Establishment And Validation Of A Predictive Model For New Onset Diabetes Mellitus After First-Attack Acute Pancreatitis

Objective:At present,there are few studies on new-onset diabetes(DM)after firstattack acute pancreatitis(AP)and the related risk factors are still not very clear.The purpose of this study is to identify risk factors and established a nomogram to predict the new onset diabetes mellitus after first-attack acute pancreatitis.Methods:Clinical data of 396 patients with AP who met the inclusion and exclusion criteria were collected at the Second Affiliated Hospital of Nanchang University from June 2016 to June 2017.Patients were randomly divided into training set and validation set.Univariate and multivariate logistic regression analysis were used to determine the independent risk factors associated with new onset DM after firstattack AP.Predictive model was established in the training set and were validated in the validation set.The performance of the predictive model was assessed by discrimination and calibration.Clinical effectiveness of nomogram under different probability thresholds was evaluated by decision curve analysis.Results:The incidence of new onset DM after first-attack AP was 8.9% in this study.Univariate and multivariate regression analyses identified the following variables as independent risk factors for new onset DM after first-attack AP: Hyperlipemia(odds ratio(OR)=6.877),ALP(OR=6.008),GGT(OR=0.025),GLU(OR=7.731),CT grade(OR=4.952),APACHE II grade(OR=3.824).Predictive models and nomograms were established using the above risk factors.External validation showed that the model has good discrimination ability.The area under the ROC curve(AUC)were 0.8983 and 0.802 in the training and validation sets,respectively.The calibration curve for the probability of first onset DM after first-attack AP showed that the nomogram predictions probability were in good agreement with actual observations probability(Hosmer-Lemeshow test: P=0.520 and 0.314 for the training and validation set,respectively).Finally,the decision curve analysis showed that the predictive nomogram has good clinical application value in 10%-65% interval.Conclusion:This study constructed and validated a prediction nomogram incorporating demographical data as well as clinical and imaging characteristics that could help identify patients who might develop DM after first-attack AP.