1. Background

There is considerable controversy regarding the risk of cardiovascular adverse events from a potential interaction between clopidogrel and proton pump inhibitors (PPIs).[1] The earliest studies were based on platelet function tests,[2,3] which were subsequently followed by studies evaluating clinical outcomes such as myocardial infarction, major cardiac adverse events and death.[4–7] However, our previous meta-analysis of cardiovascular outcomes found substantial heterogeneity and inconsistent risk amongst the studies, thus precluding any definitive conclusions on the clinical implications of concomitant clopidogrel and PPI use.[8] It is not clear if significant heterogeneity in clinical outcomes stems from poor correlation with platelet function tests, or whether there is considerable variability in the extent to which PPIs affect clopidogrel.

In November 2009, the US FDA released warnings specifically against the concomitant use of omeprazole or esomeprazole with clopidogrel, based on pharmacokinetic and platelet function data.[9] Interestingly, the FDA also stated that they did not have sufficient information on the impact of other PPIs. Nevertheless, it has been suggested that pantoprazole may be a safer option than omeprazole, based on randomized trial data where omeprazole significantly affected the antiplatelet effect of clopidogrel, whereas pantoprazole did not.[2] Given the multitude of platelet function tests, extent of individual variability and complexity of the platelet regulation pathways, the consistency of the entire dataset should be considered before deciding on one PPI or another. As such, the aim of this systematic review is to synthesize all available evidence regarding platelet function studies and the postulated interaction between clopidogrel and PPIs, focusing particularly on omeprazole, esomeprazole and pantoprazole.

2. Methods

2.1 Eligibility Criteria

We selected randomized controlled trials (RCTs) and controlled observational studies that reported laboratory values of platelet function in patients receiving clopidogrel, with and without concomitant PPI exposure.

The specific inclusion criteria for RCTs were (i) adult participants with clopidogrel exposure; (ii) randomized to treatment with PPI; (iii) control arm that was either placebo or alternative PPI; (iv) parallel group or crossover design; and (v) reporting of the quantitative data from platelet function studies.

For the observational studies, we selected non-randomized (prospective or retrospective) studies that evaluated platelet function with concomitant clopidogrel and any PPI exposure, as opposed to clopidogrel alone, or clopidogrel with different PPIs in adult patients.

We did not impose any restrictions on the type of platelet function test nor the co-morbidities of participants. However, we included only full journal publications, because the limited length of conference abstracts leads to difficulties in fully assessing the validity of the methods and results.

2.2 Search Strategy

We searched MEDLINE, EMBASE and the Cochrane Central Register of Controlled Trials with no date limitations in February 2011, using a combination of free text and indexing search terms [(proton-pump-inhibitor OR proton-pump-inhibitors) AND clopidogrel)] without any language restriction (see Supplemental Digital Content, http://links.adisonline.com/DSZ/A55). Additionally, we signed up with PubMed to receive automated electronic notification for any new articles containing the above search terms. We also checked the bibliographies of included studies and recent review articles for relevant studies.

2.3 Study Selection and Data Extraction

Two reviewers (CSK and YKL) scanned all titles and abstracts for studies that potentially met the inclusion criteria, and excluded any articles that clearly did not fulfil the selection criteria. Full reports (where available) of potentially relevant trials and studies were retrieved and independently checked by the two reviewers. We then independently collected information on study design, drug exposure, study location, characteristics of participants and relevant outcomes onto a spreadsheet. Where there was any uncertainty or discrepancies, the article was discussed between the two reviewers to determine if the studies should be included. We also contacted authors if there were any areas that required clarification.

2.4 Validity Assessment and Evaluation of Study Characteristics

Patient selection, drug exposure, outcome ascertainment (methods used in platelet function tests) and blinding were considered in the assessment of risk of bias. We also looked at additional factors related to study design and methodology including characteristics of recruited participants, loss to follow-up, and dosing regimens.

