Background: The central nervous system is a potential reservoir of HIV. Differences exist in CSF drug penetration between antiretrovirals (ARVs), which may influence the development of resistant virus.  CSF concentrations have been described for some ARVs but not for multiple drugs in a large patient population.

Methods: Patients receiving HAART had concomitant CSF and plasma samples collected. Samples were analyzed for plasma concentrations using LC/MS/MS. In patients with detectable virus, virtual phenotype was performed on both CSF and plasma.

Results: 63 patients (median age 39 years, 76% male, 25% IDU, 59% heterosexual, 62% previous AIDS, 78% neurologic diseases) had paired CSF/plasma samples collected at a median of 8 weeks from starting last ARV regimen (IQR 3-25). At time of sample collection median CD4 count, HIV-1 RNA in plasma and CSF were 111/mL (53-275), 3.34 log₁₀/mL (2.23-4.84), and 2.00 log₁₀/mL (1.90-3.11), respectively. Samples included 356 drug concentrations (178 each in plasma and CSF). CSF concentrations were undetectable for didanosine, efavirenz, and nelfinavir. Only 1 of 8 samples for ritonavir had detectable CSF levels.  CSF concentrations were highest for nevirapine with a median CSF/plasma ratio of 0.63 (n=16), followed by lamivudine (0.23; n=55), stavudine (0.20; n=31), and indinavir (0.11; n=18). Virtual phenotype in CSF and plasma was available for 40 patients. In 18 patients, differences in fold-change resistance between the CSF and the plasma virus were noted for at least one drug. Factors associated with having differences in fold-change resistance were number of drug exposures (p=0.02) and presence of neurologic disease (p=0.05). A significant association was found between duration of therapy and fold-change resistance in CSF (p at Spearman’s=0.02) and plasma (p=0.002).

Conclusions: Differences in CSF penetration exist between ARVs, with a number of drugs having limited CSF concentrations. Different resistance profiles exist in CSF isolates compared to those in plasma.  Treatment decisions for CNS manifestations may require knowledge of drug penetration and susceptibility of HIV in CSF.