期中分析的条件把握度及样本含量再估计

*:(BK2008449):,Emai:[e-mail]*(210029)陈建平魏永越陈峰于浩!,B,,,,,,,;,,,,!B(interimanalysis),,,,,,∀#,,,(reestimate),,,,,(dataandsafetymonitoringboard,DSMB)∃1%1.(conditionalpower,CP),21,,2!,1∀,N,N/2,(1)#(informationtime),,n,n/2,:#=n/N(1),0&#&1,#=01,1#(2)BzN:zN=42/N=N42N(2)SN=N,VN=Var(SN)=N2Var()=42N,:zN=(Sn+SN-Sn)/42N=Sn/42N+(SN-Sn)/42N(3),Sn/42N=n42N=42/nnN=z(#)#(4)z(#)(#)Lan&Zucker(1993)∃2%Lan&Wittes(1988)∃3%Sn/42N(Brownianmotion)B(#)=z(#)#,:Var{B(#)}=#Var{z(#)}=#(5)B:Cov{B(#i),B(#j)}=Cov{Sni/42N,Snj/42N}=#i(6)#i#j,#i&#j(3)∃4,5%,B(1)-B(#),E{B(1)}-E{B(#)}=(1#)∃∋361∋20108274Var{B(1)-B(#)}=Var{B(1)}+Var{B(#)}-2Cov{B(1),B(#)}=1-#B(1)-B(#)~N{(1-#)∃,1-#}B(#)B,B(1)-B(#)BB(1)~N{B(#)+(1-#)∃,1-#}(7),∃,∃:∃^=z^(N)=42/N=42/nNn=z##=B##(8)1,B(1)()H0B(1),(7),B(1)z1-!*/2:CP=1-%z1-!*/2-[B(#)+(1-#)∃]1-#(9)z1-!*/2!2.(8)(9),:CP=1-%z1-!*/2-[z(#)#+(1-#)z(#)/#]1-#=1-%z1-!*/2-42/n/nN1-nN(10),N(10),::N,n,5000,n+1,power;T,:,CP&power,;CP(power,;:Nmax=5000,Nmin=n+1N)=(Nmax+Nmin)/2,CP|CP-power|&T,,N),;:CPpower,,Nmax=N);CPpower,,Nmin=N),2()(trospiumchloride,TC,l20mgbid)(detrusorinstability,DI)∃6%(ml),208,10426,373m,l,10526ml,80%#=52/208=025,z#=|X1#-X2#|(1n1+1n2)2=373(126+126)∗105262=1278:B#=z##=0639,∃^=B##=0639025=2556(9):CP=1-%2178-[0639+(1-025)∗2556](1-025)=1-%{-0436}=0669,Pocockz1-!*/22178,∋362∋ChineseJournalofHealthStatistics,Aug2010,Vo.l27,No.4()669%,(669%)(80%),,,80%,53~5000,01%,274,,222,80%1.,,,,,,2.BB,B,,,,,3.,B,,!(PocockPetoOF)!(LanandDeMets)∃7,8%,Pocock4.,,,,,;,,,,,,;(,5000),;,,5.,1:1k:1,TheConditionalPowerofInterimAnalysisandSampleSizeReestimationChenJianping,WeiYongyue,ChenFeng,etal.,DepartmentofEpidemiology&Biostatistics,NanjingMedicalUniversity(210029),NanjingAbstract!ObjectiveTointroducetheprincipleofconditionalpowerofinterimanalysisandthemethodofreestimatingsamplesizeinresponsetothedatabeingcollectedinclinicaltrials.MethodsTheconditionalpowerofrejectinganullhypotheses,definedoverinformationtime,isaBrownianmotion,andcanbecalculatedasaconditionalprobabilityundertheBrownianmotionmode.lThesamplesizerequiredinthesecondstagecanalsobereestimatedbasedontheconditionalpower.Themethodwasillustratedbyarealnormallydistributeddata.ResultsIfwehavelimitedinformationabouttheunderlyingparametersbeforeclinicaltria,lthecalculatedsamplesizewouldbeinaccurate.Wecanusethedataintheexistingstudy,toreassesstheparametersand,ofcourse,thesamplesizerequiredinnextstage.ConclusionThemethodintroducedinthepapercanguaranteetheexpectedpoweroftestbyreestimatingsamplesizebasedontheconditionalpower.Keywords!Clinicaltria;lInterimanalysis;ConditionalPower;Samplesize;Bvalue1.SlutskyAS,LaveryJV.Datasafetyandmonitoringboards.NEnglJMed,2004,350:11431147.2.LanKK,ZuckerDM.Sequentialmonitoringofclinicaltrials:TheroleofinformationandBrownianmotion.StatMed,1993,12:753765.3.LanKK,WittesJ.TheBvalue:Atoolformonitoringdata.Biometrics,1988,44:579585.4.ProschanMA,LanKK,WittesJT.StatisticalMonitoringofClinicalTrials:AUnifiedApproach(StatisticsforBiologyandHealth).NewYork,N.Y.:Springer,2006:4366.5.MoyeLA.StatisticalMonitoringofClinicalTrials:FundamentalsforInvestigators.NewYork,N.Y.:Springer,2006:143168.6.CardozoL,ChappleCR,ToozsHobsonP,eta.lEfficacyoftrospiumchlorideinpatientswithdetrusorinstability:aplacebocontrolled,randomized,doubleblind,multicentreclinicaltria.lBJUInternationa,l2000,85:659664.7.,.:.,2000,17(2):111113.8.LanKK,DemetsDL.Discretesequentialboundariesforclinicaltrials.Biometrika,1983,70:659663.∋363∋20108274