病例对照与回顾性队列研究

Robert E.Fontaine,MD,MScBao-Ping Zhu,MD,MSUS Expert Advisors to Chinese FETP病例对照与回忆性队列研究病例对照与回忆性队列研究Case-ControlandRetrospectiveCase-ControlandRetrospectiveCohortStudiesCohortStudiesEpidemiology:The“CDC ApproachC=Count2D=DivideC=CompareProposed by:William Foege,MD,MPH,Director,US CDC(1977-1983)William Foege,MD,MPHCase-controlCohortIndividualsInterventionRetrospectiveProspectiveDescriptivePopulationsAnalyticalObservationalCase-seriesCross-sectionalEcologicClinical trials流行病学研究流行病学研究 Epidemiological studies有了初步假设，并已经进行了以下步骤。下一有了初步假设，并已经进行了以下步骤。下一步要做什么？步要做什么？You have a hypothesis and the following.What next?疾病特征疾病特征 Disease Characteristics+时间分布时间分布 Time distribution+地区分布地区分布 Place distribution+人群分布人群分布 Person distribution+通过访谈建立假设通过访谈建立假设 Hypothesis generating interviews+考虑异常值考虑异常值 Outliers=得到关于暴露的假设得到关于暴露的假设 possible exposure针对假设的暴露因素，对率进行比较针对假设的暴露因素，对率进行比较WeWecompareratesbyourhypothesizedcompareratesbyourhypothesizedexposuresexposuresWhat do we need to do this?对可疑的暴露进行准确的测量对可疑的暴露进行准确的测量 We obtain accurate measurements on the exposure of interest访谈访谈 Interviews查阅记录查阅记录 Written records测量测量 Measurements实验室检测实验室检测 Laboratory tests环境数据环境数据 Environmental data在病例对照和回忆队列研究中，从疾病开始，回溯暴露在病例对照和回忆队列研究中，从疾病开始，回溯暴露 Incase-controlandretrospectivecohortstudiesIncase-controlandretrospectivecohortstudieswebeginwiththediseaseandlookbacktothewebeginwiththediseaseandlookbacktotheexposure.exposure.Exposure DiseaseTime Referent exposure periodStudy begins在人群中，某种暴露与疾病的关系在人群中，某种暴露与疾病的关系 The relationship between exposure and cases in a populationDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计罹患率罹患率AR x+ABA+BA/(A+B)Reference 参照参照-CDC+DC/(C+D)Total 合计合计A+CB+D A+B+C+D人群罹患率人群罹患率 Population AR=(A+C)/(A+B+C+D)根据人群数据进行一些根本的流行病学比较根据人群数据进行一些根本的流行病学比较 BasicepidemiologiccomparisonsfromBasicepidemiologiccomparisonsfromthepopulationdatathepopulationdataRisk Difference:A/(A+B)C/(C+D)Risk Ratio:A/(A+B)/C/(C+D)=(AC+AD)/(AC+BC)Odds Ratio:(A/B)/(C/D)=(A/C)/(B/D)=AD/BCDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计罹患率罹患率AR x+ABA+BA/(A+B)Reference 参照参照-CDC+DC/(C+D)Total 合计合计A+CB+D A+B+C+DDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计 AR x+50950100050/1000Reference 参照参照-10990100010/1000Total 合计合计601940200060/2000Risk difference:50/1000 10/1000=40/1000Risk Ratio:50/1000/10/1000=5.0根据人群数据进行一些根本的流行病学比较根据人群数据进行一些根本的流行病学比较 BasicepidemiologiccomparisonsfromBasicepidemiologiccomparisonsfromthepopulationdatathepopulationdata从同一从同一2X2表也可以计算奥值和奥值比表也可以计算奥值和奥值比 Odds and odds ratios may be calculated from the same 2x2 tableDisease 疾病疾病Exposure 暴露暴露+-疾病奥值疾病奥值Odds of disease+ABA/B-(Reference 参照参照)CDC/D暴露奥值暴露奥值Odds of exposureA/CB/D暴露的奥值比暴露的奥值比 