Data Source

The data source for this study was the National Health Insurance Research Database (NHIRD), which covers 99.7% of the population (nearly 23 million people) in Taiwan. The NHIRD includes enrollment and claims files. The enrollment files contain individual subscription information and demographic factors, including sex, date of birth, type of beneficiaries, and location. The claims files contain comprehensive records of inpatient care, ambulatory care, pharmacy store, dental care, and Chinese medicine services, including date of service, International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes, claimed medical expenses, and the copayment amount for each encounter. This NHIRD dataset was provided by the National Health Research Institutes, and individual and provider identifiers have been encrypted in order to protect privacy and confidentiality. The study protocol was approved by the institutional review board of Taipei Veterans General Hospital (No 201006015IC).

Study Group and Cohort Definition

We conducted a retrospective cohort study with a 6-year observation period (2006 to 2011). We identified 10,322 patients who (1) underwent LES or PES placement between January 1, 2007, and December 31, 2010; (2) had a diagnosis of type 2 diabetes mellitus (ICD-9-CM codes 250.x0 and 250.x2) before the first DES deployment; and (3) had at least one prescription for a hypoglycemic agent during a 1-year period prior to the first placement. Anti-diabetic medication can only be prescribed while patients fulfilled any one of following: 1) Fasting plasma glucose level ≥ 126 mg/dL; 2) Plasma glucose ≥ 200 mg/dL two hours after a 75 g oral glucose load as in a glucose tolerance test; 3) Symptoms of high blood sugar and casual plasma glucose ≥ 200 mg/dL or HbA₁c≥ 6.5 under benefit package of NHI. We defined an individual’s index date as the date of discharge after the first DES implantation, and an individual was followed up for 1 year after the index date. Patients were excluded if (1) they did not have an ambulatory visit or receive a hypoglycemic agent within 90 days after the index date; (2) they experienced revascularization, ACS, or death within 7 days after the index date; (3) they underwent stent implantation within 1 year before the index date; or (4) they were implanted with both an LES and a PES concurrently at the index hospitalization. A total of 9,805 subjects were included in this analysis.

Exposure

An individual’s exposure to CLO and PPIs within 90 days after the index date was identified using computer-based prescription claims; patients were classified into CLO (n = 8,856) or CLO plus PPIs (n = 949) groups. The algorithm to identify these drugs is described in the appendix.

Outcomes and Covariates

Outcomes of this study consisted of two elements: time to event and a censoring indicator. Time to event represents the number of days from the index date to the date of the earliest of the following events: (1) the end of the observation and (2) the occurrence of target events, including revascularization (including PCI or coronary artery bypass graft [CABG] surgery) and ACS (ICD-9-CM codes 410.xx, 411.xx, and 414.9). The repeat revascularization was defined as PCI (ICD-9-CM codes 36.0–36.09) and CABG (ICD-9-CM codes 36.1–36.19). The observational period began at the cohort index date and continued until the first occurrence of any major adverse cardiac event or up to 1 year of follow up. Each patient had nine sets of outcomes given three endpoints (3 months, 6 months, and 1 year) and three target events. If the earliest event was the occurrence of a target event, and this record was not censored.

We considered the following covariates: age, sex, drug use after discharge (metformin, sulfonylurea, insulin, β-blocker, calcium-channel blocker, angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers, lipid-lowering agents, and aspirin), and comorbidities during the year before the index date (congestive heart failure, ACS, renal disease, peripheral vascular disease, cerebrovascular disease, and chronic pulmonary disease), as defined by Romano et al.[18] We further adjusted for a history of CABG during the year before the index date and the intensity of the use of medical services during the index hospitalization, including the number of days spent in the hospital and the number of stents received.

Statistical Analysis

We used the chi-square test to examine the association between two groups (CLO, and CLO plus PPIs) for categorical variables and t test for continuous variables. Multivariate-adjusted hazard ratios (HRs) were estimated using Cox proportional hazards models. We adjusted for potential confounders in the multivariable Cox models based on reported risk factors for cardiac events. The propensity score regression adjustment was also used to balance the distribution of confounders between the two groups (CLO group and CLO plus PPIs group) by summarizing all covariate information into a single probability and to simulate randomization.[19] Moreover, we further conducted a stratified analysis according to a history of ACS. All statistical tests were two sided, with an alpha level of 0.05, and the confidence intervals (CIs) were 95%. Data management and statistical analyses were performed using SAS version 9.2.

