Abstract
Objective
To review the literature on neurocognitive measures as risk markers for schizophrenia and to present data from the Perth family study of schizophrenia. Of all the risk markers that have been identified, the most promising are deficits in sustained attention.
Method
Inclusion in the review was determined by whether the research addressed a number of key questions: methods of assessing sustained attention; evidence of sustained attention deficits in patients and first-degree relatives including children; the importance of attentional dysfunction in the schizophrenic process and functional outcome; and the biological basis of sustained attention deficits.
Results
Sustained attention deficits are evident in both patients and a proportion of their first-degree relatives, a finding replicated in preliminary data from the Perth family study. The literature suggests that the attention deficit is a stable enduring trait that is independent of clinical state. The neural basis of the deficit may be a functional disconnection between prefrontal and parietal cortex. Attention impairment is an important predictor of functional outcome in patients and the development of social dysfunction in adulthood in the at-risk offspring of patients. However, sustained attention deficits that are measured in childhood results in an unacceptable high false-positive rate (21%) when predicting which at-risk offspring of parents with schizophrenia will develop a schizophrenia spectrum disorder, although the overall classification accuracy (78%) is impressive.
Conclusions
The main findings are that sustained attention deficits are important risk markers for schizophrenia but need to be supplemented by other neurocognitive risk markers to improve predictive accuracy.
There are two approaches that have been used to define at-risk groups for schizophrenia in the neurocognitive area. The first is to assess first-degree relatives of patients with a diagnosis of schizophrenia. Two types of studies fall into this category: (i) prospective studies, which take the form of investigations of children who are at risk of schizophrenia because one parent, usually the mother, has a diagnosis of schizophrenia; and (ii) cross-sectional or family studies that investigate the adult siblings or parents of patients with a diagnosis of schizophrenia. The second approach is to assess individuals who are defined as being at risk on the basis of their scores on various psychometric measures such as the Chapman Psychosis Proneness Scales [1] or the Raine Schizotypal Personality Questionnaire (SPQ) Scale [2]. Both strategies seek evidence of deficits on neurocognitive measures in at-risk populations that parallel those observed in schizophrenia patients. The present paper will focus on the first strategy; that is, examining at-risk adult populations defined on the basis of their relationship to a family member who has a diagnosis of schizophrenia.
Why search for neurocognitive factors as phenotypic markers? This strategy has partly arisen due to a lack of success in linkage studies using a clinical phenotype. There are probably many reasons why linkage studies using the clinical phenotype have not lived up to their original promise, but one possibility is that the signs and symptoms of the disorder may be rather remote from the effects of the genotype itself. Therefore, in recent years, many have argued that it could be more productive to focus on functions that are more closely tied to brain mechanisms, but nonetheless functions that are both clinically and biologically plausible [3]. Neurocognitive functions fall into this category. These functions may be more closely related to the mechanisms of gene action, for instance, the neural networks that underpin high level cognition. These functions may also be more sensitive to the underlying genetic variations in the pedigrees rather than the clinical phenotypes (although this strategy is still to be proven).
There are a number of promising neurocognitive risk factors for schizophrenia under investigation. One of the most promising areas is the performance by first-degree relatives of patients on tasks that probe a number of cognitive domains. Examples include: sustained attention (which will be a focus of this paper); eye-movement disorders [4–7]; spatial working memory [8]; verbal memory [9] and executive functions (i.e. those functions that are implemented by the prefrontal cortex and networks linked to prefrontal cortex), such as abstraction [9]; verbal fluency [10]; and inhibition of prepotent responses (responses that would normally have a high probability of occurrence on the basis of the presented stimulus [11,12]). There are also three electrophysiological measures derived from brain event-related potentials (ERP) exhibiting promise as risk factors. These are the P50 test-conditioning ratio [13–16], P300 amplitude [17,18] and P300 latency [19–21]. The most promising of all these neurocognitive risk factors is sustained attention.
