Fifty five studies (70 comparisons) were included, of which 93 per cent were RCTs. Mean sample size was 623 (range 29 to 4,249). RCT quality was generally low: 59 per cent clearly reported allocation concealment; 52 per cent conducted blinded outcome assessment; 64 per cent handled withdrawals appropriately; and only five per cent blinded providers, recipients and assessors.
Components of care: Studies that reviewed professional roles were significantly more likely to be associated with a positive outcome than other studies (OR 3.8. 95% CI: 1.08, 14.51). No other statistically significant findings were reported.
Treatment type: No statistically significant difference in the likelihood of a positive outcome was found between comparisons using the following categories of intervention: antidepressant medication (AD); cognitive behavioural therapy (CBT); AD combined with CBT; and problem solving/other psycho-education.
Study intentions: Studies that provided training and feedback (n=14) were significantly less likely to be associated with a positive outcome than other studies (OR 8.00, 95% CI: 1.64, 39.14, p=0.006). No other statistically significant findings were reported.
Detailed features of delivery: Studies that took patient preferences into account (n=5) were significantly more likely to be associated with a positive outcome than other studies (p=0.01). Studies where regular scheduled feedback was provided to general practitioners from a person monitoring care (n=11) were also significantly more likely to be associated with a positive outcome than other studies (p=0.02), as were studies (n=7) where the care monitor provided CBT or problem-solving training (p=0.006).
Discipline of the person delivering care: There was a non-significant trend for provider type to affect outcomes. A series of additional analyses concerning provider type were reported in the review.
No systematic association was detected between intervention intensity or sample size and review findings. Other analyses were also reported in the review.