Statistical analysis

All statistical analysis was performed using IBM SPSS 24 (Chicago, IL). ANCOVA was used to determine main effects of HIV, cognition, and the interaction term between HIV and cognition on CSF biomarker levels, including p-Tau₁₈₁, t-Tau, Aβ42, IP-10, IL-8, fractalkine, TNF-α, and IL-10. Student’s T-tests were used to analyze differences in log₁₀-transformed plasma and CSF viral load between HIV-NC and HAND, and Chi-squared test was used to analyze proportion of HIV-NC and HAND subjects with undetectable levels of CSF viral load.

To calculate Z-scores for IL-8 and IP-10, mean and standard deviation values from a combined cohort of NC_young and NC_older were calculated for each cytokine, and every individual’s IL-8 and IP-10 levels were transformed into Z-scores³². Z-score differences between IL-8 and IP-10 were calculated by subtracting the former’s Z-score by the latter’s Z-score³³.

Finally, for HIV-infected patients who underwent serial CSF collections, each individual’s biomarker levels from the follow-up visits were divided by his or her baseline CSF levels to determine fold-change in IL-8, IP10, p-Tau₁₈₁, and Aβ42 levels. The fold-change values then underwent log₂-transformation to give equal weight to increases and decreases. Because viral load was already analyzed after log₁₀ transformation, the change in plasma viral load was represented by the difference in serial log-transformed values. Pearson’s correlation analysis was used to determine relationship between longitudinal changes in these biomarkers.