How Mad in America Misrepresents the State of Psychiatric Genetics
Robert Whitaker at Mad in America (MIA) recently published a blog post – “Putting JAMA Psychiatry and MIA to the Genetic Test” (March 13, 2024) – that has left me perturbed. MIA, as many readers will be aware, is a website devoted to critical coverage of psychiatry. I have previously written about how the website is committed to portraying psychiatry as a failed discipline. Whitaker’s new post is prompted by a recently published viewpoint article by Kenneth Kendler in JAMA Psychiatry (which I also discussed in a prior post). Whitaker believes that this presents an opportunity for him to highlight MIA’s rigorous and unbiased coverage of psychiatric genetics over the years. The only problem is that MIA’s coverage of psychiatric genetics is anything but rigorous and unbiased. It is highly selective, and it distorts or ignores basic facts about genetics that are necessary to keep one tethered to a sensible scientific worldview. The average MIA reader doesn’t have the necessary background knowledge, familiarity with the scientific literature, or access to the relevant articles to identify the problems.
Whitaker writes:
“Thus, you can see, in our MIA Report and Kendler’s piece, two discordant conclusions regarding research on the “genetics” of schizophrenia. And here is the opportunity: if we review the evidence that is present in each of these two publications, we can assess whether Mad in America readers—or JAMA Psychiatry readers—are being provided with the most robust, and up-to-date, review of the scientific literature.”
His choice to focus on that particular article by Kendler is also highly unusual because it is not a scientific review of the genetics of schizophrenia. It is a brief article addressing a philosophical question. It only has 7 references. Kendler’s references a large genetic study of schizophrenia in support of his assertion that genetic risk factors causally and substantially influence liability for schizophrenia, but that’s basically it.
In the discussion that follows, I will demonstrate what has gone wrong with Mad in America’s coverage. First, I summarize the current state of scientific knowledge with regards to schizophrenia. Being familiar with these basic facts provides the proper context for interpreting individual research findings. In the second step, I will go through a list of research findings as described by Whitaker, and I will point out important scientific aspects that have either been left out or that, if taken into account, will help us better place the findings in the overall literature.
Step One: The Fundamentals of Schizophrenia Genetics
Schizophrenia has a high heritability in the range of 60-80%. This is a highly robust and well-replicated finding of psychiatric genetics, with convergence of estimates from studies using many different methods and populations. Heritability is a population-level statistic. It refers to the degree of variation in a trait in a population that is due to genetic variation between individuals in that population. I believe it is legitimate to debate how to best interpret the concept, to ask what heritability says about genetic risk at the individual level, and whether it is sufficient for us to characterize the condition as being a “genetic disorder” (and indeed, I have discussed these issue in a previous post). There is nonetheless broad scientific agreement that high heritability indicates a major role for genetic risk factors. The genetic factors are nonspecific, they are not determinative, they are not sufficient, their expression can only be understood in the context of gene-environment interactions, but they are substantial and they exist. To dispute this is to place oneself far outside the scientific consensus and on the fringes of scientific respectability. (On the topic of heritability skepticism, here is an excellent review of Jay Joseph’s work by Eric Turkheimer.)
There is broad scientific agreement that high heritability indicates a major role for genetic risk factors. The genetic factors are nonspecific, they are not determinative, they are not sufficient, their expression can only be understood in the context of gene-environment interactions, but they are substantial and they exist. To dispute this is to place oneself far outside the scientific consensus and on the fringes of scientific respectability.
Twin studies repeatedly show that identical twins are at high risk of developing schizophrenia, anywhere from 15% to 50% (compared to ~1% risk in the general population and 3-20% in non-identical twins). At present, genetic influence remains the most viable and plausible explanation of this increased risk in identical twins.
Aggregate genetic influence arises not only from the direct effects of specific genetic variants in isolation, but also from all the complicated ways in which the genes interact with each other.
There are daunting scientific challenges involved in translating aggregate genetic influence (represented, e.g., by family history and heritability) into genetic effects described in terms of the effects of specific genetic variants. The problem is one of scientific coherence. How much of this overall genetic influence can be made coherent and tractable? This has been described in the literature as the problem of “missing heritability.”
The largest GWAS of schizophrenia conducted to date (76K individuals with schizophrenia and 243K control individuals) revealed common variant associations at 287 distinct genomic loci. Common variants each confer a tiny increase in the risk of developing a condition.
SNP-based heritability, the proportion of variance in risk that is attributable to common genetic variants, is estimated from GWAS to be 24% for schizophrenia. (SNP-based heritability doesn’t capture rare and ultra-rare genetic variation, which also contribute to heritability at the population level.)
About a third of the SNP heritability is captured by the polygenic risk score (PRS), which adds up the effect sizes of all the SNPs that have been identified. PRS currently explains approximately 8% of the variance in independent samples.
