Why autism should not be treated as a single condition

From the outside, autism can be difficult to understand. The unique sensory world of an autistic person can mean that a joyful event—like a Christmas party—can become nightmare of noise, lights, jostling and invisible social rules. Yet sometimes autism offers huge upsides, such as giftedness in music, maths, or art.

These difficulties mean that autistic people can often find the world stressful and challenging. Hiding their differences, a state known as masking, requires enormous energy. Overloaded children may manage their distress through puzzling behaviour such as meltdowns, or repetitive movements such as rocking or hand flapping. It can also create “burnout”, where a mismatch between abilities and the demands of the world creates intense physical and mental exhaustion.

Robert F. Kennedy junior, America’s health secretary, thinks that autism has become an “epidemic” in his country. His concern stems from figures from the Centres for Disease Control and Prevention, which shows that the condition now affects 32 per 1,000 eight-year-old children in America (see chart). That is in contrast, he says, with the near-absence of the condition in his childhood. Mr Kennedy was born in the 1950s, and studies estimate a prevalence of autism of around two to four per 10,000 in the 1960s.

The increase seems striking. But much of it is an artefact of the widening definition of autism over recent decades, increased awareness of the condition, and improved and earlier detection. Much, but not necessarily all. While Mr Kennedy and some other politicians mis-state the risks of painkillers in pregnancy, or advance the myth that vaccines cause autism, scientists have been investigating the genetic and environmental causes behind the condition, to work out if there might be as-yet-unknown factors that are contributing to the rise in diagnoses.

An improved biological understanding of autism could also help clarify whether distinct conditions are being unhelpfully grouped under too broad a label. The clinical definition of autism spectrum disorder (ASD) encompasses a wide variety of people, with hugely varying symptoms. People who appear to have relatively mild social limitations, but who can otherwise live regular, independent lives, can be diagnosed with ASD alongside those with profound intellectual disabilities, who need full-time support and care. One way for autistic people to benefit from more appropriate support or, in some cases, treatment, is to find better ways of distinguishing between different forms of the condition. Emerging research on the genetics of autistic people suggests a way forward: stop thinking of ASD as a single condition.

Autism is rooted in atypical brain development that begins very early in life, even in the womb. It affects the structure and sizes of some brain regions as well as how brain cells form, organise and communicate. This results in too much, or too little, communication between them.

The condition has a strong genetic component. Heritability is estimated to be more than 80%, meaning that these differences are the main reason some people have a higher risk of autism than others. In a small number of cases this can make it easy to find a cause for a person’s condition. A single variant of a gene—say, a duplication of a DNA stretch or a mutation that makes the gene stop working—can be enough “for people not to speak and to have difficulties in social interaction”, says Thomas Bourgeron, a geneticist at the Institut Pasteur in Paris. Children with mutations in these “high-impact” genes often have a plethora of other diagnoses too, from epilepsy to intellectual disability.

But these variants, which can either be inherited or arise randomly in the sperm or the egg before a person is conceived, are rare. At most they account for a fifth of all autism diagnoses. It is thought that most autism derives from far more common genetic variants found widely in the general population. Each variant may only slightly increase a person’s risk of autism, but when a child inherits many from both parents, they can cross the threshold for an autism diagnosis, says Dr Bourgeron. In other words, two people, each carrying some autism-linked variants and perhaps showing some autism-like traits, may end up “pooling” variants in their children.

As well as helping to identify people with the condition, genetics offer tantalising clues as to how autism may play out on a biological level. High-impact variants—such as SHANK3 and NLGN3—are often found in genes involved with how neurons send messages to each other. Some directly code for proteins operating at the junctions that connect neurons to each other, whereas other genes regulate how and when those proteins are produced. Scientists consequently believe that connections within the brain must play a role in producing autistic traits.

