Abstract
The Western Australian Centre for Clinical Research in Neuropsychiatry (CCRN) is one of the youngest psychiatric research institutions in this country. It was established in 1994, as a jointly managed research facility of the University of Western Australia Department of Psychiatry and Behavioural Science (now School of Psychiatry and Clinical Neurosciences) and Graylands Hospital in Perth. The formal inauguration by the Minister for Health of Western Australia took place in 1996. The CCRN is physically located at the Gascoyne House on the grounds of the hospital, a spacious heritage building which also contains the WA Psychiatric Library and the offices of the postgraduate training section of the WA branch of the College.
As a prologue to a brief account of its history, development and present state, I feel compelled to start on a personal note (something I generally avoid). Since my move to Perth, in late 1993, I have often been asked: ‘Why Western Australia?’ The question has always baffled me with its implied suggestion that, in the eyes of well-meaning colleagues and friends, I had committed an act of eccentricity. The explanation of my decision is fairly simple. Since my first visit to Australia in 1989 (which had been arranged on the initiative of Scott Henderson), I felt attracted by the academic environment I saw and by the opportunities for research that could allow me to pursue interests and ideas I had ‘filed away’ after the completion of the program of schizophrenia studies during my 12 years with the World Health Organization. I took my bearings with some care, having received two generous offers of senior academic posts, and finally decided – not without soul searching – that Western Australia was closest to my idea of a good environment for psychiatric research. It had splendid population databases that could be harnessed to a program of studies using epidemiological ‘triangulation’ to define targets for biological research. I also felt that the University of Western Australia, less affected than other institutions by the entrepreneurial model that was being imposed on academia in the early 1990s, could provide the right balance between research, teaching and clinical work. By and large, these expectations have been borne out, and 10 years since settling in Western Australia, I feel validated in my original decision.
Soon after arrival in Western Australia, my idea of creating a dedicated centre for psychiatric research met with enthusiastic support at all levels. Allen German, then Dean of Medicine; Peter Burvill, Head of the University Department of Psychiatry, and June McDonald, then Executive Director of Graylands Hospital, were tremendously helpful in the process of laying down the administrative and financial groundwork for the new centre. The plan for the CCRN was approved by the University and I was appointed as its Director in early 1995. Within a few months, we were able to create and advertise senior research positions. It took less than a year to recruit a first-class team of chief investigators which included Joachim Hallmayer (whom I knew from my year at Stanford University, 1991–1992), Patricia Michie (previously at Macquarie University), Matt Martin-Iverson (from the University of Alberta in Edmonton, Canada), David Castle (previously at the Institute of Psychiatry in London), and Vera Morgan (from the University of South Australia). An increasing number of research assistants, PhD students, and technical support staff followed suit. Several visiting researchers from overseas, including Irving Gottesman, Bob Cloninger and Tom McNeil, as well as leading Australian investigators, such as David Copolov and John McGrath, gave us valuable advice in the process of formulating our research program. A highly productive partnership was established with the Department of Psychology, which enabled joint academic appointments across the disciplines, as well as jointly supervised PhD students. By the end of 1995 we were in a position to begin work on several research projects that marked the start of what we hoped would be a coherent, long-term program of research with a focus on schizophrenia, autism and related disorders. An additional encouragement came from the World Health Organization, which designated the CCRN as part of the WHO Australian Collaborating Centre for Research and Training in Mental Health, with the specific responsibility of serving as a reference and training centre for WHO diagnostic instruments, such as the SCAN. In this role, the CCRN participated actively in the National Health and Medical Research Council (NHMRC) Network for Brain Research in the Mental Disorders (1994–2001).
Since its establishment, CCRN has enjoyed the continuous and reliable support of the Graylands Hospital and the Department of Health (people that I wish to mention especially are Geoff Smith, George Lipton, Russell McKenny, Jim Thompson, Gary Wallace, and, more recently Aaron Groves and Peter Wynn Owen), which provided funding for a number of key staff positions, as well as equipment and general infrastructure.
Schizophrenia research at CCRN: concepts and hypotheses
A research program that has a chance of success must be self-propelled by shared interests, expertise and dedication of a core group of people who want to work together and are prepared to spend long hours of ‘handson’ involvement in tackling a problem they deem important. To our newly formed core group, the important problem was schizophrenia. It affects approximately 1% of the population worldwide and accounts for 2.3% of the global burden of disease. In Australia, the point prevalence is 5 per 1000 population aged 15–54, and some 1500–2000 new cases are diagnosed each year. The discovery of more effective treatments and prevention strategies in schizophrenia depends critically on progress in the understanding of its aetiology.
