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


A systematic review and meta-analysis was conducted in accordance with the checklist outlined by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement (Supporting Information Data S1).[27] The systematic review protocol is registered on PROSPERO: International prospective register of systematic reviews (ID number CRD42016042187).[28]

Search Strategy

The electronic databases of MEDLINE (via Ovid), EMBASE, Cochrane Central Register of Controlled Trials (CENTRAL), Scopus, Web of Science, and ClinicalTrials.gov were systematically searched from inception to October 2017. Search strings included broad medical subject headings and controlled search terms relating to the outcome of adverse cardiovascular events in combination with the prescription, dispensation or use of PPIs as an intervention. The electronic search string was adapted for each database and database-specific filtres were used. The search strategy conducted for MEDLINE (via Ovid) is included in Supporting Information Data S2. Reference lists of relevant systematic reviews, letters and other nonprimary data sources captured by the search strategy were also screened to allow for identification of relevant studies that may have otherwise been missed.

Inclusion and Exclusion Criteria

Original, peer-reviewed, research studies using human subjects and published in English were eligible for inclusion. Studies were included if primary data indicated exposure to PPIs, along with new onset or worsening cardiovascular disease of any subtype. Randomised controlled trials comparing PPI to active interventions (ie, anything other than placebo) were excluded. Studies involving participants aged under 18 years and studies including PPI concomitant therapy without assessing PPI as an independent risk factor (most commonly, in combination with anti-platelet therapy) were excluded. Case studies were also excluded. Conference abstracts and studies where full articles could not be retrieved were not initially excluded from the search strategy but later excluded from the meta-analysis. Qualitative data from these studies in included in Supporting Information Data S5.

Study Selection

Identified studies were imported into EndNote® and duplicates deleted. Two investigators (RB and RK) independently screened all titles and abstracts returned by the search strategy. Two investigators (RB and RK) independently assessed all full text copies of potentially relevant studies returned by title and abstract screening. Both title and abstract screen and full text screen were performed with both authors viewing every abstract or full text returned to minimise human error. A third investigator (DL) resolved any discordance in assessments.

Data Collection

Two investigators (RB and RK) independently used a standardised, pre-piloted data extraction form to retrieve relevant quantitative data from included studies. This data form was electronically generated using Google Sheets. The data form extracted study characteristics (author, study design, country), type of PPI and ascertainment of exposure, type of adverse cardiovascular event and ascertainment of outcome, measures of association, follow-up, and adjustment for confounders. The data extraction form was based on the minimum requirements recommended in the Cochrane Handbook for Systematic Reviews.[29]

Risk of Bias Assessment

Two investigators (RB and RK) assessed risk of bias, and a third investigator (IH) resolved discrepancies by consensus. The Newcastle-Ottawa Scale was used to assess the methodological quality of observational studies in terms of validating participant selection, population comparability, and outcome/exposure assessment.[30] The Cochrane Tool for Assessing Risk of Bias was used to assess the methodological quality of randomised controlled trials.[31] Although abstracts were encapsulated in the initial search strategy, abstracts were excluded from meta-analysis as risk of bias could not be accurately assessed using data presented.

Data Synthesis and Analysis

Measures of association, including odds ratios (ORs), hazard ratios (HRs), and relative risks (RRs), with 95% confidence intervals (CIs), were extracted from included studies. Like previously published meta-analyses,[23,32,33] adjusted ORs, HRs and RRs were pooled because adverse cardiovascular events were rare in the component studies. Data extracted from randomised controlled trials were pooled separately to data yielded from observational studies. Review Manager 5.3[34] software was used to calculate the weighted mean, variance of the overall effect, 95% CI and P–value, and generate forest plots for exposure-outcome comparisons in each dataset. The generic inverse variance method included measures of association that had been adjusted. A random-effect model was used in the setting of significant methodological and statistical heterogeneity across included studies, and the I 2 statistic was used to determine the degree to which the effect estimate varied.