Seventeen RCTs (645 patients) were included.
Quality of studies.
None of the RCTs met all of the Consolidated Standards of Reporting Trials (CONSORT) criteria for reporting results from RCTs. There were several methodological problems: the method of randomisation, or how the randomisation was concealed, was not always described; descriptions of power calculations were lacking; and the studies predominantly had a small sample size (only 2 RCTs included more than 25 patients).
Antidepressant versus placebo (4 RCTs).
The 4 RCTs were of four different drugs. Two RCTs found a significant benefit; one small RCT (47 women) found that oestrogen (5 to 25 mg/day) increased the recovery rates after 12 weeks (44 versus 0%; ARD 44%, 95% CI: 25, 63); the other small RCT (20 patients) found that viqualine (50 mg/day) increased the recovery rates after 4 weeks (50 versus 0%; ARD 50%, 95% CI: 19, 81).
Comparison of two active treatments (4 RCTs).
One RCT (123 patients) found no significant difference between venlafaxine (37.5 mg to 300 mg/day) and paroxetine (20 to 40 mg/day) in the recovery rates at 4 weeks; the recovery rates were 44 and 29%, respectively (ARD 15%, 95% CI: -2, +32). Most of the patients had been on selective serotonin re-uptake inhibitors. Other studies were not considered to be relevant to current practice. The recovery rate on placebo (3 RCTs, 38 patients) was 0%.
Antidepressant plus augmenter versus antidepressant plus placebo (8 RCTs).
Antidepressant plus lithium (300 to 1,200 mg/day) increased the recovery rates at 2 weeks in comparison with antidepressant alone; the ARD (2 RCTs, 50 patients) was 25% (95% CI: 2, 49). No significant heterogeneity was found (P=0.44).
A meta-analysis found no significant difference between antidepressant plus pindolol (7.5 mg/day) and antidepressant alone in the recovery rates at 10 days to 4 weeks; the ARD (2 RCTs, 96 patients) was 8% (95% CI: -6, 21). It was reported that the chi-squared test for heterogeneity showed little evidence of heterogeneity (P=0.07). One additional RCT of pindolol reported no recoveries and could not, therefore, be included in the meta-analysis.
The recovery rate on placebo (8 RCTs, 107 patients) was 14.4% (95% CI: 7.9%, 23.4%).
Augmentation without a placebo (3 RCTs).
These RCTs did not have a placebo arm and were not commented on in the text of the review.
No RCTs of psychotherapy were identified.