A multicentre study for clinical phenotype prediction in juvenile dermatomyositis: categorical principal component analysis-based hierarchical clustering.

Abstract

Introduction: Juvenile dermatomyositis (JDM) is the most common inflammatory myopathy in childhood and exhibits a heterogeneous disease course. This study aimed to analyse and identify phenotypic clusters by examining the laboratory findings, nailfold capillaroscopy results, and myositis-specific autoantibodies (MSAs) in patients with JDM.

Material and methods: This retrospective cohort study included data from patients with JDM treated at the Paediatric Rheumatology Departments of 14 advanced health centres in Turkey. A categorical principal component analysis (CATPCA)-based hierarchical cluster analysis method was employed for clustering.

Results: A total of 176 JDM patients were enrolled, and 5 phenotypic clusters were identified using 23 categorical variables. These clusters were interpreted as follows: Cluster A with severe muscle weakness and oesophageal involvement requiring intensive immunosuppressive treatment; Cluster B with amyopathic/hypomyopathic patients; Cluster C with skin manifestations and lung involvement; Cluster D with complicated skin manifestations; and Cluster E with classic JDM. The clinical and laboratory findings and treatments of these 5 clusters were compared. Fatigue, myalgia, photosensitivity, Raynaud's phenomenon, and the use of pulse glucocorticosteroids, intravenous immunoglobulin, and cyclophosphamide treatments differed between the groups (p < 0.001, p = 0.002, p = 0.015, p = 0.036, p = 0.002, p = 0.006, and p = 0.024, respectively). Myositis-specific autoantibodies results were available for 119 patients (65.3%). The most frequent MSAs were antinuclear matrix protein 2 (26.1%) and anti-transcription intermediary factor 1 (20.9%). However, no significant differences were found in MSAs or nailfold capillaroscopy findings.

Conclusions: We identified 5 clusters based on patient symptoms and findings. The identification of these 5 clusters can guide more effective treatment strategies in clinical practice. Additionally, these approaches may contribute to improving patients' quality of life and long-term outcomes by increasing the feasibility of individualised treatment.