April marks Autism Acceptance Month, previously known as Autism Awareness Month. Autism spectrum disorder (ASD), or autism, refers to a range of common neurodevelopmental conditions characterized by social and communication challenges and pervasive repetitive behaviors. In December 2021 the Centers for Disease Control and Prevention (CDC) released new data showing that the number of children diagnosed with ASD in the US has increased from one in 52 to one in 44. Given the common nature of autism, and based on feedback from the autism community, several autism societies have shifted their messaging in the past couple of years from “awareness” to “acceptance” in order to focus on inclusiveness and advocacy for the human and civil rights of all people with ASD, promoting full acceptance and appreciation of the many contributions of autistic people.
Despite the high prevalence of ASD, challenges remain concerning its diagnosis, in particular early diagnosis. Early treatment of ASD helps with the development of communication, social, and learning abilities, and even with motor skills, vastly improving patients’ long-term outcomes and quality of life, and placing high importance on early diagnosis. Key opinion leaders interviewed by GlobalData agreed that improved early diagnosis is important, noting that the average age of diagnosis is too high given the benefits of early intervention. Research suggests on average, there is a three-year delay between initial parental concern and an actual diagnosis of ASD, and delays in diagnosis have been exacerbated by the COVID-19 pandemic.
At the recent American Academy of Neurology (AAN) 2022 Annual Meeting this month, data was presented from a pivotal trial of Canvas Dx, a Software as a Medical Device (SaMD) that aids in the diagnosis of ASD in patients between 18 and 72 months in a primary care setting. Canvas Dx, developed by Cognoa, uses artificial intelligence (AI) to process three inputs: a brief caregiver questionnaire, a questionnaire completed by a video analyst based on two 90 second videos of the child in a home setting, and a physician questionnaire completed during either a virtual or in-person meeting with the child. It then produces one of three outputs: positive for ASD, negative for ASD, or indeterminate in cases that may be particularly complex, thereby reducing the risk of a false classification. In the pivotal, double-blind study (NCT04151290), Canvas Dx produced either a positive or negative output with high accuracy compared to a specialist diagnosis using the DSM-V criteria, with a positive predictive value of 80.8% and a negative predictive value of 98.3%. Additionally, 91.0% of patients who received an indeterminate output were found to have one or more neurodevelopmental disorders, including ASD and non-ASD conditions, following specialist diagnosis. Despite the impressive predictive values, particularly the negative predictive value indicating very few false negatives, most of the participants (68.2%) received the indeterminate outcome and would therefore require further specialist diagnosis, potentially limiting the usefulness of Canvas Dx.
The results of the study demonstrated that Canvas Dx has the ability to aid in the accurate diagnosis of ASD in a primary care setting. This could help reduce the need for timely referrals to specialists for diagnosis, which should result in earlier diagnosis and therefore earlier interventions that can have maximum benefit. In the trial, the average age of patients who received a positive ASD diagnosis was 2.8 years, significantly earlier than the current average age of diagnosis, which is four years and three months. Not only can using Canvas Dx result in earlier diagnosis, but it can also result in more efficient use of specialist resources if less time is required purely for diagnosis. Additionally, the study found that patient assessment by a physician either in person or remotely did not affect the results, however, it should be noted that only 13.9% of the physician interviews were carried out remotely. While it is appealing to be able to use this technology remotely, thus removing additional potential barriers to diagnosis, a larger sample size of remote evaluation would be useful to fully validate this ability. Furthermore, across the whole trial, there was also no evidence of inconsistency regarding a patient’s race, ethnicity, sex, geography, household income, or parental education.
In support of the results, the FDA gave marketing authorization to Canvas Dx in June 2021. Wider implementation of this novel technology should allow for earlier diagnosis and interventions, which will likely result in improved quality of life for people with ASD; however, limitations remain. Given the importance of early diagnosis, there are a variety of other technologies available and being developed to aid in the diagnosis of ASD. For example, EarliTec Diagnostics is developing a device to capture looking behavior, an indicator of a neurodevelopmental disability, to aid in diagnosing autism in children ages 16 to 30 months. Stanford University is developing an autism-specific AI algorithm for early ASD diagnosis, using brain “fingerprints” of autism generated through the analysis of patterns of neural activity using functional magnetic resonance scans. Additionally, there is a large amount of research into genetic testing for diagnosing ASD as well as the development of new and improved early screening tools and questionnaires. Ultimately, it is likely that a combination of different techniques and technologies will be important for improving early diagnosis of ASD, rather than relying on one specific method.