The prevalence of HIV-associated neurocognitive disorder (HAND) is increasing despite the widespread use of combination antiretroviral therapy (ART). Initial reports suggest that the use of antiretrovirals with good central nervous system (CNS) penetration leads to better neurocognitive outcomes, but this has not yet been confirmed in a large cohort study or randomised controlled trial. There is emerging evidence that high CNS concentrations of some antiretrovirals are potentially neurotoxic and may be associated with the development of HAND. Antiretroviral CNS exposure is ideally determined by determining the ratio of cerebrospinal fluid (CSF):plasma area under the curve of unbound drug, but usually only total drug concentrations are measured and the ratio of CSF:plasma drug concentration is done at a single time point, which can result in misclassifying CNS penetration ability. Efavirenz was previously thought to have poor CNS penetration, measured by the CSF:plasma ratio (0.87 %), but when unbound concentrations were measured it was found to have good CNS penetration (85 %). Indinavir and efavirenz are the only antiretroviral drugs for which CNS area under the concentration–time curves using unbound plasma and CSF concentrations has been calculated. Patient data to support the contribution of blood–brain barrier transporter polymorphisms to CNS antiretroviral concentrations are currently limited and lack power to detect true associations. Correlations between CNS antiretroviral exposure and effect is multifaceted, and to accurately predict CNS effects there is a need to develop a sophisticated intra-brain pharmacokinetic–pharmacodynamic–pharmacogenetic model that includes transporters as well as the influence of HIV.