This is a retrospective multicenter registry of patients with coronary bifurcation lesions who underwent percutaneous coronary intervention with drug-eluting stent. A total of 2,897 patients from 18 major coronary intervention centers between January 2003 and December 2009 were enrolled. The inclusion criteria were as follows: 1) coronary bifurcation lesions treated with drug-eluting stent only; and 2) MV diameter of 2.5 mm and SB diameter of 2.3 mm confirmed by core laboratory quantitative coronary analysis. The decisions regarding whether or not to perform POT were made by the respective operators. A bifurcation lesion with a large SB will have a large discrepancy between the proximal and distal MV according to Murray's law.7) This type of lesion is the optimal target of POT. In order to assess the effects of POT on procedural and clinical outcomes, we additionally selected bifurcation lesions with an SB reference diameter ≥2.5 mm on core-lab quantitative coronary angiography analysis. A total of 1,191 patients were ultimately included in this study. The local Institutional Review Board at each hospital approved this study, and the need for informed consent was waived.

Decisions regarding the treatment strategy, stenting technique, selection of drug-eluting stent type, and use of glycoprotein IIb/IIIa receptor inhibitors or intravascular ultrasound were all made at the operator's discretion. The POT was performed in one of two ways, as follows: 1) immediately after MV stenting with a nominal pressure, the stent balloon was pulled back to post-dilate the proximal MV just proximal to the carina with a high pressure targeting 0.25–0.5 mm overexpansion above the nominal balloon diameter; 2) after MV stenting, the MV stent was post-dilated immediately proximal to the carina using a short, non-compliant balloon sized for the proximal MV reference diameter.

Data were collected using a Web-based reporting system. Additional information was obtained from the medical records or by telephone contact, if necessary. Coronary angiograms were reviewed and analyzed quantitatively by an independent core laboratory. Angiographic and procedural characteristics of all cine-angiograms were reviewed and analyzed at the angiographic core laboratory using an automated edge-detection system (Centricity CA1000; GE; Waukesha, WI, USA) and standard definitions.8) Bifurcations were classified according to the Medina classification, in which the proximal MV, distal MV, and SB components of the bifurcation are each assigned a score of 1 or 0 depending on the presence or absence of >50% stenosis, respectively.9) Target bifurcation was defined as lesions including MV (<5 mm proximal to SB take off and <5 mm distal to SB take off) and SB ostium (<5 mm distal to take off). Quantitative coronary angiographic analysis was independently rechecked off-line by two experienced analysts based on previous study.10)

Deaths were considered to be of cardiac origin unless there was documentation of otherwise. Myocardial infarction was defined as the presence of electrocardiography findings indicative of ischemia that were unrelated to the index procedure, as well as chest discomfort associated with elevated creatinine kinase MB fraction or troponin-T/troponin I. The primary outcome was major adverse cardiac events (MACEs), which were defined as a composite of cardiac death, myocardial infarction, or target lesion revascularization (TLR) during follow-up. The secondary outcomes included individual components of the primary endpoint, all-cause death, and stent thrombosis.

Comparisons for continuous variables were made using the t-test or the Wilcoxon rank-sum test when applicable. Results are presented as mean±standard deviation or median with interquartile range. The differences between groups were evaluated using χ² or Fisher's exact test for categorical data. All outcomes of interest for analysis are survival data — all cause death, cardiac death, MI, stent thrombosis, TLR (MV, SB ostium, target bifurcation), and MACE. To estimate hazard rates for all-cause death, we used a multivariable Cox model and considered all covariates that were either statistically suggestive with (p<0.2) on univariate analysis or clinically relevant. To estimate hazard rates for other survival events with competing risks, we used the Fin and Gray's competing risk proportional hazard model; 1) cardiac death (competing risk=non-cardiac death); 2) MI (competing risk=death); 3) TLR and its components (competing risk=death); and 4) stent thrombosis (competing risk=death). We then advocated the propensity score method to balance the treatment groups and compared if confounding effects in our retrospective study might have produced bias in the estimated hazard rates obtained above. Propensity scores were estimated using multiple logistic-regression analysis. We developed a full non-parsimonious model, which includes all variables in Tables 1 and 2, as well as the baseline variables of quantitative coronary angiographic analysis in Table 3. One-to-many (1:N) propensity score matching with non-fixed matching ratio was performed using the nearest neighbor matching method, which is a greedy algorithm among patients with an individual propensity score. The covariate balance achieved by matching was assessed by calculating the absolute standardized mean differences of <10% and variance ratio of 1.0 in all covariates between the POT and non-POT groups. In the propensity score-matched population, continuous variables were compared using a paired t-test or the Wilcoxon signed-rank test, as appropriate. Categorical variables were compared using McNemar's or Bowker's test of symmetry, as appropriate. The reduction in the risk of negative outcome was estimated using a clustered Cox regression model or an extension of a competing risk proportional hazard model for the propensity matched clustered data and compared with the results from the original full data. The cumulative incidence functions of the POT and non-POT groups for MACE were obtained from the competing risk model and compared by a Gray's test. Statistical analyses were performed using SAS 9.2 (SAS Institute Inc., Cary, NC, USA). All tests were two-tailed, and p<0.05 were considered statistically significant.

