隊列研究CohortStudy

隊列研究

Cohortstudy

Incidencestudies

Longitudinalstudies

Follow-upstudies

Prospectivestudies發病率研究縱向研究隨訪研究前瞻性研究OutlineExampleDesignandimplementMeasuresofRiskDatasortandanalysisBiasandControlAdvantage，disadvantage概述實例研究設計與實施資料的整理與分析偏倚及控制優缺點隊列研究

Cohortstudy隊列研究簡史定義與目的原理與類型HistoryofcohortstudyDefinitionandPurposePrincipleandTypes歷史History起源

OriginsGraunt(17thcentury)usedcross-sectionalmortalitydatatoreconstructlifehistoryusinglife-tablemethodsGraunt(17世紀)用橫斷面死亡資料，用壽命表方法構建生命史Farr(19thcentury)advancedtheuseoflife-tablemethodsaanindicatorofpopulationhealthFarr（19世紀）將壽命表方法發展為人群健康的指標Insuranceindustrystudy1870–1899

保險業研究1870–1899Tuberculosis(20thcentury)

結核（20世紀）WHFrostperformedthefirstretrospectivecohortstudyinacohortof132homeswithtuberculosisWHFrost在有結核的132個家庭的隊列開展第一個回顧性隊列研究Usedperson-yearstoestimateattackrates用人年來估計罹患率歷史HistoryWHFrostinitiatedprospectivecohortstudyoftuberculosisinWilliamsoncounty,TennesseeWHFrost在德州啟動有關結核的前瞻性隊列研究歷史HistoryFraminghamstudyofcardiovasculardisease,1948Japaneseatomicbombsurvivors,1946Britishphysicianstudy,1950sColoradoPlateauuraniumminers,1950sAniline-dyeoccupationalcohort,1954Asbestosexposureandlungcancermortality1965RetrospectivecohortstudiesProspectivecohortstudiesNursesHealthstudy,1976topresentBritishphysicianstudyMulti-centerAIDSCohortStudyMACS,1984–1999

CurrentStudies

Anepidemiologicdesigninwhichtheincidenceofadisease(orcondition)iscomparedamongexposedandunexposedindividuals

是比較暴露與非暴力人群發病率的一種流行病學設計

定義What’thecohortstudyCohortStudyBeginwithdisease-freepatientsClassifypatientsasexposed/unexposedRecordoutcomesinbothgroupsCompareoutcomesusingrelativerisk從沒有疾病人群開始

將研究對象分為暴露與非暴露組記錄兩組結局用相對危險度比較結果CohortStudy

KeyPoint:Presenceorabsenceofriskfactorisdeterminedbeforeoutcomeoccurs.Whatisacohort?

隊列Cohort

-Latinwordforoneofthe10divisionsofaRomanlegionAgroupofindividuals

sharingsameexperiencefollowedupforaspecifiedperiodoftimeExamplesbirthcohortoccupationalcohortchemicalplantworkers隊列拉丁語原意是指古羅馬軍團中的一個分隊一組人群有共同經歷隨訪一特定時間例如出生隊列職業隊列暴露指接觸過某種物質、具備某種特征或處于某種狀態。危險因素泛指能引起某特定不良結局,或使其發生的概率增加的因子，包括個人行為、生活方式、環境和遺傳等多方面的因素。ExposurecontactsomematerialofbeinginsomestatusRiskfactorisavariableassociatedwithanincreasedriskofdiseaseorinfection.Includingbehavior,lifestyle,environmentalandgeneticfactorandsoon.

目的

Purpose

Describeincidenceofoutcomesovertime,naturalhistoryofdiseaseDeterminecausalrelationshipsbetweenthoseoutcomesandexposures(riskorprognostic)factors.

