Neuropsychological Impairments in Schizophrenia and Psychotic Bipolar Disorder: Findings from the Bipolar-Schizophrenia Network on Intermediate Phenotypes (B-SNIP) Study

Patients were referred by mental health providers or recruited through advertisements and talks at community organizations and support groups. Patients with a history of psychotic symptoms were recruited if they had at least one available first-degree relative 15–65 years of age willing to participate in the study. Probands were required to have a DSM-IV diagnosis of schizophrenia, schizoaffective disorder, or bipolar disorder with a history of psychosis, determined at consensus diagnostic meetings after review of data gathered using the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID) (37), information about the proband’s medical and psychiatric history obtained from relatives, and available medical charts. Clinical symptom ratings were assessed using the Positive and Negative Syndrome Scale (PANSS) (38), the Montgomery-Åsberg Depression Rating Scale (MADRS) (39), and the Young Mania Rating Scale (40), and functional status was assessed with the Social Functioning Scale (41). The Schizo-Bipolar Scale (18), which assesses a dimension of illness from prototypical bipolar disorder to prototypical schizophrenia, was also completed. To maintain reliability in ratings, diagnosticians from all sites underwent initial training and reviewed cases during monthly cross-site diagnostic meetings.

First-degree relatives were assessed with the SCID and the Structured Interview for DSM-IV Personality (SIDP-IV) (42). As done in previous work (43), our a priori plan was to cast a broad net for ascertaining the presence of relevant elevated personality features by identifying individuals within one criterion of a cluster A (odd or eccentric) or cluster B (dramatic, emotional, or erratic) diagnosis. Given this project’s primary focus on identifying psychosis-related features, the psychotic-like features that characterize cluster A personality disorders were the primary traits of interest. Secondarily, cluster B traits were evaluated because of their potential overlap with bipolar symptoms.

Healthy volunteers were recruited through print and electronic media and research registries. Healthy comparison subjects were required to have no personal history of a psychotic disorder or recurrent depression (based on the SCID and consensus review) and no known immediate family history of these disorders.

All participants had no history of seizures or head injury with loss of consciousness >10 minutes; had a negative urine drug screen for common drugs of abuse on the day of testing; had no diagnosis of substance abuse in the past 30 days or substance dependence in the past 6 months; had no change in medication (and were generally clinically stable over the past month); had no history of systemic medical or neurological disorder likely to affect cognitive abilities; had an age-corrected Wide-Range Achievement Test, 4th edition, reading test standard score >65; and were sufficiently fluent in English to complete neuropsychological testing.

The demographic and clinical characteristics of probands and first-degree relatives are summarized in Tables 1 and 2. To address statistically significant group differences for sex distribution and age in proband and relative groups, age- and sex-stratified normative data (44) were used to compute subtest and composite scores for each participant on the BACS. Subtest and composite scores were also computed based on age- and sex-stratified data from the present comparison group. Composite scores anchored to B-SNIP controls and Keefe norms (44) were highly correlated (r values, >0.98). To be consistent with the existing literature and to facilitate direct comparison with previous studies, performance was referenced to published normative data (44). To address the uneven distribution of race among groups, race was used as a covariate in all analyses. All major findings were unchanged after including race as a covariate. Estimated marginal means are presented in the figures.

Antipsychotic dosage in chlorpromazine equivalents (45), benztropine (anticholinergic) dosage, and the presence or absence of current antipsychotics, mood stabilizers, and antidepressants were minimally related to composite scores on the BACS across all proband groups (r values, <0.21; see Table 1 and the data supplement that accompanies the online edition of this article). A previous history of substance abuse or dependence was also minimally related to performance on the BACS (r values, <0.18). Site effects were nonsignificant. Clinical ratings of psychosis and mood symptoms had minimal associations with data from the BACS (r values, <−0.19). As each of these parameters accounted for less than 5% of the variance in data from the BACS, they were not used as covariates in the analyses reported below. (Additional details about the study sample and treatment history are available in the data supplement.)

