Billet publié sur Twitter/X le 23/05/2013 01:51

Résumé

Although written a little before the anticipated DSM-5 publication date (American Psychiatric Association (2013), this JCPP editorial will be the first to appear in a post-DSM-IV world. By that time we will know which proposed changes will have made it in to the final document. Whatever its ultimate form, DSM-5 will receive bouquets from many, but brickbats from more – inevitable given the need to satisfy so many competing demands (Stringaris, 2013)! True to the word of the APA, the process has been rigorously evidence based. As part of this process work group members have undertaken a monumental task in identifying, reading, meta-analysing and interpreting thousands of studies. A truly heroic and generous work for the greater good carried out conscientiously and skilfully by some of the world's leading clinicians and scientists. We owe each and every work group member a great debt of gratitude. But why go to so much trouble! From the point of view of the clinical research scientist, the answer is as simple as its logic is compelling (Sonuga-Barke, 1998). First, we need more effective treatments for childhood disorders. Second, necessary therapeutic innovation is most likely if we understand what ‘causes’ disorders, so treatments can be more effectively targeted. Third, diagnostic categories function as the bridge of meaning allowing effective communication between the lab and the clinic. They both furnish scientists with their clinical unit of analysis and allow clinicians to situate new scientific discoveries within the coordinates of their everyday practice. Fourth, from this perspective, science-led therapeutic innovation is most likely if diagnostic models map accurately onto disorder pathophysiology. Fifth, if emerging evidence suggests a mismatch between diagnostic and pathophysiologic models, then the former need updating. In principle this sounds like a straight-forward task – take the evidence, compare it to DSM-IV, revise where necessary to bring the latter into line with the former in DSM-5. The reality is of course much more messy and complicated. For instance, while it may seem that the accumulated evidence of the last 20 years supports a case for fundamental change in diagnostic model, such change may simply not be practicable or expedient. Take the thorny issue of whether DSM-5 should replace the current categorical model with a dimensional one. The former (at least implicitly) promoting the notion that diagnostic constructs reflect (at least approximately) real normality/disorder boundaries – or (to paraphrase Plato) the joints of nature. Some strong and influential voices have called for a dimensional revolution in the DSM. Despite this, it seems likely that DSM-5 will stick with a categorical model while making a nod towards the value of dimensional approaches with the proposed inclusion of supplementary cross cutting dimensions (Kraemer, 2007). This decision is interesting and can be seen as flying in the face of the DSM-5's evidence-based principles. This is because compelling evidence for the dimensional nature of many mental disorders has emerged since the publication of DSM-IV. Distributions of symptoms within populations rarely show the sort of extreme right ‘hump’ suggestive of a cluster of extreme scorers (Larsson, Anckarsater, Rastam, Chang, & Lichtenstein, 2012). Furthermore, taxometric techniques which test for structural discontinuities in apparently continuous distributions have confirmed the essentially dimensional nature of a number of disorders (Coghill & Sonuga-Barke, 2012). For instance, ADHD appears to be a genuinely dimensional trait rather than a category. There are a number of important practical considerations that will have fed into the eventual decision whether DSM-5 should, taking this evidence on board, move towards a dimensional approach or stick with the categorical model. First, categorical systems both (a) fulfil the practical decision-making needs of clinicians, and (b) cut with the grain of human cognition – humans appear to find it easier to think in terms of categories than dimensions. Second, moving to a dimensional model would represent a seismic shift: in practice, that may be too disconcerting for clinicians and researchers if taken all at once. Third, while one may see the logic of such a move in principle, coming up with a workable dimensional model has proven difficult in practice. Fourth, introducing dimensions for some disorders but keeping categories for others will introduce a degree of incoherence in the DSM framework – which could undermine its credibility. On balance, given these considerations, the likely decision to stick with categories for diagnosis, even where disorders are dimensional in nature, seems prudent. However, this does leave us with the problem of how we tackle sub-threshold cases; illustrated by the study of Balazs et al. (2013)1 in the current issue where they found that sub-threshold depression and anxiety are linked to increased suicide risk. From the researchers' perspective, supplementing categorical models with disorder dimensions represents a sensible stepping stone to future revisions. There is a second, and perhaps more fundamental challenge to DSM-5's potential to provide the crucial lab-to-clinic bridge needed to promote translational science. This relates to whether diagnostic categories reflect the pathophysiological structure of disorder. At its inception the ambition was to ground the DSM revision process in emerging neurobiological models of mental disorder. Early on in the process, however, it was judged that insufficient progress had been made in clinical neuroscience for this vision to realised. On balance this was the right call. However, it would be a mistake to think that recent advances in the neuroscience of childhood disorder do not have a lot to say about the scientific veracity of current diagnostic models. Indeed, accumulating evidence suggests that current diagnostic boundaries are not, in fact, optimally mapped onto the pathophysiological structure of child and adolescent psychopathology. First there is growing evidence of pathogenic heterogeneity within disorders (e.g., Pelphrey, Shultz, Hudac, & Wyk, 2011) – suggesting that different individuals with the same diagnoses can differ markedly from one another in terms of aetiology and pathophysiology – perhaps to the extent of providing evidence for an etiologic or pathophysiologic subtypes. The study of Sjowall et al. (2013) in the current issue illustrates this at the neuropsychological level. ADHD and control individuals were tested on a broad battery of tests of putative ADHD-related deficits. A number of deficits predicted ADHD group membership independently of one another – some patients showing deficits in relation to one set of neuropsychological functions while others displayed no deficits on these but substantial deficits in other areas. This diagnosis-pathophysiology mismap also can be seen in emerging evidence of pathogenic overlap between different disorders (e.g., McGrath et al., 2011). Indeed separate disorders with strikingly different clinical presentations may share many etiological factors in common (e.g., Smoller et al., 2013) with this lack of causal specificity appears to extend from originating causes in mediating brain processes. Thus recent breakthroughs in genetics and neuroscience present us with a major conundrum. Put simply, while DSM-5 categories are likely to be more useful clinically, growing evidence of etiological and pathophysiological heterogeneity and lack of specificity increasingly challenges the neurobiological/etiological relevance of current diagnostic boundaries. The problem is that mapping diagnoses more directly onto an etiological and pathophysiological structures may not increase, and may very well reduce, their clinical relevance and utility. At the same time, some degree of mapping from clinical constructs to underlying causal structures seems essential if we are to make progress in translational approaches to therapeutic innovation. If, when and how this will be achieved remains uncertain. One practical way forward would be to continue to work within the new DSM-5 structures while identifying the way that common pathophysiological dimensions (e.g., executive dysfunction, altered reinforcement learning, emotion processing deficits) operate within and between disorder boundaries. More fundamentally, the growing awareness that neurodevelopment plays a key role in driving common clinical outcomes from different causal factors and also different outcomes from the same factors, means that a developmentally-based taxonomy may ultimately be required. However, given the radical nature of such a proposition, and the programme of longitudinal research required before it can be implemented, such an innovation will require many years of concerted research action: Anyone for DSM-6?