Should We Use the Suspected Blood Indicator in Wireless Capsule Endoscopy? A Prospective Analysis of 17689 Frames from the Royal London Hospital

Abstract

Background: The suspected blood indicator (SBI) feature of capsule endoscopy (CE) was initially developed as a potential method to rapidly screen for intestinal lesions with active or recent bleeding. Unfortunately previous studies have demonstrated that CE SBI performance characteristics are suboptimal: in some studies, sensitivities vary from 64% (suspected Crohns) to 41% (Anaemia). Other studies have shown even lower sensitivities (37%) and specificities (59%) for lesions felt to have bleeding potential. Limited data exists correlating individual SBI frame analysis with CE indication and findings, or on assessing the time burden associated with image analysis. Aim: To determine the performance characteristics of SBI images by indication and findings and to calculate time burdens associated with reading the SBI in CE. Method: 17689 sequential SBI positive images from 347 capsule endoscopies performed at a single tertiary center were analysed and divided into 2 categories: images that were consistent with true red coloration (True positive - TP) and images where no red coloration was seen (False positive - FP). Data was correlated with gender, indications and findings and summary statistics calculated. Results: The mean number of SBI images per CE was 46 (range 8-60)being lowest for hematochezia (8) and highest for OGIB (60). The average time burden of analysing SBI images was 11.5 minutes (range 2-15 min). The global positive predictive value (PPV) for SBI images was 60%. The lowest PPV ranged from 51% for abdominal pain to a high of 64% for hematochezia. CE Indications included: Iron deficiency (49%), Crohns (15%), Obscure GI bleed (25%), Celiac disease (3%), Polyposis syndromes(1%). CE findings included: Angioectasia (50%), Normal (17%), Inflammation (12%), Crohns (8%), Polyps/tumor(6%). Conclusions: The low overall positive predictive values of the SBI of 60% suggests that current routine use of the tool may be unhelpful. Moreover the additional time burden up to 11.5 minutes associated with SBI image analysis (at 15 second/image) adds significantly to the time taken to read CEs.