Advertisement

Proton pump inhibitors and the risk of cardiovascular events and cardiovascular mortality: A systematic review and meta-analysis of observational studies

Published:October 01, 2022DOI:https://doi.org/10.1016/j.ejim.2022.09.021

      Highlights

      • Recent large studies substantially changed the observational evidence about the cardiovascular effect of proton pump inhibitors (PPIs).
      • PPI intake does not increase the risk of first cardiovascular events.
      • Previous reports of increased cardiovascular mortality can largely be explained by publication bias and observational study design biases.

      Abstract

      Background and Aims

      Observational research has indicated that proton pump inhibitors (PPIs) might increase the long-term risk of cardiovascular events. This study evaluated the evidence from observational studies for an effect of PPI monotherapy on the risk of incident cardiovascular events and cardiovascular mortality.

      Methods

      The databases MEDLINE, EMBASE, and Scopus were systematically searched up to September 2021. The primary outcome was first cardiovascular event, i.e. first myocardial infarction or first ischaemic stroke. The secondary outcome was cardiovascular mortality. Studies were included following a detailed risk of bias assessment with the ROBINS-I tool. Sensitivity and bias analyses adjusted for potential publication bias, immortal time bias, and unmeasured confounding.

      Results

      We included ten studies with 75,371 first cardiovascular events, as well as seven studies on cardiovascular mortality with 50,329 cardiovascular deaths in total. The pooled hazard ratios (HRs) for PPI use and cardiovascular events were 1.05 with a 95% confidence interval of (0.96; 1.15) before and 0.99 (0.93; 1.04) after adjusting for observational study design bias. The pooled HRs for PPI use and cardiovascular mortality were 1.27 (1.11; 1.44) before and 1.06 (0.96; 1.16) after adjusting for publication bias and observational study design bias.

      Conclusion

      It is questionable, whether PPI monotherapy constitutes a cardiovascular risk factor.

      Keywords

      1. Introduction

      1.1 Rationale

      Proton pump inhibitors (PPIs) are widely used to treat disorders characterized by excessive gastric acid production [
      • Rückert-Eheberg I.M.
      • Nolde M.
      • Ahn N.
      • Tauscher M.
      • Gerlach R.
      • Güntner F.
      • et al.
      Who gets prescriptions for proton pump inhibitors and why? A drug-utilization study with claims data in Bavaria, Germany, 2010–2018.
      ] and have been sold over-the-counter for more than one decade. Alongside, PPIs are used for gastroprotection in patients on dual antiplatelet therapy consisting of asprin in combination with a P2Y12 inhibitor such as clopidogrel, prasugrel or ticagrelor to prevent secondary myocardial infarctions and ischaemic strokes. Two different questions, therefore, arise regarding a potentially increased cardiovascular risk associated with PPI intake. The effect of PPI intake on secondary events as part of dual antiplatelet therapy is a question of short-term effects in a high-risk population and is examined most appropriately by clinical trials [
      • Pang J.
      • Wu Q.
      • Zhang Z.
      • Zheng T.Z.
      • Xiang Q.
      • Zhang P.
      • et al.
      Efficacy and safety of clopidogrel only vs. clopidogrel added proton pump inhibitors in the treatment of patients with coronary heart disease after percutaneous coronary intervention: a systematic review and meta-analysis.
      ,
      • Moayyedi P.
      • Eikelboom J.W.
      • Bosch J.
      • Connolly S.J.
      • Dyal L.
      • Shestakovska O.
      • et al.
      Safety of proton pump inhibitors based on a large, multi-year, randomized trial of patients receiving rivaroxaban or Aspirin.
      ]. The effect of PPI intake as a treatment of gastroesophageal diseases on primary events is a question of long-term effects in a low-risk population requiring both a large study population and long study period and thus best addressed by an observational study design.
      Unfortunately, observational studies examining the effect of PPI intake on cardiovascular outcomes are especially prone to bias as PPI intake might be associated with cardiovascular morbidity. Associations between PPI intake and cardiovascular outcomes could therefore indicate a causal effect of PPI intake or stem from residual confounding. Since the most recent meta-analyses [
      • Li S.
      • Liu F.
      • Chen C.
      • Zhu W.
      • Ma J.
      • Hu J.
      • et al.
      Real-world relationship between proton pump inhibitors and cerebro-cardiovascular outcomes independent of clopidogrel.
      ,
      • Batchelor R.
      • Kumar R.
      • Gilmartin-Thomas J.F.M.
      • Hopper I.
      • Kemp W.
      • Liew D.
      Systematic review with meta-analysis: risk of adverse cardiovascular events with proton pump inhibitors independent of clopidogrel.
      ,
      • Farhat N.
      • Fortin Y.
      • Haddad N.
      • Birkett N.
      • Mattison D.R.
      • Momoli F.
      • et al.
      Systematic review and meta-analysis of adverse cardiovascular events associated with proton pump inhibitors used alone or in combination with antiplatelet agents.
      ,
      • Shiraev T.P.
      • Bullen A.
      Proton pump inhibitors and cardiovascular events: a systematic review.
      ] found a higher risk of cardiovascular outcomes associated with PPI therapy, several large observational studies have been published that analysed this association in more detail.
      It is therefore time to have an updated look at the evidence regarding the relationship between PPI therapy and incident cardiovascular outcomes and to apply a rigorous risk of bias assessment of the included studies [
      • Sterne J.A.
      • Hernán M.A.
      • Reeves B.C.
      • Savović J.
      • Berkman N.D.
      • Viswanathan M.
      • et al.
      ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions.
      ].

      1.2 Objectives

      We performed a systematic review and meta-analysis to evaluate the effect of PPI therapy on the risk of first acute cardiovascular events, i.e. first myocardial infarction or first ischaemic stroke. In addition, we examined the effect on cardiovascular mortality.

      2. Methods

      2.1 Eligibility criteria

      2.1.1 Population

      We included observational studies in populations free of prevalent cardiovascular disease at inclusion for the analysis of first myocardial infarction. Likewise, populations free of prevalent cerebrovascular disease at inclusion were considered for the analysis of first stroke. For the analysis of cardiovascular mortality, we included study populations with and without prior cardiovascular events.

      2.1.2 Intervention

      The intervention under investigation was intake of PPIs (ATC Code A02BC). We included studies addressing an as-started [
      • Franklin J.M.
      • Patorno E.
      • Desai R.J.
      • Glynn R.J.
      • Martin D.
      • Quinto K.
      • et al.
      Emulating randomized clinical trials with nonrandomized real-world evidence studies: first results from the RCT DUPLICATE initiative.
      ] effect as well as studies addressing an on-treatment [
      • Franklin J.M.
      • Patorno E.
      • Desai R.J.
      • Glynn R.J.
      • Martin D.
      • Quinto K.
      • et al.
      Emulating randomized clinical trials with nonrandomized real-world evidence studies: first results from the RCT DUPLICATE initiative.
      ] effect. The as-started effect (also known as intention-to-treat analysis) is the effect of the initial treatment, regardless of treatment continuation. It assumes an irreversible long-term effect of a point treatment. The on-treatment effect (similar to a per-protocol analysis) is the effect of continuous treatment and assumes a reversible effect of treatment. More generally, under a reversible effect model time under risk is attributed to the current exposure, whereas in an irreversible effect model all time under risk after a point treatment is attributed to this baseline exposure.

      2.1.3 Comparators

      We included effect estimates comparing PPI intake versus H2RA (histamine-2 receptor antagonist; ATC Code A02BA) intake as an active comparator [
      • Lund J.L.
      • Richardson D.B.
      • Stürmer T.
      The active comparator, new user study design in pharmacoepidemiology: historical foundations and contemporary application.
      ] as well as estimates comparing PPI intake versus no intake.

      2.1.4 Outcomes

      The primary outcomes were first myocardial infarction and first ischaemic stroke. The secondary outcomes were the combined outcome of incident cardiovascular events (i.e. the combination of first myocardial infarction or first ischaemic stroke) and cardiovascular mortality.

