Fifteen RCTs (2,373 participants, range 15 to 833) were included. Methods of randomisation and allocation concealment were adequate in four studies, as was blinding in 10 studies and reporting of withdrawals in 12 studies. In the other studies these factors were unclear or inadequate.
There was no significant difference between antidepressants and placebo in rates of clinical response (five RCTs, Ι²=69%), remission (four RCTs, Ι²=51%) and affective switch (six RCTs, Ι²=0%). Similarly, there was no significant difference between antidepressants and other medications in rates of clinical response (four RCTs, Ι²=19%), remission (two RCTs, Ι²=0%), affective switch (three RCTs, Ι²=0%) and tolerability (five RCTs).
When antidepressants were compared head-to-head, there was no significant difference in rates of clinical response between bupropion and other antidepressants (venlafaxine, sertraline, desipramine and idazoxan) (three RCTs) and between moclobemide and imipramine (one RCT). One RCT reported higher rates of clinical response with tranylcypromine than with imipramine (75% versus 36%) but affective switch rates were comparable.
One RCT compared clinical remission rates in bupropion (37%), venlafaxine (34%) and sertraline (25%). When bupropion was compared with other antidepressants (sertraline, venlafaxine, desipramine) for affective switch rates, the risk was significantly lower in the bupropion group (RR 0.34, 95% CI 0.13 to 0.88; two RCTs, Ι²=0%). There was no significant difference between bupropion and other antidepressants in tolerability (two RCTs).
Findings of subgroup and sensitivity analyses were reported.