Treatment and intention-to-treat propensity score analysis to evaluate the impact of video-assisted thoracic surgery on 90-day mortality after anatomical resection for lung cancer

Treatment and intenon‐to‐treat propensity score analysis to evaluate the impact of video‐ assisted thoracic surgery on 90‐day mortality aer anatomical resecon for lung cancer Authors JoseLuisRecuero‐Díaz¹†,IñigoRoyo‐Crespo¹,DavidGómezde‐Antonio²,SergiCall³,Borja Aguinagalde⁴,MaríaTeresaGómez‐Hernández⁵,JorgeHernández‐Ferrández⁶,DavidSánchez‐ Lorente⁷,JulioSesma‐Romero⁸,EduardoRivo⁹,NicolásMoreno‐Mata¹⁰,andRaulEmbun¹,¹¹*† onbehalfoftheSpanishGroupofVideo‐assistedThoracicSurgery(GEVATS) †Theseauthorscontributedequallytothiswork. Affiliaons 1. DepartmentofThoracicSurgery,HospitalUniversitarioMiguelServet,HospitalClínico UniversitarioLozanoBlesa,IISAragón,Zaragoza,Spain 2. DepartmentofThoracicSurgery,HospitalUniversitarioPuertadeHierroMajadahonda, Madrid,Spain 3. DepartmentofThoracicSurgery,HospitalUniversitariMútuaTerrasa,Universitatde Barcelona,Terrasa,Barcelona,Spain 4. DepartmentofThoracicSurgery,HospitalUniversitariodeDonosa,SanSebasán‐ Donosa,Spain 5. DepartmentofThoracicSurgery,HospitalUniversitariodeSalamanca,IBSAL, UniversidaddeSalamanca,Salamanca,Spain 6. DepartmentofThoracicSurgery,HospitalUniversitarioSagratCor,Barcelona,Spain 7. DepartmentofThoracicSurgery,HospitalClínicdeBarcelona,InstutoRespiratorio, UniversityofBarcelona,Barcelona,Spain 8. DepartmentofThoracicSurgery,HospitalGeneralUniversitarioAlicante,Alicante, Spain 9. DepartmentofThoracicSurgery,HospitalUniversitarioSanagodeCompostela, SanagodeCompostela,Spain 10. DepartmentofThoracicSurgery,HospitalRamónyCajal,Madrid,Spain 11. DepartmentofSurgery,FacultyofMedicine,UniversityofZaragoza,Zaragoza,Spain *Corresponding author This is a pre-copyedited, author-produced version of an article accepted for publication in European Journal of Cardiothoracic Surgery following peer review. The version of recordJose Luis Recuero-Díaz, Iñigo Royo-Crespo, David Gómez de-Antonio, Sergi Call, Borja Aguinagalde, María Teresa Gómez-Hernández, Jorge Hernández-Ferrández, David Sánchez-Lorente, Julio Sesma-Romero, Eduardo Rivo, Nicolás Moreno-Mata, Raul Embun, on behalf of the Spanish Group of Video-assisted Thoracic Surgery (GEVATS), Treatment and intention-to-treat propensity score analysis to evaluate the impact of video-assisted thoracic surgery on 90-day mortality after anatomical resection for lung cancer, European Journal of Cardio-Thoracic Surgery, Volume 62, Issue 3, September 2022, ezac122 is available online at: https://doi.org/10.1093/ejcts/ezac122. RaulEmbun DepartmentofThoracicSurgery HospitalUniversitarioMiguelServet PaseoIsabellaCatólica1,Zaragoza50009,Spain Tel:+34‐635492179 Email:raulem[email protected]  Abstract OBJECTIVES:Theaimofthisstudywastoknowthetreatmenteffectofvideo‐assistedthoracic surgery(VATS)on90‐daymortalityaeranatomicallungreseconbasedonanaonwide cohort. METHODS:Thisisamulcentreprospecvecohortof2721anatomicalreseconsforlung cancerfromDecember2016toMarch2018.Treatmentandintenon‐to‐treat(ITT)analyses wereperformedaerinverseprobabilityscoreweighnganddifferentpropensityscore matchingalgorithms.Covariatebalancewasassessedbystandardizedmeandifferences.The esmatorsreportedweretheaveragetreatmenteffect,theaveragetreatmenteffectonthe treatedandoddsraosaercondionallogiscmodelswith95%confidenceintervals.The unconfoundednessassumponwasevaluatedbysensivityanalysisforaveragetreatment effect(c‐dependence)andaveragetreatmenteffectonthetreated(Γ). RESULTS:VATSwastheinialapproachin1911paents(70.2%),though273cases(14.3%)had