RDoC at 10: changing the discourse for psychopathology

From 1990 to 2010, researchers lamented the problems of comorbid conditions and heterogenous syndromes in psychiatric diagnoses. A volume from the American Psychiatric Association detailing a research agenda for DSM‐5 collectively captured these views, as well as the accompanying lack of progress to connect integrative neuroscience with psychiatric diagnoses ¹ .

Amid growing concerns, the Research Domain Criteria (RDoC) project was proposed by the US National Institute of Mental Health (NIMH) to offer an alternative strategy ² . The NIMH convened researchers and stakeholders to identify psychopathology‐relevant constructs from the experimental literature. Conditions for empirical support were set, including well‐elucidated neural circuits and demonstrated validity for functional behavior. Relevance for human suffering was also a requisite.

Since its introduction, RDoC facilitated a new scientific discourse about precision medicine in psychiatry, as evidenced by thousands of citations in studies or commentaries and as a keyword in over 400 National Institutes of Health (NIH)‐funded grants. It cleared a path for alternative designs for translational research, offered new tactical approaches to explicate disruptions in psychopathology mechanisms, and invited robust international dialog ³ .

An RDoC Unit at NIMH now helps advance the aims of the project in many ways. For example, sponsoring meetings to update original domains and constructs, to evaluate relevant tasks for RDoC research, and to strengthen integration of developmental processes and environmental events. An internal working group adds to the discourse, and National Advisory to Mental Health Council subcommittees provide oversight.

Before RDoC, diagnoses constructed of clinically‐observed symptoms (DSM/ICD) were the routine to define patient groups studied in translational research. Now, the assumption that clinical syndromes will be validated by corresponding internal mechanisms in a one‐to‐one way is accepted as untenable. RDoC offers new directions to understand psychopathology as dimensional deviations from normal performance (neural systems and behaviors). It captures mechanisms inherent to normal‐range functioning, and then determines how disruptions correspond to psychopathology. It expands consideration of what constitutes an independent or dependent (outcome) variable.

RDoC research calls for samples of patients broadly exhibiting a range of related symptom patterns, primarily focusing on connecting a neural mechanism with behavior, while tracking variations in co‐occurring diagnoses, degrees of functional impairment, and levels of subjective distress. RDoC research does not require DSM/ICD diagnoses to select patient groups, and allows focus in a study on one or more such clinical syndromes.

An RDoC strategy gaining considerable traction is to pool a broader band of participants with a range of potentially related syndromes, and then study dimensions of reliable biomechanisms as independent (or predictor) variables in relation to outcome. For example, researchers have redistributed patients with various anxiety and mood‐related diagnoses into quintiles based on psychophysiological response elicited by startle, and the regrouping predicted a number of psychopathological indices, including reports of distress and transdiagnostic severity ⁴ . P3 amplitude was graduated: lower among those reporting the most distress and demonstrating more comorbid disorders, and higher for those reporting less distress and more circumscribed symptom patterns, with control participants placing in the cen‐ter quintile, suggesting clinical relevance for fear responses that deviate in either direction ⁴ .

Another study found that this inverse pattern held when dividing participants in quintiles based on amygdala response using a different task, comparing emotional and neutral images during functional magnetic resonance imaging ⁵ . Participants with the least differential (emotional minus neutral) self‐reported higher trauma risk scores, whereas those with the largest differential reported lower trauma scores, with controls again placing in the middle quintile.

That fear circuitry figured more prominently in these studies for more circumscribed conditions (lower transdiagnostic severity and less complex trauma histories, respectively) is clinically relevant, suggesting when exposure therapies would be more efficacious. In contrast, those with a broader symptom range (more comorbidity, more complex trauma) may have more elaborate disruptions among internal mechanisms. This adds to our understanding of poorer outcomes and higher dropout rates for exposure treatment in people with multiple trauma post‐traumatic stress disorder.

RDoC has also helped to change the way clinical trials are conducted, including those for regulatory approval. Regulatory bodies, once bound to DSM disorders as the standard for outcome, now allow indications targeting transdiagnostic constructs such as anhedonia, cognitive functions, and suicidal behaviors ⁶ . For example, patterns of activity in ventral striatum, implicated in reward anticipation, have been investigated as primary outcome measure in the development of therapeutic agents for anhedonia in a mixed patient group of depressive and anxiety disorders ⁷ . Other researchers reported wider‐spread activations in cognitive control regions, specifically bilateral parietal cortex activity in pediatric anxiety, and are now investigating these brain‐behavior dimensions to predict cognitive behavioral therapy outcomes ⁸ .

RDoC introduced structure by grouping constructs within superordinate domains (rows), and suggesting units of analysis ranging from neuroscience to behavior (columns). Constructs and elements of the matrix were offered as exemplars, with expectations for change with accumulation of new findings. However, the format was susceptible to the interpretation that the framework incorporated a goal to curate a fixed set of constructs that constrained research. We have previously clarified that RDoC can serve instead as a nomological network to theoretically organize psychopathology data ³ . Advances in computational neuroscience offer dynamic ways to model hypothesized brain‐behavior mechanisms. These models could use RDoC as a scaffold to extend data‐driven approaches to identify new clinical phenotypes.

Models of failure might detail the implications of disruptions in one or more internal mechanisms ⁹ . It could be that a failure in either one of two distinct mechanisms leads to a similar clinical presentation, or that a poor clinical prognosis requires multiple failures. Another possibility is that a mechanism could fail (e.g., glutamatergic pathways in the amygdala associated with fear acquisition), yet clinical manifestation of that failure could be shielded by compensatory mechanisms (e.g., ventromedial prefrontal cortex circuits associated with fear extinction).

Exemplar RDoC dimensions offer a platform for a first generation of mathematical models to integrate data describing relations of neural circuits and behavior. Such developments can advance multiple aspects of the needs for precision diagnosis. Experimental paradigms developed with computational models are showing the potential to delineate specific aspects of behavior, and to relate these various aspects to their implementation, coordinated by increasingly well‐specified brain circuits ¹⁰ . This development capitalizes on the RDoC principle to explain relations across units of analysis in order to clarify psychophysiological constructs and the critical relationships among interrelating response systems.

Over the first ten years, RDoC provided a structure to unharness researchers from diagnostic categories for funding applications and treatment approval in order to open new avenues for discovery. Going forward, RDoC may provide ways to leverage the development of computational models of psychopathological systems that integrate neural and psychological mechanisms with developmental processes and environmental influences.