A meta-analysis was carried out of 42 published randomized controlled studies comparing the selective serotonin reuptake inhibitors (SSRIs) with the tricyclic antidepressants (TCAs) that measured discontinuation rates for side effects and lack of efficacy by treatment group in order to compare the discontinuation rates for side effects and lack of efficacy. These discontinuation rates were pooled to produce the main outcome measure. Seven studies were placebo controlled and the discontinuation rates in these studies were also pooled in a separate analysis. Significantly fewer patients receiving SSRIs discontinued treatment because of side effects (14.9%) compared with those receiving TCAs (19%) (p < 0.01). There was also a significant difference in discontinuation rates due to side effects in the placebo- and TCA-controlled studies analysed separately, SSRIs (19%) compared with TCAs (27%) (p < 0.01). In both analyses a similar proportion of patients discontinued for lack of efficacy on SSRIs and TCAs. There is a significant and clinically important advantage for the SSRIs compared with the TCAs in the acceptability of treatment measured by the number of discontinuations due to side effects reported in published studies. The risk-benefit calculation favours the SSRIs since there were similar levels of efficacy but more discontinuations with the TCAs. The selection of an antidepressant for first-line treatment requires critical evaluation of the full risk-benefit equation.