Audio Transcript | The Need for Artificial Intelligence in Pharmacovigilance with Dr. Robert Ball
CDER Small Business and Industry Assistance Chronicles Podcast
Listen to the The Need for Artificial Intelligence in Pharmacovigilance and the Emerging Drug Safety Technology Program audio podcast.
Transcript
Dr. Weber: Welcome to the CDER Small Business and Industry Assistance (SBIA) Chronicles Podcast.
Today’s topic: The Need for Artificial Intelligence in Pharmacovigilance
My name is Dr. Ellicia Weber and today we are joined by Dr. Robert Ball, Deputy Director of the Office of Surveillance and Epidemiology within FDA’s Center for Drug Evaluation and Research. Dr. Ball has over a decade of experience researching and developing artificial intelligence, or AI, systems, in pharmacovigilance.
Thank you for joining us Dr. Ball! I’d love to start off hearing about your experience in AI and pharmacovigilance.
Dr. Ball: Thank you for having me today. Going way back, in college I studied math and computer science and afterwards worked as a computer programmer before attending medical school. Since then, much of my work has been at the interface of computational science and medicine. More than 10 years ago now, I got interested in applying natural language processing to narratives in vaccine adverse event reports submitted to FDA and have continued that work as the AI technology has advanced.
Dr. Weber: FDA recently announced the Emerging Drug Safety Technology Program, which is meant to facilitate discussion and mutual learning about the application of AI to pharmacovigilance. But we’ll get to that in a bit. Let’s start with pharmacovigilance and how it can benefit from AI.
Dr. Ball: Right. After a drug has been approved, new safety issues may arise. Pharmacovigilance involves all the scientific and data gathering activities related to the detection, assessment, and understanding of these adverse events. This process includes a wide range of scientific methods and data sources.
So, what if we could analyze this vast amount of information faster? AI offers the potential to do just that. In fact, one area the pharmaceutical industry is exploring is the use of AI to process Individual Case Safety Reports, or ICSRs. These reports serve as an important source of new safety information and are required by regulators globally.
By applying AI, we could process data from multiple sources rapidly, potentially helping to identify adverse events more efficiently. AI may also be able to help companies prepare ICSRs, allowing them to be reported and evaluated faster.
Dr. Weber: Why ICSRs?
Dr. Ball: ICSRs serve as an early warning system of safety signals, especially for rare events. With the increasing number and variety of data sources that need evaluation, more and more ICSRs are being submitted every day. In fact, every year FDA receives nearly two million reports from industry to the FDA Adverse Event Reporting System, or FAERS. Beyond that, several hundred thousand reports are submitted from the public directly to FDA. Due to the sheer volume of reports, case processing is a big challenge.
AI could be beneficial to this process for both FDA and industry. It might be able to help automate tasks like identifying cases required to be submitted to the FDA or determining the seriousness of events. This could not only save time, but it could also allow our safety evaluators to focus more on complex tasks that have a public health impact, rather than on extracting and organizing information from ICSRs.
Dr. Weber: Sounds like human oversight remains crucial, is that an accurate statement?
Dr. Ball: Absolutely! While AI technology holds great promise for pharmacovigilance and may be able to play a significant role, it’s essential not to overlook the importance of human expertise. There are concerns about potential shortcomings in AI models that could lead them to miss crucial information. So human oversight remains indispensable.
The challenge here is finding the right balance between including a “human in the loop” as we say, to ensure quality without compromising efficiency gains from AI algorithms. Said another way, a human expert should not do the work that a machine can do well and efficiently, and a machine should not do poorly the work that the human expert can do well.
For some tasks, like identifying the reporter’s occupation during ICSR intake, AI might be adequate, but an example of a task where human expertise is important is case causality assessment. This process involves expert judgement and the application of highly specialized knowledge.
AI models are not yet able to replicate or make as nuanced judgements as a human expert, which can lead to inadequate performance.
Because of this, quality checks will need to be implemented to ensure that a combined human-AI system is at least as effective as those relying solely on humans.
Dr. Weber: How will the FDA strike the right balance between leveraging AI’s potential in pharmacovigilance and ensuring its safe application?
Dr. Ball: Yeah that’s an important question. One of the ways we’re looking to strike that balance is by collaborating and engaging in dialogue with groups outside of the agency. For example, we’re currently involved with international colleagues through initiatives such as the Council for International Organizations of Medical Sciences (or CIOMS) working group on AI in pharmacovigilance, which is developing principles for the use of AI in pharmacovigilance.
