Background:  With the advent of combination ART, CD4 and plasma viral load natural history have been altered, which consequently has led to equivocal associations between clinical variables and cognitive performance in cross-sectional studies. The current study examined the association between CD4 counts and plasma viral load modeled over a 12-month period on 18 normalized cognitive tests in a cohort of HIV-infected patients with the expectation that worse clinical history would predict worse cognitive performance.

Methods:  This study included 81 ethnically diverse HIV-infected patients (53% male; 42% black) with a mean CD4 of 428 (range 7 to 1518) copies/mL. At the time of cognitive testing, 38% had detectable plasma viral load. A linear random effects model was used to predict subject-specific CD4 counts at testing and the slope of CD4 leading up to testing. Due to the non-normal distribution of plasma viral load, measures were categorized as detectable and undetectable (<75). Plasma viral load trajectory patterns were then derived and participants were placed into 1 of 3 groups:  1, undetectable at time of testing and during the year; 2, detectable at time of testing and during the year; or 3, undetectable at time of testing but detectable during the year. We examined the relationships between the clinical trend measures and each cognitive performance score using multivariable linear regression (p <0.05 were considered significant).

Results:  Of 4 attention measures, 3 were positively related to predicted CD4 counts at testing (average of 0.14 higher z score/100 unit higher CD4 count, p <0.05). Of 5 executive measures, 4 were related to the predicted CD4 count (average of 0.24/100 unit higher CD4 count); 3 of 5 scores to the slope of CD4 trajectories (average 0.75 higher z score per 10 unit CD4 increase/month), p <0.05); 2 of 5 scores were higher (1.3 z scores, p <0.05) among patients in plasma viral load group 1 compared to those in group 3; and 2 of 5 were found to be higher among those in group 3 compared to group 2 (1.1 z scores). Effect sizes were essentially the same when models were adjusted for medication type and ethnicity. Associations with the CD4 and plasma viral load measures were not found in motor function, language, or learning and memory domains.

Conclusions:  Recent clinical history can be an important predictor of current cognitive dysfunction and, if validated, clinical variables and cognitive dysfunction models may dramatically improve our understanding of disease evolution and progression.