Most studies of autism spectrum disorder (ASD) have focused on children; as the condition is a neurodevelopmental disorder, it is necessary to initiate studies to determine its cause as early in development as possible. However, this has led to a situation where the neurobiological basis of ASD in adults is poorly understood. Researchers still don’t really know whether abnormalities detected early in life persist, change, predict outcome or can be used to help identify autism. Moreover, the exclusive focus on infants means there are few potential biomarkers that might allow clinicians to individualise diagnostic or treatment approaches to help identify autism in adulthood.
But evidence reported in a paper entitled ‘Autism in adults. New biological findings and their translational implications to the cost of clinical services’, Brain Res. (2010) by Murphy, DGM, et al., has shown that brain anatomy and aging in people with ASD is significantly different when compared with controls.
Go deeper with GlobalData
These findings have made possible the development of a new type of brain scan using support vector machine (SVM) software that takes 15 minutes and was able to diagnose autism in adults with more than 90% accuracy in the research setting. Led by Dr Christine Ecker, of the department of forensic and neurodevelopmental sciences at the Institute of Psychiatry (IoP) at King’s College London, the Autism Imaging Multicentre Study (AIMS), which led to the creation of this new autism scanning software, is a collaborative research project between King’s College London and the universities of Oxford and Cambridge. It has been in progress for four years and 200 people have so far been scanned – 100 with autism and 100 healthy controls.
“There have been a lot of developments recently in terms of analytical techniques, and it has now become possible to train a computer programme to dissociate autism from controls on the basis of brain anatomy,” Ecker says. “At the moment, the diagnosis of autism is time-consuming and expensive, so if we had some sort of biomarker or simple biological test that we could use, it would really make everybody’s lives a lot easier.”
The study is essentially seeking to substantiate the clinical diagnosis for autism by biological measures. And as the disease has significant economic consequences for society, there is little doubt that an objective biomarker for ASD would be of immense value.
“Currently, it takes a whole day to diagnose someone, and because a team of clinicians is required, it is very expensive; it is estimated to cost around £2,000,” Ecker explains. “Furthermore, the diagnosis relies solely on clinical opinions and behavioural observations.”
How well do you really know your competitors?
Access the most comprehensive Company Profiles on the market, powered by GlobalData. Save hours of research. Gain competitive edge.
Thank you!
Your download email will arrive shortly
Not ready to buy yet? Download a free sample
We are confident about the unique quality of our Company Profiles. However, we want you to make the most beneficial decision for your business, so we offer a free sample that you can download by submitting the below form
By GlobalDataThis can be problematic when the tests are carried out on older patients of up to 50-60 years of age. “We also invite parents to give interviews about the behaviour of their children, but often parents are not available anymore to tell us what their son or daughter were like when they were younger,” Ecker says, “so to utilise additional biomarkers for the condition would really make it a lot easier to diagnose patients.”
There is an increasing recognition in the scientific community that researchers need to ‘translate’ lab-based findings into clinical settings, which is largely due to the economic consequences of diseases such as ASD. On top of the cost for diagnosis, the lifetime cost for someone with autism is estimated to be as much as £4.7m per person, most of this arising from healthcare needs (Murphy et al, 2010).
Moreover, the diagnosis of ASD is also often complicated by diagnostic ‘overshadowing’. For example, 25-50% of autism sufferers have attention deficit hyperactivity disorder (ADHD) symptoms that merit clinical treatment. Also, 25%-50% of people with ADHD are severely socially disabled and/or have at least mild ASD symptoms (Murphy et al, 2010).
Due to the above factors, the patient waiting lists for autism diagnosis are continuing to rise and the demand for skilled clinicians is growing. This has been compounded by increased awareness of the condition. Indeed, the number of people seen each year by the national assessment services for adults with ASD at the South London and Maudsley NHS Foundation Trust has increased approximately five-fold between 2005 and 2010 (see Figure 1).
Yet, although all of these patients have been extensively ‘triaged’, a huge proportion of them do not have ASD, which has significant cost implications. By contrast, there are many people suffering from autism in the community who are unaware of their condition, and thus find it hard to access services. Doctors need to be able to rapidly ‘triage’ people who do not require an expert assessment for ASD in order to offer more appropriate services all round.
Ecker is keen to note that the 15-minute brain scan is not intended as a replacement for the lengthy autism diagnosis process currently being used: “We are saying it shouldn’t be used instead of the conventional diagnosis. It should just provide some more information that can guide or facilitate the conventional diagnosis.”
