| Background and aims:Variceal hemorrhage is one of the most and severe complications of liver cirrhosis and portal hypertension. Although the effective therapy of varices has improved over several decades, the mortality during variceal hemorrhage episode in patients with portal hypertension remains high. Patients with a hepatic venous pressure gradient (HVPG) greater than 12mmHg (HVPG>12mmHg) have been identified as being at a higher risk for the first hemorrhage episode. In primary prophylaxis, HVPG of up to 12mmHg after chronic treatment decreases the risk for hemorrhage from gastroesophageal varices. However, HVPG measurement for predicting variceal hemorrhage is an invasive method. Therefore, the development of a simple and less-invasive method to predict variceal hemorrhage accurately is necessary for and will be useful in the management of gastroesophageal varices in liver cirrhosis. The current study was designed to find an accurate and objective method to predict HVPG measurement and variceal hemorrhage.Methods:150 hepatitis B patients with liver cirrhosis were enrolled and analyzed retrospectively. The patients were randomly divided into two groups, the experiment group and the validation group. In the experiment group of 75 patients, every patient underwent transjugular HVPG measurement, contrast enhanced computed tomography (CECT) and upper endoscopy. Liver volume, spleen volume, liver/spleen volume ratio, the internal diameters of portal vein and splenic vein, the presence of ascites were measured by CECT. Upper endoscopy was used to classify gastroesophageal varices. Liver function, prothrombin time (PT) and platelet count were tested by laboratory examinations. The Child-Pugh score for each patient was calculated. We assessed the ability of all variables to predict HVPG measurement and variceal hemorrhage. The variables with the highest correlation with the HVPG and the highest ability to predict the HVPG were applied to construct a model to predict HVPG>12mmHg and variceal hemorrhage by multivariable linear regression model. The validation group with 75 patients was used to verify the predictive equation.Results:1. Liver/spleen volume ratio had the best performance on predicting HVPG>12mmHg among all of the continuous variables. In the experiment group, the result of correlation analysis showed that liver volume, spleen volume, liver/spleen volume ratio and platelet count were significantly correlated to the HVPG measurement. The internal diameters of portal and splenic vein, Child-Pugh score and PT were moderately correlated to the HVPG. As for the capacity and performance of continuous variables correlated to HVPG measurement in predicting HVPG>12 mmHg, among the continuous variables, liver volume, spleen volume, liver/spleen volume ratio, internal diameter of portal vein and platelet count had a higher capability than the other continuous variables. In particular, the liver/spleen volume ratio had the best predicting performance.2. Classification of varices and presence of ascites had good performance on predicting HVPG>12mmHg as categorical variables. Mann-Whitney U test of categorical variable, in the experiment group, shown that the presence of ascites was significantly correlated to the transjugular HVPG (z=-3.995, p<0.01). On spearman’s correlation analysis, the classification of varices showed a significant association with transjugular HVPG measurement (0.757, p<0.01). Therefore, classification of varices and the presence of ascites could contribute to predict HVPG measurement.3.Inter-observer agreement of liver and spleen volume was higher than internal diameter of portal vein and spleen vein. The ICC of liver volume, spleen volume, internal diameters of portal vein and splenic vein were 0.987 (p<0.001),0.992 (p<0.001),0.753(p<0.001) and 0.749(p<0.001) respectively. The kappa value of presence of ascites was 0.917(p<0.001). Therefore, The inter-observer agreement of presence of ascites, liver volume and spleen volume were higher than the internal diameters of portal vein and spleen vein.4. Construction of model for predicting HVPG>12mmHg in liver cirrhosis patients. The predictive model combined with liver/spleen volume ratio and classification of varices was constructed for predicting the HVPG measurement and the variceal hemorrhage. The AUC of this predictive equation was 0.919 (95% CI:0.839-0.999, P<0.001). The values of sensitivity, specificity, positive and negative predictive value were 92.9%,87.0%,89.7% and 90.9% respectively. The following equation was: HVPG score=13.651-6.187*ln(liver/spleen volume)+2.755*(classification of varices) [classification of varices:absence-0, small-1, large-2]Conclusions:The model of combining liver/spleen volume ratio and classification of varices can accurately predict the HVPG measurement and variceal hemorrhage in hepatitis B patients with cirrhosis. |