描述隨著時間變化結果的發生率，疾病自然史確定結果與暴露（危險或預后）因素間的因果關系目的PurposeDetermineaprognosisEvaluatenewtherapiesanddiagnosticsEvaluatescreeningprocedures確定預后評價新的治療與診斷方法評價篩檢過程FraminghamstudyofcardiovasculardiseaseIndividuals30–62yearsoldincommunityatriskfordiseaseFramingham,MA,1948topresentExamplesofcohortstudyGoaltoelucidatethenaturalhistoryofHIV/AIDS5000gaymen,volunteers5citiesinUS1984–Extensiveevaluations

QuestionnairePhysicalexaminationLaboratorytestingMulti-CenteredAIDSCohortStudySource:partiallyadaptedfromWHO,1993R1R0A1A0IncidencerateamongexposedIR1=A1(no.exposedcases)

/R1(totalperson-timeexposed)IncidencerateamongunexposedIR0=A0(no.unexposedcases)/R0(totalperson-timeunexposed)Incidencerateratio(exposedvs.unexposed)=IR1/IR0=(A1/A0)/(R1/R0)PrincipleofaCohortStudyCharacteristics

特點ObservationalCategorybyexposureNeedcontrolProspective觀察性研究根據暴露分組需要對照前瞻性類型TypesBasedonrecruittimeofstudysubjectProspectivestudyv.s.RetrospectiveStudyBasedonthetypeofcohortFixedcohortv.s.Dynamiccohort

依據研究對象召集的時間前瞻性研究VS回顧性依據隊列的類型固定隊列VS動態隊列

類型TypesCombinedwith

case-controlstudy

Nestedcase-controlstudycase-cohortstudy與病例對照研究結合

巢式病例對照研究病例隊列研究

ProspectivecohortstudytimeExposureStudystartsDiseaseoccurrence依據研究對象召集的時間分類ProspectivecohortstudyIdentifycohortinthepresentDetermineexposurestatusorpossibleexplanatory/prognosticfactorsInthepresentorinthefuture現在確定隊列確定暴露狀態或可能暴露/預后因素在現在或將來

依據研究對象召集的時間分類Follow-uptoidentifyoutcomeQuestioncanbeincidenceorriskofoutcomeQuestioncanbewhatfactorsassociatedwithoutcomeAscertainmentofoutcomedoneinfuture隨訪確定結局可以是發病率或結局的風險可以是什么因素與結局有關確認將來發生的結局

依據研究對象召集的時間分類ProspectivecohortstudyProspectiveCohortExampleQuestion:Arenon-steroidalanti-inflammatorydrugsariskfactorforGIbleeds?問題:非甾體固醇抗炎藥是否為胃腸道出血的危險因素?ExampleIdentifycohort:newdiagnosesofrheumatoidarthritisfromOctober17,2002toOctober17,2003Determineexposurestatus:identifypatientsprescribedNSAID’sandthosewhoarenot確定隊列:2002.10-2003.10診斷的類風濕關節炎患者確定暴露狀態:確定開與未開NSAID’s藥物處方的病人Determineoutcomes:follow-upallpatientsfor1year–identifyhowmanyGIbleedsthereareineachsub-cohortorexposuregroup確定結局:

隨訪所有的病人1年—確定在暴露及非暴露組出現多少胃腸道出血的病人ExampleProspectiveCohortStudiesAdvantagetimesequencestrengthens

