K-24
Construction of a Human Adverse Effects Database for Modeling Quantitative Structure-Activity Relationships (QSARs)
N. L. Kruhlak1, E. J. Matthews1, J. L. Weaver2, R. D. Benz1, J. F. Contrera1, 1FDA/CDER/OPS/ICSAS, Rockville, MD, 2FDA/CDER/OTR/DAPR, Silver Spring, MD
DA/CDER's Informatics and Computational Safety Analysis Staff (ICSAS) is conducting a pilot study to determine whether computational toxicology methods can be developed to predict the potential human adverse effects of pharmaceuticals based on post-marketing report data. The adverse effect data chosen for this investigation were extracted from FDA/CDER's Spontaneous Reporting System (SRS) database and included over 1.6 million adverse drug reaction (ADR) reports for 8620 drugs and biologics listed for 1191 Coding Symbols for Thesaurus of Adverse Reaction (COSTAR) terms. The ICSAS Adverse Effects database contains the generic names and molecular structures for a subset of 1515 organic chemicals that are suitable for modeling with commercially available QSAR software packages. ADR reports from the SRS database were pooled for the first five years that a pharmaceutical was marketed. To estimate patient exposure during this period, shipping units for each pharmaceutical were used as a denominator, resulting in the creation of the Adverse Effect Index (AEI), where AEI = (# ADR reports/# shipping units) × 1,000,000. A pharmaceutical defined as active for a single COSTAR endpoint is characterized by 4 ADR reports, >20,000 shipping units, and an AEI 4.0 during the first five years of marketing. Furthermore, test chemicals evaluated as active must contain a statistically significant structural alert at two or more toxicologically related endpoints. We also report the use of a combination QSAR module which pools observations from two or more toxicologically related COSTAR term endpoints to provide signal enhancement for poorly represented adverse effects.
2004 FDA Science Forum | FDA Chapter, Sigma Xi | CFSAN | FDA
Last updated on 2004-APR-02 by frf