Objective: There is substantial heterogeneity among the phenotypes of patients with anti-melanoma differentiation-associated gene 5 antibody-positive (anti-MDA5+) dermatomyositis (DM), hindering disease assessment and management. This study aimed to identify distinct phenotype groups in patients with anti-MDA5+ DM and to determine the utility of these phenotypes in predicting patient outcomes.
Methods: A total of 265 patients with anti-MDA5+ DM were retrospectively enrolled in the study. An unsupervised hierarchical cluster analysis was performed to characterize the different phenotypes.
Results: Patients were stratified into 3 clusters characterized by markedly different features and outcomes. Cluster 1 (n = 108 patients) was characterized by mild risk of rapidly progressive interstitial lung disease (RPILD), with the cumulative incidence of non-RPILD being 85.2%. Cluster 2 (n = 72 patients) was characterized by moderate risk of RPILD, with the cumulative incidence of non-RPILPD being 73.6%. Patients in cluster 3 (n = 85 patients), which was characterized by a high risk of RPILD and a cumulative non-RPILD incidence of 32.9%, were more likely than patients in the other 2 subgroups to have anti-Ro 52 antibodies in conjunction with high titers of anti-MDA5 antibodies. All-cause mortality rates of 60%, 9.7%, and 3.7% were determined for clusters 3, 2, and 1, respectively (P < 0.0001). Decision tree analysis led to the development of a simple algorithm for anti-MDA5+ DM patient classification that included the following 8 variables: age >50 years, disease course of <3 months, myasthenia (proximal muscle weakness), arthritis, C-reactive protein level, creatine kinase level, anti-Ro 52 antibody titer, and anti-MDA5 antibody titer. This algorithm placed patients in the appropriate cluster with 78.5% accuracy in the development cohort and 70.0% accuracy in the external validation cohort.
Conclusion: Cluster analysis identified 3 distinct clinical patterns and outcomes in our large cohort of anti-MDA5+ DM patients. Classification of DM patients into phenotype subgroups with prognostic values may help physicians improve the efficacy of clinical decision-making.
© 2022 American College of Rheumatology.