On the 20th of September this year, representatives of the organisations participating in the WHO PIDM met virtually to discuss signal detection and how they may better collaborate to strengthen pharmacovigilance. Chaired by the pharmacovigilance team of WHO, the meeting also covered how safety-related messaging and communication could be improved in terms of content and timeliness of delivery. While traditional methods were discussed for both themes, the main topic of the day was how social media may be harnessed to make signal detection and communication of medicine safety more effective.
Johan Ellenius of Uppsala Monitoring Centre (UMC) spoke on the efficacy of social media for signal detection compared to more traditional means, that is, disproportionality analysis of spontaneous report data. To do this, he revisited the findings of the WEB-RADR study of 2019, where they built several algorithms (involving Facebook and Twitter) to identify mentions of drugs and adverse events for signal detection. As part of his talk, Ellenius gave some anecdotal evidence from an informal investigation on the use of AI in the form of ChatGPT3.5 and ChatGPT4 to detect drug names and potential side effects of these drugs from social media posts. Their results for this analysis illustrated the potential of this new technology to analyse social media.
However, the conclusion from the WEB-RADR study, that social media based on Twitter and Facebook as a resource for broad-ranging signal detection is not worthwhile, is still valid as the number of different drugs that people discuss on those platforms was found to be quite limited. However, evidence from social media could serve as a complement to spontaneous report data by providing more context on the reports, such as the impact on patients’ quality of life. Ellenius added that further exploration of AI and its use in analysing social media posts in this field is ripe for further research.
This sentiment was reflected in other talks in the meeting by the pharmacovigilance team of WHO on their use of Pulsar and PEEK, two social media listening tools, to detect potential medicine and vaccine side effects more quickly than traditional pharmacovigilance methods. They used the example of the launch of the COVID-19 vaccine in the UK, where they used Pulsar to gauge the public’s reaction to the vaccine and developed a social media communication strategy on this data to help curb the spread of vaccine misinformation and vaccine hesitancy going forward.
They also worked with PEEK to develop an ‘early warning’, event-based surveillance system that serves as a complement to information from other routine surveillance systems. Continuing with the COVID-19 vaccine as an example, this system detected a potential side-effect of nephrotic syndrome following the administration of the second dose of the COVID-19 vaccine in early 2022. Interestingly, UMC researchers had detected a similar signal in VigiBase nearly a year before, where IgA nephropathy was linked to the COVID-19 vaccine. The alert provided by this PEEK early warning system gave more context on this signal, helping them see the bigger clinical picture for this side effect and the vaccine.
The final talk of the meeting, held by UMC researcher Annette Rudolph, which covered this case from UMC’s perspective, came to the same conclusion as the first talk, that social media is best suited as a complement to classical signal detection methods, and further process development and implementation strategies are required to make this work effectively.