Statistical analysis
Continuous variables are presented using mean and standard deviation or median and interquartile range. Categorical data are presented as numbers and percentages. Random-effect meta-analysis was performed because of the observational design of the included studies. HRs and 95%CI adjusted by Cox multivariate analysis was considered the effect size, which evaluate the risk of all-cause mortality between CRT-D and CRT-P. Heterogeneity among studies was assessed by the Cochran’s Q test and was considered significant when P < 0.10. The I2 statistic was used to evaluate the proportion of variance caused by heterogeneity in observed variance. Subgroup analyses were performed in accordance to the differences of follow-up duration, indication of CRT-D, region of research center and year of publication between studies. Sensitivity analysis was performed by removing one study at a time. Publication bias was estimated using funnel plots and the regression test of Egger or Begg. Statistical analyses were all performed with STATA 14 (Stata Corp LP, College Station, Texas).