An analysis of variance (ANOVA) was used to compare cognitive performance in schizophrenia and bipolar probands, the two primary and larger proband groups, with performance in healthy comparison subjects. Simple contrasts were used to clarify significant findings in omnibus testing using a Hochberg correction for multiple comparisons (49). Profile analyses were conducted using a repeated-measures multivariate analysis of variance (MANOVA), with the group-by-subtest interaction being the key statistical test.

Schizoaffective manic and schizoaffective depressed probands were considered together with schizophrenia and bipolar probands in an ANOVA. Next, the association between performance on the BACS and Schizo-Bipolar Scale score was examined in the combined proband sample.

A heritability analysis to estimate familiality of cognitive function was performed using the SOLAR (Sequential Oligogenic Linkage Analysis Routine) software package (50). In a design such as ours, an estimate of familiality (h²) represents the portion of phenotypic variance accounted for by family membership. To test for the significance of familiality, a maximum likelihood ratio test compared a model in which phenotypic variation is explained by family membership to a model assuming that no variation is explained by familial factors. A correction was applied to account for ascertainment bias, since families were recruited through the identification of a psychotic proband and not a representative community sample (51). Because of the larger sample sizes and the primary focus on capturing both the traditional diagnostic dichotomy of primary interest and the prototypical domains anchoring a mood-psychosis dimension (52), familiality estimates and group comparisons among relatives were restricted to schizophrenia and bipolar disorder.

An ANOVA was used to compare the cognitive performance of relatives of probands in the two primary groups with that of healthy comparison subjects before taking into consideration elevated personality traits among relatives. Simple contrasts were used to clarify significant findings in omnibus testing, using a Hochberg correction for multiple comparisons (49).

Among relatives of schizophrenia and bipolar probands with no history of psychosis, a two-way ANOVA was used to assess the relationship of proband diagnosis and cluster A personality traits with cognitive dysfunction. Because psychosis-related traits were of primary interest, relatives who met criteria for cluster A traits were included in the cluster A group regardless of their cluster B traits. In secondary analyses, composite scores on the BACS from individuals with cluster A traits were compared with scores from individuals with cluster B traits within and across diagnostic category. This analysis was repeated after excluding those who met criteria for both cluster A and cluster B traits (nine relatives of schizophrenia probands and six relatives of bipolar probands), with no change in the findings. Finally, nonpsychotic relatives with neither cluster A nor cluster B traits were compared with healthy comparison subjects and relatives with elevated personality traits of interest.

Repeated-measures MANOVA testing for profile differences indicated a significant group-by-subtest interaction (F=9.69, df=10, 1580, p<0.001). However, when the comparison group was excluded, the group-by-subtest interaction for proband groups was not significant, indicating that their pattern of performance did not differ across cognitive domains (Figure 2).

Schizoaffective probands were significantly less impaired than schizophrenia probands (F=4.51, df=1, 456, p=0.03) and were more impaired than bipolar probands (F=12.97, df=1, 390, p=0.008). Differences between the schizoaffective subtypes (depressed or manic) were not significant.

To characterize neurocognitive function along a bipolar-schizophrenia dimension, composite scores from the BACS were examined in relation to scores on the Schizo-Bipolar Scale (18). As illustrated in Figure 3, overall cognitive performance declined as affective features became less prominent and the persistence of psychosis became more prominent (r=−0.25, p<0.001). The progressive reduction in cognitive scores from prototypical bipolar disorder to prototypical schizophrenia and the intermediate scores of the schizoaffective case subjects support a continuum model of cognitive deficits from schizophrenia to bipolar disorder.

There were no significant differences in composite scores on the BACS for first-degree relatives with a history of psychosis compared with their respective probands. Only relatives with no history of psychotic symptoms were included in the following analyses. There was a significant group difference in composite scores on the BACS among the relative groups and the healthy comparison group (F=7.02, df=2, 800, p=0.001). Simple contrasts indicated that relatives of schizophrenia probands had significant impairments compared with healthy comparison subjects, but there was no significant difference between relatives of psychotic bipolar probands and healthy comparison subjects (Figure 3). As with probands, the relative groups did not differ in their pattern of neuropsychological strengths and weaknesses (see the online data supplement).

