Maarten Cozijnsen

45 Chapter 2 Development and validation of the MINI-index Discussion We developed the MINI index, a Mucosal Inflammation- Non-Invasive index that strongly correlates with SESCD and that can accurately assess mucosal inflammation. The MINI index was generated on a large prospective cohort of pediatric CD, and validated on three independent prospective cohorts. A cutoff of <8 best balanced sensitivity and specificity in reflecting. A cutoff of <6 had a higher PPV to reflect MH (86%) and ≥8 was most accurate to diagnose mucosal inflammation (PPV 90%). Although the index was significantly more accurate than FC (p=0.013), its clinical benefit over FC was modest for the entire dataset. However, one of the largest weaknesses of FC is the low discriminatory accuracy in the gray range of 100-599 μg/g, which may reflect severe inflammation in some patients or near mucosal healing in others. The MINI index significantly improved the utility of FC in this gray zone, adding to the correct classification of at least half of these patients with a number needed to screen of 9. Furthermore, the items added to FC are collected in routine clinical practice, and thus the use of the index should be relatively easy and intuitive. In our study, the PCDAI correlated moderately with SESCD (rho=0.59) and higher than previously reported in children (0.3–0.45) 9,11 . Of the individual items, the stooling and weight items were most significant, and abdominal pain and fatigue were not. The correlation between CRP and SESCD was 0.59, slightly higher than the range of previous reports (i.e. rho~0.12-0.54) 4,6,9,12–14,18 . In the multi-variable models, CRP had little contribution to the overall fit, due to strong collinearity with FC and ESR. The latter was similarly crowded out in the presence of the other biomarkers. Guided by strong judgmental input from our international advisory board, we incorporated both CRP and ESR with low weights. Indeed, a similar study to ours among adult CD patients 24 and a recent post-hoc analysis of the CALM randomized controlled trial of 244 adults with CD 25 both affirmed that adding CRP to FC increases the accuracy of detecting MH. Another recent retrospective study reported improved accuracy in diagnosing CD among 128 children with elevated FC levels by considering ESR, CRP and albumin as additional markers to FC. 26 We used a blended mathematical–judgmental clinimetric approach. An analytic account of the clinical phenomena that is observed, judged and decided by clinicians and patients themselves, is often missing from psychometric outcome measures 27 . Mathematical modeling alone is not a sine-qua-non for accuracy, since the psychometric approach has been criticized for lack of face validity and sensibility in developing scales 28 . To optimize the utility of scores in day-to-day clinical practice, a thoughtful combination of both mathematical