Background/Purpose: Interstitial lung disease (ILD) is one of the leading causes of mortality in patients with polymyositis or dermatomyositis (PM/DM). Since clinical courses and outcomes of ILD are highly variable among PM/DM patients with ILD, disease subsetting is essential in deciding management regimens, by offering opportunity to identify patients who have a greater risk of mortality, especially early in the disease course. In this regard, myositis-specific autoantibodies (MSAs) correlate with unique subsets of PM/DM-associated ILD, but potential utility of other serum biomarkers routinely measured in clinical practice, such as CRP and ferritin, in predicting prognosis is also reported. The aim of this study is to establish predictive modeling of mortality in patients with PM/DM-associated ILD using a large cohort data.

Methods: This study was conducted using a database of a multicenter retrospective cohort of patients with PM/DM-ILD (JAMI cohort), which involved 44 institutions across Japan. We enrolled 487 patients based on adult-onset definite or probable PM/DM including clinically amyopathic DM (CADM), ILD confirmed by imaging study, and availability of serum samples at diagnosis. Anti-melanoma differentiation-associated gene 5 (MDA5) and anti-amynoacyl tRNA synthetase (ARS) antibodies were detected by enzyme linked immunosorbent assay and RNA immunoprecipitation, respectively. CRP, ferritin, KL-6 and surfactant protein-D (SP-D) were chosen as serum biomarkers for PM/DM-ILD. Independent risk factors for all-cause mortality were identified by Cox regression analysis using MSAs and serum biomarkers, including CRP, ferritin, KL-6 and surfactant protein-D (SP-D)  as explanatory variables. The backward selection method was applied; explanatory variables were eliminated when p value was > 0.10.

Results: The overall survival rate was 83% at 1 year. The survival rate was significantly lower in patients with anti-MDA5 than in those without (P<0.0001). The cut-off values of individual serum biomarkers for predicting mortality determined by the receiver operating characteristic curve were CRP ≥1 mg/dl, ferritin ≥500 ng/ml, KL-6 ≥1000 mg/dl and SP-D <100 ng/ml. The presence of anti-MDA5 (hazard ratio [HR] = 3.0, 95% confidence interval 1.6-5.7), CRP ≥1 mg/dl (HR = 2.4, 1.4-4.0), KL-6 ≥1000 mg/dl (HR = 2.0,  1.3-3.3) and ferritin ≥500 ng/ml (HR = 1.8, 1.0-3.2) were identified as risk factors for poor prognosis. Our modeling showed that the predicted mortality rates were 1.8%, 8%, 22%, 44% and 54% in patients with zero, one, two, three and all four risk factor score, respectively (Figure 1).

Conclusion: We successfully generated predictive modeling of mortality in patients with PM/DM-associated ILD using convenient serum biomarkers. This model is potentially useful in identifying the patients with high mortality risk, which apparently require intensive treatment