Risk of Adverse Cardiovascular Events With Proton Pump Inhibitors Independent of Clopidogrel

Systematic Review With Meta-analysis

Riley Batchelor; Radya Kumar; Julia F. M. Gilmartin-Thomas; Ingrid Hopper; William Kemp; Danny Liew


Aliment Pharmacol Ther. 2018;48(8):780-796. 

In This Article


Principal Findings

The results of this systematic review of 22 studies do not show consistent evidence of increased risk of major adverse cardiovascular events with PPI therapy independent of a pharmacokinetic interaction with clopidogrel. A meta-analysis of results from eight randomised trials with low heterogeneity (4124 participants) found no association, while a separate meta-analysis of heterogeneous non-randomised data (443 284 participants) found a positive association. Given the limitations of observational data, these results may offer some reassurance when considering the cardiovascular safety of proton pump inhibitor monotherapy, in the context of emerging concerns about the general appropriateness of their widespread use.[1–5]

This is the most current meta-analysis examining PPI monotherapy and cardiovascular risk and is, to our knowledge, the only study to compare the results of both observational and randomised studies. The findings from our pooled analysis of trial data differed from those of Sun et al,[57] who found a 70% increased risk of major adverse cardiovascular events with PPI therapy in a meta-analysis of 16 randomised controlled trials. However, Sun et al. included different control groups, including histamine–2–receptor antagonists and surgery. Within this study, subgroup analysis demonstrated that histamine–2–receptor antagonists were strongly associated with a decreased risk of adverse cardiovascular outcomes as compared to PPI, with a reduction in risk with histamine–2–receptor antagonist therapy observed across all included studies.[57] The aetiology of the cardioprotective effect of histamine–2–receptor antagonists as compared to PPI observed is unclear and may merit further study. Given this result, however, inclusion of studies that compared histamine–2–receptor antagonist therapy with PPI in lieu of placebo significantly influenced the gross summary effects estimate.

Our findings from the pooled analysis of non-randomised data were in accord with those of Kwok et al,[23] who found in a subgroup meta-analysis of observational studies that PPIs were associated with increased cardiovascular risk independent of clopidogrel. This study builds upon the meta-analysis conducted by Kwok et al,[23] with the inclusion of two large observational studies[25,26] that post-dated the Kwok et al study, in addition to analysis of randomised controlled trials.

Potential Pathogenic Mechanisms by Which PPIs Increase Cardiovascular Risk

PPIs may attenuate clopidogrel's anti-platelet effects by inhibiting CYP2C19, which metabolises clopidogrel to its active metabolites.[16–19] However, PPIs may exert adverse cardiovascular effects beyond this pharmacokinetic drug-drug interaction, as evidenced by: the association between cardiovascular risk and PPIs that are not known to inhibit CYP2C19 (eg, pantoprazole) among patients taking concomitant clopidogrel[58] and the increased cardiovascular risk associated with PPIs among patients taking anti-platelet agents that are not dependent on CYP2C19 (eg, aspirin[59] and ticagrelor).[60]

Several pathogenic mechanisms by which PPIs may directly affect the risk of major adverse cardiovascular events have been proposed. Nitric oxide (NO) causes vasodilation and reduces platelet aggregation, which is essential for cardiovascular homoeostasis.[61] Asymmetric dimethylarginine (ADMA) is an endogenous competitive inhibitor of NO synthase.[62] Increased plasma ADMA levels are therefore associated with an increased risk of adverse cardiovascular events.[63] Ghebremariam et al[64] suggested that PPIs directly inhibit the activity of dimethylargininase (DDAH), an enzyme which metabolises ADMA, thereby increasing plasma ADMA levels and impairing vascular NO synthesis. The same authors conducted a subsequent cross-over study examining vascular function (using the EndoPAT device)[65] and plasma ADMA levels over a 4–week period of PPI exposure.[66] Although PPI use was associated with an increase in plasma ADMA, no adverse effect on vascular function was evident.[66] A further study conducted by Kruszelnicka et al[67] suggested that PPI use did not appear to increase plasma ADMA in a study cohort of nondiabetic men with coronary artery disease. Due to this conflicting evidence, the association between PPIs and subsequent dysregulation of NO as a potential explanation for the observed increase in cardiovascular risk remains controversial.

An alternative mechanism for the observed association between PPI use and cardiovascular risk is esomeprazole accelerating endothelial senescence, via inhibition of proteostasis and accumulation of protein aggregates linked to endothelial dysfunction.[68] This mechanism would also explain an association between PPI use and risk of chronic kidney disease[69] and vascular dementia.[70]

A final suggested mechanism involves lowering and severe depletion of plasma magnesium, which is crucial to cardiac electrophysiology, leading to cardiac arrhythmias and congestive heart failure.[71,72] Electrolyte imbalances such as hypomagnesaemia have been observed to occur with PPI use.[10]

Potential Explanation of the Findings Noted in the Meta-analysis of Non-randomised Data

Randomised controlled trials control for confounding and bias better than observational studies. The trials included in the present meta-analysis in this review were generally well conducted with reasonable reporting of adverse events. A 25% increase in relative risk was observed across observational studies included in this systematic review.