      2.1.5 Information sources, search strategy and study selection

      We searched for peer-reviewed studies in English language in the PubMed / MEDLINE, EMBASE, and Scopus electronic databases from their respective inception dates until 16 September 2021. The search strings used for each of the databases can be found in Supplementary Table S1.
      Two authors (MN, IR) independently screened all titles and abstracts after initial removal of duplicates. Original research articles reporting treatment effect estimates and meeting our eligibility criteria were included. Single-case studies, cross-sectional studies, case-control studies without density sampling and randomized controlled trials were disregarded. Then, two authors (MN, IR) independently performed full-text reviews to decide on the inclusion of studies for the detailed risk of bias assessment. Studies were excluded if they had an unsuitable study design, inept selection of treatment controls or used a qualitative study design. All discrepancies were resolved by consensus.

      2.2 Risk of bias assessment and data extraction

      The risk of methodological bias was assessed by two review authors (MN, IR) independently, using the “Risk Of Bias In Non-randomised Studies - of Interventions“ (ROBINS-I) [
      • Sterne J.A.
      • Hernán M.A.
      • Reeves B.C.
      • Savović J.
      • Berkman N.D.
      • Viswanathan M.
      • et al.
      ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions.
      ] tool. This tool draws on the concept of considering each study as an emulated target trial [
      • Hernán M.A.
      • Robins JM.
      Using big data to emulate a target trial when a randomized trial is not available.
      ]. In this context, risk of bias is separately judged in seven domains using signalling questions and the ratings within each domain are carried forward to an overall risk of bias judgement. Disagreements were resolved by discussion. Details about the reasons that lead to attributing overall serious risk of bias to individual studies are given in Supplementary Table S2.
      We performed double data extraction for details on the study design and on the statistical analysis. Where data were missing or unclear, we contacted the corresponding author. The authors of one article [
      • Brown J.P.
      • Tazare J.R.
      • Williamson E.
      • Mansfield K.E.
      • Evans S.J.
      • Tomlinson L.A.
      • et al.
      Proton pump inhibitors and risk of all-cause and cause-specific mortality: a cohort study.
      ] provided additional information upon request. For one study [
      • Adelborg K.
      • Sundbøll J.
      • Schmidt M.
      • Bøtker H.E.
      • Weiss N.S.
      • Pedersen L.
      • Sørensen HT.
      Use of histamine H(2) receptor antagonists and outcomes in patients with heart failure: a nationwide population-based cohort study.
      ] we had to estimate the number of events. Where studies reported multiple effect estimates, we used data from the analysis with the lowest risk of bias and the longest follow up time. We used adjusted hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) to present and synthesize the results.

      2.3 Statistical analysis

      We used the Hartung-Knapp-Sidik-Jonkman random-effects meta-analysis approach with inverse-variance weighting which showed to produce adequate standard errors even when the number of studies is small [
      • Veroniki A.A.
      • Jackson D.
      • Bender R.
      • Kuss O.
      • Langan D.
      • Higgins J.P.T.
      • et al.
      Methods to calculate uncertainty in the estimated overall effect size from a random-effects meta-analysis.
      ,
      • IntHout J.
      • Ioannidis J.P.A.
      • Borm GF.
      The Hartung-Knapp-Sidik-Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian-Laird method.
      ] to estimate the between-study variance (τ2) and combine study-specific log HRs. We estimated pooled HRs, with corresponding 95% CIs and 95% prediction intervals (PIs) [
      • IntHout J.
      • Ioannidis J.P.A.
      • Rovers M.M.
      • Goeman JJ.
      Plea for routinely presenting prediction intervals in meta-analysis.
      ]. The 95% CI from a random-effects model contains highly probable values for the pooled HR. The 95% PI estimates where the true HR is to be expected in 95% of future studies under similar conditions factoring in the variability of the effect over different settings [
      • Higgins J.P.T.
      • Thompson S.G.
      • Spiegelhalter D.J.
      A re-evaluation of random-effects meta-analysis.
      ]. We reported the percentage of total variation due to heterogeneity (I2). Cochran's Q statistic was used to test for between-study heterogeneity.
      Subgroup analyses examined the effects of geographic region (Asia, Europe, United States), follow up time (up to or more than 5 years), study size (up to or more than 2,000 events), proportion of prevalent cardiovascular disease at study inclusion (up to or more than 20%), use of an active comparator / new user design (yes, no) and risk of bias (moderate, serious). All statistical tests were two-sided. The statistical software R (version 4.1.2, Foundation for Statistical Computing, Vienna, Austria; packages meta, metafor, metamisc [
      • Debray T.P.A.
      • Moons K.G.M.
      • Riley R.D.
      Detecting small-study effects and funnel plot asymmetry in meta-analysis of survival data: a comparison of new and existing tests.
      ], and metasens [
      • Schwarzer G.
      • Carpenter J.R.
      • Rücker G.
      Small-study effects in meta-analysis.
      ]) was used.

      2.4 Sensitivity and bias analyses

      Random-effects meta-analysis of observational studies can produce biased estimates of pooled effect sizes if the synthesized individual studies are subject to unmeasured confounding or selection bias [
      • Egger M.
      • Schneider M.D.
      • Smith G.
      Spurious precision? Meta-analysis of observational studies.
      ]. Thus, in order to detect outliers and influential studies we analysed Baujat plots [
      • Baujat B.
      • Mahé C.
      • Pignon J.P.
      • Hill C.
      A graphical method for exploring heterogeneity in meta-analyses: application to a meta-analysis of 65 trials.
      ,
      • Schmid C.H.
      • Stijnen T.
      • White I.R.
      • White I.
      Handbook of Meta-Analysis.
      ]. We then examined possible effects due to inclusion of small studies, selective publication of positive findings, and sensitivity to unobserved confounding and selection bias. Publication bias and small study effects (funnel plot asymmetry) were examined using the regression-based tests proposed by Debray et al. [
      • Debray T.P.A.
      • Moons K.G.M.
      • Riley R.D.
      Detecting small-study effects and funnel plot asymmetry in meta-analysis of survival data: a comparison of new and existing tests.
      ]. If the tests indicated bias, we applied the trim-and-fill method [
      • Schwarzer G.
      • Carpenter J.R.
      • Rücker G.
      Small-study effects in meta-analysis.
      ], the Copas selection model [
      • Schwarzer G.
      • Carpenter J.R.
      • Rücker G.
      Small-study effects in meta-analysis.
      ] and adjusted for small study bias using the Rücker regression-based shrinkage estimator [
      • Schwarzer G.
      • Carpenter J.R.
      • Rücker G.
      Small-study effects in meta-analysis.
      ].
      In addition, we quantified bias introduced by analytical and clinical study design choices using meta-regression. Due to the small number of studies we calculated the p-value for the meta-regression model using a permutation test [
      • Higgins J.P.T.
      • Thompson S.G.
      Controlling the risk of spurious findings from meta-regression.
      ]. Following a two-stage approach [
      • Mathur M.B.
      • VanderWeele T.J.
      Methods to address confounding and other biases in meta-analyses: review and recommendations.
      ], we further adjusted each study individually for bias introduced by publication bias and study design choices. Finally, random-effects meta-analysis of these bias-adjusted HRs was performed to estimate a pooled bias-adjusted HR.

