Predictive and Diagnostic Radiomic Signatures in Non-Small Cell Lung Cancer (NSCLC)on Immunotherapy

PROJECT SUMMARY We propose to identify novel radiomic signatures of anti-programmed death ligand 1 (PDL1)/PD1 therapy response for non-small cell lung cancer (NSCLC) and evaluate how these signatures can augment established biomarkers. Immunotherapy has been rapidly integrated into NSCLC management due to dramatically improved response rates compared to conventional cytotoxic therapy and is now also accepted as 1st line therapy for selected populations. While stratification of patients based on tumor expression of PDL1 has improved therapy response rates, up to 30-40% of NSCLC patients still fail 1st line therapy with these agents, suggesting that new strategies are needed to more accurately select patients likely to benefit. While a radiomic approach has yet to be fully studied in the context of NSCLC immunotherapy, early evidence, including our preliminary data, suggests that radiomic features extracted from routine computed tomography (CT) capture important characteristics of the tumor phenotype, including vascular structure, intra-tumor heterogeneity, and immune infiltration of the tumor microenvironment, which could provide a powerful phenotypic approach to augment established biomarkers for anti-PDL1/PD1 therapy. We propose to perform the largest radiomics study conducted to date on immunotherapy for NSCLC, leveraging CT data from an existing institutional database (n=2095 patients) which includes biocorrelates of patients treated with anti-PD1/PDL1 therapy agents, and an on-going ECOG-ACRIN multi- institutional trial (n=846) to be used for independent validation. By pursuing this research, we will therefore aim to address this fundamental question: Can radiomic signatures augment established biomarkers, such as PDL1 expression, in predicting which patients are likely to benefit most from anti-PD1/PDL1 therapy? While most radiomics studies to date have focused on anti-PD1/PDL1 therapy for NSCLC in the non-1st line setting, we will seek to discover radiomic signatures specifically for 1st versus later line of immunotherapy, and we will examine such signatures both at baseline, prior to the initiation of therapy, as well as longitudinally during the course of therapy in association to tumor response, progression-free and overall survival. We will further correlate these signatures with known biomarkers of anti-PDL1 therapy response, including PDL1 expression, tumor mutational burden (TMB), circulating (ct)-DNA, and tumor-infiltrating lymphocytes (TILS), to better understand how radiomics can augment these established and emerging biomarkers in predicting anti- PD1/PDL1 therapy response. To discover these radiomic signatures, we will leverage the Cancer Phenomics Toolkit (CapTK), an open-source and highly-standardized software developed by our group, and will utilize a novel radiomic feature standardization approach, allowing us to incorporate CT scans acquired by variable acquisition. Together, these approaches will result in robust phenotypic radiomic signatures that will enable a more informed clinical management of patients selected for anti-PD1/PDL1 therapy by identifying more nearly effective and earlier therapy options.