1
系統性文獻回顧及統合分析介紹 系統性文獻回顧及統合分析介紹 系統性文獻回顧及統合分析介紹 系統性文獻回顧及統合分析介紹
高雄榮民總醫院 高雄榮民總醫院高雄榮民總醫院 高雄榮民總醫院 高雄榮民總醫院 高雄榮民總醫院 高雄榮民總醫院
高雄榮民總醫院 藥劑部藥劑部藥劑部藥劑部藥劑部藥劑部藥劑部藥劑部 林佩津總藥師 林佩津總藥師 林佩津總藥師 林佩津總藥師 林佩津總藥師 林佩津總藥師林佩津總藥師 林佩津總藥師
Outline
系統性文獻回顧及統合分析簡介
Steps for systematic reviewSteps for systematic review
Data analysis, present and interpret resultsData analysis, present and interpret results
系統性文獻回顧及統合分析之臨床重要性
系統性文獻回顧及統合分析之臨床應用分享
課程目標:
讓學員了解系統性文獻回顧及統合分析之差異
讓學員了解臨床上如何利用整合分析研究
3
Number of publications about meta
Number of publications about meta - - analysis analysis 、 、 、 、 、 、 、 、 systematic review or evidence
systematic review or evidence - - based medicine based medicine
text word “text word “systematic reviewsystematic review””
medical subject (MESH) heading “medical subject (MESH) heading “evidence-evidence-based medicine”based medicine” and “and “metameta--analysisanalysis as topicas topic””
publication type as “publication type as “metameta--analysisanalysis””
Lin PC et al. FIP 2008,
Lin PC et al. FIP 2008, PIPI--PP--022022
50 1050 2050 3050 4050 5050 6050 7050 8050
before 1988 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
4
系統性文獻回顧及統合分析簡介 系統性文獻回顧及統合分析簡介 系統性文獻回顧及統合分析簡介 系統性文獻回顧及統合分析簡介 (systematic review and
meta-analysis)
5
What is meta
What is meta - - analysis? analysis?
Systematic reviewSystematic review
a clearly stated set of a clearly stated set of objectivesobjectives with with prepre--defined eligibility defined eligibility criteria for studies
criteria for studies
an an explicit, explicit, reproduciblereproducible methodologymethodology
a systematica systematic searchsearch that attempts to identify that attempts to identify allall studiesstudies that that would meet the eligibility criteria
would meet the eligibility criteria
an an assessment of the assessment of the validityvalidity of the findings of the included of the findings of the included studies, for example through the assessment of risk of bias studies, for example through the assessment of risk of bias
a systematica systematic presentation, and synthesispresentation, and synthesis, of the , of the characteristics and findings of the included studies characteristics and findings of the included studies
MetaMeta--analysisanalysis
statistical procedurestatistical procedure for combining data from independent for combining data from independent studies
studies
Cochrane Handbook for Systematic Reviews of Interventions
Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March Version 5.1.0 [updated March 2011]. The Cochrane Collaboration, 2011. Available from www.coch
2011]. The Cochrane Collaboration, 2011. Available from www.cochranerane--handbook.org. handbook.org.
Ann Intern Med. 2011;155:839-847.6
7
2010/11/18
7
How to conduct a meta
How to conduct a meta - - analysis analysis
1996: QUOR1996: QUOROOMM
quality of reporting meta-quality of reporting meta-analysisanalysis
2009/7: PRISMA Ann Intern Med 2009; 151(4):2642009/7: PRISMA Ann Intern Med 2009; 151(4):264--99
Preferred reporting items for systematic reviews and meta-Preferred reporting items for systematic reviews and meta- analysis
analysis
Avoiding publication bias, language bias or outcome Avoiding publication bias, language bias or outcome measure etc.
measure etc.
