Objective: To compare two definitions of neurocognitive impairment (NCI) in a large clinical trial of effectively-treated HIV-infected adults at baseline.
Methods: Hopkins Verbal Learning test-Revised (HVLT-R), Colour Trail (CTT) and Grooved Pegboard (GPT) tests were applied exploring five cognitive domains. Raw scores were transformed into Z-scores and NCI defined as summary NPZ-5 score one standard deviation below the mean of the normative dataset (i.e. <-1SD) or Z-scores <-1SD in at least two individual domains (categorical scale). Principal component analysis (PCA) was performed to explore the contribution of individual tests to the total variance.
Results: Mean NPZ-5 score was -0.72 (SD 0.98) and 178/548 (32%) participants had NPZ-5 scores <-1SD. When impairment was defined as <-1SD in at least two individual tests, 283 (52%) patients were impaired. Strong correlations between the two components of the HVLT-R test (learning/recall) (r = 0.73), and the CTT and (attention/executive functioning) (r = 0.66) were observed. PCA showed a clustering with three components accounting for 88% of the total variance. When patients who scored <-1SD only in two correlated tests were considered as not impaired, prevalence of NCI was 43%. When correlated test scores were averaged, 36% of participants had NPZ-3 scores <-1SD and 32% underperformed in at least two individual tests.
Conclusion: Controlling for differential contribution of individual test-scores on the overall performance and the level of correlation between components of the test battery used appear to be important when testing cognitive function. These two factors are likely to affect both summary scores and categorical scales in defining cognitive impairment.
Trial registration: EUDRACT: 2007-006448-23 and ISRCTN04857074.