Background: A high prevalence of mild cognitive impairment (CI) has been reported in HIV-infected individuals on combination antiretroviral therapy (cART). An accurate screening tool would be useful to identify those at risk of HIV-associated CI who should undergo neuropsychological assessment (NPA). We assessed diagnostic characteristics of four often advocated screening tools: Mini Mental State Examination (MMSE), HIV Dementia Scale (HDS), Montreal Cognitive Assessment (MoCA), and a 3-item questionnaire published by Simioni et al (Simioni questionnaire).
Methods: Each screening tool and NPA were applied to a subset of HIV-uninfected and HIV-infected men on cART and with undetectable viremia ≥12 months participating in the AGEhIV Cohort Study and enrolled in a nested cognitive substudy. Two methods were used to diagnose HIV-associated CI based on NPA outcomes: Frascati criteria as published by Antinori et al, and Multivariate Normative Comparison (MNC, comparing the cognitive profile of each HIV-infected individual with the overall cognitive profile of all HIV-negatives). Abnormal screening scores were defined by the following classical thresholds: MMSE ≤24/30, MoCA ≤25/30, HDS ≤10/16, HDS ≤14/16 and Simioni questionnaire ≥1/3 “yes, definitely”. Diagnostic characteristics and receiver operating characteristic (ROC) analyses of screening tools were assessed using Frascati and MNC as the gold standards. Optimal thresholds were identified by calculating Youden index.
Results: HIV-positive and HIV-negative groups were highly comparable regarding age (median age: 54 years), sexual preference (92% MSM), educational level, subjective cognitive complaints, and depressive symptoms. HIV-infected men were infected for a median of 13.5 years and had undetectable viremia for a median of 8.3 years. Median nadir and current CD4-count were 170 and 625 cells/mm3. None of the screening tools showed statistically significant between-group differences; sensitivity and specificity were moderate at best (see table). The ROC area under the curve of MMSE, HDS, and MoCA was 0.63, 0.61 and 0.71, and 0.70, 0.67 and 0.58 using Frascati and MNC as gold standard respectively. No large improvements in sensitivity or specificity were seen using optimal thresholds (see table).
Conclusions: Each of the four screening instruments performed poorly in detecting HIV-associated CI. Cognitive deficits in well-suppressed HIV infection are subtle, and no screening instrument so far seems optimal for use in clinical practice.