Preliminary analyses were conducted to test for the effect of demographics (sex, race, and age) on cognitive performance. These analyses revealed no significant relations between these factors and performance on any of the cognitive measures. There was also no evidence of an adverse medication effect on cognitive performance.
Analyses were also conducted to compare the subjects who underwent the Time 2 (T2) assessment with those who did not. The results revealed no differences for any cognitive or symptoms score, with the exception of Positive symptoms at baseline, which were marginally higher in those who did not participate in the second assessment, t(87) = 1.97, p = .05.
Finally, comparisons were conducted to determine whether there were any diagnostic group differences in age, gender ratio, or an index of socioeconomic status (maternal and paternal education). These comparisons yielded no trends toward significant differences. For all of the analyses described below, there was no significant main or interactive effect of sex when it was included as an independent variable. Thus, we report here on the analyses without sex as a predictor.
Diagnostic Group Differences in Symptoms
Multivariate analysis of variance (MANOVA) was conducted to compare the diagnostic groups on the 3 symptom scores. This analytic procedure is most appropriate for testing group differences in sets of correlated dependent measures. Symptom scores, by diagnostic group, are illustrated in the bar graph in . As would be expected, there was a robust difference among the groups, F(Wilke's Lambda) (6, 168) = 20.30, p < .001. The univariate tests for positive symptoms, F(2, 86) = 31.52, p < .001, negative symptoms, F(2, 86) = 31.41, MSE < 12.559, p = .001, and disorganized symptoms, F(2, 86) = 62.72, p < .001, were all highly statistically significant. T-test comparisons among the pairs of groups revealed that the SPD group scored significantly higher than the NC and OD groups on all 3 symptom dimensions, with all p-values < .001. The OD group scored above the NC groups on all 3 symptom dimensions, with all p-values < .01.
Mean (SD error bars) Symptom Scores by Diagnostic Group.
Diagnostic Group Differences in Intellectual Performance
Wechsler intellectual subtest scores by diagnostic group are presented in . A MANOVA conducted on the scores yielded a significant effect of diagnostic group, F(Wilke's Lambda) (10, 162) = 2.35, p = .02. Univariate ANOVAs indicated diagnostic group differences in performance on the Arithmetic subtest, F(2, 85) = 4.35, p = .02, with the SPD group scoring below the NC group (p < .05), but no other significant group differences. The univariate tests revealed no significant differences among the groups in performance on the Vocabulary, Similarities, Picture Completion, or Block Design subtests.
Mean (SD error bars) Intellectual Subtest Scores by Diagnostic Group.
Wechsler Memory Scales
MANOVA revealed no significant difference among the diagnostic groups in performance on the WMS subtests; however, as illustrates, there was a consistent trend toward poorer performance by both the SPD and OD groups on many of the subtests. Comparison of the SPD and NC groups did show a significant deficit in the SPD group, t(48) = 1.81, p < .05, on delayed recall on the Family Pictures subtest. No other paired group comparison was statistically significant.
Mean (SD error bars) Wechsler Memory Scales Subtest Scores by Diagnostic Group.
The Relation Between Cognitive Performance and Baseline Symptom Ratings
The correlations among the cognitive and symptom scores for the entire sample, including the NC group, are presented in . As illustrated, the strongest relations are between the cognitive measures and the rating of negative symptoms, with poorer performance associated with more severe symptoms.
Correlations Among Cognitive Scores and Symptom Ratings
Regression analyses were conducted to examine the relation between cognitive scores and the symptom dimension scores (described above) for the SPD and OD groups. (The normal comparison sample was excluded from these analyses due to the low rate of symptoms in this group.) The 3 symptom scores (positive, negative, and disorganized) served as dependent variables in separate regression analyses, and the cognitive scores from each measure were the predictors, entered in a block. Analyses were first conducted for the combined samples, then separately for the SPD and OD groups.
For the combined groups, Wechsler Intelligence Scale subtest scores significantly predicted baseline ratings for disorganized symptoms, R = .45 (adjusted R2 = .14), F(5, 65) = 3.24, p <.01, and negative symptoms, R = .45 (adjusted R2 = .14), F(5, 65) = 3.21, p < .01. Significant individual predictors for negative symptoms were Picture Completion, t = –2.45, p < .02, and Arithmetic, t = –2.06, p < .05. For disorganized symptoms the Arithmetic subtest was significant, t = –3.57, p < .01.
Regression analyses for the separate groups showed that the Wechsler Intelligence Scale subtest scores significantly predicted negative symptoms at baseline for the SPD group, R = .71 (adjusted R2 = .35), F(5, 27) = 5.40, p < .001. Among the individual predictors, lower scores on the Picture Completion, t = –2.74, p < .01, and Similarities, t = –2.56, p < .01, subtests were associated with more severe negative symptoms. Subtest scores did not predict positive or disorganized symptoms for the SPD group, and none of the regression analyses were significant for the OD group alone.
For the analyses of WMS scores, the Family Pictures I and II scores were averaged and used as a single predictor, due to their high intercorrelation. The same set of regression analyses was then conducted with the scores from the WMS, and no significant relation was observed for the combined SPD and OD groups. For the SPD group, however, the memory scores were significant predictors of disorganized symptoms ratings, R = .56 (adjusted R2 = .28), F(4, 27) = 2.58, p < .05. Among the individual predictors, lower Letter-Number Sequencing subtest scores were associated with higher ratings for disorganized symptoms, t = –2.67, p < .02. Subtest scores did not predict positive or negative symptoms for the SPD group, and none of the regression analyses were significant for the OD group.
The Relation Between Cognitive Performance and Follow-Up Symptom Ratings
At this writing, all follow-up assessments have not been completed, so the findings of these analyses should be considered tentative. In testing the prediction of follow-up symptom scores, regression analyses were conducted with statistical control for baseline symptom ratings. Thus, the T1 symptom score was entered first, then the cognitive scores.
For the combined SPD and OD groups, regression analyses showed that Wechsler Intelligence Scale subtest scores were significant predictors of positive symptoms, R = .58 (adjusted R2 = .21), F(6, 33) = 2.81, p < .03, and negative symptoms, R = .63 (adjusted R2 = .29), F(6, 33) = 3.66, p < .01, but not disorganized symptoms. Marginally significant individual predictors were Similarities for negative symptoms, t = –2.02, p = .05, and Arithmetic for positive symptoms, t = –2.03, p = .05. Wechsler subtest performance did not, however, predict symptom scores when regression analyses were conducted on the separate groups. It should be noted that the statistical power for these analyses of follow-up data was low due to the small samples when the groups were analyzed separately.
Regression analyses using the WMS scores as predictors yielded no significant equations, indicating that performance on these measures did not predict follow-up symptoms when controlling for baseline levels.