Background: Advances in therapeutics and treatment strategies for Rheumatoid Arthritis (RA) have improved clinical outcomes. Although these advances also impact the well-being as shown in many patient-reported outcomes, still a sizeable number of patients in clinical remission report a reduced well-being.

Objectives: To explore factors that contribute to well-being in patients with early RA.

Methods: Patients from the 2-year pragmatic treat-to-target Care in Early Rheumatoid Arthritis (CareRA) trial were included. Patients were treated intensively, with a combination of csDMARDs and glucocorticoid remission induction schemes, except one group treated with MTX monotherapy.

Eight different validated questionnaires including the Arthritis Self-Efficacy Scale (ASES), the multidimensional Fatigue Inventory (MFI), the Pittsburgh Sleep Quality Index (PSQI) the Revised Illness Perception Questionnaire (IPQ-R), the Utrecht Coping List (UCL), the Short Form 36 (SF-36), RA Quality of Life questionnaire (RA-QOL) and the Social Support List (SSL) were taken. Questionnaires were obtained at baseline, at week 16, 52 and 104 except for the IPQ and UCL, which were only taken at baseline and week 16.

Three patients` groups were created including all patients, patients in remission (DAS28crp < 2.6) and not in remission. Regression models were constructed to define well-being at week 16, 52 and 104. The Patient Global Assessment (PGA) on a Visual Analogue Scale 0-100 (VAS) was chosen as a proxy for well-being (score 0-100). As predictors, all subscales of the 8 validated questionnaires, summing to 84 variables, with and without the VAS for Pain (VAS-Pain) were used in 18 models (3 patient groups, 3 time points, with/without VAS-Pain) in total. Data reduction used forward, backward and stepwise selection based on the Aikake information criteria. Data was checked for influential observations by Cook’s distance and for multicollinearity by variance inflation factors (threshold = 5). Influential observations were removed one observation every time. Highly correlated variables were deleted by backward selection (α=5%). Missing data was handled by multiple imputation using CART with 15 iterations.

Results: In total, 379 patients were included. Table 1 gives the number of variables and the associated R ² . In the 9 models defining well-being without VAS-Pain, 53 variables were used at least once. Most common variables were bodily pain (n=8) and social function (n=5) of the SF-36, and positive emotions (n=4) of the SSL. In the 9 models with VAS-Pain, 31 variables were used at least once. Most common variables were vitality (n=3) and social function (n=3) of the SF-36, and identity (n=3) of the IPQ-R. Model content was heterogenous regarding patient population and time.

Conclusion: Well-being is apparently difficult to define uniformly as many factors contribute to it. As already known, well-being, defined by PGA, and VAS-Pain are highly associated, even in patients in remission where pain levels should be theoretically lower. Other well-being definitions could lead to different results and should be further explored.

Disclosure of Interests: Diederik De Cock: None declared, Tianna Poffe: None declared, Geert Verbeke: None declared, Veerle Stouten: None declared, Sofia Pazmino: None declared, Delphine Bertrand: None declared, Johan Joly: None declared, Rene Westhovens Grant/research support from: Celltrion Inc, Galapagos, Gilead, Consultant of: Celltrion Inc, Galapagos, Gilead, Speakers bureau: Celltrion Inc, Galapagos, Gilead, Patrick Verschueren Grant/research support from: Pfizer unrestricted chair of early RA research, Speakers bureau: various companies