Ioannis Iakovou, Thomas Schmidt, Erminio Bonizzoni, Lei Ge, Giuseppe M. Sangiorgi, Goran Stankovic, Flavio Airoldi, Alaide Chieffo, Matteo Montorfano, Mauro Carlino, Iassen Michev, Nicola Corvaja, Carlo Briguori, Ulrich Gerckens, Eberhard Grube, Antonio Colombo. Incidence, Predictors, and Outcome of Thrombosis After Successful Implantation of Drug-Eluting Stents. JAMA. 2005;293(17):2126–2130. doi:10.1001/jama.293.17.2126
Author Affiliations: Centro Cuore Columbus
and San Raffaele Hospital, Milan, Italy (Drs Iakovou, Ge, Sangiorgi, Stankovic,
Airoldi, Chieffo, Montorfano, Carlino, Michev, Corvaja, Briguori, and Colombo);
Mediolanum Cardio Research, Milan, Italy (Dr Sangiorgi); Department of Cardiology,
Klinikum Siegburg Rhein-Sieg GmbH, Siegburg, Germany (Drs Schmidt, Gerckens,
and Grube); and Institute of Medical Statistics and Biometry, University of
Milan, Milan, Italy (Dr Bonizzoni).
Context Traditionally, stent thrombosis has been regarded as a complication
of percutaneous coronary interventions during the first 30 postprocedural
days. However, delayed endothelialization associated with the implantation
of drug-eluting stents may extend the risk of thrombosis beyond 30 days. Data
are limited regarding the risks and the impact of this phenomenon outside
Objective To evaluate the incidence, predictors, and clinical outcome of stent
thrombosis after implantation of sirolimus-eluting and paclitaxel-eluting
stents in routine clinical practice.
Design, Setting, and Patients Prospective observational cohort study conducted at 1 academic hospital
and 2 community hospitals in Germany and Italy. A total of 2229 consecutive
patients underwent successful implantation of sirolimus-eluting (1062 patients,
1996 lesions, 2272 stents) or paclitaxel-eluting (1167 patients, 1801 lesions,
2223 stents) stents between April 2002 and January 2004.
Interventions Implantation of a drug-eluting stent (sirolimus or paclitaxel). All
patients were pretreated with ticlopidine or clopidogrel and aspirin. Aspirin
was continued indefinitely and clopidogrel or ticlopidine for at least 3 months
after sirolimus-eluting and for at least 6 months after paclitaxel-eluting
Main Outcome Measures Subacute thrombosis (from procedure end through 30 days), late thrombosis
(>30 days), and cumulative stent thrombosis.
Results At 9-month follow-up, 29 patients (1.3%) had stent thrombosis (9 [0.8%]
with sirolimus and 20 [1.7%] with paclitaxel; P = .09).
Fourteen patients had subacute thrombosis (0.6%) and 15 patients had late
thrombosis (0.7%). Among these 29 patients, 13 died (case fatality rate, 45%).
Independent predictors of stent thrombosis were premature antiplatelet therapy
discontinuation (hazard ratio [HR], 89.78; 95% CI, 29.90-269.60; P<.001), renal failure (HR, 6.49; 95% CI,
2.60-16.15; P<.001), bifurcation lesions (HR, 6.42;
95% CI, 2.93-14.07; P<.001), diabetes (HR, 3.71;
95% CI, 1.74-7.89; P = .001), and a lower
ejection fraction (HR, 1.09; 95% CI, 1.05-1.36; P<.001 for each 10% decrease).
Conclusions The cumulative incidence of stent thrombosis 9 months after successful
drug-eluting stent implantation in consecutive “real-world” patients
was substantially higher than the rate reported in clinical trials. Premature
antiplatelet therapy discontinuation, renal failure, bifurcation lesions,
diabetes, and low ejection fraction were identified as predictors of thrombotic
Despite major improvements in antiplatelet therapy, thrombotic events
remain the primary cause of death after percutaneous coronary interventions.1,2 Sirolimus-eluting stents and polymer-based
paclitaxel-eluting stents have been shown to reduce neointimal hyperplasia
and risk of restenosis without increasing the risk of stent thrombosis.3- 7 Operators
are now using drug-eluting stents for a wide variety of clinical and anatomic
situations, many of which have not been evaluated in randomized studies.8- 10 We analyzed the incidence,
predictors, and clinical outcome of stent thrombosis at 9-month follow-up
in an observational cohort study.
We identified 2229 consecutive patients who underwent successful implantation
with sirolimus-eluting stents (1062 patients, 1996 lesions, 2272 stents) or
paclitaxel-eluting stents (1167 patients, 1801 lesions, 2223 stents) between
April 2002 and January 2004. Patients were treated at 2 community hospitals
or 1 academic hospital in Germany and Italy. Patients with ST-elevation acute
myocardial infarction (MI) less than 48 hours before the procedure, with intraprocedural
stent thrombosis, and those treated with both types of stents were excluded.
