THE PROMISE OF PRECISION MEDICINE

Defining diseases in terms of biological markers of a mechanism of disease (also known as biomarkers), parsing out who’s ill from who’s not based on variations on those markers, and establishing a treatment plan that targets the mechanism of disease indicated by those markers makes a lot of sense. It does so especially for a medical speciality like oncology that deals with a class of diseases (cancers) that can be characterized by aberrations at single, specific, and targetable level (the cell) and mechanism (gene expression). This is what precision medicine is all about: identifying diseases in terms of biomarkers that can guide targeted therapies based on a molecular understanding of drug action. Having such biomarkers and targeted therapies could allow tailoring interventions to individuals in ways that improve effectiveness, reduce side-effects, and avoid inappropriate treatment by identifying those people likely to respond positively and negatively to specific treatments—an approach to assessment that has been called ‘theranostics’.

Wouldn’t it be great if other medical specialities, especially those struggling with things like the non-specificity and heterogeneity of symptoms and uncertain treatment efficacy like psychiatry, could benefit from what precision medicine has to offer? And psychiatry has already taken a step toward precision medicine through the adoption of the research framework called research domain criteria (RDoC). But the hope for precision through biomarkers may be nothing but a pipe dream because it skirts some deep problems in the nature of psychiatric disorders (Tabb and Lemoine [2021]). So, is precision psychiatry a dead-on-arrival project? If it’s done the way we tend to do it now, it probably has fatal flaws—but perhaps these can be avoided if it is done right. Here, I will walk you through the argument in my article about why a precision psychiatry is not, in principle, unachievable.

There are several deep problems that threaten to make precision psychiatry unachievable, but the one that seems to me to be particularly difficult is the assumption that precision medicine is necessarily a reductionist endeavour that aims to identify low-level causal factors and that, consequently, a precision psychiatry must be reductionist as well. We can call this the assumption of ‘vertical precision’. Vertical precision is thinking about the causality of mechanistic biomarkers in terms of the most paradigmatically biological trait that is assumed to cause the disease in a bottom-up way (for example, HER2/neu gene amplification leading to HER2 protein overexpression and spread of cancer cells in HER2 breast cancers). In the case of psychiatry, this implies that the most paradigmatically non-biological traits (for example, suicidal ideation) must be accounted for by a low-level biological marker. This is the kind of causality that seems to be the focus in precision medicine, but this is precisely the kind of causality that seems improbable for most psychiatric disorders.

But there are simpler, less reductive ways of thinking about causation, ways that lend themselves to an alternative view of mechanisms and (bio)markers in precision medicine that I call ‘horizontal precision’. Horizontal precision encourages the search for mechanistic biomarkers at any level from genes to social interaction. This view borrows from the framework of mechanistically mediated effects of Carl Craver and William Bechtel (Craver and Bechtel [2007]) and the interventionist model of causation of Kenneth Kendler (Kendler and Campbell [2009]), which taken together allow us to conceive of a non-reductionist precision psychiatry research programme. Before outlining this, let me explain why some of the deep problems turn out to be not so deep when adopting the horizontal view.

The first deep problem is ontological. It is often claimed that the limited efficacy of pharmacotherapy for certain classes of psychiatric disorders is due to the fact that these disorders are multi-level entities, configured across all the levels of the biopsychosocial hierarchy, and that they therefore require interventions acting in a coordinated fashion at multiple sites (for example, using both pharmacotherapy and psychotherapy, intervening in the family and social milieu, and so on). If so, there is no way that any low-level biomarker will be enough to guide treatment of an entity as complex as a mental disorder. Now suppose that this is wrong, that mental disorders really are only a matter of dysfunctions in your brain. Even so, the brain is not a simple organ. It is immensely complex, has its own internal hierarchical structure, and may allow for many different pathways to the same behaviourally and experientially expressed affliction.

The second deep problem is technical. If mental disorders are not localized in the brain, this might simply mean that such disorders are outside the purview of a precision medicine that is methodologically committed to ‘omics’ approaches—that is, approaches that measure biological parameters such as genomics, proteomics, and metabolomics. Even if we were to be generous and accept that markers for precision could be developed for non-biological factors, what would this look like? Biomarkers play a complex role in precision medicine. They are both markers of pathology with diagnostic utility and targetable mechanisms of pathology with therapeutic utility. They are identified based on differential responses to a treatment directly targeting a marker that is understood to be part of the mechanism of pathology. Available biomarkers in psychiatry don’t have the specificity that would warrant running the kind of clinical trial needed to establish their utility.

However, I argue that the ontological problem can be solved with horizontal precision, while a solution to the technical problem might be achieved with ‘personomics’ (Ziegelstein [2015])—that is, the comprehensive and global assessment of a set of person-level features that characterize an individual.

