Background/Purpose: Diagnosis of RA is difficult in American Indian (AI) patients who often have atypical autoantibodies (e.g. ANA) with erosive arthritis. Soluble mediators may serve as alternative markers for earlier identification of RA from other conditions affecting AI patients. This study characterized soluble mediator profiles of AI RA patients to identify pathogenic mechanisms of disease and develop early diagnostic algorithms for use by healthcare providers to direct patients toward individualized therapies.

Methods: The serum levels of RF, anti-CCP, and 26 soluble mediators from AI patients with RA (n=59), arthralgias (n=28), polyarthritis/osteoarthritis (n=31), or AI healthy controls (n=116) were measured using ELISA or multiplex assays. Comparisons were made using Kruskal-Wallis testing with Dunn’s multiple comparison. To identify factors that distinguished AI RA patients from healthy controls or patients with other forms of arthritis, random forest modeling was performed using autoantibodies and soluble mediators that differed significantly between the groups. One model used autoantibodies, a second used soluble mediators and a third used both to classify AI RA patients from other AI arthritis groups. Each model was developed using 300 trees, √x variables at each split (where x=number of incorporated variables), and a terminal node size of 1. Top predictive classifiers were identified using the prediction step of VSURF function in R 3.3.0. Sensitivity, specificity, positive and negative predictive values were calculated based on 50 random forests using the top predictive identifiers for each model.

Results: When comparing AI RA patients and controls, the soluble mediator model outperformed the autoantibody model with an accuracy of 93.2% vs 88.1%. The combined model outperformed the autoantibody model with an accuracy of 92.7%. IL-8, TGF-beta, SCF, MCP-3, and IL-1RA were identified as the top predictive classifiers when using the soluble mediator model. IL-8, RF(IgM), anti-CCP, and TGF-beta were identified as the top predictive classifiers when using the combined model. When comparing RA patients to AI patients from other arthritis groups the combined model outperformed the autoantibody model (82.5% vs 78.3%) with anti-CCP, RF (IgM), SCF, resistin, and TNFRII as top predictive classifiers. The soluble mediator model did not differentiate patients with RA from the other arthritis groups, suggesting some AI patients with polyarthritis have similar immune dysfunction without meeting classification criteria. IL-8 levels best distinguished patients and controls and SCF levels best distinguished RA patients from other diseases.

Conclusion: Adding soluble mediators to traditional autoantibodies improves the accuracy of diagnostic modeling in AI RA patients when compared to AI controls or AI patients with other forms of arthritis. These data support the hypothesis that soluble mediator profiles of AI RA patients can be used to identify mechanisms of disease and to develop early diagnostic algorithms for use by healthcare providers to direct patients toward individualized therapies. The study was funded by Native American Research Centers for Health.

« Back to 2016 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/diagnostic-modeling-of-rheumatoid-arthritis-in-oklahoma-tribal-members-using-soluble-mediators/