May 3, 2022
Written by: Margaret Gardner
Neuroscientists have spent countless hours looking for exactly where in the vast, complicated network of the human brain abnormalities lead to different psychiatric symptoms. For example, does overactivity of the prefrontal cortex lead to the strict, self-imposed dietary restrictions suffered by those with anorexia nervosa1? Do altered connections between the nucleus accumbens and its neighboring regions reduce the ability of those with depression to feel pleasure2? But does each aspect or symptom of a mental illness actually have a specific cause in the brain, and if not, should we redefine or regroup these symptoms to better reflect our brains’ natural neural functions?
Researchers typically try to uncover the causes of mental illness by studying people with a specific psychiatric diagnosis and comparing their brain function, mood, and behavior to those of healthy people. However, there can be a great deal of heterogeneity – or differences in an individual’s symptoms – within each diagnosis. For example, one person with anorexia may occasionally binge-eat and then try to compensate for it by purging, while another may not; however, because both eat less than they should for fear of gaining weight, they receive the same diagnosis of anorexia. Why is that? Do these two people have the same brain abnormalities causing them to have an unrealistic perception of their own weight? Perhaps yes, but perhaps no.
On the other hand, there are many symptoms that are common across different psychiatric diagnoses, such as anhedonia, which is a loss of interest in or enjoyment of activities like hobbies or time with friends and family. Anhedonia is common in depression, but also in schizophrenia and substance use disorders. Is the anhedonia felt by someone with alcohol use disorder caused by a different neural mechanism than the anhedonia of a person with major depressive disorder? Maybe, or maybe not.
Both questions are part of bigger concerns about how, where, and in whom we look for the brain regions and circuits that can trigger different psychiatric symptoms, and how we label the abnormalities we find. Many different frameworks have been used to define and categorize the brain’s many functions, from cognition to emotion regulation to sensory perception, as well as the psychiatric symptoms that arise when these functions go wrong. An ideal framework would be specific enough that these mental constructs (the tasks or functions that the brain performs), are clearly and specifically related to their origins in the brain. Therefore, by knowing what mental construct a patient struggles with, doctors would know what and where in the brain to treat. While this “ideal framework” has yet to be defined, many have been proposed to relate mental constructs and abnormalities to each other and to the brain. Each framework explored here has different strengths and weaknesses – some leaning more “brain-based” while others aim more to describe the presentation of symptoms – but all provide useful ways of thinking about and studying the relationship between brain function and mental illnesses.
DSM
The reason that two people who restrict their eating have the same anorexia diagnosis, even if one binges and the other doesn’t, is that both meet the criteria outlined in the Diagnostic and Statistical Manual of Mental Disorders (DSM). Currently in its 5th edition, the DSM is often referred to as the “bible” of psychiatry and outlines different criteria – or symptoms a patient must have – to be diagnosed with any of its 70+ diagnoses3.
These diagnostic criteria are agreed upon by expert panels of psychiatrists who use their experience and the latest research to define disorders that are both specific enough to group people and flexible enough to account for some individual differences. Let’s quickly step outside the brain for an example of how this works. Say you want to classify books based on what it felt like to read them: And then There Were None was suspenseful, The Catcher in the Rye was angsty, and The Lord of the Rings was epic and adventurous. But what about Lord of the Flies? It’s suspenseful, angsty, and adventurous. We’d have to define some cutoff in order to say whether the boys’ struggle to survive and sadistic infighting are more like Holden Caulfield’s irritable misadventures or Christie’s characters’ attempts to name the murderer in their midst. The authors of the DSM had to make similar decisions to group a great of psychiatric symptoms into categories that are both distinct and meaningful.
Take for example Bipolar I Disorder, which affects approximately 0.6% of the US population every year4. To be diagnosed with Bipolar I, a person must experience a manic episode lasting at least 1 week. But what is a manic episode? Depending on the person, it will involve feelings of euphoria or irritability, in addition to being more active and energetic, plus at least 3 of 7 other possible symptoms. Ultimately this means that while everyone with a bipolar diagnosis has experienced mania, one person may have felt euphoric, needed less sleep than normal, and been more talkative and easily distracted during their manic episode, while another person may have felt irritable and self-important, needed less sleep, took more risks than usual, and had racing thought and lots of energy. Meanwhile, a third person may have had all the same symptoms but only for 4 days rather than a full week, in which case the DSM would classify their episode as “hypomanic” and a diagnosis of Bipolar II disorder may be more appropriate. Clearly the criteria in the DSM can be confusing and may feel somewhat arbitrary.
