What ClinOps Wants To Know About AI

By Dan Schell, Chief Editor, Clinical Leader

What is it that you don’t already know about AI? After all, it’s all we hear about in the clinical space these days. And if you don’t know something about it, or you’re unsure how you can use it, you can always ... well, ask an AI.
I remember my early years as a writer at this company covering “cutting-edge” stuff like an amazing, high-density new bar code symbology (what we now refer to as a QR code) or this short-range radio-wave tech with a weird name — Bluetooth. Over the years, these technologies have become ubiquitous. But as we all know, AI’s ascent is moving much more rapidly.
In January, our Clinical Leader Live was titled, “AI In Action: Transforming Clinical Trials.” Compared to other live events we’ve held; this one killed it with 781 registrants and 479 attendees. To put it mildly — interest was high.
As part of the event, registrants filled out a form that asked questions such as what they wanted to learn from the webinar and what their biggest barrier was to adopting AI for their clinical trial processes. After recently arriving back from the SCOPE Summit where I was inundated with discussions around AI, I thought I’d be a glutton for punishment and go back and take a closer look at the data from those Clinical Leader Live registrants.
The following are the five most common themes of questions that registrants had when signing up for the live event. This isn’t all of them, and I’ve edited some of these for clarity. Not surprisingly, there were a lot of questions about how to become trained in AI. That reminded me of a slide that I saw at a lot of SCOPE presentations this year touting an upcoming AI Training Academy SCOPE is working on in conjunction with Barnett International. And speaking of SCOPE and AI, it’s notable that this year’s Annual Participant Engagement Award went to Grove AI, a newcomer to the industry that developed a human-like AI agent that “calls, answers questions, prescreens, schedules appointments and delivers more qualified and informed participants to sites, and then engages them along the way — all while providing critical insights to sites.” That definitely falls under the category of AI examples that are actually in use today.
Examples of AI In Clinical Trials:
- Can you give an example of current tools that are already in use for improving the daily data management and study start up process?
- Do you have examples of the use of AI enabling sites to get set up faster or enroll faster?
- With all of the historical data from decades of clinical trials, are there any examples of using AI and RWD to reduce either the number of patients and/or trials?
Training On AI:
- What would a clinical research professional (with experience in clinical PM for example) need to learn in order to develop the skills to take on the role of the person who integrates AI into clinical research operations?
- Seems that risk management/risk assessment will be a key driver of the success of AI within the various clinical study disciplines (data management, recruiting, analysis, etc.). Consequently, how will risk-management disciplines need to evolve?
- 25 years ago, I was the only clinical person in the company who was interested in clinical trial technology, but IT insisted that was their role. Is this a different time and place (i.e., who "owns" AI)?
- As a small biotech I would like for us to begin implementing AI in our studies, however, I am unsure how we can do this in small incremental ways so that the company can learn and grow with its acceptance of AI. Thoughts/suggestions?
- What applications would the FDA suggest for beginners to begin getting comfortable with AI use in Clinical Research?
- What is the most effective, reliable way to learn about AI to be applied to clinical trials for independent staff?
Applications Of AI:
- Will AI-generated data replace feasibility questionnaires?
- Can AI predict the percentage of non-enrolling sites in a trial?
- How will AI impact clinical trial monitoring?
- Could you elaborate a bit more on how AI works in indication selection?
- Do you have ideas for how AI can be applied for protocol deviations?
- How can we use generative AI to build study applications?
- How can AI be used for data cleaning or safety reviews?
AI’s Shortcomings/Fears:
- Does AI generate better results overall, or does it fail to pick up on important details (such as study team motivation or how well the PI and study coordinator work together)?
- What are the concerns about AI hallucinations misdirecting organizations?
- Based on the data capabilities of AI, wouldn’t AI replace most data management Roles?
- Will IRBs delay trials that include some element of AI?
- How should we look at confidential data being used to train these models? How do we ensure AI models do not breach PHI in the warehouses in which they are stored?
Regulatory Issues With AI:
- How does the FDA view clinical studies that utilize AI during their conduct? Does the FDA scrutinize such studies since it’s such a gray area? What sort of documentation is expected?
- Has the FDA noticed more use of AI in clinical research in any specific therapeutic areas over others? Maybe in oncology?
- How do you perceive the approach toward AI implementation in ClinOps between the U.S. and EU (both in terms of industry and regulatory bodies)?