Clinical plausibility of sustained attention deficits in schizophrenia
Does sustained attention, as a risk indicator, have clinical plausibility? The concept of attention is a multifaceted one covering a variety of functions. While this diversity has led some theorists to question the utility of the concept, it has led others ‘to meet the challenge of understanding the similarities and differences between the varieties of attention’ [22, p. 3]. Current cognitive models distinguish at least three separate components to attention: (i) sustained attention; (ii) selective attention; and (iii) attentional control (although Zubin provided a similar working definition of attention back in 1975! [23]). All three components are critical to maintaining coherent behaviour in the face of multiple action or response alternatives, and it is this idea that provides considerable clinical plausibility to the notion that attention may be implicated in schizophrenia. Sustained attention, for example, ensures that the goals of behaviour are maintained over time. Selective attention ensures that only relevant external or internal stimuli gain access to the action system at any particular point in time. Attentional control ensures that the system can deal with competing demands of concurrent information processing activities [22]. The proposal that impairments of attention and of sustained attention in particular are fundamental to the disorder of schizophrenia has a long history in schizophrenia. For instance, Kraepelin noted that ‘it is quite common for patients with schizophrenia to lose both the inclination and the ability of their own initiative to maintain fixed attention for any length of time’ [24, pp. 5–6].
Assessment of sustained attention using continuous performance tasks
The continuous performance task (CPT) is the most common method of assessing sustained attention. The CPT is often described as assessing vigilance, and the terms vigilance and sustained attention are used interchangeably in the current literature. The older vigilance literature emphasised tasks (watches) of long duration, which allowed an examination of the decline in performance over a period of time. It is now known that very long versions of the CPT are not required to demonstrate sustained attention deficits in patients with schizophrenia and, in general, there is no marked decline in performance with task duration; that is, the deficit appears to be present at the onset of the task [25–27].
There are a number of variants of visual CPT. All assess the ability to discriminate targets from non-targets over a large number of trials but vary on stimulus parameters as well as the degree of perceptual and cognitive processing load.
The ‘X’ version, the simplest version of the CPT, presents the subject with a series of letters one at a time on a computer screen. The task is to press a button every time a particular target stimulus, for instance, an ‘X’, is detected.
The ‘AX’ version, a slightly more difficult version of the CPT, in which a series of letters are again presented (as described previously) and the task is to press a button every time a particular sequence, ‘A’ followed by ‘X’, occurs.
The identical pairs or ‘IP’ version usually consists of numbers, or shapes, and the target is defined by consecutive numbers (or shapes) which are identical. For the numbers version, task difficulty can be manipulated by increasing the number of digits presented on each trial. This task increases working memory load as each stimulus must be maintained in working memory and monitored to allow for comparison with the immediately following stimulus, and the contents of working memory updated following the presentation of each new stimulus. In general, because of the rapid delivery rate of stimuli (1 or 2 s onset to onset) used in most versions of the identical pairs version of the CPT, it is likely that the demands on the maintenance function of working memory are minimal in these tasks in contrast to the monitoring and updating functions. These operations place a higher load on working memory than either the ‘X’ or ‘AX’ versions.
The degraded stimulus X-version of the CPT consists of single digits (0–9) presented at a rapid rate in a degraded form with the target being ‘0’. The CPT-DS version challenges the sustained attention system by increasing the perceptual load (feature extraction and stimulus identification).
The degraded stimulus A-X version of the CPT in which digits are presented in a degraded form and a target defined by the number ‘9’ preceded by ‘1’ [28].
Versions of these tasks for use with children have been developed [29]; for instance, ‘X’ and ‘IP’ versions, based on playing cards with the target defined by a particular card (an ‘X’ version), or by consecutive cards that are not only identical in number but also in suit (an ‘IP’ version). A small number of studies report the use of auditory CPT tasks [9,30,31].
The early literature on CPT tasks in schizophrenia was directed at determining whether specific cognitive deficits would enable identification of at-risk individuals and possibly lead to the development of intervention strategies [32–34]. However, the aim of a great deal of current research is to determine the viability of CPT performance deficits as a phenotypic marker of vulnerability to schizophrenia for linkage studies, using a number of criteria including sensitivity, specificity, independence of medication, independence of clinical state and occurrence in asymptomatic at-risk populations. The evidence for this is very promising [3,35].
