The RDoC scheme can be represented as a two-dimensional matrix (Table I). The rows represent the “dimensions of observable behavior and neuroblological measures” specified in Goal 1.4 of the NIMH Strategic Plan. These dimensions are referred to as “constructs” to represent their status as concepts regarding brain organization and functioning that evolve with advances in research. In turn, constructs are grouped under five superordinate domains of activity, which reflect a conceptual typology of functions as well as empirical relationships among activity in related brain circuits.
Table I. Research Domain Criteria Matrix. “Circuits” can refer to measurements of particular circuits as studied by neuroimaging techniques, and/or other measures validated by animal models or functional neuroimaging (eg, emotion-modulated startle, (more ...)
The columns of the matrix represent various units (or levels) of analysis that can be used to measure the various constructs, with the former term preferred to emphasize the integrative approach. The units of analysis are as follows: genes, molecules, cells, circuits, physiology, behavior, and self-report. Genes, molecules, and cells are self-apparent (although in many cases, direct assessment of molecules and cells in functioning humans remains problematic). The “Circuits” unit of analysis refers to measures that can index the activity of neural circuits, either through functional neuroimaging or through recordings previously validated as circuit indices (eg, fear-potentiated startle). “Physiology” refers to well-established measures that have been validated in assessing various constructs, but that do not measure circuit activity directly (eg, heart rate, Cortisol). “Behavior” may refer either to systematically observed behavior or to performance on a behavioral task such as working memory. There is also a separate column for paradigms, in which scientific tasks that are especially useful for the study of the construct are noted.
The cells at the intersections of constructs and columns are populated by research findings. Overall, the RDoC matrix structure is intended to facilitate integration of existing research findings and foster the identification of gaps in the knowledge base that represent promising areas for integrative research.
How will the RDoC matrix actually function as a classification system for experimental purposes? For perspective, it may be pointed out that the current system imposes three constraints upon the independent variable (ie, group classification) in psychiatric studies: first, symptoms are the unit of analysis that must be utilized; second, particular constellations of symptoms must be employed (ie, the DSM poly thetic criteria or their ICD equivalents); and third, the symptoms must be employed (with rare exceptions) simply to render a binary, diagnosis present/absent decision rather than being quantified in any way. RDoC is intended to free investigators from these constraints. An element from any unit of analysis may be the independent variable. In a study of working memory, performance on a working memory task could be the independent variable (possibly stratified by particular genetic polymorphisms), and activation of relevant working memory areas (as measured by fMRI) and real-world functional capacity might be dependent variables. As another example, patients presenting with internalizing (mood or anxiety) disorders might be classified along a dimension of their overall symptom reports of distress (but independent of DSM diagnosis), and fear circuit activation in some relevant task (eg, imagery, film clips) might be assessed in order to test the hypotheses that increasing severity and/or chronicity of distress are associated with hyporeactivity in fear activation circuits. In each case, the independent variable cannot be assigned until after the experimental procedures are conducted; because the independent variable is dimensional, however, this does not necessarily pose problems in statistical power or matching subjects in groups. As these examples imply, the choice of which units of analysis to use as independent and dependent variables depends upon the research question.
Particularly in the early phases of studies using the RDoC approach, it may be heuristic for investigators to report the number of participants in study samples who meet diagnostic criteria for various DSM primary diagnoses in order to facilitate comparisons with traditional and RDoC classification. However, it should be noted that one major emphasis of Strategic Aim 1.4 is to delineate the entire range of a particular dimension, notably including patients who fall short of traditional diagnostic criteria or who may have an NOS (Not Otherwise Specified) diagnosis. Thus, including only those subjects who meet criteria for designated DSM/ICD disorders (even if more than one) is not a wholly satisfactory approach in the RDoC perspective. One of the inherent problems with the categorical approach is that, in spite of the acknowledged heterogeneity that is apparent in virtually all clinical diagnoses, the consequent analysis implicitly involves the notion of a unitary entity that has a “point” Expected Value and “normal” variance on any given measure. Findings of group differences then imply that all patients are impaired compared with normal control subjects on some measure—doubly misleading in that: (i) at least some patients are not so impaired, and it would be important to know why; and (ii) impairment in patients with NOS or forme fruste conditions may be proportionately smaller and/or less severe, and excluding these patients obscures an explication of potentially relevant dimensions and also obviates attention to clinically relevant dysfunction.
A commonly asked question is whether including patients from widely disparate diagnoses (eg, a working memory study Including patients with primary diagnoses of psychotic disorders, internalizing disorders, and externalizing disorders) would result in such excessive variance as to be meaningless. Initially, at least, this appears to be a legitimate concern. The typical situation would be that patients presenting for treatment at a given type of clinic—psychotic disorders, anxiety/mood disorders—would represent the sampling frame for a given study, thus maximizing relevant variance while avoiding “apples versus oranges” comparisons. Eventually, as the circuits and measurements are better understood, it may be productive to make these kinds of comparisons. For instance, in recent years it has become common to consider whether clinical depression is present as a comorbid syndrome in schizophrenia, for example.6
Using symptom-based criteria, it is difficult to know whether such symptoms are due to “depression” pathology or to “schizophrenia” pathology. However, measures that have been validated to assess relevant circuit functions (whether in cognition, reward circuit activity, or arousal systems) may provide a heuristic to move forward in addressing such important clinical questions.