Our shared concern was that schizophrenia research is ridden with conceptual fallacies. The most common one (illustrated by hundreds of journal publications) is to select a particular, isolated aspect of the syndrome – be it a symptom, a neuropsychological dysfunction, a neurotransmitter abnormality, a neuroanatomical feature, or a genetic polymorphism – and to demonstrate a difference at p < 0.05 in a woefully small sample of patients, as compared to a control sample of convenience. The brain function, structure, or candidate gene for that matter – supposedly affected by the dysfunction – is then inferred to be involved in the causation of the disorder. This kind of research, in my view, leads nowhere and only serves the proliferation of publications which, in their totality, suggest that hardly any brain region, neurophysiological function, or chromosome, is spared in schizophrenia. This kind of piecemeal approach cannot be of much help in trying to understand an extremely complex disorder. The loss of the ‘big picture’ in the hunt for marginally significant results, using the glamorous technologies of the day, has led over the last decades to numerous ‘breakthroughs’ in the understanding of the disease, which have often proved to be illusory. But what is the ‘big picture’? My personal predilection is to turn periodically to the great clinicians of the past, like Kraepelin and Bleuler, or to contemporary thinkers like Meehl, who provide a bird's eye perspective on the complex ‘gestalt’ of the disorder. Barring some attempts (proceeding from a rather superficial treatment of the clinical facts and research evidence) to discard the very notion of schizophrenia as an artificial construct or a ‘scientific delusion’, the clinical entity of schizophrenia appears to be a loose cluster of quite heterogeneous symptoms and traits, held together by a ‘deep structure’ (Meehl) which, for the time being, eludes our understanding. The current diagnostic criteria of DSM-IV and ICD-10, although useful for many purposes, reflect surface features of the disorder and may not provide the most appropriate phenotype for aetiological research. An underlying aetiological heterogeneity, anticipated by Kraepelin and Bleuler, is likely to be an important part of the explanation for the many non-replications of initially promising findings of genetic linkage and association studies conducted in recent years. But genetic heterogeneity is common in many diseases and schizophrenia is unlikely to be a special case in that respect. What makes schizophrenia refractory to the available methods of dealing with such heterogeneity is the confounding effect of a phenotype based solely on conspicuous symptoms and behaviours. These clinical manifestations may represent a ‘common final pathway’ for a group of brain disorders whose pathogenesis involves interactions of polygenic influences and environmental risk factors operating on brain maturational processes. Though many investigators and clinicians would readily subscribe to such a view, in the prevailing practice of schizophrenia research the diagnostic phenotype continues to be treated as a monolithic entity. This results in samples of cases and families that contain admixtures of different variants of the disorder, where disparate ‘biological signals’ tend to cancel each other, yielding disappointingly small effect sizes for any variable of possible pathogenetic interest.
The endophenotype approach
How to deal with this ‘double whammy’ of aetiological heterogeneity and phenotypic fuzziness? The strategies of ‘lumping’ or ‘splitting’ syndromes and diagnostic categories are a time-honoured approach in search for biologically meaningful disease entities. On the strength of the available evidence, schizophrenia is at present a good candidate for some splitting. However, multiple past attempts at rearranging the clinical symptoms into different principal components or clusters have achieved little – simply because symptoms are surface features, remote from the primary site of action of the putative biological causes. An alternative approach is provided by Gottesman's concept of ‘endophenotypes’, that is traits associated with brain function or structure, that are stable, objectively measurable, heritable, and independent of the clinical disorder but statistically correlated with its transmission within families.
At the time we were planning our research program, there were reports of several promising leads in this area, including the P50 event-related brain potential, the smooth pursuit eye movements, and some of the neurocognitive tasks of sustained attention which had performed surprisingly well in prospective studies of children at high genetic risk, born to mothers with schizophrenia. Our reading of the literature suggested that the cognitive deficits in schizophrenia, first described by Kraepelin as a core feature of dementia praecox, could be the prime target for an endophenotype approach. We constructed a fairly inclusive battery of putative endophenotypes, which comprised, in addition to neuropsychological tasks and event-related potentials, also measures of personality traits, soft neurological signs and structural MRI. We did not regard this set of measurements as a substitute for the assessment of schizophrenia symptoms but rather as a dimension complementing a fine-grain evaluation of the clinical phenomenology of the disorder. For the latter purpose we selected the Schedules for Clinical Assessment in Neuropsychiatry (SCAN), which was the successor of the WHO Present State Examination (PSE) interview, a well-tested instrument in multicountry clinical and epidemiological research. Our first NHMRC grant application, entitled ‘Exploring heterogeneity in schizophrenia’, was successful (as were three subsequent grants) and set our team on 7 years of uninterrupted, labour-intensive data collection. Anyone who has been involved in a similar kind of work involving patients suffering from schizophrenia and their families, will be familiar with the tribulations of the effort, which requires persistence, tact and discipline on the part of research staff, as well as understanding, motivation and patience on the part of those consenting to participate as research subjects. Apart from a small number of flat refusals, or of withdrawals after initial consent, the cooperation of patients and families in the project, which involved lengthy testing sessions and interviews, has exceeded our initial expectations. To date, we have enrolled some 200 families with one or more members affected with schizophrenia, as well as 130 normal controls (over 700 individuals in total). Of these, 116 families comprising 161 affected individuals and 298 first-degree relatives, have been evaluated on multiple measures of clinical symptomatology and neurocognitive performance, including attention, memory, speed of information processing and IQ; personality traits; neurophysiological indicators such as P50, P300, MMN, and saccadic eye movements.
