Objectives: Anti-melanoma differentiation-associated gene 5 (MDA5) positive DM is a life-threatening disease often complicated with rapidly progressive interstitial lung disease (ILD). This study aimed to establish and validate a clinical prediction model for 6-month all-cause mortality in Chinese patients with anti-MDA5 positive DM-ILD.
Methods: We conducted a retrospective observational study using a single-centre derivation cohort and a multicentre validation cohort. Hospitalized DM patients with positive anti-MDA5 antibody and ILD course ≤3 months on admission were included. Patients' baseline characteristics were described and compared between the deceased and survivors by univariable Cox regression. Optimal cut-off values were defined by the 'survminer' R package for significant continuous variables. Independent prognostic factors were determined by the final multivariable Cox regression model chosen by backward stepwise algorithm, which could be reproduced in both cohorts. The Kaplan-Meier survival analyses based on the derived predictor were conducted.
Results: A total of 184 and 81 eligible patients were included with a cumulative 40.8 and 40.7% 6-month mortality in the derivation and validation cohorts, respectively. Based on multivariable Cox regression, the prognostic factor at baseline was identified and validated as three-category forced vital capacity (FVC)%: FVC% ≥50%, FVC% <50%, unable to perform. This significantly distinguishes three risk stages with mortalities of 15.3, 46.8, 97.4% in the derivation cohort, and 14.9, 58.3, 86.4 in the validation cohort, respectively (all P <0.05).
Conclusion: The validated FVC%-based categorical predictor in anti-MDA5 positive DM-ILD is helpful for risk stratification in clinical practice and might facilitate cohort enrichment for future trials.
Keywords: anti-MDA5 antibody; dermatomyositis; forced vital capacity; interstitial lung disease; prognosis.
© The Author(s) 2021. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For permissions, please email: journals.permissions@oup.com.