Uncertain Answers — Proton-Pump Inhibition in the ICU
Samuel M. Brown
N Engl J Med Published June 14, 2024
DOI: 10.1056/NEJMe2405782
A critical care clinician’s job includes administering timely and tailored antibiotics, expediting appropriate surgery, supporting a patient’s beleaguered body and mind until they are ready to recover, and preventing complications in this vulnerable state. This is surprisingly strenuous work, even if it is often, appropriately, protocolized. The aphorism “an ounce of prevention is worth a pound of cure” persists because it feels true. So, when trial data suggest that an ounce of prevention may actually kill you, it cuts hard against the grain — which brings us to the question of gastric acid suppression among patients undergoing mechanical ventilation in the intensive care unit (ICU).
Previous randomized trials of proton-pump inhibitors involving critically ill patients and meta-analyses of these trials have been unable to support confident conclusions regarding an effect of such treatment on mortality. Meanwhile, observational data have suggested the possible association between proton-pump inhibitors and an increased risk of pneumonia and Clostridioides difficile infection. The results of two previous trials, SUP-ICU1 and the cluster-randomized PEPTIC2 (which compared proton-pump inhibitors with placebo and H2-receptor blockade, respectively), suggested that the effects on mortality may differ according to patients’ disease severity — and that the drugs were potentially less safe in more severely ill patients. These observations have continued to vex the critical care community regarding the safety of proton-pump inhibitors in the ICU.
Cook and colleagues3 now report in the Journal the results of the large Reevaluating the Inhibition of Stress Erosions (REVISE) trial of the proton-pump inhibitor pantoprazole to prevent gastrointestinal bleeding among patients undergoing invasive mechanical ventilation. In their evaluation of 4821 patients, the investigators found a small but clear reduction in the risk of clinically important gastrointestinal bleeding at 90 days (the primary efficacy end point) with pantoprazole as compared with placebo. In addition, the findings regarding death at 90 days (the primary safety end point) and the secondary end points of ventilator-associated pneumonia and C. difficile infection were similar in the two groups. Although firm conclusions cannot yet be drawn, the baseline severity of illness did not appear to be associated with a difference in either bleeding or mortality. The REVISE trial was well designed and executed, with generalizable eligibility criteria and excellent experimental separation.
An overlapping group of investigators simultaneously report the results of a rigorous meta-analysis4 that functionally combines data from the REVISE trial with the previous findings of the SUP-ICU trial (in which approximately three quarters of the patients would have met the REVISE criteria). These data confirmed the beneficial effect of pantoprazole on gastrointestinal bleeding, did not support an association with pneumonia or C. difficileinfection, and confirmed the existence of treatment-effect heterogeneity on mortality.5However, the results suggested decreased mortality among less severely ill patients and confidence intervals that included unity for higher mortality among more severely ill patients.
Meta-analyses of low-quality underlying trials are often a feckless exercise in structured nihilism. But large, high-quality trials can be meaningfully summarized through meta-analysis. When such trials are of similar quality and design, meta-analyses can often bear greater inferential weight than the individual trials. In the case of mortality, this meta-analysis meaningfully combines the findings of the SUP-ICU and REVISE trials.
The modest contributions that proton-pump inhibitors may make to reducing mortality are dwarfed by the much more potent causal factors — the critical illness process, previous health status, and decisions to limit treatment — that are at play among patients undergoing mechanical ventilation. The proposed constituent safety events (pneumonia and C. difficileinfection) that may contribute to mortality were rare and did not differ significantly between groups in either the REVISE trial or the meta-analysis. These findings are typical of safety signals, which are difficult to detect in even high-quality trials, given the low event rates and substantial causal noise. However, the lack of a substantial signal is reassuring that the hypothesized mortality difference in more severely ill patients is a chance finding.
So, what do we know from these collective data? Proton-pump inhibitors slightly but significantly decrease the risk of important gastrointestinal bleeding and have a decent chance of slightly decreasing mortality in less severely ill patients during mechanical ventilation. Moreover, we can be certain that proton-pump inhibitors do not decrease — and may slightly increase — mortality in severely ill patients. Given the challenges that are inherent in trials involving severely ill patients in whom no meaningful mortality benefit is likely, this is all we will know for the foreseeable future.
Personally, with no partisan fervor or false certainty, I plan to prescribe prophylactic proton-pump inhibitors to patients during mechanical ventilation if they have an APACHE II score of less than 25 (or a reasonable equivalent, since the prognostic scores that were used in the REVISE and SUP-ICU trials are no longer well calibrated). For sicker patients, I would probably reserve the use of proton-pump inhibitors for those who are being treated with antiplatelet agents, especially in the presence of therapeutic anticoagulants.
