Neuropsychological risk indicators for schizophrenia: a preliminary study of female relatives of schizophrenic and bipolar probands

Abstract

Evidence of subtle neuropsychological deficits in relatives of schizophrenic probands (REL-SZs) suggests that these are risk indicators for schizophrenia, but little is known about whether neuropsychological performance in REL-SZs differs from that in other groups of relatives. We compared neuropsychological function in female REL-SZs (n=39), relatives of primarily psychotic bipolar disorder probands (REL-BPs; n=15), and a normal control group (n=44). After adjustment for expected intellectual ability (based on reading recognition), REL-SZs showed deficits in verbal and visual memory (Wechsler Memory Scale-Revised logical memories, visual reproductions), and auditory attention (dichotic digits) compared with either REL-BPs or control subjects. Memory, but not dichotic listening differences remained significant after adjusting for current IQ; however, average effect sizes after controlling for either reading or IQ were roughly comparable for these three parameters (d=0.80, 0.71, and 0.69, respectively). REL-BPs and control subjects showed little difference. Although both schizophrenic and bipolar patients often manifest neuropsychological dysfunction, these preliminary findings indicate subtle neuropsychological deficits only in REL-SZs. Such differences suggest different underlying processes; neuropsychological impairment may, in part, reflect an expression of genetic liability to schizophrenia but not bipolar disorder. Replication with a larger REL-BP sample and with male relatives is needed to evaluate the generalizability of the results.

Introduction

There is now widespread evidence of neuropsychological impairment in schizophrenia (Saykin et al., 1991, Bilder, 1992, Hoff et al., 1992a, Seidman et al., 1992, Heaton et al., 1994), including deficits that are present prior to and at the onset of illness (Aylward et al., 1984, Hoff et al., 1992b) and often persist even after patients are relatively remitted (Spohn and Strauss, 1989, Sweeney et al., 1991). Neuropsychological deficits are also found in adult biological relatives of schizophrenic probands (REL-SZs) in comparison with normal volunteers (Mirsky et al., 1992, Cannon et al., 1994, Keefe et al., 1994, Faraone et al., 1995). Although milder in relatives than in schizophrenic patients, these deficits tend to be most prominent in the same neuropsychological domains in both relatives and patients (Kremen et al., 1994). We refer to deficits or impairments to indicate significantly worse performance than that of a comparison group, even though the group mean performance of REL-SZs is within normal limits.

These data support the notion that neuropsychological impairment is a genetically mediated risk indicator for schizophrenia on the basis of two criteria (Kremen et al., 1992). First, deficits are relatively stable or trait-like in patients and do not appear to be simply a consequence of acute exacerbations of the illness. Second, deficits are found in mildly ill, non-psychotic relatives and even psychiatrically well relatives compared with normal control subjects. The second criterion is related to the sensitivity of neuropsychological risk indicators for schizophrenia. To be considered a valid risk indicator, however, it is important that a characteristic also be milder or less common in other disorders or in the relatives of individuals with other disorders (Kremen et al., 1992). This third criterion is related to the specificity of neuropsychological risk indicators for schizophrenia.

The term `risk', as used in psychopathology research, can have several different meanings. A common usage is probabilistic. For example, stating that an individual is at risk for a disorder asserts that the probability that the individual will eventually develop the disorder is greater than the probability of developing it in the general population. We are not studying risk in this sense because almost all of our study participants have lived well beyond the peak risk period for developing schizophrenia. Rather, we consider neuropsychological deficits to be putative `endophenotypes', a term referring to characteristics that are causally closer to the pathogenic genotype than the disease phenotype itself. Because schizophrenia is a complex disorder, some individuals with significant vulnerability may never develop the disorder because of the presence of protective factors or the absence of other risk factors. Thus, someone may carry a latent liability for schizophrenia, but may never become ill (Meehl, 1990).

Comparisons between patient groups are, of course, important as well, but differences in chronicity, stage of illness, age of onset, and types of medication often make it difficult to fully understand the factors affecting similarities or differences in cognitive function between schizophrenic and other patient groups. Consequently, the study of non-psychotic relatives is a valuable strategy for augmenting the study of patients. Differences may, in fact, emerge between relatives even when no difference is present in patients. For example, eye tracking dysfunction may be present in bipolar patients due to the effects of lithium treatment, whereas it is more common in REL-SZs than in REL-BPs (Levy et al., 1983). Comparing relatives of individuals with different disorders is particularly important if the characteristic is purported to be a genetically-mediated risk indicator.