Statistical modelling of a multivariate phenotype
A difficult question in any research design involving multidomain or multilevel measurements is how to treat statistically the many variables assessed. If the aim is to use multiple measurements of brain function (or dysfunction) in order to tap into complex interactive neural networks, then we need means of identifying composite patterns of dysfunction, rather than single, unrelated deficits. Such composite interactive patterns should be capable of accounting for a greater proportion of the variance than the sum of individual components. In the search for an analytical tool that could take us some way towards this goal, we settled on the grade of membership (GoM) model, developed in the 1970s by Max Woodbury at the Center for Demographic Studies of Duke University, North Carolina. GoM is not yet part of the ‘mainstream’ statistical armamentarium, but I had prior familiarity with its potential, owing to a collaboration with Woodbury dating back to my work at WHO, which had resulted in two published papers on the classification of schizophrenia syndromes. Basically, GoM is a derivative of latent structure analysis, which had been introduced in the social sciences in the 1950s by Paul Lazarsfeld, and is closely related to the better known latent class analysis (LCA). Like LCA, GoM partitions the data matrix into latent classes, or ‘pure types’ by iteratively computing multiple regression relationships among all variables and estimating the maximum likelihood fit to alternative models consisting of varying numbers of pure types. Unlike LCA, where latent classes are defined as crisp partitions, GoM pure types are fuzzy sets, allowing their members to be simultaneously represented on more than one pure type by grade of membership coefficients which are constrained to sum up to 1.0. Thus, an individual may be characterized as ‘expressing’ say, 0.45 of pure type A, 0.10 of pure type B, and so forth.
The attraction of the GoM model is that it mirrors the inherent ‘fuzziness’ of psychiatric classification, where boundaries between syndromes tend to be poorly demarcated or may not exist at all. Our general working model of schizophrenia is one of a phenomenological ‘generic group’ (Jaspers), possibly comprising several partially overlapping subtypes of underlying pathophysiology and causal pathways. GoM appeared to us as a tool that might enable the teasing out of latent patterns of cognitive deficit and the exploration of their genetic basis. We conducted consecutive GoM analyses (varying the input parameters) of large datasets of neurocognitive measurements and behavioural attributes, obtained from patients, their first-degree relatives and a number of control subjects. The statistically most parsimonious solution invariably yielded five pure types, of which two consisted entirely of normal controls and unaffected relatives, while each of the remaining three picked up a proportion of the patients with schizophrenia and their siblings or parents, of whom some were clinically unaffected and others had been diagnosed with either schizophrenia spectrum or other disorders. On examination, these pure types represented different configurations and degrees of severity of cognitive deficits involving attention, executive functions, verbal memory and IQ, as well as personality traits. We finally collapsed the three types into two contrasting groups: one with severe and pervasive cognitive deficit and another with less pronounced, patchy deficits but with marked deviations in the personality traits assessed. The patients falling into these two groups did not differ on duration of previous illness, which suggested that the two types were unlikely to represent different stages of progression of a single disease process. Moreover, the pattern of pervasive cognitive deficit showed significant familial aggregation (many of the clinically unaffected relatives exhibited attenuated deficits similar to those manifest in the patients), while the pattern of milder, patchy deficits did not.
