Opportunity in Diversity: Merging Healthcare Databases in Europe for Safety Signal Detection in Pharmacovigilance

Background: The EU-ADR Project (Early Detection of Adverse Drug Events by Integrative Mining of Clinical Records and Biomedical Knowledge) explores the potential of combining EHR databases for drug safety signal detection. Objectives: One of EU-ADR’s first tasks is to develop and validate data mining techniques for signal detection. The association between NSAIDs and upper gastrointestinal bleeding (UGIB) was the first true positive signal in the validation exercise. Methods: Signal detection system in EU-ADR is based on mining of anonymous EHRs of over 30 M patients from 8 European DBs. The DBs are kept local and are linked through a virtual network. A dynamic cohort approach was adopted, with each DB specifying eligibility requirements for entry into the cohort study population. Using the EU-ADR Jerboa# software, data was extracted locally and pooled to allow data mining. Results: Crude incidence rates for UGIB ranged between 39.7 and 110.5 per 100,000 person-years in the 8 DBs. Out of a total of 37,753 first occurrences of UGIB during follow-up, 2,741 (72.6%) were found to be associated with the use of NSAIDs. Relative risks of NSAIDs ranged from 2.1 to 5.0, consistent with what has previously been cited in literature. Each DB was able to find the association between NSAIDs and UGIB. Conclusions: Testing of a true positive signal showed that all databases were valid in finding this signal upon automated mining.