Correspondence

Gastrointestinal Bleeding in Critically Ill Patients

N Engl J Med 1994; 331:51-53July 7, 1994

Article

To the Editor:

In a well-designed study Cook et al. (Feb. 10 issue)1 purport to identify risk factors for gastrointestinal bleeding in critically ill patients. “Critically ill patients” is not rigorously defined, but one may infer from the study design that they included patients admitted to a medical-surgical intensive care unit at one of the four study hospitals. However, of the 2828 patients eligible for the study, nearly half (48.5 percent) were admitted because of cardiovascular surgery. Such patients typically have a brief course in the intensive care unit and a low risk of complications. Relatively few patients were admitted because of nonsurgical catastrophes such as sepsis (1.6 percent), cardiovascular disease (6.3 percent), or respiratory disease (12.1 percent). The latter patients, who commonly account for the majority of patients admitted to medical intensive care units, may have a prolonged course and a high risk of complications. In addition, the 113 patients (3.5 percent of the initial population) who were excluded because of a “hopeless prognosis” may have had a relatively high risk of bleeding. Thus, although the study successfully identifies risk factors for gastrointestinal bleeding in the study population, the results may not be readily generalizable to patients who are critically ill with nonsurgical conditions. Therefore, the sweeping conclusion that “Few critically ill patients have clinically important gastrointestinal bleeding” is overly broad for the data presented.

Charles L. Bardes, M.D.
Mary E. Charlson, M.D.
Cornell University Medical College, New York, NY 10021

1 References

1. 1

   Cook DJ, Fuller HD, Guyatt GH, et al. Risk factors for gastrointestinal bleeding in critically ill patients. N Engl J Med 1994;330:377-381
   Full Text | Web of Science | Medline

To the Editor:

. . . I disagree on three basic points with Dr. Peterson's editorial1 on the study by Cook et al. First, the study's findings cannot be used to limit prophylactic therapy to patients without respiratory failure or coagulopathy, since the investigators did not study enough patients with sepsis, hypotension, organ failure, head injury, or trauma to permit comment on treatment for these conditions. Second, the administration of H₂-blockers intravenously is not labor intensive, especially when compared with the evaluation of hemorrhage. Third, patients admitted to most American intensive care units today have more severe disease than those admitted years ago. If we want to keep the incidence of hemorrhage from stress ulcers from increasing to more than 8 percent,2-4 we need to provide prophylactic treatment (a bad term, since 25 percent of our patients have blood in their gastric aspirates) to patients with established risk factors.

I agree with Peterson that we need to understand more about this disease. Studies that examine patients without a need for prophylaxis, that exclude from treatment all patients with blood in their gastrointestinal tract, and that do not routinely use endoscopy to correlate hemorrhage with the number of stress ulcers or other lesions will not advance our understanding of this disease.

Louis F. Martin, M.D.
Louisiana State University Medical Center, New Orleans, LA 70112

4 References

1. 1

   Peterson WL. Prevention of upper gastrointestinal bleeding. N Engl J Med 1994;330:428-429
   Full Text | Web of Science | Medline

2. 2

   Martin LF. Supportive management of acute erosive gastritis. In: Fry DE, ed. Multiple system organ failure. St. Louis: Mosby-Year Book, 1992:279-90.
   

3. 3

   Martin LF. Stress ulcers are common after aortic surgery: endoscopic evaluation of prophylactic therapy. Am Surg 1994;60:169-174
   Web of Science | Medline

4. 4

   Martin LF, Booth FVM, Reines HD, et al. Stress ulcers and organ failure in intubated patients in surgical intensive care units. Ann Surg 1992;215:332-337
   CrossRef | Web of Science | Medline

To the Editor:

Cook et al. found the presence of a coagulopathy to be a risk factor for serious hemorrhage. They defined a coagulopathy as a platelet count of less than 50,000 per cubic millimeter, an international normalized ratio greater than 1.5, or a partial-thromboplastin time greater than 2.0 times the control value. The use of heparin or warfarin was not found to be a significant risk factor. These findings seem contradictory, given that the goal of heparin or warfarin therapy is an elevated partial-thromboplastin time or an elevated international normalized ratio.

The difference in risk factors between the group with coagulopathy and the group given heparin or warfarin must be an unexplained or spontaneous elevation in the partial-thromboplastin time or international normalized ratio (independent of hepatic failure, which was not found to be a significant risk factor). This difference in risk factors suggests that measurement of fibrin-split products or other variables associated with spontaneous coagulopathies may be a better marker of a patient's risk of bleeding.

Mark J. Shumate, M.D.
Emory University, Decatur, GA 30033

To the Editor:

In the observational analysis by Cook et al. of risk factors for gastrointestinal bleeding in critically ill patients, the statement concerning withholding prophylaxis in patients at low risk is not supported by their data. The authors conclude that prophylaxis is unnecessary since over 900 patients would require treatment to prevent each episode of bleeding. This argument is flawed because it considers only the number of patients treated to prevent an occurrence, not the magnitude (or costs) of the outcomes prevented. The authors' data document the danger of clinically important bleeding: the mortality was 48.5 percent. The costs, although not stated, were probably high as well. However, the statements that prophylaxis is expensive and potentially dangerous are poorly supported. The studies1,2 cited by the authors are not applicable to withholding therapy from patients at low risk.

