8 Comments
User's avatar
Scott Alexander's avatar

I can't read the full Torrey article, but I'm confused by this and think it might reflect a misunderstanding of genetics.

It sounds like it's expecting there to be one gene which causes schizophrenia, which is no longer how people conceive of genetic conditions working (except a few very simple ones like sickle cell). Rather, there's some portion of the risk which is genetic (in schizophrenia, ~80%), distributed across hundreds or thousands of different genes (including the 300 Torrey mentions and others that we haven't found yet because their effect size is too small).

It's not correct to say that risk genes "play a role in the clinical expression of most diseases but do not cause the disease in the absence of other factors" , except insofar as no single risk gene causes it because you need many of them together.

Maybe Torrey addresses this in his article, but we don't need to speculate about whether schizophrenia is genetic or not - we can just use twin studies, which find that it mostly is. This doesn't mean there's no other factors that affect it, but it puts it firmly in the category of other mostly genetic things like height, which depends primarily on your parents' heights but can be affected by nutritional deficiencies, bone injuries and random embryological events.

Torrey's been pushing various weird infectious theories of schizophrenia for a long time, and confusing "there's no single gene that causes 100% of conditions" with "it's not mostly genetic" is a classic tactic for people who want to do this kind of thing - but although there's a little bit of residual variance for infections to explain, I honestly expected better of him and I'm surprised this argument made it into a journal in 2024.

Awais Aftab's avatar

Torrey does discuss heritability studies. I can email you the pdf if you like.

I think “heritability” gets misleading pretty fast. It’s an abstraction at the population level. Heritability is neither genetic causation nor is it proportion of risk applicable to an individual. Many conditions we don’t usually think of as “genetically caused” can show high heritability. Heritability estimates of COPD, eg, often reach 60%.

While 80% heritability sounds impressive, it’s sobering to look at actual numbers. Take Hilker et al, 2018 (https://www.sciencedirect.com/science/article/pii/S0006322317319054 ), a recent well-designed twin study of SZ using nationwide Danish register. They calculate a heritability of 79%, matching older estimates; they have a total of 81 monozygotic twin pairs in which someone has SZ: in 12 pairs both have SZ and in 69 pairs only one has SZ. If you are a monozygotic twin and you have SZ, there is only a 15% chance that both of you have SZ. The probandwise concordance rate of SZ is 33% in monozygotic twins and 7% in dizygotic twins. There is no way to explain this if SZ is 80% “caused by genetics.”

It gets trickier if we say that SZ is caused by thousands of genes of very tiny effect each. The more common variants there are, the greater the proportion of the population that will have them, but only 1% develop SZ and majority of them don’t have a family history of SZ. It’s clear that SZ is highly polygenic but the hundreds/thousands of genes involved appear to represent some sort of nonspecific vulnerability to psychopathology, rather than a causative link to SZ (a vulnerability that is neither necessary nor sufficient).

Scott Alexander's avatar

My impression is that most studies find twin concordance rates for SZ to be 40 - 50%.

I'm not sure of this, but I think in rare conditions, twin concordance rate should always be lower than heritability because of regression to the mean. That is, since schizophrenia is rare, we should expect the average person who gets it to have both a very poor roll of the genetic dice *and* a very poor roll of the environmental dice. That means that even if the genetic dice are more important, the average person who gets exactly the same roll of the genetic dice of a schizophrenic, but a random roll of the environmental dice, won't have schizophrenia.

You could do the same thing with environment - the average person who gets the exact same environment as a schizophrenic (whatever that means - not actually study-able) won't get schizophrenia, because they're genetically different. This wouldn't mean "it's not really environment, it's truly genetic and environment is only a 'risk factor'". It's just how variance works.

My impression is that the heritability of COPD is more like 30 - 40%, and that this is unsurprising - partly because everything is partly genetic, and partly because COPD is often caused by smoking and smoking status is partly genetic.

I agree that there's no clear distinction between "genetic disease" and "non-genetic disease" given that almost every disease has both sets of risk factors (including "purely genetic" diseases like sickle cell which have variable penetrance), but I think Torrey is doing an unfair thing of saying "So we all agree that means we can call the category all diseases are in 'non-genetic', and focus on the environmental factors", which we definitely can't.

Awais Aftab's avatar

40-50% concordance rate is the usual textbook figure, based on older studies with some methodological limitations (such as a selection bias towards twin pairs in which both have SZ); nationwide registries overcome some of those limitations.

What should count as a “genetic disorder” is not well defined. If you think any disorder with high heritability is a genetic disorder, sure, we can consider SZ a genetic disorder in that particular sense. But it remains the case that aside from rare variants, there are no common variants of large effect, most of these variants are nonspecific, and that these variants are not essential to develop SZ. They are more like risk factors. For any given person with SZ, they may have none or very few of these common variants or they may have a lot of them.

I don’t agree with Torrey about the infectious/environmental etiology part.

Scott Alexander's avatar

I don't know if we actually disagree or are just using different words, but I don't agree if by "risk factor" you mean something that "isn't the real cause".

Suppose someone got obese because they ate ten donuts a day for 5 years. If they'd only eaten donut #5000, and nothing else, they would not have gotten obese. So is donut #5000 a "cause" or a "risk factor" for their obesity? Is their donut-eating overall a"cause" or a "risk factor"? This is how I would think of arguments about whether X gene is a "cause" or a "risk factor" vs. whether the disease is genetic or not.

Sofia Jeppsson's avatar

Some people in the mental health field really get so hung up on the long game that they become positively hostile to doing what we can here and now. Like, people who are totally opposed to offering any kind of medication-free treatment. They're all like psychosis is in the brain, pure talk therapies were a failed nineteen sixties experiment that we shouldn't repeat, we just gotta do more brain research and make better meds.

Ok. But here and now, there are a substantial portion of people who are "non-responders", or who can't find any medication with tolerable side-effects. So what should we do with that group here and now? Try to find alternative approaches that at least HELPS, even if no one believes anymore that you can sit down and talk about your mother for a hundred hours and then you turn normal? Or should we just tell them to be patient and wait for some future scientific breakthrough that may or may not come?