P212: AI Automating Multimodal Evidence Integration for Drug Safety Signal Causality Assessment
Poster Presenter
Tarek Hammad
Vice President, Global Head of Medical Safety, Marketed Products and PDT, PSPV
Takeda Pharmaceuticals United States
Objectives
To automate a recently published holistic causality assessment framework for enhanced efficiency, accuracy, and standardization in evaluating causality of drug safety signals.
Method
Design of a Proof of Concept plan to create an AI-based algorithm to replicate the Hammad-Afsar framework, using multimodal evidence integration and weighted probabilistic approaches to automate causality assessments of drug safety signals, will be trained and validated using historical PV data.
Results
The proposed AI tool is expected to significantly reduce time required for causality assessments, achieving an estimated 40% reduction in manual effort (approximately 1,256 hours saved annually based on historical data of the average number of safety signal encountered per year in a pharmaceutical company). The algorithm aims to integrate structured and unstructured data from individual reports and aggregate safety datasets, supporting consistent, scalable, and unbiased signal confirmation.
Conclusion
The AI-enhanced causality assessment method, utilizing the holistic approach detailed in the recently published Hammad-Afsar framework*, addresses limitations in current pharmacovigilance practices by automating evidence integration and standardizing application of weighted evidence. This approach would improve regulatory compliance, optimize resource utilization, and ensure systematic, objective evaluations of drug safety signals. These advancements are poised to transform causality assessment processes in drug safety management.
* Hammad TA et al (2023), Aspects to consider in causality assessment of safety signals: broadening the thought process. Front. Drug. Saf. Regul. 3:1193413. doi: 10.3389/fdsfr.2023.1193413.
