TL;DR: Mental disorders are commonly defined as “brain disorders” in the literature, and the current funding structure in psychiatry and clinical psychology is strongly based on this notion. At least for the most common mental disorders, however, this notion remains speculation, and is not supported by strong evidence. I created a 10-week syllabus / reading list while writing this blog post available on the Open Science Framework, and hope it will be useful especially for students who are interested in the topic of what mental disorders are.
The brilliant Vaughan Bell wrote a piece in September 2017 that received a lot of attention, entitled “Why we need to get better at critiquing psychiatric diagnosis“. Bell raises a ton of lucid points I very much agree with, the most important one being that psychiatric diagnoses are very diverse. This means that folks criticizing psychiatric diagnoses broadly, or diagnostic manuals in general, most certainly took the mantra Fuck Nuance a bit too far.
I will refrain from reiterating the numerous arguments I agree with, and instead focus on one point that deserves being discussed from a somewhat different angle: biological markers for DSM diagnoses. The section of Bell’s blog relevant here is his justified criticism of the statement that “There are no biological tests for psychiatric diagnoses”. This and similar blanket statements are false: the DSM mentions numerous biological markers, for various conditions. So let us add some context here, and then provide a more nuanced argument that I hope Bell would agree with.
We live in an age of rampant biological reductionism, which is obvious in many contemporary scientific publications. Steven Hyman, Director of the National Institute of Mental Health (NIMH) from 1996 to 2001, stated in 1998 that “Mental illnesses are real, diagnosable, treatable brain disorders”1. This is a commonly accepted notion in psychiatry and medicine, and mental disorders are often defined as brain disorders. For instance, Zhang et al. (2005) introduce Major Depressive Disorder (MDD) as “a brain disorder with 2%–19% population prevalence and 40%–70% heritability” (~470 citations); Bewernick et al. (2010) inform us that “major depression is the most common serious brain disorder with a lifetime prevalence of up to 17%” (~550 citations); and Olesen et al. (2011) include mental disorders in their article entitled “The economic cost of brain disorders in Europe” (~800 citations). In 1997, the then Director of the National Institute of Drug Abuse (NIDA) published a paper in Science entitled “Addiction is a brain disease, and it matters”2, and today, the very first sentence of the Wikipedia article on addiction states: “Addiction is a brain disorder”.
In one of the most influential papers in contemporary psychiatry (~2500 citations), Thomas Insel (NIMH Director from 2002 to 2015) and colleagues introduce the core principles of the novel Research Domain Criteria (RDoC) framework:
“First, the RDoC framework conceptualizes mental illnesses as brain disorders. In contrast to neurological disorders with identifiable lesions, mental disorders can be addressed as disorders of brain circuits. Second, RDoC classification assumes that the dysfunction in neural circuits can be identified with the tools of clinical neuroscience, including electrophysiology, functional neuroimaging, and new methods for quantifying connections in vivo. Third, the RDoC framework assumes that data from genetics and clinical neuroscience will yield biosignatures that will augment clinical symptoms and signs for clinical management.”
Insel is a bit of a persistent offender in terms of naive reductionism. In 2015, he authored a paper in Science entitled “Brain disorders? Precisely”. At the same time, the NIMH under his leadership announced in 2014 that they would stop funding clinical trials that did not contain some form of biological investigation (preferably neuroimaging); the caption of the first image in the article summarizes the core issue well: “Thomas Insel wants studies to identify the biological mechanisms that underlie psychiatric symptoms”. We published a very brief letter in Nature criticizing NIMH’s decision, of course without much success.
In any case, such extreme statements score a 100 out of 100 points on the biological reductionism scale, and ignore a century of well-established research that the strongest correlates to at least some common mental illnesses like MDD are psychological and social3. Such strong assumptions about the importance of biology not only presume that “complex phenomena are ultimately derived from a single primary principle”, but also “mind-body dualism, the doctrine that separates the mental from the somatic”, as described in detail by Engel in 1977.
In sum, while the idea the mental disorders are brain disorders is certainly a position one can have, the fact of the matter is that, despite many decades of considerable research efforts into uncovering underlying biological mechanisms, we have not identified specific and reliable markers for many of the most prevalent mental disorders. As discussed by Engel, the biomedical model has become a cultural imperative — a dogma.