tobeconvertedtothoracotomy.Ninety‐daymortalityrateswere:treatmentanalysis(VATS 1.16%vsopen3.9%,P<0.001),ITTanalysis(VATS1.78%vsopen3.36%,P=0.012).Aer inverseprobabilityscoreweighngandpropensityscorematching,inthetreatmentanalysis, VATSmeantabsoluteriskreduconsbetween2.25%and2.96%andrelaveriskreducons between65%and70%(OR=0.34,95%confidenceinterval0.15–0.79,allP‐values<0.004). However,alltheesmatorsturnedouttobenon‐significantintheITTanalyses.Ahigh sensivitytounobservableconfounderswasproved(c‐dependence0.135,Γ=1.5). CONCLUSIONS:VATScanreducetheriskof90‐daymortalityaeranatomicallungresecon. However,theimplicaonsofconversiontothoracotomy,comparingITTversustreatment analysis,andthepotenalimpactofhiddenbiasshoulddeservefurtheraenoninthefuture. Keywords:Video‐assistedthoracicsurgery,90‐Daymortality,Intenon‐to‐treatanalysis, Thoracicsurgery,Anatomicallungresecon,Lungcancer  Abbreviaons ATE‐Averagetreatmenteffect ATT‐Averagetreatmenteffectonthetreated CIs‐Confidenceintervals GEVATS‐SpanishGroupofVideo‐assistedThoracicSurgery ITT‐Intenon‐to‐treat IPSW‐Inverseprobabilityscoreweighng PSM‐Propensityscorematching SMD‐Standardizedmeandifferences VATS‐Video‐assistedthoracicsurgery  Introducon Althoughtherearemulpleretrospecveseriesthathaveshownalowerrateofcomplicaons andpostoperavestayinpaentsoperatedonbyvideo‐assistedthoracicsurgery(VATS),there aremorediscrepanciesontheimpactthatthesurgicalapproachcouldhaveonpostoperave mortality.Inthisregard,arecentrandomizedclinicaltrial,designedtoevaluateshort‐termand oncologicefficacyofVATS,failedtodemonstratedifferencesinpostoperavemortality betweenVATSandthoracotomy[1]. Regardingpostoperavemortality,recentserieshaveshownthat90‐daymortalitycould double30‐dayorin‐hospitalmortalityaerlungresecon[2–4].Althoughmostseries comparingVATSwiththoracotomydonotmenon90‐daymortality,someimportantstudies havenotshownsignificantdifferencesineitherin‐hospital,30‐dayor90‐daymortality[5–7].A recentpublicaonfromtheESTSdatabasedidshowsignificantdifferencesinmortalityat discharge(VATS1%vsthoracotomy1.9%,P=0.020),notmenoningmortalityat90days[8]. Fewstudieshaveperformedanintenon‐to‐treat(ITT)analysiscomparingVATSand thoracotomyforlungresecon,andevenlesshavecomparedtheresultsobtaineddepending onthestrategyofanalysisperformed[9,10].However,sinceanon‐negligibleproporonof paentsundergoingVATSmustbefinallyconvertedtothoracotomy,ITTanalysisseemstobe themostappropriatestrategytoevaluatetreatmenteffectsrelatedtoVATSinarealscenario [11]. InviewoftheconflicngresultsandtheunderreportedITTanalysisintheliterature,the objecveofourstudywastodeterminetheimpactofthesurgicalapproachon90‐day mortality,comparingtreatmentandITTanalysisinpaentswhounderwentandanatomical lungreseconforlungcancerinthenaonalcohortbytheSpanishGroupofVideo‐assisted ThoracicSurgery(GEVATS)[12].  Materials and Methods Ethical statement Thisprojectwasapprovedbyallthelocalethicscommieesandinformedconsentwas obtainedfromtherecruitedpaentstousetheirclinicaldataforscienficpurposes(Approval byEthicsCommieeofAragonHealthResearchInstuteon20May2015PI15/0072). Data source TheGEVATSoftheSpanishSocietyofThoracicSurgeryisaprospecvevoluntarymulcentre observaonalstudywithatotalof33ThoracicSurgeryDepartmentparcipang.TheCentres werenotselectedbasedontheirexperienceinVATSoranyothercriterion.Thecohort includedpaentswhounderwentananatomicallungreseconfrom20December2016to20 March2018.TheGEVATSobjecvesweretoknowtheimpactofsurgicalapproachonshort‐ andlong‐termoutcomes.ThemethodoftheGEVATS,includingsamplesizejusficaonand theauditprocessperformed,wasrecentlypublished[12]. PaentallocaonintoVATSorthoracotomydependedonclinicalpracce,experienceand