Table 2.
Methods used for meta-analysis (n = 152)
| Characteristic | Description |
|---|---|
| Methods approach | Classic: 151 (99.3%) |
| Bayesian: 1 (0.7%) | |
| Modela | Random-effects: 141 (93.4%) |
| Fixed-effects: 7 (4.6%) | |
| Other: 2 (1.3%) | |
| Not reported: 7 (4.6%) | |
| Variance estimator (for random-effect metanalysis,n = 141) | DerSimonian and Laird: 30 (21.3%) |
| Hartung-Knapp-Sidik-Jonkman: 4 (2.8%) | |
| Restricted maximum-likelihood: 1 (0.7%) | |
| Not reported: 106 (75.2%) | |
| Transformation | Freeman-Tukey double arcsine: 32 (21.1%) |
| Logit: 5 (3.3%) | |
| Log: 4 (2.6%) | |
| Raw: 2 (1.3%) | |
| Arcsine: 1 (0.7%) | |
| Arcsine square roots: 1 (0.7%) | |
| Not reported: 107 (70.4%) | |
| Heterogeneity assessmenta | Subgroup analysis: 89 (58.6%) |
| Meta-regression: 57 (37.5%) | |
| I2: 144 (94.7%) | |
| Galbraith plot: 4 (2.6%) | |
| Other (eg. influence analysis, outliers): 54 (35.5%) | |
| Publication bias | Begg’s test: 26 (17.1%) |
| Egger test: 54 (35.5%) | |
| Funnel plot: 56 (36.8%) | |
| Doi plot: 4 (2.6%) | |
| Trim and fill: 7 (4.6%) | |
| LFK index: 4 (2.6%) | |
| Not reported: 79 (52.0%) | |
| Prediction interval | Yes: 3 (2.0%) |
| Not reported: 149 (98.0%) | |
| Softwarea | STATA: 83 (54.6%) |
| R: 29 (19.1%) | |
| Comprehensive Meta-analysis: 14 (9.2%) | |
| MetaXL: 11 (7.2%) | |
| MedCalc: 5 (3.3%) | |
| Review Manager: 3 (2.0%) | |
| Open Metanalyst: 3 (2.0%) | |
| StatsDirect: 3 (2.0%) | |
| MedScale: 1 (0.7%) | |
| Not reported: 5 (3.3%) |
aAdds to more than 100% because some reviews were counted in more than one option