PPIs Signal Mortality Risk in ICU Stress Ulcer Prophylaxis

For ICU patients on mechanical ventilation, the safest type of stress ulcer prophylaxis was left unclear by the international PEPTIC trial.

In-hospital mortality rates trended higher with a strategy giving preference to proton pump inhibitors (PPIs) at 18.3% versus 17.5% with histamine-2 receptor blocker (H2RB)-preferential prophylaxis (P=0.054), whereas clinically-important upper GI bleeding was less common with the PPI strategy (1.3% vs 1.8%, RR 0.73, P=0.009).

These findings, however, might have been affected by the pragmatic, open-label trial's allowance of crossover, leading to 20% of the H2RB group getting PPIs, Paul Young, PhD, of Wellington Hospital in Wellington South, New Zealand, and colleagues reported.

The study was presented at the Critical Care Reviews Meeting 2020 in Belfast, Northern Ireland and simultaneously published online in JAMA.

"Although the trial was powered for a difference of 2.4% in 90-day mortality, the smaller difference of 0.8% would be meaningful, if real, given that hundreds of thousands of critically ill patients are at risk annually. This would be the case even given that fewer patients assigned to receive the proton pump inhibitor strategy experienced clinically important upper gastrointestinal bleeding," argued an accompanying editorial by Todd Rice, MD, of Vanderbilt University Medical Center in Nashville, Tennessee, and colleagues.

While a post hoc analysis by treatment adherence suggested similar findings, "the motivation for the trial was based on safety concerns of proton pump inhibitors, not a hypothesized benefit of histamine-2 receptor blockers. As such, this direction of bias should raise concerns," they added.

Other studies have estimated 2.6% of adults acutely admitted to the ICU developed clinically significant upper GI bleeding and 70% were prescribed stress ulcer prophylaxis to forestall this.

PEPTIC attempted to address the safety questions with 26,982 patients enrolled in an open-label cluster crossover design at 50 ICUs in Australia, Canada, England, Ireland, and New Zealand. The mean age in both treatment groups was just over 58, about 64% of both groups were male, and the burden of comorbidities at baseline was comparable.

All patients required invasive mechanical ventilation within 24 hours of ICU admission and were randomized by site to preferential use of PPIs versus preferential use of H2RBs. Each strategy was used sequentially for 6 months, in any order.

Of patients treated at sites randomized to PPIs, an estimated 4.1% received H2RBs; of those treated at sites randomized to H2RBs, 20.1% received PPIs.

In the primary outcome, the risk ratio for in-hospital death by day 90 was 1.05 for PPIs versus H2RBs (95% CI 1.0-1.10) and an absolute risk difference of 0.93 percentage points (P=0.054).

New-onset Clostridium difficile infection diagnosed in the ICU was reported for 0.30% in the PPI group and 0.43% in the H2RB group (RR 0.74, 95% CI 0.51-1.09, P=0.13).

As for pneumonia, the investigators said the possibility that PPIs may increase mortality attributable to hospital-acquired pneumonia cannot be excluded because pneumonia rates were not measured. They pointed out however, that ICU and hospital lengths of stay and ventilation time did not significantly differ by treatment group, suggesting that such pneumonia episodes may be minor.

One patient in the PPI group had an allergic reaction to omeprazole (Prilosec, Losec, Zegerid) and switched to ranitidine (Zantac).

They editorialists cautioned that the combination of a pragmatic open-label cluster crossover design and incomplete data "confounds interpretation of the results and leaves the clinician unsure of the best way to optimize benefit and avoid harm when deciding on stress ulcer prophylaxis for individual critically ill patients."

The authors also acknowledged several study limitations, including the exclusion of some patients with a diagnosis of upper GI bleeding at ICU admission that might actually have been lower GI bleeding. And, some diagnosed with upper GI bleeding in the ICU may have already been bleeding at admission.

In addition, only data from the index hospitalization were included. Because mortality data were obtained from registries, they may have contained random errors. And since clinicians and research staff were aware of treatment assignments, bias may have affected the interpretations of secondary outcomes such as upper GI bleeds.

Finally, although clinicians used various drugs and administration routes, thereby increasing the generalizability of the results, a trial using different combinations of drugs or different routes could potentially have yielded different findings.