Comparison of Probe-Based Confocal Laser Endomicroscopy With Virtual Chromoendoscopy for Classification of Colon Polyps

BACKGROUND & AIMS: Probe-based confocal laser endomicroscopy (pCLE) allows in vivo imaging of tissue at micron resolution. Virtual chromoendoscopy systems, such as Fujinon intelligent color enhancement and narrow band imaging, also have potential to differentiate neoplastic colorectal lesions. The accuracy of these systems in clinical practice is, however, unknown. Our primary aim was to compare sensitivity and specificity of pCLE to virtual chromoendoscopy for classification of colorectal polyps using histopathology as a gold standard. A secondary aim was to compare sensitivity and specificity of pCLE to virtual chromoendoscopy using a modified gold standard that assumed that all polyps >= 10 mm had malignant potential and were considered neoplastic or high risk. METHODS: Patients underwent colonoscopy using high-resolution colonoscopes. The surface pit pattern was determined with NBI or FICE in all patients. Confocal images were recorded and subsequently analyzed offline, blinded to the endoscopic characteristics and histopathology. Each polyp was diagnosed as benign or neoplastic based on confocal features according to modified Mainz criteria. RESULTS: A total of 119 polyps (81 neoplastic, 38 hyperplastic) from 75 patients was assessed. The pCLE had higher sensitivity compared to virtual chromoendoscopy when considering histopathology as gold standard (91% vs 77%; P = .010) and modified gold standard (88% vs 76%; P = .037). There was no statistically significant difference in specificity between pCLE and virtual chromoendoscopy when considering histopathology or modified gold standard. CONCLUSIONS: Confocal endomicroscopy demonstrated higher sensitivity with similar specificity in classification of colorectal polyps. These new methods may replace the need for ex vivo histological confirmation of small polyps, but further studies are warranted.