Data Mining and Signal Detection in CFSAN's Adverse Event Reporting System
View Presentation View Presentation
*Stuart Jay Chirtel, FDA\CFSAN\OFDCER 

Keywords: signal detection, data mining, disproportionality analysis

CFSAN receives passive adverse event reports related to foods, dietary supplements and cosmetics from individuals, industry and other sources. These data are analyzed for suspicious or unexpected reporting patterns using a variety of techniques including several types of disproportionality analyses. The goal of the analysis is to detect higher-than-expected numbers of reports of product-symptom combinations. Due to the lack of information on product usage and reporting, signal detection techniques rely on deviations in the symptom profile of a specific product compared to that of the database as a whole. The choice of the comparison products for the analysis will have profound effects on the signal scores. Examples of disproportionality analyses of CFSAN-regulated products, such as the dietary supplement Hydroxycut, using several disproportionality techniques will be described.