The pharmaceutical industry is expanding its use of artificial intelligence (AI) and other emerging technology across the drug product lifecycle. The Emerging Drug Safety Technology Meeting (EDSTM) program offers applicants with an approved application, and/or other relevant parties (e.g., academia, contract research organizations, pharmacovigilance (PV) vendors, software developments) who meet the eligibility and selection criteria, with an opportunity to meet with CDER staff to discuss their research, development, and use of AI and other emerging technology in PV. PV is the science and activities relating to the detection, assessment, understanding, and prevention of an adverse event (AE) or any other medicine/vaccine related problem. All medicines and vaccines undergo rigorous FDA testing for safety and efficacy through clinical trials before they are authorized for use.

In this CDER Conversation, Robert Ball, M.D., M.P.H., ScM, Deputy Director of CDER’s Office of Surveillance and Epidemiology, explains the EDSTM program and the application of AI in PV based on his experience of more than a decade researching and developing AI systems in PV.

What is the EDSTM program seeking to accomplish?

In the initial phase, EDSTMs will provide a forum to facilitate mutual learning and discussion of participants’ application of emerging technology in pharmacovigilance (PV). FDA plans to leverage these learnings to inform potential regulatory and policy approaches within PV and consider providing regulatory advice on specific technologies to facilitate their adoption, when appropriate.

How have you observed artificial intelligence (AI) being used for PV and what are the challenges?

At present, one way the pharmaceutical industry has described its application of AI is for processing of individual case safety reports (ICSRs). Case processing includes intake, evaluation, follow up, and the submission of ICSRs to regulatory authorities. There is great interest in applying AI to case causality assessment, which is the determination of whether there is a reasonable possibility that a drug caused the reported AE. Case causality assessment is much more challenging than case processing because it involves more expert judgment and the application of highly specialized knowledge.

Currently AI cannot replicate or make as nuanced judgments as a human expert, leading to inadequate performance in ICSR evaluation, such as in predicting whether a drug may have caused an AE. In other tasks, such as identifying the reporter information during ICSR intake, AI might be adequate, though its performance still is not at the threshold needed for full automation.

The use of emerging technology, such as AI, has the potential to bring about transformative benefits for the way organizations approach PV by lowering administrative burdens and costs, and improving the efficiency and effectiveness of safety surveillance. However, accomplishing this may involve substantial investment in technologies that still bear significant amounts of uncertainty around aspects such as the credibility and trustworthiness of AI models, and how they may be integrated into existing PV workflows.

Through the EDSTM program, CDER set out to establish a learning environment, which will enable CDER and participants to engage in mutual learning of how emerging technology can be applied in PV. We expect that the knowledge gained through EDSTMs will help inform potential future regulatory and policy approaches related to the challenges surrounding emerging technology and help foster confidence in their continued investment.