P213: Future Fit Pharmacovigilance - Deploying AI and ML Under Regulated Environment
Poster Presenter
Anuradha Roy
Senior Vice President
One Vector United States
Objectives
As FDA inspections sharpen their focus on data integrity, real-time reporting, and digital compliance, life sciences must transition from reactive compliance to proactive safety governance, this is possible by using AI & automation. This abstract provides an structured approach and design for AI strategy, data, infrastructure, and operations to implement AI tools for PV.
Method
Combining automation, real-world evidence, and predictive intelligence with structured oversight, explainability, and lifecycle monitoring moves reactive compliance to proactive risk leadership for Pharmacovigilance scientists. The approach is first create an AI roadmap based on the user needs, existing process, data foundation, security requirements , cost and benefit analysis. Identify a mix of few simple and complex use cases to implement a pilot in collaboration with a technology partner. AI & automation can dramatically strengthen signal detection, compliance monitoring, and documentation consistency when embedded within a structured governance framework.
Results
An analysis of top FDA observations indicate common deficiencies namely late/ incomplete Adverse Event Reporting, poor SOP adherence, lack of written procedures, signal oversight, data integrity issues, vendor oversight lapses, non-compliance with electronic records regulations.
Possible cause of these deficiencies are manual processes, paper-based processes or siloed systems, inadequate tracking of changes made, etc. Many of these deficiencies can be addressed by automation and embedding AI / ML tools in operational processes.
AI & automation can dramatically strengthen signal detection, compliance monitoring, and documentation consistency. But only when embedded within a structured governance framework. A scalable model focusses not only on automation but on a governed approach to AI/ ML use case integration.
The next challenge is operationalizing ICSR intake & triage, case processing, signal detection, etc., into standardized, reliable, repeatable, scalable PV systems that run in production and deliver measurable value, while keeping human in the loop. Data sources are identified, pipelines are automated, data drift and quality issues are tracked, triaged and resolved. Data is then migrated to production environments. After production roll out data is monitored, maintenance and adjusted as needed. Governance, controls and policies are put in place to ensure audit trails, security, privacy and access controls are met.
Conclusion
Leveraging AI tools to fast track and optimize PV operations is no longer a luxury but is existential. Combining automation, real-world evidence, and predictive intelligence with structured oversight, explainability, and lifecycle monitoring will move from reactive compliance to proactive risk leadership.
AI is transforming PV by automating adverse drug reaction detection, enhancing signal
identification, and enabling faster, data-driven safety assessments. Key benefits include efficiency gains from reduction in manual processing and accelerating safety signal identification, improvement in data precision through duplicate detection and advanced analytics, Rela time monitoring , and providing data driven insights for decision making
Forward looking companies are partnering with Data and AI technology partners to create a structured, multi-stage roadmap that aligns strategy, data, infrastructure, and operations.
