One of the aims of the meta-analysis of clinical trials is to deter- mine the efficacy of a new type of treatment. This efficacy is commonly measured by the difference between the efficacy of a standard treatment and the new treatment. For binary data the difference can be measured by a probability difference. We investigate, by simulations and using box plots, the basic statistical properties of the point estimator of the probability difference of overall treatment effects in the meta-analysis based on multicentre trials for various chosen situations. This estimator was suggested in Dokoupilova (2011).