Meta分析的常见偏倚-周旭毓

216《》TheJuonral。rEvideneeBsaedMedieine,2002,2(4):216220·、·Meta`’2(1.;2.,510080)【】Meta,Meta。Meta、,。Meta;;;R195.lA1675144(2002)04021605BiasinmetaanalysisZhouXuyu,FangJiqian1.Medieall〕ramrionritute,2.DePartmentofMediealStaristicsadnEPidemiolo、SunaYtSennUivesriyt,Cualhou10080,P.R.ChinaIAbstract]Metaanalysi。。houdb。,iewed。。。。。bsevration。tudyofth。。videnee.Bia,。fte。。ause,。onnietingersultsOfthemetaanalysisandtheratensitsinternalvalidityandreliability.Thereaeratleasttheretypesofbiasinvolvedinme*aanalysis:samplingbias,seleetionbiasandwithinstudybias.Inthisartiele,theeausesandeonsequeneesofbiaswouldbeintordueedbriefly,andthemethodsofdeteetingandeorreetingofrpublieationbiaswouldbediseussed.KeyWordsmetaanalysis;bias;publieationbias:funnelplotMeta,Meta,。、、。Meta(samplingbias)、(seleetionbias)(withinstudybiases)。,。1.1Meta:(1%9),,,,,Meta;,,,。*,,74(510080)。eTl:87330445,87330805Email:[e-mail]。,,:l)(publieati。。bias);2)(idnexibas);3)(Searchibas);4)(refeercneibas)(Citationbias);5)(Englishlanguagebias);6)、“”(fugitiveliterature)(biasinprovision。fdata);7)2(multiplyusedSubjeetSbias)(multiplepublieationbias)。、、,.Meta217,Meta。,MetaMeta。,,,。,,(、、),。2。1.2Meta,(inelusionCirteabias)(Seleetorbias)。,、、、、、、,。Meta,。,1992,LeizoorviezNumrhoamde《》(BM)J《》(aLcnet)Meta,1984-x991,【2”。Leizorovie::Meta,;NumrhoamedMeta。Egge:,`41。Numrhoamed、,,,。eLiozorvicz,,,。,EggerNumrhoamde。,,。,、,Meta。,,,、、、。。,Meta。Meta,;,Meta(orbusr)。1.3Meta,:1)Meta(extareto:bias);2)、(ibasinSeoirngstudyquality);3)Meta(eropirtgnibas),,,。,Meta,、、。,,,。。,。“”。2,。“”“”(null)。,、。,。1991Easethtorok19841987285,154281,.Meta,,,Meta,,,,,,,,。()。1,,;,。,,,。、,。,,,。,。93,60.4%,13145,34.3%。,(、、),(OR=2.32,95%Cl1.254.28)〔5’。:1),,;2),,。,;3);;4),,;5)。Meta,Meta,Meta,,,Meta。,iSmesMeta,Meta(1.16,95%cI1.061.27);(IntenrationalCaneerResearehDataBank)()Meta(1.06,95%Cl0.971.15)`6,。,。、。。2.,(funnelplots,funnelgarphs),Lighi1984`,’。,。1-..。:{`.0.060.020.020.060.060.020.020.06;:r,ans(z99)s`a:e,25:3912.21997,Egge:,,“`。:Meta(standardonmraldeviate,sND)(preeision),k,t、:`i,sND。/、,=1,,SND,SD=abx。,x0,,sND,y。,。,SND。,Meta,.Meta219,,aO,,,。,a,a,。2,,,。,a95%CI,aO,,。。,。t`。`Metai,*it,r(t`i)/,_(tj/)(1/)l“`.。t`。t`’:`。(t*i,,K(K1)/2。`vj(ij),。。`,Q,12345678910---...123456789107=—.,5)2;:e。112.(2002).statd,20:62,Egger8Meta。4Meta,a.90%CIO,0.1,Meta。4,Meta,3Metaa90%:CIo,0.1,Meta。19931996《JAMA》、《BMJ》、《LANCET》、《Annals。fxntemalMedieine》4Meta,38%Meta,Meta,。CcohraneMeta13%{’。,Egger,,。,。2.3Begg1994Kendall,5tau’9’,Z1.96,0.05,;z1.96,0.05,。eBgg,,,、。,。2.4(failsafenumber)Rosenthal1979,“0]。Meta,,()Meta。!P,2,…,、MetaK,,,、+:,…,+、N。Z`Z(P`),i=],2,…,K+N,s:Z`,u=,Z`S+NK。+N〕Z(0.05)1.645Meta0.05,。,N,U0。,Meta(0.05)()K,220,MetaN(,“)“=(S/1.645)2K0.01,、=(S/2.33)2K。,Z,。,,。,,,,〔。:P,,,。DaerBgegMeta,,{`“,。lGseerLJ,KN+K,g(,P,l。,0H,p(),、,。iGvenGHiGbbS,,(missingstudies),Meta“”`’41。Meta。199212,(EPA)31MetaA,Meta,。iGvensGH,Meat,1010。Meta,30%。iGvenS,。。Meta,、、、,。,。Cochrane,,,,。,()。,,,。FelsonD.Biasinmetaanalytiereseacrh[J].JClinEpidemiol,1992,45:885892IeizoroviezA,HaughMC,ChapuisFR,etal.Lowmoleeularweighthepairninpreventionofpeiroperativethrombosi,[J].BMJ,1992,305:913920NumrohamedMT,RosendaalFR,BuellerHR.wmoleeulaweighrhepairnvesrusstandadrhePanningeneralandorthoPaediesue:ametaanalysis【J」.neet,1992,340:162156EggerM,SmithDG.Metaanalysis:BiasinloeationandseleetionofstudiestJ].BMJ,1998,316:6166EasrerbrookPJ,BerlinJA,oGPalanR,etal.Publieationbiasineliniealreasecrh【J].neet,1991,337:867872SimesJRPublieationbias:heeaseofranintemationaligstryoftrials[J].JClinoneol,1986,15291541LightRJ,PillemerDB.Quantilativeporeeduers.InSumminguP:TheSeieneeofReviewingReseareh【MI.Cambirdy,MA:HavradrUniversityPress,1984EggerM,SmithGD,SehneidetM,etal.Biasinmetaanalysisdeteetedbyasimple,phiealtestIJI.BMJ,1997,315:629-634BeggCB,MazumdarM.operatingeharaeteirstiesofarankcorrelationtestforpublicationbias[J].Biometires,1994,50:10881099RosenthalRl’he“6ledrawerporblem”andtolearneeofrnullresultsIJ].psyeholBull,1979,86:638642ThorntonA,IJeeP.Publieationbiasinmetaanalysis:itseausesandeonsequenees[J].JClinEpidemiol,2000,53:207216DearKBG,BeggCB.AnapproaehforassessingPublieationbiaspiror10performingametaanalysis[J].StatSei,1992,7:237-245GleserIJ,01kin,1.ModelforestimatingthenumberofunpublishedstudiesIJ].StatMed,1996,15:24932507GivenGF,SmithDD,wTeedieRL.Publieationbiasinmeta-analysis:aBayesiandataauglnentationaPProaehtoaeeountforissuesexempliifedinthepassivesmokingdebate【J].tSatcSi,1997.12:221250lIll0[]]l4l13[]]