First, a big congratulations to the great folks at the FDA who got this out the door! Here are four key points from the FDA’s discussion paper on Using AI & ML in the Development of Drug & Biological Products [1]:
The FDA is working across centers. The paper was authored by the FDA’s drug center (CDER) in collaboration with the biologics center (CBER) and the medical device center’s Digital Health Center of Excellence (CDRH/DHCoE). Scientists talking to scientists across the FDA is a good thing: It promotes more holistic, consistent, and efficient approaches agency wide. Reinforcing this, don’t miss footnote 4 on page 1, which talks about how the principles in this document can apply to combination products, companion devices, and complementary devices.
The FDA is pulling back the curtain on how its own experts talk about AI/ML in drug development. Take section II.C, for example, where the FDA breaks down the use of AI/ML in clinical research into 8 broad categories. If you’re putting together a study that involves AI/ML, show the FDA that you’re a pro and describe your proposed study using the same categories the FDA’s own experts used here. And don’t miss the glossary! Use these terms when you communicate with the FDA and reference the discussion paper when you do.
The FDA wants more decentralized clinical trials. There are other documents out there about this right now, and the early mention on page 5 of this paper is notable. This topic just keeps coming up. It’s definitely a “tailwind topic” at the agency right now: If you have an opportunity to frame your next proposal to FDA in the context of driving towards the future of decentralized clinical trials, that could help push you ahead.
And finally, the Big Takeaway: The FDA is inviting you to help drive a larger effort. This paper is generating a dialog before the FDA gets too far down the path of deciding whether or not to put formal policy language together. Yeah, a discussion paper is not guidance, but, at least in my experience, the FDA doesn’t go to this much trouble unless they’re testing the waters and seriously considering putting something like a guidance together down the road. If you have policy ideas you want to see the FDA implement in the future, This Is Your Chance. Get your comments into the Federal Register so the FDA can use them!
Oh – and don’t miss this tidbit from the announcement: “In 2021, more than 100 drug and biologic applications submitted to the FDA included AI/ML components.” [2]
References
Reuse
Citation
@online{o'leary2023,
author = {O’Leary, Brendan},
title = {Four Things to Know about the {FDA’s} New Discussion Paper on
{AI/ML} in Drug Development},
date = {2023-05-10},
url = {https://www.boleary.com/blog/posts/202305-aiml-in-drug-development/},
langid = {en}
}