THU0238\u2005CLINICAL EFFICACY OF LEFLUNOMIDE/HYDROXYCHLOROQUINE COMBINATION THERAPY IN PRIMARY SJOGREN’S SYNDROME IS PREDICTED BY SERUM PROTEOME BIOMARKERS – RESULTS FROM REPURPSS-I

Abstract

Background: Despite major efforts to treat primary Sjogren’s syndrome (pSS) advances in pSS treatment remain disappointing. Treatment strategies effective in other immunological conditions lack an effect in pSS. Hence, there is a huge unmet need in finding an effective treatment, let alone to have the ability to predict who responds to what treatment. Based on the unique immunosuppressive properties of leflunomide (LEF) and hydroxychloroquine (HCQ), we recently conducted a randomized, double–blind, placebo-controlled proof of concept study. Objectives: To evaluate the efficacy, safety and tolerability of LEF/HCQ therapy in patients with pSS and to predict treatment response by molecular profiling using baseline serum proteomics. Methods: Clinically active (European SS disease activity index ESSDAI score ≥5) pSS patients (n=29) were randomized to receive LEF 20 mg daily and HCQ 200 mg twice daily or placebo/placebo (2 verum:1 placebo) for 24 weeks. Clinical and safety outcomes were assessed at baseline, 8, 16 and 24 weeks. Clinical response was defined by a decrease in ESSDAI of ≥3 points at 24 weeks. In addition, at baseline 386 proteins involved in inflammation, immune response, metabolism and cardio-metabolism were measured in serum of all patients using the Olink platform (for panels, see Olink website). Results: Overall, LEF/HCQ was safe and well-tolerated and significantly reduced ESSDAI scores, the primary endpoint (p<0.0001). Furthermore, LEF/HCQ treatment was associated with improvement of oral dryness, ESSPRI, Physician’s and Patient’s Global Assessment, serum IgG, IgM rheumatoid factor, C3 and C4 levels (all at least p<0.05), which was not observed in the placebo group. Clinical response was observed in 11 out of 21 patients receiving treatment, providing a unique opportunity to examine biomarkers to predict response. However, except for C3 levels, clinical markers at baseline were not significantly different between responders and non-responders, underscoring the challenges in prediction of therapy response. Olink proteomic analysis revealed 43 significantly differentially expressed proteins between responders and non-responders. Unsupervised hierarchical clustering using the most differentially-expressed analytes (p-value based) demonstrated distinct serum proteomes in responders vs. non-responders. Using multidimensional scaling, based on all measured analytes, a high degree of dissimilarity of responders and non-responders was observed. Next a random forest machine-learning model was established to identify the proteins that best predict response to treatment using the baseline serum proteome. The results were validated using 500 leave-7-out iterations. The highest mean accuracy (84%) was achieved using a set of 16 differentially expressed analytes, indicating a 96% chance to robustly predict responders and a 74% chance to correctly identify non-responders. Conclusion: We demonstrated a clinical response in pSS patients treated with the combination of leflunomide and hydroxychloroquine1. Strikingly, a set of 16 circulating proteins predicts response to therapy with clinically meaningful accuracy. Given the exciting but preliminary nature of these observations a follow-up RCT aimed at replication of these results is warranted. Reference: [1] Clinical Efficacy of Leflunomide/Hydroxychloroquine Combination Therapy in Patients with Primary Sjogren’s Syndrome: Results of a Placebo-Controlled Double-Blind Randomized Clinical Trial. Radstake T, et al. Abstract L10 ACR 2018. Disclosure of Interests: None declared