ContentslistsavailableatScienceDirect
Pathology
–
Research
and
Practice
j o ur na l h o me p a g e :w w w . e l s e v i e r . c o m / l o c a t e / p r p
Original
article
Her2
immunohistochemical
evaluation
by
traditional
microscopy
and
by
digital
analysis,
and
the
consequences
for
FISH
testing
Thomas
Marcuzzo
a,
Fabiola
Giudici
a,∗,
Elisa
Ober
b,
Clara
Rizzardi
a,b,
Cristina
Bottin
a,
Fabrizio
Zanconati
a,baDepartmentofMedical,SurgicalandHealthScience–UniversityofTrieste,StradadiFiume447,34149Trieste,Italy
bUCOdiAnatomiaeIstologiaPatologica–AziendaSanitariaUniversitariaIntegratadiTrieste,StradadiFiume447,34149Trieste,Italy
a
r
t
i
c
l
e
i
n
f
o
Articlehistory: Received30June2015 Keywords: Breastcancer Her2 Digitalanalysis Immunohistochemistry Fluorescencea
b
s
t
r
a
c
t
Aim:Her2proteinisthekeymarkerdeterminingthechoiceofHerceptintherapyafteradiagnosisof breastcancer.Itsevaluationismadeinmostlaboratoriesbyimmunohistochemistry,andinterpretedby apathologistusinganopticalmicroscope,aprocesssubjecttointer-observervariability,particularlyfor samplesscoredasequivocal(2+).Softwareanalysisproductshavebeenintroduced,seekingtoreducethis variability.Inthisstudy,wecomparedtheresultsofbothtraditionalevaluationandaspecificsoftware package(VISIAImaging)tothosefromfluorescentinsituhybridization(FISH).
Materialsandmethods:Weselected176casesofinvasivebreastcancersampledduring2012–2014that wereclassifiedasequivocalafterevaluationofHer2immunohistochemistry,andthatwerealsoevaluated byFISH.EachtissueslidewasscannedwithadigitalD-SightFluo2.0microscopeandanalysedwithVISIA ImagingS.r.l.software.Thefinalresultswerecategorisedasfollows:negative(0–1+),equivocal(2+),or positive(3+).TheneachresultwascomparedtothatobtainedbyFISH.
Result:Thedigitalmethodconfirmed85samples(48.3%)asequivocal(2+),while23(15.1%)were reclas-sifiedasnegative(1+)and44(28.9%)aspositive(3+).Ofthe176cases,24(13.6%)werenotsuitable fordigitalanalysis(inadequate).Of67reclassifiedcases(1+or3+),62wereinagreementwithFISH results(concordancerate92.5%).Thesensitivityandspecificityofthedigitalmethodwere100%and82%, respectively.
Conclusion:Theapplicationofthisanalysissoftwareledtoanimprovementintheinterpretationofcases classifiedasequivocal,decreasingtheneedforFISHandincreasingdiagnosticcertainty.
©2016TheAuthors.PublishedbyElsevierGmbH.ThisisanopenaccessarticleundertheCCBY license(http://creativecommons.org/licenses/by/4.0/).
1. Introduction
Theexpressionofthehumanepidermalgrowthfactor2(Her2)
membraneproteinisakeymarkerintheprognosticandpredictive
assessmentofbreastcancer[1].Itsevaluationisgenerallymade
usingimmunohistochemicaltestsinterpretedbyapathologist,and
usinginsitu hybridisationtesting foruncertaincases[2].
How-ever,apathologist’sevaluationofopticalmicroscopyisasubjective
parameter,opentointer-observervariability[3].Manyeffortshave
beenmadetofindamoreobjectiveprocedure,includingdigital
analysis[4–8].Inthisstudy,adigitalanalysissoftwareproductfor
Her2immunohistochemicalevaluationwastested.Thesoftware
usesslideimagespreviouslyacquiredwithadigitalmicroscopeand
∗ Correspondingauthor.
E-mailaddress:fgiudici@units.it(F.Giudici).
semiautomaticanalysisofthevirtualslide,classifyingthesample
aspositive(3+),negative(0–1+),orequivocal(2+).Weevaluated
whethertheapplicationofthissoftwareanalysiscouldleadtoan
improvementintheinterpretationofequivocalcases,decreasing
theuseofFISH.Weexaminedtheconcordanceratebetweenthe
resultsobtainedbydigitalanalysisandbyFISH.
2. Materialsandmethods
We selected 176 cases of invasive breast cancer diagnosed
during2012–2014,classifiedasequivocalafterHer2
immunohisto-chemicalslideevaluation.AconfirmingFISHtestwasmadeonlyon
equivocalcases[1].Allslideswereselectedfromthearchiveofthe
InstituteofPathologicalAnatomyandHistologyofthe
Academic-Hospitalpublic-servicecorporation“OspedaliRiuniti”ofTrieste,
Italy.Immunohistochemistrywasperformedafterprevioustissue
fixation(10%neutralbufferedformalin)andparaffinembedding.
http://dx.doi.org/10.1016/j.prp.2016.07.016
Table1
Immunohistochemicalprocedure.