Odds ratio exposure=(A/C)/(B/D)=AD/BC疾病的奥值比疾病的奥值比 Odds ratio disease=(A/B)/(C/D)=(A/D)/(B/C)=AD/BC暴露奥值与疾病奥值完全相等暴露奥值与疾病奥值完全相等 Odds ratios from exposure or disease odds are identicalDisease 疾病疾病Exposure 暴露暴露+-Odds of disease+509505/95=1/19Reference 参照参照-109901/99Odds of exposure5/195/99Odds Ratio exposure=(5)/(95/99)=5*99/95=99/19Odds Ratio disease=(1/19)/(1/99)=99/19OR也是病例与非病例的交叉乘积比也是病例与非病例的交叉乘积比 The OR can be calculated by the cross product of cases and not illDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+ABA+BReference 参照参照-CDC+DTotal 合计合计A+CB+DA+B+C+DADBCOR=(A/B)/(C/D)=(A/C)/(B/D)=RR是疾病与合计之间的交叉乘积比是疾病与合计之间的交叉乘积比 The RR can be calculated by the cross product of cases and totalsDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+ABA+BReference 参照参照-CDC+DTotal 合计合计A+CB+DA+B+C+DRR=A/(A+B)/C/(C+D)=A(C+D)/C(A+B)=AC+ADAC+BC奥值比与危险比之间的关系奥值比与危险比之间的关系 Relationship of the odds ratio to the risk ratioWhen RR 1,OR RRWhen RR 1,OR RRWhen RR=1,OR=RR=101RR ORRROR假设为稀有疾病，假设为稀有疾病，ACAC很小，很小，OROR RRIftheRRIfthediseaseisrarethetermACissmallandthediseaseisrarethetermACissmallandtheORandRRarenearlyidenticalORandRRarenearlyidenticalDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计 AR x+ABA+BA/(A+B)Reference 参照参照-CDC+DC/(C+D)Total 合计合计A+CB+DA+B+C+DRisk Ratio:A/(A+B)/C/(C+D)=(AC+AD)/(AC+BC)Odds Ratio:(A/B)/(C/D)=(A/C)/(B/D)=AD/BCAC+ADAC+BCADBCRRORDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+59951000Reference 参照参照-19991000Total 合计合计a+cb+da+b+c+dDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+509501000Reference 参照参照-109901000Total 合计合计a+cb+da+b+c+d5(999)1(995)5+5(999)5+1(995)50(990)10(950)500+50(990)500+10(950)499599550001000495009500=5.02=5.0=5.2=5.05000010000假设为稀有疾病，假设为稀有疾病，ACAC很小，很小，OROR RRIftheRRIfthediseaseisrarethetermACissmallandthediseaseisrarethetermACissmallandtheORandRRarenearlyidenticalORandRRarenearlyidentical病例对照与回忆性队列研究的比较病例对照与回忆性队列研究的比较HowareHowarecasecontrolstudiesandretrospectivecasecontrolstudiesandretrospectivecohortstudysimilar?cohortstudysimilar?Disease 疾病疾病Exposure 暴露暴露+-Total 合计合计+ABA+BReference 参照参照-CDC+DTotal 合计合计A+CB+DA+B+C+D回忆性队列研究从暴露开始进行抽样回忆性队列研究从暴露开始进行抽样 AretrospectivecohortstudysamplesAretrospectivecohortstudysamplesbyexposurebyexposureDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+AB(A+B)kReference 参照参照-CD(C+D)jTotal 合计合计k=proportion of exposed persons sampledj=proportion of unexposed persons sampledThis results in the following distribution inside the table.Disease 疾病疾病Exposure 暴露暴露+-Total 合计合计+AkBk(A+B)kReference 参照参照-CjDj(C+D)jTotal 合计合计A+CB+DA+B+C+D抽样比例约除，得到人群的抽样比例约除，得到人群的RD,RR,OR The sample fractions cancel leaving us with the RD,RR,and OR of the populationDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+AkBk(A+B)kReference 参照参照-CjDj(C+D)jTotal 合计合计A+CB+DA+B+C+DRD=Ak/(A+B)k-Cj/(C+D)j=A/(A+B)-C/(C+D)RR=Ak/(A+B)k/Cj/(C+D)j=A/(A+B)/C/(C+D)=(AC+AD)/(AC+BC)OR=(Ak/Bk)/(Cj/Dj)=(A/B)/(C/D)=AD/BC病例对照研究是从疾病状态开始抽样病例对照研究是从疾病状态开始抽样 A case control study samples by disease status.Disease 疾病疾病Exposure 暴露暴露+-Total 合计合计+A+BReference 参照参照-C+DTotal 合计合计(A+C)k(B+D)jA+B+C+Dk=proportion of diseased persons sampledj=proportion of well persons sampledDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+AkBjA+BReference 参照参照-CkDjC+DTotal 合计合计(A+C)k(B+D)jA+B+C+D病例对照研究是从疾病状态开始抽样病例对照研究是从疾病状态开始抽样 A case control study samples by disease status.抽样比例约除，得到人群的抽样比例约除，得到人群的ORAgain,the sampling fractions cancel leaving the OR of the populationDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+AkBjA+BReference 参照参照-CkDjC+DTotal 合计合计(A+C)k(B+D)jA+B+C+DOR=(Ak/Bj)/(Ck/Dj)=ADjk/BCjk=AD/BC假定在人群中随机抽取样本作为对照假定在人群中随机抽取样本作为对照 Suppose that you use as controls a sample of the population.Disease 疾病疾病Exposure 暴露暴露+-Total 合计合计+AkBj(A+B)jReference 参照参照-CkDj(C+D)jTotal 合计合计(A+C)k(B+D)j(A+B+C+D)j这种研究称为这种研究称为“疾病疾病-参比研究参比研究ThisisaThisisacase-base(orcase-referent,case-case-base(orcase-referent,case-cohort)studycohort)studyDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+AkB(A+B)jReference 参照参照-CkD(C+D)jTotal 合计合计(A+C)k(B+D)(A+B+C+D)j此时，交叉积比是人群此时，交叉积比是人群RR，而非，而非OR Notice that the cross product is now a RR and not an ORDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+AkB(A+B)jReference 参照参照-CkD(C+D)jTotal 合计合计(A+C)k(B+D)(A+B+C+D)jCross Product=Ak(C+D)j/Ck(A+B)j=Ak(C+D)j/Ck(A+B)j=A/(A+B)/C/(C+D)=A(C+D)/C(A+B)=RR病例对照，或回忆性队列研究？要考虑：时间，花费，病例对照，或回忆性队列研究？要考虑：时间，花费，精确度精确度Case-controlorretrospectivecohort?Case-controlorretrospectivecohort?Time,cost,andaccuracyTime,cost,andaccuracy50 cases among 200 persons50 cases among 2000 persons500 cases among 2000 persons病例对照研究的主要步骤病例对照研究的主要步骤 Major steps in case-control study定义并选择病例定义并选择病例 Define and select cases Define and select cases定义并选择对照定义并选择对照 Define and select Define and select controlscontrols确定暴露确定暴露 Ascertain exposures Ascertain exposures比较病例和对照的暴露史比较病例和对照的暴露史 Compare exposure Compare exposure in cases and controlsin cases and controls对照代表病例所来自的人群对照代表病例所来自的人群 A control is a representative of the population from which the cases came.Disease 疾病疾病Exposure 暴露暴露CaseControlTotal 合计合计+aba+b-cdc+dTotal 合计合计a+cb+da+b+c+d不要与不要与实验研究中的研究中的对照照组相混淆相混淆 Do not confuse with experimental controls.病例定义病例定义 Case definition20072007年北京市朝阳区常住人口中实验室确诊的脑膜炎球菌疾病。年北京市朝阳区常住人口中实验室确诊的脑膜炎球菌疾病。“Laboratory-confirmed meningococcal disease during 2007“Laboratory-confirmed meningococcal disease during 2007 in regular residents of Chaoyang District,Beijing.in regular residents of Chaoyang District,Beijing.