Fraud Is Bad, Studying Fraud Is Hard

Recently reported in Science: a German hematologist with a bibliography of 347 papers has been shown to have included fraudulent data in 52 of these papers, with another 42 raising strong suspicions of fraud [1]. It appears this gentleman will be joining the list of notorious researchers who, motivated presumably by unrestrained ambition, falsified and manipulated data in numerous publications before being discovered and exposed. Names such as William Summerlin, John Darsee, Robert Slutsky and Roger Poisson will forever remind us that those who enter the realm of scientific research are not always inspired solely by the desire to seek truth [2, 3]. Each new instance of research fraud that makes its way into the public consciousness eats away at public confidence in the scientific enterprise. This is particularly true for fraud in clinical investigations because the results of clinical research have a more direct impact on the treatment of patients. For example, much more media attention was devoted to Roger Poisson, who falsified data on women with breast cancer to make them appear eligible for clinical trials, than to any prior uncovering of fraud. As a consequence, this case had a more explosive impact than others, garnering the attention of the U.S. Congress and ultimately affecting the careers of senior scientists and leading to changes in the management of sponsored research [3]. It is surely in the best interest of science that fraudulent results be detected before they can be disseminated. One might hope that the peer review process would identify such distortions, but major fraud seems unlikely to be readily detected by the typically limited attention given by peer review. Horton, commenting on another recent case of data falsification in a breast cancer trial performed in South Africa [4], suggests that increasing the burden on the review process to improve the chance of detecting fraud is not warranted [5]. Those in the best position to ferret out data falsification or other manipulation prior to publication of results are probably the scientific collaborators of the individual attempting to perpetrate fraud. These individuals have in-depth knowledge of the investigation in progress and an extended time period during which they may become aware of the acts of fraud. Data from a review of cases of alleged misconduct in academic institutions during 1980 to 1987 support this speculation; half of the 26 cases analyzed were initially detected by coworkers, while only 3 cases were identified by editorial peer review [6]. The paper by Ranstam et al. in this issue of Controlled Clinical Trials suggests the frightening scenario that fraud in clinical research may be much more prevalent than