Autism therapy at crossroads

Abstract

Various studies that have questioned the scientific evidence for established autism treatments have attracted both criticisms but also inspired new research to apply treatment more efficiently.

Research into therapies to treat autism is in a commotion after several studies demonstrated a lack of reliable evidence for the efficacy of longstanding treatments, especially non‐pharmaceutical interventions (NPIs). While these studies have generated some upheaval, they have also inspired new approaches to treatments that better account for the broad spectrum of conditions lumped under the umbrella term of autism rather than shoehorning a one size fits all strategy.

The problem for autism and its treatment is that there is no defined physiological basis for diagnosis comparable with other major neurological conditions, such as Parkinson's disease or Alzheimer's. Instead, diagnosis is based on observed behavioral, developmental, and physical characteristics, such as facial expressions, that become apparent during a child's first few years. Diagnosis and treatment are not made easier by the fact that there are no clear genetic, environmental, developmental, or other possible causes of autism spectrum disorder (ASD).

The problem for autism and its treatment is that there is no defined physiological basis for diagnosis comparable with other major neurological conditions …

This is also the cause of ongoing debate over the merits of pharmacological interventions which, given the absence of clear clinical targets such as a gene or neurological pathway, are often prescribed to treat symptoms rather than causes. First are antipsychotic drugs developed primarily to treat other conditions such as schizophrenia, that are prescribed only when patients exhibit more extreme behaviors. There are drugs for treating depression and anxiety, which are given alongside non‐NPIs. Third are drugs to treat the hyperactivity and shortage of attention span often associated with autism, but critics argue that long‐term improvements are more likely to come from NPIs. Finally, there are drugs to treat insomnia that can accompany autism. This is somewhat controversial because of the risk of dependency and side effects, but there is plenty of evidence supporting a link between duration and quality of sleep and cognitive performance.

… as often in the history of medicine, practitioners of established therapies will strongly react against anything that threatens their livelihood and the foundation of their expertise.

Non‐pharmaceutical interventions though are the primary therapies for autism and the subject of recent controversy. Meanwhile, some providers of traditional NPIs have responded to the accusations by pointing to long established records of success in some cases, especially for applied behavior analysis (ABA). Actually, there should be no contradiction in that some established therapies may work while lacking sufficiently rigorous evidence from properly conducted trials. But, as often in the history of medicine, practitioners of established therapies will strongly react against anything that threatens their livelihood and the foundation of their expertise.

A recent parallel is the story of statins, the cholesterol‐lowering drugs prescribed to decrease the risk of cardiovascular attacks or strokes. Concerns about the widespread use of statins came to the surface around 2015 when some doctors argued that benefits had been exaggerated and side effects downplayed (Demasi, 2018). Others countered that statins make a significant contribution to reducing cardiac events among people considered at risk (Krishnamurthy et al, 2022). What is relevant to the autism debate was the reaction of statin advocates defending a practice in which they had invested their reputation, along with a pharmaceutical industry that collectively had earned US$1 trillion from the drug by 2020, making it one of the most profitable drugs ever. A common element of the statin and autism stories is that although one is a drug and the other involves NPIs, both revolve as much around the quality of the evidence as over efficacy.

Standard of care

Established NPIs for autism come under five broad headings: early intervention, developmental, behavioral, educational, and sensory integration therapy (SST). As these titles suggest, there is considerable overlap, and more than one may be applied simultaneously in a given case.

Early intervention is just the idea that treatment of whatever kind should begin as soon as possible in a child's life. The developmental approach is the broadest category encompassing motor skills, speech/language, recreational therapy such as play, social interaction, and occupational therapy associated with everyday tasks such as dressing and eating. The third category, ABA, is the longest established autism treatment. Children are rewarded for desired behaviors but not for negative ones. This could be called the “carrot without stick” approach widely practiced these days by parents in general.