Sustained attention deficits in patients and first degree relatives
Patients with schizophrenia have been shown to be impaired on the X and AX versions of the CPT (as described earlier) compared with controls and also compared to other patient groups (reviewed by Cornblatt and Kielp [35]). Patients with an affective disorder also display CPT deficits but the nature of the attentional impairment is different from that observed in patients with schizophrenia. Affective patients are only impaired on the shape version of the CPT-IP and their performance is characterised by a high false alarm rate. In contrast, schizophrenia patients are impaired on both number and shape versions of CPT-IP and their performance is distinguished by a low hit rate and a high rate of random false alarms [25]. Research has shown that the CPT deficit in schizophrenia is independent of psychotic state [36,37] and is not a secondary effect of chronicity, severity of illness or hospitalisation [38,39]. It has been reported that while patients in remission show improved performance, they do not perform at the level of unaffected individuals; that is, there is a residual deficit [40–43]. Medication improves performance but, once again, does not bring it up to the level of controls [39,44,45].
Performance deficits on the CPT are also evident in first-degree relatives: unaffected siblings and parents as well as children of patients with schizophrenia. However, more demanding tasks are required to demonstrate deficits in unaffected relatives. For example, the X and AX undegraded versions of the CPT task do not produce impaired performance in adult relatives [9] or in children who have one or more parents with schizophrenia [29]. In contrast, the identical pairs or degraded stimulus versions have been shown to be sensitive enough to detect subtle sustained attention deficits in at-risk individuals, both first-degree relatives of patients with schizophrenia (CPT-IP: [34, 36,46]; CPT-DS: [26]) as well as individuals identified as at risk using psychometric methods [47]. Similarly, a difficult auditory version of the CPT also distinguishes at-risk individuals (one parent with schizophrenia) from a matched control group [31].
Risch's risk ratio (λ) provides an estimate of the extent to which CPT performance is useful for linkage studies of schizophrenia. λ varies as a function of the cut-off used for defining deficit and is computed by dividing the frequency with which the deficit occurs in first-degree relatives by the frequency in healthy controls. For very stringent cut-off values, λ has been estimated to be in excess of 15 for the undegraded version of the CPT [28] and 30 for both the degraded A-X CPT [28] and the attentional deviance index (ADI), which is derived by Cornblatt [3] and basely largely on measures from the CPT IP [48]. Furthermore, Faraone and colleagues reviewed published data and concluded that the ADI is potentially more useful for linkage analysis than other measures such as assessments of schizotypal personality disorder, eye-tracking dysfunction, allusive thinking, auditory evoked potentials, neuropsychological impairment and neurological signs [3].
In terms of positive predictability (i.e. success in predicting whether an at-risk individual will develop schizophrenia), measures of attention deficits derived from CPT are less successful, although it is clear that a developmental deficit is predictive of aspects of adolescent and adult social functioning. The New York High-Risk Project has shown that a sustained attention deficit (measured using the ADI) is evident in at-risk children at 7 years of age and is stable throughout development in the at-risk group. The ADI measured at age 7 or 9 was predictive of global adjustment measured at adolescence in the at-risk children [32] and of social isolation in adulthood [33,34]. However, the ADI measured in childhood predicted adolescent and adult functioning only in the group who were at risk for schizophrenia and not in a group at risk for affective disorders or in the control group of children whose parents have no psychiatric history. Freedman and colleagues [49], working with an expanded at-risk sample made up of two separate cohorts from the New York High-Risk Project, demonstrated that prediction of social dysfunction in adulthood could be improved by combining the ADI measured in childhood with the physical anhedonia scale from Chapman's Psychosis Proneness scales [3] measured at adolescence. Using path analysis, the ADI predicted two (suspicious solitude and lack of empathy) of three components of a social dysfunction index directly, but for two of the components (suspicious solitude and social insecurity), prediction was enhanced when the potentiating value of the physical anhedonia scale was taken into account. Nonetheless, the ADI had the largest total effect of any of the predictors and the largest effect was on suspicious solitude, the outcome most closely tied to schizoid personality.
