Retrospective study: analysis of risk factors and construction of prediction model for pancreatic steatosis

BACKGROUND: Pancreatic steatosis, closely related to various metabolic diseases has no unified standard for the diagnosis. The purpose of this study was to construct and validate a risk prediction model of pancreatic steatosis. METHODS: 854 cases admitted to Xi'an Mei Zhao physical examination center were divided into modeling set (n = 598) and validation set (n = 256) according to the ratio of 7:3, and the risk prediction model was constructed and validated based on the risk factors. RESULTS: Of 598 cases in modeling set, 232 were in the pancreatic steatosis group and 366 were in the non pancreatic steatosis group. There were significantly statistical differences in Age, Gender, Body Mass Index (BMI), Glucose (GLU), Total Cholesterol (TC), Triglycerides (TG), High-Density Lipoprotein (HDL), Low-Density Lipoprotein (LDL), Aspartate Aminotransferase (AST), Alanine Aminotransferase (ALT), Ratio of Aspartate Aminotransferase to Alanine Aminotransferase (AST/ALT), Non-Alcoholic Fatty Liver Disease (NAFLD), Pancreas computed tomography (CT) value, Spleen CT value, Liver CT value as well as Pancreas/Spleen CT value (P < 0.05) between pancreatic steatosis cases and non pancreatic steatosis cases. Cox regression showed that Age (97.5% CI 0.981-1.030), BMI (97.5% CI 1.198-1.436), GLU (97.5% CI 0.921-1.521), TC (97.5% CI 0.800-2.130), HDL (97.5% CI 0.074-0.649), LDL (97.5% CI 0.603-1.877), AST/ALT (97.5% CI 0.241-0.917), Pancreas CT value (97.5% CI 0.788-0.904), Liver CT value (97.5% CI 0.962-1.015), and Pancreas/Spleen CT value (97.5% CI 0.001-0.649) were independent risk factors for pancreatic steatosis (P < 0.05). Based on above variables, a pancreatic steatosis model was successfully constructed, the validation results showed the model had good credibility and predictive value. CONCLUSION: Age, BMI, GLU, TC, HDL, LDL, AST/ALT, Pancreas CT value, Liver CT value, Pancreas/Spleen CT value are risk factors for pancreatic steatosis, and this model can identify the patients at high risk of pancreatic steatosis.