      3. Results

      3.1 Systematic review and qualitative analysis

      After removal of duplicate publications our literature search identified a total of 7,038 publications (Fig. 1). Examination of titles and abstracts and a full-text review of 39 studies left us with 17 studies for a detailed risk of bias assessment [
      • Sterne J.A.
      • Hernán M.A.
      • Reeves B.C.
      • Savović J.
      • Berkman N.D.
      • Viswanathan M.
      • et al.
      ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions.
      ]. Only studies with moderate or serious risk of bias were included in the analyses. Studies with critical risk of bias were excluded (Supplementary Table S3). Two studies [
      • Nolde M.
      • Ahn N.
      • Dreischulte T.
      • Rückert-Eheberg I.M.
      • Güntner F.
      • Günter A.
      • et al.
      The long-term risk for myocardial infarction or stroke after proton pump inhibitor therapy (2008–2018).
      ,
      • Sehested T.S.G.
      • Gerds T.A.
      • Fosbøl E.L.
      • Hansen P.W.
      • Charlot M.G.
      • Carlson N.
      • et al.
      Long-term use of proton pump inhibitors, dose-response relationship and associated risk of ischemic stroke and myocardial infarction.
      ] estimated the effect on myocardial infarction and ischaemic stroke separately in their study populations. Rooney et al. [
      • Rooney M.R.
      • Bell E.J.
      • Alonso A.
      • Pankow J.S.
      • Demmer R.T.
      • Rudser K.D.
      • et al.
      Proton pump inhibitor use, hypomagnesemia and risk of cardiovascular diseases: the atherosclerosis risk in communities (ARIC) study.
      ] estimated the effect on ischaemic stroke and cardiovascular mortality in one study population, while Landi et al. [
      • Landi S.N.
      • Sandler R.S.
      • Pate V.
      • Lund J.L.
      No increase in risk of acute myocardial infarction in privately insured adults prescribed proton pump inhibitors vs histamine-2 receptor antagonists (2002–2014).
      ] estimated the effect on myocardial infarction in two separate study populations. Thus, in total, five estimates regarding myocardial infarction [
      • Nolde M.
      • Ahn N.
      • Dreischulte T.
      • Rückert-Eheberg I.M.
      • Güntner F.
      • Günter A.
      • et al.
      The long-term risk for myocardial infarction or stroke after proton pump inhibitor therapy (2008–2018).
      ,
      • Sehested T.S.G.
      • Gerds T.A.
      • Fosbøl E.L.
      • Hansen P.W.
      • Charlot M.G.
      • Carlson N.
      • et al.
      Long-term use of proton pump inhibitors, dose-response relationship and associated risk of ischemic stroke and myocardial infarction.
      ,
      • Landi S.N.
      • Sandler R.S.
      • Pate V.
      • Lund J.L.
      No increase in risk of acute myocardial infarction in privately insured adults prescribed proton pump inhibitors vs histamine-2 receptor antagonists (2002–2014).
      ,
      • Shih C.J.
      • Chen Y.T.
      • Ou S.M.
      • Li S.Y.
      • Chen T.J.
      • Wang SJ.
      Proton pump inhibitor use represents an independent risk factor for myocardial infarction.
      ] and five estimates regarding ischaemic stroke [
      • Nolde M.
      • Ahn N.
      • Dreischulte T.
      • Rückert-Eheberg I.M.
      • Güntner F.
      • Günter A.
      • et al.
      The long-term risk for myocardial infarction or stroke after proton pump inhibitor therapy (2008–2018).
      ,
      • Sehested T.S.G.
      • Gerds T.A.
      • Fosbøl E.L.
      • Hansen P.W.
      • Charlot M.G.
      • Carlson N.
      • et al.
      Long-term use of proton pump inhibitors, dose-response relationship and associated risk of ischemic stroke and myocardial infarction.
      ,
      • Rooney M.R.
      • Bell E.J.
      • Alonso A.
      • Pankow J.S.
      • Demmer R.T.
      • Rudser K.D.
      • et al.
      Proton pump inhibitor use, hypomagnesemia and risk of cardiovascular diseases: the atherosclerosis risk in communities (ARIC) study.
      ,
      • Nguyen L.H.
      • Lochhead P.
      • Joshi A.D.
      • Cao Y.
      • Ma W.
      • Khalili H.
      • et al.
      No significant association between proton pump inhibitor use and risk of stroke after adjustment for lifestyle factors and indication.
      ,
      • Wang Y.F.
      • Chen Y.T.
      • Luo J.C.
      • Chen T.J.
      • Wu J.C.
      • Wang S.J.
      Proton-pump inhibitor use and the risk of first-time ischemic stroke in the general population: a nationwide population-based study.
      ] were combined for the analysis of acute cardiovascular events (24,547 cases of first ischaemic stroke and 50,824 cases of first myocardial infarction). In the analysis of cardiovascular mortality, the estimates of seven studies with 50,329 cardiovascular deaths were included [
      • Brown J.P.
      • Tazare J.R.
      • Williamson E.
      • Mansfield K.E.
      • Evans S.J.
      • Tomlinson L.A.
      • et al.
      Proton pump inhibitors and risk of all-cause and cause-specific mortality: a cohort study.
      ,
      • Adelborg K.
      • Sundbøll J.
      • Schmidt M.
      • Bøtker H.E.
      • Weiss N.S.
      • Pedersen L.
      • Sørensen HT.
      Use of histamine H(2) receptor antagonists and outcomes in patients with heart failure: a nationwide population-based cohort study.
      ,
      • Rooney M.R.
      • Bell E.J.
      • Alonso A.
      • Pankow J.S.
      • Demmer R.T.
      • Rudser K.D.
      • et al.
      Proton pump inhibitor use, hypomagnesemia and risk of cardiovascular diseases: the atherosclerosis risk in communities (ARIC) study.
      ,
      • de Francisco
      • Angel L.M.
      • Varas J.
      • Ramos R.
      • Merello J.I.
      • Canaud B.
      • Stuard S.
      • et al.
      Proton pump inhibitor usage and the risk of mortality in hemodialysis patients.
      ,
      • He Q.
      • Xia B.
      • Meng W.
      • Fan D.
      • Kuo Z.C.
      • Huang J.
      • et al.
      No associations between regular use of proton pump inhibitors and risk of all-cause and cause-specific mortality: a population-based cohort of 0.44 million participants.
      ,
      • Shah N.H.
      • LePendu P.
      • Bauer-Mehren A.
      • Ghebremariam Y.T.
      • Iyer S.V.
      • Marcus J.
      • et al.
      Proton pump inhibitor usage and the risk of myocardial infarction in the general population.
      ,
      • Xie Y.
      • Bowe B.
      • Yan Y.
      • Xian H.
      • Li T.
      • Al-Aly Z.
      Estimates of all cause mortality and cause specific mortality associated with proton pump inhibitors among US veterans: cohort study.
      ].
      Fig. 1
      Fig. 1PRISMA Flowchart Flowchart of the inclusion of studies in the review. Four studies each yielded two separate effect estimates.
      Studies differed in size (58–28,207 events), study design (comparator H2RA or non-user; new use or prevalent use), duration of follow-up (maximum follow-up between 4 and 231 months), study population characteristics (age and sex structure, prevalence of comorbidities), and statistical analysis (reversible/irreversible effect model, adjusted covariates). Detailed characteristics of all included studies are shown in Tables 1 and 2. Among studies with a cardiovascular mortality endpoint there was large variation of the proportion of prevalent cardiovascular disease at study inclusion and some studies [
      • Adelborg K.
      • Sundbøll J.
      • Schmidt M.
      • Bøtker H.E.
      • Weiss N.S.
      • Pedersen L.
      • Sørensen HT.
      Use of histamine H(2) receptor antagonists and outcomes in patients with heart failure: a nationwide population-based cohort study.
      ,
      • de Francisco
      • Angel L.M.
      • Varas J.
      • Ramos R.
      • Merello J.I.
      • Canaud B.
      • Stuard S.
      • et al.
      Proton pump inhibitor usage and the risk of mortality in hemodialysis patients.
      ,
      • Shah N.H.
      • LePendu P.
      • Bauer-Mehren A.
      • Ghebremariam Y.T.
      • Iyer S.V.
      • Marcus J.
      • et al.
      Proton pump inhibitor usage and the risk of myocardial infarction in the general population.
      ] were undertaken in clinical populations with high cardiovascular morbidity. In particular, the two large studies [
      • Brown J.P.
      • Tazare J.R.
      • Williamson E.
      • Mansfield K.E.
      • Evans S.J.
      • Tomlinson L.A.
      • et al.
      Proton pump inhibitors and risk of all-cause and cause-specific mortality: a cohort study.
      ,
      • Xie Y.
      • Bowe B.
      • Yan Y.
      • Xian H.
      • Li T.
      • Al-Aly Z.
      Estimates of all cause mortality and cause specific mortality associated with proton pump inhibitors among US veterans: cohort study.
      ] on cardiovascular mortality examined all-cause mortality and reported cause-specific mortality as part of their subgroup analyses. This broader scope also meant that eligibility criteria were not tailored specifically for studying cardiovascular mortality and patients with prevalent cardiovascular disease were included, which possibly biased effect estimates due to confounding by indication.
      Table 1Characteristics of included studies (ordered by outcome, year).
      Author, yearData SourceContinent, CountryOutcome,No. casesComparatorMax follow-upHR (95% CI)Robins-IEffect modelStatistical ModelNew user design
      Nolde,

