Interobserver variation in the interpretation of magnetic resonance enterography in Crohn's disease

Objectives: To evaluate interobserver variability for diagnosis of disease presence and extent of small bowel and colonic Crohn's disease using MR enterography (MRE) Methods: Data from the first 73 consecutive patients (mean age 32, 33F, 28 new diagnosis, 45 suspected relapse) recruited to a multicentre, prospective diagnostic accuracy trial evaluating MRE for small bowel Crohn's disease were each read independently by three (from a pool of 20) radiologists. Radiologists documented presence and segmental location of small bowel Crohn's disease and recorded morphological mural/extramural parameters for involved segments. Per patient percentage agreement for disease presence and extent were calculated against an outcome based construct reference standard (averaged between pairs of readers). Prevalence-adjusted bias-adjusted Kappa (PABAK) was calculated. Results: Agreement for small bowel disease presence for new diagnosis/relapsed patients was 68%(lc = 0.36)/ 78% (lc = 0.56) and 43%(lc = 0.14)/ 53% for disease extent (lc = 0.07), respectively. For disease presence, all three radiologists agreed correctly with the reference standard in 41/59 (69%) of patients with small bowel involvement, and in 8/14 (57%) cases of without small bowel disease. Agreement was highest for multisegment disease, greater than 5 cm in length, with mural thickness>6 mm, and increased mural T2 signal. Agreement for colonic disease presence was 61% (lc = 0.21 fair agreement) for new diagnosis/ 60% (lc = 0.20, slight agreement) for relapsed patients. Conclusion: There is a reasonable agreement between radiologists for small bowel disease presence using MRE for newly diagnosed Crohn's disease, and patients with suspected relapse, respectively. Agreement is lower for disease extent. Advances in knowledge: There is reasonable agreement between radiologists for small bowel disease presence using MRE for newly diagnosed (68%) Crohn's disease, and patients with suspected relapse (78%). Agreement is lower for disease extent (43% new diagnosis and 53% suspected relapse).