PPIs associated with MIs???

There was a recent article published online and hitting the press regarding a possible association between the use of proton-pump inhibitors (PPIs) and subsequent MIs (see PPI and MI PLoS 2015 in dropbox, or DOI:10.1371/journal.pone.0124653​). The article was unusual in that it used a new technique for medical research: data-mining.  The bottom line is that if there were a real association between PPI use and MIs, this would have profound medical implications, since over 113 million PPI scripts are filled annually around the world with over $13 billion in sales, and in the US in 2009 21 million people filled at least one PPI script (the 3rd highest seller in the US). details:

--they looked at over 16 million clinical documents on 2.9 million people to assess PPI use and cardiovascular risk. There were 2 large data sources for the data mining (Stanford, and Practice Fusion, Inc) and one prospective source for a survival analysis. For the data-mining, they electronically evaluated clinic notes to search for patients with a GERD diagnosis, use of meds, and subsequent notes with diagnoses of cardiovascular disease

--overall, 32,363 patients were identified (mean age 55, 44% male, 50% white/27% unknown, 6% on clopidogrel). they identified a similar number of propensity score-matched controls. mean followup 2.1 years

--results:

    --overall, patients with GERD and on PPIs had an adjusted odds ratio of 1.16 (1.09-1.24) for association with MI. H2 blockers were not associated with increased cardiovasc risk

    --survival analysis in the prospective cohort followed 5.2 years found a 2-fold increased association with cardiovascular mortality [HR=2.00 (1.07-3.78, p=0.031)]. No association was found for H2-blockers

    ​--the PPI association was independent of the concommitant use of clopidogrel (used in only 5.9% of the population)

although this data-mining is potentially a powerful tool to look at possible real associations, it is much less persuasive than the usual organized epidemiologic studies in terms of the types and quality of the data collected.  I bring up this article for several reasons. 

1. as I have highlighted in many of my older blogs (prior to the BMJ posts), I am very concerned about the over-use of PPIs. It used to be said that treatment for GERD (the most common indication for PPIs) could be either step-up (start with calcium, then go to H2 blocker, then to PPI as needed for symptom control) or step-down therapy (hit hard with PPI, then wean down to H2 blocker or calcium as tolerated). not so surprisingly, it is rare that we in primary care actually do the step-down (after all, the patient is doing well on the PPI, there are lots of other issues to address in the confines of a quick primary care visit, so dealing with stepping down therapy just isn't the priority). but PPIs are very powerful drugs, and there are known and potential problems associated with their use. The documented adverse effects include those associated with infections (since neutralizing the gastric acid eliminates one of the barriers to infection), malabsorption (since stomach acidity is important for some nutrient absorption), and potentially other issues. In terms of infections, there are pretty good data that prolonged PPI use is associated with increased risk of C. difficile infections, other enteric infections (salmonella, campylobacter), and community-acquired pneumonia (though data here are a bit mixed). Malabsorption has been found for iron, vitamin B12, magnesium and calcium, and the possible association of PPIs with hip, spine and wrist fractures (and an FDA warning about this). And one of the other concerns is the profound hypergastinemia, with potential risk of colon cancer (not found so far) and atrophic gastritis (which may be more common in those who are H Pylori positive, and could potentially lead to gastric cancer). And potential drug interactions (eg with clopidogrel) or other unusual adverse effects (acute interstitial nephritis)

2. there are many plausible explanations of an association between PPI use and MI, including:

    --PPIs inhibit the activity of dimethylarginine dimethylaminohydrolase (DDAH), which is responsible for 80% of the clearance of asymmetric dimethylarginine (ADMA), which is an endogenous molecule which inhibits the activity of nitric oxide synthase (NOS), which might impair endothelial function, increase vascular resistance, and promote inflammation and thrombosis. (this is the explanation offered by the authors of  the study, though the data are based on animal studies and cultured human endothelial cells)

    --perhaps there is an associated chronic inflammatory state created by the increase in GI infections from PPI use, and this inflammatory state is associated with CAD (as found with rheumatoid arthritis, in a couple of new studies on psoriasis, perhaps in HIV??)

    --or, the explanation I favor, is that a small but significant proportion of patients being treated for GERD in fact have GERD-like symptoms from CAD which are falsely ascribed to GERD. I did just see a patient who epitomized this: he had GERD-like symptoms, treated by H2 blockers which did not work, augmented to a PPI which still did not work. And on further questioning, his GERD symptoms were exertional. He will be getting worked-up for this, but he presents the case of approximately 10% of angina patients present with predominantly GI symptoms, including typical GERD symptoms (even more classical than his, with postprandial and not exertional symptoms). therefore, the relationship of PPI to MI may be really from the error of using PPIs to treat GI symptoms which really come from the heart...

so, I bring this up mostly as a means to reinforce my sense that we use too many PPIs (my other big concern is overuse of NSAIDs, but that is a side issue here...). And also to highlight a new mechanism in medical research for doing quick and dirty epidemiologic associations through data-mining.