Arnaub Chatterjee Talks AI, Patient Recruitment, and Real-world Data

Welcome to the Clinical Minds “Innovator Insights Corner,” where we’ll be sharing fascinating stories, perspectives, and predictions from the guests of ‘from Dreamers to Disruptors’, a podcast by Medidata exploring life sciences innovation and the visionaries behind it all.

Data is the foundation of clinical trials, flowing through all parts of a study and beyond. Patients share it, sites gather it, and researchers use it to design the treatments of the future. A major task of clinical research and study design is harnessing AI and other emerging technologies to develop new, more efficient ways to collect, move, transform, and analyze data. By doing so, we can create better outcomes for patients worldwide.

Arnaub Chatterjee, Datavant’s GM and President, Life Sciences, Ecosystem and Public Sector, is one of the foremost experts in clinical data management. With experience working in both the private and public healthcare sectors, and as a teacher at Harvard Medical School, he has unparalleled insights into how ethical AI adoption and a longitudinal approach to data can improve patient recruitment and unlock the next generation of life-saving treatments.

Tokenization and Real-world Data

Traditionally, patient data from clinical trials has been siloed. The patient takes part in the study, and when it comes to an end, that’s it—any involvement that person has with future studies would essentially be a reset.

Pharmaceutical companies recognize the value in breaking down these barriers to build a picture of a patient’s medical journey over an extended time: their medical histories, the different studies they’ve taken part in, and their experiences and outcomes after the trial is complete. Harnessed and standardized in the right way, this longitudinal, real-world data promises to unlock deep insights that a lone trial cannot, and identify the best patients to recruit for future studies.

“If we stopped thinking about trials in a siloed or a function-based manner, we'd have a long view on that patient's entire lifespan, of how they're performing before and after,” says Arnaub. “If you can [secure] patient consent early in the trial, and they opt in to linking their data, that gives the sponsor a chance to collect that patient's data before, during, and after that trial. That works for everything outside of that clinical trial: the patient's experience with an insurance company, a lab or diagnostics company, and the medical records from a certain hospital system.”

“What we call post-trial data, or real-world data, gives us meaningful opportunities to understand if a drug is performing successfully after its launch and, if not, why.”

– Arnaub Chatterjee

“The goal is to make commercial data more usable, reliable, and impactful, through technologies like tokenization, which connects important real-world data usage patterns to clinical evidence while protecting patient anonymity,” adds Anthony.

Pharma companies aren’t the only beneficiaries of tokenization and real-world and historical data incorporation. The patients who volunteer for clinical trials are keen on having their information be put to the best use possible—to bring hope to themselves and others who suffer from the same medical conditions. By consenting to the wider sharing of their data, they’re extending the value and use of that data far beyond a single trial.

The Patient Recruitment Challenge

Patient recruitment is an enduring challenge to delivering successful clinical trials. If we can’t identify and enlist the right candidates for a study, then we'll struggle to gather the necessary evidence to create new treatments. “Right now, we have, I think, 6,500 ongoing clinical trial studies in our market, globally,” says Arnuab. “And the vast majority of them are going to under-recruit.”

For optimal recruitment, we need to be able to identify patients who are suitable for a trial, ensure they have the information and tools they need to consent and participate, and connect them to suitable sites. We also need to engage diverse populations, in order to design effective treatments that work for everyone.

“To give patients the opportunity to participate in research, we have to overcome a technology problem, a data problem, and a patient empowerment problem.”

– Arnaub Chatterjee

Streamlining the clinical trial experience is paramount, but, as Anthony says, “for all of the great technology that we make, that we think should simplify trials or ease the burden on patients and sites, many times the feedback coming back from sites is just, ‘I now have another technology I have to use. It's not standardized across my sponsors or my trials.’” Technological solutions to patient recruitment must be designed with input from sites and patients to make sure they represent a genuine benefit to the people who'll be using them.

AI is proving to be a powerful tool in tackling patient recruitment issues, analyzing data to identify suitable populations and matching them with the right sites. This helps trials to meet their recruitment goals and deliver high-quality, reliable results. But to get the most out of AI, we must employ it thoughtfully and responsibly.

Effective, Ethical AI Adoption

AI has the potential to transform clinical trials from the ground up, and Arnaub points to an acceleration in adoption that shows no signs of slowing down. To get the most out of these emerging technologies, we need to not only identify the opportunities to leverage them in clinical research, but to also be conscious of their limitations and understand how to mitigate them.

“As we're getting closer and closer to leveraging data in very creative ways, I think we have to ask ourselves some questions,” says Arnaub. “The biggest thing for me on AI guardrails is around model calibration: the outputs that are getting used, and whether those are point-in-time outputs based on a data set that is clearly biased. We must continue to calibrate our models and adjust to the ways providers treat patients, and the ways that patient populations change over time. That's probably the most important thing, because we have to keep understanding that the dynamics of the disease or the patient population will change.”

For instance, AI could be a game-changing tool in identifying the diverse populations we need for successful clinical trials. “But, to the extent that we use AI in clinical research, the foundations of those AI learnings are fundamentally not diverse, because we don't have a history of diverse trials to draw on,” says Arnaub.

“You have to have the broader network, the long-term consent, the follow-up, the real-world evidence, the pre-clinical and post-clinical data, if you want to [build] an algorithm that's going to get smarter and smarter.”

– Anthony Costello

“Those studies have to be recruited, they have to be run, they have to be closed, the data has to be shared, the AI has to be updated, in a continuous fashion,” he continues. While working to refine our trials and AI tools, we must be conscious of possible bias in AI and consider how to work with those limitations until we can overcome them.

As AI adoption continues to accelerate, we need to be vigilant about its outputs and the data it is trained on. With care and consideration, AI will help usher in a new era of clinical trials that offer better results for researchers and vastly improved outcomes for patients everywhere.

Listen to Arnaub Chatterjee and Anthony Costello’s full conversation in from Dreamers to Disruptors Episode 4 to discover how life sciences is adapting to the new AI status quo, the creative ways we’re harnessing data, and how we’re tackling the patient recruitment challenge head on.