Meta-analysis: The Effects of Proton Pump Inhibitors on Cardiovascular Events and Mortality in Patients Receiving Clopidogrel

C. S. Kwok; Y. K. Loke

Aliment Pharmacol Ther. 2010;31(8):810-823.

Abstract and Introduction

Abstract

Background Recent studies have suggested an adverse interaction between proton pump inhibitors (PPI) and clopidogrel.
Aim To perform a meta-analysis of cardiovascular outcomes and mortality in patients taking clopidogrel, with and without concomitant PPI.
Methods We searched MEDLINE, EMBASE, Cochrane Controlled Trials Register in October 2009, and checked conference abstracts for randomized and nonrandomized studies that reported the risk of cardiovascular events and mortality with PPI exposure in patients taking clopidogrel. We performed random effects meta-analysis, stratified by study design and assessed heterogeneity using the I ² statistic.
Results Our review included 23 studies covering 93 278 patients. There was substantial heterogeneity in the meta-analyses of major cardiovascular events (19 studies, I ² = 79%) or myocardial infarction (12 studies, I ² = 77%). Analysis of propensity-matched or randomized trial participants showed no associated cardiovascular risk with PPIs, whereas other observational studies generally showed a significant association. Meta-analysis of 13 studies showed no significant association between PPI use and overall mortality (RR 1.09, 95% CI: 0.94–1.26, P = 0.23, I ² = 60%).
Conclusion As there are conflicting and inconsistent data regarding the adverse clopidogrel–PPI interaction, clinicians should focus on potential harm from ulcers/haemorrhage before deciding to omit PPIs in patients taking clopidogrel.

Introduction

There is currently significant controversy regarding cardiovascular (CV) adverse events arising from a potential drug interaction with concomitant use of clopidogrel and proton pump inhibitors (PPI).^[1–3] On one hand, a number of recent commentaries from experts in the field have argued that there is no evidence that this potential drug interaction carries any major clinical impact.^[4,5] In contrast, the US Food and Drug Administration (FDA) recently issued an updated statement cautioning against concomitant clopidogrel and PPI use^[6] and this stance appears to be shared by some other researchers who believe that there is a genuine clinical problem.^[7]

Given that there has now been a rapid profusion of diverse study designs reporting widely varying results, it seems likely that different opinions may be formed depending on the selection of studies that one has access to. In this complex situation, an up-to-date systematic review of all currently available data may help clarify things. Moreover, when data from many different types of studies are available, the Cochrane Adverse Methods Group recommends that such an analysis can be usefully stratified according to type of study design.^[8] This is particularly important when the risk or level of bias varies with the methods and design used in particular studies.

As such, the aim of our meta-analysis was to determine the effect of concomitant PPIs and clopidogrel use on CV outcomes and mortality, stratified by study design.

1 2 3 4 5

Next Section

Cite this: Meta-analysis: The Effects of Proton Pump Inhibitors on Cardiovascular Events and Mortality in Patients Receiving Clopidogrel - Medscape - Apr 01, 2010.

