Two clinical subgroups of immune checkpoint inhibitor-induced inflammatory arthritis determined by latent class analysis

Abstract

Background

Immune checkpoint inhibitors (ICI) for cancer treatment can cause inflammatory arthritis (IA). Since ICI-IA has a unique pathogenesis, applying categories of traditional IA may be of limited use.

Methods

Participants were ≥18 years old, treated with anti-PD-1, anti-PD-L1, and/or anti-CTLA-4 agents, and had ICI-IA diagnosed by a rheumatologist. We clustered patients using latent class analysis (LCA) applied with phenotypic data from the baseline rheumatology visit. The Bayesian Information Criteria (BIC) was used to select the number of phenotypes. We compared demographics, cancer type and treatments, and IA clinical features and treatments between the estimated phenotypes. Finally, we explored differences in cytokine levels and the presence of shared epitope between the groups.

Results

Twenty variables were used to estimate latent classes. Two distinct phenotypes were indicated by the BIC; 77 patients were estimated to be the first phenotype and 49 in the second phenotype. The statistically significant features that distinguished the phenotypes included higher levels of all components of the CDAI, more stiffness, and more small and upper extremity joint involvement for phenotype 2. Patients in phenotype 2 were marginally more likely to require steroids during their course. There were no significant differences in cancer type, stage or ICI treatment between the phenotype groups. Baseline levels of VEGF-A were higher in phenotype 2.

Conclusions

Two separate phenotypes of ICI-IA were identified using LCA, the second having a more severe polyarthritis at baseline and involving the upper extremities. These subgroups provide an opportunity to identify corresponding biomarkers to predict disease outcomes.