A consecutive series of 96 patients with HIV-associated dementia treated with HAART were studied to identify specific clinical factors associated with an improved response to therapy. The Memorial Sloan-Kettering dementia severity scale and the HIV Dementia Scale were used to assess outcomes. Of 30 patients meeting the inclusion criteria with adequate follow-up, 60% improved neurologically and 40% progressed. There was a trend toward improvement associated with plasma viral suppression, whereas progression was strongly associated with injection drug use history (odds ratio, 13.3). Age, ethnicity, gender, adherence, and predicted CNS penetrance of HAART were not associated with improved outcomes.