Abstract
Objective
To describe the characteristics of schizophrenia relevant to conducting indicated preventive interventions.
Method
A systematic review of the literature informed by experiences at the Personal Assistance and Crisis Evaluation (PACE) clinic.
Results
Primary prevention requires a sophisticated knowledge of key causal risk factors relevant to the expression of a disorder. The causal risk factors most useful from an intervention standpoint may turn out to be somewhat removed from the neurobiology of the disorder and may even be relatively non-specific, so that tackling them could reduce the risks for a range of mental disorders. The frontier for more specific prevention in schizophrenia and related psychosis is currently represented by indicated preventive interventions for subthreshold symptoms. Again, these may be relatively broad spectrum early in the prepsychotic phase but more proximal to onset, greater treatment specificity can be explored. However, this can be viewed more as preventively orientated treatment rather than primary prevention per se. Early detection of first episode psychosis and optimal intensive treatment of first episodes and the critical early years after diagnosis also represent increasingly attractive preventive foci in psychotic disorders.
Conclusion
As evidence accumulates, implementation of evidence-based practice in real work settings is a major challenge as it is throughout the mental health service system. The momentum of preventively orientated treatment must be maintained through the 2nd National Mental Health Strategy and in the face of recent misleading polemic regarding the treatability of psychotic disorders, especially schizophrenia. The evidence demonstrates that schizophrenia and related disorders have never been more treatable.
The Global Burden of Disease report revealed that several mental disorders feature in the top 10 contributors to health burden, with schizophrenia, bipolar disorder, depression and the psychotic forms of these disorders making a large contribution to disability around the world [1]. However, it is only in the last 5–10 years that the idea of early intervention and prevention of schizophrenia and psychotic disorders has seriously surfaced. Over a relatively brief period, preventive interventions have gained a large amount of momentum. Although principally focused on prevalence reduction rather than incidence reduction, there is increasing confidence that significant progress can now be made.
However, as recently as 1997, DeLisi published a review entitled 100 Years of Dementia Praecox, in which there was very little mention of prevention or even therapeutic optimism [2]. This review reflected how the possibilities for prevention over the last one hundred years have been hampered by the manner in which the disorder is conceptualised. In fact, Kraepelin's concept, which brought together a whole range of disparate syndromes, on the basis of poor outcome and failure to recover, to comprise schizophrenia, does not leave much room for thinking preventively about the disorder. These days, it is better appreciated that there are a range of presentations of the disorder with a range of associated outcomes that are by no means fixed. In 1999, Andreasen pointed out that the early course of schizophrenia was the most fulminant phase and the period of greatest severity and disability [3]. In fact, after that, things improve for the great majority of patients, although many people do continue to suffer a reduced quality of life [4].
One of the problems with schizophrenia is that researchers still tend to think that the prototypical disorder actually exists as a single disease entity, even though the illness consists of many syndromes or descriptive categories and almost certainly has a myriad of underlying aetiopathologies. When a patient presents with a first episode of psychosis, only approximately 55–60% will meet the criteria for any form of schizophrenia. After another 3 years, that percentage will have changed with additions from other categories, while conversely, others move out of the schizophrenia spectrum [5]. These facts need to be taken into account when looking at risk factors (for onset, recovery and remission) and prevention of the disorder.
Mrazek and Haggerty provided a more modern framework for the prevention of mental disorders, describing three categories: universal, selective and indicated prevention [6]. These authors highlighted that, in the case of schizophrenia and psychotic disorders, the best we can aim for at this stage is indicated prevention:
Indicated preventive interventions for mental disorders are targeted to high-risk individuals who are identified as having minimal but detectable signs or symptoms foreshadowing mental disorder, or biological markers indicating predisposition for mental disorder, but who do not meet DSM-III-R diagnostic levels at the current time.
Kraemer et al. clearly defined what constitutes a risk factor or a marker and described a framework for prevention [7]. They highlighted that the key criterion for a risk factor is that it precedes the outcome of interest. If it does not, it is simply a correlate, a consequence or a concomitant of the outcome. Therefore, first, the outcome of interest needs to be defined. For example, is it the first psychotic episode, or is it the first time the patient met the criteria for the diagnosis of schizophrenia? What then happens if the diagnosis changes to bipolar disorder for example? It is possible to study the onset process and calculate risks of transition for different descriptive diagnostic targets. Schizophrenia needs to be considered as a more distal predictive target than psychosis because it becomes a more stable category with the passage of time. However, multiple syndromal targets are possible from a predictive standpoint.
