Statistical methods

Continuous variables were expressed as means with standard deviations, whereas categorical variables were presented as numbers with percentages. Continuous variables were compared using the Student t test. Categorical variables were compared using the chi-square test. Survival times were censored at the date of the end point or last follow-up. The Kaplan-Meier product-limit estimator and the log-rank test were used to analyze continuous survival times, and the mixed-effects Cox proportional hazards regression modeling was used to test the interaction of treatment assignment and subgroup factors, as well as multivariate modeling of risk factors. Covariates for the adjustment were selected using the stepwise Akaike information criterion (AIC) method. We included as covariates all the variables found to be statistically significant (p < 0.05) in the univariate analysis of the multivariable analysis or variables known to be clinically important, excluding those with multicollinearity with others. The proportional hazard assumption of each variable was tested on the basis of Schoenfeld residuals. The model included the available patient characteristics, clinical risk factors, lesion characteristics, and procedural data. To minimize bias by indication and missing values, an inverse-probability treatment-weighted (IPTW) cohort was created using the “twang package” after multiple imputation of missing values by the “MICE package” in the R program (The R Foundation for Statistical Computing, Vienna, Austria; http://www.R-project.org). We provided a pooled result from 5 different datasets by multiple imputation. A two-sided p value of < 0.05 was considered indicative of a statistically significant difference. Statistical analyses were performed using R programming version 3.4.2 (The R Foundation for Statistical Computing). All the statistical analyses were performed by a professional statistician (S.H.K.).