More common, but less dramatic, variants also offer hints. In a study of more than 46,000 Danes, around 40% of whom were autistic, Jakob Grove, a mathematician working in bioinformatics at Aarhus University, flagged several genes—such as KCNN2 and FEZF2—that are primarily active in the amygdala, a region of the brain responsible for fear, anxiety and social communication; the hippocampus, which is critical for memory; and the neocortex, which is involved with sensory perception. All of these can be affected in autistic people (although Dr Grove is hesitant to pin too much on genes that increase risk by only a small amount).

With hundreds of genes, and thousands of variants, thought to be implicated, autism is likely to involve many neurobiological perturbations in the brain. Some research points to dysfunction in the production of dopamine in the brains of some autistic people. Another idea is that problems with the social-reward system might underlie other forms of the condition. This huge variability in genetics and symptoms raises a question—what if autism is not a single condition, but several?

Autism has been subdivided before. When the condition was first officially recognised in 1980 and included in the Diagnostic and Statistical Manual of Mental Disorders (DSM)—sometimes called the “Bible of psychology”—it was a narrow diagnosis focused on young children who seemed not to respond to social engagement.

But clinicians found that too restrictive. To capture a larger, more heterogeneous group who were showing autistic traits, an update to the DSM in 1994 created five categories: classic autism, Asperger’s syndrome (defined by poor social communication but no delay in language development), childhood disintegrative disorder (in which young children regress developmentally and lose previously acquired skills), pervasive developmental disorder not otherwise specified (for people who fulfilled some, but not all of the diagnostic criteria for autism) and Rett’s syndrome (a condition controlled by a single gene variant that mostly affects girls).

These categories were ultimately ditched, however. Not only did the number and severity of autistic symptoms a person had overlap between the groups, the groups themselves did not do a useful job of predicting how an individual’s condition would eventually develop. By 2013 the condition had been redefined again, as the present-day ASD.

That has not stopped biologists from continuing to look for patterns among people with ASD. Researchers at Princeton University and the Flatiron Institute in New York, who looked at genetic and behavioural data from more than 5,000 autistic Americans, recently showed that it is possible to break ASD into four subcategories, each with its own genetic profile affecting development. The team first analysed the behavioural symptoms and development of the people in the cohort. Four separate groups of people emerged whose symptoms clustered together. Then they found that the groups differed genetically, too—not just in which genetic variants they carried but when those genes were active throughout prenatal development and into childhood. They published their findings in Nature Genetics in July.

One category, which they called “broadly affected”, included people with profound challenges across all autism-linked traits: they had developmental delays, showed limited and repetitive behaviour, were anxious and struggled severely with social communication. They were also more likely to carry rare genetic mutations. A “moderate challenges” group captured people who seemed to struggle the least, and a “mixed” profile included those with developmental delays and impaired social communication but who showed little anxiety and disruptive behaviour.

Last, there was the “social/behavioural” category. Although many autism-linked genes are switched on during developmental windows in the womb, the genes involved in the “social/behavioural” category often do not switch on until after the person is born and some continue to increase in activity into adolescence.

These children often grow up hitting their developmental milestones at the same time as their neurotypical peers and they are diagnosed later, too, says Natalie Sauerwald, a computational biologist at the Flatiron Institute who co-led the work. They also tend to meet the diagnostic criteria for attention-deficit hyperactivity disorder (ADHD) and severe depression. This “social/behavioural” group also seems to fit with a developmental and genetic profile noticed in a separate study published in Nature in October.

This kind of research, into potential subtypes of ASD, will help to identify coherent, biologically informed ways to make sense of the spectrum’s enormous diversity. Knowing that some autistic people are at risk for ADHD or mental-health concerns, for example, can guide decisions over schooling, and allow for better support to be offered earlier.

But genetics will never be enough to explain why autism develops in the way it does. A person’s environment also matters. The idea of environmental influences for autism became tainted after Andrew Wakefield, a British doctor, claimed (wrongly) in 1998 that the measles, mumps and rubella (MMR) vaccine was a cause of autism. This led to years of wasted research trying to find a link that did not exist. So when Mr Kennedy, himself a campaigner against vaccines, announced that the National Institutes of Health would embark on a $50m programme called the Autism Data Science Initiative (ADSI) to get to the bottom of autism’s environmental causes, many researchers were worried.

