ABSTRACT
BACKGROUND: This study strives to define prognostic models for outcome after surgery for malignant pancreatic neuroendocrine tumors. METHODS: Forty-one patients were included. Prognostic models for mortality and disease recurrence were developed with multivariate binary logistic regression. RESULTS: The proposed prognostic model for tumor recurrence risk after surgery in percentage (AUROC = 0.774, 95%CI = 0.611-0.937) is: Risk in % = Exp(Y)/(1 + Exp(Y)), with Y = -4.360 + (0.015 × tumor diameter in cm) + (0.010 × preoperative platelet count in thousand/µl) + (1.077 × distant metastases, if yes = 1; if no = 0) + (-0.026 × Ki-67-positive cells in %) + (-1.086 × upper abdominal pain, if yes = 1; if no = 0). The proposed prognostic model for observed 3-year survival probability after surgery in % (AUROC = 0.932, 95%CI = 0.857-0.999) is: Survival probability in % = Exp(Y)/(1 + Exp(Y)), with Y = -12.492 + (0.054 × preoperative platelet count in thousand/µl) + (0.112 × minimal distance of the resection margin from the tumor in mm) + (-1.574 × number of positive lymph nodes) + (2.292 × histological tumor infiltration, if yes = 1; if no = 0) CONCLUSIONS: The platelet count was identified as a relevant risk factor. Proposed prognostic models with good model-fit display properties that indicate potential clinical usefulness.