While much attention has been paid of late by regulatory authorities to the matter of design requirements for what is expected to become an increasingly technologically integrated, artificial intelligence (AI)–assisted clinical trial infrastructure than is currently the status quo, the evidence that following the existing guidance will improve operational outcomes is lacking. At least, that’s the view of the author of a peer-reviewed article set to appear in a forthcoming issue of ACRP’s Clinical Researcher journal. Meanwhile the contributor of another accepted article notes that no federal regulation currently requires clinical trial sites to inventory AI tools, conduct pre-deployment risk assessments, or establish clear lines of accountability when those tools fail and, in any event, a third author posits that AI is not replacing clinical data managers as some may worry, but elevating the role to that of clinical data scientists.
Such wide-ranging perspectives on issues arising from the expanding applications of AI within the clinical research enterprise come as a new analysis from the Tufts Center for the Study of Drug Development and Medable shows that AI agents unequivocally accelerate clinical trials and improve staff productivity, delivering net financial gains as high as $21 million per drug development program (64x for Phase II and 82x for Phase III modeled return on investment). Tufts analyzed key activities in oncology trials that directly affect timelines, resource utilization, and efficiency, and found that these AI-related savings reflect clinical research associate time that could be reallocated to other studies. But impressive dollar figures aside, many questions linger about the practical implications of AI use in clinical trial settings, especially at the site level.
In “A Case for Integrated AI-Assisted Clinical Trial Infrastructure,” expected to appear in the December issue of Clinical Researcher, Ervin Ilia Mazniku, MD, PhD, Founder and CEO of KLINiSOL, writes that running clinical trials typically means seven or more discrete software systems, each with its own login, data model, validation ruleset, and vendor contract, and no common data architecture. However, he examines how a prospective study comparing operational outcomes seen from traditional practices to those seen from using integrated, Fast Healthcare Interoperability Resources–based technology “could show the gap is far narrower than architecture suggests, or that time saved on data entry is dwarfed by validation and change-control burden.” It could just as easily support the opposite, he cautions, saying, “The measurement has not yet been made, and from claims alone we cannot say which is more likely.”
Integration is not self-evidently correct, Mazniku notes. “It concentrates vendor dependence, raises the stakes of a security failure, imposes uneven transition costs, and introduces model-behavior questions that existing validation frameworks address imperfectly,” he adds. “These are reasons the question deserves an empirical answer rather than an architectural one.”
Meanwhile, in “Governing Artificial Intelligence in Clinical Trials: An Ethical Framework for Site-Level Implementation,” also slated for the December Clinical Researcher, Leslie Byatt, MBA, MSML, PMP, CCRC, FACRP, Director of Clinical Trials Operations at the University of New Mexico Hospital, writes that AI is being embedded in clinical trial workflows faster than governance frameworks exist to manage it. “Sponsors, contract research organizations, and third-party vendors deploy AI-enabled tools—often without explicit disclosure to the site—in functions that directly affect participant safety, data integrity, regulatory compliance, and equity,” she notes. “No federal regulation currently requires clinical trial sites to inventory AI tools, conduct pre-deployment risk assessments, or establish clear lines of accountability when those tools fail. The gap is real, and it belongs to sites to fill.”
And in “From Clinical Data Manager to Clinical Data Scientist: Reframing the Clinical Data Manager Role in the Era of Artificial Intelligence,” Shruthi Gajjala, PharmD, MS, CCDM, PMP, a clinical data management professional, writes that her specialty is undergoing a profound transformation. “Machine learning and broader AI capabilities are moving from conference talking points into day-to-day operations, automating high-volume tasks, identifying risks earlier, and shifting the data manager’s work from manual reconciliation toward oversight, interpretation, and strategy,” she explains. “My commentary considers where machine learning is genuinely entering clinical data management today, what it does and does not change about the role, and the skills clinical data professionals will need as the profession evolves from data management into clinical data science.”
Edited by Gary Cramer
