Abstract
Antiretroviral therapy (ART) drugs vary in their distribution into cerebrospinal fluid (CSF), which can be estimated using the central nervous system (CNS) penetration effectiveness (CPE) score. Although higher CPE has been associated with lower CSF HIV RNA levels, its relationship to CSF inflammation is less clear. We investigated associations between CPE and three CSF immune biomarkers (CXCL10, TNF-α, and IL-6) in 275 virally suppressed people with HIV (PWH) on three-drug ART regimens using a training-validation design. Participants were randomized into training (TG, n = 144) and validation (VG, n = 131) groups with similar demographics, ART characteristics, and CPE scores. The CSF levels of the biomarkers were quantified by bead suspension array-based immunoassays. In both groups, higher CPE correlated with lower levels of CXCL10 (TG: r = -0.31, p < 0.001; VG: r = -0.30, p < 0.001) and TNF-α (TG: r = -0.19, p = 0.04; VG: r = -0.18, p = 0.03), with remarkably similar effect size. CPE did not correlate with IL-6 in either group. Multivariable models confirmed the associations between higher CPE and both lower CXCL10 (R(2) = 0.16, p < 0.001) and TNF-α (R(2) = 0.07, p = 0.02) in CSF, and supported the relative resistance of IL-6 to ART effects. During suppressive ART, regimens that achieve higher concentrations in the CNS may better reduce some indicators of CSF inflammation (CXCL10 and TNF-α, closely related to the interferon pathway), but they may not fully normalize the neuroimmune environment (IL-6). Distinct ART regimens may produce different neuroimmune signatures, potentially contributing to heterogeneous patterns of brain injury.