Directly administered antiretroviral therapy (DAART) is one approach to improving adherence to among human immunodeficiency virus (HIV)-infected drug users. We evaluated the essential features of a community-based DAART intervention in a randomized, controlled trial of DAART versus self-administered therapy. Of the initial 72 subjects, 78% were racial minorities, and 32% were women. Social and medical comorbidities among subjects included homelessness (35% of subjects), lack of interpersonal support (86%), major depression (57%), and alcoholism (36%). At baseline, the median CD4+ cell count was 403 cells/mL and the median HIV-1 RNA load was 146,333 copies/mL (log10 5.31 copies/mL). During the prior 6 months, 33% of subjects had missed a medical appointment, and 47% had visited an emergency department. Although most subjects (67%) preferred to take their own medications, 76% would accept DAART if it were made compulsory. A methadone clinic was the DAART venue acceptable to the fewest subjects (36%), and a mobile syringe-exchange program was acceptable to the most subjects (83%). Adherence was higher for supervised than for unsupervised medication administration (P<.0001), a finding that supports use of daily supervision of once-daily regimens. Moreover, DAART should incorporate enhanced elements such as convenience, flexibility, confidentiality, cues and reminders, responsive pharmacy and medical services, and specialized training for staff.