“All of the autism researchers who went into it were nervous,” says Judith Miller, a psychologist at the Children’s Hospital of Philadelphia who is leading one of the ADSI projects. So far, however, the research is running smoothly. Her project tracks children born as early as 2008 who have moved through the hospital and been screened for autism—roughly 104,000 in total, 4,000 of whom are autistic—and includes some data on genetics and maternal health during pregnancy, as well as information on the quality of air, water and green space in the children’s living environments.

Dr Miller hopes her new project will be able to find correlations between environmental exposures and autism that can then be more rigorously tested. Although some factors—notably maternal health, air pollution and hormone-disrupting chemicals—have observational evidence behind them already, collecting data that accounts for many of these risk factors at once, as well as genetics, could reveal new relationships and explain why some people might be affected more than others. “What we’re trying to do is figure out, can we hone in on what might be hereditary and then can we boost that prediction with some of these other factors,” she says.

It is easy to imagine how environmental factors might fit in with the genetic contribution, says Zeyan Liew, an epidemiologist at Yale University. For example, higher parental age—arguably an environmental factor—has been linked to a child’s chance of being autistic, yet the mechanism is thought to be genetic. As people age, their sperm and egg cells accrue new mutations. Such mutations could then confer autism on any resulting child. According to a review in Molecular Psychiatry in 2022, particles of air pollution or some heavy metals might similarly induce new mutations, either in the sperm and egg or in the early embryo, by damaging DNA or disrupting its natural repair processes.

Another possibility is that an environmental trigger turns up or down the activity of a gene linked to autism, a so-called epigenetic effect. One of the most well-supported environmental causes of autism is exposure to valproate, an epilepsy medication, in the womb—and valproate is a known instigator of epigenetic changes. Valproate caused a scandal around the world because pregnant women taking it were not warned that their fetuses could be harmed. Or a straightforward environmental exposure could be to blame. For example, although a causal link has not been established, a meta-analysis of 36 studies from 2021 found that mothers who experience a serious fever during pregnancy more often have autistic children. Many possibilities like this are known from observational data, animal studies and knowledge of developmental mechanisms. But conclusive evidence that any of these factors is causative in humans is lacking.

Treating autism as a single condition might explain why it is proving hard for researchers to draw firm conclusions—if autism actually breaks down into different categories, and a particular factor only affects one of them, then a study that lumps all autistic people together might never see a signal, says Dr Liew. “If you can refine your phenotype, you have a better chance of finding the cause.”

Splitting the spectrum up, though, is not universally welcomed. Some worry that it will increase stigma and exclusion for those with the most severe impairments; others worry about the opposite problem, that those with fewer visible difficulties will be marginalised and will not be able to gain access to support.

Meanwhile some parents argue that categories can be helpful, because of the different challenges involved. Jill Escher, a mother of two profoundly affected adult children and the co-founder of the National Council on Severe Autism, a charity in America, says many in her community look after people who will need round-the-clock care for the rest of their lives. They often also display complex behaviours, including self-harm, aggression, and the destruction of property.

Beyond behavioural and educational support for autistic children to help overcome specific difficulties such as speech, movement and emotional regulation, researchers are also looking for pharmacological approaches to modify some of the core symptoms, such as impairments to social communication and repetitive thoughts or habits. So far no approved medication does this. Some treatments, such as risperidone, an anti-psychotic, can help with the behavioural symptoms such as irritability, aggression and self-injurious behaviour. Getting better treatment demands a more sophisticated understanding of the biology of autism. That, in turns, means building a better understanding of both genetic and environmental causes.

Autism advocates have long argued—correctly—that society needs to do more to adapt schools, workplaces and public spaces so autistic people can thrive in them. Equally, however, better interventions will offer some people more autonomy, agency and greater well-being in a challenging world. ■