrightTM.Accuracyof pathologicinterpretationofcolorectalpolypsbygeneral pathologistsincommunitypractice.GastrointestEndosc. 1999;50:468–74.

13.KhalidO,RadaidehS,CummingsOW,O’BrienMJ,Goldblum JR,RexDK.Reinterpretationofhistologyofproximalcolon polypscalledhyperplasticin2001.WorldJGastroenterol. 2009;15:3767–70.

14.WongNA,HuntLP,NovelliMR,ShepherdNA,WarrenBF. Observeragreementinthediagnosisofserratedpolypsofthe largebowel.Histopathology.2009;55:63–6.