      2021 (MI)
      Claims DataEurope,

      Germany
      MI,

      4,606
      H2RA120

      months
      0.96

      (0.80-1.16)
      ModerateIrreversibleWeighted CoxYes
      Landi,

      2018 (a)
      Claims Data (Truven Marketscan Commercial)North America,

      US
      MI,

      21,670
      H2RA36

      months
      0.94

      (0.88-0.99)
      ModerateIrreversibleWeighted CoxYes
      Landi,

      2018 (b)
      Claims Data (Truven Marketscan Medicare)North America,

      US
      MI,

      23,556
      H2RA36

      months
      0.96

      (0.92-1.01)
      ModerateIrreversibleWeighted CoxYes
      Sehested,

      2018 (MI)
      Linked

      registers
      Europe,

      Denmark
      MI,

      863
      Non-user12

      months
      1.12

      (0.95-1.27)
      SeriousIrreversibleMulti-variable CoxNo
      Shih,

      2014
      Claims DataAsia,

      Taiwan
      MI,

      129
      Non-user4

      months
      1.58

      (1.11-2.25)
      SeriousIrreversibleMatched CoxYes
      Nolde,

      2021 (IS)
      Claims DataEurope,

      Germany
      IS,

      18,393
      H2RA120

      months
      0.98

      (0.89-1.08)
      ModerateIrreversibleWeighted CoxYes
      Rooney,

      2020 (IS)
      Cohort StudyNorth

      America,

      US
      IS,

      122
      Non-user84

      months
      0.92

      (0.60-1.44)
      SeriousIrreversibleMulti-variable CoxNo
      Nguyen,

      2018
      Cohort Studies

      (Nurses’

      Health Study & Health

      Professionals Follow-up Study)
      North America,

      US
      IS,

      2,599
      Non-user144

      months
      1.08

      (0.91-1.27)
      SeriousReversibleTime-varying CoxNo
      Sehested,

      2018 (IS)
      Linked

      registers
      Europe,

      Denmark
      IS,

      1,198
      Non-user12

      months
      1.20

      (1.06-1.36)
      SeriousIrreversibleMulti-variable CoxNo
      Wang,

      2017
      EHRAsia,

      Taiwan
      IS,

      2,235
      H2RA143

      months
      1.11

      (1.02-1.21)
      ModerateIrreversibleMatched CoxYes

      (30 days

      washout)
      Brown,

      2021
      EHR

      (General Practice

      Research Database)
      Europe,

      UK
      CVM,

      28,207
      H2RA231

      months
      1.14

      (1.07-1.22)
      ModerateIrreversibleWeighted CoxYes
      He,

      2021
      EHR

      (UK Biobank)
      Europe,

      UK
      CVM,

      352
      H2RA

      (regular use)
      121

      months
      1.26

      (0.89-1.79)
      ModerateIrreversibleMulti-variable CoxNo
      Rooney,

      2020 (CVM)
      Cohort StudyNorth

      America,

      US
      CVM,

      121
      Non-user84

      months
      1.36

      (0.87-2.12)
      SeriousIrreversibleMulti-variable CoxNo
      Xie,

      2019
      EHR

      (US Department of Veterans Affairs)
      North

      America,

      US
      CVM,

      18,148
      H2RA120

      months
      1.25

      (1.10-1.44)
      ModerateIrreversibleWeighted CoxYes
      Adelborg,

      2018
      Linked

      registers
      Europe,

      Denmark
      CVM,

      3,220*
      H2RA60

      months
      1.23

      (1.08-1.41)
      SeriousIrreversibleMatched CoxYes
      De Francisco,

      2018
      EHR

      (European Clinical Database)
      Europe,

      Spain
      CVM,

      223
      Non-user30

      months
      1.67

      (1.03-2.71)
      SeriousIrreversibleMatched CoxNo
      Shah,

      2015
      Cohort StudyNorth

      America,

      US
      CVM,

      58
      Non-user96

      months
      2.00

      (1.07-3.78)
      SeriousIrreversibleMulti-variable CoxNo
      HR: hazard ratio; MI: myocardial infarction; IS: ischaemic stroke; CVM: cardiovascular mortality;
      EHR: Electronic health records; H2RA: H2 receptor antagonists
      Weighted Cox: Cox regression model using balancing weights to adjust for baseline confounding
      * estimated
      Table 2Additional characteristics of included studies (ordered by outcome, year).
      Author, yearStudy PopulationPREV CVDAdjusted CovariatesRemarksThe seven domains of the ROBINS-I
      CFSELINTDEVMISSOUTREP
      Nolde,

      2021 (MI)
      Age ≥ 18NoAge, Sex, COME, COMO, TI, ASP, CLOo++++++
      Landi,

      2018 (a)
      Working population;

      age 18–65
      1.4 % strokeAge, Sex, COME, COMO, TI, CLO, health care utilizationAdditional on-treatment effect estimateo++++++
      Landi,

      2018 (b)
      Retired population; age ≥ 658.9 % strokeAge, Sex, COME, COMO, TI, CLO, health care utilizationAdditional on-treatment effect estimateo++++++
      Sehested,

      2018 (MI)
      Patients after elective upper endoscopy;

      age 30–99
      NoAge, Sex, COME, COMO, TI, ASP, SESIn patients after upper endoscopy-++++++
      Shih,

      2014
      Age 18–8012.4 % CEVDAge, Sex, COME, COMO, ASP, CLO, SES, health care utilizationo-++++o
      Nolde,

      2021 (IS)
      Age ≥ 18NoAge, Sex, COME, COMO, TI, ASP, CLOo++++++
      Rooney,

      2020 (IS)
      5th visit of a cohort study;

      age 69–90
      NoAge, Sex, Race, Education, COME, Diabetes, Livestyle, Lab---+o+o
      Nguyen,

      2018
      Age 30–755.0 %

      CVD
      Age, Sex, COME, COMO, TI, ASP, LivestyleOnly study assuming reversible causal effect-oo+o++
      Sehested,

      2018 (IS)
      Patients after elective upper endoscopy;

      age 30–99
      NoAge, Sex, COME, COMO, TI, ASP, SESIn patients after upper endoscopy-++++++
      Wang,

      2017
      Age ≥ 2017.4 % CAD

      1.2 %

      MI
      Age, Sex, COME, COMO, ASP, CLO, SES, health care utilizationoo+++++
      Brown,

      2021
      Age ≥ 188.2 % CHD;

      4.5 %

      CEVD
      Age, Sex, COME, COMO, ASP, CLO, Livestyle, SES, health care utilizationo++++++
      He,

      2021
      Age 37–73NoAge, Sex, Race, COME, COMO, TI, ASP, SES, Education, Livestyleooo+o++
      Rooney,

      2020 (CVM)
      5th visit of a cohort study;

      age 69–90
      NoAge, Sex, Race, Education, COME, Diabetes, Livestyle, Lab---+o+o
      Xie,

      2019
      Veterans;

      96% male
      25.2 %

      CVD
      Age, Sex, Race, COME, COMO, TI, Livestyle, SES, health care utilization, Laboooo+++
      Adelborg,

      2018
      Patients hospitalized with first-time heart failure;

      mostly older
      53.6 %

      CAD;

      14.8 %

      stroke
      Age, Sex, COME, COMO, TI, SESIn patients with heart failureo-o++++
      De Francisco,

      2018
      Hemodialysis

      Patients;

      age ≥ 18,

      mostly older
      40.0 %

      on anti-platelets
      Age, Sex, COME, COMO, Lab, ASP/CLO combinedIn patients on hemodialysis--++o++
      Shah,

      2015
      Patients after non-emergent elective coronary angiogram76% CADAge, Sex, Race, COME, COMO, Lab, LivestyleIn patients after coronary angiogram---+o+o
      MI: myocardial infarction; IS: ischaemic stroke; CVM: cardiovascular mortality;
      PREV CVD: prevalent cardiovascular disease; CF: confounding; SEL: selection; INT: intervention; DEV: deviations; MISS: missing; OUT: outcome; REP: reporting; MI: myocardial infarction; CEVD cerebrovascular disease; CVD cardiovascular disease; CHD: coronary heart disease; CAD: coronary artery disease; COME: comedications; COMO: comorbidities; TI: treatment indications; SES: socio-economic status; ASP: aspirin; CLO: clopidogrel;
      +: low risk of bias; o: moderate risk of bias; -: serious risk of bias

