2023-11-05-世界卫生组织-Value_of_social_media_and_other_online_listening_posts_in_Pharmacovigilance_13页_671kb

WHO-UMC Pharmacovigilance Meeting Summary: Value of Social Media and Other Online Listening Posts

Background and Objective

The meeting explored the role of social media and online listening posts in pharmacovigilance (PV), building on pandemic-related experiences. It addressed signal detection gaps, data quality issues, and the need for improved safety communication. The WHO Programme for International Drug Monitoring (PIDM), supported by the Uppsala Monitoring Centre (UMC), has historically focused on spontaneous reporting, but social media provided complementary insights during COVID-19, despite limitations in broad signal detection.

Methods for Signal Detection

Traditional methods include disproportionality analysis in the UMC database Vigibase, which detects disproportionate reporting patterns. Advanced tools developed by UMC include:

• Vigipoint: Detects signals in subgroups (e.g., elderly patients) for drugs like ceftriaxone.
• Vigirank: Incorporates additional data (e.g., narrative availability, reporting source) for enhanced prediction.
• Vigigroup: Groups similar reports for cluster analysis to avoid fragmented assessments.
  Social media analysis via projects like Web-RADR showed inferior performance compared to spontaneous reports for broad signal detection, but AI tools (e.g., ChatGPT variants) may help refine signals and detect personal experiences missed in traditional reports.

Case Study: COVID-19 Signal Detection

During the pandemic, WHO PVG used social media platforms (e.g., Pulsar, PEEK) to monitor and verify potential signals, such as:

• Anaphylaxis rumors after mRNA vaccines, addressed through social media sentiment analysis.
• Glomerulonephritis/nephrotic syndrome signals with mRNA vaccines (Pfizer and Moderna), involving 56 cases identified through UMC analysis, showing dose-dependent onset.

Discussion and Findings

Participants shared lessons from COVID response:

• 83% faced challenges verifying misinformation via social media and traditional channels.
• 74% did not use social media for signal detection, citing resource constraints or AI needs.
• Key suggestions include training on credibility assessment, safety communication, social media monitoring, and preparing for new interventions with AI tools.

Conclusions

• Spontaneous reporting and social media offer complementary approaches; classic methods remain gold standard for rare signals.
• AI advancements hold promise for improving social media analysis.
• Future efforts should focus on ethical AI use, training, and tools for real-time risk communication to enhance global PV operations.