Abstract
Introduction Remarkable progress has been made in the field of cancer therapy in recent years owing to the development of immune checkpoint inhibitors (ICIs); however, controlling immune-related adverse events (irAEs) remains challenging for treatment completion. This is the first study to visualize the irAE profiles of ICIs using self-organizing maps (SOM) and to combine this with decision tree analysis. The purpose of this study is to identify adverse events from a wide variety of irAEs in eight ICIs that can be useful for early detection. Methods Three anti-programmed death-1, three anti-programmed death-ligand 1, and two anti-cytotoxic T-lymphocyte antigen-4 antibodies were analyzed. Reported irAEs extracted from the Japanese Adverse Drug Event Report (JADER) database were analyzed based on the preferred term in the Medical Dictionary for Regulatory Activities. SOM was applied using the SOM package in R (version 4.1.2; R Foundation for Statistical Computing, Vienna, Austria). Results The JADER database registered 880,999 reports published between April 2004 and February 2024. The numbers of irAEs reported for atezolizumab, avelumab, cemiplimab, durvalumab, ipilimumab, nivolumab, pembrolizumab, and tremelimumab were 3797, 361, 17, 2554, 9315, 16,574, 11,487, and 196, respectively. After ICIs were classified using the SOM, they were adapted for decision tree analysis. The eight ICIs were divided into four groups based on the reported rates of type 1 diabetes mellitus and hematological disorders. Conclusion Our findings provide a reference for healthcare providers to predict irAE characteristics induced by ICIs in patients, thereby facilitating effective cancer treatment.