Statistical analysis
Traditional pairwise meta-analysis with Review Manager 5.4.1For binary
data, odds ratio (OR) and 95% CI are used as effect size indicators.
OR<1 indicates that one intervention may be worse than another
intervention; OR>1 indicates that one intervention may be
better than another intervention; 95%CI containing 1 means that the
difference is not statistically significant. For continuous data,
standard mean difference (SMD) and 95% CI were used as effect size
indicators. SMD<0 means that one intervention may be worse
than another; SMD>0 means that one intervention may be
better than another; 95%CI containing 0 means that the difference is
not statistically significant. In direct Meta analysis, Q test and I2
index are used to evaluate the heterogeneity of each effect size. If
P>0.1 and I2<50%, it indicates that the results
of each study have good homogeneity, then the fixed effects model is
used. If P≤0.1 and/or I2≥50%, the results of the study are
statistically heterogeneous, and a random effects model is used.
Network meta-analysis with
software
ADDIS 1.16.8
The Node-split model is used to test the consistency in the network
meta-analysis. If there is no statistical difference between the studies
within the subgroup (P>0.05), it indicates that the
heterogeneity of the included studies is small, so the consistency model
is used for analysis; otherwise, an inconsistency model is used for
analysis. A ranking probability table is used to rank the pros and cons
of intervention measures (the value indicates the probability of
intervention measures in the nth position). Regarding the main
indicators of this article, the higher the ranking, the better.