How AI is Being Used to Prevent Clinical Trial Mismatches

One of FDA’s main focuses at the moment is streamlining the drug approval process. In June of this year, the agency announced a new pilot program designed specifically to shorten the early-stage drug trial timelines.

A key factor of this is the growing acceptance of RWD in trial design. The agency hopes that by allowing RWD collected by qualified research institutions, drug developers will not only be able to move through early stages faster, but also reduce the need for late-stage trials as well.

Pharmaceutical Executive spoke with Jen Lamppa, VP of commercial strategy, how RWD impacts trial design, specifically during the patient eligibility screening phase. The addition of this data is expected to impact recruitment timelines, although it may have other unintended impacts as well. As such, pharma and biotech companies will likely need to adjust their recruitment models to properly take advantage of the new environment.

Pharmaceutical Executive: How is AI being used to prevent clinical trial mismatches?
Jen Lamppa: AI can be applied very effectively in this space — particularly to accelerate analysis of unstructured clinical data. In administrative data such as claims, the bulk of information is in structured formats: diagnosis codes, procedure codes, medication codes. But in clinical data, where much of the real value sits — the value most adjacent to what is collected in a clinical study — a significant portion is unstructured. It lives in clinical notes, radiology reports, and images. AI is being applied broadly to extract insights from that unstructured layer, and then to validate that those insights are accurate.

The ability to harness AI to accelerate specific analytics — including modeling and sensitivity analyses — is critical to actually putting real-world data to use for its intended purposes. The traditional approach has always been manual adjudication and validation, which does not scale. That limitation has been widely recognized for some time, and there will always be a place for manual review in data management and validation. But the application of AI to reduce the manual burden, compress timelines, and ultimately improve the accuracy of what is being collected on patients is foundational to the broader use of real-world data — particularly complex clinical real-world data.