Genetic validation of the neurocognitive endophenotype
The critical test for the GoM subtypes as endophenotypes complementing the clinical entity of schizophrenia should be a demonstration of their criterion validity – do they identify biologically discrete patterns? As a first step in a series of validation studies, we explored the genetic linkage of the two contrasting GoM groups to regions on chromosomes 6, 10 and 22 which had been previously implicated by other investigators as possibly linked to schizophrenia. Using a subset of our families, which were stratified for liability by a combination of clinical diagnosis and the composite GoM trait, we obtained suggestive lod scores (in excess of 2.0) on chromosome 6p for the pervasive deficit subtype, while no linkage emerged for the patchy deficit subtype. A parallel analysis of individual measures of deficit separately, for example attention dysfunction, verbal memory, or IQ, only resulted in low positive lod scores in the same region on chromosome 6. This suggested that the composite trait might indeed be assessing an interactive functional network [1]. Stronger supporting evidence emerged recently from the linkage analysis of the complete genome scan of an enlarged sample of our families, using a more refined version of the composite neurocognitive phenotypes. This time we obtained genome-wide significant lod scores in excess of 3.6 for the pervasive deficit subtype in a narrow region on chromosome 6p and on chromosome 10q. A complementary GoM analysis of chromosome 6, designed by Elizabeth Corder at Duke University, demonstrated that sibling pairs sharing the pattern of cognitive deficit also exhibit highly significant sharing of identical alleles for several markers on 6p. Lod scores of this size, coupled with converging identity-by-state allele sharing, justify genomic searches for specific genes that may be involved in causal pathways of disorder. Our preliminary screen of the human genome databases pointed to several plausible candidates within the two linkage regions, involved in synaptogenesis, dendritic pruning, and regulation of glutamatergic signalling. However, we are only at the beginning of a long new phase of research. Proving a causal role for a candidate gene is a formidable task requiring replication of the fist set of results on independent samples of families and case-control series, refinement of the region, positional cloning and sequencing of the candidate genes, and ultimately demonstrating a functional role for a mutation in brain tissue expression studies and animal experiments.
Neurophysiology and neuropsychology research
In the context of our overall research strategy of detailed, multidomain assessment of neurocognition in schizophrenia, a number of studies have been initiated by Pat Michie and Johanna Badcock, targeting specific deficits of working memory, inhibitory processes, and speed of information processing. Inhibitory deficits, frequently reported in schizophrenia, are associated with activities of prefrontal cortex and related networks. An understanding of intentional inhibitory control requires knowledge of how actions are planned and initiated and the executive components involved in stopping these actions. In a visual choice reaction time (‘go’) task, patients with schizophrenia, a comparison group with affective psychoses, and healthy controls attempted to inhibit their responses to the ‘go’ task when an auditory ‘stop’ signal was heard. Three movement-related cortical potentials preceding voluntary, self-initiated action were recorded. Compared to controls, both patient groups were impaired in their ability to inhibit a response across a range of stop-signal delays. However, the mechanism underlying this impairment in schizophrenia patients was markedly different from that of patients with affective psychoses and involved a selective difficulty in reliably triggering the inhibitory response, suggesting a dysfunction in the supplementary motor area.
Another study is focusing on the subgroup of ‘highfunctioning’ people with schizophrenia who show preserved intellectual ability, as assessed by IQ measures. A detailed characterization of the cognitive abilities of this group may be important with regard to the planning of appropriate treatment and management for schizophrenia patients with minimal cognitive impairment. Since it is widely believed that cognitive processing speed is the basis of individual differences in ‘intelligence’, the objective of this study was to stratify a sample of schizophrenia patients into subgroups with preserved, deteriorated and compromised intellect according to IQ assessment in order to examine whether these groups differ in specific abilities or general cognitive processing speed. The latter was measured with an inspection time task, designed in collaboration with Mike Anderson at the University of Western Australia (UWA) Department of Psychology. The patients who displayed no evidence of IQ decline, were found to show relatively better performance on measures of executive function, compared with patients in the deteriorated and compromised subgroups. The surprising finding was that their average inspection times were as slow as those in the deteriorated or compromised groups, indicating that high functioning patients are impaired in terms of cognitive processing speed. These findings suggest that schizophrenia patients with preserved general ability may nonetheless experience enduring cognitive difficulties related to speed of processing that need to be taken into account when planning rehabilitation interventions.
One of the priority areas of research at CCRN has been the study of event-related brain potentials (ERP) in patients with psychoses, their unaffected biological relatives, and ‘normal’ control subjects. Several studies, conducted under the supervision of Pat Michie over the past several years (including some excellent research by PhD student Juanita Todd), have resulted in novel findings, published in a number of journal articles and presented at international conferences. One of the most consistent findings in schizophrenia research over the past decade is a reduction in the amplitude of an early auditory event-related brain potential known as mismatch negativity (MMN). MMN is generated whenever a deviant sound occurs in a background of repetitive auditory stimulation. The reduced amplitude of MMN in patients with schizophrenia is most robust for deviant sounds that differ in duration relative to background sounds, but similar findings have been observed for sounds that are deviant in frequency. In a first investigation of its kind, Pat Michie and collaborators found that asymptomatic first degree relatives of schizophrenia patients showed a similar reduction in MMN amplitude to duration deviants (there were no significant differences between patients and relatives). These findings suggest that reduced MMN amplitude may be an endophenotype marker of the genetic predisposition to schizophrenia.