√
Antigenretrieval:applicationofaheatprocedurewithPT-Link instrument(Bioptica)andW-CAPEDTABufferpH8.0(Bioptica): Preheat:heatingofthebuffer,itstemperatureincreasesfromroom
temperatureto65◦Cin20min(min).
Retrieval:thetemperatureincreasesto95◦Candbecomesstablefor
20min;
Cool:coolingofthebufferfrom95◦Cto65◦Cin20min; √
Endogenousperoxidaseinhibition:H2O2at3%(Bioptica)for10min;
√
Primaryantibodyincubation:cloneMIB-1(Bioptica),1:200dilution for20min;
√
Secondarysystem:UltravisionQuantolargevolumedetectionkit-HRP (ThermoScientific):
Enhancerfor10min; Polymerfor10min; √
Chromogen:3,3diaminobenzidine(DABQuanto)for5min
(ThermoScientific); √
Counterstaining:Mayerhaematoxylinfor3min; √
Dehydration:arisingsequenceofalcohol; √
Clarificationwithxylene;
√
Assembling.
Allimmunohistochemicaltestswereperformedusinganautomatic
stainer(Lab Vision,ThermoScientific),following theprocedure
showninTable1.Eachstepwasfollowedbywashingwith
Tris-bufferedsalinecontainingTween20(ThermoScientific).AfterJuly
2014,anotherwashingstepwithrunningwaterwasaddedafter
counterstaining,toimprovethehaematoxylinresults.Everyslide
had been evaluated by a pathologist onan optical microscope
withalowmagnificationandclassifiedasequivocal,inagreement
withthemostrecentguidelinesoftheAmericanSocietyof
Clini-calOncologyandtheCollegeofAmericanPathologists(ASCO-CAP
2013).Cases evaluatedbeforethe2013 guidelinescut-off date,
which had been classified using the previousset of guidelines
[9],wererevaluated,andallwereconfirmedasequivocal.Each
slideunderwentdigitalacquisition,usingaD-SightFluo2.0
dig-italmicroscope(A.MenariniDiagnostics,Firenze,Italy)andwas
analysedwithVISIAImagings.r.l.software(version2.5.0.1,San
Gio-vanniValdarno,Italy).Thedigitizingofasliderequiresacquisition
withalow-magnificationlens(4×)tocreateaslidepreview,on
whichtheareaofinteresttobescannedwithahigh-magnification
lens(20×)isselected.Withthisprocedure,adigitalequivalentof
theslideisobtained, andit canbedisplayedat any
magnifica-tionfrom0to20×,maintaininganexcellentimagequality.The
importanceofusingahigh-magnificationlens isthatthe
analy-sissoftwarerequiresahigh-resolutionimage(notlessthan20×
magnification).EveryequivocalcaseunderwentaFISHconfirming
test(PathVysionHer2DNAProbeKitII,AbbotMolecular)evaluated
using a fluorescent microscope,following international criteria
(ASCO-CAP2013).Resultscollectedbeforetheadoptionofthese
guidelineswere reevaluated, and no changes wereseen in the
finaldiagnosisforanycase.AfterJuly2014,fluorescentslideswere
scannedwithaD-SightFluo2.0digitalmicroscope,andanalysed
withspecificsoftware.
Digitalslidesweredividedintocategoriesbasedon:
• Typeofspecimen:Surgicalsampleorbiopsy.
Thedifferenceinthequantityoftumoursamplethatcanbe
foundinthetwodifferenttypesofspecimenisimportant,asisthe
subsequenthandlingofsamplesduringallthestepspreceedingthe
immunohistochemicalstainingtoachieveaperfectreactionquality
[10].
• Suitabilityfordigitalanalysis:Inadequateoradequate.
Table2
Parametersfordigitalanalysis.
Parameter Range Function
Onlycellswithnucleus ON/OFF Countsonlythemembrane associatedwithadetected nucleus.
Threshold 0–255 Detachmentbetween backgroundandnucleionthe haematoxylinchannel. Minimumarea 0–6000 Minimumareaunderwhich
theobjectisidentifiedas nucleus.
Crumbling 0–255 Settlesthesensibilityonthe membraneseparation. Threshold+vs− 0–255 Settlesthethresholdof
membranecompletenessto classifycellsasnegative(0)or positive(1+,2+,3+).
Threshold2+vs3+ 0–255 Settlestheaverageintensityof completemembrane classificationasstrong(3+)or moderate(2+).
Someoftheslidesinourcollectiondidnotmeetthe
require-mentsoftheanalysissoftware,mostlyduetopoorqualityofthe
immunohistochemicalsignal.Thesecaseswereclassifiedas
inad-equate(technicalissues).
