Agreement between Blinded and Non-Blinded Reviewers of Capsule Endoscopy (CE) Rapid Videos

Abstract

Purpose: Capsule Endoscopy has become the primary diagnostic test for small bowel disease. Its accuracy is dependent upon reader interpretation. Aim: To compare the agreement level between an experienced investigator, non-blinded to the patient medical history and current condition, to a second, less experienced, blinded reviewer. Methods: Capsule studies were identified from Given database of funded studies. Studies were included in this analysis if the study reported a blinded, second reading of CE videos and had at least 10 adult enrollees. CE findings were categorized into major findings including presence of tumors/polyps, bleeding and/or ulceration or minor findings erosions, erythema, celiac, diverticula, nematodes, etc. Statistics included descriptive (mean STD, range), calculation of 95% CI based normal distribution, and ANOVA. Results: 767 patients, 382 (49.8%) men and 385 (50.2%) women, enrolled in 28 studies were analyzed. There were 883 pathological findings (498 major and 385 minor) in 530 patients and 237 videos were normal. Overall there was agreement in 661 of 767 patients (86.2%, 83.5 to 88.5% CI) comprised of 237 normal's and 424 abnormal. Calculated kappa for agreement level is 0.706 (p < 0.01). The overall agreement for major findings is 84.1% (CI 80.6 to 87.2%) vs. 53.8% (CI 48.6 to 58.8%) overall for minor findings (p < 0.01). There was disagreement in 106 patients (13.8%, CI 11.5 to 16.5%). There was no significant difference between disagreements regarding presence of a major or minor lesion. Conclusions: 1) Agreement level between experienced, non-blinded reviewer and a less experienced blinded reviewer on the presence of major lesions is high (Kappa > 0.7, p < 0.01); 2) Proportion of agreement between reviewers is significantly higher for major over minor lesions (p < 0.01); 3) Disagreement between reviewers occurred in 106/767 (13.8%) patients with no significant difference between major and minor lesions. Summary: CE interpretation for major lesions shows high reproducibility between experienced reviewers' non-blinded to patient history and current condition and less experienced, blinded reviewers.