Novel scoring system for recurrence risk classification of surgically resected G1/2 pancreatic neuroendocrine tumors - Retrospective cohort study

Background: Existing staging systems for pancreatic neuroendocrine neoplasms only provide accuracy in stratifying disease severity without enough sensitivity in prognosis predicting. Previously presented models mainly set overall survival as endpoint but ignore the importance of recurrence. Methods: Univariate and multivariate analyses were retrospectively conducted on the potential prognostic factors of 245 patients who underwent curable surgery of G1/2 pancreatic neuroendocrine tumors from December 2002 to May 2018 in our institute. Proposed model based on statistically significant factors were tested for recurrence risk classification estimation validity as measured by discrimination (receiver operator characteristic [ROC] curve and Harrell's c-index [HCI]) and calibration. Results: Multivariate analyses found lymph node metastasis, larger tumor size and grade 2 independent risk factors of disease-free survival (DFS). The novel scoring system for recurrence risk classification sorted patients into three groups: score < 15.4 for low risk (79.0%, 3- and 5-year recurrence risk 0.8% and 4.3%, median DFS not reached), score > 15.4, < 24.5 for intermediate risk (20.8%, 3- and 5-year recurrence risk 11.6% and 21.4%, median DFS 70 months) and score > 24.5 for high risk (10.2%, 3- and 5-year recurrence risk 37.3% and 68.7%, median DFS 49 months).The area under curve (AUC) and HCI of our system were 0.871 and 0.929, superior to those of European Neuroendocrine Tumor Society (AUC 0.833, HCI 0.806). Conclusion: The presented system can be utilized to identify G1/2 pancreatic neuroendocrine tumors patients with high recurrence risk, which might be appropriate for peri-operative adjuvant therapy.