• Profile:A,B,C,D,E,F,G.
Theanalysissoftwareusesasequenceofparametersthatallow
evaluationofstainingintensityandofthepercentageofthecancer
cellmembranethatisstained(Table2).Anaccurateanalysisofthe
sliderequiresthattheseparametersbecarefullychosentoavoid
misclassificationofcells.Thisisdirectlyaffectedbythequalityof
theimmunohistochemicalreaction,andbytheintensityofboththe
haematoxylinstainingofthenucleiandthe3,3diaminobenzidine
(DAB) staining and its distribution.Specific profiles have been
developedbasedonslideswiththesamestainingquality,with
ref-erencetoarepresentativeslide,fromAtoG.Everyscannedslide
wasevaluatedatlowmagnificationandenteredinthe
appropri-ateanalysiscategory.Aspecificprofile(F)wasdevelopedonlyfor
biopsy,evenifthestainingoftheslideisidenticaltothesurgical
specimenthatunderwentBprofileanalysis.Thisisduetothe
differ-entnatureofthetwosamples,biopsieshavefewertissuequantity
anddifferenthistologicaltissuetreatment(fixationand
process-ing)comparedtosurgicalspecimens,causingstainingdifferences
thattheanalysissoftwarecandetect,astheyarenotvisibletothe
nakedeye.Itisreallyimportanttoseparatethetwoprofilestoavoid
misclassificationoftheslide.Overall,sevendifferentprofileswere
defined(Table3),eachwithitsrepresentativeslide(Image1),to
maketheanalysisreproducibleandreliable.
• Stainingdistribution:Homogeneous,heterogeneous.
DABstainingintensitydistributionisanessentialparameterin
thechoiceoftheanalysisarea.Eachslidewasvisualizedatlow
magnification(max4×)andthestainingintensitywasevaluated.
Ifthedistributionwashomogeneous,theanalysisareawaschosen
randomlyontheslide.Ifthedistributionwasheterogeneous,two
separateareaswerechosen,oneinthemostintenseregion(hot
spot),anotherintheweakest,containingapproximatelyan
equiv-alentnumberofcells.Theseareaswereanalysedseparatelyanda
finalaverageresultfrombothwasconsidered.
Whentheseparameterswereset,theimageanalysiswas
per-formedusinga20×magnification.Itwasdecidedtolimitthearea
ofanalysisto20mm2(4×field),duetothetime-consumingnature
Table3
Profileforthedigitalanalysis.
Profile Parameters Value
A
Greynuclei,weak intensity Representativeslide B13-1404I
Onlycellswithnucleus Threshold Minimumarea Crumbling Threshold+vs− Threshold2+vs3+ OFF 1 10 25 137 77 B
Intensebluenuclei Representativeslide B14-8580/1B
Onlycellswithnucleus Threshold Minimumarea Crumbling Threshold+vs− Threshold2+vs3+ OFF 40 120 25 167 87 C
Faintslide,softDAB Representativeslide B12-772E
Onlycellswithnucleus Threshold Minimumarea Crumbling Threshold+vs− Threshold2+vs3+ OFF 1 10 25 172 102 D Faintslide, well-definedDAB Representativeslide B12-9154G
Onlycellswithnucleus Threshold Minimumarea Crumbling Threshold+vs− Threshold2+vs3+ OFF 1 20 10 187 117 E Grey/bluenuclei, cytoplasmicor intenseDAB Representativeslide B12-13575/2C
Onlycellswithnucleus Threshold Minimumarea Crumbling Threshold+vs− Threshold2+vs3+ OFF 1 10 25 140 65 F
Biopsy,Intenseblue nuclei
Representativeslide B14-9699
Onlycellswithnucleus Threshold Minimumarea Crumbling Threshold+vs− Threshold2+vs3+ OFF 40 10 10 140 60 G
Strongintensitymix slide(A/E) Representativeslide B12-15122/2o B14-1761
Onlycellswithnucleus Threshold Minimumarea Crumbling Threshold+vs− Threshold2+vs3+ OFF 1 10 25 97 35
havetakenanexcessiveamountoftime,andthenumberofcellsin
anareaof20mm2wasexpectedtobehighenoughtobe
represen-tativeofthewholesection.Thiswasdemonstratedtobethecase
(seeResultsbelow).Theanalysiswasbasedontheidentification
ofDAB-positivecellmembranesandonthepercentageofthecell
membranestained.Sampleswereclassifiedintofourcategories,as
follows:
• 0Negative:nopresenceofmembranestaining,theanalysisfinds
onlythehaematoxylinstainingofthenuclei.
• 1+Negative:weakandincompletestaining.
• 2+Equivocal:moderateorincompletemembranestaining.
• 3+Positive:strongandcompletemembranestaining.
