- RDoC Matrix
- Example Studies
- Developmental and Environmental Aspects
- Process and Final Product
Draft 3.1: June, 2011
Over the past several decades, an increasingly comprehensive body of research in genetics, neuroscience, and behavioral science has transformed our understanding of how the brain produces adaptive behavior, and the ways in which normal functioning becomes disrupted in various forms of mental disorders. In order to speed the translation of this new knowledge to clinical issues, the NIMH included in its new strategic plan Strategy 1.4: “Develop, for research purposes, new ways of classifying mental disorders based on dimensions of observable behavior and neurobiological measures.” (For the full text, see http://www.nimh.nih.gov/about/strategic-planning-reports/index.shtml#strategic-objective1). The implementation of this strategy has been named the Research Domain Criteria Project (RDoC). The purpose of this document is to describe the RDoC project in order to acquaint the field with its nature and direction, and to facilitate commentary from scientists and other interested stakeholders regarding both general and specific aspects of the RDoC approach.
Across all areas of medicine, research in genomics, cell biology, and pathophysiology is revolutionizing diagnosis and treatment. In disorders as diverse as cancer, heart disease, diabetes, and inflammatory bowel disease, the discovery of identifiable subtypes within broad clinical phenotypes has led to more specific, more effective treatments or identification of new targets for prevention. Research in mental disorders is also developing quickly: Novel data about genomic factors and the role of particular brain circuits are reported almost monthly. However, new findings on mental disorders have had limited clinical impact, partly because they map only moderately onto current diagnostic categories for mental illness. Thus, some of the risk genes for psychotic disorders appear to be associated with both schizophrenia and bipolar disorder and the same prefrontal region has been implicated in depression and PTSD. In contrast to cancer and heart disease, where research has identified subtypes of common disease, it appears that the biological findings with mental disorders are relatively non-specific; could specificity in fact exist, but not for the currently recognized clinical categories? This question leads to a consideration of how current categories were derived.
Currently, diagnosis in mental disorders is based on clinical observation and patients’ phenomenological symptom reports. This system, implemented with the innovative Diagnostic and Statistical Manual-III (DSM-III) in 1980 and refined in the current DSM-IV-TR (Text Revision), has served well to improve diagnostic reliability in both clinical practice and research. The diagnostic categories represented in the DSM-IV and the International Classification of Diseases-10 (ICD-10, containing virtually identical disorder codes) remain the contemporary consensus standard for how mental disorders are diagnosed and treated, and are formally implemented in insurance billing, FDA requirements for drug trials, and many other institutional usages. By default, current diagnoses have also become the predominant standard for reviewing and awarding research grants.
However, in antedating contemporary neuroscience research, the current diagnostic system is not informed by recent breakthroughs in genetics; and molecular, cellular and systems neuroscience. Indeed, it would have been surprising if the clusters of complex behaviors identified clinically were to map on a one-to-one basis onto specific genes or neurobiological systems. As it turns out, most genetic findings and neural circuit maps appear either to link to many different currently recognized syndromes or to distinct subgroups within syndromes. If we assume that the clinical syndromes based on subjective symptoms are unique and unitary disorders, we undercut the power of biology to identify illnesses linked to pathophysiology and we limit the development of more specific treatments. Imagine treating all chest pain as a single syndrome without the advantage of EKG, imaging, and plasma enzymes. In the diagnosis of mental disorders when all we had were subjective complaints (cf. chest pain), a diagnostic system limited to clinical presentation could confer reliability and consistency but not validity. To date, there has been general consensus that the science is not yet well enough developed to permit neuroscience-based classification. However, at some point, it is necessary to instantiate such approaches if the field is ever to reach the point where advances in genomics, pathophysiology, and behavioral science can inform diagnosis in a meaningful way. RDoC represents the beginning of such a long-term project.
RDoC is intended as a framework to guide classification of patients for research studies, not as an immediately useful clinical tool. While the hope is that a new way forward for clinical diagnosis will emerge sooner rather than later, the initial steps must be to build a sufficient research foundation that can eventually inform the best approaches for clinical diagnosis and treatment. It is hoped that by creating a framework that interfaces directly with genomics, neuroscience, and behavioral science, progress in explicating etiology and suggesting new treatments will be markedly facilitated.
RDoC will follow three guiding principles, all diverging from current diagnostic approaches.
- First, RDoC is conceived as a dimensional system (reflecting, e.g., circuit-level measurements, behavioral activity, etc.) spanning the range from normal to abnormal. As with dimensions like hypertension or cholesterolemia in other areas of medicine, this approach incurs both the problem and advantage of defining cutpoints for the definition and extent of pathology – e.g., mild, moderate, and severe. (To the extent that DSM-V introduces dimensions in addition to classes, the crosswalks to RDoC dimensions may be enhanced.)
