Abstract
Spiced with a little polemic and philosophy, Rosenman and Anderson (this issue p.509–514) reprise the tenacious question of the validity of psychiatric diagnosis and the relationship between clinical phenotypes and putative underlying ‘processes’. Having spent most of the 1980s immersed in this issue myself, trapped in a maze of historical literature and beguiled by the ghosts of schizophrenia past, I read this paper with a real sense of nostalgia. The novel angle here though is the contention that the fortunes of the early intervention paradigm, and specifically prodromal or pre-psychotic intervention, depend on an adherence to a narrow 19th-century disease model of psychiatry, which rigidly equates syndromes with underlying disease entities: a model which, although fatally flawed, dies very hard. The crux of the authors’ argument is that ‘prodromal diagnosis’ as they characterize it, is an invalid extension of an invalid model. They assert that a hypothetical, heavily deterministic and as yet undiscovered ‘process’ is seen by proponents of pre-psychotic treatment as a necessary target for interventions aimed at reducing risk of progression to a more serious stage.
The authors’ concerns are similar to those which led Alison Yung and I to eschew at the outset the prodromal label for this pre-psychotic or sub-threshold stage and to coin the new term ‘at-risk mental state’ (ARMS) [1] and later the ‘ultra-high risk’ (UHR) state [2,3]. This was intended to capture the high level of risk for progression conferred by the phenotype we described and validated in prediction studies, but on the other hand to underscore that progression was by no means inevitable, or even the most common outcome, and there were many false positives. At the time, most clinicians believed in a highly deterministic model in which ‘a trace of schizophrenia is schizophrenia’ and that once the ‘process’ started it rarely if ever receded or stopped. Some still do. This was unreconstructed Kraepelin, and a model that I sought at the outset of my career to challenge [4,5]. I suspect that Rosenman and Anderson are very much in sympathy with this.
We subsequently have elaborated a broader ‘clinical staging’ model [6,7] which acknowledges the fluidity of the clinical phenotype, recognizes that it evolves over time and will have a variable and as yet unclear relationship with underlying biomarkers and any putative ‘processes’. The latter term is a very loaded one in psychiatry, and is perhaps better avoided, since it contains so many assumptions and connotations. A more scientific approach would seek to separate biomarkers which reflect prior and underlying vulnerability (endophenotypes) from those which indicate either the ‘substrates’ and possibly drivers of emerging symptoms, or the secondary consequences of progressive illness. A clinical staging model enables this to be carried out systematically. All of these biomarkers cross current syndromal diagnostic boundaries to varying degrees, so syndromes and stages form a two-dimensional matrix. The clinical staging model proposes an arbitrary series of categorical stages linked to the degree of clarity, specificity and extension of the clinical phenotype. It imposes structure on what are obviously dimensional phenomena but from a perspective orthogonal to the traditional diagnostic categories. The model allows us to explore the relationship between elements of the clinical phenotype (putative core symptoms versus ‘noise’) at each stage and the prediction of remission or progression to a more severe stage. The relationship of the clinical phenotype with biomarkers within each stage and their role in prediction can also be understood and the strength and topography of underlying neurobiological changes (‘process’) may be better characterized as a ‘biosignature’ [8]. This is a long way from the inevitability of ‘deterioration’, since at every stage except the last, only a proportion/subset will progress to the next.
Pre-emptive intervention trials aimed at prevention of onset, relapse and chronicity can be much better configured around a clinical staging framework, which is explicitly heuristic and open to evolution and redesign as new data accumulate. It does not pay excessive respect to the traditional syndromes and indeed cuts across them in the spirit of the future DSM-V. The traditional syndromes are more characteristic of the mid- to late stages of illness and crystallize long after the need for care is first evident. The initial stages, when ‘need for care’ first manifests, are very broad and shallow clinical phenotypes with unstable admixtures of salient symptoms and ‘noise’. From these, resolution and recovery may often occur, but so too multiple relatively discrete exit syndromes will differentiate as targets, though combinations of ‘comorbidity’ or blends are to be frequently expected. I would therefore agree that the prodromal term does conspire to confuse the early intervention task with its deterministic undertones, so it is important to construct an approach that does no harm and yet can do much good for patients to prevent a range of ‘bad outcomes’. In fact, it is in part fear of this latent deterministic attitude and all that goes with it that drives some of the controversy and resistance to early intervention, especially in the earliest clinical pre-psychotic stage. There is a valid fear that people at this stage will be over-treated and stigmatized. The clinical staging model, which is sensitive to the risk/benefit ratio at each stage, seeks to deal with this issue by requiring that the treatment in earlier stages is both safer and should be shown to be more effective if such treatment is delivered earlier rather than later. Examples of this in psychosis are cognitive behavioural therapy (CBT) and omega-3 fatty acids which have efficacy in the UHR stage but less so (alone) in later stages, and low dose antipsychotic medications and vocational interventions in first-episode psychosis. Clozapine is reserved for a later stage precisely because the change in risk/benefit ratio which it reflects is acceptable by that point. Early intervention (EI) and preventive strategies in psychotic and mood disorders can be seen as no more than attempting to reduce the risks of progression from one stage to the next (especially early on) within this clinical staging model.