Thisclassificationis equivalenttothat recommended inthe
most recent published guidelines [1]. Every cell is categorized
accordingtothesecriteriaandthesoftwarereportsthe
percent-ageof cellsineachcategory fromthetotalnumber ofanalysed
cells.Thefinalresultisscoredbasedonapositivitycut-offsetat
10%ofthetotal cellnumber,asnegative(0–1+),equivocal(2+),
orpositive(3+).Resultsfromeachcasewerecomparedtothose
obtainedbyearlierFISHtests,whichareconsideredthediagnostic
goldstandard.
• Numberofcells:Adequate,inadequate.
Whenthetotalnumberofanalysedcellswaslessthan10,000,
thewholeanalysiswasconsideredunreliable.Weobservedthat
with fewer cells the positive/negativepercentage wasstrongly
influencedbythenumberofcellsanalysed,withsmalldifferences
leadingtolargechanges.Becausethisvariationstartedtostabilize
above10,000cells,weusedthis numberasareliability
thresh-old.Ifasingleareadidnotcontainenoughcells,theanalysiswas
extendeduntiltheminimumthresholdwasreached.Ifinsufficient
cellswerepresentwithinthewholesection,theslidewas
consid-eredinadequate(<10,000).Totestthesuitabilityoftheselected
threshold,weexaminedthetotalnumberofcellsanalysedforall
areasofeachslide.Alloftheanalysedareaswereincludedinthis
examination,thusoneforeveryhomogeneouscase(129)andtwo
foreachheterogeneouscase(23).Furthermore,wealsoexamined
cellnumbersonslidesscoredasinadequate(<10,000),forafinal
totalof191analysedareas.
3. FISH
Aspreviouslydescribed,theconfirmingFISHtestneedsits
anal-ysiswithafluorescentmicroscope.Theanalysisstartedwithan
observationoftheslidebythepathologistwithaDAPIfilter.This
allowedhimtoevaluatethenucleimorphologyandeventualnuclei
overlapping,choosingthesuitableone.ThentheDAPIfiltermust
beswitchedwiththeSpectrumOrangetodetectHer2/neusignals
andcountthem,losingthenuclearmorphologicaldetailthatthe
observermustkeepinmind.Thenthefilterisswitchedagainto
theSpectrumGreen,tocountCep17signals.Boththecountswere
signedandtheobservermovedonanothernucleus.Thisprocess
continueduntiltheachievementoftheminimalnucleinumber,not
lessthan20nuclei[1].AfterJuly2014,theevaluationoftheFISH
resultswasmadeusingthedigitalequivalentoftheslide:weused
theD-SightFluo2.0digitalmicroscopetocreatedigitalfluorescent
slide,andthenanalysedthemwiththespecificanalysissoftware
madebyVISIA.Thissoftwareisanupgradeofthesameusedina
recentarticle[11]aboutstandardizationofcompletelyautomated
FISHanalysis.Theanalysiswasmadeonselectedareasafter
exam-inationofallfluorescentslides.Thesoftwarescannedtheselected
fieldsat 60× magnification.A mercurylamp poweredthelight
directedonthesampleandaDAPI
(4,6-diamidino-2-phenylindol)-stainedimagewasacquiredtoidentifynuclei,usinganautomated
perfectfocussystem.Thesoftwareproceededtoscansignalsfrom
probesforbothHer2/neu(SpectrumOrange)andCEP
(Spectrum-Green),usingappropriatefilters,andcollectingdatafromagreater
numberofZ-levelscomparedtotheDAPIimages.Levelnumber
isadjustable,butforthesesamplesitwasfixedat13,separated
fromeachotherby1m.Thesoftwarecompressedthescanned
imagesintoonethattheusercanscrollthrough,removing
back-groundfrombothprobes.Thepathologistgainedtheadvantagesof
observingtheslideonascreen,andnotdirectlyonthefluorescent
microscope,withahighimagequalityincludingallfiltersand
with-outneedoffocusing[12].Thesoftwareautomaticallyhighlighted
therecognizablenucleiandtheredandgreenprobes.The
pathol-ogisthadtocheckeverynucleusandeverysignaltocontrolthe
suitabilityoftheanalysis.Ifneeded,hecoulderase,modifyoradd
nucleiorprobessignalswithadedicatedmanualcontrolsystem.
Thesoftwarecontinuouslyupdateditsfinalresultsbasedoneach
probeandnucleusacceptedbythepathologist.Thisallowedafast,
effective,andreproducibleanalysisoftheselectednuclei.Itwas
alsopossibletostorehigh-qualityimagesofeachsample,avoiding
lossofsampleinformationduetofluorescencedecay.Furthermore,
digital analysisallowedtheobservationofa greater numberof
nucleicomparedtodirectopticalfluorescentmicroscopy
evalua-tion.Themostrecentguidelines[1]changedtheminimumnumber
ofnucleithata pathologistmustcounttogivea reliableresult,
Image1. Representativeslides.
thisreduction.Eachofthe49Her2FISHdigitisedcaseswascounted
multipletimes,startingfromaminimumof20andincreasingthe
numberevaluateduntilthemaximumanalysablenucleiwithinthe
selectedareaswasachieved(20,30,40,50,60,max).Thehighest
numberofanalysablenucleirangedfrom65to384.Resultsusing
differentcellcountswerecompared,todetermineifithad
influ-encedthefinalscoring.Inthesecases,thefinalresultwasobtained
byagreementoftwopathologists,expertsinfluorescence
interpre-tation.
