P233. How accurate are capsule endoscopy scoring systems in Crohn's disease?
G. Holleran1,2, B. Hall1,2, M. Hussey2, O. Thornton2, M. Dobson2, D. McNamara1,2, 1Trinity College Dublin, Clinical Medicine, Dublin, Ireland, 2AMNCH, Department of Gastroenterology, Dublin, Ireland
Capsule endoscopy (CE) is a useful diagnostic modality in small bowel Crohn's disease (CD) and can be more sensitive than other imaging techniques. However, reports are observer dependent and findings can be non-specific. 2 scoring systems, the Lewis score (LS), and the CE CD Activity Index (CECDAI) have been developed to grade severity of CD. The Harvey Bradshaw Index (HBI) is a subjective scoring system, and the C-reactive protein (CRP) is a biochemical marker of inflammation.
The aims were to determine the inter-observer agreement on the presence and severity of small bowel CD using the LS and CECDAI, and to assess the correlation of these scoring systems with the HBI, and CRP levels.
A review of 28 CE procedures in patients with suspected CD was undertaken. Patients completed a HBI prior to CE, and CRP levels were recorded within 2 mths. 2 observers, reported the CEs and assigned a LS and CECDAI score to each. Reports were compared to determine inter-observer agreement, using previously defined thresholds for mild and severe disease, of 135 and 790 for LS, and 3.5 and 5.8 for CECDAI. Correlation of scoring systems with the HBI and CRP level was also assessed. The Pearson correlation coefficient was used to compare findings. Linear regression analysis was performed to validate new thresholds for CECDAI scores.
CD was present in 68% (n = 19), with inter-observer agreement of 95%. A LS was recorded in 28 patients, with a mean of 782 (0–5362) and CECDAIs were recorded in 26, with a mean of 8.42 (0–34). The inter-observer agreement for both the LS and CECDAI was strongly positive, 0.921 (LS1 vs LS2) and 0.854 (CECDAI1 vs CECDAI2) respectively. Overall there was a positive correlation between the LS and CECDAI, 0.853. There was a discrepancy between LS and CECDAI for severe CD (0.263), employing current CECDAI thresholds (65% and 11% severe CD on CECDAI and LS respectively). The LS and CECDAI had no correlation with the HBI (−0.065 and −0.301) or CRP (−0.152 and −0.100) respectively. Linear regression analysis of our data showed that predefined thresholds of 135 and 790 for LS corresponded to CECDAI levels of 4 and 23.5.
There is good inter-observer agreement on the presence of CD using both scoring systems. There was no correlation with either the HBI or CRP, suggesting neither are reliable predictors of small bowel CD. The previous CECDAI severity threshold of 5.8 appears too low. We have identified a higher threshold of 23.5 for severe inflammation which may be more useful in guiding clinical management.