[Fraudulent studies, unpublished data and their effect on the development of guidelines and evidence-based recommendations].

Publication bias and scientific fraud are major threats to valid guidelines, especially guideline recommendations. Both topics have so far not been investigated systematically. It is assumed that guideline authors underestimate the potential influences on the validity of their recommendations. On the basis of questions and the example of depression guidelines the potential influence of publication bias will be outlined and discussed. Formal consensus processes which are used to phrase and grade recommendations might protect guideline recommendations against the influence of publication bias and might thereby imply one major difference to systematic reviews. Based on the examples of Werner Bezwoda and Scott S. Reuben, who fabricated studies in breast cancer and analgesia, it can be estimated that scientific fraud has no or only minimal effect on the recommendations given in clinical guidelines. Either the fraudulent work is in line with other (true) studies, which consequently leads only to an overestimation of the quantity of evidence in the guideline. Or the faked study is the only piece of evidence available, which guideline authors would then regard as an insufficient basis for a clinical recommendation. Although publication bias and scientific fraud had no influence on the example guidelines, guideline authors should be aware of this possibility and control this problem by systematically reviewing the evidence, critically appraising primary studies, and formulating prudent recommendations.