Objectives: Sjögren's disease (SjD) is clinically and biologically heterogeneous, and no immunomodulatory drug has yet demonstrated efficacy in phase 3 trials. We previously identified 4 transcriptomic endotypes in SjD patients using whole-blood RNA sequencing. We hypothesised that these endotypes may predict differential therapeutic responses.
Methods: We analysed clinical, biological, and transcriptomic data from 3 randomised controlled trials evaluating hydroxychloroquine-leflunomide (HCQ-LEF; n = 18; RepurpSS-I trial), rituximab (RTX; n = 56; TRACTISS trial), and abatacept (n = 117). Patients were assigned to the 4 endotypes using semisupervised uniform manifold approximation and projection combined with a support vector machine model. Demographics, disease activity, and therapeutic response, as defined by the Sjögren Tool for Assessing Response index, were compared across clusters in both pooled and individual trial analyses.
Results: Of 170 patients, 81, 24, 80, and 6 were classified into clusters 1 to 4, respectively. European Alliance of Associations for Rheumatology (EULAR) Sjögren's Syndrome Patient Reported Index scores were comparable across clusters, while the EULAR Sjögren's Syndrome Disease Activity Index was significantly higher in clusters 3 and 4 vs clusters 1 and 2 (P = .003). In pooled analyses, patients in cluster 1 had significantly greater response rates with active treatment vs placebo (61.5% vs 32.6%: P = .016), with a similar trend in the RTX trial. In contrast, patients in cluster 3 benefitted from HCQ-LEF, whereas cluster 2 (healthy-like patients) showed no significant response to any therapy.
Conclusions: Transcriptomic stratification of SjD patients revealed differential responses to HCQ-LEF and RTX. Notably, healthy-like patients exhibited minimal treatment response, suggesting they may be unsuitable candidates for future therapeutic trials.
Copyright © 2025 European Alliance of Associations for Rheumatology (EULAR). Published by Elsevier B.V. All rights reserved.