inferenceaboutcausemoreaccurate

measurementofrisk

factors(donotneedto

reconstructpastexposures)morecomplete

measurementofconfounding優點因果推斷的時間順序強危險因素測量跟準確(不需要夠姜過去的暴露)更準確的測量混雜因素Disadvantage:expensiveandtimeconsuming(inceptionandfollow-up)largenumbersrequiredtostudyrareoutcomesdifficulttostudychronicdiseaseswithlonglatency缺點:昂貴,費時間(啟動及隨訪)研究罕見結局是需大樣本難與研究慢性潛伏期長的疾病ProspectiveCohortStudies前瞻性隊列研究應用條件明確地檢驗假設所研究疾病的發生率較高,一般不低于5‰明確規定暴露因素和結局變量可靠的測量手段足夠的觀察人群和暴露情況能完成隨訪的人群足夠的人、財、物力RetrospectivecohortstudytimeExposureStudystartsDiseaseoccurrenceHistorical(Retrospective)CohortStudyIdentifycohortinthepastE.g.,throughrecordsoradministrativedatabasesDetermineexposureorprognosticfactorsinthepastAgain,recordsordatabasesIdentifyoutcomeOutcomecanbeidentifiedinpastorpresentOutcomemustbeafterprevioustwostepsHistoricalCohortExampleQuestion:Arenon-steroidalanti-inflammatorydrugsariskfactorforGIbleeds?問題:非甾體固醇抗炎藥是否為胃腸道出血的危險因素?HistoricalCohortExampleIdentifycohort–e.g.,peoplediagnosedwithrheumatoidarthritisbetweenJanuaryandDecember1992Determineexposurestatus.WhichofthesepatientswereprescribedNSAID’s?Determineoutcome.Didpatientdevelopkidneydiseaseduringthefiveyearsafterinceptionintocohort?(follow-uplasteduntilDecember1997)Historical(Retrospective)CohortStudies

Advantage:cohorteasiertoassemble(inceptionperiodinpast)baselinemeasurementsalreadyavailablefollow-upperiodalreadytakenplacelesscostlyandtime-consuming優點:隊列易召集(在過去起始)基線測量已經可以利用已經隨訪耗費少,省時間Disadvantage:nocontroloverthequalityofpastmeasurementsincompletedatasetscontrolforconfoundingmaybeincomplete缺點無法控制過去暴露的質量資料不全混雜因素控制不全Historical(Retrospective)CohortStudiesMixedCohortExampleQuestion:Arenon-steroidalanti-inflammatorydrugsariskfactorforGIbleeds?Identifycohort(patientswhowerediagnosedwithrheumatoidarthritisbetweenOctober2000andSeptember2001)WhichofthesepatientsareprescribedNSAID’s?HowmanyofthesepatientswillexperienceGIbleedsduringnextfiveyears?FixedCohortStartEndCohortStudyDesignTypesFixedCohort

AgroupofindividualsrecruitedandenrolledatauniformpointinthenaturalhistoryofadiseaseorbysomedefiningeventCohortdoesnottakeonnewmembersafteritisassembledExamplesPatientsadmittedtotheERwithacuteMISurvivorsofHiroshimabombingsChildrenborntoHIV-infectedmothersTypesDynamiccohortAgroupofindividualsrecruitedandenrolledthroughamechanismthatallowsforinandoutmigrationofpeopleDefinedbycharacteristicotherthandisease,e.g.,geographiclocation,administrativeunitDynamicpopulationExamplesFraminghamStudyKaiserPermanenteDynamicCohortStartEnd二硫化碳長期低劑量的暴露與冠心病的關系

第二節實例研究二硫化碳（CS2）神經系統毒物，抑制酶的活性，影響脂蛋白代謝，造成心血管疾病長期接觸低濃度CS2可引起慢性中毒和動脈粥樣硬化短時間接觸高濃度的CS2蒸氣可急性中毒研究因素長期低劑量的CS2暴露定義在有CS2暴露但不至引起急性中毒的車間工作>5年20世紀60年代芬蘭職業衛生研究所Hernberg和Tolonen教授做的前瞻性隊列研究二、確定研究結局

心肌梗死血壓變化心電圖的改變心絞痛發作三、確定研究現場和人群

暴露組

1942—1967年某粘纖廠25至64歲，343名男性工人有5年以上CS2暴露史對照組

年齡±3歲出生地區相同工種的體力消耗相當在同一城市的造紙廠隨機選擇的343名男性工人

四、資料收集

查閱檔案記錄用藥情況、既往車間CS2的濃度等詢問

姓名、性別、年齡、工種及工作年限、吸煙、業余時間的體力活動情況實驗室檢查

血糖、血脂、血清膽固醇水平、血壓、心電圖、心臟大小、體重及車間CS2濃度的動態變化

五、資料分析表4-1暴露組和對照組的心肌梗死發生率及RRCS2暴露組發生心肌梗死的RR為3.57，兩組致死性心肌梗死發生率和總的心肌梗死發生率差異有顯著性

CS2在不同臨床類型冠心病的發生中作用程度不同臨床類型RRAR心肌梗塞3.575.25致死性心肌梗塞4.693.21非致死性心肌梗塞2.742.04心絞痛1.8911.6心電圖冠心樣改變1.46.1表4-2CS2與不同臨床類型冠心病的RR和AR比較六、結論

