One hundred and twenty-five RCTs were included in the review. Data were missing on drop-out rates in 5 studies, and on efficacy in 33 studies.
Ninety-eight studies were included in the analysis of efficacy (SSRIs and related drugs versus others; n=5,044 for treatment with an SSRI or related drug and n=4,510 for treatment with an alternative antidepressant).
One hundred and twenty-three studies were included in the analysis of drop-out rate (n=7,032 for treatment with an SSRI or related drug and n=6,334 for treatment with an alternative antidepressant).
Tricyclic antidepressants appear slightly more efficacious than SSRIs or related drugs, although this effect is of uncertain practical importance. SES for SSRIs and related drugs versus alternative antidepressants was 0.034 (95% CI: -0.007, 0.075; Q=149.34, d.f.=98, p<0.001), which was not statistically significant.
The result was fairly robust to assumptions on inclusion: SES for SSRIs alone, compared with tricyclics, was 0.030 (95% CI: -0.018, 0.092; Q=88.64, d.f.=66, p=0.03), which was not statistically significant.
Results were also robust to the type of analysis used: the SES for SSRIs and related antidepressants, using a random-effects model, was 0.045 (95% CI: -0.010, 0.101) whilst the SES for SSRIs alone versus tri-cyclic antidepressants was 0.042 (95% CI: -0.020, 0.104).
In the group of studies comparing SSRIs with tricyclic antidepressants in in-patients, the SES using a random-effects model was 0.10 (95% CI: -0.072, 0.272; Q=49.1, d.f.=22, p=0.008).
SSRIs and related drugs are slightly better tolerated than tricyclic antidepressants, reducing the risk of drop-out by about 4% during 6 weeks of treatment in double-blind randomised trials.
The drop-out rate for participants treated with an SSRI or related drug, compared to that of participants treated with an alternative antidepressant, was 27.7% (1,948 out of 7,032 participants) versus 32.7% (2,072 out of 6,334 participants); RR was 0.87 (95% CI: 0.80, 0.95).
These results were robust to assumptions on inclusion (RR 0.88, 95% CI: 0.83, 0.95; Q=104.8, d.f.=86, p=0.08). The results for both analyses were also robust to the type of model used: similar findings were obtained for both analyses when a fixed-effect model was used.
Lofepramine appears similar in efficacy and tolerability to alternative antidepressants.
Monoamine oxidase inhibitors other than moclobemide, have considerable side-effects and are unlikely to be appropriate for initiation in general practice. Moclobemide, however, is less prone to dangerous side-effects and appears to have similar tolerability and efficacy to other antidepressants.
There is a substantial range of toxicity associated with different antidepressants as currently used in primary care. The SSRIs and lofepramine are associated with the smallest risk of fatal poisoning.