Objective-To examine associations between plasma cystatin C and neurocognitive impairment (NCI), and its performance as a diagnostic marker before and during initial antiretroviral therapy (ART). Methods-Multivariable logistic regression and generalized estimating equations (GEE) examined associations with NCI, determined by neuropsychological measurements, in participants of a 48-week randomized clinical trial of initial ART. Receiver operator characteristic (ROC) curves examined diagnostic models of NCI. Results-Cystatin C was associated with NCI before ART (OR 3.4 [95%C.I., 1.2, 9.4] for each 2-fold increase in baseline levels) and during 48 weeks of ART, in models that excluded baseline measurements (OR 3.0 [1.2, 7.8] for each 2-fold increase in time-updated levels). The strength of association increased with more severe impairment using HIV-Associated Neurocognitive Disorder criteria (OR 2.2 [0.8, 6.0] with asymptomatic neurocognitive impairment, and OR 4.0 [1.5, 11.0] with mild neurocognitive disorder or HIV associated dementia vs. no impairment, for each 2-fold increase in time-updated levels), or by Global Deficit Scores (OR 2.6 [1.1, 6.3] with mild impairment, and OR 4.6 [1.1, 18.9] with moderate or severe impairment vs. no impairment).