Is Schizophrenia as ontologically defined as the Flu is? There have been plenty of cases of it being misused as a political diagnosis, in the Soviet Union for instance, and even if there's a known ontology it would still be equivocation if it's never clinically tested for and the diagnosis is still only based on symptoms.
I’m surprised that despite a dozen arguments on this list, not one of them is the arguably-correct one: that heritability might be a bad match for human intuitions of what “mostly caused by” means. (e.g. the low concordance probability of schizophrenia in twins, which is counterintuitive. We can respond to this either by saying concordance probabilities are a bad operationalization of “caused by genes,” or by saying the biggest cause of schizophrenia is “shitty luck” or “nonlinear responses to inputs that don’t add up to 1.” Of course, this buys you a whole heck of a lot of other paradoxes instead.)
Polygenic screening doesn't prevent schizophrenia in a meaningful way, though. It just prevents those embryos with high risk from developing into people (modulo your beliefs about when personhood begins).
...Who the hell has been saying epilepsy isn't genetic? It's pretty damn obviously genetic. Barring the ones from people getting whacked on the head really hard.
> By comparison, you can very clearly halve your children’s risk of schizophrenia through polygenic selection, which costs only a few hundred dollars if you’re already doing IVF.
Last time I looked into it, polygenic risk scores for schizophrenia were only able explain ~2% of variance of schizophrenia. It’s really hard to recruit enough schizophrenics for variants in a GWAS to reach genome wide significance. And even with GWAS, you can’t measure things like copy number variants or rare variants. I think it’s at least another few decades until we can screen out schizophrenic embryos.
I'm curious how the part about "schizophrenia is so famously hard to define that we’re not even sure it exists" squares with "studies from 1970s China find exactly the same heritability as Western studies". If it's that hard to define, isn't it likely they're using different definitions in different studies, making direct comparisons pretty hard?
Regarding argument number 3 specifically, neuroscientist Kevin Mitchell wrote a great book called Innate in which he explains how many mental disorders seem to be caused by non-specific genetic mutations that increase developmental noise. His book is full of insights.
Schizophrenia serves an important social function by generating people who assert and act according to blatantly false ideas for non-predatory reasons:
* Based function: This allows sociopaths to prototype new con games while scapegoating non allied persons who act similarly enough to be confused for them
* Woke function (theoretical): This allows neurotypicals to discern deeply entrenched con games and the people who play them
I see that much of the discussion about causation in this context is still based on verbal arguments and common sense. On the other hand I keep hearing that DAGs and the other machinery developed by Pearl and others is widely applied to epidemiology. Did anyone try to throw a bunch of potential risk factors, genetic or not, into a causal discovery framework to see what happens?
The other environmental risk factors for schizophrenia are equally hard to change. Poverty? Okay, don’t be poor, thanks for the important life advice. Social defeat? “Doctor, are you saying I have to never let anyone defeat me?” “Yes, it’s my official medical recommendation that you become invincible.”
"This is both totally antithetical to the spirit of real clinical medicine as it’s practiced, and ethically odious to anyone who has witnessed the side effects of antipsychotics first-hand. "
You've perhaps already written it, but an essay on side effects of psychiatric drugs (or even of drugs in general) would be interesting and useful, and a good counterweight to "just take some pills, it's no big deal" thinking.
I more or less agree with all of your arguments presented here, but I disagree with the conclusion. I don't think the solution is to assert that "genetics cause schizophrenia", but rather as a society we should be a whole lot more careful about using the word "causes" in scientific research (including smoking/lung cancer).
I understand your frustration with applying a different bar to smoking and psychological diseases, but one has to start *somewhere* to enact societal change in the level of confidence that scientists imply when they are speaking with the lay public. If someone is a psychology researcher they probably will start with psychology stuff. A cancer researcher may simultaneously be pushing back against the "smoking causes lung cancer" narrative in a way that you don't see because your day job doesn't involve following prominent cancer researchers who are labelled (possibly inappropriately) as Big Tobacco puppets.
There are some things that scientists are quite confident about (nearly all of mathematics, most of physics, a lot of chemistry, some molecular biology, etc.), and there are some things that we have really good correlational data on but don't fully understand (nearly all of psychology, etc.). For the latter set of things, it is very likely that one day, as you suggest in this article, we'll find that "computational center" and it will all make causal sense. At that time, we'll find that genetics were not causal, just risk factors like in your kidney example, and we'll be able to create a very clear causal chain for each specific instance of schizophrenia. If we say "genetics cause schizophrenia" today, then when that future day comes we'll have to say, "sorry, we were dumbing it down for you, but now we know for sure what causes it" and no one will believe you (fool me once, shame on you, fool me twice, shame on me).
I personally find that way too many scientists are far too eager to assign causation long before they should. IMO, we shouldn't say we know causation until we can predict future outcomes with near 100% certainty (like we can with all of mathematics, most of physics, etc.) At the moment, our predictive capacity for both lung cancer and schizophrenia is appallingly low and doesn't get anywhere near reaching the bar where I think using the word "causal" is appropriate.
Leaving that aside, the whole concept of disease and causality is complicated, and the very idea of something "causing" a disease is context-dependent.
For example, every child born in the U.S. gets tested for a condition called PKU. If you look at a diet Coke can, you will see a warning that it “contains phenylalanine.” PKU is a genetic disease, and the people who have it have a mutation in the gene for the enzyme that allows them to digest phenylalanine.
But take a step back. If instead being a tiny percent of the population, if the mutation were prevalent – for example, like the mutation that allows us to digest milk – then it wouldn’t even be considered a disease. It would be considered normal, and phenylalanine would simply be considered to be a toxin.
Similarly, if instead of 1% of the population, 99% of the population had the CCR5 mutation that made you immune to HIV, then HIV would not be an infectious disease but a genetic disease instead.
Diseases are intellectual constructs. They’re like boundaries of countries. Sometimes they’re based on the contours of natural geography, much like borders that wind along rivers or stop at mountains. Other times, they’re arbitrary, like vertical lines dividing some of the American Midwestern states. What’s the dividing line between pneumonia vs. bronchitis? What’s the difference between a heart attack and angina? Where we divide the diseases into categories and how many divisions we draw, as physicians, are not God-given. They’re man-made.
Schizophrenia is particularly germane example because it is a mix of what are probably very different processes. Catatonic schizophrenia is very different from classic forms. Doctors tend to lump very different diseases into one category if (as was true for a long time for schizophrenia) there isn't much you can do about it. Generally, diseases are categorized by what therapy they respond to, because pragmatically, if you're going to treat the patients the same, you lump them together. And if not, you separate them. (In some instances, especially in psychiatry, diagnosis can be driven by reimbursement considerations too.)
Here is an example of a disease that some pharma companies invented: high blood pressure. Until the 1970s, there was a great deal of debate over whether hypertension was a disease that needed to be treated, especially in the elderly. Some physicians thought that as we aged, and as our blood vessels became stiffer, hypertension was a natural effort by the body to maintain blood flow to all the important parts of the body. It wasn’t until a strenuous effort by Merck and other companies to change people’s minds that it became accepted that hypertension had to be treated.
I should note that it is not clear still how, when, and with what we should treat hypertension. As more data comes out, it is becoming clear that some drugs work better than others (and some are actually harmful). Not too long ago, the guidelines were changed to make treatment of elderly patients less aggressive. The complexity of the consensus guidelines suggests that there are still many unanswered questions. The very fact that there are guidelines means the answer is no obvious - only when there is uncertainty is a guideline needed.
Not sure it matters much to the bottom line, but I’d be cautious about using the terminology associated with smoking risks as the sort of default baseline. Seems to me like this is an atypical situation, where there were enormous policy influences on the language and terminology associated with smoking; there was a huge (justified and good) campaign to discourage smoking, and the public health community was heavily involved in the messaging in ways that shouldn’t necessarily be assumed to be normative for every other terminological situation. I’m not saying that the scientific community wouldn’t have reached the same “smoking causes cancer” formulation even in the absence of the unusual public health campaign surrounding the issue; just that it would probably be safer to use a different and less policy-laden example as a semantic baseline for this sort of thing.
I'm far more hereditarian about psychological and behavioral traits than most liberals, so it's funny for me to take the non-hereditarian side of this one, but I'll at least offer some thoughts.
The reason we readily say that smoke has a causal, not merely a correlative, relation to cancer is not because of the statistics (although they are important). It's because we have a mechanistic explanation of the causal pathway. Tobacco has over 70 carcinogens that directly cause DNA damage, in addition to oxidative stress and chronic inflammation.
Now, of course, whatever genes do is also ultimately mechanistic, but here's where the similarity ends. We have no mechanistic account of how genes might cause schizophrenia. Of course, one might protest that that's just a statement about our knowledge. But it's worse than that. It just doesn't seem likely that whatever is going on neurologically has a causal pathway anything as straightforward as the effects of carcinogens. The interactions with the environment seem like they're far more intricately involved. It's as if smoking caused cancer, but only at certain altitudes, within certain temperature zones, when you ate certain foods, etc., etc. If the latter were the case, we'd be much more hesitant to just say "smoking causes cancer."
The key issue at the heart of the debate: whether as describing it as mostly genetic, we can then use genetic scores to motivate clinical action. See Kendler et al., 2023: Relationship of Family Genetic Risk Score With Diagnostic Trajectory in a Swedish National Sample of Incident Cases of Major Depression, Bipolar Disorder, Other Nonaffective Psychosis, and Schizophrenia
Scott, obesity is also 80% genetic. Can I hear you say "fine, obesity is also genetic"? Do you bite this bullet?
Of course, it does not match what we colloquially mean by "genetic" at all, since obesity is 100% environmentally malleable. Can we *please* adopt a more sophisticated frame than "is genetic" vs "is not genetic"?
You are modelling the world as if everything is a sum of small contributing factors. But it's not; sometimes there are AND gates. Sometimes you need to have the genes AND to eat the right food. It is not possible, conceptually, to assign a "percent heritability" in such a situation.
1. Out of curiosity. Kidney disease heritability. "We report heritability estimates of the CKD-defining traits eGFR (44%), UAE (20%), and UACR (19%), for the kidney biomarkers serum urea (31%), Scr (37%), and uric acid (48%), and for the serum electrolytes potassium (28%), calcium (27%), and sodium (22%) " They checked and basically every kind of family member with the disease elevates your risk, even "genetically unrelated" ones like spouses (assortative mating or maybe shared environment). https://www.sciencedirect.com/science/article/pii/S0272638620311628 It doesn't seem to be impressively heritable, but there's obviously some genetic causes of this, whatever they might be exactly.
2. There are some single mutations that cause schizophrenia alone, or schizophrenia like traits at least. https://www.pnas.org/doi/abs/10.1073/pnas.2112560118 "A total of 48.2% of individuals with severe, extremely treatment-resistant schizophrenia carried at least one rare, damaging missense or loss-of-function variant in intolerant genes compared to 29.8% of typical schizophrenia individuals".
3. Family studies yielding estimate of heritability, shared environment etc. tell you where to look for causes. They give you the overall importance of some group of causes, but doesn't tell you what they are except in a broad sense. In other words, they tell you where to dig for more. They also give you a decent idea about what kinds of interventions are going to be effective. When heritability is very high, it means that genetic selection will be effective. It scales linearly with the square root of the variance. If "family environment" doesn't show up or is very minor, then trying to modify risk factors in that area will not be effective unless you somehow introduce a new kind of thing that doesn't already vary between families. https://www.emilkirkegaard.com/p/high-heritability-does-mean-that
4. Assortative mating and measurement error are usually quite large downwards biases in family studies. This is particularly the case for self-report symptoms of mental health vs. more objective data (e.g. clinician ratings). I wrote a summary of such research. https://www.emilkirkegaard.com/p/heritabilities-are-usually-underestimated
You say it yourself, schizophrenia is quite hard to neatly define. So much so that perhaps it doesn't exist / isn't one unified thing.
I think most of these writers who are arguing that schizophrenia isn't genetic, and some of those arguing that schizophrenia shouldn't be prevented (only treated), they mean to instead argue "schizophrenia isn't real".
They don't make that argument, and instead get wrapped up in denying the practical facets you lay out, because they are the sorts of people who are not comfortable rejecting the existence of schizophrenia, for social-norm reasons. The people who feel "woke" social norms to treat-not-prevent are also the ones who have a cognitive bias towards saying it's intervention-able (not genetic).
But strip away their social and cognitive biases, and they probably would wish to unpack "what is schizophrenia actually? It's not well defined enough, there's more to learn. "
Regarding 6B, genes could be a proxy for a more satisfying cause of schizophrenia. It is not hard to think of ways this could work. As you say, one is that genes could make people more vulnerable to some virus that attacks the brain and causes schizophrenia. There are many other possibilities of this kind. What they all have in common is that you can think of them as genes that change the person’s environment, causing it to be more schizophrenogenic. (For instance the person in your example with the bad immune system lived in an environment that contained more dangerous viruses.) Here’s a list:
-Genes that cause someone to be physically unattractive, or to have other characteristics that make it less likely they will receive social approval: They’ve got poor gross motor coordination and are bad at playground games, and so they get less acceptance and approval from peers. They’re slow to learn to read, and have an especially hard time sitting still and shutting up in school, and so get less approval from their teachers. Kids with these genes live in an environment with fewer smiles per year.
-Genes that cause some disability that interferes with the person making sense of their environment, and is subtle enough not to get diagnosed in most people: prosopagnosia, difficulties putting your thoughts into words, some spacial orientation problem that makes you get lost easily. People with these genes live in a world that’s more confusing and harder to navigate.
-Genes for some as-yet-unrecognized disorder that’s like PKU (phenylketonuria), but instead of causing severe developmental delay causes, by age 20 or so, whatever sort of brain damage underlies schizophrenia. People with these genes live in a world where ordinary foods cause brain damage.
-Genes that cause people to have unusually strong reactions to stressors. For these kids, seeing a squirrel get run over is as upsetting as seeing a hanging would be for others. People with these genes live in a world with an unusually large number of truly traumatic events.
The idea that social problems are harder to fix than biology is a prime example of Americentrism. Other countries have made far more progress than the US in reducing drug abuse, poverty or social conflict (i.e. instances likely to lead to social defeat). I think this is because the US is unusually individualistic and conservative (and Conservative) for a developed country.
I also think it's rather flippant to suggest people undergo IVF to address genetic problems. I have two daughters by IVF. IVF is very rough on the mother, far more so than the side effects of schizophrenia drugs (which I've also had the misfortune of experiencing). This may improve with better IVF procedures, but we can also discover better drugs.
I know this got addressed in a paragraph above, but didn't Scott just get finished with demonstrating (with stats and maths) that, because of the low incidence of schizophrenia, the causes in any specific individual with schizophrenia will mostly be environmental.
