Background/aims: The variations in C-reactive protein (CRP) levels have been reported to have prognostic significance in decompensated cirrhotic patients. We aimed to provide an external validation of a prognostic model combining model for end-stage liver disease (MELD) and 'sustained high CRP levels' as main variables and to optimize the model to the context of liver transplantation by focusing on 3-month mortality with no consideration of severe chronic extrahepatic diseases.
Patients and methods: Data from cirrhotic patients enrolled in the CANONIC study were collected. Multivariate analyses used the competing risk model. The prognostic performance [area under receiver operating characteristic curve (AUROC)] of the model incorporating CRP variations within 15 days was compared with that of the MELD score alone.
Results: 583 decompensated cirrhotic patients with Child-Pugh more than B7 and serial CRP measures available were included. Of these, 111 patients had baseline CRP at least 29 mg/l and 60 still had CRP at least 29 mg/l at day 15±6 (group A). Multivariate analysis (competing risk) identified three predictors of 3-month mortality: high MELD score [hazard ratio (HR)=1.14; 95% confidence intervals (CI): 1.11-1.17, P<0.001], age (HR=1.04; 95% CI: 1.02-1.06, P<0.001), and group A (HR=1.69; 95% CI: 1.01-2.81, P=0.046). The performance of the three variables taken together for predicting 3-month mortality was 0.796 (AUROC), which was significantly higher than that of the MELD score (AUROC=0.769; P=0.019).
Conclusion: In Child-Pugh higher than B7 cirrhotic patients with decompensation, prognostic models incorporating variations in CRP within 15 days and age predict 3-month mortality better than the MELD score alone. Such models would improve the ranking of candidates for liver transplantation by differentiating the severe patients with persistent systemic inflammation and intermediate MELD scores.