Development of a Novel Circulating Autoantibody Biomarker Panel for the Identification of Patients with 'Actionable' Pulmonary Nodules

Simple Summary Circulating biomarkers for the identification of patients with "actionable" nodules may increase screening uptake and decrease false-positive rates associated with low-dose computed tomography (LDCT). Novel autoantibody biomarkers were identified utilizing a HuProt (TM) protein microarray. Luminex assays were developed for the targeted measurement of identified biomarkers within a large Biomarker Development Cohort (n = 841). Each individual biomarker's performance was assessed. The Biomarker Development Cohort was split into three separate cohorts: Training, Validation 1, and Validation 2. Utilizing a Training cohort, a random forest model for identifying patients with "actionable" nodules from those with "non-actionable" nodules was built. The random forest model performance characteristics were determined for both a Validation 1 and the Validation 2 cohort. From these steps we have developed a risk-stratification method that assesses circulating levels of a panel of novel autoantibody biomarkers to serve as a companion diagnostic method for lung cancer screening. Due to poor compliance and uptake of LDCT screening among high-risk populations, lung cancer is often diagnosed in advanced stages where treatment is rarely curative. Based upon the American College of Radiology's Lung Imaging and Reporting Data System (Lung-RADS) 80-90% of patients screened will have clinically "non-actionable" nodules (Lung-RADS 1 or 2), and those harboring larger, clinically "actionable" nodules (Lung-RADS 3 or 4) have a significantly greater risk of lung cancer. The development of a companion diagnostic method capable of identifying patients likely to have a clinically actionable nodule identified during LDCT is anticipated to improve accessibility and uptake of the paradigm and improve early detection rates. Using protein microarrays, we identified 501 circulating targets with differential immunoreactivities against cohorts characterized as possessing either actionable (n = 42) or non-actionable (n = 20) solid pulmonary nodules, per Lung-RADS guidelines. Quantitative assays were assembled on the Luminex platform for the 26 most promising targets. These assays were used to measure serum autoantibody levels in 841 patients, consisting of benign (BN; n = 101), early-stage non-small cell lung cancer (NSCLC; n = 245), other early-stage malignancies within the lung (n = 29), and individuals meeting United States Preventative Screening Task Force (USPSTF) screening inclusion criteria with both actionable (n = 87) and non-actionable radiologic findings (n = 379). These 841 patients were randomly split into three cohorts: Training, Validation 1, and Validation 2. Of the 26 candidate biomarkers tested, 17 differentiated patients with actionable nodules from those with non-actionable nodules. A random forest model consisting of six autoantibody (Annexin 2, DCD, MID1IP1, PNMA1, TAF10, ZNF696) biomarkers was developed to optimize our classification performance; it possessed a positive predictive value (PPV) of 61.4%/61.0% and negative predictive value (NPV) of 95.7%/83.9% against Validation cohorts 1 and 2, respectively. This panel may improve patient selection methods for lung cancer screening, serving to greatly reduce the futile screening rate while also improving accessibility to the paradigm for underserved populations.