The authors concluded that a variety of strategies significantly improve adherence to highly active antiretroviral therapy in adults. It is unclear which specific components of the interventions are most effective. Although the review was generally well-conducted, these conclusions may require some caution in interpretation given the poor quality of the outcomes assessment in many of the primary studies.

To assess the effectiveness of behavioural interventions in improving adherence to highly active antiretroviral therapy (HAART).

MEDLINE, PsycINFO, ERIC, EMBASE, the Cochrane Library and CRISP were searched from 1996 to 2005. Experts in the field were consulted and a call for studies was posted on an HAART electronic mailing list (http://mailman1.u.washington.edu/mailman/listinfo/haart_ adherence_ research). The reference lists of relevant articles were handsearched. There were no language restrictions.

Randomised controlled trials (RCTs) of behavioural interventions to address HAART adherence were eligible for inclusion. The interventions were classified in the review as didactic provision of generic information, interactive discussion of patient-specific information, behavioural strategies and external reminders. Most of the studies used a combination of these, delivered to individuals and/or groups by health professionals, over half of whom were researchers rather than regular clinic staff. The intensity of the intervention varied widely, with a median of two intervention sessions (range: 1 to 54) and median duration of 70 days (range: 1 day to 1 year). Control group conditions included wait list, minimal or standard care and/or other measures (e.g. educational videotapes).

Only studies with individuals as the unit of randomisation were eligible: they were required to target participants aged over 18 years. Several studies in the review were restricted to participants with risk markers for non-adherence. A median of 75% of participants were male and a median of 53% were men who have sex with men. Most studies (84%) were set in human immunodeficiency virus (HIV) primary care out-patient clinics, 74% in the USA. The majority of participants in the American studies were from ethnic minorities.

Eligible studies were required to report outcome data on either adherence to therapy or HIV-1 RNA VL (i.e. the percentage of participants either attaining 95% or better adherence, or with an undetectable assay according to the measure used in the primary study, respectively). Adherence was measured by self-report or by electronic data monitoring in the included studies. Some studies conducted assessment only immediately after the intervention, while others conducted up to three follow-up assessments, the first follow up being at a median of 56 days post-intervention (range: 14 to 365). The median recall period for adherence was 7 days (range: 3 to 30). Studies were excluded if sufficient information could not be obtained to compute an effect size.

The authors did not state how the papers were selected for the review, or how many reviewers performed the selection.

The following components of study quality were assessed: sample size, duration of follow-up, retention (overall and by trial arm), handling of missing data and method of measuring adherence.

The authors did not state how the validity assessment was performed.

Odds ratios (ORs), with 95% confidence intervals (CIs), were calculated from the numbers of events in the control and intervention groups of each study. Outcomes in the review were based on the first follow-up period reported or (if no follow-up data were reported) immediately post intervention. Where studies used multiple interventions, only one intervention arm (the most comprehensive or the most consistent with other studies) was reported in the review.

Pairs of reviewers independently abstracted the data from the included studies using standardised data extraction forms. Authors were contacted and asked to provide outcomes data in the standardised form used in the review. Data provided by authors were used in preference to published data.

The data were pooled using the inverse variance of the logged ORs. Pooled ORs were converted back to unlogged ORs and 95% CIs. Heterogeneity was examined by means of the Q statistic. Both fixed-effect and random-effects models were used, with similar results obtained. The results of random-effects models were reported in the review. Sensitivity analyses were conducted to examine the effect of each individual study on the pooled effect estimate, and also to examine whether characteristics of the study, population, outcome measures, methodology or intervention were associated with the size of the treatment effect. Funnel plots and a linear regression test were used to check for publication bias.

Nineteen RCTs (n=1,839) were included in the review.

Median retention rates immediately after the intervention and at first follow-up were 80% (range: 40 to 100) and 70% (range: 55 to 100), respectively. One third of studies imputed missing data or treated missing values as failures, while the rest excluded participants with missing data from the analysis. Over half of the studies relied solely on self-reported adherence and several did not conduct multiple or long-term assessment of the outcomes.

When 18 RCTs (n=1,633) were pooled, participants in the intervention groups were significantly more likely to achieve at least 95% adherence to therapy than those in the control groups (OR 1.50, 95% CI: 1.16, 1.94). There was no statistically significant heterogeneity in this finding and the exclusion of individual studies did not change its significance.

When 14 RCTs (n=1,247) were pooled, there was no statistically significant difference between the intervention and control groups in the likelihood of achieving an undetectable viral load, though a non significant trend favoured the intervention group (OR 1.25, 95% CI: 0.99, 1.59). There was no statistically significant heterogeneity in this finding and the exclusion of individual studies did not substantially change it.

Sensitivity analyses were performed. Stratification by study characteristics resulted in a significantly larger effect size in studies with recall periods of 2 weeks or 1 month than those with recall periods of less than 7 days (p<0.05). No other stratification analyses found significant differences between the groups for either outcome, though power was limited. There was a trend for larger effect sizes in studies in which the intervention included didactic information, those including interactive discussion, non-USA studies, and those in which the data came from the first follow-up (as opposed to immediately post intervention). There was no indication that self-report of adherence inflated the intervention effect, nor that adherence was higher in studies that were restricted to participants perceived as at increased risk of non-adherence.

No evidence of publication bias was detected.

A variety of strategies have a significant positive effect on adherence to HAART in adults. It is unclear which specific components of interventions are the most effective.

The review question and inclusion criteria were clear, the literature search was extensive, and relevant details of the included studies were provided. Two reviewers independently extracted the data, but it was unclear whether similar steps were taken to reduce the risk of error and bias in the study selection and validity assessment processes. Appropriate statistical methods were used to pool the studies, test for heterogeneity and explore the effect of individual study characteristics, and potential sources of bias were well-addressed in the text. Despite some limitations in the reporting of methods, the review was generally well-conducted. However, the authors' conclusions may require some caution in interpretation, owing to the limited quality of the outcomes assessment in the primary studies, with half of the studies relying on self-reported data and several lacking long-term or repeated outcomes assessment.

Practice: The authors stated that it is important to provide patients with basic information about HAART and also to engage them in discussion to address negative cognitive factors, lack of motivation and unrealistic expectations.

Research: The authors stated that future studies should investigate which components of the interventions are the most effective, and also evaluate the effects on adherence of provider characteristics, medication regimen and the wider context (e.g. accessibility of clinic). Multiple assessment and long-term follow-up would be valuable. The effects of variables such as indicators of viral resistance and medication regimens should be considered. Studies of suitable interventions and implementation strategies in resource-constrained settings are also needed.

National Institutes of Health, grant numbers 2 RO1 MH 58986 and F31 MH71179; Minority Dissertation Fellowship.

Amico KR, Harman JJ, Johnson BT. Efficacy of antiretroviral therapy adherence interventions: a research synthesis of trials, 1996 to 2004. J Acquir Immune Defic Syndr. 2006;41:285–297.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.