beliefsofeachoftheparcipangsurgeonsanddepartments. Inthisprospecveobservaonalcohortstudy,wetriedtospecificallyelucidatetheimpactof thesurgicalapproachon90‐daymortalityaeranatomicallungreseconforlungcancer.For thispurpose,thosepaentswithadiagnosisotherthanlungcarcinomaandthosewho underwentpneumonectomyorextendedlungreseconwereexcluded.Extendedresecon wasconsideredincaseofchestwall,diaphragmorsleeveresecon.Ourmanuscriptis reportedaccordingtotheSTROBErecommendaonsandESTSStascalPrimerforpropensity scoreanalysis. Stascal analysis Adoubletypeofanalysiswascarriedout,treatmentandITT,dependingontheneedof conversionfromVATStoopenthroughouttheprocedure.Theassociaonofthesurgical approach(openversusVATS)asthetreatmentvariable,withbaselineoncologicandsurgical variablesthatcouldinfluenceontheoutcomevariable(90‐daymortality)andtheapproachto bechosen,wasanalysedbytwo‐tailedstascalhypothesistesng,usingMann–Whitneyand Chi‐squaretestsandstandardizedmeandifferences(SMD).ThosevariableswithaP‐valueof lessthan0.2and/orstandardizeddifferencesgreaterthan0.1werethecovariatesusedto buildthepropensityscoretocorrectforseleconbias. Missingdataweredealtbycasewisedeleonanalysiswhenlessthan5%ofpaentshad incompleteregistries. Thepropensityscorewasesmatedbyalogitmodelandtheoverlapassumponwasassessed ondensityplotsfortreatmentandITTanalysis.Thetreatmenteffectswereevaluatedbasedon theinverseprobabilityscoreweighng(IPSW)andthepropensityscorematching(PSM) throughthenearest‐neighbormethodwithandwithoutreplacement,usingdifferentcalliper widths(0.035,0.05and0.1)andmatchingraos(1:1,1:2and1:3). CovariatebalancedwasassessedbySMD,beforeandaerweighngormatching.SMDless than0.1or0.05wereconsideredgoodorexcellent,respecvely,toexcluderesidualimbalance [14].BalanceofcovariateswasdisplayedondotplotsforIPSWandPSM,separately. Thetreatmenteffectswereesmatedbyweightedmeanandmatchingoutcomemodels reporngtheaveragetreatmenteffect(ATE),basedonthedifferenceinpotenaloutcome means,andtheaveragetreatmenteffectonthetreated(ATT).ATEandATTwerereportedas absoluteriskreducons(IPSWandPSM)andrelaveriskreducons(IPSW).Inaddion,in caseofPSM1:1withoutreplacementandcalliper0.035,acondionallogiscfixed‐effects regressionmodelwasconducted.Inthiscase,treatmenteffectswerereportedasoddsraos. TheimpactofsurgeonexperienceinVATSprocedures(≤50versus>50cases),surgeonseniority (residentversusfaculty<10yearsversusfaculty10–20yearsversusfaculty>20years),surgical volumeandVATSratebydepartment(discretevariables)wereusedtoadjusttheoddsraos reportedbythecondionallogiscmodelspreviouslydescribed.Tocomputesurgicalvolume andVATSratebyinstuonthroughoutthe15‐monthrecruitmentperiod,weconsideredthe reportssubmiedbytheHeadsoftheAdministraveDepartmentsfromeachInstuon. Thesereportswereusedintheauditprocesswepreviouslypublished[12].95%confidence intervals(CIs)werecalculatedfromrobuststandarderrorsandP‐valueslessthan0.05were consideredstascallysignificant. ThecondionalparaldependencemethodproposedbyMastenetal.wasusedtoevaluate thesensivityofconclusionsabouttheATE.BoundsontheATEgivenasetofc‐dependence values(between0and1)andthebreakdownpoint(maximumvalueofthec‐dependence parameterunderwhichtheconclusionsllholds)werereported.Inaddion,theimpactof hiddenbiasontheATTesmatoraerPSM(1:1,calliper0.035,noreplacement)wasassessed withtheboundingapproachproposedbyRosenbaum.TheΓparameterandcorrespondingP‐ valueswereusedtomeasurethesensivityofATTtounobservableconfounders. TheTreatmentEffectsSuiteinStata/MP16.0andtheStatapackagesStddiff,Psmatch2, Calipmatch,TesensivityandMhboundswereusedforthestascalanalysis.TablauDesktop 2020.3.1wasusedforplotrepresentaonofthecovariatebalance.  