Running the numbers with Scott's methodology by way of example, this means that you could screen 500 people for schizophrenia risk factors using a microarray or something, set the cutoff at a 0.95 genetic risk factor, wind up with 25 'candidate' schizophrenics (of which 4 will actually develop schizophrenia) and still miss one actual schizophrenic.
Which, in this example, means that your expensive microarray test has an 84% false positive rate, and a 20% false negative rate in practice.
And if you bump your threshold to 0.99, then all you get is 7 candidate schizophrenics, of which (in this example) none happened to actually develop the condition.
Doesn't this put screening for purely genetic factors for schizophrenia firmly in the 'scientifically true but useless' category?
A distinction I've been thinking about which I've been unsure about but which seems reasonable/important to me:
Schizophrenia is related to psychosis-propensity, arguably caused by psychosis-propensity but also arguably distinct from it in that schizophrenics tend to get convinced they are being persecuted and then retreat into themselves.
It seems to me that one could hypothesize that the mostly-genetic, biological brain-brokenness causes the psychosis propensity, but then the other aspects of schizophrenia are caused by a psychosocial dynamic where e.g. outsiders react badly to the hallucinations and therefore confirm the schizophrenic's suspicions that they are in on the conspiracy.
If true, I don't know how this affects the conclusions in your article. I think it depends on how damaging the psychosis is on its own, vs how damaging the downstream social dynamic is. If most of the damage is with the psychosis-proneness then this is just kind of nitpicky. But if most of the damage is with the social dynamic, then e.g. it raises the question of whether one could cure most of schizophrenia by just averting that social dynamic, and therefore it starts seeming reasonable to think of schizophrenia as environmentally Caused.
While one needs to understand that there are a lot of philosophical nitpicks and therefore it's hard to describe anything as true, I also think it's worth clearly mapping out the philosophical nitpicks. For example, if they don't get clearly mapped out then people might not realize they are there (try telling people schizophrenia is 80% heritable and then ask them about the twin concordance rate), and even if they do realize they are there, mapping them out is the first step to generating theories about how to most accurately take them into account.
I think it also deescalates conflicts. If you've found an important-seeming nuance that doesn't get described very clearly, you might want to generate clearer descriptions. This could lead to some cooperate work if people assist you, but it could also lead to ideological conflict if people dismiss it as philosophical nitpicks and you then start writing big books about it
Re your thoughts on viruses: in my experience teaching students, when talk about causality, they talk about levers they think we can pull on. They’re not interested in causal inference for levers we cannot pull. Many people don’t know about polygenic embryo selection as a lever you can pull or have convinced themselves it’s immoral to pull on it. It’s not a consistent definition of causality, but it’s a sensible position to look for levers. You just happen to have a niche position on which levers are easy to pull, the rest of the disagreements come from that.
Of course a widespread virus being necessary would change our idea of which lever to pull first, ie people are trying vaccines for Epstein Barr now rather than do embryo selection for genetic risk of multiple sclerosis. Because it’s much easier and proven and (for some) not as ethically dubious
'In a society where everyone inhaled radioactive dust all day, smoking would be an utterly insignificant cause of lung cancer compared to all the radioactive dust inhalation'
Slight nitpick - cigarette smoke IS 'radioactive dust', see e.g. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2672370/, where they estimate a smoker may be exposed to an additional 10 Sv, or 1000 rem, in 25 years.
This discussion feels a bit confused. You and your opponents certainly would both agree on a sufficently precisce statement of the claim in statistical language. So when you say you're arguing about whether "schizophrenia is largely genetic" you aren't arguing about whether the facts comport with some agreed on meaning of that phrase but whether that phrase has a meaning which is made true by those facts. And yet, you aren't actually just making a boring corpus data/polling based claim that most competant speakers do give the phrase such a meaning.
I feel like what's going on is the same thing that goes wrong in debates about whether trans-women are women. The people in the debate think they are arguing over a fact about the external world like whether smoking causes cancer but in fact, as you pointed out in a great piece, they are really making normative arguments about what would be the more desierable way to use language (should we assign a meaning to woman/schizophrenia is genetic that includes...?).
In other words, what you described as why you care is actually the argument: using the word the way your opponents prefer causes these harms because it misleads people about what responses will be effective. I agree and find that relatively convincing but that's the actual debate and it would be better to phrase the other stuff as: responses to potential arguments that the usage you favor is also misleading/confusing.
It seems like there's a lot of muddled discussion over "cause." Causal inference is an entire field with a pretty simple and well-hashed-out definition of a "causal effect." From Miguel Hernan's "Causal Inference: What If?":
>We compare (usually only mentally) the outcome when an action A is taken versus the outcome when the action A is withheld. If the two outcomes differ, we say that the action A has a causal effect, causative or preventive, on the outcome. Otherwise, we say that the action A has no causal effect on the outcome
Concretely: Alice smokes cigarettes, and gets lung cancer. If she would not have gotten lung cancer in the counterfactual world where all else was the same, then cigarettes caused Alice's lung cancer.
Of course, barring a time machine or an ersatz replacement (e.g. a cross-over study with a washout for a drug, or a twin study), we usually can't determine individual causal effects. Sometimes individual causal effects are not even that interesting. Maybe Bob developed lung cancer after being hit by a stray gamma ray.
However, we can estimate average causal effects at the population level with a variety of tools, ranging from excellent (e.g. an RCT) to sometimes-questionable (e.g. a haphazardly-specified regression model applied to survey data of questionable accuracy).
> Actually it’s more complicated than this: if you look at the simulation argument from last time, it’s hopefully obvious that every case is caused both by genetic and by environmental factors, or that it’s not really meaningful to try to disentangle them. I don’t know whether real schizophrenia is like this.
Real schizophrenia is definitely not like the simulation in a relevant way which I will detail below:
Your model involved assigning two types of risk levels to people, summing them, and then diagnosing scizophrenia if the summed risk exceeded a threshold. The threshold was higher than the maximum value of any individual risk component, and therefore it wasn't possible for any component to be solely responsible for a schizophrenia diagnosis. (But, by design, it was easy for one component to contribute five times as much "problem" as the other component.)
The problem is that your risk levels were ordinal numbers, not cardinal numbers. This could only be valid as a model of schizophrenia if schizophrenia were diagnosed solely by comparison to everybody else. Objective criteria, such as hallucinations, invalidate the assumption.
(It's also weird that your model involves adding an ordinal genetic risk to an incommensurable ordinal environmental risk. We try to avoid adding ordinal numbers. By assuming that genetic and environmental risk are normally distributed, you can convert centiles to standard deviations. But I think you'd end up unable to say whether a standard deviation of environmental risk was more or fewer units of risk than a standard deviation of genetic risk, which is kind of the opposite of the point of the simulation.)
In reality, we are surely interested in a model of schizophrenia where you get small numeric contributions from many causes and schizophrenia happens when a numeric threshold is exceeded, not where you get large comparable-but-not-theoretically-measurable contributions and schizophrenia happens when a comparative threshold is exceeded (that is to say, when a certain number of people are doing worse than you).
When I went to a class/group mental health thing for ADHD a few years ago, one of the things I learned was that ADHD is considered "epigenetic": you need the genes, but the genes have to be activated by environmental factors (lead exposure was mentioned) before you really "get" ADHD. Would calling schizophrenia "epigenetic" satisfy both sides of the argument? Also, is this something people still think is true about ADHD?
> Would we really? If schizophrenia was caused by an omnipresent virus, but the only people who got it, got it entirely because of genetic differences from the rest of the population, why would we call the virus “the cause” rather than the genes? Is it just because we’re used to calling pathogens the causes of things from our long history with infectious disease? It might be equally reasonable in this case to think of schizophrenia as a genetic immunodeficiency. I don’t know exactly how to think about things like this.
I mean, it kind of sounds like you do know how to think about things like this, but you don't want to do it.
Suppose there were oxygen lying around everywhere (as in fact there is!), and some people had genes that caused them, when exposed to oxygen, to harness it to cells in their circulatory system and generate energy in specific areas through controlled chemical reactions, while other people had genes that caused them, when exposed to oxygen, to catch fire and burn to death.
Would that be a genetic problem? Could we fairly attribute it to the genes?
Of course; everything you ever do is caused by your genes, because there is no theoretical alternative. The range of possible reactions of an organism to different environments is one more thing that is fully specified by its genes.
> even though smoking causes 80% of lung cancer, if Alice gets lung cancer and Bob smokes exactly the same as Alice, Bob will have a much lower than 80% chance of getting lung cancer. The solution isn’t to stop describing smoking as causing lung cancer, it’s to get better intuitions.
This example makes me queasy; the two 80%s do not refer to related numbers, and neither of them is related to the problem you're discussing in the post. I want to think about the link between smoking and lung cancer in terms of risk increase, something along the lines of "in these two otherwise comparable groups of nonsmokers and habitual smokers, the smokers get five times as much lung cancer". It's never relevant how much of lung cancer is caused by smoking; you'd need to normalize that against the number of people who smoke. This is just the confusion between P(smoker | lung cancer), which in the example is at least 80% (though actually more†), and P(lung cancer | smoker), which is Bob's chance of getting lung cancer from being a smoker.
But while confusing the two directions of a conditional probability is a common problem that occurs when people try to think about probability, it is not the same as what is otherwise the focus of this post, which is the meaning of claims like "80% of the variance in random variable X is explained by random variable Y, a component of X".
This makes the thesis "this is equally true of everything else" a bit weird. The smoking / cancer example provided in support isn't analogous to the schizophrenia / heritability claim and doesn't transfer. It could be an example illustrating that we need to get better intuitions about everything, and, sure, I guess that would be nice...
† Why more? By assumption, 80% of lung cancer patients developed their cancer by smoking, and they are all, by definition, smokers. This gives us a lower bound of 80%. But there will be some smokers who developed lung cancer even though their smoking didn't contribute to it, the same way nonsmokers develop lung cancer without any contribution from smoking, and those smokers-with-lung-cancer-unrelated-to-smoking will raise the rate above 80%.
None of those arguments make up the true rejection. Instead, like you say, if genetics could be admitted as the cause, then it would likely be possible to describe any workable solution as EUGENICS, which is obviously far too horrible to contemplate.
I don’t see any particular issues with this statement. Many forms of epilepsy are highly heritable. I’m currently trying to solve my husband’s and sister-in-law’s epilepsy and there’s a ton of data that indicate that at least propensity for seizures is highly genetically determined.
"This is possible, but I think it’s false. We know many apparent risk factors for schizophrenia: cannabis use, birth canal asphyxia, social defeat, toxoplasma infection, poor prenatal nutrition. If you try to combine these into one big picture, it doesn’t work."
What the - cocial defeat causes schizophrenia! How? Why? Is this another case of X is not about X, it's about signalling?
> The only thing that makes sense to me (and I’m definitely not an expert, and E. Fuller Torrey is, so take this with a grain of salt) is thinking of schizophrenia as cumulative damage to some abstract computational organ.
If that's on the right track, then lots and lots of genes would presumably be relevant, including glutamate receptors, myelination, dendrite structure, synapse structure, neuron resting potential, and on and on.
well-argued as usual. You address only briefly our weird reluctance to acknowledge that schizophrenia in fact does have a large genetic component. This reflects something about current society... in past eras, genetic contributions to behavior were taken for granted and often exaggerated. What's the deal here?
Much of what you say about the comparison with smoking causes cancer seems obviously right. But I guess the meat (or perhaps, heat) of this argument is about what happens downstream of the statement, "schizophrenia is caused by X": where money for researching therapies ought to go, or what advice to (potential) sufferers should be.
The statement "schizophrenia is mostly genetic" is, on the face of it, a suggestion to direct research dollars to genetics research; for sufferers it is perhaps perceived by Fuller as a counsel of despair. (I wonder if that is a motivator for him making the argument?)
Two things in your post leapt out at me as worthy of comment:
(1) "Kidney disease is just a catch-all term for “anything causing your kidney to work less well than it should”. If you examined its heritability, it would probably be substantially genetic."
Assuming this is true, this seems like an argument against saying "schizophrenia is mainly genetic," because that's not something we commonly say about kidney disease (is it?).
(2) "schizophrenia is caused by some extremely common virus...everybody has all the time...immune system controls the virus...genes for bad immune systems...virus causes schizophrenia."
I just wanted to raise the obvious analogy of HPV here. Once it was discovered that HPV was a key factor in cervical cancer, research moved to HPV prevention, culminating with the HPV vaccine, which as far as I know is now doing an admirable job of reducing infections and the downstream cancers. This seems to me to illustrate what I was saying above: the argument about what we should *say* causes condition X is really an argument about how we should go about researching and treating condition X.
Back on a more general level, the issue you point up of different common medical terms for diseases/conditions/syndromes being on completely different levels of the symptom-condition-cause spectrum just makes any generalisation hard to do. Particularly for something like schizophrenia, where it's still unknown whether that word refers to a causally disparate collection of different symptoms; a single, fairly well-defined syndrome; or a monocausal disease. Any conclusion you might draw about how to talk about particular kinds of diseases may or may not apply to schizophrenia...
I am sad to see this bunch of non-arguments that neither grasp the philosophical question, the science communication problem or the scientific problem.
The philosophical question is, when multiple prerequisites are necessary, but neither is on their own sufficient, what do we consider the cause of something? This is a persistent problem in medicine. When an old person dies of a rhinovirus infection, what is the cause of death? The rhinovirus or old age? A young person would have surived the virus without any problem whatsoever, so it is reasonable to say old age, but then again, without the virus, the person might still be alive, so it would be equally reasonable to say the virus. The sad fact is that there is no scientific way of assigning what amounts to moral blame. Scott would like to use the numbers™, but heritability is a statistical expression of variance, not quantitative causal contribution or population attributable fraction, it rests mostly on outdated evidence and is misleading regarding the preventative effects environmental interventions may have. Neither of these points is adressed in the post.
The science communication question is whether we should advertise heritability estimates as relevant. The original article claims we should not, and nothing Scott argues changes that very much. Environmental interventions still can have large effects on incidence in diseases with high heritability. Scott ignores this for some reason.
Lastly, the scientific problem involves the high heritability estimates of twin studies, which were not replicated by GWAS. It is yet unclear whether they will be replicated by whole-genome sequencing studies, but the likelihood is low in my opinion (WGS studies claim better sensitivity to rare variants, but this is unlikely a major contributor do disease burden due to selection pressure). Polygenic scores do not explain much more than the GWAS estimates, either. The missing heritability problem is easily resolved when assuming twin studies simply give overestimates due to GIGO. This has been argued for a decade now (doi: http://dx.doi.org/10.2190/HS.43.2.f). So what are we left with here, exactly? An isolated demand to strongly believe in outdated results based on poor methodology. This is motivated reasoning at its finest.
