Twenty RCTs (22 comparisons, 2,772 participants) reported in 51 publications were included in the review. Sample sizes ranged from 15 to 635. Risk of bias was evaluated as low for four studies, unclear for 10 studies and high for six studies.
SSRI versus Placebo (nine RCTs): SSRI showed significantly better response (BR 1.49, 95% CI 1.29 to 1.71; Ι²=0%) and remission (BR 1.53, 95% CI 1.29 to 1.80; Ι²=0%) outcomes compared to placebo. There were no significant differences for drop-outs and adverse events. Statistically significant heterogeneity was identified for drop-outs (Ι²=52%)
TCA versus Placebo (seven RCTs): TCA showed significantly better response (BR 1.74, 95% CI 1.50 to 2.02; Ι²=0%) and remission (BR 1.77, 95% CI 1.24 to 2.53; Ι²=51%) outcomes compared to placebo. Statistically significant heterogeneity was identified for remission. Adverse events were significantly higher in the TCA group compared to the placebo group (OR 2.87, 95% CI 1.83 to 4.50; Ι²=20%). There were no significant differences between TCA and placebo for drop-outs.
SSRI versus TCA (six RCTs): There was no statistically significant difference between SSRI and TCA for response or remission. SSRI showed significantly lower drop-out (OR 0.41, 95% CI 0.19 to 0.86; Ι²= 71%) and adverse event (OR 0.48, 95% CI 0.32 to 0.70; Ι²= 1%) outcomes compared to TCA. Statistically significant heterogeneity was identified for drop-outs.
Subgroup analysis showed no significant differences for any outcomes between types of diagnosis (TCA studies only). Only one comparison showed a statistically significant difference between studies with an unclear or high risk of bias compared with a low risk of bias (this included only one study judged to have low risk of bias).
No indication for publication bias was identified using the Egger's test and visual examination of funnel plots.