{"architecture":{"backbone":"GPT-4 class models via API in the paper; open models can be substituted","family":"LLM pipeline (retrieval + criterion-level reasoning)","input":"patient summary (free text) + trial eligibility criteria","output":"criterion-level eligibility labels with explanations; ranked trials"},"article":null,"cancer_slugs":["pan-cancer"],"category":"clinical-LLM","confidence":"high","datasets":[{"name":"ClinicalTrials.gov","note":"trial corpus the system retrieves from","role":"evaluation","slug":"clinicaltrials-gov"}],"developer":"NIH National Library of Medicine (with National Cancer Institute collaborators)","evaluation":[{"benchmark":"SIGIR / TREC clinical-trial cohorts","external":true,"metric":"criterion-level accuracy vs. experts","source":"https://www.nature.com/articles/s41467-024-53081-z","value":"87.3% (paper)"}],"hf":null,"kind":"tool","license":"public domain (NIH code) \u2014 depends on the backing LLM","limitations":"- LLM errors on nuanced criteria (lab thresholds, prior therapies); human review required.\n- Privacy: patient text leaves the institution when using hosted models.","links":{"demo":null,"docs":null,"doi":"10.1038/s41467-024-53081-z","github":"https://github.com/ncbi-nlp/TrialGPT","huggingface":null,"paper":"https://www.nature.com/articles/s41467-024-53081-z","pmid":null},"modalities":["clinical-text"],"name":"TrialGPT","notable_uses":"In a Nature Communications (2024) evaluation, its criterion-level judgments reached 87.3% accuracy \u2014 close to physician experts \u2014 and a pilot user study showed a 42.6% reduction in screening time. The system was piloted on oncology trials managed at or related to the National Cancer Institute.","openness":"api-only","regulatory":{"intended_use_en":"Screening aid for trial coordinators; eligibility is confirmed by the study team.","intended_use_pl":"Pomoc przesiewowa dla koordynator\u00f3w bada\u0144; kwalifikacj\u0119 potwierdza zesp\u00f3\u0142 badania.","source_url":"https://github.com/ncbi-nlp/TrialGPT","status":"research-only"},"regulatory_status":"research-only","release_date":"2024-11-18","run_snippet":"","settings":["clinical-trials"],"slug":"trialgpt","sources":[{"label":"Jin Q et al. Matching patients to clinical trials with large language models. Nat Commun 2024","url":"https://www.nature.com/articles/s41467-024-53081-z"},{"label":"GitHub \u2014 ncbi-nlp/TrialGPT","url":"https://github.com/ncbi-nlp/TrialGPT"}],"summary":"Three-stage LLM framework (retrieval, criterion-level matching, ranking) that matches a patient summary to clinical trials from ClinicalTrials.gov; in the paper it cut clinician screening time by more than 40% in a pilot user study.","tasks":["clinical-trial-matching","information-extraction"],"training":{"summary_en":"None. TrialGPT performs zero-shot patient-to-trial matching by prompting a general-purpose large language model through the OpenAI or Azure API (the repository documents runs with GPT-4 / GPT-4-turbo); no model is trained or fine-tuned, so behaviour depends on the LLM that is plugged in.","summary_pl":"Brak. TrialGPT dopasowuje pacjenta do bada\u0144 w trybie zero-shot, odpytuj\u0105c og\u00f3lnego przeznaczenia du\u017cy model j\u0119zykowy przez API OpenAI lub Azure (repozytorium dokumentuje uruchomienia na GPT-4 / GPT-4-turbo); \u017caden model nie jest trenowany ani dostrajany, wi\u0119c zachowanie zale\u017cy od pod\u0142\u0105czonego modelu."},"updated_at":"2026-09-06T15:20:45.311358","url":"/ai-oncology/models/trialgpt","usage":{"library":"Python (repo) + LLM API","notes_en":"The code is open, but results depend on the LLM you plug in; patient data sent to a commercial API needs a compliant deployment."},"verified_at":"2026-09-05T22:26:00.504750","verified_by":"editorial","version":null,"what_it_does":"TrialGPT is a large-language-model framework developed at the US National Institutes of Health that matches patient records to clinical trials. It analyses eligibility criterion by criterion, explains its reasoning in natural language, and ranks candidate trials for each patient."}