A systematic search found 108 meta-analyses of the use of antidepressants in depressive disorders. Defining newer antidepressants as those introduced since the early 1980s, 18 meta-analyses were selected as being informative about their relative efficacy and tolerability in comparative randomised controlled studies (RCTs). Findings with higher confidence include: little difference in efficacy between most new and old antidepressants; superior efficacy of serotonin and noradrenaline re-uptake inhibitors (SNRIs) over selective serotonin re-uptake inhibitors (SSRIs); a slower onset of therapeutic action of fluoxetine over other SSRIs; a different side effect profile of SSRIs to TCAs with superior general tolerability of SSRIs over TCAs; poorer tolerability of fluvoxamine than other SSRIs in a within group comparison; no increased the risk of suicidal acts or ideation in fluoxetine compared with TCAs (or placebo) in low-risk patients. Findings with a lower level of confidence include: greater efficacy of TCAs than SSRIs in in-patients; greater efficacy of amitriptyline than SSRIs; better tolerability of moclobemide than TCAs; no demonstrable difference in tolerability between SSRIs and TCAs in the elderly; no better tolerability of fluvoxamine than TCAs; better tolerability of dothiepin (dosulepin) than SSRIs; better tolerability of sertraline and greater frequency of agitation on fluoxetine than other SSRIs in a within group comparison. In general, the meta-analyses were of uneven quality, as were the studies included, which limits the confidence in many of the results. Generalising from mostly short-term randomised controlled studies to clinical practice requires caution.