The data from the New York High-Risk cohorts suggest that the ADI measured in childhood does not predict psychosis in adulthood, although it may well be predictive of a schizophrenia spectrum disorder. Erlenmeyer-Kimling and coworkers report that there is no link between ADI and psychosis for the first cohort recruited in 1971–1972, not even in the at-risk group for schizophrenia [50]. For the second independent cohort recruited in 1977–1979, Cornblatt et al. report that performance on the most difficult version of the CPT-IP (4-digit version) measured in late childhood and early adolescence does distinguish between the small number of at-risk offspring (n = 6) who subsequently developed a schizophrenia spectrum disorder (schizophrenia, schizoaffective disorder, schizotypal and paranoid personality disorders) and other at-risk offspring (n = 15) [51]. The prespectrum group performed significantly worse on the CPT-IP than the remainder of the high-risk group (and the group at risk for affective disorder and a no-risk group) at both ages. The investigaters do not present separate data for those offspring (n = 2) who subsequently developed schizophrenia or schizoaffective disorder. Attention performance in the prespectrum group was not affected by the onset of the illness, suggesting that the sustained attention deficit is an enduring trait that precedes the illness by many years and is not related to acute episodes. Furthermore, attention was independent of clinical ratings of behavioural abnormalities (based on ratings from Global Adjustment Scales) as demonstrated by different developmental trajectories of the two measures.
Discriminant function analysis indicated that sustained attention measured in late childhood is a reasonably accurate predictor of an adult schizophrenia spectrum disorder, giving 78% correct classification overall with a sensitivity of 67% and specificity of 79%. Adding sustained attention measured in late adolescence did not markedly improve the accuracy of classification. Cornblatt and her colleagues suggest that adding other neurocognitive measures might improve the unacceptably high false-positive rate of 21%. Their own data [51] suggest that adding ratings of global adjustment taken in late childhood to attentional measures resulted in only a minor improvement in prediction accuracy. It is only when adjustment ratings in the mid-teens are added that there is a nontrivial improvement in classification accuracy (overall accuracy of prediction of 89%, sensitivity of 83%, and specificity of 90%) but the mid-teens may be too late to institute any effective interventions in the at-risk group.
Therefore, there is consistent evidence that those at risk of schizophrenia exhibit attentional dysfunction when sufficiently challenging sustained attention tasks are used. In addition, children at increased risk of schizophrenia exhibit sustained attention deficits in childhood, deficits that are predictive directly of some types of social dysfunction in adulthood. The observation that attentional dysfunction in childhood predicts physical anhedonia in adolescence which in turn predicts adult outcomes, points to a central role for attentional dysfunction in the development of schizophrenia-related personality traits. Those offspring who go on to develop a schizophrenia spectrum disorder exhibit a marked and stable impairment of attention from childhood but attentional deficits are not exacerbated by the onset of illness and are independent of clinical ratings. It is important to note that these relationships between attentional dysfunction in childhood and poor adult outcome have only been observed in those who have a genetic liability for schizophrenia.
The role of attentional dysfunction in the schizophrenic process and functional outcome
There are two models of how attention deficits relate to the clinical features of schizophrenia: (i) attention dysfunction leads directly to the symptoms of the disorder; and (ii) there is a primary defect at cellular level that causes both the attention dysfunction and symptoms, but the two are relatively independent and only loosely correlated. In the past, most interest has focused on the first model; that a core deficit in attentional processes can provide an explanation of the clinical manifestations of schizophrenia [23,52,53]. Early theoretical models of symptoms focused on attentional abnormalities as an explanation of the positive symptoms of schizophrenia, particularly during the early stages of the disorder [52]. However, subsequent research found either a relationship between attention deficits and negative symptoms [54] or, more commonly, no evidence of a relationship between impaired attention and clinical symptoms [55–57].
In recent years, and in line with the second model outlined earlier, there has been increasing recognition that cognitive impairments should be considered a fundamental feature of the psychopathology of schizophrenia and, in this respect, as important in characterising the disorder as the range of positive and negative symptoms. It is also becoming increasingly evident that neurocognitive deficits in the domains of attention and memory have considerable impact on the daily lives of patients by restricting their functional capacity and contributing in a major way to their social disability. These deficits are more strongly predictive of functional outcome than psychotic symptoms and in many instances, but not universally, negative symptoms as well [58]. Conventional neuroleptics do not usually have a major impact on neurocognitive deficits despite pronounced effects on psychotic symptoms. While atypical antipsychotics are more effective in ameliorating cognitive impairments, residual impairments remain [59].
