如何报告系统综述和meta分析

如何报告系统综述和meta分析北京大学循证医学中心詹思延2006-12-14北京(AnnSurg2005;241:450–459)ResultsoftheQUOROM-basedchecklistassessment:scoringthequalityofreviewsontreatmentandpreventionofbreastandcolorectalcancersQualityclassesbytheGRADEworkinggroup„High(A+,A–):Furtherresearchisveryunlikelytochangeourconfidenceintheestimateofeffect.„Moderate(B+,B–):Furtherresearchislikelytohaveanimportantimpactonourconfidenceintheestimateofeffectandmaychangetheestimate.„Low(C+,C–):Furtherresearchisverylikelytohaveanimportantimpactonourconfidenceintheestimateofeffectandislikelytochangetheestimate.„Verylow(D):Anyestimateofeffectisveryuncertain.国内期刊发表的Meta分析存在的主要问题„文献查全率不高„没有列出被排除的试验„病人的特征范围、诊断标准和治疗范围不明确„对资料的可合并性的检验较差„对潜在偏倚的控制和检测不足„统计分析不规范„缺少对原始研究的质量评价„未改变方法进行敏感度分析„缺少对发表偏倚的检测„缺少对结果应用价值的评估(孙福红,魏水易,2002)„由于meta分析是对原各研究结果的统计合成，它不仅不能排除原始研究中存在的偏倚，而且在文献查找和选择过程中，如果处理不当，还会引入新的偏倚，导致合并后的结果歪曲了真实的情况。„在撰写和报告meta分析论文时应当详细描述研究过程，使读者能够评估研究的质量。QUOROM声明„1996年10月，由临床流行病学家、临床学家、统计学家、从事meta分析的研究人员以及来自英国和北美对meta分析感兴趣的编辑共30人，组成专家小组，对RCT的meta分析报告质量进行了方法学的评价，并提出了统一报告格式。Lancet1999;354:1896–900MOOSE„1997年4月，由美国疾病预防控制中心资助，召集了27名来自临床实践、现场干预、统计学、流行病学、社会科学以及生物医学编辑等方面的专家组成专题研究小组，讨论并制定了流行病学中观察性研究的meta分析报告规范。Meta-analysisofObservationalStudiesinEpidemiologyAProposalforReportingDonnaF.Stroup,JesseA.Berlin,SallyC.Morton,etal，fortheMeta-analysisOfObservationalStudiesinEpidemiology(MOOSE)GroupJAMA.2000;283:2008-2012论文报告规范的制定情况1993-94SORTRCT1994AsilomarRCT1995-96CONSORTRCT1996-99QUOROMSR/M-AofRCTs1997-00MOOSEM-Aofobsstudies1999-01CONSORTIIRCT2000-03STARDDiagnostic2003-04TRENDNonRCT/Behavioural2001-05REMARKPrognostic(cancer)2004-??STROBEObservational2005-??QUOROMIISR/M-AofRCTs声明解读„题目：能鉴定出是否为RCT的meta分析或系统综述QUOROM(Circulation.2005;112:1945-1952.)„摘要：使用结构化的格式„目的：明确描述临床问题Theeffectoffishoilonheartrate(HR),amajorriskfactorforsuddendeath,isnotwellestablished.Wecalculatedthiseffectinameta-analysisofrandomized,double-blind,placebo-controlledtrialsinhumans.„资料来源：列出文献数据库和其他信息来源RandomizedtrialsoffishoilthatevaluatedHRwereidentifiedthroughMEDLINE(1966throughJanuary2005),hand-searchingofreferences,andcontactwithinvestigatorsforunpublishedresults.QUOROM„摘要：使用结构化的格式„综述方法：概括研究选择的标准(如对象、干预、结局和研究设计)，详细描述真实性评价、资料提取和数据定量合成的方法，以及研究的特征，使读者能够重复„Twoinvestigatorsindependentlyextractedtrialdata.„Apooledestimatewascalculatedfromrandom-effectsmeta-analysis.Predefinedstratifiedmeta-analysesandmeta-regressionwereusedtoexplorepotentialheterogeneity.QUOROM„摘要：使用结构化的格式„结果：描述纳入与排除的RCT的特征，给出定性、定量的分析结果(例如点估计值及可信区间)及亚组分析结果„Of197identifiedarticles,30metinclusioncriteria.