Implementation of an automated signal detection method in the French pharmacovigilance database: a feasibility study

In France, early detection of adverse effects does not currently involve any automatic signal detection method. The present objective was to assess the feasibility and measure the potential benefit of the incorporation of an automatic signal detection tool (GPS(pH0)) in the French pharmacovigilance system.

GPS(pH0) was first applied to the data collected from 1 January 2000 to 31 December 2008 and then to the data collected from 1 January 2000 to 31 March 2009. A total of 1,414 original signals were detected. They were shared out for further expertise among 32 centres, i.e. the 31 Regional Pharmacovigilance Centres and the French medicine agency (AFSSAPS) pharmacovigilance department.

The participating centres (n = 28) analysed 1,292 signals in May 2009. Overall, 277 signals whether known or unknown were thus considered worth following up. Half of the other 893 categorised signals were "well-known" (35.7%) and non-interpretable/non-pertinent signals (36.6%); 4% were not categorised because of a lack of time. Analysis of the signals was time-consuming, but the working time estimated by the participants was highly variable (median time: 6 h; minimum: 2 h maximum: 26 h).

The results of this study are in favour of the integration of an automated signal detection tool to complement the current pharmacovigilance activities. The Anatomic Therapeutic Chemical for drug classification poses difficulties in many situations; the international proprietary name might be more efficient. The variability observed in the time needed for analysis suggests that a standardised methodology should be employed. Overall, the findings of this prospective study will contribute to refining the signal management procedure to be implemented in the future.