Twenty-five RCTs were included in the review (1,197 participants; range 12 to 120); 21 were included in the meta-analysis. Five RCTs reported adequate sequence generation, eight had sufficient blinding, 13 had adequate completeness of outcome data. Study duration ranged from four to 52 weeks.
Antidepressant use was associated with significantly better response than placebo at four to five weeks (OR 1.93, 95% CI 1.15 to 3.42; five RCTs), six to eight weeks (OR 2.25, 95% CI 1.38 to 3.67; 12 RCTs) and nine to 18 weeks (OR 2.71, 95% CI 1.50 to 4.91; seven RCTs). Some heterogeneity was indicated at six to eight weeks (Ι²=40%) but not at the other time points. Sensitivity analyses showed similar results, except at four to five weeks when excluding trials at high risk of bias and when including trials that used a narrow definition of depression; in these cases no significant treatment effect was observed.
Drop-outs were significantly greater with antidepressants compared with placebo at nine to 18 weeks (OR 2.09, 95% CI 1.02 to 4.31; six RCTs) but were similar at the other time points.
All four studies that reported quality of life outcomes indicated greater improvements with antidepressant than placebo. One of three studies reported significantly greater improvement with antidepressant than placebo.
Dry mouth, nausea, dizziness, sexual dysfunction, hypotension and headache were more likely to be reported with antidepressant than placebo.
The funnel plot was asymmetrical which indicated the possibility of publication bias. Subgroup analyses were reported: tricyclic antidepressants showed a significant effect at all time points whereas selective serotonin reuptake inhibitors only showed a significant effect at nine to 18 weeks.