Background: To analyze the risk factors of disease progression in patients with hypertriglyceridemic acute pancreatitis (HTG-AP), and to establish a prediction model for ultra-early (24h) severity of the disease.
Methods: A total of 205 eligible patients diagnosed with HTG-AP who were hospitalized in Zhuhai people's Hospital (Zhuhai hospital affiliated with Jinan University) were included. Among them,147 patients were enrolled in the modular group and 58 patients were enrolled in the validation group.Based on R language analysis (version 3.6.3), the R package glmnet package was used to conduct Lasso regression analysis on variables (P<0.05) with inter-group differences and select the optimal predictive variables which can predict disease progression in early. The R package (rms package) was used and the optimal prediction variables were analyzed by Logistic regression. Then, the Decision Curve Analysis (DCA) was drawn to further verify the benefits of clinical patients in the model. Finally, external verification was carried out to further assess the predictive efficiency and generalizability of the risk prediction model.
Results: (1).There were no statistical difference in the basic data (including sex,age,basic disease) between mild group and severe group in the modular group(P>0.05). (2).There were statistical difference in the laboratory results (including GLU,Cr,PCT,TG,CHOL and D-dimer) between mild group and severe group in the modular group(P<0.05). (3).The results of multivariate Lasso regression indicated that GLU,Cr,PCT,TG,CHOL and D-dimer were independent predictors of hypertriglyceridemic acute pancreatitis severity. The optimal prediction variables were analyzed by Logistic regression. (4).Nomogram was drawn by rms package. The AUC was 0.788 and the C-index was 0.788. The C-index of internal verification was C-index=0.788(95%CI：0.785-0.791). Decision Curve Analysis showed that the prediction model had better clinical usefulness than active treatment or negative treatment if the threshold probability is within 12%-99%. (5).In the validation group, the AUC and C-index of external verification were 0.821(95%CI:0.714-0.928). Decision Curve Analysis showed that the prediction model had better clinical usefulness than active treatment or negative treatment if the threshold probability is within 10%-99%.
Conclusions: TG, CHOL, D-dimer and PCT within 24 hours of HTG-AP onset were the risk factors which could aggravate disease progression. The early prediction model based on the four risk factors could relatively accurately predict exacerbation trend in early phase of hypertriglyceridemic acute pancreatitis.