Symptomatic overlap between affective disorders and schizophrenia has long been noted. Even the psychotic, or "positive," symptoms most closely identified with schizophrenia—Schneider’s first-rank symptoms—also occur in patients with affective disorder +(1). This overlap and other considerations have led some investigators to challenge the dichotomous Kraepelinian classification of the psychoses +(2, +3). Recently, overlap has also been noted in the findings from genetic linkage studies of bipolar disorder and schizophrenia. Four chromosomal regions that could potentially harbor susceptibility genes for both disorders have been identified: 10p12–13, 13q32, 18p11.2, and 22q11–13 (reviewed in reference +4). In spite of the limitation imposed by the generally low resolution of linkage data, this overlap has renewed interest in the hypothesis that there are susceptibility genes that contribute to the psychotic manifestations of both bipolar disorder and schizophrenia.

If psychosis genes exist, they may lead to familial aggregation of psychotic symptoms in subjects with familial bipolar disorder. There are suggestions from prior studies that psychotic symptoms cluster in some families with affective disorder, although the evidence to date has been limited. To our knowledge, only one prior study +(5) has specifically addressed the hypothesis that psychotic bipolar disorder aggregates in families; the finding was negative in a small study group. In another study +(6), in which bipolar disorder and major depression were grouped together, relatives of probands with psychotic affective disorder had a higher rate of psychotic affective disorder than did relatives of nonpsychotic probands, although the difference was not statistically significant. Two studies +(7, +8) have provided evidence of familial aggregation of psychotic symptoms in major depression, although in only one of these +(7) did the finding reach significance. A follow-up study +(9) failed to replicate this significant finding.

We have previously used clinical features in an effort to define a more genetically homogenous group of bipolar disorder families. Bipolar II disorder +(10) and comorbid panic disorder +(11) have been investigated within our group and have been found to enhance evidence for linkage to chromosome 18q. Here we assess the familial aggregation of psychotic symptoms within bipolar disorder pedigrees to determine whether this feature may increase homogeneity and inform the discussion of possible overlap in genetic susceptibility to bipolar disorder and schizophrenia.

Aspects of family ascertainment have been described in detail previously +(12). Probands were selected from inpatients and outpatients screened in Baltimore and Iowa City and from individuals who contacted us to volunteer. The ascertainment criteria required a treated proband with bipolar I disorder who had two first-degree relatives who were affected with bipolar I disorder, bipolar II disorder with at least two major depressive episodes, recurrent unipolar depression, or schizoaffective disorder, manic type. In addition, the families had to be unilineal: only one parent could be affected and the unaffected parent could not have any affected first-degree relatives. Five families included in the analysis had only two affected members each. They were ascertained and studied because they had initially appeared to meet the inclusion criterion of three affected members according to family history. Because many useful clinical and genotype data on these families had already been obtained, we included them in the genome-wide scan analysis under way by our group. The 65 families in the genome-wide scan provide the basis for the analyses in this paper.

All available first- and second-degree relatives 14 years old or older, as well as probands themselves, were interviewed by psychiatrists using the Schedule for Affective Disorders and Schizophrenia—Lifetime Version (SADS-L) +(13). After complete description of the study to the subjects, written informed consent was obtained. Collateral information from multiple family informants was obtained, and medical records were added whenever possible. Diagnoses of affective disorders were made independently by two psychiatrists using a best-estimate procedure and the Research Diagnostic Criteria (RDC) +(14).

First-degree relatives with diagnoses of major affective disorder—bipolar I disorder, bipolar II disorder with at least two major depressive episodes, schizoaffective disorder, manic type, and recurrent unipolar depression—were analyzed in this study. Schizoaffective disorder, manic type, and recurrent unipolar depression were included in addition to bipolar disorder because of genetic epidemiological evidence suggesting that liability to these disorders may be higher than normal in both familial schizophrenia and familial bipolar disorder (see Discussion). In addition, analyses using only relatives with bipolar I disorder were performed. No subjects in the data set were diagnosed with schizophrenia or schizoaffective disorder, depressed type. The presence of psychotic symptoms—defined as hallucinations or delusions (mood-congruent and/or mood-incongruent)—during mania and during depression were assessed by using the SADS-L interview, which elicits information regarding the most severe manic and most severe depressive episode. A total of 45 probands and 52 first-degree relatives were identified as psychotic on the basis of the SADS-L interview alone. Data from family informants and medical records were reviewed to increase the sensitivity of detection of psychotic symptoms. This resulted in the identification of an additional two probands and three relatives as psychotic.