2.5 Data Analysis

We aimed to perform a meta-analysis of standardized mean differences in platelet reactivity between groups with differing or no PPI exposure. If there was considerable variation in reporting of outcomes amongst studies such that meta-analysis was not straightforward, then we aimed to perform only a narrative synthesis with the main platelet function data presented in tabular format. In view of the diversity of the platelet function tests, we summarized the interpretation of the co-administration of PPIs in the following ways:

  • 1)|‘significant interaction’ where the PPI led to a statistically significant reduction in the antiplatelet effect of clopidogrel;

  • 2)|‘no significant difference’ where the PPI altered (increased or reduced) the antiplatelet effect of clopidogrel but not to a statistically significant extent;

  • 3)|‘enhancement of effect’ where the PPI significantly increased the antiplatelet effect of clopidogrel.

Interpretation of the key platelet function tests and the interaction are summarized below.

  • • Adenosine diphosphate (ADP)-aggregation or maximal platelet aggregation (MPA): increase shows that PPI has a significant interaction with clopidogrel.

  • • Vasodilator-stimulated phosphoprotein (VASP)-platelet reactivity index (PRI): increase shows PPI has a significant interaction with clopidogrel.

  • • VerifyNow® ADP (Accumetrics, San Diego, CA, USA) —platelet reactivity units (PRU): increase with PPI shows a significant interaction with clopidogrel.

  • • Inhibition of platelet aggregation: decrease shows PPI has a significant interaction with clopidogrel.

3. Results

3.1 Study Design and Characteristics

The search yielded 19 relevant studies (from 18 journal articles) with 4693 patients, and the process of selection is shown in figure 1. The studies included consisted of seven RCTs,[2,3,10–14] six crossover trials,[4,15–18] three cohort studies[19–21] and three cross-sectional studies.[22–24] The mean age was 53.0 years and the majority (66.3%) of the participants were men. Most of the studies involved patients who had existing atherosclerotic disease but four studies recruited healthy volunteers,[15,17,18] while one study recruited a mixture of both.[14] The main characteristics of the included studies are described in table I.

Fig. 1
figure 1

Flow diagram of the process of article selection for the systematic review. PPIs = proton pump inhibitors.

Full size image
Table I
figure Tab1

Study design and participant characteristics

Full size image

Five studies reported platelet function results of patients receiving PPIs but did not specify exactly which individual PPIs were used.[4,19,21,23–24] The individual PPIs studied and validity assessment are shown in table II.

Table II
figure Tab2

Validity assessment of the studies

Full size image

3.1.1 Randomization and Blinding of Studies

The quality of the included studies varied. Thirteen studies were controlled clinical trials,[2–4,10–18] with two trials using blinding of participants and investigators.[3,4] One open-label trial stated that details of the sample and patients were concealed from the laboratory physicians.[12] Methods of randomization and allocation concealment were seldom described in the trials, and only two observational studies used statistical adjustment to reduce the risk of confounding.[20,24] While most studies did not comment on blinding of laboratory measures, risk of bias may be low if objective methods were deployed in computing platelet function.

Loss to follow-up was described in four studies, with two studies each reporting 16 patients whose outcomes were not assessed at follow-up,[3,15] while in the other two studies, four and three patients, respectively, were not followed up.[11,13]

3.1.2 Verification of Drug Exposure and Assessment of Outcomes

Verification of drug exposure in the non-randomized studies was particularly problematic, with considerable variation in clopidogrel exposure and very limited data on the dose and duration of the PPIs. Only one trial carried out pill counts to check that the patients were compliant with therapy.[12]

3.2 Comparison of Platelet Function among Individual Proton Pump Inhibitors (PPIs) and Placebo

The actual values of the platelet function tests with respect to different PPIs are shown in table III. Due to the diverse and inconsistent nature of the reporting, we were unable to replicate some of the statistical analyses, and could not proceed with meta-analysis. As such, we provide a summary of the key findings with regards to ome-prazole, esomeprazole and pantoprazole based on the presence or absence of statistically significant interactions.