PRS doesn’t have enough predictive power to be routinely used in the clinic, however, it is a valuable tool because it is a marker of genetic risk: the highest percentile of PRS has an odds ratio of 39 for schizophrenia compared to the lowest percentile of PRS. (Schizophrenia PRS is available for commercial use for the purposes of polygenic embryo selection – something I discussed in a prior post.)
PRS only takes into account the common variants that we have identified as significantly associated with a condition. There are hundreds and thousands of genes with tiny effects contributing to SNP-based heritability that we have not yet identified.
Whole-exome and whole-genome sequencing studies show that about 4% of individuals with schizophrenia have deletions or duplications of relatively large segments of DNA (copy number variants) or mutations of individual coding nucleotides (single nucleotide variants). These variants are rare but have larger effects than common variants.
In contrast to the above, Whitaker encourages skepticism about, if not outright denies, the 60-80% heritability of schizophrenia, and asserts that “gene variants account for perhaps 2% to 3% of the risk.” Aggregate genetic influences and genetic influences in terms of the effects of specific identifiable variants are constantly conflated.
Step Two: Scrutinizing MIA’s Presentation of Research Results
Whitaker refers to an MIA report by Peter Simons, their front-page editor and science writer, in which “Simons detailed Torrey’s review of the genetics research, and then also reviewed many of the research findings that can be found in our science archives on this topic. In particular, he cited studies that assessed the “impact” of risk genes on the likelihood of developing schizophrenia. Here are the principal articles he referred to:”
I will now go through this list of articles. The texts in italics are direct quotes.
2014: A study in Lancet Psychiatry found that environmental risk factors (i.e., perinatal brain insults, cannabis use, neurotrauma, psychotrauma, urbanicity, and migration) were a “major risk factor for early onset schizophrenia,” while “polygenic genome-wide association study risk scores did not have any detectable effects on schizophrenia phenotypes.”
What this leaves out:
Polygenic schizophrenia risk scores reproduced the association with schizophrenia diagnosis highly significantly (up to p=1·15×10−50).
PRS captures only a small subset of the overall genetic influence. And this is an older study, with a smaller set of identified genes.
PRS was associated with schizophrenia diagnosis but not with age of onset, age at prodrome, symptom expression, and socioeconomic parameters. Researchers interpreted this finding to suggest genes associated with diagnosis status are not the same genes associated with these aspects: “SNP variants associated with disease risk as aggregated into genome-wide association study-derived polygenic schizophrenia risk scores might not always overlap with risk variants associated with specific syndrome domains. Moreover, schizophrenia risk score-based analyses might not be optimised for study of association with disease-relevant phenotypes… the genetic effect is probably highly specific, and definition of biological disease subgroups or syndromes rather than building on the heterogeneous clinical construct “schizophrenia” will be indispensable for successful genome-wide association studies in the future. By contrast, the effect of environmental factors is enormous but rather non-specific, and will exert its share of damage in any genetically predisposed individual.”
2015: In Molecular Psychiatry, researchers concluded that “the current empirical evidence strongly supports the idea that the historical candidate gene literature yielded no robust and replicable insights into the etiology of schizophrenia.”
What this leaves out:
It is commonly accepted that the candidate gene era research was poorly conceived and didn’t generate valid or replicable findings. The problems of the candidate gene era, however, don’t carry over to GWAS, which is based on methods that are scientifically sound and whose results hold up well. Here is Ken Kendler (2022) on the candidate gene era and the transition to GWAS:
“… the ill‐conceived candidate gene era, where the field flaunted well‐understood rules of multiple testing. Furthermore, it was imagined that genes involved in the structure of neurotransmitter receptors and/or the uptake or degradation of these neurotransmitters were true candidate genes. However, these genes were not involved in the etiology of the disorders but in the action of pharmacological treatments – a classic category mistake. In a triumph of exuberance over common sense, these studies also yielded almost entirely false‐positive findings.
Then came the more mundane and much more effortful brute force method of genome‐wide association studies (GWAS). This humbler method was based on the fact that, for psychiatric disorders, we knew next to nothing about specific etiologic mechanisms of illness. The field properly did its multiple testing homework. What was unknown was the expected effect sizes of the risk alleles. Not surprisingly, initial estimates here were far too optimistic. The first studies with sample sizes thought to be adequate were entirely negative. Then something unexpected happened. The field abandoned important parts of the more typical academic model of inter‐group competition for a model of inter‐group cooperation, forming the Psychiatric Genomic Consortium. Positive results, that have replicated well, finally began to flow in, first as a trickle and then as a cascade.”
2017: In Biological Psychiatry, researchers concluded that “taken as a group, schizophrenia candidate genes are no more associated with schizophrenia than random sets of control genes.”
