Data analysis A data extraction form was developed to collate information from each RCT identified through the above process. Multiple outcome measures derived from the same sample were synthesized so that each study contributed only one overall effect size to the analysis. The effect size metric used in this metaanalysis was the standardized mean difference (SMD, or Cohen's d). Before computing the average effect size for each study outcome, a positive or negative sign was assigned to each individual effect size. If means and standard deviations (SD) or Cohen's d were not reported, then odds ratios (OR) were used to convert to Cohen's d. The weighted average effect size and corresponding 95% confidence intervals (95% CI) were then calculated for each study outcome. If the interval did not contain the value d = 0, the null hypothesis that no relationship existed in the population was rejected and the d index was considered significant. A random effects model was then applied. Cohen [ 31] provides guidelines for interpreting d effect sizes (i.e. SMD). He recommends assigning qualitative descriptors as follows: 0.20-0.49 as small, 0.50-0.79 as medium and 0.80 and above as large. Heterogeneity among primary outcome studies and moderator variables was assessed using the I2 statistic [ 32] . This metaanalysis targets specifically work on illicit drugs. For this reason, when studies involved multiple substances, we analysed only the outcome for illicit drugs. The overall effect