Background/Purpose: Use of immune checkpoint inhibitors (ICIs) can lead to a broad spectrum of immune-related adverse events (irAEs) affecting multiple organ systems. No clear pattern for organs and systems most likely to be affected by simultaneous toxicity has emerged in the published reports of multisystem irAEs. Understanding co-occurrence patterns and prognostic outcomes of irAEs is critical for immunotherapy management. Herein, we focused our analysis on patients from a multi-ethnic Southeast Asian cohort with the intention of identifying distinct phenotypes of irAEs associated with survival after ICIs.

Methods: In total, 870 cancer patients treated with ICIs were recruited into this prospective, multi-institution cohort study of which 341 patients had irAE(s). We used an unbiased approach to identify patients with different patterns of irAEs and to see if these clusters predict outcome. The irAE manifestations were considered as binary (yes/no) variables and used in a latent class analysis (LCA) to identify different phenotypes of multisystem irAEs. Multivariable Cox regression was performed to determine which phenotypic cluster was associated with shorter progression-free and overall survival (OS). An odds ratio (OR) cutoff of ≥ 15.0 was used to define irAEs that drove the clustering. irAEs do not have to occur within the same timeframe to be considered as being multisystem.

Results: Three clusters were identified using LCA (Table 1). All patients in Cluster 1 (C1) (n = 118) had dermatologic (100%) irAEs. With respect to Cluster C1, Cluster 2 (C2) (n = 40) had predominant hepatic (32.5%, OR = 116.3) and pulmonary (77.5%, OR = 785.8) irAEs while Cluster 3 (C3) (n = 183) was characterized by multisystem irAEs such as cardiac (7.7%, OR = 20.3), neurologic (7.7%, OR = 20.3), cytokine release syndrome (12.0%, OR = 33.0), rheumatic (16.9%, OR = 49.0) and hepatic (17.5%, OR = 50.8) toxicities. Comparing patients without irAEs (n = 526), C1, C2 and C3, they were significantly different in terms of concomitant use of proton pump inhibitors, week 6 neutrophil count, week 6 neutrophil-to-lymphocyte ratio, number of cycles of ICI, cumulative dose of ICI and cancer treatment responses. Multivariable COX regression analysis revealed that C2 had the best OS. Compared to C2, no irAE [adjusted hazard ratio (HR) = 6.3, 95% confidence interval (CI) 2.1-18.5, p < 0.001], C1 (adjusted HR = 3.5, 95% CI 1.1-11.1, p = 0.036) and C3 (adjusted HR = 4.2, 95% CI 1.4-12.4, p = 0.010) were associated with shorter OS after adjusting for body mass index, performance status, primary diagnosis and number of cycles of ICI (Figure 1).

Conclusion: Our study reveals differences in patterns of clustering of irAEs. In particular, C3 tends to have toxicities in multiple organ systems and is associated with shorter OS. In addition, we found significant survival benefits for C1 and C2. This may suggest differences in terms of the interactions between the degree of immune activation, immunosuppression and treatment responses that will require further studies to elucidate.

Table 1. Clustering of irAEs in various organ systems using LCA.

Table 2. Clinical characteristics of patients with and without irAEs.

Figure 3. Kaplan–Meier curves for OS according to the different phenotypic clusters.

« Back to ACR Convergence 2025

ACR Meeting Abstracts - https://acrabstracts.org/abstract/latent-class-analysis-identifies-distinct-phenotypes-of-multisystem-immune-related-adverse-events-predictive-of-survival-after-immune-checkpoint-inhibitors/