In line with a recent Cochrane review,43and in an attempt to address the complexity of clinical
heterogeneity of interventions, subgroup analyses were conducted to compare psychosocial interventions with (1) treatment as usual (TAU); (2) education or information; (3) HIV infection testing and counselling; (4) control interventions of lesser time or intensity with OST; and (5) control interventions of lesser time or intensity without OST. As follow-up duration may affect intervention effectiveness, further subgroup analyses were conducted where possible, comparing length of time, in months, from the end of the intervention to the final follow-up of included trials (i.e.≤3 months, 4–6 months and≥9 months follow-up).
Injecting risk behaviour and sexual risk behaviour (as described inEligibility) were used as outcomes of interest for meta-analysis. The Cochrane handbook (section 16.5) stresses that trials with multiple treatment arms‘that compare more than two intervention groups need to be treated with care’to avoid‘making multiple pair-wise comparisons between all possible pairs of intervention groups’in the meta-analysis.62
Therefore, where trials included in the meta-analysis had more than one intervention group, data from the most relevant psychosocial intervention to address the aims of the systematic review were compared with the control intervention in the meta-analysis. For Boothet al.,68the most relevant intervention condition was
considered TAU plus HIV/HCV infection counselling and education, rather than TAU plus a therapeutic alliance to facilitate treatment entry. For Sterket al.,81the enhanced negotiation intervention was considered
more relevant to the aims of the systematic review than the enhanced motivation intervention. For Schroder et al.,71TAU plus weekly CBT plus CM (CBT
+CM) was considered superior to both weekly CBT plus
DETERMINING THE EVIDENCE BASE
NIHR Journals Library www.journalslibrary.nihr.ac.uk
non-contingent vouchers and CM plus weekly group therapy and, therefore, was selected as the intervention condition for the meta-analysis. Goet al.82conducted a multilevel intervention using a 2 × 2 (four-arm) factorial
design consisting of (1) standard of care (i.e. HIV infection testing and counselling); (2) a structural-level community stigma reduction programme; (3) individual-level post-test counselling and skill-building support groups; and (4) both individual-and structural-level activities. For the purpose of this systematic review, the individual-level post-test counselling and skill-building support groups will be compared with individual standard of care.
Random-effects models were applied to compare the following outcomes of interest for meta-analysis by type of control intervention, and by type of control intervention and length of follow-up post intervention: any injecting risk behaviour (seeFigures 3and4), including sharing of needle/syringes (seeFigures 5and6), or other injecting equipment (seeFigures 7and8) and frequency of injecting (seeFigures 9and10), reported separately or as an aggregated outcome; and any sexual risk behaviour (seeFigures 11and12), including unprotected sex (seeFigures 13and14) or number of sexual partners (seeFigure 15), reported separately or as an aggregated outcome.
Results
Study selection and assessment
The electronic database searches to 26 May 2015 resulted in 2493 citations; an additional 77 citations were identified from 1 January 2015 to 9 December 2016 (Figure 1). One additional manuscript was identified from hand-searching other reviews’reference lists. After removal of duplicates, 1903 citations remained. In total 1771 abstracts were excluded as they did not meet eligibility criteria and 132 abstracts were selected for full-text assessment, including four related manuscripts referenced in these selected texts.83–86Eighty-nine articles
were excluded for the following reasons: they were not RCTs (n=34);47,68,87–118the outcomes of interest for this
review were not assessed or presented at follow-up (n=29);86,119–146outcomes were not presented by PWID
(n=6);147–152the number of PWID was not reported (n=4);153–156or the intervention studied was not
psychosocial (n=4).157–160Additionally, 10 manuscripts were excluded as the results did not compare
intervention groups161–166or did not evaluate the effect of the intervention.167–170One further manuscript
was excluded because the same psychosocial intervention was delivered to each treatment group, and the difference between treatment groups was receipt of a coupon for 90 days of free methadone maintenance treatment (MMT).171One manuscript published in Chinese was excluded.172
In total, 42 manuscripts from 32 trials were eligible,22,68–77,81–85,173–19841 originating from the electronic database
searches, and one from hand-searching recent reviews and review of reviews178was added as a result of this
process. Twenty-four trials were included in the meta-analysis.22,68–73,75,76,81,82,173,175,179,180,182,186,188,189,192–195,197
The reasons for excluding eight trials from the meta-analysis were not providing the number of PWID for control and intervention groups at follow-up,84,174,176,177,196only providing risk ratios,178outcome combined
HIV infections with sexually transmitted infections77and data for
‘unsafe injection practices’were presented
only at baseline.74
Quality and publication bias assessment
The summary of authors’judgements about the quality of each trial included in the systematic review is described inFigure 2. The risk of bias varied between trials. Incomplete outcome data was the most common risk of bias found in the trials included in the review, but selective outcome reporting also contributed to potential risk of bias for some trials. Other potential sources of bias included altering randomisation protocols depending on the number of participants enrolled on a particular day;182statistically significant differences
between groups at baseline in the injecting subscale;173variation in the TAU group across sites;68possible
crossover contamination between groups;69,75,179,180,182,194a high proportion of excluded individuals with
the excluded individuals differing significantly to those included;71and large variations reported in the
follow-up period.74
DOI: 10.3310/hta21720 HEALTH TECHNOLOGY ASSESSMENT 2017 VOL. 21 NO. 72
© Queen’s Printer and Controller of HMSO 2017. This work was produced by Gilchristet al.under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.