Psychometric considerations in using error-related brain activity as a biomarker in psychotic disorders.

Abstract

Psychotic disorders are characterized by profoundly blunted neural responses to errors, as indicated by reductions in two event-related potential (ERP) components: the error-related negativity (ERN) and error positivity (Pe). The potential utility of the ERN and Pe as biomarkers for psychotic disorders is currently limited, however, by an incomplete understanding of their psychometric properties. To address this gap in the literature, we considered the reliability and validity of these measures in both healthy individuals (n = 52) and patients with psychotic illness (n = 84) across two experimental paradigms that have been used in previous studies in schizophrenia: a flankers task and a picture/word matching task. Internal consistency reliability was higher on the flankers compared to the picture/word task overall. On the flankers task, fair internal consistency was achieved among patients with relatively few trials (ERN = five trials, Pe = 12 trials). The number of available error trials influenced reliability among patients more than among healthy individuals, and on the picture/word task more than the flankers task. Moderate convergent validity for the ERN and Pe was observed across tasks in both the patient and healthy groups. ERPs on the flankers task exhibited external validity, and were related to several clinical characteristics, including diagnosis, negative symptom severity, rehospitalization, employment, and neuroticism; associations with the picture/word task were generally weaker. These data indicate that task differences can strongly affect psychometric properties of error-related neural activity indices in healthy and patient populations.