What Autism Heritability Estimates Do and Don’t Tell a Family

Where the figure comes from

Fifty-five percent. That was the heritability of autism estimated by a 2011 twin study, the largest population-based study of its kind to use contemporary diagnostic standards, and the number was low enough to unsettle a field that had spent two decades assuming a much higher one. Pauline Chaste and Marion Leboyer, psychiatric geneticists at INSERM and Paris-Est University in Créteil, set the estimate in context in their 2012 review of autism risk factors in the journal Dialogues in Clinical Neuroscience.

Heritability describes a population rather than a person. It estimates how much of the variation among individuals in a particular group, at a particular time, can be attributed to variation in their genes. Move to a different population, or to the same one a generation later, and the estimate moves with it.

Autism clusters in families, and the clustering has been measured for decades. Siblings of a child with autism carry a recurrence risk for pervasive developmental disorder of 2 to 8 percent, rising to between 12 and 20 percent once siblings showing impairment in one or two of autism’s three symptom domains are included. Earlier twin studies attributed that family aggregation to shared genes rather than to the shared household, and autistic traits measured across the general population proved heritable at a comparable level, with estimates ranging from 40 to 80 percent.

The 2011 estimate of 55 percent drew criticism on two counts. The confidence interval around its odds ratio ran from 9 to 81 percent, a range wide enough to accommodate almost any conclusion, and the participation rate was low. Chaste and Leboyer, whose review surveys the field narratively and states no search protocol, nonetheless treat the study’s design as the strongest of the set and the size of autism’s genetic share as an open question.

A separate twin study published the year before found identical twins more alike than fraternal twins for autism, attention deficit hyperactivity disorder, developmental coordination disorder and tic disorder, with the cross-disorder resemblance behaving differently in the two kinds of twin pair. Whatever the genetic contribution consists of, it does not appear to respect the boundaries between diagnoses.

What the figure says about one family

A heritability estimate performs arithmetic on the differences between people, and the result cannot be redistributed back to any one of them. Fifty-five percent does not mean that 55 percent of a particular child’s autism is genetic and the remainder something else. No portion of any individual’s condition is assigned by the calculation at all.

The same limit governs the environmental share. Chaste and Leboyer argue that the evidence for environmental contribution is substantial enough to warrant far more research than it has received, and the argument concerns populations. It reaches no verdict on a particular pregnancy or a particular household, and a parent reading it is not being handed a place to look in their own history.

A high heritability also carries no implication about what can be changed. The statistic reports the sources of variation under whichever conditions happened to prevail while the measurement was taken; alter those conditions and both the variation and the estimate can shift. Nothing in a heritability figure speaks to whether a child’s difficulties will respond to teaching, therapy or accommodation, and nothing in it functions as a forecast.

The question the review finds genuinely interesting is how genetic vulnerability and environmental exposure combine, and the honest answer in 2012 was that most of the evidence came from animals. Mice carrying a mutation affecting the TSC2 gene showed a lack of normal social approach only when their mothers had undergone immune activation during pregnancy. Human family and population studies had not yet been extended to test such interactions, and detecting an interaction requires more statistical power than detecting genetic or environmental effects on their own.

Two prospective cohorts were launched to supply that power, each following 100,000 children from conception, one in the United States to the age of twenty-one and the other to the age of seven. Studies built on that scale report slowly, and any settled account of how genes and circumstance combine in autism will arrive on their schedule rather than sooner. The 55 percent you may have read was an estimate of variation within a population, criticised in its own field for an interval wide enough to accommodate almost any conclusion, and it has never described your child.