Statistical analysis
Measurement data are reported as the mean ± standard deviation. The
analysis used a random effects model (inverse-variance method). For
short term categorical outcomes, risk difference with 95% confidence
interval were used, as many studies have zero events in both sides. For
continuous outcomes, standardized mean difference (SMD) and 95%CI were
used. For late outcomes, incidence rate ratio and 95%CI were estimated
from the total number of events observed within a treatment group out of
the total person-time of follow-up for that treatment
group8. Meta-regression was used to assess the effect
of sample size, age, gender, chronic obstructive pulmonary disease, New
York Heart Association (NYHA), previous stroke/transient ischemic attack
(TIA) and redo surgery on the primary outcomes and secondary outcome
(perioperative mortality).
Hypothesis testing for equivalence was set at the 2-tailed 0.05 level.
Analyses and data modeling were performed with R-project (version 3.3.3
R project for Statistical Computing), following packages were used:
‘metafor’, ‘stats’, and ‘graphics’ for data visualization.
.