Session (CTI)

The panel will discuss global regulatory guidance on the use of artificial intelligence (AI) to support regulatory decision-making. The increasing prevalence of AI and its intersection with real-world data (RWD) is fundamentally changing how evidence is generated and assessed. AI’s ultimate value relies on extracting meaningful insights from complex RWD sources. A vast amount of this RWD exists in unstructured formats such as clinician notes and pathology reports. There is a pressing need to curate this data using natural language processing and other AI techniques. These techniques must be of sufficient quality so that the resulting analytic dataset can inform regulatory decision-making. The moderator will provide an overview of global guidances from authorities such as EMA, NICE, and CDA and they compare. Following the introduction, panelists will debate how best to operationalize the guidance, what is still needed from a regulatory perspective, and what biopharma needs in terms of assurances from data providers to meet regulatory expectations. The panelists will provide two differing company approaches to dissecting and implementing FDA’s 7-step process for establishing and assessing the credibility of an AI model for a specific context of use and the challenges they have encountered. A key piece of the discussion will focus on assessing data quality for data generated through this approach. Panelists will also discuss the need for more information from regulatory authorities on how to best operationalize certain aspects of the guidances, how this scheme may conflict with existing regulatory schemes such as 21 CFR Part 11, and what is needed from an international convergence or harmonization perspective. This session will equip HEOR professionals with multiple perspectives on how to navigate the technical and regulatory landscape of AI abstraction and curation approaches, ensuring that evidence generation strategies and HTA submissions create the best possible body of evidence.