Upper gastrointestinal bleeding is a leading cause of death in patients with liver cirrhosis. In most cases haemorrhage originates from oesophageal varices or from congestive gastropathy, and the evaluation of the bleeding risk is based on oesophagogastroduodenoscopic data. The aim of this prospective study was to determine whether the measurement of portal flow velocity by Duplex-Doppler, compared with endoscopic data, can help in detecting patients with cirrhosis at risk of bleeding. One hundred and seventy-three patients underwent endoscopy to ascertain the size of the varices and the severity of congestive gastropathy. For each patient maximal portal flow velocity measurements were obtained. No difference in portal flow velocity was observed between patients with or without oesophageal varices or congestive gastropathy. During a 2-year observation period, 27 patients (15.6%) had at least one episode of acute digestive bleeding. Stepwise multiple logistic regression analysis demonstrated a correlation between oesophageal varices and congestive gastropathy endoscopic grading and the incidence of bleeding; only the former was entered into the final regression equation (p < 0.001). No relationship between the max portal flow velocity value and incidence of bleeding was found. This study shows that portal flow velocity is unrelated to the degree of the endoscopic abnormalities in patients with liver cirrhosis and that it has no value in the identification of patients with cirrhosis at risk of upper gastrointestinal bleeding.