Which brings me to the last point, "if you are already doing IVF" - no, most people are not "already doing IVF". If you need IVF you either missed the on-ramp to having kids or there is something biologically wrong to begin with (some might argue that the former is an expression of the latter, but I digress). IVF is a disagreeable procedure for women to go through. It likely leads to inferior pergnancy outcomes compared to, oh, just having sex. Preimplantation genetic testing is an experimental procedure with little long term data that barely passes ethical muster in obvious cases of monogenetic diseases. Recommending PID to a patient willy-nilly based on an abstract desire to improve their offspring might (at least ought to) get your clinical license yanked.
Polygenic screening was never proven effecntive in a clincal trial. The 50% reduction in risk you were sold is not based on RCT evidence. Selecting against schizophrenia might select for something else, for better or worse. Your mileage may still vary overwhelmingly.
I used to appreciate this blog because of the thoughtful discussion of medical evidence it offered on psychiatric treatments and diseases. Reading this series of articles both suggests that these skills do not translate well to other fields of medicine and casts doubt on the quality of previous discussions.
> if we knew for sure that all of the genes involved in schizophrenia were for an autoimmune condition, we might say “schizophrenia is an autoimmune disease”.
My understanding of polygenic scores is that it is a statistical method.
My reluctance for a (voluntary) eugenic solution depending on these scoring is, that we don't understand these genes and if there is a side effect? It is like deleting programming code without knowing if it maybe also had beneficial effects? Maybe in this autoimmune genes are also supergood for surviving covid or preventing allergies. Maybe they are involved in creativity and erradicating slightly crazy genes would lead to a boring dystopia of bureaucrats out of tune with the spirit world?
My favourite example is that wearing earrings is very heritable in modern human population. Based on whether person has two X chromosomes or not, one can predict quite well whether they are wearing earrings or not.
>You also can’t point to any individual lung cancer patient and say “smoking caused this person’s lung cancer”.
Actually we can! Tobacco smoke causes characteristic types of mutations in cells, and if we see those mutations in someone's cancer, it's pretty certain that it was caused by smoking.
The natural way to describe such a situation would be: the virus causes schizophrenia, and certain people's genes cause them to get schizophrenia.
Does that distinction help for the issue at hand? I think it does help clarify what we mean by causation. As for what we mean by "genetic", not directly, but it does point toward a more precise way of describing the situation: "genetically mediated". Would we be having this argument if the question was about schizophrenia being genetically mediated? I don't think so.
I'll note that the vast majority of smokers don't get lung cancer (LC), so even the perfectly reasonable sounding "smoking causes lung cancer" is suspect. It's true that smoking massively increases the odds of LC (i've seen ORs from 20x-40x), the base rate is so low that even for smokers the lifetime risk of LC is less than 1% (at least by my quick and sloppy estimates from CDC/NIH numbers. My hope is that by being wrong on the internet I can get someone else to do this, and even if I'm off by 50x I'm still right.)
The only use I've gotten out of calling things 'causal' is to indicate that modifying one variable will meaningfully affect another variable. It's about actions one can take that are expected to do the thing you think they should. I'm okay with saying "smoking causes lung cancer" because I believe that modifying smoking habits will clearly and directly decrease LC risk, whatever mechanisms or moderators or whatever are "actually" responsible.
From my perspective, we are in this odd position where we know that a substantial aggregate genetic influence on schizophrenia exists (and that such substantial aggregate influences exist for nearly all complex disorders, even if not to the same extent), but we seem unable to make it tangible. The influence dissolves into genes of tinier & tinier effects, and this seriously limits what we can practically do with all this. It’s obvious that hundreds and thousands of gene influence the risk of developing schizophrenia.
We need to distinguish between several different strands of the debate here. “Are genetic influences in aggregate a cause of schizophrenia?” is one question (and I’d say yes, in the same sense of cause as smoking causes lung cancer) but that is a different question from whether “Is the etiology of schizophrenia 80% genetic?” The latter question is rather ambiguous but that’s how many ppl talk about schizophrenia. My view is that it is 80% genetic only in the strict sense of the heritability statistic for a population, and that the heritability statistic becomes kinda meaningless if you try to turn it into a statement about what component of “etiology” (etiology ≠ phenotype variation) is genetic.
A second thing we need to distinguish is that we mean different things by the term “genetic disorder.” It could mean something very general, like any disorder where a substantial portion of the risk/vulnerability comes from aggregate genetic influences. A second more specific sense is where identification of causal genetic variants provides a coherent way to understand the mechanisms or pathophysiology of a condition. Schizophrenia genetics program was kinda based on the hope that we could move from the former sense to the latter sense, and it has become rather clear by now that this is not likely to happen. The influence of any particular gene is too small & current polygenic scores explain like 6-7% of risk. What people are hoping now is that genetic associations can provide clues to understanding what pathways are involved, like disruption of synapses or immune dysregulation. I don’t mind if we consider schizophrenia as a genetic disorder in the first sense (which means many other complex disorders are also genetic) but we shouldn’t simply assume that because it is genetic, the latter sense is also true.
Agree that preventing SCZ is good, and polygenic screening probably helps more than other known interventions, and most of the arguments you reject ought to be rejected, except #7. Violating assumptions of heritability estimates is not a small point.
First, the limitations of adoption studies are not just the importance of the intrauterine environment. An adoptee's adult behavior is caused by a combination of 1. intrauterine environment, 2. early-life experiences before adoption, 3. post-adoption environment, 4. direct genetic causal effects on behavior, and 5. indirect genetic causes on behavior (e.g. gene → physical appearance → another trait). I would not confidently rule out any of these factors without experimental evidence directly controlling for them, which in the case of gene-by-environment interactions (#5) is essentially impossible.
For twin studies, it feels extremely implausible that the equal environment assumption is even close to true. Glancing at the paper you cited, they seem to find that the ways in which identical twins' environments are more similar only have small relationships with behavior. However, the value of this analysis depends on accurately identifying how identical twin's environments are similar. It seems this paper neglected some plausible ways in which identical twin's environments are more similar than fraternal twins (genes → attractiveness → people being nice to you, genes → athleticism → being on a sports team, genes → disease susceptibility → feeling sick). If some of these mechanistic explanations are correct (likely in combination or for particular sub-groups), it wouldn’t be reasonable to say that genes cause schizophrenia.
If you want to say X causes Y, a randomized controlled trial is a nearly universally accepted way to do so. With smoking, we have controlled trials in animals and some limited smoking-cessation interventions in humans. Also as you say, we have mechanistic evidence about mutations & that adds confidence. Sure, this is an example of imperfect evidence that most people accept and act upon, but it also feels like much stronger evidence than the purely genetic basis of schizophrenia.
The reason I feel strongly about this is that a blanket acceptance of heritability estimates leads to assertions about the innate intellectual ability of races. In this case, the arguments for the violation of heritability estimates are extremely plausible (e.g. in an adoption study, black kid gets adopted into a white family, but is still treated as black by society, and consequently his environment comes with him).
PS I love this blog & massively respect what you do.
“But as we often discuss here, this is backwards - society is fixed and biology is mutable.”
…the examples in the 2014 piece are fine, but come on – “society is fixed while biology is mutable” is just as wrong as the opposite statement.
The boring but truer statement is: both society and biology are mutable to some extent, but it varies to which extent. The variation is partly based on which society we are talking about, partly based on which biological factor/s we are talking about, and partly based on what kind of problem/s we are talking about.
I'll never forget how startled I was when I learned in college that no one with congenital or early blindness has ever been diagnosed as schizophrenic. A quick search just now confirmed that this is still believed to be true. Are we any closer to understanding why? Has investigation of this oddity yielded any progress toward understanding the cause(s) of schizophrenia?
> We know many apparent risk factors for schizophrenia: [...] social defeat
"Social defeat" struck me as the odd one out in that list. I clicked on the link, which was just to the wikipedia article on the concept, and it in turn had this in its list of sources:
That merely proposed the hypothesis that social defeat can increase the risk, and presented data in support. It didn't claim that we "know" it's a risk factor. Instead, it ends with suggestions for how to test the hypothesis in the face of the ethical problem of subjecting humans to social defeat and the inability to test for schizophrenia in other species of animals it is permitted to experiment on.
You seem to be completely right, but you make it a hard read. Why not start with "People in schizophrenia research are adamantly opposed to saying it's a largely genetic disease, even though the evidence says it is. Why is that?" You probably do want to put the long-winded analysis at the bottom as a massive footnote to the first sentence, but the interesting part is the psycho-politics of why people resist considering schizophrenia a largely genetic condition. And indeed, you barely scratch the surface of that. I'd think that if you did as much careful sniffing of how people talked about schizophrenia as you did dissecting the analysis of the data, you'd get a pretty good idea.
I came across something recently called the "eggshell skull rule" in Law which I feel is relevant to describing human intuitions about causation that other comments have mentioned. The idea behind the rule as described by wikipedia is:
"If a person had a skull as delicate as that of the shell of an egg, and a tortfeasor who was unaware of the condition injured that person's head, causing the skull unexpectedly to break, the defendant would be held liable for all damages resulting from the wrongful contact, even if the tortfeasor did not intend to cause such a severe injury. "
So if the genetic correlates of schizophrenia are ultimately best described as vulnerabilities -- things that do require one or more other significant factors to result in schizophrenia -- then I suspect that many people would intuitively find it much more comfortable to just describe them as genetic vulnerabilities rather than as genetic causes. And that's even if the most straightforward path to preventing schizophrenia were still to prevent or correct the genes somehow.
Second, anecdotally, I narrowly avoided a schizophrenia diagnosis a couple years ago (and did get a schizotypal plus transient psychosis diagnosis) despite having symptoms that, based on most descriptions I've read, do not fit schizophrenia well at all in my opinion: chronic pain, chronic headaches, fatigue, pins-and-needles, dizziness, difficulty thinking, and involuntary muscle contraction, all of which are weirdly exacerbated by common allergens. I specifically did not at the time have visual or auditory hallucinations, delusions, or disorganized thinking (by my estimate still, years later, though the psychiatrists considered my pain to be hallucinatory and my belief that it was associated with common allergens to be delusional) And based on that experience of how easily schizophrenia can be "read into" various symptoms as well as stories I've come across like this one: https://www.washingtonpost.com/wellness/2023/06/01/schizophrenia-autoimmune-lupus-psychiatry/ where a woman spent twenty years with a schizophrenia diagnosis when all her symptoms were caused by lupus and enormously improved for the first time by finally, y'know, treating the lupus, I've come to doubt the usefulness of schizophrenia as a label.
When schizophrenia is in practice reduced to "What might be interpreted as psychosis plus general impairment," as seems to be the case to me, while at the same time being thought of as something basically genetic, then there seems to be enormous potential for it to act as a diagnostic thought-stopping cliche. Maybe at a high level "what might be interpreted as psychosis plus general impaiment" is best correlated with genetics, but if it's actually the case that there are dozens of specific genetic conditions or genetic vulnerabilities each of which manifests apparent psychosis plus general impairment under certain conditions, in response to certain things, or in the absence of certain treatments, then wouldn't it be better to not treat schizophrenia as a monolithic "thing" with a "cause" at all, but rather as characteristic presentation of many things, most of which are ultimately genetically-based?
So I guess I'm on the side of: "The studies are probably right, but in a philosophical sense we shouldn’t describe it as mostly genetic."
I assume the real issue is that “genetic” is ambiguous between things like sickle cell or Tays-Sachs where there is a very specific gene variant whose presence or absence is definitive, and things that involve complex interactions between many genetic variants and the environment. It’s not like the former. And we don’t have good use to make of the latter category yet, so it isn’t one that we have broadly developed useful vocabulary for describing.
If people need a low tech method, not reproducing with people who have a family history of schizophrenia should help. On the other hand, if the percentage of schizophrenia is the same worldwide, it doesn't seem to help much, or is harder than it sounds, or helps the same amount everywhere.
I really do understand why you're so defensive of polygenic screening. I wouldn't judge any individual parents for choosing the option, and obviously in 2024 you are in no meaningful sense affecting the genetic fabric of society by doing so. But all the same, I worry about what it does to humanity in the long term if it becomes normalised, and I think you're wrong not to.
To get it out of the way there's the obvious thing where if it becomes possible, bigots might select against having (non-cripplingly) autistic, or gay, or etc. children in sufficient numbers to endanger the long-term survival of those traits. But that's not even what I'm most worried about. I'm worried that trying to do polygenic screening with our current half-baked understanding of genetics is the equivalent of doing advanced chemistry with a kitchen blender.
That is to say: can we reliably select against schizophrenia without selecting against various non-harmful forms of neurodiversity? Or indeed, random unrelated benign traits which happen to be genetically correlated with schizophrenia but have to do with the shape of your nose or left-handedness or something? On a societal level I worry about this a lot. I find it *most* concerning with people selecting against severe autism than against schizophrenia, because severe autism is more clearly on the same directly-related spectrum as personality types which humanity would be immeasurably poorer for not having. But A) I'm not sure schizophrenia isn't also correlated with positive /benign traits, and B) I don't think I trust society to be responsible with these tools if they get normalised, and *only* use them to screen out "safe" conditions.
My family has a history of schizophrenia, including members of both my immediate and extended family. We also have a history of high academic achievement, which includes advanced degrees in hard sciences and at least one department head at a major US university. I have noticed other families of high academic achievers with schizophrenic children.
Is there any data that links schizophrenia and high academic achievement within a family? I feel like I have seen this mentioned before. If so, is there any possibility that selecting against schizophrenia would select against gifted people as well?
You are ignoring the massive role that psychological injuries play in people later developing symptoms of psychosis. People with 5 Adverse Childhood Experiences are anywhere from 60 to 150 times more likely to have these symptoms, than people with no ACE's. https://pubmed.ncbi.nlm.nih.gov/17586579/ Schizophrenia is hugely influenced by being emotionally neglected, sexually abused, having a raging alcoholic as a parent, and other emotional wounds. It is not fundamentally in the genes.
I suppose one argument for not saying schizophrenia is genetic is to prevent that from becoming a dogma that can't easily be displaced? Like suppose some researcher has a theory about the cause or a treatment for schizophrenia but it's not linked to genes in any clear way and so when she goes to apply for funding she gets rejected because "everybody knows it's genetic".
You probably aren't arguing that schizophrenia should be specifically and always described as a genetic disease you just don't want people to be in denial about the role of genetics but perhaps (and to be clear I haven't read them) they are arguing against a more extreme version of your position.
Is it fair to say that genes do not cause two-leggedness? The heritability is near zero, on the same meaning of heritability. For bonus-points, y'all obviously agree accent is more heritable and therefore more genetic than two-leggedness, right?
Can't have any isolated demands for rigor here, and in certain communities any demand for statistical rigor is isolated, so please everybody get with the program.