The next step is to examine the statistical significance and gauge whether the association is by chance or is statistically meaningful. If it is statistically significant, questions need to be addressed that relate to the potency of the different risk factors in terms of odds ratios or relative risks and whether the influence of the risk factor varies (i.e. does it change over the course of the person's life?). For example, gender and season of birth are fixed risk factors, while cannabis use and stress levels would be considered variable risk factors. Within this group of variable risk factors, can it be shown that altering the level of the risk factor in turn alters the risk of the disorder? If so, it can be considered a causal risk factor. If not, it should be considered a variable marker. The key point is that markers, both fixed and variable, are very useful for defining populations at risk but are not as useful for defining interventions to reduce the risk. Causal risk factors need to be identified to accomplish this goal. An example comes from the HIV/AIDS area: while HIV is the fundamental cause of the disorder, needle sharing, for example, is a causal risk factor. Successful intervention in this area has been shown to reduce the risk for the disease.
Mrazek and Haggerty also emphasise the need to define the greatest points of purchase for different disorders in terms of prevention and the phases of life in which they might operate [6]. They stress that risk factors may be differentially malleable at different phases of life and that there are sensitive periods when risk factors contribute most to aetiology. For example, in the early course of psychotic disorders, it is known that, due to the developmental phase of those experiencing the disorder, there is an increased risk of developing disability, which may be able to be influenced [8,9]. Mrazek and Haggerty also note the need to define prospectively the length of prodromes (the specific features distinguishing them from normal behaviour) and precursors conferring the heightened risk for developing the full-blown disorder [6]. Such knowledge really represents the bedrock of indicated prevention.
Eaton and Harrison suggested that the bulk of prevention efforts for some disorders, depending on incidence rates, should be focused around the onset stage [10]. Cook and Sackett published a formula that provided the answer to the question ‘how many people need be given a particular intervention for a particular disease, to prevent one negative outcome?’ [11]. This is the ‘number needed to treat’ (NNT) statistic, which can be calculated from clinical trial data and has been used extensively in the Cochrane collaboration. If the intervention is very powerful and effective, then the number of people that need to be treated to prevent one adverse outcome is low. If the intervention is weak or diffuse, then that number goes up substantially. However, even if the risk factors are not specific and the intervention only has a weak effect on the specific disorder in question, there may be other disorders for which the risk could be reduced. An example of this is childhood sexual abuse. Hypothetically, if the level of that risk factor could be reduced, this may prevent a small number of cases of psychosis, but a much larger number of cases of borderline personality or depression might be averted.
When examining potential risk factors for psychotic disorders, investigations may need to go back to birth, or even earlier. Woods examined the evidence supporting neurodevelopmental versus neurodegenerative models of schizophrenia [12]. While the neurodevelopmental model has been the dominant paradigm in relation to schizophrenia in recent years, evidence has been emerging recently in support of a neurodegenerative or neuroprogressive process, albeit as a putative secondary phase. The original neurodevelopmental model proposed that the members of the future schizophrenia population are different at birth in some way and diverge during the period at risk because of forces such as anomalous brain development and, ultimately, cross the threshold leading to the manifestation of their disorder. This is a very deterministic model and most unlikely to be correct. It is more likely that patients developing schizophrenia are vulnerable on a variety of levels (largely, but maybe not exclusively, biological) with risk factors triggering a precursor syndrome. The precursor symptoms are less tightly tied to the disorder than the concept of prodrome (which really means the early phase of disorder is usually only diagnosed retrospectively). Other risk factors could then be seen as having the capacity to translate the precursor syndrome into the fully fledged diagnosis or disorder. It is most unlikely that people who develop schizophrenia have a ‘built-in computer virus’ that they will express at some point in time in the manner of a time bomb or the dreaded Y2K bug. The vulnerability must be more widely distributed in the population, perhaps not across the whole population but certainly encompassing a larger subgroup than those who actually express it. Indeed, the vulnerability to isolated psychotic symptoms of uncertain clinical significance seems to be relatively widespread [13]. Those people who do succumb to the disorder express it in a variety of ways, depending on their exposure to different risk factors and other pathoplastic variables.
A variety of risk factors and markers for schizophrenia are currently being studied. Some of these include:
The effects of famine on women during pregnancy. Susser et al. noted that famine increases the risk of schizophrenia [16].
The issue of delayed childhood development and childhood abnormalities and whether these are markers or risk factors [17].
Minor physical anomalies. Difficulties in measuring this marker exist, however, the methodology is improving [18].
