Seventy studies were included in the review, of which 46 RCTs (n=11,553) assessed efficacy or effectiveness. Twenty-two of these (n=5,333) compared SSRIs with other SSRIs and 24 (n=6,220) compared an SSRI with another second-generation antidepressant. Twenty-four observational or placebo-controlled studies were used in the analysis of adverse events (number of participants not given).
SSRI compared with another SSRI.
Twenty-one studies were considered to be of a fair quality and one study was rated as good. Three studies were classed as effectiveness trials. The majority of the efficacy trials and all effectiveness trials reported no statistically significant differences on any outcome measure. Two meta-analyses were conducted. Six studies were pooled (774 patients) to compare fluoxetine with paroxetine, demonstrating no statistically significant differences between the drugs (relative benefit 1.09, 95% CI: 0.97, 1.21). No major publication bias was found in the Begg adjusted rank order correlation test (Kendall tau 0.3; P=0.47) and the Egger regression approach (intercept 2.107, 95% CI: -0.237, 4.450). Five studies were pooled (1,190 patients) to compare fluoxetine with sertraline, demonstrating a modest statistically significant treatment benefit for sertraline (relative benefit 1.10, 95% CI: 1.01, 1.20). Again, no publication bias was observed in the Begg adjusted rank order correlation test (Kendall tau 0.2; P=0.82) or the Egger regression approach (intercept 0.799, 95% CI: -6.069, 7.668).
In the 9 studies that measured health-related quality of life, good to fair evidence suggested that SSRIs did not differ substantially in terms of improvements on this outcome. Sleep quality was reported to be better in fluvoxamine-treated patients in the one trial that compared this drug with fluoxetine.
The majority of trials found no difference in terms of speed of the patient's response to SSRIs.
SSRIs compared with other second-generation antidepressants.
All 24 trials were classified as efficacy studies. Twenty-one trials were recorded as being of a fair quality and 3 were rated good. The majority of the trials demonstrated similar efficacy between the drugs. There was some evidence in favour of a higher response rate for venlafaxine-treated patients; however, few individual trials reached statistical significance. One meta-analysis was conducted. Six studies were pooled (1,340 patients) to compare venlafaxine with fluoxetine, demonstrating a modest statistically significant treatment benefit for venlafaxine (relative benefit 1.12, 95% CI: 1.02, 1.23) on the HAM-D scale. There was no observed publication bias in the Begg adjusted rank order correlation test (Kendall tau 0.067; P>0.99) or Egger regression approach (intercept 0.141 (95% CI: -4.158, 4.440).
In the 5 trials that assessed quality of life or health-related functional capacity, no significant differences in overall quality of life were reported. Mirtazapine-treated patients experienced better sleep quality in one further trial that compared this drug with sertraline.
No statistically significant differences in response rates were noted in the 20 trials that measured this outcome.
The overall incidence of adverse events was similar amongst the drugs examined. However, there were noticeable differences in the nature of these adverse events. The authors advised cautious interpretation of these results given the differences in the assessment and reporting techniques.