Clinical Trial Matching
Clinical trial matching is traditionally a manual and time-intensive process. Clinicians must sift through hundreds of trial protocols to determine eligibility, and delays or limited screening scopes can mean missed opportunities for patients seeking promising new treatments. Preliminary use of AI in clinical trial matching was accomplished through Trial GPT, a proof-of-concept system developed by researchers at NIH. Trial GPT boasted an 87.3% accuracy in matching while reducing screening time by 42.6%1. While these results are encouraging, the last year has seen significant advancements in the reasoning abilities of LLMs which can be leveraged to increase accuracy and provide more nuanced explanations.
To this end, CAAI is exploring how an advanced trial matching system that leverages the latest AI tools and techniques. The system uses API calls to pull clinical trial data from clinicaltrials.gov, applying an initial filter based on key criteria such as diagnosis, age, sex, and geographic location. Once a relevant set of trials is identified, patient information is compared against each trial’s eligibility requirements. Data and the employed reasoning model are hosted and operated out of UK-owned NIST-compliant infrastructure. Using LLMs with advanced reasoning capabilities, the system not only validates match accuracy but also generates detailed explanations. These explanations support clinicians in making informed decisions and help patients better understand why certain trials are, or are not, appropriate for them. This goal of this effort is to reduce the time clinicians spend searching manually and help match patients with trials more efficiently.
This work is ongoing. In collaboration with the Markey Cancer Center’s Molecular Tumor Board, the team is refining the tool to ensure accuracy, transparency, and seamless integration within existing workflows.
[1] Jin, Q., Wang, Z., Floudas, C.S. et al. Matching patients to clinical trials with large language models. Nat Commun 15, 9074 (2024). https://doi.org/10.1038/s41467-024-53081-z
From <https://www.nature.com/articles/s41467-024-53081-z#citeas>
Photo courtesy of NIH/NLM https://www.ncbi.nlm.nih.gov/research/trialgpt/
