For comparison purposes, these analyses were repeated for the duplicate pairs in which neither video was presented with clinical details. Analysis of variance with terms for investigator and pair type were used to compare absolute differences. Reliability ratios for the UCEIS and overall severity, and intraobserver agreement at the descriptor level, were calculated as described previously. Bowker’s test for symmetry 11 tested for presentation order effects (ie, impact of viewing videos with clinical details before or after the blinded

version) on responses to descriptors. Two additional methods for calculating the Natural Product Library UCEIS were examined: 1. A normalized sum was used, in which descriptors were combined so as to contribute equally, as one-half “vascular pattern” plus one-third “bleeding” and one-third “erosions and ulcers”; the range of normalized UCEIS scores was then 0 to 3, with 17 possible scores. The design of this study did not permit a direct evaluation of the UCEIS in terms of sensitivity to change between videos at the individual patient level. Nevertheless, the data can be analyzed to assess the power of differentiation across patients (videos). All possible pairings of the 57 videos were formed, for a total of 1596 distinct pairings. Each video was evaluated by between 6 and 15 investigators in the main analysis

set. For each pair, mean differences in the UCEIS and overall endoscopic severity on the VAS, and 2-sample t tests for differences between videos for evaluation of overall severity on selleck screening library the VAS and the UCEIS were calculated. Proportions of significantly different scores (confirmed Dichloromethane dehalogenase by t tests) were studied globally and as a function of the difference in endoscopic severity on the VAS. To compare the UCEIS with established clinical measures for UC, Spearman

rank correlation tests were performed between the UCEIS and full Mayo score, partial Mayo score (excluding endoscopic evaluation),12 stool frequency/rectal bleeding, and patient functional assessment. Statistical analyses were performed using Statistical Analysis System (SAS, Cary, NC) software version 9.2. Twenty-nine investigators from 14 countries were screened for participation in the study. Eleven of the 29 succeeded on first qualification and 14 on their second attempt. One investigator failed both times, and 3 were withdrawn due to noncompliance with procedures, resulting in a total of 25 investigators (11 from North America, 9 from Central Europe, and 5 from Western Europe; see Acknowledgments). In total, 698 of the planned 700 evaluations were performed. Each video was assessed by 6 to 15 investigators. The response rate was 100% for assessment of overall severity on the VAS and for all descriptors of these 698 evaluations. The analyses that follow exclude 50 videos from the second evaluation of repeat pairs and the 100 evaluations used for clinical details/no clinical details evaluation, unless stated otherwise.