Among the 1,191 eligible patients, POT was performed in 252 patients (21.2%). Eighty-one of these patients were treated with a short, non-compliant balloon. There was no significant difference in the clinical characteristics between the POT and non-POT groups, except for the proportion of dyslipidemia (Table 1). The angiographic and procedural characteristics are shown in Table 2. Compared with patients in the non-POT group, those in the POT group had a higher incidence of left main bifurcation, non-true bifurcation, SB predilation, use of intravascular ultrasound, one stent technique, and large stent diameter (Table 2). After propensity score matching, there were no significant differences in the clinical, angiographic, and procedural characteristics between the two groups.

After propensity score matching, there were no significant differences in results of quantitative coronary angiographic analysis between the two groups (Table 3). Notably, the manual lymph drainage (MLD) of the proximal MV was similar between the two groups, both in the total population and in the propensity score-matched population. After POT, the MV proximal, middle and distal MLD, SB ostial, and distal MLD were similar between the two groups. These results of quantitative coronary angiographic analysis were maintained after POT in the propensity-matched populations.

The median follow-up duration was 37.3 (interquartile range, 26.0–51.8) months. In the total population, the incidence of MACEs was significantly lower in the POT group than it was in the non-POT group (POT vs. non-POT; 4.8% vs. 11.7%, p=0.006, Table 4, Figure 1A). This result was mostly driven by the lower incidence of TLR (3.6% vs. 9.5%, p=0.008), mostly in MV, in the POT group compared to that in the non-POT group. After 1:N propensity-score matching, the POT was associated with a lower incidence of MACE (3.8% vs. 11.8%, hazard ratio [HR], 0.34; 95% confidence interval [CI], 0.17–0.69; p=0.003, Table 4, Figure 1B) and TLR (3.3% vs. 9.5%, HR, 0.37; 95% CI, 0.17–0.78; p=0.01), mostly in MV.

We calculated the unadjusted HR for MACE and TLR in various complex subgroups in order to determine whether the observed outcomes related to POT (in the overall populations) were consistent. Compared with the non-POT group, the association of POT with better MACE outcomes was consistent across various subgroups. Of note, POT was not associated with a reduction in MACE in patients who were treated with kissing ballooning. In contrast, POT was associated with a reduction in MACE in those treated without kissing ballooning (p for interaction=0.052, Figure 2). Similarly, the association of POT with better TLR outcomes was consistent across various subgroups without a significant interaction. However, there was a significant interaction between the use of POT and TLR according to the presence or absence of final kissing ballooning (p for interaction=0.03, Figure 3).

In patients who did not undergo kissing ballooning, the proximal MV MLD was significantly greater in the POT group compared to that in the non-POT group (3.46±0.56 mm, 3.27±0.61 mm, p=0.001). In contrast, there was no difference between the two groups in patients treated with kissing ballooning (3.59±0.58 mm, 3.61±0.60 mm, p=0.76). POT was associated with a significantly lower risk of MACE and TLR than was the non-POT group (MACE: 2.0% vs. 10.5%, HR, 0.20; 95% CI, 0.06–0.63; p=0.01 and TLR: 0.7% vs. 8.5%, HR, 0.08; 95% CI, 0.01–0.57; p=0.01). This result was consistent after additional propensity-score matching between the POT and non-POT groups (Table 5).