Saphenous vein grafts (SVG) remain the commonest conduits used in surgical revascularization for severe coronary atherosclerosis. However, SVG patency after coronary artery bypass grafting remains suboptimal; in a large randomized controlled trial the per-patient angiographic vein graft failure rate was >40% by 18 months (1). Guideline-based recommendations consider percutaneous coronary intervention (PCI) with balloon angioplasty and stenting of diseased SVGs to be a reasonable treatment in selected patients, given the risks of repeat cardiac surgery, and the practice is now widely employed (2). However, clinical outcomes are worse after SVG PCI than after PCI of native coronary arteries (3). The saphenous vein graft is more prone to thrombosis compared with native coronary arteries, a propensity that might be mitigated by long-term dual antiplatelet therapy. Current practice patterns with regard to antiplatelet therapy in patients undergoing SVG PCI are informed largely by data extrapolated from clinical trials and observational studies of PCI of native coronary arteries.

Although early termination or noncompliance with dual antiplatelet therapy has been shown to be associated with an increase in adverse cardiovascular events, including death and myocardial infarction (MI), the pattern of adverse events after clopidogrel cessation with longer durations of therapy remains less clear (4). We hypothesized that discontinuation of clopidogrel therapy after SVG PCI, including after 1 year of treatment as recommended by current guidelines, might be associated with a short-term increase in death or MI.

We identified consecutive patients undergoing SVG PCI at the Regional Cardiac Catheterization laboratory located at the Kaiser Permanente Los Angeles Medical Center from January 1, 2000, to December 31, 2009. Kaiser Permanente is an integrated pre-paid health plan providing comprehensive care to more than 3 million members in Southern California. The health plan owns and operates medical centers, ambulatory care facilities, pharmacies, and laboratories. The Regional Cardiac Catheterization laboratory serves as the primary referral site for 8 acute care hospitals in the region.

All patients in the cohort underwent SVG PCI and received either a drug-eluting (DES) or bare-metal stent (BMS). The choice of stent type was at the discretion of the operator. Heparin with or without a glycoprotein IIb/IIIa inhibitor or bivalirudin was used in all cases. Clopidogrel 300 to 600 mg was administered before PCI. Patients who had an adverse event while receiving clopidogrel therapy were excluded in the primary analyses.

Baseline demographic data were obtained for review from patient medical records and health plan databases. For each patient, data on age, sex, stent size, stent diameter, site of stent deployment, other medical history, and procedural data were obtained from health plan databases. Laboratory data were obtained at the time of cardiac catheterization or the most recent value before the procedure.

Clopidogrel use was obtained by review of prescription records through an electronic pharmacy prescription database. The pharmacy database links all health plan pharmacies and captures all prescriptions dispensed. Subjects in the study obtained medications from health plan pharmacies; those without prescription records available (1.8%) were excluded. The number of prescriptions filled, the daily dose, and the number of pills dispensed for each prescription beginning on the day of discharge after index PCI were used to calculate the duration of clopidogrel use. All patients received 75 mg daily of clopidogrel post-PCI. The largest quantity dispensed was 100 tablets, requiring patients to return to the pharmacy for additional medication. Noncompliance was defined as a gap in clopidogrel therapy of >5 days. In the primary analysis, the last date of clopidogrel use was based on the date of the last prescription and number of tablets dispensed.

Adjunctive pharmacotherapy with statins, beta-blockers, angiotensin-converting enzyme inhibitors, and angiotensin II receptor blockers were also determined by electronic pharmacy prescription records. Use of these medications was defined as 1 or more prescriptions for each drug after the PCI date. Aspirin use was not available through the electronic prescription database. Nevertheless, all patients were instructed to take aspirin 325 mg daily for 90 days and then 81 mg daily indefinitely, regardless of stent type.

The primary endpoint of the study was a composite of all-cause mortality or nonfatal MI. Vital status was obtained from the medical records, health plan databases, and MORTLINK, which is a validated death database of the residents of California maintained by the state. Follow-up MI data were obtained with International Classification of Disease-9th Revision-Clinical Modification codes for MI (410.xx) and at least 1 abnormal cardiac biomarker (e.g., Troponin I, creatine kinase, or creatine kinase-myocardial band). Patients were censored for loss of health plan membership, a subsequent PCI procedure for an indication other than MI, and at the end of the follow-up period. Hospital stays outside the healthcare network were also captured through administrative claims data. The follow-up period ended January 1, 2010.