Overall, the DSM is known to have both a great deal of variability within disorders and overlap between them. This is important because it isn’t clear whether there are brain-based differences in those who meet Bipolar I criteria vs Bipolar II criteria, or how similar brain dysfunction is in those who have anhedonia in schizophrenia and those who experience anhedonia in major depressive disorder. Without the answers to these questions, treating the root cause of a person’s symptoms can be much trickier. Therefore, while the DSM has been and remains useful in guiding psychiatric classification, many agree there are likely other frameworks that more closely mirror the brain’s function5.
RDoC
One framework designed to map mental constructs more directly onto the brain’s natural organization is the Research Domain Criteria (RDoC) from the National Institute of Mental Health (NIMH). To create the RDoC framework, experts grouped mental constructs into 6 domains based on the latest knowledge about each construct and the systems in the brain that produce it. For example, the Social Processes domain includes the mental construct of self-knowledge, which is what RDoC calls the “ability to make judgments about one’s current cognitive or emotional internal states, traits, and/or abilities”6. This construct is defined not just by how humans experience it, which may include things like emotional awareness or self-consciousness, but also by the types of neurons and brain regions that current data suggests are involved in giving us these abilities. The goal of this type of framework is to be data-driven, or to incorporate as much biological knowledge as possible rather than relying on people’s own descriptions of how they think and feel to define constructs. The data that is incorporated into RDoC classifications also spans many “levels” of brain structure and activity, from genes and molecules to regions and circuits that seem to all play a role in producing each specific construct. RDoC is therefore not a collection of diagnoses and, as its name implies, is only used for research. It is, however, intended to be used in investigating the causes of mental illnesses and perhaps eventually influence diagnostic criteria once even stronger brain-symptom relationships have been established.
RDoC is also dimensional, meaning that unlike the DSM, which only includes mental constructs when they become abnormal, RDoC attempts to group and describe mental constructs as they are in healthy states and across any illnesses. For example, you might want to study auditory hallucinations and hypothesize that they an abnormal manifestation of Perception, one of the RDoC’s cognitive constructs. Using this dimensional approach, you would study perception in a wide range of people to see if there are any similarities among those who have auditory hallucinations; regardless of whether they meet DSM criteria for schizophrenia, PTSD, dissociative identity disorder, or no disorders at all, we’d expect to see the same perceptual alterations in everyone experiencing auditory hallucinations. By casting such a broad net, dimensional studies have a better chance of uncovering the brain abnormalities causing auditory hallucinations specifically and will be relevant to any individuals who experience them, regardless of what other symptoms or diagnoses they may have.
Going back to our book-grouping example, we can now apply the same type of methods that RDoC uses to classify mental constructs: data-driven and dimensional. Right away, there are many objective pieces of information about each book that you could use to group them, such as author, length, year of publication, or weeks on the New York Times best-seller list. But grouping just on these facts – or a combination of them – would miss out on other important, subjective aspects about a book. For example, reading Neil Gaiman’s dark, mythical fantasy American Gods is a very different experience than reading his biography of 80’s band Duran Duran. Conversely, would it really be meaningful to group Pride and Prejudice with 1984, even though they both remind you of high school English class? These are the kinds of decisions that the experts constructing RDoC had to make in integrating the latest neuroscience research about how the brain works with how different mental constructs seem to present themselves in illness and everyday life.
RDoC’s dimensional approach is slightly less relevant to psychiatry than the DSM, where constructs are grouped by the patterns in which they tend to go awry in real people. However, RDoC’s attempt to define constructs that represent unique and specific functions of the brain’s component parts is useful in grounding psychiatric concepts in measurable – and ultimately treatable – brain processes.
fMRI-Driven Framework
While RDoC domains are based in neurobiological research, the experts compiling these studies still had to integrate information from across many subfields of neuroscience and make assumptions about it to build a clear framework. A recent project led by Elizabeth Beam of Stanford University tried to remove as much human decision-making as possible by building a framework from statistical groupings of functional magnetic resonance (fMRI) papers’ results7. fMRI uses magnets to approximate brain activity in human subjects by measuring the changes in blood flow throughout the brain.
