Background Adherence is key to the success of antiretroviral therapy. Enhanced partner support might benefit patients with previous treatment failure. We aimed to assess whether an enhanced partner-based support intervention with modified directly observed therapy would improve outcomes with second-line therapy in HIV-infected patients for whom first-line therapy had failed. Methods We did a multicentre, international, randomised clinical trial at nine sites in Botswana, Brazil, Haiti, Peru, South Africa, Uganda, Zambia, and Zimbabwe. Participants aged 18 years or older for whom first-line therapy had failed, with HIV RNA concentrations greater than 1000 copies per mL and with a willing partner, were randomly assigned (1:1), via computergenerated randomisation, to receive partner-based modified directly observed therapy or standard of care. Randomisation was stratified by screening HIV RNA concentration (Ã¢â€°Â¤10 000 copies per mL vs >10 000 copies per mL). Participants and site investigators were not masked to group assignment. Primary outcome was confirmed virological failure (viral load >400 copies per mL) by week 48. Analysis was by intention to treat. This trial is registered with ClinicalTrials.gov, number NCT00608569. Findings Between April 23, 2009, and Sept 29, 2011, we randomly assigned 259 participants to the modified directly observed therapy group (n=129) or the standard-of-care group (n=130). 34 (26%) participants in the modified directly observed therapy group achieved the primary endpoint of virological failure by week 48 compared with 23 (18%) participants in the standard-of-care group. The Kaplan-Meier estimated cumulative probability of virological failure by week 48 was 25Ã‚Â·1% (95% CI 17Ã‚Â·7.32Ã‚Â·4) in the modified directly observed therapy group and 17Ã‚Â·3% (10Ã‚Â·8.23Ã‚Â·7) in the standard-of-care group, for a weighted difference in standard of care versus modified directly observed therapy of 6Ã‚Â·6% (95% CI 16Ã‚Â·5% to 3Ã‚Â·2%; p=0Ã‚Â·19). 36 (14%) participants reported at least one grade 3 or higher adverse event or laboratory abnormality (n=21 in the modified directly observed therapy group and n=15 in the standard-of-care group). Interpretation Partner-based training with modified directly observed therapy had no effect on virological suppression. The intervention does not therefore seem to be a promising strategy to increase adherence. Intensive follow-up with clinic staff might be a viable approach in this setting. Funding AIDS Clinical Trials Group and the National Institute of Allergy and Infectious Diseases, US National Institutes of Health.