Preventing Scientific Fraud

IN RESPONSE:

Dr. Noble feels that journals should do more to prevent scientific fraud. He proposes that journals do an in-depth investigation of randomly chosen articles. Research misconduct is almost always detected inside the institution. In 1988, Dr. Rennie proposed doing such a random audit of accepted manuscripts as an experiment, the results to be published only in aggregate (1, 2). The idea was to measure the frequency of the grossest forms of misconduct, for example, how often researchers sent in papers based entirely on fabricated patients, as had happened in the cases of Soman and of Darsee (3). Dr. Rennie proposed that acceptance for publication would be conditional on the authors allowing the auditors full access to their places of work.

The problem with making such a proposal routine is that editors would face insuperable barriers in access to research institutions and have no power or expertise (nor time or money) to carry out adequate investigations. In addition, the pretest probability of out-and-out fraud is doubtless very low. Editors' lack of forensic skills would guarantee that routinely testing for fraud would have a low sensitivity (4), so the probability of detecting fraud would be extremely low, and we would miss many cases. Moreover, the scheme would have costs, starting with Dr. Noble's proposed fee and the random selection of a target, which would undoubtedly create an atmosphere of distrust. In the end, we agree with Kennedy (5) that routinely auditing a random selection of all papers is a bad plan.

In fact, most journals investigate the validity of articles that they intend to publish. Both the Journal of the American Medical Association and Annalsof Internal Medicine do a detailed evaluation of the statistical methods and often ask for the study protocol. However, we see no workable alternative to starting with the assumption that authors are trying to offer a faithful depiction of the facts. We see our main mission as presenting research reports in a form that makes it easy for scientific peers to evaluate them and try to reproduce the findings, which is the best, and certainly the most practical, way to detect results that are wrong for whatever reason. The true test of any discovery begins with publication.

We thank Dr. Garfield and his colleagues for reminding all of us that the SCI is another way to find articles that have been retracted. We agree that present-day authors have no excuse for citing a retracted article. Still, they do.

Dr. Garrel and her colleagues have informed the research community that their university has investigated the 6 articles written by Dr. Poehlman during his tenure at the University of Montréal. The university has fulfilled its obligation, and we have fulfilled ours by publishing its findings. May all universities and journals take their responsibilities so seriously.