I'm a public health professional. I think there is no difference between cause and risk-factor for scientists or professionals, but there is a big difference when creating public messages. Many people interpret "cause" to mean 100% correlation, as in, smoking 100% always causes cancer. When you say "risk-factor" people are better at understanding that there is some amount of probability around the factor. So generally you see a preference for saying risk-factor.
I think this is the core of the failure in communication. The human mind does not seek truth, it seeks satisfaction.
For a person not affected by schizophrenia, it's perfectly normal to think that "Schizophrenia is a disease that crazy* people have, and runs in the family" (*I'm not endorsing that word, but it is in common usage)
If you are affected, or work on it, that's not nearly good enough. A satisfying definition would include the idea that it's a bucket of conditions, not just one thing, and that it can cause a range of symptoms of varying degree.
When you work on a problem, you want to be able to work backwards and forwards with different suppositions to see what works.
Cause is like Proof - it's a place where you can stop thinking about something. If you say "I have evidence of something", that means you may have to keep looking. If you say "I have PROOF", then you can stop looking.
----
I wonder how the various parties would engage with the terms Major Contributing Factor vs Minor Contributing Factor.
It is unfortunate there is no special word like "causes" in English that means "always causes but is not always the cause of." Successfully feeding someone into a wood-chipper always causes death, but death is not always caused by being fed into a wood-chipper. You could say genes cause schizophrenia, but don't wood-chipper cause schizophrenia.
>it’s a million times more than the drug abuse and social defeat issues
A million is a lot, this is meant rhetorically, like "a bazillion"...right? After the "Unintuitive Properties" post, I don't totally trust my surprise at seeing 10^6 mixed in with the other numbers.
Not directly relevant, but thank you for at least mentioning COPD. Everybody thinks tobacco will give you cancer, and it totally can, but it's way more likely to give you the other thing which will also kill you but in a very different (still horrible) way. We stress cancer way more, probably because cancer is an easier and more familiar thing to understand than an umbrella category for two overlapping disease states.
I think that the resistance to genetics as a cause for diseases is because it is very unsatisfying to blame genetics.
If the cause of your disease is environmental and beyond your control, you can always blame society. You are the victim of systemic injustice because society did not protect you well enough.
If the cause is environmental and thought to be within your control, you can always be blamed yourself. You should have known better than to smoke.
If the cause is genetic, that does not make for a very satisfying narrative outside of Nazi circles. Blaming your parents for not using screening or CRISPR is frowned upon, so you can just call it bad luck.
To the degree genetics as a cause is actionable, it would proscribe actions (gene screenings, selective abortions, DNA edits) which would make huge parts of society uncomfortable.
Didn't you write an article about how mainstream media rarely lies and apply a standard of "lying" under which one could claim the national enquirer rarely lies? (In that they didn't lie if they accurately reported what it was their source said to them without comment on the accuracy of that source)
There’s a disease that occurs only in people with a known, single gene (actually one allele). Because we know that gene, we know exactly what’s happening when it kills someone.
And yet it doesn’t kill everyone it could; people with the bum gene are still alive. Though if it did wipe them all out, we’d probably *still* blame the vector, not the victims’ genes.
The disease is mad cow, and the cause is ingesting misfolded prions. We are so impressed by a concrete explanation that most people don’t even know every victim (save one, IIRC) was genetically identical in one spot. If that’s not a genetic disease, how is schizophrenia?
Nobody wants a GWAS disease. It’s too fuzzy and it smacks of either missing something on the causal side or the definition being too broad. Like other unsatisfactorily explained diseases (MS, chronic fatigue, etc.), it’s better to admit ignorance than blame a thousand insignificant genes in the victim affecting an entirely hypothetical no-voices system in his brain.
Genetic screening for dominant genes whose pathological effects appear in adulthood reduces the frequency of those genes. That is an unalloyed good only if we certain than any possible beneficial effect of those genes is outweighed by its pathological effects. But we can't be certain of that.
Imagine a future ultracontagious virus that kills everybody who doesn't have the gene for Huntington's Chorea (WIV is working on that right now -- maybe...). With 100% screening of embryos for the gene for Huntington's Chorea, human beings go extinct.
"health care should be about treating schizophrenia, not preventing it"
I was trying to figure out why anyone would oppose polygenic screening and this sounds like the real reason. A doctor or therapist can't make a career transition to genetic screening tech, so they correctly see "we could cure schizophrenia" as a threat to their employment.
Genetics has nothing to do with schizophrenia, save for growing up among the deviations from “society’ as it is accepted and wanting to be ‘someone else’. Some young actors have been diagnosed schizophrenic after taking on a role and keeping it. You can pretend to be someone else and end up really believing it. To someone who craves attention, they can create as many different “voices” as they want. Not being accepted in society is what keeps them so sad and hides away in the group they created for themselves. When someone is sick or not all there as with drugs, some sayings can stay with the person. I have helped a person see where there voices originally came from. A friend kept telling her she never listens so she wasn’t. They have to be active in their life and at least able to laugh once in a while to be able to even talk to them about it.
Ash Clif_high. He says he has the ‘schizophrenic gene’. But he is not schizophrenic . Basically if you cannot get them to create their life, they will just stay apathetic and reactionary. The medication for psychotics is a tranquilizer,......this keeps them quiet and depressed unless they can form some life while on them. Drugs is definitely not the answer. They need acceptance and confidence and acknowledgement. Cal mag and B vitamins really help.
Once argued with someone who said it was "reductive" to say that "genes are among the causes of risk of skin cancer." And at that point I gave up because at a certain point demands for further nuance just become a reason not to have thoughts.
Is there really a clear definition of "schizophrenia"? This was all very interesting -- no irony -- but without a clear definition of the supposed disease, talking about its "causes" seems somewhat beside the point.
I have a question: if schizophrenia is mostly genetic, then why is there so little evidence of it in ancient texts? (Whereas other mental illnesses are described in Greek and Roman sources etc.) Doesn't that suggest that there are factors new in the environment that cause it (perhaps in interactions with genes or recent gene mutations). As well, all twin studies are compromised by the shared womb and in the case of adoption, by shared adoption trauma, and while genes run in families, so do environmental factors, stress levels, nutritional factors, parenting styles, and intergenerational traumas.
From Iain McGilchrist (The Master and His Emissary, page 404):
"... schizophrenia has in fact increased in tandem with industrialization and modernity. In England, schizophrenia was rare indeed, if it existed at all, before the eighteenth century, but increased dramatically in prevalence with industrialization. Similar trends can be observed in Ireland, Italy, and the United States, and elsewhere ... What is beyond reasonable doubt, however, since it has been established by repeated research over at least half a century, is that schizophrenia increased pari passu with industrialization; that the form in which schizophrenia exists is more severe and has a clearly worse outcome in Western countries; and that, as recent research confirms, prevalence by country increases in proportion to the degree that the country is "developed", which in practice means Westernized. Descriptions of melancholia, or of manic-depressive (now called bipolar), are immediately recognizable in accounts from ancient Egypt, Greece and Rome, yet there are no descriptions of schizophrenia ... The relative risk of developing schizophrenia in an urban rather than a rural setting is nearly double, and the evidence suggests that is it more likely that the urban environment causes psychosis than that high-risk individuals migrate to urban areas."
I can't believe that you would reduce THE Ronald Fisher to a tobacco statistician. Back in the day it might have been a reasonable position to say that the causation hasn't been proven to a high degree of confidence. Just because it turned out to be true, doesn't mean it was wrong to caution against premature regulation!
> But age is certainly more than correlational with flu
Minor but: but your example is wrong. If you changed age and held everything else equal then you wouldn’t get the flu less. It’s because age is correlated with other health properties (like a weaker immune system) and you’re assuming these change too when you change age.
I haven’t read the Aftab post or Torrey paper, but it seems that something is being assumed at the outset of the arguments you’re responding to—and your responses to them—that hasn’t been established yet, as far as I’m aware; namely, that there are “genes that cause schizophrenia” (as stated in Argument 3). Has this actually been established?
I’m reading Chris Palmer’s book, Brain Energy and he has a chapter on genetics in which he claims that since the human genome project, we haven’t been able to find specific genes related to mental disorders, though researchers have tried very hard to do so. My understanding based on my reading of Palmer’s theory is that epigenetics—the turning on or off of genes that are mostly common in all of us—are what produce psychiatric symptoms when epigenetic regulators like mitochondria are defective. Interested in your (or anyone’s) thoughts on this, as well as the whole “mental illnesses are metabolic disorders of the brain” idea promoted by Palmer. After all, the satisfying nexus, or the thing that “all schizophrenia causes have in common” might be, according to this idea, mitochondrial dysregulation—which happens to be treatable by various methods, including (but not limited to) medication.
Thanks, from a schizoaffective with his illness in remission for 5 years!
In my opinion, this all comes back to the irresponsible shirking of physical brain scans which we have the technology and enough information to do. It's frankly bureaucratic stupidity from my seat.
But if you want my wild stab, I think it's environment poisoning. Some persons are more inclined to sensitivity writ large. It's not necessarily a weakness toward anything, but all things. You can really see this in autism more clearly which is highly related to schizophrenia.
So causal things more apt to create disability would include, in no particular order, metal poisoning, mother stress, mother drugs, father age, heavy toxins/molds, toxin history of mother and father, nutrition, lack of natural sunlight, lack of functional digestion, metabolism dysregulation, working your body (mostly mother) in regularwork performance .. and yes, I'll risk it potentially even up to and including getting poked with a needle up to around age 2 or 3 in particular.
"This is possible, but I think it’s false. We know many apparent risk factors for schizophrenia: cannabis use, birth canal asphyxia, social defeat, toxoplasma infection, poor prenatal nutrition. If you try to combine these into one big picture, it doesn’t work."
There's a theory that does: Cell Danger Response Theory by Robert Naviaux. All the above are related to over/under-engagement of the healing process in response to stress/trauma. He even lists toxo as a possible disorder from the CDR malfunctions. In short: When you have a chronic stressor in your life for over 6 months or of sufficient magnitude your body cannot complete the healing process and it gets stuck at one of 3 sequential stages. 1 Cell shutdown and hardening, 2 cell replication , 3 cell differentiation and establishment of connections to other cells.
As the 334th comment I expect this to get buried, but in any case there is a literature you may find helpful regarding the erotetic nature of explanation: We generally aren't explaining things in extreme specificity and/or extreme universality (a God's eye view). Instead, we are evidentially-limited beings with goals, and we choose the scope and granularity to explain something to a situated human for a specific purpose.
That provides a framework to understand the fact that the same question may have more than one answer. It also allows one to make systematic sense of why some causally-relevant factors are placed in the background rather than called "causes," and why the background and foreground can change.
We take the background for granted. Those conditions are "normal" while the thing we call the cause is the intrusion on the normal state, or a violation of the state we prefer (in other words "normal" can refer to frequency of occurrence or normative preference). The distinction between cause and background condition is often NOT in the modal logic of implication. It is rather based on our own interests and preferences about where to intervene to change things, or our preference for assigning responsibility/blame as agents.
It had never occured to me that biology was sometimes easier to fix than environment...
Now it see it everywhere.
For example obesity is both a stigma and a medical condition surrounded by truckload of environmental (if not moral) advice and one a someone invents a drug that just fixes obesity and it is not a problem anymore
I don’t think the cause | risk factor distinction is explained well using the language of necessary and sufficient conditions, but here’s an alternative:
Causes are events which lead to the dice rolling (for a given outcome to happen). A given cause doesn’t have to be the *only* way for the event to happen.
Risk factors influence the odds of the roll. Alternatively, risk factors can also refer to any change in the success criteria. Smoking causes lung cancer, but being a heavy smoker is also a risk factor for lung cancer because repeated rolling of the dice increases the probability that one will hit.
My gut says this definition is pretty close, but I’m concerned that (e.g.) smoking would, given a deep enough understanding of the underlying biomechanics, only be a risk factor. Maybe the more expert commenters can evaluate better.
I stumbled onto this Substack just today. I perceive many readers/commentators are way smarter than me. I only want to say, I had a terrible schizophrenic sister as well as a brother with bipolar disorder. None of this information was around forty years ago. My grandfather committed suicide (in his younger life he had electroshock therapy for severe depression). My dad wasn’t right but it was the 50’s and 60’s. I have considered these disorders hereditary based on my personal experience (and some reading). As a mom, I made sure my kids (now fine adults) knew what to look for if they felt/believed such and such. I wanted them to know our family history and be aware because it seems to me once the disease is in full swing, the patient becomes the proverbial ‘I’m not mentally ill’ person who is very hard to help. It’s a terrible burden I would advise for genetic testing when available.
Is Schizophrenia as ontologically defined as the Flu is? There have been plenty of cases of it being misused as a political diagnosis, in the Soviet Union for instance, and even if there's a known ontology it would still be equivocation if it's never clinically tested for and the diagnosis is still only based on symptoms.
I’m surprised that despite a dozen arguments on this list, not one of them is the arguably-correct one: that heritability might be a bad match for human intuitions of what “mostly caused by” means. (e.g. the low concordance probability of schizophrenia in twins, which is counterintuitive. We can respond to this either by saying concordance probabilities are a bad operationalization of “caused by genes,” or by saying the biggest cause of schizophrenia is “shitty luck” or “nonlinear responses to inputs that don’t add up to 1.” Of course, this buys you a whole heck of a lot of other paradoxes instead.)
Polygenic screening doesn't prevent schizophrenia in a meaningful way, though. It just prevents those embryos with high risk from developing into people (modulo your beliefs about when personhood begins).
...Who the hell has been saying epilepsy isn't genetic? It's pretty damn obviously genetic. Barring the ones from people getting whacked on the head really hard.
> By comparison, you can very clearly halve your children’s risk of schizophrenia through polygenic selection, which costs only a few hundred dollars if you’re already doing IVF.
Last time I looked into it, polygenic risk scores for schizophrenia were only able explain ~2% of variance of schizophrenia. It’s really hard to recruit enough schizophrenics for variants in a GWAS to reach genome wide significance. And even with GWAS, you can’t measure things like copy number variants or rare variants. I think it’s at least another few decades until we can screen out schizophrenic embryos.
Is Argument #9 even true? Did anyone NOT call epilepsy genetic?
I'm curious how the part about "schizophrenia is so famously hard to define that we’re not even sure it exists" squares with "studies from 1970s China find exactly the same heritability as Western studies". If it's that hard to define, isn't it likely they're using different definitions in different studies, making direct comparisons pretty hard?
Regarding argument number 3 specifically, neuroscientist Kevin Mitchell wrote a great book called Innate in which he explains how many mental disorders seem to be caused by non-specific genetic mutations that increase developmental noise. His book is full of insights.
Schizophrenia serves an important social function by generating people who assert and act according to blatantly false ideas for non-predatory reasons:
* Based function: This allows sociopaths to prototype new con games while scapegoating non allied persons who act similarly enough to be confused for them
* Woke function (theoretical): This allows neurotypicals to discern deeply entrenched con games and the people who play them
Turkheimer … didn't he write something similar about the intelligence not being genetic either?
