临床试验中中心效应的评价及处理方法

CN3421206PR,ISSN100922501http:PP(9):1073-107620042072272004208226,,,,,,:Tel:[phone]Email:[e-mail],,210029,,,Breslow2Day,CMH,logistic,Breslow2Day,,CMH;;Breslow2Day;CMH;logistic;logistic:R51:A:100922501(2004)0921073204,IIIII[1],,,,,,[2,3],,,,,,Breslow2Day;,CMH(Co2chran2Mental2Haenszel)[4]Breslow2Day,CMH;logistic15II,48,24(ITT),(1)15(%)11491079.17241562112.502422571175.00241281316.672432461275.00241473012.502442561170.83241012208.332453561066.67248114120.83242CMH2.1,,Breslow2Day(homogeneity)(degreeoffreedom,DF)-12,SAS[5]Breslow2Day:2=2.1743,DF=4,P=0.70374a,b,c,d,n,jj,:ja:mj=(aj+bj)(aj+cj)nj3701j=(aj+bj)(cj+dj)(aj+cj)(bj+dj)n2j(nj-1)CMH2(Cochran2Mantel2HaenszelStatistics):2CMH=(aj-mj)2j12,CMH2=83.7494,DF=1,P0.001,,,2.2,,Breslow2DayCMH(),,CMH2,12CMH2=89.3660,P0.0001,3LogisticBreslow2Day,,CMH,,logistic[6](y=0)(y=1),,021(),treatcenter,logistic,(2)25(dummyvariables)centercnt1cnt2cnt3cnt4cnt5110000201000300100400010500001lo2gistic,(loglikeli2hoodratio),logistic:logitP=+treatlogistic:logitp=+treat+1cnt1+2cnt2+3cnt3+4cnt44cnt1cnt45SAS,(3)3logisticlogisticWald2PlogisticWald2P-1.01640.207124.07330.0001-1.01160.206424.01430.0001Treat2.82520.335071.12900.00012.81310.333471.20030.0001cnt1-0.13180.32520.16430.6852cnt2-0.13180.32520.16430.6852cnt30.00010.32470.00000.9997cnt40.26330.32530.65550.4181-118.153-118.5472=-2(118.153-118.547)=0.789,df=4,P=0.9399,4,,2=0.789,DF=4,P=0.939,,(Wald2=71.1290,P0.0001),4cnt1cnt4,14514,4,4logistic,,(ordinalresponse),,logistic,logistic[6](ordinallogistic4701ChinJClinPharmacolTher2004Sep;9(9)regression)yk,k1,2,,klogistic:logitPj=logit[P(yj|x)]=-j+pi=1ixi,j=1,2,,k-1:P(yj|x)=11+exp-j+pi=1ixi(cumulativeoddsmod2el),(k-1)+p,ji(j=1,2,k-1,i=1,2,,p)aj(j=1,2,,k-1),k-1;ik=2,logis2tic,logisticlogistic(4)4logisticlogisticWald2PlogisticWald2P1-1.33890.1736259.48500.0001-1.33460.173259.40100.000120.33970.15594.74740.02930.33790.15564.71650.029931.56830.180275.75920.00011.56620.179875.87290.0001Treat-1.34870.146185.23080.0001-1.34160.145884.70360.0001cnt10.18050.24250.55370.4568cnt2-0.05590.24170.05350.8171cnt30.08370.24200.11960.7295cnt4-0.04090.24170.02870.8655-277.938-278.4332=0.991,df=4,P=0.9112:,2=0.991,P=0.9112,5,,,,,(),,,meta(metaanalysis)(weightedaverage),,,[6,7]Logistic,(co2variate),,,(baseline),CMHmeta,,logistic,,,(interac2tion)[8],logistic1.[S].20002.[S].20023ICHE9.StatisticalPrinciplesforclinicaltrials[S].19984CochranWG.Somemethodsofstrengtheningthecommon2tests[J].Biometrics,1954;10(2):417-515ME,Stokes.CategoricaldataanalysisusingtheSASsystem[M].NorthCarolina:SASInstituteInc,1997:39-566SuttonAJ,AbramsKR,JonesDR,SheldonTA,SongF.MethodsforMeta2analysisinmedicalresearch[M].JohnWiley:Chichester,20007SuttonAJ,AbramsKR,JonesDR.Anillustratedguidetothemethodsofmeta2analysis[J].JEvalClinPract,2000;7(2):135-488.[M].:,2001:83-11357012004Sep;9(9)Statisticalapproachesforevaluatinganddealingwithcentereffectsinmulti2centerclinicaltrialYUHao,CHENFengDepartmentofBiostatistics,NanjingMedicalUniversity,Nanjing210029,Jiangsu,ChinaABSTRACTTosearchthestatisticalapproachesforevaluatinganddealingwiththecentereffectsinmulti2centerclinicaltrial.Breslow2Daytestwasusedforevalu2atingthedifferencesofthevirtualratingamongthecenterswhoseresponseswerebinary,Cochran2Mantel2Haenszeltestfordealingwiththecentereffectswhoseresponseswerebinaryorordinal,andlogisticregressionforevaluat2inganddealingwiththecentereffectssimultaneously.TheresultsshowedthatBreslow2Daytestisinvalidinevaluatingthecentereffectswhoseresponseisordinal,andCochran2Mantel2Haenszeltestcannotbeusedfordealingwiththeeffectsofothercovariatesmeanwhile.KEYWORDSmulti2centerclinicaltrial;centeref2fects;Breslow2Daytest;Cochran2Mantel2Haenszeltests;logisticregression;ordinallogisticregression2005,ABCISSN100221523,CN6121397PR,,52253,BM4334,5.00():1(712083):0910231661503159543;E2mail:[e-mail]2005,19986,,,CN4221626PR,,1688,8,96,382325,:2;:430014;:02728277858082835077;E2mail:[e-mail]@periodicals.net.cn();http:PPzgys.periodicals.net.cnP,6701ChinJClinPharmacolTher2004Sep;9(9)