Predicting Liver Metastasis in Pancreatic Neuroendocrine Tumor After Surgery: A Population-Based Study

Abstract

Background The occurrence of liver metastasis in pancreatic neuroendocrine tumor (pNET) after primary site surgery significantly hampers the improvement of patient’s overall survival (OS). Therefore, it is necessary for early detection of metastatic lesions. However, the relationship between clinical variables and the liver metastasis potential remains obscure. Methods Detailed information of pNET patients received primary site surgery was retrieved from the Surveillance, Epidemiology and End Results (SEER) database between 2010 and 2019. Univariate and multivariate logistic regression analysis were recruited to generate independent risk factors of liver metastasis to construct a model presented as a nomogram using training cohort of SEER database. Moreover, a testing cohort from SEER database and a cohort of 96 patients from Second Affiliated Hospital of Zhejiang University School of Medicine were further recruited for internal and external verification respectively. The receiver operating characteristic curve, calibration curve, decision curve analysis (DCA) and clinical impact curve (CIC) were used to evaluate the accuracy, reliability and clinical application value respectively. The risk subgroups were finally generated according to the score of the nomogram. Results 2458 pNET patients were included in the present study. And 1638 of them were assigned as training cohort and 820 of them were assigned as testing cohort. Tumor size, AJCC T stage, functional status and other site metastases were considered as independent risk factors of liver metastasis via multivariate logistic regression analysis (all, p < 0.05). Our nomogram showed an excellent accuracy with the area under curve (AUC) of 0.821 in training cohort, 0.766 in testing cohort and 0.817 in validation cohort, respectively. Moreover, the calibration curve, DCA and CIC indicated a better net benefit and clinical application value in training cohort, testing cohort and validation cohort compared with single variate. Finally, pNET patients could be classified into low, medium and high risk of liver metastasis. Patients with high risk of liver metastasis showed a significant poorer OS compare to other two groups. Conclusion Tumor size can be an important predictor of liver metastasis in pNET patients. The nomogram we established could predict liver metastasis of pNET patients after surgery accurately.