Although highly active antiretroviral therapy (HAART) can reverse HIV-related neurocognitive impairment (NCI), neuropsychologic (NP) deficits may persist in a substantial proportion of patients despite antiretroviral treatment. We assessed the prevalence and predictors of persistent NP deficits despite long-term HAART in patients with HIV-related NCI. A group of 94 patients with HIV-related NCI underwent 2 to 7 serial NP batteries, neurologic examination, and brain imaging studies. Patients received HAART for a mean of 63 (range: 6-127) months. According to NP assessment results, patients were considered to have reversible or persistent NP deficits. Kaplan-Meier analyses and Cox proportional hazards models were used to analyze time to first evidence of NP deficit reversion. Persistent NP deficits were observed in 59 (62.8%) patients. Age, gender, Centers for Disease Control and Prevention stage, risk category, CD4 cell count, plasma viral load, and use of central nervous system-penetrating drugs were not associated with persistent NP deficits. By contrast, patients with persistent NP deficits were less educated and showed poorer baseline performances in NP measures exploring concentration and speed of mental processing, memory, and mental flexibility. In multivariable analyses, only the baseline severity of NCI, as measured by the composite NPZ8 global score (odds ratio = 3.07, 95% confidence interval: 1.54 to 6.08; P = 0.001) remained significantly associated with persistent NP deficits. The severity of NCI at HAART initiation seems to be the strongest predictor of persistent NP deficits despite long-term HAART. Our data indicate that HAART should be initiated as soon as NCI is diagnosed to avoid potentially irreversible neurologic damage.