Premature and permanent thienopyridine discontinuation after drug-eluting stent (DES) implantation conveys a risk for stent thrombosis ((1),(2),(3),4), probably on the basis of increased platelet reactivity. Therefore, double antiplatelet therapy (DAT) is recommended for at least 6 to 12 months (5). Although the risk is highest during the first month after DES implantation and still high within the first 6 months ((4),6), a safe time period for antiplatelet therapy discontinuation (ATD) has not yet been defined ((7),8). Furthermore, information is scanty about the risk of temporary ATD during the first year that may occur in different scenarios, such as life-threatening hemorrhage, surgical intervention, errors in medical prescription, or lack of adherence. However, the current evidence has been interpreted as implying that any interruption can be dangerous ((6),9).

The ACDC (Adherence to Treatment of Coronary Patients After a Catheterization With DES Implantation) is a prospective cohort study addressing the background, incidence, potential predictors, and safety of ATD during the first year after DES implantation (10). It showed that over 14% of patients who received DES interrupted at least 1 antiplatelet drug during the first year after DES implantation, in most cases clopidogrel, usually temporarily, and that discontinuation was most often based on patient decision or medical decisions not associated with major bleeding events or major surgical procedures.

In the present paper, we assess the risk associated with ATD in the ACDC cohort, and specifically, temporary discontinuation, during the first year of DES implantation in terms of cardiac mortality or acute coronary syndrome (ACS).

Methods of the ACDC study have been described elsewhere (10). All patients receiving at least 1 DES between January 28, 2008, and April 28, 2008, were recruited by clinical investigators in 29 participating hospitals from Spain. Local investigators were specifically trained and actively participated in the draft of the study protocol.

Study variables included data related to coronary angiography, cardiovascular risk factors, cardiovascular history, complications during admission, and medications at discharge. In addition, psychosocial variables and several hospital characteristics were recorded.

A quality control was performed to ensure consecutive inclusion and quality of data collection in 28 of 29 centers. This quality control led to include retrospectively 75 patients who had been missed by local investigators and to review all the data entered in 5 centers where more than 5 errors/patients were detected.

All patients who signed informed consent were interviewed by phone by trained researchers at 3, 6, 9, and 12 months using a standardized questionnaire to determine: 1) vital status; 2) current medications (patients were asked to collect all their current medications and to read out every brand name); 3) medications temporally or permanently interrupted since the previous phone call; 4) reason and duration of discontinuation; and 5) hospital readmissions. For patients who died, a close relative was interviewed. In case of readmission, clinical records were reviewed at the corresponding center and centrally checked by the main investigator team to establish the reason for readmission and medications during hospitalization and at discharge.

From the phone interviews and the review of clinical records, the following data were assessed for each patient: the approximate date of ATD, the antiplatelet drug that had been interrupted (clopidogrel, aspirin, or both), and in the case of resuming the antiplatelet drug, the date of resumption. Thus, the approximate number of days of discontinuation could be determined in each patient.

The main outcomes of interest were ACS and cardiac death. Both were identified from clinical records by the main investigator team, who was blind to the DAT status at the time of endpoint adjudication. ACS required an increase of cardiac necrosis biomarkers above the upper limit for each local laboratory plus either suggestive symptoms or electrocardiogram changes. Cardiac death was considered in cases of ACS, congestive heart failure, or unexpected death not clearly secondary to a noncardiac cause. The events were adjudicated by the main coordinator team with use of the original source documents.

Descriptive data are presented as mean ± SD or proportions for individual characteristics.

To explore the association between ATD and 1-year cardiac mortality or ACS, we used survival analysis, patients being censored at the time of the first of the 2 events. Other causes of censoring were death from noncardiovascular causes, bypass surgery, and loss to follow-up. We employed extended Cox regression modeling, introducing the covariate ATD as time dependent. We introduced the variable ATD using a step function that equals 0 all the time the individual is taking clopidogrel and aspirin, and equals 1 when the individual is not taking clopidogrel and aspirin. We also explored the specific effect of interrupting aspirin or clopidogrel, or both, assigning a different value to each category. Additionally, we explored the specific risk of temporary ATD (i.e., the interruption of antiplatelet therapy and subsequent resumption without the occurrence of any new revascularization or cardiovascular event) by censoring, at the time of ATD, patients who interrupted any drug without resumption.