Applied behavior analysis in turn is broken down into styles, one being discrete trial training (DTT), where actions such as washing hands or brushing teeth are taken apart into component parts that can be practiced in isolation and then integrated into various practical activities. Second under the ABA category is Pivotal Response Training (PRT), which teaches core or “pivotal” skills from which other tasks can be learned. This might include conventions associated with starting a conversation, or asking for something rather than avoiding the task, or just resorting to grunting or pointing.

The Education category, often considered under the heading Treatment and Education of Autistic and Related Communication‐Handicapped Children (TEACCH), is meant to embrace a consistent “whole‐life” program tailored to individuals, based more on visual and social than language‐based learning. Children are continually assessed for development so that teachers can adjust the curriculum or classroom. Finally, there is Sensory Integration Therapy (SIT), which involves modulating reactions to sounds, sight, and other stimuli. This is either to avoid excessive response, sometimes called sensory overload, or to help children with autism overcome hyperresponsiveness when they fail to exhibit normal reactions to stimuli.

The controversy

The controversy over not so much the therapies themselves, but the evidence supporting them, dates back to May 2011, when a systematic review reported a lack of robust scientific evidence for established NPIs (Warren et al, 2011). This involved a search of Medline, PsycINFO, and ERIC (Education Resources Information Centre), identifying 34 separate studies from 2000 to May 2010 deemed good enough to merit consideration. Of these, only one was rated as good quality, 10 as fair quality, and 23 as poor. The strength of the evidence overall was ranked from insufficient to low. The study did find that ABA and some forms of early behavioral intervention were associated with improvements in ASDs, but there were concerns over the methodology of the studies. For other therapies, it was impossible to find any clear indication of improvements.

This was followed by a drip feed of other studies casting doubt on the evidence base for established ASD therapies before matters came to a boil in 2020 when Australian autism specialist Andrew Whitehouse authored one of the largest metareviews so far. Whitehouse, Professor of Autism Research at the Telethon Kids Institute and at the University of Western Australia, had been appointed by the Australian Federal Government to chair a group of research leaders assessing worldwide evidence for ASD interventions.

The report, available on request from AutismCRC (Interventions for children on the autism spectrum | Autism CRC), a body established in 2013 under the Commonwealth Government's Cooperative Research Centre (CRC) Program, was quite hard‐hitting. Perhaps this is why the report attracted a lot of flak from practitioners in the field. Whitehouse himself faced threats of legal action from professional bodies around the world and even some harassment of his family. He admitted this took him by surprise since he felt he was merely following the science, and that it took him some time to come to terms with it. “I think our review was the first to collate all of the evidence and put it in front of everyone in such stark terms,” Whitehouse said. “Challenging the status quo can be confronting for many, but that doesn't mean that it's wrong or not worth doing. As clinicians and scientists, our first obligation always is to the safety and wellbeing of kids and families – everything else is secondary to that. This means that we have to be brave to confront where we are currently falling down and use that as a first step to getting better. Highlighting the poor evidence underpinning many autism interventions is that first step.”

Around the same time, a US study of comparable scope and design called Project AIM (Autism Intervention Meta‐analysis) came to similar conclusions yet did not seem to attract the same vitriol from within the profession (Sandbank et al, 2020). The study aggregated evidence relating to seven types of early intervention, including SIT and TEACCH, and found positive effects only for two of the categories under the headings of naturalistic developmental behavioral interventions (NDBI) and early behavioral interventions when the analysis was confined to studies with randomized controlled trial (RCT) designs.

We were seeking simple answers to an issue that is highly complex and dependent on many contextual factors.

Project AIM's Lead Investigator Michael Sandbank from The University of North Carolina at Chapel Hill expressed similar hopes to Whitehouse over the eventual outcome of his study. “My hope is that they are pushing the field to conduct more rigorous tests of interventions,” he commented. “Running a rigorous experimental study with low risk of bias is difficult, and in the past, I think that difficulty led many in the field to accept lower quality evidence as sufficient. In the past decade or so, we've seen that it is indeed possible to conduct more rigorous tests of interventions, so standards are increasing.”