It is important to understand the extent and nature of the neurocognitive deficits in schizophrenia as they may act as rate-limiting factors in the patients' ability to acquire new skills in rehabilitation programs. For instance, deficits on sustained attention tasks are predictive of the patients' ability to acquire and re-learn psychosocial skills [reviewed by Keefe et al. 59]. This is not surprising given the complexity and sometimes subtle nature of the information that a participant must detect and respond to in most social interactions (a real-life analogue of a sustained attention task involving detection of subtle transient cues).
Therefore, it is evident that sustained attention impairments are a key feature of the schizophrenic disorder and understanding the nature of these impairments has direct implications for the treatment and management of the patient. It is therefore somewhat surprising that there have been few attempts to remediate sustained attention deficits. One of the few studies in this area by Medalia and colleagues used a computerised remediation system to train patients on a number of component operations that are important in sustained attention (arousal, alertness, scanning, target detection and rapid information processing) [60]. Significant improvement in post-CPT performance (relative to pre-CPT performance) was observed as well as in the Brief Psychiatric Rating Scale in the intervention group relative to a control group. This preliminary study needs replication with a better design in terms of the comparison control group, which in the Medalia et al. study was essentially a no-treatment control group.
Biological plausibility of sustained attention deficits
Neuropsychological data from brain-injured patients that suggest a special role for right anterior regions in sustained attention [61], has received considerable support from functional neuroimaging studies (positron emission tomography, single photon emission computed tomography and functional magnetic resonance imaging) of intact individuals [62]. However, neuroimaging data reveal the extent of the network of activated areas includes not only right dorsolateral and inferior prefrontal regions, but also posterior parietal cortex, superior temporal gyrus, anterior cingulate, basal ganglia and thalamic nuclei [reviewed by Morecraft et al. 63]. These areas overlap substantially with a number of models of the functional architecture of visual attention [63–65].
The most frequently investigated CPT paradigms in neuroimaging studies are the CPT-DS [66], CPT-AX [67], and an auditory equivalent of the CPT-X paradigm [68]. There is a single published study of controls using a variant of the CPT-IP [69], the CPT-double (targets were consecutive occurrences of the letter T) and only one published report (other than pilot data [35]) using the CPT-IP [70] as an activation paradigm. However, the latter presents data only for a comparison of a number version with a shapes version of the task and no data are presented comparing either task with a baseline condition. Hence, there is no information available on the network of areas activated by the CPT-IP numbers version of sustained attention tasks. The precise network of areas activated by a CPT task varies somewhat across different reports, but apparent inconsistencies may be attributable to the type of CPT paradigm used to generate activation and the degree of spatial resolution and sensitivity of the imaging technique or analyses.
Functional neuroimaging techniques provide a rather direct means to investigate the dysfunctional neural circuits that underpin sustained attention deficits in patients with schizophrenia and those at risk. A number of studies of patients have shown promise in this respect but, as yet, there are no published reports of data from first-degree relatives. Relative to healthy controls, patients show reduced activation in frontal areas (measured either in absolute terms [71] or as a ratio of frontal to posterior activation [71,72]); the anterior cingulate [66,73]; caudate [73]; striatum in a subset of patients [73]; and thalamic nuclei [73]. However, there are a number of reports that find increased activation in patients over parietal [66,71,72] and/or temporal areas [72,73]. These findings suggest an abnormal pattern of fronto–posterior interaction in schizophrenia in sustained attention tasks, comparable to the disturbed fronto–temporal interaction that has been observed in verbal fluency tasks [74] and verbal memory tasks [75].
In summary, the available data suggest that sustained attention tasks activate at minimum a network involving right dorsolateral and inferior prefrontal regions, posterior parietal cortex, inferior temporal regions, anterior cingulate, basal ganglia and thalamic nuclei, and that these regional specific activations are abnormal in patients, with generally reduced activation in anterior areas and excessive activation in parieto–temporal areas. Importantly, most of these areas have also been implicated in the pathophysiology of schizophrenia, with reports of morphometric changes in the anterior cingulate [76], MRI evidence of thalamic abnormalities [77], basal ganglia volume enlargement [78] and volume decrease in frontal lobes [79], although the latter is not universally observed.