      3.2 Meta-analysis

      3.2.1 Cardiovascular events

      The random-effects meta-analysis yielded pooled HRs of 1.05 with a 95% confidence interval of (0.83; 1.32) (Fig. 2a) for first myocardial infarction, 1.08 (0.97; 1.20) (Fig. 2b) for first ischaemic stroke and 1.05 (0.96; 1.15) (Fig. 2c) for first cardiovascular events. All CIs and all PIs included the null. There was moderate to substantial heterogeneity (46.9%-71.2%) between studies. Stratified analyses suggested that studies with a serious risk of bias (HR 1.16), small studies (HR 1.18), and studies conducted in an Asian population (HR 1.13) resulted in higher risk estimates (Table 3).
      Fig. 2
      Fig. 2(a–d). Forest plot of random-effects meta-analyses for (a) myocardial infarction, (b) ischaemic stroke, (c) acute cardiovascular events, (d) cardiovascular mortality Study-specific hazard ratios (HR) are represented by black diamonds (with their 95% confidence interval [CI] as error bars). HRs were combined using a Hartung-Knapp-Sidik-Jonkman random-effects model, yielding a pooled HR and its 95% CI and 95% prediction interval. The dotted line represents the pooled HR. Two-sided P value for between-study heterogeneity based on Cochran Q statistic.
      Table 3Subgroup meta-analyses.
      Cardiovascular Events (10 studies)
      SubgroupNo. of studiesHR (95% CI)I2, %Τ2P
      Risk of bias (ROBINS-I)

      moderate

      serious


      5

      5


      0.99 (0.91–1.07)

      1.16 (0.96–1.40)


      63.0

      23.3


      0.0029

      0.0185
      0.031

      Follow-up time

      short studies (≤ 5 years)

      long studies (> 5 years)


      5

      5


      1.09 (0.87–1.37)

      1.04 (0.95–1.13)


      82.7

      18.1


      0.0294

      0.0024
      0.53

      Number of events

      small studies (≤ 2000)

      big studies (> 2000)


      4

      6


      1.18 (1.08–1.29)

      0.98 (0.95–1.01)


      59.0

      32.2


      0.0249

      0.0030
      0.062

      Geographic region

      Asia

      Europe

      US


      2

      4

      4


      1.13 (1.04–1.23)

      1.06 (0.99–1.13)

      0.96 (0.92–0.99)


      72.4

      62.4

      0


      0.0402

      0.0073

      0.0017
      0.084

      Cardiovascular Mortality (7 studies)
      SubgroupNo. of studiesHR (95% CI)I2, %Τ2P
      Risk of Bias (ROBINS-I)

      moderate

      serious


      3

      4


      1.17 (1.03–1.34)

      1.39 (1.02–1.89)


      0

      14.1


      0.0320

      0.0195
      0.097

      Active comparator / new user design

      yes

      no


      3

      4


      1.18 (1.04–1.34)

      1.46 (1.08–1.99)


      3.3

      0


      0.0010

      0.0142
      0.033

      Prevalent cardiovascular disease

      ≤ 20%

      > 20%


      3

      3


      1.16 (1.01–1.32)

      1.27 (0.98–1.66)


      0

      0


      0.0017

      0.0095
      0.16

      Follow-up time

      short studies (≤ 5 years)

      long studies (> 5 years)


      2

      5


      1.32 (0.26–6.78)

      1.25 (1.05–1.50)


      29.9

      18.6


      0.0195

      0.0173
      0.72

      Number of events

      small studies (≤ 2000)

      big studies (> 2000)


      4

      3


      1.46 (1.08–1.99)

      1.18 (1.04–1.34)


      0

      3.3


      0.0142

      0.0010
      0.033

      Geographic region

      Europe

      US


      4

      3


      1.22 (1.02–1.45)

      1.36 (0.84–2.21)


      11.2

      5.3


      0.0098

      0.0242
      0.37

      CI: confidence interval; HR: hazard ratio (calculated in the Hartung-Knapp-Sidik-Jonkman random-effects model); I2: percentage of total variance explained by τ2; τ2: between-study variance; P: p-value of Q test for subgroup differences.

      3.2.2 Cardiovascular mortality

      The pooled HR for PPI use and cardiovascular mortality was 1.27 (1.11; 1.44) (Fig. 2d). Heterogeneity between studies was low (17.9%). In the stratified analyses, we found smaller effect estimates in large studies and in studies following an active comparator new user design [
      • Lund J.L.
      • Richardson D.B.
      • Stürmer T.
      The active comparator, new user study design in pharmacoepidemiology: historical foundations and contemporary application.
      ] (Table 3).

      3.3 Sensitivity and bias analyses

      3.3.1 Cardiovascular events

      The Baujat plot confirmed that the studies of Landi et al. [
      • Landi S.N.
      • Sandler R.S.
      • Pate V.
      • Lund J.L.
      No increase in risk of acute myocardial infarction in privately insured adults prescribed proton pump inhibitors vs histamine-2 receptor antagonists (2002–2014).
      ] had the largest influence on the effect estimate (Supplementary Fig. S1).
      Analysis of funnel plot asymmetry (Supplementary Fig. S2), Egger test (p-value 0.069) and Debray test (p-value 0.095) (Supplementary Table S4) showed little evidence for small study bias. According to the contour-enhanced funnel plot (Supplementary Fig. S2) the reported estimates were sufficiently explained under the null.
      Meta-regression analysis estimated the effect of study design choices summarized by the ROBINS-I assessment (moderate or serious) to 0.16 on the log(HR) scale with a standard error of 0.06 and a p-value of 0.030. We adjusted the reported HRs and CIs accordingly (Supplementary Table S5). Meta-analysis of these bias-adjusted HRs yielded a pooled bias-adjusted HR of 0.99 (0.93; 1.04) (Fig. 3a).
      Fig. 3
      Fig. 3(a,b). Forest plot of bias-adjusted random-effects meta-analyses for (a) acute cardiovascular events, (b) cardiovascular mortality Study-specific hazard ratios (HR) are represented by black diamonds (with their 95% confidence interval [CI] as error bars). HRs were combined using a Hartung-Knapp-Sidik-Jonkman random-effects model, yielding a mean hazard ratio and its 95% CI and 95% prediction interval. The dotted line represents the pooled HR. Two-sided P value for between-study heterogeneity based on Cochran Q statistic.(a) HR and 95%-CI adjusted for bias introduced by study design (serious risk of bias) (b) HR and 95%-CI adjusted for publication bias and bias introduced by study design (prevalent cardiovascular disease, prevalent use, non-user control).

      3.3.2 Cardiovascular mortality

      The Baujat plot (Supplementary Fig. S3) identified the study of Brown et al. [
      • Brown J.P.
      • Tazare J.R.
      • Williamson E.
      • Mansfield K.E.
      • Evans S.J.
      • Tomlinson L.A.
      • et al.
      Proton pump inhibitors and risk of all-cause and cause-specific mortality: a cohort study.
      ] as the most influential study for the effect estimate. Despite the small number of studies included, analysis of funnel plot asymmetry (Supplementary Fig. S4), Egger test (p-value 0.005) and Debray test (p-value 0.039) (Supplementary Table S4) showed strong evidence for small study bias. We calculated pooled HRs adjusted for small study bias (Supplementary Table S6) using the trim-and-fill method [HR 1.19 (1.01; 1.39)] (Supplementary Fig. S5), Copas selection model [HR 1.17 (1.09; 1.25)] and Rücker's shrinkage procedure [HR 1.16 (1.07; 1.26)] (Supplementary Fig. S6).
      Multiple meta-regression analysis estimated the bias introduced on the log(HR) scale by deviating from an active comparator, new user design to 0.21 (standard error 0.07) and by including patients with prevalent cardiovascular disease to 0.0025 (standard error 0.0009) per 1% increase (Supplementary Fig. S7) with a p-value of 0.019 for the meta-regression model. Reported HRs and CIs were adjusted for both publication bias according to Rücker's shrinkage procedure and design bias (Supplementary Table S5). Meta-analysis of these bias-adjusted HRs yielded a bias-adjusted pooled HR of 1.06 (0.96; 1.16) (Fig. 3b).

