Signal Detection of Potentially Fraudulent Activity in Clinical Trials

• Activity Number: 217
• Type: Topic Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Richard C. Zink*+
• Companies: JMP Life Sciences, SAS Institute
• Keywords: Clinical Trial ; Fraud Detection ; Graphics ; Multiplicity

Abstract

It is currently believed that fraud in clinical trials is a relatively rare phenomenon with the proportion of investigators committing fraudulent acts below one percent (Buyse et al, 1999). However, instances of fraudulent activity are likely to have gone undiagnosed due to a lack of available statistical and graphical tools, and the past over-reliance on the manual monitoring of data to assess trial quality. In this talk, we briefly review some analyses to identify patient- and investigator-perpetrated fraud. Further, we describe an approach to screen the study database to identify potential fraud and other quality issues using volcano plots. We illustrate the use of volcano plots with analyses of digit preference and visit attendance from a clinical trial of delayed cerebral vasospasm in patients who experienced an aneurysmal subarachnoid hemorrhage.