4. Results
4.1. Cellnumberdistribution
Cellnumbers withintheanalysed areas rangedfrom987to
127,933,witha meanof35,483.26;thestandarddeviationwas
25,844.91.Furthermore,50% ofthesamplescontained between
16,793and46,304cells.
4.2. Generalresults(Table4)
Ofthe176totalcases, 24wereinadequate(12 fortechnical
issuesand12forcellnumbers<10,000),85(48.3%)wereconfirmed
asequivocal,23 and44 wereclassifiedasnegative(13.1%) and
positive(25.0%),respectively.
In comparison tothe gold standard method of FISH the 85
equivocalcasesconfirmedasequivocal,67 caseswerecorrectly
reclassifiedaspositiveornegative,whilst5caseswerediscordant
comparedtoFISH.Thesecaseswereclassifiedasamplifiedby
dig-italanalysisandnotamplifiedbyFISH.Ofallourclinicalrecords
(176)thedigitalanalysisof62casesisconcordantwiththeFISH
Table4
CharacteristicsofslidesclassificationafterHER2immunohistochemistrydigital analysis(176cases). Characteristic N % DiagnosticCategories Inadequate 24 14% Negative(1+/0) 23 13% Positive(3+) 44 25% Equivocal(2+) 85 48% Total 176 100% TypeofSpecimen SurgicalSample 132 75% CoreBiopsy 44 25% Total 176 100%
SuitabilityforIHCdigitalanalysis
Inadequate 24 14% Adequate 152 86% Total 176 100% StainingDistributiona Homogeneous 139 85% Heterogeneous 25 15% Total 164 100%
ConcordanceRatebetweenIHC DigitalAnalysisandFISHtestb
FISHconcordant 62 41%
FISHdiscordant 5 3%
Equivocal(2+) 85 56%
Total 152 100%
aExcluded12casesfortheimpossibilitytoevaluatethedistribution. bExcluded24inadequatecases.
withFISHfindingsThesefivecaseswillbedescribedindetailinthe
discussion.
The24samples(14%ofcases)thatwereinadequatefordigital
analysiswereexcludedfromsomeofthefollowinginvestigations,
decreasingtheclinicalrecordsto152.
4.3. Sensitivityandspecificityofthedigitalmethod(Table5)
InthecomparisonbetweenFISHanalysisanddigitalanalysis
weexcludedthe85equivocal(2+)casesandalsothe24
inade-quateones.Thereforetheanalysisisfocusedon67cases,ofwhich
39showedamplification byFISH,while 28werenot amplified.
WiththedigitalIHCevaluation,inmostcasestherewas
concor-dancewiththeFISHevaluation(62casesof67),exceptforfive
cases(falsepositives)forwhichtheinstrumentidentified
hyper-expressionwhereitwasnotamplified(positivepredictivevalueof
thedigitalmethodVPP=89%[95%CI78–95%]).Thesensitivityand
specificityofthedigitalmethodwere100%(95%CI[93–100%])and
82%(95%CI[71–90%]),respectively.
Table5
ComparisonbetweenFISHAnalysisandDigitalAnalysis(excluded85equivocaland 24inadequatecases).
DigitalIHCAnalyis
Positive(3+) Negative(1+) Total
FISHAnalyis Amplified 39 0 39
NotAmplified 5 23 28
Total 44 23 67
4.4. Comparisonbetweentypesofspecimen(Table6)
Therecordsincluded132(75%)surgicalspecimensand44(25%)
biopsies.Thewerenostatisticaldifferencesbetweenthe
propor-tionsofinadequatesamplesinbiopsiesandinsurgicalspecimens
(18%vs12%,p=0.47).However,therewereanumberofsamples
thatremainedequivocalafterdigitalanalysis,andmoreofthem
werefoundamongsurgicalspecimensthaninbiopsies.
Evaluat-ingtheconcordancepercentage withFISHanalysiswefoundit
tobeslightly higheramongbiopsysamplesbut notstatistically
significant(50%vs38%,p=0.27).
4.5. Comparisonbetweenstainingdistributions(Table7)
Among the analysis of 164 samples (12 were excluded for
theimpossibilitytoevaluatethedistribution),139(85%)showed
homogeneousstaining.Furthermore, inboth homogeneousand
heterogeneousstaining categories,there wasa highnumber of
equivocalcases.Focusingonthenon-equivocalcases,in
homoge-neoussamples,thepercentageofFISH-discordantcaseswas2%(1
case in58),whileinheterogeneouscases,whichwerefewerin
number,thediscordantcasepercentagewas17%(p=0.005).The
singlehomogenousdiscordantcasewasabiopsy,whileallfour
heterogeneousdiscordantcasesweresurgicalspecimens.
