Background Some guidelines recommended two-drug antiretroviral therapies as alternative regimens to triple therapy in selected patients with the aim of reducing drug burden and toxicity and preserving future treatment options. We aimed to assess the efficacy and safety of dual-therapy versus triple therapy as first-line treatment or in treatment simplification. Methods For this systematic review and meta-analysis, we searched Medline, Embase (via OVID), the Cochrane Trial Registry, and major conference proceedings for randomised trials published between Jan 1, 2008, and Dec 31, 2015. We included studies comparing dual-therapy (from two independent classes) antiretroviral regimens as a first-line or a switch strategy (in virologically suppressed individuals) with standard triple-drug regimens. Our primary outcome was the risk of virological failure (non-completion=failure) at the 48 week timepoint. We did a random-effect meta-analysis to pool the relative risk (RR) or odds ratio (OR) for each of the outcomes. Findings For the primary outcome, we included 21 studies (11 first-line and ten switch studies), providing data for 4821 individuals (2478 in dual-therapy groups and 2343 in control groups). Overall, the RR of failure with dual-therapy compared with triple-therapy (control) groups was 1·14 (95% CI 0·91–1·43). In first-line studies, the RR of failure for dual-therapy versus control groups was 1·17 (0·94–1·47; I2=51%), which reduced to 1·05 (0·86–1·28; I2=26%) on exclusion of maraviroc-containing studies. In switch studies, the RR of failure for dual-therapy versus control groups was 1·21 (0·72–2·02; I2=67%), which reduced to 1·13 (0·64–1·99; I2=61%) after exclusion of maraviroc-containing studies. In patients with a baseline viral load of more than 100 000 copies per mL, RR of failure for dual-therapy versus control groups was 1·24 (1·03–1·49), which reduced to 1·18 (0·94–1·47) on excluding maraviroc-containing studies. We recorded the ORs for dual-therapy versus control groups for serious adverse events (1·16 [0·92–1·48]), adverse events (0·82 [0·52–1·28]), and mutations (2·11 [1·32–3·36]). Interpretation Dual therapy, especially with regimens excluding maraviroc, could be safe and efficacious, particularly in patients with baseline viral loads of less than 100 000 copies per mL. However, dual therapy seems to have a greater risk of selecting resistance mutations compared with standard triple therapy. Funding None.