Abstract
Background. Accurate diagnosis and clinical management of indeterminate biliary strictures are often a challenge. Tissue confirmation modalities during Endoscopic Retrograde Cholangiopancreatography (ERCP) suffer from low sensitivity and poor diagnostic accuracy. Probe-based confocal laser endomicroscopy (pCLE) has been shown to be sensitive for malignant strictures characterization (98%) but lacks specificity (67%) due to inflammatory conditions inducing false positives. Methods. Six pCLE experts validated the Paris Classification, designed for diagnosing inflammatory biliary strictures, using a set of 40 pCLE sequences obtained during the prospective registry (19 inflammatory, 6 benign, and 15 malignant). The 4 criteria used included (1) multiple thin white bands, (2) dark granular pattern with scales, (3) increased space between scales, and (4) thickened reticular structures. Interobserver agreement was further calculated on a separate set of 18 pCLE sequences. Results. Overall accuracy was 82.5% (n = 40 retrospectively diagnosed) versus 81% (n = 89 prospectively collected) for the registry, resulting in a sensitivity of 81.2% (versus 98% for the prospective study) and a specificity of 83.3% (versus 67% for the prospective study). The corresponding interobserver agreement for 18 pCLE clips was fair (k = 0.37). Conclusion. Specificity of pCLE using the Paris Classification for the characterization of indeterminate bile duct stricture was increased, without impacting the overall accuracy.