Nineteen studies were included (N = 14002 subjects).
Individual counselling: 3 studies including 2 RCTs (390 subjects) and one pre- and post-test (402 subjects).
Group interventions: 9 studies including 8 RCTs (1581 subjects) and 1 non-randomised (289 subjects).
HIV testing and counselling: 2 RCTs (422 subjects).
Street outreach: 2 RCTs (9251 subjects), 1 pre and post test (276 subjects), and 1 non-equivalent control (461 subjects).
Social intervention: 1 non-randomised study (975 subjects).
Individual counselling: one RCT reported less injection risk behaviour in intervention group but no differences in sexual practices. The other RCT reported sizeable injection risk behaviour and some sexual risk behaviour in both groups.
Group interventions: 2 of 4 RCTs studying behaviour reported less risk taking behaviour in the experimental groups, with the other 2 RCTs reporting reductions in both groups. 3 of 4 RCTs assessing knowledge reported increased knowledge in both groups. Studies had the following limitations: short period of follow-up; potential selection bias; and low baseline rates of drug use and sex.
HIV testing and counselling: Both RCTs reported decreased risk taking behaviour (injection and sexual) in all groups.
Street outreach: Both RCTs reported reductions in risk behaviour (injection and sexual). Limitations included poor follow-up rates.
Social intervention: non-randomised study with low follow-up rate.
The authors consider that the 4 studies without design limitations that reported the intervention to be more than the comparison treatment evaluated more intensive and sustained interventions and were conducted with stable and more motivated subjects. The small number of studies involved precluded investigation of this hypothesis.