MULTICENTER STUDY OF INTEROBSERVER AGREEMENT IN DESCRIBING CAPSULE ENDOSCOPY FINDINGS

Background and Aim: Few studies have specifically addressed the interobserver agreement in describing the lesions identified during CE. The aim of our study is to evaluate interobserver agreement in identification and interpretation CE findings. Materials and Methods: The study was a multicentre study. A group of 8 exspert endoscopists (more than 50 CE videos read) participated in the study. We choose to visualise a short segment (20 seconds) of CE. An external author, non participant in the reading of the tapes, prepared the cases to view that were recorded on DVD. This author was the only one to know the final diagnosis of such patients and he is considered as the “gold standard”. The videos contained different lesions in 80% of the cases and the remaining 20% were negative. In order to obtain uniformity in acquisition, interpretation and reporting the results of CE examinations all readers used a validated standard terminology (CEST). Each endoscopist reported the presence of any lesions in the segment observed and, if positive, described the size, colour, shape and the final diagnosis of the finding using the standardised terminology. Interobserver agreement was assessed using kappa statistics, a measure of agreement that can be applied to different diagnostic categories. Results: Agreement for the ability to identify lesions showed moderate agreement (kappa 0.55, range between 0.39 to 0.92). The kappa for the presence of active bleeding was 0.58 (range between 0.34 to 0.98). For the shape of the lesions agreement was moderate (k 0.50, range between 0.41 to 0.74). The colour of the lesion identified showed the same level of agreement (k 0.53, range between 0.47 to 0.76). The size of the lesions was the category with worst agreement, with a k value of 0.37 (range from 0.24 to 0.58). Finally agreement for the diagnosis was similar to the other categories with a k level of 0.47 (range from 0.43 to 0.66). Conclusions: In the current study we found a moderate level of agreement for reading CE and this level is similar for the majority of the considered categories. This is the first CE agreement study in wich a enougth number of lesions for statistical analysis was reviewed and all authors used CEST terminology for reporting CE findings.