In an attempt to improve the prognostic capacity of Child-Pugh score in nonbleeding cirrhotics, 110 consecutive in-patients without gastrointestinal hemorrhage at admission were studied and followed up for 24 months or until death. Fifty-five of the 110 patients (50%) died during this period. Mean survival time was 18.8 +/- 1.4 months (mean +/- SEM). In addition to Child-Pugh score, eight variables, including anthropometric nutritional parameters, routine renal function tests, and alcoholism markers, were recorded at admission. The ability of these variables to improve the prognostic capacity of the Child-Pugh score was assessed with the proportional hazard Cox's regression procedure, using a stepwise method for covariate selection, after including the Child-Pugh score at the first step. Thus, in addition to Child-Pugh score (beta = 0.302), three variables were included in the final model: serum urea (beta = 0.113), MCV (beta = 0.027), and mid-arm muscle circumference (beta = -0.025). According to the contribution of each of these factors to the model, a prognostic index was obtained to estimate survival in the individual patient. An assessment of the predictive power of the model was made by means of a split-sample technique. The prognostic index described in this study may contribute to improve the selection of nonbleeding patients with advanced cirrhosis to receive specific therapies such as transplantation. However, its true clinical relevance will be established only by prospectively comparing its prognostic value with that of the Child-Pugh score in a new sample of patients.