A population-based study of the drug interaction between proton pump inhibitors and clopidogrel

Setting

We conducted a population-based nested case–control study among Ontario residents aged 66 years or older who were discharged from hospital between Apr. 1, 2002, and Dec. 31, 2007, after treatment for acute myocardial infarction. These individuals had universal access to hospital care, physicians' services and prescription drug coverage. The study was approved by the Research Ethics Board of Sunnybrook Health Sciences Centre.

Data sources

We examined the computerized prescription records of the Ontario Public Drug Program, which contains comprehensive records of prescription medications dispensed to Ontario residents 65 years of age or older. We identified hospital admissions using the Canadian Institute for Health Information Discharge Abstract Database, which contains detailed diagnostic and procedural information about hospital admissions. We used the Ontario Health Insurance Plan database to identify claims for inpatient and outpatient physician services. We obtained basic demographic information, including date of death, from the Registered Persons Database, which contains a unique entry for each Ontario resident who has ever received a health card. Researchers routinely use these databases to study drug safety,18^–20 including the clinical consequences of drug–drug interactions.21^–23 We linked the databases in an anonymous fashion using encrypted 10-digit health-card numbers.

Identification of patients and outcomes

The study design is outlined in Figure 1. We established a cohort of patients aged 66 years or older who filled a prescription for clopidogrel within 3 days after hospital discharge following treatment for acute myocardial infarction. The date of discharge from hospital served as the date of cohort entry. We excluded patients who had received clopidogrel, ticlopidine or dipyridamole in the year before admission to hospital. We also excluded patients in long-term care facilities and those who received, within 90 days before or after the index date, proton pump inhibitor combination products used to eradicate Helicobacter pylori.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1: Study design. The solid black line indicates the period following hospital discharge after treatment of acute myocardial infarction. We excluded patients who did not receive a prescription for clopidogrel within 3 days of the discharge date (upward arrow). Among the remaining patients, cases were those readmitted because of myocardial infarction within 90 days after discharge. We attempted to identify 3 matched controls, with no myocardial infarction before the same date (the index date, indicated by downward arrow) for each case. Use of proton pump inhibitors before the index date was categorized as current (within 30 days), previous (31–90 days) or remote (91–180 days).

We followed patients who were receiving clopidogrel for a maximum of 90 days after hospital discharge or until readmission because of acute myocardial infarction. We identified patients with continuous use of clopidogrel using an adherence algorithm, which included patients with prescription refills of the drug at intervals not exceeding 1.2 times the number of days' supply of the preceding prescription. We identified hospital admissions using codes from the International Classification of Disease and Related Health Problems, 10th revision, for acute myocardial infarction (I21, I22). We designated the index date as the date of hospital readmission with acute myocardial infarction. For patients who had multiple readmissions with acute myocardial infarction during the study period, we considered only the first readmission.

We defined cases as patients who died or were readmitted for myocardial infarction within 90 days after the initial hospital discharge. We used random sampling with replacement24 from the same cohort of patients to identify the controls. These individuals were at risk but were not readmitted because of myocardial infarction before the index date. Controls were matched to cases on age (born within 3 years), receipt of percutaneous coronary intervention in hospital (defined as Canadian Classification of Procedure codes 1IJ50 or 1IJ57), date of hospital discharge (within 4 days) and predicted probability of short-term mortality (within 0.05 of that of the corresponding case), determined using a previously validated cardiac risk prediction model.25 When the number of controls who could be matched to a case was fewer than 3, we used only those controls and did not alter the matching algorithm.

Exposure to proton pump inhibitors

We used prescription records to ascertain exposure to proton pump inhibitors during clopidogrel therapy. We categorized use of proton pump inhibitors according to the most proximate prescription as either current (within 30 days before the index date), previous (31–90 days before the index date) or remote (91–180 days before the index date). We reasoned that a causal association between use of a proton pump inhibitor and recurrent myocardial infarction would manifest attenuating effects with more distant exposures.