      4. Discussion

      Our study adds information to the safety evaluation of PPIs, a question of high clinical relevance [
      • Manolis A.A.
      • Manolis T.A.
      • Melita H.
      • Katsiki N.
      • Manolis A.S.
      Proton pump inhibitors and cardiovascular adverse effects: real or surreal worries?.
      ,
      • Veettil S.K.
      • Sadoyu S.
      • Bald E.M.
      • Chandran V.P.
      • Khuu S.A.T.
      • Pitak P.
      • et al.
      Association of proton-pump inhibitor use with adverse health outcomes: a systematic umbrella review of meta-analyses of cohort studies and randomised controlled trials.
      ], as PPIs are amongst the most frequently used medications [
      • Rückert-Eheberg I.M.
      • Nolde M.
      • Ahn N.
      • Tauscher M.
      • Gerlach R.
      • Güntner F.
      • et al.
      Who gets prescriptions for proton pump inhibitors and why? A drug-utilization study with claims data in Bavaria, Germany, 2010–2018.
      ]. This meta-analysis combined ten studies on cardiovascular events including 24,547 cases of first ischaemic stroke and 50,824 cases of first myocardial infarction, as well as seven studies on cardiovascular mortality with 50,329 cardiovascular deaths in total. The pooled HRs for PPI use and cardiovascular events were 1.05 (0.96; 1.15) before and 0.99 (0.93; 1.04) after adjusting for observational study design bias. The pooled HRs for PPI use and cardiovascular mortality were 1.27 (1.11; 1.44) before and 1.06 (0.96; 1.16) after adjusting for publication bias and observational study design bias.
      An effect of PPI intake on cardiovascular events has been discussed for more than a decade. PPIs, especially omeprazole, seem to attenuate clopidogrel's antiplatelet effects by inhibiting CYP2C19, which metabolises clopidogrel to its active metabolites [
      • Pang J.
      • Wu Q.
      • Zhang Z.
      • Zheng T.Z.
      • Xiang Q.
      • Zhang P.
      • et al.
      Efficacy and safety of clopidogrel only vs. clopidogrel added proton pump inhibitors in the treatment of patients with coronary heart disease after percutaneous coronary intervention: a systematic review and meta-analysis.
      ]. Besides that, several mechanisms have been suggested, by which PPIs might directly affect cardiovascular risk via impaired vascular endothelial function [
      • Ghebremariam Y.T.
      • LePendu P.
      • Lee J.C.
      • Erlanson D.A.
      • Slaviero A.
      • Shah N.H.
      • et al.
      Unexpected effect of proton pump inhibitors: elevation of the cardiovascular risk factor asymmetric dimethylarginine.
      ,
      • Nolde M.
      • Bahls M.
      • Friedrich N.
      • Dörr M.
      • Dreischulte T.
      • Felix S.B.
      • et al.
      Association of proton pump inhibitor use with endothelial function and metabolites of the nitric oxide pathway: a cross-sectional study.
      ] or accelerated endothelial aging [
      • Yepuri G.
      • Sukhovershin R.
      • Nazari-Shafti T.Z.
      • Petrascheck M.
      • Ghebre Y.T.
      • Cooke J.P.
      Proton pump inhibitors accelerate endothelial senescence.
      ]. Evidence for an effect independent of clopidogrel inhibition was conflicting between randomized trials and observational studies [
      • Batchelor R.
      • Kumar R.
      • Gilmartin-Thomas J.F.M.
      • Hopper I.
      • Kemp W.
      • Liew D.
      Systematic review with meta-analysis: risk of adverse cardiovascular events with proton pump inhibitors independent of clopidogrel.
      ]. While randomized trials showed no differences between PPI users and placebo-users [
      • Moayyedi P.
      • Eikelboom J.W.
      • Bosch J.
      • Connolly S.J.
      • Dyal L.
      • Shestakovska O.
      • et al.
      Safety of proton pump inhibitors based on a large, multi-year, randomized trial of patients receiving rivaroxaban or Aspirin.
      ,
      • Batchelor R.
      • Kumar R.
      • Gilmartin-Thomas J.F.M.
      • Hopper I.
      • Kemp W.
      • Liew D.
      Systematic review with meta-analysis: risk of adverse cardiovascular events with proton pump inhibitors independent of clopidogrel.
      ], observational studies indicated a potentially increased cardiovascular risk for PPI users [
      • Li S.
      • Liu F.
      • Chen C.
      • Zhu W.
      • Ma J.
      • Hu J.
      • et al.
      Real-world relationship between proton pump inhibitors and cerebro-cardiovascular outcomes independent of clopidogrel.
      ,
      • Batchelor R.
      • Kumar R.
      • Gilmartin-Thomas J.F.M.
      • Hopper I.
      • Kemp W.
      • Liew D.
      Systematic review with meta-analysis: risk of adverse cardiovascular events with proton pump inhibitors independent of clopidogrel.
      ]. This seemed concerning, as observational studies are better suited to detect long-term effects and the combined CI of (0.25; 5.73) from randomized trials [
      • Batchelor R.
      • Kumar R.
      • Gilmartin-Thomas J.F.M.
      • Hopper I.
      • Kemp W.
      • Liew D.
      Systematic review with meta-analysis: risk of adverse cardiovascular events with proton pump inhibitors independent of clopidogrel.
      ] could not reject the observational effect estimate.
      Our analyses showed that this discrepancy can be resolved by adjusting for observational study design bias. First, we minimized the effects of clopidogrel inhibition by observing first cardiovascular events only. Second, we analysed each study's risk of bias in detail using the ROBINS-I tool [
      • Sterne J.A.
      • Hernán M.A.
      • Reeves B.C.
      • Savović J.
      • Berkman N.D.
      • Viswanathan M.
      • et al.
      ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions.
      ] and excluded studies with critical risk of bias. Finally, we adjusted individual study estimates for bias introduced by study design choices and combined the adjusted estimates to the pooled bias-adjusted HR of 0.99 (0.93; 1.04), which coincides with estimates from randomized trials [
      • Moayyedi P.
      • Eikelboom J.W.
      • Bosch J.
      • Connolly S.J.
      • Dyal L.
      • Shestakovska O.
      • et al.
      Safety of proton pump inhibitors based on a large, multi-year, randomized trial of patients receiving rivaroxaban or Aspirin.
      ]. Especially, risk of bias in individual studies could have been reduced by applying a new user design, as prevalent PPI therapy might be a sign of a pre-existing cardiovascular condition and the inclusion of prevalent PPI users would therefore introduce indication bias. By design, studies on cardiovascular mortality included patients with prior cardiovascular disease, which increased the potential for bias due to interaction with clopidogrel and confounding by indication. Furthermore, we found considerable publication bias among studies on cardiovascular mortality. After adjusting individual study estimates for publication and study design bias we yielded a bias-adjusted pooled HR of 1.06 (0.96; 1.16) consistent with the analysis of cardiovascular events.
      Although we did not find an overall effect of PPI therapy on the risk of cardiovascular events, the subgroup analysis revealed that the two studies [
      • Shih C.J.
      • Chen Y.T.
      • Ou S.M.
      • Li S.Y.
      • Chen T.J.
      • Wang SJ.
      Proton pump inhibitor use represents an independent risk factor for myocardial infarction.
      ,
      • Wang Y.F.
      • Chen Y.T.
      • Luo J.C.
      • Chen T.J.
      • Wu J.C.
      • Wang S.J.
      Proton-pump inhibitor use and the risk of first-time ischemic stroke in the general population: a nationwide population-based study.
      ] in Asian populations reported substantially higher effect estimates than studies from other regions. Unfortunately, the number of studies was not sufficient to decide, whether this is pure coincidence or actual effect modification.
      The limitations of this meta-analysis stem mostly from the limitations of the data used in the individual studies. Especially, exposure to PPI therapy was usually identified using dispensed prescriptions and use of over-the-counter medications and combination products was not captured. PPI therapy was considered a point treatment. Effects of long-term intake or cumulative dose-dependent effects were not accessible. Long-term and high-dose users of PPIs were part of the analyses, but their cardiovascular risk might have been diluted by mostly short-term and low-dose PPI users.
      In conclusion, this qualitative and quantitative synthesis of all available prospective observational studies suggests that PPI intake as a limited treatment of gastroesophageal diseases does not increase the risk of first cardiovascular events. Reports of increased cardiovascular mortality can largely be explained by publication bias and observational study design biases, such as indication bias and unmeasured confounding. In combination with results from randomized trials it seems therefore questionable, whether PPI intake constitutes a cardiovascular risk factor independent of any possible interaction with clopidogrel. Further studies might investigate the cardiovascular risk of PPI therapy in Asian populations.