Now, there is nothing wrong with biological psychiatry or related disciplines: trying to understand the biology of mental illness is as important as trying to understand other aspects, and I personally know many fantastic and thorough researchers doing amazing work in this field — some are close colleagues and friends. But that is not the point here. The point is that important conclusions drawn regularly by prominent researchers in the field do not follow from data, and if anybody would argue that “depression is a social disorder” (in contrast to a “brain disorder”), I would push back equally hard: things are not that simple.
And because they are not so simple, Don Robinaugh and I are guest editing a special issue in BMC Medicine on entitled “Complexity in Mental Health Research: Theory, Method, and Empirical Contributions”.
So what do researchers and NIMH Directors mean when they refer to mental disorders as brain disorders? Curiously, the term is rarely defined. But there are many causal relationships one can envision between the brain and mental disorders, and without any specification, the notion that “depression is a brain disorder” is unfalsifiable — and, therefore, unscientific. As Bennett & Hacker4 asserted in 2003, “The vast majority of neuroscientists [do not] offer any explanations of what they mean by ascribing psychological attributes to the brain”. 15 years later, this situation has, if anything, worsened.
So what is a brain disorder, then? Explanatory reductionism requires the biological identification of constructs independently of psychological variables, or, as discussed clearly in a paper forthcoming in BBS by Denny Borsboom, Angélique Cramer, and Annemarie Kalis5: “explanatory reductionism […] requires […] one to identify the hypothesized brain disorders as brain disorders” (my highlight). In other words, it is insufficient to merely identify correlates for DSM diagnoses, because some correlates always exist, and because correlates are not sufficiently informative about whether they are “realizations, causes, or effects” of symptoms (quote from Borsboom et al.).
This notion of brain disorders is related to the idea that disorders are natural kinds, clear-cut categories in nature that humans discover (this brief letter briefly explains kind essentialism that relates to such nosological entities).
“A precondition for the existence of pure forms of illness in the psychic field must be that a particular cause produces material changes in certain parts of the brain; the expression of these changes in a straight continuous line shall then represent the clinical symptoms. Thus the belief in the existence of forms of illness is inseparably tied to the belief in the existence of a tangible anatomical basis, no matter whether in this connection we are dealing with broad structural changes or microchemical or other types of functional changes, but in any case with localized and possibly intelligible processes.”
Going back to Bell’s blog post, he describes “Neurocognitive disorder due to prion disease” (p634) for which clear biomarkers have been identified, and which requires a brain scan or blood test. This is a good example of a brain disorder that clearly meets the criteria discussed above, because the disorder is identifiable via the brain, which is presently not the case for prevalent mental disorders. And maybe things will change — and I would be the first one to throw a party once we found out that depression causally originates in the brain, because it would likely enable us to treat patients much more efficaciously — but until then, the notion that MDD is a brain disorder is a hypothesis, and not a fact6.
This resonates with the main conclusion the United Nations put forward about depression on World Health Day 2017: “the dominant biomedical narrative of depression” is based on “biased and selective use of research outcomes” that “must be abandoned”. We need to move from “focusing on chemical imbalances” to focusing more on “power imbalances”.7.
Biological tests for mental disorders
Let’s revisit the argument above that Bell rejected: “There are no biological tests for psychiatric diagnoses”. I agree that this statement is false, but I’m not sure that it is the most interesting version of this statement to discuss. So let me add some nuance to the argument: while the advent of biological psychiatry half a century ago came with the promise that we would identify specific and reliable biological markers for most common psychiatric disorders — and while much of the contemporary funding infrastructure is based on this notion that is strongly focused on biology — we have failed to identify specific and reliable markers for many of the highly prevalent mental disorders.
This seems a much more interesting point to raise. As a thought experiment, choose one of the most prevalent mental disorders, such as MDD, phobias, or generalized anxiety disorder, and take a dataset with 200 healthy participants, and 200 people with this mental illness; the dataset should include a wide range of social, psychological, and biological variables. Now your challenge is to pick one single variable that correctly classifies at least 60% of the participants. You would not pick 5-HTTLPR (the serotonin-transporter-linked polymorphic region that has been implicated in MDD), or hippocampal volume, or inflammation markers, because associations are generally extremely weak (and also unspecific). You would choose psychological or social variables to try to maximize your chances for prediction, such as adverse life events.
