ACT: What are some key regulatory requirements or challenges that need to be addressed when using real-world data (RWD)?
Gliklich: Before I dive into what’s good and bad, I would say that the agency has been very forward thinking in publishing really clear guidance documents on real-world data. They’ve been very clear that data needs to be fit-for-purpose and that can be a little bit in the eye of the beholder, but if you really come down to it, it’s a pragmatic way to think about and I like that, but the two requirements that are really necessary are that the data that’s used, readable data for submission is traceable, traceable from the source to the database. Every transformation, everything that’s done with it, what the source ID is, all of that has to be captured, and then it needs to be auditable, so that’s part of the agency’s requirements. That means that an inspector can go to a clinical site, look at the electronic medical record and confirm that what’s on the screen is in the database. If you can meet those two things, plus getting appropriate IRB approvals and consent and those types of things, you’ll meet the regulatory bar. Those are the two key things, and they’re hard for a lot of groups collecting real-world data because traceability can be hard, and auditability can be hard as well.
One of the areas, though, that I think the guidance is and so forth are still behind, is where technology is advancing really rapidly, and large language models are a good example. It’s hard for guidance documents to keep up with what’s almost generalized AI. I’m not criticizing them, but they’re going to need to learn to keep up to some extent, because those tools; people are going to want to use those tools in clinical trials as well.