Whitehouse agreed. “I think the experience has highlighted the real challenges we experience in the autism field, particularly in terms of how entrenched different models of practice currently are,” he added. “It has also been mentioned to me that other scientists have observed the fall out and sought to challenge their own views regarding how they view the evidence base. That is a very big positive.”

Outcomes and future aims

Another positive outcome is that the analyses conducted separately under Sandbank and Whitehouse do seem to confirm the same treatment categories with the greatest evidence of efficacy, that is NDBI and early development. Yet, Whitehouse stressed that this was just a starting point. “This supporting evidence simply indicates that these intervention approaches can be considered by a clinician for a given child and family. The exact choice of which of these is most appropriate will depend on the child, their family and the environment in which they live,” he cautioned.

This should set the tone for future studies into the relationship between therapies and individual variations in the severity and nature of ASD. “Most of my research right now centres on meta‐analysis of prior studies,” Whitehouse explained. “Consequently, after we publish our new results, we'll likely examine whether child language ability when intervention begins affects the extent to which children benefit from certain types of interventions, using the data we have collected from prior studies. In the past, we have found that children that enter intervention with greater language ability tend to benefit more from certain types of interventions, suggesting that they have a developmental foundation upon which intervention can more effectively build. This suggests that children with limited language skills may require a different intervention approach, or that all children could benefit from communicative support (such as the provision of adaptive and augmentative communication supports) at the beginning of intervention to help them better access learning opportunities.”

In any case, the argument seems to be over the effectiveness of clinical studies rather than the underlying therapies…

Whitehouse added that past studies had failed to yield solid evidence of efficacy partly because they were analyzing treatments applied too indiscriminately. “We were seeking simple answers to an issue that is highly complex and dependent on many contextual factors,” he explained. “Any view that ‘one type of therapy is always better than another’ simply cannot be considered an evidence‐based approach, because therapy choices are highly dependent on the child, their family, and the environment they live in.”

More discriminating studies of ASD therapies would need to use statistical techniques such as moderator analysis to distinguish more precisely between impacts on different subgroups across ASD. “By applying moderator analyses, we can determine whether participants benefit more or less from intervention based on their individual characteristics, such as age when intervention starts, cognitive ability, and language ability,” Whitehouse explained. “We can apply moderator analyses in the context of meta‐analysis to look at these associations across studies, and often those analyses can shed light on associations that would not have been apparent in small primary studies.”

Treatment efficiency

Another aspect is the ability to adapt therapies on an ongoing basis, both to reflect a child's progress and to better match that individual. This is especially important for ASD where it can be hard to find the best therapy from the outset. “We can also use SMART (Sequential Multiple Assignment Randomized Trials) studies as a way to rigorously test whether adapted interventions will be more helpful to those who do not benefit from the initial approach,” Whitehouse commented. “There are a few good examples of these in autism intervention literature, but we need more so that we can better understand how to effectively adapt intervention when such adaptation is needed.” SMART is a pipeline approach to clinical studies where participants are randomized to one of two or more intervention options at successive stages. Each stage of randomization allows researchers to compare the relative effectiveness and then aggregate the results across the whole pipeline.

Such studies have already challenged previous assumptions, such as the idea that treatments can only be most effective if they start early, and that they need to be intensive in terms of hours spent. “What we are seeing in our own analyses is that age at intervention does not appear to be associated with intervention effectiveness, at least up to age 8 (when data collection tends to stop),” Whitehouse said. “So, while children who are diagnosed and supported early certainly benefit from intervention, child age does not appear to place a ceiling on the extent to which intervention can be helpful. We are also seeing that the amount of intervention, in terms of total hours provided, does not appear to influence the effectiveness of intervention, regardless of type of approach, so families can feel confident working with clinicians to design intervention plans that fit within their family routines rather than feeling pressured to pursue intensive interventions that might be stressful for their child and themselves.”

