Construction and Validation of Risk Factors and Prognostic Model for Liver Metastasis of Pancreatic Cancer

Abstract

Background Pancreatic cancer (PC) is a common malignancy that often metastasizes to the liver. The presence of liver metastasis (LM) in PC significantly impacts treatment selection and prognosis, but factors affecting the occurrence and prognosis of pancreatic cancer with liver metastasis (PCLM) are not well described. Methods Patients diagnosed with PC between 2010 and 2015 were selected from the Surveillance Epidemiology and End Results (SEER) database. Independent risk factors for PCLM were identified using univariable and multivariable logistic regression. Independent prognostic factors affecting the overall survival (OS) of PCLM patients were analyzed by univariate and multivariate Cox regression, and two nomograms were constructed to predict the risk and prognosis of PCLM. Nomograms were evaluated by receiver operating characteristic (ROC) analysis, C-index, calibration plots, and decision curve analysis (DCA). Results Multivariate logistic regression showed that age, primary site, grade, histological subtype, N stage, radiotherapy, surgery, bone metastasis, and lung metastasis were independent risk factors for PCLM. Multivariable COX regression showed that age, grade, histological subtype, surgery, radiotherapy, chemotherapy and lung metastasis were independent prognostic factors for PCLM. Diagnostic and prognostic nomograms were constructed based on the ROC curves, C-index, calibration curves and DCA curves, and both nomograms showed good predictive performance and clinical utility. Conclusion The two nomograms constructed in this study exhibit good predictive performance in the risk and prognosis of PCLM and may thus serve as a guide for future clinical management of PCLM.