Results Atotalof3533paentswererecruited,including1917VATScases(54.3%).Aerexclusionof paentswithadiagnosisdifferenttolungcancer(448paents,12.7%),pneumonectomy(236 paents,6.7%)andextendedresecon(165,4.7%),2721paents(77%oftheenreGEVATS cohort)mettheinclusioncriteria.Thetypesofreseconsincludedwere:2444lobectomies (90%),111bilobectomies(4%)and166anatomicalsegmentectomies(6%). VATSwastheinialapproachin1911paents(70.2%),represenngthetreatmentgroupin theITTanalysis.However,273cases(14.3%)hadtobeconvertedtothoracotomyand, therefore,theVATSarminthetreatmentanalysisconsistedof1638paents(60.2%).The unadjustedanalysisshowedanimportantassociaonbetweenin‐hospitalmortalityand90‐day mortalitywiththesurgicalapproachperformed(Table1).Theoutcomevariable(90‐day mortality)wasmissingin11cases(0.4%)thatwerenotconsideredintheanalysisoftreatment effects.Thepercentagesofmissingvalueswerenegligibleforalltheconfoundersexceptfor DLCO.ThemainanalysisexcludedDLCOasconfounder,so14variableswereusedtobuildthe propensityscoreinacomplete‐caseanalysissinceonly1.8%ofthepaentshadmissingvalues insomeofthecovariatesincluded(Table2).Thepropensityscoredistribuontoprovethe overlapassumponisshowninadensityplot(Figure1).Only9paentsintheVATSgroup (0.6%)hadapropensityscorehigherthanthemaximumpropensityscorevalueintheopen group,whileonly2paentsintheopengroup(0.2%)hadapropensityscorelowerthanthe minimumvalueintheVATSgroup. AerIPSW,theraosofVATStothoracotomywere1339:1324paentsinthetreatment analysisand1347:1317paentsintheITTanalysis.Covariatebalancewasproved(Figure2). PSM1:1withoutreplacementandacalliperwidthof0.035yielded872matchedpairsinthe treatmentanalysisand705matchedpairsintheITTanalysis.Mostofthecovariatesreached anexcellentbalanceaermatching(Table3).Theproporonsofcasesmatchedto1control were54.2%(treatmentanalysis)and37.6%(ITTanalysis). AerPSMwithreplacement,2664paents(1608VATSand1056open)inthetreatment analysisand2663paents(1877VATSand786open)intheITTanalysiswerematchedto1,2 or3counterparts.Independentlyofthematchingraosandcalliperwidths,mostofthe covariatesreachedanexcellentbalance(Figure3). Basedonthetreatmentanalysis,VATSwasconsistentlyassociatedtoalower90‐daymortality rateaccordingtotheATEandATTesmators,whateverthepropensityscoretechniqueor algorithmused.However,inthecaseofITTanalysis,theslllowermortalityaerVATSwas associatedtonon‐significantATEandATT,intermsofabsoluteandrelaveriskreducons,in allthecases(Table4). InthecondionallogiscmodelaerPSM1:1(noreplacementandcalliperwidth0.035),the resultsobtainedaertreatmentanalysis(OR=0.38;95%CI0.20–0.73;P=0.004)andITT analysis(OR=0.71;95%CI0.38–1.33;P=0.283)wereconsistentwiththeATEandATT esmators.Aeradjusngthefixed‐effectmodelsbysurgeonexperience(numberofVATS proceduresandseniority)anddepartmentexperience(surgicalvolumeandVATSrate)theodds raosweresimilartothoseinthenon‐adjustedmatchedtreatmentanalysis(OR=0.34;95%CI 0.15–0.79;P=0.012)andlower,thoughsllnon‐significant,intheITTanalysis(OR=0.41;95% CI0.13–1.25;P=0.117). Inthesensivityanalysis,theclosetozeroc‐dependencevaluesunderwhichtheconclusions sllheld(treatmentanalysis0.135andITTanalysis0.09)meantahighsensivityoftheATE esmatortohiddenbias(Figure4).Inaddion,themaximumΓvalueatwhichsignificantP‐ valueswereobtained(Γ=1.5)showedthattheATTesmaonofthisstudyissensivetobias (unobservedvariables)abletoincreasetheoddsofreceivingVATSinmorethan50%. Inthemainanalysis,wedidnotconsiderDLCObecauseofthehighrateofmissingvalues (15.2%).However,inasecondaryanalysisincludingDLCOasacovariatetobuildthepropensity score,theATEandATTesmatorsaerIPSWwereequivalenttothoseobtainedinthemain analysis.Moreover,thesensivityanalysisshowedanevenhigherpotenalimpactof unmeasuredorunobservableconfounders(Γ=1.15).  