Statistical analysis

We used conditional logistic regression to estimate the odds ratio (OR) for the association between reinfarction and exposure to a proton pump inhibitor, using as the reference group patients with no prescription for a proton pump inhibitor in the previous year. We adjusted for age, sex, income quintile (estimated from the residential postal code), Charlson comorbidity index,26 length of stay in hospital during the first admission for myocardial infarction, and 9 medical conditions previously shown to correlate with short-term mortality following acute myocardial infarction and identified at the index hospital admission (diabetes with complications, dysrhythmias, pulmonary edema, cardiogenic shock, acute renal insufficiency, chronic renal insufficiency, congestive heart failure and cerebrovascular disease.)25 We also adjusted for use of other commonly used cardiovascular medications, other cytochrome P450 2C19 inhibitors or inducers, and other cytochrome P450 3A4 inhibitors or inducers27 (Appendix 1, available at www.cmaj.ca/cgi/content/full/cmaj.082001/DC1) between hospital discharge and the reference date. In the secondary analysis examining use of histamine H₂-receptor antagonists, we also adjusted for exposure to proton pump inhibitors between hospital discharge and the index date.

We conducted several additional analyses to test the robustness of our findings. We altered the definition of “current use” to include any prescription for a proton pump inhibitor between hospital discharge and the index date, and we conducted analyses of recurrent myocardial infarction and death within 1 year of hospital discharge. We examined use of the histamine H₂-receptor antagonists ranitidine, famotidine and nizatidine as a tracer exposure. These drugs have clinical indications similar to those of proton pump inhibitors, yet they do not inhibit cytochrome P450 2C19. To test the specificity of our primary findings, we replicated our analysis in a cohort of patients who did not receive a prescription for clopidogrel within 90 days of discharge following acute myocardial infarction. Finally, we conducted a stratified analysis of the risk of recurrent myocardial infarction during treatment with pantoprazole or with other proton pump inhibitors. Pantoprazole does not inhibit cytochrome P450 2C19 and therefore should not interfere with the metabolic activation of clopidogrel, whereas other proton pump inhibitors or their primary metabolites do inhibit cytochrome P450 2C19 and can be expected to attenuate clopidogrel's beneficial effects.13

Limitations

Some limitations of our research merit emphasis. We had no data on important cardiac risk factors such as smoking status, blood pressure and lipoproteins. Importantly, because our data sets included only drugs listed on the provincial formulary, we could not identify use of nonprescription medications, particularly over-the-counter ASA. Although we matched cases and controls on important predictors of outcome, some imbalance was evident in their measured characteristics (Table 1). However, this observation is expected and does not threaten our primary conclusion, because we found no association between proton pump inhibitor therapy and reinfarction in several sensitivity analyses, including an analysis of patients not receiving clopidogrel. Miscoding is a potential threat to the validity of any observational study, but previous validation studies have indicated that coding for acute myocardial infarction in Ontario's administrative databases is very good.30^,31 Finally, because some patients take proton pump inhibitors intermittently, misclassification of exposure status is possible. However, with our study design, such misclassification would likely attenuate estimates of effect.

Conclusion

In summary, we found that, among patients taking clopidogrel following acute myocardial infarction, the concomitant use of a proton pump inhibitor that inhibits cytochrome P450 2C19 (omeprazole, lansoprazole or rabeprazole) was associated with an increased risk of recurrent myocardial infarction. This effect, which was not seen with pantoprazole therapy, presumably reflects inhibition of the metabolic bioactivation of clopidogrel. Our findings highlight a widely unappreciated, common and completely avoidable drug interaction in a population at high risk of recurrent coronary events. Pending further data regarding the clinical significance of drug interactions with clopidogrel, we believe that concomitant treatment with clopidogrel and proton pump inhibitors other than pantoprazole should be minimized when possible. Ranitidine or another H₂-receptor antagonist may be an appropriate alternative for patients who require acid-lowering therapy. If a proton pump inhibitor is required, pantoprazole should be used preferentially in patients who are also receiving clopidogrel.

@@ See related commentary by Lau and Gurbel, page 699