A modified test for small-study effects in meta-analyses of controlled trials with binary endpoints.

Area of Science:

• Biostatistics
• Medical Research Methodology
• Evidence Synthesis

Background:

• Publication bias can distort meta-analysis findings.
• Funnel plots are commonly used to visually assess bias, but formal tests are needed.
• Existing statistical tests for funnel plot asymmetry can have high false-positive rates with binary outcomes.

Purpose of the Study:

• To develop and evaluate a modified linear regression test for funnel plot asymmetry.
• To address limitations of existing tests in detecting publication bias for binary data.
• To improve the reliability of meta-analysis by refining bias detection methods.

Main Methods:

• Development of a modified linear regression test using efficient score and Fisher's information.
• Simulation analyses comparing the new test with existing methods (e.g., Egger's test).
• Evaluation based on characteristics of published controlled trials, including treatment effects, event counts, and trial sizes.

Main Results:

• The modified test demonstrates a false-positive rate close to the nominal level with little or no between-trial heterogeneity.
• The new test maintains similar statistical power to the original linear regression test ('Egger' test) under these conditions.
• No proposed test uniformly performs well when substantial between-trial heterogeneity is present.

Conclusions:

• The developed modified linear regression test is a more reliable tool for detecting publication bias in meta-analyses with binary outcomes, particularly when heterogeneity is low.
• This advancement enhances the accuracy of evidence synthesis by providing a robust method for bias assessment.
• Further research may be needed to address limitations in highly heterogeneous meta-analyses.