对照定义对照定义 Control definition20072007年北京市朝阳区常住人口中未发生脑膜炎球菌疾病的居民。年北京市朝阳区常住人口中未发生脑膜炎球菌疾病的居民。“Regular residents of Chaoyang District,Beijing“Regular residents of Chaoyang District,Beijing during 1997 and not suffering from meningococcal during 1997 and not suffering from meningococcal disease.disease.选择对照的原那么选择对照的原那么PrinciplesofPrinciplesofControlSelectionControlSelection与病例来源于同一人群与病例来源于同一人群FromsametargetFromsametargetpopulationascasespopulationascases对照的选择应该独立于暴露状态对照的选择应该独立于暴露状态SelectedSelectedindependentlyofexposurestatus.independentlyofexposurestatus.如果对照发病，那么他们就会是病例如果对照发病，那么他们就会是病例IftheyhadIftheyhaddevelopedillness,theywouldhavebeenadevelopedillness,theywouldhavebeenacase.case.信息与病例具有可比性信息与病例具有可比性ComparableinformationComparableinformationtocasestocases对照与病例应该有相同的暴露时机对照与病例应该有相同的暴露时机TheymustTheymusthavethesameopportunityforexposureasahavethesameopportunityforexposureasacasecase病例与对照具有相同的入选标准病例与对照具有相同的入选标准UsethesameUsethesameinclusionandexclusioncriteria.inclusionandexclusioncriteria.病例对照研究中，设立对照组的目的病例对照研究中，设立对照组的目的 Case-control study:Purpose of the comparison(control)group“设立对照组的目的是提供病例的源人群中设立对照组的目的是提供病例的源人群中暴露因素分布的估计。暴露因素分布的估计。“The comparison(Control)group“The comparison(Control)group serves to provide an estimate of the serves to provide an estimate of the exposure distribution in the study exposure distribution in the study base from which the cases originate.base from which the cases originate.-Rothman KJ,1986-Rothman KJ,1986Extra Slides横断面研究Cross-Sectional Studies横断面研究与病例对照、回忆性队列研究有何不同？横断面研究与病例对照、回忆性队列研究有何不同？Whatisthedifferencebetweenacase-control,Whatisthedifferencebetweenacase-control,aretrospectivecohort,andacross-sectionalaretrospectivecohort,andacross-sectionalstudy?study?Disease 疾病疾病Exposure 暴露暴露+-Total 合计合计+AAA+BReference 参照参照-CDC+DTotal 合计合计A+CB+DA+B+C+D人群中疾病与暴露的分布人群中疾病与暴露的分布病例对照或回忆性队列研究那么根据疾病或暴露状病例对照或回忆性队列研究那么根据疾病或暴露状态来抽样态来抽样Case-controlandretrospectiveCase-controlandretrospectivecohortstudiessamplebycaseorexposurecohortstudiessamplebycaseorexposurestatusstatusDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+AjBj(A+B)jReference 参照参照-CjDj(C+D)jTotal 合计合计(A+C)j(B+D)j(A+B+C+D)j横断面研究：从全人群中进行抽样横断面研究：从全人群中进行抽样 In a cross-sectional study the entire population is sampledDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+ABA+BReference 参照参照-CDC+DTotal 合计合计A+CB+D(A+B+C+D)j抽样后再去调查，确定疾病和暴露状态抽样后再去调查，确定疾病和暴露状态 Disease or exposure status are determined during or after the samplingDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+?Reference 参照参照-?Total 合计合计?