Introduction: Meta-analyses are increasingly being used to assess the treatment effects. However, care is required in the interpretation of their results.

Objective: To demonstrate how purely chance effects can influence the results of meta-analyses to give results, which could have a major public impact.

Methods: Participants in the Edinburgh Stroke Course rolled dice to represent patients in the treatment and control groups of randomised trials. Whenever the dice showed a "six" a death was recorded. The results were combined in a meta-analysis and predefined and post hoc subgroup analyses performed. A random list of all the trials was produced and publication bias simulated by assuming the first 70% of positive trials and the first 40% of negative trials from this list were published.

Results/Discussion: Omitted at the request of the authors (to be distributed at meeting).