FDA Safety Alerts - Looking for Early Signals in Social Media

February 01, 2017

FDA Safety Alerts - Looking for Early Signals in Social Media: Collaborative Research by Epidemico, Booz Allen, FDA, and WHO-UMC

Previous studies have demonstrated how social media data can be used to augment existing pharmacovigilance processes by providing patient-generated insights on safety issues1, 2. Our most recently published paper, "Evaluation of Facebook and Twitter Monitoring to Detect Safety Signals for Medical Products: An Analysis of Recent FDA Safety Alerts", featured in Drug Safety, expands on this work by investigating whether social media can be used to identify rare and serious safety signals earlier than traditional reporting channels.

In this retrospective study, we worked closely with collaborators from the US FDA and Uppsala Monitoring Centre (WHO-UMC) to assess causality in posts with resemblance to adverse events (Proto-AEs), and to explore the temporal relationship between relevant Proto-AEs and 10 established MedWatch alerts.

Product-event pairs selected for analysis
Table 1: Product-event pairs selected for analysis

Using our MedWatcher Social system, we automatically collected, classified, and coded publicly available posts from Twitter and Facebook from up to two years prior to when each safety alert had been established by FDA.13 posts were determined to be relevant and were manually reviewed according to WHO-UMC's established criteria for causality assessment. Ultimately, six of the 13 posts that represented two of the safety alerts (dronedarone-vasculitis and Banana Boat Sunscreen-skin burns) received high causality scores: possible, probable, or certain. Dates of these more probable social media cases were then compared to both the date that the first report was received by FAERS and the date when the FDA reported signal detection. In one of these cases, the first report occurred in social media prior to signal detection from FAERS (Banana Boat Sunscreen-skin burns), while the other case occurred first in FAERS (dronedarone-vasculitis).

This study applied strict criteria to determine the value of social media data -- both in our use of UMC's causality assessment methodology and in our decision to focus on serious and sometimes rare adverse events. Despite the strict criteria, there were earlier signals in social media for rare and serious adverse events that we were able to detect. This research complements our previous study2, which showed highest volume of social media Proto-AEs discussing general disorders that are typically less serious (e.g. abdominal symptoms), upon which the question remained whether social media could be helpful in detecting more serious and rare safety issues as well. Beyond the results, this study furthered our understanding of the kind of AEs that are typically discussed by patients in social media and the language that patients use to describe these AEs, underscoring the potential value in operationalizing the monitoring of social media in pharmacovigilance in addition to spontaneous reports.

1  Powell, G.E., Seifart, H.A., Reblin, T., Burstein, P.J., Blowers, J., Menius, J.A…Dasgupta, N. (2016). Social Media Listening for Routine Post-Marketing Safety Surveillance. Drug Safety, 39(5), 443-454. doi:10.1007/s40264-015-0385-6
2 Freifeld, C.C., Brownstein, J.S., Menone, C.M., Bao, W., Filice, R., Kass-Hout, T., & Dasgupta, N. (2014). Digital Drug Safety Surveillance: Monitoring Pharmaceutical Products in Twitter. Drug Safety, 37(5), 343-350. doi:10.1007/s40264-014-0155-x