Are there DSM diagnoses where biological markers do very well in terms of prediction? Absolutely, see the example of neurocognitive disorder due to prion disease discussed above, but that’s not the point here. The point is that mental illnesses are broadly referred to as brain disorders in influential contemporary publications, with the notion that the disorders causally originate in the brain8, when there is little evidence to support this strong stance for many common mental disorders.
Four problems with the notion of mental disorders of brain disorders
In my opinion, there are 4 core issues with papers claiming that “X is a marker for mental disorder Y”: false positive findings, the lack of specificity of disorder-marker associations, the lack of strength of disorder-marker associations, and social factors in nosology.
1. Power, null-findings, and replicability
Many study results — especially those published in highly respectable journals during the early days of neuroimaging where 5 participants were considered sufficient for authors to draw sweeping conclusions — have a low probability to be replicable. In a recent analysis of 10,000 cognitive neuroscience papers, Denes Szucs & John Ioannidis concluded that the “false report probability is likely to exceed 50% for the whole literature”. As main reasons for replicability issues in the field, the authors discuss resource investment (you really need to get a paper out of an expensive MRI study) and analytic flexibility (data can always be analyzed until you find something, and keeping samples small is actually beneficial for fishing); see this 2015 paper by Cyril Pernet and Jean-Baptiste Poline for a critical discussion of reproducibility in the neuroimaging literature. In a related publication, Alice Carter and colleagues showed that neuroscience papers submitted to Nature Neuroscience are often not even sufficiently powered to detect large effects (which we rarely expect), and that only about 10% of the papers report formal power calculations. And Veronika Müller et al. (2016) meta-analyzed 57 functional MDD neuroimaging studies, with 99 individual experiments in 1058 patients; no consistent results emerged9.
This is not only a problem of studies in the early 2000s. This recent study, for instance, tested whether 171 serum proteins differed between 687 individuals with current depression and 420 healthy subjects. While the abstract concludes that 28 of the 171 analytes differed significantly between cases and controls, the results show that none of the effects survives correction for multiple testing. Now, I don’t mind exploratory papers, and I’m not saying that none of the markers will replicate in other studies … but there is substantial concern about false positive findings, and not reporting the results of the corrected analyses in the abstract is indicative of the state the field is in10.
2. Lack of specificity
The second problem is that results are often not specific to one disorder. The death of a loved one is not a social marker for MDD (it is also implicated on many other mental disorders), the same way glutamate neurotransmission is not a biological marker for MDD since dysregulations have also been implicated in the etiology of schizophrenia, OCD, and anxiety disorders12.
It is common to only investigate one disorder and compare it to healthy controls, and then draw specific inferences about that disorder. For instance, take this 2014 paper in Psychological Medicine, entitled “Brain grey matter abnormalities in medication-free patients with major depressive disorder: a meta-analysis”. The authors analyzed 14 datasets comprising 400 depressed patients and 424 healthy controls, and concluded that “the present study identified grey matter reduction in the prefrontal–limbic network in MDD. The subgroup meta-analysis results suggest that an increased right thalamus volume might be a trait directly related to MDD”. This does not follow, in the same way we should not conclude that increased prefrontal volume is a “trait directly related to being a human with above average shoe size[depression]”, compared to beavers[controls], when we haven’t tested humans with below average shoe size in the study[schizophrenia, PTSD]. Maybe it’s not shoe size that is related to increased volume, but being human?
In our response to a paper that identified reduced hippocampal volume as a biological marker for MDD13, we highlighted that reduced hippocampal volume is also related to schizophrenia, post-traumatic stress disorder, chronic alcoholism, epilepsy, Alzheimer’s disease, Huntington’s disease, ageing, lack of exercise, and many others14 … which means hippocampal volume cannot possibly ‘mark’ MDD; it marks pretty much everything, really.