When compared with healthy comparison subjects, relatives exhibited significant impairments on the BACS composite scores when either cluster A or cluster B features were present. Among those who had elevated axis II traits (either cluster A or cluster B), there was no difference on the composite score between relatives of schizophrenia probands and relatives of bipolar probands. As illustrated in Figure 4, the interaction between axis II status (those with elevated cluster A or B traits versus those with neither traits) and proband diagnosis (schizophrenia versus bipolar) was significant (F=4.05, df=1, 505, p=0.05). This interaction was characterized by significant cognitive deficits in relatives of schizophrenia probands with no personal history of psychosis, even when cluster A and cluster B traits were present (F=8.68, df=1, 507, p=0.003). In contrast, in relatives of bipolar probands, cognitive deficits were observed when either elevated cluster A or cluster B traits were present, but not when axis II traits were nominal. This pattern could not be attributed to differential rates of axis I disorders because rates of disorders such as depression and anxiety disorders were similar in these relative groups. (Correlational analyses were also used to examine personality traits as continuous variables [number of criteria for any cluster A or B disorder]; these findings are presented in the online data supplement.)

Among relatives of both schizophrenia and bipolar probands, those with cluster A traits exhibited cognitive impairments similar to those seen in relatives with cluster B traits. This pattern parallels previous findings of medium to large effect sizes for cognitive deficits across a range of cluster A and B disorders, including schizotypal (61), antisocial (62), and borderline (63, 64) personality disorders. The main factor leading to greater overall cognitive deficits in relatives of schizophrenia probands than in relatives of bipolar probands was that relatives of schizophrenia probands as a group demonstrated cognitive deficits regardless of whether cluster A or B personality traits were present, while relatives of bipolar probands did not. This pattern suggests that cognitive deficits in families with a schizophrenia proband are transmitted at least partially independently from factors associated with schizotypal and other personality disorder traits, while in relatives of probands with bipolar disorder, cognitive deficits are more closely linked with elevated cluster A or cluster B personality traits. This may point to broader qualitative differences in the selectivity or penetrance of familial risk mechanisms affecting cognition across psychotic disorders.

The observation of significant familiality of cognitive function in schizophrenia is consistent with several previous studies (24, 28, 29, 33, 34), and our findings provide new evidence for a similar pattern in bipolar disorder with a history of psychosis. First, it is noteworthy that familiality estimates were somewhat larger for single word reading than for scores on the BACS. This may reflect a more substantial shared family environment component in which some parents provide a more enriching educational environment. Second, familiality estimates for individual neuropsychological subtests were variable and were generally lower than those observed for the BACS composite score. Individual tests focus more narrowly on specific cognitive processes, while composite scores integrate several cognitive skills regardless of their specific nature. As reflected in the history of intelligence test development, composite measures are often more closely related to functional ability in day-to-day life. Some researchers have suggested that genes with a broad impact on brain development and function may have a general impact on cognition (65, 66). It remains unclear whether examining aggregate effects of multiple genetic factors that affect scores (67) or individual genetic effects on specific cognitive functions will best advance gene discovery in psychotic disorders. Both approaches have potential, as particular genes may be tightly linked to a specific cognitive deficit while global deficits may represent a final common pathway of multiple factors that can be understood using systems biology approaches. The comparative advantages of these approaches remain to be clarified, but our findings in this study support the view that measures of generalized cognitive deficits can be useful for understanding familial factors contributing to bipolar disorder with psychosis (13, 57) as well as schizophrenia (6, 68). Based on the interaction between axis II status and cognitive performance, in which only bipolar relatives with elevated cluster A or cluster B traits exhibited cognitive deficits, a potentially promising phenotyping strategy for tracking risk mechanisms in bipolar disorder might lie in evaluating the co-occurrence of personality traits and cognitive dysfunction.