Beam and her colleagues looked across over 18,000 fMRI studies’ measures of regional brain activation and compared how often the articles included terms for mental constructs like “decision making” and “emotion”. Using statistics, the team was able to map which mental construct terms were most strongly and specifically associated with activation in different brain regions. For example, if an article shows activation in the amygdala, it’s likely to also include the word “fear” more than one would expect just any old fMRI article to. Brain regions that were associated with the same mental construct words were then grouped into circuits.
Finally, to select how many domains to organize these circuits into, the researchers tried clustering them into anywhere from 5 to 25 groups and tested each set to see how accurately they could predict the use of terms within a domain based on brain activation and vice versa. Ultimately the researchers settled on 6 domains, each representing a distinct pattern of brain activation and associated mental constructs.
Bringing this back to our literary analogy, this would mean that we’re classifying books (mental constructs) on their association with one reasonably objective criterion, such as the frequency with which authors use different words. To do this, we’d measure which books use which words (i.e. which brain regions are activated) more than expected, then see if we can predict a book from its words used and vice versa. Frankenstein, for instance, probably includes words like “monster”, “experiment”, and “Switzerland” more than other books, and a book that uses those words a lot is probably Frankenstein. Our model will then group these books into domains based on shared word associations. In the end, Frankenstein could be grouped with books like War and Peace, in a 19th-Century Classics domain, or it could be grouped with Dracula and Do Androids Dream of Electric Sheep? in an Existential Monsters domain. The goal is for the data to drive the groupings, and then it’s up to us to make sense of why different books (or constructs) are sorted the way they are.
After building their framework, the researchers also applied similar methods to map the RDoC and DSM domains onto brain activation reported in fMRI papers. Of course, the fMRI-driven framework matched brain activation most closely since that’s literally how it was made. However, researchers were curious to see how the DSM and RDoC compared to their framework and to each other. Overall, the DSM and RDoC domains were more likely to group mental constructs that don’t share activation patterns or to split up those that do, with the DSM matching fMRI activations the least.
By grouping mental constructs on shared brain activation, the fMRI-driven framework suggests that some constructs may be more or less similar than one might think based on the DSM, RDoC, or our own intuition. For example, constructs like fear and positive/negative emotions were grouped into the fMRI-derived Memory domain, suggesting that these phenomena – which are grouped separately in both RDoC and the DSM – may be best thought about and studied in relation to memory and to each other.
What’s Next?
While the labels and groups we apply to mental constructs and psychiatric symptoms may seem arbitrary, the way these ideas are framed affects how researchers and clinicians form their hypotheses and decide treatments. While each of the frameworks discussed has its own pros and cons, it is important that the field keeps working to refine and remap the relationships among mental functions that are important to disease, as well as their origins in the brain. It is only with this shared, meaningful vocabulary that researchers can study, uncover, and eventually repair the underlying causes of psychiatric symptoms.
References
1. Boehm I, Geisler D, King JA, et al. Increased resting state functional connectivity in the fronto-parietal and default mode network in anorexia nervosa. Front Behav Neurosci. 2014;8(OCT):346. doi:10.3389/FNBEH.2014.00346/BIBTEX
2. Liu R, Wang Y, Chen X, Zhang Z, Xiao L, Zhou Y. Anhedonia correlates with functional connectivity of the nucleus accumbens subregions in patients with major depressive disorder. NeuroImage Clin. 2021;30. doi:10.1016/J.NICL.2021.102599
3. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders (DSM-5-TR). Published 2021. Accessed April 24, 2022. https://psychiatry.org/psychiatrists/practice/dsm
4. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders : DSM-5-TR. (American Psychiatric Association, ed.).; 2022.
5. Insel T, Cuthbert B, Garvey M, et al. Research Domain Criteria (RDoC): Toward a New Classification Framework for Research on Mental Disorders. https://doi.org/101176/appi.ajp201009091379. 2010;167(7):748-751. doi:10.1176/APPI.AJP.2010.09091379
6. National Institute of Mental Health. RDoC: Self-Knowledge. Accessed April 24, 2022. https://www.nimh.nih.gov/research/research-funded-by-nimh/rdoc/constructs/self-knowledge#cell_221
7. Beam E, Potts C, Poldrack RA, Etkin A. A data-driven framework for mapping domains of human neurobiology. Nat Neurosci 2021 2412. 2021;24(12):1733-1744. doi:10.1038/s41593-021-00948-9
PennNeuroKnow 