The integrity of the data from clinical trials and of its use is an essential element of the scientific method, and of the trust one can have in this method. There are many examples of fraud, and they recur regularly. The objective of this round table was to work on the definition of fraud, on its recognition and prevention especially in the institutional system. Fraud involves an active decision to cheat, and ranges from trying to hide incompetence to wholesale invention of data, patients or studies. Its frequency is difficult to evaluate but might be as high as 1% of all studies or publications. Fraud can involve ethics (post-hoc IRB [institutional review board] approval, IRB requests not applied, lack of consent), or any of the steps from realisation to interpretation of studies or trials. Identification of fraud is made harder by the usual risk for the whistle-blowers, who must be protected. Seeking fraud is implicit in Good Clinical Practices (GCP) that all industry sponsors must apply, but that are less often applied by institutional sponsors. It might be useful to install procedures to detect fraud in studies, especially institutional. Various statistical methods can be used to identify unusual data patterns that could suggest fraud. Once fraud is identified, its management is often not foreseen. Here again, clear procedures or recommendations would be of help.