Comparison of hepatic venous pressure gradient and two models of end-stage liver disease for predicting the survival in patients with decompensated liver cirrhosis

Goals: We evaluated the efficacy of initial and follow-up hepatic venous pressure gradient (HVPG), models of end-stage liver disease (MELD), and MELD-Na for predicting the survival of patients with decompensated liver cirrhosis (LC). Background: MELD with/without Na score and HVPG have been important predictors of mortality in patients with LC. Study: Between January 2006 and 2011, a total of 57 patients with decompensated LC, all of whom underwent >2 HVPG measurements for the confirmation of propranolol dosing, were enrolled. MELD and MELD-Na scores were calculated on the day of HVPG measurement. The prognostic accuracy of the initial and follow-up HVPG, MELD, and MELD-Na were analyzed, and independent factors for mortality were evaluated. Results: Ten patients (17.5%) died from LC. Initial HVPG (0.883), initial MELD-Na (0.877), follow-up HVPG (0.829), and follow-up MELD-Na (0.802) showed good area under the receiver operating characteristic curve scores in predicting 1-year mortality. In predicting 2-year mortality, only follow-up HVPG (0.821, cut-off value 18 mm Hg) showed good score. Overall area under the receiver operating characteristic curves (initial and follow-up) were 0.843 and 0.864 in HVPG, 0.721 and 0.674 in MELD, and 0.762 and 0.715 in MELD-Na, respectively. In the Cox regression analysis, only follow-up HVPG (P=0.02; odds ratio, 1.11) was associated with mortality. Conclusions: The efficacy of HVPG for predicting mortality is excellent compared with that of MELD or MELD-Na. Therefore, aside from the confirmation of adequate propranolol dosing, HVPG may be needed for predicting the survival of patients with decompensated LC.