Abstract and Introduction
Methods
Results
Discussion
Conclusions
References

Table 1. Study design and characteristics
Study	Design; country and setting	No. patients	Mean age	% Male	Selection criteria	Outcome and follow-up
Banjerjee et al.^[12]	Retrospective cohort study in the US	197	NS	96.7	Patients on clopidogrel following insertion of drug eluting stent	MACE at 582 days
Bhatt et al.^[13]	Prospective, double-blind placebo- controlled multicentre RCT	3627	67.2	70.7	Age >21 years with ACS or coronary stent requiring clopidogrel and aspirin for next 12 months Excluded those with significant GI or bleeding history, current use of gastroprotective drugs or anticoagulants	MACE, MI for median of 133 days
Ching et al.^[14]	Retrospective cohort study in the US	1192	N.S.	N.S.	Patients with PCI	MACE, mortality at 9 months
Dunn et al.^[15]	Post hoc analysis of patients randomized to 12 months of clopidogrel & aspirin in US RCT	1053	61.5	71	Patients undergoing PCI. Excluded those with contraindications to antiplatelets and anticoagulants	MACE at 1 year
Gaspar et al.^[16]	Retrospective cohort study in Portugal	922	N.S.	N.S.	Patients admitted with ACS discharged with aspirin and clopidogrel	MACE at 6 months
Gupta et al.^[17]	Retrospective study at a single hospital in the US	315	62	N.S.	All PCI patients who were discharged on clopidogrel	MACE and mortality in 4 years
Hall et al.^[18]	Retrospective cohort study using healthcare database in the US	10 703	60	62.5	Patients who had received coronary stents and registered in cardiac catheterization database, Intermountain Healthcare	MACE, MI, mortality at 1 year
Ho et al.^[19]	Retrospective cohort and nested case control study in 127 hospitals in the US	8205	67	98.5	All patient with acute MI or unstable angina discharged with clopidogrel	MACE and mortality between Oct 2003 and Jan 2006
Jarai et al.^[20]	Retrospective cohort study in Austria	1082 (propensity matched arm)	N.S.	N.S.	Patients on clopidogrel following PCI. Propensity score analysis was used in matching PPI users and non-users	MACE and death at 723 days
Juurlink et al.^[21]	Nested-case control study in Ontario, Canada	2791	77 median	54	Patients age >65 who filled prescription of clopidogrel after hospital discharge for MI. Excluded patients who received clopidogrel, ticlopidine or dipyridamole before admission and those in long-term care facilities	MI and mortality up to 3 months
O'Donoghue et al.^[22]	Double-blind RCT in hospitals in the US, Europe and other parts of world	6795	60.5	73.9	Patients undergoing PCI who were randomized to the clopidogrel arm or the trial. Excluded those with risk of bleeding, anaemia, thrombocytopenia, pathological intracranial finding or thienopyridine within 5 days before randomization	MACE, MI and mortality
Pezalla et al.^[23]	Retrospective cohort study in the US	1010	57.7	67.3	Patients <65 who were adherent to clopidogrel therapy; underlying diagnoses were not stated	MI only at 1 year
Ramirez et al.^[24]	Retrospective cohort study based on registry data of hospital patients in Pittsburgh, US	535	N.S.	N.S.	Patients on clopidogrel following PCI	Death, MI, MACE (MI/death) at 1 year
Rassen et al.^[25]	Retrospective cohort study based on 3 insurance databases in the US and Canada, covering patients hospitalized for ACS	18 565	76.1	48.2	Patient on clopidogrel following PCI for ACS, propensity-matched	MI, MACE (MI/death), death at 6 months
Sarafoff et al.^[26]	Retrospective cohort study in Germany	2025	66.8	N.S.	Patients on clopidogrel randomly selected from database of those who had undergone PCI	Death, MI, MACE (MI/death) at 1 year
Simon et al.^[27] and Simon et al.^[28]	Cohort study based on French Registry of patients post-MI	2208, with Simon et al.^[28] = 2178	66.2	70.6	Patients age >18 with serum markers of AMI or ECG pathological Q waves, ST elevation, ST depression	MACE at 1 year;^[27] Death at 1 year^[28]
Stanek et al.^[29]	Retrospective cohort study in the US based on medical and pharmacy claims database	16 690	66	68.9	Patient with clopidogrel treatments following coronary stenting	MACE at 1 year
Stockl et al.^[30]	Retrospective cohort study of patients enrolled in health insurance plan in the US	2066	N.S.	N.S.	Patents discharged from hospital with clopidogrel after MI or coronary stent placement, matched 1:1 for CV risk with propensity scoring method between patients with and without PPI	MI up to 360 days
Sweeny et al.^[31]	Retrospective cohort study based on NY state interventional database	8311	N.S.	N.S.	Patients undergoing PCI with drug eluting stents during study period	Mortality at over 2 years
Torgusen et al.^[32]	Retrospective cohort study in the US	1896	N.S.	N.S.	Patients undergoing PCI with drug-eluting stents discharged with aspirin and clopidogrel	MACE at 12 months
Yasuda et al.^[33]	Retrospective cohort study in Japan	243	68	75.3	Consecutive patients who had undergone PCI. Excluded if cardiogenic death within 1 week, malignancy, chronic haemodialysis and patients on warfarin or single platelet therapy, transferred to other hospitals	MI, unclear follow-up period
Zairis et al.^[34]	Post hoc analysis of prospective cohort study in Greece	612	N.S.	N.S.	Patients with coronary stenting for stable or unstable coronary disease	MI and mortality at 1 year

