Factors associated with interstitial lung disease among patients with idiopathic inflammatory myopathies

Abstract Objectives Idiopathic inflammatory myopathies (IIM) are heterogeneous disorders that often affect the lungs as IIM-associated interstitial lung disease (IIM-ILD). We used Meta-ANalysis of Transethnic Associations (MANTRA) and machine learning methods to evaluate predictors of IIM-ILD. Methods Subjects (N = 450) were enrolled in studies at the National Institutes of Health. We studied adult (N = 262) and juvenile (N = 188) dermatomyositis (N = 276), polymyositis (N = 109), and overlap myositis (N = 61) patients with clinical, autoantibody, human leucocyte antigen (HLA), and single nucleotide polymorphism (SNP) data, with (N = 162) or without (N = 288) ILD. Logistic regression and MANTRA analyses were used to evaluate the associations of SNPs (Muc5b rs35705950, TOLLIP rs5743890 and rs3750920, TLR5 rs5744168, and TERT rs2736100) previously identified as risks for idiopathic pulmonary fibrosis (IPF). Classification and Regression Tree (CART) and gradient boosting machine learning were used to simultaneously evaluate the clinical, autoantibody, HLA, and SNP data for their relative predictive power of ILD. Results Smoking status, older age, African American heritage, and certain HLA genes were associated with IIM-ILD, but anti-synthetase, myositis-associated, and anti-MDA5 autoantibodies showed the strongest risk associations, with an increased odds of ILD by up to 20-fold. Conversely, anti-signal recognition particle, anti-TIF1 (P155/140), and anti-NXP2 autoantibodies showed the strongest protective effects, with decreased odds of ILD by up to 40%. The effects of some HLA allele groups and IPF SNPs on ILD were inconsistent and weaker. Conclusions This sample of IIM patients showed autoantibodies to be the strongest predictive or protective factors for ILD, yet the full range of associations of IIM-ILD remain undefined.