In an exciting advancement for clinical research, researchers from the National Institutes of Health (NIH) have created an artificial intelligence (AI) algorithm designed to expedite the matching of potential volunteers with pertinent clinical research trials found on ClinicalTrials.gov.
This AI algorithm, called TrialGPT, can effectively identify suitable clinical trials for which an individual qualifies and offer a summary that clearly details how that individual satisfies the enrollment criteria for the study. The researchers concluded that this tool could assist clinicians in navigating the extensive and constantly evolving array of clinical trials available to their patients, potentially resulting in enhanced clinical trial enrollment and quicker advancements in medical research.
A group of researchers from NIH’s National Library of Medicine (NLM) and National Cancer Institute utilized the capabilities of large language models (LLMs) to create an innovative framework for TrialGPT to facilitate the clinical trial matching process. TrialGPT initially processes a patient’s summary, which includes relevant medical and demographic data.
The algorithm then recognizes applicable clinical trials from ClinicalTrials.gov for which a patient qualifies while omitting trials for which they do not qualify. Afterward, TrialGPT clarifies how the individual meets the study enrollment criteria. The resulting output is a ranked and annotated list of clinical trials based on relevance and eligibility, which clinicians can utilize to discuss clinical trial options with their patients.
“Machine learning and AI technology have held promise in matching patients with clinical trials, but their practical application across diverse populations still needed exploration,” said NLM Acting Director, Stephen Sherry, PhD. “This study shows we can responsibly leverage AI technology so physicians can connect their patients to a relevant clinical trial that may be of interest to them with even more speed and efficiency.”
To evaluate how effectively TrialGPT forecasted whether a patient fulfilled a particular requirement for a clinical trial, the researchers compared TrialGPT’s findings with those of three human clinicians who reviewed over 1,000 pairs of patients and criteria. They discovered that TrialGPT reached an accuracy level nearly identical to that of the clinicians.
Moreover, the researchers conducted a preliminary user study, where they instructed two human clinicians to examine six de-identified patient summaries and align them with six clinical trials. For each patient-trial combination, one clinician was tasked with manually reviewing the patient summaries, verifying if the individual was eligible, and determining their potential qualification for the trial. In contrast, another clinician assessed the patient’s eligibility for the same patient-trial combination using TrialGPT.
The researchers observed that with the use of TrialGPT, clinicians reduced their time spent on patient screening by 40% while maintaining the same accuracy level. Clinical trials yield significant medical advancements that enhance health, and prospective participants frequently discover these opportunities through their healthcare providers.
However, the process of identifying appropriate clinical trials for interested individuals is often laborious and requires considerable resources, which can impede vital medical research.
“Our study shows that TrialGPT could help clinicians connect their patients to clinical trial opportunities more efficiently and save precious time that can be better spent on harder tasks that require human expertise,” said NLM Senior Investigator and corresponding author of the study, Zhiyong Lu, PhD.
Due to the promising benchmarking outcomes, the research team has recently been chosen for The Director’s Challenge Innovation Award to evaluate the model’s effectiveness and fairness in real-world clinical environments. The researchers hope that this project can enhance the efficiency of clinical trial recruitment and aid in eliminating obstacles to participation for groups that are underrepresented in clinical studies.
Journal reference:
- Qiao Jin, Zifeng Wang, Charalampos S. Floudas, Fangyuan Chen, Changlin Gong, Dara Bracken-Clarke, Elisabetta Xue, Yifan Yang, Jimeng Sun & Zhiyong Lu. Matching patients to clinical trials with large language models. Nature Communications, 2024; DOI: 10.1038/s41467-024-53081-z