summarizes the baseline demographic and clinical characteristics of the study participants. The healthy comparison and high-risk groups did not differ significantly on age at testing, WRAT-3 reading score, years of education, parental socioeconomic status, handedness, or ethnicity; however, the groups differed significantly on estimated current IQ, race, and gender ratio. The healthy comparison group had a higher mean estimated current IQ and a lower proportion of white males compared with the clinical high-risk group.
Baseline Demographic and Clinical Characteristics of Participants in a Study of the Effect of Neurocognition on Social and Role Functioning in Individuals at Clinical High Risk for Psychosis
The clinical high-risk group demonstrated significant impairments in social (Cohen’s d=2.01) and role (Cohen’s d=1.78) functioning relative to the healthy comparison group (). Social functioning of the clinical high-risk patients ranged from good to an inability to function socially, and role functioning ranged from good to extreme dysfunction. Although significantly correlated with each other, the social and role measures appear to be independent constructs sharing some method variance (healthy comparison group: r=0.37, p<0.001; clinical high-risk group: r=0.29, p<0.001).
At the time of testing, a majority of the clinical high-risk patients (55.9%) were not receiving any medication. The remaining patients (44.1%) were receiving atypical antipsychotics, antidepressants, mood stabilizers, stimulants, and/or anxiolytics. There were no significant effects of medication type or status (medication naive or treated) for clinical high-risk patients on any of the demographic characteristics or neurocognitive domains reported here.
illustrates the mean performance across the eight neurocognitive domains for the clinical high-risk group relative to the healthy comparison group. The results of the MANCOVA indicated a significant main effect for neurocognitive performance (Wilks’s lambda=0.88; F=3.73, df=7, 186, p<0.01). Gender was a significant covariate (Wilks’s lambda=0.86; F=4.19, df=7, 186, p<0.001), as female patients exhibited poorer global neurocognitive performance than male patients. MANCOVA also revealed that the neurocognitive profile of the clinical high-risk group deviated significantly from flatness, as indicated by the significant interaction between group and neurocognitive performance (Wilks’s lambda=0.88; F=3.47, df=7, 186, p<0.01). The clinical high-risk group showed a global neurocognitive impairment (mean=−0.63, SD=0.99) as compared to the healthy comparison group (mean=0.00, SD=0.61). Post hoc paired t tests in the clinical high-risk group indicated that processing speed (t=4.31, df=126, p<0.001) and verbal memory (t=4.91, df=126, p<0.001) were significantly more impaired relative to the mean of all the other neurocognitive domains.
FIGURE 1 Mean Performance on Eight Neurocognitive Domains, Relative to Healthy Comparison Subjects, Among Patients at Clinical High Risk for Developing Psychosisa
summarizes the individual univariate ANCOVAs for each cognitive domain. Individual ANCOVAs revealed that, relative to the healthy comparison group, the clinical high-risk group was significantly impaired in all eight cognitive domains. There was a significant effect of gender on attention (F=6.00, df=1, 193, p<0.05), visuospatial processing (F=22.04, df=1, 203, p<0.001), and language (F=4.89, df=1, 202, p<0.05), as female patients exhibited poorer neurocognitive performance than male patients in these domains.
Neurocognitive Performance in Patients at Clinical High Risk for Developing Psychosis (N=127)
Relationships Between Cognition and Social and Role Functioning
As shown in , global neurocognition was a significant predictor of social and role functioning at baseline. The global neurocomposite score accounted for 8% and 5% of the variance for social and role functioning, respectively. Among the domain-specific neurocognitive scores, processing speed was a significant predictor of social and role functioning at baseline. Processing speed predicted 10% and 7% of the variance for social and role functioning, respectively. This relationship was independent of positive symptoms, as follow-up regression analyses indicated no significant relationships between total attenuated positive symptoms and either social or role functioning.
Linear Regression Results for the Effects of Neurocognitive Performance on Social and Role Functioning at Baseline in Individuals at Clinical High Risk for Psychosis
Given that processing speed was significantly associated with social and role functioning, additional analyses were performed to determine whether processing speed scores could accurately determine functional status when the latter was dichotomized as good or poor functioning. Of the 127 clinical high-risk patients, 83 (65.4%) were in the poor social functioning group and 85 (66.9%) were in the poor role functioning group. Of the 80 healthy comparison subjects, 74 (94.9%) were in the good social functioning group and 73 (93.6%) were in the good role functioning group. The final logistic regression model indicated that participants with higher processing speed scores had a significantly lower likelihood of poor social functioning (odds ratio=0.60, 95% confidence interval [CI]=0.48–0.76; Wald’s χ2=18.57, df=1, p<0.001) and poor role functioning (odds ratio=0.62, 95% CI=0.50–0.78; df=1, p<0.001) after controlling for gender and race. The final models accounted for 19% and 18% of the pseudovariance (Nagelkerke’s R2) for social and role functioning, respectively. Both models were well calibrated according to the Hosmer-Lemeshow test.
Logistic regression was performed to determine whether neurocognition and functioning could discriminate between clinical high-risk patients and healthy comparison subjects. Processing speed (odds ratio=0.56, 95% CI=0.36–0.87; Wald’s χ2=6.72, df=1, p<0.01), social functional status (odds ratio=34.51, 95% CI=8.98–132.70; Wald’s χ2=26.56, df=1, p<0.001), and role functional status (odds ratio=20.88, 95% CI=6.58–66.27; Wald’s χ2=26.60, df=1, p<0.001) were significant predictors of group membership. The final model was well calibrated and accounted for 71% of the pseudovariance. Participants with impaired processing speed and poor functioning were more likely in the clinical high-risk group.