Gender-Related "Higher-Than-Expected" Drug-Event Combinations in Spontaneous Adverse Drug Event Reports
Ana Szarfman, M.D., Ph.D., QMR, Office of Biostatistics, CDER, FDA
The large Food and Drug Administration (FDA) Spontaneous Reporting System (SRS) database is the most important of its kind in the U.S. Its primary objective is for the study of adverse drug effects and it is used to identify and document rare events not detected during randomized controlled clinical trials. It contains over 1.8 million records collected between 1968 to the present. We used the full SRS database to apply a new empirical Bayesian (EB) data mining method developed by William DuMouchel (The American Statistician, August 1999). We detect Gender-Related Signal Scores (SS) by computing the EB estimates separately for each gender and then comparing them visually by gender, drug, classes of drugs, events, and groups of events. Our studies show that women have higher SS than men in association with severe adverse drug events, including "torsades de pointes", "QT prolongation", "agranulocytosis", "bleeding", "renal events", "pancreatitis", and "liver events" with a higher number of drugs. This higher association is independent of drug use by gender. In contrast, "anaphylaxis", Stevens Johnson Syndrome", or "toxic epidermal necrolysis" do not show that higher SS are associated with a specific gender. Our results support the existence of higher SS in women in association with severe adverse drug events with old and newly marketed drugs. *Supported with a grant from OWH, OC, FDA.