Both inpatient and outpatient medical records were reviewed system-wide by 2 investigators (A.S. and S.B.) for bleeding events (gastrointestinal, intracranial, or other) or the administration of blood products associated with termination of clopidogrel therapy and a primary endpoint.

Continuous variables were reported as mean ± SD or median and interquartile range if the distribution was not normally distributed. The Student t test was used for normally distributed variables, and the Wilcoxon rank sum test was used for non-normally distributed variables. Categorical variables are reported as percentage and absolute values with comparisons made with the chi-square test or Fisher exact test where appropriate.

A multivariable Cox regression model was constructed to obtain risk-adjusted instantaneous incidence rates of death or MI after stopping clopidogrel with kernel hazard functions, including by stent type. The following covariates were included in the model: age, sex, comorbidities, clopidogrel duration and noncompliance, stent type and diameter, PCI period (year of index PCI: early, 2000 to 2004; late, 2005 to 2009), and use of concomitant medications (beta-blockers, statins, and others). The proportionality assumption was tested for Schoenfeld residuals and satisfied. Failure plots by clopidogrel duration and stent type were created according to the method of Kaplan-Meier. The strata of clopidogrel duration were compared with the log-rank test.

A piecewise exponential survival model was used to generate adjusted incidence rate ratios (IRRs) and 95% confidence intervals (CIs) comparing the 0- to 90-day and 91- to 365-day intervals after clopidogrel cessation. The independent variable of interest was the risk of adverse events in the 0- to 90-day interval after clopidogrel cessation compared with the subsequent 91- to 365-day period. The covariates included in this model were those noted in the preceding text. We also used Poisson regression to model the relationship. The findings were consistent with the primary results and are not further reported. Adjusted IRR were calculated for pre-specified subgroups: clopidogrel duration, sex, diabetes status, PCI period, stent type, and stent diameter. A p value for interaction was obtained with first-order interaction terms in the adjusted regression model.

In pre-specified sensitivity analyses, we excluded patients with bleeding events associated with clopidogrel cessation to assess the influence of such events on the primary endpoint. In a subsequent analysis, we also excluded patients with a history of gastrointestinal bleeding.

All analyses were performed with STATA statistical software (version 12, Stata Corp., College Station, Texas). Kaiser Permanente Southern California Institutional Review Board approval was received with waiver of the requirement for written informed consent.

A total of 603 patients underwent SVG PCI during the study period. Baseline characteristics are displayed in (Table 1). The mean age was 68 ± 10 years, and 18% were female. There were 1.5 ± 0.8 stents/person implanted in the study cohort. The stent types used were 50% BMS and 50% DES (27% paclitaxel-eluting, 22% sirolimus-eluting, and 2% everolimus-eluting stents). The distribution of clopidogrel use post-PCI was as follows: <6 months, 45%; 6 months to 1 year, 14%; >1 year 41%. The mean duration of clopidogrel therapy after SVG PCI was 270 days (interquartile range, 59 to 900 days). The mean duration of follow-up after the index SVG PCI was 3.1 years and the median was 3.4 years.

Among the 603 patients who underwent SVG PCI during the study period, 411 were event-free at the time of clopidogrel cessation and had discontinued clopidogrel therapy before the study end date. The characteristics of these patients are also shown in (Table 1). Mean follow-up after stopping clopidogrel treatment was 543 days, and median was 720 days; there were 223,324 person-days of follow-up. The composite primary endpoint of all-cause mortality or nonfatal MI occurred in 29.7% (n = 122) of patients.

