Despite improved stent implantation technologies and more effective antiplatelet regimens, stent thrombosis (ST) continues to occur with an estimated incidence varying between 1% and 5% (1- 6). This wide variability in the estimated incidence strongly indicates the multifactorial nature of the phenomenon of ST.

Several observational studies have identified a number of clinical, angiographic, and procedural determinants of ST. These studies are hampered, however, by small sample size, retrospective character of study design, and variation in the definition of ST (1- 5,7). Consequently, less-frequent but clinically meaningful risk factors could not be explored, and the multivariate analysis model was overfitted in the majority of studies. Furthermore, these studies did not focus on possible differences in underlying pathophysiological mechanisms between: 1) different indications of stent implantation (stable angina vs. acute coronary syndromes [ACS]); and 2) early versus late ST.

Because recognition of risk factors attributable to ST might help to improve prognosis by the development of a risk-stratification model, we sought to identify predictors of early and late ST and to determine predictors of ST in different populations (stable angina and ACS, non–ST-segment elevated myocardial infarction [NSTEMI] and ST-segment elevated myocardial infarction [STEMI]) in a “real world” of mixed bare-metal stent (BMS) and drug-eluting stent (DES) use. Given the detailed character of our study, we were also able to study the impact of the clopidogrel discontinuation on the occurrence of ST at the different time points after the index procedure.

The Dutch stent thrombosis registry is a large-scale, multi-centre study conducted in 3 high-volume centers in the Netherlands (>2,500 interventions/center/year). All consecutive patients with an angiographically confirmed ST presenting to the participating centers from January 2004 to February 2007 were enrolled. Stent thrombosis was defined according to the Academic Research Consortium criteria for “definite” ST (8). Clinical criteria consisted of a new episode of chest pain and/or ischemic electrocardiographic changes and/or increase of cardiac biomarkers release. Angiographic criteria consisted of partial or complete occlusion within the previously implanted stent with evidence of fresh thrombus. On the basis of the elapsed time since stent implantation, ST was classified as acute (intraprocedural or within 24 h of the procedure), subacute (from 24 h to 30 days), late (>30 days to 1 year), or very late (>1 year). Acute and subacute ST were also defined as early ST. Likewise, late and very late ST were defined as late ST.

Patients who underwent percutaneous coronary intervention (PCI) with stent implantation but with no evidence of ST during follow-up were recruited and served as control subjects in a 2:1 ratio. Control subjects were individually matched to case subjects by the following criteria: 1) similar indication (either stable angina or ACS [NSTEMI, STEMI]) for the index-PCI procedure; 2) same date of the index PCI procedure (±1 day in a minority of STEMI patients); and 3) same performing institution.

The PCI was performed by standard techniques via the femoral approach in most cases. During PCI, patients were anticoagulated with 70 IU/kg of unfractionated heparin. All patients were treated with aspirin (80 to 100 mg) before PCI and were loaded with clopidogrel (300 to 600 mg) if they were not taking maintenance therapy. Aspirin was continued indefinitely. The recommended duration of clopidogrel therapy after the index PCI varied from 4 weeks after BMS implantation during elective angioplasty to 3 to 12 months with DES. For patients presenting with ACS, 12 months clopidogrel therapy was recommended, regardless of the stent type used. The use of adjunctive devices (e.g., thrombus aspiration catheter) or glycoprotein IIb/IIIa therapy was at the operators' discretion.

Detailed data on patient, angiographic, and procedural characteristics for both the cases and control subjects were collected. Comprehensive information about the use of antithrombotic therapy (i.e., aspirin, clopidogrel, Coumadin) at the time of the index PCI was also collected. The duration of clopidogrel use as well as aspirin compliance after patient discharge was assessed with telephonic patient interview as well as data from pharmacy records (the date of clopidogrel dispensed and the number of days supplied for each dispense). In case of disagreement, the pharmacy data were used for the analysis.

