Background/aims: Prognostic studies on cirrhosis are needed, since several attempts to obtain better survival predictors than the empirically derived Child-Pugh's score have failed.
Methods: Four hundred and ninety-four consecutive in-patients with cirrhosis at the first episode of decompensation (ascites, jaundice, encephalopathy) and/or of digestive haemorrhage from ruptured oesophageal varices were followed from admission (1983-1989) to 1993, studying the relationship between 26 prognostic variables and survival. Three prognostic models were constructed using Cox's regression model and the Receiver Operating Characteristic (ROC) analysis was used to compare their predictive ability.
Results: During follow-up 351 patients died (median cumulative survival 1.82 years). Child-Pugh's score (explicative variable of the first Cox's model), albumin and encephalopathy among the 5 Child-Pugh variables (second model), and oesophageal varices haemorrhage and 3 biochemical indexes among the 7 significant variables on univariate analysis (third model) correlated with survival. The area under the ROC curve of the first model did not significantly differ from that for the other 2 models.
Conclusions: The Child-Pugh's score is still the best and simplest index for assessing the prognosis of liver cirrhosis.