MACE, major adverse cardiac events; PCI, percutaneous coronary intervention; CV, cardiovascular; MI, myocardial infarction; ACS, acute coronary syndrome; PPI, proton pump inhibitor.

Table 2. Study outcomes, interventions and quality
Study	PPI exposure and ascertainment	Lost to follow-up	Definition of outcome and ascertainment	Risk of bias
Banjerjee et al.^[12]	Any PPI, no other details	Not specified	MACE defined as death, MI, revascularization and stroke. Monitoring not stated	Abstract only, no further details
Bhatt et al.^[13]	Omeprazole	Not specified	MACE was prespecified composite of CV death, nonfatal MI, CABG or PCI, or ischaemic stroke, adjudicated by independent committee	Abstract only, low risk of confounding as there were no baseline differences
Ching et al.^[14]	Esomeprazole, lansoprazole, omeprazole, pantoprazole	None	MACE defined as recurrent MI, revascularization and death. Monitoring not stated	Abstract only, analysis was adjusted for age, gender, glycoprotein IIb/IIIa inhibitors and stent size; no further details
Dunn et al.^[15]	Any PPI, no other details	38 (3.6%) losses	MACE was prespecified as death, MI or stroke and adjudicated by blinded independent committee	Despite adjustment for confounding, PPI exposure alone without clopidogrel still showed significantly increased risk above that of patients taking neither PPI nor clopidogrel
Gaspar et al.^[16]	Omeprazole, rabeprazole and lansoprazole were considered (pantoprazole excluded as investigators judged it to be metabolized by another route). Prescription and clinical records used to define exposure	None	Not specified	Abstract only, baseline differences were present, multivariate analysis was performed to assess effect of PPI exposure. Authors concede that compliance was not assessed and prescription records may be incomplete
Gupta et al.^[17]	Rabeprazole, omeprazole and lansoprazole for unclear duration, data obtained from discharge summary	None. Mean follow-up of 50 months	Reviewed medical records. MACE defined as composite of death, nonfatal MI and target vessel failure	Risk of confounding as baseline differences were present; adjusted risk ratios were presented. Duration of exposure and treatment compliance unclear
Hall et al.^[18]	Omeprazole, lanosprazole, pantoprazole	Not specified	MACE defined as death/MI, monitoring not stated	Abstract only; adjusted for confounders such as comorbid conditions and concomitant medications. Despite adjustment, PPI exposure alone without clopidogrel still showed significantly increased risk above that of patients taking neither PPI nor clopidogrel, thus suggesting possibility of residual confounding
Ho et al.^[19]	Omeprazole, rabeprazole, lansoprazole and pantoprazole for unclear duration based on pharmacy dispensing records	Vital status available for all between October 2003 and September 2006, with median follow-up of 521 days	A vital status file and chart review (including ICD-9 codes) using Veterans Administration database for mortality and MACE (death or rehospitalization for ACS)	Risk of confounding as baseline differences were present; this was adjusted for using multivariate logistic regression
Jarai et al.^[20]	Any PPI, no other details	Not specified	MACE consisted of death or target vessel revascularization. Monitoring not stated	Abstract only, no information on exposure and outcome ascertainment, propensity score was used in patient matching
Juurlink et al.^[21]	Pantoprazole and any other PPI, on Ontario drug prescriptions computerized record	None	Outcomes checked on Health Insurance (using ICD codes) and Health Registered Person databases	Data were adjusted for demographic differences, concomitant medication and comorbidity. Authors state limitations that include possible miscoding of outcome, and inability to verify compliance with exposure
