Neurological Signs and Cognitive Performance Distinguish Between Adolescents With and Without Psychosis

After complete description of the study to the subjects and their legal guardians, we obtained written informed consent for study participation from the legal guardian and assent from the subjects in accordance with the policies and procedures of the Colorado Multiple Institutional Review Board.

The psychotic group (N=19) consisted of seven subjects with schizophrenia (one female), four subjects with schizoaffective disorder (one female), and eight subjects with bipolar disorder with psychotic features (four females). All subjects with bipolar disorder with psychotic features were euthymic (i.e., not in a manic, hypomanic, manic, or mixed episode and also not psychotic) at the time of study assessments. The mean age of onset for first psychiatric diagnosis in the psychotic group was 6.9 (SD=2.5) years (range=4 to 11 years) and included attention-deficit hyperactivity disorder (42.1%), schizophrenia (10.5%), bipolar disorder (type unspecified or “not otherwise specified,” 31.6%), “aggressiveness” (5.3%), and no formal diagnosis (10.5%). The mean age of onset for study diagnosis was 9.7 (SD=2.8) years (range=4 to 13 years).

Among the psychotic subjects, 31.6% had a history of maternal drug use during pregnancy, including exposure to stimulants (21.1%), alcohol (21.1%), and nicotine (5.3%), alone or in combination, and 16 of 19 subjects (84.2%) had a history of treatment with stimulant medications prior to or since the onset of study diagnosis. At the time of NES assessment, medications used among subjects with psychosis included antipsychotics (84.2%), lithium in any formulation (36.8%), anticonvulsant mood-stabilizers (42.1%), antidepressants (36.8%), stimulants (31.6%), and benzodiazepines (21.1%). At the time of IQ assessment, medications used among subjects with psychosis included antipsychotics (78.9%), lithium in any formulation (42.1%), anticonvulsant mood-stabilizers (47.4%), antidepressants (42.1%), stimulants (36.8%), and benzodiazepines (15.8%). Subjects with bipolar disorder with psychotic features differed from those with schizophrenia/schizoaffective disorder with respect to the frequency of treatment with lithium (75% versus 18%, Fisher’s exact p<0.03) and anticonvulsant mood-stabilizers (88% versus 18%, Fisher’s exact p<0.005).

There was no effect of diagnosis on Pubertal Development Scale (PDS) scores (F=0.6, df=2.32, p=0.57). Post-hoc comparison using Tukey’s Honest Significant Difference (HSD) test for unequal sample sizes demonstrated no differences in PDS scores between diagnostic subgroups.

Subtle neurological signs were significantly more common and more severe among adolescents with psychosis than among similarly aged healthy comparison subjects. The neurological domains demonstrating the widest divergence between the subjects with and without psychosis included primitive reflexes, complex motor sequencing, motor coordination, and short-term memory. Total NES scores did not distinguish between subjects with schizophrenia/schizoaffective disorder and bipolar disorder with psychotic features, but NES scores in both groups differed significantly from those of the comparison subjects. Additionally, adolescents with psychosis failed to demonstrate age-expected reductions of NES scores.

These findings suggest that neurological dysfunction as assessed by the NES is strongly associated with the presence of psychosis, regardless of whether that psychosis is persistent or mood episode-related. Although the possibility of medication effect on neurological signs in this population may be of immediate concern, the regression analyses performed here suggest that this is an unlikely explanation for this finding. This suggestion is supported by the lack of association between NES scores and antipsychotic medication use reported elsewhere. 2, 14 – 16

Similarly, IQ scores were lower among subjects with psychosis than among the comparison subjects. Although the mean full scale IQ for the psychosis group was within the normal ranges described for these measures, population-based studies demonstrate higher mean IQ levels (∼ 1 SD) in the Denver metropolitan region than in the Wechsler scale standardization sample. 17 The IQ scores of the psychotic adolescents in this study therefore reflect a greater departure from expected performance than might otherwise be apparent by these scores. Since deterioration of IQ in childhood-onset schizophrenia may begin up to 2 years prior to psychosis onset and stabilize within 2 years thereafter, 18 it is also possible that among the subjects with psychosis studied here, and particularly those with schizophrenia/schizoaffective disorder, IQ may not yet have reached its final (and possibly even lower) level.

In contrast to neurological signs, and similar to IQ findings among adults with these conditions, 19, 20 there was a significant effect of diagnosis on IQ: scores were lowest among subjects with schizophrenia/schizoaffective disorder, highest among the healthy comparison subjects, and intermediate among subjects with bipolar disorder with psychotic features. Simple logistic regression did not identify a predictive relationship of FSIQ, VIQ, or PIQ on psychotropic medication treatment status. Although it is possible that treatment history may have influenced the IQ differences between adolescents with schizophrenia/schizoaffective disorder and bipolar disorder with psychotic features observed here, data adequate to the task of examining this issue statistically were not available for analyses. However, Zammit et al., 21 in a population-based sample of more than 50,000 male subjects, identified lower IQ as a risk factor for schizophrenia but not bipolar disorder. This suggests that cognitive differences between these conditions antedate both their onset and treatment with psychotropic medications. This finding supports framing IQ differences between schizophrenia/schizoaffective disorder and bipolar disorder with psychotic features subjects observed in the present study as reflective of differences in neurodevelopment between these disorders rather than as a consequence of them or their treatments.