Prepulse inhibition and animal studies
Collateral research, conducted by Matt Martin-Iverson and several PhD students, is focused on the prepulse inhibition (PPI) of the acoustic startle reflex, which provides a model of the putative sensory gating deficit in schizophrenia. PPI is of interest because it is reduced in patients with schizophrenia, schizophreniform and schizotypal disorders, and may also be attenuated in posttraumatic stress disorder. Animal studies on treatment of neonate rodents with stress hormones have shown that activation of glucocorticoid receptors prior to development of the blood–brain barrier produces a schizophrenia-like reduction in PPI in the animals when tested as adults. A similar effect was produced by raising the rodents in social isolation during their juvenile stage of development. In addition, we found that PPI may be increased by cannabinoids in rats, but that this effect depends on the level of stress hormones. With low activation of glucocorticoid receptors (receptors for cortisol and corticosterone), cannabinoids increase PPI, but with high activation of these stress hormone receptors, cannabinoids decrease PPI. Clinical research at CCRN has shown that effects of chronic cannabis use in both patients with schizophrenia and healthy volunteers depend upon the attentional demands of the task performed during PPI measurement. Presentation of these data at the British Association for Psychopharmacology meeting won the PhD student Karina Kedzior the ‘best clinical research presentation’ award. The results support the view that cannabinoids may have a neuroleptic-like effect, and that psychosis associated with chronic cannabis use may be a ‘withdrawal supersensitivity’ effect, as observed with typical antipsychotic drugs. Currently, there is a debate in the literature as to whether or not PPI is differentially affected by typical and atypical antipsychotic drugs. Research at CCRN favours the hypothesis that deficits in PPI are absent in patients treated with atypical antipsychotics, but are present in those treated with typical antipsychotics.
Autism research
The development of this area of research at CCRN was largely driven by Joachim Hallmayer, who brought with him an already established framework of international collaboration that involved Stanford University and the University of Utah in what was to become a large, international data collection on families with two or more autistic children. There is compelling evidence that the causes of autism are genetic. The familial recurrence risk (for a second child to be affected, if the family already has an autistic child) is in the range between 2% and 7%, which translates into a 10- to 100-fold increase in risk compared to the general population. Case-finding for families with two or more autistic children included all of Australia. Two CCRN researchers spent several months on an interstate expedition which ascertained over 50 families with affected sib-pairs and a similar number of families with discordant sib-pairs. A full genome screen and genetic linkage analysis was conducted in 90 US and Australian multiplex families, with a follow-up in 49 additional families. Discordant sib-pairs served as controls. The analysis revealed excess sharing of ‘identity by descent’ alleles, which was estimated to result from involvement of a large number of genes (at least 15) on different chromosomes, each of small effect. These findings, published in 1999 [2], suggested an extremely complex genetic basis for autism. Clearly, positional cloning of such multiple susceptibility loci by linkage analysis would be a formidable task and alternative approaches will be necessary.
One such approach, similar to the strategy adopted for schizophrenia, is to identify and characterize specific neurocognitive impairments associated with autism, as well as other biological markers that could serve as correlated phenotypes. An ongoing study, involving collaboration with the Department of Psychology (Murray Maybery) and the University of Oxford (Dorothy Bishop), is targeting four areas of neurocognitive functioning shown to differentiate individuals with autism from matched controls. The data collection part of the study has just been completed and data analyses have begun. Another study, conducted by PhD student Emma Glasson, focused on obstetric complications in children with autism. A first step in the project was to establish a retrospective and prospective Western Australian register of pervasive developmental disorders, centralizing diagnostic information from various services. The next step was to link all the cases to the Maternal and Child Health Research Database (MCHRDB). Cases were compared to their siblings and to a population-based control group (n = 1313). Compared to the controls, the autism cases had significantly older parents, were more likely to be first-born, to have experienced fetal distress and to be delivered by caesarean section. In addition, she studied the atypical pattern of head growth that had previously been described in autistic children. In 366 cases of autism, she found that the majority (75%) had a head circumference greater than the population mean by half a standard deviation. Head growth varied with age, being normal at birth, accelerating during childhood and decelerating toward adolescence. It remains to be seen, however, whether neurodevelopmental indices, such as atypical head growth, could be used as physical phenotype markers.
Epidemiology of psychotic disorders and associated risk factors
Population-based epidemiological research is a major component of the CCRN work plan and is likely to remain so. In the last 7 years we have embarked on several major studies, some of which have already borne fruit, and others are steadily advancing to the data analysis and writing up stage.
