Understanding and Preserving Data Flow Integrity in AI-Assisted Clinical Trials

Clinical trial sponsors must shift from viewing data integrity as a database-focused concern to an end-to-end information flow issue to manage the risks of AI integration. As AI tools are increasingly adopted by CROs and study partners for tasks like data cleaning and medical review, maintaining regulatory compliance requires strict adherence to ALCOA+ principles and robust oversight of outsourced activities. This approach is essential for meeting updated FDA and ICH expectations regarding trial reconstruction and sponsor accountability in an increasingly automated research environment.

Dawn Wydner, Ph.D., of ELIQUENT Life Sciences, emphasizes that data integrity in modern clinical research depends on the seamless flow of information across the entire trial life cycle, from site capture to regulatory submission. This flow is increasingly complicated by AI, which expands the scope of governance to include how information is processed, influenced, or altered across both internal and outsourced activities. Under current FDA and ICH E6(R3) expectations, sponsors must ensure that data remains attributable, legible, contemporaneous, original, and accurate (ALCOA+), even when processed through AI-assisted interpretations or cloud environments managed by CROs.

The practical challenge for pharma and biotech companies lies in maintaining control as data moves across diverse stakeholders, including CROs, central labs, and technology providers. Regulatory guidance, such as the FDA’s 2024 update on electronic systems, underscores that while sponsors can delegate activities, they cannot delegate the ultimate responsibility for trial integrity. This necessitates risk-based governance that covers source data, derived data sets, and AI-assisted summaries. Without clear traceability and documented review of AI work products, sponsors risk being unable to reconstruct trial decisions or defend study results during regulatory inspections.

A significant blind spot identified in the sector is the use of generative AI by CRO teams, medical writers, and data managers outside of validated clinical systems. These tools are frequently used to draft narratives, analyze data sets, and prepare inspection materials, yet they often lack formal sponsor oversight. To mitigate these risks, organizations must establish enforceable boundaries around AI use, including secure platforms and approved vendor lists. Ensuring that AI accelerates trial execution without creating undocumented decision pathways is critical for preserving the reliability of essential records and maintaining participant confidentiality and blinding protections.