The efficacy of cimetidine in the prevention and treatment of stress-induced gastroduodenal lesions was evaluated in a randomized, double-blind, placebo-controlled study in which serial endoscopy was used to examine patients without clinical evidence of bleeding who were admitted to a medical intensive care unit. Endoscopy showed that 14 of 21 patients treated with cimetidine, compared with 5 of 18 patients treated with placebo, had normal or improved gastroduodenal mucosa (p less than 0.05). Endoscopic signs of bleeding cleared or did not develop in 20 patients treated with cimetidine and in 11 patients treated with placebo (p less than 0.01). Significantly fewer blood transfusions were given to patients with endoscopic signs of bleeding in the cimetidine-treated group (0.5 +/- 0.3 [SE] units) than in placebo-treated patients (4.5 +/- 1.5 units; p less than 0.05). The mortality rate was not statistically different between treatment groups. By preventing established gastroduodenal stress lesions from progressing in severity, cimetidine diminished both bleeding and the need for transfusions.