Key points are not available for this paper at this time.
Immune checkpoint inhibitors (ICIs) have revolutionized cancer therapy, however, their use is limited by heterogeneous and unpredictable immune-related adverse events (irAEs), which can progress to life-threatening conditions requiring intensive care unit (ICU) admission. Reliable biomarkers for predicting and stratifying ICU-level irAEs are urgently needed to improve immunotherapy safety and critical care management. Here, we performed comprehensive mass spectrometry-based proteomic profiling to identify plasma biomarkers for the prediction and monitoring of irAEs in 65 patients receiving ICI treatment. Our analysis identified 217 differentially abundant proteins and four co-expression modules related to humoral (antibody-mediated) and cellular (T cell-mediated) immunity spanning mild to severe irAEs. Through feature selection and cross-validation with proteomics and ELISA data, we identified two key proteins, IL1RL1 and FABP3, as potential biomarkers for irAE risk. In addition, we developed a plasma proteomic machine learning model (ProIRAE) that demonstrated high and robust predictive performance with area under the receiver-operating characteristic curve (AUROC) values of 0.929 and 0.766 for identifying patients at risk of developing irAEs, and AUROC values of 0.978 and 1.000 for predicting severe irAEs in the discovery and independent validation cohorts, respectively. Collectively, our study provides a valuable plasma proteomic atlas of ICI-related irAEs. The ProIRAE model offers a non-invasive tool for the detection and severity stratification of irAEs, with a great potential to improve precision monitoring and management of immunotherapy complications in critical care settings.
Yan et al. (Sat,) studied this question.