INTRODUCTION: Some observational studies suggest a cardiovascular risk from proton-pump inhibitor (PPI) treatment, but observational data may be subject to bias. We conducted a meta-analysis of randomized controlled trial data on cardiovascular events during PPI treatment.

METHODS: Manufacturers of PPIs provided data from their PPI clinical trial programs. We included randomized trials with at least 100 subjects, treatment duration >30 days, and a non-PPI comparator (active or placebo). We obtained person-time of exposure per trial, to calculate summary incidence rate ratios (primary analysis) and incident rate differences (secondary analysis). Our primary composite outcome was major adverse cardiovascular events-plus (MACE+), which included nonfatal myocardial infarction, nonfatal stroke, fatal cardiovascular adverse events, hospitalization for unstable angina, or coronary revascularization; events were adjudicated blindly.

RESULTS: Cardiovascular outcomes were infrequent in randomized trials of PPIs, and our primary analysis found no overall association (summary incident rate ratio, MACE+ events, PPI:placebo, 0.72) (95% confidence interval 0.42-1.26). There was some heterogeneity by indication category, and by the incidence rate difference metric, gastroesophageal reflux disorder trials had 1.04 (95% confidence interval 0.58-1.50) excess MACE+ events per 100 person-years of treatment vs placebo. Comparisons with active controls generally showed positive incidence rate differences with PPI treatment.

DISCUSSION: Overall, we found no association of cardiovascular events with PPI treatment. Cardiovascular events appeared more frequent with PPI treatment in gastroesophageal reflux disorder trials, but results from this subgroup should be interpreted with the limitations of the analysis in mind, including sparse outcome counts and lack of individual patient data.

Cardiology rater

Reassuring data about the cardiovascular safety of PPI.

Gastroenterology rater

Reassuring information but the data arising from pharmaceutical industry-based data lessens the impact.

Internal Medicine rater

Too few events of interest with 95% confidence intervals suggesting high uncertainly of the main finding - cannot exclude a clinically relevant benefit or harm.