Rates of cognitive impairment in a South African cohort of people with HIV: variation by definitional criteria and lack of association with neuroimaging biomarkers.

There is wide variation in the reported prevalence of cognitive impairment in people with HIV (PWH). Part of this variation may be attributable to different studies using different methods of combining neuropsychological test scores to classify participants as either cognitively impaired or unimpaired. Our aim was to determine, in a South African cohort of PWH (N = 148), (a) how much variation in reported rates was due to method used to define cognitive impairment and (b) which method correlated best with MRI biomarkers of HIV-related brain pathology. Participants completed detailed neuropsychological assessment and underwent 3 T structural MRI and diffusion tensor imaging (DTI). We used the neuropsychological data to investigate 20 different methods of determining HIV-associated cognitive impairment. We used the neuroimaging data to obtain volumes for cortical and subcortical grey matter and total white matter and DTI metrics for several white matter tracts. Applying each of the 20 methods to the cognitive dataset resulted in a wide variation (20-97%) in estimated rates of impairment. Logistic regression models showed no method was associated with HIV-related neuroimaging abnormalities as measured by structural volumes or DTI metrics. We conclude that for the population from which this sample was drawn, much of the variation in reported rates of cognitive impairment in PWH is due to the method of classification used, and that none of these methods accurately reflects biological effects of HIV in the brain. We suggest that defining HIV-associated cognitive impairment using neuropsychological test performance only is insufficient; pre-morbid functioning, co-morbidities, cognitive symptoms, and functional impairment should always be considered.