http://www.prisma-http://www.prisma-statement.org/ statement.org/
7
2011/9/30
2010/11/18
8
QUOR QUOR O O M M
Lancet 1999; 354:1896-900 Potentially relevant RCTs
Potentially relevant RCTs identified identified and screened for retrieval (n=
and screened for retrieval (n= ------))
RCTs retrieved for
RCTs retrieved for more detailed more detailed evaluation
evaluation (n= ---(n= ---))
Potentially appropriate RCTs
Potentially appropriate RCTs to be to be included in the meta
included in the meta--analysis (n= analysis (n= ----))
RCTs included
RCTs included in metain meta-analysis -analysis (n=
(n= ------))
With usable information With usable information by by
outcome (n=
outcome (n= ------))
RCTs excluded, with RCTs excluded, with
reasons (n=
reasons (n= ------))
RCTs excluded, with RCTs excluded, with
reasons (n=
reasons (n= ------))
RCTs excluded from meta RCTs excluded from meta-- analysis, with reasons (n=
analysis, with reasons (n= ------))
RCTs withdrawn, by outcome, RCTs withdrawn, by outcome,
with reasons (n=
with reasons (n= ------))
Progress through the Progress through the stages of a meta
stages of a meta-- analysis for RCTs analysis for RCTs
2011/9/30 9
2010/11/18
9
2010/09/04
9
10
2010/11/18
10
2010/09/04
10
11
How How to conduct a meta to conduct a meta - - analysis analysis
Steps for systematic review
Steps for systematic review
Formulate review questionFormulate review question
Define inclusion and exclusion criteriaDefine inclusion and exclusion criteria
Locate studiesLocate studies
Select studiesSelect studies
Extract dataExtract data
Assess study qualityAssess study quality
Data analysis, present and interpret results
Data analysis, present and interpret results
Egger M et al.
Egger M et al. Systematic Reviews in Health Care: MetaSystematic Reviews in Health Care: Meta--analysis in analysis in context. BMJ, 2rd edition 2005 context. BMJ, 2rd edition 2005
12
How How to conduct a meta to conduct a meta - - analysis analysis
Steps for systematic reviewSteps for systematic review
Formulate review questionFormulate review question
Define inclusion and exclusion criteriaDefine inclusion and exclusion criteria
Locate studiesLocate studies
Search strategy (reproducible)Search strategy (reproducible)
Key words Key words
MeSH termsMeSH terms
Databases Databases
PubMed or MedlinePubMed or Medline
EMBASEEMBASE
CINHALCINHAL
PEDroPEDro
Conference abstract (dependent)Conference abstract (dependent)
Reference lists of the identified articlesReference lists of the identified articles
Hand searchHand search
Registered centersRegistered centers
13
RCT RCT 註冊平台 註冊平台 註冊平台 註冊平台 註冊平台 註冊平台 註冊平台 註冊平台
International Clinical Trials Registry PlatformInternational Clinical Trials Registry Platform (ICTRP)
(ICTRP) http://www.who.int/ictrp/en/http://www.who.int/ictrp/en/
世界衛生組織下設的臨床試驗註冊平台，其設置的重要任務在使臨 世界衛生組織下設的臨床試驗註冊平台，其設置的重要任務在使臨 床試驗的進行更為透明化，並提供各權益相關人所需的有關資訊，
床試驗的進行更為透明化，並提供各權益相關人所需的有關資訊，
更重要是減少性質相近或相同的試驗重複投進行的浪費 更重要是減少性質相近或相同的試驗重複投進行的浪費
ClinicalTrials.govClinicalTrials.gov http://www.clinicaltrials.gov/http://www.clinicaltrials.gov/
由美國國家衛生研究院下之國家醫學圖書館所設，專門收錄由衛生 由美國國家衛生研究院下之國家醫學圖書館所設，專門收錄由衛生 院、聯邦政府、地方政府及民間私立機構進行中的臨床試驗登錄資 院、聯邦政府、地方政府及民間私立機構進行中的臨床試驗登錄資 料料
James Lind Library http://www.jameslindlibrary.org/James Lind Library http://www.jameslindlibrary.org/
