A randomised trial compared two instruments for assessing self-reported adherence to antiretroviral medications: (1) a day-by-day recall instrument that elicited the number of missed doses in each of the prior three days (3-day instrument; n=64) and (2) a general recall instrument that elicited an estimate of proportion of pills taken during the prior seven days (7-day instrument; n=70). Adherence was measured at study visits over 12 months among participants in a clinical trial assessing treatment strategies for individuals with virologic failure and multidrug-resistant HIV. Participants had a median (interquartile range) of 133 (41-264) CD4 cells/ml(3) and a median of 10 major HIV resistance mutations at baseline. Mean adherence levels were 90-98% throughout the study. There was a greater trend in the likelihood of 100% adherence when measured by the 3-day versus the 7-day instrument (odds ratio (OR)=1.45; p=0.06). The likelihood of consistent 100% adherence measured by either instrument decreased over time (p<0.001). Participants reporting 100% adherence at more than half of study visits had better virologic and immunologic outcomes at month-12 compared to those reporting 100% adherence at half or fewer visits (HIV RNA decline of 0.96 versus 0.51 log, respectively, p=0.02; and CD4 cell increase of 51.0 versus 17.8 cells, p=0.04). This study demonstrated the utility of the general 7-day recall adherence self-report instrument as well as the 3-day day-by-day recall adherence self-report instrument for measuring antiretroviral adherence. Self-reported adherence was significantly associated with virologic and immunologic outcomes in this population with advanced drug-resistant HIV disease.