      5. Registration and protocol

      This review was registered at the PROSPERO database (CRD42020197513). It was designed and conducted in accordance with the Cochrane Handbook of Systematic Reviews [
      • Higgins J.P.T.
      • Green S.
      Cochrane handbook for systematic reviews of interventions.
      ] and has been authored according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guideline [
      • Page M.J.
      • McKenzie J.E.
      • Bossuyt P.M.
      • Boutron I.
      • Hoffmann T.C.
      • Mulrow C.D.
      • et al.
      The PRISMA 2020 statement: an updated guideline for reporting systematic reviews.
      ].

      Sources of funding

      This study was supported by the Innovation Committee at the Federal Joint Committee (Gemeinsamer Bundesausschuss, G-BA), the highest decision-making body of the joint self-government of physicians, dentists, hospitals, and health insurance funds in Germany (grant no. 01VSF18013).

      Availability of data and materials

      Data derived from public domain resources

      Declaration of Competing Interest

      UA is member of the advisory board of Vetter Pharma-Fertigung GmbH & Co. KG, Ravensburg, Germany.

      Acknowledgments

      We thank Jeremy P. Brown for disclosing the number of events in his study [
      • Brown J.P.
      • Tazare J.R.
      • Williamson E.
      • Mansfield K.E.
      • Evans S.J.
      • Tomlinson L.A.
      • et al.
      Proton pump inhibitors and risk of all-cause and cause-specific mortality: a cohort study.
      ] upon request.

