A serum proteomics‐based predictive model for steroid‐resistant immune checkpoint inhibitor associated myocarditis

Background Immune checkpoint inhibitor–associated myocarditis (ICIAM) is one of the immune–related adverse events (irAEs) with low incidence but extremely high mortality. High–dose glucocorticoid pulse therapy remains the recommended initial treatment for patients with ICIAM. However, an inadequate response to corticosteroid therapy is observed in some patients, who are subsequently classified as having steroid‐resistant immune checkpoint inhibitor‐associated myocarditis (srICIAM). Early combination of second‐line immunosuppression may represent an effective therapeutic strategy. However, Biomarkers capable of identifying such patients before treatment are currently lacking. We therefore sought to characterize circulating inflammatory proteins associated with corticosteroid responsiveness and to derive a preliminary serum‐based prediction model for steroid‐resistant ICIAM (srICIAM). Methods Serum specimens obtained from patients with ICIAM at baseline and after methylprednisolone treatment were analyzed to quantify circulating immune‐related proteins using an antibody‐based proteomic assay. We developed a preliminary model for the early identification of patients at risk of srICIAM using proteins selected by three complementary machine–learning algorithms and evaluated its discriminatory performance in an internal hold‐out cohort. Results Pretreatment serum samples from 47 patients with ICIAM were used to establish a protein–based prediction model, with 33 patients assigned to model development and 14 to internal validation. Patients with a high srICIAM predictive score (≥.5625) based on CXCL9, TRANCE and MMP1 were more likely to show an inadequate early response to corticosteroid therapy and may warrant closer clinical monitoring and further evaluation. The analysis also revealed that several serum immune proteins decreased in patients with ICIAM after methylprednisolone therapy. Serum CXCL9 levels were positively correlated with lactate dehydrogenase (LDH) and cardiac troponin T (cTnT). Conclusion Serum immune proteins were dynamically altered in patients with ICIAM after methylprednisolone therapy. The preliminary score may enable earlier recognition of patients at risk of an inadequate corticosteroid response.