The Long-term Risk for Myocardial Infarction or Stroke After Proton Pump Inhibitor Therapy (2008–2018)

Michael Nolde; Nayeon Ahn; Tobias Dreischulte; Ina-Maria Rückert-Eheberg; Florian Güntner; Alexander Günter; Roman Gerlach; Martin Tauscher; Ute Amann; Jakob Linseisen; Christa Meisinger; Sebastian-Edgar Baumeister


Aliment Pharmacol Ther. 2021;54(8):1033-1040. 

In This Article


We estimated the long-term effect of PPI compared to H2RA therapy on MI and IS risk in adults without pre-existing cardiovascular disease. Our analyses do not indicate an increased risk of MI or IS in the first decade after PPI therapy. PPIs are among the most frequently used medications.[1] This makes their safety an important clinical concern.

Our study adds information to the safety evaluation of PPIs. Large and well-controlled cohort studies had linked PPI use to an increased risk of MI and IS,[2–6] but more recently these concerns have been attenuated. In a large study using administrative claims data from commercial and Medicare Supplemental plans, researchers found no increase in risk of primary MI during PPI intake of up to 3 years.[12] In addition, a large randomized trial with 17 598 participants comparing pantoprazole intake vs placebo over 3 years showed no increase in the overall cardiovascular risk.[11] This study included patients with stable cardiovascular disease and peripheral artery disease, while our study included subjects without the history of cardiovascular conditions. Also, an analysis of 68 514 women enrolled in the Nurses' Health Study found no increased risk for primary IS in prevalent users of PPIs.[13]

Given that some patients use PPIs for many years,[34] an observational window of more than 3 years might be necessary, especially for patients without prevalent cardiovascular disease at baseline, if PPI intake was to cause vascular damage and thereby increase cardiovascular risk.

The study has several limitations. First, exposure was identified using dispensed prescriptions. Use of over-the-counter (OTC) medications and combination products (ATC Code A02BD) was not included in our exposure definition. This means that prevalent OTC users might have been included in our cohort, regardless of the 1-year exclusion period before study entry. We assumed that therapy is usually initiated by a physician, and therefore new users were well captured by our approach.

Another limitation is that our estimates relied on the assumption that the measured baseline covariates were sufficient to adjust for confounding. Large-scale randomized trials provide the most reliable evidence to detect small to moderate effects, while observational studies always remain under the risk of unadjusted confounding.

Only a minority of PPI and H2RA initiators had any condition that would indicate therapy at the start of treatment. This made control for confounding by indication more difficult. At the same time, it made it suitable to create a second control group of non-initiators that did not start any treatment at all. Indication bias might have played different roles at different times of our study. During the initial phase of our observation period PPIs were regarded as the more modern and effective drug, but later on concerns about the safety of PPIs were raised following a FDA warning regarding a clopidogrel-omeprazole interaction in 2009.[35] We addressed these issues by analysing 97 pre-selected negative control (tracer) outcomes and found it unlikely that unmeasured or residual confounding would bias the estimates away from the null.

We performed an 'as-started'[19] analysis capturing the long-term effect of PPI therapy understood as a point treatment and thereby neglecting any effects of dose or duration of PPI intake. This approach is ideal to address the long-term effect of PPI therapy, if intake causes irreversible vascular damage, but has limited power to detect effects, if only high-dose or long-term intake were to raise the cardiovascular risk.

Finally, many H2RA initiators switched to PPIs later on and H2RA initiators were much more likely to switch to a PPI than vice versa. Switching medication does bias the estimate towards the null. However, this problem did not affect the comparison with non-initiators, which resulted in similar estimates.

Despite these limitations, this study represents a significant contribution to the literature on the safety of PPIs. The large cohort size, the very good capture of acute cardiovascular outcomes in hospital records, the use of an active comparator in combination with high-dimensional IPT-weighting for confounding control and an extensive analysis of negative control outcomes make this study a highly reliable source of information with a moderate risk of bias.[36]

In summary, for patients with no history of MI or IS PPI therapy does not appear to increase the risk of MI or IS in the first decade after treatment initiation; any presumed effect is moderate, at most. Thus, physicians and patients should not avoid starting an indicated PPI therapy because of concerns related to increased cardiovascular risk. Further studies should examine the effects of long-term and high-dose intake of PPIs.