Meta-analysis in Medical Research: The handbook for the understanding and practice of meta-analysis

109 Pages·2005·1.5 MB·English

by  Gioacchino Leandro

P1:FCH/FFX P2:FCH/FFX QC:FCH/FFX T1:FCH
BY035-FM BY035-Leandro-v1 October5,2004 15:39
Meta-analysis
in Medical Research
Thehandbookforthe
understandingand
practiceofmeta-analysis
i
P1:FCH/FFX P2:FCH/FFX QC:FCH/FFX T1:FCH
BY035-FM BY035-Leandro-v1 October5,2004 15:39
Tothosewhohavehelpedmebecome
whoIam,eveniftheyexpectedmore
IamverygratefultoAlbertoPilottoandAngeloAndriulliformany
stimulatingandthoughtfuldiscussionsonanumberoftheoreticaland
practicalaspectsofmeta-analysis.
IamindebtedtoGiorgioGasparro,DPman,forhispatiencewithmy
requests.
Aheart-feltthankstoAndrewKBurroughs,beaconofknowledgeand
wisdomformanypeople. Withouthisadviceandsupport,thiswork
wouldhaveneverbeenwritten.
iii
P1:FCH/FFX P2:FCH/FFX QC:FCH/FFX T1:FCH
BY035-FM BY035-Leandro-v1 October5,2004 15:39
Meta-analysis
in Medical Research
The handbook for the
understanding and
practice of meta-analysis
Gioacchino Leandro
Gastroenterologist,Biostatistician
CastellanaGrotte,Italy
FOREWORDBY
Giuseppe Gallus
HeadoftheInstituteofMedicalStatistics
UniversityofMilan,Italy
v
P1:FCH/FFX P2:FCH/FFX QC:FCH/FFX T1:FCH
BY035-FM BY035-Leandro-v1 October5,2004 15:39
(cid:1)c 2005GioacchinoLeandro
PublishedbyBlackwellPublishingLtd.
BMJBooksisanimprintoftheBMJPublishingGroupLimited,usedunderlicence
BlackwellPublishing,Inc.,350MainStreet,Malden,Massachusetts02148-5020,USA
BlackwellPublishingLtd,9600GarsingtonRoad,OxfordOX42DQ,UK
BlackwellPublishingAsiaPtyLtd,550SwanstonStreet,Carlton,Victoria3053,Australia
TherightoftheAuthortobeidentifiedastheAuthorofthisWorkhasbeenassertedin
accordancewiththeCopyright,DesignsandPatentsAct1988.
Allrightsreserved.Nopartofthispublicationmaybereproduced,storedinaretrievalsystem,
ortransmitted,inanyformorbyanymeans,electronic,mechanical,photocopying,recording
orotherwise,exceptaspermittedbytheUKCopyright,DesignsandPatentsAct1988,
withoutthepriorpermissionofthepublisher.
Firstpublished2005
LibraryofCongressCataloging-in-PublicationData
Leandro,Gioacchino.
Meta-analysisinmedicalresearch: thehandbookfortheunderstandingandpracticeof
meta-analysis/GioacchinoLeandro.
p. ; cm
Includesbibliographicalreferencesandindex.
ISBN1-4051-2733-3(alk. paper)
1. Meta-analysis. 2. Medicine–Research–Evaluation.
[DNLM: 1. BiomedicalResearch–methods. 2. Meta-analysis. 3. Bias(Epidemiology)
4. PublicationBias. WA950L437m2004] I.Title.
R853.M48L43 2004
610(cid:1).72–dc22
2004017776
ISBN1-4051-2733-3
AcataloguerecordforthistitleisavailablefromtheBritishLibrary
Setin10/13.5SabonbyTechBooks,India
PrintedandboundinIndiabyReplikaPressPvt. Ltd.
CommissioningEditor: AlisonBrown
EditorialAssistant: ClaireBonnett
DevelopmentEditor: FionaPattison
ProductionController: KateCharman
ForfurtherinformationonBlackwellPublishing,visitourwebsite:
http://www.blackwellpublishing.com
Thepublisher’spolicyistousepermanentpaperfrommillsthatoperateasustainableforestry
policy,andwhichhasbeenmanufacturedfrompulpprocessedusingacid-freeandelementary
chlorine-freepractices. Furthermore,thepublisherensuresthatthetextpaperandcoverboard
usedhavemetacceptableenvironmentaccrediationstandards.
vi
P1:FCH/FFX P2:FCH/FFX QC:FCH/FFX T1:FCH
BY035-FM BY035-Leandro-v1 October5,2004 15:39
Contents
Foreword ix
Preface xi
Introduction 1
A brief history 4
Aims 5
How to Plan a Meta-Analysis 7
Defining the outcomes 8
Choosing the characteristics of the trials that one
wants to select 8
The ‘grey literature’ 9
Finding and evaluating the articles 10
Statistical procedures 13
Interpretation of results 14
Bias in Meta-Analytical Research 15
Sampling bias 16
Selection bias 18
Within study bias 19
Other biases 19
Appendix 20
Statistical Procedures 23
Fixed effects models 24
Random effects models 29
vii
P1:FCH/FFX P2:FCH/FFX QC:FCH/FFX T1:FCH
BY035-FM BY035-Leandro-v1 October5,2004 15:39
viii Contents
Working Procedures 33
Accuracy of the data 34
Evaluation of the numeric output for each trial 35
Evaluation of the pooled effect 38
Additional considerations on heterogeneity 39
Quantifying heterogeneity: the I2 index 42
Tests for publication bias 42
Number needed to treat (NNT) 43
Graphical representation 44
How to Read, Evaluate and Present a Meta-Analysis 57
How to read and evaluate a meta-analysis 58
