2.3. Statistical analysis
Data for the outcomes of interest are presented in numbers and
percentages and mean values with a 95% confidence interval (CI) were
also considered. The measurement of the weighted prevalence of outcomes
was performed and is presented in forest plots. Moreover, the conversion
of the median for continuous values (e.g., age and length of stay) into the mean
was applied as needed. Furthermore, the I2 statistic
was used to quantify the total variability among eligible studies,
ascribable to heterogeneity, and reported in percentages with a 95% CI.
The I2 covers the spectrum of 0% indicating no
heterogeneity, 25% to 49% as low heterogeneity, 50% to 74% as
moderate heterogeneity, and ≥75% as high heterogeneity, according to
published guidelines [13, 14]. The random-effects model, where the
I2 value lies above 50%, was utilized, and a
fixed-effects model (Mantel – Haenszel method) was used when
significant heterogeneity was absent. Regarding the inconsistency among
some of the results of observational studies including stent type (Thoraflex
Hybrid, E-Vita, Cronus, Frozenix, and mixed/other), location of the
study center (Asia, North America, and Europe), and
elective/non-elective procedure, subgroup analyses were performed.
Publication bias was evaluated by Egger’s test and visualized by the
funnel plots, and additionally, statistical tests for its asymmetry were
employed. The statistical significance was reached by P-values below
0.05. All the statistical analyses were performed using R (version
4.1.2) and STATA (V.16.0) softwares.