Maybe the most relevant paper for a lack of specificity was published in 2017 by Emma Sprooten et al.15; the authors’ meta-analyses revealed that, across 21,427 participants in 537 task-fMRI studies, no consistent brain differences emerged across schizophrenia, bipolar disorder, major depressive disorder, anxiety disorders, and obsessive compulsive disorder. Sprooten et al. conclude that their “whole-brain studies did not identify any significant effect of diagnosis or RDoC domain or construct. These results resonate with prior reports of common brain structural and genetic underpinnings across these disorders and caution against attributing undue specificity to brain functional changes when forming explanatory models of psychiatric disorders.”
3. Lack of strength
Finally, there might be cases with a specific relationship between marker X and common mental disorder Y. In these cases, the relationship is often so small that it is clinically irrelevant. Going back to the example of reduced hippocampal volume in patients with MDD discussed above, the authors only reported a p-value and an effect size in their paper, so we calculated prediction accuracy in our letter, which was about 52.6% (without taking into account important covariates that are known to correlate with reduced hippocampal volume; see point 2 above). That means you need several thousand participants to obtain a prediction accuracy that is only slightly above chance, and probably not above chance after partialing out some basic covariates.
These facts did not reduce the biomarker enthusiasm. One of the authors concluded in an interview that the paper “resolves for good the issue that persistent experiences of depression hurts the brain”, and another team of researchers carried out a clinical depression trial in which they gave depressed patients drugs to increase their hippocampal volume.
Genetic studies have also often revealed associations that are both weak and unspecific. After a decade of GWAS null-findings for depression16, recent papers have identified a few SNPs that sometimes replicated across one or two datasets. But the explained variance is so low that it is usually not even mentioned in GWAS papers on depression17. It is also relevant here to keep in mind that GWAS studies do usually not partial out important covariates … so if a study does explain 2% or 5% on Major Depression, one explanation is that it explains variance of correlated phenotypes18.
4. Social factors in nosology
DSM or ICD phenotypes and disorders are often based on intuitions, experiences, historical decisions, and path dependence. That means we consider phenotypes as brain disorders that aren’t really phenotypes in the first place, but fairly arbitrary categorizations (some, like MDD, more arbitrary/heterogeneous/unreliable/invalid than others).
Kenneth Kendler summarized this beautifully in his recent World Psychiatry paper, in two quotes:
“Imagine turning the clock back ten thousand years and allowing human civilization […]. Then we wait till [a] psychiatry-like discipline decides to write a diagnostic manual [and] repeat this experiment 100 times […] What will we find? My intuition (and those of many with whom I have shared this thought experiment) is that a substantial proportion of our current categories will not be represented reliably in these manuals.”
“What would have happened if Kraepelin stayed in Wundt’s laboratory, as he wanted, and never went on to his psychiatric career? What if Wernicke, the one genuine competitor with Kraepelin for prominence in Germany psychiatry at the turn of the 20th century, had not died from a bicycle accident at the age of 52 in 1905? What if Spitzer really liked psychoanalysis and never got involved in psychiatric nosology? One can plausibly argue that, if any of these events had occurred, DSM-5 and/or ICD-10 would be meaningfully different from what they are now.”
If mental disorders are not natural kinds, not clear categories in the universe we discover, but at least partially influenced by social decisions, then the notion the specific DSM diagnoses are specific brain disorders is pretty problematic19.
In an insightful paper in 2005, Kendler summarized five requirements to determine if a scientific result qualifies for the conclusion “X is a marker for Y” 20. He argues that X qualifies as a marker for disorder Y if 1) the association is strong; 2) the association is specific; 3) the association is noncontingent, i.e. the relationship between X and Y is not dependent on other factors, particularly exposure to a specific environment or on the presence of other markers; 4) there is causal proximity of X to Y, i.e. X is directly and immediately related to Y; and 5) X needs to be the appropriate level of explanation for Y — nobody would study depression at the level of quarks.
How many common mental disorders fulfil these criteria? And if there are so few, why have scientists and funding agencies settled on calling mental disorders brain disorders?
I see two ways in which we can improve the situation in the field.
