临床研究中的统计学简介.pptx

Introduction of Biomedical Statistics in Clinical ResearchJiang Yong MD, MPH, PhDClinical Trial Research Center,Beijing Tiantan Hospital, Capital Medical UniversityOutlineStudy designStudy design methodsClinical trialSamplingBias and ErrorReliability and ValidityData TypesIntroduction of Biostatistical MethodsStatistical significanceAppropriate statistical tests and result interpretationHow to work with statistician Study designStrength of Study Design: The Evidence PyramidFind the highest strength of evidence available to answer your clinical question.Study DesignSelecting the “Best” study designFactors that must be considered Resources: $ cost-effectiveMan power TimeEthnic problemsPatient burdenClinical trial -basic principalsRandomizationRandom sampling – representative, generalizationRandom grouping – comparability ControlBlindingRandomization methodAim: to ensure the comparability of each groupsMaximize /wiki/Statistical_powerstatistical /wiki/Statistical_powerpowerMinimize selection biasMinimize allocation bias (or confounding)Methods:Tossing a coinRandom number tableComputerized random number generator (SAS/Excel/…)Using blocking and/or stratified randomization to avoid imbalanceControlAim : To ensure the comparabilityClassification according to designConcurrent randomized controlBefore-after control / cross-over controlMatching controlNon-randomized controlHistorical controlClassification according to intervention measuresPlacebo controlEffective controlBlindingMasking the identity of the assigned interventionsAim: to avoid potential bias caused by conscious or subconscious factorsClassification:Single blind: patientDouble blind: patient + investigatorTriple blind: patient + investigator + statisticianHow to Blind PatientsPlacebo pill of same size, color, shape as treatmentsham operation sham device such as sham acupunctureCan Blinding Always be Done?In some studies it may be impossible (or unethical) to blinda treatment may have characteristic side effects