Background Immune-related adverse events (irAEs) induced by immune checkpoint inhibitor (ICI) therapy can involve multiple organ systems, of which cutaneous irAEs (c-irAEs) are the most common.^1–5 Understanding co-occurrence patterns and prognostic outcomes of irAEs is critical for immunotherapy management. However, previous studies have been limited by cohort size, thus limiting generalizability.^6–8 In addition, the prior study approaches primarily utilized pairwise comparison analyses, which limited the examination of irAE relationships to two organ systems at a time. In this study, we leverage a population-level database and compare results to a multi-institutional cohort from a tertiary-level academic healthcare system to investigate the co-occurrence patterns irAEs and their impact on immunotherapy outcomes through non-negative matrix factorization (NMF),⁹ which allows for multi-organ analyses.

Methods After propensity-matching in a 1:2 ratio based on demographic and cancer-related variables and exclusion of ambiguous cases, the study analyses included 9,273 patients in the Mass General Brigham Healthcare System (MGB) cohort and 23,689 patients in the TriNetX network (figure 1). The identification of irAEs was based on ICD codes. Pairwise correlation analysis and NMF were conducted to investigate the co-occurrence patterns. Multivariate landmark analyses were conducted to evaluate the associated impact on overall survival, adjusting for demographics, cancer type, and ICI type.

Results Characteristics of the TriNetX and MGB are shown in (table 1). Pairwise co-occurrence analyses showed patients with c-irAEs were at increased risk of developing an irAE in nine of the eleven organ systems evaluated. The co-occurrence of c-irAEs and non-cutaneous irAEs (nc-irAEs) was associated with improved survival (HR:0.68, CI, 0.61–0.76; p<0.001) (table 2). NMF identified four unique patient clusters, of which three were consistent between the TriNetX and MGB cohorts: c-irAEs, endocrine irAEs, and multiple internal organ irAEs (comprising mostly neurologic, respiratory, gastrointestinal irAEs) (figure 2). The Cutaneous and endocrine clusters showed strongly favorable prognoses across all landmark times (table 3). The endocrine-dominant cluster displayed a better prognosis (MGB: HR=0.48, p<0.001; TriNetX: HR=0.58, p<0.001) compared to the Cutaneous-dominant cluster (MGB: HR=0.55, p<0.001; TriNetX: HR=0.65, p<0.001).

Conclusions Our study demonstrates that patients who develop c-irAEs are at significantly increased risk of developing toxicities in other organs. This emphasizes the importance of monitoring, diagnosing, and managing c-irAEs given their valuable prognostic benefit. In addition, we specifically found a significant survival benefit among patients who develop cutaneous and endocrine irAEs. This may suggest underlying mechanisms that differ from other ICI and organ system interactions, which will require future studies to elucidate.

References

1. Ramos-Casals M, Brahmer JR, Callahan MK, et al. Immune-related adverse events of checkpoint inhibitors. Nat Rev Dis Primers. 2020;6(1):1–21. Doi:10.1038/s41572–020-0160–6

2. Puzanov I, Diab A, Abdallah K, et al. Managing toxicities associated with immune checkpoint inhibitors: consensus recommendations from the Society for Immunotherapy of Cancer (SITC) Toxicity Management Working Group. J Immunother Cancer. 2017;5(1):95. Doi:10.1186/S40425–017-0300-Z

3. Brahmer JR, Lacchetti C, Schneider BJ, et al. Management of immune-related adverse events in patients treated with immune checkpoint inhibitor therapy: American society of clinical oncology clinical practice guideline. Journal of Clinical Oncology. 2018;36(17):1714–1768. Doi:10.1200/JCO.2017.77.6385

4. Le TK, Brown I, Goldberg R, et al. Cutaneous Toxicities Associated with Immune Checkpoint Inhibitors: An Observational, Pharmacovigilance Study. J Invest Dermatol. 2022;142(11):2896–2908.e4. doi:10.1016/J.JID.2022.04.020

5. Wongvibulsin S, Pahalyants V, Kalinich M, et al. Epidemiology and risk factors for the development of cutaneous toxicities in patients treated with immune-checkpoint inhibitors: A United States population-level analysis. J Am Acad Dermatol. 2022;86(3):563–572. Doi:10.1016/J.JAAD.2021.03.094

6. Chieng JHL, Htet ZW, Zhao JJ, et al. Clinical Presentation of Immune-Related Endocrine Adverse Events during Immune Checkpoint Inhibitor Treatment. Cancers (Basel). 2022;14(11):2687. Doi:10.3390/CANCERS14112687

7. Yamada K, Nakamura M, Yamamura T, et al. Clinical characteristics of gastrointestinal immune-related adverse events of immune checkpoint inhibitors and their association with survival. World J Gastroenterol. 2021;27(41):7190. Doi:10.3748/WJG.V27.I41.7190

8. Asdourian MS, Shah N, Jacoby T V, et al. Development of multiple cutaneous immune-related adverse events among cancer patients after immune checkpoint blockade. J Am Acad Dermatol. 2023;88(2):485–487. Doi:10.1016/j.jaad.2022.06.030

9. Pedregosa FABIANPEDREGOSA F, Michel V, Grisel OLIVIERGRISEL O, et al. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research. 2011;12(85):2825–2830. Accessed February 24, 2023. http://jmlr.org/papers/v12/pedregosa11a.html

Ethics Approval Reviewed and approved by Mass General Brigham IRB (Protocol #2020P002179)

• Download figure
• Open in new tab
• Download powerpoint

Abstract 1241 Figure 1

The study population

• Download figure
• Open in new tab
• Download powerpoint

Abstract 1241 Figure 2

The results of NMF and hierarchical clustering on the consensus matrix. (A, B) The NMF results and clusters on the TriNetX cohort; C-D: The NMF results and clusters on the MGB cohort. The NMF decomposed the irAE count matrix into two low-rank matrices, representing the organ-level irAE factors (referred to as ‘basis’; A, C) and the weights of irAE factors that comprise each patient (referred to as ‘weight’; B and D), respectively. For the basis matrix, rows are organ systems, and columns are the irAE factors, each named by the dominant organ systems. For the weight matrix, columns are patients, rows are irAE factors, and the clustering results are presented at the top. Similar patterns were observed between the two basis matrices (A, C). Patients in both cohorts could be categorized into four groups (B, D) Cutaneous, Cutaneous + Internal