Collectively, the present findings suggest that NES and IQ reflect dependent domains in which aberrant neurodevelopment occurs among adolescents with psychosis. The utility of including neurological and cognitive assessments in the evaluation of adolescents with these conditions is evidenced by the robust prediction of group membership (psychotic versus nonpsychotic) using the combination of total NES and FSIQ scores: the discrimination between schizophrenia/schizoaffective disorder and never mentally ill subjects was excellent, with no member of either group misclassified as a member of the other.

The present findings also suggest that diagnosis (i.e., disorders with persistent psychosis versus disorders with mood episode-related psychosis) may differentially influence development in these domains of neuropsychiatric function. If similar findings are observed in a larger sample of adolescent subjects with schizophrenia/schizoaffective disorder and bipolar disorder with psychotic features, inclusion of neurological and cognitive examinations in the clinical assessment of adolescents with psychotic disorders might improve diagnostic accuracy. However, this suggestion is necessarily speculative, as the present data indicate that NES and IQ alone are not sufficient to the task of discriminating fully between bipolar disorder with psychotic features and the other diagnostic groups included herein. The single subject in each of those two groups that was diagnostically misclassified by the model was predicted to be a member of the bipolar disorder with psychotic features group. Additionally, three of the eight subjects in the bipolar disorder with psychotic features group were diagnostically misclassified, with two identified as members of the schizophrenia/schizoaffective disorder group and one identified as never mentally ill. These observations suggest that inclusion of additional factors in the model is needed to separate the bipolar disorder with psychotic features subjects from those with schizophrenia/schizoaffective disorder and healthy comparison subjects. Alternatively, if NES and IQ assessments are more strongly related to the neurobiology of these conditions than are phenomenological psychiatric diagnostic assessments, then cross-sectional K-SADS-PL and clinical interview-based diagnoses may not be the optimal method by which to discriminate between adolescents with these diagnoses. 22 – 24

Several limitations of the present study require consideration. It is possible that examiner knowledge of subject diagnosis may contribute to NES scores, and that examiner-subject interactions may influence the performance of subjects on both neurological and cognitive assessment measures. To the latter point, the blinded cognitive assessment by a trained technician is consistent with the methodology of other studies of this population and should minimize the potential effect of clinician bias on subject performance. The structured administration and anchored scoring methods of the NES limit the likelihood of clinician influence on subject performance, although this concern cannot be assuaged entirely. However, the magnitude of the NES differences between adolescents with and without psychosis observed here, as well as the comparability of total NES scores among subjects with and without psychosis to those reported in previous studies, 2, 3 argue against evaluator bias as an explanation for the present findings.

The present findings also argue against psychotropic medication administration as an explanation for differences in NES and IQ scores between subjects with and without psychosis, as well as for differences in IQ scores between the psychotic subgroups. However, the absence of treatment-naïve subjects precludes a more definitive assessment of this issue.

The subject groups were not identical with respect to gender, particularly for diagnostic subgroup comparisons. Though there is no a priori rationale for suggesting an effect of gender on NES score, 2, 14, 25 careful matching of subjects is needed in future studies in order to more fully evaluate the effect, if any, of gender in this context.

The size of the present sample precludes evaluation of a larger number of predictor variables, including medications (individually or by class), gender, and socioeconomic status, and also reduces the power needed to investigate significant predictor variables with smaller effect sizes. Similarly, the nature of the cognitive assessments employed in the present study precludes evaluation of the relationships, if any, between specific domains of cognition and neurological dysfunction. Further studies of a larger sample of adolescents with psychotic disorders inclusive of these variables, including longitudinal studies with periodic diagnostic reassessment, symptom severity assessments, and structural or functional neuroimaging using magnetic resonance imaging, neurophysiological assessments (e.g., electroencephalography, evoked or event-related potentials, and/or magnetoencephalography) are needed to address these issues.

This work was supported by a grant from NIMH (R01 MH63442) and a grant from the Developmental Psychobiology Research Group (DPRG) of the University of Colorado Health Sciences Center, Denver, Colorado.

Received February 1, 2006; revised May 13, 2006; accepted August 21, 2006. From the University of Colorado School of Medicine, Denver, Colorado. Address correspondence to Dr. Arciniegas, Neuropsychiatry Service, University of Colorado School of Medicine, Campus Box C268-25, 4200 East Ninth Avenue, Denver, CO 80262; [email protected] (e-mail)