From June 2008 to May 2010, we prospectively identified 734 patients who were treated by percutaneous coronary intervention (PCI) with either EES or R-ZES for chronic stable coronary artery disease or acute coronary syndrome, including myocardial infarction (MI) with or without ST-segment elevation. Patients who were treated with at least one DES were enrolled in the ‘DES registry of the Seoul National University Bundang Hospital (DES-SNUBH)’ at the catheterization laboratory. There were no restrictions or exclusion criteria for enrollment regarding the reference vessel diameter, total numbers of treated lesion or vessels, numbers of stents implanted, lesion length, referral diagnosis, or comorbidities. Out of 734 patients, 55 patients who were treated with both EES and R-ZES were excluded from the clinical and angiographic outcomes analyses. Therefore, the main analyzed cohort was the remaining 679 patients with 866 lesions (EES 405 patients with 500 lesions; R-ZES 274 patients with 366 lesions) from the registry. Among the main analyzed cohort, 138 patients of EES group (34.2%) and 114 patients of R-ZES group (41.6%) were also enrolled to other Korean multicenter registries (the EXCELLENT and RESOLUTE-Korea registries) under same inclusion and no exclusion criteria [14]. All patients provided written informed consent, and the study protocol was approved by the institutional review board of Seoul National University Bundang Hospital.

Coronary interventions were performed according to current standard techniques. The choice of the stent, predilatation, post-stenting adjunctive balloon inflation, and the use of intravascular ultrasound or glycoprotein IIb/IIIa inhibitors were all left to the operators’ discretion. All patients received at least 100 mg of aspirin before the procedure. A loading dose of 300 to 600 mg of clopidogrel was administered to all patients who were not on clopidogrel prior to the procedure. After the procedure, all patients were given aspirin (at least 100 mg/day) indefinitely and clopidogrel (75 mg/day) for at least 1 year after index procedure. During the procedure, unfractionated heparin at a dose of 70 to 100U per kilogram of body weight was administered to achieve and maintain an activated clotting time of more than 250 seconds.

Clinical follow-up after PCI was recommended at 1 month, 6 months, and 1 year and annually thereafter. Angiographic follow-up was routinely recommended at 9 months, post-PCI or earlier if patients complained of any ischemic symptoms or if noninvasive evaluation suggested the presence of ischemia. The follow-up angiography was performed systematically and in a routine manner, unless the patient rejected the follow-up procedure or the patient’s condition was not suitable to undergo coronary angiography. All patients enrolled in this registry provided a written informed consent not only for the enrollment but also for the follow-up angiography. Coronary angiography obtained at baseline, immediately postprocedure, and at follow-up were analyzed quantitatively (by quantitative coronary angiography [QCA]) using the Cardiovascular Angiography Analysis System (CAAS) II (Pie Medical Imaging, Maastricht, Netherlands) by an experienced technician who was not aware of the study purpose. The minimum lumen diameter (MLD), reference vessel diameter, percent stenosis, acute gain (defined as the difference in MLD before and after stent implantation), late lumen loss (defined as the difference between the postprocedure and follow-up MLD), and binary restenosis (defined as stenosis of 50% or more at follow-up angiography) were measured [15]. All QCA measurements of the target lesion were obtained in the in-stent zone, and over the entire segment including the stent and its 5-mm proximal and distal margins (in-segment zone).

The stent-related clinical outcome was target lesion failure (TLF), defined as a composite of cardiac death, MI (not clearly attributed to a nontarget vessel), or a clinically indicated target lesion revascularization by percutaneous or surgical methods [16]. The patient-related clinical outcome was patient-oriented composite outcome (POCO) which included all-cause mortality, any MI (including nontarget vessel territory), and any revascularization (including all target and nontarget vessels, regardless of percutaneous or surgical methods) [16]. Other clinical outcomes, including MI, stent thrombosis, target lesion revascularization, and target vessel revascularization were defined as according to Academic Research Consortium (ARC) definitions [16]. Patients with complex lesions (considered an off-label indication for use of both DES) were defined as having at least one of the following characteristics: serum creatinine concentration ≥140 umol/L (1.6 mg/dL); left ventricular ejection fraction (LVEF) < 30%; an acute MI within the previous 72 hours; more than one lesion per vessel; two or more vessels treated with a stent; a lesion greater than 27 mm; or a bifurcated lesion, bypass graft, in-stent restenosis, unprotected left main coronary artery, presence of thrombus, or total occlusion [11, 17].

