Picking pairs: People with autism are up to 11 times more likely than their typical peers to choose a partner on the spectrum.

Henrik Sorensen / Getty Images

The tendency of people with autism to partner with others on the spectrum may raise the condition’s prevalence within families and across the population, according to a new study¹.

Researchers used genetic modeling to estimate the impact of this ‘nonrandom mating’ pattern among individuals with autism. The approach uses equations to deduce how genetic traits move through populations.

Nonrandom mating may boost autism prevalence by up to 50 percent within one generation, the approach revealed. The results were published 12 October in JAMA Psychiatry.

The real-world effects of this phenomenon are likely to be modest, however. “It’s not at the level that would be a huge cause for concern,” says lead investigator Naomi Wray, professor of statistical genetics at the University of Queensland in Brisbane, Australia.

Wray’s study builds on another team’s analysis of medical records from more than 700,000 individuals in the Swedish National Patient Register, going back to 1973. That work found that people with any of 11 psychiatric conditions, including autism, attention deficit hyperactivity disorder and schizophrenia, tend to choose partners who share their condition². People with autism are up to 11 times more likely than their typical peers to choose a partner on the spectrum.

Wray and her colleagues fed the Swedish data into a computer model to calculate how nonrandom mating affects the prevalence of psychiatric conditions, as well as their heritability — that is, the degree to which they run in families. Nonrandom mating can render certain psychiatric conditions more heritable because it concentrates genetic variants associated with conditions.

Maximum effect:

One generation of nonrandom mating increases autism’s heritability by around 2 percent, the researchers found. The effect on autism prevalence at a population level could be more substantial, however.

Wray’s team predicts that nonrandom mating could trigger a 50 percent uptick in autism prevalence from one generation to the next. The prevalence would level off after nine generations, stabilizing at 2.4 times its original rate.

The study “clarifies for the field what the magnitude of these influences might be,” says John Constantino, professor of psychiatry and pediatrics at Washington University in St. Louis, who was not involved in the work. “I think it’s a very valuable article that way.”

However, the approach relies on assumptions that may not hold up outside the laboratory.

“What they’re trying to do is look at the biggest possible effects that assortative mating may have on genetics,” says Paul O’Reilly, senior lecturer in statistical genetics at King’s College London, who was not involved in the study.

Real time:

For instance, the approach assumes an abrupt switch from completely random mating to the relatively high degree of nonrandom mating from one generation to the next. In reality, some degree of nonrandom mating has probably gone on for generations.

It is impossible to know whether nonrandom mating contributed to the increase in autism prevalence over the past several decades, Wray says. But increased geographic mobility and a growing arsenal of social media apps allow people to connect with more potential romantic partners than ever before, so it’s likely to play a bigger — albeit still minor — role in the future.

The model may also not capture the unique features of autism, Constantino says, such as the fact that a genetic factor linked to autism may not confer equal risk for the condition in boys and girls. Evidence suggests that girls require a bigger genetic hit than boys do to show autism traits. “That is a nuance of autism that’s different” from other psychiatric conditions, he says.

Wray says the model accounts for this difference by using a sex-specific distribution of autism traits.

But she agrees that models do not always capture the nuances of real life. “At the end of the day, I think any results that come out of this sort of modeling have to be interpreted with caution,” she says. “But I still maintain that if you don’t try to model these boundaries, then you know nothing.”