From an initial cohort of 124 patients, 92 underwent baseline LS, SS, HVPG measurements and upper gastrointestinal endoscopy at enrolment and then followed-up for 2 years or until the occurrence of the first CD. Univariate and multivariate logistic regression models were used for determining judgement criteria associated parameters. Accuracy of predictive factors was evaluated using c statistic. The final model was internally validated using the bootstrap method.
During follow-up, 30 out 92 (32.6%) patients developed CD. At univariate analysis varices at enrolment, all non-invasive parameters, HVPG, and model for end-stage liver disease (MELD) resulted clinical predictors of CD. At multivariate analysis only SS (p = 0.0001) and MELD (p = 0.014) resulted as predictive factors. A decision algorithm based on the results of a predictive model was proposed to detect patients with low risk of decompensation.
This study shows that in compensated cirrhotic patients a SS and MELD predictive model represents an accurate predictor of CD with accuracy at least equivalent to that of HVPG. If confirmed by further studies, SS and MELD could represent valid alternatives to HVPG as prognostic indicator of CD in HCV-related cirrhosis.