TrialGPT FAQ

TrialGPT is an end-to-end framework designed for patient-to-trial matching, leveraging large language models (LLMs). It consists of three main modules: TrialGPT-Retrieval for filtering and retrieving relevant clinical trials, TrialGPT-Matching for determining patient eligibility against trial criteria with truthful explanations, and TrialGPT-Ranking for scoring and ranking the trials based on patient suitability. Given a patient summary, TrialGPT can return a list of clinical trial recommendations from a large collection of candidates, based on the patient’s eligibility.

TrialGPT operates through three sequential components:

• TrialGPT-Retrieval: This module uses LLMs to generate keywords from patient summaries to filter out the most irrelevant clinical trials from an initial, large pool using a combination of lexical and semantic retrieval methods.
• TrialGPT-Matching: Analyzes patient-trial pairs on a criterion-by-criterion basis, providing explanations of patient eligibility and locating relevant sentences in patient notes for each criterion.
• TrialGPT-Ranking: Aggregates criterion-level predictions to compute trial-level scores, which are used to rank the clinical trials by the extent of patient eligibility.

The targeted users of TrialGPT include clinical trial recruiters, research teams involved in trial matching, and healthcare professionals who facilitate patient recruitment. It is designed to assist domain experts by enhancing efficiency and reducing time spent on patient-trial matching.

While TrialGPT can empower individual patients to identify potential clinical trials they may qualify for, its primary design focuses on supporting clinical and research staff to facilitate patient matching. Patients might access its benefits through healthcare intermediaries at this point.

Yes, the TrialGPT framework is publicly accessible. The code and related resources can be found on its GitHub repository: https://github.com/ncbi-nlp/TrialGPT.

No. Although one can potentially use the official code to deploy a web-based system for TrialGPT, we do not provide an official implementation.

TrialGPT is developed by researchers from the Division of Intramural Research, National Library of Medicine, National Institutes of Health.

Unlike traditional embedding-based methods such as DeepEnroll [1] and COMPOSE [2] which require massive training data and are not explainable, TrialGPT can perform zero-shot and explainable patient-to-trial matching without requiring any training data. Unlike structuring-based methods such as Criteria2Query [3], TrialGPT does not depend on the ontology mapping of clinical trial criteria, which is a non-trivial task itself.

References

1. Zhang X, Xiao C, Glass LM, et al. DeepEnroll: Patient-Trial Matching with Deep Embedding and Entailment Prediction. Proceedings of The Web Conference 2020. New York, NY, USA: Association for Computing Machinery 2020:1029–37
2. Gao J, Xiao C, Glass LM, et al. COMPOSE: Cross-Modal Pseudo-Siamese Network for Patient Trial Matching. Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York, NY, USA: Association for Computing Machinery 2020:803–12
3. Fang Y, Idnay B, Sun Y, et al. Combining human and machine intelligence for clinical trial eligibility querying. Journal of the American Medical Informatics Association. 2022;29:1161–71. doi: 10.1093/jamia/ocac051

Please cite TrialGPT as:

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. Nat Commun. 2024;15:9074. doi: 10.1038/s41467-024-53081-z

Given the promising benchmarking results, the research team was recently selected for The Director’s Challenge Innovation Award to further assess the model’s performance and effectiveness in real-world clinical settings. The researchers anticipate that this work could make clinical trial recruitment more effective and help reduce barriers to participation for patients seeking to join clinical research.