Prognostic value of C-reactive protein levels in patients with cirrhosis.

I dentifying cirrhosis with a poor short-term prognosis remains crucial for improving the allocation of liver grafts. The purpose of this study was to assess the prognostic value of a model combining the variation of C-reactive protein (CRP) levels within 15 days, the Model for End-Stage Liver Disease (MELD) score, and the presence of comorbidities in patients with decompensated cirrhosis with a Child-Pugh score > B7 and to test the relevance of this model in patients with compensated cirrhosis. We collected data for cirrhotic patients without hepatocellular carcinoma, extrahepatic malignancy, human immunodeficiency virus infection, organ transplantation, seen between January 2010 and December 2011. Multivariate analyses of predictors of 3-month mortality used Cox models adjusted with the age-adjusted Charlson comorbidity index. The prognostic performance [area under receiver operating characteristic curves (AUROCs)] of the 3-variable model was compared to that of the MELD score. The 241 patients who met the inclusion criteria included 109 patients with a Child-Pugh score > B7 who were hospitalized for decompensation. In these patients with severe cases, the 3-month mortality was independently predicted by the MELD score [hazard ratio (HR), 1.10; 95% confidence interval (CI), 1.05-1.14; P < 0.001] and a CRP level > 32 mg/L at the baseline and on day 15 (HR, 2.21; 95% CI, 1.03-4.76; P = 0.042). This model was better than MELD alone (AUROC, 0.789 versus 0.734; P = 0.043). In the whole population with cirrhosis, the 3-month mortality was also predicted by high MELD scores (HR, 1.11; 95% CI, 1.07-1.16; P < 0.001) and a CRP level > 10 mg/L at the baseline and on day 15 (HR, 2.89; 95% CI, 1.29-6.48; P < 0.001), but the AUROCs of the 3-variable model and the MELD score alone were no longer significantly different (0.89 versus 0.88, not significant). In conclusion, prognostic models incorporating variations in CRP predict 3-month mortality in patients with cirrhosis. Such models are particularly relevant for patients with decompensated cirrhosis but provide a limited increase in prediction in comparison with the MELD score in the whole population with cirrhosis.