Background & Aims: Differentiating pancreas cyst subtype is a fundamental step to develop management for patients with incidental pancreatic cystic lesions (PCL), but current standard of care has suboptimal accuracy. We evaluated the diagnostic accuracy of EUS-nCLE for diagnosing PCLs in a prospective study.
Methods: Consecutive patients with at least one PCL measuring ≥ 1.8 cm and having a definitive histopathological diagnosis were selected from two prospective observational studies: (1) INDEX, a single-center study conducted from 2015 to 2018 (NCT02516488, n=145, and (2) CLIMB, an ongoing multi-center study initiated in 2018 (NCT03492151, n=275; currently 12 US tertiary centers). EUS-nCLE patterns were interpreted and documented in real-time by the performing endosonographer during the index procedure. We conducted analyses to assess the accuracy of nCLE in the differentiation (mucinous vs. non-mucinous) and specific diagnoses of PCLs. Other diagnostic modalities, including cyst fluid next-generation sequencing (NGS), glucose, CEA, and cytology, were included in the analysis when performed for the evaluation of PCLs. Diagnostic parameters are reported with 95% confidence interval (CI). NGS involved analysis of neoplasm associated variants in 48 genes.
Results: A total of 194 participants were included, with a mean PCL diameter of 39.1±18.4 mm. One hundred eleven (57.2%) of the PCLs were mucinous based on histopathology. At least 87% had at least 1 worrisome feature based on the Fukuoka criteria. The mean duration of EUS-nCLE was 7 ± 3 minutes.
The sensitivity, specificity, and accuracy of EUS-nCLE in diagnosing specific PCL types are shown in Table 1. EUS-nCLE differentiated mucinous cysts with 96% accuracy (95% CI 93-99%), IPMNs with 93% accuracy (95% CI 89-96%), and 93% (95% CI 89-96%) for mucinous cystic neoplasms (MCNs). The accuracy was similarly excellent for serous cyst adenomas (SCA) at 99% (95% CI 96-100%), cystic-neuroendocrine tumors (cystic-NET) at 99% (95% CI 96-100%), and for pseudocysts at 97% (95% CI 93-99%).
The diagnostic accuracies (Table 2) for differentiating mucinous vs. non-mucinous lesions were as follows: 96% (95% CI 93-99%) for EUS-nCLE (n=194), 86% (95% CI 80-91%) for NGS (n=140), and 80% (95% CI 74-85%) for cyst fluid glucose with CEA and/or cytology (n=193). When considering cysts as mucinous if at least one parameter was consistent with a mucinous lesion, the accuracy for CEA, cytology, and/or glucose was 80% (95% CI 74-85%).
Conclusions: In this large prospective study, real-time interpretation of EUS-nCLE PCL patterns can distinguish between mucinous and non-mucinous cysts and diagnose specific PCL types with high accuracy. Future research should investigate methods to optimize endoscopists' proficiency in CLE image acquisition and interpretation, with the goal of optimizing the management of patients with PCLs.

<b>Table 1.</b> Diagnostic performance of real-time EUS-nCLE expert interpretation of PCLs compared with definitive histopathology

Table 1. Diagnostic performance of real-time EUS-nCLE expert interpretation of PCLs compared with definitive histopathology

<b>Table 2. </b>Diagnostic classification of mucinous vs. non-mucinous PCLs compared with definitive histopathology

Table 2. Diagnostic classification of mucinous vs. non-mucinous PCLs compared with definitive histopathology

Samer El-Dika

Stanford University

Appearances