Publication bias is a type of systematic error when synthesizing evidence that cannot represent the underlying truth. Clinical studies with favorable results are more likely published and thus exaggerate the synthesized evidence in meta-analyses. The trim-and-fill method is a popular tool to detect and adjust for publication bias. Simulation studies have been performed to assess this method, but they may not fully represent realistic settings about publication bias. Based on real-world meta-analyses, this article provides practical guidelines and recommendations for using the trim-and-fill method. We used a worked illustrative example to demonstrate the idea of the trim-and-fill method, and we reviewed three estimators (R ₀, L ₀, and Q ₀) for imputing missing studies. A resampling method was proposed to calculate P values for all 3 estimators. We also summarized available meta-analysis software programs for implementing the trim-and-fill method. Moreover, we applied the method to 29,932 meta-analyses from the Cochrane Database of Systematic Reviews, and empirically evaluated its overall performance. We carefully explored potential issues occurred in our analysis. The estimators L ₀ and Q ₀ detected at least one missing study in more meta-analyses than R ₀, while Q ₀ often imputed more missing studies than L ₀. After adding imputed missing studies, the significance of heterogeneity and overall effect sizes changed in many meta-analyses. All estimators generally converged fast. However, L ₀ and Q ₀ failed to converge in a few meta-analyses that contained studies with identical effect sizes. Also, P values produced by different estimators could yield different conclusions of publication bias significance. Outliers and the pre-specified direction of missing studies could have influential impact on the trim-and-fill results. Meta-analysts are recommended to perform the trim-and-fill method with great caution when using meta-analysis software programs. Some default settings (e.g., the choice of estimators and the direction of missing studies) in the programs may not be optimal for a certain meta-analysis; they should be determined on a case-by-case basis. Sensitivity analyses are encouraged to examine effects of different estimators and outlying studies. Also, the trim-and-fill estimator should be routinely reported in meta-analyses, because the results depend highly on it.

Epidemiologic research often involves meta-analyses of proportions. Conventional two-step methods first transform each study’s proportion and subsequently perform a meta-analysis on the transformed scale. They suffer from several important limitations: the log and logit transformations impractically treat within-study variances as fixed, known values and require ad hoc corrections for zero counts; the results from arcsine-based transformations may lack interpretability. Generalized linear mixed models (GLMMs) have been recommended in meta-analyses as a one-step approach to fully accounting for within-study uncertainties. However, they are seldom used in current practice to synthesize proportions. This article summarizes various methods for meta-analyses of proportions, illustrates their implementations, and explores their performance using real and simulated datasets. In general, GLMMs led to smaller biases and mean squared errors and higher coverage probabilities than two-step methods. Many software programs are readily available to implement these methods.

It is common to measure continuous outcomes using different scales (eg, quality of life, severity of anxiety or depression), therefore these outcomes need to be standardized before pooling in a meta-analysis. Common methods of standardization include using the standardized mean difference, the odds ratio derived from continuous data, the minimally important difference, and the ratio of means. Other ways of making data more meaningful to end users include transforming standardized effects back to original scales and transforming odds ratios to absolute effects using an assumed baseline risk. For these methods to be valid, the scales or instruments being combined across studies need to have assessed the same or a similar construct

Meta‐analyses have been increasingly used to synthesize proportions (eg, disease prevalence) from multiple studies in recent years. Arcsine‐based transformations, especially the Freeman–Tukey double‐arcsine transformation, are popular tools for stabilizing the variance of each study's proportion in two‐step meta‐analysis methods. Although they offer some benefits over the conventional logit transformation, they also suffer from several important limitations (eg, lack of interpretability) and may lead to misleading conclusions. Generalized linear mixed models and Bayesian models are intuitive one‐step alternative approaches, and can be readily implemented via many software programs. This article explains various pros and cons of the arcsine‐based transformations, and discusses the alternatives that may be generally superior to the currently popular practice.