I see that much of the discussion about causation in this context is still based on verbal arguments and common sense. On the other hand I keep hearing that DAGs and the other machinery developed by Pearl and others is widely applied to epidemiology. Did anyone try to throw a bunch of potential risk factors, genetic or not, into a causal discovery framework to see what happens?
The other environmental risk factors for schizophrenia are equally hard to change. Poverty? Okay, don’t be poor, thanks for the important life advice. Social defeat? “Doctor, are you saying I have to never let anyone defeat me?” “Yes, it’s my official medical recommendation that you become invincible.”
Keep writing paragraphs like that.
Re 6B: A good analogy might be peanut allergies. Is it meaningful to argue the anaphylaxis was caused by peanuts instead of the allergy?
"This is both totally antithetical to the spirit of real clinical medicine as it’s practiced, and ethically odious to anyone who has witnessed the side effects of antipsychotics first-hand. "
You've perhaps already written it, but an essay on side effects of psychiatric drugs (or even of drugs in general) would be interesting and useful, and a good counterweight to "just take some pills, it's no big deal" thinking.
Contrary opinion:
http://unwashedgenes.blogspot.com/2021/01/the-genetics-of-schizophrenia.html?m=1
I more or less agree with all of your arguments presented here, but I disagree with the conclusion. I don't think the solution is to assert that "genetics cause schizophrenia", but rather as a society we should be a whole lot more careful about using the word "causes" in scientific research (including smoking/lung cancer).
I understand your frustration with applying a different bar to smoking and psychological diseases, but one has to start *somewhere* to enact societal change in the level of confidence that scientists imply when they are speaking with the lay public. If someone is a psychology researcher they probably will start with psychology stuff. A cancer researcher may simultaneously be pushing back against the "smoking causes lung cancer" narrative in a way that you don't see because your day job doesn't involve following prominent cancer researchers who are labelled (possibly inappropriately) as Big Tobacco puppets.
There are some things that scientists are quite confident about (nearly all of mathematics, most of physics, a lot of chemistry, some molecular biology, etc.), and there are some things that we have really good correlational data on but don't fully understand (nearly all of psychology, etc.). For the latter set of things, it is very likely that one day, as you suggest in this article, we'll find that "computational center" and it will all make causal sense. At that time, we'll find that genetics were not causal, just risk factors like in your kidney example, and we'll be able to create a very clear causal chain for each specific instance of schizophrenia. If we say "genetics cause schizophrenia" today, then when that future day comes we'll have to say, "sorry, we were dumbing it down for you, but now we know for sure what causes it" and no one will believe you (fool me once, shame on you, fool me twice, shame on me).
I personally find that way too many scientists are far too eager to assign causation long before they should. IMO, we shouldn't say we know causation until we can predict future outcomes with near 100% certainty (like we can with all of mathematics, most of physics, etc.) At the moment, our predictive capacity for both lung cancer and schizophrenia is appallingly low and doesn't get anywhere near reaching the bar where I think using the word "causal" is appropriate.
First of all, some schizophrenia is probably caused by antibodies - an immune reaction. (https://www.psych.ox.ac.uk/news/new-study-finds-antibodies-that-may-be-the-cause-of-schizophrenia-in-some-patients)
Leaving that aside, the whole concept of disease and causality is complicated, and the very idea of something "causing" a disease is context-dependent.
For example, every child born in the U.S. gets tested for a condition called PKU. If you look at a diet Coke can, you will see a warning that it “contains phenylalanine.” PKU is a genetic disease, and the people who have it have a mutation in the gene for the enzyme that allows them to digest phenylalanine.
But take a step back. If instead being a tiny percent of the population, if the mutation were prevalent – for example, like the mutation that allows us to digest milk – then it wouldn’t even be considered a disease. It would be considered normal, and phenylalanine would simply be considered to be a toxin.
Similarly, if instead of 1% of the population, 99% of the population had the CCR5 mutation that made you immune to HIV, then HIV would not be an infectious disease but a genetic disease instead.
Diseases are intellectual constructs. They’re like boundaries of countries. Sometimes they’re based on the contours of natural geography, much like borders that wind along rivers or stop at mountains. Other times, they’re arbitrary, like vertical lines dividing some of the American Midwestern states. What’s the dividing line between pneumonia vs. bronchitis? What’s the difference between a heart attack and angina? Where we divide the diseases into categories and how many divisions we draw, as physicians, are not God-given. They’re man-made.
Schizophrenia is particularly germane example because it is a mix of what are probably very different processes. Catatonic schizophrenia is very different from classic forms. Doctors tend to lump very different diseases into one category if (as was true for a long time for schizophrenia) there isn't much you can do about it. Generally, diseases are categorized by what therapy they respond to, because pragmatically, if you're going to treat the patients the same, you lump them together. And if not, you separate them. (In some instances, especially in psychiatry, diagnosis can be driven by reimbursement considerations too.)
Here is an example of a disease that some pharma companies invented: high blood pressure. Until the 1970s, there was a great deal of debate over whether hypertension was a disease that needed to be treated, especially in the elderly. Some physicians thought that as we aged, and as our blood vessels became stiffer, hypertension was a natural effort by the body to maintain blood flow to all the important parts of the body. It wasn’t until a strenuous effort by Merck and other companies to change people’s minds that it became accepted that hypertension had to be treated.
I should note that it is not clear still how, when, and with what we should treat hypertension. As more data comes out, it is becoming clear that some drugs work better than others (and some are actually harmful). Not too long ago, the guidelines were changed to make treatment of elderly patients less aggressive. The complexity of the consensus guidelines suggests that there are still many unanswered questions. The very fact that there are guidelines means the answer is no obvious - only when there is uncertainty is a guideline needed.
I have a post expanding on this at: https://clinicaltrialist.com/2017/11/13/whats-in-a-name/
Not sure it matters much to the bottom line, but I’d be cautious about using the terminology associated with smoking risks as the sort of default baseline. Seems to me like this is an atypical situation, where there were enormous policy influences on the language and terminology associated with smoking; there was a huge (justified and good) campaign to discourage smoking, and the public health community was heavily involved in the messaging in ways that shouldn’t necessarily be assumed to be normative for every other terminological situation. I’m not saying that the scientific community wouldn’t have reached the same “smoking causes cancer” formulation even in the absence of the unusual public health campaign surrounding the issue; just that it would probably be safer to use a different and less policy-laden example as a semantic baseline for this sort of thing.
I'm far more hereditarian about psychological and behavioral traits than most liberals, so it's funny for me to take the non-hereditarian side of this one, but I'll at least offer some thoughts.
The reason we readily say that smoke has a causal, not merely a correlative, relation to cancer is not because of the statistics (although they are important). It's because we have a mechanistic explanation of the causal pathway. Tobacco has over 70 carcinogens that directly cause DNA damage, in addition to oxidative stress and chronic inflammation.
Now, of course, whatever genes do is also ultimately mechanistic, but here's where the similarity ends. We have no mechanistic account of how genes might cause schizophrenia. Of course, one might protest that that's just a statement about our knowledge. But it's worse than that. It just doesn't seem likely that whatever is going on neurologically has a causal pathway anything as straightforward as the effects of carcinogens. The interactions with the environment seem like they're far more intricately involved. It's as if smoking caused cancer, but only at certain altitudes, within certain temperature zones, when you ate certain foods, etc., etc. If the latter were the case, we'd be much more hesitant to just say "smoking causes cancer."
Does "social defeat" mean, like... getting bullied at school? (Or at work, I guess?)
Getting bullied is a (small) risk factor for schizophrenia?
The key issue at the heart of the debate: whether as describing it as mostly genetic, we can then use genetic scores to motivate clinical action. See Kendler et al., 2023: Relationship of Family Genetic Risk Score With Diagnostic Trajectory in a Swedish National Sample of Incident Cases of Major Depression, Bipolar Disorder, Other Nonaffective Psychosis, and Schizophrenia
Scott, obesity is also 80% genetic. Can I hear you say "fine, obesity is also genetic"? Do you bite this bullet?
Of course, it does not match what we colloquially mean by "genetic" at all, since obesity is 100% environmentally malleable. Can we *please* adopt a more sophisticated frame than "is genetic" vs "is not genetic"?
You are modelling the world as if everything is a sum of small contributing factors. But it's not; sometimes there are AND gates. Sometimes you need to have the genes AND to eat the right food. It is not possible, conceptually, to assign a "percent heritability" in such a situation.
1. Out of curiosity. Kidney disease heritability. "We report heritability estimates of the CKD-defining traits eGFR (44%), UAE (20%), and UACR (19%), for the kidney biomarkers serum urea (31%), Scr (37%), and uric acid (48%), and for the serum electrolytes potassium (28%), calcium (27%), and sodium (22%) " They checked and basically every kind of family member with the disease elevates your risk, even "genetically unrelated" ones like spouses (assortative mating or maybe shared environment). https://www.sciencedirect.com/science/article/pii/S0272638620311628 It doesn't seem to be impressively heritable, but there's obviously some genetic causes of this, whatever they might be exactly.
2. There are some single mutations that cause schizophrenia alone, or schizophrenia like traits at least. https://www.pnas.org/doi/abs/10.1073/pnas.2112560118 "A total of 48.2% of individuals with severe, extremely treatment-resistant schizophrenia carried at least one rare, damaging missense or loss-of-function variant in intolerant genes compared to 29.8% of typical schizophrenia individuals".
3. Family studies yielding estimate of heritability, shared environment etc. tell you where to look for causes. They give you the overall importance of some group of causes, but doesn't tell you what they are except in a broad sense. In other words, they tell you where to dig for more. They also give you a decent idea about what kinds of interventions are going to be effective. When heritability is very high, it means that genetic selection will be effective. It scales linearly with the square root of the variance. If "family environment" doesn't show up or is very minor, then trying to modify risk factors in that area will not be effective unless you somehow introduce a new kind of thing that doesn't already vary between families. https://www.emilkirkegaard.com/p/high-heritability-does-mean-that
4. Assortative mating and measurement error are usually quite large downwards biases in family studies. This is particularly the case for self-report symptoms of mental health vs. more objective data (e.g. clinician ratings). I wrote a summary of such research. https://www.emilkirkegaard.com/p/heritabilities-are-usually-underestimated
You say it yourself, schizophrenia is quite hard to neatly define. So much so that perhaps it doesn't exist / isn't one unified thing.
I think most of these writers who are arguing that schizophrenia isn't genetic, and some of those arguing that schizophrenia shouldn't be prevented (only treated), they mean to instead argue "schizophrenia isn't real".
They don't make that argument, and instead get wrapped up in denying the practical facets you lay out, because they are the sorts of people who are not comfortable rejecting the existence of schizophrenia, for social-norm reasons. The people who feel "woke" social norms to treat-not-prevent are also the ones who have a cognitive bias towards saying it's intervention-able (not genetic).
But strip away their social and cognitive biases, and they probably would wish to unpack "what is schizophrenia actually? It's not well defined enough, there's more to learn. "
Regarding 6B, genes could be a proxy for a more satisfying cause of schizophrenia. It is not hard to think of ways this could work. As you say, one is that genes could make people more vulnerable to some virus that attacks the brain and causes schizophrenia. There are many other possibilities of this kind. What they all have in common is that you can think of them as genes that change the person’s environment, causing it to be more schizophrenogenic. (For instance the person in your example with the bad immune system lived in an environment that contained more dangerous viruses.) Here’s a list:
-Genes that cause someone to be physically unattractive, or to have other characteristics that make it less likely they will receive social approval: They’ve got poor gross motor coordination and are bad at playground games, and so they get less acceptance and approval from peers. They’re slow to learn to read, and have an especially hard time sitting still and shutting up in school, and so get less approval from their teachers. Kids with these genes live in an environment with fewer smiles per year.
-Genes that cause some disability that interferes with the person making sense of their environment, and is subtle enough not to get diagnosed in most people: prosopagnosia, difficulties putting your thoughts into words, some spacial orientation problem that makes you get lost easily. People with these genes live in a world that’s more confusing and harder to navigate.
-Genes for some as-yet-unrecognized disorder that’s like PKU (phenylketonuria), but instead of causing severe developmental delay causes, by age 20 or so, whatever sort of brain damage underlies schizophrenia. People with these genes live in a world where ordinary foods cause brain damage.
-Genes that cause people to have unusually strong reactions to stressors. For these kids, seeing a squirrel get run over is as upsetting as seeing a hanging would be for others. People with these genes live in a world with an unusually large number of truly traumatic events.
The idea that social problems are harder to fix than biology is a prime example of Americentrism. Other countries have made far more progress than the US in reducing drug abuse, poverty or social conflict (i.e. instances likely to lead to social defeat). I think this is because the US is unusually individualistic and conservative (and Conservative) for a developed country.
I also think it's rather flippant to suggest people undergo IVF to address genetic problems. I have two daughters by IVF. IVF is very rough on the mother, far more so than the side effects of schizophrenia drugs (which I've also had the misfortune of experiencing). This may improve with better IVF procedures, but we can also discover better drugs.
I know this got addressed in a paragraph above, but didn't Scott just get finished with demonstrating (with stats and maths) that, because of the low incidence of schizophrenia, the causes in any specific individual with schizophrenia will mostly be environmental.
Running the numbers with Scott's methodology by way of example, this means that you could screen 500 people for schizophrenia risk factors using a microarray or something, set the cutoff at a 0.95 genetic risk factor, wind up with 25 'candidate' schizophrenics (of which 4 will actually develop schizophrenia) and still miss one actual schizophrenic.
Which, in this example, means that your expensive microarray test has an 84% false positive rate, and a 20% false negative rate in practice.
And if you bump your threshold to 0.99, then all you get is 7 candidate schizophrenics, of which (in this example) none happened to actually develop the condition.
Doesn't this put screening for purely genetic factors for schizophrenia firmly in the 'scientifically true but useless' category?
A distinction I've been thinking about which I've been unsure about but which seems reasonable/important to me:
Schizophrenia is related to psychosis-propensity, arguably caused by psychosis-propensity but also arguably distinct from it in that schizophrenics tend to get convinced they are being persecuted and then retreat into themselves.
It seems to me that one could hypothesize that the mostly-genetic, biological brain-brokenness causes the psychosis propensity, but then the other aspects of schizophrenia are caused by a psychosocial dynamic where e.g. outsiders react badly to the hallucinations and therefore confirm the schizophrenic's suspicions that they are in on the conspiracy.
If true, I don't know how this affects the conclusions in your article. I think it depends on how damaging the psychosis is on its own, vs how damaging the downstream social dynamic is. If most of the damage is with the psychosis-proneness then this is just kind of nitpicky. But if most of the damage is with the social dynamic, then e.g. it raises the question of whether one could cure most of schizophrenia by just averting that social dynamic, and therefore it starts seeming reasonable to think of schizophrenia as environmentally Caused.
