A review of publication bias in the gastroenterology literature.

In systematic reviews and meta-analyses, publication bias is particularly problematic, given that combining only statistically significant outcomes is likely to overestimate the true effect of an intervention since non-significant findings have been omitted. We examined practices for evaluating publication bias from gastroenterology literature. We performed a PubMed search to identify systematic reviews published in American Journal of Gastroenterology, Gut, and Gastroenterology from 2005 to 2015. Of the 304 found, 215 studies were eligible for inclusion based on relevant study characteristics. There were 190 systematic reviews which used at least one method to evaluate publication bias and/or included ten or more primary studies. There were 115/190 (60.53%) systematic reviews which used at least one method to evaluate publication bias. Most (105/115, 91.27%) qualified reviews used at least one method to evaluate publication bias and 78/115 (67.83%) used a combination of methods. The most common methods were funnel plot (100/115, 86.96%), Egger's regression (67/115, 58.26%), and Begg's (28/115, 24.35%). Of the 115 reviews that performed evaluations, 26 (22.61%) conducted these analyses with fewer than ten primary studies, and a minority (24/115, 20.87%) reached the conclusion that publication bias was present in their work. While methods to assess publication bias were frequently noted among qualified systematic reviews, these methods are limited in value and could be improved by incorporating approaches that assess the degree of publication bias severity.