How the software works
While it is abundantly clear that an objective biomarker for identifying ASD would be hugely beneficial, there had been little investigation into the potential of brain imaging until recently. Ecker’s study, however, is beginning to prove its value. She and her team investigated the diagnostic value of gray and white matter anatomical MRI scans using SVM software.
“It is a computer programme, which can be trained to find a pattern of regions in the brain that can be used to dissociate people with autism from healthy controls,” she explains. “It works in an automatic fashion so we have to train the computer programme first to find those regions, some of which are bigger, some smaller. It establishes the network of regions in a very well characterised sample. Once the computer has identified this pattern we can then scan someone new where we don’t know whether it’s a patient or a control, and the computer programme will tell us whether this new brain is more similar to someone with or without autism.”
Once the brief scan is complete, the programme takes a couple of hours to translate the findings into data. “We are trying to make it run a bit faster, but there are limitations,” Ecker adds.
The new scan clearly has great potential, and is able to correctly classify individuals with ASD with 90% specificity and sensitivity in the research setting. But researchers still don’t know how well the test will perform in the clinical setting. “So far we’ve only looked at very typical cases and very typical controls,” Ecker says. “But some people who come through the clinic have all sorts of co-morbidities, such as ADHD or social anxiety disorder, and we don’t know how well it will cope with these.”
AIMS researchers have thus far only looked at people with high-functioning autism, meaning they did not have a learning disability. “We don’t really know yet how well it works with people that do have a learning disability,” says Ecker. “We are also looking into distinguishing people with Asperger’s syndrome from those with high-functioning autism.”
The technique has been subject to scepticism in other areas. “Some people were worried about the costs, because scanning is usually quite expensive,” Ecker notes. “But we just acquire structural scans and it only takes 15 minutes, so it’s going to be between £150-£200 – not too bad.”
The researchers have also been criticised for the high rate of false positives that would result if the method was used as a population screening tool. Ecker responds: “This tool was not designed to scan the whole population of the UK; if there’s no suspicion of a diagnosis, why would we scan someone? It’s really just designed to scan people at risk, or to confirm the conventional diagnosis or if informants are not present, which would of course reduce the number of false positives drastically.”
The future
Although the technique was designed for adults, Ecker hopes it will be applied to children. “If you look at the research on brain anatomy and ASD, the differences are more pronounced during childhood than during adulthood when the brain has matured,” she says. “We haven’t looked at the data ourselves because we mainly work with adults, but other publications have looked at kids using a similar technique and they found high accuracies of 85%-90%.”
The new method has had extremely positive feedback from patients, far beyond Ecker’s expectations: “At first we were a bit worried, thinking that people would not like the scanner as it can feel a little claustrophobic and is quite noisy, but we’ve been contacted by so many people giving us positive feedback – they really would like to see a biological measure confirming their diagnosis.”
On top of this, doctors are welcoming the potential of brain imaging. “At the moment, it’s a little bit difficult because even with a team of expert clinicians that are trained to spot autism reliably and accurately, it is often not a black and white diagnosis,” Ecker explains. “It can often be difficult to spot simply because it’s such a wide-spectrum condition.
“When there’s a suspicion of autism, GPs refer patients to adult specialist services. So if we could offer GPs a biomarker to test for ASD, it would make it a lot easier. I guess anything we can do to facilitate a diagnosis would be welcome.”
It will still be a few years before brain scanning for autism appears in the clinical setting, but this is currently being tested by Ecker and her team. They are scanning everybody who comes through their clinic regardless of whether they end up with a diagnosis or not and comparing the performance of their classifier with the clinical diagnosis. “This is to see how well it works in the clinical setting. We are going to have the data in the next two years,” Ecker says. “If it works out, we could potentially use it in the clinical setting in the next three to four years.”
Most hospitals already have a clinical MRI scanner so, according to Ecker, it will be an easy process to integrate the technique into healthcare settings. “The only problem is that technically, it’s not so easy to pool the data across scanners because every one is unique and you get different sorts of data from different scanners,” she acknowledges.
“For our study, we developed techniques that allowed us to combine data across scanners, but these are specialist acquisition techniques. It’s not impossible but it does require more pilot testing.
“If we had a biological test for autism, it could be used in various ways. Ultimately we hope that it will improve access to treatment and interventions.”
