Maxime Verhoeven

190 Chapter 10 cohort (i.e., the 1/3 of the MMC data not used for model development). External validation was performed in the ACURA cohort by applying the derived formula. Agreement of the different DAS-OST indices and DAS28 was determined in the internal- and external validation cohorts by using the Bland-Altman plot method, calculating the SD of the difference (i.e., measurement error), and by random two-way mixed effect intra-class correlation coefficient (ICC). 11 The area under the receiver operating characteristic curve (AU-ROC) was calculated to assess overall discrimination of DAS-OST for established definitions of remission, LDA and HDA, based on DAS28 and Boolean remission criteria (i.e., SJC≤1, TJC≤1, patient VAS≤10 and taking an CRP≤10mg/L, as estimated from the established ESR cut-off, see Supplementary Data S1). Furthermore, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of DAS-OST for these established definitions were determined using the same cut-offs for DAS-OST as used for DAS28 (i.e., remission= DAS-OST≤2.6, LDA= DAS-OST≤3.2, HDA= DAS-OST>5.1). Only for Boolean remission a cut-off was (pre-)defined in the development cohort using Youden’s index. All tests were two-sided, a p-value ≤0.05 was considered as statistically significant. No missing data was imputed and analyses were performed in SAS version 9.4. RESULTS Characteristics of the two patient cohorts Data of 3358 observations were extracted, without missing values, from the medical records of the MMC, including 1505 unique RA patients. A random sample of 2/3 of the data (i.e., 2238 observations) was used as development cohort and the remaining data (i.e., 1120 observations) was used as internal validation cohort. The external validation cohort included 168 unique RA patients. Due to missing values of OST-score and/or DAS28, data of 151 RA patients was used for the analyses. Patients’ demographic and clinical data are shown in Table 1. Overall, disease activity (i.e., DAS28, its components and OST-score) was statistically significantly higher in the ACURA cohort compared to the MMC cohorts, i.e., development and internal validation (p<0.01 and p<0.01, respectively, for all variables). Other statistical significant differences between the MMC cohorts and the ACURA cohort were shown for age, disease duration, and seropositivity (p<0.01, p<0.01 and p<0.01, respectively, for all variables).