665. Assessing cognitive impairment in schizophrenia (CIAS) clinical trials: meta-analysis of CANTAB computerized assessments of MATRICS domains

Abstract

Abstract Background Cognitive impairment is a core symptom in people with schizophrenia and represents an important treatment target. The MATRICS process identified seven cognitive domains particularly relevant to CIAS and recommended pencil and paper neuropsychological assessments of these domains. However, computerized tests offer important operational advantages for clinical trials. A key requirement for the selection of cognitive tasks is sensitivity to impairment in patients with schizophrenia. We focus on data from the CANTAB cognitive battery, which has been used for more than 30 years, generating a substantial volume of case-control data, and has previously been used in clinical trials targetting CIAS. Aims & Objectives This meta-analysis aims to quantify case-control differences across seven computerized cognitive tasks assessing MATRICS domains and evaluate the impact of patient and study characteristics on performance using meta-regression to understand the robustness of the observed effects. Method Seven CANTAB tasks were analyzed, corresponding MATRICS cognitive domains: (Speed of processing – Choice Reaction Time (RT), Attention/vigilance - Rapid Visual Processing (RVP), Working memory - Spatial Span (SSP), Visual learning – Paired Associates Learning (PAL), Verbal learning - Verbal Recognition Memory (VRM), Social cognition (Emotion Recognition) and Reasoning and problem solving - Stockings of Cambridge (SOC)). Following PRISMA guidelines, publications were identified through PubMed and Google Scholar databases. Inclusion criteria were: 1) DSM or ICD for diagnosis 2) Includes a healthy comparison group 3) Used CANTAB to assess cognition 4) Reported sufficient descriptive data to conduct meta-analysis 5) Published in a peer-reviewed journal. Studies were evaluated for eligibility and quality by two independent raters. Descriptive statistics and relevant covariates (e.g. mean age and sex) were extracted. Meta-analysis and meta-regression were conducted using a random effects model, accounting for expected shared variance within studies using multiple tasks. We explored both task-specific and cross-task case-control differences. The effect of age, sex and the inclusion of patients with First Episode Psychosis (FEP) were evaluated for their moderating effect on the magnitude of overall case-control difference. Results 79 eligible publications were identified, with between 10 and 277 participants with schizophrenia per study and between 12 and 207 healthy controls. Significant case-control differences were observed in all CANTAB tasks, with moderate to large effect sizes. Aggregating across tasks there was a large overall case-control difference in performance, with significant heterogeneity. This heterogeneity did not introduce systematic bias, and exclusion of four outlier comparisons (Cook’s distance >1) did not change the results. There was no significant effect of CANTAB task or available demographic and clinical variables. Discussion & Conclusions CANTAB computerised tasks are sensitive to impairments in patients with schizophrenia across the MATRICS cognitive domains, with moderate to large effect sizes, demonstrating sensitivity to CIAS. Significant heterogeneity was observed which was not accounted for by task or demographic or clinical factors reported in these studies, suggesting that performance is globally affected across domains implying that a briefer cognitive battery may have equivalent sensitivity with reduced patient burden.