trimandfill

Trim-and-fill is a statistical method used in meta-analysis to adjust for publication bias by estimating and imputing missing studies to achieve a more symmetrical funnel plot and to obtain an adjusted overall effect size. The method, introduced by Duval and Tweedie in 2000, treats funnel plot asymmetry as evidence of missing studies on one side of the plot, usually due to non-publication of small or negative studies.

Procedure: The observed effects and standard errors are used to construct a funnel plot. In the trim step, small studies are removed from the side causing asymmetry to identify the funnel’s center and estimate how many studies are missing. In the fill step, those missing studies are imputed by reflecting the trimmed ones about the center, yielding a symmetric dataset. A meta-analysis is then performed on the observed and imputed studies to obtain an adjusted effect and a revised funnel plot.

Output and interpretation: The method provides an adjusted effect size and confidence interval, reflecting what might be obtained if the missing studies were included. It is commonly used as a sensitivity analysis to assess the potential impact of publication bias on meta-analytic conclusions.

Limitations: Trim-and-fill relies on the assumption that asymmetry is primarily due to publication bias; other factors such as true heterogeneity, study quality, or selection biases can also cause asymmetry. It does not prove bias and can be misleading if heterogeneity is substantial or if the missingness mechanism is not accurately captured. It is one of several tools for exploring funnel plot asymmetry and should be interpreted cautiously.