All patients were pretreated with ticlopidine or clopidogrel and aspirin;
a loading dose of 300 mg of clopidogrel was given to patients not previously
taking the agent. Aspirin was continued indefinitely and clopidogrel or ticlopidine
for at least 3 months after sirolimus-eluting stent implantation and for at
least 6 months after paclitaxel-eluting stent implantation. Stent implantation
methods have been described previously.11 Glycoprotein
IIb/IIIa inhibitors were administered at the physician’s discretion.
Standard qualitative and quantitative analyses and definitions were used for
the angiographic analysis.12
All patients signed an informed consent document and local institutional
review boards approved the study as planned.
Stent thrombosis was determined as the occurrence of any of the following
events: angiographic documentation of partial or total stent occlusion detected
within 30 days of the procedure (an acute clinical ischemic event in addition
to angiographic documentation had to be present when the event occurred after
30 days), or sudden cardiac death or postprocedural MI after successful stent
implantation not clearly attributable to another coronary lesion. Stent thrombosis
cases were categorized according to the timing of occurrence into subacute
(from procedure end through 30 days) and late (>30 days).
Major adverse cardiac events were defined as death (all-cause), Q-wave
MI, target lesion revascularization, and target vessel revascularization.
Differences in proportions were tested with the χ2 or
Fisher exact test. SAS version 8.2 (SAS Institute Inc, Cary, NC) was used
for data analysis.
A total sample of 2300 observations was computed to achieve 80% power
at a 2-sided .05 significance level to detect a hazard ratio (HR) equal to
or greater than 3.0 with a Cox regression of the log HR on a binary risk factor
with a 25% or greater prevalence. The sample size was adjusted for an anticipated
event rate of 1.5%.
Relationships of event incidence to covariates were investigated with
univariate Cox regression models. The proportional hazard assumption was checked
for all screened covariates and no relevant violations were found. The predictive
robustness of univariate findings was subsequently tested by means of a bootstrap
subset selection method in which multivariable Cox regression analysis using
a stepwise elimination process was repeated for each of 1000 bootstrap samples.13 The relative frequency of selection of “important”
variables was used as a criterion for inclusion of predictors in the final
multivariable model. Variables were retained if the selection frequency of
the bootstrap samples was at least 50% or if they were clinically relevant.
Because of the small absolute number of events, we performed an internal validation
process to test model adequacy and quantify “overfitting.”14
The proportion of total variability was estimated by means of the Nagelkerke
Index (pseudo R2), which was calculated
when the models were fitted to 1000 bootstrap replications (training sample)
and to the original data (test sample). The Nagelkerke Index obtained from
each bootstrap sample was then subtracted from the initial index value of
the original population. The average of the differences was considered as
a measure of optimism in the model fit, where optimism is the proportion of
variability ascribed to overfitting. Finally, a corrected index was calculated
by subtracting the average of the optimism estimates from the original Nagelkerke
Index. The estimates of slope shrinkage, which were used to identify overparameterized
models, were obtained for the multivariable models using the same bootstrapping
validation process. Thus we were reassured that, although based on few events,
overfitting did not substantially bias our final model.
Baseline, angiographic, and procedural characteristics are shown in Table 1. All stents were deployed successfully.
There were no significant differences between the 2 stent groups regarding
procedural complications and in-hospital outcome. The rates for Q-wave and
non–Q-wave MI were 0.3% and 9%, respectively. There were 4 in-hospital
deaths, of which 2 were determined to be caused by stent thrombosis.
At 9-month follow-up (available in all patients), 29 patients (1.3%)
had stent thrombosis (9 [0.8%] in the sirolimus group and 20 [1.7%] in the
paclitaxel group; P = .09). Fourteen patients
had subacute thrombosis (0.6%), 4 in the sirolimus group and 10 in the paclitaxel
group (0.4% vs 0.8%; P = .19) and 15 patients
had late thrombosis (0.7%), 5 in the sirolimus group and 10 in the paclitaxel
group (0.5% vs 0.8%; P = .30).
A total of 71% (10/14) of the subacute cases occurred within 1 week
of the procedure (median, 4 days) and 53% (8/15) of the late thrombosis cases
occurred within 3 months of the procedure (median, 57 days). Seven cases (24%)
presented as death, 20 (69%) as nonfatal MI, and 2 (7%) as unstable angina.
At follow-up the case-fatality rate—including death at presentation—was
45% (13/29). Angiographic documentation of thrombosis was available in 12
of the 22 patients (55%) with stent thrombosis (including both patients that
presented with unstable angina) that did not present with death. In-hospital
data were available for all patients who presented with acute MI and had no
angiographic evidence of stent thrombosis.
At 9-month follow-up, target lesion revascularization was performed
in 141 patients (6.3%) (58 [5.5%] in the sirolimus group vs 83 [7.1%] in the
paclitaxel group; P = .13). Major adverse
cardiac events occurred in 242 patients (10.8%) (109 [10.3%] in the sirolimus
group vs 133 [11.4%] in the paclitaxel group; P = .42).
