Meta-analysis: Difference between revisions
imported>Robert Badgett |
imported>Robert Badgett |
||
| Line 6: | Line 6: | ||
==Methods of meta-analysis== | ==Methods of meta-analysis== | ||
===Displaying results=== | ===Displaying results=== | ||
Study results may be grouped and displayed with a Forest plot.[[Image: | Study results may be grouped and displayed with a Forest plot.[[Image:Forest plot of target HbA1c for Diabetes Type II showing heterogeneity.jpg|right|thumb|350px|{{#ifexist:Template:Forest plot of target HbA1c for Diabetes Type II showing heterogeneity.jpg/credit|{{Forest plot of target HbA1c for Diabetes Type II showing heterogeneity.jpg/credit}}<br/>|}}'''Forest Plot''' showing [[meta-analysis]] of [[randomized controlled trial]]s of differing target glucose control and mortality for [[diabetes mellitus type 2]]. Note the heterogeneity (P<0.05 and high I<sup>2</sup> in circled in red) due to increased death when the [[glycosylated hemoglobin A]] (Hb A1c) target was 6.0% in the ACCORD trial<ref name="pmid18539917">{{cite journal |author=Gerstein HC, Miller ME, Byington RP, ''et al'' |title=Effects of intensive glucose lowering in type 2 diabetes |journal=N. Engl. J. Med. |volume=358 |issue=24 |pages=2545–59 |year=2008 |month=June |pmid=18539917 |doi=10.1056/NEJMoa0802743 |url=http://content.nejm.org/cgi/pmidlookup?view=short&pmid=18539917&promo=ONFLNS19 |issn=}}</ref>]] | ||
===Measuring consistency of study results=== | ===Measuring consistency of study results=== | ||
Consistency can be statistically tested using either the Cochran's ''Q'' or ''I<sup>2</sup>''.<ref name="pmid12958120">{{cite journal |author=Higgins JP, Thompson SG, Deeks JJ, Altman DG |title=Measuring inconsistency in meta-analyses |journal=BMJ |volume=327 |issue=7414 |pages=557–60 |year=2003 |month=September |pmid=12958120 |pmc=192859 |doi=10.1136/bmj.327.7414.557 |url=http://bmj.com/cgi/pmidlookup?view=long&pmid=12958120 |issn=}}</ref> The ''I<sup>2</sup>'' is the "percentage of total variation across studies that is due to heterogeneity rather than chance."<ref name="pmid12958120"/> These numbers are usually displayed for each group of studies on a Forest plot. | Consistency can be statistically tested using either the Cochran's ''Q'' or ''I<sup>2</sup>''.<ref name="pmid12958120">{{cite journal |author=Higgins JP, Thompson SG, Deeks JJ, Altman DG |title=Measuring inconsistency in meta-analyses |journal=BMJ |volume=327 |issue=7414 |pages=557–60 |year=2003 |month=September |pmid=12958120 |pmc=192859 |doi=10.1136/bmj.327.7414.557 |url=http://bmj.com/cgi/pmidlookup?view=long&pmid=12958120 |issn=}}</ref> The ''I<sup>2</sup>'' is the "percentage of total variation across studies that is due to heterogeneity rather than chance."<ref name="pmid12958120"/> These numbers are usually displayed for each group of studies on a Forest plot. | ||
Revision as of 12:22, 23 October 2008
Meta-analysis is defined as "a quantitative method of combining the results of independent studies (usually drawn from the published literature) and synthesizing summaries and conclusions which may be used to evaluate therapeutic effectiveness, plan new studies, etc., with application chiefly in the areas of research and medicine."[1]
A meta-analyses is a subset of systematic reviews in which the results of the studies are numerically pooled.
Methods of meta-analysis
Displaying results
Study results may be grouped and displayed with a Forest plot.
Forest Plot showing meta-analysis of randomized controlled trials of differing target glucose control and mortality for diabetes mellitus type 2. Note the heterogeneity (P<0.05 and high I2 in circled in red) due to increased death when the glycosylated hemoglobin A (Hb A1c) target was 6.0% in the ACCORD trial[2]
Measuring consistency of study results
Consistency can be statistically tested using either the Cochran's Q or I2.[3] The I2 is the "percentage of total variation across studies that is due to heterogeneity rather than chance."[3] These numbers are usually displayed for each group of studies on a Forest plot.
In interpreting of the Cochran's Q, heterogeneity exists if its p-value is < 0.05.
