OBJECTIVES: While traditional signal detection methods in pharmacovigilance are based on individual case safety reports, the use of social media data is emerging among regulators and industry. This study aims at assessing the reliability of signals from web-based patient's forums in France on 3 products (teriflunomide, insuline glargine, zolpidem). Signals detected from this source were compared with those from adverse events reporting system. METHODS: Data sources are patient's medical forums in France and the WHO adverse events reporting system (VigiBase®). Data were extracted from the Detec't database, composed of messages posted on patient's forums between 2005 and 2015. All drug-event pairs from the selected forums were included. Several metrics were used to define signal of disproportional reporting (PRR, ROR, EBGM, IC, RFET). Comparison of signals detected from the forums to signals detected in VigiBase® was done by describing the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and ROC curves. RESULTS: The number of posts from the patient's forums containing teriflunomide, insuline glargine, zolpidem and with the mention of intake were respectively: 102, 3326, 4584. The number of unique drug – adverse event pairs were respectively 33, 194, 318. The number of detected signals, according to all methods used for disproportionality, ranges, respectively, from 4 to 12, 21 to 48, and 23 to 95. Comparison analysis shows that, according to metrics used, sensitivity ranges from 29.1 to 50.6 %, specificity from 86.1 to 95.5 %, PPV from 51.2 to 75.4 %, NPV from 68.5 to 91.6 %, accuracy from 68 to 87.7%. CONCLUSIONS: Web-based signal detection using forums data in France appears to have good performance features compared to adverse events reporting system. Further analysis will explore time difference between the date of signals from the forums and date of signals from Vigibase.