- Second, RDoC is agnostic about current disorder categories. The intent is to generate classifications stemming from basic behavioral neuroscience. Rather than starting with an illness definition and seeking its neurobiological underpinnings, RDoC begins with current understandings of behavior-brain relationships and links them to clinical phenomena.
- Third, RDoC will use several different units of analysis in defining constructs for study (e.g., imaging, physiological activity, behavior, and self-reports of symptoms). Indeed, RDoC, as a research framework, has been developed with the explicit goal of permitting investigators to choose an independent variable from one of several different units of analysis. The details of this approach are explained next.
The RDoC research framework can be considered as a matrix whose rows correspond to specified dimensions of function; these are explicitly termed “Constructs,” i.e., a concept summarizing data about a specified functional dimension of behavior (and implementing genes and circuits) that is subject to continual refinement with advances in science. Constructs represent the fundamental unit of analysis in this system, and it is anticipated that most studies would focus on one construct (or perhaps compare two constructs on relevant measures). Related constructs are grouped into major Domains of functioning, reflecting contemporary thinking about major aspects of motivation, cognition, and social behavior; the five domains are Negative Valence Systems (i.e., systems for aversive motivation), Positive Valence Systems, Cognitive Systems, Systems for Social Processes, and Arousal/Regulatory Systems. The columns of the matrix represent different classes of variables (or units of analysis) used to study the domains/constructs. Seven such classes have been specified; these are genes, molecules, cells, neural circuits, physiology (e.g. cortisol, heart rate, startle reflex), behaviors, and self-reports. Circuits represent the core aspect of these classes of variables – both because they are central to the various biological and behavioral levels of analysis, and because they are used to constrain the number of constructs that are defined. Investigators can select any level of analysis to be the independent variable for classification (or multiple levels in some cases, e.g., behavioral functioning stratified by a genetic polymorphism), and dependent variables can be selected from multiple columns. In addition, since constructs are typically studied in the context of particular scientific paradigms, a column for “paradigms” has been added; obviously, however, paradigms do not represent units of analysis.
Three criteria guided the selection of the draft list of candidate constructs presented here. First, the inclusion of a construct was constrained by whether a particular brain circuit or area could reasonably be specified that implements that dimension of behavior. Given the complexity of the brain and of behavior, this was more ambiguous in some cases than others; some constructs, such as attention, reflect activity spread relatively diffusely over many brain areas, while attachment behavior may similarly reflect neurotransmitter and hormonal functions (e.g., oxytocin) acting at disparate locations throughout the brain. Second, an attempt was made to maintain a reasonable “grain size” that would permit a tractable listing of the major functional dimensions of behavior. While it is recognized that there may be important and meaningful sub-constructs that could be considered (e.g., various types of aggression), an overly specified list could result in an unwieldy and excessively long listing. Third, the constructs are based on current literatures that have provided a neurobehavioral research base for each of the entries.
The draft RDoC matrix is listed in the table below, followed by examples of how the classification system might be used and several points of clarification. Dimensional constructs are listed in the rows. Below the matrix, several constructs are listed to provide examples of brain circuits and/or neurotransmitters that help define and constrain each one, along with a brief indication, where appropriate, of constructs representing the dimension’s opposite pole; note that listings of circuit components and neurotransmitters are meant to be illustrative, not exhaustive. Constructs are grouped into five major domain areas as listed above. It is important to emphasize that these particular domains and constructs are simply starting points that are not definitive or set in concrete. We expect these to change dynamically with input from the field, and as future research is conducted. The keys here are the overall framework that we are suggesting, and the process for its development.
Draft Research Domain Criteria Matrix
|Domains/Constructs||————— Units of Analysis —————|
|Negative Valence Systems|
|Positive Valence Systems|
Cognitive (effortful) control
|Systems for Social Processes|
Affiliation and attachment
Perception and Understanding of Self
Perception and Understanding of Others
|Arousal and Regulatory Systems|
Notes regarding the Units of Analysis
- “Circuits” can refer to measurements of particular circuits as studied by neuroimaging techniques, and/or other measures validated by animal models or functional neuroimaging (e.g., emotion-modulated startle, event-related potentials).
- “Physiology” refers to measures that are well-established indices of certain constructs, but that do necessarily not tap circuits directly (e.g., heart rate, event-related potentials).
- “Behavior” can refer variously to behavioral tasks (e.g., a working memory task), or to behavioral observations.
