With one of the highest, if not the highest, attrition rates in drug development, neuropsychiatry has had a rough ride in the first two decades of the 21st Century. In the two areas I have worked in the most – cognitive impairment and affective disorders (depression, anxiety) – it is only recently that we have seen late-stage movement followed by some approvals (depression), but with no symptomatic drugs passing muster for cognitive impairment (despite recent approvals for disease-modifying agents in Alzheimer’s disease).

This resulting ‘desert’ led a number of companies (for example Pfizer and AstraZeneca) to exit CNS drug development altogether, and although some have returned, many have stayed away. For those that remained, and for biotechs pursuing the challenge, there has been an increasing focus on ‘precision psychiatry ’: (1) a broadly reductionist approach to drug development, outlined in the figure.

Figure. The Research Domain Criteria (RDoC) Initiative framework. The RDoC framework “provides an organizational structure for research that considers mental health and psychopathology in the context of major domains of basic human neurobehavioral functioning, rather than within established diagnostic categories”. (2)

Reductionism as a principle is seen at a number of points within this paradigm. Firstly, traditional psychiatric disorders, for example, generalized anxiety disorder, are dispensed with to focus on individual domains and constructs, as outlined by the Research Domain Criteria (RDoC) Initiative (2). Domains encompass major systems of emotion, cognition, motivation, and social behaviours, and are then divided into constructs, which are behavioural elements, processes, mechanisms, and responses, some of which are recognisable as symptoms a patient may experience and report; others are more complex and could encompass a number of different symptoms. The RDoC approach then seeks to understand how each construct is comprised based on genetic, neurocircuit, behavioural, and self-report assessments.

This approach is critical to neuropsychiatry, because it takes a construct (for example potential threat [anxiety], which sits in the ‘negative valence systems’ domain), and seeks to understand how it manifests symptomatically and behaviourally (measurable in patients), how these relate to aberrations in brain circuits (which may also be measurable), and finally entertains the possibility that there may be genetics underpinning such aberrations (although the importance of environmental influences is also encompassed).

This approach is revolutionary in that it seeks to deconstruct traditional psychiatric diagnoses to examine components (constructs) that exist across traditional diagnoses and attempt to understand their underlying neurobiology. In doing so, the hope is that we can deliver a more precise approach to their treatment: the development of therapeutic interventions (pharmacologic, psychotherapeutic, digital) aimed at these specific constructs. It recognises fundamental concerns over systems such as the Diagnostic and Statistical Manual of Mental Disorders (DSM) (3), namely that a majority of patients meet the diagnostic criteria for more than one condition, and that the diagnoses are not related to known aberrations in neurobiology. An example would be the high frequency of anxiety symptoms in patients with major depression (circa two-thirds to three-quarters, depending on the definition of ‘anxiety’), leading some experts to believe that the conditions may not even be separable. As one group commented: “In terms of nosology, high comorbidity suggests that some unitary conditions have been split into multiple diagnoses, which co-occur frequently as a result, indicating the need to redraw boundaries between disorders” (4). This is less of a problem for RDoC because it “…explicitly focuses on the complex overlapping multidimensionality of mental illness” (5).

The focus on understanding the association between individual constructs/symptoms and the underlying neurobiology is a highly laudable approach to more rationally target pathways that drive mental health symptoms. But concerns have been raised.

Firstly, patients don’t present with single symptoms very often, if at all. Indeed, in the latest incarnation of the DSM – DSM-V-TR – there is only one recognised clinical condition with a single symptom (Trichotillomania: Recurrent pulling out of one’s hair, resulting in hair loss). This then is a utility question: would a practitioner be meeting a patient’s needs if they focused on a single symptom alone? It is however important to recognise that RDoC, although it does not provide a classification system intended for immediate clinical use, focuses on constructs rather than individual symptoms per se: there is room under the term ‘anxiety’ for a number of symptoms that result from that construct. In RDoC, there is an attempt to classify a core phenomenon (such as anxiety), inclusive of the range of symptoms it could encompass (such as avoidance) without assumption about such a hierarchy. For example, if ‘anxiety’ leads to greater psychic features (such as worry) and greater somatic symptoms (racing heart rate, dizziness, sweating), there isn’t any assumption on the ‘chicken and egg’ nature of these sets of symptoms, or individual symptoms. However not only are symptoms often inter-related, but so are constructs. Back to our example of the association of low mood and anxiety. Indeed, symptoms are often interrelated, as witnessed through many symptom network analyses, which seek to determine whether symptom structures within conditions are hierarchic or not (does low mood for example lead to other symptoms of depression such as anhedonia [loss of pleasure in activities?]).

Secondly, to what extent can these constructs be attributed to an individual network in what is, a highly complex system? This of course may depend on which construct is under question. In fact, there is increasing evidence from functional MRI and EEG studies that where aberrations are occurring in the brain that are related to these constructs can be identified, although these may be occurring in more than one circuit.

Finally, there is – what is to me – one of the biggest challenges that the precision psychiatry approach faces. Identification of circuits' underlying constructs only provides the first layer of what can be targeted through pharmacology. The layer of molecular targets is of course deeper, and the approach to precision psychiatry requires the identification of drug targets that relate to the circuit. It is evident of course that those targets may also exist in other circuits and may have similar or different neuronal effects. Furthermore, it is not clear that modulation of a single target may be sufficient to induce the desired effect, or of course, that simple inhibition or potentiation is what is required. It has been posited that combination approaches may well be needed to produce a significant clinical response even within a single construct; this is even more likely when considering the multiple co-morbid symptoms/behavioural changes that need addressing in many patients (this looks very likely in cognitive impairment, where deficits are often seen across multiple cognitive domains).

This also raises an important question: are highly targeted drugs likely to be the most effective in neuropsychiatry? Many of those currently being prescribed could better be classified as ‘dirty’ in terms of their pharmacologic profiles, and ‘promiscuous’ in the effects they produce neurologically. Yet they have clinical benefits for patients, albeit with side effects that also reflect their promiscuous behaviour. This has led some to conclude that their effectiveness is not despite their dirtiness, but because of it (6). While the often multiple “off-target” effects of drugs may not contribute significantly (if at all) to their efficacy, for some it is either plausible or even demonstrable that they are contributors.

Returning to the title of this piece, reductionism – whether for the analysis of patients’ behaviours and symptoms or in the analysis of neurologic circuits and targets within them – may reap rich rewards but has potential pitfalls. Although more precise pharmacologic targeting may unlock previously unfound efficacy, it assumes that broader targeting would not have greater benefits, either for an individual construct or holistically for a patient: an observation that may not be a truism.

References

1. Williams LM et al. CNS Spectr; 2023 Sep 7: 1–14

2. https://www.nimh.nih.gov/research/research-funded-by-nimh/rdoc

3. https://www.psychiatry.org/psychiatrists/practice/dsm

4. Kotov R et al. Journal of Abnormal Psychology 2017; 126: 454–77

5. Clark LA et al. Psychological Science in the Public Interest 2017; 18: 72–145

6. Sadri A. J Med Che. 2023; 66: 12651–77