东南大学流行病学.ppt

东南大学流行病学2OverviewDesignofcase-controlstudiesBasicconceptOR=RRExclusive(traditional)Inclusive(case-cohort)Concurrent(density)Case-to-caseCase-crossoverWhy case-control studyInacohort study,youneedalargenumberofthesubjectstoobtainasufficientnumberofcase,especiallyifyouareinterestedinararedisease.GastriccancerincidenceinJapanesemale:128.5/100,000personyearAcase-control study ismoreefficientintermsofstudyoperation,time,andcost.Case-control study-SequenceofdeterminingexposureandoutcomestatusStep1:Determine and select cases of your research interestStep2:Selection of appropriate controlsStep3:Determine exposure status in both cases and controlsCase ascertainmentWhatisthedefinitionofthecaseCancer(clinicallyPathologically)Viruscarriers(Asymptomaticpatients)YouneedtoscreentheantibodyIncludingdeceasedcasesYouhavetodescribethefollowingpoints,thedefinitionwhen,where&howtoselectCasedefinitionConceptualdefinitionObesitydefinedasbodyfatpercentage33%OperationaldefinitionBodyMassIndex30Casedefinition:IssuesCasedefinitionshouldavoidmisclassificationForexample:Sinhaetal(2008)AnemiawasdefinedasHemoglobin keptinbothgroupsORestimatesrelativerisk48Cohort studyCases exposed End of studyExposed population(E)Unexposed population(U)Cases unexposed Still at risk Still at risk TimeRodrigues L et al.Int J Epidemiol.1990;19:205-13.Cases Sample of source populationStill at risk Concurrent design:density case control 49Concurrent design:density case control ControlsselectedfromthosestillatriskwhenacaseoccursControlcanlaterbecomeacaseNotviceversa-casesnolongeratriskControlswholaterbecomecaseskeptinbothgroupsControlsrepresentpersonyearsatriskexperienceamongexposedandunexposedMatchedanalysisontimeofselectionORestimatestherateratio50How to select controls to estimate the respective measure of associationMeasureDesignFormulationAlternativeformulationControlstobesampledfromRiskratioInclusiveCe/NeCu/NuCe/CuNe/NuRateratioConcurrentCe/pyareCu/pyaruCe/Cupyare/pyaruOddsratioExclusiveCe/(Ne-Ce)Cu/(Nu-Cu)Ce/Cu(Ne-Ce)/(Nu-Cu)Rodrigues L et al.Int J Epidemiol.1990;19:205-13.TotalstudypopulationregardlessofpastorfuturediseasestatusPeoplecurrentlyatriskPeopledisease-freethroughoutstudyperiod51Rarediseases:allgivesimilarresultsNon-recurrentdiseasewithhighincidence-Casecohortdesign(inclusive):ORrelativeRiskRecurrentcommondisease-Densitycasecontroldesign(concurrent):OR relativeRateProbabilityoreffectofexposurechangesalongtime-Densitycasecontroldesign:ORrelative RateNoneedtoquantify-traditionaldesignWhat design and when52Relationship between OR and RR,according to the primary attack rate(AR)Acknowledgements:OlivierlePolain,EPIETCohort15HPALondonEpidemiologyUnit,UK53Cases detected by surveillance systemsNon-randomselectionprocess:Hostfactors(eg.asymptomaticinfections)DifferenthealthcareseekingbehaviourIncompletelabinvestigationIncompletereportingDifferentialrecallBetweenreportedandnotreportedcasesBetweencasesandcontrols54Case-to-case approachSamedisease,differentsubtypes/clones:SerotypesPhagetypesAntibioticresistancepatternsControls=caseswithnonepidemicsubtypesfromsamesourcepopulationsamesusceptibility(underlyingdiseases)includedascasesiftheyhadtheoutbreakstrainreadilyavailableReducesselectionANDrecallbiasFood-exposurecollectedbeforestatusisknown55Case-crossover designSamepersontakenasitsowncontrol-Nobetween-personsconfoundingMatcheddesign:CompareexposureinariskperiodtooneormorecontrolperiodsOnlypairsofdiscordantperiodsusedintheanalysisAcutediseasesExposuremustvaryovertimeshortinductionandtransienteffectsensitivetorecallbiasReferenceperiod “Washout”period Currentperiod ExposureStudyCasesMatchedpairs1Discordant0,12Discordant1,03Concordant1,14Concordant0,0Case-crossover design57Folicacidantagonists(FAA)inpregnancyandcongenitalcardiovasculardefects(CCD)Case-crossover approachCase:Woman who had a child with CCD(N=3870)Exposure:FAA during 2nd&3rd month of pregnancyControl:Woman who had a child without CCD(N=8387)OR=1.0(0.5-2.0)OR=0.3(0.2-0.6)Case-time control OR=1/0.3=2.9(1.2-7.2)-2 -1 1 2 3 4 5 6 7 8 9Cases:-2 -1 1 2 3 4 5 6 7 8 9Controls:ControlperiodRiskperiodDeliveryHernandez-DiazS.Am J Epidemiol 2003;158:385-39158ConclusionsIfyoudonotneedthatORestimatescorrectlytheRR-“traditionaldesign”Otherwise,ifyouneedOR RR-identifythebestdesignforeachsituationIfitisdifficulttofindappropriatecontrolsCasetocaseCase-crossoverQuestionAssignedreadings,session2:SchultzKF,GrimesDA.Case-controlstudies:researchinreverse.Lancet2002;359:431-4.OR=RRPleaselistsomevariantsofcasecontrolstudy.