Hybrid test for publication bias in meta-analysis

Area of Science:

• Biostatistics
• Medical Research Methodology
• Evidence Synthesis

Background:

• Publication bias, where study results influence publication, can lead to skewed meta-analyses.
• Existing statistical tests for publication bias often rely on specific assumptions, limiting their power in diverse situations.
• Selecting the optimal test for real-world meta-analyses is challenging due to the difficulty in identifying the exact bias mechanism.

Purpose of the Study:

• To develop a novel hybrid statistical test for publication bias.
• To create a test that maintains high power across various publication bias mechanisms.
• To offer a more robust method for detecting publication bias in meta-analyses.

Main Methods:

• A hybrid statistical test was developed by synthesizing various existing tests.
• The performance of the hybrid test was evaluated through simulation studies.
• The hybrid test was compared against established methods like regression, rank tests, and trim-and-fill using real-world meta-analyses.

Main Results:

• The proposed hybrid test demonstrated superior performance compared to existing methods.
• The hybrid test maintained relatively high statistical power across different simulated publication bias mechanisms.
• Validation using three real-world meta-analyses confirmed the hybrid test's effectiveness.

Conclusions:

• The novel hybrid test offers a more reliable approach to detecting publication bias in meta-analyses.
• This method enhances the accuracy of evidence synthesis by mitigating the impact of suppressed studies.
• The hybrid test provides a versatile tool for researchers dealing with potential publication bias in their work.