We first estimated the crude effect of ATD, including it in the model as a single variable (i.e., crude estimate). To estimate the adjusted effect of ATD on cardiac death or ACS, we considered those variables that may potentially be common causes of exposure (i.e., ATD) and outcome (11) as candidate confounders. We considered factors related with the global patient risk, factors related with the severity and natural history of the coronary disease, and factors related with the hospital where the patient underwent the procedure (i.e., teaching hospital and mean number of patients receiving stents in 1 year). All these factors were finally retained if their inclusion modified the coefficient of the effect of the exposure >10%. Proportional hazards assumption was tested for each variable by plots (log (time) versus log [−log (survival)] stratified by the variable).

Several patients died of cardiac causes during the follow-up. In this case, although researchers interviewed their relatives, the possibility of information bias concerning the discontinuation of AT before dying was plausible. Thus, we performed a simulation analysis as sensitivity analysis. The objective was to quantify the impact of a potential information bias. In cases of death from cardiac causes, we simulated different rates of ATD preceding death: 0%, 10%, 25%, 50%, and 100%. Patients who died of cardiac causes were randomly selected for ATD, assuming a Bernoulli distribution. The interval between ATD and death was simulated to be 7 and 15 days. The interval of 7 days was chosen as this was the actual median number of days without DAT in those patients with temporary discontinuation. The previous Cox model was employed to include the variable ATD, again as a time-dependent covariate, but assuming the new ATD distribution. We made 15 iterations to estimate the risk of ATD.

Twenty of 1,985 patients included in the ACDC study died during admission. Thus, there were 1,965 candidates to follow-up. At least 1 time-point follow-up (3, 6, 9, or 12 months) could be assessed in 1,622 (82.5%). In 1,536 patients, follow-up could be achieved for the 4 time points. In the rest, only 2 or 3 follow-up time points were available.

(Figure 57_gr1) shows the cumulative incidence of the endpoint cardiac mortality or ACS along the 4 time-point study period and its relationship with previous ATD. Eighty-seven of 1,622 patients (5.4%; 95% confidence interval [CI]: 4.3 to 6.6) had a major cardiac event during the first year after DES, but only in 7 of these 87 (8%; 95% CI: 3.3 to 15.9) was a history of ATD recorded. These 7 patients had ACS, and there was important variability concerning the drug interrupted, the moment and duration of interruption, and the time from ATD to event (Table 1). The rate of ATD in patients with events was similar to that among the 1,535 patients without events or who had been censored before the end of follow-up (n = 165, 10.7%; 95% CI: 9.2 to 12.4; p = 0.23). Overall, 172 (10.6%) patients had interrupted at least 1 antiplatelet drug, most of them (n = 111, 64.5%) temporarily: 79 clopidogrel (31 temporarily), 38 aspirin (27 temporarily), and 55 both drugs (53 temporarily). The median number of days without DAT in those who resumed was 7 (interquartile range: 5 to 8.5). The rate of ATD varied slightly across the 4 study intervals: 2.16%, 1.7%, 3.6%, and 3.6%, respectively. Only 1 patient interrupted DAT during the first month after DES implantation. It was to prevent hemorrhagic risk during an admission for endocarditis, and the patient died of sepsis several days later.

(Table 2) shows the baseline characteristics of the study population according to the ATD status. Patients who interrupted DAT had a lower rate of prior coronary angioplasty and a higher rate of comorbidities such as chronic obstructive pulmonary disease, chronic renal impairment, and previous major hemorrhage. There were no relevant differences between both groups concerning psychosocial characteristics. Finally, patients who interrupted DAT were attended less often in more active centers (i.e., with higher rates of patients receiving stents in 1 year).

