Episodic memory dysfunction in individuals at high-risk of psychosis: a systematic review of neuropsychological and neurofunctional studies.

Cognitive impairment is one of the key features of schizophrenia, with the largest effect sizes identified for verbal learning and memory, however little is known about its features in the time that precedes psychosis onset. Here we review a total of thirty-two studies that examined memory and learning in populations at clinical and genetic high-risk for psychosis. These studies can be divided into three different categories based on their design. Some were cross-sectional and examined neuropsychological differences between high-risk individuals and healthy controls. A second type of studies included a clinical follow-up that permitted dividing participants based on their outcome to examine abnormalities specific of subsequent transition to psychosis as well as the inclusion of cognitive data in regression models for psychosis prediction. A third type of studies had a longitudinal design with measures repeated at two or more time points in order to examine the course of cognitive functions over time. We also reviewed all neurofunctional studies investigating subjects at risk for psychosis and focused on brain alterations associated with the above neuropsychological impairments. Results of cross-sectional studies revealed impairments in verbal learning and memory as well as executive function/working memory, attention and processing speed; in most of these studies, performance of individuals at clinical or genetic high-risk was intermediate between that of healthy controls and first episode patients. Neurofunctional investigations revealed altered brain functioning in the neural circuits underlying memory and learning processes. Results are less consistent in terms of clearly identifying cognitive differences and their progression over time between individuals subsequently developing psychosis and those remaining non-psychotic. However, studies that included cognitive variables in regression models or prediction algorithms suggest that some areas of cognition, particularly verbal memory, can increase the accuracy obtained in the identification of individuals developing psychosis beyond that based purely on psychopathological measures, suggesting that the inclusion of neurocognitive tests of domains for which there is evidence of prediction potential could be useful in a stepwise assessment of risk.