Background: Statistical tests for the detection of publication bias are used routinely in systematic reviews: However, the uptake of approaches to model publication bias has been minimal in applications partly due to the complexity of the methods and the lack of user-friendly software to conduct the analyses. An interesting approach to model publication bias are selectivity models developed by Copas and colleagues [1,2]. In selectivity models, a single source of bias (publication bias) is modelled by linking two regression equations (a selection equation and an equation for the treatment outcome). Objectives: Empirical evaluation of the Copas selection model for publication bias compared to the well-known trim-and-fill method [3]. Methods: Application of Copas selection model and trim-and-fill to a set of 157 meta-analyses with binary outcomes. Key features in the comparison are the bias-adjusted treatment estimates and the number of missing studies predicted by the two methods. The statistical package R has been utilised in this evaluation using the publicly available R package meta and an additional R package to fit the Copas selection model. Results: The evaluation of the 157 meta-analyses is currently underway. Results will be presented at the Cochrane Colloquium. References: [1] John Copas and Jian Qing Shi, ‘‘Meta-analysis, funnel plots and sensitivity analysis’’, Biostatistics, vol. 1, pp. 247-262, 2000. [2] J. B. Copas and J. Q. Shi, ‘‘A sensitivity analysis for publication bias in systematic reviews’’, Statistical Methods in Medical Research, vol. 10, pp. 251-265, 2001. [3] Sue Duval and Richard Tweedie, ‘‘A nonparametric ‘‘Trim and Fill’’ method of accounting for publication bias in metaanalysis’’, Journal of the American Statistical Association, vol. 95, pp. 89-98, 2000.