This study was derived from an observational cohort consisting of 8943 patients with 3VD. These patients were prospectively and consecutively enrolled at Fu Wai Hospital from April 2004 to February 2011 (Beijing, China). The definition of 3VD was angiographic narrowing of ≥ 50% in the left circumflex, left anterior descending, and right coronary arteries, with or without involvement of the left main artery. In this study, the inclusion criteria were as follows: (1) patients who were ≥ 75 years of age; (2) patients who were willing to be followed up; and (3) patients with complete survival data. The final analysis included 711 patients with 3VD who were eligible. Because we obtained data of this post hoc analysis from a subpopulation of the 3VD cohort, the sample size was determined by the number of patients who met the inclusion criteria instead of being calculated in advance. The choice of therapeutic strategy was based on contemporary clinical guidelines, the judgement of cardiologist teams, and the patient's personal preference [11, 12]. Before the PCI procedure, patients were treated with aspirin and clopidogrel (300 mg or 600 mg, loading dose), and they received standard dual antiplatelet therapy for no less than 1 year after the procedure. The choice of equipment, drugs, and techniques during PCI was at the discretion of the operators. For patients who underwent CABG, standard bypass techniques were used with preferably grafting of the left internal mammary artery to the left anterior descending artery. Experienced surgeons performed the procedure either using on-pump or off-pump surgical techniques on the basis of their individual preference. The ethics committee of Fu Wai Hospital approved this study and it adhered to the Declaration of Helsinki. Informed consent was provided by all participants.

We obtained information on the medical history and in-hospital data through the electronic record system of our hospital. Patients were followed up by certificated clinical research coordinators through telephone, follow-up letter, or outpatient visit. The last follow-up was finished on March 2016. All-cause death was the primary endpoint. Major adverse cardiac and cerebrovascular events (MACCE, which consisted of unplanned revascularization, stroke, myocardial infarction and all-cause death) and cardiac death were the secondary endpoints. Unless unequivocal non-cardiac causes could be established, all deaths were considered cardiac deaths.

Data are shown as means with standard deviations to describe continuous variables of baseline characteristics. Percentages and frequencies are shown for categorical variables. Differences in categorical variables of baseline characteristics were compared with the Pearson chi-square test and Fisher's exact test. Differences in continuous variables of baseline characteristics were compared with ANOVA and the Kruskal–Wallis test. A survival curve was shown by using the Kaplan–Meier method and compared by using the log-rank test. Variables that were considered clinically relevant or showed a significant univariate relationship with outcomes were incorporated into multivariate Cox analysis. Variables for adjustment included clinical presentation, age, sex, body mass index, diabetes, hypertension, hyperlipidemia, chronic kidney disease, peripheral artery disease, previous stroke, SYNTAX score, previous myocardial infarction, smoker, left ventricular ejection fraction, and left main disease. Interaction was tested by using the Cox regression model to assess the effects of therapeutic strategies in subgroups. Statistical significance was defined as a two-sided α = 0.05. The statistical analyses mentioned above were performed by using IBM® SPSS® v25.0.0.0 software (IBM Inc., Armonk, NY, USA). Moreover, we performed competing risk regression analysis with sub-distribution hazard models for cardiac death and other cardiovascular endpoints, considering competing risks for non-cardiac death and all-cause death, respectively. Inverse probability of treatment weighting (IPTW) regression analysis based on propensity score was also performed in order to better balance the discrepancies among groups. Standardized mean differences were compared to evaluate the degree of baseline variable balance. A standardized mean difference less than 0.1 was regarded as a high degree of balance. The covariates included in propensity score model were as follows: age, sex, BMI, diabetes, hypertension, hyperlipidemia, previous MI, previous stroke, chronic kidney disease, peripheral artery disease, smoker, clinical presentation, left main disease, LVEF and SYNTAX score. R software (version 4.0.2) was used to perform competing risk and IPTW regression analysis.