James Lind Library
James Lind Library 係由係由CochraneCochrane Collaboration發起人之一的英國Collaboration發起人之一的英國 爵士爵士Iain Chalmer教授所率領的編輯團隊建置，此一圖書資料建置Iain Chalmer教授所率領的編輯團隊建置，此一圖書資料建置 主要的目的在呼籲重視臨床治療的公平試驗
主要的目的在呼籲重視臨床治療的公平試驗(fair test)(fair test)原則，以確保原則，以確保 治療的真正成效，圖書資料有包括對公平試驗的解釋定義及研究設 治療的真正成效，圖書資料有包括對公平試驗的解釋定義及研究設 計等，並已有簡體中文版的翻譯文章可供閱讀，並收錄大部分的參 計等，並已有簡體中文版的翻譯文章可供閱讀，並收錄大部分的參 考文獻免費供線上瀏覽，值得有興趣者參考
考文獻免費供線上瀏覽，值得有興趣者參考
國衛院實證臨床指引平台
14
RCT RCT 註冊平台 註冊平台 註冊平台 註冊平台 註冊平台 註冊平台 註冊平台 註冊平台
The Cochrane Central Register of Controlled Trials (Clinical The Cochrane Central Register of Controlled Trials (Clinical Trials
Trials)) http://www.mrw.interscience.wiley.com/cochrane/cochhttp://www.mrw.interscience.wiley.com/cochrane/coch rane_clcentral_articles_fs.html
rane_clcentral_articles_fs.html
簡稱簡稱CENTRAL之臨床試驗註冊資料庫，收錄詳細的臨床試驗主CENTRAL之臨床試驗註冊資料庫，收錄詳細的臨床試驗主 題、文獻書目、摘要，不含全文，資料來源包含發表於
題、文獻書目、摘要，不含全文，資料來源包含發表於 MEDLINE
MEDLINE、、EMBASEEMBASE及其他期刊等資料庫，以及未發表的參考及其他期刊等資料庫，以及未發表的參考 文獻來源如僅於研討會發表之文獻報告，約
文獻來源如僅於研討會發表之文獻報告，約3/5的資料由3/5的資料由 MEDLINE
MEDLINE取得，其他則由包括各取得，其他則由包括各Cochrane Review Groups依其Cochrane Review Groups依其 主要興趣領域所收集的臨床試驗文獻，整個收到的資料再由
主要興趣領域所收集的臨床試驗文獻，整個收到的資料再由US US Cochrane Center
Cochrane Center作統整再納入作統整再納入Cochrane LibraryCochrane Library
The Society for Clinical TrialsThe Society for Clinical Trials http://www.sctweb.org/http://www.sctweb.org/
與與Clinical Trials相關之研究發展、設計之國際組織Clinical Trials相關之研究發展、設計之國際組織; 亦提供臨床; 亦提供臨床 試驗摘要查詢功能
試驗摘要查詢功能
聯合人體試驗委員會聯合人體試驗委員會 http://www.jirb.org.tw/Default.aspxhttp://www.jirb.org.tw/Default.aspx
聯合人體試驗委員會為由北榮、台大、長庚、三總、成大等五 聯合人體試驗委員會為由北榮、台大、長庚、三總、成大等五 家醫學中心共同成立目的為加速國內臨床試驗審查速度及提升 家醫學中心共同成立目的為加速國內臨床試驗審查速度及提升 臨床試驗水準。該委員會網頁上有許多臨床試驗及
臨床試驗水準。該委員會網頁上有許多臨床試驗及IRB審查的資IRB審查的資 訊或訓練課程講義可供下載參考
訊或訓練課程講義可供下載參考 國衛院實證臨床指引平台
16
How How to conduct a meta to conduct a meta - - analysis analysis
Steps for systematic review
Steps for systematic review
Formulate review questionFormulate review question
Define inclusion and exclusion criteriaDefine inclusion and exclusion criteria
Locate studiesLocate studies
Select studiesSelect studies
At least 2 reviewers independentlyAt least 2 reviewers independently evaluated each evaluated each
identified study and abstracted relevant characteristics identified study and abstracted relevant characteristics
Contact authorsContact authors for the unpublished data for the unpublished data
17
How How to conduct a meta to conduct a meta - - analysis analysis
Steps for systematic reviewSteps for systematic review
Formulate review questionFormulate review question
Define inclusion and exclusion criteriaDefine inclusion and exclusion criteria
Locate studiesLocate studies
Select studiesSelect studies
Extract dataExtract data
Assess study qualityAssess study quality
Prepare tables comparing studies with respect to:Prepare tables comparing studies with respect to:
Authors, YearAuthors, Year
Patients (setting)Patients (setting)
InterventionIntervention
ComparisonComparison
Outcome (results)Outcome (results)
Study qualityStudy quality
2011/9/30 18
Quality
Quality of included studies of included studies
Scale (score) approachScale (score) approach
Jadad scaleJadad scale
Chalmers TCChalmers TC
------
Components approachComponents approach
Scales or components?Scales or components?