O'Donoghue et al.^[22]	Omeprazole, pantoprazole, esomeprazole, lansoprazole at time or randomization, and at various intervals of follow-up	None	MACE defined as CV death, nonfatal MI, or nonfatal stroke. Independent committee adjudicated events	Propensity score was used to adjust for covariates. Variations in PPI exposure with time were checked in sensitivity analysis – no material effect on results was detected
Pezalla et al.^[23]	Any PPI use ascertained through pharmacy claims database	None	Health insurance and pharmacy database, using ICD codes for MI	Potential confounding as baseline differences were present, and authors attempted to correct for only a few comorbid factors
Ramirez et al.^[24]	Any PPI, no other details	None	Not specified	Abstract only, no information on intervention and outcome ascertainment, risk of confounding as potential baseline differences were present
Rassen et al.^[25]	Omeprazole, esomeprazole, lansoprazole, pantoprazole, or rabeprazole, recorded on pharmacy database	82% were eligible for matched propensity analysis, median follow-up was 29–30 days, with 2% completing maximum follow-up of 180 days	Study endpoints were ascertained up to 180 days of follow-up. MI recorded via discharge coding in health insurance database; death through vital statistics and government agencies	Propensity score was used to match patients
Sarafoff et al.^[26]	Any PPI, no other details	None	Not specified	High; abstract only, no information on intervention and outcome ascertainment, risk of confounding as baseline differences were present
Simon et al.^[27] and Simon et al.^[28]	Omeprazole and PPI for unspecified duration	222 were not included	Investigators contacted patients' physicians, the patients or their family and registry offices at places of birth. Blinded independent committee adjudicated events, endpoints ascertained at 1 year	Uncertain about ascertainment of PPI exposure; but there was blinding of outcome ascertainment
Stanek et al.^[29]	Omeprazole, esomeprazole, pantoprazole and lansoprazole based on prescription claims	Not specified	Outcomes recorded via ICD-9 codes on claims database	Abstract only, no information on compliance. Risk of confounding was addressed in multivariate analysis
Stockl et al.^[30]	Any PPI, no other details	None	Not specified	Abstract only, patients were propensity score-matched based on CV risk; no further details
Sweeny et al.^[31]	Esomeprazole, lansoprazole, omeprazole and pantoprazole	37 excluded from analysis, mean follow-up of 2 years	New York State Interventional database and Social Security Death Index	Abstract only, outcome ascertainment was done, no information on checking of drug exposure, but multivariate adjustment was done for confounders
Torgusen et al.^[32]	Any PPI, no other details	None	MACE defined as death, MI or revascularization. Monitoring not stated	Abstract only, author stated that PPI-exposed patients were more complex and sicker than unexposed, and used multivariate regression to adjust
Yasuda et al.^[33]	Lansoprazole, omeprazole, rabeprazole for unspecified duration; review of hospital records	None	Review of hospital records and coronary angiography at 6–16 months after stent placement	Intervention exposure and outcomes were not ascertained, risk of confounding as baseline differences were present
Zairis et al^[34]	Any PPI, no other details	None	Not specified	High, abstract only, no information on intervention and outcome ascertainment, low risk of confounding because of baseline difference

MACE, major adverse cardiac events; ICD, International Classification of Diseases; CV, cardiovascular; MI, myocardial infarction; ACS, acute coronary syndrome; PPI, proton pump inhibitor.