Work

Quantifying Cancer Clinical Trial Eligibility Using Artificial Intelligence-Based Matching

Year: 2026

Type: preprint

Abstract: PURPOSE To develop and validate an artificial intelligence-enabled platform that combines efficient screening with explicit eligibility logic by converting unstructured cancer trial criteria into reusable structured queries, and illustrates potential eligibility using synthetic patients. METHODS We downloaded actively recruiting US interventional treatment trials for advanced/metastatic breast cancer, colon cancer, and non-small cell lung cancer from ClinicalTrials.gov . Medical oncologists created 24 synthetic patient vignettes. A large language model converted trial eligibility criteria into Structured Query Language (SQL) code and patient information into structured records, enabling rule-based matching without further LLM inference. Cancer details and treatment history were considered, but not laboratory results or comorbidities. Validation included physician editing of generated eligibility code for 30 trials and independent assessments by two physicians, followed by consensus adjudication, for eligibility across ten trials. We descriptively summarized potential eligibility by trial characteristics and when varying age, ECOG performance status, sex, and ZIP code. RESULTS Of 833 candidate trials, 746 met inclusion criteria. Performing matching for all 746 trials took a median of 0.459 seconds per patient, excluding patient-data preparation. Against physician consensus, automated matching agreed on 230/240 patient-trial pairs (95.8%; sensitivity 83.3%, specificity 100%). Across synthetic patients, potentially eligible trials ranged from 10 to 258 without geographic restrictions. Counts decreased with worse performance status and geographic restriction and were relatively stable across ages 30–90. Later-phase, randomized, and molecularly selective trials generally matched fewer vignettes. CONCLUSION LLM-generated SQL enables rapid, explicit rule-based screening for potential trial eligibility. Evaluation in synthetic vignettes supports feasibility, while missed matches underscore the need for further validation.

Source: medRxiv

DOI: 10.64898/2026.06.03.26354859