- “Self-reports” refer to interview scales, questionnaires, or other instruments that may encompass normal-range and/or abnormal aspects of the dimension of interest.
Examples of Constructs (individual entries) within Domains (boldface)
Negative Valence Systems
- Fear (opposite pole, – fearlessness): amygdala, hippocampus, interactions with ventromedial PFC
- Potential threat: HPA axis, BNST, hippocampus; CRF, cortisol
Positive Valence Systems
- Approach motivation (opposite pole – anhedonia): mesolimbic dopamine pathway
- Habit-based behavior (including OCD spectrum): orbitofrontal cortex, thalamus, dorsal striatum
- Working memory: dorsolateral PFC, other areas in PFC
- Cognitive (Effortful) control (opposite pole – impulsivity, disinhibition, externalizing): anterior cingulate gyrus, various areas of medial and lateral PFC
Systems for Social Processes
- Social dominance: distributed cortical activity, mesolimbic dopamine systems; testosterone, serotonin
- Facial expression recognition: ventral visual stream, fusiform gyrus
- Self-representational circuits: dorsal & posterior ACC, insula
- Stress regulation: raphe nuclei circuits; serotonin
- Facilitated stimulus processing: locus coeruleus circuit; norepinephrine
- Readiness for stimulus processing and responding: brain resting state network
- PFC: Pre-frontal cortex
- HPA: hypothalamic-pituitary axis
- BNST: bed nucleus of the stria terminalis
- CRF: corticotrophin releasing factor
- OCD: obsessive-compulsive disorder
- ACC: anterior cingulate cortex
Given that RDoC is a classification framework, how might the scheme work in actual practice, given the goals of (1) permitting widely differing independent variables and (2) implementing a dimensional system that allows variance extending down into what would be regarded as sub-threshold psychopathology? Two general approaches are as follows. The first is to include all patients presenting for treatment at a given type of treatment facility, as in the second example below; the statistical approach then becomes one of regression. The second approach is to specify a particular criterion for selecting multiple groups – e.g., patients who score more than one standard deviation below the mean on a cognitive task, patients who show significant activation in a specified brain area on a neuroimaging task – and compare these to other patients not meeting the criterion and/or to a non-clinical control group. In any case, exclusions for co-morbid conditions would be expected to be much less stringent (although the usual exclusions such as other medical or neurological disorders, extreme substance abuse, etc. could still apply). Manuscripts submitted under RDoC will be expected to state how many patients were screened for inclusion in the study, and the reasons for exclusion.
Two example studies are listed in order to illustrate the types of studies that might be conducted within the RDoC framework. For clarity, the variables used to classify subjects are reiterated at the end of each example.
- Recent studies have shown that a number of genes reported to confer risk for schizophrenia, such as DISC1 (“Disrupted in schizophrenia”) and neuregulin, actually appear to be similar in risk for unipolar and bipolar mood disorders. These findings are consistent with a number of recent papers questioning the classical Kraepelinian distinction between schizophrenia and bipolar disorder; however, little data are available to evaluate psychotic disorders as a spectrum since studies almost always focus on one or the other, and patients falling short of DSM/ICD criteria are excluded. Thus, in one potential design, inclusion criteria might simply consist of all patients seen for evaluation at a psychotic disorders treatment unit. The independent variable might comprise two groups of patients: One group would be positive and the other negative for one or more risk gene configurations (SNP or CNV), with the groups matched on demographics such as age, sex, and education. Dependent variables could be responses to a set of cognitive paradigms, and clinical status on a variety of symptom measures. Analyses would be conducted to compare the pattern of differences in responses to the cognitive or emotional tasks in patients who are positive and negative for the risk configurations. The results of studies of this type could contribute to knowledge about the particular types and severity of behavioral and/or neurobiological deficits that tend to be associated with a given risk gene; in turn, such results could help build a foundation to study mechanisms by which a particular candidate gene contributes to adverse effects. Eventually such research might lead to redefining how psychotic disorders are conceptualized. Classification variables: In this example, the domain under study is Cognition (possibly comparing two to three constructs such as working memory versus declarative memory). The independent variable for classification is the risk gene configuration(s), and the dependent variables comprise performance on the various cognitive tasks. (It is possible that DSM diagnosis, or some other set of psychiatric symptoms, might serve as a second independent, between-subjects variable; however, an emphasis on studying mechanisms would dictate that the sample not be constrained to patients with only schizophrenia or bipolar disorder – i.e., inclusion criteria should incorporate those with schizoaffective disorder, delusional disorder, psychotic disorder NOS, etc.)