Angiograms of both the cases and the control subjects were reviewed by 2 experienced interventional cardiologists who were blinded to the objectives of this study and outcome data. Calcification was identified as readily apparent radio-opacities within the vascular wall. Angiographic thrombus was defined as a filling defect seen in multiple projections surrounded by contrast in the absence of calcification. Angiographically visible, uncovered dissections were graded according to the modified classification of the Heart, Lung, and Blood Institute (9). Given the relatively low incidence of coronary dissections, the patient population was analyzed on the basis of the presence or absence of any dissection type. The sizing of the implanted coronary stent(s) was evaluated by visual estimate. It is important to note that the stent deployment was judged on the basis of angiographic appearance and not by quantitative coronary analysis or intravascular ultrasound (IVUS) analysis. Undersizing of the coronary stent(s) was considered significant as 1 of the following criteria was met: 1) the stent to the reference segment diameter ratio was <1; 2) inappropriate alignment of the coronary stent with the coronary vessel wall; and 3) mismatch in post-deployment stent dimensions in relation to the proximal and distal segments of the target vessel.

The presence of intermediate coronary artery disease (CAD), defined as a visually estimated percentage of coronary stenosis of ≥50% but ≤70%, proximal and distal to the stented segment(s) of the target vessel were also scored.

Continuous variables were presented as mean ± SD and were compared with the Student t test or Mann-Whitney U test. The chi-square or Fisher exact test was used to analyze differences in categorical variables. Conditional logistic regression analysis was performed to determine independent predictors of ST. Selected variables were first entered into the univariate analysis. Variables with p < 0.05 by univariate analysis were then entered in the conditional logistic regression analysis for identification of predictors of ST.

To study the impact of the determinant “lack of clopidogrel therapy at different time points after the index PCI (<24 h, 1 to ≤30 days, >30 days to 6 months, and beyond)” a multivariable Cox proportional hazards model was created with “lack of clopidogrel therapy at the time of ST” as time varying covariates. For this analysis, it was assumed that the time-interval between index PCI and the “virtual ST” of the control patients was exactly the same as that for its matched case. All patients with ST and their matched control subjects were also further subdivided into 3 groups: 1) those who were taking clopidogrel at the (“virtual”) time of ST; 2) those who had discontinued the clopidogrel within 14 days before the ST; and 3) those who had discontinued the clopidogrel longer than 14 days before the ST (it is important to note that we also made an attempt to perform a sensitivity analysis with varying time-intervals for the time between clopidogrel cessation and the ST; however, due to the relatively small patient groups with late and very late ST who discontinued the clopidogrel treatment, no reliable estimates could be provided when the time interval between clopidogrel cessation and the occurrence of ST exceeded 14 days; for the first 14 days, the hazard ratio [HR] remained relatively constant [data not shown]).

The independent “baseline” predictors of ST from conditional logistic regression analysis were also included in this model. All tests were 2-tailed and used a p value <0.05 to characterize statistical significance.

During the study period, a total of 21,009 patients underwent stent implantation in the participating hospitals. A total of 31,065 stents were implanted (19,840 BMS and 11,225 DES). As expected, there were significant differences in age, prevalence of cardiovascular risk factors, lesion characteristics, and the prevalence of other clinical comorbidities between patients who received a BMS and patients who received a DES (Table 1). During a median follow-up of 30.9 months (25th to 75th percentiles: 23.6 to 41.9 months), 437 patients (2.1%) presented with an angiographic confirmed ST. According to the different categories of ST, 140 (32.0%) were acute ST, 180 (41.2%) were subacute ST, 58 (13.3%) were late ST (36 within 6 months), and 59 were very late (13.5%). Two-hundred seventy STs were related to a BMS (cumulative incidence: 2.2%), 152 stent thromboses were related to a DES (cumulative incidence: 2.0%), and 15 were related to both a BMS and a DES stent (mixed use, cumulative incidence: 1.8%). The cumulative incidence of ST was not significantly different between the 2 different types of coronary stents: p = 0.38).

We were able to match 866 control subjects to the case subjects (99.0%). Detailed clinical, procedural, and angiographic features of the patients with ST and the matched control subjects are described in (Table 2).

