Researchers have created an AI algorithm that can effectively find suitable clinical trials for individuals and provide a clear summary of how they meet enrollment requirements. This tool aims to simplify the process for both doctors and patients to discover appropriate clinical trial options.
Researchers from the National Institutes of Health (NIH) have introduced an artificial intelligence (AI) algorithm designed to enhance the efficiency of connecting potential participants with applicable clinical research trials available on ClinicalTrials.gov. A study published in Nature Communications revealed that this AI algorithm, named TrialGPT, can accurately pinpoint relevant clinical trials for which individuals qualify and present a summary detailing how they satisfy the enrollment criteria. The findings suggest that this innovation could assist healthcare providers in navigating the extensive and continuously evolving landscape of clinical trials available to patients, potentially increasing trial enrollment rates and accelerating advancements in medical research.
A team from NIH’s National Library of Medicine (NLM) and National Cancer Institute utilized large language models (LLMs) to construct an innovative framework for TrialGPT, enhancing the clinical trial matching process. Initially, TrialGPT analyzes a patient’s summary, which includes important medical and demographic details. The algorithm then filters through clinical trials listed on ClinicalTrials.gov to identify which ones the patient is suitable for while excluding unsuitable options. Subsequently, TrialGPT articulates how the individual meets the study’s enrollment criteria. The result is a ranked list of clinical trials annotated based on relevance and eligibility, which clinicians can use to discuss available opportunities with their patients.
“Machine learning and AI technology have shown promise in pairing patients with clinical trials, yet their practical application among diverse populations required further investigation,” explained NLM Acting Director, Stephen Sherry, PhD. “This study demonstrates that we can responsibly utilize AI technology, enabling physicians to connect their patients to relevant clinical trials more swiftly and efficiently.”
To evaluate how well TrialGPT predicted a patient’s eligibility for specific clinical trial requirements, researchers compared the algorithm’s results with those of three human clinicians who assessed more than 1,000 patient-criterion pairs. They discovered that TrialGPT’s accuracy was nearly on par with that of the clinicians.
Furthermore, the researchers carried out a pilot user study, where two human clinicians reviewed six anonymous patient summaries and matched them to six clinical trials. In each patient-trial pair, one clinician manually reviewed the patient summaries and determined eligibility, while the other clinician used TrialGPT for the same purpose. Results indicated that clinicians utilizing TrialGPT spent 40% less time screening patients while maintaining similar accuracy levels.
Clinical trials play a significant role in discovering vital medical insights that enhance health outcomes, and potential participants often learn about these opportunities via their healthcare providers. Nevertheless, the search for fitting clinical trials can be a time-consuming and resource-heavy process, which may hinder crucial medical research.
“Our study illustrates that TrialGPT has the potential to assist clinicians in linking their patients with clinical trial opportunities more effectively, thus saving valuable time that could be redirected to more complex tasks requiring human expertise,” stated NLM Senior Investigator and lead author of the study, Zhiyong Lu, PhD.
Following the encouraging results, the research team has been honored with The Director’s Challenge Innovation Award to further evaluate the model’s performance and fairness in real-world clinical applications. They hope that this work will enhance the effectiveness of clinical trial recruitment and lower barriers to participation for underrepresented groups in clinical research.
The study was co-authored by collaborators from Albert Einstein College of Medicine in New York City, the University of Pittsburgh, the University of Illinois Urbana-Champaign, and the University of Maryland, College Park.
