BACKGROUND: Directly observed therapy has been recommended to improve adherence for patients with HIV infection who are on highly active antiretroviral therapy, but the benefit and cost-effectiveness of this approach has not been established conclusively. We did a systematic review and meta-analysis of randomised trials of directly observed versus self-administered antiretroviral treatment. METHODS: We did duplicate searches of databases (from inception to July 27, 2009), searchable websites of major HIV conferences (up to July, 2009), and lay publications and websites (March-July, 2009) to identify randomised trials assessing directly observed therapy to promote adherence to antiretroviral therapy in adults. Our primary outcome was virological suppression at study completion. We calculated relative risks (95% CIs), and pooled estimates using a random-effects method. FINDINGS: 12 studies met our inclusion criteria; four of these were done in groups that were judged to be at high risk of poor adherence (drug users and homeless people). Ten studies reported on the primary outcome (n=1862 participants); we calculated a pooled relative risk of 1.04 (95% CI 0.91-1.20, p=0.55), and noted moderate heterogeneity between the studies (I(2)= 53.8%, 95% CI 0-75.7, p=0.0247) for directly observed versus self-administered treatment. INTERPRETATION: Directly observed antiretroviral therapy seems to offer no benefit over self-administered treatment, which calls into question the use of such an approach to support adherence in the general patient population. FUNDING: None.