長期低劑量與冠心病發病和死亡存在因果關系CS2所致的冠心病，以致死性心肌梗死為主措施芬蘭當局已于1972年把CS2的車間最高容許濃度從20ppm降至10ppmIdentifyexposuresIdefineoutcomesIdentifystudyfieldandpopulationSamplesizeDatacollectioanandFollow-upbothgroupsQualitycontrol確定研究因素確定研究結局確定研究現場與研究人群確定樣本量資料的收集與隨訪質量控制第三節設計和實施DesignandimplementIdentifyExposure

確定研究因素MainexposureBasedondescriptivestudyandcase-controlstudyFactormayeffectoutcomeconfounder,demographycharacteristic

主要暴露因素在描述性研究和病例對照研究的基礎上確定可能影響結局的因素混雜因素人口學特征等MeasuringExposureContent-Natureoftheexposure;biologicmechanismsQualityContinuous-e.g.,serumcholesterolPeriodic-e.g.,cigarettes,sexualcontactsSingular-e.g.,nuclearexposureQuantityContinuousandperiodicexposuresmustbequantifiedDose-responserelationshipMeasuringExposureMeasurementsInterviewMedicalexamBloodtestsorotherspecimens,Biomarkers,OtherlaboratorytestsRecords,medicalrecordsSamplestorageEnvironmentsurveillancedataMeasuringExposureMeasuringexposureisoneofthefundamentalactivitiesofacohortstudyExposuremeasurementmustbecomparableforallmembersofthecohortCarefullydefinedinadvanceofstudySpecificattentionshouldbegiventotheaccuracyandprecisionofproposedmeasurementsPilotstudiesoftenneededOutcomeDefinitionExpectedresultsoffollow-upPrimaryoutcome-themaineventthatwillberelatedtotheexposureFailure-timeoutcomesDeathDiseaseoccurrenceRepeatedmeasures

OutcomeDefinitionSecondaryoutcomes-othereventsthatareofinterestandmaycorroboratethefindingsofthemainoutcomeUsingexistedinternationalstandardExamplesofOutcomesorDiseasesLungCancerHeartDiseaseMotorvehicleinjuryHIVinfectionDiabetesDiphtheriaDefineConfoundingDefineconfoundinginthestudyControlconfoundingindesigndatacollectiondataanalysis

StudyPopulationDefinePopulationatRiskusinginclusioncriteriaIndividualswithoutcomeofinterestattimeofscreeningandenrollmentarenoteligibleforstudySelectionofexposedpopulationThegeneralpopulation(i.e.,theoutcomeofinteresthasahighincidencerate)Specialexposuregroups(e.g.,smokers,X-rayworkersSpecialoccupationalexposuregroup(uraniumminersorasbestosworkers.Cooperation,goodrecord,regu.Exam,easyfollow-up)SelectionofexposedpopulationSpecialresourcegroups(e.g.,alumni,physicians,nurses,insured)Geographicallyorfacility-definedgroups(e.g.,ThreeMileIsland,hospitalswithspecializedmaternitycare)SelectionofNonexposedgroup

GeneralpopulationAccordingtotheexposurestatusgroupcanbedividedintosubgroups(exposure+andexposure-)Specificcomparisongroup Example:foreffectofradiationifradiologistscohortgroupcomparisongroupmaybeinternalizes. Textileworkersforasbestosworkers,