1. What are mental disorder? A syllabus
While nearly all of us reject Cartesian dualism, the distinction of disorders into 1) organic / brain-based disorders vs. 2) functional / mind-based disorders has somehow remained deeply engrained in psychiatry21. Many researchers across numerous disciplines such as philosophy of science, biology, psychiatry, medicine, psychology, and epidemiology have thought thoroughly about the question what psychiatric disorders are, and about relationships between psychopathology and biology, in publications that are too rarely acknowledged or discussed in the contemporary literature. So I sat down and created a syllabus of my favorite papers for a reading group (enough for about 1 semester) on the Open Science Framework, and hope this will be useful. This syllabus was created with feedback from, among others (in no particular order): Don Robinaugh, Ken Kendler, Lucy Robinson, Gary Brown, Rory Byrne, and Nicholas van Dam.
The syllabus covers different theories about the nature of mental disorders, which – among others — have been conceptualized as brain disorders, emergent properties resulting from complex biopsychosocial processes, and dysregulated adaptive processes22. These topics are embedded in the larger question what kinds of things psychiatric diagnoses are: natural kinds that exist in the world independently of human observation and need to be discovered (realism); practical kinds, a notion for which it does not matter if disorders are ‘real’ as long as they are clinically useful (instrumentalism); or social kinds that are constructed.
2. There is nothing wrong with biology, but biology alone ignores important aspects
I am in no way arguing against biology, or against any of the disciplines interested in biological explanations. I am also not arguing that there are no biological causes for mental processes, or no biological liabilities. And although adverse life events explain about two orders of magnitude more variance on depression or PTSD than biological variables, I would never suggest to stop funding research on biological factors. That would be silly. What I am saying is that there is no sufficient evidence for the notion that most mental disorders are brain disorders; that naive biological reductionism and the reification of mental disorders as brain disorders is dangerous for progress in science; and that the strong consequences biological reductionism has had on the funding landscape are unfortunate, because they are inherently unscientific.
Clarity in all the confusion would emerge from clear concepts, and I would be happy if authors stated what they mean when they use the term “brain disorder” in the context of mental illness. Without definitions we have no clarity, and without clarity we cannot falsify scientific ideas. And if your notion is merely that there are going to be some non-disorder-specific changes in the brain as a result of a disorder23 — the weakest stance on the biological reductionism scale — why use the term brain disorder in the first place?
Conceptual clarity seems a big issue in the field … when I recently asked researchers who work in computational psychiatry to define computational psychiatry, I got vastly different definitions. Tobias Hauser nailed it, with both tweet and gif:
Although admittedly, you would get the same for comparably new fields such as evolutionary psychology.
How to move forward? In my personal opinion, main insights in the last decades in physics, environmental sciences, biology, and medicine are all derived from embracing complexity. Stuff is really complicated, and human stuff is among the most complicated stuff there is. From this perspective, it seems silly to assume that mental disorders are simple and clearcut, with specific causes and etiologies and expressions. And complexity for psychopathology entails research that combines multiple levels of analyses, including behavioral, psychological, environmental, social, and biological factors. Such data-driven approaches provide important ways forward, and may eventually lead to categories or dimensions with higher clinical utility. Nicholas van Dam showed in a recent paper how one could go about this.
It seems like even the most arduous proponents of naive reductionism can have a change of heart. Insel, for instance, concluded in a 2017 interview in WIRED:
“I spent 13 years at NIMH really pushing on the neuroscience and genetics of mental disorders, and when I look back on that I realize that while I think I succeeded at getting lots of really cool papers published by cool scientists at fairly large costs—I think $20 billion—I don’t think we moved the needle in reducing suicide, reducing hospitalizations, improving recovery for the tens of millions of people who have mental illness. I hold myself accountable for that.”
This kind of insight gives hope for the future.
In my own view, some DSM diagnoses are more useful than others. While I know next to nothing about a patient’s symptoms, etiology, or clinical course if you tell me that he or she is diagnosed with MDD, specific phobias are more informative, and therefore more clinically useful. And it doesn’t matter if we find biomarkers for dog phobia or not, or if it is an actual natural kind – a true category in nature – or not: we know what treatments work, and what treatments do not work, and we know that the treatments that do work work fairly well. This makes it a useful diagnosis. This pragmatist understanding of mental disorders matches the one Bell concludes his blog post with:
“Finally, I think we’d be better off if we treated diagnoses more like tools, and less like ideologies. They may be more or less helpful in different situations, and at different times, and for different people, and we should strive to ensure a range of options are available to people who need them, both diagnostic and non-diagnostic. Each tested and refined with science, meaning, lived experience, and ethics.”