Table III
figure Tab3

Summary of platelet function results with different categories of PPI exposurea

Full size image

3.2.1 Omeprazole versus no PPI/Placebo Studies

Nine studies evaluated the platelet function of patients on omeprazole compared with placebo.[3,10,11,15,17–20,22] A significant interaction was reported in five studies,[3,11,15,18,22] whereas no significant difference was demonstrated in four studies, possibly because of limited sample sizes.[10,17,19,20]

3.2.2 Pantoprazole versus no PPI/Placebo Studies

Six studies investigated the platelet function effect of treatment with pantoprazole.[10,15,19,20,22,23] No significant difference with pantoprazole was detected in five studies,[10,19,20,22,23] whereas one study[15] found a significant interaction using ADP-aggregation, but not with the VASP-PRI.

3.2.3 Esomeprazole versus no PPI/Placebo Studies

Six studies evaluated the antiplatelet effects of treatment with esomeprazole.[11,12,14,19,22,23] None of the studies were able to demonstrate a statistically significant interaction.

3.2.4 PPIs as a Class versus no PPI/ Placebo Studies

Five studies reported platelet function on a wider group of patients with exposure to PPIs as a class.[4,19,21,23,24] Two of the five suggested that there was a significant interaction,[21,24] whereas the remaining three studies reported nonsignificant differences in platelet activity comparing patients with any PPI exposure versus no PPI exposure.[4,19,23]

3.2.5 Pantoprazole versus Omeprazole/Esomeprazole Studies

There were six studies that evaluated both omeprazole and pantoprazole,[2,10,16,19,20,22] with three studies reporting significantly greater impact on platelet function in omeprazole-treated patients compared with pantoprazole.[2,16,22] In one of these studies, significant interaction was noted using VASP-PRI, but not with MPA testing.[2] Of the three studies reporting no significant difference in platelet function between omeprazole-and pantoprazole-treated patients, two were limited by the small sample size of omeprazole recipients,[19,20] and were thus potentially underpowered to detect a difference. The third study was a randomized trial with 20 patients each in the omeprazole and pantoprazole arms, and we estimated the mean difference in percentage MPA values between the two PPIs to be −0.80% (95% CI −3.82, 2.22).[10] The upper bound of the confidence intervals suggests that omeprazole is unlikely to increase platelet aggregation by more than an absolute value of 2.2% compared with pantoprazole.

Three studies that compared esomeprazole-with pantoprazole-treated patients reported no significant difference in platelet activity between the two PPIs.[19,22,23]

4. Discussion

Our systematic review demonstrates the heterogeneous, inconclusive nature of evidence regarding the impact of PPIs on the antiplatelet efficacy of clopidogrel. Hence, the postulated interaction may not be as firmly established as initially thought. This is particularly true for esomeprazole, for which none of the six platelet-function studies demonstrated a significant interaction, despite the specific warning from the regulatory authority regarding esomeprazole. There is a larger body of evidence implicating omeprazole in a significant interaction with clopidogrel, although there are still a few small studies that failed to consistently demonstrate a significant problem with omeprazole. However, much of the available platelet function data indicate that pantoprazole is less likely to be affected by the interaction, even though differences in platelet function have not always been consistently demonstrated between omeprazole/ esomeprazole compared with pantoprazole in certain studies (possibly due to some studies being underpowered). The correlation between platelet function studies and clinical cardiovascular outcomes with PPIs also remains unclear.

The strengths of our systematic review stem from the comprehensive search and detailed extraction of platelet function results. In contrast, publication bias favouring the early dissemination of striking or significant findings can potentially lead to unreliable conclusions. This is a particular problem if single studies are considered in isolation, rather than as a complete dataset where consistency of evidence can be checked. The considerable heterogeneity in platelet function data demonstrated here is in line with the results of our previous meta-analysis that showed inconclusive results for clinical outcomes.[8]

There are a number of potential explanations for these inconsistent findings. The participants differed considerably amongst studies, ranging from relatively young, healthy volunteers[14,15,17,18] to older patients with stroke or coronary stents, as well as various ethnic groups. There was marked variation in study design, extent of drug exposure and laboratory measures of platelet function tests. Genetic differences in cytochrome P450 (CYP) 2C19 and paraoxonase-1[25,26] (as well as other yet to be identified metabolic pathways) could have an important role. The apparent lack of interaction between clopidogrel and esomeprazole (the S-enantiomer of omeprazole that shares similar metabolic pathways) is a key feature in this systematic review, and raises important questions about the hypothesized interaction and proposed biological mechanisms.