Statistical analyses were performed in three parts. The primary analysis was to compare the angiographic outcomes at 9 months and the secondary analyses were to compare the clinical outcomes. Comparison of angiographic outcomes between two groups was performed in the crude population. Clinical outcome analysis was perfomed on 1) the crude population and 2) the propensity score matched population. Event rates of clinical outcomes were expressed as incidence density (i.e. Kaplan-Meier estimates) and the log-rank or Breslow test was used to compare between-group differences. For the subgroup analysis of TLF, Cox proportional hazard model was used to calculate the hazard ratio of EES compared with R-ZES, and interaction p values between treatment and each subgroup. Since differences in baseline characteristics could impact the development of clinical outcomes, a 1:1 matched analysis without replacement was performed using propensity score which was calculated from the model with 23 covariates (Additional file 1: Table S1). The log-odds of the probability that a patient received an R-ZES were modeled as a function of the confounders. A caliper width of 0.6 SDs was used because this value has been shown to eliminate almost 90% of the bias in the observed confounders [18]. Success of the propensity score matching was assessed by calculating percentage standardized differences of the baseline characteristics. A less than 10% difference supports the assumption of a balance between matched groups [19]. Propensity score adjusted stratified Cox proportional hazard regression model was fitted to compare the clinical outcomes of the matched EES and R-ZES groups. All probability values were two-sided and p-values < 0.05 were considered statistically significant. The statistical package SPSS, version 18.0 (SPSS Inc., Chicago, IL, USA) and R programming language, version 2.12.2 (R Foundation for Statistical Computing) were used for statistical analyses.

During the study period, a total of 1,508 patients received PCI in our institution. One hundred and eighty-one patients were treated with bare-metal stent or balloon angioplasty. Of the remaining 1,327 patients who were treated with DES, 734 patients (55.3%) were treated with EES or R-ZES. Fifty-five patients were excluded since they were treated with EES and R-ZES simultaneously. The flow of participants for the study is presented in Figure 1. The cohort used for the main analysis consisted of 679 patients with 866 lesions (EES 500 lesions/405 patients, R-ZES 366 lesions/274 patients). The baseline clinical characteristics were similar in both groups, except that the proportion of patients with multivessel disease was higher in R-ZES group (Table 1). A total of 501 of 679 (73.8%) patients were classified as complex, which was distributed similarly in both group. EES was more frequently implanted into the left main coronary lesion and the in-stent restenosis lesion than R-ZES; however, R-ZES was more often used for small vessel lesion. As a consequence, the mean stent diameter was significantly smaller in the R-ZES group. The mean in-stent and in-segment MLDs after the procedure were also smaller in R-ZES group than in EES group, despite the similar acute gains in the two groups (Table 2).

Angiographic follow-up at 9 months was available in 445 patients with 564 lesions (65.5%) (Figure 1). Baseline clinical and angiographic characteristics of the patient cohort undergoing QCA analysis at 9 months are listed in Additional file 1: Table S2 and were similar between groups. In-segment late loss was 0.23 ± 0.52 mm for EES and 0.29 ± 0.64 mm for R-ZES (p = 0.248). In-stent late loss showed similar findings (0.24 ± 0.53 mm for EES vs. 0.29 ± 0.58 mm for R-ZES, p = 0.267). In addition, the rate of binary restenosis (5.8% vs. 6.8%, p = 0.716) did not show between-group differences, despite significantly smaller reference vessel diameter in the R-ZES gropup (2.88 ± 0.60 mm vs. 2.77 ± 0.50 mm, p = 0.038) (Table 3).