SelectionofNonexposedgroup

StudyPopulationsExamplesFraminghamstudyofcardiovasculardiseaseIndividuals30–62yearsoldincommunityatriskfordiseaseFramingham,MA,1948topresentFraminghamStudyCohortAssemblyNo.MenNo.WomenTotalRandomSample3,0743,4336,507Respondents2,0242,4454,469Volunteers312428740RespondentsfreeofCHD19752,4184,393VolunteersfreeofCHD307427734TotalfreeofCHD2,2822,8455,127StudyPopulationsMACSMulti-CenteredAIDSCohortStudyGoaltoelucidatethenaturalhistoryofHIV/AIDS5000gaymen,volunteers5citiesinUS1984–1999ExtensiveevaluationsQuestionnairePhysicalexaminationLaboratorytestingRepository

SampleSizeIncidenceingeneralpopulationIncidenceinexposedpopulationSignificantlevelPower(1-)EstimatedofRRFollow-upPurposeoffollowingupTracksubjectsinbothexposedandnon-exposedgroupDefineoutcomeeventsFurthercollectdatainexposuresandconfounding

Follow-upCompletenessandnon-participation90%ruleofthumbAllsubjectsmusthaveanequallikelihoodfordetectingtheoutcomeDiseaseascertainmentmustbecomparablebetweentheexposedandunexposedsubjectsNumberofvisitsReasonsforadditionalevaluationsFollow-upmechanismsActivePassiveFollow-upFollow-upPassiveSurveillanceHospitalsDiseaseRegistriesClinicsorphysicianofficesSurveillancesystems,e.g.,NationalDeathIndex,CDCreportableconditionsActivesurveillanceSystematicevaluationsforoutcomeofinterestRegulartimeintervalsInallstudysubjects Regardlessofactiveorpassivesurveillance,thepersonsevaluatingsubjectsmustbeblindedtoexposurestatusFollow-upFollowup

WhentostartfollowNeedtohavetimeaftertheexposureforthediseasetodevelop.Theinductionperiod+incubationFollowupperiodHowlongtofollowupIndividualsfollowuptooutcome(disease,death,syndrome)CohortsfollowupexpectedresultsFollowupintervaldependontheincidence,latency.THEISSUEOFFOLLOW-UPSomecohort(retrospectiveorprospective)studiesextendoverlongperiodsoftime.Difficulttotrackindividualsandevents.Ifalargeproportionofparticipantsarelosttofollow-upthevalidityofthestudymaybequestion—follow-upbiasIflosstofollow-upisdifferentialbetweencohortgroups,and/orforreasonsrelatedtoboththeexposureandoutcome,thevalidityofthestudymaybeinquestion.THEISSUEOFFOLLOW-UP調查員選擇調查員培訓制定調查員手冊監督ChoseofinvestigatorTrainingofinvestigatorFormulationbrochureforinvestigatorMonitor質量控制QualityInsurance

第三節設計與實施BasicmodelfordatasortingPersontimecalculateRatecalculateEffectestimation

第四節資料的整理和分析Datasortingandanalysis資料的基本整理模式人時的計算率的計算效應估計一、資料的基本整理模式

Basicmodelofdatasorting病例case非病例noncase合計total暴露組exposureaba+b=n1非暴露組Nonexposurecdc+d=n0合計a+c=m1b+d=m0a+b+c+d=t暴露組發病率=a/n1Incidenceofexposedgroup非暴露組發病率=c/n0Incidenceofnon-exposed

表4-3隊列研究資料歸納整理表Datasorttableforcohortstudy二、人時的計算

Calculateforperson-time

精確法近似法壽命表法ExactmethodApproximationmethodLifeexpectancymethod三、率的計算

calculateofrate累積發病率發病密度標化死亡比標化比例死亡比CumulativeincidenceIncidencedensityStandardizedmortalityratio,SMRStandardizedproportionalmortalityratio,SPMR

變化范圍0-1

range:0-1

適用條件樣本大

suitablefor:largesample

人口穩定

stablepopulation

整齊的資料

evendata

報告時必須注明時間長短Notethetimeperiodwhenreport累積發病率

(cumulativeincidence,CI)CI=觀察期內發病（或死亡）Numberofnewcase(ordeath)

觀察開始時的人口數

Numberofpersonsenteringobservation

發病密度

(incidencedensity)

