| Background and Objective: Portal vein thrombosis(PVT)is one of the common complications in patients with cirrhosis.When patients with cirrhosis suffer from PVT,venous blood flow resistance increases,portal vein pressure increases,esophageal and gastric fundus varices are more likely to bleed,aggravate liver function injury,and seriously threaten patients’ lives.Therefore,more attention should be paid to PVT patients with esophagogastric fundus varices.In addition,the treatment of PVT in cirrhosis is an important clinical problem,including anticoagulation and interventional thrombolysis.At present,there is no conclusion on the start and end time of treatment at home and abroad.Therefore,it is particularly important to predict the therapeutic effect of PVT patients with cirrhosis.In view of the above problems,this study aims to predict the therapeutic efficacy of patients with cirrhosis of PVT and the bleeding risk when complicated with esophagogastric varices through machine learning and other methods,so as to provide new ideas for the study of PVT and lay a research foundation for this field.Methods: From January 2016 to January 2022,461 PVT patients with cirrhosis in the First Hospital of Lanzhou University were incorporated.After excluding 128 patients,333 patients were included for the retrospective analysis of PVT bleeding in liver cirrhosis.After exclusion of 235 patients,226 patients were included for the prediction of treatment efficacy.To predict the Esophagogastric variceal bleeding(EGVB)in PVT patients with cirrhosis,patients were divided into esophagogastric variceal group and non-esophagogastric variceal group,and the esophagogastric variceal group was further divided into rupture bleeding group and non-bleeding group.Through univariate analysis,LASSO regression,logistic regression and nomogram construction,baseline,blood routine,biochemical and coagulation indexes of each group were analyzed.C-index,Area under the curve(AUC),calibration chart,and decision curve were used to evaluate the predictive efficacy and clinical benefit of the model.To predict the therapeutic effect of PVT in liver cirrhosis,patients were divided into Transjugular intrahepatic portasystemic shunt(TIPS)group,low molecular weight heparin(LMWH)group,low molecular weight heparin sequential rivaroxaban group and no treatment group.The baseline,blood routine,biochemical indexes,coagulation indexes,and imaging examinations of each group were analyzed by univariate analysis to establish a prediction model for the treatment efficacy of PVT patients with cirrhosis.Use accuracy,sensitivity,and specificity to evaluate the predictive efficacy of the model.Results:1.Prediction of cirrhotic PVT with EGVB:(1)Compared with the non-intravenous veins,the red blood cells,hematocrit,hemoglobin,alkaline phosphatase,globulin,blood lipids,creatine kinase,lactate dehydrogenase,calcium,fibrinogen,antithrombin and carcinoembryonic antigen are lower,albumin/globulin,urea,sodium,and osmotic pressure are higher.The differences are statistically significant(P<0.05).(2)Other causes,decreased red blood cells,hemoglobin,hematocrit,globulin,high-density lipoprotein cholesterol,creatine kinase,calcium,and increased urea were more prone to bleeding compared with the non-bleeding group.The differences were statistically significant(P<0.05).Multivariate logistic regression analysis revealed that etiology(viral hepatitis B vs other OR: 4.917,95% CI: 1.677-14.419),decreased hematocrit(OR:0.892,95% CI: 0.846-0.941),decreased high-density lipoprotein cholesterol(OR:0.198,95%CI: 0.058-0.669)were independent risk factors for EGVB in PVT patients with cirrhosis,and a nomogram was constructed to predict the bleeding feasibility of patients,showing good consistency and discrimination(AUC=0.835,95%CI: 0.770-0.835).The 10-fold cross validation(C-index=0.827)and Hosmer-Lemeshow test(P=0.215)suggested that the model had high accuracy and good fit,the calibration curve suggested that it had good stability,and the decision curve suggested that it had clinical practicability.2.Prediction of treatment response in cirrhotic patients with PVT:(1)TIPS treatment of PVT: After 12 weeks of treatment,compared with the ineffective treatment group,the red blood cells,hemoglobin,hematocrit,absolute eosinophil count,creatine kinase isoenzyme,low-density lipoprotein cholesterol in the recanalization group were lower,and α-L-fucosidase was higher.After 24 weeks of treatment,direct bilirubin in the recanalization group was lower than that in the ineffective group,and the difference was statistically significant(P<0.05).(2)In the treatment of PVT with low molecular weight heparin: after 12 weeks of treatment,thrombosis was more likely to be recanalized within 3 months after diagnosis of PVT.Compared with the ineffective treatment group,the absolute value of lymphocytes in the recanalization group was higher.After 24 weeks of treatment,compared with the ineffective treatment group,the C-reactive protein and creatine kinase isoenzyme in the recanalization group were lower,and the percentage and absolute value of lymphocytes were higher,and the difference was statistically significant(P<0.05).(3)Low molecular weight heparin sequential rivaroxaban treatment of PVT: after 12 weeks of treatment,compared with the treatment ineffective group,the percentage of lymphocytes and total bile acids in the recanalization group were higher,and the percentage of neutrophils,red blood cell distribution width variation,fibrinogen content,systemic immune inflammation index,and neutrophil-to-lymphocyte ratio were lower.After 24 weeks of treatment,the percentage of neutrophils,the variation of red blood cell distribution width,and the neutrophil-to-lymphocyte ratio in the recanalization group were lower than those in the treatment ineffective group,and the percentage of lymphocytes,total bile acid,creatinine,and uric acid were higher than those in the treatment ineffective group.The differences were statistically significant(P<0.05).(4)Untreated PVT: after 12 weeks,PVT patients with other causes were more likely to be recanalized.Compared with the progressive group,the glucose,creatine kinase,partial thromboplastin time,plasma prothrombin ratio and potassium were lower and phosphorus was higher in the recanalization group.Compared with the partial recanalization group,the mean platelet volume was lower and D-dimer was higher in the complete recanalization group.After24 weeks,total bile acid was higher in the unchanged group,urea/creatinine was the highest in the partial recanalization group,alkaline phosphatase,calcium,α-Lfucosidase,and anion gap were higher in the complete recanalization group,and homocysteine was the highest in the progression group.The differences were statistically significant(P<0.05).(5)Based on the above indicators,the Bp neural network optimized by genetic algorithm was used to establish prediction models.The prediction effect was good at 12 weeks(accuracy 83.33%-95.83%),but the prediction effect after 24 weeks was mediocre(accuracy 61.54%-85.71%).Conclusions: 1.The independent risk factors of EGVB in cirrhotic PVT patients were other causes(including alcoholic,drug-induced and unexplained cirrhosis),hematocrit and high-density lipoprotein cholesterol.The established prediction model had good predictive value.It can provide a reference for medical staff to screen patients with high bleeding risk and treat them accordingly.2.Based on the general data and serological data of patients,the model was established to predict the changes of thrombosis in liver cirrhosis patients with PVT after 12 weeks and 24 weeks without treatment,and the overall prediction effect of the model was good.3.Six prediction models of different treatment efficacy of liver cirrhosis PVT were established based on the general data and serological data of patients,including two prediction models using TIPS,two prediction models using low molecular weight heparin,and two prediction models using low molecular weight heparin sequential rivaroxaban.The therapeutic efficacy of patients after 12 weeks was better than that after 24 weeks.It can help medical staff to choose the treatment course and treatment methods for patients with liver cirrhosis and PVT. |
PDF Full Text Request