Health Technol Assess. 1999;3(12):i-iv, 1-98
21
2010/11/18
21
22
2010/11/18
22
Study quality
Study quality – – Jadad score Jadad score
Modified from Controlled Clin Trials. 1996; 17:1-12
23
BMJ. 2008 Jun 14;336(7657):1359-61
24
Aliment Pharmacol Ther 2002; 16: 1623–1632.
How How to conduct a meta to conduct a meta - - analysis analysis
Steps for systematic review
Steps for systematic review
Formulate review questionFormulate review question
Define inclusion and exclusion criteriaDefine inclusion and exclusion criteria
Locate studiesLocate studies
Select studiesSelect studies
Extract dataExtract data
Assess study qualityAssess study quality
Data analysis, present and interpret results
Data analysis, present and interpret results
How How to conduct a meta to conduct a meta - - analysis analysis
Data analysis, present and interpret results
Data analysis, present and interpret results
Systematic review or meta-Systematic review or meta-analysisanalysis
Moher D et al. Arch Pediatr Adolesc Med 1998;152:915-20
How How to conduct a meta to conduct a meta - - analysis analysis
Data analysis, present and interpret results
Data analysis, present and interpret results
Systematic review or metaSystematic review or meta--analysisanalysis
Treatment effect (effect size)Treatment effect (effect size)
Continuous data: standardized (differences in means)Continuous data: standardized (differences in means)
Dichotomous dataDichotomous data
Difference scale: risk differenceDifference scale: risk difference
Ratio scale: odds ratio, risk ratio (hazard ratio)Ratio scale: odds ratio, risk ratio (hazard ratio)
Point estimate (forest plot) or cumulative forest plotPoint estimate (forest plot) or cumulative forest plot
28
Calculation of RR (relative risk, risk ratio)
Experimental event rate (EER) = a/(a+b)
Experimental event rate (EER) = a/(a+b)
實驗組事件比率
實驗組事件比率
（發生率）（發生率）
；治療組風險；治療組風險
Control event rate (CER) = c/(c+d)
Control event rate (CER) = c/(c+d)
對照組事件比率
對照組事件比率
（發生率）（發生率）
；對照組風險；對照組風險
RR= EER/CER= (a/a+b)/(c/c+d) RR= EER/CER= (a/a+b)/(c/c+d)
Treatment Event
Positive Negative Exposed (experimental) a b
Not exposed c d
29
Calculation of OR (odds ratio; relative odds)
Experimental event Odds (EEO)Experimental event Odds (EEO)
實驗組中發生目標疾病的勝算實驗組中發生目標疾病的勝算
EEO= a/bEEO= a/b
Control event Odds (CEO)Control event Odds (CEO)
控制組中發生目標疾病的勝算控制組中發生目標疾病的勝算
CEO= c/dCEO= c/d
OR= EEO/CEO = (a/b)/(c/d)= ad/bcOR= EEO/CEO = (a/b)/(c/d)= ad/bc
Treatment Event
Positive Negative Exposed (experimental) a b
Not exposed c d
Forest plot for summary data (RD)
Crit Care Med 2010; 38:1197
Crit Care Med 2010; 38:1197–1205–1205
Publication Bias in Meta-Analysis – Prevention, Assessment and Adjustments Edited by H.R. Rothstein, A.J. Sutton and M.