Let's now turn then to focus a little more specifically on EI. The world is complex and our ability to understand it and predict events is limited. The Nobel prize-winning economist and polymath Herbert Simon [9] pointed out that our rationality is ‘bounded’. We invent rules of thumb or heuristics to enable us to reduce complexity, so even though the decisions are not necessarily the best ones, we can at least function in the world. The key test is: do these rules of thumb work? The heuristics (just one or two in essence) we have used to date for treatment decisions and predicting outcome in psychosis, derived from the neo-Kraepelinian model, have been far too rigid and needed to be seriously loosened [4]. Rosenman and Anderson would probably agree. They simply don't work well enough for aetiological or treatment purposes [4,5]. Clinical staging and early intervention seek to provide a new set of heuristics which can be tested. Their ultimate test is empirical: do they work? The authors point out that early intervention in psychosis has been harshly criticized (by a diverse range of people ranging from the traditional anti-psychiatry forces to a small cadre of academic psychiatrists [10,11]). There is no need to be ‘shocked’ by this criticism, since some of it derives from genuine scientific scepticism and some is simply the inertia of the status quo and late adoption. However, some of the more vitriolic contributions are clearly spawned by those whom Naomi Oreskes and Eric Conway have termed ‘merchants of doubt’ [12] acting on behalf of vested interests. Interestingly, we have not seen the EI strategy challenged in this emotive yet often calculating manner in other branches of health care, even where benefits are less achievable. It is worth pointing out that very few consumers, families, most other doctors, health bureaucrats or politicians are among the critics.
This is because early intervention for potentially serious health problems is axiomatic for obvious reasons, which the general public readily understand. This places the onus of proof appropriately on critics to somehow justify the converse scenario of late or delayed intervention (debate about how to achieve early intervention is a different and second order issue). The mean period of delay in treating first-episode psychosis for standard adult services still sits at over one year and it was this shocking statistic and the human suffering that lay behind it that really helped to catalyse the EI reforms which began in the early 1990s. In psychosis, EI was an idea that was mooted in the 1920s [13], but never seriously pursued until the 1990s. It has since attracted widespread support from the public, from policy makers and from leading clinicians and researchers all over the world, though policy and rhetoric have so far outstripped investment in this country and others. This means that most psychotic patients still receive late, sparse and sporadic intervention in under-resourced generic services. As explained above, EI has been devised as a reform paradigm to challenge the assumptions that Rosenman and Anderson have also questioned. While it is very much a work in progress which is committed to evidence-based principles, it is crucial to understand that evidence flows from as well as drives and guides the direction of reform. Despite some confusing atmospherics in their critique, Rosenman and Anderson seem to be generally in favour of early intervention for people who are experiencing mental ill-health and its consequences, including psychotic symptoms, but genuinely worried that the wrong conceptual approach will undermine this endeavour. Let us consider what EI is trying to achieve and how it is performing and evolving.
There are several targets for EI in psychosis, which can be seen as a prototype for early intervention in other potentially severe and enduring disorders. It is not necessarily preventive in a primary sense, or curative, but pre-emptive [14–16]. So the judgement as to whether it is effective or not rests on whether the course of illness can be modified and disease burden ‘averted’, and not whether cases can be ‘cured’. Firstly, the reduction of treatment delay and duration of untreated psychosis (DUP) for first-episode psychosis is now backed by solid evidence from the TIPS project in Norway, which has shown that functional outcome and especially negative symptoms are improved over 5 years [17], and recovery rates are improved over 10 years by reducing the DUP [18]. Secondly, specialized and personalized intervention programmes do pre-empt much of the personal and social impact and modify the early course of disease, producing better outcomes compared to standard care. The evidence now suggests that such care needs to be provided for at least a subset for more than 2 years and up to 5 or more to maintain the initial gains produced by EI [11,19,20]. EI of this type for first-episode psychosis is highly cost-effective and frees up precious funds for other programmes in mental health care [21–23]. So rather than diverting resources away from other patients, as claimed by one group of critics, EI is able to shrink the proportion of those with heavy long-term demand for care and free up resources to care for the remainder who become disabled despite EI. This is why governments are taking the EI model much more seriously.
