The Role of Data Audits in Detecting Scientific Misconduct

Abstract

To evaluate the extent of the problem of scientific misconduct in investigational drug trials, we reviewed data from 1955 routine audits conducted by the US Food and Drug Administration (FDA) from June 1977 to April 1988. Serious deficiencies were detected in 12% of audits prior to October 1985, but in only 7% since that date. At the same time, there was no evidence of a decline over time in the rate of detection of many categories of deficiencies, and some investigators were able to continue to participate in drug trials after flagrant violations of recognized norms of research. The data auditing program should be continued, but additional measures are needed to regulate misconduct. These must be tailored to the variety of causes of misconduct, ranging from negligence to fraud. Possible additional approaches could include certifying the competence of potential investigators; peer-reviewed, competitive application for the opportunity to conduct FDA-authorized clinical trials; limiting an investigator's level of participation in clinical trials; penalizing manufacturers who fail to detect their investigators' misconduct; and permitting the FDA to suspend investigators prior to a hearing. Measures taken should maximize public utility at the least economic cost to society and should be evaluated thoroughly.

Assessing the Integrity of Publicly Funded Research

“…Of the roughly one-in-four researchers who were aware of misconduct (27%), 47% said that the cases were not public knowledge (4). • Audits of research procedures and results have turned up "significant problems" or "major deviations" a levels that range at and above the 10% level (5)(6)(7)(8). These results do not correlate directly with FFP, since they do not take into account whether discrepancies result from deliberate actions.…”

“Reproducible” Research in Mathematical Sciences Requires Changes in our Peer Review Culture and Modernization of our Current Publication Approach

“…The good news is that multiple studies show that there is no evidence that the failure to “reproduce” is the result of scientific misconduct (see Fanelli 2018 and reference therein). Less than 2% of scientists are fabricating or falsifying data (Shapiro and Charrow 1989, Steneck 2006, Fanelli 2009). Metaresearch studies suggest that the “reproducibility” crisis results from heterogenous factors distributed across different areas of science that include bias in hypothesis testing (Fanelli 2010, Fanelli et al 2017), poor experimental design, statistical problems (John et al 2012, Button et al 2013) and lack of standardization to report and share experimental protocols, results and data (Wicherts et al 2006, Nature Editorial 2014, Stark 2018).…”

Rationale, objectives, and interpretation of randomized controlled trials

“…In a review of approximately 2,000 audits conducted between 1977 and 1988, serious deficiencies were discovered in 12% [30]. The most prevalent problems related to patient consent, drug accountability, protocol nonadherence, and inaccurate records.…”

“…The most prevalent problems related to patient consent, drug accountability, protocol nonadherence, and inaccurate records. In addition, over the past decade, there has been a disturbing number of reports of scientific misconduct, including data fabrication and plagiarism [30][31][32]. Any audit plan needs to concentrate on key information and to distinguish between important and unimportant errors.…”