The overall incidence rate of death or MI was 0.54 (95% CI: 0.45 to 0.64)/1,000 person-days. Incidence rates for the 0- to 90-day, 91- to 365-day, and 1- to 2-year intervals after clopidogrel cessation are shown in (Table 2). The rate of death or MI was greatest in the first 90 days after stopping clopidogrel therapy, corresponding to an approximately 2.6-fold increase (IRR: 2.58; 95% CI: 1.64 to 4.07) in the incidence of death or MI in the first 90 days post-clopidogrel cessation compared with the subsequent 91 to 365 days. Risk-adjusted instantaneous incidence rates/person-day for death or MI after stopping clopidogrel are shown in (Figure 14_gr1). The risk of an adverse event seems to be greatest early after cessation of clopidogrel therapy, regardless of clopidogrel duration and stent type. The findings of an increased risk of adverse events associated with the 0 to 90-day interval compared with the 91 to 365-day interval were similar when death alone was evaluated (IRR: 2.33; 95% CI: 1.32 to 4.11).

We further explored the distribution of adverse events after clopidogrel cessation by duration of clopidogrel therapy (Figure 14_gr2). Regardless of the duration of therapy, approximately 40% to 60% of the adverse events in the first year after cessation of clopidogrel occur within the initial 90 days. This relationship was consistent for patients treated with clopidogrel for <6 months, 6 months to 1 year, 1 to 2 years, or >2 years after SVG PCI. Moreover, the early hazard associated with stopping clopidogrel therapy was consistent across multiple subgroups (Table 3). In particular, there was no significant difference between BMS and DES for long-term death/MI after clopidogrel cessation (Figure 14_gr3). The 2-year event rates after clopidogrel cessation in the BMS and DES groups were 29% and 34%, respectively (log-rank p = 0.45).

In sensitivity analyses, the results were qualitatively similar after removing patients with a bleeding event leading to clopidogrel cessation. There were 8 (1.3%) patients with a primary endpoint who discontinued clopidogrel therapy related to a bleeding event. The incidence rates for the 0- to 90-day interval and 91- to 365-day intervals excluding such patients were 1.25 (95% CI: 0.92 to 1.69) and 0.55 (95% CI: 0.34 to 0.89), respectively; the corresponding IRR was 2.27 (95% CI: 1.90 to 2.68). The results were also qualitatively similar after removing patients with a history of gastrointestinal bleeding, which was noted in 10% of the cohort. The IRR was 2.30 (95% CI: 1.90 to 2.79) after excluding patients with a history of gastrointestinal bleeding. We also calculated the incidence rate for the 0- to 90-day interval after excluding patients with <30 days of clopidogrel therapy, a group that might be at heightened risk for noncompliance and therefore adverse events. After excluding such patients, incidence rates for death or MI at 0 to 90 and 91 to 365 days were 1.12 (95% CI: 0.78 to 1.61) and 0.44 (95% CI: 0.31 to 0.63), respectively; the corresponding IRR was 2.05 (95% CI: 1.21 to 3.47).

In secondary analyses, we analyzed the full study cohort (n = 603) and observed that longer duration of clopidogrel therapy was associated with a lower rate of long-term death or MI post-PCI (Figure 14_gr4). The cumulative 5-year event rate was lowest in patients treated with clopidogrel for >2 years (p <0.001). The event rate at 5 years post-PCI was 54.3% for 0 to 6 months of clopidogrel therapy, 52.3% for 6 months to 1 year of therapy, 42.7% for 1 to 2 years of therapy, and 37.3% for >2 years of clopidogrel therapy (log-rank p < 0.001). The adjusted hazard ratio for >2 years of clopidogrel therapy compared with <6 months was 0.48 (95% CI: 0.33 to 0.70; p <0.001), corresponding to a significant 52% reduction in death or MI.

Little is known about the role of dual antiplatelet therapy after SVG PCI. To our knowledge, this is the first study to report on the pattern of adverse events after clopidogrel cessation in patients undergoing SVG PCI. We observed an increase in death and MI shortly after clopidogrel cessation, regardless of the duration of clopidogrel therapy. The clustering of events was observed to be greatest 0 to 90 days after stopping clopidogrel therapy. There was an approximately 2.6-fold increase in the incidence of death or MI in the first 90 days post-clopidogrel cessation, compared with the subsequent 91 to 365 days. This relationship was apparent across a broad range of treatment durations. Among patients treated with clopidogrel for <6 months, 6 months to 1 year, 1 to 2 years, or >2 years, there was a consistent early hazard with discontinuation of clopidogrel therapy. Hence, regardless of the treatment duration, approximately 40% to 60% of the adverse events within the first 1 year after stopping clopidogrel occurred in the first 90 days.

