BACKGROUND: Structured treatment interruption (STI) of antiretroviral therapy (ART) could potentially reduce cost and toxicity, but clinical efficacy requires evaluation. METHODS: An assessment of fixed-duration STI was nested in DART, a multicentre trial comparing strategies for monitoring ART in Uganda and Zimbabwe (ISCRTN 13968779). Of 3316 ART-naive symptomatic adults with CD4 cell count < 200 cells/microl at ART initiation, 813 with > or = 300 cells/microl after 48 or 72 weeks underwent a second randomization to either STI, cycles of 12 weeks on/off (408), or continuous ART (CT; 405). RESULTS: Median age at STI/CT randomization was 37 years (range, 19-67) and CD4 cell count 358 cells/microl (range, 300-1054). A second review terminated the STI/CT randomisation on 15 March 2006, and participants changed to CT. Median follow-up was 51 weeks (range, 0-85): 99% and 50% of time was spent on ART in CT and STI, respectively. First new World Health Organization (WHO) stage 4 events or death occurred more frequently in STI (24; 6.4/100 person-years) than CT (9; 2.4/100 person-years) (hazard ratio, 2.73; 95% confidence interval, 1.27-5.88; P = 0.007); oesophageal candidiasis being the most frequent event (STI, 13; CT, 3). Nine (1%) participants died (STI, 5; CT, 4). There was no difference in time to first serious adverse event (P = 0.78), although ART change owing to toxicity occurred more with CT (10; 2.6/100 person-years) than with STI (2; 0.5/100 person-years) (P = 0.02). CONCLUSIONS: Although absolute rates of WHO stage 4 events/death were low, 12 week STIS initiated at a CD4 cell count >/= 300 cells/microl resulted in a greater than twofold increased relative rate of disease progression compared with continuous therapy in adult Africans initiating ART with advanced disease, and cannot be recommended.