Development and Validation of a Magnetic Resonance Index for Assessing Fistulas in Patients With Crohn's Disease

Abstract

Background & aims: There is no validated magnetic resonance imaging (MRI) index for assessment of perianal fistulas in patients with Crohn's disease (CD). We developed and internally validated a new instrument.

Methods: We used paired baseline and week-24 MRI scans from 160 participants in a randomized placebo-controlled trial of stem cell therapy for patients with perianal fistulizing CD. Four radiologists scored disease activity using index items identified during previous studies and exploratory items. Reliability was assessed using intraclass correlation coefficients. We developed an index using backward elimination linear regression analysis, in which potential independent variables were items having intraclass correlation coefficients of at least 0.4 and the dependent variable was perianal fistulizing disease activity, measured on a 100-mm visual analogue scale. The final model was internally validated using the .632 bootstrap method to correct model optimism and quantify calibration accuracy. We evaluated responsiveness of the index by assessing longitudinal validity and estimating standardized effect sizes.

Results: We developed the magnetic resonance novel index for fistula imaging in CD (MAGNIFI-CD) using 6 items. The optimism-corrected R² of the model was 0.71, which was comparable to R² for the original sample (0.74). The calibration slope for the model was 0.98. Compared with the original and modified versions of the Van Assche Index, the MAGNIFI-CD had improved operating characteristics. Estimates of intraclass correlation coefficients for MAGNIFI-CD, the modified Van Assche Index, and Van Assche Index were 0.85 (95% confidence interval [CI], 0.77-0.90), 0.81 (95% CI, 0.74-0.86), and 0.81 (95% CI, 0.71-0.86) for intra-rater reliability, and 0.74 (95% CI, 0.63-0.80), 0.67 (95% CI, 0.55-0.75) and 0.68 (95% CI, 0.56-0.77) for inter-rater reliability. Corresponding standardized effect size estimates were 1.02 (95% CI, 0.65-1.39), 0.84 (95% CI, 0.48-1.21), and 0.68 (95% CI, 0.33-1.03).

Conclusions: We developed an index called the MAGNIFI-CD, which is based on 6 items. It assesses MRI data and determines perianal fistulizing CD activity with improved operating characteristics compared to previous indices. This index may be used as an outcome measure in clinical trials comparing treatment effects in patients with perianal fistulizing CD. Although the performance of the MAGNIFI-CD indicates its stability and reasonable external validity, external validation is needed.

Keywords: ICCs; VAS; healing; remission.