Background: Two models are mostly used to predict survival in cirrhosis: the Child-Pugh score (CP score) and the model for end-stage liver disease score (MELD score).
Aims: The aim of this study is to evaluate the CP score and the MELD score for short- and long-term prognosis in cirrhosis, as well as CP-creatinine score, MELD-Na score, and UKELD score.
Methods: One thousand and forty-seven patients from five referral centers were included: men/women: 620/427, median age: 58 years (IQR 48-66), median follow-up: 33 months (IQR 12-74), CP (A/B/C): 493/357/147, CP score: 7 (IQR 5-9), MELD score: 12 (IQR 9-16). The performance of each score was evaluated by the Cox hazard model in terms of their: discrimination ability (C-index and Somer's D) and calibration (3, 12 months). Internal validation was done with bootstrapping (100 samples).
Results: Three hundred and fifty-two patients (33.6%) died. All scores were significantly associated with overall mortality, when assessed by univariate Cox analysis. CP-creatinine score performed significantly better than all other scores [bootstrap C-index 0.672, 95% CI 0.642-0.703, bootstrap Somer's D 0.344 (0.285-0.401)], apart from CP score, which showed similar performance. Inclusion in the multivariable Cox model of age together with CP-creatinine score improved the discriminative ability of the model [bootstrap C-index (95% CI) 0.700 (0.661-0.740)]. In terms of calibration, CP-creatinine score was the best for both 3- and 12-month survival in the total population.
Conclusions: CP score and CP-creatinine score have better prognostic value compared to MELD score, MELD-Na score, and UKELD score for predicting short- and long-term mortality in patients with stable cirrhosis.
Keywords: Cirrhosis; Cox proportional hazards models; Liver diseases; Prognosis; Survival.