While one needs to understand that there are a lot of philosophical nitpicks and therefore it's hard to describe anything as true, I also think it's worth clearly mapping out the philosophical nitpicks. For example, if they don't get clearly mapped out then people might not realize they are there (try telling people schizophrenia is 80% heritable and then ask them about the twin concordance rate), and even if they do realize they are there, mapping them out is the first step to generating theories about how to most accurately take them into account.
I think it also deescalates conflicts. If you've found an important-seeming nuance that doesn't get described very clearly, you might want to generate clearer descriptions. This could lead to some cooperate work if people assist you, but it could also lead to ideological conflict if people dismiss it as philosophical nitpicks and you then start writing big books about it
Would "genetics cause height" work better here?
Re your thoughts on viruses: in my experience teaching students, when talk about causality, they talk about levers they think we can pull on. They’re not interested in causal inference for levers we cannot pull. Many people don’t know about polygenic embryo selection as a lever you can pull or have convinced themselves it’s immoral to pull on it. It’s not a consistent definition of causality, but it’s a sensible position to look for levers. You just happen to have a niche position on which levers are easy to pull, the rest of the disagreements come from that.
Of course a widespread virus being necessary would change our idea of which lever to pull first, ie people are trying vaccines for Epstein Barr now rather than do embryo selection for genetic risk of multiple sclerosis. Because it’s much easier and proven and (for some) not as ethically dubious
Great blog post!
'In a society where everyone inhaled radioactive dust all day, smoking would be an utterly insignificant cause of lung cancer compared to all the radioactive dust inhalation'
Slight nitpick - cigarette smoke IS 'radioactive dust', see e.g. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2672370/, where they estimate a smoker may be exposed to an additional 10 Sv, or 1000 rem, in 25 years.
This discussion feels a bit confused. You and your opponents certainly would both agree on a sufficently precisce statement of the claim in statistical language. So when you say you're arguing about whether "schizophrenia is largely genetic" you aren't arguing about whether the facts comport with some agreed on meaning of that phrase but whether that phrase has a meaning which is made true by those facts. And yet, you aren't actually just making a boring corpus data/polling based claim that most competant speakers do give the phrase such a meaning.
I feel like what's going on is the same thing that goes wrong in debates about whether trans-women are women. The people in the debate think they are arguing over a fact about the external world like whether smoking causes cancer but in fact, as you pointed out in a great piece, they are really making normative arguments about what would be the more desierable way to use language (should we assign a meaning to woman/schizophrenia is genetic that includes...?).
In other words, what you described as why you care is actually the argument: using the word the way your opponents prefer causes these harms because it misleads people about what responses will be effective. I agree and find that relatively convincing but that's the actual debate and it would be better to phrase the other stuff as: responses to potential arguments that the usage you favor is also misleading/confusing.
It seems like there's a lot of muddled discussion over "cause." Causal inference is an entire field with a pretty simple and well-hashed-out definition of a "causal effect." From Miguel Hernan's "Causal Inference: What If?":
>We compare (usually only mentally) the outcome when an action A is taken versus the outcome when the action A is withheld. If the two outcomes differ, we say that the action A has a causal effect, causative or preventive, on the outcome. Otherwise, we say that the action A has no causal effect on the outcome
Concretely: Alice smokes cigarettes, and gets lung cancer. If she would not have gotten lung cancer in the counterfactual world where all else was the same, then cigarettes caused Alice's lung cancer.
Of course, barring a time machine or an ersatz replacement (e.g. a cross-over study with a washout for a drug, or a twin study), we usually can't determine individual causal effects. Sometimes individual causal effects are not even that interesting. Maybe Bob developed lung cancer after being hit by a stray gamma ray.
However, we can estimate average causal effects at the population level with a variety of tools, ranging from excellent (e.g. an RCT) to sometimes-questionable (e.g. a haphazardly-specified regression model applied to survey data of questionable accuracy).
I found Miguel Hernan's article "Does Water Kill?" (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5207342/) to be very useful tool for clearing up how I think about causes, and it certainly applies here.
> Actually it’s more complicated than this: if you look at the simulation argument from last time, it’s hopefully obvious that every case is caused both by genetic and by environmental factors, or that it’s not really meaningful to try to disentangle them. I don’t know whether real schizophrenia is like this.
Real schizophrenia is definitely not like the simulation in a relevant way which I will detail below:
Your model involved assigning two types of risk levels to people, summing them, and then diagnosing scizophrenia if the summed risk exceeded a threshold. The threshold was higher than the maximum value of any individual risk component, and therefore it wasn't possible for any component to be solely responsible for a schizophrenia diagnosis. (But, by design, it was easy for one component to contribute five times as much "problem" as the other component.)
The problem is that your risk levels were ordinal numbers, not cardinal numbers. This could only be valid as a model of schizophrenia if schizophrenia were diagnosed solely by comparison to everybody else. Objective criteria, such as hallucinations, invalidate the assumption.
(It's also weird that your model involves adding an ordinal genetic risk to an incommensurable ordinal environmental risk. We try to avoid adding ordinal numbers. By assuming that genetic and environmental risk are normally distributed, you can convert centiles to standard deviations. But I think you'd end up unable to say whether a standard deviation of environmental risk was more or fewer units of risk than a standard deviation of genetic risk, which is kind of the opposite of the point of the simulation.)
In reality, we are surely interested in a model of schizophrenia where you get small numeric contributions from many causes and schizophrenia happens when a numeric threshold is exceeded, not where you get large comparable-but-not-theoretically-measurable contributions and schizophrenia happens when a comparative threshold is exceeded (that is to say, when a certain number of people are doing worse than you).
When I went to a class/group mental health thing for ADHD a few years ago, one of the things I learned was that ADHD is considered "epigenetic": you need the genes, but the genes have to be activated by environmental factors (lead exposure was mentioned) before you really "get" ADHD. Would calling schizophrenia "epigenetic" satisfy both sides of the argument? Also, is this something people still think is true about ADHD?
> Would we really? If schizophrenia was caused by an omnipresent virus, but the only people who got it, got it entirely because of genetic differences from the rest of the population, why would we call the virus “the cause” rather than the genes? Is it just because we’re used to calling pathogens the causes of things from our long history with infectious disease? It might be equally reasonable in this case to think of schizophrenia as a genetic immunodeficiency. I don’t know exactly how to think about things like this.
I mean, it kind of sounds like you do know how to think about things like this, but you don't want to do it.
Suppose there were oxygen lying around everywhere (as in fact there is!), and some people had genes that caused them, when exposed to oxygen, to harness it to cells in their circulatory system and generate energy in specific areas through controlled chemical reactions, while other people had genes that caused them, when exposed to oxygen, to catch fire and burn to death.
Would that be a genetic problem? Could we fairly attribute it to the genes?
Of course; everything you ever do is caused by your genes, because there is no theoretical alternative. The range of possible reactions of an organism to different environments is one more thing that is fully specified by its genes.
>schizophrenia is so famously hard to define that we’re not even sure it exists
This was news to me! You kind of just dropped a bombshell there and kept walking.
> even though smoking causes 80% of lung cancer, if Alice gets lung cancer and Bob smokes exactly the same as Alice, Bob will have a much lower than 80% chance of getting lung cancer. The solution isn’t to stop describing smoking as causing lung cancer, it’s to get better intuitions.
This example makes me queasy; the two 80%s do not refer to related numbers, and neither of them is related to the problem you're discussing in the post. I want to think about the link between smoking and lung cancer in terms of risk increase, something along the lines of "in these two otherwise comparable groups of nonsmokers and habitual smokers, the smokers get five times as much lung cancer". It's never relevant how much of lung cancer is caused by smoking; you'd need to normalize that against the number of people who smoke. This is just the confusion between P(smoker | lung cancer), which in the example is at least 80% (though actually more†), and P(lung cancer | smoker), which is Bob's chance of getting lung cancer from being a smoker.
But while confusing the two directions of a conditional probability is a common problem that occurs when people try to think about probability, it is not the same as what is otherwise the focus of this post, which is the meaning of claims like "80% of the variance in random variable X is explained by random variable Y, a component of X".
This makes the thesis "this is equally true of everything else" a bit weird. The smoking / cancer example provided in support isn't analogous to the schizophrenia / heritability claim and doesn't transfer. It could be an example illustrating that we need to get better intuitions about everything, and, sure, I guess that would be nice...
† Why more? By assumption, 80% of lung cancer patients developed their cancer by smoking, and they are all, by definition, smokers. This gives us a lower bound of 80%. But there will be some smokers who developed lung cancer even though their smoking didn't contribute to it, the same way nonsmokers develop lung cancer without any contribution from smoking, and those smokers-with-lung-cancer-unrelated-to-smoking will raise the rate above 80%.
Is age a risk factor for the flu? It's a risk factor for *dying from* the flu, but is it one for getting it?
None of those arguments make up the true rejection. Instead, like you say, if genetics could be admitted as the cause, then it would likely be possible to describe any workable solution as EUGENICS, which is obviously far too horrible to contemplate.
> Fine. Epilepsy is also genetic.
I don’t see any particular issues with this statement. Many forms of epilepsy are highly heritable. I’m currently trying to solve my husband’s and sister-in-law’s epilepsy and there’s a ton of data that indicate that at least propensity for seizures is highly genetically determined.
"This is possible, but I think it’s false. We know many apparent risk factors for schizophrenia: cannabis use, birth canal asphyxia, social defeat, toxoplasma infection, poor prenatal nutrition. If you try to combine these into one big picture, it doesn’t work."
What the - cocial defeat causes schizophrenia! How? Why? Is this another case of X is not about X, it's about signalling?
> The only thing that makes sense to me (and I’m definitely not an expert, and E. Fuller Torrey is, so take this with a grain of salt) is thinking of schizophrenia as cumulative damage to some abstract computational organ.
My pet theory is that schizophrenia arises more specifically from a deficiency in long-range cortex-to-cortex communication. I wrote about it (poorly) at https://www.lesswrong.com/posts/H2epKysvFgPcTwC2f/schizophrenia-as-a-deficiency-in-long-range-cortex-to-cortex and then later https://www.lesswrong.com/posts/tgaD4YnpGBhGGbAy5/model-of-psychosis-take-2
If that's on the right track, then lots and lots of genes would presumably be relevant, including glutamate receptors, myelination, dendrite structure, synapse structure, neuron resting potential, and on and on.
well-argued as usual. You address only briefly our weird reluctance to acknowledge that schizophrenia in fact does have a large genetic component. This reflects something about current society... in past eras, genetic contributions to behavior were taken for granted and often exaggerated. What's the deal here?
Much of what you say about the comparison with smoking causes cancer seems obviously right. But I guess the meat (or perhaps, heat) of this argument is about what happens downstream of the statement, "schizophrenia is caused by X": where money for researching therapies ought to go, or what advice to (potential) sufferers should be.
The statement "schizophrenia is mostly genetic" is, on the face of it, a suggestion to direct research dollars to genetics research; for sufferers it is perhaps perceived by Fuller as a counsel of despair. (I wonder if that is a motivator for him making the argument?)
Two things in your post leapt out at me as worthy of comment:
(1) "Kidney disease is just a catch-all term for “anything causing your kidney to work less well than it should”. If you examined its heritability, it would probably be substantially genetic."
Assuming this is true, this seems like an argument against saying "schizophrenia is mainly genetic," because that's not something we commonly say about kidney disease (is it?).
(2) "schizophrenia is caused by some extremely common virus...everybody has all the time...immune system controls the virus...genes for bad immune systems...virus causes schizophrenia."
I just wanted to raise the obvious analogy of HPV here. Once it was discovered that HPV was a key factor in cervical cancer, research moved to HPV prevention, culminating with the HPV vaccine, which as far as I know is now doing an admirable job of reducing infections and the downstream cancers. This seems to me to illustrate what I was saying above: the argument about what we should *say* causes condition X is really an argument about how we should go about researching and treating condition X.
Back on a more general level, the issue you point up of different common medical terms for diseases/conditions/syndromes being on completely different levels of the symptom-condition-cause spectrum just makes any generalisation hard to do. Particularly for something like schizophrenia, where it's still unknown whether that word refers to a causally disparate collection of different symptoms; a single, fairly well-defined syndrome; or a monocausal disease. Any conclusion you might draw about how to talk about particular kinds of diseases may or may not apply to schizophrenia...
I am sad to see this bunch of non-arguments that neither grasp the philosophical question, the science communication problem or the scientific problem.
The philosophical question is, when multiple prerequisites are necessary, but neither is on their own sufficient, what do we consider the cause of something? This is a persistent problem in medicine. When an old person dies of a rhinovirus infection, what is the cause of death? The rhinovirus or old age? A young person would have surived the virus without any problem whatsoever, so it is reasonable to say old age, but then again, without the virus, the person might still be alive, so it would be equally reasonable to say the virus. The sad fact is that there is no scientific way of assigning what amounts to moral blame. Scott would like to use the numbers™, but heritability is a statistical expression of variance, not quantitative causal contribution or population attributable fraction, it rests mostly on outdated evidence and is misleading regarding the preventative effects environmental interventions may have. Neither of these points is adressed in the post.
The science communication question is whether we should advertise heritability estimates as relevant. The original article claims we should not, and nothing Scott argues changes that very much. Environmental interventions still can have large effects on incidence in diseases with high heritability. Scott ignores this for some reason.
Lastly, the scientific problem involves the high heritability estimates of twin studies, which were not replicated by GWAS. It is yet unclear whether they will be replicated by whole-genome sequencing studies, but the likelihood is low in my opinion (WGS studies claim better sensitivity to rare variants, but this is unlikely a major contributor do disease burden due to selection pressure). Polygenic scores do not explain much more than the GWAS estimates, either. The missing heritability problem is easily resolved when assuming twin studies simply give overestimates due to GIGO. This has been argued for a decade now (doi: http://dx.doi.org/10.2190/HS.43.2.f). So what are we left with here, exactly? An isolated demand to strongly believe in outdated results based on poor methodology. This is motivated reasoning at its finest.
Which brings me to the last point, "if you are already doing IVF" - no, most people are not "already doing IVF". If you need IVF you either missed the on-ramp to having kids or there is something biologically wrong to begin with (some might argue that the former is an expression of the latter, but I digress). IVF is a disagreeable procedure for women to go through. It likely leads to inferior pergnancy outcomes compared to, oh, just having sex. Preimplantation genetic testing is an experimental procedure with little long term data that barely passes ethical muster in obvious cases of monogenetic diseases. Recommending PID to a patient willy-nilly based on an abstract desire to improve their offspring might (at least ought to) get your clinical license yanked.
Polygenic screening was never proven effecntive in a clincal trial. The 50% reduction in risk you were sold is not based on RCT evidence. Selecting against schizophrenia might select for something else, for better or worse. Your mileage may still vary overwhelmingly.