Another way we’re seeking to do this is by engaging in direct dialogue with industry. To that end, we’ve launched the Emerging Drug Safety Technology Program to help foster discussions with participants on their use of emerging technology in pharmacovigilance.
Dr. Weber: Dr. Ball, can you tell us more about the Emerging Drug Safety Technology Program?
Dr. Ball: I’d love to. The Emerging Drug Safety Technology Program, or EDSTP, is an exciting initiative we’re launching here at CDER. It’s designed to foster mutual learning and discussion between FDA and industry, as well as other parties supporting industry’s pharmacovigilance activities, such as academia and software developers, through the use of a new meeting pathway we’re calling the Emerging Drug Safety Technology Meetings.
We’re inviting these groups to request a meeting with us so we can have conversations with them about their research, development, and implementation of cutting-edge technology in pharmacovigilance, in particular AI. In the early stages, these meetings will be focused on mutual learning – sharing ideas, discussing challenges. This will allow CDER to learn from their experiences firsthand.
What we learn from these conversations will inform our potential regulatory and policy approaches within pharmacovigilance and help us consider providing regulatory advice on specific technologies to help facilitate their adoption, when appropriate.
Interested parties who are looking to learn more about the program can visit our website, where we describe how to request a meeting.
Dr. Weber: What type of information does FDA hope to gain from the Emerging Drug Safety Technology Program?
Dr. Ball: Through the EDSTP we’re hoping to gain a deeper understanding of how the pharmaceutical industry is exploring or applying AI and other emerging technologies to pharmacovigilance. We touched on a few examples earlier, such as using AI for ICSR processing or causality assessment. We’re also interested in learning about other potential applications that we haven’t thought of. For instance, how might AI be used to analyze social media or scientific literature to identify potential drug safety issues?
Beyond the specific use cases, we’re also very interested in understanding how industry is working to ensure that the AI models they develop are credible and trustworthy. For example, how are they planning to incorporate human-led governance? What are they doing to ensure data quality and mitigate bias? And what is their plan for validating their model’s performance? Those are a few areas we’re hoping to explore through our meetings.
Dr. Weber: Is this program only focused on the use of AI in pharmacovigilance?
Dr. Ball: AI is currently in the spotlight because of the amount of attention and investment it has received, alongside its potential for use in pharmacovigilance. However, our program actually covers a wide range of emerging technologies and data sources in pharmacovigilance. Eligible groups developing new technologies beyond AI are welcome to apply.
Dr. Weber: What will participants gain from the program?
Dr. Ball: Through the meetings, we want to establish a learning environment, which will enable all parties to engage in mutual learning of how emerging technology can be applied to pharmacovigilance.
We expect that the knowledge gained through the EDSTP meetings will help inform potential future regulatory and policy approaches related to the challenges surrounding emerging technology and help foster confidence in their continued investment.
Keeping in mind that there may be substantial investments in technologies that still bear significant amounts of uncertainty, the use of emerging technology, such as AI, has the potential to bring about transformative benefits for the way organizations approach pharmacovigilance by lowering administrative burdens and costs, and improving the efficiency and effectiveness of safety surveillance.
Dr. Weber: It sounds like a win-win! Thank you so much for speaking with us today, Dr. Ball. Are there any final words that you would like to share with our audience?
Dr. Ball: Well, thank you for having me, and I guess I’d just like to say that while the EDSTP is focused on pharmacovigilance, FDA is actively also working many other areas to understand how AI can best be used throughout the drug product lifecycle.
We also recognize that technology is evolving rapidly – what we understand today may change in the very near future.
We really hope that industry and other relevant parties will engage with CDER via the Emerging Drug Safety Technology Program to facilitate mutual learning of how and when these technologies may be applied to pharmacovigilance and how challenges may be addressed.
Dr. Weber: Details about participation in the Emerging Drug Safety Technology Program are linked from our episode webpage for this podcast.
FDA will grant the meeting requests quarterly each calendar year for a total of up to nine participants in a 12-month period for the initial phase. The deadline for the first round of submissions was October 1, 2024, but parties who are interested in requesting a meeting for the next round should submit their requests by January 1, 2025. Participation will involve meeting the eligibility and selection criteria and submitting a meeting package via email.
You can find a link to the full SBIA Chronicles article at fda.gov/cdersbiachronicles. Also visit fda.gov/cdersbia to stay connected with upcoming webinars and conferences, sign up for SBIA email updates, and follow SBIA on LinkedIn. Thanks for tuning in!