Statistics Glossary → Trim-and-fill
Trim-and-fillconcept
A meta-analysis method that uses funnel-plot asymmetry to estimate how many studies may be missing and gives an adjusted pooled effect after adding them.
Under publication bias, small studies with non-significant results go missing and one side of the funnel plot empties out. Duval and Tweedie's method temporarily 'trims' the extreme small studies causing the asymmetry, re-estimates the centre from the symmetric remainder and estimates the number of missing studies (with the L₀ or R₀ estimator). It then 'fills' in mirror images of the trimmed studies around the centre and recomputes the pooled effect with the enlarged set.
The method assumes that publication bias is the only cause of the asymmetry, whereas genuine heterogeneity, small studies conducted in different populations or poor quality can also produce it. The filled studies are not real data. Trim-and-fill should therefore be read not as a correction but as a sensitivity analysis showing how much the result depends on publication bias.
An everyday example
The effect of a drug is pooled in a meta-analysis of 15 studies, giving an average effect of d = 0.50. The lower-left corner of the funnel plot is empty: small studies with null findings seem to be unpublished. Trim-and-fill estimates 4 missing studies and the adjusted effect is d = 0.35; there is still an effect, but it may not be as large as first estimated.
In MerQur
The 'Publication bias: Trim-and-fill' option in Meta-Analysis gives the number of missing studies estimated with Duval and Tweedie's L₀ estimator and the adjusted pooled estimate.