Based On Preoperative Serological Indexes And Objective Imaging Indicators,a Model For Predicting The Efficacy Of The First TACE In Patients With Liver Cancer Was Established And Verified

Objectives:Based on preoperative serological indexes and objective imaging indexes,an easy-to-use prediction model was established to predict the efficacy of the first hepatic arterial chemoembolization in patients with intermediate-stage hepatocellular carcinoma,which was used to screen out suitable patients with intermediate-stage hepatocellular carcinoma before the first hepatic arterial chemoembolization.Methods:Patients who underwent digital subtraction angiography examination in the Department of Gastroenterology or Interventional Surgery of the First Affiliated Hospital of Nanchang University from January 2012 to January 2022 were selected.A total of 4482 patients who received digital subtraction angiography were retrospectively screened.A total of 427 hepatocellular carcinoma patients who received the first transcatheter arterial chemoembolization treatment were selected for analysis,and randomly divided into training group and validation group.Minimal absolute contraction and selection operator,univariate and multivariate logistic regression analysis were used to screen predictors.Build candidate predictive models based on predictors.Use the Generalized Variance Inflation Factor and correlation analysis to avoid multicollinearity in model building.The performance of candidate prediction models and other grading systems was evaluated and compared by constructing subject working curves,calibration curves,Hosmer–Lemeshow goodness-of-fit test,and Akaike Information Standard,and the best model was selected.Finally,R language is used to draw the Norman diagram of the best model,and the decision curve analysis of the model is carried out to evaluate the clinical application value of the best model.According to the best model,hepatocellular carcinoma patients who received the first transcatheter arterial chemoembolization treatment were grouped into risk groups,and trend test analysis was performed on the risk groups.Results:1.General data statistical results:the ineffective rate of the first transcatheter arterial chemoembolization in 427 liver cancer patients was 45.4%.A total of 427patients were randomly divided into training(n=350)and validation(n=70).Except for K~+,there were no statistically significant differences in demographics,objective imaging features,and hepatocellular carcinoma staging between the training and validation groups(P≥0.05).2.Screening of predictors:albumin,maximum tumor diameter,hepatic arteriovenous fistula and tumor envelope integrity were statistically significant through least absolute shrinkage and selection operator regression,univariate and multivariate logistics regression analysis(P<0.05).In univariate logistics regression analysis,direct bilirubin showed critical statistical significance(P=0.06).Therefore,the above five variables become candidates for the new predictive model.3.Construction of candidate prediction model:Based on the results of multivariate logistic regression analysis,albumin,maximum tumor diameter,hepatic arteriovenous fistula and tumor envelope integrity were used to establish the model 2.And try to add direct bilirubin as one of the predictors to build model 1.For ease of calculation,we set albumin and direct bilirubin in model 1 as categorical variables to build model 3.Generalized variance inflation factor and correlation analysis showed that the correlation between the predictors of each model was weak.4.Evaluation and comparison of the performance of candidate prediction models and other grading systems:The Receiver Operating Characteristic Curve shows that the area under the receiver operating characteristic of these candidate models is greater than 0.7 in the training group and the validation group,while the area under the receiver operating characteristic of Barcelona Clinic Liver Cancer stage,Child-Pugh grade and Albumin-bilirubin score is 0.5-0.7.The calibration curve shows that in the verification group,the calibration slopes of Barcelona Clinic Liver Cancer stage,Child-Pugh grade and Albumin-bilirubin score are 1.959,-2.296 and0.008,respectively,while the calibration slopes of model 1,model 2 and model 3 are1.210,1.175 and 1.176,respectively.These results show that the candidate predictive models have better discrimination and calibration ability than other grading systems.In addition,the conversion of continuous variables albumin and direct bilirubin to binary variables may result in data loss,but model 3’s discrimination and calibration ability are no worse than those of model 1 and model 2.The Akaike Information Criterion values of model 1,model 2,model 3,Barcelona Clinic Liver Cancer stage,Child-Pugh grade and Albumin-bilirubin score were 419.87,419.31,419.62,458.70,468.80 and 461.09,respectively,and the p-values of HL test>were 0.05.Since model3 performed well and was simpler to implement than models 1 and 2,we chose this model as the final model and named model 3 as T(tumor size)A(Hepatic arteriovenous fistula)C(integrity of tumor capsules)F(liver function)model.5.Model visualization and clinical application:Use R language to draw the Norman diagram of the TACF scoring model.Decision curve analysis.results show that the TACF scoring model achieves a larger net benefit than other grading systems.According to the TACF model,patients were divided into low-risk(score<145),intermediate-risk(score 145-194),and high-risk(score>194)using isometric assignments.The transcatheter arterial chemoembolization ineffective rates were26.3%,40.2%and 76.8%,respectively.The trend test analysis of risk groups shows that there are significant differences between groups.Conclusions:1.The TACF prediction model integrates 3 objective imaging features and 2liver function indicators.The model has good discrimination and calibration ability.2.The TACF prediction model is a risk stratification management tool,which divides hepatocellular carcinoma patients who undergo transcatheter arterial chemoembolization for the first time into low-risk group,intermediate-risk group and high-risk group,which can accurately identify hepatocellular carcinoma patients who do not respond well to the first transcatheter arterial chemoembolization treatment,and can be used simply and accurately to guide clinical decision-making.