We conducted this study using the Coronary Bifurcation Stenting (COBIS) Registry. The methods and patient population of the COBIS registry have been described previously (7). Briefly, the COBIS registry is a retrospective multicenter registry dedicated to bifurcation lesion PCI that includes data for patients treated at 16 major coronary intervention centers in Korea. Inclusion criteria were: 1) coronary bifurcation lesions which were treated solely with at least 1 drug-eluting stent (DES) between January 2004 and June 2006; and 2) a main vessel diameter ≥ 2.5 mm and side branch diameter ≥ 2.0 mm. Exclusion criteria were the presence of cardiogenic shock, ST-segment elevation myocardial infarction (MI) within 48 hr, and left main coronary artery bifurcation lesions. Among 1,668 patients, a total of 753 patients were available for analysis of in-hospital bleeding. The registry was sponsored by the Korean Society of Interventional Cardiology.

The vascular approach method was decided by the operator's discretion for each patient. Intravascular ultrasound (IVUS) guided stenting was encouraged to achieve optimal stent expansion and lesion coverage. Stent type was selected per the treating physician's discretion at the time of the study, namely, sirolimus-eluting Cypher stent (Cordis, Johnson & Johnson, New Brunswick, NJ, USA), paclitaxel-eluting Taxus stent (Boston Scientific, Natick, MA, USA), and other. Stent techniques to treat coronary bifurcation lesions included: main vessel stenting only, T-stenting, crush, V-stenting, culotte, and other. Final kissing ballooning (FKB) was performed in cases with suboptimal results after crossover stenting at the ostium of the side branch and, in cases with 2-stent implantations.

All patients were administered loading doses of aspirin (300 mg) and clopidogrel (300 to 600 mg) or ticlopidine (500 mg) before coronary intervention, unless they had previously received these antiplatelet medications. During the procedure, anticoagulation therapy was performed according to current practice guidelines by the Korean Society of Interventional Cardiology (8), and use of glycoprotein IIb/IIIa inhibitors was determined by the operator. After procedures, aspirin was prescribed indefinitely and clopidogrel or ticlopidine was prescribed for at least 6 months.

Demographic, clinical, angiographic, procedural, and outcome data were collected with the use of a web-based reporting system. Additional information was obtained by further inquiries into medical records or telephone contact if necessary. All baseline and procedural cine coronary angiograms were reviewed and qualitatively analyzed at the angiographic core laboratory in the Cardiac and Vascular Center, Samsung Medical Center, Seoul, Korea. All patients were evaluated by a clinic visit or by phone at 1, 3, 6, and 12 months and annually thereafter.

Baseline bifurcation anatomy was assessed according to the Medina classification (9). Medina type (1.1.1), (1.0.1), and (0.1.1) lesions were defined as true bifurcation lesions. Angiographic success was defined as the achievement of a Thrombolysis In Myocardial Infarction (TIMI) flow grade of 3 with a final residual stenosis of < 30% for the main vessel or < 50% for the side branch by visual estimation. Procedural success was defined as angiographic success without in-hospital death, MI, or emergent coronary artery bypass surgery. After PCI, elevations of the creatine kinase-myocardial band greater than 3 times the upper limit of normal, assuming a normal baseline biomarker value, were considered indicative of peri-procedural MI (10).

The peri-procedural period included the first 48 hr after PCI. Peri-procedural complications were defined as the occurrence of coronary arterial dissection or abrupt closure during PCI, cardiac death, emergent bypass surgery, coronary perforation or cardiac tamponade during PCI. Death that could not be attributed to non-cardiac etiology was considered cardiac death. MI was diagnosed by electrocardiographic changes and/or a rise and fall of cardiac biomarkers in the presence of ischemic symptoms that were not related to the index procedure during follow-up. Target lesion revascularization (TLR) was defined as any repeated PCI or bypass grafting surgery to treat a luminal re-narrowing in-stent or within a 5 mm border adjacent to the stent. Target vessel revascularization (TVR) was repeat revascularization of the target vessel by PCI or bypass graft surgery. Stent thrombosis was adjudicated according to Academic Research Consortium definitions as definite, probable, or possible (11). Bleeding events were evaluated during hospitalization and characterized according to TIMI study criteria (TIMI major, TIMI minor bleeding), in addition to any reported bleeding (12). Additionally, the amount of bleeding was measured by changes in hemoglobin or hematocrit pre- and post-intervention.

