Close this search box.

The State of AI in Clinical Development

ACT: Where does the clinical research industry currently stand when it comes to integrating artificial intelligence (AI)?

Stephen Pyke: It’s one of a small number of questions which I get asked a lot. I’m sure AI seems to be the acronym on everyone’s lips right now. Maybe I’ll begin by standing back a little bit from clinical trials and talk more broadly about AI and clinical development and how it fits into the whole drug development process. Artificial intelligence is not new, let’s begin there. It’s easy to think that AI came into existence when ChatGPT emerged right at the back end of 2022. Of course, that’s far from true. In fact, AI probably has existed since the 1950s. Alan Turing, who many people may be familiar with, was one of the early proponents and advocates for artificial intelligence; very influential figure. And since the 1950s, we’ve seen steadily accelerating progress with AI based technologies. So, we saw machine learning, we saw deep learning. Most recently, we’ve seen large language models, and then there are image and video and speech-based equivalents. So, the technology has been around for 75 years, albeit, its capabilities, it’s expanding and advancing most recently, very rapidly, indeed. And that’s helpful to remember as a context, because the next thing I wanted to mention is some of what of the regulators think about all this and of course, given that history, they have had time to think about it; this is not new to them. And, and indeed, they published a paper 2023, which spoke about the submissions they’ve been receiving for drugs, which have been supported and touched in some way by AI. And it’s really striking, I read the paper and was astonished to read that in just one year in 2021, 132 submissions to CDER (Center for Drug Evaluation and Research) and CBER (Center for Biologics Evaluation and Research. So that’s just the US agency (FDA) had included some aspect of AI in the submission, which is already a pretty big number. And then you remind yourself that this is before large language models emerge. So, one has to assume that number is going to accelerate further in the next few years. Even so, although the regulators have had time to think about it, and they have held workshops, they’ve held advisory committees, they’ve been engaging with clinical trial and other drug development participants. There’s not much by way of formal guidance as of yet. We’ve heard them speak and we’ve heard about some of the areas that they’re concerned about. And I suppose all of this leads up to say, in that context, although we’ve seen AI start to make inroads into drug development, and to some extent into clinical trials and clinical development, I think there’s a long way to go. And I think it may happen a little bit more slowly than you would expect because of the lack of formal regulatory guidance and the natural concern of the sponsors of new drugs. Obviously, they don’t want to take risks with their assets, but neither do they want to take risks with patients. And of course the patients are the most important stakeholder in this whole process. AI is a new technology in its current format. It’s building on strong foundations, but there’s a lot that is still new and unknown. And we want to make sure that we use it wisely, carefully, responsibly, ethically, and so on.