Objectives: Rituximab is emerging as a promising therapeutic option for systemic sclerosis-associated interstitial lung disease (SSc-ILD). However, little is known about factors that predict the efficacy of rituximab in SSc-ILD.
Methods: A post-hoc analysis was performed on prospective data from 48 patients with SSc-ILD in the double-blind, randomized, placebo-controlled DESIRES trial. A total of 28 baseline factors were selected as candidates to predict the efficacy of rituximab on the percentage of predicted forced vital capacity (ppFVC) at 24 weeks. A machine learning causal tree algorithm was used to explore the combination of predictors to identify subpopulations with a good response to rituximab.
Results: Serum levels of C-reactive protein (CRP) and Krebs von den Lungen-6 (KL-6) were selected as branches of the decision tree to stratify patients into three subpopulations. In the subpopulation with serum CRP levels ≥0.055 mg/dl, ΔppFVC was significantly higher in the rituximab group than in the placebo group [difference 8.01% (95% CI: 4.40%, 11.62%)]. In the subpopulation with serum CRP levels <0.055 mg/dl and serum KL-6 levels ≥364 U/ml, ΔppFVC was comparable between the two groups [difference 2.47% (95% CI: -1.99%, 6.92%)]. In the subpopulation with serum CRP levels <0.055 mg/dl and serum KL-6 levels <364 U/ml, ΔppFVC was significantly lower in rituximab than in placebo [difference -6.85% (95% CI: -10.80%, -2.91%)].
Conclusion: Even slight elevations in serum CRP levels are associated with the improvement in ppFVC and may serve as predictors of rituximab efficacy in SSc-ILD.
Keywords: interstitial lung disease; machine learning; rituximab; systemic sclerosis.
© The Author(s) 2024. Published by Oxford University Press on behalf of the British Society for Rheumatology.