Ab0019 autoantibody Biomarkers of Immune-Related Adverse Events and Survival in Melanoma Patients Receiving Immune Checkpoint Inhibitor Therapy

Abstract

BackgroundImmune checkpoint inhibitor (ICI) therapy has greatly improved the prognosis of advanced cancers but is often complicated by immune-related adverse events (irAEs). Biomarkers of ICI safety and efficacy are needed.ObjectivesTo identify autoantibodies (AAbs) associated with irAEs and cancer outcomes in ICI-treated patients.MethodsThis analysis used clinical data and biospecimens from a prospective trial of 60 patients with advanced melanoma who received 2-4 doses of combination ICI (anti-PD1 + anti-CTLA4) followed by anti-PD1 monotherapy.[1]Patients were followed for three years for irAE development, cancer progression, and death. Plasma was collected at baseline and six weeks after ICI initiation and assayed for AAbs using PhIP-Seq with a human peptide library spanning the entire human proteome. Baseline levels of AAbs with a ≥2 fold-change in ≥10 patients at follow-up (n = 833 AAbs) were concurrently included in a bootstrapped LASSO Cox proportional hazards model (1000 replications; α = 0.5; λ chosen by cross-validation for each replication) that was separately run for each outcome: ≥1 severe irAE (CTCAE grade ≥ 3), progression-free survival (PFS), and overall survival (OS). The proportion of times each AAb was replicated was reported as its “validation frequency.”[2]Hazard ratio confidence intervals were reported after individually calculating bias-corrected and accelerated (BCa) bootstrap intervals (1000 replications) for each AAb. The AAbs with the highest validation frequencies were used for Partial Least Squares Discriminant Analysis (PLS-DA) clustering.ResultsAmong the 60 patients, 47% developed a severe irAE, 25% progressed, and 23% died.Table 1shows the six AAbs with the highest validation frequencies for severe irAE, PFS, and OS. PLS-DA was able to cluster patients by the development of severe irAEs using only six AAbs with a 10-fold cross-validated AUROC of 0.86 ± 0.01 (p <.001) (Figure 1). PLS-DA performed similarly well for PFS and OS.Table 1reflects that higher levels of the identified AAbs at baseline were associated with worse PFS and OS (HR > 1); this pattern remained true across the 25 AAbs with the highest validation frequencies for PFS (21/25, 84%) and OS (23/25, 92%). Conversely, amongst the top AAbs linked to severe irAEs, lower levels in the majority (15/25, 60%) of the identified AAbs were associated with this outcome (HR < 1). Most of the identified top AAbs were unique to each outcome, but there was some overlap: PFS and OS shared seven AAbs (28%), PFS and severe irAEs shared two AAbs (8%), and OS and severe irAEs shared one AAb (4%). No AAbs overlapped between more than two outcomes.ConclusionThese results suggest that specific autoantibodies may be useful biomarkers for predicting outcomes in melanoma patients receiving ICI therapy. However, these results require validation.References[1]Postow MA, et al. Journal of Clinical Oncology 2022;40:1059-1067[2]Yang JJ, et al. Blood 2012;120:4197-4204Table 1.Hazard ratios and 95% confidence intervals for the top six autoantibodies for each outcome≥1 Severe irAE (n = 60, events = 28)Progression-Free Survival (PFS) (n = 60, events = 15)Overall Survival (OS) (n = 60, events = 14)AAbHazard Ratio (95% CI)VF*sVF**AAbHazard Ratio (95% CI)VFsVFAAbHazard Ratio (95% CI)VFsVFAKAP91.25(1.08-1.42)76.4%100ACAN1.72(1.21-2.35)66.3%100KHSRP1.34(1.13-1.65)52.8%100KTN11.26(1.02-1.50)60.0%79POLR2A1.29(1.04-1.57)62.7%95EN21.31(1.05-1.61)51.8%98GCN10.66(0.45-0.95)54.4%78ZGRF11.47(1.03-1.81)50.9%77SBNO21.33(1.04-1.65)50.0%95FCHO21.21(0.98-1.48)49.5%71DYNC2H11.39(0.97-1.85)45.2%68PAK41.46(1.16-1.87)43.5%82NRXN20.72(0.44-0.95)44.2%58TRIOBP1.26(1.05-1.43)41.3%62DYNC2H11.35(0.95-1.78)39.0%74EPC11.28(1.09-1.53)43.2%57SDCCAG81.40(1.08-1.74)39.4%59ACAN1.43(0.71-2.11)34.0%64*VF = Validation Frequency (proportion of times each AAb was replicated by the LASSO model)**sVF = Scaled Validation Frequency (validation frequency scaled to highest value for each outcome)Acknowledgements:NIL.Disclosure of InterestsCarlos Aude: None declared, Nilasha Ghosh Grant/research support from: Hospital for Special Surgery, Deanna Jannat-Khah Shareholder of: Walgreens Boots Alliance, AstraZeneca, and Cytodyn, Grant/research support from: Hospital for Special Surgery, Karmela Kim Chan: None declared, Michael Postow Consultant of: BMS, Merck, Novartis, Eisai, Pfizer, and Chugai, Grant/research support from: RGenix, Infinity, BMS, Merck, and Novartis, H Larman Consultant of: Scientific Advisory Board member for TScan Therapeutics, Employee of: Founder of ImmuneID, Portal Bioscience, and Alchemab, Anne Bass Grant/research support from: Hospital for Special Surgery, Memorial Sloan Kettering Cancer Center, and the Rheumatology Research Foundation.