Current issues of ACP Journal Club are published in Annals of Internal Medicine


Stress ulcer prophylaxis in critically ill patients: a meta-analysis

ACP J Club. 1992 July-Aug;117:15. doi:10.7326/ACPJC-1992-117-1-015

Source Citation

Cook DJ, Witt LG, Cook RJ, Guyatt GH. Stress ulcer prophylaxis in the critically ill: a meta-analysis. Am J Med. 1991 Nov;91:519-27.



To investigate the effect of drugs for prevention of stress ulceration on overt bleeding and mortality in critically ill patients.

Data sources

Citations to 1990 were retrieved from MEDLINE and EMBASE databases using the key and text words hemorrhage (gastrointestinal), critical care, and clinical trials. Further trials were sought from the bibliographies of relevant articles and from pharmaceutical companies.

Study selection

42 randomized clinical trials that compared prophylactic regimens for overt and clinically important bleeding in the critically ill were selected from 168 articles.

Data extraction

Data on study validity, patient characteristics, drug regimens, and outcomes were independently evaluated. Overt bleeding was defined as hematemesis, bloody gastric aspirate, melena, or hematochezia. Clinically important bleeding was defined as overt bleeding associated with hemodynamic changes or the need for transfusion.

Main results

A total of 4409 patients were evaluated in this overview. Compared with placebo, the incidence of overt bleeding was reduced by antacids (common odds ratio [COR], 0.40; 95% CI, 0.20 to 0.79) and by histamine-2-receptor (H2) antagonists (COR, 0.29; CI, 0.17 to 0.45). Antacids were as effective in reducing overt bleeding as sucralfate (COR, 1.19; CI, 0.62 to 2.29) and prostaglandins (COR, 0.99). H2-antagonists reduced overt bleeding events more than antacids (COR, 0.56; CI, 0.33 to 0.97) but not more than pirenzepine (COR, 3.78; CI, 1.33 to 14.45) or sucralfate (COR, 1.77; CI, 0.78 to 3.96). Compared with placebo, H2-antagonists reduced the incidence of clinically important bleeding (COR, 0.35; CI, 0.15 to 0.76) as did antacids (COR, 0.35; CI, 0.08 to 1.33; not significant). CIs for CORs all included unity in comparisons of the incidence of clinically important bleeding between antacids and sucralfate (COR, 0.65) or prostaglandins (COR, 1.04), or when H2-antagonists were compared with antacids (COR, 0.84) or with sucralfate (COR, 0.95). Mortality did not differ significantly in any comparison of medications or placebo.


The incidence of overt gastrointestinal bleeding in critically ill patients is reduced by pirenzepine, histamine-2-receptor antagonists, or antacids. Histamine-2-receptor antagonists also reduce the incidence of clinically important bleeding. Stress ulcer prophylaxis does not affect mortality.

Source of funding: No external funding.

Address for article reprint: Dr. D.J. Cook, Department of Medicine, Division of Critical Care, St. Joseph's Hospital, 50 Charlton Avenue East, Hamilton, Ontario L8N 4A6.


The meta-analysis of stress ulcer prophylaxis by Cook and colleagues compared with that by Tryba (1), abstracted in the January/February 1992 issue of ACP Journal Club, raises several issues. Both articles conclude that antacids and H2-antagonists significantly reduce overt bleeding when compared with no prophylaxis, reiterating the conclusions of earlier meta-analyses by Lacroix and colleagues (2) and Shuman and colleagues (3). Newer agents (sucralfate, pirenzepine, and the prostaglandins) were also evaluated by both Cook and colleagues and by Tryba. From their analyses of published data, they disagree about the efficacy of sucralfate and pirenzepine, and their differences may be of substantial clinical importance. Cook finds a trend favoring antacids over sucralfate for overt bleeding, and Tryba does not. Cook finds a trend that favors sucralfate over antacids and H2-antagonists for lower mortality, and Tryba reports that sucralfate significantly reduces mortality. If the latter is true, should not all patients receive sucralfate?

Meta-analyses can disagree for many possible reasons that are related to the quality of each effort. The control of biases in the selection of papers and extraction of data is most important (4). The databases of Cook and colleagues and of Tryba are not identical. Further, neither article presents all of the results from each study in ways that allow the reader to decide which is correct. They should resolve the issue forthwith by comparing their data and reporting the results.

Thomas C. Chalmers, MD
Harvard School of Public Health Boston, Massachusetts