We agree with the recent paper of Infante-Rivard, Esnaola and Villeneuve, who recommend that the Pugh score should be used as a standard of reference when evaluating new prognostic tests in patients with cirrhosis (1). We have previously published two studies identifying the total serum bilirubin/y -glutamyl transpeptidase ratio (2) and serum fibronectin (3) as independent prognostic tests in patients with cirrhosis. The y-glutamyl transpeptidase did not have an independent value when serum fibronectin was entered in a multidimensional prognostic analysis but fibronectin and age had independent value, even after including the variables of the Child-Pugh classification (3). We would like to demonstrate that serum fibronectin and age had independent prognostic values in comparison to Pugh score as used by Infante-Rivard, Esnaola and Villeneuve.
For this purpose we have reanalyzed a total of 114 cirrhotic patients who have been prospectively studied and described in detail elsewhere (3). We used the Cox model to estimate the prognostic value at 1 year of the Pugh score exactly as defined in Model 2 of Infante-Rivard, Esnaola and Villeneuve (1). Then, we used this Pugh score together with fibronectin in a second model and together with fibronectin and age in a third model. Results showed (Table ) that serum fibronectin and age added significant prognostic value to Pugh score alone (p < 0.0001). The best prognostic index was obtained by fibronectin plus age with Pugh score. Figure shows that the receiver operating characteristic (ROC) curve of fibronectin plus age with Pugh score, and the ROC curve of fibronectin with Pugh score, were better than the ROC curve of Pugh score alone. In our population, these "statistical" differences are of practical value as for the same true positive rate of 80%, the false positive rate is 30% in the model with fibronectin, age and Pugh score in comparison to a false positive rate of 60% for the model with Pugh score alone (Figure ). Thus, we think that serum fibronectin and age in this group of alcoholic cirrhotic patients added significant prognostic value to Pugh score.