There has been far less empirical evidence of neuropsychological risk indicators for schizophrenia in REL-SZs vs. relatives of psychiatric comparison groups than in REL-SZs vs. a normal comparison group (Kremen et al., 1994). High-risk studies of children of schizophrenic parents have shown reduced primacy on a digit span task with distraction (Harvey et al., 1981) and impairment on a composite measure of attention that included a visual continuous performance test (CPT; Cornblatt and Erlenmeyer-Kimling, 1985) in comparison with children of mood disorder parents. It should be noted that not all high-risk studies have included psychiatric comparison groups. It has also been suggested that the stress or disruption of being raised by a psychiatrically ill parent may make a non-specific contribution to developmental lags in offspring (Sameroff and Seifer, 1990); this could, in turn, obscure differences that were specific to risk for schizophrenia. One unpublished report showed impaired attention on visual CPTs and impaired memory in adult REL-SZs compared with REL-BPs (Moldin et al., 1995).

The aforementioned studies generally focused on a few tests and did not use comprehensive neuropsychological test batteries that cover a wide range of functions. We previously examined neuropsychological function in normal control subjects and non-psychotic adult REL-SZs using an extensive test battery. Relatives under 60 years of age performed more poorly than control subjects in verbal memory, abstraction-executive function, auditory attention, mental control-encoding, and general verbal ability (Faraone et al., 1995, Faraone et al., 1996). We identified the first three of these neuropsychological functions as putative risk indicators for schizophrenia because, in addition to lower group means, there was also greater variability and/or a higher proportion impaired in the relatives than in the control subjects (Faraone et al., 1995). A study of sex differences in an expanded sample of these adult relatives (n=54; 72 control subjects) showed that in comparison to same-sex control subjects, male relatives had more motor function deficits, whereas female relatives had significant deficits in verbal memory and, to a lesser extent, in mental control-encoding and auditory attention (Kremen et al., 1997).

In the present study, we used the same battery to examine whether neuropsychological risk indicators for schizophrenia are more prominent in REL-SZs than in relatives of individuals with another major mental illness. We compared the performance of individuals in the expanded sample of REL-SZs referred to in the previous paragraph with adult REL-BPs. Most of the bipolar probands had psychotic features.

There are several advantages of using relatives of bipolar patients with psychotic features as a psychiatric comparison group for the study of cognitive dysfunction in REL-SZs. It is well established that genetic factors play an important role in bipolar disorder (Faraone et al., 1990). Concordance rates for monozygotic bipolar and schizophrenic twins are comparable; estimates are approximately 50% for schizophrenic (Gottesman and Shields, 1982) and 58–74% for bipolar (Tsuang and Faraone, 1990) twin pairs approximately. Risk of bipolar illness to first-degree relatives is 5% (Tsuang and Faraone, 1990), similar to risk estimates for first-degree relatives of schizophrenic patients of less than 10% (Gottesman and Shields, 1982). Psychotic mood disorder patients have a sex distribution that is more similar to that of schizophrenia as compared with non-psychotic unipolar depression (Faraone et al., 1990, Schatzberg and Rothschild, 1992). Neuroimaging studies provide evidence for structural brain abnormalities in bipolar patients (Marsh et al., 1996, Pearlson et al., 1997). Mood disorder patients do frequently manifest neuropsychological deficits, and in some cases, these deficits are similar in type and severity to that of schizophrenic patients (Jones et al., 1988, Coffman et al., 1990, Hoff et al., 1990, Morice, 1990, Lesser et al., 1991, Goldberg et al., 1993, Jeste et al., 1996). As has been found for schizophrenia, there is evidence of poor premorbid adjustment in some psychotic mood disorders (Sands and Harrow, 1994). Although mood disorders generally have a better outcome and more episodic course than schizophrenia, an estimated 20–30% of bipolar patients continue to experience mood lability and social-occupational dysfunction outside of manic or depressive episodes (American Psychiatric Association, 1994). Bipolar illness with psychotic features also gives the closest phenomenological approximation to schizophrenia. Thus, it provides the closest and most stringent comparison for examining putative genetically mediated neuropsychological risk indicators.

As noted above, the importance of genetic factors and the presence of neuropsychological dysfunction are two of several areas of similarity between schizophrenia and psychotic bipolar disorder. Although both clinicians and researchers tend to think of neuropsychological dysfunction as more central to schizophrenia, little is actually known about the processes underlying neuropsychological dysfunction in schizophrenia vs. bipolar disorder. The presence of neuropsychological deficits in REL-SZs suggests that such deficits reflect genetically mediated indicators of vulnerability to schizophrenia. The few findings comparing cognitive function in REL-SZs and REL-BPs suggest that this may not be the case for bipolar disorder. If cognitive deficits are part of the genetic aspects of schizophrenia, but not bipolar disorder, then we would expect REL-SZs, but not REL-BPs, to manifest subtle neuropsychological impairment.