Longitudinal correlates of HIV RNA levels in 2,207 CSF specimens.

Type: Poster
Title: Longitudinal correlates of HIV RNA levels in 2,207 CSF specimens.
Authors: Letendre SL, Vaida F, Ellis R, Clifford D, Gelman B, Marra C, McArthur J, McCutchan A, Simpson D, Grant I, and the CHARTER Group
Date: 03-05-2012
Abstract:Background: Few large projects have evaluated the factors that influence the level of HIV RNA (viral load) in cerebrospinal fluid (CSF) over time. Persistent HIV RNA in the central nervous system (CNS) during ART has implications for HIV eradication. We evaluated the correlates of CSF viral load in 413 multiply sampled subjects who were taking ART at all visits. Methods: All subjects participated in CHARTER, a 6-site, US cohort. Four hundred and thirteen subjects underwent venous and lumbar puncture every 6 months, were assessed at least 3 times, and took ART at all visits (N = 2207). HIV RNA levels were measured by commercial RT-PCR (lower limit of quantitation [LLQ] 50 copies/mL). Other lab data were measured by routine clinical methods. ART distribution into the CNS was estimated by the 2010 CNS penetration-effectiveness (CPE) method. ART adherence was estimated by the ACTG 4-day method. The analysis used longitudinal mixed effects logistic regression. Correlation within subjects was modeled via a subject-specific random effect. The multivariable models were built using backward elimination. The initial model included all variables significant at the 0.2 level in univariable analyses. The variables not significant at the 0.1 level were sequentially eliminated from the model. Results: Mean duration of observation was 33.7 months (range 12 to 84). Thirty-three percent of plasma specimens and 12% of CSF specimens were >LLQ. Twenty variables were associated CSF viral loads >LLQ over time, including higher plasma viral load (p <0.0001), lower current CD4+ cell count (p <0.0001), higher CD8+ cell count (p <0.0001), higher CSF white blood cells (WBC) (p <0.0001), shorter duration of ART (p <0.0001), lower CPE value (p = 0.001), worse adherence (p = 0.002), and protease inhibitor (PI) use (p = 0.006). The best multivariable model included plasma viral load (odds ratio [OR] 18.0), CD4+ (OR 0.9), and CD8+ cell counts (OR 1.1), CSF WBC (OR 1.1), CPE (OR 0.7), and PI use (OR 3.3) (model R2 = 0.22, p < .0001). Conclusions: In this large, longitudinal analysis of treated individuals, many variables were associated with detectable CSF viral loads over time. Only 6 of these were present in multivariable models, including some with important implications. First, suppression of plasma viral load is the most influential factor for suppression of CSF viral load. Second, CD8+ T cells appear to influence HIV RNA levels in CSF by a mechanism distinct from CD4+ T cells. Third, protease inhibitors may influence CSF viral loads by mechanisms that are not adequately captured by the CPE method.