Background: It is currently impossible to predict the prognosis of patients with juvenile dermatomyositis (JDM). The aim of this study was to find clinical features most strongly associated with outcome variables in JDM as a first step towards tailor-made treatment.
Methods: In a large, prospectively followed, multicenter cohort study of 340 patients with JDM, each contributing multiple visits, a Bayesian model of disease activity was developed, using the four continuous outcome variables creatine kinase (CK), childhood myositis assessment score (CMAS), manual muscle testing of 8 muscle groups (MMT8) and the physician's global assessment of disease activity (PGA). Covariates were clinical signs and symptoms. Correlations among visits of the same patient were resolved by introducing subject-specific random effects.
Results: Myalgia and dysphonia were associated with worse disease activity according to all outcome variables. Periorbital rash, rash on the trunk, rash over large joints, nail fold changes and facial swelling were associated with higher PGA. Notably, periorbital rash was also associated with higher CK and lower CMAS and nail fold changes with lower CMAS. Contractures were associated with lower CMAS and MMT8 and higher PGA. Patients with higher CMAS exhibited a higher MMT8 as well. PGA had the highest probability among the four outcome variables of being abnormal even if the other three outcome variables were normal.
Conclusions: The signs and symptoms associated with disease activity could be used to stratify patients and adapt treatment plans to disease activity. The correlation between CMAS and MMT8 and the unique information captured by PGA implied that PGA should be maintained as an outcome variable, whereas CMAS and MMT8 might be simplified.
Keywords: Bayesian model; Clinical associations; Disease activity; Juvenile dermatomyositis; Longitudinal data.