Proton pump inhibitors (PPIs) perennially rank among the top three classes of oral medications by sales in the United States. In the U.S., 10 to 15 percent of adults have used a PPI within the last year and eight to 10 percent have used a PPI within the last 30 days. Yet simultaneously there seems to be tremendous concern regarding the safety of PPIs: articles appear nearly daily describing increased risk for heart attack, kidney disease, fracture, C. difficile infection, dementia and even all-cause death. Popularized by the lay press, these findings cause understandable consternation to our patients. What are we to do?
Understand the limitations of these studies
PPI adverse effect studies are generally conducted by retrieving retrospective data from a hospital system or an insurer. Simply put, the researcher compares the rate of bad outcome Y among those who received a PPI to the rate among similar patients who did not.
There is a major challenge inherent in this time-honored study design: patients who use PPIs differ at baseline from those who do not. Patients who use PPIs are older, have more comorbidities, are more likely to be hospitalized and are more likely to have hospitalizations that are long in duration with a high severity of illness. In every conceivable category, the patients who use PPIs are sicker. No one maintains that these differences were caused by the PPIs because they were present at baseline (i.e., before the patient took the PPI). But the baseline differences must be dealt with because sicker patients are also more likely to have bad outcome Y which would bias the study results. So, statistics are used to adjust for measurable baseline differences between the patients who did and did not receive PPIs, to equalize the baseline disparities between groups. The challenge comes because differences that are unmeasured, such as the differences in severity of diabetes or in the intensity of hospitalizations, cannot be adjusted for. These leftover differences (called “residual confounding”) tend to increase the strength of the association between the PPI and bad outcome Y. In other words, residual confounding generates bias against PPIs. When datasets are large and observed associations are weak, small amounts of residual confounding create false associations.
Large retrospective studies publish relative risks but patients should care about absolute risks. Physicians play an important patient education role in this respect. If a PPI triples the relative risk of bad outcome Y but the absolute incidence of Y is one per billion then the absolute excess risk of the PPI is (1 x 3) – 1 = 2 per billion — an insignificantly small number. (In fairness this same point could be made for some drug studies purporting to show a strong relative benefit. If the study is large, the absolute benefit may be meaningless). PPI adverse effect studies are very large and consistently show very weak relative risks which translates into exceedingly small absolute risks.
The problem with PPIs is not that they are dangerous but that they are overprescribed.
Risk for enteric infections is an exception, with excess relative risks of two to six-fold. Yet even for enteric infections this translates into small absolute risks of three to 200 per 100,000 patient-years. Small absolute risks are hard to weigh. Putting it in perspective, this risk is comparable to the annual risk of a car accident resulting in hospitalization. Will that make you stop driving?
Stop PPIs that are not indicated
The problem with PPIs is not that they are dangerous but that they are overprescribed. Many patients receive a PPI inappropriately for stress ulcer prophylaxis in the hospital, and then have the PPI inappropriately continued as an outpatient. Patients with uncomplicated gastroesophageal reflux disease (GERD) symptoms who report improvement after PPIs usually do very well with rare/intermittent PPIs. Countless patients take long-term PPIs for non-specific upper abdominal symptoms or for exclusively atypical symptoms of GERD. These patients are taking drugs that are no better than placebos. Gastroenterologists are certainly not the only physicians guilty of overprescribing PPIs but we may be in a unique position to take patients off PPIs.
Tapering PPIs takes time and effort, but it is time well spent. Rare side effects occur from PPIs like they do from all drugs (e.g., hypomagnesemia or acute interstitial nephritis). These are one in a million events but, if it happens to your patient who took a long-term PPI for a vague indication, you will regret it. Polypharmacy, which PPIs contribute to substantially, has serious consequences. Polypharmacy leads to medication errors, drug-drug interactions and demeans the medicines that actually matter. When PPIs are not indicated, they should be stopped.
There are sound reasons to believe that most of the adverse effects associated with PPIs are not in fact caused by PPIs. If PPIs do have serious adverse effects (e.g., increased risk for enteric infections) the absolute risks involved are so low that fear of these risks should not drive the decision of whether or not to take a PPI. When PPIs are prescribed appropriately, their benefits exceed any real or potential risks. When PPIs are not prescribed appropriately, they have no benefits and should be aggressively discontinued.
Dr. Freedberg consults for Pfizer.