Modelling publication bias in meta-analysis: a review

Area of Science:

• Biostatistics
• Medical Research Methodology

Background:

• Meta-analysis is a crucial tool for synthesizing evidence from multiple studies.
• Publication bias, favoring statistically significant results, can compromise the accuracy of meta-analyses.
• Uncritical combination of published studies may lead to overly optimistic conclusions.

Purpose of the Study:

• To review methods for assessing the impact of publication bias in meta-analyses.
• To promote awareness and use of sensitivity analysis techniques for publication bias.
• To provide a research agenda for addressing publication bias in meta-analysis.

Main Methods:

• Review of existing statistical methods for sensitivity analysis.
• Focus on techniques to assess the potential impact of missing non-significant studies.
• Exploration of approaches beyond simple detection of publication bias.

Main Results:

• Several sensitivity analysis methods can be employed to evaluate publication bias.
• These methods help quantify the potential distortion of meta-analysis results.
• Current statistical tests for publication bias do not offer solutions for proceeding when bias is suspected.

Conclusions:

• Sensitivity analysis is essential for robust meta-analysis when publication bias is a concern.
• Further development and application of these methods are needed.
• Addressing publication bias is critical for reliable evidence synthesis in research.