A defense of proton-pump inhibitors
Data on their risks may seem persuasive, but one expert is not convinced.
In the eyes of Paul Moayyedi, MB ChB, PhD, MPH, the researchers who study the risks of proton-pump inhibitors (PPIs) are somewhat like the Kardashian family.
Reality television values drama over substance, he explained. “My view is that, unfortunately, science is moving in this direction too, in search of sensationalism rather than careful thought about what the data mean,” said Dr. Moayyedi, who is a professor of gastroenterology at McMaster University in Hamilton, Ontario.
He reviewed the body of research on risks of PPIs, and the shortcomings he has identified, during his talk at the annual meeting of the American College of Gastroenterology in Philadelphia in October.
“There are a number of potential long-term adverse events with PPIs,” said Dr. Moayyedi. “I will focus on those with the most data.”
Pneumonia tops that list, with both the most and the oldest research showing an association with PPIs. A study published in JAMA in 2004 first highlighted higher rates of community-acquired pneumonia among patients on PPIs.
“Since then there have been a number” of similar publications, he said, citing six case-control studies, three cohort studies, and three systematic reviews. “Basically, whichever systematic review you looked at, there were a lot of patients (anywhere between a million and 2 million participants in these studies), a lot of heterogeneity, but the finding was consistent: odds ratio of about 1.3 for developing pneumonia if you were taking PPIs.” Based on these findings, the authors of at least one of the reviews urged caution in the use of acid-suppression drugs, he noted.
“The data sound quite persuasive. Why am I not excited by it? I'd like to borrow a quote from Linda Evangelista, who famously said, ‘Supermodels don't wake up for less than $10,000 a day,’” said Dr. Moayyedi. “What epidemiologists would say is that they don't wake up for an odds ratio of less than 2.”
Smaller odds ratios suggest that the results may be due to confounding, he explained. “The biggest problem with all these data, and it's very consistent, as I will show you, is that the sicker patients are the ones that are prescribed PPIs.”
As an example, he cited a study published in JAMA in 2009 with 42,093 patients, about half of them on PPIs. It found a hospital-acquired pneumonia rate of 4.9% with PPIs versus 2.0% without. However, Dr. Moayyedi noted, “Those on acid suppression were much more likely to be diabetic, much more likely to have renal failure, much more likely to have cancer, COPD, MI.”
The studies are adjusted to deal with these associations, he acknowledged. “Of course, these researchers are not dumb. They realized this and they adjusted for confounders.”
However, in multiple studies, such adjustment dramatically decreased the association between PPIs and pneumonia, from between 2 and 3 to about 1.2 to 1.5.
“It's still statistically significant, but given that these are all database studies, where they're not designed to look at all possible known confounders, and certainly can't look at unknown confounders, it's very likely that these various modest odds ratios are not real, and that in reality, they're just residual confounding, and there is really no association between PPI and pneumonia,” said Dr. Moayyedi. “If you get a big drop between the unadjusted and adjusted, and the adjusted is close to 1, you've got to be very suspicious about the results.”
The same problem arises in the data indicating that PPIs increase fracture risk. “Again there's a lot of data: a systematic review of case-control cohort studies, again roughly a million people in case-control cohort studies, and again a statistically significant odds ratio. But if you look at the unadjusted odds ratio from these studies: The odds ratio is 1.3. When you adjust for confounding factors, it's 1.15, which is just statistically significant,” he said.
Clostridium difficile has also been associated with PPIs. “What about C. diff? It's the same sort of story,” said Dr. Moayyedi, citing a systematic review published in PLOS One in 2012. “Their conclusion is that if you do rigorous assessment of the data, then the association of PPI and C. diff does not support a cause-and-effect relationship,” he reported.
That finding got much less attention than the research uncovering risks of PPIs, according to Dr. Moayyedi. “Notice the other studies were published in JAMA. This is a much lower-impact journal, because it's not as exciting,” he said. “It's back to that reality TV culture. If you're sensational, you'll get published in a high impact; if you're not, you won't.”
Dr. Moayyedi and colleagues are soon to add their own contribution to the literature on the topic, focusing on the apparent cardiovascular risk of PPIs, which has been highlighted by studies finding increased adjusted hazards of myocardial infarction in patients taking a PPI with clopidogrel.
The main focus of the new study he has been working on, the COMPASS trial, was randomizing more than 27,000 patients with coronary artery disease to rivaroxaban or aspirin. Dr. Moayyedi's component of the study was to randomize a portion of those patients to PPI or placebo, principally to study whether a PPI reduced the bleeding risk associated with rivaroxaban and aspirin.
“We have the data, but unfortunately it's currently undergoing review for publication, and I can't tell you what the results are, but you'll know soon,” he said. He did promise that the trial will provide some randomized controlled answers to the questions raised by retrospective research on PPIs.
In the meantime, he'd like clinicians to look at the existing data with a cool eye. “The number needed to harm in a year in a 50-year-old woman that you prescribe these drugs to is … 700 for pneumonia, 2,000 for fracture, 4,000 for C. diff in the community. So really we're talking about a lot of heat for very little risk.”
Current practice includes both over- and under-prescription of PPIs, according to Dr. Moayyedi. “We need to be mindful of that. We really need to get this message out to our patients and really forcefully indicate that they need to take a PPI if it is clinically indicated,” he said.
He also had a message for the scientists putting out research on PPIs, which he again phrased as an appropriated quote, this time from Archimedes. “‘Give me a fulcrum and I can move the world.’ What I would say is ‘Give me a database and I can show you an association. Most of them will not be causal.’”