I used to appreciate this blog because of the thoughtful discussion of medical evidence it offered on psychiatric treatments and diseases. Reading this series of articles both suggests that these skills do not translate well to other fields of medicine and casts doubt on the quality of previous discussions.
Shower thoughts:
1. Gene X is highly correlated with Bad Thing A
2. Everyone starts screening for Gene X
3. It turns out Gene X basically makes people more in touch with reality
4. Being in touch with reality causes Stress Type 1 because our society sucks
5. Stress Type 1 causes Bad Thing A
6. Much fewer people now experience Stress Type 1, since they're delusional enough to ignore how much society sucks
7. Bad Thing A has been virtually eradicated
8. Success?
> if we knew for sure that all of the genes involved in schizophrenia were for an autoimmune condition, we might say “schizophrenia is an autoimmune disease”.
My understanding of polygenic scores is that it is a statistical method.
My reluctance for a (voluntary) eugenic solution depending on these scoring is, that we don't understand these genes and if there is a side effect? It is like deleting programming code without knowing if it maybe also had beneficial effects? Maybe in this autoimmune genes are also supergood for surviving covid or preventing allergies. Maybe they are involved in creativity and erradicating slightly crazy genes would lead to a boring dystopia of bureaucrats out of tune with the spirit world?
It's very puzzling. Here we have claims that it's
A) the same prevalence in nearly every society studied
B) higher in African-Americans
What? How could both be simultaneously true?
Possibilities ...
- African-Americans are subject to higher stress due to being discriminated against (environmental effect)?
- Some sort of local parasite that people of European ancestry have evolved resistance to ? (Difficulty: we're talking about America here...)
- Desire to emigrate being influenced by same alleles as schizophrenia?
- Discriminatory diagnosis by psychiatrists ? (E.g. marginal cases more likely to be diagnosed if patient is both (a) black and (b) in America)?
My favourite example is that wearing earrings is very heritable in modern human population. Based on whether person has two X chromosomes or not, one can predict quite well whether they are wearing earrings or not.
>You also can’t point to any individual lung cancer patient and say “smoking caused this person’s lung cancer”.
Actually we can! Tobacco smoke causes characteristic types of mutations in cells, and if we see those mutations in someone's cancer, it's pretty certain that it was caused by smoking.
See https://pubmed.ncbi.nlm.nih.gov/27811275/ "Mutational signatures associated with tobacco smoking in human cancer"
Re: footnote 3 (ubiquitous virus plus genes):
The natural way to describe such a situation would be: the virus causes schizophrenia, and certain people's genes cause them to get schizophrenia.
Does that distinction help for the issue at hand? I think it does help clarify what we mean by causation. As for what we mean by "genetic", not directly, but it does point toward a more precise way of describing the situation: "genetically mediated". Would we be having this argument if the question was about schizophrenia being genetically mediated? I don't think so.
I'll note that the vast majority of smokers don't get lung cancer (LC), so even the perfectly reasonable sounding "smoking causes lung cancer" is suspect. It's true that smoking massively increases the odds of LC (i've seen ORs from 20x-40x), the base rate is so low that even for smokers the lifetime risk of LC is less than 1% (at least by my quick and sloppy estimates from CDC/NIH numbers. My hope is that by being wrong on the internet I can get someone else to do this, and even if I'm off by 50x I'm still right.)
The only use I've gotten out of calling things 'causal' is to indicate that modifying one variable will meaningfully affect another variable. It's about actions one can take that are expected to do the thing you think they should. I'm okay with saying "smoking causes lung cancer" because I believe that modifying smoking habits will clearly and directly decrease LC risk, whatever mechanisms or moderators or whatever are "actually" responsible.
Scott, if you haven’t already seen the reply I got from Steven Hyman to my posts, that will interest you: https://www.psychiatrymargins.com/p/a-note-from-hyman-on-the-genetic
From my perspective, we are in this odd position where we know that a substantial aggregate genetic influence on schizophrenia exists (and that such substantial aggregate influences exist for nearly all complex disorders, even if not to the same extent), but we seem unable to make it tangible. The influence dissolves into genes of tinier & tinier effects, and this seriously limits what we can practically do with all this. It’s obvious that hundreds and thousands of gene influence the risk of developing schizophrenia.
We need to distinguish between several different strands of the debate here. “Are genetic influences in aggregate a cause of schizophrenia?” is one question (and I’d say yes, in the same sense of cause as smoking causes lung cancer) but that is a different question from whether “Is the etiology of schizophrenia 80% genetic?” The latter question is rather ambiguous but that’s how many ppl talk about schizophrenia. My view is that it is 80% genetic only in the strict sense of the heritability statistic for a population, and that the heritability statistic becomes kinda meaningless if you try to turn it into a statement about what component of “etiology” (etiology ≠ phenotype variation) is genetic.
A second thing we need to distinguish is that we mean different things by the term “genetic disorder.” It could mean something very general, like any disorder where a substantial portion of the risk/vulnerability comes from aggregate genetic influences. A second more specific sense is where identification of causal genetic variants provides a coherent way to understand the mechanisms or pathophysiology of a condition. Schizophrenia genetics program was kinda based on the hope that we could move from the former sense to the latter sense, and it has become rather clear by now that this is not likely to happen. The influence of any particular gene is too small & current polygenic scores explain like 6-7% of risk. What people are hoping now is that genetic associations can provide clues to understanding what pathways are involved, like disruption of synapses or immune dysregulation. I don’t mind if we consider schizophrenia as a genetic disorder in the first sense (which means many other complex disorders are also genetic) but we shouldn’t simply assume that because it is genetic, the latter sense is also true.
Agree that preventing SCZ is good, and polygenic screening probably helps more than other known interventions, and most of the arguments you reject ought to be rejected, except #7. Violating assumptions of heritability estimates is not a small point.
First, the limitations of adoption studies are not just the importance of the intrauterine environment. An adoptee's adult behavior is caused by a combination of 1. intrauterine environment, 2. early-life experiences before adoption, 3. post-adoption environment, 4. direct genetic causal effects on behavior, and 5. indirect genetic causes on behavior (e.g. gene → physical appearance → another trait). I would not confidently rule out any of these factors without experimental evidence directly controlling for them, which in the case of gene-by-environment interactions (#5) is essentially impossible.
For twin studies, it feels extremely implausible that the equal environment assumption is even close to true. Glancing at the paper you cited, they seem to find that the ways in which identical twins' environments are more similar only have small relationships with behavior. However, the value of this analysis depends on accurately identifying how identical twin's environments are similar. It seems this paper neglected some plausible ways in which identical twin's environments are more similar than fraternal twins (genes → attractiveness → people being nice to you, genes → athleticism → being on a sports team, genes → disease susceptibility → feeling sick). If some of these mechanistic explanations are correct (likely in combination or for particular sub-groups), it wouldn’t be reasonable to say that genes cause schizophrenia.
If you want to say X causes Y, a randomized controlled trial is a nearly universally accepted way to do so. With smoking, we have controlled trials in animals and some limited smoking-cessation interventions in humans. Also as you say, we have mechanistic evidence about mutations & that adds confidence. Sure, this is an example of imperfect evidence that most people accept and act upon, but it also feels like much stronger evidence than the purely genetic basis of schizophrenia.
The reason I feel strongly about this is that a blanket acceptance of heritability estimates leads to assertions about the innate intellectual ability of races. In this case, the arguments for the violation of heritability estimates are extremely plausible (e.g. in an adoption study, black kid gets adopted into a white family, but is still treated as black by society, and consequently his environment comes with him).
PS I love this blog & massively respect what you do.
“But as we often discuss here, this is backwards - society is fixed and biology is mutable.”
…the examples in the 2014 piece are fine, but come on – “society is fixed while biology is mutable” is just as wrong as the opposite statement.
The boring but truer statement is: both society and biology are mutable to some extent, but it varies to which extent. The variation is partly based on which society we are talking about, partly based on which biological factor/s we are talking about, and partly based on what kind of problem/s we are talking about.
I'll never forget how startled I was when I learned in college that no one with congenital or early blindness has ever been diagnosed as schizophrenic. A quick search just now confirmed that this is still believed to be true. Are we any closer to understanding why? Has investigation of this oddity yielded any progress toward understanding the cause(s) of schizophrenia?
> We know many apparent risk factors for schizophrenia: [...] social defeat
"Social defeat" struck me as the odd one out in that list. I clicked on the link, which was just to the wikipedia article on the concept, and it in turn had this in its list of sources:
https://www.cambridge.org/core/journals/the-british-journal-of-psychiatry/article/social-defeat-risk-factor-for-schizophrenia/ABB536863ACE99B8B4D7D1CCE18B49C5
That merely proposed the hypothesis that social defeat can increase the risk, and presented data in support. It didn't claim that we "know" it's a risk factor. Instead, it ends with suggestions for how to test the hypothesis in the face of the ethical problem of subjecting humans to social defeat and the inability to test for schizophrenia in other species of animals it is permitted to experiment on.
You seem to be completely right, but you make it a hard read. Why not start with "People in schizophrenia research are adamantly opposed to saying it's a largely genetic disease, even though the evidence says it is. Why is that?" You probably do want to put the long-winded analysis at the bottom as a massive footnote to the first sentence, but the interesting part is the psycho-politics of why people resist considering schizophrenia a largely genetic condition. And indeed, you barely scratch the surface of that. I'd think that if you did as much careful sniffing of how people talked about schizophrenia as you did dissecting the analysis of the data, you'd get a pretty good idea.
I came across something recently called the "eggshell skull rule" in Law which I feel is relevant to describing human intuitions about causation that other comments have mentioned. The idea behind the rule as described by wikipedia is:
"If a person had a skull as delicate as that of the shell of an egg, and a tortfeasor who was unaware of the condition injured that person's head, causing the skull unexpectedly to break, the defendant would be held liable for all damages resulting from the wrongful contact, even if the tortfeasor did not intend to cause such a severe injury. "
So if the genetic correlates of schizophrenia are ultimately best described as vulnerabilities -- things that do require one or more other significant factors to result in schizophrenia -- then I suspect that many people would intuitively find it much more comfortable to just describe them as genetic vulnerabilities rather than as genetic causes. And that's even if the most straightforward path to preventing schizophrenia were still to prevent or correct the genes somehow.
Second, anecdotally, I narrowly avoided a schizophrenia diagnosis a couple years ago (and did get a schizotypal plus transient psychosis diagnosis) despite having symptoms that, based on most descriptions I've read, do not fit schizophrenia well at all in my opinion: chronic pain, chronic headaches, fatigue, pins-and-needles, dizziness, difficulty thinking, and involuntary muscle contraction, all of which are weirdly exacerbated by common allergens. I specifically did not at the time have visual or auditory hallucinations, delusions, or disorganized thinking (by my estimate still, years later, though the psychiatrists considered my pain to be hallucinatory and my belief that it was associated with common allergens to be delusional) And based on that experience of how easily schizophrenia can be "read into" various symptoms as well as stories I've come across like this one: https://www.washingtonpost.com/wellness/2023/06/01/schizophrenia-autoimmune-lupus-psychiatry/ where a woman spent twenty years with a schizophrenia diagnosis when all her symptoms were caused by lupus and enormously improved for the first time by finally, y'know, treating the lupus, I've come to doubt the usefulness of schizophrenia as a label.
When schizophrenia is in practice reduced to "What might be interpreted as psychosis plus general impairment," as seems to be the case to me, while at the same time being thought of as something basically genetic, then there seems to be enormous potential for it to act as a diagnostic thought-stopping cliche. Maybe at a high level "what might be interpreted as psychosis plus general impaiment" is best correlated with genetics, but if it's actually the case that there are dozens of specific genetic conditions or genetic vulnerabilities each of which manifests apparent psychosis plus general impairment under certain conditions, in response to certain things, or in the absence of certain treatments, then wouldn't it be better to not treat schizophrenia as a monolithic "thing" with a "cause" at all, but rather as characteristic presentation of many things, most of which are ultimately genetically-based?
So I guess I'm on the side of: "The studies are probably right, but in a philosophical sense we shouldn’t describe it as mostly genetic."
I assume the real issue is that “genetic” is ambiguous between things like sickle cell or Tays-Sachs where there is a very specific gene variant whose presence or absence is definitive, and things that involve complex interactions between many genetic variants and the environment. It’s not like the former. And we don’t have good use to make of the latter category yet, so it isn’t one that we have broadly developed useful vocabulary for describing.
If people need a low tech method, not reproducing with people who have a family history of schizophrenia should help. On the other hand, if the percentage of schizophrenia is the same worldwide, it doesn't seem to help much, or is harder than it sounds, or helps the same amount everywhere.
I'm still worried about the eugenics thing.
I really do understand why you're so defensive of polygenic screening. I wouldn't judge any individual parents for choosing the option, and obviously in 2024 you are in no meaningful sense affecting the genetic fabric of society by doing so. But all the same, I worry about what it does to humanity in the long term if it becomes normalised, and I think you're wrong not to.
To get it out of the way there's the obvious thing where if it becomes possible, bigots might select against having (non-cripplingly) autistic, or gay, or etc. children in sufficient numbers to endanger the long-term survival of those traits. But that's not even what I'm most worried about. I'm worried that trying to do polygenic screening with our current half-baked understanding of genetics is the equivalent of doing advanced chemistry with a kitchen blender.
That is to say: can we reliably select against schizophrenia without selecting against various non-harmful forms of neurodiversity? Or indeed, random unrelated benign traits which happen to be genetically correlated with schizophrenia but have to do with the shape of your nose or left-handedness or something? On a societal level I worry about this a lot. I find it *most* concerning with people selecting against severe autism than against schizophrenia, because severe autism is more clearly on the same directly-related spectrum as personality types which humanity would be immeasurably poorer for not having. But A) I'm not sure schizophrenia isn't also correlated with positive /benign traits, and B) I don't think I trust society to be responsible with these tools if they get normalised, and *only* use them to screen out "safe" conditions.
My family has a history of schizophrenia, including members of both my immediate and extended family. We also have a history of high academic achievement, which includes advanced degrees in hard sciences and at least one department head at a major US university. I have noticed other families of high academic achievers with schizophrenic children.
Is there any data that links schizophrenia and high academic achievement within a family? I feel like I have seen this mentioned before. If so, is there any possibility that selecting against schizophrenia would select against gifted people as well?
You are ignoring the massive role that psychological injuries play in people later developing symptoms of psychosis. People with 5 Adverse Childhood Experiences are anywhere from 60 to 150 times more likely to have these symptoms, than people with no ACE's. https://pubmed.ncbi.nlm.nih.gov/17586579/ Schizophrenia is hugely influenced by being emotionally neglected, sexually abused, having a raging alcoholic as a parent, and other emotional wounds. It is not fundamentally in the genes.
