Measure | Bayesian meta-analysis | Frequentist meta-analysis | |
---|---|---|---|
DerSimonian-Laird approach | Sidik-Jonkman approach | ||
Overall measure of treatment effect: pooled risk ratio (95% interval)* | 0.25 (0.08-0.47) | 0.42 (0.29-0.59) | 0.37 (0.21-0.65) |
Measures of between-study heterogeneity | |||
Between-study variance (log risk ratio scale) (95% interval)* | 0.64 (0.06-1.75) | 0.02 (SE 0.19) | 0.27 (SE 0.26) |
Q-test p value | - | 0.4029 | 0.4029 |
95% prediction interval for risk ratio | 0.02-1.34 | 0.24-0.69† | 0.09-1.45† |
P (predicted risk ratio < 1) | 0.96 | - | - |
I2 statistic (95% interval), %* | 67.4 (1.6-96.7) | 4 (0-64) | 41.26 (0-77) |
Note: SE = standard error.
*95% credible interval for the Bayesian meta-analysis, 95% confidence interval for the frequentist meta-analyses.
†Calculated by means of the methods described by Higgins and colleagues.45