Menu Back to Poster-Presentations-Details

W 28: Development of a Matching Dictionary Between Lay and Corresponding Scientific Terms to Detect Web Reported Adverse Events

Poster Presenter

Manon Exposito

Pharmacist Project Manager
Universal Medica
France

Objectives

To develop a tool to detect public web reported adverses events (AEs) formulated in lay language via the use of a specifically designed dictionnary

Method

ORAL PRESENTATION SCHEDULED: Session 2B at 12:20- 12:30 PM

First, we translated PT MedDRA terms in French lay language, Then, after analysis of the actual messages, the opposite approach was executed. The result was the set up of a bidirectional dictionary.The validity of the dictionary was checked.

Results

About 25,000 terms in lay language have been linked to 1,000 PT MedDRA terms. The approach has been implemented upon request of an official body for three products (A, B, and C) with well-known safety profiles as described in the FAERS and MedEffet databases. An algorithm was designed to capture the adverse events in the messages and to translate the latter with the help of the dictionary. More than 4,000 web messages from close to 300 websites were analysed. 1. For Product A, the 15 leading adverse events described in the reports of the authorities have been detected on the Web messages and were correctly translated in PT MedDRA terms. Quantitatively, the frequency of the occurrence of adverse events displayed differences between lay and official reporting. Subjective and harmless side effects (e.g. weight increased) were reported more frequently on the Web. The results obtained with the algorithm were monitored by analysing all Web comments related to the Product A. Adverse events of nearly 90% of messages citing the Product A were correctly translated. Noteworthy, the algorithm was able to detect an unlisted event in the European SPC. 2. For Product B, adverse events described in the reports of the authorities have been detected on the Web messages. In addition, the algorithm was able to detect several cases of misuse (1.5%). 3. For the Product C, adverse events described in the reports of the authorities have been detected on Web messages. In addition, the algorithm was able to detect numerous off-label prescriptions (32%).

Conclusion

The use of our unique matching dictionary of patient lay expressions/wordings and PT MedDRA terms for our tool identifying adverse events on Web messages allowed us to reliably detect known adverse events but also new unlisted adverse events in the SPC and pharmacovigilance circumstances. Unlike scientific language, lay language is constantly changing which requires a work intensive and continuous enhancement of the dictionary by a group of professionals in pharmacovigilance. The use of this entirely original dictionary has enabled the development of a reliable tool not only for the characterization of adverse events reported on the Web messages but also to early detect misuse and off-label use making it a useful and valuable patient centric additional approach for pharmacovigilance activities. Additional authors : Katia HIMEUR, Rafi MARDACHTI.