Pearson’s chi-square analysis, Fisher’s exact test, and Student’s t test were used to test differences between the psychotic and nonpsychotic proband groups on diagnostic, demographic, and clinical variables. Three tests of the hypothesis of familial aggregation of psychotic symptoms were performed. The first was a comparison of the proportion of psychotic probands having at least one psychotic relative with affective disorder to the proportion of nonpsychotic probands with at least one such relative; the chi-square test was employed for this comparison. In the second analysis, the odds of being psychotic for first-degree affectively ill relatives of psychotic probands were compared to the odds for those relatives of nonpsychotic probands. This was implemented by using the generalized estimating equation approach +(15), which uses logistic regression but has the added advantage of taking into account potential correlation between observations when multiple members of the same family are considered. In these analyses we controlled for sex, age at interview, and duration of illness. A third analysis of familial clustering of psychotic symptoms was performed by testing the distribution of psychosis among the affectively ill subjects (including probands) across all 65 families. The null hypothesis of no familial aggregation of psychotic symptoms was tested by using Tarone’s one-sided score test for binomial distributions +(16). A low and nonsignificant score indicates the expected random distribution of a variable, in this case psychosis, whereas a high and significant score suggests that the variable clusters more in some families than in others. We repeated these three analyses using only bipolar I disorder as the affected phenotype.

The subjects included 65 bipolar I disorder probands and 202 affected first-degree relatives. The diagnoses for the affected relatives were as follows: bipolar I disorder, N=60; schizoaffective disorder, manic type, N=6; bipolar II disorder, N=77; and recurrent unipolar depression, N=59. Among the subjects with bipolar I disorder, 71.2% (89 of 125) had psychosis, whereas the figures were 3.9% (three of 77) for bipolar II disorder and 6.8% (four of 59) for recurrent unipolar depression. By definition, all six subjects with schizoaffective disorder, manic type, had psychosis. There were 47 probands with psychotic symptoms and 18 probands with no psychotic symptoms. The psychotic probands had 146 first-degree relatives with affective disorder, and the nonpsychotic probands had 56 such relatives. The two sets of families did not substantially differ in number of first-degree relatives interviewed per family, number of affected first-degree relatives per family, or age of relatives at the time of interview (+Table 1). The proportion of relatives with affective disorder whose specific diagnosis was bipolar I disorder was higher in the families of psychotic probands than in the families of nonpsychotic probands: 50 (34.2%) of 146 versus 10 (17.9%) of 56 (χ²=5.21, df=1, p=0.02). The affectively ill relatives of the psychotic probands included six individuals diagnosed with schizoaffective disorder, manic type, while no one with this diagnosis was found among the relatives of nonpsychotic probands; this difference was not significant (six of 146 versus none of 56; Fisher’s exact test, p=0.15). The affected relatives of the psychotic probands had significantly more hospitalizations (mean=2.23, SD=4.86) than did the affected relatives of the nonpsychotic probands (mean=0.81, SD=2.01) (t=2.91, df=201, p=0.004). No statistically significant differences were found between affected members of the families of the psychotic and nonpsychotic probands in the following severity indicators: age at onset of affective disorder, duration of longest mood disorder episode, rate of attempted suicide, and rates of alcoholism and substance abuse.

Among the families of the psychotic probands, 63.8% contained at least one first-degree relative with psychotic affective disorder (+Table 1). Among the families of the nonpsychotic probands, the proportion with psychotic affective disorder in at least one relative was significantly lower, 27.8%. We also did the analysis by considering only subjects with bipolar I disorder as affected. Of the families with psychotic probands, 48.9% contained at least one first-degree relative with psychotic bipolar I disorder, compared to 22.2% of the families of nonpsychotic probands (+Table 1).

Among all first-degree relatives with major affective disorder, 27.2% (55 of 202) had a lifetime history of psychotic symptoms during an affective episode. In the families of psychotic probands, 33.6% (49 of 146) of the relatives with major affective disorder had psychotic symptoms, compared to only 10.7% (six of 56) of those in nonpsychotic families; the difference was statistically significant (+Table 2). Psychosis was a feature of 74.0% (37 of 50) of the relatives with bipolar I disorder in the families of the psychotic probands and of 50.0% (five of 10) of the relatives with bipolar I disorder in the families of the nonpsychotic probands. This difference was not statistically significant.

Of all probands and their first-degree affected relatives, 38.2% (102 of 267) were psychotic. Analysis of familial clustering was performed to determine whether the distribution of psychotic individuals was uniform throughout the 65 families (the null hypothesis) or, conversely, whether these individuals clustered in some families but not in others. The results indicated significant clustering (+Table 3). A subset of 31 families were found to contain two or more psychotic affected subjects; 30 of these families had a psychotic proband. Ten families with three or more psychotic affected members were identified; all of these included a psychotic proband. When we considered only subjects with bipolar I disorder as affected, the evidence for the clustering of psychotic disorder was stronger than in the analysis of all subjects with major affective disorder (+Table 4).

