Maxime Verhoeven

173 Utility of the HandScan in monitoring disease activity in RA period of 6 months. Disease activity was measured at each visit, first with the HandScan and shortly after with DAS28. The following baseline data was collected: age, gender, BMI, smoking status, alcohol use, rheumatoid factor (RF) status, and anti-cyclic citrullinated peptide antibody (anti-CCP) status. DAS28 (and its components) and OST-scores were collected at every visit, whereas the functional ability and quality of life were assessed at baseline and every 3 months thereafter, using respectively the Health Assessment Questionnaire (HAQ) and EuroQol five dimensions questionnaire (EQ5D-5L). Statistical analysis Baseline characteristics and treatment response were described for all patients, and stratified by RA stage (early or established; csDMARD therapy or TNFi therapy). Data of early and established RA were combined to obtain a more adequate sample size. The effect of RA stage was taken into account in all model based analyses (e.g., see explanation of the mixed effect models and the binary logistic regression models further in this section). 10 Pearson or Spearman correlation coefficients, depending on distribution of the data, of DAS28, SJC, TJC, HAQ and EQ5D-5L, with OST-score were calculated for all patients, both concurrently, as well as with time-lags to explore the crude associations of OST-score over time with other frequently used outcome measures. To determine whether changes in OST-scores are related to changes in DAS28 in individual patients, an autoregressive mixed effects model with a random intercept at patient level was used. 11 The outcome variable was DAS28; independent variables were OST-score, visit month, RA stage and DAS28 at previous visit (i.e., autoregressor). The same analyses were performed for SJC (square root transformed), TJC (square root transformed), HAQ and EQ5D-5L as respective outcome variables. It was also explored whether RA stage (early vs. established; csDMARD vs. TNFi) modified the association between OST-score and the outcomes by adding the interaction term, e.g., OST-score*RA stage. Binary logistic regression was used to test the predictive value of short-term (i.e., 1 month) OST-score change together with baseline OST-score for the outcome EULAR good response (y/n), and ACR50 response (y/n) at 3 or 6 months. Baseline DAS28, and short-term (i.e., 1 month) DAS28 change were also evaluated in a similar separate analysis for comparison with the former model. This analysis was also adjusted for RA stage (early vs. established) as the initiated therapy differ (csDMARD vs. TNFi), and it was tested if RA stage modified the association between changes in OST-score and outcome (i.e., adding the interaction term OST-score*RA stage). Because of the exploratory nature of this study, no power calculation was performed. The statistical analyses were performed in SAS version 9.4. All tested were two-sided and a p-value ≤0.05 was considered statistically significant. Seven of 64 patients had missing information on DAS28 and/or OST-score, but only at the 6 months visit. As mixed model analysis, using all longitudinally available data of the patients, is robust 9