An endophenotype describes a trait marker that is directly associated with underlying vulnerability yet distinct from the disease phenotype itself (1). The identification of an endophenotype for bipolar disorder might facilitate early detection and treatment and provide a marker for genetic research. Bipolar disorder is highly heritable; family members of patients with bipolar disorder are at high risk of bipolar disorder and other affective disorders (2). Therefore, a useful endophenotype could be expressed in this population.

Within the neuropsychological domain, deficits in cognitive flexibility and verbal learning have been described in bipolar disorder (3) and unipolar depression (4). In bipolar disorder, these deficits may be trait-related and indicate functional neuropathology of the frontal and temporal lobe, respectively (3). We have examined these two cognitive domains in relatives of patients with bipolar disorder to investigate their potential as cognitive endophenotypes. Cognitive flexibility was assessed with the intradimensional/extradimensional shift task (5), in which a series of stages assess different components of attentional flexibility, including reversal learning, set formation, and extradimensional ("set") shifting. The task was developed from the "gold-standard" Wisconsin Card Sorting Test of frontal lobe integrity (6). Mnemonic function was measured with the California Verbal Learning Test (7), in which learning, recall, and recognition of a 16-item shopping list are assessed. In addition to healthy comparison subjects, euthymic patients with recurrent major depression formed a clinical comparison group, given the frequency of unipolar mood disorder in bipolar families.

Twenty-seven first-degree relatives of patients with bipolar I disorder, 15 euthymic patients with unipolar depression, and 47 healthy comparison subjects with no family history of psychiatric illness were given the California Verbal Learning Test (7) and the intradimensional/extradimensional shift task (www.camcog.com). The groups did not differ significantly in age, education, or gender, although the proportion of female participants was greater in the group of euthymic patients with unipolar depression (t1). Relatives of patients with bipolar I disorder were contacted through 18 patients with DSM-IV-diagnosed bipolar I disorder. They were 10 parents, 12 siblings, and five children (age range=17–68 years).

Previous psychiatric hospitalization or a current DSM-IV mood episode were exclusion criteria for relatives of patients with bipolar I disorder, but several subjects reported previous affective disorders: major depressive episode (five subjects), bipolar II disorder (one subject), hypomania (one subject), and subclinical depression (four subjects). Sixteen relatives reported no psychiatric history.

Euthymia was defined as a Hamilton Depression Rating Scale (8) score <9 and Young Mania Rating Scale (9) score <9 for the week before testing. Six euthymic patients with unipolar depression were receiving antidepressant medication (selective serotonin reuptake inhibitors [SSRI] [N=2], tricyclics [N=3], SSRI/serotonin-norepinephrine reuptake inhibitor combination [N=1]), and seven euthymic patients had a family history of affective disorder. After complete description of the study, written informed consent was obtained from all participating subjects.

In the California Verbal Learning Test, a 16-item shopping list is read to the subject over five learning trials. Immediate free recall, delayed (20 minutes) free recall, and recognition are assessed. Data were analyzed with repeated-measures analysis of variance (ANOVA) with learning trial as a five-level within-subjects factor. A series of one-way ANOVAs were used to test group differences in immediate recall, delayed recall, and recognition. Information was missing on five relatives of patients with bipolar I disorder because of protocol amendment.

The intradimensional/extradimensional shift task is a nine-stage visual discrimination task using multidimensional stimuli. At each stage, the subject must reach a criterion of six consecutive correct responses within 50 trials to progress to the next stage (else the task is aborted). Completion rates (chi-square test) and errors to criterion (one-way ANOVAs) were used in the data analysis. Error rates were square-root transformed for parametric analysis and adjusted for subjects failing to complete the task. Exclusion of noncompleters is overly conservative and reduces degrees of freedom at later stages. Noncompleters were assigned 25 errors to subsequent conceptual stages (intradimensional or extradimensional shifting), and their average number of reversal errors was used for subsequent reversal stages.

California Verbal Learning Test

Repeated-measures ANOVA revealed a highly significant effect of trial over the five learning blocks (F=184.2, df=4, 324, p<0.0001) but no effect of group (F=1.41, df=2, 81, p=0.25) or group-by-time interaction (F=0.96, df=8, 234, p<0.47). One-way ANOVAs also indicated no significant group effects on immediate or delayed recall or recognition (t2). In the group of relatives of patients with bipolar I disorder, higher scores on the Hamilton depression scale were associated with lower California Verbal Learning Test learning scores (trials 1–5 total) (r=–0.46, df=22, p=0.03). The association with Hamilton depression scale scores was not significant in the group of euthymic patients with unipolar depression (r=–0.35, df=15, p<0.21), although these correlations did not differ significantly between the two groups (Fisher’s test, z=–0.391, p>0.05).

