Back to the past: REVISEing Stress Ulcer Prophylaxis

Backward reasoning and lack of ambition are the sins of the recent NEJM study and accompanying meta-analysis

I was surprised and frustrated when I saw the recent pantoprazole RCT to prevent stress ulcers during mechanical ventilation. That was indeed a moment of mixed feelings.

It was a surprise because I thought this tedious topic was over a long ago. "It’s so last week…", my daughter would sigh ironically as a witty teenager. And I got frustrated because the study lacked the ambition to move forward. I'd ask the authors: OK, I will prescribe pantoprazole for everyone right after intubation. But are all ventilated patients at the same risk? and: Would you help me navigate the risk factors to individualize benefits? If they don't, they will disappoint a decades-long consumer of their research.

Emmett "Doc” Brown when Marty McFly told him in the future scientists will get stuck for 30 years doing the same research

I think I got this discussion from the beginning. Stress ulcer bleeding and even deaths were a real problem in the ICU in the 80’s and 90’s. I remember clearly a day in 1996 and another day in 1999 when we lost patients to this terrible complication. I probably lost others, but I can recall those two because I felt it was so unfair and unexpected.

Proton pump inhibitors (PPI) were welcomed as the most promising medicine to prevent that kind of unexpected death and eventually became the standard of care. However, we also knew that higher gastric pH meant gastric contamination with enteral gram-negative flora, microaspiration of gram-negative rods, and ventilator-associated pneumonia.

The idea was common sense and has somehow waned. A few years ago, discussing how to implement genomic identification to track an Acinetobacter outbreak, I posited that Acinetobacter could come from the patient's gut, not only from the hands of the workers. It caused agitation in the room. There was an ethos of finding someone to blame. We decided to culture nasogastric aspirates of some patients and guess what: all tested patients with Acinetobacter VAP had the same bacteria in their stomachs. Moreover, their stomachs were alkaline. It was nice to see VAP pathophysiology working.

The pathophysiology of stress ulcers is also long-known. The luminal pH theory of gastric ulceration recedes to Schartz's aphorism “no acid, no ulcer”. It still stands, of course. But it is not enough. We all have acid in the stomach, and luminal pH gets very low daily. There has to be something else to induce mucosal erosion, ulceration, and bleeding.

Critical illness may induce gastric mucosal damage by adding two other factors. The stomach's inner lining has a sophisticated way of protecting the cells from the luminal acid (pH=2.0) by creating a thin mucous layer with pH=7.0. However, the delicate mechanism depends on adequate microperfusion. There is no news here: 30 years ago gastric intramucosal pH was already used as a surrogate for perfusion. If your patient has signals of hypoperfusion such as prolonged capillary refill time, be certain that the gastric inner mucus layer is dissolving and the outer cells will need protection.

The other factor has an iatrogenic flavor. Keeping the patient in a prolonged fast state causes the luminal pH to drop and may deprive the outer cells of nutrients. On the other hand, if you feed the patient through the mouth or stomach you will revert the process. Moreover, enteral formulas' neutral or alkaline pH offers direct buffering, and gastric perfusion is enhanced when food is inside the organ. At least I was told so in physiology classes.

Not surprisingly, the incidence of mucosal bleeding is higher in the first week of critical illness when shock and prolonged fast should be more frequent. Why not study the effect of one-week vs. one-month stress ulcer prophylaxis (SUP)? That would be so helpful.

Moreover, low pH is necessary for mucosal injury. A beautiful observational study by Mesquita et al. puts all the pieces together by showing in the context of critical care low pH is required for gastric mucous injury. I always imagined an RCT testing the hypothesis that controlling gastric luminal pH reduces bleeding, instead of giving PPI for everyone reduces bleeding.

This is why I said REVISE lacked ambition. The authors lost the opportunity to add nuance to the topic. Instead, they repeated their selves. In 1994, the same group conducted a large cohort to study risk factors for gastrointestinal bleeding in the ICU. They have identified a few, including shock and enteral nutrition, but also reported mechanical ventilation as the #1 risk factor. This is the point to be revised. Is there a CAUSAL pathway connecting mechanical ventilation to gastrointestinal bleeding? Or is the need for mechanical ventilation an aggregator of causal factors? I’d love to see how MV correlates with factors with clearer causal implications. A simple reanalysis of the 1994 dataset would help us understand better. However, the trialists opted for an uncritical revision of the practical rule “if the patient is in MV give gastric protection” they proposed in 1994.

Backward reasoning:

The act of testing a hypothesis without thinking about biological plausibility, finding a spurious association, and then performing post-hoc speculative reasoning about biological plausibility

The papers also bring an inelegant instance of backward reasoning. They compared outcomes between severity-wise subgroups without explaining why it would make sense in human biology.

Let’s think about it. You can have two patients with the same SAPS or APACHE score but with different combinations of diseases or physiological parameters compounding the number. Hence you can’t treat both patients as having the same risk, because you don’t know what pathophysiological processes are running in each one.

But the authors did it anyway: they took a continuous variable containing unmeasurable information causally related or not to the outcome, dichotomized (ugh) it to create two groups, compared the groups, and only then tried to explain the biology of a difference in the group’s outcomes. It was like throwing a brick upwards and then calling the whole world to watch the brick return from the sky to smash your head. One can feel the uneasiness in the meta-analysis Discussion section. It’s the pure juice of backward reasoning.

Hypotheses must make biological sense BEFORE they are tested in an RCT, for the Canadian taxpayers‘ sake!

I conclude that like a time machine, REVISE took us back to the nineties. How about that? That was a nice time. I am nostalgic about the analogical world, free of the mental burden of being online all the time. It’s hard to imagine today that as a Med student in 1994, my access to NEJM was an expensive quarterly Portuguese-language edition with a few selected articles. Today it looks like we’ve entered an over-access dystopia. So why not embark on the time journey?

I remember when Marty McFly traveled 30 years back in time to the year 1955 and met the young Doc. I finish this essay envisioning a scene that’s not in the movie. I saw Marty explaining, to Doc’s painful disappointment, that after 30 years scientists are still milking the same cow.

Thanks for reading!