What this leaves out:
Candidate genes, as above
2019: Researchers reviewing “genome-wide association studies,” which have found statistically significant associations between large gene sets and a schizophrenia diagnosis, determined that these associations explained 2.28% of the risk that a person will be diagnosed with schizophrenia. That left other risk factors, such as the environment’s impact on biology, emotional trauma, and childhood experiences constituting nearly 98% of the risk. The study was published in Neuropsychopharmacology.
What this leaves out:
It misrepresents what the study was intended to examine: “Our study aimed to quantitatively assess whether specific gene sets, centred on putative core genes, make a larger-than-expected contribution to polygenic risk. We focused on eight gene sets, six of which are centred on genes previously implicated in schizophrenia risk.”
There is no claim by the researchers that these gene sets with putative core genes represent the entirety of genetic influence of schizophrenia. In fact, they explicitly say that their results support the omnigenic model according to which “most genes expressed in cells that are relevant to the biology of an illness contribute to heritability and PRSs because of the likely interaction of multiple signalling pathways within cells that support their biological functions. In the light of this omnigenic hypothesis, implicating a greater number of SNPs than the ordinary polygenic model would suggest, our results support the hypothesis by demonstrating a weak polygenic effect extant in every random subset of genes.”
The interpretation that “other risk factors, such as the environment’s impact on biology, emotional trauma, and childhood experiences” constitute “nearly 98% of the risk” is laughably false. It is the sort of amateur error that someone who-doesn’t-even-know-what-they-don’t-know would make. It is based on the mistaken idea that the gene sets centered on putative core genes represent the entirety of the genetic influence of schizophrenia.
The interpretation that “other risk factors, such as the environment’s impact on biology, emotional trauma, and childhood experiences” constitute “nearly 98% of the risk” is laughably false. It is the sort of amateur error that someone who-doesn’t-even-know-what-they-don’t-know would make.
There is nothing in the study to support this. The inference that 98% risk is non-genetic (i.e. environmental) is not something stated or even implied by the authors of the 2019 paper, because those authors are serious experts who know what they are talking about.
2020: In Schizophrenia Bulletin, researchers reported that sequencing the exome (a particular set of genetic information) provided no clinically relevant data for making a schizophrenia diagnosis. “The main conclusion of this investigation is a negative one,” they wrote. “The diagnostic yield for exome sequencing of known neuropsychiatric genes in this sample is about 1%.”
What this leaves out
The study did find several variants that would be categorized as pathogenic.
Exome sequencing is used to identify rare and very rare genetic variants that are present in a very small number of people. It is no surprise that the diagnostic yield will be low.
2020: Researchers from the Netherlands, writing in Schizophrenia Bulletin, reported that familial and environmental factors account for most of the risk of developing schizophrenia, and that genetics constitute a risk factor of about 0.5%
What this leaves out:
From the study: “In the combined model, familial and environmental factors explained around 17% of the variance in mental health, of which around 5% was explained by age and sex, 30% by social circumstances, 16% by pain, 22% by environmental risk factors, 24% by family history, and 3% by PRS for schizophrenia (PRS-SZ).”
Family history, of course, is an index for aggregate genetic risk! It is absurd to exclude family history from “genetics.”
Qualification by the authors: “These results cannot be interpreted as showing that genetic factors are not important. Indeed, all measures of environmental and social circumstances that were used may in fact reflect, to a degree, genetic effects. Conversely, measures of family history also mediate environmental effects…”
2021: A study of 50,000 people, which was published in the Journal of Affective Disorders, failed to find any genes that influenced mental illness. “The results obtained from this study are completely negative,” the authors wrote. “No gene is formally statistically significant after correction for multiple testing, and even those which are ranked highest and lowest do not include any which could be regarded as being biologically plausible candidates.”
What this leaves out:
This is an exome sequencing study looking at rare variants.
It is also notable that the discussion of the environment shies away from emphasizing major risk factors such as cannabis use and infections (including prenatal) while hyping up factors like emotional trauma and family dynamics, presumably because the typical MIA reader is more receptive to that narrative.
Whitaker writes:
“And here is why this is important: the corrective to the “fact” that schizophrenia is in large part a “genetic” disease leads to a reconceptualization of the nature of the disorder, and what treatments may be of most help. The genetics conception leads to a search for treatments that can, in some manner, counteract the impact of the “faulty” genes. The conception also suggests it is bound to be a permanent state—you can’t change a person’s genetic makeup when hundreds or thousands of genes are said to confer the increased risk. But if environmental factors confer most of the risk, then it is reasonable to develop treatments that are responsive to those factors. This conception also provides hope for a full recovery from an episode of “schizophrenia.””
