资源描述:
《problems and issues in adaptive clinical trial design》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、ProblemsandissuesinadaptiveclinicaltrialdesignGordonLanStatisticalResearchCenterPfizerInc.ASANJ-ChapterSeminarPiscataway,NJJune4,2002Outline:1.Information,samplesize,trendofthedataandconditionalpower(MaxHalperin)2.Re-estimationofsamplesize3.Over-rulingofasequentialboundary29Ada
2、ptive(Flexible)designsModificationofthedesignofanexperimentbasedonaccrueddatahasbeeninpracticeforhundreds,ifnotthousands,ofyearsinmedicalresearch.Inthepast,wehaveatendencytoadoptstatisticalproceduresavailableintheliteratureandapplythemdirectlytothedesignofclinicaltrials.However
3、,sincethoseprocedureswerenotmotivatedbyclinicaltrialpractice,theymaynotbethebesttoolstohandlecertainsituations.Themajorconcernaboutunblindingduringaninterimanalysisisthepotentialbiasintroducedbyachangeinclinicalpracticeresultingfromfeedbackfromtheanalysis.Topreservethecredibili
4、tyofthestudy,resultsoftheunblindedanalysisshouldnotbemadeavailabletoanyonedirectlyinvolvedinmanagingthestudy.DSMB/FDA29Howdowedesignaclinicaltrialforanewtreatment?H0:Newtreatment=StandardtreatmentHa1:Newtreatmentis“betterthan”thestandardtreatmentHa2:Newtreatmentis“worsethan”the
5、standardtreatmentá1=P(AcceptHa1
6、H0)á2=P(AcceptHa2
7、H0)á=á1+á2Fixedversussequentialdesign.(1)á1=0.025,á2=0.025.(2)á1=0.025,á2=0.2.Incase(2),á=0.025+0.2=0.225.Isthestudycredible?Example:Z=2.14,one-sidedp-value=0.016.Doubledone-sidedp-value=2x0.016=0.032.Two-sidedp-value=?29*0.025*
8、****0.02529Onesample,determinationofsamplesizeForagivenvalueofm,solvem=0.50.20.10.050.01®0N=36225900360090000®¥29Intheone-samplecasewithXN(m,1),wesolveforNfromIntheone-samplecasewithXN(m,s2),wesolveforNfromForcomparisonsoftwomeans,wesolveNfromForcomparisonsoftwosurvivaldistribu
9、tionsusingthelogranktest,wesolveD(expectednumberofevents)fromWewillgiveabriefintroductiontothestatisticalproceduresforclinicaltrialdesignandinterimdataanalysisfortheone-samplecasewithXN(m,1).Thesameidea(withminimalmodifications)appliestothetwo-samplecomparisons.29The“trend”ofth
10、edataZ(1),Z(2),…….Astochasticprocess.GivenZt,EC[Z1]=?V