Is the HIV dementia scale a reliable tool for assessing HIV-relatedneurocognitive decline?

Abstract

Background: The HIV Dementia scale (HDS) has been recommended as a cross-sectional screen for HIVassociated neurocognitive disorder (HAND) but its longitudinal usefulness has not been optimally established. Method: 55 HIV+ participants underwent baseline and follow-up HDS screening after an average of 3.9 (SD=1.1) months. They had completed standard neuropsychological (NP) evaluations within the last 6-18 months to calculate a baseline HAND rate (49.1%) and gold standard cognitive change performance: 12.7% showed mild to moderate decline compared to normative standards for NP change; 80% confidence interval; 1-tailed, the rest of the sample was cognitively stable. After normalizing HDS raw-scores, standard regression-based change scores were developed to quantify HDS-based decline, which corrected for practice effect, regression towards the mean, and yielded individual predictions of neurocognitive change. Clinically significant decline was defined as a z-score outside the 80% confidence interval; 1-tailed. Baseline HIV disease and laboratory data were collected. Results: The magnitude of HDS reliability was very large r=0.76 (p<0.0001). HDS-testing found that 21.8% significantly declined. Compared to gold standard NP decline, the HDS showed 57% sensitivity and 82% specificity. Only participants (n=4) that declined moderately (median HDS-change z-score=-2 SD below mean of zero; at least 3-4 points decline in raw HDS-score) were congruently identified. Standard regression-based change scores did not operate optimally at smaller magnitudes of decline. HAND diagnosis (gold standard) at baseline (p<0.03) and more severe HAND were associated with greater chance of decline (p<0.03). No baseline HIV biomarkers were associated with decline. Conclusions: The HDS is a reliable screen to detect at least moderate neurocognitive decline in individuals with MND and HAD. Other screening instruments are needed to detect milder levels of neurocognitive decline. Alternatively, non-parametric statistical modelling is needed to improve predictions of individual cognitive change on such scales that typically have a restricted range of values. © 2013 Lu GM, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. © 2013 Lu GM, et al.