Prognostic factors of the outcome of upper gastrointestinal bleeding in patients with cirrhosis are insufficiently defined. Pertinent clinical, biochemical, and endoscopic data of 332 upper gastrointestinal bleedings in 268 patients with cirrhosis observed in the participating centers during 31 months were recorded. Clinical data were analyzed until 40 days after bleeding. A further set of 82 bleedings was used as a validation group. Ninety-two of the 268 patients died within the time of the study, and 28 of the 82 patients of the validation group died. According to a stepwise logistic regression analysis, s-creatinine, ascites on admission, previous diagnosis of hepatocellular carcinoma, s-bilirubin, prothrombin index, varices as definite or probable source of bleeding, gender, and presentation with hemathemesis were the best set of covariates for predicting outcome. From them a prognostic index was developed and validated in the 82 further bleedings. Sensitivity and specificity in the cumulated training and test sets were 75 and 80%, respectively. In the present material, the prognostic index was significantly more efficient than Child-Pugh score or the prognostic index proposed by Garden et al. These data show that it is possible to predict the outcome of upper gastrointestinal bleeding in cirrhosis on the basis of few easily available data. The prognostic index we proposed and validated may become useful to predict the outcome of a bleeding and to select or stratify patients in clinical trials.