The Low Prevalence Study
Following several years of discussion and consultation, a major decision was made by the Australian Health Ministers' Advisory Council in late 1994. The rational planning of mental health services, in accordance with the adopted National Strategy, required an epidemiological database, and such information could only be provided by a nationally representative survey. A core group of individuals, including Harvey Whiteford, Gavin Andrews, Scott Henderson and Wayne Hall, formulated the broad parameters of a survey that would encompass three interrelated, yet relatively independent studies: a representative national sample of the adult population, a catchment area-based investigation of psychotic (‘low prevalence’) disorders and a school-based sample of children and adolescents. I was invited to submit proposals for the design of the low prevalence arm of this national endeavour to the Steering Committee, magisterially chaired by Scott Henderson. It was to be a major undertaking. Intense discussions with Helen Herrman, John McGrath, Scott Henderson and many others helped toward agreeing on a plan that envisaged selection of geographical catchment areas in the Australian Capital Territory, Queensland, Victoria and Western Australia for a two-phase survey involving a census and screening of all contacts with the mental services and with selected general practices, to be followed by structured interviewing of a sample of people with psychotic illnesses. The CCRN was entrusted with the coordination of the multicentre research network, and our team, including David Castle, Vera Morgan, Anna Wattereus, and a number of field interviewers, worked over the next 2 years in close contact with John McGrath, Helen Herrman, Carol Harvey, Ailsa Korten, Mandy Evans, Vaughan Carr, Oye Gureje and many others.
The results of the Low Prevalence Study have been published in a major report and a number of journal articles. Apart from generating point prevalence data for psychotic disorders in urban areas in Australia (a weighted mean of 4.7 per 1000 aged 18–64), the study was unique in ascertaining symptom profiles, rates of functional impairments and disability, indices of quality of life, substance use comorbidity, service utilization patterns and side effects of medication. Subsequent economic analyses coordinated by Vaughan Carr provided estimates of direct and indirect costs associated with psychotic disorders.
One of the ‘best selling’ products of the study was the research instrument, the Diagnostic Interview for Psychosis (DIP), which was designed by our group with substantive inputs from all the investigators and many subsequent technical refinements provided by Vera Morgan, Anna Wattereus and Helen Stain. After the completion of the study, the DIP has remained in considerable demand nationally and internationally, and CCRN continues to provide training in its use through workshops conducted by Helen Stain.
Record linkage epidemiology
Advances in the past decade in record linkage software, combined with the availability of comprehensive population-based registers of long-standing in Western Australia, have provided us with access to a precious research resource that is practically unique for Australia and comparable to the population databases existing in Scandinavian countries.
In recent years, we have embarked on several record linkage studies of psychiatric morbidity, with a major focus on psychotic disorders. While this reflects to some extent the research interests and expertise of the investigators, it is also a response to the very large burden of disability associated with psychotic illness, its emotional and financial costs, as well as the strain it imposes on the health care system.
Mortality and medical comorbidity in the mentally ill
The first series of studies was a collaboration with the UWA Department of Public Health and involved linkage between the psychiatric case register (the Mental Health Information System, or MHIS) and the information-rich WA hospital morbidity and mortality registers. The objective was to study the distribution of the entire range of mortality and physical morbidity among people with mental disorders over the last 20 years, and to better inform public health programs of the health status of this population. Key figures in this research were David Lawrence, a PhD student jointly supervised by D'Arcy Holman and myself, and John Bass, who designed and supervised the electronic linkage of data. The findings, published in several journal papers and a major report, were not unexpected but the totality of the picture was disturbing. We found that the physical health status of psychiatric patients was exceptionally poor, people with schizophrenia and bipolar disorder faring worst. Compared to the general population, there were significantly higher rates of mortality from all causes of death (but particularly suicide and ischaemic heart disease) and clear evidence of suboptimal medical care. While mortality from ischaemic heart disease in the general population had been decreasing appreciably over the last two decades, psychiatric patients were an exception from this trend. Their death rates from heart disease had remained stationary or, in women with mental illness, increased substantially. Our analyses indicated that the root of the problem was not so much under-detection or missed diagnoses but under-treatment of disease. For example, patients with psychotic disorders and diagnosed heart disease had much lower rates of revascularization procedures than people with similar problems but without a diagnosis of psychosis. The ethical and service organization implications of this series of studies were compelling, and we made several representations to the Minister for Health which were received with due concern, though the results of all that are yet to be seen. Overall, these studies demonstrated the power of record linkage epidemiology to highlight an often neglected dimension of the problem of mental illness that, in principle, is tractable and amenable to intervention.
