Critical illness can disrupt local and systemic mechanisms that protect against upper gastrointestinal bleeding, a condition that may be associated with increased mortality,1 particularly among patients receiving extracorporeal life support.2 On the basis of randomized trials performed over a period of 40 years,3 most guidelines recommend preventive therapy with either histamine H2–receptor antagonists or proton-pump inhibitors (PPIs) for patients in the intensive care unit (ICU) who are at risk for the development of stress ulceration and bleeding.
A review of this approach is timely, owing to a number of considerations. First, contemporary observations suggest that stress-ulcer bleeding may be less prevalent than it was in the past, probably as a result of improved care in the ICU. Second, the overall benefit from PPIs may be reduced by adverse events that are associated with the use of these agents, including nosocomial pneumonia, Clostridium difficile enteritis, and myocardial ischemia.1,4 Third, administration of enteral feeding in parallel with bleeding prophylaxis may reduce the risk of gastrointestinal bleeding, possibly obscuring the benefits of PPIs; conversely, the combination may increase the risk of nosocomial pneumonia.5
Krag and colleagues address the value of prophylaxis for gastrointestinal bleeding in the ICU in their report, now published in the Journal,6 of a multicenter European trial in which 3298 adult patients were randomly assigned to receive daily, single-bolus, intravenous pantoprazole (40 mg) or placebo during their nonelective ICU stay. The patient population was at high risk for clinically important gastrointestinal bleeding because of a history of liver disease, coagulopathy, shock, treatment with anticoagulant agents, renal replacement therapy, or mechanical ventilation that was expected to last for more than 24 hours.
There was no significant difference between the pantoprazole group and the placebo group in the rate of the primary outcome of death by 90 days after randomization (31.1% and 30.4%, respectively). This finding is similar to the mortality rates found in meta-analyses of similar trials.7,8 However, the observed mortality rate was higher in the placebo group of the current trial than in the cohorts that were used to calculate the sample size for this trial.1 It also remains unknown whether there could have been a survival benefit among patients in the ICU who were at especially high risk for death.
Similarly, no difference was found with regard to the secondary, unvalidated composite outcome of clinically important gastrointestinal bleeding, new-onset pneumonia, C. difficile infection, or acute myocardial ischemia (21.9% in the pantoprazole group and 22.6% in the placebo group). This composite outcome was designed to capture the net clinical effect of PPI use, but it is difficult to interpret. It encompassed some outcomes that may improve and others that may worsen in association with the intervention. Furthermore, the individual components of the outcome have disparate pathophysiological mechanisms, different expected incidences, and different risks attributable to treatment.9
Although the trial was not powered to address each component of the combined secondary outcome, thus limiting conclusions, useful new information can be drawn from the findings. First, clinically important gastrointestinal bleeding occurred in 2.5% of the patients in the pantoprazole group and in 4.2% of the patients in the placebo group (no P value was reported, since there was no adjustment for multiple comparisons). The magnitude of this possible benefit of pantoprazole is within the range of differences between PPI treatment and placebo for gastrointestinal bleeding prophylaxis that was shown in a recent meta-analysis,10 although the rate of clinically significant bleeding in the placebo group in this trial was greater than the upper 95% confidence limit of the bleeding rate in a cohort that was analyzed to plan the trial. Moreover, these results highlight the low prevalence of upper gastrointestinal bleeding in modern ICUs. Second, the combined rates of the infectious outcomes of pneumonia and C. difficile infection were essentially identical in the two groups (16.8% in the pantoprazole group and 16.9% in the placebo group). These analyses were not stratified according to whether enteral nutrition was administered, which could have modified the outcomes, but the percentages of patients receiving enteral nutrition were balanced between the two groups, with 56 to 85% of the patients receiving it at some point during their ICU stay.
In our view, the take-home message from this trial is that, given the low incidence of clinically important upper gastrointestinal bleeding in the ICU, prophylaxis with a PPI, if initiated, should be reserved for seriously ill patients who are at high risk for this complication. (An important problem, of course, is that the definition of a high risk of bleeding varies between guidelines and needs to be standardized.3) Although 90-day mortality was not affected by pantoprazole administration in the current trial, the between-group difference in the rate of important upper gastrointestinal bleeding may still support the recommendation of using a prophylactic PPI, particularly given the absence of a difference in the rate of adverse events between the pantoprazole group and the placebo group. We base this view mainly on the admittedly small (1.7-percentage-point) difference in bleeding rates between the groups. Additional data are needed to determine the clinical effects of prophylaxis for gastrointestinal bleeding in the ICU, especially in groups of patients who are at very high risk for this complication, and to quantify any protective or harmful effects attributable to the coadministration of enteral nutrition.