We retrospectively reviewed 298 charts in order to evaluate the efficiency of a protocol used to prevent gastrointestinal bleeding among ICU patients. The protocol included the use of an antacid (186 patients), iv administration of cimetidine (66 patients), or both drugs when the combination was needed because of a persistently low gastric pH after antacid (28 cases). In 18 cases the implementation of this protocol was stopped when enteral feeding through a nasogastric tube was started. All four groups were homogenous for average age and the presence of risk factors at admission as well as at the time of bleeding. Nevertheless the percentage of gastric bleeding during ICU stay (coffee-ground vomitus haematemesis and/or melaena) widely varied: 5% for the antacid group; 15% in the cimetidine group; 25% in the "both" group and 56% in the "enteral" group. Possible explanations for these differences are discussed. The results support the use of treatment protocols in order to prevent gastrointestinal bleeding in patients with risk factors who are admitted to ICU.