Premium
Selected control events and reporting odds ratio in signal detection methodology
Author(s) -
Ooba Nobuhiro,
Kubota Kiyoshi
Publication year - 2010
Publication title -
pharmacoepidemiology and drug safety
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.023
H-Index - 96
eISSN - 1099-1557
pISSN - 1053-8569
DOI - 10.1002/pds.2014
Subject(s) - medicine , confidence interval , odds ratio , incidence (geometry) , pharmacovigilance , event (particle physics) , control (management) , statistics , pharmacoepidemiology , meddra , demography , adverse effect , artificial intelligence , computer science , pharmacology , mathematics , quantum mechanics , sociology , medical prescription , physics , geometry
Purpose To know whether the reporting odds ratio (ROR) using “control events” can detect signals hidden behind striking reports on one or more particular events. Methods We used data of 956 drug use investigations (DUIs) conducted between 1970 and 1998 in Japan and domestic spontaneous reports (SRs) between 1998 and 2008. The event terms in DUIs were converted to the preferred terms in Medical Dictionary for Regulatory Activities (MedDRA). We calculated the incidence proportion for various events and selected 20 “control events” with a relatively constant incidence proportion across DUIs and also reported regularly to the spontaneous reporting system. A “signal” was generated for the drug‐event combination when the lower limit of 95% confidence interval of the ROR exceeded 1. We also compared the ROR in SRs with the RR in DUIs. Results The “control events” accounted for 18.2% of all reports. The ROR using “control events” may detect some hidden signals for a drug with the proportion of “control events” lower than the average. The median of the ratios of the ROR using “control events” to RR was around the unity indicating that “control events” roughly represented the exposure distribution though the range of the ratios was so diverse that the individual ROR might not be regarded as the estimate of RR. Conclusions The use of the ROR with “control events” may give an adjunctive to the traditional signal detection methods to find a signal hidden behind some major events. Copyright © 2010 John Wiley & Sons, Ltd.