Clinical Utility of a Combined Biomarker Approach to Diagnose Lung Cancer

Vanderbilt University Medical Center, Nashville TN

Investigators

Linked publications & trials

Abstract

Lung cancer remains the number one cancer killer in the United States and clinically useful biomarkers are needed to improve early detection and diagnosis. The objectives of this proposal for our continuing Clinical Validation Center are to push early lung cancer detection biomarkers into clinical practice while continuing to serve as a core resource to the EDRN, as well as to our academic and industry partners. Our overall objective is to demonstrate that biospecimen and imaging biomarkers will provide clinical utility to diagnose lung cancer by reducing the number of invasive procedures performed for benign disease and the time to diagnosis for cancer. Aim 1 will seek to demonstrate clinical utility of a combined biomarker and radiomic approach for providing IPN diagnoses. We will expand the existing IPN specimen and imaging biorepository available to the NCI and scientific community, demonstrate the clinical utility of combination biospecimen and radiomic biomarkers, and validate additional candidate lung cancer risk biomarkers.  We hypothesize that a cancer, benign, and radiomic biomarker approach will decrease invasive procedures and the time to diagnosis in an IPN population.  In Aim 1a we will validate the combined approach of Histoplasmosis EIA benign biomarker (MiraVista), hs- CYFRA 21-1 cancer biomarker and radiomic biomarker (HealthMyne) in the EDRN Lung Team Project 2 (LTP2) and National Lung Screening Trial (NLST) cohorts.  In Aim 1b we will validate candidate blood and epithelial biomarkers in Phase 2 and 3 prospective-specimen collection and retrospective-blinded-evaluation (PRoBE) design studies for the early diagnosis of lung cancer and determine suitability for a clinical utility trial.  In Aim 2 we will validate radiomic risk assessment platforms in IPNs and conduct a pilot clinical implementation trial in screening discovered IPNs. We will leverage the robust bioinformatics infrastructure at Vanderbilt University Medical Center to capture and deidentify 800 thoracic CT scans in patients with IPNs. A Lung Cancer Prediction Convolutional Neural Network (LCP-CNN) and the HealthMyne radiomic model will be compared to each other and against the Lung-RADS categories. We will perform a prospective pilot evaluation of the best performing model in Lung-RADS category 3 and 4 IPNs. To accomplish Aim 2 we will: 1) compare the accuracy of LCP-CNN and HealthMyne radiomics 2) determine the LCP-CCN’s ability to reclassify nodules in screening patients in a prospective clinical implementation pilot study. In aim 3 we will test the hypothesis that the addition of a radiomic machine learning- based biomarker for pulmonary nodule lung cancer risk prediction compared to usual care improves patient management in a pragmatic, prospective randomized clinical trial at VUMC, Meharry, and Washington University in St. Louis. At the completion of this proposal, we will have 1) evaluated clinical utility of combining lung cancer biospecimen and imaging biomarkers, 2) developed a platform within current practice to present an imaging biomarker approach to improve IPN risk assessment, and 3) enhanced the biorepository resource for the EDRN and collaborative use.