Background: Neuropsychological tests are commonly used in psychopharmacological research to understand the nature and the magnitude of cognitive effects of licensed and novel compounds. While the science of cognitive change assessment has advanced considerably, statistical techniques used to guide inferences about differences in cognitive change have not been considered in the same detail, especially in light of recent advances in modeling repeated data.
Methods: Data from a randomized, placebo-controlled, crossover study of the effect of an acute dose of lorazepam on cognitive function in healthy adults were analyzed using five statistical approaches (paired sample t-test, area under curve (AUC), repeated measures analysis of variance (ANOVA), change from baseline, and linear mixed models (LMM)). Results of significance tests and effect sizes were compared at maximum concentration (C(max)) and over the ascending slopes of cognitive performance.
Results: LMM approaches were superior to other statistical approaches with respect to results of significance testing and magnitudes of estimated effect size change.
Conclusions: Results of this study suggest that employment of LMM, which permit examination of specific fixed effects (e.g., time, treatment, treatment x time) and that are not confounded by between-subject variability, provide a sensitive approach to detecting the cognitive effects of pharmacologic challenges, even with small samples.