Sunday, January 04, 2015

Is Mental Illness Really Genetic?

For a long time, we've been told, insistently, that this or that mental disorder, such as schizophrenia, depression, autism, Substance Use Disorder, etc., is at least X% genetic, based on "heritability" analysis. Usually, we're talking about twin studies. Somehow, people have it in their minds that twin studies are a kind of Gold Standard. They're actually more of an aluminum standard, or a lead standard.

Some of the worst scientific "studies" in the world involve twins. The very worst are represented in the work of the avowed eugenicist Franz Kallmann, who "proved" the genetic basis of homoesexuality in 1952. The disturbing, one might even say delusional, bias that runs throughout Kallmann's work makes one wonder how anybody ever took him seriously. But even if we discard the work of Kallmann (which anyone should), we're left with tons of other twin studies, some good in quality, most rather obviously poor (for reasons of lack of researcher blinding and small sample size, among other concerns). Psychologist Jay Joseph and others have pointed out the serious methodological issue that underlies all twin studies, which is the "equal environment assumption." Basically, the assumption is that twins are raised the same way as other children (even non-twins). The advocates of twin studies essentially say that if a particular trait shows a high rate of concordance in monozygotic (identical) twins relative to that trait's co-occurrence in dizygotic (fraternal) twins, the only possible reason for the difference is genetics, since monozygotic twins have exactly the same DNA whereas dizygotic twins do not. (The same-DNA hypothesis is also wrong, on a micro level, but we'll ignore that for now.) It should be obvious, up front, that hundreds of thousands of variables go into raising a pair of children. (You know this, if you have children.) To say you've controlled for everything in the world except genes is ludicrous. It's particuarly ridiculous in the case of twins, since we know twins are celebrated (constantly, in obvious and not-so-obvious ways) as the wonderful accidents of biology they are. They're raised differently from other kids, and as much as some parents may try to accentuate the individual uniquenesses of the siblings, nonetheless the twins aren't fooled: they know they're twins.

Because the equal environment assumption is not satisfied, twin studies must be viewed with caution. That's a kind way of saying they're sketchy; not to be trusted at all, really. Why? Because of the built-in potential for the swamping of genetic effects by uncontrolled environmental variables. As Jay Joseph says, "Many critics have argued that it is likely that identical-fraternal comparisons capture nothing more than identical pairs’ more similar treatment, greater environmental similarity, stronger attachment and emotional bond, and greater levels of identity confusion (feeling like they are two halves of the same whole)." The potential for amplification of environmental effects is much greater in twin-rearing than in the raising of other children. If we find concordances in this or that trait (whether it's schizophrenia or the ability to eat French fries), are we to be shocked? Really? Of course there are concordances. How could there not be?

But another problem with twin studies is that the results, quite often, just aren't that impressive. Although early twin studies found concordance rates for dementia praecox in identical twins as high as 86%, subsequent studies have found much lower rates, and in fact, the later the study, the lower the rate. By the late 1980s (with schizophrenia much more precisely defined, in the DSM), some studies were reporting pairwise concordance rates of under 20% for schizophrenia in twins. The largest such study (involving over 16,000 subjects) found a rate of just 11%. If monozygous twins were truly identical genetically, and if schizophrenia were truly genetic, the occurrence of schizohrenia in one twin should mandate its occurrence in the other twin at a rate approaching 100%, not 11%. Yet somehow researchers focus narrowly on the 11% number, rather than feeling a need to explain the missing 89%.

Fortunately, we now have better, more powerful genetic techniques at our disposal for looking at the heritability of things we supposedly "know" are genetic: schizophrenia, depression, alcoholism, autism. But there's just one problem. When we do genome-wide association studies to find genetic features that can predict particular traits-of-interest, we come up emptyhanded. More on that in another post.

This post is derived, in part, from Appendix A of a mental health memoir I'm working on. Stay tuned for information on how to obtain free excerpts from the book.