The following has been proposed for interpreting I2:[3]
- Low heterogeneity is I2 = 25%
- Moderate heterogeneity is I2 = 50%
- High heterogeneity is I2 = 75%
or according to the Handbook of the Cochrane Collaboration:[4]
- 0%-40%: might not be important
- 30%-60%: may represent moderate heterogeneity
- 50%-90%: may represent substantial heterogeneity
- 75%-100%: considerable heterogeneity
Variations on meta-analysis
Individual patient data meta-analysis
Individual patient data meta-analysis (IPD meta-analysis) may have more long lasting results than other meta-analyses.[5]
Network meta-analysis
A network meta-analysis pools studies in order to compare to treatments that have not been directly compared.[6][7]
Factors associated with higher quality meta-analyses
Meta-analyses by the Cochrane Collaboration tend to be of higher quality.[8]
Individual data meta-analyses, in which the records from individual patients are pooled together into one dataset, tend to have more stable conclusions.[5]
Factors associated with lower quality meta-analyses
About a third of meta-analyses that happen to precede large randomized controlled trials will conflict with the results of the trial.[9]
Conflict of interest
Meta-analyses produced with a conflict of interest are more likely to interpret results as positive.[10]
Publication bias
Publication bias against negative studies may threaten the validity of meta-analyses that are positive and all the studies included within the meta-analysis are small.[11][12]
In performing a meta-analyses, a file drawer[13] or a funnel plot analysis[12][14] may help detect underlying publication bias among the studies in the meta-analysis.
Outcome reporting bias
Meta-analyses in which a smaller proportion of included trials provide raw data for inclusion in the meta-analysis are more likely to be positive.[15] This may be due a bias against reporting negative results.[16]
References
- ↑ National Library of Medicine. Meta-analysis. Retrieved on 2007-12-06.
- ↑ Gerstein HC, Miller ME, Byington RP, et al (June 2008). "Effects of intensive glucose lowering in type 2 diabetes". N. Engl. J. Med. 358 (24): 2545–59. DOI:10.1056/NEJMoa0802743. PMID 18539917. Research Blogging.
- ↑ 3.0 3.1 3.2 Higgins JP, Thompson SG, Deeks JJ, Altman DG (September 2003). "Measuring inconsistency in meta-analyses". BMJ 327 (7414): 557–60. DOI:10.1136/bmj.327.7414.557. PMID 12958120. PMC 192859. Research Blogging.
- ↑ Higgins JPT, Green S:Cochrane Handbook for Systematic Reviews of Interventions. Cochrane Collaboration (2008). Retrieved on 2008-10-23.
- ↑ 5.0 5.1 Poynard T, Munteanu M, Ratziu V, et al (June 2002). "Truth survival in clinical research: an evidence-based requiem?". Ann. Intern. Med. 136 (12): 888–95. PMID 12069563. [e]
Cite error: Invalid
<ref>tag; name "pmid12069563" defined multiple times with different content - ↑ Lumley T (August 2002). "Network meta-analysis for indirect treatment comparisons". Stat Med 21 (16): 2313–24. DOI:10.1002/sim.1201. PMID 12210616. Research Blogging.
- ↑ Salanti G, Kavvoura FK, Ioannidis JP (April 2008). "Exploring the geometry of treatment networks". Ann. Intern. Med. 148 (7): 544–53. PMID 18378949. [e]
- ↑ Olsen O, Middleton P, Ezzo J, et al (2001). "Quality of Cochrane reviews: assessment of sample from 1998". BMJ 323 (7317): 829–32. PMID 11597965. [e]
- ↑ LeLorier J, Grégoire G, Benhaddad A, Lapierre J, Derderian F (1997). "Discrepancies between meta-analyses and subsequent large randomized, controlled trials". N. Engl. J. Med. 337 (8): 536–42. PMID 9262498. [e]
- ↑ Veronica Yank, Drummond Rennie, and Lisa A Bero, “Financial ties and concordance between results and conclusions in meta-analyses: retrospective cohort study,” BMJ 335, no. 7631 (December 8, 2007), http://www.bmj.com/cgi/content/abstract/335/7631/1202 (accessed December 7, 2007).
- ↑ Sutton AJ, Duval SJ, Tweedie RL, Abrams KR, Jones DR (2000). "Empirical assessment of effect of publication bias on meta-analyses". BMJ 320 (7249): 1574–7. PMID 10845965. [e]
- ↑ 12.0 12.1 Egger M, Davey Smith G, Schneider M, Minder C (1997). "Bias in meta-analysis detected by a simple, graphical test". BMJ 315 (7109): 629–34. PMID 9310563. [e]
Cite error: Invalid
<ref>tag; name "pmid9310563" defined multiple times with different content - ↑ Pham B et al. (2001). "Is there a "best" way to detect and minimize publication bias? An empirical evaluation". Evaluation & the Health Professions 24: 109–25. PMID 11523382. [e]
- ↑ Terrin N et al. (2005). "In an empirical evaluation of the funnel plot, researchers could not visually identify publication bias". J Clin Epidemiol 58: 894–901. DOI:10.1016/j.jclinepi.2005.01.006. PMID 16085192. Research Blogging.
- ↑ Furukawa TA, Watanabe N, Omori IM, Montori VM, Guyatt GH (February 2007). "Association between unreported outcomes and effect size estimates in Cochrane meta-analyses". JAMA 297 (5): 468–70. DOI:10.1001/jama.297.5.468-b. PMID 17284696. Research Blogging.
- ↑ Chan AW, Hróbjartsson A, Haahr MT, Gøtzsche PC, Altman DG (May 2004). "Empirical evidence for selective reporting of outcomes in randomized trials: comparison of protocols to published articles". JAMA 291 (20): 2457–65. DOI:10.1001/jama.291.20.2457. PMID 15161896. Research Blogging.