Meta分析的统计学基础专家讲座.pptx

Meta－分析统计学基础一、概述什么是循证医学Evidence-basedmedicine(EBM)循证医学是在对个体病人制订临床决议时应基于当前最正确科学研究结果。循证医学是最正确证据、临床经验和病人价值有机结合。用X线进行乳腺癌筛查MammographyforbreastcancerisanestablishedscreeningmethodTheauthorsfoundthatnotrialdatawereofhighqualityTwowereofmediumquality,andtherestwerepoorqualityorflawed.Whentheresultsofthetwomediumqualitytrialswerecombined,theriskratiowas1.00(95%CI0.96,1.05)Theyconcludedthat“screeningforbreastcancerwithmammographyisunjustified”TheUSPreventiveServicesTaskForcereviewedthesamesetoftrialsThemeta-analysisperformedfortheUSPSTFonthemostcurrentpublisheddatafoundthatthepooledeffectsizeofthecombinedtrialswassizableandstatisticallysignificant:thesummaryrelativerisk(RR)ofbreastcancerdeathamongwomenrandomizedtoscreeninginseventrialsthatincludedwomenolderthan50was0.77(95percentCI,0.67-0.89).TheUSPSTFrecommendsscreeningmammography,withorwithoutclinicalbreastexamination,every1-2yearsforwomenaged40andolder.被动吸烟危害Hackshawetal.conductedaverycomprehensivesystematicreviewin1997什么是循证医学循证方法两个关键方面循证医学证据分级系统性综述循证医学与传统医疗实践四个主要区分(1992JAMA)Meta-analysisSystematicreviews/meta-analysesindexedinPubMed–10yearsHowtoreadasystematicreview?EvaluationofqualityofprimarystudiessetssystematicreviewsapartfromtraditionalreviewsEmpiricresearchshowsthatnotallSRsassessstudyquality:240SRsfromjournals:48%assessedquality(Moher1999)480SRsinDARE:52%assessedquality(Petticrew1999)50SRsonasthma:28%reportedvalidityassessmentcriteria(Jadad)Berkeley,onqualityofSRsonHIV,publishedduringQualityassessmentdonein56%ofreviewsTestingforheterogeneitydonein56%ofreviewsNotallSRswithsignificantheterogeneityexploredreasonsforitManyreviewsdidnotstatetherationaleforchoiceofmodelsusedforcombiningdataOnly12%ofreviewsevaluatedpublicationbias近年Meta分析文件中存在主要问题（到5月）Meta分析流程图异质性检验执行情况发表偏倚说明及评定所以当你找到所需系统性综述时，你有责任去评价它质量。系统性综述与Meta-analyses质量评价标准二、Metaanalysis原理和基本思想在用样本信息推断总体参数时，是存在抽样误差，而且抽样误差大小与样本量大小相关。统计学用抽样分布理论来描述样本统计量改变规律。从一个均数为1.5，标准差为0.7正态总体中进行随机抽样，样本量分别为20，50，100，200，300，500，1000，不一样样本量均进行20次抽样，共得到140个样本。分别计算每个样本均数，标准差和标准误。以样本均数为横