Funnel Plots and Publication Bias

Detecting missing studies graphically

A funnel plot graphs each study's effect size against its precision. Without bias, studies scatter symmetrically like an inverted funnel. When symmetry breaks — a corner where small, non-significant studies should appear stays empty — publication bias or small-study effects may be at work. Egger's regression test quantifies the asymmetry numerically, and the trim-and-fill method imputes missing studies to yield an adjusted effect estimate.

Core Concept: What Is a Funnel Plot?

A funnel plot represents each primary study in a meta-analysis along two axes: the horizontal axis shows the effect size (such as standardized mean difference or log odds ratio) and the vertical axis shows a measure of precision based on sample size or the inverse of standard error. Large, precise studies cluster near the top; small, uncertain studies spread along the base. When there is no bias, the points form a symmetric inverted funnel or triangle around the pooled effect line. This symmetry is the visual signal that synthesized evidence has not been filtered by publication processes.

Reading Asymmetry and Statistical Tests

When publication bias is present, small studies with non-significant results tend not to be published, so one lower corner of the funnel plot stays empty and the graph looks lopsided. Assessing asymmetry by eye is subjective, which is why regression-based approaches such as Egger's test are used. Egger's test regresses the standardized effect size on its standard deviation; when the intercept departs significantly from zero, asymmetry is flagged. Trim-and-fill analysis then adds hypothetical missing studies to make the funnel symmetric and reports an adjusted pooled effect estimate, giving a sense of how robust the original conclusion is.

Concrete Example: Antidepressant Efficacy

Imagine a meta-analysis of 30 antidepressant trials. When the funnel plot is drawn, large studies with positive outcomes cluster near the apex, while the lower corner for small negative-result studies is nearly empty. Egger's test returns p = 0.02. Trim-and-fill adds 8 hypothetical missing studies and brings the pooled effect size noticeably below the original estimate. This finding alerts the researcher that the true treatment effect may be weaker than initially reported, prompting sensitivity analyses and a more cautious interpretation of the pooled evidence.

Common Pitfalls and Good Practice

Funnel plot asymmetry does not exclusively signal publication bias; it can also reflect genuine heterogeneity, methodological differences between small and large studies, or true small-study effects. In meta-analyses with fewer than ten studies, statistical power is low and false-negative results are likely. A significant Egger test is a warning signal, not definitive proof of bias. Contour-enhanced funnel plots help clarify which region of significance drives the asymmetry. Best practice combines funnel plot inspection with grey literature searching, prospective registration on PROSPERO, and transparent reporting — all of which reduce the underlying problem rather than merely detecting it.

Key terms

Publication Bias: Systematic distortion arising because studies with significant results are more likely to be published.
Funnel Plot: Scatter plot of each study's effect size against its precision; symmetry indicates absence of bias.
Egger's Test: Regression-based statistical test that quantifies funnel plot asymmetry numerically.
Trim-and-Fill: Method that imputes hypothetical missing studies to restore symmetry and yield an adjusted effect estimate.
Small-Study Effect: Tendency of small-sample studies to report different effect sizes than larger ones.