Psychometric characteristics of the MATRICS Consensus Cognitive Battery in a large pooled cohort of stable schizophrenia patients

• Published in: Schizophrenia research Vol. 190; pp. 172 - 179
• Main Authors: Georgiades, Anastasia, Davis, Vicki G., Atkins, Alexandra S., Khan, Anzalee, Walker, Trina W., Loebel, Antony, Haig, George, Hilt, Dana C., Dunayevich, Eduardo, Umbricht, Daniel, Sand, Michael, Keefe, Richard S.E.
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.12.2017
• Subjects: Adult; Age Factors; Cognition; Computer Simulation; Female; Humans; Male; MATRICS Consensus Cognitive Battery; Neuropsychological Tests; Psychiatry; Psychometrics; Regression Analysis; Reproducibility of Results; Schizophrenia; Schizophrenia - diagnosis; Schizophrenic Psychology; Sex Factors; Schizophrenia; MATRICS Consensus Cognitive Battery; Cognition
• DOI: 10.1016/j.schres.2017.03.040
• ISSN: 0920-9964; 1573-2509

The MATRICS Consensus Cognitive Battery (MCCB) was developed to assess cognitive treatment effects in schizophrenia clinical trials, and is considered the FDA gold standard outcome measure for that purpose. The aim of the present study was to establish pre-treatment psychometric characteristics of the MCCB in a large pooled sample. The dataset included 2616 stable schizophrenia patients enrolled in 15 different clinical trials between 2007 and 2016 within the United States (94%) and Canada (6%). The MCCB was administered twice prior to the initiation of treatment in 1908 patients. Test-retest reliability and practice effects of the cognitive composite score, the neurocognitive composite score, which excludes the domain Social Cognition, and the subtests/domains were examined using Intra-Class Correlations (ICC) and Cohen's d. Simulated regression models explored which domains explained the greatest portion of variance in composite scores. Test-retest reliability was high (ICC=0.88) for both composite scores. Practice effects were small for the cognitive (d=0.15) and neurocognitive (d=0.17) composites. Simulated bootstrap regression analyses revealed that 3 of the 7 domains explained 86% of the variance for both composite scores. The domains that entered most frequently in the top 3 positions of the regression models were Speed of Processing, Working Memory, and Visual Learning. Findings provide definitive psychometric characteristics and a benchmark comparison for clinical trials using the MCCB. The test-retest reliability of the MCCB composite scores is considered excellent and the learning effects are small, fulfilling two of the key criteria for outcome measures in cognition clinical trials.