Risk Prediction Modeling Based on a Combination of Initial Serum Biomarker Levels in Polymyositis/Dermatomyositis–Associated Interstitial Lung Disease

OBJECTIVE: To establish predictive models for mortality in patients with polymyositis/dermatomyositis-associated interstitial lung disease (PM/DM-ILD) using a combination of initial serum biomarker levels. METHODS: The Multicenter Retrospective Cohort of Japanese Patients with Myositis-Associated ILD (JAMI) database of 497 incident cases of PM/DM-ILD was used as a derivation cohort, and 111 cases were additionally collected as a validation cohort. Risk factors predictive of all-cause mortality were identified by univariate and multivariable Cox regression analyses using candidate serum biomarkers as explanatory variables. The predictive models for mortality were generated in patients with and those without anti-melanoma differentiation-associated gene 5 (MDA-5) antibody, using a combination of risk factors. Cumulative survival rates were assessed using Kaplan-Meier analysis, and were compared between subgroups using the Breslow test. RESULTS: In the derivation cohort, C-reactive protein (CRP) and Krebs von den Lungen 6 (KL-6) levels were identified as independent risk factors for mortality in both anti-MDA-5-positive and anti-MDA-5-negative patients. We then developed a prediction model based on anti-MDA-5 antibody status, CRP level, and KL-6 level, termed the "MCK model," to identify patients at low (<15%), moderate (15-50%), or high (≥50%) risk of mortality, based on the number of risk factors. The MCK model successfully differentiated cumulative survival rates in anti-MDA-5-positive patients (P < 0.01 for low versus moderate risk and P = 0.03 for moderate versus high risk) and in anti-MDA-5-negative patients (P < 0.001 for low versus moderate risk). The utility of the MCK model was replicated in the validation cohort. CONCLUSION: Our findings indicate that an evidence-based risk prediction model using CRP and KL-6 levels combined with anti-MDA-5 antibody status might be useful for predicting prognosis in patients with PM/DM-ILD.