I’m pleased to have this opportunity to help kick off this year’s Global Summit in Regulatory Science. I’m sorry I can’t be with you in person.
I’ve had the chance to participate in several of these meetings, going back to 2016, and I’m always impressed by how this group focuses on, and embraces the most cutting edge topics in regulatory science. That’s as it should be, because as regulators we need to stay ahead of the scientific curve to be able to respond effectively to new challenges with appropriate technological knowledge and capacity.
This meeting offers the opportunity for regulators and the agencies we represent to share our knowledge and experiences. And it allows us to collaborate in applying the best regulatory science to emerging regulatory and health challenges in order to protect the populations we serve.
This year’s agenda continues in this tradition, with discussions focused on a variety of emerging technologies and their application to the regulation of food and drug safety -- from regulatory apps, to new approach methods (NAMS), to nanotechnology.
I’m particularly pleased that the FDA and the Product Quality Research Institute (PQRI) are leading a workshop focused on the future of artificial intelligence and the regulatory framework surrounding the use of AI in pharmaceutical manufacturing.
While for many lay people, AI has just recently emerged as a hot issue, as regulators we’ve been keeping up with the science of data science for years, trying to anticipate and harness it’s potential. We know, for instance, that we’re likely to see substantial innovations in pharmaceutical manufacturing as a result of AI, including an impact on process measurement, modeling, and control, among other issues. And we also understand that some of these developments are likely to challenge approaches we’ve taken in the past.
At the FDA, there are a number of ways we’re already working to apply AI and machine learning technology to our work. For example, our National Center for Toxicological Research (NCTR) supports our Center for Drug Evaluation and Research’s AI needs via the IND Smart Template System and AI models for drug safety review. An NCTR study, AI4PharmcoVig, is applying an AI model for document screening, classification, and processing to enhance pharmacovigilance. And our Office of Minority Health & Health Equity (OMHHE) is engaged in examining ethnic and racial disparities in critical care delivery to heart-failure patients using AI and real-world data. It's just one example of how AI has the potential to make RWE more accessible and increase the ability to use all clinical evidence, RWE and otherwise, for POC clinical decision-making.
Of course, AI is just one of the many technological tools with tremendous recognizable potential, as well as the potential for scientific and regulatory challenges. The U.S. Congress has recognized the importance of this area as well, by providing us with some important additional authorities. The FDA Modernization Act 2.0, for instance, is helping facilitate the FDA’s adoption of alternative methods in reviewing food and drugs. Similarly, we are implementing the Modernization of Cosmetics Regulation Act of 2022, which expands the FDA’s authority to implement emerging technologies to assess safety in this under-regulated area.
Without a doubt, the application of advances in data and information is one of the biggest changes with some of the greatest promise I’ve seen during my long career in medicine and science. The new technologies that account for what has been called the fourth industrial revolution offer regulators numerous opportunities to strengthen our work. If we use digital technology to its full capacity, it can be central to everything we do, from reviewing medical product submissions, to protecting America's supply chain, to keeping track of quality systems in the industries we regulate, to application of data science to improve our own operations.
To maximize the potential of these technologies, we must evaluate them to ensure they will improve our processes and bring value. For example, one role of tools for Artificial Intelligence, Machine Learning, and automation products is to improve productivity as part of our background computing environment.
The fundamental issue is that the best measure of successful information technology implementation is that the enterprise is more efficient, and the basic institutional mission is accomplished with better outcomes at a lower cost.
Just as we’re digesting the tremendous impact of AI, we have quantum computing on the horizon. While I’m far from a technical expert, I believe that quantum will have an enormous impact on our understanding products we regulate, its early impact may be on the ability to use quantum to protect our data through encryption methods that are impenetrable from our adversaries.
As I’ve already mentioned, AI and associated technologies will provide enormous support to our workforce on certain tasks. As with clinician documentation burden, I believe that large language models and related technology will free up our staff to spend less time on rote documentation and more time using their brains to make better decisions and explaining those decisions to others. It can aid, rather than replace them, thereby increasing efficiency. As with many of your organizations, our most precious resource is our workforce. And it’s important that they have the freedom and support to make good decisions, while spending as little time as possible on repetitive documentation.
Another area of potential benefit involves the use of AI in industries we directly regulate. We’re making good progress there, building on the excellent framework created by our Center for Devices over five years ago. This approach makes the critical point that as hard as it is to develop an algorithm and calculate its operating characteristics, the post market phase is even more challenging. Left alone in a clinical environment, an algorithm will deteriorate unless constantly tuned using methods that create valid samples for discrimination and calibration of the predictions from the algorithms.
This means that all members of the ecosystem, including companies developing these algorithms and health systems that plan to use them, should be developing systematic approaches that comprehensively measure clinical outcomes over time in representative populations, or we run the risk of harm from predictions that stray from the original reported results. As opposed to what might be called traditional devices that operate the same way every time, AI algorithms are only as good as the ability of the algorithm to make an accurate prediction at the time the prediction is needed.
A final area involves the regulation of these technologies, such as generative AI, themselves, as opposed to regulating their use in medical devices. This is still an uncharted area, filled with political and global economic implications. While our agencies are not the primary driver here, I’ll predict that many people will point to the areas of medical information and decision support as critical bellwether issues in regulation of generative AI.
During your meeting today, you’ll have the opportunity to discuss many of these and other exciting developments, such as a project funded by the FDA Perinatal Health Center for Excellence to study predictive toxicology models of drug-placental permeability using 3D-fingerprints and machine learning, or our work on computational modelling to determine the structure of cell wall synthesis enzyme with inhibitor complexes against antimicrobial resistance and multidrug resistance bacteria.
I look forward to your comments about these developments, as well as your own advances, and most importantly, your thoughts on our future areas of focus.
I hope to see you in person at next year’s meeting in Little Rock, Arkansas, and I wish you all an enjoyable and productive meeting.