The challenge to the status quo has not been welcomed by everyone though. In particular, it has met resistance among practitioners of ABA therapy as developed by Norwegian‐American psychologist Ole Ivar Løvaas in the 1980s at the University of California, Los Angeles (Løvaas, 1987). Michelle Carney, ABA Program Consultant at Autism Outreach in the USA, argued there was not any evidential crisis in autism treatment research, largely because ABA is the only recognized treatment. “Dr. Lovaas' study in 1986 recovered almost half of the children in his study using 40+ hours of ABA, which is still the only recognized and proven treatment for autism,” she insisted.

… in lieu of clear genetic factors, there is mounting evidence for the role of viral infections of mothers during pregnancy.

Indeed, ABA was the one therapy that faired positively in the meta‐analyses but with the important caveat that a one size fits all approaches fails to embrace the full spectrum of disorders. In any case, the argument seems to be over the effectiveness of clinical studies rather than the underlying therapies, with the main call being for the ability to adapt during the course of treatment as the child develops.

Causes and therapeutic options

Treatment of autism may have to adapt anyway with emerging insights into probable genetic or environmental causes. Indeed, in lieu of clear genetic factors, there is mounting evidence for the role of viral infections in mothers during pregnancy. This comes under the heading of maternal immune activation (MIA), which is mediated not by infectious viruses directly but by the activation of inflammatory pathways resulting in increased levels of cytokines and chemokines. Those, the authors assert, can cross the placental and blood–brain barriers to alter foetal neural development and cause conditions such as autism, schizophrenia, ADHD, and depression (Massrali et al, 2022).

However, there are many cases of maternal infection where the developing foetus remains unaffected. The study attempted to unpick this dichotomy by noting that neurodevelopment in the foetus could be linked to changes in the microbiome; indeed, the microbiome has been separately associated with immune function abnormalities that can lead to autism. If true, this could inform treatment or prevention through drugs that mediate MIA and retard immune molecules crossing the placental and blood–brain barriers. Similarly, therapies that address microbiome dysfunction might also be effective. Accordingly, there is plenty of scope for further research at the molecular level in addition to NPIs.

There is also ongoing research into the genetic background of autism, which has recently hit a snag in the UK, demonstrating how difficult it is to conduct studies on neurological disorders that carry a risk of stigma for those affected. The UK's autism genetics study called Spectrum 10 K, based at the University of Cambridge and led by Simon Baron‐Cohen, Director of the Autism Research Centre there, aims to sequence genomes from 10,000 people diagnosed with autism to search for genetic risk factors. But recruitment was paused in August 2021 after concerns raised by autistic advocates and some researchers about the study's goals and practices, despite approval from research ethics committees in the UK.

Among the concerns raised were fears that data could be misappropriated and that the perspectives of autism sufferers had not been properly taken into account. There was also a widespread feeling, voiced for example by Ilina Singh, professor of neuroscience and society at the University of Oxford, that research funding failed to prioritize ethics or engage with the public.

Eventually, the study was resumed in 2022 after a review, but the process highlighted growing disagreement over the definition of autism in the first place and whether the term “autism” is appropriate for all cases including mild ones. Whitehouse also acknowledged growing awareness of the need to avoid stigmatizing children with a condition that is complex and ill‐defined, which itself can be counterproductive. “Interventions are shifting to be more affirming of autistic identity, focused on societal acceptance,” he said. “For example, clinicians do not attempt to interrupt or decrease nonharmful repetitive behaviours but instead work to educate others to promote acceptance […].” Caught between the risk of stigmatization, vested interests in particular therapies, and a lack of understanding of the causes in the first place, autism research and therapy development has to navigate difficult seas.

EMBO reports (2023) 24: e56915