Diagnosis of gastric intraepithelial neoplasia by narrow-band imaging and confocal laser endomicroscopy.

Aim: To evaluate the diagnosis of different differentiated gastric intraepithelial neoplasia (IN) by magnification endoscopy combined with narrow-band imaging (ME-NBI) and confocal laser endomicroscopy (CLE).

Methods: Eligible patients with suspected gastric IN lesions previously diagnosed by endoscopy in secondary hospitals and scheduled for further diagnosis and treatment were recruited for this study. Excluded from the study were patients who had liver cirrhosis, impaired renal function, acute gastrointestinal (GI) bleeding, coagulopathy, esophageal varices, jaundice, and GI post-surgery. Also excluded were those who were pregnant, breastfeeding, were younger than 18 years old, or were unable to provide informed consent. All patients had all mucus and bile cleared from their stomachs. They then received upper GI endoscopy. When a mucosal lesion is found during observation with white-light imaging, the lesion is visualized using maximal magnification, employing gradual movement of the tip of the endoscope to bring the image into focus. Saved images are analyzed. Confocal images were evaluated by two endoscopists (Huang J and Li MY), who were familiar with CLE, blinded to the related information about the lesions, and asked to classify each lesion as either a low grade dysplasia (LGD) or high grade dysplasia (HGD) according to given criteria. The results were compared with the final histopathologic diagnosis. ME-NBI images were evaluated by two endoscopists (Lu ZS and Ling-Hu EQ) who were familiar with NBI, blinded to the related information about the lesions and CLE images, and were asked to classify each lesion as a LGD or HGD according to the "microvascular pattern and surface pattern" classification system. The results were compared with the final histopathologic diagnosis.

Results: The study included 32 pathology-proven low grade gastric IN and 26 pathology-proven high grade gastric IN that were detected with any of the modalities. CLE and ME-NBI enabled clear visualization of the vascular microsurface patterns and microvascular structures of the gastric mucosa. The accuracy of the CLE and the ME-NBI diagnosis was 88% (95% CI: 78%-98%) and 81% (95% CI: 69%-93%), respectively. The kappa coefficient of agreement between the histopathology and the in vivo CLE imaging was 0.755; between the histopathology and the in vivo CLE imaging was 0.615. McNemar's test (binomial distribution used) indicated that the agreement was significant (P < 0.05). When patients were diagnosed by ME-NBI with CLE, the overall accuracy of the diagnosis was 86.21% (95% CI: 73%-96%), and the kappa coefficient of agreement was 0.713, according to McNemar's test (P < 0.05).

Conclusion: Higher diagnostic accuracy, sensitivity and specificity of CLE over ME-NBI indicate the feasibility of these two techniques for the efficacious diagnostic classification of gastric IN.