Intradimensional/Extradimensional Shift Task

Six of 27 relatives of patients with bipolar I disorder failed to complete the task (two intradimensional shifting, two extradimensional shifting, two extradimensional reversal), and four of 15 euthymic patients failed to complete the task (one compound discrimination, one intradimensional reversal, two extradimensional shifting). All comparison subjects completed the task. Task completion rates differed significantly between groups (χ²=12.8, df=2, p=0.002), and at the extradimensional shifting stage specifically (χ²=6.23, df=2, p<0.05). Analysis of error rates confirmed a significant group difference in total task errors, which was also significant at the extradimensional shifting stage (t2). The main effects of group remained significant after we covaried for gender (total errors F=4.01, df=2, 85, p=0.02, and extradimensional shifting errors F=5.99, df=2, 85, p=0.004). There were no group differences in reversal errors or intradimensional shifting errors. Post hoc comparisons using Tukey’s honestly significant difference indicated that relatives of patients with bipolar I disorder made more extradimensional shifting errors than comparison subjects (p=0.02) and euthymic patients made more extradimensional shifting errors than comparison subjects (p<0.03). The subset of relatives of patients with bipolar I disorder without a personal history of affective disturbance (N=16) also showed more extradimensional shifting errors than comparison subjects (the relatives’ mean=6.8, SD=9.5) (t=2.55, df=61, p=0.01). There was no difference in the euthymic patients receiving (N=6) or not receiving (N=9) antidepressant medication (t=0.61, df=13, p<0.56). Total errors did not correlate with Hamilton depression scale scores in the relatives of patients with bipolar I disorder (r=0.18, df=27, p<0.37) or euthymic patients (r=0.12, df=15, p=0.68).

A neuropsychological endophenotype for a disorder should be associated with trait dysfunction in affected probands and be present in unaffected first-degree relatives of probands (1). In the present study, both relatives of patients with bipolar I disorder and euthymic patients with unipolar depression were significantly impaired on the intradimensional/extradimensional shift task, making more errors at the extradimensional shifting stage specifically. Performance in the two groups resembled that of euthymic patients with bipolar disorder in our previous research (10) (t2). These data identify impaired attentional shifting as a putative endophenotype for mood disorder. Set-shifting deficits were previously reported in mania (3) and unipolar depression (4). The deficit may be associated with a functional neuropathology in the lateral prefrontal cortex. The extradimensional shifting stage is selectively impaired by lateral prefrontal cortex lesions (5), and this region shows neuropathological abnormalities in bipolar disorder (3).

These relatives and patients performed at the level of comparison subjects on the California Verbal Learning Test. Verbal learning may represent a sensitive state marker in bipolar disorder rather than a trait marker (10). This interpretation is supported by these data, where California Verbal Learning Test performance correlated with Hamilton depression scale ratings in the relatives. In a previous study that reported California Verbal Learning Test deficits in patients with remitted unipolar depression (11), Hamilton depression scale scores remained mildly inflated (mean=5.9). The present data show that verbal learning performance may be unimpaired in an asymptomatic group of patients with remitted unipolar depression.

Investigation of first-degree relatives may be limited by several factors. Exclusion of relatives with any psychiatric history could result in selection of a supernormal group. We included relatives with a history of minor affective disturbance to maximize sensitivity, although these subjects were not disproportionately impaired and the extradimensional shifting impairment remained significant in the relatives with no personal history. The level of genetic vulnerability may also vary across studies. Investigation of obligate carriers (unaffected relatives with both an affected child and an affected parent) and families with multiple probands would increase sensitivity to a cognitive endophenotype in further research.

Received June 4, 2004; revisions received Nov. 4, 2004, and Jan. 6, 2005; accepted Feb. 1, 2005. From the Department of Experimental Psychology, University of Cambridge; and the Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford, U.K. Address correspondence and reprint requests to Dr. Clark, Department of Experimental Psychology, University of Cambridge, Downing St., Cambridge CB2 3EB, U.K.; lc260@cam.ac.uk (e-mail). Funded by a Medical Research Council (U.K.) studentship (Dr. Clark). The authors thank Dr. Zubin Bhagwagar for patient referral and the subjects and their families for participation.