Objective: Replicating HIV-1 in the brain is present in HIV encephalitis (HIVE) and microglial nodule encephalitis (MGNE) and is putatively linked with HIV-associated neurocognitive disorders (HAND). A cliniconeurovirological correlation was conducted to elucidate the relationship between brain viral load and clinical phenotype. Subjects and assays: HIV gag/pol RNA and DNA copies were quantified with reverse transcriptase-polymerase chain reaction or polymerase chain reaction in 148 HAART-era brain specimens. Comparison with HAND, HIVE, and MGNE and correlation with neuropsychological (NP) test scores were done using one-way ANOVA with Tukey-Kramer and Spearman tests, respectively. Results: Brain HIV RNA was higher in subjects with HAND plus HIVE versus without HAND (delta = 2.48 log10 units, n = 27 versus 36, P , 0.001). In HAND without HIVE or MGNE, brain HIV RNA was not significantly different versus without HAND (P = 0.314). Worse NP scores correlated significantly with higher HIV RNA and interferon responses in brain specimens (P , 0.001) but not with HIV RNA levels in premortem blood plasma (n = 114) or cerebrospinal fluid (n = 104). In subjects with MGNE, brain HIV RNA was slightly higher versus without MGNE (P , 0.01) and much lower versus with HIVE (P , 0.001). Conclusions: Brain HIV RNA and to a lesser extent HIV DNA are correlated with worse NP performance in the 6 months before death. Linkage occurs primarily in patients with HIVE and MGNE, and these patients could obtain added NP improvement by further reducing brain HIV while on HAART. Patients not in those groups are less certain to obtain added NP benefit. Copyright © 2012 by Lippincott Williams & Wilkins.