Impact of tumor size and location on endoscopic ultrasound-guided sampling of pancreatic neuroendocrine tumors: A recursive partitioning analysis.

BACKGROUND: Current guidelines provide weak recommendations to treat small (<2 cm) non-functional pancreatic neuroendocrine tumors with low Ki-67 proliferation index either by resection or clinical follow-up. However, there is a lack of consensus regarding the minimal size of pNET, which allows EUS-guided biopsy with high enough diagnostic accuracy for stratification. METHODS: We conducted a retrospective, bicentric analysis of patients who had undergone EUS-guided pNET sampling in two tertiary care Endoscopy Units in Germany and Poland. Using a recursive partitioning of the tree-aided model, we aimed to stratify the probability of successful EUS-guided biopsy of pNET lesions according to their size and location. RESULTS: In our pNET cohort, successful histological confirmation of a pNET diagnosis was achieved in 59/69 (85.5%) cases at the initial EUS-guided biopsy. In 41 patients with a pNET size less than 18.5 mm, the EUS-guided first biopsy was successful in 90.2%. In 16 of these patients with smaller lesions, EUS-guided sampling was 100% in very small (less than 11 mm) and extremely small lesions (less than 8 mm). The biopsy success rate was 100% in tail lesions in the size range between ≥5.95 and <8.1 mm but only 33.3% independent of the investigator in pancreatic head or body, with an error rate of 11.2% CONCLUSION: Using a recursive partitioning of the tree-aided stratification model, we demonstrate for the first time that in balancing risks and benefits, very small pNETs (<1 cm) in the tail of the pancreas should be sampled under EUS-guidance.