Preliminary sustained attention data from the Perth cohort
Over the past three years, the University of Western Australia Centre for Clinical Research in Neuropsychiatry (CCRN) at Graylands Hospital in Perth, Western Australia, has been collecting data on the identical pairs version (four digit version) and the degraded stimulus version of the CPT in a large family study. While recruitment of patients and family members has been strong (in excess of 100 people for both categories for some measures), the recruitment of normal controls has been somewhat slower (50–60; Table 1). All controls are screened for psychopathology and family history of psychosis. Two problems exist with the recruitment to date. First, the family members consist of adult siblings and parents of people with schizophrenia, which gives an older age range for family members. Second, most of the patients recruited are male, most of the family members are female and the controls have approximately equal numbers of males and females. These problems have been dealt with by using analysis of covariance to adjust for age, gender, education levels and the National Adult Reading Test (NART), as a measure of premorbid intellectual functioning.
Details of participants in the Centre for Clinical Research cohort completing the identical pairs (CPT-IP) and degraded stimulus (CPT-DS) versions of the continuous performance task. Mean (± SD) age, years of formal education and estimated premorbid IQ based on the National Adult Reading Test (NART) are shown as well the number of males and females in each group
The CPT-IP task utilised four-digit numbers (1–9) presented with an exposure time of 200 ms, a stimulus onset asynchrony (onset to onset or SOA) interval of 2 s. Subjects were instructed to press a button as rapidly as possible to targets of consecutive identical four-digit numbers. The task was made more difficult by having not only target trials but catch trials as well, where the consecutive four-digit numbers differ by one digit only. Performance was measured via hit rate (i.e. correctly detected targets) and false alarm rate to catch trials (the number of catch trials on which subjects respond incorrectly) and the discrimination index, dL, and bias index, CL, based on the logistic distribution functions [81].
The data from the identical pairs test showed a large separation in terms of dL, hit rate, false alarm rate and mean reaction time to hits between the three groups (patients, asymptomatic and clinically healthy family members, and controls; Table 2) and between the family members and controls for dL and false alarm rate only (Table 2).
Adjusted mean (± SD) hit rate, false alarm rate, discrimination (dL) and bias indices (CL), and reaction time of probands, family members (FM) and controls on the Clinical Research cohort completing the identical pairs CPT (CPT-IP) after controlling for the covariates of age, gender, number of years of formal education and National Adult Reading Test (NART)
Bonferroni adjusted probability (p) value for multiple comparisons.
For the degraded stimulus version of the task, the stimuli were single digits (0–9) presented for 50 msec with a 1 s interval SOA. Stimuli were degraded by removing 60% of the pixels at random. The task for subjects was to press a button as rapidly as possible when they detected the digit ‘0’. Results (Table 3) show a differentiation between probands compared to family members and controls but the differentiation between family members and controls was not significant for any measure.
Adjusted mean (± SD) hit rate, false alarm rate, discrimination (dL) and bias indices (CL), and reaction time of probands, family members (FM) and controls on the Centre for Clinical Research cohort completing the degraded stimulus CPT (CPT-DS) after controlling for the covariates of age, gender, number of years of formal education and National Adult Reading Test (NART)
Bonferroni adjusted probability (p) value for multiple comparisons.
The results to date extend the findings from a number of laboratories that family members of people with schizophrenia (an at-risk population) on average perform more poorly on the CPT task than controls who have no family history of psychoses. This is particularly the case for the CPT-IP.
In addition to neurocognitive measures of sustained attention performance derived from the IP and DS versions of the CPT, measures of immediate and delayed verbal recall (using the Rey Auditory Verbal Learning Task or RAVLT) and verbal fluency (using the FAS version) have also been obtained from the Perth cohort recruited at the Centre for Clinical Research in Neuropsychiatry (Table 4). This cohort of patients, family members and controls is being assessed currently on an inhibition task, the antisaccade eye-movement task, as well as three electrophyisological measures, P50 test/conditioning ratio, mismatch negativity (MMN) and P300 amplitude and latency.