How to present a meta-analysis 59
Examples 61
How to Use the Program 63
General information 64
Installation of the program 64
Structure of the program 65
Use of the program 69
Getting started 70
References 87
Glossary 89
Index 93
P1:FCH/FFX P2:FCH/FFX QC:FCH/FFX T1:FCH
BY035-FM BY035-Leandro-v1 October5,2004 15:39
Foreword
Intheircurrentmedicalpractice,physiciansneedtobeupdatedwiththere-
sultsofthemostimportantclinicalstudies. Inaddition,theyareoftenasked
to be part of trials and to evaluate the results of new diagnostic tools and
experimentaldata. Agoodknowledgeofprinciplesandthemethodologyof
statisticsisthereforemandatoryforallphysicians.
Althoughtherehasbeenanincreasinginterestinevidence-basedpractice
of medicine in the past few years, the data interpretation is still performed
badlyandstatisticsareoftenmisused. Amongtheprincipalreasonsforthis
isthelackofknowledgeconcerningtheuseofmathematicalandstatistical
tools, and a consequent tendency to avoid them if possible. Thus, those
physicianswhoaremorefamiliarwiththistypeofstatistics,andwhoknow
what is required in good medical practice and how to communicate with
theircolleagues,areofgreathelp.
This guide to meta-analysis, together with the included statistical soft-
ware, is the work of a clinician and it deals with the difficulties mentioned
above.
The reader is introduced to the principal aspects of meta-analysis using
a simple terminology, which does not imply a superficial approach. The
most common errors encountered in the practical application of statistical
procedures are discussed. The program is designed for easy use, which
willhelpclinicianstoutilizemeta-analysis. Thehandbookcontainsashort
sectiondedicatedtothestatisticalproceduresapplied. Abasicknowledgeof
statisticalmethodsisanessentialstartingpoint,butthismanualcanalsobe
ofgreatinterestandprovideusefulinformationforthosewhohavegreater
experience and are consolidating their statistical knowledge. It is therefore
logicalthattheauthor’schoicehasbeentoconcentrateonpracticalguidelines
ratherthandescribingstatisticalmethods.
ix
P1:FCH/FFX P2:FCH/FFX QC:FCH/FFX T1:FCH
BY035-FM BY035-Leandro-v1 October5,2004 15:39
x Foreword
I am therefore very pleased to write a Foreword for this handbook of
meta-analysis,withcertaintythatitwillbeasuccess.
ProfessorGiuseppeGallus
HeadoftheInstituteofMedicalStatistics
UniversityofMilan,Italy
P1:FCH/FFX P2:FCH/FFX QC:FCH/FFX T1:FCH
BY035-FM BY035-Leandro-v1 October5,2004 15:39
Preface
The aim of this handbook is to provide the clinical researcher, who is tra-
ditionallynotfamiliarwithmathematicsandstatistics,withthebasicprin-
ciplesnecessarytounderstandthepowerofmeta-analysisandtheessential
instrumentstoperformagoodclinicalstudy. ForthispurposeIhavechosen
simpleanddirectlanguage, usingeasyconceptsandexamplestoenablean
intuitive understanding of the problems, at the same time trying to avoid
over-simplification.
For these reasons this manual is not intended as a textbook on meta-
analysis, but rather a learning tool or a guide for the clinical researcher to
understandthebasicprinciplesofstatisticssothatlateritmaybehelpfulfor
correctuseofmeta-analysis.
Thisbookshouldenablethereadertoperformameta-analysis,reportits
results in the correct way and also help understand and critically evaluate
meta-analysis published in scientific journals. For further information the
readerisreferredtoothertextsorarticles.
Thesoftwareincludediseasytouseanditcontainsinnovativefunctions
(i.e. movinginformationfromprogramtoprograminWindows),additional
procedures(cumulativemeta-analysis,numberneededtotreat(NNT),pub-
licationbiasassessment,testforasymmetryoffunnelplot)andgraphicsnot
available in other programs (Galbraith’s plots, funnel plot and its test for
asymmetry). The plot for the sub-group analysis, especially for sensitivity
analysis, allows a better understanding of the effect of therapies or proce-
dures on sub-groups in clinical trials. These features make the program a
completeandusefultool.
Thechapteronstatisticalanalysisexplainsthecomputationalmethodsin
detail using simple terms and examples. The expert reader will find in this
sectiontheformalaspectsofthestatisticalprocedureused,whereasabegin-
ner will learn several concepts which will be very useful in understanding
meta-analysis.
xi