School performance and social class. It has been suggested that at least for a subgroup, high social class, as well as low social class, is associated with increased risk [19].
Urban/rural birth. Mortensen and colleagues reported that although genetic factors have the highest relative risk, a weaker factor such as urban birth, because of its wide distribution in the population, has a more important population-attributable risk than does genetics [20]. This is a watershed finding. A similar phenomenon is responsible for the fact that depression, generally a less severe illness, is responsible for a greater proportion of the total burden of disease than is psychosis.
Family environment. There is some evidence suggesting that young people who are at increased genetic risk of schizophrenia have an increased risk of expressing it if they are reared in a dysfunctional family environment [21].
Substance abuse. Evidence suggests that cannabis use seems to be associated with an approximate doubling of the risk of onset of schizophrenia [22].
Stress. Evidence suggests that stress would be a very good candidate risk factor for onset and also for intervention [23].
Perinatal brain damage. The population attributable risk of this particular variable has been reported at 7% [17].
Therefore, although genetics provides information about the underlying basis of the disorder such that it may be possible to intervene at the genetic level at some point, at the moment genes constitute fixed risk factors or fixed markers. Until they can be manipulated, they are not of much practical use. Furthermore, the failure to find a clear genetic pattern for risk has proven to be a major disappointment in schizophrenia research. Conversely, understanding how the non-genetic risk factors contribute to the way schizophrenia is expressed may provide insights into reducing the risk. Currently, clinical epidemiology is the most appropriate scientific model guiding the prevention of these illnesses.
Van Os et al. reported that the same markers and risk factors that may influence onset might also affect the course of the disorder (e.g. premorbid functioning, ventricular volume, genetic risk, life events or ethnicity) [24]. Indeed, these risk factors can be shown to operate in both areas. Further, they noted that there was little evidence that any risk factor was specific to any diagnostic category within the functional psychoses. However, the pattern and severity of risk factors does vary as a function of baseline psychopathology. That is, some risk factors have stronger associations with, for example, non-affective psychosis as compared with affective psychosis.
The question of how to predict within a clinical population who exactly is at increased risk of an early transition to psychosis is a critical one. The uniform pattern has been that, on average, the first symptom (as with all major mental disorders) precedes the first diagnosis (or the first reaching of the diagnostic threshold) by a few years. Häfner et al. highlighted that the difficulty of the clinical task in schizophrenia was that the initial symptoms were very non-specific [8]. This is related to the broader issue of comorbidity in psychiatry and the essentially dimensional nature of psychopathology. The symptoms are generally negative or depressive and have very little specificity at the earliest stage. It is, therefore, difficult to aim to prevent one particular disorder and this suggests the potential value of a broader youth mental health focus. For example, we carried out a survey of DSM-IIIR prodromal symptoms in a high school population and found approximately 10–15% of senior high school students could have been regarded as prodromal for schizophrenia, if these symptoms are indeed genuinely prodromal (which is most unlikely) [25]. There is clearly a need for greater specificity in prediction, something which is impossible in a single-step process. While failing adjustment in a previously functional young person or a drop from a previous level of functioning is a first level of screening, there appears to be a need for a second level of screening, with a range of disorders as targets for the second screening step.
Dr Allison Yung, Ms Lisa Phillips and I, along with other colleagues, have developed the Personal Assistance and Crisis Evaluation (PACE) Clinic in Melbourne, which is conducting a series of studies on young people who have been referred to the clinic. We are trying to improve the prediction of transition to psychosis and also trying to develop interventions for this phase of illness. The studies are based on the idea of sequential screening, using a combination of markers to define the group at increased risk. One of the main issues encountered has been defining the exit point. That is, when is a person regarded as psychotic and when does a person actually meet the threshold for diagnosis? To some extent, this is arbitrary because many patients dimensionally slide into psychosis over a period of time. It is generally a slow process, through which an operational definition of psychosis is ultimately attained. At the PACE clinic, this threshold was empirically defined as when neuroleptics were indicated on a clinical basis: arbitrarily, 1 week of sustained positive symptoms.
Studies conducted at the PACE clinic have examined some historical and neurodevelopmental variables, particularly measuring obstetric complications, developmental milestones and childhood behaviour, to see whether they predicted an increased risk of developing sustained psychosis. Results have indicated that this was not the case, probably because they are general risk factors or markers for a range of psychopathology that was also present in the PACE sample. That is, those patients who remained nonpsychotic generally met criteria for other axis I disorders. However, older maternal age (> 30 years) was found to be associated with an increased risk of becoming psychotic within the next 12 months [Crump N. unpublished data].
