Maarten Cozijnsen

40 Results Derivation of the MINI index We first constructed a regression model where the SESCD served as the dependent variable and the explanatory variables included the total PCDAI score, CRP and FC. The PCDAI and FC were associated with SESCD (both p<0.001), while CRP was not (p=0.32) (R 2 =0.45). Next, we aimed to identify key items of the PCDAI reflecting mucosal inflammation in a model with and without CRP and FC (Supplementary table 1). We judgmentally excluded the items ‘well-being’ which is poorly defined and lacks reliability, ‘abdominal examination’ which lacks reliability and ‘height velocity’ which is an important determinant of mucosal inflammation but has poor responsiveness over time and not relevant to adolescents who completed their growth period. These three items were previously proven to be redundant in a multivariable regression analysis of the PCDAI 23 . The stooling item of PCDAI and FC were strongly associated with SESCD in all models (Supplementary table 1). Hypoalbuminemia (<3 g/dL) was associated with low SESCD in the multivariate analyses, counterintuitive to the expected direction and thus was excluded – in univariate analyses the association was opposite, as expected. We constructed further models where we substituted the laboratory PCDAI items (i.e. hematocrit, albumin and ESR) with their absolute values (rather than the categorized values), but this did not improve the model fit (R 2= 0.476). We also analyzed models with additional laboratory measures associated with inflammation (i.e. platelets, white blood cells) but again without an added value (R 2= 0.469). ESR and the weight items of the PCDAI were significant in some of the models, especially without FC or CRP (Supplementary table 1). Indeed, FC, CRP and ESR, were in collinearity with each other and FC crowded out the association of ESR and CRP. Nonetheless, the advisory board reached consensus to retain CRP in order to improve face and content validity, considering the extensive literature and vast clinical experience demonstrating the importance of CRP in reflecting mucosal inflammation in CD, and since at times only CRP and not ESR is available or vice versa. To construct an intuitive tool, we grouped continuous laboratory measures (FC, CRP and ESR) into categories by plotting the variable against categories of endoscopic disease severity (Figure 1). The advisory board opted to exclude the weight item since it is a longitudinal measure, has limited responsiveness to change, it can be artificially affected by other factors like steroids and its contribution to the overall model fit was negligible (Supplementary table 2).