Baseline Predictors of Difficult-To-Treat Rheumatoid Arthritis Across Adapted Definitions: Insights from the OBRI-RA Registry

Objectives In the first phase of this study, 4 adapted definitions of difficult-to-treat rheumatoid arthritis (D2T-RA) were evaluated within the Ontario Best Practices Research Initiative (OBRI), differing by criteria for treatment failure, active or symptomatic disease, and clinician perception (Figure). These definitions showed heterogeneity in prevalence, with broader definitions capturing patients with higher disease activity and worse function. Building on those findings, this analysis examined baseline clinical and demographic characteristics associated with D2T-RA and assessed whether similar predictors emerged across the 4 definitions. Figure. Comparative framework and predictive performance of four adapted definitions of difficult-to-treat rheumatoid arthritis (D2T-RA) . The left panel outlines the criteria used in each definition, while the right panel displays receiver operating characteristic (ROC) curves from multivariable logistic regression models predicting D2T-RA status. Model discrimination was fair across definitions (AUC 0.66-0.71). Abbreviations: b/tsDMARD, biologic targeted synthetic DMARD; CDAI, Clinical Disease Activity Index; DAS28. Disease Activity Score-28; ESR, erythrocyte sedimentation rate; CRP, C-reactive protein; HAQ, Health Assessment Questionnaire; MD, physician; PT, patient. Methods Data were derived from the Ontario Best Practices Research Initiative (OBRI), a longitudinal, real-world registry of patients with rheumatoid arthritis (RA). Multivariable logistic regression analyses were conducted among participants with established RA who initiated their first advanced therapy (biologic or targeted synthetic DMARD) after enrollment. Candidate predictors were identified a priori based on existing literature encompassing demographic, clinical, serologic, and treatment-related variables. Separate regression models were developed for each of the 4 adapted definitions of D2T-RA. Results A total of 507 patients were included in the complete case analysis. The mean (SD) age was 56.9 (12.2) years, the mean disease duration was 7.6 (9.1) years, and 80.3% were female. The prevalence of difficult-to-treat RA (D2T-RA) ranged from 5.7% (Definition 4) to 14.4% (Definition 2). In multivariable models, female sex independently predicted D2T-RA under Definitions 2 (OR 2.46, 95% CI 1.10-5.49) and 3 (OR 3.42, 95% CI 1.11-10.5), and higher HAQ-pain scores were consistently associated with D2T-RA across inclusive definitions (OR 1.60-2.00, p < 0.05). Radiographic erosions were significant only under Definition 3 (OR 2.03, 95% CI 1.02-4.03). No predictors reached significance for the restrictive Definition 4 aside from borderline methotrexate use (p = 0.053). Disease duration, seropositivity, smoking, obesity, and baseline disease activity were not independently associated with D2T-RA. Model performance was fair across definitions (AUC 0.68-0.74) (Figure). Conclusion Female sex and higher pain-related functional impairment consistently predicted D2T-RA across the more inclusive definitions, suggesting that sex/gender considerations and pain perception are important factors to consider in treatment refractoriness. In contrast, restrictive definitions based solely on treatment count demonstrated limited discriminative value. These findings suggest that adapted D2T-RA definitions capture different facets of the disease spectrum, and their use should be guided by the specific clinical or research context—whether the goal is to identify persistent refractory inflammatory disease or to characterize the broader, multidimensional burden encountered in real-world practice.