P156 Interobserver variation of visual assessment vs. measuring ulcer size in Crohn’s disease using the RAPID9 Software in PillCam COLON2

S. Quah1, A. Gorelik2, E. Tsoi3, E. Chow4, S. Sivanesan4, G. Brown5, S. Anwar6, A. Taylor3, F. Macrae1

1The Royal Melbourne Hospital, Department of Colorectal Medicine and Genetics, Melbourne, Australia, 2Australian Catholic University, Department of Biostatistics, Melbourne, Australia, 3St Vincent’s Hospital, Department of Gastroenterology, Melbourne, Australia, 4The Royal Melbourne Hospital, Department of Gastroenterology, Melbourne, Australia, 5Epworth Healthcare, Department of Gastroenterology, Melbourne, Australia, 6Western Health, Department of Gastroenterology, Melbourne, Australia

Background

Ulcer size is important when assessing mucosal healing in Crohn’s disease (CD) and commonly assessed with visual estimation. There is a new built-in measurement tool in the RAPID9 reader software used in colon capsule endoscopy (CCE). We aimed to assess the agreement between visually estimating ulcer size vs. using the estimation tool and the interobserver agreement for both methods independently using PillCam Colon studies in patients with Crohn’s disease.

Methods

7 experts of PillCam readers were recruited to assess CCE in this study. 7 CCE recordings were thumbnailed for the presence of ulcers. A total of 43 ulcers were consecutively selected for this study. The readers were provided with an instruction manual which specified that they were to first visually estimate the ulcer and then subsequently measure the ulcer with the size estimation tool in RAPID9 software. Two measurements were taken, A – longest distance of the ulcer end to end, and B – longest distance that is perpendicular to A. Results were analysed using Kendall’s Coefficient of Concordance (Kendall’s W).

Results

When comparing visual assessment of ulcer size vs. using the estimation tool, both measurements A and B demonstrated statistically significant agreement (p of A = 0.000, p of B = 0.000; W of A=0.90, W of B=0.82). For measurement A, our results demonstrated higher interobserver agreement when assessing ulcer size using the estimation tool (W=0.84) as compared with visually assessing ulcer size (W=0.80). This contrasts with measurement B where there was higher interobserver agreement when visually assessing ulcer size (W=0.75) as compared with using the measurement tool (W=0.73). Nevertheless, there was no statistically significant variability amongst measurement A nor B (p = 0.000) in visual assessment and measurement tool.

Conclusion

The conventional visual estimation of ulcer size in CD using CCE was just as accurate as the measurement tool. Furthermore, consistency between readers was demonstrated with the high level of interobserver agreement for the measurement tool and visual estimation. This novel finding is reassuring given the decade of capsule studies done with older software, where a measurement tool was not available. Further study should be conducted to assess and consolidate the validity of the estimation tool in assessing ulceration and mucosal healing in CD with a larger number and variety of lesions studied.