The incidence of stent thrombosis according to selected patient characteristics
and the univariate predictors of cumulative stent thrombosis are shown in Table 2. Five of 17 patients with premature antiplatelet
therapy discontinuation had stent thrombosis; in 1 of them only clopidogrel
was discontinued. Independent predictors of subacute, late, and cumulative
stent thrombosis are shown in Table 3.
The key predictors of stent thrombosis were premature antiplatelet therapy
discontinuation, renal failure, bifurcation lesions, diabetes, and low ejection
fraction. For subacute thrombosis, stent length was also a predictor: for
each 1-mm increase in length, there was 1.03 times greater risk of thrombosis.
In a large cohort of consecutive patients undergoing drug-eluting stent
implantation, we noted a 9-month cumulative stent thrombosis incidence of
1.3%, substantially higher than rates reported in major clinical trials (0.4%
at 1 year for sirolimus and 0.6% at 9 months for paclitaxel).3,5 With
widespread availability of drug-eluting stents, the scope of percutaneous
coronary intervention has been expanded to more complex lesions and patients.
In our study, 27% of the population had diabetes and 79% of the lesions were
complex. The clinical consequences of stent thrombosis were severe, with a
case-fatality rate of 45%.
Similar to previous reports of both bare-metal stents2,16,17 and
drug-eluting stents,18,19 our
study found that premature discontinuation of antiplatelet therapy was the
most important predictor of stent thrombosis after implantation. Thrombosis
occurred in 29% of patients who prematurely discontinued dual antiplatelet
therapy, making treatment adherence of paramount importance. Nonetheless,
the absolute numbers of events in our cohort were small, leading to wide confidence
intervals for estimation of effect magnitude.
In addition to premature discontinuation of antiplatelet therapy, other
key predictors of stent thrombosis were renal failure, bifurcation lesions,
diabetes, low ejection fraction, and, for subacute thrombosis, stent length.
Renal failure has been linked to cardiac disease with microvascular
and metabolic abnormalities that may predispose to thrombus formation.20,21 Renal failure also has been associated
by numerous studies with an increased mortality rate despite successful coronary
Regarding bifurcational lesion location, pathology studies have suggested
that arterial branch points are foci of low shear and low flow velocity and
are sites predisposed to the development of atherosclerotic plaque, thrombus,
and inflammation.25- 27 The
observed associations of diabetes,28- 32 low
ejection fraction,2,33 and stent
length2 with stent thrombosis are also consistent
with previous reports.
In our study, stent type did not emerge as an independent predictor
of thrombosis. Of concern, however, we found an almost double incidence of
stent thrombosis after paclitaxel compared with sirolimus stent implantation.
These findings are consistent with those of the ISAR-DESIRE (Intracoronary
Stenting and Antithrombotic Regimen–Drug-Eluting Stents for In-Stent
Restenosis) trial of drug-eluting stents for in-stent restenosis, in which
sirolimus-eluting stents tended to be associated with a better outcome than
paclitaxel-eluting stents.6 However, without
larger numbers it is difficult to make firmconclusions; subsequent randomized
trials will be critically important.
A limitation of the current study was the lack of angiographic documentation
in all cases adjudicated as stent thrombosis. Nevertheless, the definition
of stent thrombosis that we proposed is similar to the one used in important
recent studies extended beyond 30 days.3,19
In conclusion, the incidence of stent thrombosis at 9 months after successful
drug-eluting stent implantation in consecutive real-world patients was 1.3%.
Premature antiplatelet therapy discontinuation, bifurcation lesions, and low
ejection fraction were identified as independent predictors of subacute, late,
and cumulative stent thrombosis. In addition, stent length was also recognized
as a predictor of subacute thrombosis, and renal failure and diabetes as predictors
of both subacute and cumulative stent thrombosis. The clinical consequences
were death in 45% of patients and nonfatal MI in the majority of the others.
Corresponding Author: Antonio Colombo, MD,
EMO Centro Cuore Columbus, 48 Via M Buonarroti, 20145 Milan, Italy (firstname.lastname@example.org).
Author Contributions: Dr Colombo had full access
to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Iakovou, Schmidt,
Acquisition of data: Ge, Sangiorgi, Stankovic,
Airoldi, Chieffo, Montorfano, Carlino, Michev, Corvaja, Gerckens, Grube.
Analysis and interpretation of data: Iakovou,
Drafting of the manuscript: Iakovou, Colombo.
Critical revision of the manuscript for important
intellectual content: Schmidt, Bonizzoni, Ge, Sangiorgi, Stankovic,
Airoldi, Chieffo, Montorfano, Carlino, Michev, Corvaja, Briguori, Gerckens,
Statistical analysis: Iakovou, Bonizzoni.
Study supervision: Schmidt, Sangiorgi, Chieffo,
Montorfano, Carlino, Briguori, Gerckens, Grube, Colombo.
Financial Disclosures: None reported.
Previous Presentation: This study was presented
at the American Heart Association Scientific Sessions; November 7-10, 2004;
New Orleans, La.