Borenstein
2011/9/30
How How to conduct a meta to conduct a meta - - analysis analysis
Data analysis, present and interpret results
Data analysis, present and interpret results
Systematic review or meta-Systematic review or meta-analysisanalysis
Treatment effect (effect size)Treatment effect (effect size)
Continuous data: standardized (differences in means)Continuous data: standardized (differences in means)
Dichotomous dataDichotomous data
Difference scale: risk differenceDifference scale: risk difference
Ratio scale: odds ratio, risk ratio (hazard ratio)Ratio scale: odds ratio, risk ratio (hazard ratio)
Point estimate (forest plot) or cumulative forest plotPoint estimate (forest plot) or cumulative forest plot
Crit Care Med 2010; 38:1197–1205
33
4. Perform meta
4. Perform meta - - analysis analysis
Moher D et al. Arch Pediatr Adolesc Med 1998;152:915-20
34
How How to conduct a meta to conduct a meta - - analysis analysis
Data analysis, present and interpret results
Data analysis, present and interpret results
Systematic review or metaSystematic review or meta--analysisanalysis
Treatment effect (effect size)Treatment effect (effect size)
Software for metaSoftware for meta--analysisanalysis
Comprehensive metaComprehensive meta--analysis, STATA, Metaxis or analysis, STATA, Metaxis or MetaWin
MetaWin
Free software (Review Manager 5)Free software (Review Manager 5)
BMC Medical Research Methodology
BMC Medical Research Methodology 2007, 72007, 7:40:40
35
How How to conduct a meta to conduct a meta - - analysis analysis
Data analysis and present results
Data analysis and present results
Systematic review or metaSystematic review or meta--analysisanalysis
Treatment effect (effect size)Treatment effect (effect size)
Software for metaSoftware for meta--analysisanalysis
HeterogeneityHeterogeneity
VisualVisual
Statistical testsStatistical tests
Chi-Chi-square testsquare test for heterogeneity (Cochran Q test ) for heterogeneity (Cochran Q test )
II22 statisticsstatistics
Explore the potential risks for heterogeneityExplore the potential risks for heterogeneity
Heterogeneity-I
2test
0-40%: might not be important
30-60%: may represent moderate heterogeneity
50-90%: may represent substantial heterogeneity
75-100%: considerable heterogeneity
Fixed effect model vs random effect model
Cochrane Handbook for Systematic Reviews of Interventions. 2009
37
Forest Plot &
Forest Plot &
heterogeneity heterogeneity
analysis analysis
Publication Bias in Meta-Analysis – Prevention, Assessment and Adjustments Edited by H.R. Rothstein, A.J. Sutton and M.
Borenstein
2011/9/30 38
Check for heterogeneity Check for heterogeneity
If significant heterogeneity is found:
If significant heterogeneity is found:
Find out what factors might explain the Find out what factors might explain the heterogeneity
heterogeneity
Can decide not to combine the dataCan decide not to combine the data
If no heterogeneity:
If no heterogeneity:
Can perform metaCan perform meta--analysis and generate a analysis and generate a common, summary effect measure
common, summary effect measure
39
How How to conduct a meta to conduct a meta - - analysis analysis
Data analysis and present results
Data analysis and present results
Systematic review or metaSystematic review or meta--analysisanalysis
Treatment effect (effect size)Treatment effect (effect size)
Software for metaSoftware for meta--analysisanalysis
HeterogeneityHeterogeneity
VisualVisual
Statistical testsStatistical tests
Explore the potential risks for heterogeneityExplore the potential risks for heterogeneity
Sensitivity analysisSensitivity analysis
L’L’Abbe plotAbbe plot
Subgroup (stratified) analysisSubgroup (stratified) analysis
Meta-Meta-regressionregression
Supplementary File 2: Sensitivity analysis - Pooled effect estimates for stress related upper GI bleeding
Drop out study