4.6. Comparisonamongthreeyears(Table8)
Wedividedthefinalresultsintothethreeyearsofthestudy,and
examinedthemseparately.Thenumberofcasesclassifiedas
equiv-ocaldecreasedovertime,from50%oftotalcasesin2012to42%
in2014.Thepercentageofcaseswithnon-equivocalfinalresults,
concordantwithFISH,increasedfrom22%in2012to44%in2014.
However,asmallpercentageofdiscordantcaseswaspresent
dur-ingallyears.Thefractionofcasesscoredasinadequatefortechnical
reasonsdecreasedfrom12%to5%duringthethreeyearsexamined
(Image2).
Table6
DigitalAnalysisaccordingtoTypeofSpecimens.
Characteristic SurgicalSample CoreBiopsy p-value p-trend
DiagnosticCategories 0,43 Inadequate 16(12%) 8(18%) 0,47 Negative(1+/0) 16(12%) 7(16%) 0,69 Positive(3+) 32(24%) 12(27%) 0,84 Equivocal(2+) 68(52%) 17(39%) 0,19 Total 132 44
ConcordanceRatebetweenDigitalAnalysisandFISHtesta SurgicalSample CoreBiopsy p-value p-trend
FISHconcordant 44(38%) 18(50%) 0,27 0,44
FISHdiscordant 4(3%) 1(3%) 1
Equivocal(2+) 68(59%) 17(47%) 0,31
Total 116 36
Table7
DigitalAnalysisaccordingtoStainingDistributions.
Characteristic Homogeneous Heterogeneous p-value p-trend
DiagnosticCategories p=0.83 Inadequate 10(8%) 2(8%) 1 Negative(1+/0) 21(16%) 2(8%) 0,45 Positive(3+) 37(29%) 7(28%) 1 Equivocal(2+) 71(55%) 14(56%) 1 Total 139 25
ConcordanceRatebetweenDigitalAnalysisandFISHtest* Homogeneous Heterogeneous p-value p-trend
FISHconcordant 57(44%) 5(22%) 0,07 <0.001
FISHdiscordant 1(1%) 4(17%) 0.005
Equivocal(2+) 71(55%) 14(61%) 0,77
Total 129 23
* Inadequatecaseswereexcluded.
Table8
Resultcomparison:2012–2014.
D-sightresults 2012 2013 2014
InadequatecellN◦<10.000 6 12% 2 3% 4 6%
Inadequatefortechnicalpitfall 6 12% 3 5% 3 5%
Equivocal 26 50% 33 52% 26 42%
FISHconcordant 11 22% 24 38% 27 44%
FISHdiscordant 2 4% 1 2% 2 3%
Totalamount 51 100.0% 63 100.0% 62 100.0%
4.7. FISH
Ofthe49casesanalysedbyFISHwiththedigitalprocedure,41
showednovariationinresultsbasedonthedifferentnumbersof
totalnucleicounted.Theremainingeightcasesdidshow
differ-encesbetweenresultsobtainedfromcountingthefirst20nuclei
andfromoneothertotalcountcategory.Moreover,weobserved
thatonlythe40-nucleicountsgavethecorrectresultinallcases.
5. Discussion
WecomparedtheresultsoftraditionalHer2IHCevaluationwith
aspecificsoftwarepackage(VISIAImaging)forHer2IHCanalysis,
usingFISHasgoldstandard.Toclearlyinvestigateiftheapplication
ofdigitalIHCanalysissoftwarecouldleadtoadecreaseofFISHtest
andmorerapid,reproducibleandreliableIHCresults,wediscuss
eachstepseparately.
5.1. Cellnumberdistribution
Aminimumof10,000cellsappearedtobeanidealcut-offfor
reliabilityofthisanalyticdata,confirmingearlieranecdotal
evi-dence.Indeed,morethat75%ofcasesinthisstudyprovidedhigher
cellnumbers,and50%ofsamplescontainedbetween16,000and
46,000cells.Therefore,weconsidertheanalysiswehaveconducted
tobereliable.Resultfromcaseswithfewercellnumbershouldnot
beconsidered.
5.2. Generalresults
Somesamplesmaynotbesuitablefordigitalanalysis,dueto
characteristicsofthesample,asin thecase oflow cellnumber
previouslydescribed,ortotechnicallimitations.Itisimportantto
separatethesetworeasons,asthefirstisanindependentvariable
(becausewecannotdeterminethenumberofcancercellspresent
inasample)whiletheotherisdeterminedbythequalityofthe
immunohistochemicalassay.Wehavedemonstratedthatresults
in35%ofallequivocalcases,withevaluationbyopticalmicroscopy,
couldbebetterdefinedbydigitalanalysisavoidingFISH.Inaroutine
useperspectivethissystemwilldecreasebyathirdthe
fluores-cenceconfirmingtest.Thediagnosticcontributionthatthedigital
methodbringsonthismatterisundoubtedlyrelevant.