Independent clinical, procedural, and angiographic predictors of ST when comparing all cases with all matched control subjects in multivariate analysis are depicted in (Figure 1). Cessation of clopidogrel at various time points, undersizing, present malignant disease, the presence of intermediate CAD proximal to the culprit lesion, suboptimal procedural result (Thrombolysis In Myocardial Infarction [TIMI] flow grade <3 after PCI), uncovered dissection, bifurcation stenting, left ventricular ejection fraction (LVEF) <30%, peripheral artery disease, the presence of intermediate CAD distal to the culprit lesion, no aspirin therapy, diabetes mellitus, use of any DES, and younger age were associated with the occurrence of ST.

With regard to the different indications for stent implantation, the cumulative incidence of ST in patients who underwent an elective PCI for the indication stable angina pectoris was low (113 of 11.207 patients, cumulative incidence: 1.00%). However, the cumulative incidence of ST was higher when the indication for index stent implantation was unstable angina/NSTEMI (72 of 3,960 patients, cumulative incidence: 1.8%) and STEMI (252 of 5,842, cumulative incidence: 4.3%). The timing of ST was also significantly different throughout the different indications for stent implantation (p = 0.0003). The STEMI patients experienced an early ST significantly more often (79% of all STs in STEMI patients were early vs. 65% of all STs in stable angina and NSTEMI patients), whereas the proportion of late and very late ST was higher in the stable angina and NSTEMI group.

Determinants of ST for the different indications of index stent implantation are depicted in (Figure 2). Independent factors that predispose to the development of ST in patients undergoing elective PCI with stent implantation for the indication stable angina were undersizing, the presence of intermediate CAD proximal to the culprit lesion, malignant disease, suboptimal procedural result (TIMI flow grade <3 after PCI), LVEF <30%, uncovered dissection, multivessel disease, left descending coronary artery stenting, and long total stent length. Predictors of ST in the setting of ACS (including STEMI) as the indication for index PCI were undersizing, suboptimal procedural result (TIMI flow grade <3 after PCI), uncovered dissection, the presence of intermediate CAD proximal to the culprit lesion, bifurcation lesion, any DES, no aspirin therapy, LVEF <30%, the presence of intermediate CAD distal to the culprit lesion, and multivessel disease. Of note, periprocedural use of glycoprotein IIb/IIIa therapy for the indication ACS (including STEMI) was associated with a reduction of ST. It is important to note that the time-varying covariable “cessation of clopidogrel” was not included in these multivariate models.

Almost 75% of the STs occurred within 30 days after stent implantation. (Figure 3) displays the independent predictors of early ST (≤30 days after the index PCI) with associated odds ratios and 95% confidence interval (CI) and the independent predictors of late ST (>30 days after the index PCI). Early predictors of ST included undersizing, uncovered dissection, suboptimal procedural result (TIMI flow grade <3 after PCI), the presence of intermediate CAD proximal to the culprit lesion, present malignant disease, no aspirin, LVEF <30%, bifurcation lesion, the presence of intermediate CAD distal to the culprit lesion, any DES, and total number of stents. Glycoprotein IIb/IIIa was protective for the occurrence of early ST.

The following determinants were independently associated with the occurrence of late ST: undersizing, present malignant disease, the presence of intermediate CAD proximal to the culprit lesion, peripheral artery disease, diabetes mellitus, bifurcation lesions, long total stent length, and younger age.

Again, it is important to note that the time-varying covariable “cessation of clopidogrel” was not included in these multivariate models.

The proportion of cases and matched control subjects that were receiving clopidogrel therapy for the time-frames of the different categories of ST is presented in (Figure 4). In detail, a total of 134 (30.7%) cases were not taking clopidogrel therapy at the time of the ST. Of these, in 9 of 140 (6.4%) patients presenting with an acute ST, the clopidogrel was erroneously not initiated; 30 of 179 (16.7%) patients with a subacute ST had discontinued the clopidogrel for a median of 5 days (interquartile range: 3 to 7 days); 39 of 58 (67.2%) patients with a late ST had discontinued the clopidogrel for a median of 13 days (interquartile range: 7 to 61 days); and 56 of 59 (94.9%) patients with a very late ST had discontinued the clopidogrel for a median of 200 days (interquartile range: 23 to 981 days).