I suppose one argument for not saying schizophrenia is genetic is to prevent that from becoming a dogma that can't easily be displaced? Like suppose some researcher has a theory about the cause or a treatment for schizophrenia but it's not linked to genes in any clear way and so when she goes to apply for funding she gets rejected because "everybody knows it's genetic".
You probably aren't arguing that schizophrenia should be specifically and always described as a genetic disease you just don't want people to be in denial about the role of genetics but perhaps (and to be clear I haven't read them) they are arguing against a more extreme version of your position.
Is it fair to say that genes do not cause two-leggedness? The heritability is near zero, on the same meaning of heritability. For bonus-points, y'all obviously agree accent is more heritable and therefore more genetic than two-leggedness, right?
Can't have any isolated demands for rigor here, and in certain communities any demand for statistical rigor is isolated, so please everybody get with the program.
I'm a public health professional. I think there is no difference between cause and risk-factor for scientists or professionals, but there is a big difference when creating public messages. Many people interpret "cause" to mean 100% correlation, as in, smoking 100% always causes cancer. When you say "risk-factor" people are better at understanding that there is some amount of probability around the factor. So generally you see a preference for saying risk-factor.
> "more satisfying cause of schizophrenia"
I think this is the core of the failure in communication. The human mind does not seek truth, it seeks satisfaction.
For a person not affected by schizophrenia, it's perfectly normal to think that "Schizophrenia is a disease that crazy* people have, and runs in the family" (*I'm not endorsing that word, but it is in common usage)
If you are affected, or work on it, that's not nearly good enough. A satisfying definition would include the idea that it's a bucket of conditions, not just one thing, and that it can cause a range of symptoms of varying degree.
When you work on a problem, you want to be able to work backwards and forwards with different suppositions to see what works.
Cause is like Proof - it's a place where you can stop thinking about something. If you say "I have evidence of something", that means you may have to keep looking. If you say "I have PROOF", then you can stop looking.
----
I wonder how the various parties would engage with the terms Major Contributing Factor vs Minor Contributing Factor.
It is unfortunate there is no special word like "causes" in English that means "always causes but is not always the cause of." Successfully feeding someone into a wood-chipper always causes death, but death is not always caused by being fed into a wood-chipper. You could say genes cause schizophrenia, but don't wood-chipper cause schizophrenia.
How does “the prevalence of schizophrenia is the same across societies” square with the large differences in schizophrenia rates by race?
>it’s a million times more than the drug abuse and social defeat issues
A million is a lot, this is meant rhetorically, like "a bazillion"...right? After the "Unintuitive Properties" post, I don't totally trust my surprise at seeing 10^6 mixed in with the other numbers.
Risk factors are not consistently defined. Here's a real science article making the point in a fun way: https://www.bmj.com/content/355/bmj.i6536
The tldr is that there's four common meanings for risk factor.
- diagnosis: something that can tell us if you have a disease
- prognosis: something that can predict later development of a disease
- treatment effect: something that will casually change whether you develop a disease
- aetiology: something involved biologically in the disease pathway
Great post, Scott!
typo: reminds of them of
Not directly relevant, but thank you for at least mentioning COPD. Everybody thinks tobacco will give you cancer, and it totally can, but it's way more likely to give you the other thing which will also kill you but in a very different (still horrible) way. We stress cancer way more, probably because cancer is an easier and more familiar thing to understand than an umbrella category for two overlapping disease states.
I think that the resistance to genetics as a cause for diseases is because it is very unsatisfying to blame genetics.
If the cause of your disease is environmental and beyond your control, you can always blame society. You are the victim of systemic injustice because society did not protect you well enough.
If the cause is environmental and thought to be within your control, you can always be blamed yourself. You should have known better than to smoke.
If the cause is genetic, that does not make for a very satisfying narrative outside of Nazi circles. Blaming your parents for not using screening or CRISPR is frowned upon, so you can just call it bad luck.
To the degree genetics as a cause is actionable, it would proscribe actions (gene screenings, selective abortions, DNA edits) which would make huge parts of society uncomfortable.
Didn't you write an article about how mainstream media rarely lies and apply a standard of "lying" under which one could claim the national enquirer rarely lies? (In that they didn't lie if they accurately reported what it was their source said to them without comment on the accuracy of that source)
There’s a disease that occurs only in people with a known, single gene (actually one allele). Because we know that gene, we know exactly what’s happening when it kills someone.
And yet it doesn’t kill everyone it could; people with the bum gene are still alive. Though if it did wipe them all out, we’d probably *still* blame the vector, not the victims’ genes.
The disease is mad cow, and the cause is ingesting misfolded prions. We are so impressed by a concrete explanation that most people don’t even know every victim (save one, IIRC) was genetically identical in one spot. If that’s not a genetic disease, how is schizophrenia?
Nobody wants a GWAS disease. It’s too fuzzy and it smacks of either missing something on the causal side or the definition being too broad. Like other unsatisfactorily explained diseases (MS, chronic fatigue, etc.), it’s better to admit ignorance than blame a thousand insignificant genes in the victim affecting an entirely hypothetical no-voices system in his brain.
Genetic screening for dominant genes whose pathological effects appear in adulthood reduces the frequency of those genes. That is an unalloyed good only if we certain than any possible beneficial effect of those genes is outweighed by its pathological effects. But we can't be certain of that.
Imagine a future ultracontagious virus that kills everybody who doesn't have the gene for Huntington's Chorea (WIV is working on that right now -- maybe...). With 100% screening of embryos for the gene for Huntington's Chorea, human beings go extinct.
"health care should be about treating schizophrenia, not preventing it"
I was trying to figure out why anyone would oppose polygenic screening and this sounds like the real reason. A doctor or therapist can't make a career transition to genetic screening tech, so they correctly see "we could cure schizophrenia" as a threat to their employment.
How do most of these arguments not also apply to IQ being genetic?
Genetics has nothing to do with schizophrenia, save for growing up among the deviations from “society’ as it is accepted and wanting to be ‘someone else’. Some young actors have been diagnosed schizophrenic after taking on a role and keeping it. You can pretend to be someone else and end up really believing it. To someone who craves attention, they can create as many different “voices” as they want. Not being accepted in society is what keeps them so sad and hides away in the group they created for themselves. When someone is sick or not all there as with drugs, some sayings can stay with the person. I have helped a person see where there voices originally came from. A friend kept telling her she never listens so she wasn’t. They have to be active in their life and at least able to laugh once in a while to be able to even talk to them about it.
Ash Clif_high. He says he has the ‘schizophrenic gene’. But he is not schizophrenic . Basically if you cannot get them to create their life, they will just stay apathetic and reactionary. The medication for psychotics is a tranquilizer,......this keeps them quiet and depressed unless they can form some life while on them. Drugs is definitely not the answer. They need acceptance and confidence and acknowledgement. Cal mag and B vitamins really help.
Once argued with someone who said it was "reductive" to say that "genes are among the causes of risk of skin cancer." And at that point I gave up because at a certain point demands for further nuance just become a reason not to have thoughts.
Is there really a clear definition of "schizophrenia"? This was all very interesting -- no irony -- but without a clear definition of the supposed disease, talking about its "causes" seems somewhat beside the point.
I have a question: if schizophrenia is mostly genetic, then why is there so little evidence of it in ancient texts? (Whereas other mental illnesses are described in Greek and Roman sources etc.) Doesn't that suggest that there are factors new in the environment that cause it (perhaps in interactions with genes or recent gene mutations). As well, all twin studies are compromised by the shared womb and in the case of adoption, by shared adoption trauma, and while genes run in families, so do environmental factors, stress levels, nutritional factors, parenting styles, and intergenerational traumas.
From Iain McGilchrist (The Master and His Emissary, page 404):
"... schizophrenia has in fact increased in tandem with industrialization and modernity. In England, schizophrenia was rare indeed, if it existed at all, before the eighteenth century, but increased dramatically in prevalence with industrialization. Similar trends can be observed in Ireland, Italy, and the United States, and elsewhere ... What is beyond reasonable doubt, however, since it has been established by repeated research over at least half a century, is that schizophrenia increased pari passu with industrialization; that the form in which schizophrenia exists is more severe and has a clearly worse outcome in Western countries; and that, as recent research confirms, prevalence by country increases in proportion to the degree that the country is "developed", which in practice means Westernized. Descriptions of melancholia, or of manic-depressive (now called bipolar), are immediately recognizable in accounts from ancient Egypt, Greece and Rome, yet there are no descriptions of schizophrenia ... The relative risk of developing schizophrenia in an urban rather than a rural setting is nearly double, and the evidence suggests that is it more likely that the urban environment causes psychosis than that high-risk individuals migrate to urban areas."
I can't believe that you would reduce THE Ronald Fisher to a tobacco statistician. Back in the day it might have been a reasonable position to say that the causation hasn't been proven to a high degree of confidence. Just because it turned out to be true, doesn't mean it was wrong to caution against premature regulation!
> But age is certainly more than correlational with flu
Minor but: but your example is wrong. If you changed age and held everything else equal then you wouldn’t get the flu less. It’s because age is correlated with other health properties (like a weaker immune system) and you’re assuming these change too when you change age.
I haven’t read the Aftab post or Torrey paper, but it seems that something is being assumed at the outset of the arguments you’re responding to—and your responses to them—that hasn’t been established yet, as far as I’m aware; namely, that there are “genes that cause schizophrenia” (as stated in Argument 3). Has this actually been established?
I’m reading Chris Palmer’s book, Brain Energy and he has a chapter on genetics in which he claims that since the human genome project, we haven’t been able to find specific genes related to mental disorders, though researchers have tried very hard to do so. My understanding based on my reading of Palmer’s theory is that epigenetics—the turning on or off of genes that are mostly common in all of us—are what produce psychiatric symptoms when epigenetic regulators like mitochondria are defective. Interested in your (or anyone’s) thoughts on this, as well as the whole “mental illnesses are metabolic disorders of the brain” idea promoted by Palmer. After all, the satisfying nexus, or the thing that “all schizophrenia causes have in common” might be, according to this idea, mitochondrial dysregulation—which happens to be treatable by various methods, including (but not limited to) medication.
Thanks, from a schizoaffective with his illness in remission for 5 years!
You say "If cannabis really is involved in schizophrenia, it’s probably 1-3% of the variance - but nobody is entirely sure it’s involved"; Do you think that the risk of psychosis from cannabis use is currently exaggerated by the media? (E.g. perhaps https://www.usatoday.com/story/life/health-wellness/2024/01/31/what-is-cannabis-induced-psychosis-california-stabbing-reveals-dangers/72411293007/).
In my opinion, this all comes back to the irresponsible shirking of physical brain scans which we have the technology and enough information to do. It's frankly bureaucratic stupidity from my seat.
But if you want my wild stab, I think it's environment poisoning. Some persons are more inclined to sensitivity writ large. It's not necessarily a weakness toward anything, but all things. You can really see this in autism more clearly which is highly related to schizophrenia.
So causal things more apt to create disability would include, in no particular order, metal poisoning, mother stress, mother drugs, father age, heavy toxins/molds, toxin history of mother and father, nutrition, lack of natural sunlight, lack of functional digestion, metabolism dysregulation, working your body (mostly mother) in regularwork performance .. and yes, I'll risk it potentially even up to and including getting poked with a needle up to around age 2 or 3 in particular.
"This is possible, but I think it’s false. We know many apparent risk factors for schizophrenia: cannabis use, birth canal asphyxia, social defeat, toxoplasma infection, poor prenatal nutrition. If you try to combine these into one big picture, it doesn’t work."
There's a theory that does: Cell Danger Response Theory by Robert Naviaux. All the above are related to over/under-engagement of the healing process in response to stress/trauma. He even lists toxo as a possible disorder from the CDR malfunctions. In short: When you have a chronic stressor in your life for over 6 months or of sufficient magnitude your body cannot complete the healing process and it gets stuck at one of 3 sequential stages. 1 Cell shutdown and hardening, 2 cell replication , 3 cell differentiation and establishment of connections to other cells.
As the 334th comment I expect this to get buried, but in any case there is a literature you may find helpful regarding the erotetic nature of explanation: We generally aren't explaining things in extreme specificity and/or extreme universality (a God's eye view). Instead, we are evidentially-limited beings with goals, and we choose the scope and granularity to explain something to a situated human for a specific purpose.
That provides a framework to understand the fact that the same question may have more than one answer. It also allows one to make systematic sense of why some causally-relevant factors are placed in the background rather than called "causes," and why the background and foreground can change.
We take the background for granted. Those conditions are "normal" while the thing we call the cause is the intrusion on the normal state, or a violation of the state we prefer (in other words "normal" can refer to frequency of occurrence or normative preference). The distinction between cause and background condition is often NOT in the modal logic of implication. It is rather based on our own interests and preferences about where to intervene to change things, or our preference for assigning responsibility/blame as agents.
Van Fraassen is typically presented as the classic argument for erotetic explanation. But I found Hart and Honore to be more eye-opening. If you've never read their work on causation and the law, I highly recommend it. A place to start: https://scholarship.law.vanderbilt.edu/cgi/viewcontent.cgi?article=3749&context=vlr
It had never occured to me that biology was sometimes easier to fix than environment...
Now it see it everywhere.
For example obesity is both a stigma and a medical condition surrounded by truckload of environmental (if not moral) advice and one a someone invents a drug that just fixes obesity and it is not a problem anymore
I don’t think the cause | risk factor distinction is explained well using the language of necessary and sufficient conditions, but here’s an alternative:
Causes are events which lead to the dice rolling (for a given outcome to happen). A given cause doesn’t have to be the *only* way for the event to happen.
Risk factors influence the odds of the roll. Alternatively, risk factors can also refer to any change in the success criteria. Smoking causes lung cancer, but being a heavy smoker is also a risk factor for lung cancer because repeated rolling of the dice increases the probability that one will hit.
My gut says this definition is pretty close, but I’m concerned that (e.g.) smoking would, given a deep enough understanding of the underlying biomechanics, only be a risk factor. Maybe the more expert commenters can evaluate better.
I stumbled onto this Substack just today. I perceive many readers/commentators are way smarter than me. I only want to say, I had a terrible schizophrenic sister as well as a brother with bipolar disorder. None of this information was around forty years ago. My grandfather committed suicide (in his younger life he had electroshock therapy for severe depression). My dad wasn’t right but it was the 50’s and 60’s. I have considered these disorders hereditary based on my personal experience (and some reading). As a mom, I made sure my kids (now fine adults) knew what to look for if they felt/believed such and such. I wanted them to know our family history and be aware because it seems to me once the disease is in full swing, the patient becomes the proverbial ‘I’m not mentally ill’ person who is very hard to help. It’s a terrible burden I would advise for genetic testing when available.
“ You can never 100% prove a number with no assumptions.”
This is borderline Liars Paradox.