Association analysis of drug side effects using visual grouping techniques
Naohiro Masukawa, Systems Engineer, JMP Japan Division, SAS Institute Japan Co., Ltd.
Association analysis is used to find combinations of products that are highly related among those purchased by consumers in stores, but it can also be applied to find patterns of side effects of drugs. This analysis can tell you which combinations of side effects are likely to occur in patients who have been administered a certain drug. JMP Pro 13 has an "association analysis" platform, which uses an a priori algorithm to present combinations of side effects that are frequent, even with a large amount of side effect data and a wide variety of side effects. In addition, this platform has an option to perform singular value decomposition (SVD) on matrices of transactions (patients) and items (side effects), which allows you to group patients with similar patterns of side effects and visually group the side effects themselves by reducing the matrix to a lower dimension and illustrating it. Although singular value decomposition is a grouping method different from association analysis, there are cases where it can find associations that cannot be found with association analysis, so it is useful to combine the results of both analyses. In this presentation, we will present an example of analyzing the association of side effects in patients administered a specific drug using the Japan Adverse Drug Reaction Database (JADER) published by the Pharmaceuticals and Medical Devices Agency (PMDA), using the "Association Analysis" platform, and demonstrate the usefulness of examining the singular value decomposition report.
