Risk factors for disease severity in hypertriglyceridemic acute pancreatitis: a single-center retrospective study

Hypertriglyceridemia is an increasingly common cause of acute pancreatitis (AP). Patients with hypertriglyceridemic acute pancreatitis (HTG-AP) have higher complication and mortality rates compared to those with other etiologies. Early prediction of disease severity remains challenging due to the lack of readily available tools specifically for HTG-AP.

This was a single-center retrospective cohort study. A total of 214 HTG-AP patients were classified into mild acute pancreatitis (MAP, n = 106) and moderately severe/severe acute pancreatitis (MSAP/SAP, n = 108) groups based on the revised Atlanta criteria. Clinical characteristics and laboratory parameters were compared between the two groups. Binary logistic regression analysis and ROC analysis were performed to identify risk factors and develop a combined predictive model. Bootstrap analysis was performed for internal validation, and calibration curves were utilized to evaluate model calibration.

The MSAP/SAP group exhibited elevated triglyceride (TG), amylase (AMY), blood glucose (GLU), C-reactive protein (CRP), and white blood cell (WBC) levels, but lower serum calcium (Ca²⁺) and apolipoprotein A1 (ApoA1) levels. Binary logistic regression analysis identified several independent risk factors for MSAP/SAP: TG, CRP, WBC, and Ca²⁺. The combined predictive model achieved an area under the curve (AUC) of 0.837. At the optimal cut-off value of 0.48, the sensitivity and specificity of the combined predictive indicator were 77.4% and 75.0%, respectively. Bootstrap validation demonstrated that the 95% confidence intervals for the regression coefficients of TG, Ca²⁺, CRP, and WBC did not include zero. The Hosmer-Lemeshow goodness-of-fit test showed a p-value of 0.312 (>0.05).

Elevated TG, CRP, and WBC levels, as well as decreased Ca²⁺, are independent risk factors for severe HTG-AP. A combined model based on these readily available early parameters demonstrates robust predictive performance, stability, and calibration.