Genetic causation in mental health and behaviour /cont./

Let’s take another look at heritability and genetic causation. I go on ranting on this topic as we so often hear that this or that is “highly heritable” or “mostly caused by genes”. For complex conditions, these claims are made based on estimates of heritability (see last week’s post).

Heritability is generally calculated as the difference in concordance (i.e. that both twins share a feature) between monozygotic pairs (who share 100% of their genes or almost) and dizygotic pairs (who are no more alike than any pair of siblings, sharing on average 50% of genetic variation). Tripartiate models assign percents of the population variance to genetic, shared environment or unique environment effects. Shared environment is non-genetic effects that make the twins more similar, unique that which makes the individuals more different.

But let’s focus on one aspect of twin research: monozygotic pairs who are concordant or discordant for a feature. They share almost all their genetic information, even if each individual may have unique de novo-mutations. If one of the twins has a certain feature, what is then the probability that his/her sibling also has it? This is answered by the pairwise or proband-wise concordance rates. Pairwise looks at each pair just once, while proband-wise counts concordant pairs twice (as both individuals with the feature are counted once as proband and once as sibling). The latter is often preferred for statistical reasons, but the former gives us a more accurate view to understand causation.

Pairwise concordance looks at pairs in which at least one of the twins is affected. Given that the co-twin will have the same genes and, to a fair extent, the same environment, his/her probability of having the phenomenon under study will be an easy-to-grasp measure of its overall predictability, the degree to which it is caused by shared factors.

First, we will see that the overall prevalence of the phenomenon in the population will influence the results. Naomi Wray and Robert Maier put it like this in a recent review:

“For example, for a disease with lifetime prevalence of 1% and heritability of 80%, less than half of cases are expected to report family history when considering all first, second, and third generation relatives (Yang and co-workers, 2010). Likewise, for the same disease more than 60% of monozygotic twins are expected to be discordant for disease status (Smith, 1970).“

Sebastian Lundström recently mapped all twin pairs with an autism spectrum disorder in the Child and Adolescent Twin Study in Sweden (CATSS). Autism is often cited as one of the most heritable conditions in psychiatry, and indeed among all complex conditions in medicine, with a heritability of 80%. How many affected monozygotic siblings of a proband with autism did we find? Thirty-nine percent. This means that the causation from genes to mental disorder is not straight-forward. It is wrong to assume that genetic causes of autism will add up to explain 80% of all autism, or that a person with autism is doomed by his/her genes to go on having autism in 80% of cases, or to 80% of the present degree.

But genes do matter. Not only did they rise the risk of autism 30 times or more, they also predicted other types of early-onset mental health problem in nine out of ten siblings. If you share your gene set-up and early environment with someone who has autism, you have a small chance of being completely healthy.

One possible explanation to this finding is that genes interact on a “final common pathway”, which can give rise to many different forms of health problems. I will go into this possibility at length later in the series.

Take-home message: genes matter, but genetic causation is much less clear-cut than assumed in statements such as “autism is 80% genetic”. The enigma of the “missing heritability” is not so enigmatic after all, but results from a simplified translation of heritability estimates to causal pathways.

Heritability estimates: do they equal degree of genetic causation?

This is the first entry in a series on causation, mental problems and criminality that will run on the blog for the next month or so as Forensic Fridays. This first entry will deal with genetics and what can be inferred from heritability estimates. The series will start broadly with causation in neuroscience and behavioural sciences and later focus on forensic psychiatry and the use of mental health expertise in courts of law, where sometimes diverging understandings of causation come to the fore.

Almost every type of behaviour, personality trait or liability for mental health problem that has been subjected to family studies (esp twin studies), has been shown to be to some extent under genetic influence. The proportion of the population variance ascribable to genetic effects is usually between 40 and 80%, with the higher end of the range reserved for narrow phenotypes such as autism, schizophrenia or bipolar disorder. Personality traits typically come in around 50%, while some behaviour patterns are higher, like persistent aggressive antisocial behaviours at 65% (Burt and coworkers, 2009). The great value of this research is that it has proved beyond any reasonable doubt that life is unjust (as if we hadn't guessed before, but the blank-slate model was actually commonly defended not that long ago ). Humans indeed come into the world differently equipped and have different possibilities of abstaining from substance abuse, violence and other destructive behaviours. Let this inspire some humility in all of us.

But then things get less clear-cut. Very few gene variants that account for more than a percent of the variance have been identified, and, even when combined, such effects amount to far less than the heritability calculated in family studies. This is referred to as the “missing heritability”. But why has it been so hard to identify the responsible gene variants? There is a high degree of complexity in the integrations between genes (meaning that each carries a small effect size on the group level but may exert large effects in specific families or individuals), there are epigenetic modifications, gene x environment interactions and many other possible explanations.

But did a heritability of, say, 65%, ever mean that 65% of the causation behind the phenomenon under study was genetic? No, it didn’t, it just said that 65% of the difference in liability across a given population at a  given time could statistically be referred to genetic effects. Consider this: over history, sociocultural changes have had very large effects on, for example, violence. The introduction of a strong state in the 15-17th centuries reduced violence by factors of 10 to 100. During such a time frame, very little genetic change happened. Yet the spectrum of violence in society changed dramatically.

Or take the example of height. If nutrition is lacking, people will be of short stature. Then, genetic effects will be smaller, as the general growth is hampered by an environmental factor and the overall variance is constrained. But when nutrition is adequate, the role of genes in the total variance will be larger. Children of tall parents will be much taller than children of shorter parents. But wouldn’t this also mean that nutrition is the main cause of someone attaining his/her adult height? 

So, for this week, heritability means that the variation of something in a population is under genetic influence - but it does not measure the degree of genetic causation for the phenomenon per se. Therefore, the missing heritability may not be such an enigma after all. There was never any reason to believe that it would be easy to identify the genes behind non-Mendelian, partly heritable behaviours or health problems.