The unadjusted global risk (hazard ratio [HR]) of cardiac death or ACS associated with ATD was 1.93 (95% CI: 0.87 to 4.28; p = 0.1): 1.95 (95% CI: 0.47 to 7.99; p = 0.35) for isolated aspirin discontinuation, 1.34 (95% CI: 0.32 to 5.5; p = 0.68) for isolated clopidogrel discontinuation, and 2.71 (95% CI: 0.84 to 8.72) for DAT discontinuation. When adjusting for potential confounders, the association remained nonsignificant (Table 3). The same was true when assessing the risk of the isolated temporary ATD by censoring the patients who did not resume DAT at the time of ATD: HR: 0.86 (95% CI: 0.21 to 3.6; p = 0.83).

(Figure 57_gr2) shows the HR and 95% CI of cardiac death or ACS when simulating, in those patients who died of cardiac reasons, a rate of ATD 7 days before dying of 0%, 10%, 25%, 50%, 75%, and 100%. The risk of major cardiac events associated with ATD would have been statistically significant if at least 18.8% of patients who died had interrupted DAT 7 days before dying (HR: 2.04; 95% CI: 1.01 to 4.1). In other words, in our sample, we would have had to misclassify at least 18.8% who died of cardiac causes concerning ATD (i.e., false negatives) to conclude that there was not a statistically significant risk of cardiac events associated with ATD when actually there was. Similar results were obtained simulating the interruption of DAT 15 days before dying (data not shown).

The ACDC study shows that although ATD during the first year after DES implantation is not exceptional, in most instances, it was a temporal interruption, the antiplatelet medication being resumed in the following days (median: 7 days). Most importantly, it was not necessarily followed by major cardiovascular events, at least in patients who interrupted DAT later than 1 month after stenting, which was the most common situation in our population.

Interruption of antiplatelet drugs has been shown to be deleterious in several contexts (12), particularly DES thrombosis after clopidogrel discontinuation ((1),(4),13). However, the impact of discontinuation is less clear after the study by Kimura et al. (14), which casts doubts about the actual role of thienopyridines associated with aspirin for preventing stent thrombosis. Likewise, in another registry (4), clopidogrel discontinuation was not associated with stent thrombosis when it occurred later than 6 months after DES implantation. Although our study was not aimed at assessing stent thrombosis, the tight relationship between stent thrombosis and cardiac death and ACS ((4),(15),16) leads to similar conclusions regarding the risk assessment of ATD.

The absence of risk of major cardiovascular events associated with ATD found in the present study may be explained by several reasons. First, the rate of ATD during the first month after stenting, the period with the highest risk for ATD ((2),3), was negligible. Second, most patients discontinued only clopidogrel or aspirin, and discontinuation was temporary in most, its median duration being 7 days. This is in agreement with recent studies in which a risk of stent thrombosis associated with isolated clopidogrel discontinuation was not shown (14) or was limited to the early period of the treatment ((4),17), and where the highest rate of stent thrombosis was usually detected beyond 1 week after discontinuation ((4),14). And, third, although in our study the rate of patients undergoing DES implantation in the context of ACS reached 58%, this rate is lower than in other series that have recorded devastating consequences of clopidogrel discontinuation (3), which is probably the context in which DAT is more advantageous. In fact, a recent study about clopidogrel discontinuation after ACS showed a higher risk of death or nonfatal myocardial infarction in those patients who definitely interrupted clopidogrel therapy (18).

Bias may play a role in the results. Specifically, the presence of information bias concerning antiplatelet use could, if extreme, invalidate the results. This risk is especially high in those patients who died, and thus the interviewee had to be a relative. Therefore, we simulated that the information obtained from the relative was wrong in the sense of “favoring” a potential underlying relationship between ATD and major cardiac events. Considering the most unfavorable scenario of misclassification of ATD in those patients who died from cardiac causes, at least 18.8% wrong ATD categorizations would have been needed to falsely conclude an absence of significant risk associated with ATD. Although we believe that such a high rate of misclassification is unlikely, even in this worst scenario, the risk of serious events associated with ATD would have ranged from an HR of 1.01 to an HR of 4.1, which is a far less devastating effect than reported in other studies ((2),4). The bias of wrong ATD classification in those patients who did not die seems less likely. Conversely, it could also be possible that the high rate of adverse events associated with clopidogrel discontinuation observed in other studies (3) was partially due to a bias in the sense of detecting more ATD in those patients with events.