5.3. Sensitivityandspecificityofthedigitalmethod
Digitaltestappearedtohave100%sensitivity,duetotheabsence
offalsenegatives.However,thepresenceofFISH-discordantcases
cannotbeignored.Inthewholesetofsamplestherewerefive
false-positiveswhichdecreasedthetestspecificityto82%.
5.4. Comparisonbetweentypesofspecimen
Theinadequatepercentageissimilarinbothcategories,evenif
itwasslightlyhigherinbiopsies.ConsideringthefiveFISH
discor-dantcasesindetail,fourofthemweresurgicalspecimens,while
onewasabiopsy.Biopsiesrepresentasmallernumberofsamples
comparedtosurgicalspecimens,buthaveaslightlyhigher
frac-tionofinadequatesamples.Thisisinherenttothesamples,which
presentcomplexchallengestobothpre-analyticstandardization
andimmunohistochemicalmethods.Furthermore,mostbiopsies
samplescontainevidentartefactsofover-orunder-stainingdue
tosamplehandling,oftenfocused onbordersorseen in
small-dimensionsamples.Thedifferencebetweenbiopsiesandsurgical
samplesimmunohistochemistry couldbe of great influence for
somemarkersandthereisvarianceevenforHer2[13].This
dif-ferencehasbeenreported alsoinotherstudies onothertissue
samples[14].Guidelinesrecommendalargegroupofsituationsin
whichIHCtestonasinglebiopsysampleisnotadequateandnew
Her2testonthesurgicalspecimenmustbedone[1].Thecertainty
ofcorrecttissuehandlingandofIHCmethoddecreasestherisks
ofalteredstainingandimprovesdigital analysis.Arecentstudy
underlinedthefactthatapercentageofsamplesscoredas
posi-tive(3+)bythepathologist,usingopticalmicroscopy,werefound
nottobeamplifiedafterFISHtest[15].Thisshowsthatthereare
fewsampleswithhighmembranestainingthatdonothavegene
amplification.Allthesefactorleadtodiscrepancieswhenabiopsy
isanalysed,evenwithautomateddigitalsoftware.
5.5. Comparisonbetweenstainingdistribution
Heterogeneous cases present different staining intensities
withinthesamesample.
Frequentlyitisunclearifvariationinintensityisduetofocal
hyperexpressionorisrelated toartefacts (id:delaysinfixation,
overfixation,automatic stainermalfunction,ortootherreasons
relatedtopreanalyticalandanalyticalphases[10,16]).Inall
het-erogeneouscasesweselectedtwoareaswithdifferentialstaining
containingcomparablecellnumbers.ConsideringthefiveFISH
dis-cordantcases,fourofthemaresurgicalspecimensthatpresenta
heterogeneousstainingdistribution.Heterogeneouscasesshould
beinterpretedwithcaution.
5.6. Comparisonamongtheyears
Duringthethreeyearsofthisstudy,thequalityofthe
immuno-histochemicalassay changed, due toa strictstandardisation of
pre-analysisprocessesandincreasedcontrolofthe
immunohisto-chemicalmethods.Coldischemiaandformalinfixationtimewere
carefully standardized as suggested by internationalguidelines
[1,9],antibodyspecificitywasperiodicallytestedonpositiveand
negativecontroltissue, andtheanalytic activitywassubmitted
toexternalqualitycontrol(NordiQC).Particularlyusefulwasthe
standardizationofhaematoxylincounterstaining.Agood-quality
sampleis essential toperforma reliabledigital analysis.Lower
samplequalityrisksinaccuratedigitalresults.
5.7. FISH
Weobservedthatlimitingthetotalevaluatednucleito20could
leadtodoubtfulclassifications,butalsotoextendthecounttoan
excessivelyhighnucleinumberdidnotimprovethereliabilityof
results.Wesuggestthat40istheidealnucleinumberforareliable
count.
6. Conclusion
ThisanalysissoftwareofimmunohistochemistryHer2slidehas
significantvalueandutilityintheprognostic-predictiveevaluation
ofbreast cancer.Itgives resultsneededtopatienttherapyvery
rapidly,avoidingtheuseofFISHonathirdoftotalcases.Digital
analysisisalsosimpletoapply,highlysensitive,andreproducible
whenproperlyarranged.However,therisksofmisclassification,
although low, are not entirelyremovable, and we believe that
heterogeneous-stainingcasesshouldbeconfirmedbyFISHtesting.
