Scientists at the National Institutes of Health (NIH) have engineered an innovative artificial intelligence (AI) system aimed at optimizing the matching of prospective trial participants with appropriate clinical research studies listed on ClinicalTrials.gov.
A paper in Nature Communications highlights the efficiency of TrialGPT, the AI tool capable of pinpointing eligible trials for individuals and offering a detailed summary that illustrates how each person aligns with the study requirements. This advancement promises to aid doctors in managing the expansive and dynamic array of clinical trials, potentially enhancing the recruitment process and accelerating progress in medical research.
Collaborators from the NIH’s National Library of Medicine (NLM) and National Cancer Institute have utilized large language models to create a cutting-edge framework for TrialGPT, focused on streamlining trial matching. TrialGPT begins by analyzing a patient summary with relevant health and demographic data, identifies suitable trials from ClinicalTrials.gov, and clarifies the match criteria. The outcome is a prioritized list of trials, sorted by relevance and eligibility, assisting healthcare providers in discussing trial options with patients.
“While machine learning and AI have shown potential in trial matching, practical implementation across varied groups remained unexplored,” remarked NLM’s Acting Director, Stephen Sherry, PhD. “Our findings suggest AI can be employed judiciously to assist physicians in swiftly connecting patients to trials that may interest them.”
To verify TrialGPT’s precision in eligibility assessments, researchers compared its findings with evaluations by three clinicians across over 1,000 patient-criterion pairs. It was discovered that TrialGPT’s accuracy was comparable to that of the medical professionals.
In addition, a pilot user study involving two clinicians reviewing six anonymized patient records led to a time savings of 40% when using TrialGPT, without compromising accuracy in identifying trial matches.
Discoveries from clinical trials are crucial for advancing healthcare, with patient recruitment often occurring through healthcare professionals. However, identifying suitable trials can be tedious and resource-consuming, occasionally slowing down crucial medical research.
“Our research suggests that TrialGPT not only speeds up patient-trial matching but also allows clinicians to dedicate more time to complex tasks needing human expertise,” stated NLM Senior Investigator Zhiyong Lu, PhD.
Following encouraging results, the research team received The Director’s Challenge Innovation Award to further evaluate the model’s effectiveness and impartiality in actual clinical environments. This initiative hopes to improve recruitment efficiency and address underrepresentation in clinical research participation.
Contributors to the study included experts from Albert Einstein College of Medicine, University of Pittsburgh, University of Illinois Urbana-Champaign, and University of Maryland, College Park.
NLM excels in biomedical informatics and data science as the largest medical library globally, facilitating the storage, retrieval, preservation, and dissemination of health information. Each year, NLM’s resources and tools are globally accessed billions of times, serving a wide array of informational needs in molecular biology, health services, and beyond.
The NIH, the principal U.S. agency for medical research, includes 27 Centers and Institutes, under the U.S. Department of Health and Human Services. It spearheads vital research aimed at uncovering the causes, treatments, and cures for prevalent and rare diseases alike.