International Validation of a Nomogram to Predict Recurrence after Resection of Grade 1 and 2 Nonfunctioning Pancreatic Neuroendocrine Tumors

Background: Despite the low recurrence rate of resected nonfunctional pancreatic neuroendocrine tumors (NF-pNETs), nearly all patients undergo long-term surveillance. A prediction model for recurrence may help select patients for less intensive surveillance or identify patients for adjuvant therapy. The objective of this study was to assess the external validity of a recently published model predicting recurrence within 5 years after surgery for NF-pNET in an international cohort. This prediction model includes tumor grade, lymph node status and perineural invasion as predictors. Methods: Retrospectively, data were collected from 7 international referral centers on patients who underwent resection for a grade 1–2 NF-pNET between 1992 and 2018. Model performance was evaluated by calibration statistics, Harrel’s C-statistic, and area under the curve (AUC) of the receiver operating characteristic curve for 5-year recurrence-free survival (RFS). A sub-analysis was performed in pNETs >2 cm. The model was improved to stratify patients into 3 risk groups (low, medium, high) for recurrence. Results: Overall, 342 patients were included in the validation cohort with a 5-year RFS of 83% (95% confidence interval [CI]: 78–88%). Fifty-eight patients (17%) developed a recurrence. Calibration showed an intercept of 0 and a slope of 0.74. The C-statistic was 0.77 (95% CI: 0.70–0.83), and the AUC for the prediction of 5-year RFS was 0.74. The prediction model had a better performance in tumors >2 cm (C-statistic 0.80). Conclusions: External validity of this prediction model for recurrence after curative surgery for grade 1–2 NF-pNET showed accurate overall performance using 3 easily accessible parameters. This model is available via www.pancreascalculator.com.