While this argument has an emotional appeal, it makes little logical sense. Even if genetic factors confer a greater risk than environmental factors, it would still be reasonable to develop treatments that target environmental factors. The existence of genetic risk factors doesn’t preclude environmental interventions from having a large therapeutic effect. Furthermore, genetic risk (even if it is greater than environmental risk) doesn’t preclude full recovery from the condition in question. It is one thing to recognize that people mistakenly interpret genetic risk as implying that the condition is fixed and permanent and to suggest that we need to be mindful of connotations; it’s entirely another thing to accept that these false connotations are true and to actively foster the view that hope and recovery depend on a rejection of the existence of genetic risk.
It is one thing to recognize that people mistakenly interpret genetic risk as implying that the condition is fixed and permanent and to suggest that we need to be mindful of connotations; it’s entirely another thing to accept that these false connotations are true and to actively foster the view that hope and recovery depend on a rejection of the existence of genetic risk.
I have previously written about how antipsychiatry can be understood, in part, as an analogue of populism (and not the virtuous kind!). The movement expresses an anti-establishment sentiment in medicine/psychiatry, it pits the “people” against the “psychiatric elite,” it presents the psychiatric profession as homogenous, corrupt, and self-serving, and it offers overly simplistic answers to complex questions. Whitaker’s attempt to portray JAMA Psychiatry as untrustworthy and Mad in America as a trustworthy source of scientific information is a move right out of the populist playbook of delegitimizing scientific expertise. The tragedy is that the average MIA reader is ill-equipped to recognize or resist this. Writing this will not win me any friends. I am courting animosity and hit pieces; it is dangerous to poke the bear. But I feel an overwhelming obligation to point out this troubling dynamic in a landscape already fraught with epistemic pitfalls.
Journalists and laypeople are not experts dueling it out with other experts. In the absence of formal scientific training, they are often approaching complicated issues from a place of profound ignorance about the fundamentals of the field. I am of the view that those engaged in the task of communicating science to the public have a responsibility to accurately convey the scientific consensus (or, in the absence of a scientific consensus, the general opinion held by actual experts in the field) even when the writers are endorsing a view that goes against that consensus. It is perfectly legitimate to highlight research that challenges the mainstream view or that shows why some experts disagree, but it is highly problematic to present unorthodox views to the public as established truths when most experts disagree with them and do not find them persuasive.
See also:
As someone who has no scientific training and is interested in the genetics of psychiatric conditions, I appreciate your clear writing on this subject. It’s a particularly confusing subject. I intend to read some of your earlier posts.
I do feel that you are missing something, though. Maybe MIA is going to extremes concerning genetics, but they’re only doing what psychiatrists have always done. You say MIA “offers overly simplistic answers to complex questions.” But that’s just what the earlier researchers did, referring to schizophrenia as if it were a “genetic disorder” and not being caused by many factors. Before that, many claimed that it was caused by bad mothering, and there was a backlash to that—mothers got angry!—so they switched to, no, it’s entirely a brain disease. Unfortunately, everyone—scientists too!—tends to cling to theories that focus on just one cause. I guess that’s changing in psychiatry, but it’s still got a long way to go.
You also say that MIA “presents the psychiatric profession as homogenous, corrupt, and self-serving.” I don’t think the profession is homogenous these days, rather in a state of confusion! but maybe that’s because it’s in a state of change, and maybe that’s good.
As for corrupt, do you really think psychiatrists are any less corrupt than anyone else? Is ‘corrupt’ the right word for a profession that has been influenced so much by the money of the pharmaceutical companies? Would you call Freud and all who followed him corrupt in denying the evidence of incest and claiming, without evidence, that very young children fantasize about sex with their parents? Seems corrupt to me. Psychiatry has a history of corruption, as MIA, among others, usefully points out.
Thank you very much for this thorough critique of Whitaker’s misleading pronouncements about the genetics of schizophrenia—his latest salvo in the anti-psychiatry campaign he and his MIA army have been waging for some time. I am not qualified to evaluate the complexities of the genetics or the statistical arguments. But I have observed the damage that the anti-medical view of mental illness can do. Of course questions of etiology are infinitely complicated (and not resolvable in my lifetime). These are not just academic questions, however. “Have a better childhood” or “wait for a society to be more humanely organized” don’t cut it as solutions faced with acute mental suffering. The consequences of accepting Whitaker’s anti-medical propaganda can be dire. He positions himself in part as a crusader against the greed of Big Pharma, but he too has a dog in the fight. I don’t claim to know his motives, but Whitaker's livelihood is dependent on promoting his own ideological position. He is a journalist and the exposé is his bread and butter: An important function, but not when the baby goes the way of the bathwater and no realistic alternative is on offer.
Perhaps it is human nature to take a dim view of doctors—like undertakers, they inhabit a realm we would all rather avoid. They become associated with illness and pain and so end up being blamed for its existence.