Interaction between CLO and PPIs

When CLO is administered as a prodrug, 85% of it is inactivated by plasma esterases and the remaining 15% undergoes liver metabolism by the cytochrome P450 system to generate an active metabolite. The isoenzyme CYP2C19 plays an important role in CLO activation.[21] All anti-diabetic medication included in our study may not alter clopidogrel effect. [22] It has been shown that among CYP450 isoforms, a CYP2C19 loss-of-function variant is associated with a reduced response to CLO.[23] The prevalence of CYP2C19 loss-of-function alleles is much higher among Taiwanese than among other populations.[24] This might imply that a drug—drug interaction exists between CLO and PPIs. Matetzky et al reported that the lowest response to CLO in patients undergoing stenting for ST-elevated MI was associated with a 40% probability of a recurrent cardiovascular event within 6 months.[25] In patients undergoing PCI with DES placement, a low response to CLO was significantly associated with an increased risk of stent thrombosis.[26] After adjusting for risk factors such as age, sex, drug use after discharge, and comorbidity, concomitant use of PPI also predicted rehospitalization for ACS after DES deployment in our study. This suggests the possibility that a low response to CLO exists in patients with concomitant PPI use.

Possible Explanations for the Observed Increased ACS Risk

In subgroups of patients with diabetes in randomized clinical trials, no significant difference in ACS was observed between LES and PES among diabetic patients.[27],[28] In the ISAR-DESIRE 2 trial, there were no significant differences in the MI rates between LES and PES among diabetic patients.[29] Although these studies demonstrated no significant difference in MI rates between LES and PES, the present study shows that MI rates are significantly increased in diabetic patients using CLO plus PPI after LES implantation. There are several possible explanations for the increased MI associated with concomitant treatment with CLO and PPIs after LES implantation. First, PPI therapy was independently associated with increased residual platelet reactivity and a higher frequency of high residual platelet reactivity (HRPR) and decreased the CLO inhibitory effect among patients with sirolimus-eluting stent implantation.[30] HRPR with CLO therapy has been associated with an increased risk of cardiovascular events after PCI.[31] Second, rapamycin can increase the activity of tissue factor, a key trigger for the coagulation cascade, in endothelial cells.[32] Jiang et al reported that platelet-endothelial adhesion was enhanced by rapamycin.[33] In contrast, paclitaxel significantly inhibited collagen-induced platelet aggregation in vitro.[34] The attenuated antiplatelet effect of CLO by PPI and increased platelet—endothelial adhesion by rapamycin might contribute to the increased MI rate with LES implantation. However, our study demonstrated that PPIs substantially affected the clinical outcomes of patients treated with LES. Prospective randomized studies are thus warranted to determine whether the findings from the present analysis of diabetic patients are generalizable to all stents eluting rapamycin analogs. In addition, further studies are required to understand the mechanisms underlying these findings so that more effective therapies can be developed for high-risk diabetic patients.

The present study has several limitations. First, the diagnoses in national health insurance claims primarily serve administrative-billing purposes and do not undergo verification for scientific purposes. However, subjects with PCI were ascertained by hospitalization for ACS and by stent implantation procedures, which were very reliable. Second, information regarding patient adherence or self-paid medications is not available. Although this could have lead to misclassification and biased the results towards a null effect, it would not have changed the clinically significant result regarding the increased risk of rehospitalization for ACS associated with CLO and PPI use. Third, death records were not available in the NHIRD, but we have captured important clinical outcomes of rehospitalization of ACS. Fourth, all the PPIs are available as prescription drugs in Taiwan. The possibility of misclassification of exposure to PPIs may be small in this study. Fifth, data regarding personal habits such as smoking, the severity of coronary artery disease, characteristics of coronary stenotic lesions (ex. instent stenosis, instent thrombosis or coronary lesions outside the initially implanted stent), and stent length and diameter were not available from this dataset. Also, NHIRD contained code of disease but did not contain important information like blood HbA1c, glucose levels, or platelet aggregometry. Therefore, we cannot provide information about these data correlate with clinical outcome or effect of PPI on the Adenosine diphosphate (ADP) induced platelet aggregation. In our study, PPI users were older, more female, and had more previous heart failure, ACS, cancer and liver disease, irrespective LES or PES, whereas non-PPI users took more often aspirin, ACEI and lipid lowering agents. Although propensity scoring and Charlson comorbidity index were included in the multivariate analysis, these factors could not be strongly excluded associated with worse outcomes. Therefore, further randomized study will be necessary to determine the effect of PPIs on CLO in different DESs. However, this study provides benchmark data for patients treated with different DESs and the inclusion of consecutive patients referred for PCI and stent implantation reflects a “real-world” situation and therefore strengthens our results.

Clinical Implications

If a diabetic patient is at low risk of GI tract bleeding, it is best to avoid or decrease the dosage of PPIs used in combination with CLO, as the risk of negative interaction is greater than the risk of GI tract bleeding. Because there is no substantive evidence that PPIs attenuate the therapeutic effect of prasugrel or ticagrelor, these inhibitors might be considered if diabetic patients are at moderate or high risk of GI tract bleeding should it be necessary to use PPIs. Platelet function testing might be considered in specific high-risk patients (e.g. history of MI or high GI bleeding risk).