Discussion OurmainfindingwasthatVATScanreduce90‐daymortalityaerananatomicallungresecon forlungcancer.However,thisbeneficialeffectdecreasedandturnedouttobenon‐significant whenanITTanalysiswasperformed.Noteworthy,thesefindingswereconsistentalongallthe propensityscorealgorithmscarriedout.Finally,thehighsensivityofouranalysistohidden biashighlightsthatthepotenalimpactofunmeasuredandunobservableconfoundersin observaonalstudiescomparingVATSandopensurgerycouldbedeterminant. Theresultsofthefewstudiescomparingtheimpactofthesurgicalapproachon90‐day mortalityleadtocontradictoryconclusions.Limitedstascalpowersecondarytosmall samplesizesandsurprisinglylowmortalityratescouldpartlyexplainnon‐conclusivefindings [1,6,7].However,whenanalysingtheresultsobtainedbylargeinstuonalcohorts,significant stascaldifferencescouldonlycorrespondtoclinicallytrivialtreatmenteffect[15]. Inthethoracicsurgeryliterature,usefulrecommendaonshavebeenmadetoimprove reporngbasedonthepropensityscoreanalysis,sincethisanalycapproachisnotas straighorwardasregressionanalysis[16].ConsideringtherecommendaonsbytheEuropean SocietyofThoracicSurgeonsinitsstascalprimerreport,wecarriedoutthe2advised methods,IPSWandPSMwithcalliper[13].Inaddion,toassesstheconsistencyofourresults, weputintopraccedifferentmatchingalgorithms(treatedtocontrolraos,calliperwidths andreplacementversusnoreplacementmatching). Theuseof2propensityscorestechniquescomparingoutcomesaerVATSoropenlung reseconisanexcepon[10].Pagèsetal.publishedtheresultsbyaFrenchnaonwidestudy evaluangtheimpactofsurgicalapproachonshort‐andlong‐termoutcomesaerlobectomy forlungcancer.Inthislargecohort(n=24,811)recruitedthroughoutan8‐yearperiod,only 4.9%ofthecaseswereoperatedonbyVATS.Itisnoteworthythatthesignificancelevel obtainedinafewoutcomesdifferedbetweenthepropensityscorestrategiesused(IPSW versusPSM).However,regarding30‐daymortality,bothIPSWandPSMfailedtodemonstratea significantreduconaerVATS(PSMOR=0.89,95%CI0.45–1.81;IPSWOR=0.74,95%CI 0.37–1.45). TheimportanceofanITTstrategywhenevaluangtheimpactofthesurgicalapproachwas highlightedinoneofthefirstmeta‐analysespublishedcomparingVATSandthoracotomy[17]. However,theinformaonaboutconversionfromVATStoopensurgeryisnotalwaysregistered, evenincaseofnaonalandinternaonalregistries[6,8,10,18,19].Consequently,ITTanalysis isnotawidepracceinthethoracicsurgeryliteratureand,therefore,thebenefitconferredto VATSwhenevaluangtreatmenteffectscouldbemisleading.Infact,thehighertherateof conversionfromVATStoopenthehigherthepotenaldiscrepancybetweentreatmentandITT analyses.Inthisregard,despiteahigherrateofVATSinourcohort(60.2%)thantheone reportedbymanynaonwideandmulcentreregistries,theproporonofconversionweare reporng(14.3%)isnotablysuperiorcomparedtosomerecognizedseriesrangingfrom2%to 9%[11,20,21].Riskofconversioninarecentmetanalysiswas9.6%(95%CI:6.6–13.9%)[22]. Wecannotconcludeaboutthereasonsforourhigherconversionrate.However,naonwide representaveness,dataauditthatinvolvedalltheparcipangcentres,orjustamaerof disposionofsurgeons,morepronetostarngtheprocedurebyVATSinourcountry,couldbe someofthehypotheses.Inaddion,althoughanyonecouldexpectthathigherVATSratesare paralleltolowerconversionratesbecauseofexperienceacquision,thisreasoningcouldbe conflicngdependingonthescenarioandthetypeofcohortinqueson,forexamplesingle centreversusmulcentre.Inthisregard,ourrelavelyhighconversionratecouldsimply representaconsequenceofourdirectlyproporonalhighVATSrate. AlthoughsomestudiesarguethatconversionfromVATStoopendoesnotentailasurgical failure,otherserieshaveassociatedconversiontoworseoutcomesintermsofpostoperave morbidity[20,23].Ourstudywouldsupportadetrimentaleffectassociatedtoconversion. However,moredetailedandspecificanalysescomparingconvertedVATStostraight