Clinical characterization of immune checkpoint inhibitor-induced myocarditis and the triple M overlap syndrome

Background: Immune checkpoint inhibitor (ICI)-induced myocarditis (MC) is one of the most fatal immune-related adverse events (irAEs) in cancer patients, often overlapping with myositis (MS) and myasthenia gravis (MG), forming the fulminant triad of Triple M Overlap Syndrome (TMOS). In this study, we report the largest clinical series of this rare fatal syndrome, aimed at delineating clinical features, fatality predictors, and temporal trends. Methods: We surveyed the WHO Vigibase pharmacovigilance database for cases of ICI-MC, MS, and MG in cancer cases through January 1, 2025. Seven groups emerged: MC alone, MS alone, MG alone, MC+MS, MC+MG, MS+MG, and TMOS. A machine learning (ML) model using the XGBoost algorithm was constructed using a subset (n = 858) of ICI-MC with complete data availability (age, sex, co-reactions, cancer/ICI type, and MC timing) for MC fatality prediction with an 80/20% data split for training and internal testing. An external public real-world dataset (n = 28) of ICI-MC was used for independent validation of our fatality ML prediction model. Results: Among a total of 4,950 ICI-MC/MS/MG cases, we identified 2,641 ICI-MC cases, of which 1,911 (72.4%) were MC alone and 730 (27.6%) were overlapping with MS and/or MG (MC+MS = 364, 13.8%; MC+MG = 159, 6%; TMOS = 207, 7.8%). TMOS occurred predominantly in melanoma (35.8%) and was more likely in males (64.7% vs 52.9%, p-value = 0.0049) treated with ICI dual therapy (25.1% vs 20.4%, p-value = 0.0030) compared with MC alone. Hepatitis was the most common irAE co-occurring in cases with TMOS (n = 30, 14.5%). MC-specific fatality rates were higher in TMOS (38%) compared to MC alone (21.2%), MC+MS (22.5%), or MC+MG (25.7%). MC alone had a later onset from ICI start than MC+MS, MC+MG, and TMOS (median time-to-MC: 60.8, 27, 27, and 26 days, respectively; p < 0.05). Early-onset MC within the first month of ICI initiation was independently associated with increased MC fatality after adjustment for age, ICI regimens, cancer type, and co-reactions (≤1 vs 1-3 months - OR: 0.41, 95% CI [0.22-0.73], p-value = 0.0036; ≤1 vs 3-12 months - OR: 0.44, 95% CI [0.21-0.86], p-value = 0.0212). Our final MC fatality classifier achieved an AUC of 0.79, 0.75, and 0.85 with the training (n = 686), internal testing (n = 172), and external independent validation (n = 28) datasets, respectively, with the top predictive features for MC fatality being early-onset MC occurrence (within the first month of ICI start) and cardiorespiratory co-reactions. Conclusion: This is the largest global dataset to characterize TMOS and MC fatality. Our data shows that TMOS represents a uniquely fatal phenotype of ICI-MC with a higher tendency in cancers with wider ICI use. Regarding MC fatality, we show MC timing to be a critical determinant of fatality, specifically within the first month of ICI start, demonstrating the need for refined screening and monitoring strategies within this time window.