BACKGROUND: Stopping antiretroviral therapy in patients with HIV-1 infection can reduce costs and side-effects, but carries the risk of increased immune suppression and emergence of resistance. METHODS: 430 patients with CD4-positive T-lymphocyte (CD4) counts greater than 350 cells per muL, and viral load less than 50 copies per mL were randomised to continued therapy (n=146) or scheduled treatment interruptions (n=284). Median time on randomised treatment was 21.9 months (range 16.4-25.3). Primary endpoints were proportion of patients with viral load less than 50 copies per mL at the end of the trial, and amount of drugs used. Analysis was intention-to-treat. This study is registered at ClinicalTrials.gov with the identifier NCT00113126. FINDINGS: Drug savings in the scheduled treatment interruption group, compared with continuous treatment, amounted to 61.5%. 257 of 284 (90.5%) patients in the scheduled treatment interruption group reached a viral load less than 50 copies per mL, compared with 134 of 146 (91.8%) in the continued treatment group (difference 1.3%, 95% CI-4.3 to 6.9, p=0.90). No AIDS-defining events occurred. Diarrhoea and neuropathy were more frequent with continuous treatment; candidiasis was more frequent with scheduled treatment interruption. Ten patients (2.3%) had resistance mutations, with no significant differences between groups. INTERPRETATION: Drug savings with scheduled treatment interruption were substantial, and no evidence of increased treatment resistance emerged. Treatment-related adverse events were more frequent with continuous treatment, but low CD4 counts and minor manifestations of HIV infection were more frequent with scheduled treatment interruption.