Masahiko Gosho1, Kazushi Maruo2, Keisuke Tada3, Akihiro Hirakawa4. 1. Department of Clinical Trial and Clinical Epidemiology, Faculty of Medicine, University of Tsukuba, 1-1-1, Tennodai, Tsukuba, Ibaraki, 305-8575, Japan. mgosho@md.tsukuba.ac.jp. 2. Department of Clinical Epidemiology, Translational Medical Center, National Center of Neurology and Psychiatry, 4-1-1 Ogawa-Higashi, Kodaira, Tokyo, 187-8551, Japan. 3. Biostatistics & Programming, Sanofi K. K, Tokyo Opera City Tower, 3-20-2, Nishi Shinjuku, Tokyo, 163-1488, Japan. 4. Center for Advanced Medicine and Clinical Research, Nagoya University Hospital, 65 Tsurumai-cho, Showa-ku, Nagoya, 466-8560, Japan.
Abstract
PURPOSE: We proposed a statistical criterion to detect drug-drug interactions causing adverse drug reactions in spontaneous reporting systems. METHODS: The used criterion quantitatively measures the discrepancy between the observed and expected number of adverse events via chi-square statistics. We compared the performance of our method with that of Norén et al. (Stat Med 2008; 27 (16): 3057-3070) through a simulation study. RESULTS: When the number of events for a combination of two drugs was equal to or lower than two, the false positive rate for our method ranged from 0.01 to 0.08, whereas the rate for Norén's method ranged from 0.01 to 0.06. The sensitivity for our method ranged from 0.09 to 0.29, whereas the sensitivity for Norén's method ranged from 0.03 to 0.24. The area-under-the-receiver operating characteristic curve for our method was significantly larger than that for Norén's methods regardless of simulation settings. The proposed method was also applied to the Food and Drug Administration Adverse Event Reporting System database, and a recognized drug-drug interaction was detected. CONCLUSIONS: The proposed criterion controlled false positives at an acceptable level and had higher sensitivity than that of Norén's method had when events were rare.
PURPOSE: We proposed a statistical criterion to detect drug-drug interactions causing adverse drug reactions in spontaneous reporting systems. METHODS: The used criterion quantitatively measures the discrepancy between the observed and expected number of adverse events via chi-square statistics. We compared the performance of our method with that of Norén et al. (Stat Med 2008; 27 (16): 3057-3070) through a simulation study. RESULTS: When the number of events for a combination of two drugs was equal to or lower than two, the false positive rate for our method ranged from 0.01 to 0.08, whereas the rate for Norén's method ranged from 0.01 to 0.06. The sensitivity for our method ranged from 0.09 to 0.29, whereas the sensitivity for Norén's method ranged from 0.03 to 0.24. The area-under-the-receiver operating characteristic curve for our method was significantly larger than that for Norén's methods regardless of simulation settings. The proposed method was also applied to the Food and Drug Administration Adverse Event Reporting System database, and a recognized drug-drug interaction was detected. CONCLUSIONS: The proposed criterion controlled false positives at an acceptable level and had higher sensitivity than that of Norén's method had when events were rare.
Keywords:
Adverse event reporting system; False positive; Sensitivity; Signal detection
Authors: A Bate; M Lindquist; I R Edwards; S Olsson; R Orre; A Lansner; R M De Freitas Journal: Eur J Clin Pharmacol Date: 1998-06 Impact factor: 2.953
Authors: Srinivasan V Iyer; Rave Harpaz; Paea LePendu; Anna Bauer-Mehren; Nigam H Shah Journal: J Am Med Inform Assoc Date: 2013-10-24 Impact factor: 4.497
Authors: David J Graham; Judy A Staffa; Deborah Shatin; Susan E Andrade; Stephanie D Schech; Lois La Grenade; Jerry H Gurwitz; K Arnold Chan; Michael J Goodman; Richard Platt Journal: JAMA Date: 2004-11-22 Impact factor: 56.272