POS0833 A RETROSPECTIVE COHORT STUDY IN CHINESE PATIENTS WITH ADULT POLYMYOSITIS AND DERMATOMYOSITIS: RISK OF COMORBIDITIES AND SUBCLASSIFICATION USING MACHINE LEARNING

Abstract

Background:Idiopathic inflammatory myopathy (IIM), also known as myositis, refers to a group of heterogeneous disorders including polymyositis (PM), dermatomyositis (DM), inclusion body myositis and immune-mediated necrotising myopathy. Phenotype, pathogenesis, and prognosis vary due to multi-organ involvement and comorbidities. With the clinical application of MSAs, a new classification system for myositis was explored to reduce confusion between subgroups. But it is far from showing the full picture of myositis due to high heterogeneity. Therefore, it is necessary to systematically evaluate the relevant risk factors of myositis for ILD, other rheumatic diseases, and malignancy for better clinical vigilance. And further exploring the subclassification of myositis is critical.Objectives:To identify the risk factors in Chinese patients with adult polymyositis and dermatomyositis for their comorbidities and explore a subclassification system.Methods:Clinical records of 397 patients with idiopathic inflammatory myopathies were retrospectively reviewed. Logistic regression was used to identify potential risk factors for interstitial lung disease (ILD), other rheumatic diseases, and malignancy after bivariate analysis. Hierarchical clustering and decisional tree were utilized to identify subgroups and explore a subclassification system.Results:A total of 119 polymyositis and 191 dermatomyositis patients were included. Anti-PM/Scl, anti-Ro52, anti-aminoacyl-tRNA synthetase and anti-MDA5 (adjusted odds ratios (AOR)=4.779, 1.917, 5.092 and 7.714 respectively) antibodies were risks (p<0.05), whereas overlapping malignancy was protective (AOR=0.107; p=0.002) for ILD across polymyositis, dermatomyositis and the total group. In subgroup models, Raynaud’s phenomenon, arthralgia and semi-quantitative anti-nuclear antibody (AOR=51.233, 4.261, 3.047 respectively) were risks for other overlapping rheumatic diseases (p<0.05). For overlapping malignancy, male and anti-TIF1γ antibodies (AOR=2.533, 16.949) were risks (p<0.05), whereas disease duration and combination of ILD (AOR=0.954, 0.106) were protective in the total group (p<0.05); while anti-NXP2 antibodies were identified as risk factors (AOR=73.152; p=0.038) in polymyositis. Hierarchical clustering suggested a subclassification with 6 subgroups: malignancy overlapping dermatomyositis, classical dermatomyositis, polymyositis with severe muscle involvement, dermatomyositis with ILD, polymyositis with ILD, and overlapping of myositis with other rheumatic diseases according to the characteristics of grouped patients. Accuracy of the classification and regression trees model was 0.768 (95% CI 0.711 to 0.819) on training set and 0.633 (95%CI 0.499 to 0.754) on test set.Conclusion:Accompanying ILD, other rheumatic diseases and malignancy are strongly associated with clinical manifestation and myositis-specific or myositis-associated autoantibodies among Chinese polymyositis and dermatomyositis patients. The subclassification system proposed a more precise phenotype defining toward stratified treatments.Acknowledgements:The study was supported by the Natural Science Foundation of China [No. 81803932] and the Natural Science Foundation of Guangdong Province [No. 2018030310025 and 2017A030313868]. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscriptDisclosure of Interests:None declared