Commentary|Videos|December 2, 2024

Using Artificial Intelligence to Manage Clinical Trial Conduct

In this video interview, Jeff Sidell, PhD, chief technology officer, Advarra, talks AI and how it can automate repetitive processes.

In a recent video interview with Applied Clinical Trials, Jeff Sidell, PhD, chief technology officer, Advarra, discussed current challenges in clinical research technology. He highlighted the overabundance of systems and emphasized the need for integration to streamline processes and reduce administrative overhead. Sidell also discussed the potential of AI, particularly predictive analytics, to automate repetitive tasks, enhancing efficiency and quality.

A transcript of Sidell’s conversation with ACT can be found below.

ACT: What are some unique areas of clinical research that could benefit from the use of artificial intelligence (AI)?

Sidell: I know that there's been a lot of attention paid to AI in the actual research cycle, that's not what Advarra is focused on. We're focused on the conduct of the clinical trials, all the way from submission to the IRB (institutional review board) through the execution of the trial and oversight by the sponsors of the sites and the subjects. If you think about it, that whole process is a series of smaller processes. It's people doing things somewhat repetitively. Clinical trials are not a factory, they're not banging out motor cars, but it is nonetheless repetitive processes, or similar processes. That's an area where artificial intelligence, particularly predictive analytics, can come in and automate, or semi-automate the process, and the efficiency gains effectiveness. A lot of people think you're going to make it faster, and my job is going to go away, or something like that. It does make the process more efficient. It allows people to do more with the same number of resources, but it also will typically increase the quality of the work that's done and it's not that the machine learning models don't make mistakes, but they are incredibly dependable—if it's been trained correctly, if the model is a good model—then they will repeatedly do what they were trained to do over and over and over and they'll do it well. They don't have bad days. They don't come in sick or hungover or something like that. They do have to be kept up to date, though. There is a maintenance aspect to introducing machine learning models into processes, but that's where I see the greatest, ripe for opportunity area, because it is such a process-heavy industry. Those processes can be made much, much, much more efficient.


Related to this article

© Grispb - © Grispb - stock.adobe.com
Mistral and OpenAI produced identical overall eligibility labels across 200 synthetic patient-study assessments, with criterion-level disagreements concentrated in inclusion criteria where models switch between “Uncertain” and “Not included” when records lack sufficient evidence, so model choice should be driven by deployment architecture, control, and cost rather than assumed interchangeability.
The Human Side of Clinical Trials: Before Burden Becomes Risk, Capturing Friction in Real Time
In Episode 4 of The Human Side of Clinical Trials, host Brian S. McGowan, PhD, FACEHP, chief learning officer and co-founder of ArcheMedX, Inc., speaks with Jason Riis, PhD, CEO and chief behavioral scientist at Behavioralize, about how capturing friction closer to the moment it happens can reveal unclear instructions, technology friction, and workarounds that retrospective surveys and debriefs tend to miss.
Building AI Assistants Clinical Trial Teams Can Trust: Q&A with E.K. Koh, Suvoda
In this Q&A, E.K. Koh, chief product officer at Suvoda, discusses what separates a trustworthy AI assistant from one that adds risk, why curated clinical knowledge matters more than general-purpose flexibility, and what safeguards must be in place when assistants move from answering questions to taking actions in clinical trial systems.