Further, neurocognitive and structural central nervous system variables have been studied with a range of neuropsychological testing and neuroimaging studies carried out with the assistance of the patients of the PACE clinic, in collaboration with colleagues from the Mental Health Research Institute of Victoria. It was expected, in line with the neurodevelopmental theory, that volume reduction within the hippocampus would help to predict who might be at greater risk of transition to psychosis. In fact, the (statistically) significantly larger hippocampi were associated with a greater risk of transition to psychosis. This counter-intuitive finding is now the subject of further research, which has subsequently shown that progressive volume change in the hippocampus is associated with the appearance of psychotic symptoms. Neurocognitive measures have been studied by Brewer [26] in a cross-sectional study and the results support the idea that the PACE group (which contains 40% of people who are at increased risk of early transition) have abnormalities halfway between normal controls and first episode psychosis patients [26]. The predictive analysis of these data for transition, however, has not proven useful to date [26].
The PACE clinic has also undertaken a randomised control trial that evaluated whether it was possible to reduce the risk (attenuate the severity or delay the onset) of psychosis in young people at high risk of schizophrenia and similar disorders [27]. The trial has been underway for more than 2 years and recruitment is completed with approximately 30 people randomly assigned to one of two groups. The first group (controls) received supportive psychosocial treatment, including the use of antidepressants where necessary. The second group (treatment) received a more specific intervention package, including low-dose risperidone for 6 months and a cognitive intervention (developed at the PACE clinic) for at least 3–6 months. This group was also allowed to have antidepressants if major depression was present. Results from the 6-month treatment phase have shown that three of 31 people in the treatment group (approximately 10%) have become psychotic, while 10 of 28 people in the control group (approximately 36%) became psychotic. When examining the data in terms of ‘compliance’, it was found that only one person from the treatment group became psychotic while receiving the medication and attending for the psychological interventions. The more-specific forms of treatment, cognitive therapy and novel antipsychotic medication therefore seemed to offer an advantage, at least in delaying psychosis. Interestingly, the group who refused randomisation to the trial has also had a reduced rate of transition (although, of great concern, two of these people committed suicide). There is another five-site study underway in the US looking at this issue, using olanzapine and a somewhat different study design. This highlights the increasing amount of interest in this field and the desirability for patients to receive specific treatment at this stage. However, it may well be that some patients will respond to psychosocial treatment alone or to other forms of biological therapy.
Another important issue to consider is early detection and the idea that shortening the duration of the declared illness (in other words, shortening the duration of untreated psychosis) might be a very useful preventive strategy. Wyatt reported that patients who received delayed neuroleptics, in particular, did much worse than patients who received neuroleptics earlier in the course [28]. It is, however, a complicated issue as sceptics note that people who do worse tend to present later and the two variables may be confounded [29]. It is true that symptom changes in people who have a more insidious onset of the disorder are less likely to be as salient to themselves or others and therefore these people are more likely to present and receive treatment at a later stage. However, insidious onset cannot fully explain the relationship between the duration of untreated psychosis and outcome. In any case, delayed treatment for psychosis might be the very mechanism whereby insidious onset produces a worse prognosis. This area, in particular, requires further investigation to first, determine what the strength of the association would be and second, determine whether the duration of untreated psychosis is reduced and results in an improvement in outcome.
If we examine statistically the amount of outcome variance explained by a delay in treatment alone, the maximum is approximately 16% for 12-month outcome. If we put other predictive variables into a regression analysis first, the outcome variance that is accounted for by delayed treatment reduces to approximately 10%. This is still quite a reasonable amount of outcome variance, so it is worth pursuing [30].
A comparison of patients treated at the Early Psychosis Prevention and Intervention Centre (EPPIC) before and after the development and implementation of a more intensive psychosocial program led to the conclusion that there was some effect of an enhanced treatment program if the illness was detected within 6 months [31]. If the duration of untreated psychosis was less than 1 month, patients did well even with the less-intensive program. It is possible that many of those patients would have had a good outcome irrespective of what was provided and at least some may have remitted without treatment. However, if patients had a duration of untreated psychosis of 6 months or more, even the more intensive psychosocial program did not improve their outcome. Therefore, if the duration of untreated psychosis was from 1 to 6 months, there was a significant improvement in outcome with more intensive phase-specific intervention. It must be noted that there were methodological weaknesses in this study (i.e. it was not controlled or randomised), however, the results are suggestive that some people can be returned to a better level of functioning with early and optimal intervention.