The primary objective of our study was to compare long-term major adverse cardiac events (MACE) between patients treated by TR and TF approaches. MACE was defined as the occurrence of cardiac death, MI, and TLR in hospital and during follow-up.

Patient demographics, clinical, angiographic, and procedural characteristics were described in the TR and TF groups using means for continuous variables and frequencies for categorical variables. Proportions were compared between groups using chi-square test and Fisher's exact test as appropriate. Group means were compared using Student's t test. Cumulative event rates were estimated by the Kaplan-Meier method and assessed by the log-rank statistic. Differences in the incidence rates of individual clinical outcomes and composite outcomes between TR and TF groups were investigated by multivariable Cox proportional hazards models after confirming that the proportional hazards assumption was met. Adjusted hazard ratios and their 95% confidence intervals were estimated with potential confounders entered as covariates (age, sex, diabetes, chronic renal failure, clinical presentation of acute coronary syndrome, FKB, periprocedural complications, and TR approach). To reduce selection bias for approach type and potential confounding, a 1:1 matched propensity score analysis was performed. Propensity scores were calculated using all variables listed in Table 1, and estimated using multiple logistic regression analysis. The TR patients were ordered and sequentially matched to the nearest unmatched TF patients according to the estimated propensity score. We assessed the balance in baseline covariates between the 2 groups in a propensity score-matched population. The discrimination and calibration abilities of the propensity score model were assessed using the c-statistic and the Hosmer-Lemeshow statistic. Additionally, to determine whether similar outcomes between the 2 groups were consistent in the various subgroups, we performed subgroup analyses. The ten pre-specified subgroups were age (<75 or ≥75 yr), sex, diabetes mellitus, acute coronary syndrome versus stable angina patients, and left ventricular ejection fraction (<50 or ≥50% by 2D-echocardiography), true bifurcation, guiding catheter (≤6 Fr or >6 Fr), IVUS, 2-stent technique, and FKB. All analyses were conducted using SPSS software, version 17 (SPSS Inc., Chicago, IL, USA), and 2-tailed P values of 0.05 or less were considered statistically significant.

This study was approved by the institutional review board (IRB) of Samsung Medical Center (IRB approval number: 2007-04-042). In addition, the local IRB at each participating hospital approved this study and waived the requirement for informed consent for access to each institutional PCI registry.

Overall Population: Baseline clinical characteristics are represented in Table 1. Significant differences in clinical characteristics were found between patients treated by the TR approach and those treated by the TF approach. Overall, patients in the TR group were less likely to have dyslipidemia, peripheral artery occlusive disease, and chronic renal failure, and were less likely to present with acute coronary syndrome.

Angiographic and procedural characteristics are shown in Table 1. There were also significant differences in angiographic and procedural characteristics between the two groups. Types of bifurcation defined by Medina classification were significantly different between the two groups. The prevalence of true bifurcation and glycoprotein IIb/IIIa inhibitor administration tended to be higher in the TR group but the differences were not statistically significant. In contrast, the rates of IVUS guidance, and use of the 2-stent technique were significantly higher in patients treated with a TF approach. FKB, however, was performed more frequently with TR access. For TR patients receiving 2-stents for PCI (10.9%, 55 patients), bifurcation stent techniques included T-stenting (85.5%), crush (7.2%), culotte (1.8%), and V-stenting (5.5%). Among TF bifurcation cases treated with a 2-stent strategy (20.3%, 236 patients), techniques were T-stenting (39.8%), crush (40.3%), culotte (3.0%), and V-stenting (16.9%). Total stent length in the main vessel was longer in the TF than TR group.

