P103 Minimal Clinically Important Difference (MCID) for the Ulcerative Colitis Patient-Reported Outcomes Signs and Symptoms (UC-PRO/SS) Tool
Pulley, J.(1);Galloway, D.(1);Oh, Y.(2);Devine, J.(3);Petrek, H.(2);Khavarian, A.(2);Matsui, A.(3);
(1)Roche Products Ltd, Biostatistics, Welwyn Garden City, United Kingdom;(2)Genentech, Product Development, South San Francisco, United States;(3)Genentech, Patient-Centered Outcomes Research, South San Francisco, United States
Background
The UC-PRO/SS tool developed by the IBD PRO Consortium is undergoing FDA Drug Development Tool qualification (COA DDT# 000040). Derivation of minimal clinically important difference (MCID) thresholds for improvement in bowel and functional signs and symptoms domain scores are presented here.
Methods
MCID estimates for bowel and functional domains were derived using anchor- and distribution-based approaches applied to data from three studies of Etrolizumab. Initial MCIDs were developed using Mayo Clinic Score (MCS) and Inflammatory Bowel Disease Questionnaire (IBDQ) response thresholds as anchors; Cronbach’s alpha SEM and Cohen’s moderate effect size statistics were calculated. MCIDs were then cross-validated internally and externally, i.e. within and across target populations, using clinical (modified and partial MCS) and patient-centered (IBDQ “general wellness” item 10 and EQ5D VAS) measures, to support generalizability. CDF plots with pooled data provide further empirical support.
Results
Study populations had mean ages of 39.2-40.5 years; were 54.5-58% male; 77.5-93.9% white; and had mean disease duration of 8.75-8.98 years. Six points was selected as the bowel domain threshold as change of this magnitude aligns with the anchor settings (5.89, 6.93, and 7.15 for IBDQ and 5.63, 6.12, and 7.63 for MCS) and exceeds measurement error as shown by distributional methods (Cronbach’s alpha SEM 2.071, 2.160, and 2.167, Cohen’s moderate effect size range 2.086, 2.135, and 2.191). A two-point threshold was similarly selected for the functional domain based on anchor settings (1.77, 1.98, and 2.53 for IBDQ and 1.56, 1.68, and 2.01 for MCS) and distributional methods (Cronbach’s alpha SEM 1.336,1.140, and 1.417, Cohen’s moderate effect size range 1.204, 1.270, and 1.309). External, cross-validation showed numerically robust differences (no overlapping CI’s) between the group achieving the MCID versus not across clinical and PRO variables for both domains. Magnitudes of change on IBDQ item 10 and EQ5D VAS variables scores exceeded the ½ SD for interpreting moderate levels of PRO score change. For the bowel domain, separation of CDF curves between responder groups showed cumulative percentages of approximately 60% and 70% based on IBDQ and MCS anchors, respectively, at the thresholds indicating change of a six-point magnitude. For the functional domain, separation between CDF curves at the 2-point threshold occurred at points indicating cumulative proportions of response greater than 50% using both anchors.
Conclusion
MCIDs of 6 and 2 points for the bowel and functional domains of the UC-PRO/SS are robustly defined and cross-validated, allowing interpretation of clinical improvement in symptoms among moderate to severe adult UC patients.