诊断试验(英文)

AnalysisandevaluationoftheevidenceofdiagnostictestClinicalTrailStudyCenterCaoSumeiDiagnostictestarenotjustaboutdiagnosisScreeningDeterminingseverityOptimallytherapyPrognosisMonitorExampleCarotidultrasoundcantellyoutheseverityofthepatient’scarotidstenosisCarotidultrasoundcantellyouthepatient’sprognosisforstrokeanddeathCarotidultrasoundcanpredictyourpatient’slikelyresponsivenesstotherapyBasicprinciplesofconductingdiagnosticstudiesApplythegoldstandardtodeterminewhetherornotthetargetconditionispresentGoldstandard:ThemostrecognizedstandardforcliniciantodiagnosethetargetconditionPathologicalmeasurementOperationfindingSpecialimagingdetectionLong-termfollow-upRecognizedstandardWhatifyourtestismoregoldthanthestandardMayleadtounderestimateofthediagnosticpoweroftheevaluatingOnestrategyfordealingwiththisproblemistouselong-termfollow-upasagoldstandardToWhomShouldtheGoldStandardBeApplied?toeveryoneselectiveperformingthegoldstandardonpatientsmayresultin“verificationbias”or“workupbias”RecruityourparticipantsRecruitthetarget-negativeandtarget-positiveparticipantsidentifiedbygoldstandardcharacteristicofthosetowhomyouwillwanttoapplythetestinclinicalpracticeIncludingabroadspectrumofthediseasedcase：frommildlytoseverelycontrol：abroadspectrumofcompetingconditionsAnalternativeapproachisthatrecruitingaconsecutivesampleofpatientsMeasurementproceduresSpecifyingtesttechniqueReproducibilityBlindingoftheindividualconductingorinterpretingthetesttothegoldstandardSelectstatisticalprocedureCalculatingsamplesize212(1)ZSenSenn222(1)ZSpeSpenExample:Assumingasensitivityof80%,specificityof60%ofultrasonographyfordiagnosisofcholecystolithiasis.Howmanysamplesareneeded?2122220.05,1.96(0.80,0.60,0.10(1.96)(0.80)(10.80)62(0.10)(1.96)(0.60)(10.60)93(0.10)ZSenSpenn双侧），设例例ResultevaluationindexExample:126patientsunderwentindependent,blindBNPmeasurementandechocardiographyfordiagnosisofLVD.sensitivity：a/(a+c)=35/40=0.88,or88%specificity：d/(b+d)=29/86=0.34,or34%positivepredictivevalue(PPV)：a/(a+b)=35/92=0.38,or38%negativepredictivevalue(NPV)：d/(c+d)=29/34=0.85,or85%prevalence:(a+c)/(a+b+c+d)=40/126=0.32,or32%Pre-testodds:pre-testprobability/(1-pre-testprobability)=32%/68%=0.47positivelikelihoodratio(LR+):Sen/1-Spe=88%/(100%-34%)=1.3MultilevellikelihoodratiosStabilityoftheindexStable:Sen,SpeRelativelystable:LR+,LR-Unstable:PPV,NPV,prevalence:Receiveroperatingcharacteristiccurves(ROC)Itillustratestheperformanceofadiagnostictestwhenyouselectdifferentcut-pointstodistinguish“normal”from“abnormal”ItdemonstratesthefactthatanyincreaseinsensitivitywillbeaccompaniedbyadecreaseinspecificityThecloserthecurvegetstotheupperleftcornerofthedisplay,themoretheoverallaccuracyofthetestThecloserthecurvecomestothe45-degreediagonaloftheROCspace,thelessaccuratethetestTheareaunderthecurveprovidesanoverallmeasureofatest’saccuracyFigAROCforBNPasadiagnostictestforLVDParalleltestAtestBtestResult+++++++-ReductionmissdiagnosisExcludesomediseaseWhenprevalenceislow,astheprimaryscreeningmethod(1)SenSenASenASenBSpeSpeASpeBSerialtestAtestBtestResult++++-+--Sen=SenA×SenBSpe=SpeA+(1-SpeA)SpeBMisdiagnosismaycausenuisanceeffectConfirmatorydiagnosisSerialtestwithenzymelabeledcompoundassayfordiagnosisofmyocardialinfarctionEnzymelabeledcompoundassaySenSpeCPK9657SGOT9174LDH8791910.76(1)10.57(10.57)0.740.8920.89(10.89)0.910.99SenSenASenBSenCSpeSpeASpeASpeBSpeSpeMultivariateanalysisSENSPEsinglevariableanalysismarkermethodsSEN(%)spe(%)cutoffAREAaELISA90.788.80.19150.926bELISA77.373.20.20350.802cELISA74.270.90.09050.762dELISA78.481.61.080.836eELISA90.784.40.3560.932fELISA84.581.60.7990.899multivariateanalysisusinglogisticregressionCombinedmarkersSEN（％）SPE（％）AREAaandb88.882.50.926aandc87.782.50.927aande91.690.070.974aandf95.590.070.967aandd87.285.6.0936bandc78.876.30.837bandd87.786.60.934bande83.882.50.900bandf82.781.40.863Candd87.785.60.946Cande88.385.60.926Candf81.679.40.854dande89.486.60.946dandf88.386.60.952eandf87.785.60.933PredictiontheprobabilityofadiseaseLogit(P)=-0.934+4.797xa+2.203xeAvoidingoverfittingOverfittingoccurswhenacomputermodelidentifiesa“chance”patternthatdiscriminatescancerpatientsfromnon-cancerpatients,perfectlyfittingthatdatasetbutnotreproducibleinotherdatasets.Onewaytoavoidingoverfittingistorandomlysplitthedataintoseparatetrainingandtestsamples.TheEBMstepsfordiagnostictestsLookingforthemostsuitablestudypapersaccordingtotheclinicalquestionBringforwardthequestioninclinicExample2:ifdetectionofserumforritincandiagnoseIrondeficiencyanemia?Searchthecomputerinformationusingtheappositekeyword“diagnoseIrondeficiencyanemia”and“diagnostictest”and“human”EvaluationofthescientificityofthepapersIfcomparedwiththegoldstandardindependentlyandblindlyExample2:IronstainwithmyeloidbiopsyIfdetectedwiththecontroltestforeveryquizzeeGoldstandardTotal（No.）Results+-Newdiagnostictest+351550sensitivity=46%Newdiagnostictest-40460500specificity=96.8%VerificationbiasGoldstandardTotal（