The median follow-up duration after index procedure was 1014 days (33 months, interquartile range 27.0–38.0 months). There were no significant differences in Kaplan-Meier estimates of TLF (7.5% in the EES group vs. 7.9% in the R-ZES group, p = 0.578) or POCO (22.8% vs. 20.1%, p = 0.888). A total of 9 patients developed ARC-defined definite or probable stent thrombosis. Kaplan-Meier estimates of definite or probable stent thrombosis were not significantly different between the groups (1.5% vs. 1.8%; p = 0.741) (Figure 2 and Table 4). Detailed descriptions of the cases are presented in Additional file 1: Table S3.

Exploratory subgroup analyses regarding TLF were performed according to the presence of diabetes, acute myocardial infarction (< 72 hours), multivessel PCI, long lesion (≥ 28 mm), and small vessel (< 2.75 mm). There were no significant differences between EES and R-ZES, and the results were consistent across all subgroups, with no significant interaction p values (Figure 3).

Matching by propensity score with caliper width of 0.6 SDs yielded 249 EES patients matched to 249 R-ZES patients. Standardized differences of baseline clinical and angiographic characteristics were less than 10%, and both groups were more balanced than before matching, with the exception of bifurcation lesion (percent standardized difference 12.91%, Additional file 1: Table S4). The comparable incidences of clinical outcomes were corroborated in propensity score matched group analysis. The adjusted hazard ratio for TLF and POCO were 0.875 (95% CI 0.427 - 1.793; p = 0.715) and 1.029 (95% CI 0.642 - 1.650; p = 0.904), respectively, for EES over R-ZES (Table 5). Among the propensity-score matched population, 325 patients (65.3% of 498 patients) performed 9-month follow-up angiography. Angiographic outcomes including in-stent and in-segment late loss, and the rate of binary restenosis were comparable between the two groups (data not shown).

Some limitations of this study should be considered. First, this study was a non-randomized observational study using a single center registry data with a relatively small number of patients and was not sufficiently powered to compare clinical outcomes between two stent groups due to limited sample size and low event rates. Second, even though we used propensity score matching and stratified Cox proportional hazard regression modeling to minimize the allocation bias and control for potential confounding variables, the possibilities of uncontrolled and unknown confounding factors need to be considered. Third, because data were from observational registries, the clinical events may not have been captured with scrutiny and patient follow-up may not have been as tight as would be in RCTs. Although we analyzed clinical outcomes up to 3 years of follow-up, a considerable number was lost to follow-up during the second year of follow-up. This can also be another source of bias and may explain lower TLF rate compared with that of RESOLUTE All Comer study at 2 year. However, censoring distributions of the two groups for the clinical outcomes were similar and little influence on the comparisons of two groups is expected. Fourth, the criteria of complex patients or lesional characteristics could not include homogeneous patient population. For example, a patient with severe renal insufficiency which has (at the same time) an acute MI due to left main disease is really different from a patient which receives a stent in an obtuse marginal artery because of in-stent re-stenosis, but both patients would be represented as the same complex group. Fifth, QCA analysis was not performed by an independent core laboratory, but rather by a single analyst. Although the analyst was blinded to the allocated stent group, systematic error cannot be completely excluded. In addition, there were no available data about intravascular ultrasound guidance PCI or the use of post-adjunctive balloon which is important for the optimal procedure. At last, a relatively high number of clinical events occurred around the ninth month during the follow-up, when the follow-up angiography was performed. It has been demonstrated that mandatory angiographic follow-up increases rates of revascularization [20]. We also speculate that the so-called ‘oculo-stenotic reflex’ may have contributed to the revascularization rates in this study. Therefore the present results cannot be directly extrapolated to routine clinical practice where mandatory follow-up angiography is not performed.