After applying the exact elapsed time-frame between the index PCI and ST for every case to their matched control subjects, we show that a significant higher proportion of control subjects were taking clopidogrel therapy at the “virtual time” of occurrence of ST (Figure 4). The lack of clopidogrel therapy at the time of ST in the first 30 days after the index PCI was strongly associated with ST (HR: 36.5, 95% CI: 8.0 to 167.8). Likewise, the lack of clopidogrel therapy at the time of ST between 30 days and 6 months after the index PCI was also linked to the occurrence of ST (HR: 4.6, 95% CI: 1.4 to 15.3). Multivariate Cox proportional hazard analysis also found that discontinuation of clopidogrel therapy after 6 months from the index stent implantation was a predictor of ST (HR: 5.9, 95% CI: 1.7 to 19.8).

Alternatively, given the fact that clopidogrel irreversibly inhibits the human platelet throughout its entire lifespan (10 to 12 days), we hypothesized that cessation of clopidogrel <14 days before the ST would reveal the temporal relationship between the discontinuation of clopidogrel and ST. After introducing this time-varying covariate in the multivariate Cox proportional hazard model, “cessation of clopidogrel within 14 days before ST” in the first 30 days after the index PCI emerged as a highly significant predictor of ST (HR: 36.9, 95% CI: 7.9 to 173.3). Similarly, “cessation of clopidogrel within 14 days before ST” between 30 days and 6 months after the index PCI was also independently associated with the occurrence of ST (HR: 21, 95% CI: 2.2 to 198.3). The number of events beyond the 6-month time-frame after stent implantation was too small to reliably estimate the impact of “cessation of clopidogrel and the subsequent occurrence of ST within 14 days” on the occurrence of ST in this subcategory of patients.

Another important finding relates to the magnitude of impact of “cessation of clopidogrel and the subsequent occurrence of ST within 14 days” between DES and BMS. The risk for ST associated with “cessation of clopidogrel and the subsequent occurrence of ST within 14 days” was significantly higher in patients who had received a DES as compared with those who had received a BMS (odds ratio for DES: 1.88, 95% CI: 1.21 to 2.94, p = 0.0052).

A significantly higher percentage of the patients with ST did not use aspirin therapy at the time of the index as compared with their matched control (13.1% vs. 4.5%, p < 0.0001). The predominant reasons for no aspirin treatment were Coumadin use in 85 of 96 (84.4%) patients and allergy to aspirin in 7 of 96 (7.3%) patients. Multivariate logistic regression analysis identified that the absence of aspirin therapy was also a strong independent predictor of ST (HR: 1.91 95% CI: 1.01 to 3.88, p = 0.0487).

Given the devastating consequences of ST, great efforts should be directed to identify those patients at highest risk, who would probably benefit most from an alternative strategy. The findings of the present study add considerably to the understanding of the profiles of patients at high risk for ST.

Although previous studies have already recognized multiple risk factors that confer a significant risk of ST (1- 3,5,7,10), many of these studies have limitations, mostly related to an overall small sample size with a limited number of cases. Also, the identified determinants of ST represent those that have been looked for, and many theoretical likely factors (such as present malignancy, severity of atherosclerotic disease, aspirin use) are not investigated in most published studies.

The case-control design of our study as well as the acquisition of very detailed data on medical history, medication use, and angiographic characteristics enabled us to comprehensively examine the most important risk factors that are associated with ST. Moreover, the large sample size allowed identification of relatively infrequent determinants.

The highly variable duration of clopidogrel use throughout the years of patient recruitment (2004 to 2007) allowed us to comprehensively study the impact of early clopidogrel cessation after stent implantation. As expected, lack of clopidogrel therapy at the time of the ST in the first 6 months after the index PCI was identified as the strongest independent predictor of ST, and this observation is in line with some (1- 2,11) but not all (3,7) previous reports. Also, we predefined a likely “vulnerable time-frame” between cessation of clopidogrel and the occurrence of ST (<14 days) and demonstrated that this is the period that patients are at the highest risk for ST, especially when the discontinuation of clopidogrel was within the first 6 months after stent implantation.

