Feasibility and accuracy of confocal endomicroscopy in comparison with narrow-band imaging and chromoendoscopy for the differentiation of colorectal lesions.

ABSTRACT

OBJECTIVES: Several advanced imaging techniques have been developed to improve differentiation of colorectal lesions. These techniques need to be assessed for both feasibility and accuracy in order to prove their value in daily clinical practice. The current study assessed the feasibility of probe-based confocal laser endomicroscopy (pCLE) in acquiring videos of sufficient quality. Furthermore, the accuracy of pCLE for the differentiation of colorectal lesions was assessed and compared with narrow-band imaging (NBI) and chromoendoscopy (CE). METHODS: Consecutive patients scheduled for surveillance colonoscopy at our centre were included. All procedures were performed by two expert colonoscopists, who previously participated in studies evaluating pCLE, NBI, and CE. All detected lesions during colonoscopy were differentiated real-time with NBI and CE for Kudo pit pattern. Lesions were also assessed real-time for vascular pattern intensity (VPI) during NBI. Subsequently, pCLE videos of each lesion were acquired and biopsies were taken for histopathology. All pCLE videos were assessed post-hoc for the percentage of time demonstrating sufficient image quality (i.e., depicting at least one crypt or vessel). Finally, pCLE videos were assessed post-hoc for diagnostic accuracy by two experts. RESULTS: A total of 154 lesions detected in 64 patients were included. Accuracy of Kudo pit pattern with NBI for predicting neoplasia (88.7%) was significantly better than accuracy of VPI (77.5%, P = 0.05) but not significantly different from CE (89.3%, P = 0.125). During pCLE, no histology was shown at all on the video in 19 lesions (12%). The mean time to acquire a pCLE video of the remaining 135 lesions was 50 seconds (s.d. 47) per lesion. The median percentage demonstrating sufficient quality per video was 40.5% (interquartile range 21.2-67.0). Accuracy of pCLE for both observers (66.7 and 71.9%) was significantly lower than accuracy of CE (P < 0.001) and NBI (P < 0.001). CONCLUSIONS: Video acquisition with pCLE could not be achieved in a small number of lesions. The majority of pCLE videos demonstrated insufficient quality in more than half of the time recorded. Moreover, post-hoc accuracy of pCLE was significantly lower in comparison with real-time accuracy of CE and NBI. Future research should assess whether further increase in experience could improve pCLE video acquisition and determine the real-time accuracy of pCLE for differentiating colorectal lesions.