Method:
1). Patient: This retrospective study included 321 children diagnosed with acute lymphoblastic leukemia (ALL) at Beijing Children’s Hospital between November 2016 and May 2023. All patients were identified from the hospital’s case management system. ALL diagnosis was confirmed through bone marrow morphology. Acute pancreatitis was defined based on the presence of at least two out of three criteria: acute gastrointestinal symptoms, serum lipase or amylase levels at least three times the upper limit of normal, and characteristic imaging findings. Patients were excluded if they had pre-existing organ failure, a history of pancreatic surgery, prior pancreatitis, liver cirrhosis, hepatitis, or incomplete information in the case management system. The control group consisted of patients without pancreatitis. The study was deemed minimal risk by the ethics committee, and therefore informed consent was waived.
2). Data collection: demographic and other characteristics of the patients were collected, including age, gender, weight, Immunophenotyping, risk classification, chemotherapy phase, cumulative dose of asparaginase administered (usually Pegaspargase in our center), MRD level after induction therapy, presence of infection prior to chemotherapy, presence of pancreatic morphologic abnormalities as indicated by ultrasound, liver and kidney function, lipids, bilirubin, pancreatin, C-reactive protein, etc.
3). Statistical analysis: To build the predictive model, variables with over 20% missing data were excluded. The remaining missing data were addressed using multiple imputation with the “mice” package in R software. Patients were randomly divided into training (80%) and validation (20%) cohorts. The predictive model was developed using the training data and then evaluated for accuracy using the validation data. Descriptive statistics for continuous variables included mean and standard deviation or median and interquartile range, while categorical variables were presented as percentages. Chi-square or Fisher’s exact test was used to compare categorical variables between groups, while t-tests were employed for continuous variables.
Independent risk factors for pancreatitis were screened by lasso regression which were further output for the construction of nomogram model. To prepare the independent variables for analysis, disordered categorical variables were converted into dummy variables. Continuous variables underwent standardization before being included in the regression analysis. The predictive model’s performance was evaluated using the area under the receiver operating characteristic curve (AUC). An AUC exceeding 0.7 indicated a useful model, while an AUC between 0.8 and 0.9 suggested high diagnostic accuracy.[9, 10] To assess the calibration of the prediction model, a calibration curve was generated. [11]
Statistical significance was defined as a p-value less than 0.05, and all analyses were performed using R version 4.0.3.
The local ethics committee of Beijing Children’s Hospital approved this study.