quality of sycommonly used are those developed byChalmers et al9and Jadad et al10Perhapsmore important than the scale used iswhether a scale has been used at all Onceaquality score has been assigned t ID: 897155
Download Pdf The PPT/PDF document "complete coverage of allrelevant studies..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1 complete coverage of allrelevant studies
complete coverage of allrelevant studies,look for the presence of heterogeneity,and explore the robustness of the main findings usingsensitivity analysis.1What is...? seriesSecond editionFor further titles in the series,visit:www.whatisseries.co.ukIain K CrombiePhD FFPHMProfessorof Public Health,University of DundeeHuw TO Davies quality of sy commonly used are those developed byChalmers et al9and Jadad et al.10Perhapsmore important than the scal
2 e used iswhether a scale has been used a
e used iswhether a scale has been used at all. Onceaquality score has been assigned, theimpact of excluding low quality studiescan be assessed bysensitivity analysis(see below).Calculating effect sizesClinical trials commonly present theirresults as the frequency of some outcome(such as a heart attack or death) in theintervention groups and the controlgroup. For meta-analysis these are usually hanged by varying the approach toaggregation. A good
3 sensitivity analysiswill explore, among
sensitivity analysiswill explore, among other things, the without sensitivity analyses, the readerhas tomake guesses about the likelyimpact of these important factors on thekey findings.4What ismeta-analysis? eatment effect. The aggregateeffect size obtained by combining all thestudies is usuallydisplayed as a diamond. HeterogeneityAmajor concern about meta-analyses isthe extent to which they mix studies thatare different in kind (heterogeneity
4 ). One Balcon (propanolol)14/56 (25.0)15
). One Balcon (propanolol)14/56 (25.0)15/58 (25.9)Ð0.25.5 an overall summary measure ofeffect size. Unfortunately, there is noreliable objective measure to decide whenpooling is appropriate. Thus, a rule ofthumb is given above. The technique ofmeta-regression is introduced because itprovides one way of overcoming theproblem of heterogeneity.Meta-regressionWhen heterogeneity is detected, it isimportant to investigate what may havecaused it. Meta-
5 regression is a techniquewhich allows re
regression is a techniquewhich allows researchers to explore whichtypes of patient-specific factors or studydesign factors contribute to theheterogeneity. The simplest type of meta- 1. Davies HT, Crombie IK, Tavakoli M. When can odds ratiosmislead? BMJ1998; 316:989Ð991.12. Deeks JJ, Altman DG, Bradburn MJ. Statistical methods forexamining heterogeneity and combining results from severalstudies in meta-analysis. In: Egger M, Smith GD, Altman DG(e
6 ds). Systematic Reviews in Health Care:
ds). Systematic Reviews in Health Care: Meta-analysis incontext,2nd edn. London: BMJ Publishing Group, 2001.13. Thornton A, Lee P. Publication bias in meta-analysis: itscauses and consequences. JClin Epidemiol2000; 53:207Ð216.14. Egger M, Davey Smith G, Schneider M, Minder C. Bias inmeta-analysis detected byasimple, graphical test. BMJ1997;315:629Ð634.15. Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L.Comparison of twomethods todetect publ
7 ication bias inmeta-analysis. JAMA2006;
ication bias inmeta-analysis. JAMA2006; 295:676Ð680.16. Sterne JA, Egger M. Funnel plots for detecting bias inmeta-analysis: guidelines on choice of axis. JClin Epidemiol2001; 54:1046Ð1055.17. Egger M, Smith GD. Misleading meta-analysis. BMJ1995;311:752Ð754.18. Sterne JA, Egger M, Smith GD. Systematic reviews inhealth care: Investigating and dealing with publication andother biases in meta-analysis. BMJ2001; 323:101Ð105.19. Lau J, Ioannidis JP,
8 Terrin N, Schmid CH, Olkin I. The caseof
Terrin N, Schmid CH, Olkin I. The caseof the misleading funnel plot. BMJ2006; 333:597Ð600.20. Terrin N, Schmid CH, Lau J. In an empirical evaluation ofhe funnel plot, researchers could not visually identifypublication bias. JClin Epidemiol2005; 58:894Ð901.21. Lewis S, Clarke M. Forest plots: trying to see the wood andthe trees. BMJ2001; 322:1479Ð1480.22. Huque MF. Experiences with meta-analysis in NDAsubmissions. Proc Biopharm Section Am Stat As