Triple M overlap syndrome (TMOS): Evaluating immune checkpoint inhibitor-related overlap syndrome of myocarditis, myositis and myasthenia gravis using an international pharmacovigilance database.

2643 Background: Myocarditis (MC) secondary to ICIs is a severe immune-related adverse event (irAE) and may co-occur with myositis (MYS) and myasthenia gravis (MG), presenting as the triple M overlap syndrome (TMOS). Limited data have been reported on TMOS as a fatal irAE complication. We sought to understand the clinical complexities of TMOS and applied machine learning approaches using data from Vigibase, the WHO global database of individual case safety reports (ICSR). Methods: VigiBasewas used to extract data (cut off Jan 11, 2023) from ICSR for ICI-treated patients (pts) with MYS or MG either alone or co-occurring with MC. Pts were categorized into three groups: MG + MC, MYS + MC, and TMOS. Pearson Chi-squared test, Wilcoxon rank-sum/ANOVA test, and log-rank test were used to compare categorical, continuous, and time-to-event data. A machine learning model using the Weighted Subspace Random Forest (WSRF) classification was constructed to predict the occurrence of TMOS using available clinical features. Results: 1726 cases with ICI-induced MYS or MG were retrieved, of which 358 (20.7%) pts had co-occurring MC. Median age was 71 years (IQR: 65-77), with 66.7% males. MYS + MC, MG + MC, and TMOS were reported in 196 (54.7%), 59 (16.5%), and 103 (28.8%) pts, respectively. TMOS had a higher proportion of anti-PD-1/CTLA-4 combinations compared to MG + MC or MYS + MC (29.1% vs. 11.9% vs. 26%, p < 0.001). Melanoma was the most common cancer type (n = 107; 29.9%) and had a higher frequency of TMOS (39%) compared to MG + MC (29%) or MYS + MC (26%). There was no significant difference in median time-to-MC occurrence after adjustment for treatment duration between all three groups (TMOS: 24.5, MG + MC: 26, MYS + MC: 23 days, p = 0.81). Concurrent ICI-induced hepatitis had a significantly higher frequency (p = 0.035) in TMOS compared to the other two groups (20.4% vs 10.2% vs 10.2%). Fatality rates were significantly higher for TMOS compared to MC with MYS or MG (TMOS: 43.8% vs MC + MYS/MG: 29.8%, p = 0.033). WSRF modeling demonstrated an accuracy score of 0.87 (95% CI: 0.82 – 0.91, sensitivity = 0.66, specificity = 0.96) in the training set and 0.68 (95% CI: 0.57 – 0.77, sensitivity = 0.23, specificity = 0.81) in the testing set with longer treatment duration and older age having the highest association with TMOS occurrence. Conclusions: This is the largest dataset to date demonstrating that TMOS is associated with significantly higher mortality than MC with MYS or MG with risk factors of older age and treatment with doublet ICI, especially in pts with melanoma. These findings underscore the urgency of further research to identify the underlying biology, scale, and risk factors, as well as explore early immunosuppressive strategies in addition to steroids for improved outcomes in ICI-related TMOS.