thoracotomycouldgenerateinsighulknowledgeabouttheimplicaonsofconversionitself andthemostappropriatedisposiontowardsstarngtheprocedurebyVATS. ITTanalysisisclaimedasthegoldstandardinrandomizedcontroltrialstoreflectapragmac clinicalscenario,tomaintainprognoscbalancegeneratedfromtheoriginalrandomizaonand togiveanunbiasedesmateoftreatmenteffect.However,weconsideredimportantatthis stagewhenmostofthestudiesdealingwithVATSaresllreporngtheirresultsbasedona treatmentanalysis,toincludethislessconservavestrategyofanalysissothatwecouldsll compareourfindingswiththecoetaneousliterature.Toourknowledge,thisisthefirst manuscriptthatevaluatetheimpactofthesurgicalapproachonshort‐termmortalityaer lungresecon,comparingatreatmentandITTanalysis. Limitaons Themaindrawbackinmulcentrevoluntaryregistriesisrelatedtoseleconbiasanddata quality.Althoughthedetailsofourauditwerepreviouslyreported,wecannotrejectresidual biasatthisstage[12]. Aspreviouslypublished,tocalculateoursamplesize,weconsidered2%absoluterisk differencein90‐daymortalityasrelevantinpracce(4%vs2%),andanexpected25%ofcases byVATS[12].Becauseoftheunderesmated90‐daymortalityratedifferencebetweenVATS andOpen(1.16%vs3.9%)andthehigher‐than‐expectedVATSrate(60%),thestascalpower ofour'posive'conclusioninthetreatmentanalysis(99%)ishigherthantheconvenonal 80%.However,intheITTanalysis,thelowerdifferenceinmortalitybetweenbothapproaches (1.78%vs3.36%)andthelowerproporonofcontrolpaents(29%)couldbethereasonsfora limitedpower(69%)ofour'negave'results. Missingvaluesareanothercornerstoneinobservaonalstudies.AlthoughDLCOwastheonly covariatewithamissingrateofhigherthan2%(415/2721paents15%),thisrespiratory parameterhasbeenprovedasoneofthemostdeterminantriskfactorsaerlungresecon, eveninourowncohort[24–26].Despiteequivalentresultsinaposthocanalysisbasedonthe subgroupofpaentswiththisvaluepresent,confoundingbiascannotbeexcluded. DespitethenaonalrepresentavenessoftheGEVATScohort,sinceitwouldhaveincluded 50%oftheanatomicallungreseconsperformedinSpainoverthe15‐monthrecruitment period,thehighauditedrateofVATSregistered(60%)couldcompromisetheexportaonof ourresultstoothernaonalorlocalscenarioswithanimportantdiscrepancyinVATS implementaon[12].However,theequipoiseinourstudyintermsoftheproporonof paentsineachgroupcomparedtootherseriesmakesitlesslikelythateitherouropengroup representsastronglyselectedcolleconofchallengingcasesorourVATSgroupastrongly selectedcolleconoffavourablecases[6,10,27]. PSMisconsideredtojeopardizethegeneralizaonofitsconclusionswhenmatchedsample sizeislowerthan50%oftheoriginalsample.InourstudybycarryingoutPSM1:1without replacementandcalliperwidth0.035,theproporonofmatchedsamplewas64%and52% aertreatmentandITTanalysis,respecvely.Nevertheless,theresultsobtainedaersuch PSMalgorithmswereequivalenttothosereachedaermakinguseofallthesamplewith otherpropensityscoretechniques,whichseemstopreserveourconclusionatamoregeneral seng. Finally,hiddenbiasisaubiquitousproblemineveryobservaonalstudy,evenincaseofa rigorouspropensityscoreanalysis,andsensivitytesttriestogetabeerunderstandingofthis obstacletoobtainreliableesmators.However,thistypeofanalysisisexceponalinthoracic surgery[10].ThestudypublishedbyPagèsetal.fromtheEpithordatabasereportedahigh sensivitytounobservableconfounderswhenesmangpostoperavedeath(Γ=1.6).Our resultswereverysimilar(Γ=1.5),evenaerincludingDLCOasacovariateinthepropensity scoreanalysis(Γ=1.2,thecloserto1themoresensiveouresmaonstohiddenbias). Consequently,thereliabilityofouresmaonsshouldbecauouslyadmiedunlmore robustevidence.  