2.4 Statistical analysis
We conducted the review using the Stata 12.0 and RevMan 5.3 software.
Comparisons of the dichotomous variables were performed using odds ratio
(OR) with 95% confidence interval (CI). For continuous variables, the
standard mean differences (Std. MD) with 95% CI were used. Z-test was
performed to determine the statistical significance of results with
statistical significance defined as P<0.05 .
Heterogeneity between articles was assessed by Cochrane’s Q-test. Data
with low heterogeneity (P≥0.10 andI2<50% ) were analyzed by a
fixed-effects model while a random-effect model was used for data with
high heterogeneity (P<0.10 andI2>50% ). To explore sources of
heterogeneity the leave-one-out sensitivity analysis was also conducted.
Furthermore, a funnel plot was used to evaluate the publication bias
when applicable.