Publication bias in neuroimaging research: implications for meta-analyses

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Neuroscience

Background:

• Neuroimaging and neuroscience databases facilitate meta-analysis of activation results.
• Publication bias, the tendency for studies with significant findings to be published over those with null results, has not been adequately addressed in neuroimaging meta-analysis.
• This bias can skew the interpretation of collective research findings.

Purpose of the Study:

• To investigate the presence and extent of publication bias in functional magnetic resonance imaging (fMRI) studies of working memory.
• To examine this bias specifically within frontal lobe regions (Brodmann Areas 4, 6, 8, 9, 10, 37, 45, 46, and 47).

Main Methods:

• Selected 74 studies on working memory fMRI in the frontal lobes.
• Calculated Pearson's r correlation coefficient and Cohen's d effect sizes for each study's activation.
• Employed regression-based techniques, including Egger's regression, Macaskill's regression, and the 'Trim and Fill' method, to assess publication bias by comparing effect sizes to study sample sizes.

Main Results:

• Significant evidence of publication bias was identified across the analyzed body of literature (p < 0.01 for all tests).
• The identified publication bias was not dependent on specific working memory tasks or frontal lobe sub-regions.
• These findings suggest a generalizable issue within the broader brain imaging literature.

Conclusions:

• Publication bias is a significant concern in neuroimaging research, particularly in fMRI studies of working memory.
• The identified bias can influence the accuracy of meta-analyses and the interpretation of collective findings.
• Researchers and database curators should consider and account for potential publication bias when synthesizing neuroimaging data.