Background — Adherence is key to antiretroviral therapy (ART) success. Enhanced partner support may benefit patients with prior treatment failure. Methods — We conducted a 1:1 randomized trial of a partner-based modified directly observed therapy (mDOT) compared with standard of care (SOC) at 9 sites in 8 countries. Participants had failed a first-line regimen with HIV RNA >1000 copies/mL and a willing partner. Randomization was computer generated and balanced by site. Participants and site investigators were not masked to group assignment. ART included lopinavir/ritonavir (400/100 mg) twice daily and emtricitabine/tenofovir disoproxil fumarate (200/300 mg) once daily. Trained partners observed one ART dose daily ≥5 days/week for 24 weeks. Primary outcome was HIV RNA >400 copies/mL before or at week 48 and adherence measured with microelectronic monitors was a secondary outcome. Findings — We randomized 129 participants to mDOT and 128 to SOC, 130 (51%) males, 204 (79%) of African origin, 52 (20%) Latino, with median age 38 years. Partners were parents, 57 (22%), spouses 55 (21%), siblings 50 (19%), friends 41 (16%), and others 54 (21%). Primary outcome occurred in 26% (34/129) of mDOT and 18% (23/128) of SOC participants at week 48 (p=0.13). Median adherence was similar [Q1: 95% vs. 96% p=0.38, Q2: 91% vs. 94% p=0.40, Q3: 90% vs. 93% p=0.17, Q4: 90% vs. 93% p=0.36] in mDOT and SOC, respectively. Interpretation —This intervention had no effect on outcomes. Potential reasons include study visits maximizing adherence in both groups and control partners already providing sufficient support. Partner-based training with mDOT does not appear promising to enhance adherence. Intensive follow-up with clinic staff may be a viable strategy in this setting.