JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Comparison of two methods to detect publication bias in meta-analysis

  • 0Centre for Biostatistics and Genetic Epidemiology, Department of Health Sciences, University of Leicester, Leicester, England. jlp9@leicester.ac.uk
Jama +

|

February 10, 2006

|

February 10, 2006

Related Experiment Videos

Contact us if these videos are not relevant.

View abstract on PubMed

Summary

This summary is machine-generated.

A new regression test based on sample size demonstrates appropriate type I error rates for detecting publication bias when using the natural log of the odds ratio (lnOR). This alternative test performs comparably to Egger's regression test, offering a reliable method for meta-analysis.

Area Of Science

  • Biostatistics
  • Medical Research Methodology
  • Meta-Analysis

Background

  • Egger's regression test is a common tool for identifying publication bias in meta-analyses.
  • The reliability of Egger's test and funnel plots is questioned when the natural log of the odds ratio (lnOR) is the summary estimate.
  • Publication bias can distort the findings of meta-analyses, impacting evidence-based medicine.

Purpose Of The Study

  • To compare the performance of Egger's regression test against a novel sample size-based regression test.
  • To evaluate these tests specifically when the natural log of the odds ratio (lnOR) serves as the summary estimate.
  • To assess the accuracy and power of each test in detecting publication bias.

Main Methods

  • Simulated meta-analyses were conducted under various scenarios, including the presence and absence of publication bias and between-study heterogeneity.
  • Type I error rates (false positives) were calculated for both Egger's regression test and the alternative sample size-based regression test.
  • The power (true positive rate) of each test to detect existing publication bias was assessed.

Main Results

  • Egger's regression test exhibited higher Type I error rates compared to the alternative regression test.
  • The alternative regression test maintained appropriate Type I error rates across different odds ratios, study numbers, and heterogeneity levels.
  • Under low between-study heterogeneity, the alternative test showed comparable power to Egger's test in detecting publication bias.

Conclusions

  • The alternative regression test demonstrates appropriate Type I error rates and reduces the correlation between lnOR and its variance.
  • This sample size-based regression test is recommended as a replacement for Egger's regression test when using lnOR as the summary estimate.
  • The findings support the use of the alternative test for more reliable publication bias detection in meta-analyses with lnOR estimates.

Related Experiment Videos

Contact us if these videos are not relevant.

Related Concept Videos