變化范圍0-∞range:0-∞

適用條件觀察時間長

Suitablefor:longobservation

人口不穩定

Unstablepopulation

存在失訪

withlostfollowup

資料不很整齊

datanotevenID=觀察期內發病（或死亡）人數Numberofnew(death)cases觀察人時Observedpersontime標化死亡比

(standardizedmortalityratio,SMR)變化范圍0-∞

range：

0-∞適用條件結局事件的發生率低

suitablefor：lowincidenceofoutcome

不宜直接計算率時

notsuitablecalculateratedirectly

SMR=研究人群觀察期內發病（或死亡）人數Numberofnew(death)casesinstudiedpopulation標準人口預期發病（或死亡）人數Numberofexpected(death)casesinstandardpopulation

全人口某病的發病（死亡）率×觀察人口數Newcases(death)ofadiseaseintotalpopulationXNumberofobservedpersons預期發病（死亡）數的計算：Expectedcases(death):SMR的意義

被研究人群發生（死于）某病的危險性是標準人群的多少倍

Thetimesofriskofadisease(ordeath)instudypopulationoverstandardpopulationSMR=1研究人群某病發病（死亡）危險=標準人群

IftheSMRisquotedasaratioandisequalto1.0,thenthismeansthenumberofobserveddeathsequalsthatofexpectedcases.

SMR>1研究人群某病發病（死亡）危險>標準人群，是標準人群的SMR倍

Ifhigherthan1.0,thenthereisahighernumberofdeathsthanisexpected,istheSMRtimesofstandardpopulation

SMR<1

研究人群某病發病（死亡）危險<標準人群標化比例死亡比

（standardizedproportionalmortalityratio,SPMR）

變化范圍0~∞

適用條件不能得到歷年人口資料僅有死亡人數、原因、日期和年齡

SPMR=ActuallydeathNo.ExpecteddeathNo.Range0~∞Suitablefor:noyearlypopulationdata,onlythereisdeathNo.,cause,dateandageofdeath

預期死亡數計算（ExpecteddeathNo）：全人口中某病因死亡數全部死亡數×某單位實際全部死亡數Deathduetoadisease

TotalNo.ofDeath×ActuallydeathNo.insomedepartment率的顯著性檢驗

significancetestofrateU檢驗

Utest

直接概率法

Probablenumbermethod

二項分布檢驗

Binomialdistribution

泊松分布檢驗

Poissondistribution

2檢驗

2test

計分檢驗scoretest

四、效應的估計

TheEstimationofEffectRelativeRisk(RR)AttributableRisk(AR)ARPercent(AR%)PopulationAR(PAR)PARPercent(PAR%)Doseresponserelationship

相對危險度歸因危險度歸因危險度百分比人群歸因危險度人群歸因危險度百分比劑量反應關系相對危險度（RelativeRiskRR）

意義

implicationE發病或死亡的危險是ē的多少倍

thetimesoftheprobabilityofthediseaseordeathoccurringintheexposedgroupversusanon-exposedgroup.RR值暴露的效應暴露與結局關聯強度

RReffectofexposedtheassociationbetweenexposureandoutcome暴露組率Rateinexposed非暴露組率Rateinnonexposed意義

implication

吸煙者因肺癌死亡的危險是非吸煙者的10.7倍

Smokerswouldbe10.7timesaslikelyasnon-smokerstodieoflungcancer

吸煙者因心血管疾病死亡的危險是非吸煙者的1.7Smokerswouldbe1.7timesaslikelyasnon-smokerstodevelopcardiovasculardiseaseCardiovasculardisease170.321.7表4-4吸煙者與非吸煙者死于不同疾病的RRRRofdeathfromdifferentdiseasesinsmokerandnonsmoker

LungcancerDisease296.7550.12Smoker4.69Non-smoker10.7RR（1/10萬人年）

表4-5RR與關聯強度

RRandstrengthenofassociation

很強verystrengthen10~＜0.1

強Strengthen3.0~9.90.1~0.3

中Middle

1.5~2.90.4~0.6弱Weak

1.2~1.40.7~0.8

無Noassociation1.0~1.10.9~1.0關聯強度RRstrengthenofassociationRR的95%CI

RR95%CI

反自然對數即為RR95%CI

Theanti-logarithmoflnisRR95%CI

Woolf法()dcbaRRVar1111+++=ln()RRVarRRln96.1±lnln歸因危險度(AttributableRisk,AR)

