Can the EU-ADR Database Network Detect Drug Safety Signals Faster than Spontaneous Reporting System?

Background: Several ongoing initiatives are exploring if mining electronic health records (EHRs) may fasten the process of drug safety signal detection. We investigated if the signal concerning rofecoxib and acute myocardial infarction could have been identified in EU-ADR database network faster than spontaneous reporting system (SRS), and earlier than rofecoxib withdrawal (30th September 2004). Methods: EU-ADR distributed network comprises of seven EHR databases covering approximately a population of almost 30 million persons from four European Countries during the years 1996–2010. Harmonized data extraction and analysis has been conducted in all databases through custom-built software Jerboa, which allows for data aggregation and elaboration while databases remained locally. A signal was defined as statistically significant (p-value <0.05) and at least two-fold increase in the risk for AMI with rofecoxib, as compared to all the other drugs, using Longitudinal Gamma Poission Shrinkage. Concerning spontaneous reporting system, FDA AERS database and as EB05 = 2 as threshold for signal detection have been considered. Stratifying by consecutive quadrimestre till rofecoxib withdrawal, we measured the first point in time in which signal concerning rofecoxib and AMI would have been identified in EU-ADR network and FDAAERS. On average, one year time gap should be also considered for ta transferring in EU-ADR network. Results: A total of 685 cases exposed to rofecoxib were captured in EUADR network during the years 2000–2004. Since the third quadrimestre of 2000, EU-ADR network would have been able to identify a strong signal concerning rofecoxib and AMI as 847 609 person-days of exposure to this drug was available. In the same quadrimestre a Relative Risk = 4.46 (95%Confidence Interval: 2.84-6.72) was found. With respect to FDA database a strong signal could not be identified. The number of reports of AMI due to rofecoxib increased substantially after the drug withdrawal as a results of lawyers reporting. Discussion: The use of a system such as EU-ADR could have led to a faster rofecoxib removal from the market. In general, these systems may be helpful in the strengthening of signals identified using other sources. Conclusions: EU-ADR database network would have been able to detect strong signal concerning rofecoxib and acute myocardial infarction around three years earlier than rofecoxib withdrawal and faster than SRS. System such as EU-ADR may be a valid tool for signal strengthening concerning events such as acute myocardial infarction.