Fifteen RCTs (n=2,458) were included. These studies provided 30 treatment arms. Ten studies (n=836) evaluated SNRIs, 11 studies (n=916) evaluated SSRIs and 9 studies (n=706) evaluated TCAs.
Significant statistical heterogeneity was found for all three drug classes for remission and drop-out rates. The authors could find no apparent explanations for this heterogeneity.
Patients in the TCA and SNRI treatment arms were significantly older than patients in the SSRI treatment arms (53.7 and 52.5 years, respectively, versus 40.9 years, p<0.05). Patients in the SSRI treatment arms were significantly heavier than patients in the TCA treatment arms (75.4 kg versus 62.6 kg, p<0.05).
Remission rates: ITT analysis.
Patients in the SNRI and TCA treatment arms had significantly higher rates of remission than patients in the SSRI treatment arms (49% and 44.1%, respectively, versus 37.7%, p<0.001 for both comparisons). There was no significant difference between the SNRI and TCA treatment arms. Remission rates varied within each drug class: from 31 to 70% within the SNRI treatment arms, from 19 to 70% within SSRIs, and from 23 to 54% within TCAs.
Remission rates: per protocol analysis.
TCAs were associated with significantly higher rates of remission than SNRIs and SSRIs (69% versus 64% and 54%, respectively, p<0.001 for both comparisons). There was no significant difference between the SNRI and SSRI treatment arms.
Drop-out rates.
ITT analysis: patients in the TCA treatment arms had significantly higher drop-out rates than patients in the SSRI and SNRI treatment arms (35.7% versus 28.4% and 26.1%, respectively, p<0.05 for both comparisons). There was no significant difference between the SNRI and SSRI treatment arms.
Drop-outs due to ADRs were significantly higher in the TCA treatment arms than in the SNRI and SSRI treatment arms (19.8% versus 10.3% and 8.3%, respectively, p<0.05).
There was no significant difference between the three drug classes for drop-outs due to loss of efficacy.
ADRs.
Patients in the TCA treatment arms had the highest rates of occurrence for most selected ADRs, including dry mouth (58%), sweating (28%) and constipation (26%).
The authors stated that the funnel plot suggested the presence of publication bias but the Begg-Mazumdar test found no strong evidence for it.