“Social media serves as a relevant complementary source for pharmacovigilance, capturing patient perspectives often missed by spontaneous reporting.”

Luis Phillipe Nagem Lopes, Researcher at State University of Rio de Janeiro

Researchers are exploring how AI analytics could transform platforms like Reddit and X into large-scale drug safety monitoring tools, capable of surfacing adverse events long before they appear in traditional reporting systems.

A recent study in the journal Nature Health analysed over 400,000 Reddit posts discussing semaglutide and tirzepatide use. Along with the well-known gastrointestinal side effects, researchers also discovered less common side effects like menstrual irregularities and chills in these Reddit comments. Most of these experiences are rarely captured in structured clinical datasets or formal AE reports. The truth is, patients are continuously generating real-world safety data online, whether or not the industry is monitoring it.

Social media monitoring introduces the possibility of near real-time signal detection from unstructured patient conversations happening at enormous scale. Innovations in natural language processing are making this more feasible. AI can scan huge volumes of posts, identify symptom patterns and detect emerging safety concerns across online communities. These insights will be particularly valuable for subjective or underreported side effects like fatigue, mood swings or sexual issues, which many patients may discuss more openly online than during clinical trials.

However, there are major limitations to this approach. Social media data is messy, incomplete and sometimes impossible to validate. Posts usually lack dosage information, medical history or any real proof of causality. Additionally, slang, sarcasm and misinformation create additional challenges for AI’s interpretations of patient discussions. Furthermore, the internet is increasingly populated by bots, how do we know these safety signals come from real humans?

Despite this approach still being in its infancy, it is indicative of the direction of travel of the PV industry. As wearable devices and digital health tools continue generating massive volumes of patient data, PV is moving towards a far more continuous model of safety monitoring.