A New Tool for Flagging Adverse Events of Potential Clinical Interest

• Activity Number: 522
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Li Zhou*+ and David Kerr
• Companies: Axio Research Inc. and Axio Research Inc.
• Keywords: adverse events ; safety ; clinical trials ; odds ratio

Abstract

Dictionary-coded adverse events (AEs) are usually tabulated by treatment arm and perused visually for treatment differences in clinical trials. To speed this process, p-values are sometimes reported for the comparisons with flags for values below a certain threshold (e.g. 0.05). Many times, however, particularly for events with relatively small event counts, the results will not necessarily be statistically significant but are still of clinical interest. A new statistic was developed to aid this review process. It uses both the p-value of the comparison and the odds ratio which represents the magnitude of differences. In its simplest form, the proposed statistic is defined as: - log10(p-value) + abs(log10(odds ratio)) The clinical reviewer can set the level of sensitivity that is used for each review by choosing a cutoff value for the proposed statistic. This statistic has good properties that go beyond what would be found using solely the p-value and is a useful tool for quick filtering of a lengthy report so that clinical reviewers can see events that would potentially warrant further investigation. The graphical representation of this statistic is also useful and presented.