Reproductive pathology in women with schizophrenia and affective disorders
The second wave of studies, capitalizing on the capacity of record linkage across population databases, was focused on the problem of early neurodevelopmental lesions that may be contributing to the risk of schizophrenia and major affective disorder. Many studies have suggested that complications of pregnancy and delivery, as well as peri- and post-natal events, such as neuroinfection and trauma, may be critical risk factors enhancing the genetic susceptibility to psychotic disorders that become manifest later in life. The evidence to date, however, is inconsistent. Many of these studies were retrospective, depending on maternal recall of obstetric complications. Several studies involving prospective assessments of birth cohorts have produced equivocal or contradictory results. We felt that linkage across the prospectively collected WA databases could provide us with an opportunity to avoid some of the methodological traps. The paradigm chosen was to examine the existing data on pregnancies and offspring outcomes of all women with schizophrenia and affective disorders in Western Australia during 1980–2000, as compared to a control sample of women without any diagnosed psychiatric disorder. The offspring of women with psychotic disorders could be assumed to be at an increased genetic risk, and any severe obstetric complications occurring in this cohort could be expected to increase the incidence of psychosis in the offspring over and above the actuarial risk attributable to the genetic background. With Vera Morgan at CCRN and Steve Zubrick and Carol Bower at the Telethon Institute for Child Health Research, we embarked on a large project (funded by the Stanley Foundation, US) which is still in progress. We identified all women with diagnoses of schizophrenia or major affective disorder who had given birth since 1980 and linked them to the research register of pregnancies (the Maternal and Child Health Research Database). Using the standardized scale of McNeil and Sjöström to score obstetric complications (for which a computer algorithm was written and validated by PhD student Li-Anne Yelachich), we found that both women with schizophrenia and women with major affective disorders experienced a higher rate of obstetric complications than the non-psychiatric control women in our sample. One of the striking findings was the significantly lower birthweight (adjusted for gestation) of the babies born to women with schizophrenia, as compared to the offspring of women with affective disorders and the control mothers. Even more importantly, we found that, in women with schizophrenia, an increased rate of obstetric complications tended to occur mainly in pregnancies conceived after the onset of psychosis. This finding argued against the hypothesis of a primary (presumably genetic) predisposition to obstetric pathology in women with schizophrenia and pointed to risk factors that may result from the actual experience of psychosis, with its attendant psychosomatic turmoil, changes in lifestyle and exposure to pharmacological intervention (in a collateral study, we are gathering data on the treatments prescribed during those index pregnancies).
Developmental outcomes in children at high genetic risk
In the second part of the study we are reconstructing the developmental histories of the children from birth to early adulthood, by sifting through a wealth of prospectively recorded paediatric, genetic, general medical and psychiatric information. Some of the early findings highlighted an unexpectedly broad spectrum of morbidity and developmental aberrations, such as a sixfold increase in the relative risk for learning disability among children of women with schizophrenia and a fivefold increase in epilepsy among children of women with bipolar disorders. In addition, there was a higher than expected incidence of some rare congenital syndromes and pervasive developmental disorders such as autism and Rett syndrome. As this cohort is now entering the age range of the highest risk for onsets of early psychosis, we will soon be in a position to test the key hypothesis about an interaction between increased genetic risk for psychoses and complications of pregnancy and delivery. Moreover, the availability of prospective data on the entire previous history of these children should enable us to link longitudinally multiple variables into developmental pathways leading from risk exposures to disease, as well as pathways that do not result in disease despite the presence of risk factors. The latter may provide clues to the phenomenon of resilience and protective factors.
Criminal offending and schizophrenia
In a third study we use record linkage to examine criminal offending in individuals with a psychiatric disorder. Since stereotypes and stigmatizing attitudes associating schizophrenia with criminality and dangerousness still prevail in the public perception, we need valid data to inform the justice system, the mental health services, as well as public debate. In addition to obtaining population-based data on the risk of criminal offending associated with specific diagnostic groups of psychiatric disorders, we are specifically interested in the temporal relationship between the age at onset of offending behaviour and the age at onset of schizophrenia as compared with age at onset of other psychiatric disorders. Western Australia has a comprehensive recording system of offences that dates back to 1946. In this study we have to date linked the psychiatric records of some 220 000 people on the MHIS with close to 390 000 individual records of various offences on the research database created by the UWA Crime Research Centre. We found that 24% of the individuals on the MHIS had at least one contact with the criminal justice system. Examination by diagnostic group revealed, predictably, that the highest rates of recorded offending were associated with drug and alcohol-related disorders (49%) and personality disorders (39%). Persons with diagnosed schizophrenia also had a higher rate of offending (32%) than the average for all psychiatric patients but the proportion of violent crime attributable to schizophrenia on a population basis was a small fraction of the total. Analysis of data on three birth cohorts (1955–1959; 1960–1964; and 1965–1969) indicated important trends. In a high proportion of the offenders with any psychiatric diagnosis, the history of offending preceded by several years their first contact with any mental health service on which the diagnosis was first made and treatment initiated. However, this temporal gap between the onset of offending and the detection and treatment of psychiatric disorder showed a steady and significant overall decline across the three birth cohorts. The only exception from the trend toward an earlier detection of psychiatric disorders was schizophrenia, where the gap between first offending and diagnosis had actually been increasing over time. This widening of the diagnostic and treatment gap was mainly attributable to a progressively earlier age at the onset of offending behaviour in individuals who later became diagnosed with schizophrenia – a phenomenon which had no parallel in any other diagnostic group. Further analyses will explore alternative hypotheses that may explain this finding, including the possibility of an association of offending with an increasingly earlier exposure to street drugs. Whatever the ultimate explanation, it is likely to have important implications for the planning of services and preventive programs of this vulnerable population group.