Under certain reasonable assumptions, prophylaxis appears cost effective, even if it prevents only 1 episode of bleeding in 1000 patients. For example, with the average wholesale costs at this institution, five days of prophylaxis cost between $12 and $59, depending on the drug and route used (Schlicht JR: personal communication). The average cost of an uncomplicated episode of gastrointestinal bleeding is between $4,000 and $9,659, and that of a complicated episode is between $5,000 and $11,381 (the low value is the average Medicare reimbursement for the respective diagnosis-related group, and the high value is our institution's charge for the same diagnosis-related group). Assuming (from the study of Cook et al.) prophylaxis with 50 percent efficacy, clinically important bleeding in 1 of 1121 patients, a ratio of two episodes of bleeding that are not clinically important for each episode that is, and a mortality rate of 48.5 percent for clinically important bleeding, prophylaxis would cost between $20,575 and $85,729 per life saved, well within the range of standard, “cost-effective” therapies. Moreover, if the bleeding rate increased to just 0.3 percent (well within the 95 percent limits of the study), prophylaxis would actually save money as well as lives.

Mark S. Roberts, M.D.
University of Pittsburgh School of Medicine, Pittsburgh, PA 15261

2 References

1. 1

   Driks MR, Craven DE, Celli BR, et al. Nosocomial pneumonia in intubated patients given sucralfate as compared with antacids or histamine type 2 blockers: the role of gastric colonization. N Engl J Med 1987;317:1376-1382
   Full Text | Web of Science | Medline

2. 2

   Lacroix J, Infante-Rivard C, Gauthier M. Upper-gastrointestinal-bleeding (UGIB) prophylaxis: a decision-making analysis and a cost benefit analysis. Clin Invest Med 1986;9:Suppl:A22-A22 abstract.
   Web of Science

Author/Editor Response

The authors reply:

To the Editor: Drs. Bardes and Charlson and Dr. Martin suggest that the generalizability of our results is limited. Unless one believes in prophylaxis for patients with hopeless prognoses, excluding those who are unlikely to survive 24 hours is prudent. We repeated our analysis after excluding patients admitted with cardiovascular surgery, and found the results unchanged. Of the 312 patients admitted with a medical diagnosis and without respiratory failure or coagulopathy, only 2 (0.6 percent; 95 percent confidence interval, 0.1 to 2.3 percent) had clinically important bleeding, whereas of 405 patients with a medical diagnosis and with either respiratory failure or coagulopathy, 18 (4.4 percent; 95 percent confidence interval, 2.7 to 6.9 percent) had such bleeding. For patients in mixed medical-surgical units, therefore, our results show a very low incidence of clinically important bleeding in all major subgroups. We did not enroll a sufficient number of patients to allow extrapolating our findings to patients in burn and trauma units.

We remind Dr. Shumate that anticoagulation was a risk factor for bleeding in the univariate analysis but was not as strong as coagulopathy, and thus failed to be an independent predictor in the multivariate analysis. This should resolve the apparent contradiction that he raises. It is not surprising that coagulopathy was a more powerful predictor than anticoagulation, since anticoagulation reflected only the intention to achieve a therapeutic goal, whereas coagulopathy reflected thrombocytopenia (platelet count, <50,000 per cubic millimeter), an international normalized ratio above 1.5, or a partial-thromboplastin time more than 2.0 times the control value. The need for mechanical ventilation and respiratory failure proved to be extraordinarily powerful predictors of bleeding, although it is possible that further predictive power could be gained through refinements such as measurements of fibrin-split products.

There is a major problem in Dr. Roberts's analysis, in that he assumes that preventing bleeding will save lives. In conducting a meta-analysis of this question we found that prophylaxis for gastrointestinal bleeding does not reduce mortality.1 Roberts's analysis, if done appropriately, would require expressing the episodes of bleeding prevented and the number of lives saved in terms of a common unit (such as the quality-adjusted life year). In an appropriate analysis, the cost per quality-adjusted life year gained would probably be far greater than the cost per life saved that Roberts has calculated. A rigorous economic analysis is needed to inform readers about the efficiency in different subgroups of prophylaxis for stress ulcer.

Deborah J. Cook, M.D.
Gordon H. Guyatt, M.D.
Lauren Griffith, M.Sc.
McMaster University, Hamilton, ON L8S 4K1, Canada

1 References

1. 1

   Cook DJ, Witt LG, Cook RJ, Guyatt GH. Stress ulcer prophylaxis in the critically ill: a meta-analysis. Am J Med 1991;91:519-527
   CrossRef | Web of Science | Medline

Citing Articles (3)

Citing Articles

1. 1

   Michael J Cawley. (2007) Mechanical Ventilation: A Tutorial for Pharmacists. Pharmacotherapy 27:2, 250-266
   CrossRef

2. 2

   J Hiramoto. (2003) Evidence-based analysis: postoperative gastric bleeding: etiology and prevention. Surgical Oncology 12:1, 9-19
   CrossRef

3. 3

   Brian L. Erstad, James M. Camamo, Michael J. Miller, Anthony M. Webber, John Fortune. (1997) Impacting cost and appropriateness of stress ulcer prophylaxis at a university medical center. Critical Care Medicine 25:10, 1678-1684
   CrossRef