Demographic and Clinical Characteristics
As shown in , the schizophrenia patient group was matched to the healthy comparison group on level of parental education, sex, and race. The patient sample tended to be somewhat older and their education levels were slightly lower than the healthy group, as reflected by statistical trend effects. Average ratings on the BPRS indicate that the patients exhibited low symptom levels overall, as might be expected of a clinically stabilized sample.
P50 Suppression Ratio and P50 Amplitudes
Grand-average ERP waveforms are presented in . As expected, schizophrenia patients exhibited impaired P50 suppression (M = .59, SD = .31) relative to healthy comparison subjects (M = .35, SD = .20) on the ratio measure (F = 6.26, df = 1, 28, p < .05). P50 suppression ratios and amplitudes obtained for the two groups are illustrated in . Although the group by stimulus interaction was not statistically significant (F = .47, df = 1, 28, p = .50), the significant P50 ratio score difference between groups can be attributed largely to differences in Stimulus 2 amplitude (schizophrenia patients: M = 2.51, SD = 1.44; healthy comparison subjects: M = 1.55, SD = 1.16; F = 3.92, df = 1, 28, p = .058) as amplitude differences to Stimulus 1 were not evident between the groups (schizophrenia patients: M = 4.56, SD = 2.63; healthy comparison subjects: M = 4.06, SD = 2.46; F = .29, df = 1, 28, p = .60).
Grand average event-related potential waveforms at the Cz recording site. Waveforms are unfiltered, and the N40 and P50 components are indicated with arrowheads.
Mean P50 suppression ratios and P50 amplitudes to paired stimuli for healthy comparison subjects and schizophrenia patients.
Evaluation of Dipole Model
To assess Gof of the proposed dipole model, the total amount of variance explained was examined. A considerable percentage of the variance was accounted for in healthy subjects: (Stimulus 1 = 94.89%, Stimulus 2 = 94.39%) and patients with schizophrenia (Stimulus 1 = 93.70%, Stimulus 2 = 91.02%). No effects involving group or stimulus approached statistical significance. To validate the adequacy of the proposed model, an anatomically-unconstrained dipole was added but was not found to contribute substantial power in explaining the total variance (i.e., less than one-half percent). Moreover, the individual contribution of each region within the proposed neural network accounted for a significantly greater proportion of the total variance than that of theoretically-unrelated dipoles (all p-values less than .05). The proposed dipole model was also found to explain significantly more of the total variance when compared to the model consisting of 8 theoretically-unrelated dipoles for Stimulus 1 (M’s = 94.26% vs. 91.90%, respectively; F = 10.07, df = 29, p < .01) and Stimulus 2 (M’s = 92.59% vs. 88.90%, respectively; F = 16.67, df = 29, p < .001). Lastly, relatively few dipoles were localized at the edge of the search window. Across the x, y, and z coordinates, 91.35 % of Stimulus 1 and 88.46 % of Stimulus 2 dipoles did not reach the search radius edge for healthy individuals. Similarly, 95.31 % of Stimulus 1 and 85.16 % of Stimulus 2 dipoles fell below the search radius limit for patients with schizophrenia.
Evaluation of Dipole Moments
Comparison of Stimulus 1 and Stimulus 2 dipole moments across the STG, hippocampus, thalamus, and DLPFC revealed a significant main effect for neural structure (F = 42.50, df = 3, 84, p < .001, ε = .43), with the hippocampus and thalamus showing the greatest source strength across both groups (see ). Consistent with the inverse-square law, this pattern of findings is to be expected as deep structures require large current flows to be detected at distant scalp sites. There were no significant interaction effects involving group, neural structure or stimulus.
Mean dipole moments to paired stimuli for the hippocampus, thalamus, dorsolateral prefrontal cortex, and superior temporal gyrus for healthy comparison subjects and schizophrenia patients.
Examination of the relationship between the ratio of the dipole moment of each neural structure and P50 suppression for healthy comparison subjects revealed that the hippocampal dipole moment ratio significantly correlated with the P50 suppression ratio (r = .69, p < .01; see , top panel). Although not statistically significant, there also was some suggestion of an association between the DLPFC dipole moment ratio and the P50 suppression ratio (r = .46, p = .10). STG and thalamic dipole moment ratios were not significantly correlated with the P50 suppression ratio (r = .41, p = .15; r = .23, p = .44, respectively).
Associations between the P50 suppression ratio and the hippocampal dipole moment ratio for healthy comparison subjects and schizophrenia patients.
A somewhat different pattern of associations emerged for schizophrenia patients. In contrast to healthy subjects, patients did not exhibit a significant correlation between the hippocampal dipole moment ratio and the P50 suppression ratio (r = .21, p = .43; see , bottom panel). A significant association was observed between the DLPFC dipole moment ratio and the P50 suppression ratio (r = .61, p < .05). Similar to healthy individuals, correlations between the P50 ratio and the STG or the thalamic dipole moment ratios were not statistically significant (r = .06, p = .82; r = .00, p = .99, respectively).
Planned comparisons were conducted to examine whether the groups differed statistically in the strength of the relationships between the dipole moment ratio for each neural source and the P50 ratio score. For the association between the hippocampal dipole moment ratio and the P50 suppression ratio, the group difference fell just short of conventional levels of statistical significance (p = .06, one-tailed). There were no significant group differences in the strength of relationships between DLPFC, STG or thalamus and the P50 ratio (all p’s > .18, one-tailed). Among healthy subjects, the strength of the associations also did not differ reliably between the four neural structures (p’s > .15 one-tailed), with the possible exception of a marginally significant effect involving the hippocampus and thalamus (p=.08, one-tailed). For patients with schizophrenia, the P50 ratio score correlated more strongly with the DLPFC dipole moment ratio than with the STG or thalamic dipole moment ratios (all p’s < .05).
In order to assess whether hippocampal and DLPFC dipole moment ratios accounted for unique variance in the P50 ratio score, hierarchical linear regression analyses were performed separately for the two groups. In each of the regression analyses, STG and thalamic dipole moment ratios were entered in the first step because both neural structures are implicated in basic auditory processing. As a second step, the hippocampal or DLPFC dipole moment ratio was entered. For healthy subjects, the regression model including the STG, thalamus, and hippocampus accounted for significantly more variance than the model including only STG and thalamus (F change (1, 10) = 7.90, p < .05). When the DLPFC dipole moment ratio was entered during the second step, a statistical trend was observed (F change (1, 10) = 3.94, p = .08). For schizophrenia patients, the regression model including the STG, thalamus, and hippocampus was not significant relative to the model involving only STG and thalamus (F change (1, 12) = 1.37, p = .26). However, the DLPFC dipole moment ratio accounted for unique variance when added to the STG and thalamic dipole moment ratios (F change (1, 12) = 7.29, p < .05).