Characterizing and Operationalizing the Third Criterion of Difficult-to-Treat Rheumatoid Arthritis: Patient and Physician Perspectives from the GISEA Study

Authors:

Abstract
Objectives: To characterise the determinants of the third EULAR criterion of difficult-to-treat rheumatoid arthritis (D2T RA)—the perception of problematic disease management—from physician and patient perspectives, evaluate their agreement, and compare the performance of multivariable prediction models with VAS-based thresholds for its identification.Methods: In this multicentre cross-sectional study of the Italian GISEA registry, 457 patients with RA who had failed ≥2 b/tsDMARD classes were included. Problematic disease management was assessed from physician, patient, and overall (physician and/or patient) perspectives. Agreement was evaluated using Cohen's kappa. Determinants were identified through univariable analyses followed by LASSO-based multivariable logistic regression. Model discrimination (AUC) and calibration were compared with ROC-derived VAS thresholds.Results: Problematic disease management was identified in 57.1%, 67.8%, and 76.4% of patients according to physician, patient, and overall definitions, respectively, with moderate physician–patient agreement (κ=0.41). Treatment failure burden, disease activity, and patient-reported outcomes were common determinants across perspectives. Physician-rated problematic disease management was more strongly associated with inflammatory burden, whereas patient-rated problematic disease management was more closely associated with mental health disorders and fibromyalgia. Multivariable models showed good discrimination (AUC 0.80–0.83) and consistently outperformed VAS-based thresholds (AUC 0.71–0.76).Conclusions: The third EULAR D2T RA criterion represents a multidimensional construct with distinct physician- and patient-driven components, supporting the coexistence of inflammatory and non-inflammatory D2T RA phenotypes. Multivariable models outperform single VAS thresholds for identifying this criterion in retrospective studies and provide a standardized, data-driven approach for its operationalization.