Thirty-one RCTs (n=4,410) were included.
When all the studies were combined (31 RCTs; 35 comparisons), the relapse rate was 18% with continuing antidepressant treatment versus 41% with treatment discontinuation. Continuing antidepressant treatment reduced the risk of relapse by 70% (SE=4) (OR 0.30, 95% CI: 0.22, 0.38, P<0.00001) compared with treatment discontinuation and this was statistically significant. There was statistically significant heterogeneity (P=0.02). The risk reduction ranged from 60 to 90% in the six separate analyses based on the length of treatment before randomisation and the duration of continuing treatment (chi-squared test for heterogeneity, P ranged from 0.03 to 0.1).
The risk reduction was similar in the first 12 months after randomisation compared with 12 to 36 months after randomisation: 81% (SE=11) versus 77% (SE=21) reduction in odds) (based on 6 trials; chi-squared test for heterogeneity, P=0.1).
The ORs for the studies grouped by type of antidepressant were not reported, though in the four groups where there was more than one trial pooled, continuing antidepressant treatment had a more favourable effect on relapse than treatment discontinuation. There was statistically significant heterogeneity (P=0.001).
There was a higher rate of trial withdrawal in patients randomised to continuing treatment (18%) compared with placebo (15%); the OR was 1.30 (95% CI: 1.07, 1.59, P=0.009). In one trial there were five suicides in 746 patients on maprotiline compared with one in 374 controls, while in another trial there was one suicide in 185 patients on sertraline compared with none in 110 controls.