Non-invasive predictive model for hepatic venous pressure gradient based on a 3-dimensional computed tomography volume rendering technology

摘要

Portal hypertension secondary to liver cirrhosis may cause a number of life-threatening complications. The rupture of gastroesophageal varices is associated with a high mortality rate of 15-30%. Hepatic venous pressure gradient (HVPG) is an accurate reflection of disease severity, however this can only be assessed via an invasive interventional procedure. The aim of the present study was to explore a non-invasive method based on 3D computed tomography (CT) volume rendering technology to accurately predict HVPG. A total of 77 patients diagnosed with liver cirrhosis underwent HVPG examination in the present study and the appropriate clinical and radiological data were retrospectively reviewed. A 3D liver and spleen volume rendering was constructed for volume measurements. All non-invasive parameters were tested using univariate analysis and the resulting variables that were statistically significant (P0.20) were used in the multivariate linear regression model. The HVPG predictive model was as follows: HVPG = 18.726 - 0.324 (albumin) + 1.57 (aminotransferase-to-platelet ratio index) + 0.004 (liver volume) (multivariate regression analysis, P=0.006). The corresponding area under receiver operating characteristic curve to identify clinically significant portal hypertension defined as HVPG 10 mmHg was 0.810 (95% confidence interval; 0.705-0.891), with an optimal cut-off value of 12.84, yielding a sensitivity of 80.36% a specificity of 76.19%. The results of the present study indicate that 3D CT volume rendering technology may have the potential to be used for non-invasive prediction of HVPG.