(A+B+C+D)j剂量反响关系剂量反响关系Dose-responserelationshipDose-responserelationship暴露水平在两个以上暴露水平在两个以上 Exposures may take on several levels of intensityDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+a3b3a3+b3+a2b2a2+b2+a1b1a1+b1Reference 参照参照-a0b0a0+b0Total 合计合计aba+b每个暴露水平都应与一个共同参照水平比较每个暴露水平都应与一个共同参照水平比较 EachlevelneedstobecomparedtotheEachlevelneedstobecomparedtothesamecommonreferencelevelsamecommonreferencelevelDisease 疾病疾病Exposure 暴露暴露+-Total 合计合计+a3b3a3+b3+a2b2a2+b2+a1b1a1+b1Reference 参照参照-a0b0a0+b0Total 合计合计ab(a+b)OR=a3b0/b3a0RR=a3/(a3+b3)/a0/(a0+b0)=a3(a0+b0)/a0(a3+b3)Disease 疾病疾病Exposure 暴露暴露+-Total 合计合计+a3b3a3+b3+a2b2a2+b2+a1b1a1+b1Reference 参照参照-a0b0a0+b0Total 合计合计ab(a+b)OR=a2b0/b2a0RR=a2/(a2+b2)/a0/(a0+b0)=a2(a0+b0)/a0(a2+b2)每个暴露水平都应与一个共同参照水平比较每个暴露水平都应与一个共同参照水平比较 EachlevelneedstobecomparedtotheEachlevelneedstobecomparedtothesamecommonreferencelevelsamecommonreferencelevel病例对照与回忆性队列研究比较Case-controlvs.RetrospectiveCohortStudies病例对照与回忆性队列研究有何类似之处？病例对照与回忆性队列研究有何类似之处？Howarecase-controlandretrospectiveHowarecase-controlandretrospectivecohortstudiessimilar?cohortstudiessimilar?病例病例Cases暴露暴露Exposure+-合计合计 Total+aba+b-cdc+d合计合计 Alla+cb+da+b+c+d累计发病病例累计发病病例 Cumulative incidence 用用2x2 表来表示表来表示 2x2 table arrangement疾病和暴露均已发生疾病和暴露均已发生 Both disease&exposure have already occurred病例对照与回忆性队列研究有何不同之处？病例对照与回忆性队列研究有何不同之处？HowdocasecontrolandretrospectiveHowdocasecontrolandretrospectivecohortstudiesdiffer?cohortstudiesdiffer?病例病例Cases暴露暴露Exposure+-合计合计 Total+aba+b-cdc+d合计合计 Alla+cb+da+b+c+d回忆性队列研究回忆性队列研究 率，率比，奥值比率，率比，奥值比 Retrospective cohort rates,rate ratios,&odds ratios病例对照研究病例对照研究 奥值，奥值比奥值，奥值比Case-control odds&odds ratios 如何从回忆性队列研究计算率，率比，奥值比如何从回忆性队列研究计算率，率比，奥值比？Howtocalculatetheattackrate,riskHowtocalculatetheattackrate,riskratio&oddsratiofromaretrospectiveratio&oddsratiofromaretrospectivecohort?cohort?病例病例 Cases暴露暴露Exposure+-合计合计 TotalAR+aba+ba/(a+b)-cdc+dc/(c+d)合计合计 Alla+cb+da+b+c+d率差率差 Risk difference:a/(a+b)c/(c+d)率比率比 Risk Ratio:a/(a+b)/c/(c+d)=(ac+ad)/(ac+bc)奥值比奥值比 Odds Ratio:(a/b)/(c/d)=(a/c)/(b/d)=ad/bc暴露与非暴露组样本大小不会影响暴露与非暴露组样本大小不会影响RR和和OR的计算的计算 Different sample sizes of exposed and unexposed will give correct risk ratios and odds ratios CasesExposure+-Total+akbk(a+b)k-cjdj(c+d)jAlla+cb+da+b+c+d Risk difference=ak/(a+b)k-cj/(c+d)j=a/(a+b)-c/(c+d)Risk ratio=ak/(a+b)k/cj/(c+d)j=a/(a+b)/c/(c+d)=(ac+ad)/(ac+bc)Odds ratio=ak/bk)/(cj/dj)=(a/b)/(c/d)=ad/bc问题：在作了病例对照研究以后，是否需要问题：在作了病例对照研究以后，是否需要再用回忆性队列研究进行验证？再用回忆性队列研究进行验证？Question:Afterconductingacase-Question:Afterconductingacase-controlstudy,isitnecessarytocontrolstudy,isitnecessarytoconductaretrospectivecohortstudyconductaretrospectivecohortstudytofurtherverifythehypothesis?tofurtherverifythehypothesis?为什么？为什么？Why or why not?