Prognosis in Patients With Cirrhosis using MELD

Clinical Biochemical Parameters and MELD Score

The MELD score was calculated according to the original formula. The MELD scores of all the patients were over 20. The patients were divided into three groups according to the MELD scores: 20-29, 30-39, and ≥40. The clinical and biochemical characteristics and MELD scores of 300 patients with AOCH are presented in Table 1 .

Relationship Between MELD Scores and 3-month Mortality

During the 3-month follow up, 223 patients died. The causes of death were all related to liver disease: 49 patients died of hepatorenal syndrome; 117 patients died of hepatic encephlophy; 44 patients died of variceal bleeding; 13 patients died of sepsis because of liver dysfunction.

Five patients bridged to liver transplantation, of whom one died during the 3-month follow-up because of liver failure. Thirty patients abandoned treatment, 13 of them due to lack of money, and 17 due to deterioration.

The overall 3-month mortality of the patients with AOCH was 74.3% ( Table 1 ). In the group of MELD score 20-29, 51 patients died (mortality: 56.0%); in the group of MELD score 30-39, 117 patients died (mortality: 76.5%), and in the group of MELD score over 40, 55 patients died (mortality: 98.2%).

Validity of MELD Score for 3-month Survival Using ROC Curve

To compare the accuracy of the MELD score as a predictor of survival with the 3-month mortality of AOCH patients, the concordance c was calculated (Fig. 1). The concordance (c) statistic was 0.782, SE = 0.028, P < 0.0001, 95% confidence interval (CI): 0.727-0.837. The best cutoff value of MELD score was 26, with a sensitivity of 92.3% and specificity of 69.9%.

The prognosis in patients with AOCH who had a MELD score over 26 was poor using the MELD scoring system; 80.7% (192/238) of patients died, whereas of those with a MELD score under 26, 50.0% (31/62) of patients died.

Validity of the models was measured by the receiver-operator characteristic (ROC) curve to predict the 3-month mortality of AOCH patients with acute-on-chronic hepatitis (AOCH). The concordance (c) statistic (equivalent to the area under the ROC curve) was 0.782, SE = 0.028, P < 0.0001, 95% confidence interval: 0.727-0.837.

The best cutoff value of MELD score was 26, with sensitivity of 92.3% and specificity of 69.9%.

Factors Influencing the Prediction Ability of MELD Scoring System

Univariate prognostic analysis of the 3-month mortality of patients with MELD score 20-29. In univariate analysis, we analyzed the prognostic value of clinical and biological variables in 3-month mortality of patients with MELD score 20-29 ( Table 2 ).

Univariate analysis showed that the following factors had significant prognostic value: age (P = 0.047), etiology (P = 0.039), serum sodium (P = 0.029) and ascites (P = 0.031). The mortality of patients with MELD score 20-29 did not seem to be associated with sex (P = 0.973).

Univariate prognostic analysis of the 3-month mortality of patients with MELD score 30-39. In univariate analysis, we analyzed the prognostic value of clinical and biological variables in 3-month mortality of patients with MELD score 30-39 ( Table 3 ).

Univariate analysis showed that the following factors had significant prognostic value: age (P = 0.046), and etiology (P = 0.042). The mortality of patients with MELD score 30-39 did not seem to be associated with sex (P = 0.897), serum sodium (P = 0.236) or ascites (P = 0.396).

Univariate prognostic analysis of the 3-month mortality of patients with MELD score ≥40. In univariate analysis, we analyzed the prognostic value of clinical and biological variables in 3-month mortality of patients with MELD score ≥40 ( Table 4 ).

Univariate analysis showed that the mortality of patients with MELD score ≥40 did not seem to be associated with sex (P = 0.846), age (P = 0.743), etiology (P = 0.851), serum sodium (P = 0.901) or ascites (P = 0.517).

Prognostic factors associated with 3-month mortality in multivariate analysis. In multivariate analyses to identify predictors of 3-month mortality ( Table 5 ), the significant factors (sex, age, etiology, serum sodium, ascites, MELD score) were entered in multivariate logistic regression model.

When the remaining predictors significant at P < 0.05 were subjected to binary logistic regression, only age (P = 0.013), etiology (P = 0.009), serum sodium (P = 0.026), ascites (P = 0.025), and MELD score (P = 0.025) were independent predictors of 3-month mortality in all patients ( Table 6 ).

Of these five factors, only MELD score (P = 0.015) was independent predictor of 3-month mortality in patients with MELD score ≥30. Conversely, multivariate analysis of these five factors in patients with MELD scores 20-29 found that age (P = 0.012), etiology (P = 0.024), serum sodium (P = 0.005) and ascites (P = 0.017) were independent predictors of 3-month mortality; MELD score was not an independent predictor of mortality in patients with MELD score 20-29 once other factors were taken into account.