Adjusted mean (± SD) recall, immediate and delayed on the Rey Auditory Verbal Learning Task (RAVLT) and the number of words produced, and number of rule breaks on verbal fluency for probands, relatives and controls after controlling for the covariates of age, gender, number of years of formal education and National Adult Reading Test
A multivariate analysis of covariance (M
For the two-group analysis, after adjusting for the covariates of age, gender and premorbid IQ (education was not required in the model), the combined dependent variables significantly discriminated between relatives and controls on the basis of Wilks' criterion (Table 5). Univariate tests indicated that, in addition to poorer discrimination performance on the CPT-IP (dL), family members had lower immediate recall on the RAVLT and produced fewer total words on verbal fluency than controls (Table 4). Based on the relative size of the standardised discriminant function coefficients (SDFC; Table 5), these variables, in addition to discrimination performance on the CPT-DS, were the largest contributors to the composite dependent variable. The Roy–Bargman stepdown procedure indicated that only discrimination performance on the CPT-IP and immediate recall on the RAVLT made unique contributions to the discrimination between relatives and controls.
Summary of multivariate analysis of covariance (Mancova) with dependent variables of dL, CL and hit reaction time for the CPT-IP and CP-DS tasks; total words recalled on the immediate and delayed test from the Rey Auditory Verbal Learning (RAVLT); and total number of words produced and number of rule breaks in the FAS verbal fluency task after controlling for the covariates of age, sex, years of education and NART. Two Mancova are summarised: (i) the comparison of three groups of patients, relatives and controls; and (ii) the comparison of relatives and controls. For each Mancova, the probability (p) values for the univariate tests for each variable are shown, together with the standardised discriminant function coefficients (SDFC)
Multivariate F-test (from Wilks' lambda) of significance (F = 8.13, df = 20,528, p < 0.001).
Multivariate F-test (from Wilks' lambda) of significance (F = 2.24, df = 10,180, p < 0.017).
Not adjusted for multiple comparisons.
For the three-group analysis, after adjusting for the covariates of age, gender, premorbid IQ and education, the combined dependent variables were again significantly related to group membership (Table 5). Inspection of the SDFC (Table 5) and the Roy– Bargmann stepdown procedure revealed that four variables made unique contributions to the composite dependent variable that best discriminates between the three groups (i.e. patients, family members and controls): (i) discrimination performance and (ii) RT on the CPT-IP; (iii) delayed (rather than immediate) recall on the RAVLT; and (iv) total words from the FAS verbal fluency. In other words, sustained attention, speed of responding, delayed recall and verbal fluency make independent contributions to group differences; however, for both group comparisons, the most important contributions were measures derived from the CPT-IP (Table 5).
Conclusion
In conclusion, sustained attention deficits are important risk markers for schizophrenia. Not only do patients exhibit impairments on relatively undemanding sustained attention tasks, a percentage of their first-degree relatives exhibit similar deficits when working memory is challenged by the sustained attention task. In patients, the deficit is independent of clinical state and in at-risk offspring who develop a spectrum disorder, the deficit is not affected by the onset of the illness. These and other findings point to the stable, trait nature of the impairment. There is emerging evidence that dysfunctional neural circuits underpin these attention impairments in patients, although little is known about the neural basis of the impairments in relatives. Evidence is also emerging demonstrating the crucial role of attention in determining outcomes for patients in terms of their functional capacity and ability to benefit from rehabilitation programs and for at-risk offspring in terms of their adult level of social functioning. However, in order to improve the capacity to predict which at-risk offspring of parents with schizophrenia will develop psychosis or a schizophrenia spectrum disorder, sustained attention measures need to be supplemented by other neurocognitive measures. The data from the Perth cohort suggest that measures of verbal memory and executive functions (verbal fluency) are two variables that should be considered (in addition to sustained attention) in the search for increased accuracy of prediction.
Footnotes
Acknowledgements
This research was supported by NH&MRC grant 960579. We are most grateful to T. Budd and P. Dragicevich for their technical support and J. Combrinck, P.W. Owen, D. Wood, B. Jansen and J. Cooper for reviews of patient diagnoses. In particular, we thank the patients and family members who participated in the research.