Background: Confocal laser endomicroscopy (CLE) can provide in vivo subcellular resolution images of esophageal lesions. However, the learning curve in interpreting CLE images of precancerous or early-stage esophageal squamous cancer is unknown. The goal of this study is to evaluate the diagnostic accuracy and inter-observer agreement for differentiating esophageal lesions in CLE images among experienced and inexperienced observers and to assess the learning curve.
Method: After a short training, 8 experienced and 14 inexperienced endoscopists evaluated in sequence 4 sets of high-quality CLE images. Their diagnoses were corrected and discussed after each set. For each image, the diagnostic results, confidence in diagnosis, quality and time to evaluate were recorded.
Results: Overall, diagnostic accuracy was greater for the second, third, fourth set of images as compared with the initial set (odds ratio [OR] 2.01, 95% CI 1.22-3.31; 7.95, 3.74-16.87; and 6.45, 3.14-13.27), respectively, with no difference between the third and fourth sets in accuracy (p = 0.67). Previous experience affected the diagnostic accuracy only in the first set of images (OR 3.70, 1.87-7.29, p<0.001). Inter-observer agreement was higher for experienced than inexperienced endoscopists (0.732 vs. 0.666, p<0.01).
Conclusion: CLE is a promising technology that can be quickly learned after a short training period; previous experience is associated with diagnostic accuracy only at the initial stage of learning.