2.5 Statistical analysis
We performed a meta-analysis of the included studies using a
random-effects model. Studies were excluded if they did not include an
outcome in each intervention group, did not have enough information
required for continuous data comparison, or shared the same patient
population. Incidence rate ratios (IRR) of events were calculated by
dividing the number of incident cases of events by the total number of
person-years follow-up. IRR were pooled after applying the
Freeman-Turkey double arcsine transformation to stabilize the variances
(9). We pooled the point estimates of IRR, odds ratio, and incidence
rate from each study using the generic inverse-variance method of Der
Simonian and Laird (10). If no event was observed, pooled IRR were
estimated using Poisson regression with random intervention effects
(11). The heterogeneity of effect size estimates across these studies
was quantified using the I2 statistic. The
I2 statistic ranges in value from 0 to 100%
(I2< 25%, low heterogeneity;
I2= 25%–50%, moderate heterogeneity; and
I2> 50%, substantial
heterogeneity). A sensitivity analysis was performed to assess the
influence of the individual studies on the overall results by omitting
one study at a time. Publication bias was assessed using a funnel plot
and the Egger’s regression test (12). (p< 0.05 was
considered significant). All data analyses were performed using the
STATA SE version 14.2.