意義implicationE與ē人群比較，所增加的疾病發生數量

TheincreasedNo.ofdiseasescomparinganexposedpopulationandanunexposedpopulation

AR值暴露因素消除后所減少的疾病數量

Eliminatingtheexposure,thereducedNo.ofdiseasesca或()1000-=-×=RRIIIRRAR010nnIIARe-=-=

意義RR吸煙對肺癌的病因學意義較大AR戒煙對心血管疾病的預防作用較大即公共衛生意義較大

表4-6RR與AR的區別ThedifferencebetweenRRandARCardiovasculardiseaseLungcancerDisease1.710.7RR126.43170.32296.7545.434.6950.12ARNon-smoker

Smoker（1/10萬人年）ImplicationRRsmokinghaslargeretiologysignificanceofonlungcancerARstopsmokinghashigherpreventablesignificanceoncardiovasculardiseases,i.e.thepublichealthsignificance歸因危險度百分比AR%

（病因分值

EtiologicfractionEF）意義implication:

暴露人群中的發病或死亡歸因于暴露的部分占全部發病或死亡的百分比

Theproportionofthecasesthattheexposurehadplayedacausalroleinitsdevelopment.

RR

或-%100%0×=eIIIeAR%1001%×-=RRAR人群歸因危險度

(populationattributablerisk,PAR)

意義暴露人群與一般人群比較，所增加的疾病發生率的大小

TheincreaseddiseaseintheexposedcomparingwithgeneralpopulationPAR值暴露因素消除后所減少的疾病數量

PARthereductioninincidenceaftereliminatetheexposure

PAR=It－I0

It：總人群率

rateintotalpopulationIo：非暴露組率

rateinNon-exposedpopulation

人群歸因危險度百分比PAR%

意義implication

PAR占總人群全部發病（或死亡）的百分比

TheproportionofPARincasesor(death)intotalpopulation

或

Pe：總人群的暴露比例

proportionofexposedintotalpopulation劑量反應關系Doseresponserelationship分析方法列出不同暴露水平下的發病率以最低暴露水平組為對照，計算各暴露水平的RR和危險度差(RD)必要時，應對率的變化作率的趨勢性檢驗

AnalysismethodListtheincidenceofdifferentexposedlevelsCalculatetheRRandARofdifferentexposedlevelsusingthelowestasreferenceCarriedouttendencytestifnecessary

結果血清膽固醇水平患冠心病的RR說明存在劑量效應關系表4-740-59歲男子按初始血清膽固醇分組冠心病6年發生情況

血清膽固醇(mmol/L)

人數

病例數

危險度

平均年發病率

RR

AR

<210

454

16

0.0352

0.0059

1.00

0.0000

210-

455

29

0.0637

0.0106

1.81

0.0285

>245

424

51

0.1203

0.0200

3.39

0.0851

合計

1333

96

0.0720

0.0120

選擇偏倚失訪偏倚信息偏倚混雜偏倚第五節偏倚及其控制Biasandcontrol

Selectionbiasfollowupbiasinformationbiasconfoundingbias一、選擇偏倚selectionbias

Subjectsaredifferentwithgeneralpopulationortargetedtotalpopulationinsomeimportantaspect,whichresultintheoutcomeofresearchbias.

研究人群在一些重要因素方面與一般人群或待研究的總體人群存在差異，而導致研究結果的偏倚。

產生原因

Causes

選擇對象的方法不當最初選定參加研究的對象中有人拒絕參加歷史性隊列研究中部分檔案丟失或記錄不全志愿者隊列研究開始時未能發現早期病人等UnsuitableSelectionmethodforsubjectTheselectedsubjectedrefusetoparticipateFilelostoruncompletedrecordinhistoricalcohortstudyVolunteercohortUnidentifiedearlypatientincohort控制

Contr