Prognostic factors in cirrhosis identified by Cox's regression model

Abstract

In a controlled clinical trial in 488 patients with chronic liver disease treated with prednisone or placebo, survival data were analyzed using Cox's proportional hazards model. A total of 162 variables were screened separately for prognostic and/or therapeutic effect by log-rank analyses, whereby 46 clinical, biochemical, serological, and histological variables were isolated. Another five variables traditionally found to be important in patients with liver disease were included. After extensively checking the assumptions of the model, the 51 variables were, by a step-wise procedure, reduced to a final model. It comprised, besides a treatment indicator, 12 variables with significant (p less than 0.05) prognostic or therapeutic effect. The following eight variables had a significant prognostic effect: sex, age, prothrombin, acetylcholinesterase, eosinophil leucocytes in liver parenchyma, liver cell necrosis, inflammation in liver connective tissue, and efferent veins in parenchymal nodules. A prognostic index based on the final model is formed allowing calculation of 5 years survival probability. The usefulness of the prognostic index was tested by a cross-validation method, and no statistical significant difference was found between the estimated and observed survivorship functions.