To the extent that these results have fallen short of the expectations of some, this could be due to the fact that the EI models tested [19,20] were more dilute than the optimal or fully fledged EI model, and that the tenure of care was foreshortened, which clearly was harmful to the sustained recovery of a significant subgroup. Some patients are relatively immune to the benefits of EI, and a subset of patients fails to make any better response than to standard care. An emerging focus for EI then is the more intensive treatment of the 20% of cases who fail to remit within the first 6 months. Early use of clozapine and intensive psychosocial interventions, notably CBT, needs to be studied more seriously [24]. Next, since most people have a need for care which precedes onset of positive symptoms and so much of the psychosocial impact embeds in this period, identifying and intervening with people with sub-threshold yet disabling clinical features prior to the onset of full threshold and sustained positive psychotic symptoms has been a growing focus for clinical research globally in recent years [25]. Finally, the recognition that other potentially disabling and enduring mental disorders emerge in the same age group as early psychosis has meant that there has been a conceptual shift and service redesign [26–30], which is congruent with the thrust of Rosenman and Anderson's paper.
There is a need to correct the rather jaundiced view that Rosenman and Anderson convey regarding the research in UHR. Drawing on community surveys [31,32] which suggest the clinical phenotype of psychosis is more widespread and often benign, they assert that the falling transition rate means that the clinical phenotype of UHR is innocuous. This is demonstrably false if the latter is ascertained in an ‘enriched’ (for risk) sample of help-seeking and distressed young people. In fact, even with a lower recent transition rate (20% or so as compared with 40% previously) the relative risk is 200 times the general population risk and these cases all manifest at least moderate functional impairment. Somewhat less enrichment in risk through changes in referral or sampling methods may indeed be part of the explanation for a lower rate of transition, as they suggest, but so could lead-time bias or more effective intervention [33]. These patients certainly do not have an innocuous condition. Even those who fail to convert to psychosis in many cases carry or develop other persistent diagnoses and their social and vocational functioning is seriously compromised. There have been at least five randomized controlled trials (RCTs) focused on UHR patients so far [34] and all bar one show a very positive benefit to risk ratio of intervention with a potent number needed to treat (NNT) of 4. Hardly wheels spinning deeper in the mud! If we had imagined 20 years ago it would be possible to delay the onset of first-episode psychosis in vulnerable symptomatic patients, this would have been seen as a genuine and sought-after advance. Of course it is still early days, with many questions to be answered, and this task needs to be approached within a broader context where other syndromes, not only sustained positive psychosis are pre-emptive targets, an approach of which Rosenman and Anderson seem to approve. Finally, one simple piece of hard data shows that the use of the sub-threshold stage as the initial target for intervention, far from undermining the effectiveness of EI for first-episode psychosis, actually enhances it. In clinics devoted to sub-threshold cases or putatively prodromal cases, a proportion (10–20% of referrals) are found to be already psychotic. If the barrier to entry had not been lowered and access provided on the basis of the need for care associated with the sub-threshold UHR phenotype, these patients would have presented much later and arguably in crisis mode as so many first-episode psychosis cases still do. The UHR channel means that more first-episode psychosis cases are detected earlier with lower DUPs and less collateral damage, at least for a subset of cases.
In summary then, the clinical staging model transcends the assumptions disputed by Rosenman and Anderson by defining a series of (admittedly arbitrary) categorical stages to allow study of the risk of progression (or not) from one stage to the next, and of interventions to reduce such risk, as well as mapping what syndromes or clinical phenotypes develop with progression and the relationship of putative endophenotypes and other biomarkers to stage and clinical phenotype. Whether specific terms like the ultra-high risk state or the psychosis risk syndrome can validly be applied depends on the degree of selective predictive power or valence for a particular exit or transition syndrome. This is the subject of research as well as intense debate in relation to DSM-V [35]. If the valence is loose or distributed, an alternative early clinical phenotype could be a common or pluripotential risk syndrome [36]. This initial stage would be characterized by nonspecific symptoms, largely distress, anxiety and depression plus social consequences or correlates and coping reactions and multiple possible exit syndromes. Conversely, some researchers believe that these early clinical phenotypes do indeed carry specific biological risk for certain disease pathways and therefore contain a particular ‘latent process’ which might correspond with more specific clinical features such as hypomanic, positive or negative symptoms. This is a testable hypothesis. It is by no means necessary, however, to believe this is true to accept the need to offer care not only for what is already manifest, but also to try to prevent a worsening or evolution of the clinical picture into something more serious or disabling, whatever labels we give to this later stage. It seems quite legitimate and not necessarily ‘baleful’ at all to have these two targets for all early patients and to be positive and hopeful but honest about this. There is absolutely no evidence that such a dual agenda disturbs patients or families as claimed by the authors. We have been doing this work for 17 years and provided the setting is youth-friendly without stigma [11], and there is open dialogue and informed consent, we have found no difficulty. If these elements are not present, however, and a simplistic deterministic biomedical approach is taken, then some of the authors’ concerns may be valid and this perhaps is the essence of their argument. I am grateful to Rosenman and Anderson for applying their minds and experience to this important challenge for psychiatry and to the Editor for the opportunity to give my response.
Footnotes
Acknowledgements