Studies usually evaluate the status of thienopyridine use at certain time intervals ((3),(4),13), and occasionally, they have analyzed it immediately before an event ((4),(14),15). In the present study, we assessed the consequences of the most common situation, that is, the temporal discontinuation of clopidogrel and/or aspirin for a few days with subsequent resumption. This has received less attention, or its evaluation has been less clear, probably because of its complexity. In addition to the problems in adherence assessment, there are difficulties of definition that hinder comparison or interpretation in different studies. For instance, in some studies (13), complete interruption is defined as any discontinuation lasting for longer than a given period, not providing further data on its actual duration. In fact, most studies lack information about the temporary or permanent character of discontinuation (18).

Temporary ATD may occur in different scenarios, more often in the context of bleeding events or invasive procedures (10). In the present work, we used extended Cox regression modeling with a time-dependent covariate, which permits us to examine the continuous risk of ATD and thus the implications of temporary ATD. Our results suggest that a discontinuation for a few days (median: 7 days) of ATD after the first month of DES implantation may be reasonably safe in terms of major cardiac events. However, the absence of a statistically significant association may have been because of insufficient power, as convincing instances of stent thrombosis shortly after ATD have been reported. Moreover, it is possible that some of the cardiac events observed in the ACDC study, with or without DAT, may have been due to stent thrombosis. In any case, our study suggests that the risk associated with temporary ATD may not be so devastating as implied in previous reports. This information could be helpful in situations with conflicting risks, such as the unexpected need for major noncardiac surgery in patients with DES, but needs further confirmation because stent thrombosis, even if rare, may have dire consequences.

A total of 343 of 1,965 (17.4%) patients were finally not followed up, in most cases because they refused to participate. Although there were no important differences between both cohorts (Online Appendix), some kind of selection bias is possible. Particularly problematic are the 23 patients from this cohort who died and in whom compliance with antiplatelet medication could not be assessed. Stent thrombosis associated with ATD cannot be ruled out as the ultimate cause of death. As previously reported (10), none of these patients had undergone invasive procedures or hospitalization for hemorrhagic complications before death, which are the most frequent circumstances we found associated with temporary ATD. However, ATD on patient initiative is still possible. As additional information, we could find out that none of these deaths occurred during the first month after stenting and that at least in 13 of them, the cause of death was not cardiac. Similarly, 105 of 1,622 patients (6.47%) were censored before completing 1 year of follow-up because of loss to follow-up, and thus their final status could not be determined. However, most of these follow-up losses occurred during the last follow-up period (i.e., from the 9th to the 12th month) when the risk associated with ATD is likely lower.

A potential recall bias may exist concerning follow-up information. However, most patients received a phone call not later than 3 months after the previous one, and this could minimize such bias. On the other hand, repeated calls may have led patients to be more compliant with medication, so that in fact, the estimated rate of ATD could be lower than it would have been in other contexts. Finally, confounding bias is also possible.

Our study has additional strengths. We systematically evaluated antiplatelet medication resumption after discontinuation with a structured survey, which is an uncommon practice in studies assessing ATD, as is the distinction between both drugs. We did a thorough quality control to ensure consecutive inclusion and quality of data collection, as previously reported (10). Finally, clinical records of readmitted patients were centrally reviewed by the main investigator team.

ATD within the first year and beyond the first month after DES is not exceptional and is usually temporary. Although further knowledge about individual risk is desirable, our results suggest that discontinuation for a few days (median: 7 days) of DAT after the first month of DES implantation may be reasonably safe in terms of major cardiac events.