We have found evidence for the familial aggregation of psychotic symptoms, defined as hallucinations or delusions, in bipolar disorder pedigrees. Four analyses support this: 1) more families with psychotic probands than with nonpsychotic probands had at least one first-degree relative with psychotic affective disorder; 2) among the first-degree affectively ill relatives of psychotic probands, there was a higher rate of psychosis than among comparable relatives of nonpsychotic probands; 3) psychotic affective disorder clustered disproportionately in a subset of families; and 4) the clustering of psychosis was also apparent when only subjects with bipolar I disorder were considered as affected.

A strength of this study is that all of the diagnostic interviews were done by psychiatrists, thus maximizing the chance of eliciting valid reports of psychotic symptoms. In addition, the reliability of the assessment of hallucinations and delusions with the SADS-L has previously been demonstrated to be excellent +(13). The test-retest reliability for the diagnosis of psychotic bipolar disorder was also found to be good to excellent in another study using the SADS-L +(17). Even so, the SADS-L assesses only the most severe manic and depressive episodes; it does not require the interviewer to ask about each lifetime episode, so that a psychotic affective episode could be missed.

This study used a narrow definition of psychosis. We did not assess formal thought disorder or disorganized behavior even though these features are often considered psychotic. Although formal thought disorder is assessed by the SADS-L, it has been shown to have lower reliability than hallucinations and delusions +(13). Prior analyses of the structure of psychosis +(18) have suggested that hallucinations and delusions may be independent of thought disorder and disorganization, however, so that our narrow definition of psychosis may be helpful in our pursuit of a genetically more homogeneous subtype of bipolar disorder.

We considered whether hallucinations and delusions were just one aspect of a broader severity that was familial. No statistically significant differences in most severity indicators were found between relatives in the families of psychotic and nonpsychotic probands. However, two significant differences were found: an excess of bipolar I disorder subjects and more hospitalizations in the families of psychotic probands. The higher rate of hospitalization is not surprising given that psychotic symptoms generally lead to impaired judgment and thus to potential dangerousness. The finding that a higher rate of psychosis is associated with a higher rate of bipolar I disorder is not unexpected given that mania is associated with a substantially higher rate of psychosis than is depression +(19). Among the relatives with psychotic bipolar I disorder, all but two experienced psychosis during mania (the two exceptions were psychotic only during depression). In order to assess whether the clustering of psychosis in families was independent of the clustering of bipolar I disorder in families, we did separate analyses in which we considered only bipolar I disorder as the affected phenotype. Analysis by family and analysis by individual relative yielded differences in the rate of psychosis between proband groups that were not statistically significant. Application of the third analytic method to the phenotype of bipolar I disorder only, however, resulted in a highly significant clustering of psychosis; this latter method had more power than the first two to detect familial aggregation of psychotic symptoms because, by including probands, it took into account more than double the number of subjects with bipolar I disorder.

These results are consistent with some +(6–+8), but not all +(5, +9), previous findings on familial aggregation of psychotic affective disorder. One study, by Winokur et al. +(5), specifically had a negative result in a study group restricted to bipolar disorder. One explanation for the difference between that result and ours may lie in the different inclusion criteria used: the Winokur et al. study used only relatives with psychiatric records (N=43), whereas we assessed all available relatives. Our study group therefore included 108 relatives without psychiatric records among 137 total relatives with bipolar disorder (after exclusion of relatives of probands with schizoaffective disorder, manic type, and recurrent unipolar depression, as in the study by Winokur et al.). The relatives with records were significantly more likely to be psychotic than those without records. Among our bipolar disorder relatives with records, 48.3% (14 of 29) were psychotic, whereas among those without records the figure was 28.7% (31 of 108) (χ²=3.97, df=1, p=0.05). Because the Winokur et al. analysis omitted those without records, many nonpsychotic relatives may have been excluded, and, in the presence of familial aggregation of psychotic symptoms, those excluded would have been disproportionately relatives of nonpsychotic probands, thus biasing the study group toward a negative finding.

Our findings should be considered in the context of other studies of several types that have yielded evidence to support the utility of subtyping by the presence of psychosis. These can be grouped into clinical studies, family studies, and biological studies.

1. Clinical studies have included those that assessed symptoms, course of illness, and treatment response. Two studies using factor analysis to assess the underlying structure of symptoms in mania have demonstrated an independent psychotic factor based on the presence of hallucinations and delusions +(20, +21). One study +(22) was an assessment of the factor structure of familial bipolar disorder and produced evidence for the three-factor model of negative, psychotic, and disorganized dimensions. Investigations on the course of illness have shown that, compared to nonpsychotic bipolar subjects, subjects with psychosis have an earlier onset +(23), more impaired functioning over time +(24–+26), and a higher rate of relapse at 4-year follow-up +(24). One study, however, did not show differences in functioning +(27). The findings of studies assessing treatment response have been more equivocal +(28–+31).

