480 Investigating genetic predisposition to immune-related adverse events in patients with cancer treated with immune checkpoint inhibitors

<h3>Background</h3> Patients with cancer treated with immune checkpoint inhibitors (ICIs) can experience immune-related adverse events (irAEs). A biomarker to identify patients likely to develop irAEs with ICI treatment could have utility for heightened surveillance, personalized therapy decisions, and drug development. Given the similar presentation of irAEs and autoimmune diseases, and the high heritability of the latter, we hypothesized that there is genetic predisposition to irAEs. In part I of this study, we developed a machine learning model of irAEs. In part II, we evaluated model performance in an independent cohort. <h3>Methods</h3> In part I, we conducted germline whole exome sequencing of peripheral blood mononuclear cell samples from 100 patients with solid tumors treated with ICIs in clinical trials at NCI. Adverse events attributed by clinical trial investigators as both treatment-related and immune-related were considered irAEs. Variant calling, annotation, and filtering were conducted to collapse functional germline variants by gene. Feature selection was followed by final modeling using a decision tree-based algorithm with 5-fold repeated stratified cross-validation and subsequent hyperparameter tuning. In part II, model performance was evaluated in an independent cohort of 30 patients treated with ICIs in the NCI-MATCH clinical trial.¹ The model was applied to whole exome sequencing data to blindly predict occurrence of an irAE. Adverse event attribution and model performance testing were conducted by external collaborators to facilitate unbiased evaluation. <h3>Results</h3> In part I, feature selection restricted the model input to the mutation status of 131 genes. The model had an overall accuracy of 43% in the part II cohort. <h3>Conclusions</h3> Low model accuracy in the independent cohort suggests that larger sample sizes and more standardized cohorts are necessary to identify genetic predisposition to irAEs. Inclusion of multiple tumor types and ICIs introduced variability. The subjectivity of adverse event attribution also added variability, underscoring the importance of systematic methods for characterizing irAEs.² Feature selection may have caused overfitting of the model. Biomarkers of irAEs with ICI treatment remain an important area for future research, particularly given the sizable population of patients treated with these products. Promising approaches are to decrease variability in studied cohorts, such as by focusing on specific tumor types, ICIs, and irAEs (particularly irAEs with clear objective definitions). Exploring variation in regulatory regions using whole genome sequencing could also be advantageous. Increased availability of sequencing data with corresponding high-quality de-identified clinical data would greatly facilitate advancements in this field. <h3>Acknowledgements</h3> This work was supported in part by appointments to the Research Participation Program at the Office of Oncologic Diseases, Center for Drug Evaluation and Research and the Office of Biostatistics and Pharmacovigilance, Center for Biologics Evaluation and Research at the FDA administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the U.S. Department of Energy and the FDA. This article reflects the views of the authors and should not be construed to represent the FDA’s view or policies. <h3>Trial Registration</h3> NCT01772004, NCT02517398, NCT02155647, NCT02465060, NCT04439214. <h3>References</h3> Azad NS, Gray RJ, Overman MJ, Schoenfeld JD, Mitchell EP, Zwiebel JA, <i>et al</i>. Nivolumab is effective in mismatch repair-deficient noncolorectal cancers: results from arm Z1D—a subprotocol of the NCI-MATCH (EAY131) study.<i> J Clin Oncol</i> 2020;<b>38</b>(3):214–222. FDA. Characterizing, Collecting, and Reporting Immune-Mediated Adverse Reactions in Cancer Immunotherapeutic Clinical Trials Guidance for Industry: Draft Guidance. 2022. [https://www.fda.gov/media/162341/download]. <h3>Ethics Approval</h3> The clinical trials in this study were conducted with approval from the institutional review board at the National Cancer Institute (protocol numbers 13-C-0063, 15-C-0012, 15-C-0179, and EAY131) and required informed consent for all patients.