0883: Predictive Significance of Serum Proteins for the Course of Systemic Sclerosis-Related Interstitial Lung Disease in the Multicenter CONQUER Cohort

Background/Purpose: The course of interstitial lung disease (ILD) in systemic sclerosis (SSc) is highly variable and difficult to predict using clinical variables alone. Therefore, there is a need for reliable biomarkers to improve prognostication. The Scleroderma Lung Study (SLS) II showed that Krebs von den Lungen 6 (KL-6), CCL18, IFNγ-inducible 10-kDa protein (IP-10), and monokine induced by IFNγ (MIG) were significant predictors of ILD progression. In this study, we aimed to validate these findings and build a multivariable model to predict the course of ILD in SSc based on longitudinal data in a US-based multicenter observational cohort.

Methods: Patients with SSc enrolled in the CONQUER cohort were included in this analysis. Enrollment in CONQUER required fulfillment of the 2013 ACR/European League Against Rheumatism classification criteria and disease duration of less than five years from the onset of the first non-Raynaud’s symptom. For the current study, we included only those patients who had interstitial lung disease (ILD) confirmed by high-resolution computed tomography (HRCT) and who were treated with mycophenolate mofetil (MMF) between the baseline and 12-month visits. The study visits were scheduled every 6 months. Blood samples were collected at baseline. Sixty-two proteins were measured as part of a multiplex assay. We used linear mixed-effects models to evaluate the predictive significance of individual cytokines for ILD progression. Follow-up FVC% was the outcome, with baseline cytokine levels, MMF treatment status, baseline FVC%, and time as fixed effects. Random intercepts and slopes accounted for between-patient variability in baseline FVC% and its change over time. A Bayesian regularized linear mixed-effects model was used to develop a prediction model that included clinical variables and serum protein levels.

Results: A total of 122 patients were included, of whom 100 (82%) were female, with a median age of 52 years and a mean disease duration of 2.7 years. Higher baseline levels of IP-10 (b = 0.0020, P = 0.013) and MIG (b = 0.0007, P = 0.023) were associated with higher FVC% over time, indicating a better ILD course. In contrast, higher KL-6 levels (b = –0.0020, P = 0.015) predicted lower FVC% over time. CCL18 was not significantly associated with the FVC% trajectory. In the exploratory analysis, α2-macroglobulin (α2M, b = -3.29, P < 0.001) and serum amyloid P component (SAP, b = -0.49, P = 0.007) also showed predictive significance. In the multivariable Bayesian model, baseline FVC% had the highest estimate among clinical variables (β = 14.03), while α2-macroglobulin had the highest among protein predictors (β = –1.39). The R2 of the model, including clinical predictors and a selected number of serum proteins, was 0.765.

Conclusion: This study confirmed the predictive significance of IP-10, MIG, and KL-6 for the course of SSc-ILD and identified α2M and SAP as potential biomarkers. We also built a multivariable model including clinical and protein variables to predict the SSc-ILD course.