Seventeen studies (4,088 patients) were included in the review comprising nine controlled studies (3,817 patients, sample sizes ranged from 19 to 2,729) and eight cohort studies (271 patients, sample sizes ranged from six to 107). One of the controlled studies was a randomised controlled trial.
Five studies were globally rated using the Quality Assessment Tool as "strong" all of which were controlled trials, a further four controlled studies and two cohort studies were rated as "moderate" and six cohort studies were allocated a "weak global rating. The authors stated that the strongest causes of bias were little reporting of randomisation, small sample sizes as a result of treatment drop-out, little use of intention-to-treat analyses and little reporting of blinding procedures.
Statistically significant benefits were found with integrated treatment programmes for the change of substance use disorder symptoms from baseline to longest follow-up (ES 0.60 95% CI 0.42 to 0.78; Q=104.39; p<0.000916 studies), and post-traumatic stress disorder symptoms (ES 0.88, 95% CI 0.66 to 1.09; Q=117.83; p<0.0001) with statistically significant heterogeneity across the studies for both outcomes.
There were no significant differences found between integrated treatment programmes and control/comparator groups at longest follow-up on substance use disorder symptoms (weighted average ES 0.10, 95% CI -0.01 to 0.21; Q=6.09; p=0.64, nine studies) and post-traumatic stress disorder symptoms (weighted average ES 0.08, 95% CI -0.03 to 0.19; Q=3.58; p<0.82).
Evaluations of the funnel plots found some evidence of publication bias across the outcomes.