Statistical Analysis

Demographic variables were analyzed with χ² test for categorical data. Continuous variables for between-group comparisons, including verbal working memory, processing speed, MMN, and MRS measures, were analyzed with analysis of variance. The group comparisons were repeated using age, sex, and smoking status. For MRS measures, voxel cerebrospinal fluid, white matter, and gray matter volume proportions were used as covariates. Pearson product moment correlations were conducted to examine the association between MMN and MRS measures with specific hypotheses regarding associations between MMN and glutamate and the ratio of glutamine to glutamate. The associations between the ratio of glutamine to glutamate, GABA, MMN, and DST performance were examined with structural equation modeling. All tests were 2 tailed, and significance was set at P ≤ .05 except for nonhypothesized tests, for which a Bonferroni correction was applied.

Structural equation modeling was used to test the effects of neurochemistry and MMN amplitude on DST performance. Models were evaluated separately in patients with schizophrenia and controls. The goodness-of-fit χ² test was used to examine model fits to the data using maximum likelihood estimation. Model fits were evaluated with the Akaike information criterion (AIC)⁵⁶ and root-mean-square error of approximation (RMSEA).⁵⁷ An RMSEA below 0.10 indicates a good fit, and an RMSEA below 0.05 indicates a very good fit. The AIC considers the complexity of the model with the goodness of fit to the sample data and penalizes overfitting, with a minimal value being the preferred model. The conceptual full and comparison models were analyzed. In the case of model comparisons between patients and controls, significant differences in the fit of one model were compared with the other model, and individual paths were allowed to vary in a stepwise manner to determine which connections contributed to the increased fit of the alternative model. The model with the best fit is presented herein, and the other models are shown in eFigure 1, eFigure 2, and eFigure 3 in the Supplement.