Clinical validation of the NeuroScreen

Abstract

The NeuroScreen comprises two easily administered components: the Brief NeuroCognitive Screen (BNCS), designed to estimate the frequency of human immunodeficiency virus (HIV)-associated cognitive disorders; and the Brief Peripheral Neuropathy Screen (BPNS), for distal sensory polyneuropathy (DSPN) in HIV. In this study, both the NeuroScreen and a more extensive standardized validation neurodiagnostic evaluation were administered to HIV-positive subjects (N = 301) enrolled in two large cohort studies at multiple sites. BNCS performance was summarized in the form of a demographically adjusted mean z-score, the NPZ3. The area under the receiver-operating characteristic (ROC) curve for the BNCS as compared to the reference standard neuropsychological (NP) evaluation was 0.74 (95% confidence interval [CI] 0.69, 0.79). Using a cut-point of -0.33 on the NPZ3 provided a correct classification rate of 68%, with roughly balanced sensitivity (65%) and specificity (72%). Under the assumption of a 30% prevalence of cognitive impairment, the calculated positive predictive value (PPV) of the BNCS was 86%. Relative to its reference standard, a modified Total Neuropathy Score (TNS) administered by a neurologist, the BPNS gave a similar correct diagnostic classification rate of 78%; sensitivity 49% [95%, 60%]; specificity 88% [95% Cl82%, 91%]. Under the assumption of a 40% prevalence of DSPN, the PPV of the BPNS was 72%. These predictive values suggest that the NeuroScreen will be useful for tracking trends in the prevalence of HIV-associated neurologic disease in large cohorts in the era of combination antiretroviral therapy. However, because it yields substantial numbers of false positives and negatives, the NeuroScreen may be less useful in evaluating individual patients.