Clinical characteristic and pathogenesis of tumor-induced acute pancreatitis: a predictive model

Abstract

Abstract Background The aim of our study was to investigate the clinical characteristics and pathogenesis of tumor-induced acute pancreatitis (AP), and to develop a reliable prediction model of the clinical features to guide the diagnosis and treatment. Methods Patients with AP between January 2013 and December 2021 were enrolled in the study and were subdivided into the tumor group and the non-tumor group. The tumor group was subdivided into three groups based on the primary sites. Characteristic parameters, laboratory and imaging results were compared between groups. Least absolute shrinkage and selection operator regression model, XGBoost and random forest model were used to select the predictors associated with tumor-induced AP. Logistic regression analysis was used to validate the performance of the selected predictors and a nomogram was established to provide individualized probability of a tumor origin for AP. Results A total amount of 8970 patients were admitted for AP during the study period, and 8637 AP patients were enrolled in the study. Of these, 100 cases (1.16%) were tumor-induced AP. The tumor group was significantly older than the non-tumor group (t = 6.050, p = 0.000). Mild AP was observed in 90 cases, moderate AP in 9 cases and severe AP in one case. Tumors respectively originated from distal bile duct (14 cases), ampulla (13 cases) and pancreas (73 cases). The median time from initial AP to tumor diagnosis was 8.57 weeks and the median number of episode was 2 in the tumor group, which significantly surpassed the non-tumor group (p = 0.000). Age, white blood cell count, percentage of neutrophils, pancreatic or bile duct dilation and recurrent attacks were selected independent predictors for tumor origin. A nomogram model based on these factors was established. Conclusion For patients with agnogenic AP, elderly man, recurrent attacks, pancreatic or bile duct dilatation and continuous no significant increase of inflammatory markers prompt to further screening of pancreatic biliary and ampulla.

Keywords

• Tumor-induced acute pancreatitis
• Characteristics
• Pathogenesis
• Predictor
• Machine learning