Skip to main content

Table 1 Estimation of the overall effectiveness and magnitude of heterogeneity

From: No role for initial severity on the efficacy of antidepressants: results of a multi-meta-analysis

Analyses Effect size (CI/CrI) Heterogeneity (CrI) Role for initial severity Respects randomisation
  RMD SMD RMD SMD RMD SMD  
Frequentist methods        
 Simple random effects meta-analysis 2.71 (1.96, 3.45) 0.33 (0.24, 0.42) 2.50 0.03    Yes
 NMA random effects analysis    2.31 0.03    Yes
 Meta-regression random effects analysis    1.57 0.02 Yes Yes Yes
 Turner et al. (random effects meta-analysis) - 0.31 (0.27, 0.35) - - - - Yes
 Khan et al., ratio of the number of early discontinued patients divided by the total number of patients in each group (chi-square is also calculated) - - - - Yes Yes No
 Kirsch et al. (weight by sample size/SD2) 1.80 0.32 - - - Yes No
 Horder et al. (random effects meta-analysis) 2.70 (1.95, 3.44) - - - Yes - Meta-analysis: yes
Meta-regression: no
 Fountoulakis et al. (sample size weighting) 2.18 0.32 - - - - No
 Fountoulakis et al. (inverse variance weighting) 2.68 0.35 - - - - No
Bayesian methods        
 Simple random effects meta-analysis 2.61 (1.94, 3.30) 0.32 (0.25, 0.40) 1.61 (0.53, 3.53) 0.02 (0.00, 0.05)    Yes
 NMA random effects analysis    0.71 (0.24, 0.98) 0.01 (0.00, 0.05)    Yes
 Simple random effects meta-regression analysis 2.77 (2.18, 3.36) 0.34 (0.27, 0.42) 0.58 (0.00, 2.13) 0.01 (0.00, 0.05) Yes No Yes
 Simple random effects meta-regression analysis using 12 different priors      Yes No Yes
 NMA random effects meta-regression analysis    0.59 (0.01, 2.19) 0.01 (0.00, 0.05) Yes No Yes
 NMA random effects meta-regression using 12 different priors      Yes No Yes
  1. Using the raw mean difference (RMD) and standardised mean difference (SMD) under both frequentist and Bayesian methods. In the frequentist approach, each variable is provided with its confidence interval (CI) and in the Bayesian methods with its credible interval (CrI). Each method is categorised according to its association with the role for initial severity and its respect to randomisation.