# Individualized prediction of post-acute pancreatitis diabetes mellitus by combining lipid metabolism and anatomical features

## TL;DR

This study identifies risk factors for diabetes after acute pancreatitis and shows how combining imaging and lipid data can help predict it early.

## Contribution

The study introduces a new quantitative model integrating lipid metabolism and pancreatic anatomy for early prediction of post-acute pancreatitis diabetes.

## Key findings

- Larger pancreaticobiliary junction angles and B-P type are significant risk factors for PPDM.
- A combined model using imaging and lipid data achieved an AUC of 0.886 for predicting PPDM.
- Early prediction allows for personalized prevention and treatment of PPDM.

## Abstract

To investigate the lipid metabolism and anatomical risk factors of post-acute pancreatitis diabetes mellitus (PPDM) and their value in individualized prediction.

A continuous retrospective analysis was conducted on 241 patients with acute pancreatitis (AP) treated in our hospital from January 2017 to December 2021. The type and angle of the pancreaticobiliary junction were measured on magnetic resonance cholangiopancreatography (MRCP) images, and baseline lipid metabolism indicators were collected. We evaluated the risk factors of PPDM using univariate and multivariate Cox proportional hazard analysis, established quantitative prediction models for PPDM, and evaluated the predictive value of lipid metabolism and features of the pancreaticobiliary junction.

Overall, 85 of 241 eligible patients (35.27%) ultimately developed PPDM. Univariate and multivariate analyses showed B-P type in pancreaticobiliary junction (p = 0.017), the angle of junction (p = 0.041), non-high-density lipoprotein (p = 0.029), alcohol index (p < 0.001), body mass index (p = 0.042), inflammatory frequency (p = 0.016), fasting blood glucose (p = 0.002), concomitant hypertension (p < 0.001) were important predictive factors for the occurrence of PPDM. The model that integrated imaging features of the pancreaticobiliary junction has a higher predictive performance than models without imaging features, with an AUC of 0.882 (95% CI, 0.836–0.930). The AUC of the combined model was 0.886 (95% CI, 0.841–0.932), and there was no statistical difference in AUC between the combined model and the pancreaticobiliary junction model (p = 0.340).

The lipid metabolism and morphological characteristics of the pancreaticobiliary junction are additional risk factors for PPDM, and the quantitative prediction model shows moderate predictive performance.

The type and angle of the pancreaticobiliary junction based on MRCP are independent predictors of PPDM, which can quantitatively predict risk in the early stage.

PPDM has an increasing incidence and poor prognosis, which requires early monitoring.Larger angles and B-P type in the pancreaticobiliary junction are risk factors for PPDM.Quantitative prediction of PPDM risk allows for early personalized prevention and treatment.

PPDM has an increasing incidence and poor prognosis, which requires early monitoring.

Larger angles and B-P type in the pancreaticobiliary junction are risk factors for PPDM.

Quantitative prediction of PPDM risk allows for early personalized prevention and treatment.