Intellectual disability and psychiatric illness
Work has commenced on a linkage between the Intellectual Disability Register and the Mental Health Information System in a collaboration involving the CCRN, the Telethon Institute for Child Health Research and the Disability Services Commission. The study will result in a statewide database identifying persons with a comorbid psychiatric disorder and intellectual disability, thereby allowing systematic investigations of this population that have not been possible to date. Findings from other studies, some dating back to early 20th century, indicate not only an overall increase in the rate of psychiatric disorders among individuals with intellectual disability, but also some very specific associations. In particular, there appears to be an elevated risk of schizophrenia, with recent studies suggesting that 3% of individuals with intellectual disability will have comorbid schizophrenia. At present, not enough is known about the nature of the association between intellectual disability and schizophrenia, and there are a number of candidate aetiological models. The aim of the study is to determine the spectrum of comorbid psychiatric morbidity among persons with mental retardation and conversely, the prevalence of comorbid intellectual disability among persons with a primary diagnosis of psychiatric disorder in Western Australia, as a basis for testing hypotheses related to specific mental disorders (in particular, schizophrenia).
Current state and future prospects
Several personnel changes have occurred at CCRN in the last 2 years. Two of our key investigators have left the centre to occupy leadership positions elsewhere – Joachim Hallmayer as Associate Professor at Stanford University and Pat Michie as Professor and Head of the School of Psychology at the University of Newcastle. Their contributions to the scope and quality of research at CCRN have been tremendous and colleagues of this calibre are difficult to replace. However, two highprofile researchers from Germany – Dieter Wildenauer and Sibylle Schwab – have recently decided to join our group in Perth. Their arrival will provide a strong impetus to our neurogenetics research, which is likely to focus increasingly on genomic studies in the next few years. Johanna Badcock, with her strong background in experimental psychology, has taken over a major supervisory role with regard to neuropsychological assessments, and Greg Price has recently been appointed as manager of the electrophysiological laboratory. Danny Rock, a senior research nurse and a PhD student, is providing support across the board in his role of clinical research manager. We have an active program of postgraduate research, closely linked to the major studies currently in progress at CCRN. At present, there are 22 PhD students involved in such research, and during the last several years nine have successfully completed their doctorates (of them, two ‘with distinction’).
The limits of this overview do not permit a truly comprehensive account of all research conducted at CCRN. For example, interesting work on functional neuroimaging; the study of phospholipid membrane metabolism using the niacin skin reaction in schizophrenia patients, biological realtives and controls; PhD studies on seasonal patterns of violent behaviour and suicide, mixed-handedness and laterality in schizophrenia; or on chromosomal anomalies and possible X-linkage within the autistic spectrum of disorders, must remain with only a brief mention here. Nor can all the researchers and supporting staff that have provided valuable input to our research over the years be listed. Such information, as well as a complete bibliography of publications and PhD theses, is available on request.
On a concluding note, I find with some satisfaction that, within a relatively short period of time, the CCRN has consolidated its infrastructure and made considerable advances in research areas such as molecular genetics, neurophysiology, neuropsychology, behavioural pharmacology, diagnostic assessment procedures and record linkage epidemiology. Several major projects, referred to above, have generated a wealth of novel data converging on a working model of schizophrenia as a complex, heterogeneous disorder, that may eventually yield to an integrated multidisciplinary approach. We are planning to increase further the size of our sample of families with schizophrenia and to develop an epidemiological sampling frame for case-control studies. An important priority will be the careful translation of some of our research findings – especially those related to the neurocognitive subtyping of schizophrenia – into prospective clinical trials of selected interventions. But an overriding concern is that future, hopefully groundbreaking, research into the nature of schizophrenia will require large-scale data collection that can only be achieved through extended collaborative networks. Both the potential and motivation for an Australia-wide collaboration in this respect are stronger than in many other parts of the world. The success of the Low-Prevalence Study, one of the largest national collaborations in the study of psychotic disorders ever undertaken, provides an ample illustration of this. Jointly with several research groups across Australia, we made in the last couple of years two attempts to obtain NHMRC program funding for a major concerted effort that would generate a national database on the genetic epidemiology of schizophrenia. Unfortunately, these attempts were unsuccessful, and, in my view, an important opportunity for Australian research was missed. Yet the prospect of creating such a network should remain high on the agenda.
Acknowledgements
Our thanks go to Vera Morgan, Matt Martin-Iverson, Johanna Badcock, Joachim Hallmayer and Danny Rock for assistance in preparing this overview.