Conclusion Havinginmindtheexclusioncriteriapreviouslyreferred(pneumonectomyandextended resecon),VATSwasconsistentlyshowntoreduce90‐daymortalityby65–70%aeran anatomicallungreseconforlungcancerintheGEVATScohort.However,drawnontheITT analysisincludingpaentsconvertedtoopensurgeryintotheVATSgroup,thisbenefit decreasedtotheextentofobtainingstascallynon‐significant,thoughpotenally underpowered,differences.UnlupcomingevidencecomparingconvertedVATStostraight thoracotomy,aneffortshouldbemadetoopmizethebeneficialeffectofVATSonshort‐term mortalityaerananatomicalreseconforlungcancer.Intheanalysisofhiddenbias,we demonstratedthatthetreatmenteffectesmatorswerehighlysensivetounobservable confounders,whichcouldonlybeovercomeinasufficientlylargemulcentrerandomizedtrial.  Acknowledgements Wethankallthelocalresearcherswhocontributedwithpaentrecruitmentanddataentry. Coordinator:RaulEmbun(UniversityHospitalMiguelServet). Scienfic commiee:DavidGómezdeAntonio(UniversityHospitalPuertadeHierro Majadahonda,Madrid);SergiCall(MútuaTerrasaUniversityHospital,UniversityofBarcelona, Terrasa,Barcelona);NicolásMoreno‐Mata(RamónyCajalUniversityHospital,Madrid); MarceloF.Jiménez(SalamancaUniversityHospital,UniversityofSalamanca,IBSAL, Salamanca);MiguelCongregado(VirgenMacarenaUniversityHospital,Seville);andSergio Bolufer‐Nadal(GeneralUniversityHospitalofAlicante). Local researchers:JoseLuisRecueroandIñigoRoyoCrespo(UniversityHospitalMiguelServet andLozanoBlesa);BorjaAguinagaldeandIkerLópezSanz(DonosaUniversityHospital,San Sebasán‐Donosa);SergioAmor‐AlonsoandFranciscoJavierMoradiellos‐Díez(Quironsalud MadridUniversityHospital,Madrid);MiguelJesúsArrarás(ValencianOncologyInstute Foundaon,Valencia);AnaIsabelBlancoOrozco(UniversityHospitalVirgendelRocío,Sevilla); MarcBoadaandDavidSánchez(HospitalClinicofBarcelona,RespiratoryInstute,Universityof Barcelona,Barcelona);AlbertoCabañeroSánchez(RamónyCajalUniversityHospital,Madrid); IsabelCalVázquezandRamónMoreno‐Balsalobre(UniversityHospitalLaPrincesa,Madrid); ÁngelCillerueloRamos(UniversityClinicalHospital,Valladolid);SilvanaCrowleyCarrasco (UniversityHospitalPuertadeHierroMajadahonda,Madrid);ElenaFernández‐Marnand FlorennoHernandoTrancho(HospitalClínicoSanCarlos,Madrid);SanagoGarcía‐Barajasand CiprianoLópezGarcía(UniversityHospitalofBadajoz,Badajoz);MariaDoloresGarcía‐Jiménez (UniversityHospitalofAlbacete,Albacete);JoseMaríaGarcía‐PrimandEduardoRivo(Sanago deCompostelaUniversityHospital,SanagodeCompostela);JoseAlbertoGarcia‐Salcedo(12 deOctubreUniversityHospital,Madrid);JuanJoséGelbenzu‐ZazpeandMaríaElenaRamírez‐ Gil(ComplejoHospitalarioofNavarre,Pamplona);CarlosFernandoGiraldo‐OspinaandRoberto MongilPoce(HospitalRegionalUniversitario,Málaga);MaríaTeresaGómezHernández (SalamancaUniversityHospital,UniversityofSalamanca,IBSAL,Salamanca);JorgeHenández andJuanJoséFiblaAlfara(SagratCorUniversityHospital,Barcelona);JenniferD.IllanaWolf (PuertadelMarHospital,Cádiz);AlbertoJaureguiAbularach(Valld'HebronUniversityHospital, Barcelona);UnaiJiménezandRafaelRojo‐Marcos(UniversityHospitalCruces,Bilbao);NéstorJ. Marnez‐Hernández(LaRiberaUniversityHospital,Alcira,Valencia);ElisabethMarnez‐Téllez andJuanCarlosTrujilloReyes(SantaCreuandSantPauHospital,AutonomousUniversityof Barcelona,Barcelona);LucíaMillaCollado(HospitalArnaudeVilanova,Lleida);SergioB. MorenoMerino(VirgenMacarenaUniversityHospital,Seville);CarmeObiols(MútuaTerrasa UniversityHospital,UniversityofBarcelona,Terrasa,Barcelona);FlorencioQuero‐Valenzuela (VirgendelasNievesHospital,Granada);RicardRamos‐Izquierdo(UniversityHospitalof Bellvitge,HospitaletdeLlobregat,Barcelona);AlbertoRodríguez‐Fuster(HospitaldelMar, HospitaldelMarMedicalResearchInstute,Barcelona);LauraSanchezMoreno(University 11. DecaluweH,PetersenRH,HansenH,PiwkowskiC,AugusnF,BrunelliAetal.;ESTS MinimallyInvasiveThoracicSurgeryInterestGroup(MITIG).Majorintraoperave complicaonsduringvideo‐assistedthoracoscopicanatomicallungresecons:an intenon‐to‐treatanalysis.EurJCardiothoracSurg2015;48:588–99. 