Propensity-Matched Population: After performing propensity score-matching for all patients, a total of 424 matched pairs of patients were created (Table 1). The c-statistic for the propensity score model was 0.869, which indicates good discrimination. There were no significant differences in the baseline clinical, angiographic, and procedural characteristics for the propensity-matched subjects except for prevalence of dyslipidemia, glycoprotein IIb/IIIa inhibitor use, and guiding catheter size.

Overall Population: Despite differences in baseline characteristics, the overall procedural success rates were high and were similar between the two approaches (Table 2). The difference between the two groups in the occurrence of peri-procedural complications and peri-procedural MI were not statistically significant, although there was a tendency for more frequent occurrence of peri-procedural MI in the TR than TF group.

Complete clinical follow-up data were obtained for 97.8% of the overall patients with a median follow-up of 672 days (interquartile range 437-965 days), although the follow-up period was longer for TR patients (median 721 days vs 661 days, P<0.01). There were no significant differences in rates of MACEs (5.2% vs 7.0%, P=0.11, Fig. 1A), or individual endpoints, including cardiac death or MI (1.8% vs 2.2%, P=0.45), and TLR (4.1% vs 5.2%, P=0.22) between the TR and TF groups, respectively. Definite and probable stent thrombosis rates were also similar between the TR and TF groups (0.6% vs 0.5%, P=0.77). However, TVR rates were lower in TR group than in TF group (4.4% vs 6.9%, P=0.04). All hazard ratios and 95% confidence intervals for clinical endpoints were subjected to multivariable adjustment (Table 3). The TR approach was not an independent predictor of MACEs, cardiac death or MI, and TLR during follow-up.

Propensity-Matched Population: Procedural success rates in the main vessel were similar between the two groups, but procedural success rates in the side branch were lower with the TR approach than with the TF approach. Periprocedural complications occurred more frequently in the TF group (1 cardiac death, 3 emergent bypass surgery, and 3 coronary arterial dissection or abrupt closure during PCI) than in the TR group (1 cardiac death) (Table 2). There were also no significant differences in rates of MACEs (4.5% vs 6.8%, P=0.09, Fig. 1B), cardiac death or MI (1.4% vs 2.4%, P=0.32), TLR (3.3% vs 5.4%, P=0.09), or stent thrombosis (0.5% vs 0.5%, P=0.87) between the TR and TF groups. However, TVR rates were lower in TR group than in TF group (3.8% vs 7.1%, P=0.02). Even after multivariable adjustment in the propensity score-matched patients, the TR approach was not an independent factor predicting endpoints during follow-up (Table 3).

Among a total of 1,668 patients, records of in-hospital bleeding events were collected in 753 patients. When bleeding events were characterized according to TIMI study criteria, no differences were observed in the rates of in-hospital bleeding events between the TR group and the TF group, even though the amount of bleeding measured by changes in hemoglobin (-0.63±1.20 g/dL vs -0.93±1.28 g/dL, P=0.01) or hematocrit (-1.97±3.53% vs -3.00±3.88%, P<0.01) was significantly smaller in the TR group than in the TF group (Table 4).

In subgroup analysis, there were no significant interactions between the effects on MACE of the vascular approach types (TR or TF) and the eight pre-specified subgroups of age (<75 or ≥75 yr), sex, diabetes mellitus, acute coronary syndrome versus stable angina patients, and left ventricular ejection fraction (<50 or ≥50% by 2D-echocardiography), guiding catheter (≤6 Fr or >6 Fr), IVUS and 2-stent technique (Fig. 2). However, there was a tendency for significant interactions to exist between the presence or absence of true bifurcation lesions and MACE (P for interaction=0.06). While the MACE rate was not significantly different between the groups in true bifurcations lesions, the MACE rate was substantially lower in the TR group than in the TF group among patients with non-true bifurcation lesions. In addition, there was a significant interaction between FKB and MACE (P for interaction=0.01). MACE rate was significantly higher in the FKB group than in non-FKB group and the MACE rate was substantially lower in the TR group than in the TF group among patients without FKB.