Objectives: Evaluation of indeterminate biliary strictures remains challenging due to limited sensitivity of endoscopic tissue sampling. Biliary probe-based confocal laser endomicroscopy (pCLE) has shown promise to detect and exclude neoplasia. However, knowledge of whether individual inflammatory criteria are more prevalent in neoplasia compared to benign strictures is limited. The objective of this work is to improve diagnosis of neoplastic and inflammatory conditions using pCLE.
Materials and methods: The charts of all patients who underwent pCLE at a single referral center between 2009 and 2015 were reviewed. ERCP reports were reviewed for eleven Miami and Paris criteria. Primary outcome was the identification of neoplasia by histopathology (defined as high-grade dysplasia and/or adenocarcinoma). To model predictors of neoplasia, we fit a binary regression model incorporating data from pCLE operating criteria, pCLE impression, and PSC status.
Results: 97 patients were identified. In the 27 patients with neoplasia, there was increasing number of Miami malignant criteria (Pearson r = 0.512, p < .001) while inflammatory criteria were less prevalent. 10% (5/51, p < .001) of patients with benign pCLE impression developed neoplasia, while 48% (22/46, p < .001) with suspicious pCLE impressions developed neoplasia. The binary regression model to predict neoplasia had a sensitivity of 83.3%, specificity of 92.5%, and overall accuracy 89.7%.
Conclusions: Presence of malignant criteria and absence of certain inflammatory criteria are more prevalent in patients with neoplasia. Our model, which weights individual imaging components, shows impressive sensitivity and specificity over prior prognostic efforts. Prospective studies will be required to evaluate this model.
Keywords: Probe-based confocal laser endomicroscopy; cholangiocarcinoma; diagnosis of malignancy; indeterminate biliary strictures; primary sclerosing cholangitis.