Revolutionizing Clinical Trial Matching with TrialGPT

Patient recruitment for clinical trials is notoriously complex, requiring careful analysis of medical records and eligibility criteria. TrialGPT addresses this challenge with a three-step AI-driven process: TrialGPT-Retrieval, which filters trials with over 90% accuracy using less than 6% of the initial dataset; TrialGPT-Matching, which predicts patient eligibility with 87.3% accuracy and provides clear, expert-level explanations; and TrialGPT-Ranking, which outperforms baseline models by 43.8% in ranking trials. Tested on synthetic patient data with over 75,000 annotations, TrialGPT demonstrated its ability to recall relevant trials and significantly reduce the manual burden for clinicians.

In a user study mimicking real-world recruitment tasks, TrialGPT cut screening times by 42.6%, showing its potential to enhance efficiency in patient recruitment. By combining advanced language models with domain-specific methodologies, TrialGPT offers a groundbreaking solution to improve access to experimental treatments and accelerate clinical trial processes.