Box 1. What is Publication Bias?

Statistically significant studies are more likely to be published, are more easily identified in the scientific literature, and are thus more likely to appear in meta-analyses compared with statistically non-significant studies. Because studies with larger effect sizes are more likely to be statistically significant than studies with smaller effect sizes (given the same sample size), studies included in meta-analyses tend to have systematically larger effect sizes than those that are not identified for inclusion in meta-analyses. Hence, publication bias refers to a systematic deviation from the truth in the results of a meta-analysis due to the higher likelihood for published studies to be included in meta-analyses than unpublished studies. Assessing publication bias using visual plots or associated statistical tests cannot conclusively determine whether included study effect sizes and potential “missing” studies overestimate the true summary effect size in a meta-analysis. Therefore, ‘risk of publication bias’ or ‘potential publication bias’—rather than simply ‘publication bias’—is what is being assessed to judge the risk on whether publication bias is present.