The TIPS project aims to reduce the duration of untreated diagnosis by delivering a public education campaign focused on early detection of psychosis (with presentations on television, in cinemas and in newspapers) for a small geographical area in Norway [32]. This study does not actually target risk factors but aims to facilitate early detection and intervention. The strategy could be considered as targeting the risk factors for delayed treatment. A similar type of intervention has also been conducted in Australia, however, the focus here has been antistigma rather than early detection. In addition, the EPPIC program has used the Norwegian approach in a limited way by initiating a health promotion campaign in a section of the clinic's catchment area. While the campaign did not actually reduce the duration of untreated psychosis, it did seem to change the make-up of the patient sample by finding more long duration cases that might never have been found otherwise. Hence, the composition of the sample was altered. Other strategies to reduce delays would be the education of general practitioners, having a more accessible mental health service and providing more acceptable pathways for people presenting for the first time.
These approaches have clarified that there is a total annual incidence of psychosis in a particular catchment area, which is classified as N, whereas only a subset, n, of this number is actually treated. Early detection methodology has assumed that N is treated, which is clearly incorrect [33]. Even in Iceland, where the population is only 250 000, Helgason acknowledges that quite a significant percentage of people who would have met diagnostic criteria were never treated [34]. The variability of the gap between N and n cannot be measured very easily. It seems that early detection campaigns not only reduce the duration of untreated psychosis, but also increase the n. If this increase in treated cases is made up of a different mix of duration of untreated psychosis cases, it will prove very difficult to interpret this design. This means that if we really need to definitively answer the question of whether early versus delayed treatment improves outcome, an ethical way of doing a randomised study must be reconsidered [27].
A final critical issue when considering prevention is the content of early treatment itself. That is, providing more intensive, phase-related, specialised treatment in the early years of illness. McGorry et al. have reported that providing this type of treatment leads to improved outcome [35]. It is also a cost-effective method, as the funding saved from inpatient costs can then be directed back into the community-based elements [36].
The need to develop new interventions and treatment technology is great, particularly in the area of adolescent and youth psychiatry, as many of the treatment methods currently being used are out of date or relatively crude. New technology and interventions are needed to help young people recover and deal with not just the illness but the social conditions that they have to cope with in today's environment. So, when we examine possible signposts for the prevention of schizophrenia and other psychoses, there is a need for precision in terminology and methodology and a need to focus on risk factors for onset, recovery and remission. With the current state of knowledge, it would seem that the emphasis should be placed on indicated prevention, although a larger scale risk factor-orientated mental health promotion campaign could also produce indirect benefits for psychotic disorders. We also need to systematically implement better systems of early detection and optimal early treatment to go beyond the first generation of reform.
Finally, we have entered an era in which schizophrenia and other psychoses have become more amenable to treatment than ever before. The paradox is that these treatments are still failing to adequately reach a proportion of people (the minority) with the most severe forms of these illnesses, an example of the ‘clinician's illusion’ [4,37]. There is a failure of and a delay in knowledge utilisation combined with the ‘clinician's illusion’ and this underpins deliberately misleading and cynically contentious interpretations of flawed data, such as the 1994 metanalyses of Hegarty et al. [38] and Andrews [39].
There is an overdue appreciation that other psychiatric disorders require much greater attention because, although they are generally less severe, their high prevalence means that they cause widespread distress and extensive burden. Logically, this should lead to a substantially increased share of the total health-care budget for psychiatric disorders [39, 41]. Andrews has argued that this is ‘too hard’ and has misleadingly and wrongly characterised schizophrenia as ‘untreatable’ [39]. He goes on to argue for the soft option of reallocation of much of the funding for these illnesses to the putatively more treatable anxiety and mood disorders. This specious argument is based on deliberately comparing apples with pears. Andrews essentially contrasts efficacy data for anxiety and mood disorders with effectiveness and real-world studies for schizophrenia and psychosis [39]. A fair contrast would be to compare efficacy studies for both. Further, the concept of schizophrenia selects in the treatment resistant subgroup of cases, whereas concepts of mood and anxiety disorder do not have this confusion of prognosis and diagnosis. This line of argument is dangerous and misleading. It has the potential to promote division within the mental health sector, already weakened in contrast to other sectors with the health field, to derail the Second National Mental Health Strategy and, further, to increase the suffering and mortality associated with psychotic disorders in Australia. It represents a new twist to the Kraepelinian curse and it poses an immediate threat to the potential of prevention approaches in this country.