Craver and Bechtel’s mechanistically mediated effects framework offers a way through the ontological problem. On this approach, we should not worry too much about being unsure of the exact causal pathway between levels (for example, how gene expression ‘pushes’ behaviour). Rather, we should try only to understand causation within the level of the mechanism (for example, how genetic information instructs cells’ functioning or how your friend rejecting you caused your social withdrawal, but not how the genetic information causes your social withdrawal). Horizontal precision, understood within this mechanistically mediated effects framework, suggests that markers can be found within levels, at any level of mechanism, because the good kind of causation that we expect from a mechanistic biomarker will not be found between levels of mechanism. And so it makes perfect sense to look for mechanistic biomarkers at multiple levels of mechanism, from the cellular all the way up to the social, and to hope for targeted treatments through interventions acting in a coordinated fashion at multiple sites. Now, for those who might worry that the horizontal view might be throwing away inter-level causation, the interventionist model of causation should help. Inter-level causation exists and still matters. However, it should be read as interventionist causation between levels of mechanism where a low-level—or high-level—cause (for example, neurotransmitters) will be considered as responsible for an effect (for example, behaviour) if an intervention that respects some standards of statistical analysis (for example, a randomized controlled trial) on the cause reliably changes the effect.

The ontological problem can be addressed with a theoretical framework such as the horizontal precision framework, because solving this problem is about opening our mind to an alternative way of thinking about what characterizes a disease or disorder. This second set of issues is technical, in the sense that it calls for us to find a way to better characterize the mechanisms of disorders on multiple levels and produce operationalizable descriptions that could become the objects of clinical trials. I suggest that ‘personomics’ could be a response to this second set of problems. The details of such a programme for precision psychiatry still need to be fleshed out, but I think that it could go like this: A programme in precision medicine is typically viewed as, first, employing an ‘omic’ approach to the identification of the markers of the pathology and, second, as using a computational approach and large data set to stratify a population based on treatment response to those markers. There are a variety of person-level traits of mental disorders that could be considered as personomic markers. This isn’t really the challenge, although moving from mere description to non-reductive mechanistic description at the personal level may require some ingenuity. Personomic markers could be constitutive features of health and disease configured at the person level in terms of experiential, developmental, psychological, social, and ethical processes, as per the biopsychosocial model (Bolton and Gillett [2019]) and as modes of self-construal as per the cultural eco-social view in psychiatry (Kirmayer [2007]). Most of the work do be done to operationalize those dimensions of personhood in terms of targetable personomic markers involves developing valid computational constructs to measure their change in response to a therapy targeting them. For that, a personomic programme could appeal to the various methods developed in the domain of digital, big data, and modelling psychiatry (Gauld et al. [2021]) (for example, natural language and digital phenotyping methods and the use of ecological momentary assessment data).

Of course, this is just to gesture at some directions that a personomic programme could take. Fleshing it out will require much further work. What I offer is an argument debunking the claim that precision psychiatry is in principle not possible. This is important for many reasons, but mainly because the project of precision medicine has many potential benefits for patients. However, if we are to pursue such a project in psychiatry, we will need research that examines all the important components of mental disorders. Such research will only be possible if the funding bodies that are currently promoting the project of precision medicine recognize the place of non-biological factors in this framework, and if those who would carry out this research appreciate how persons—with their agency, subjectivity, and lifeworld—fit into the picture of contemporary (biological) psychiatry.

References

Bolton, D. and Gillett, G. [2019]: The Biopsychosocial Model of Health and Disease: New Philosophical and Scientific Developments, Cham: Palgrave Macmillan.

Craver, C. F. and Bechtel, W. [2007]: ‘Top-Down Causation without Top-Down Causes’, Biology and Philosophy, 22, pp. 547–63.

Gauld, C., Dumas, G., Fakra, E., Mattout, J. and Micoulaud-Franchi, J.-A. [2021]: ‘Les Trois Cultures de La Psychiatrie Computationnelle’, Annales Médico-Psychologiques, Revue Psychiatrique, 179, pp. 63–71.

Kendler, K. S. and Campbell, J. [2009]: ‘Interventionist Causal Models in Psychiatry: Repositioning the Mind–Body Problem’, Psychological Medicine, 39, pp. 881–87.

Kirmayer, L. J. [2007]: ‘Psychotherapy and the Cultural Concept of the Person’, Transcultural Psychiatry, 44, pp. 232–57.

Tabb, K. and Lemoine, M. [2021]: ‘The Prospects of Precision Psychiatry’, Theoretical Medicine and Bioethics, 42, pp. 193–210.

Ziegelstein, R. C. [2015]: ‘Personomics’, JAMA Internal Medicine, 175, pp. 888–89.