And I agree with Bell that we need get better a critiquing DSM diagnoses, and that overgeneralizations will not help. My own focus has been on one specific diagnosis, namely Major Depression, and I recently summarized papers that criticize the reliability, validity, and clinical utility of MDD in a series of 18 tweets, containing what I consider to be useful references and visualizations.
- Hyman SE. NIMH during the tenure of Director Steven E. Hyman, MD: The now and future of NIMH. American Journal of Psychiatry. 1998; 155(Suppl):36–40. [PubMed: 9736863].
- Leshner AI. Science. 1997; 278:45–47. [PubMed: 9311924].
- e.g., Miller, G. A. (2010). Mistreating Psychology in the Decades of the Brain. Perspectives on Psychological Science, 5(6), 716–743. http://doi.org/10.1177/1745691610388774; or Engel, G. (1977). The need for a new medical model: a challenge for biomedicine. Science, 196(4286), 129–136. http://doi.org/10.1126/science.847460.
- Bennett, MR.; Hacker, PMS. Philosophical foundations of neuroscience. Malden, MA: Blackwell; 2003.
- Unless you have a different definition of brain disorder; please do post it below in the comments section!
- Quoted from: https://www.theguardian.com/society/2018/jan/07/is-everything-you-think-you-know-about-depression-wrong-johann-hari-lost-connections
- For instance, the “neurotrophic hypothesis of depression”; Schmaal, L., Veltman, D. J., van Erp, T. G. M., Sämann, P. G., Frodl, T., Jahanshad, N., … Hibar, D. P. (2015). Subcortical brain alterations in major depressive disorder: findings from the ENIGMA Major Depressive Disorder working group. Molecular Psychiatry, 1–7. http://doi.org/10.1038/mp.2015.69.
- If you want to know my opinion why this is the case, see here.
- In many cases, worse than social psychology, as I’ve described in a blog series on questionable clinical studies on depression.
- I want to repeat myself here: these problems are not limited to biological studies, obviously. But they are problems, and they need to be acknowledged in the light of the very strong conclusions researchers in more biological fields such as neuroimaging tend to draw
- Several references in: Fried, E. I. (2015). Problematic assumptions have slowed down depression research: why symptoms, not syndromes are the way forward. Frontiers in Psychology, 6(306), 1–11. (PDF).
- Fried, E. I., & Kievit, R. A. (2016). The volumes of subcortical regions in depressed and healthy individuals are strikingly similar: a reinterpretation of the results by Schmaal et al. Molecular Psychiatry, 21, 724–725. http://doi.org/10.1038/mp.2015.199 (PDF).
- Admittedly, the authors themselves make this clear in their original paper, which makes their own conclusions that hippocampal volume is a marker for MDD even more surprising.
- Discussed by Neuroskeptic here.
- See 7 references of null-findings in Fried, E. I. (2015). Problematic assumptions have slowed down depression research: why symptoms, not syndromes are the way forward. Frontiers in Psychology, 6(306), 1–11. http://doi.org/10.3389/fpsyg.2015.00309.
- Cai, N., Bigdeli, T. B., Kretzschmar, W., Li, Y., Liang, J., Song, L., … Flint, J. (2015). Sparse whole-genome sequencing identifies two loci for major depressive disorder. Nature, 523(7562), 588–91. http://doi.org/10.1038/nature14659.
- After all, you have a ton of power to pick up these effects if you have samples of half a million people …
- Obviously, this does not mean that patients with a given DSM diagnoses do not or cannot have neurological abnormalities
- The paper is primarily about genes, but I think this holds for any markers, no matter if they are biological, social, or environmental
- And I also use this distinction here in the blog post to delineate groups of variables such as biological, social, or psychological
- As Wakefield described then: “harmful dysfunctions”
- Hippocampal volume in birds increases in autumn and winter; https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2830249/