The observational studies included in the present meta-analysis adjusted for age, sex, diabetes, hypertension and hyperlipidaemia,[73,74] but not obesity, diet, alcohol consumption, physical exercise, insulin resistance and family history of cardiovascular disease.[75–77] Confounding by indication was also a possibility. Gastro-oesophageal reflux disease has been associated with an increased risk of adverse cardiovascular events in large observational studies,[78,79] and only two of the eight studies included in the meta-analysis of non-randomised data adjusted for the presence of gastro-oesophageal reflux disease.[25,46] In addition, residual confounding by unknown factors was also a possibility. Despite observational data suggesting an increased relative risk, participants included in observational studies were, in general, of increased cardiovascular risk as compared to the general population. This precluded calculation of an accurate absolute risk increase.

Use of aspirin by patients at high cardiovascular risk, for whom PPIs are prescribed for gastric protection, was another important potential confounder. Four of eight observational studies included in the meta-analysis accounted for aspirin use and maintained a positive association.[26,37,38,48] Charlot et al[37] undertook additional analysis with stratification by aspirin treatment, with the association between PPI use and cardiovascular risk persisting. Shih et al[26] conducted interaction tests to measure the influence of anti-platelet and nonsteroidal anti-inflammatory drugs on the demonstrated positive association. The association remained consistent, and no interaction test produced a statistically significant result, including that of aspirin.[26] Furthermore, Shah et al[25] evaluated the cardiovascular risk of histamine–2 receptor antagonists, an alternative to PPIs in reducing gastric pH and reducing risk of bleeding in aspirin patients. A lack of association was found between histamine–2 receptor antagonists and myocardial infarction (OR 0.93; 95% CI 0.86–1.02, P–value not published) within the same dataset that demonstrated a positive association with PPIs. The study conducted by Shih et al[26] noted similar findings; that use of histamine–2 receptor antagonists conferred a lower risk of myocardial infarction compared to use of PPIs. The lack of evidence of an association between histamine–2 receptor antagonists and cardiovascular risk also suggests that protopathic bias in unlikely to be present, which arises when the intervention is in response to a symptom of the undiagnosed outcome. This is relevant as dyspepsia can mimic the prodromal symptoms of ischaemic myocardial pain and has been proposed to lead to misidentification of acute coronary syndrome as gastro-oesophageal reflux disease.[80]

The potential for inaccurate diagnoses via ICD–9 codes noted in medical records was another limitation. The included observational studies did not use electrocardiogram findings or biochemical markers such as troponin, creatinine kinase or myoglobin to verify cardiovascular event diagnoses. However, because inaccurate diagnoses were likely to have equally affected both patients taking PPIs and patients not taking PPIs, this would have led to random misclassification, leading to an underestimation of the true effect of any association, rather than information bias.

Risk ratios at the lower confidence interval for summary measures of observational studies (1.11 for all cardiovascular events and 1.09 for myocardial infarction), as well as the risk ratios themselves (1.25 and 1.21 respectively) represent small effect sizes and are marginally significant. In the context of the limitations listed above, the observed association is clinically unimportant.

Finally, duration of PPI use was not consistently reported in the observational studies, which precluded analyses by length of follow-up or exposure. For the studies that noted length of PPI use, time-dependent associations with cardiovascular risk were not observed consistently.


A methodological limitation of the present systematic review and meta-analysis stemmed from heterogeneity in the nonrandomised studies. Furthermore, studies reporting odds ratios did not contain data necessary to generate time-dependent measures, and thus were pooled with hazard ratios in meta-analysis without conversion. However, given that major adverse cardiovascular events were rare in the component studies, this was unlikely to have impacted the results of meta-analysis significantly. In addition, effect size was not dependent on how the measure of effect was expressed. The pooling of different measures of effect size has also been undertaken in other large meta-analysis under similar circumstances.[6,23] Nonetheless, to address the issue, sensitivity analysis of studies reporting only hazard ratios were undertaken, the results of which were consistent with the main finding.

None of the eight randomised controlled trials included in the present review strictly excluded clopidogrel users. However, these trials minimised participation by people with known cardiovascular disease or who were using clopidogrel. At a minimum, all trials sought to differentiate patients with myocardial chest pain from reflux-related chest pain and had excluded patients on long-term NSAID or high-dose aspirin, or with uncontrolled systemic disease. Two trials strictly excluded patients with exercise tolerance tests positive for myocardial ischaemia[40] or previous history of cardiovascular disease.[45]

Conclusions and Policy Implications

The present systematic review and meta-analysis found no consistent evidence of an association between PPI use and an increased risk of cardiovascular events independent of clopidogrel. The results suggest that initiation of PPI monotherapy should not be avoided on the basis of cardiovascular risk.