Title: A NASAL CLINICAL-GENOMIC CLASSIFIER FOR ASSESSING RISK OF MALIGNANCY IN LUNG NODULES DEMONSTRATES ACCURATE PERFORMANCE INDEPENDENT OF NODULE SIZE OR CANCER STAGE
Abstract: TOPIC: Lung Cancer TYPE: Late Breaking PURPOSE: Accurate assessment of the risk of malignancy (ROM) is critical in the management of a screen-detected or incidental pulmonary nodule (PN) to minimize procedures for benign disease and for timely diagnosis and treatment of patients with lung cancer. We recently demonstrated that a clinical-genomic classifier using RNA whole-transcriptome sequencing of cells from the nasal epithelium in ever-smokers with a PN can accurately classify cancer risk.1 Now unblinded after clinical validation, we show that the classifier has excellent performance in PN independent of size or stage. METHODS: Nasal brushings from ever-smokers with a PN ≤30 mm and no history of cancer underwent RNA extraction. Patients were followed for ≥1 year for radiographic stability or until a diagnosis of benign or malignant disease (with stage). The classifier was developed in a training set of >1100 patients using whole-transcriptome sequencing and machine learning with clinical factors (age, pack-years, years-since-quit, size, spiculation).1 Two decision boundaries were chosen to maximize sensitivity and specificity for low and high risk nodules, respectively, yielding low (L), intermediate (I) and high (H) risk categories. It was validated in a set of 249 patients. Predicted performance of the classifier in a population with a 25% cancer prevalence has been reported1. We now show performance in PN <8mm and ≥8mm and lung cancers of various stages, reporting sensitivity/specificity, which are independent of prevalence, in these subgroups. RESULTS: The classifier labeled all malignant PN ≥8mm as H (66%) or I (34%), demonstrating 100% sensitivity for an L vs. not L risk classification. It labeled all malignant PN <8mm as I (67%) or L (33%), retaining a 67% sensitivity in this challenging subgroup. All benign PN <8mm were labeled L (63%) or I (37%) risk, demonstrating a 100% specificity for an H vs. not H risk classification. For benign PN ≥8mm, the majority were classified as either L (15%) or I (63%) risk, retaining a 78% specificity. Performance was similarly robust in all four stages of non-small cell lung cancer. Comparing overall performance to clinical risk models, for the low-risk classification fixed at a sensitivity of 96%, the classifier’s specificity is significantly better than Gould (p=0.019). For the high-risk classification fixed at a specificity of 90%, the classifier’s sensitivity is significantly better than Mayo (p=0.037) and Brock 1b (p=0.003). CONCLUSIONS: The nasal classifier accurately assesses ROM in a pulmonary nodule independent of size or stage, with superiority over clinical risk models. Use of the classifier to guide decision-making for ever-smokers with a lung nodule could lead to fewer unnecessary diagnostic procedures in patients without cancer and timelier treatment in patients with lung cancer. 1Mazzone, PJ. Presented at: ASCO; June 4, 2021 CLINICAL IMPLICATIONS: The nasal classifier accurately assesses ROM in a pulmonary nodule independent of size or stage, with superiority over clinical risk models. Use of the classifier to guide decision-making in clinical algorithms of care for ever-smokers with a lung nodule could lead to fewer unnecessary diagnostic procedures in patients without cancer and timelier treatment in patients with lung cancer. DISCLOSURES: Employee relationship with Veracyte, Inc Please note: 2 years Added 04/25/2021 by Sangeeta Bhorade, source=Web Response, value=Salary Employee relationship with Veracyte Please note: 2/1/2021-present Added 06/24/2021 by William Bulman, source=Web Response, value=Salary Employee relationship with Veracyte, Inc. Please note: >$100000 by Jie Ding, source=Web Response, value=Salary Removed 06/23/2021 by Jie Ding, source=Web Response Employee relationship with Veracyte, Inc. Please note: 09/2017 - present Added 06/23/2021 by Jie Ding, source=Web Response, value=Salary Employee relationship with Veracyte Please note: >$100000 by Jing Huang, source=Web Response, value=Salary Employee relationship with Veracyte Please note: >$100000 by Marla Johnson, source=Web Response, value=Salary Employee relationship with Veracyte, Inc. Please note: 2008-present Added 05/10/2021 by Giulia Kennedy, source=Web Response, value=Salary Consultant relationship with Veracyte Please note: 2019-2021 Added 06/23/2021 by Carla Lamb, source=Web Response, value=Honoraria Employee relationship with Veracyte Please note: >$100000 by Lori Lofaro, source=Web Response, value=Salary Research support to my institution relationship with Veracyte Please note: $5001 - $20000 by Peter Mazzone, source=Web Response, value=Grant/Research Support Research support to my institution relationship with DELFI Please note: 6/2021 - ongoing Added 06/24/2021 by Peter Mazzone, source=Web Response, value=Grant/Research Research support to my institution relationship with Nucleix Please note: 6/2021 - ongoing Added 06/24/2021 by Peter Mazzone, source=Web Response, value=Grant/Research Support Employee relationship with Veracyte Please note: May 2018 - present Added 06/23/2021 by JIanghan Qu, source=Web Response, value=Salary No relevant relationships by Chakravarthy Reddy, source=Web Response research relationship with Veracyte Please note: 1/1/18-present Added 06/23/2021 by Kimberly Rieger-Christ, source=Web Response, value=Grant/Research Support no disclosure on file for avi spira; Employee relationship with Veracyte Please note: 2010-present Added 06/23/2021 by Patric Walsh, source=Web Response, value=Salary Employee relationship with Johnson & Johnson Please note: >$100000 by Duncan Whitney, source=Web Response, value=Salary Employee relationship with Veracyte Please note: >$100000 by Jonathan Wilde, source=Web Response, value=Salary No relevant relationships by Shuyang Wu, source=Web Response