Ana F Macedo1, Francisco B Marques, Carlos F Ribeiro. 1. Administração Regional de Saúde do Centro, Núcleo de Farmacovigilância do Centro, Faculdade de Medicina, Faculdade de Farmácia, Universidade de Coimbra, Coimbra, Portugal. filipa@fcsaude.ubi.pt
Abstract
AIM: The usefulness of algorithms for assessing the causality of suspected adverse drug reactions (ADRs) has yet to be established and, since the validation of causality algorithms depends upon their sensitivity and specificity, our study was carried out to evaluate these measures. METHOD: In this study, an expert panel assessed causality of adverse reports by using the WHO global introspection (GI) method. The same reports were independently assessed using 15 published algorithms. The causality assessment level 'possible' was considered the lower limit for a report to be considered to be drug related. For a given algorithm, sensitivity was determined by the proportion of reports simultaneously classified as drug related by the algorithm and the GI method. Specificity was measured as the proportion of reports simultaneously considered non-drug related. The analysis was performed for the total sample and within serious or unexpected events. RESULTS: Five hundred adverse reports were studied. Algorithms presented high rates of sensitivity (average of 93%, positive predictive value of 89%) and low rates of specificity (average of 7%, negative predictive value of 31%). CONCLUSION: Decisional algorithms are sensitive methods for the detection of ADRs, but they present poor specificity. A reference method was not identified. Algorithms do not replace GI and are not definite alternatives in the individual causality assessment of suspected ADRs.
AIM: The usefulness of algorithms for assessing the causality of suspected adverse drug reactions (ADRs) has yet to be established and, since the validation of causality algorithms depends upon their sensitivity and specificity, our study was carried out to evaluate these measures. METHOD: In this study, an expert panel assessed causality of adverse reports by using the WHO global introspection (GI) method. The same reports were independently assessed using 15 published algorithms. The causality assessment level 'possible' was considered the lower limit for a report to be considered to be drug related. For a given algorithm, sensitivity was determined by the proportion of reports simultaneously classified as drug related by the algorithm and the GI method. Specificity was measured as the proportion of reports simultaneously considered non-drug related. The analysis was performed for the total sample and within serious or unexpected events. RESULTS: Five hundred adverse reports were studied. Algorithms presented high rates of sensitivity (average of 93%, positive predictive value of 89%) and low rates of specificity (average of 7%, negative predictive value of 31%). CONCLUSION: Decisional algorithms are sensitive methods for the detection of ADRs, but they present poor specificity. A reference method was not identified. Algorithms do not replace GI and are not definite alternatives in the individual causality assessment of suspected ADRs.
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: Shaun M Comfort; Bruce Donzanti; Darren Dorrell; Sunita Dhar; Chris Eden; Francis Donaldson Journal: Drug Saf Date: 2022-10-23 Impact factor: 5.228
Authors: Sven Schmiedl; Marietta Rottenkolber; Joerg Hasford; Dominik Rottenkolber; Katrin Farker; Bernd Drewelow; Marion Hippius; Karen Saljé; Petra Thürmann Journal: Drug Saf Date: 2014-04 Impact factor: 5.606
Authors: Stefanie Amelung; David Czock; Markus Thalheimer; Torsten Hoppe-Tichy; Walter E Haefeli; Hanna M Seidling Journal: J Clin Med Date: 2022-07-23 Impact factor: 4.964
Authors: Joanna E Klopotowska; Peter C Wierenga; Clementine C M Stuijt; Lambertus Arisz; Marcel G W Dijkgraaf; Paul F M Kuks; Henk Asscheman; Sophia E de Rooij; Loraine Lie-A-Huen; Susanne M Smorenburg Journal: PLoS One Date: 2013-08-05 Impact factor: 3.240