A machine learning algorithm called TrialGPT, developed by researchers at the National Institutes of Health (NIH) in Bethesda, can accurately match volunteers to clinical trials they could participate in and provide a written summary explaining why they are a good match.
Writing in the journal Nature Communications, the researchers say that their algorithm performed well on accuracy testing and could reduce patient recruitment screening time by more than 40%.
“Matching patients to suitable clinical trials can be a challenging process,” write lead author Zhiyong Lu, a senior investigator at the National Library of Medicine (NLM) at the NIH in Bethesda, and co-authors.
“This process includes analyzing a patient’s medical history, understanding the eligibility criteria of each clinical trial, and ensuring a match that satisfies both patient needs and trial requirements. As such, manually matching patients and clinical trials is often labor-intensive, time-consuming, and prone to human errors.”
To try and make the process easier and quicker for researchers, clinicians and patients, Lu and colleagues used large language artificial intelligence models to create TrialGPT, which has three different parts: TrialGPT-Retrieval, TrialGPT-Matching, and TrialGPT-Ranking.
TrialGPT-Retrieval finds clinical trials that could be suitable, TrialGPT-Matching looks at how well a given patient matches the available clinical trials and TrialGPT-Ranking creates scores and ranks trials based on how eligible a given volunteer may be.
Testing the model showed it was able to find 90% of relevant trials and the accuracy of matching was 87%. The explanations provided were also assessed and found to be accurate and easy to understand.
The ranking score created by TrialGPT was tested in comparison to human expert judgement and was well correlated. It was also 44% better at ranking and excluding trials than other available models.
The investigator also carried out a small study where two clinicians were asked to either match patient summaries to trials themselves or use TrialGPT as a decision-making tool in the same process. The clinician using TrialGPT spent 43% less time making their assessments with the same level of accuracy.
“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 Lu in a press statement.
This work is promising, but at an early stage and needs additional validation. The team that developed the algorithm have been selected for the NIH Director’s Challenge Innovation Award to help test the technology further in a real-world setting.
