ABSTRACT
AIM: To build a predictive model for PVT in cirrhotic patients. MATERIALS AND METHODS: A single centre case-control study was carried out. From the database of 1512 cirrhotic patients 94 with newly diagnosed PVT based on contrast-enhanced computed tomography were referred to the Case group. Malignant PVT was an exclusion criterion. Patients without PVT were stratified and matched according to sex, age and etiology of cirrhosis; case-control ratio was 1 : 3-4. The prevalence of PVT in the database, clinical, laboratory, instrumental parameters of the groups were evaluated. Logistic regression model was used to estimate association between variables and PVT. RESULTS: The overall prevalence of PVT was 6.2% with the highest rates among the patients with HBV infection - 16.7%, nonalcoholic steatohepatitis - 15.6%, alcohol abuse in combination with HCV infection - 11.7%. The best predictive model included variables: Child-Pugh classes B-C (coefficient of regression ß=1.853, Ñ=0.001), ascites (ß=0.460, Ñ=0.003), hepatocellular carcinoma without vascular invasion (ß=2.126, Ñ=0.0001), endoscopic band ligation (ß=0.774, Ñ=0.003), azygoportal disconnection (ß=2.734, Ñ=0.001), portal hypertensive gastropathy (ß=0.793, Ñ=0.017), portal vein diameter (ß=0.203, Ñ=0.004), and local factors - ulcerative colitis flare, Clostridium difficile enterocolitis, spontaneous bacterial peritonitis, colorectal cancer, splenectomy, cholecystectomy (ß=2.075, Ñ=0.017). The model had accuracy 85.8% (95% CI 81.7-89.4%), sensitivity - 55.1% (95% CI 43.4-66.4%), specificity - 95% (95% CI 91.6-97.3%), and AUC - 0.871 (95% CI 0.826-0.916). CONCLUSION: Child-Pugh classes B-C, severe portal hypertension, hepatocellular carcinoma without vascular invasion, and local factors were estimated as risk factors of PVT in cirrhotic patients.
