The Fidelity of Rheumatoid Arthritis Multivariate Diagnostic Biomarkers Using Discriminant Analysis and Binary Logistic Regression.

Abstract

Rheumatoid arthritis (RA) is an autoimmune inflammatory disease that causes multi-articular synovitis. The illness is characterized by worsening inflammatory synovitis, which causes joint swelling and pain. Synovitis erodes articular cartilage and marginal bone, resulting in joint deterioration. This bone injury is expected to be permanent. Cytokines play a prominent role in the etiology of RA and could be useful as early diagnostic biomarkers. This research was carried out at Riyadh's King Khalid University Hospital (KKUH). Patients were enrolled from the Rheumatology unit. Seventy-eight RA patients were recruited (67 (85.9%) females and 11 (14.1%) males). Patients were selected for participation by convenience sampling. Demographic data were collected, and disease activity measurements at 28 joints were recorded using the disease activity score (DAS-28). Age- and sex-matched controls from the general population were included in the study. A panel of 27 cytokines, chemokines, and growth factors was determined in patient and control sera. Binary logistic regression (BLR) and discriminant analysis (DA) were used to analyze the data. We show that multiple cytokine biomarker profiles successfully distinguished RA patients from healthy controls. IL-17, IL-4, and RANTES were among the most predictive variables and were the only biomarkers incorporated into both BLR and DA predictive models for pooled participants (men and women). In the women-only models, the significant cytokines incorporated in the model were IL-4, IL-17, MIP-1b, and RANTES for the BLR model and IL-4, IL-1Ra, GM-CSF, IL-17, and eotaxin for the DA model. The BLR and DA men-only models contained one cytokine each, eotaxin for BLR and platelet-derived growth factor-bb (PDGF-BB) for DA. We show that BLR has a higher fidelity in identifying RA patients than DA. We also found that the use of gender-specific models marginally improves detection fidelity, indicating a possible benefit in clinical diagnosis. More research is needed to determine whether this conclusion will hold true in various and larger patient populations.