Special Section: AI in Clinical Research Part 3
How AI Can Finally Address Notorious Site Burdens
Wendy Tate

ecent years have seen a dramatic increase in new clinical trial technologies. These tools, layered onto sites’ daily workflows, are meant to capture more high-quality data as well as alleviate the traditional burdens of trial operations.

In many ways, however, all this new technology has only added to sites’ challenges. According to a new clinical trial survey, sites are feeling overwhelmed by technology. Nearly 60% of top-performing sites say their study volume is higher than it was five years ago, in addition to juggling more technology than ever. More than half of sites report using electronic data capture (EDC), interactive voice response, and safety letter systems, while 33%-50% say they are also using electronic patient-reported outcomes (ePRO) and electronic clinical outcomes assessments (eCOA), learning management, document exchange, and site payment systems in more than three-quarters of studies.

Some of these technologies do not even provide real value. Dr. Jeff Kingsley, chief development officer at Centricity Research, invests in new solutions for his sites. However, he has seen too many offerings sold as “high-tech” fall flat. “I had a trial last year where you handed a tablet to the patient for consent, but you weren’t supposed to coach them on how to fill it in. Yet the tablet allowed the patient to sign incorrectly on the investigator line,” said Kingsley in an interview for this article. “Many of these ‘e’ products simply move from paper to digital without adding any intelligence.”

With increased sponsor and CRO-provided technologies also comes a commensurate increase in logins for site staff to remember: nearly 70% of respondents in the previously referenced survey report having six or more logins per study. And while all these technologies are undoubtedly improving the data captured by trials, site respondents overwhelmingly find sponsor technology setup and training burdensome; 67% say it’s worse than just five years ago.

“Sponsors may expect us to use eight different technologies for a single study—one for study training, one for recruitment, one for regulatory, one for IRB, one for subject registrations, one for randomization, one for the lab, one for the patient diary, and one for EDC, … plus we have our own systems including a database, CTMS, eReg, eReg binder, and eSource combining to be north of a dozen different systems for a single trial,” added Patricia Larrabee, RNNP, a technology consultant to several research organizations, in an interview for this article. “When you count up all the portals, we end up with more portals than patients in some studies!”

Yet, technology has driven some breakthrough improvements even if it hasn’t yet solved all the challenges inherent to clinical trial operations. Most notably, digital technologies enabled clinical trials to continue even as most of the planet was sequestered in their homes during the COVID-19 pandemic. Wearable sensors have enabled researchers to capture more real-world data on patients without dramatically impacting their quality of life. And password managers have helped large research companies juggle thousands of patient usernames and passwords.

It’s clear that effectively deploying technology in clinical trials is a balancing act between solving problems without adding further burden. But how do we strike that balance?

Could Technology’s Latest Darling be the “Silver Bullet”?

Artificial intelligence (AI) and large language models like generative AI (GenAI) are already being embraced across industries including life sciences and healthcare, particularly in drug discovery and data analysis. However, AI is also poised to solve many burdens clinical trial sites face—both traditional hurdles as well as new challenges created by technology.

“Early tech is expensive,” said Kingsley, who has been investing in AI for at least five years. “But if we want the research industry to improve, we should be invested in doing the things that will help us improve. One of those things is AI.”

3 New Ways AI Can Alleviate Site Burdens

1. Transfer lower-level tasks to the machine.

Trial setup is rife with administrative work, from writing contracts to setting budgets. Every site seems to have its own template language for contracts, and more than 26% of survey respondents in the 2023 survey referenced above reported that contract negotiations took eight weeks or more. And when it comes to establishing trial budgets, everyone seems to start from scratch, demonstrating a concerning gap in communication and information sharing both across and within research sites, as only 31% leverage existing budget terms.

GenAI has the power to take over (and accelerate) much of this tedious pre-work. AI can be directed to consume infinite existing forms to create a first draft of contracts, budgets, consent forms, and more. Sites could shave weeks off trial timelines using these tools, without sacrificing compliance or quality.

Larrabee is trialing GenAI to comb through data in spreadsheets, identify trends, and auto-generate certain documents and tables. “We are in our infancy but are optimistic that the future can be better with the help of advanced AI,” she said.

AI is already helping sites improve one of the more challenging processes in trial start up: patient recruitment. “We’ve been beta testing AI that finds patients out of EMRs as well as social media,” said Kingsley. “Rudimentary queries in EMR are not great; you can’t do Boolean searches, and you can only pick a top three or four search criteria. AI has no such limits and identifies many real patients the EMR does not.”

From a people perspective, these tools have the power to eliminate job roles without eliminating employees themselves. Today, administrative work is performed by entry-level employees, many of whom are highly motivated and trying to go to medical or graduate school. If AI absorbs these tedious tasks, those staff can be elevated to a higher-level of involvement in the clinical trial process, thereby elevating the entire trial. Against a backdrop of increasingly tight trial budgets, these efficiencies are especially welcome.

“AI will be hugely disruptive for data entry,” said Kingsley. “AI can read what we put into eSource and transcribe it into an EDC without even having to have an API. AI will be able to read survey responses or blood pressure output and transfer them into EDC flawlessly.”

2. Improve connectivity in the ecosystem.

Clinical research is complex, and no single technology can fully manage every element of it. Sites will need to continue working with multiple, niche technologies specifically tailored to the various needs of the trial, whether they be clinical, financial, geographic, or privacy related.

Here, AI has the power to improve integrations across the trial ecosystem, connecting platforms into a seamless experience. AI can help point data exactly where it is supposed to go, from one system to another: into data warehouses where researchers can feed it into different workflows; into communication mechanisms so various parties are being informed; and touching sponsors, sites, patients, and even regulators, streamlining the FDA submission process.

“The lack of APIs (application programming interface) is hindering progress,” explained Larrabee. “We have always been early adopters of technology, but in this industry proprietary software that does not integrate is the norm. AI could create the connective tissue that brings these systems together into a seamless, powerful ecosystem.”

AI also holds the power to improve processes onsite. Intelligent analysis might suggest a different order of operations for various trial activities. Soon, AI may even be able to help sites analyze various technologies and choose the options that are best for them.

3. Foster expertise across research and technology.

A challenge endemic to clinical research is that it asks doctors to be administrators, clinical research coordinators to be financial analysts, and many other experts deeply trained in their field to take on administrative work they weren’t formally trained to perform. AI promises to redirect these experts’ efforts back toward their desired focus: finding treatments that save lives.

“Research coordinators are people who want to be in front of other humans and build rapport with a patient, taking their blood pressure and asking about their health. Instead, we ask them to sit in front of computers doing data entry, and it worsens turnover,” said Kingsley.

Likewise, AI is also encouraging people who aren’t doctors, nurses, pharmacists, or lab technicians to enter the world of clinical research. Computer scientists and technologists are excited by the opportunity to build the AI models that will disrupt entire industries—healthcare included. AI is opening doors for these experts to join in the mission to make clinical research safer, more efficient, and more effective than ever. This virtuous cycle promises to alleviate the technological and administrative burdens facing clinical research sites, to the benefit of human health.

The Stimulus for Better Site Operations, Better Health

The technology layered onto sites’ daily workflows is intended to solve problems. But in many ways, it has only added to sites’ burdens, with site personnel—the front lines of trials working directly with patient participants—suffering the greatest strain. AI can help by executing administrative tasks, elevating site performance, connecting the ecosystem more efficiently, and catalyzing a technological revolution that will continue to improve the way we conduct clinical research. These fundamental advancements promise not only to improve timelines and costs of trials but also their outcomes, to the benefit of sponsors, patients, and human health.