Establishment and validation of early prediction model for hypertriglyceridemic severe acute pancreatitis

Abstract

Abstract Background The prevalence of hypertriglyceridaemia-induced acute pancreatitis (HTG-AP) is increasing due to improvements in living standards and dietary changes. However, currently, there is no clinical multifactor scoring system specific to HTG-AP. This study aimed to screen the predictors of HTG-SAP and combine several indicators to establish and validate a visual model for the early prediction of HTG-SAP. Methods The clinical data of 266 patients with HTG-SAP were analysed. Patients were classified into severe (N = 42) and non-severe (N = 224) groups according to the Atlanta classification criteria. Several statistical analyses, including one-way analysis, least absolute shrinkage with selection operator (LASSO) regression model, and binary logistic regression analysis, were used to evaluate the data. Results The univariate analysis showed that several factors showed no statistically significant differences, including the number of episodes of pancreatitis, abdominal pain score, and several blood diagnostic markers, such as lactate dehydrogenase (LDH), serum calcium (Ca2+), C-reactive protein (CRP), and the incidence of pleural effusion, between the two groups (P < 0.000). LASSO regression analysis identified six candidate predictors: CRP, LDH, Ca2+, procalcitonin (PCT), ascites, and Balthazar computed tomography grade. Binary logistic regression multivariate analysis showed that CRP, LDH, Ca2+, and ascites were independent predictors of HTG-SAP, and the area under the curve (AUC) values were 0.886, 0.893, 0.872, and 0.850, respectively. The AUC of the newly established HTG-SAP model was 0.960 (95% confidence interval: 0.936–0.983), which was higher than that of the bedside index for severity in acute pancreatitis (BISAP) score, modified CT severity index, Ranson score, and Japanese severity score (JSS) CT grade (AUC: 0.794, 0.796, 0.894 and 0.764, respectively). The differences were significant (P < 0.01), except for the JSS prognostic indicators (P = 0.130). The Hosmer–Lemeshow test showed that the predictive results of the model were highly consistent with the actual situation (P > 0.05). The decision curve analysis plot suggested that clinical intervention can benefit patients when the model predicts that they are at risk for developing HTG-SAP. Conclusions CRP, LDH, Ca2+, and ascites are independent predictors of HTG-SAP. The prediction model constructed based on these indicators has a high accuracy, sensitivity, consistency, and practicability in predicting HTG-SAP.