This early hazard of stopping clopidogrel therapy does not seem to be the result of noncompliance and is less likely to be the result of confounding events leading to clopidogrel cessation. In sensitivity analyses, we assessed the impact of noncompliance by excluding patients with <30 days of therapy. After excluding such patients, there was a 2-fold increase in early death or MI, which was qualitatively similar to the overall result of the primary analysis. One postulated mechanism that might confound the observed relationship and could account for the early hazard is bleeding leading to cessation of clopidogrel therapy. Bleeding events often lead to a hypercoagulable state and disruption of cardioprotective medications, including dual antiplatelet therapy, placing the patient at heightened risk for adverse cardiovascular events. Also, the need for transfusions might promote a prothrombotic state (5). There remained a 2.3-fold increase in the early hazard of death or MI, qualitatively similar to the full study cohort, after patients with a concomitant bleeding event requiring cessation or disruption of clopidogrel therapy were excluded. The results were also consistent after excluding subjects with a history of gastrointestinal bleeding and across several subgroups. In support of our findings, there is evidence that an early hazard associated with clopidogrel cessation might exist in PCI of native coronary arteries ((6),7).

It is possible that the clustering of events early after clopidogrel cessation is reflective of the increased potency of dual antiplatelet therapy at protecting from underlying thrombotic substrate compared with aspirin alone. After clopidogrel cessation, this protective effect is no longer present, which might lead to adverse events. This might be more evident in populations at high-risk for thrombotic complications, such as patients undergoing SVG PCI. This phenomenon might also be mediated by an increase in platelet reactivity, because some studies have observed an increase in platelet reactivity after discontinuation of long-term clopidogrel therapy ((8),(9),(10),11). Moreover, withdrawal of clopidogrel therapy has been associated with a significant increase in markers of inflammation (8).

Our results provide insight into the relationship between duration of clopidogrel therapy and adverse events, although this is not the primary aim of this study. To our knowledge, there are no prior reports on this topic in patients undergoing SVG PCI. These data suggest that the lowest event rates are observed with longer durations of clopidogrel therapy with the lowest rates of death or MI observed with >2 years of treatment. Nevertheless, as previously noted, even in patients treated for >2 years there was evidence of an early hazard for death or MI associated with clopidogrel cessation. In aggregate, these data support longer durations of dual antiplatelet therapy after SVG PCI.

First, this was an observational analysis, and clopidogrel use was not randomly assigned. Nevertheless, the observation of a consistent early hazard across a broad range of duration of clopidogrel therapy suggests the results are not confounded by differences in duration of therapy, noncompliance, or comorbidity severity. Second, ascertainment of the exposure, clopidogrel use, was based upon pharmacy prescription records; pill counts might have provided a more precise estimate of clopidogrel use. The use of prescriptions filled has been shown to reflect medication use by the patient with a high degree of accuracy (12). Third, aspirin use was not available through the pharmacy prescription database, and our inability to adjust for aspirin noncompliance is a limitation. However, all patients received the same written instructions with regard to aspirin therapy. It is very likely that patients refilling clopidogrel prescriptions were also compliant with aspirin therapy. Fourth, the patients included in this study were insured and received care in a large pre-paid health plan. The health plan membership is representative of the general population of the state of California except for the socioeconomic extremes. It is possible that in other populations, particularly the underinsured or non-insured, the observed pattern of events might be different and the magnitude of the early hazard associated with clopidogrel cessation might be exaggerated. Fifth, details on stent thrombosis or rates of target vessel revascularization due to MI were not available. Lastly, clopidogrel was the only adjunctive antiplatelet agent investigated. These results do not apply to newer antiplatelet agents such as prasugrel or ticagrelor.

We observed an approximately 2-fold increase in the incidence of death or MI in the initial 0 to 90 days after clopidogrel cessation compared with the subsequent follow-up period. This pattern of events was consistent among patients treated with clopidogrel for <6 months, 6 months to 1 year, 1 to 2 years, or >2 years and in subgroup analyses. These data provide new insights into the importance and possible limitations of dual antiplatelet therapy after SVG PCI. Additional studies are needed to identify treatment strategies and elucidate the mechanisms to address this effect.