The Reliability of Signal Analysis in the Adverse Events Setting of Priorities

• Activity Number: 481
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Qian Graves*+ and Peng T. Liu and Stuart Chirtel and Curtis Barton and Debra Street
• Companies: FDA/CFSAN and FDA and FDA/CFSAN and FDA/CFSAN and FDA/CFSAN
• Keywords: Numerator Analysis ; Good Signal ; Risk Order ; Adverse Drug Reactions ; Mantel-Haenszel Procedure ; Setting Priority

Abstract

The lack of exposure (i.e., denominator) information is one of the critical deficiencies in the Adverse Events Reporting System (AERS). Logically speaking, it is unlikely that one can rank the AE {adverse events} risks correctly by manipulating the numerator data alone. The purpose of this presentation is to develop the following procedure for testing the reliability of the signal analysis being currently employed: 1) construct a hypothetical population with a priori known risk order or use the real-life data obtained from one of the post-market follow-up studies, 2) censor the denominator information to mimic what occurs for the AERS, 3) apply the signal analysis in use now for analyzing the AERS database, 4) rank the AE risks based on the "relative risk signals", and 5) compare the risk order determined by the signals with the a priori known risk order.