Publication bias

Publication bias is the tendency for studies with positive or striking results to be published, and published sooner, while null results stay in the file drawer. The literature that survives is therefore not a fair sample of the research that was actually done, and any meta-analysis built on it inherits the shifted picture. The standard detection tool is the funnel plot. Each study is plotted by effect size against its precision. Small, imprecise studies scatter widely at the base, large ones cluster near the pooled estimate, and a complete literature forms a roughly symmetric funnel. A gap in one lower corner suggests missing small negative studies. Egger's test fits a regression to that asymmetry. Trim-and-fill imputes the apparently missing studies and recomputes the pooled effect. All three are weaker than they look. Egger's test has low power below roughly ten studies, and it flags asymmetry more readily when effects are ratios and events are rare. Trim-and-fill adds studies that may never have existed, so its adjusted estimate is a sensitivity check, not a correction. Asymmetry also has causes other than publication bias: real differences between small and large trials, weaker methods in small studies, or genuine heterogeneity. Prospective trial registration and pre-specified outcomes address the problem at its source better than any test applied afterwards.

Last reviewed: August 25, 2026

This definition is educational and is not medical advice, a diagnosis, or treatment. Talk to a doctor about any health decisions. Read our full medical disclaimer