Abstract
ABSTRACTObjectivesTo aid research on etiology and treatment of the heterogeneous rheumatoid arthritis (RA) population, we aimed to identify phenotypically distinct RA subsets using baseline clinical data.MethodWe collected numerical- (hematology work-up & age) and categorical variables (serology, joint location & sex) from the Electronic Health records (EHR) repository of the Leiden University Medical Center, comprising 1,387 unique first visits to the outpatient clinic. We used deep learning and clustering to identify phenotypically distinct RA subsets. To ensure the robustness of our findings, we tested a)cluster stability (1000 fold) b)physician confounding, c)association with remission and methotrexate failure, d)generalizability in second data set (the Leiden Early Arthritis clinic; n=769), and e) interaction of clusters with RA-susceptibility genesResultsWe identified four subsets of RA patients that were delineated on the following characteristics: Cluster-1) arthritis in feet, Cluster-2) seropositive oligo-articular disease, Cluster-3) seronegative hand arthritis, and Cluster-4) polyarthritis. We found a high stability (mean 78%-91%), no physician confounding, significant difference in methotrexate failure(P-value=6.1e-4) and occurrence of remission(P-value=7.4e-3), and good generalizability. The hand-cluster (3) had the most favorable outcomes (HRremission=1.65 (95%CI:1.20-2.29), HRmethotrexate=0.48 (95%CI:0.35-0.77)), particularly for the ACPA-positive patients in this cluster, while in the other clusters the ACPA-negative patients did best. In total, we identified 9 SNP interactions: Cluster-1)IL2RA, SPRED2, RUNX1, Cluster-2)COG6, GRHL2, ETS1,Cluster-3)AFF3and Cluster-4)CCL21, CTLA4.ConclusionsWe discovered four phenotypically distinct subgroups of RA at baseline that associate with clinical outcomes. Our study provides evidence for the presence of separate hand and foot subgroups.Key messagesWhat is already known about the subject?-Rheumatoid arthritis is a heterogeneous disease and clinicians have not identified the disease discerning patterns in clinical practice.-Data-driven unsupervised techniques are able to identify hidden structures in big data.What does this study add?-We identified four novel RA clusters at baseline: feet involvement, oligo-articular disease, hand involvement and polyarthritis. In particular, the hand and feet clusters show a marked difference in outcomes.-The association of both ACPA and age of onset with long-term outcomes differed between the clusters, suggesting that the association between these markers might be different depending on the RA subset.How might this impact on clinical practice or future developments?-Our clusters are a next step towards elucidating etiology and treatment of RA.-RA-patients with hand or feet involvement might need different treatment.
Publisher
Cold Spring Harbor Laboratory