Whiletheidealsamplefordigitalanalysisisasurgicalspecimen,
iftheonlyavailablesampleisabiopsy,itwillbeimportanttobe
certainofitscorrectprocessingandstaining.Ifdoubtsonits
han-dlingarepresent,FISHmayalsoberequired.IntheanalysisofFISH,
areliablycorrectevaluationmayrequireincreasingthenumberof
nucleiexaminedto40.Theapplicationofthisanalysissoftwarecan
leadtoanimprovedinterpretationofequivocalcases,decreasing
theneedforFISHevaluation,andincreasingdiagnosticcertainty
andreproducibility.
Conflictofinterest
Theauthorshavenoconflictofinteresttodeclare.
Acknowledgementsandfundinginformation
ThisworkwassupportedbygrantA.MenariniDiagnosticS.R.L.
TheEnglishinthisdocumenthasbeencheckedbyatleasttwo
professionaleditors, both nativespeakersof English.For a
cer-tificate,pleasesee:http://www.textcheck.com/certificate/185gVo.
TheauthorsthankAnaMariaGheorghe-Guttaforherprecious
con-tributionincheckingtheEnglishaccuracyofthefinalversionofthe
manuscript. References
[1]A.C.Wolff,M.E.Hammond,D.G.Hicks,etal.,Recommendationsforhuman epidermalgrowthfactorreceptor2testinginBreastcancer:AmericanSociety ofClinicalOncology/CollegeofAmericaPathologistsClinicalPractice GuidelineUpdate,J.Clin.Oncol.31(2013).
[2]C.B.Moelans,R.A.deWeger,E.VanderWall,P.J.vanDiest,Current technologiesforHer2testinginbreastcancer,Crit.Rev.Oncol.Hematol.80 (2011)380–392.
[3]I.AlHaddabi,A.Qureshi,A.Saparamadu,etal.,Inter-observeragreementin reportingHER2Neuproteinoverexpressionbyimmunohistochemistry, IndianJ.Pathol.Microbiol.57(2014)201–204.
[4]H.Masmoudi,S.M.Hewitt,N.Petrick,K.J.Myers,M.A.Gavrielides,Automated quantitativeassessmentofHER-2/neuimmunoistochemicalexpressionin breastcancer,IEEETrans.Med.Imaging28(6)(2009).
[5]H.R.Ali,M.Irwin,L.Morris,etal.,Astronomicalalgorihmsforautomated analysisoftissueproteinexpressioninbreastcancer,Br.J.Cancer108(2013) 602–612.
[6]T.Keay,C.M.Conway,N.O’Flaherty,etal.,Reproducibilityintheautomated quantitativeassessmentofHer2/neuforbreastcancer,J.Pathol.Inform.4 (2013)19.
[7]V.J.Tuominen,T.T.Tolonen,J.Isola,ImmunoMembrane:apubliclyavailable webapplicationfordigitalimageanalysisofHER2immunohistochemistry, Histopathology60(2012)758–767.
[8]G.Turashvili,S.Leung,D.Turbin,etal.,Inter-observerreproducibilityofHer2 immunohistochemicalassessmentandconcordancewithfluorescentinsitu hybridization(FISH):pathologistassessmentcomparedtoquantitativeimage analysis,BMCCancer9(2009)165.
[9]A.C.Wolff,M.E.Hammond,D.G.Hicks,etal.,AmericanSocietyofClinical Oncology/CollegeofAmericanPathologistsGuidelinerecommendationsfor humanepidermalgrowthfactorreceptor2testinginbreastcancer,Arch. Pathol.Lab.Med.25(2007).
[10]G.Bussolati,E.Leonardo,Technicalpitfallspotentiallyaffectingdiagnosesin immunohistochemistry,J.Clin.Pathol.(2008)1184–1192.
[11]E.M.J.VanderLogt,D.A.J.Kuperus,J.W.vanSetten,etal,Fullyautomated fluorescentinsituhybridization(FISH)staininganddigitalanalysisofHein breastcancer:avalidationstudy,PLoSOne10(4)(2015)e0123201,http://dx. doi.org/10.1371/journal.pone.0123201.
[13]C.S.Connor,O.W.Tawfik,A.J.Joyce,etal.,Acomparisonprognostictumor markersobtainedonimage-guidedbiopsiesandfinalsurgicalspecimens,Am. J.Surg.184(October(4))(2002)322–324.
[14]L.Taillade,F.Penault-Llorca,T.Boulet,etal.,Immunohistochemichal expressionofbiomarkers:acomparativestudybetweendiagnosticbronchial biopsiesandsurgicalspecimensofnon-small-celllungcancer,Ann.Oncol.18 (June(6))(2007)1043–1050.
[15]Z.Varga,A.Noske,C.Ramach,B.Padberg,H.Moch,AssessmentofHER2status inbreastcancer:overallpositivityrateandaccuracybyfluorescenceinsitu hybridizationandimmunohistochemistryinasingleinstitutionover12 years:aqualitycontrolstudy,BMCCancer13(615)(2013).