Equally, the validity and generalizability of the platelet function studies are important limitations within our review. One study reported by the FDA was based on healthy volunteers who were given supratherapeutic doses of omeprazole (80mg daily).[15] Detailed verification of actual drug exposure (for both PPIs and clopidogrel) was not routinely carried out except for one study that performed pill counts.[12] A recent study has suggested that up to half of the patients with ‘aspirin resistance’ may be accounted for by poor medication adherence.[27] Equally, patients who were prescribed concomitant clopidogrel and PPIs may have not been able to take their clopidogrel regularly, perhaps because they have more comorbidities and are exposed to polypharmacy. As the studies were predominantly small randomized trials or observational studies, confounding could have occurred due to baseline or unmeasured differences between participants in the comparison groups. Finally, although we initially planned on conducting a meta-analysis of differences in platelet function, we were unable to replicate some of the statistical analyses due to limitations in the format and clarity of reporting within the primary studies.

In view of these limitations, there remains considerable doubt regarding postulated interaction between PPIs and clopidogrel, particularly as the clinical implications of the platelet function tests are uncertain. It is worth noting that in one study, the maximal platelet aggregation (±SD) in PPI-exposed patients was 41.1 ±17.4% compared with no PPI exposure (40.3 ±16.6%).[24] While this was reported by the authors as being statistically significant, it seems unlikely that a 0.8% difference in platelet aggregation would lead to clinically significant cardiovascular consequences. In the recent GRAVITAS (Gauging Responsiveness with A VerifyNow assay-Impact on Thrombosis And Safety) trial, clopidogrel-resistant patients were randomized to high-dose clopidogrel with a consequential reduction of PRU to 211 units (compared with PRU of 250 units with standard-dose clopidogrel), but this did not result in any difference in major cardiovascular outcomes.[28]

The benefit : harm profile of adding PPIs to clopidogrel is best assessed through RCTs where absolute rates of cardiovascular and gastrointestinal events can be directly compared.[5] A meta-analysis that included the COGENT (Clopidogrel and the Optimization of Gastrointestinal Events Trial) found that PPIs significantly reduced gastrointestinal complications such as haemorrhage or ulcer in patients receiving dual antiplatelet therapy.[29] While the large COGENT trial is limited by early termination and use of a specific clopidogrel-omeprazole formulation, there was evidence of clear gastrointestinal benefit without any apparent cardiovascular harm. More recently, Hsu et al.[12] carried out a 6-month RCT of esomeprazole in 165 patients receiving clopidogrel monotherapy with known history of peptic uclers, and found a significantly lower rate of recurrent endoscopically diagnosed peptic ulcer in esomeprazole users (1/83) compared with the control arm (9/82).[12] Although the study was open-label, the endoscopists were blinded to treatment assignation, and cardiovascular events were independently adjudicated (4/83 with esomeprazole vs 3/82 with control).

Hence, we believe that the ideal study would be a double-blind, randomized trial in patients undergoing percutaneous coronary intervention. It is important that this study is adequately powered to compare clinical outcomes (such as myocardial infarction, gastrointestinal haemorrhage or death) with individual PPIs, and to have predefined platelet function monitoring as well as rigorous ascertainment of actual use of clopidogrel/PPI. However, such a trial may not be feasible as it is likely to need several thousand patients with extended follow-up and monitoring, particularly if there is a plan to stratify according to metabolic variation from CYP2C19 genotype and paraoxonase-1 activity.

5. Conclusions

The findings of this review suggest that there is conflicting and inconsistent evidence on the impact of PPIs on platelet function studies. At present, we caution against relying solely on studies of platelet function in determining the risk of clinical interaction. Instead, clinicians should continue to clinically assess the gastrointestinal risk of the patients and prescribe PPIs accordingly to those at high risk, given the relatively greater strength of evidence supporting the benefits of PPIs in such patients.