501 – Methods for Safety, Quality, and Benefit-Risk
Mining Association Rules in Clinical Trial Safety Data
Li Zhu
Amgen
Amy Xia
Amgen
Qi Jiang
Amgen
William Go
Amgen
Liron Walsh
Amgen
Clinical trial safety data are routinely analyzed to help the drug developer gain knowledge on the safety profile of an investigational drug, and detect potential safety signals in the pre-marketing stage. Adverse events (AEs) are typically summarized in incidence tables, tested on two-by-two contingency tables, or analyzed through drug-AE model that are based on aggregated data. However, although regularly collected in clinical trial data, the information about within-subject correlations is often ignored when AEs are analyzed in such aggregated fashions. Association rule is a typical statistical learning technique that has been commonly used for mining commercial databases. We investigate the application of association rule in mining clinical trial AE data. Our results demonstrate interesting findings on drug-AE associations in both one-to-one and one-to-many mappings, associations between AEs and patient characteristics, and a new way of grouping AEs that is alternative to traditional approaches such as MedDRA hierarchical coding or standard MedDRA queries search strategies.