Comparison of drug safety signal detection between spontaneous reporting databases and eu-adr longitudinal database network

Introduction: In recent years, signal detection (SD) and data mining research has focused on leveraging longitudinal healthcare databases to detect signals. The EU-ADR project aims to exploit different European, longitudinal, electronic healthcare records databases for SD. The aim of this study is to compare SD results between FDA/WHO SRS databases and EU-ADR database network concerning six events (bullous eruptions- BE, acute renal failure-ARF, acute myocardial infarction-AMI, anaphylactic shock-AS, rhabdomyolysis-RHABD, and upper gastrointestinal bleeding-UGIB). Methods: The FDA and WHO SRS databases were mined from 1969 2010 for the MedDRA PTs of six events. Signal thresholds used were EB05 > 2 in presence of at least 1 report. Drugs that exceeded this threshold were identified as potential signals. EU-ADR platform consist of eight databases from four countries (Denmark, Italy, Netherlands, and United Kingdom), contributing data from 1996 - 2010. A custom-built software (Jerboa) elaborates locally harmonized input data and generates aggregated data which are subsequently analysed through different statistics (i.e.Longitudinal Gamma Poisson Shrinker). Drugs with statistically significantly (P < 0.05) increased posterior expectation of RR 2 have been identified as potential signals. Based on background incidence rate, for each event we estimated the required amount of exposure to test a drug as potential signal. Drugs not reaching this threshold were not included in the analyses. Results: SRSs could explore, as potential signals, a broader number of drugs for the six events, in comparison to EU-ADR (range: 630–3393 vs. 87–856), particularly for those events commonly thought to be potentially drug-induced (i.e.BE: 2053-3393 vs. 228). On the contrary, EUADR may investigate a greater number of drugs concerning AMI (856 vs. 630–791). The highest proportion of signals detected in SRSs was reported for BE, ARF and AS, while for ARF and UGIB in EU-ADR. Conclusions: EU-ADR longitudinal database network may complement traditional spontaneous reporting system for SD, especially for those adverse events that are frequent in general population and are not commonly thought to be drug-induced.