Risk difference [95% CI], p Heterogeneity-I2
Powell -0.05 [-0.10, 0.00], 0.07 71%
Levy -0.02 [-0.05, 0.01], 0.19 26%
Phillips -0.03 [-0.08, 0.01], 0.15 65%
Azevedo -0.03 [-0.08, 0.02], 0.18 65%
Kantorova -0.06 [-0.12, 0.01], 0.08 72%
Conrad -0.06 [-0.12, 0.00], 0.10 79%
Somberg -0.06 [-0.14, 0.01], 0.05 62%
Crit Care Med 2010; 38:1197
Crit Care Med 2010; 38:1197––1205 online supplement1205 online supplement
L’Abbe plot
Crit Care Med 2010; 38:1197
Crit Care Med 2010; 38:1197––1205 online supplement1205 online supplement
Crit Care Med Crit Care Med 2010; 38:1197 2010; 38:1197–– 1205 online 1205 online supplement supplement
43
How How to conduct a meta to conduct a meta - - analysis analysis
Data analysis and present results
Data analysis and present results
Systematic review or metaSystematic review or meta--analysisanalysis
Treatment effect (effect size)Treatment effect (effect size)
Software for metaSoftware for meta--analysisanalysis
HeterogeneityHeterogeneity
Explore the potential publication biasExplore the potential publication bias
Funnel plotFunnel plot
Begg and Mazumdar’Begg and Mazumdar’s rank correlation tests rank correlation test
Egger’Egger’s regressions regression
Publication Bias in Meta-Analysis – Prevention, Assessment and Adjustments Edited by H.R. Rothstein, A.J. Sutton and M.
Borenstein
BMC Infectious Diseases 2004, 4:6
2011/9/30 46
BMC Medical Research Methodology
BMC Medical Research Methodology 2007, 2007, 77:40:40 BMC Medical Research Methodology
BMC Medical Research Methodology 2007, 72007, 7:40:40
47
系統性文獻回顧及統合分析之臨床重要性 系統性文獻回顧及統合分析之臨床重要性 系統性文獻回顧及統合分析之臨床重要性 系統性文獻回顧及統合分析之臨床重要性
臨床執行的立場
臨床執行的立場
Minimize the potential bias in traditional narrative Minimize the potential bias in traditional narrative reviews
reviews
provide more precise estimates of the effects of provide more precise estimates of the effects of health care from individual studies
health care from individual studies
Clinical practice and decision makingClinical practice and decision making
實證醫學的立場
實證醫學的立場
Level of evidenceLevel of evidence
2011/9/30 48
The importance of meta
The importance of meta - - analysis analysis
49
Circulation 2011, 123:104-12350
2011/9/30 Abstract code: PE-1351
The 4th Asian Conference of Pharmacoepidemiology, 2009
系統性文獻回顧及統合分析之 系統性文獻回顧及統合分析之 系統性文獻回顧及統合分析之 系統性文獻回顧及統合分析之
臨床應用分享
臨床應用分享 臨床應用分享
臨床應用分享
page. 57
Crit Care Med 2010; 38:1197–1205
2011/9/30 59
N Engl J Med 2008;359:1543-54.60
61
Because of the strength of the Because of the strength of the UPLIFT data, the absence of a UPLIFT data, the absence of a
strong signal related to stroke strong signal related to stroke or cardiovascular events with or cardiovascular events with
tiotropium, and
tiotropium, and the potential the potential methodologic limitations of the methodologic limitations of the
Singh meta
Singh meta--analysisanalysis, the FDA , the FDA concluded that
concluded that current data do current data do not support
not support the conclusion that the conclusion that there is an increased risk of there is an increased risk of
stroke, heart attack, or death stroke, heart attack, or death
associated with tiotropium associated with tiotropium
HandiHaler HandiHaler
2011/9/30 62
How do you think about MA How do you think about MA
homogeneous homogeneous
heterogeneity heterogeneity
2011/9/30 63
How do you think about MA How do you think about MA
Why Why
a SR be performeda SR be performed
Disadvantages in traditional narrative reviewsDisadvantages in traditional narrative reviews
Clinical practice and decision makingClinical practice and decision making
Systematic review or meta
Systematic review or meta
--
analysis in clinicalanalysis in clinical practice
practice
Appropriate typeAppropriate type--ofof--study included, good quality study included, good quality
Apply to appropriate patient (group)Apply to appropriate patient (group)
2011/9/30 64