2. Family studies have suggested shared liability between psychotic affective disorder and schizophrenia, although the evidence for overlapping liability between bipolar disorder (irrespective of psychotic symptoms) and schizophrenia is less strong. Three large data sets have been studied to address the issue of overlap between schizophrenia and psychotic affective disorder +(6, +32–+35). In all three, higher than expected rates of psychotic affective disorder in relatives of schizophrenic probands, and vice versa, were found. Methodologically rigorous studies +(36–+38), however, have not shown an excess of schizophrenic relatives in families of probands with bipolar disorder or an excess of relatives with bipolar disorder in families of probands with schizophrenia; excess rates of major depression and of schizoaffective disorder were found in both types of families.

3. Biological studies have focused on brain imaging and on neurophysiology. One study +(39) indicated anomalous brain asymmetry in the postcentral gyrus in psychotic, but not nonpsychotic, subjects with bipolar disorder. A functional imaging study +(40) demonstrated high dopamine D₂ receptor densities in patients with psychotic, but not nonpsychotic, bipolar disorder; the high levels were similar to those seen in schizophrenia patients. Evidence for shorter REM latency in psychotic than in nonpsychotic subjects with depressive episodes (some of whom had bipolar disorder) has been found +(41). In two brain imaging studies +(42, +43) subjects with psychotic affective disorder and those with schizophrenia had abnormally low volumes of similar regions: the left hippocampus in one study +(42) and the left posterior amygdala-hippocampal complex in the other +(43).

If subtyping bipolar disorder by the presence of psychosis is biologically meaningful, at least two scenarios seem plausible: there might be a vulnerability to psychosis and a separate vulnerability to bipolar disorder, or there might be a particular vulnerability to psychotic bipolar disorder. Given that psychosis is familial in our study group, either the separate vulnerabilities or the particular vulnerability may be the result of genetic influences. Either scenario would be consistent with the hypothesis that genes increase susceptibility to both psychotic bipolar disorder and to schizophrenia: the separate psychosis vulnerabilities might reflect shared psychosis genes because delusions and hallucinations occur in both disorders; the particular psychotic bipolar disorder vulnerability might reflect genes that predispose to psychotic symptoms and to affective symptoms because affective symptoms frequently co-occur with psychosis not only in bipolar disorder, but also in schizophrenia—there is a high rate of depressive symptoms in schizophrenia +(44), and this co-occurrence may be familial +(45, +46). Studying bipolar disorder families enriched for the psychotic bipolar disorder subtype could yield greater power to detect genes leading to either of the vulnerabilities described. In this regard, the chromosomal regions 10p12–13, 13q32, 18p11.2, and 22q11–13, which have been implicated in both bipolar disorder and schizophrenia +(4), are of particular interest. Further study of families with psychotic bipolar disorder should also focus on putative schizophrenia-associated biological markers, including abnormalities in evoked response, oculomotor, neuroimaging, and cognitive measures +(47).

In summary, our data demonstrate that hallucinations and delusions during affective episodes show familial aggregation in bipolar disorder pedigrees. This finding suggests the value of using the psychotic bipolar disorder subtype in genetic and biological investigations. Such investigations may inform the discussion about possible etiologic overlap between bipolar disorder and schizophrenia.

Presented in part at the VIII World Congress of Psychiatric Genetics, Versailles, France, Aug. 17–21, 2000. Received Aug. 16, 2000; revision received Jan. 12, 2001; accepted Feb. 7, 2001. From the Department of Psychiatry and Behavioral Sciences, The Johns Hopkins School of Medicine; and the Department of Biostatistics, School of Public Health, University of North Carolina, Chapel Hill. Address reprint requests to Dr. Potash, Meyer 3-181, Johns Hopkins Hospital, 600 North Wolfe St., Baltimore, MD 21287-7381; jpotash@jhmi.edu (e-mail). Supported by grants from the Theodore and Veda Stanley Foundation, the Charles A. Dana Foundation Consortium on the Genetic Basis of Manic Depressive Illness, and the National Alliance for Research on Schizophrenia and Depression; by NIMH grant MH-42243 to Dr. DePaulo; and by the Affective Disorders Fund, the George Browne Laboratory Fund, and the Alexander Wilson Schweizer Fund at Johns Hopkins University. The authors thank Drs. Francis McMahon, Susan Folstein, Phillip Slavney, and Paul McHugh; Barbara Schweizer, R.N., Jo Thomas, Erin Miller, Krista Vishio, Katie Aman, and Jen Chellis; and members of the study families who volunteered their time.