A novel and very important finding of the present study, contrary to a recently published study (7), is that the lack of clopidogrel therapy (but not necessarily cessation of clopidogrel within the 14 days preceding the ST) beyond 6 months after index PCI was a predictor of ST. A likely explanation for this difference might be that previous studies were hampered by a very low number of events (16 patients, of whom 7 did not use clopidogrel at time of ST) >6 months after coronary stent implantation (7). Nonetheless, our results should also be interpreted with caution, because only a small portion of patients with ST beyond 6 months had discontinued the clopidogrel in the 14 days preceding the ST.

It remains pure speculative as to why “lack of clopidogrel therapy after 6 months from the index stent implantation” but not “cessation of clopidogrel and the occurrence of ST (<14 days)” beyond 6 months was a predictor of late and very late ST. Perhaps the loss of protection by clopidogrel therapy rather than a “rebound in platelet reactivity” might explain these findings.

Undersizing of the coronary stent was the second strongest predictor of ST in our study. Indeed, previous studies have elucidated the importance of correct sizing of coronary stents, in particular in patients with a high thrombotic burden with subsequent vasoconstriction or severe and diffuse target vessel disease (12- 16). Previous reports have suggested that, despite improvements in techniques and materials in the last decade, the incidence of incomplete stent deployment and undersizing ranges from 20% to 30%, and this percentage is even higher when assessed by IVUS (15,17). Notwithstanding the results of previous studies suggesting that the judgment of the correct sizing and/or deployment of coronary stents is superior with the use of IVUS (18), we demonstrated that the identification of undersizing of a coronary stent by the simple means of eyeballing is a strong predictor of ST. In most of the cases, undersizing was probably due to severe calcification or related to a high thrombotic burden with subsequent vasoconstriction. However, incorrect judgment of the true coronary vessel size by the performing operator is also a likely explanation for undersizing in a considerable number of cases.

The long-term safety of DES has been a main topic of debate at recent international cardiology meetings. In our study, patients who received a DES had higher baseline risk-profiles and more complex lesion characteristics. Nonetheless, the use of a DES was not independently associated with late or very late ST. In contrast, any DES use was independently associated with ST in both the all-cases versus all-control subjects analysis as well as in the sub analysis on risk factors for early ST.

Several previous studies have elucidated the importance of mechanical (both angiographic and procedural factors) etiologies underlying early ST (12,15,17). Given the more complex lesion characteristics (an off-label DES indication) and limited flexibility of the first generation of DES, uncovered endothelial damage during DES implantation might explain this association between early ST and DES use.

Several limitations of the present study need to be acknowledged. First, notwithstanding that a case-control design enables the evaluation of rare events such as ST, a case-control design is at the same time vulnerable to several sorts of bias. Indeed, 3 control patients were excluded from analysis because they were admitted for ST in another nonparticipating hospital during follow-up. However, given the size of the control group (2:1 ratio), the random selection of control subjects and the extent of differences in clinical procedural and angiographic findings between cases and control subjects, it is unlikely that the results would have changed considerably by increasing the number of control subjects. Second, only angiographically documented cases with ST (“definite”) were reported. This might have led to an underestimation of the actual incidence of ST, because patients who had suffered from a sudden cardiac death or from a silent stent occlusion were not included in our analysis. Third, one might question our evaluation of the sizing of coronary stents by simple means of eyeballing instead of using sophisticated techniques such as IVUS or quantitative coronary analysis. Nonetheless, visual estimation was able to detect undersizing in 18.1% of the cases and in only 2.3% of the control subjects, which makes it a valuable tool.

We identified several important predictors of ST in the contemporary era of mixed DES and BMS use. Discontinuation of clopidogrel, undersizing of the coronary stent, presence of intermediate CAD proximal to the culprit lesion, and concomitant malignant disease were the strongest predictors of ST. Risk factors for ST also vary throughout the different indications for PCI (stable angina vs. ACS) and differ for the different categories of ST (early vs. late ST).