12. EmbunR,Royo‐CrespoI,RecueroDíazJL,BoluferS,CallS,CongregadoMetal.Spanish Video‐AssistedThoracicSurgeryGroup:method,auding,andinialresultsfroma naonalprospecvecohortofpaentsreceivinganatomicallungresecons.Arch Bronconeumol2020;56:718–24. 13. BenedeoU,HeadSJ,AngeliniGD,BlackstoneEH.Stascalprimer:propensityscore matchinganditsalternaves.EurJCardiothoracSurg2018;53:1112–7. 14. AusnPC.Balancediagnoscsforcomparingthedistribuonofbaselinecovariates betweentreatmentgroupsinpropensity‐scorematchedsamples.StatMed2009 10;28:3083–107. 15. AbdelsaarZM,AllenMS,ShenKR,CassiviSD,NicholsFC,WigleDAetal.Variaonin hospitaladoponratesofvideo‐assistedthoracoscopiclobectomyforlungcancerand theeffectonoutcomes.AnnThoracSurg2017;103:454–60. 16. AusnPC.Propensity‐scorematchinginthecardiovascularsurgeryliteraturefrom 2004to2006:asystemacreviewandsuggesonsforimprovement.JThorac CardiovascSurg2007;134:1128–35. 17. YanTD,BlackD,BannonPG,McCaughanBC.Systemacreviewandmeta‐analysisof randomizedandnonrandomizedtrialsonsafetyandefficacyofvideo‐assistedthoracic surgerylobectomyforearly‐stagenon‐small‐celllungcancer.JClinOncol 2009;27:2553–62. 18. PuriV,GaissertHA,WormuthDW,GroganEL,BurfeindWR,ChangACetal.Thoracic surgeonsparcipantsperformingthoracoscopiclobectomy.AnnThoracSurg 2019;107:202–8. 19. PaulS,AltorkiNK,ShengS,LeePC,HarpoleDH,OnaisMWetal.Thoracoscopic lobectomyisassociatedwithlowermorbiditythanopenlobectomy:apropensity‐ matchedanalysisfromtheSTSdatabase.JThoracCardiovascSurg2010;139:366–78. 20. BongiolaS,GonfioA,ViggianoD,BorgianniS,PoliL,CrisciRetal.;ItalianVATS Group.RiskfactorsandimpactofconversionfromVATStoopenlobectomy:analysis fromanaonaldatabase.SurgEndosc20191;33:3953–62. 21. LaursenLØ,PetersenRH,HansenHJ,JensenTK,RavnJ,KongeL.Video‐assisted thoracoscopicsurgerylobectomyforlungcancerisassociatedwithalower30‐day morbiditycomparedwithlobectomybythoracotomy.EurJCardiothoracSurg 2016;49:870–5. 22. PowerAD,MerriRE,Abdel‐RasoulM,Moffa‐BruceSD,D'SouzaDM,KneuertzPJ. Esmangtheriskofconversionfromvideo‐assistedthoracoscopiclungsurgeryto thoracotomy—asystemacreviewandmeta‐analysis.JThoracDis2021;13:812–23. 23. FourdrainA,DeDominicisF,IquilleJ,LafieS,MerluscaG,Wie‐PfisterAetal. Intraoperaveconversionduringvideo‐assistedthoracoscopydoesnotconstutea treatmentfailure.EurJCardiothoracSurg2019;55:660–5. 24. BrunelliA,DineshP,Woodcock‐ShawJ,LilechildD,PompiliC.Ninety‐daymortality aervideo‐assistedthoracoscopiclobectomy:incidenceandriskfactors.AnnThorac Surg2017;104:1020–6. 25. GómezdeAntonioD,CrowleyCarrascoS,RomeroRománA,RoyuelaA,SánchezCalle Á,ObiolsFornellCetal.Surgicalriskfollowinganatomiclungreseconinthoracic surgery:aprediconmodelderivedfromaSpanishMulcenterDatabase.Arch Bronconeumol2021Feb24:S0300‐2896(21)00070‐3. 26. AguinagaldeB,InsausA,LopezI,SanchezL,BoluferS,EmbunR.VATSlobectomy morbidityandmortalityislowerinpaentswiththesameppoDLCO:analysisofthe databaseoftheSpanishVideo‐AssistedThoracicSurgeryGroup.ArchBronconeumol 2021;57:750–6. 27. XuJ,NiH,WuY,CaoJ,HanX,LiuLetal.Perioperavecomparisonofvideo‐assisted thoracicsurgeryandopenlobectomyforpT1‐stagenon‐smallcelllungcancerpaents inChina:amul‐centerpropensityscore‐matchedanalysis.TranslLungCancerRes 2021;10:402–14. Note:Thisistheauthor'sacceptedmanuscriptversion.Thefinalpublishedarclemaycontain addionaleditsandformangchangesmadeduringtheproduconprocess.Pleasecitethe publishedversionwhenreferencingthiswork. Figure_1 Figure_2 Figure_3 Figure_4