Discussion: Bias is a threat to meta-analysis. It is obvious that including flawed data into a meta-analysis will lead to flawed conclusions of that meta-analysis. Interestingly, all known biases in original reports tend to overestimate treatment effect. The main question is: "what is the quantitative impact of a particular source of bias on the overall result of a meta-analysis?" This question has been answered for duplicate publications: the quantitative impact may be non-negligent. This question, however, has still not been answered for unpublished data. Thus, the quantitative impact of publication bias on meta-analysis remains unclear.