- A large number of studies have examined neuroimaging responses to various types of emotional challenges in patients with a particular mood or anxiety disorder, compared to non-clinical controls. Frequently, the conclusion is that disorder X is characterized by an abnormality in task Y – such as emotion regulation, activation of a particular circuit or brain area (e.g., amygdala, ventromedial PFC), or response to some emotion-related task. However, such abnormal mechanisms appear to be involved in many different disorders, while on the other hand, not all patients with a given diagnosis necessarily show the abnormality – suggesting that there are fundamental mechanisms in common across these disorders. A design to study fear circuitry might thus have as inclusion criteria all patients presenting at an anxiety disorders clinic. Classification variables: The construct of interest is Fear/extinction, in the domain of Negative Affect. The independent variable for grouping would be the extent of responding to fearful stimuli using a measure such as amygdala response (from fMRI) or fear-potentiated startle (i.e., a circuit-level variable). Dependent variables would be symptom measures on various fear and distress measures, in order to test hypotheses about mechanisms by which hyper-reactivity and hypo-reactivity to threat cues affect the nature and severity of presenting symptoms. As an outcome of such research, these results might generate predictive validity studies leading to improved treatment selection or new pharmacological targets for intervention.
Developmental and Environmental Aspects
The RDoC concept is organized around basic neural circuits, their genetic and molecular/cellular building blocks, and the dimensions of functioning that they implement. There are two highly important areas of mental disorders research that are thus not represented in the matrix per se, but are considered to be critical elements in research fostered by RDoC. These two areas are developmental aspects and interactions with the environment. The intent is that the RDoC matrix will enhance the study of both areas by promoting a systematic focus on their relationship to specific circuits and functions.
Developmental aspects. Mental disorders are increasingly viewed as neurodevelopmental disorders in one way or another. Therefore, addressing development issues across various phases of the life span represents a critical consideration that is implicit to the RDoC framework, and might be considered as a third dimension in the matrix. The types of constructs typically found in the child temperament literature are (not coincidentally) similar to the RDoC domains, and many areas of the child psychopathology literature (e.g., broadly addressed to Internalizing or Externalizing problems) serve as a more compatible model for a dimensionally-based approach compared to the highly specified categories of adult psychopathology. Four brief examples might be given of life-span goals that could be addressed within the RDoC framework: (1) Further explicate the longitudinal course of adolescent brain maturation and synaptic pruning to identify genes and circuit development factors associated with departures from normal developmental functioning, and points in prodromal stages where intervention might particularly be targeted; (2) Evaluate the extent to which the recruitment of additional cortical areas during task performance or emotional challenge in elderly subjects predicts resilience against onset or deterioration of course in mental disorders; (3) Generate improved explication of the construct of cognitive control (or effortful control), relative to disentangling current controversies regarding ADHD, juvenile bipolar disorder, conduct disorder, etc.; (4) Specify the mechanisms regarding developmental changes in systems for fear and distress across puberty (including the effects of the social environment), that could explain clinical data indicating that adolescent anxiety disorders often precede depression.
Environmental aspects. The central nervous system is exquisitely sensitive to interactions with various elements of its environment virtually from the moment of conception. The social and physical environment comprises sources of both risk and protection for many different disorders occurring at all points along the life span, and methods for studying such phenomena as gene expression, neural plasticity, and various types of learning are rapidly advancing. As with developmental aspects, environmental influences may thus be considered as another critical dimension of the RDoC matrix. The effects of a particular interaction with the environment, e.g., the effects of early child abuse, may pose risk for a wide variety of disorders. As another example, illicit drug use may cause sensitization of mesolimbic dopamine circuits that generalizes to other drugs of abuse and addictive behaviors. Thus, it is hoped that a research program organized around the relevant circuit-based dimensions that are affected, independent of a particular disorder, will accelerate knowledge regarding such environmental influences along the entire range of analysis from genes to behavior.
- As mentioned above, the current organization is focused on (and constrained by) circuit definitions in order to (1) avoid an over-specification and proliferation of constructs, and (2) provide an organizing point that facilitates the integration both of genetic, molecular, and cellular levels of analysis regarding sub-components of circuits, and of behavioral and self-report levels of analysis regarding the kinds of behaviors that circuits implement. The intent is not to arbitrarily exclude constructs, but rather to foster thinking about how constructs are related at various levels of analysis. For example, extraversion is not listed in the draft matrix, but might be considered to represent another aspect of social dominance -- in that they are both typically described in terms of activity in mesolimbic dopamine systems, and thus may reflect different aspects of what is fundamentally the same dimension.
- The framework is directed toward constructs most germane to mental disorders, and makes no claim to span the entire gamut of functional behavior. For instance, circuits relevant to thermoregulation and reproductive behavior are not included.
