Abstract
BACKGROUND AND OBJECTIVES: We investigated the relationship between cerebrospinal fluid (CSF) and plasma biomarkers of inflammation, neurodegeneration, and neurocognitive performance in people with HIV (PWH), using longitudinal samples from two previously published cohorts: ACTG A5090 (virally suppressed on antiretroviral therapy, ART) and A736 (ART-naïve or failing). METHODS: We analyzed paired CSF and plasma samples, as well as 7-domain standardized neurocognitive test scores, at baseline and 24 weeks. Biomarkers included markers of inflammation (e.g., TNF-α, IL-6, IP-10) and neurodegeneration (e.g., NFL, p-Tau217, Aβ42), which were quantified via high-sensitivity immunoassays. Associations with cognition were tested using regression, mediation, and interaction models. RESULTS: Cross-sectional analyses revealed nominal associations between inflammatory markers and cognitive performance, with plasma IL-6 and IP-10 at baseline, and CSF TNFα at week 24 showing the strongest correlations (p < 0.05, uncorrected); however, none survived correction for multiple comparisons. Conversely, higher CSF Aβ42 and plasma BDNF were positively associated with memory and executive function. Longitudinally, biomarker changes did not significantly predict change in global cognition (ΔNPZ-8); the strongest trend (p-Tau217, ρ = -0.12, p = 0.38) was not statistically significant, and multivariate models failed to identify robust predictors (R(2) < 0.15). Exploratory mediation analysis suggested CSF TNFα, but not plasma TNFα, partially mediated the effect of CSF HIV RNA on cognition (indirect β = 0.14, 95% CI: 0.045-0.235, p = 0.006), though this finding requires replication in larger cohorts. Within-compartment biomarker correlations were stronger in CSF than plasma, and cross-compartment agreement was highest for TNFα, GFAP, and NFL. ART initiation in A736 led to significant declines in CSF IL-6, IL-10, and TNFα; no changes were observed in A5090. Exploratory interaction models suggested that astrocytic activation may amplify tau-related cognitive risk, but these effects were not statistically reliable when analyses were restricted to participants with complete data for GFAP, tau biomarkers, and cognitive scores. DISCUSSION: These results suggest a potential role of CSF TNFα in mediating the neurocognitive effects of HIV and highlight compartment-specific inflammatory dynamics. Plasma TNFα, GFAP, and NFL may serve as peripheral indicators of CNS pathology, though with only moderate concordance. Astrocyte-tau interactions require cautious interpretation pending replication in larger cohorts.