      Appendix. Supplementary materials

      References

        • Rückert-Eheberg I.M.
        • Nolde M.
        • Ahn N.
        • Tauscher M.
        • Gerlach R.
        • Güntner F.
        • et al.
        Who gets prescriptions for proton pump inhibitors and why? A drug-utilization study with claims data in Bavaria, Germany, 2010–2018.
        Eur J Clin Pharmacol. 2021; https://doi.org/10.1007/s00228-021-03257-z
        • Pang J.
        • Wu Q.
        • Zhang Z.
        • Zheng T.Z.
        • Xiang Q.
        • Zhang P.
        • et al.
        Efficacy and safety of clopidogrel only vs. clopidogrel added proton pump inhibitors in the treatment of patients with coronary heart disease after percutaneous coronary intervention: a systematic review and meta-analysis.
        Int J Cardiol Heart Vasc. 2019; 23100317https://doi.org/10.1016/j.ijcha.2018.12.016
        • Moayyedi P.
        • Eikelboom J.W.
        • Bosch J.
        • Connolly S.J.
        • Dyal L.
        • Shestakovska O.
        • et al.
        Safety of proton pump inhibitors based on a large, multi-year, randomized trial of patients receiving rivaroxaban or Aspirin.
        Gastroenterology. 2019; 157 (e2.): 682-691https://doi.org/10.1053/j.gastro.2019.05.056
        • Li S.
        • Liu F.
        • Chen C.
        • Zhu W.
        • Ma J.
        • Hu J.
        • et al.
        Real-world relationship between proton pump inhibitors and cerebro-cardiovascular outcomes independent of clopidogrel.
        Int Heart J. 2019; 60: 910-918https://doi.org/10.1536/ihj.18-584
        • Batchelor R.
        • Kumar R.
        • Gilmartin-Thomas J.F.M.
        • Hopper I.
        • Kemp W.
        • Liew D.
        Systematic review with meta-analysis: risk of adverse cardiovascular events with proton pump inhibitors independent of clopidogrel.
        Aliment Pharmacol Ther. 2018; 48: 780-796https://doi.org/10.1111/apt.14955
        • Farhat N.
        • Fortin Y.
        • Haddad N.
        • Birkett N.
        • Mattison D.R.
        • Momoli F.
        • et al.
        Systematic review and meta-analysis of adverse cardiovascular events associated with proton pump inhibitors used alone or in combination with antiplatelet agents.
        Crit Rev Toxicol. 2019; 49: 215-261https://doi.org/10.1080/10408444.2019.1583167
        • Shiraev T.P.
        • Bullen A.
        Proton pump inhibitors and cardiovascular events: a systematic review.
        Heart Lung Circ. 2018; 27: 443-450https://doi.org/10.1016/j.hlc.2017.10.020
        • Sterne J.A.
        • Hernán M.A.
        • Reeves B.C.
        • Savović J.
        • Berkman N.D.
        • Viswanathan M.
        • et al.
        ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions.
        BMJ. 2016; 355: i4919https://doi.org/10.1136/bmj.i4919
        • Franklin J.M.
        • Patorno E.
        • Desai R.J.
        • Glynn R.J.
        • Martin D.
        • Quinto K.
        • et al.
        Emulating randomized clinical trials with nonrandomized real-world evidence studies: first results from the RCT DUPLICATE initiative.
        Circulation. 2021; 143: 1002-1013https://doi.org/10.1161/CIRCULATIONAHA.120.051718
        • Lund J.L.
        • Richardson D.B.
        • Stürmer T.
        The active comparator, new user study design in pharmacoepidemiology: historical foundations and contemporary application.
        Curr Epidemiol Rep. 2015; 2: 221-228https://doi.org/10.1007/s40471-015-0053-5
        • Hernán M.A.
        • Robins JM.
        Using big data to emulate a target trial when a randomized trial is not available.
        Am J Epidemiol. 2016; 183: 758-764https://doi.org/10.1093/aje/kwv254
        • Brown J.P.
        • Tazare J.R.
        • Williamson E.
        • Mansfield K.E.
        • Evans S.J.
        • Tomlinson L.A.
        • et al.
        Proton pump inhibitors and risk of all-cause and cause-specific mortality: a cohort study.
        Br J Clin Pharmacol. 2021; https://doi.org/10.1111/bcp.14728
        • Adelborg K.
        • Sundbøll J.
        • Schmidt M.
        • Bøtker H.E.
        • Weiss N.S.
        • Pedersen L.
        • Sørensen HT.
        Use of histamine H(2) receptor antagonists and outcomes in patients with heart failure: a nationwide population-based cohort study.
        Clin Epidemiol. 2018; 10: 521-530https://doi.org/10.2147/CLEP.S162909
        • Veroniki A.A.
        • Jackson D.
        • Bender R.
        • Kuss O.
        • Langan D.
        • Higgins J.P.T.
        • et al.
        Methods to calculate uncertainty in the estimated overall effect size from a random-effects meta-analysis.
        Res Synth Methods. 2019; 10: 23-43https://doi.org/10.1002/jrsm.1319
        • IntHout J.
        • Ioannidis J.P.A.
        • Borm GF.
        The Hartung-Knapp-Sidik-Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian-Laird method.
        BMC Med Res Methodol. 2014; 14: 25https://doi.org/10.1186/1471-2288-14-25
        • IntHout J.
        • Ioannidis J.P.A.
        • Rovers M.M.
        • Goeman JJ.
        Plea for routinely presenting prediction intervals in meta-analysis.
        BMJ Open. 2016; 6e010247https://doi.org/10.1136/bmjopen-2015-010247
        • Higgins J.P.T.
        • Thompson S.G.
        • Spiegelhalter D.J.
        A re-evaluation of random-effects meta-analysis.
        J R Stat Soc Ser A Stat Soc. 2009; 172: 137-159https://doi.org/10.1111/j.1467-985X.2008.00552.x
        • Debray T.P.A.
        • Moons K.G.M.
        • Riley R.D.
        Detecting small-study effects and funnel plot asymmetry in meta-analysis of survival data: a comparison of new and existing tests.
        Res Synth Methods. 2018; 9: 41-50https://doi.org/10.1002/jrsm.1266
        • Schwarzer G.
        • Carpenter J.R.
        • Rücker G.
        Small-study effects in meta-analysis.
        Meta-Analysis with R: Springer, 2015: 107-141https://doi.org/10.1007/978-3-319-21416-0_5 (Cham)
        • Egger M.
        • Schneider M.D.
        • Smith G.
        Spurious precision? Meta-analysis of observational studies.
        BMJ. 1998; 316: 140-144https://doi.org/10.1136/bmj.316.7125.140
        • Baujat B.
        • Mahé C.
        • Pignon J.P.
        • Hill C.
        A graphical method for exploring heterogeneity in meta-analyses: application to a meta-analysis of 65 trials.
        Stat Med. 2002; 21: 2641-2652https://doi.org/10.1002/sim.1221
        • Schmid C.H.
        • Stijnen T.
        • White I.R.
        • White I.
        Handbook of Meta-Analysis.
        First edition. | Boca raton : taylor and Francis, [2020]| Series: chapman & Hall/CRC handbooks of modern statistical methods. Chapman and Hall/CRC, 2020
        • Higgins J.P.T.
        • Thompson S.G.
        Controlling the risk of spurious findings from meta-regression.
        Stat Med. 2004; 23: 1663-1682https://doi.org/10.1002/sim.1752
        • Mathur M.B.
        • VanderWeele T.J.
        Methods to address confounding and other biases in meta-analyses: review and recommendations.
        Annu Rev Public Health. 2021; https://doi.org/10.1146/annurev-publhealth-051920-114020
        • Nolde M.
        • Ahn N.
        • Dreischulte T.
        • Rückert-Eheberg I.M.
        • Güntner F.
        • Günter A.
        • et al.
        The long-term risk for myocardial infarction or stroke after proton pump inhibitor therapy (2008–2018).
        Aliment Pharmacol Ther. 2021; 54: 1033-1040https://doi.org/10.1111/apt.16565
        • Sehested T.S.G.
        • Gerds T.A.
        • Fosbøl E.L.
        • Hansen P.W.
        • Charlot M.G.
        • Carlson N.
        • et al.
        Long-term use of proton pump inhibitors, dose-response relationship and associated risk of ischemic stroke and myocardial infarction.
        J Intern Med (GBR). 2018; 283: 268-281https://doi.org/10.1111/joim.12698
        • Rooney M.R.
        • Bell E.J.
        • Alonso A.
        • Pankow J.S.
        • Demmer R.T.
        • Rudser K.D.
        • et al.
        Proton pump inhibitor use, hypomagnesemia and risk of cardiovascular diseases: the atherosclerosis risk in communities (ARIC) study.
        J Clin Gastroenterol. 2021; 55: 677-683https://doi.org/10.1097/MCG.0000000000001420
        • Landi S.N.
        • Sandler R.S.
        • Pate V.
        • Lund J.L.
        No increase in risk of acute myocardial infarction in privately insured adults prescribed proton pump inhibitors vs histamine-2 receptor antagonists (2002–2014).
        Gastroenterology. 2018; 154 (e6.): 861-873https://doi.org/10.1053/j.gastro.2017.10.042
        • Shih C.J.
        • Chen Y.T.
        • Ou S.M.
        • Li S.Y.
        • Chen T.J.
        • Wang SJ.
        Proton pump inhibitor use represents an independent risk factor for myocardial infarction.
        Int. J. Cardiol. 2014; 177: 292-297https://doi.org/10.1016/j.ijcard.2014.09.036
        • Nguyen L.H.
        • Lochhead P.
        • Joshi A.D.
        • Cao Y.
        • Ma W.
        • Khalili H.
        • et al.
        No significant association between proton pump inhibitor use and risk of stroke after adjustment for lifestyle factors and indication.
        Gastroenterology. 2018; 154 (e1.): 1290-1297https://doi.org/10.1053/j.gastro.2017.12.006
        • Wang Y.F.
        • Chen Y.T.
        • Luo J.C.
        • Chen T.J.
        • Wu J.C.
        • Wang S.J.
        Proton-pump inhibitor use and the risk of first-time ischemic stroke in the general population: a nationwide population-based study.
        Am J Gastroenterol. 2017; 112: 1084-1093https://doi.org/10.1038/ajg.2017.101
        • de Francisco
        • Angel L.M.
        • Varas J.
        • Ramos R.
        • Merello J.I.
        • Canaud B.
        • Stuard S.
        • et al.
        Proton pump inhibitor usage and the risk of mortality in hemodialysis patients.
        Kidney Intl Rep. 2018; 3: 374-384https://doi.org/10.1016/j.ekir.2017.11.001
        • He Q.
        • Xia B.
        • Meng W.
        • Fan D.
        • Kuo Z.C.
        • Huang J.
        • et al.
        No associations between regular use of proton pump inhibitors and risk of all-cause and cause-specific mortality: a population-based cohort of 0.44 million participants.
        Am J Gastroenterol. 2021; https://doi.org/10.14309/ajg.0000000000001377
        • Shah N.H.
        • LePendu P.
        • Bauer-Mehren A.
        • Ghebremariam Y.T.
        • Iyer S.V.
        • Marcus J.
        • et al.
        Proton pump inhibitor usage and the risk of myocardial infarction in the general population.
        PLOS ONE. 2015; 10e0124653https://doi.org/10.1371/journal.pone.0124653
        • Xie Y.
        • Bowe B.
        • Yan Y.
        • Xian H.
        • Li T.
        • Al-Aly Z.
        Estimates of all cause mortality and cause specific mortality associated with proton pump inhibitors among US veterans: cohort study.
        BMJ. 2019; 365: l1580https://doi.org/10.1136/bmj.l1580
        • Manolis A.A.
        • Manolis T.A.
        • Melita H.
        • Katsiki N.
        • Manolis A.S.
        Proton pump inhibitors and cardiovascular adverse effects: real or surreal worries?.
        Eur J Intern Med. 2020; 72: 15-26https://doi.org/10.1016/j.ejim.2019.11.017
        • Veettil S.K.
        • Sadoyu S.
        • Bald E.M.
        • Chandran V.P.
        • Khuu S.A.T.
        • Pitak P.
        • et al.
        Association of proton-pump inhibitor use with adverse health outcomes: a systematic umbrella review of meta-analyses of cohort studies and randomised controlled trials.
        Br J Clin Pharmacol. 2021; https://doi.org/10.1111/bcp.15103
        • Ghebremariam Y.T.
        • LePendu P.
        • Lee J.C.
        • Erlanson D.A.
        • Slaviero A.
        • Shah N.H.
        • et al.
        Unexpected effect of proton pump inhibitors: elevation of the cardiovascular risk factor asymmetric dimethylarginine.
        Circulation. 2013; 128: 845-853https://doi.org/10.1161/CIRCULATIONAHA.113.003602
        • Nolde M.
        • Bahls M.
        • Friedrich N.
        • Dörr M.
        • Dreischulte T.
        • Felix S.B.
        • et al.
        Association of proton pump inhibitor use with endothelial function and metabolites of the nitric oxide pathway: a cross-sectional study.
        Pharmacother J Hum Pharmacol Drug Ther. 2021; 41: 198-204https://doi.org/10.1002/phar.2504
        • Yepuri G.
        • Sukhovershin R.
        • Nazari-Shafti T.Z.
        • Petrascheck M.
        • Ghebre Y.T.
        • Cooke J.P.
        Proton pump inhibitors accelerate endothelial senescence.
        Circ Res. 2016; 118: e36-e42https://doi.org/10.1161/CIRCRESAHA.116.308807
        • Higgins J.P.T.
        • Green S.
        Cochrane handbook for systematic reviews of interventions.
        Wiley-Blackwell, Chichester England, Hoboken NJ2008
        • Page M.J.
        • McKenzie J.E.
        • Bossuyt P.M.
        • Boutron I.
        • Hoffmann T.C.
        • Mulrow C.D.
        • et al.
        The PRISMA 2020 statement: an updated guideline for reporting systematic reviews.
        BMJ. 2021; 372: n71https://doi.org/10.1136/bmj.n71