- The number of constructs might well be viewed as sparse by many scientists. The attempt has been to include relatively high-level constructs in order to avoid an over-specification of functions that could become unwieldy and also necessitate unnecessarily frequent revisions to the list as research progresses. However, the framework is meant to foster, not discourage, research that explicates mechanisms within and across the constructs as listed. As stated above, the current framework should be viewed as a starting point and part of a work in progress.
- The complexity of the brain is such that circuits and constructs will necessarily have considerable overlap, and arbitrary separations are unavoidable. For instance, the basolateral amygdala is well-known to be involved with both threat and appetitive processing. This reflects the hierarchical nature of the nervous system, and the difficulty of creating a system that encompasses various levels in one framework. It should also be noted that some constructs, such as emotional regulation or homeostasis, are not listed here; these are considered superordinate principles of nervous system activity that operate across many different circuits.
- Research with post-mortem tissue samples may be appropriate for studies within the RDoC framework, where the hypotheses and other variables are conceived in terms of relevant domains and constructs.
- The RDoC framework is explicitly agnostic with respect to current definitions of disorders. For instance, depression as a clinical syndrome has been related to abnormal activity in the amygdala, anterior cingulate cortex, nucleus accumbens, and multiple monoamine systems, while also strongly comorbid with multiple anxiety disorders, eating disorders, etc. The idea is that studying the individual mechanisms may lead to better understanding of current disorders, or perhaps new and novel definitions of disorders, but in either case improved information about treatment choices.
- As mentioned above, the aim of RDoC is to create a framework for grouping participants in research studies, in order to create a foundational research literature that informs future versions of nosologies based upon genetics and behavioral neuroscience. RDoC is not intended for clinical diagnosis at the current time. In the future, research supported by RDoC could inform diagnostic approaches using new laboratory procedures, behavioral assessments, and novel instruments to provide enhanced treatment and prevention interventions. It is also hoped that RDoC will support enhanced development of new pharmacological and psychosocial interventions based upon neurobiological and behavioral mechanisms.
Process and Final Product
The NIMH intends that the RDoC process be as transparent as possible. An internal NIMH steering group, advised by a small group of external experts, has created the initial RDoC framework and devised the list of candidate domains, constructs, and classes of variables. NIMH issued a companion Request for Information (RFI) in the NIH Guide to seek input about all aspects of this first draft of the RDoC matrix and process. These comments were taken into account in further refining the initial version of the matrix.
A series of workshops is currently in progress as an initial step in defining the specifications for each construct. At a minimum, one workshop will be held for each of the five domains. However, in order to gain experience with the process, the first workshop focused on the construct of working memory. Each workshop involves experts from various areas that span the RDoC’s units of analysis. Participants are asked to discuss and decide upon current findings, paradigms, and procedures relevant to each level of analysis, along with critical research questions. Proceedings of each workshop are posted on the RDoC page of the NIMH web site for continuing commentary and suggestions for changes. Depending on the nature and extent of comments, a second workshop may be held to achieve consensus on final specifications.
The final specification for each construct will consist of:
- A definition of the construct’s functional aspects, summary of relevant circuitry, and relationship to other constructs;
- A list of current state-of-the art measures, paradigms, and procedures at each level of analysis;
- Current pressing research questions and issues pertaining to the construct, including one or two salient examples of the groupings of DSM/ICD categories that might be included in studies addressing these questions.
The intent of the RDoC is to accelerate the pace of new discoveries by fostering research that translates findings from basic science into new treatments addressing fundamental mechanisms that cut across current diagnostic categories. The research specifications are intended to guide investigators in conducting such integrative research by including cutting-edge variables in research applications. However, since RDoC is a research framework, use of such variables is not required; indeed, one goal is to speed the pace of new information at all levels of analysis. For this reason, RDoC will incorporate a mechanism for continual evaluation of new findings, and inclusion into the domain/construct specifications. While the exact procedures remain to be worked out, it is anticipated that the NIMH steering group will work together with subject-area experts from each of the relevant domains to accept nominations (from the evaluation team or from scientists in the field) for modifications and additional listings.
Although the formal period for commenting under the RFI has terminated, NIMH welcomes continuing commentary regarding any aspect of the RDoC project, including, but not limited to, the following points. Comments may be emailed to email@example.com.
- The overall RDoC framework, including the organization of the Domains and Constructs, and the Units of Analysis.
- Particular constructs that should be added, deleted, merged, or changed.
- Criteria for determining what constructs should be included or modified.
- “Grain size” of the constructs.
- Criteria for making changes to the domain and construct specifications.
NIMH staff look forward to working with all groups of interested stakeholders as the RDoC project is developed.