„Intheoverallpooledestimate,fishoildecreasedHRby1.6bpm(95%CI,0.6to2.5;P=0.002)comparedwithplacebo.„FishoilreducedHRby2.5bpm(P0.001)intrialswithbaselineHR≥69bpm(median)buthadlittleeffect(0.04-bpmreduction;P=0.56)intrialswithbaselineHR69bpm(Pforinteraction=0.03).QUOROM„摘要：使用结构化的格式„结论：对主要结果加以论述„Inrandomizedcontrolledtrialsinhumans,fishoilreducesHR,particularlyinthosewithhigherbaselineHRorlongertreatmentduration.Thesefindingsprovidefirmevidencethatfishoilconsumptiondirectlyorindirectlyaffectscardiacelectrophysiologyinhumans.Potentialmechanismssuchaseffectsonthesinusnode,ventricularefficiency,orautonomicfunctiondeservefurtherinvestigation.QUOROM„前言：明确描述临床问题、干预的生物学合理性和meta分析的理由QUOROM“Fattyfishandfishoilintakeisassociatedwithlowerriskofcardiacarrhythmias,including….1–8”“Experimentalstudiesin…,nonhumanprimatessuggestthatfishoilhasdirectcardiacelectrophysiologicaleffects,includingslowingoftheheartrate(HR).9–11However,sucheffectsarenotwellestablishedinhumans.”“BecausehigherHRisamajorindependentriskfactorfor…aneffectoffishoilonHRwouldbothconfirmaninfluenceoncardiacelectrophysiologyinhumansandindicateaplausiblepotentialmechanismforobservedrelations….”“Wethereforeperformedameta-analysisofrandomizedplacebocontrolledclinicaltrialstodeterminetheeffectoffishoilconsumptiononHRinhumans.”„文献检索：详细介绍信息来源（如文献数据库、注册库、个人档案、专家信息、机构、手工检索），对检索的限制（如年代、发表状态、发表语言等）„“identifiedthroughMEDLINE(1966throughFebruary2005),hand-searchingofreferencelistsofobtainedarticles,andcontactinginvestigatorsforunreportedHRdatainpublishedtrialsorforHRdatafromunpublishedtrials.”„选择：描述纳入、排除标准(定义对象、干预、主要结局和研究设计)„English-languagetrialsinhumansubjects18yearsofagewereincludediforalfishoilsupplementationwasrandomizedandchangesinHRorbaselineandfollow-upHRweremeasured;trialswithorgantransplantsubjects,cointerventionsthatcouldnotbeseparatedfromfishoiltreatment,noplacebocontrol,nonblindingofparticipants,orduration2weekswereexcluded.QUOROM方法„真实性评价：描述评价标准和过程(例如设盲的情况、质量评价方法及评价结果)„Studyqualitywasalsoindependentlyassessedby2investigatorsaccordingtothecriteriaforqualityassessmentofrandomizedclinicaltrialsdevelopedbyDelphiconsensus.50The9criteria(1a,1b,and2through8)include...Weassessedthevalidityofdataextractionbycomparingtheindependentlyabstractedresultsforconcordance,andanydiscrepancieswereresolvedbydiscussionandreviewoftheoriginalmanuscriptbythe2investigatorswhoextractedthedataor,ifnecessary,acommitteecomprisingallthenvestigators.QUOROM方法„资料提取：描述提取过程和方法(例如双人平行摘录)„Foreachofthearticlesmeetingincl