2023 World Conference on Lung Cancer (Posters)-P1.20. Leveraging NLP to Identify Biomarker-Eligible Patients

Pdf Summary

This document describes a study that aims to identify biomarker-eligible patients for a clinical trial using natural language processing (NLP) and artificial intelligence (AI). The study, called TAPUR (Targeted Agent and Profiling Utilization Registry), is a phase II prospective trial that matches clinically approved drugs to specific genomic targets in patients with advanced cancers. However, patient identification for the treatment cohorts is challenging due to the inclusion of a wide range of histologies and the complexity of identifying eligible mutations from different clinical laboratories and referral sites. <br /><br />The pilot study in this document used NLP and AI to process physician notes in the Electronic Health Record (EHR) to identify patients with relevant mutations and histology. The study included 771 living, advanced lung cancer patients who had undergone genomic biomarker testing at the Fox Chase Cancer Center. The accuracy of NLP in identifying eligible patients was found to be 94% based on a review of 10% of cases.<br /><br />Results showed that 469 out of 771 lung cancer patients had a positive biomarker assay. Among them, 118 patients had one of the nine somatic mutations chosen for the TAPUR trial, and 102 were potentially eligible for one of the TAPUR studies. The mutations identified in lung cancer patients included ATM, BRAF, BRCA1, BRCA2, CDKN2A, ERBB2, CHEK2, FGFR1, and FLT3. The study also identified the number of eligible patients for specific therapies in the open trials.<br /><br />The use of NLP analysis of physician notes allowed for the expedited identification of non-small cell lung cancer patients for complex clinical trials without the need for manual searching for genomic reports. This method has the potential to streamline the identification process for biomarker-eligible patients, contributing to the efficiency of clinical trials in lung cancer research.

Asset Subtitle

John Ruckdeschel

Meta Tag

Speaker John Ruckdeschel

Topic Pathology & Biomarkers: Artificial Intelligence in Pathology

Keywords

biomarker-eligible patients

clinical trial

natural language processing

artificial intelligence

TAPUR

phase II prospective trial

genomic targets

advanced cancers

physician notes

Electronic Health Record